proceedings - user.iiasa.ac.atjonas/cv iiasa/uploads/3rduncws 22-24sep10... · jaishankar pandey...

LVIV POLYTECHNIC NATIONAL UNIVERSITY

INTERNATIONAL INSTITUTE FOR APPLIED

SYSTEMS ANALYSIS

SYSTEMS RESEARCH INSTITUTE

Polish Academy of Sciences

3rd International Workshop on Uncertainty in Greenhouse Gas Inventories

PROCEEDINGS

LVIV POLYTECHNIC NATIONAL UNIVERSITY

Lviv, Ukraine

September 22-24, 2010

ISBN 978-966-8460-81-4

3rd International Workshop on Uncertainty in Greenhouse Gas Inventories, September 22-24, 2010, Lviv, Ukraine

Printed in the form submitted by authors.

Approved by the Scientific Committee of the 3rd International Workshop on Uncertainty in Greenhouse Gas Inventories, September 22-24, 2010, Lviv, Ukraine.

Acknowledgement Publication was partially financed by the State Department of Environment Protection in Lviv Region, Ukraine.

Printed by “Print on Denand” – FOP Soroka S.V.

© Lviv Polytechnic National University, 2010.


3

About the Workshop

The assessment of greenhouse gases (GHGs) emitted to and removed from the atmosphere is high on both political and scientific agendas internationally. Under the United Nations Framework Convention on Climate Change (UNFCCC), parties to the Convention have published national GHG inventories, or national communications to the UNFCCC, since the early 1990s. Methods for proper accounting of human-induced GHG sources and sinks at national scales have been stipulated by institutions such as the Intergovernmental Panel on Climate Change (IPCC) and many countries have been producing national assessments for well over a decade. However, as increasing international concern and cooperation aim at policy-oriented solutions to the climate change problem, a number of issues have begun to arise regarding verification and compliance under both proposed and legislated schemes meant to reduce the human-induced global climate impact.

The issues of concern at the International Workshops on Uncertainty in Greenhouse Gas Inventories − the 1st Workshop was held on September 24-25, 2004, in Warsaw, Poland; and the 2nd Workshop on September 27-28, 2007, in Laxenburg, Austria − are rooted in the level of confidence with which national emission assessments can be performed, as well as the management of uncertainty and its role in developing informed policy. The Workshops cover state-of-the-art research and developments in accounting, verifying and trading GHG emissions and provide a multidisciplinary forum for international experts to address the methodological uncertainties underlying these activities. The topics of interest center around national GHG emission inventories, bottom-up versus top-down emission analyses, signal processing and detection, verification and compliance, and emission trading schemes.

The 3rd International Workshop on Uncertainty in Greenhouse Gas Inventories took place September 22-24, 2010 at the Lviv Polytechnic National University (LPNU) in Lviv, Ukraine. This Workshop was jointly organized by the Austrian-based International Institute for Applied Systems Analysis, the Systems Research Institute of the Polish Academy of Sciences, and the Lviv Polytechnic National University in Ukraine. Main topics:

− achieving reliable national GHG inventories;

− accounting emissions across spatial scales (project, national, regional/continental);

− bottom-up versus top-down emission analyses;

− detecting and analyzing emission changes;

− reconciling short-term commitments and long-term targets;

− verification and compliance;

− trading emissions;

− communicating, negotiating and effectively using uncertainty.

Special attention was given to translating scientists’ understanding of uncertainty into options of use for policy makers to consider uncertainty in frameworks of negotiating climate change.


4

Scientific Committee

Chairmen: Rostyslav Bun (Lviv Polytechnic National University, Ukraine)

Zbigniew Nahorski (Polish Academy of Sciences, Poland)

Members: Yuri Ermoliev (International Institute for Applied Systems Analysis, Austria)

Evgueni Gordov (Siberian Center for Environmental Research & Training, Russia)

Giacomo Grassi (Joint Research Centre, Institute for Environment and Sustainability, Italy)

Mykola Gusti (Lviv Polytechnic National University, Ukraine)

Khrystyna Hamal (Boychuk) (Lviv Polytechnic National University, Ukraine) - Secretary

Javier Hanna (Reporting, Data and Analysis Programme, UNFCCC Secretariat)

Olgierd Hryniewicz (Polish Academy of Sciences, Poland)

Matthias Jonas (International Institute for Applied Systems Analysis, Austria)

Gregg Marland (Environmental Sciences Division, Oak Ridge National Laboratory, USA)

Sten Nilsson (International Institute for Applied Systems Analysis, Austria)

Krzysztof Olendrzyński (National Emission Centre, Poland)

Jaishankar Pandey (National Environmental Engineering Research Institute, India)

Jean-Daniel Paris (Laboratoire des Sciences du Climat et de l'Environnement, France)

Ariel Macaspac Penetrante (International Institute for Applied Systems Analysis, Austria)

Stefan Pickl (Universität der Bundeswehr München, Germany)

Anatoly Shvidenko (International Institute for Applied Systems Analysis, Austria)

Jochen Theloke (IER University of Stuttgart, Germany)

Isabel van den Wyngaert (Centre of Ecosystems, Alterra, The Netherlands)

William I. Zartman (John Hopkins University, USA)


5

Table of Contents

Viorel Blujdea, Giacomo Grassi, Roberto Pilli Estimating the uncertainty of the EU 15 forest CO2 sink ........................................... ..9

Keith A. Brown, Joanna MacCarthy, John D. Watterson, Jenny Thomas Uncertainties in national inventory emissions of methane from landfills: A UK

case study ……….. …….. .................................................................... ..21

Dieter Cuypers, Tom Dauwe, Kristien Aernouts, Ils Moorkens, Johan Brouwers Comparison between energy and emission data reported under the ETS and

energy balance and greenhouse gas inventory of Flanders .................... ..31

Dhari Al-Ajmi Climate Change in the Gulf Countries – Situation and Reactions ............................. ..41

Olga Diukanova, Igor Liashenko Addressing uncertainties of GhG emission abatement in Ukraine .................................. ..47

T. Ermolieva, Y. Ermoliev, M. Jonas, G. Fischer,M. Makowski, F. Wagner, W. Winiwarter A model for robust emission trading under uncertainties .......................................... ..57

Pedro Faria Uncertainty and variability In corporate GHG inventories and reporting ................. ..65

Mykola Gusti Uncertainty of BAU emissions in LULUCF sector: Sensitivity analysis of the

Global Forest Model ……………………….. ........................................ ..73

Khrystyna Hamal, Rostyslav Bun, Nestor Shpak, Olena Yaremchyshyn Spatial cadastres of GHG emissions: Accounting for uncertainty ............................. ..81

Joanna Horabik, Zbigniew Nahorski Improving resolution of spatial inventory with a statistical inference approach ....... ..91

Olgierd Hryniewicz, Zbigniew Nahorski, Joanna Horabik, Matthias Jonas Compliance for uncertain inventories: Yet another look? ....................................... ..101

Wolfram Joerss Determination of the uncertainties of the German emission inventories for

particulate matter (PM10 & PM2.5) and aerosol precursors (SO2, NOx, NH3 & NMVOC) using Monte-Carlo analysis .................................... ..109

Matthias Jonas, Volker Krey, Fabian Wagner, Gregg Marland, Zbigniew Nahorski Dealing with Uncertainty in Greenhouse Gas Inventories in an Emissions

Constrained World …………………… ............................................... ..119

Vyacheslav I. Kharuk, Maria L. Dvinskaya, Sergey T. Im The potential impact of CO2 and air temperature increases on krummholz’s

transformation into arborescent form in the southern Siberian Mountains ………………………………. ............................................ ..129


6

Ivan Lakyda Peculiarities of sequestered carbon assessment in urban forests of Kyiv ................ ..139

Lyubov Lebed‘ Reducing uncertainties in GHG inventory using modern agricultural lands

monitoring systems ……………………………… .............................. ..143

Derek Lemoine, Sabine Fuss, Jana Szolgayova, Michael Obersteiner Abatement, R&D policies, and negative emission technology in climate

mitigation strategies …………………… ............................................. ..149

Myroslava Lesiv, Andriy Bun, Mykola Medykovsky Uncertainties of results of GHG inventories: Europe 2020 ..................................... ..159

George Magalhães, Francisco do Espirito Santo Filho, João Wagner Alves, Matheus Kelson, Roberta Moraes Reducing the uncertainty of methane recovered (R) in greenhouse gas inventories

from waste sector and of adjustment factor (AF) in landfill gas projects under the clean development mechanism ............................... ..165

Gregg Marland The U.S. NRC report on monitoring and verification of national greenhouse gas

emissions inventories …………………………… ............................... ..177

Zbigniew Nahorski, Jarosław Stańczak, Piotr Pałka Multi-agent approach to simulation of the greenhouse gases emission permits

market ………………………………….. ............................................. ..183

Maria Nijnik, Guillaume Pajot Accounting for uncertainties and time preference in economic analysis of tackling

climate change through forestry ……….. ............................................. ..195

Sang-hyup Oh, Gwisuk Heo, Jin-Chun Woo Uncertainty of site-specific FOD for the national inventory

of methane emission …………………. ................................ …………..207

Jos G.J. Olivier, John A. van Aardenne, Suvi Monni, Ulrike M. Döring, Jeroen A.H.W. Peters, Greet Janssens-Maenhout Application of the IPCC uncertainty methods to EDGAR 4.1 global greenhouse

gas inventories ……………………………… ...................................... ..219

Jean Pierre Ometto, Ana Paula Dutra Aguiar, Carlos A. Nobre Reducing uncertainties on carbon emissions from tropical deforestation: Brazil

Amazon study case ………………………… ....................................... ..227

J.S. Pandey, R. Kumar, S.R. Wate , T. Chakrabarti Application of spatio-temporal emission-factors (STEFs) for carbon footprinting

of Indian coastal zones ………………………. .................................... ..233

Peter Rafaj, Markus Amann, Henning Wuester Changes in European air emissions 1970 – 2010: decomposition of determining

factors ……………………………………… ....................................... ..241


7

Rodolfo Rubén Salassa Boix The government mechanisms of environmental protection .................................... ...251

Kwang-IL Tak, Hyeon-Kyu Won, Kyeong-hak Lee, Man-Yong Shin Uncertainties of forest carbon accounting with an international application

of the carbon budget model for Canadian forest sector (CBM-CFS): South Korean case …………………… ................................................ ..259

Jochen Theloke, Folke Dettling Uncertainties implied in the country specific baselines caused by different

approaches applied for recalculating the NMVOC emissions into CO2 equivalents …………………………………........................................ ..267

Balendra Thiruchittampalam, Jochen Theloke, Melinda Uzbasich, Matthias Kopp, Rainer Friedrich Analysis and comparison of uncertainty assessment methodologies for high

resolution Greenhouse Gas emission models ....................................... ..271

Nina E. Uvarova The improvement of greenhouse gas inventory as a tool for reduction of emission

uncertainties for oil activities in Russia ............................................... ..285

Jörg Verstraete Using a fuzzy inference system for the map overlay problem ................................ ..289


8


9

Estimating the uncertainty of the EU 15 forest CO2 sink

Viorel Blujdea1, Giacomo Grassi1, Roberto Pilli1

1European Commission - Joint Research Centre, Institute for Environment and Sustainability, Climate Change Unit, TP 029, Via Fermi, 21027 Ispra (VA), Italy

[email protected]

Abstract

EU GHG inventory estimates are compiled and summed-up from its Member States individual submissions, but we only consider here CO2 emission/removal related to 5A – Forestland of old 15 member states confined to EU’s KP commitments. Current analysis is done on pools, different by that on land use subcategories, with the purpose to show eventual contribution of highly uncertain processes (harvest, disturbances). Under consideration that inputs estimates of member states sinks are rather independent due to country specific datasets and methodologies, the assessment of the EU 15 5A aggregated uncertainty considers several regimes of covariance given by non independent estimates (e.g. use of default factors), which may double it compared to independent data inputs. Level uncertainty at EU’s 5A1 sink varies between 34-37 % upon the share of the contributing pools and their uncertainty. Under current level of uncertainty the confidence in EU 5A1 sink trend is around 4,42 %, with correlations in time and across MS reducing the confidence interval. Disturbances have no important contribution to overall uncertainty.

Keywords: EU 15, forest sink, Monte Carlo simulations, pools, uncertainty

1. Introduction

EU GHG inventory is a particular case of submission by a Party under UNFCCC, as it results from summing-up the net emission/removal of each Member State of the European Union (EU MS), in a compiling bottom-up process. Meanwhile, emissions reduction commitment under Kyoto Protocol is 8 % assumed in the same way (nevertheless, it applies only to 15 MS of the former European Community). LULUCF is integral part of Annex I reporting and accounting even its inclusion generates increased overall uncertainty [1],[2],[3].

Uncertainty is the lack of confidence in a single value [4]. Sources of uncertainties in the MS GHG inventories include error related to datasets (i.e. inconsistent definitions, partially adequate proxies); incomplete or simplified models and methodologies, incomplete monitoring and reporting (i.e. partially counted); missing data; limited understanding of the processes controlling sources&sinks and GHG fluxes, and inconsistent processing and aggregating of uncertainty of the components. Gillenwater et al. (2004) divided uncertainty in GHG inventories into: 1) knowledge uncertainty caused by weak science (i.e. the case of non-CO2 GHG) and 2) estimation uncertainty containing both model and input parameters uncertainty [5]. Contrary to system variability that cannot be reduced (as inherent quantitative variation), the uncertainty arises from incapacity to determine the true value of the quantity (i.e. related to our measurement or efforts, incomplete assessment, process knowledge) and can often be reduced by further study (i.e. reduction of bias) or the system could be better described by using larger or more representative data sets (i.e. reducing random error) [6].

Currently, the LULUCF uncertainty shows similar ranges for the Annex I countries, both regarding total net emissions and GHGs, while maintains in the same order of


10

magnitude as Tier 1 assessments, although a somewhat higher trend uncertainty was found by Tier 2 method [7],[8],[9]. Uncertainty estimates provided by Parties are difficult to compare, and there is a concern that “uncertainty in the uncertainty estimates may be even larger than uncertainties in emission estimates” [10], under missing verification tools and partial accounting of land [3].

A systematic, quantitative analysis of the uncertainty shows the degree of confidence in the inventory estimates or emission reductions and should help focusing the effort for their minimization. Often it is seen as discounting method in emissions reduction projects or buffering in commercial transactions [11],[12]. The tiered approach provided by the IPCC helps quantitative assessment [11], but also raises additional issues (i.e. missing data). Approaches based on resimulations allow better characterization of GHG inventory uncertainties of level and trend [13],[14], as that can handle non-normal (and even empirical) distributions, correlations between input parameters and extreme uncertainties (as in nitrous oxide from agricultural soils).

To understand the factors affecting the EU GHG aggregated estimates, the uncertainty of emissions/removals by sources/sinks in forestland sub-categories at MS level are qualitatively analyzed, but a full-blown quantitative analysis can not be done (under lack of full transparency and especially MS’s heterogenous methodologies). Overall EU quantification of uncertainty is shown in EU NIR 2010 (available at: http://www.eea.europa.eu/publications/ european-union-greenhouse-gas-inventory-2010).

The present work focus on challenges related to assessment of the aggregated uncertainty of EU 15 sink and individual pools and the uncertainty in the trend taking into consideration an approach based on C stock change in pools (not on land use subcategories) in order to reveal the effect of forest disturbance and eventual co-variations at EU 15 level. Also, an analysis of the MS and pools for which it is important to have more accurate estimates is done. Target gas is CO2 related to 5A1 and 5A2, the most significant contributors the LULUCF sink, in order to reveal methodological issues and criticalities in the quantification of uncertainty at EU aggregated level, potential help to set priorities for the GHG inventory improvement from EU perspective and confidence in the emission reduction compliance.

2. Assumptions and methods

EU 15 GHG inventory sums up the ‘best estimates’ reported by the MS and aggregates underlying uncertainties. This way, we assume systematic errors are removed at MS level under QA/QC implemented. The Tier 1 uncertainty analysis achieved on land subcategories in EU GHG inventory 2010 weakly account for different correlations under complicate spreadsheets, whereas the possibility to easy account for correlations (i.e. within the year, temporal) is generally accepted as strong side of the re-simulation techniques (Tier 2 of IPCC based on Monte Carlo - MC techniques family). Flexibility of the MC analysis consists in fact that it offers complete information on the distribution of occurrence probabilities of the random variable, with distribution moments providing adequate metrics. Like any method, MC only provides satisfactory results if it is properly implemented under requirement that analyst to have scientific and technical understanding of the inventory. Of course, the results will only be valid to the extent that the input data, including any expert judgements, are sound [15].


11

Data gap filling. Under incomplete reporting of emission/removal, gap filing is assumed practically under IPCC Tier 1 (neutral pools for DM and SOM, poorly reported pools). Also, in order to ensure consistent representation of lands across EU 15, UK’s 5A2 estimate is analyzed as being 5A1 as to avoid any misbalance given by its sink (UK reports all forest established since 1920 as being under conversion in 5A2 with previous 1920 forests as neutral). Uncertainty associated with various pools is assumed the simple average of values submitted by the other MS. Data is provided by 2010 CRF and represents the C stock changes assimilated with the aggregated pools and sources of CO2: biomass gain, biomass loss, dead organic matter, mineral and organic soil organic mater and disturbances. This is atypical processing of data: loss refers only to harvest, while disturbances are re-computed and their contribution analyzed separately. Reporting practice shows nevertheless that the “gain” and “loss”, when reported as such, have different sources of data on the general background that the two processes have opposed determinants (one mainly natural, another one mainly anthropogenic), with suspected unreliable statistics on wood harvest. Winiwarter and Rypdal (2001) consider that discrimination of sink in the two processes is meaningless in Austria due to the kind of data available. In case stock change based reporting by a MS then only the available value is used for assessment (as net gain). Disturbance emission is not apparent in the CRFs but they were recomputed by the JRC tool [16] building on IPCC default equations between emission of N2O and CH4 reported in CRF Table 5(V) and CO2 from burning biomass (IPCC GPG 2003, table 3A.1.16).

Probability distribution curves. Analysis is made on total annual C stocks changes in the aggregated pools and sources assuming full mass balance principle with CO2 emissions. They are continuous random variables with values over intervals as inferred from their proxies (i.e. MS’s C stock changes on pools and emissions by sources). The parameters of the PDFs were set from MS reported data. In order to test the type of PDFs basic data was analyzed by Goodness-to-fit of the @Risk and shape testing facility assuming the similarity of C stock changes and emissions from sources (i.e. disturbance) to emission factors distribution. Logical assumptions were made: gain can not be negative; harvest can not be positive and sink could be negative (thus applying truncations in the re-sampling of distributions).

Co-variation of factors reflects our qualitative understanding of the relations amongst MS estimates, as far as their quantification can not be performed as there is no explicit data but heterogeneous methodologies and computation pathways available by individual MS. In running simulations it was considered both time & across MS dependent and independent estimates.

Sensitivity Analysis completes and enhances the uncertainty analysis by ranking the inputs variation on the variation of the output. Sensitivity analysis regression allows to breakdown the total’s variance on the contribution of individual inputs, thus allowing exploring and ranking the uncertain inputs and assumptions according to their effect, on a short list of influential factors. The regression sensitivity (Standard B coefficient) is a normalized parameter assessing how sensitive is the change of output in StdDev terms to a change in the StdDev of the inputs, assuming that the regression is linear and intercept X axe in the origin (0).


12

Uncertainty of the trend was computed based on simulations between base year and latest submitted year both for C stock changes in individual pool/sources and EU 15 net estimate under various regimes of correlation.

The software package, @Risk by Palisade is used to assess the propagation of uncertainties in aggregation process, as it gives streamlined operability under CRF format. Overall the computation pathway follows the arithmetic between pools and sources. Convergence of outputs (i.e. central value/average and standard deviation) to less than 1.5 % was followed. Re-sampling was fully random. Uncertainty parameter is relative (%), always computed as 2 StdDev to central expected value of the parameter assuming normally distributed errors (not as 2.5&97.5 % percentile, which could give narrower confidence intervals). In order to test the robustness of the results of the MC analysis, we performed repeated same-inputs runs and compared with EU GHG inventory estimates and EU Tier 1 uncertainty assessment on land subcategories.

3. Results and discussions

GHG inventories have be compiled with values which represent the best available estimates, affected, to the extent possible, by a known degree of variability and a low degree of uncertainty [17]. Overall, the accuracy of GHG estimates is mainly related to variability, while precision is mainly related to uncertainty. Practically, GHG inventories are not dealing with intrinsic variability specific to LULUCF as it should be dealt with by the underlying data collection procedures (e.g. by National Forest Inventories sampling).

EU aggregated estimates are often argued from comparability perspective under methodological diversity adopted by the MS, thus supplementary uncertainty is quantified at EU level despite consistent and adequate approaches at MS level. Sources for EU 15 5A sink uncertainty are propagated from MS inventories: systematic errors (i.e. omission like missing removals/emissions reported as Tier 1 or not reported; misunderstanding or wrong interpretation of certain definitions; datasets or statistics)1, missing or loose quantitative assessments of uncertainty, while also arises from aggregation process at EU (i.e. common use of default factors).

Uncertainty of annual CO2 emission/removal reported by EU 15’s MS in their NIRs varies on pools: 10-30% for biomass gains and losses, 28-107% for DOM, 20-184 % on 5A1 and 50-124 % on 5A2 lands for SOM in mineral soils, 51-78 % on 5A1 and 51-65 % on 5A2 lands for SOM in organic soils. Overall uncertainty reported by MS varies between 25-100 % in 5A1 and 28-100 % in 5A2. Uncertainty of emissions from disturbances is occasionally reported by MS, in general 100 %. In fact, the forest disturbances effect is hidden in the sink, as included in the loss. There is general lack of transparency (i.e. explicit datasets, despite often explicit methodologies in the MS’s NIRs) in reporting such emissions which does not allow understanding their overall effect on the EU sink. In fact, there are inconsistencies with other datasets (i.e. area reported under European Forest Fire Information System) which is not always explained by underlying methodologies.

1 Key methodological information and reporting completeness on land use subcategories & pools approached by the MS of EU 27 are provided in the Annual European Union greenhouse gas inventory 1990 – 2008 and inventory report 2010, available at: http://www.eea.europa.eu/publications/european-union-greenhouse-gas-inventory-2010


13

Uncertainty assessment methodologies followed by MS is error propagation method, with few exceptions (i.e. United Kingdom). Basically, sources assessed to have high uncertainty are frequently assumed as log-normal or beta distribution (having a long tail covering probabilities of values larger than the mean) [7]. C stock change factors or underlying parameters are prone to distribute asymmetrically (i.e. BEFs [18]), compared to total emissions which reflect complex processes thus compromised as symmetric distributions. Empirical PDF for C stock change factors in biomass gain could be constructed based on available proxy data (i.e. forest stands annual growth on age classes2) (Table 1). The type of distributions have to more carefully studied (e.g. by checking empirical data, especially for SOM and DOM), both for C stock change factors and total/pools C stock change, under various management approaches (i.e. growing share of old or unmanaged forests).

Table 1. Expected PDFs for C stock change factors, pools and sources

Pool/Source

Assumptions

PDF of C stock change

factors

PDF of emissions/re

moval

Issues and restrictions

Biomass growth

Symmetrical distribution of the mean value (coefficient of variation <30%)

Normal/ Lognormal

Normal / lognormal

C stock change can not be negative or nil (distributions truncated to zero as to avoid the simulation with negative values)

Biomass loss Symmetrical distribution of the mean value (coefficient of variation <30%)

Normal/ lognormal/ uniform

Normal/ lognormal/uniform

C stock change can be negative or nil, Normal distribution truncate by the value of annual growth and zero. Long tail may be assumed as collection of wood may not be fully counted (i.e. small farming, illegal cuts). If several datasets are available by the MS then they can be used to generate uncertainty.

Disturbances Assymetrical distribution, standard deviation equal or greater than 30%

Lognormal Normal/ Lognormal

C stock change can be negative or nil Estimates are computed as a product of several parameters

SOM - Organic soils

Central value could be zero (both negative to positive values), possibly asymmetric

Normal/ triangular

Normal C stock change can be negative, nil or positive

SOM - Mineral soils

Central value could be zero (both negative to positive values), possibly asymmetric

Normal/ triangular


Dead Organic Matter (litter &dead wood)

Reported values are the most expected ones and the minimum and maximum values of the range (or standard deviations)

Normal/ triangular


In theory, summing up many PDFs results in a normally distributed sum, which is the

key assumption we follow for current EU 15 5A sink analysis. Expected EU 5A1 annual sink was 281 (190 – 372) in 1990 and 294 (192 – 396) th.

GgCO2 in 2008 (Table 2). Assuming totally independent estimates of MS,

2 derived from European Forest Yield Tables Database: http://afoludata.jrc.ec.europa.eu/index.php/public_area/data_and_tools


14

the aggregated 5A1 sink’s relative uncertainty in 1990 was 34 % (range between 25÷49%) and 37 % in 2008 (ranging from 26÷54%) and it practically maintains in this range for each individual year since 1990. The EU aggregated sink was normally distributed, but positively skewed; the long right tail may reflect the susceptibility to increased uncertainty (nevertheless, values with very low occurrence probability). By contrary, PDFs of the relative uncertainty of aggregated C stock change in each pool are negatively skewed. The CDFs show long tails toward small values (but less than 2.5 % frequency) which may reflect extremes simulated values of C stock changes and inaccurate and abnormal input data. Also, they all show very high kurtosis (> 250) with fat tails within the confidence interval, practically low even distributions (likely extreme values explained by the wide range of the inputs: some MS estimates and their uncertainty). Consistently, relative uncertainty of C stock changes in each aggregate pool distribute normally, but in all cases there is very narrow 95 % confidence interval (maximum 5 percentage points for both time correlated and not correlated estimates). PDFs of DOM and SOM mineral soils also expand in the negative domain (although values show frequency less than 2.5 %), with other pools being either totally positive or negative.

Table 2. Simulated uncertainties in aggregated EU 15 5A1 sink (2008, independent estimates)

Parameters

Estimated changes in C stock (Gg C) Overall

Sink (gG

CO2)

Biom Gain

Biom Loss

DOM SOM

Mineral SOM

Organic

Actual Distur-bances

Expected sink 155815 -65617 2850 13709 -4321 -1905 293053

Standard Deviation 14689 6190 1015 4561 1083 652 52285

Aggregated uncertainty

19% (18-20)

19% (16-17)

69% (68-71)

69% (68-71)

50% (50-51)

69% (68-71)

37% (26-54)

Sensitivity knowledge shows the sources undermining the EU 5A sink accuracy and

allows deeper understanding of the behavior of this sink system. EU 15 5A sink and its relative uncertainty are sensitive in opposite way to change of inputs, with a decrease of relative uncertainty with increasing MS’s estimates. Most important changes are given by biomass gain and biomass loss (i.e. Italy, France, Finland) with smaller contribution of SOM mineral (i.e. Italy, Finland) and organic soils (i.e. Sweden, Finland). Sensitivity of the relative uncertainty to MS pool’s changes is low (R2<0.5).

Under erratic time and space occurrence of emissions from disturbance, there is apparent low effect on EU 15 sink uncertainty under low amount of CO2 involved. Annual average CO2 emission from forest fire is some 9.5 Gg, roughly representing less 1 % of 5A1 sink. Assuming there is no any CO2 emission from forest fires, total EU 15 sink uncertainty is ~ 2 percentage points less (with narrower 95% interval than in case of reported actual emissions). By contrary, assuming that maximum CO2 emissions


15

occurring in each MS over 1990-2008 would occur simultaneously in a single year (i.e. in 2008 with total emissions of 22 th. Gg CO2), the sink of that particular year would decrease by some 3 % and expected relative uncertainty would increase by less than 2 percentage points (with 95% interval slowly widens on both sides, but especially toward higher values by some 5 percentage points).

Fact is that MS annual estimates may not be totally independent at EU. Co-variation is caused by the MS use of IPCC default or other factors (e.g. BEFs, root-to-shoot; wood density, allometric equations). Notably, there is no increase of the overall level uncertainty in case of 100 % time correlated estimates for same MS (i.e. 1990 and 2008), but the full correlation between both across MS and times lead to double central values of uncertainty of 78 % in 5A1 (range between 42÷285%) and 29 % in 5A2 (23÷41%). Various correlations would bring also change of frequency pattern, with decreasing probability for central expected value (Figure 1). Nevertheless, a thorough check of the NIRs information (i.e. datasets, methodologies, tiers) provides the likeliness of very low correlation pattern for the annual and trend emissions and associated uncertainty within EU 15 5A sink (Table 3), thus lowest uncertainty values of assessed ranges are considered realistic.

Figure 1. Distribution functions for independent and correlated estimates of the 5A1 sink in 2008

Reporting under 5A2 is largely inconsistent (e.g. time series start in various years),

thus current estimates may be unrealistic, but the uncertainty estimate could still be valuable. For 1990 it is 33 % (ranges between 24-46 %) in 1990 and 16 % (14-19 %) in 2008, with no difference between correlated or non-correlated estimates (for a MS in time). As in 5A1, C sink is sensitive to the main sink contributors’ change, while the relative uncertainty is decreasing with changes of the sink. 5A2 sink seems not affected by disturbance as they are reported under 5A1. Comparatively, in 5A2, there is much higher influence of SOM and DOM on the total uncertainty. Annual sink is in 2008 some 46 MtCO2 (with a 95% confidence interval of 30-63 Mt CO2).

0

0.01

0.02

0.03

0 100 200 300 400 500 600

5A1 sink (th. Gg CO2)

Fre

qu

en

cy

Correlatedest imates

Independentest imates


16

Table 3. Possible range of dependency, co-variation for various pools in the EU 15 MS

Parameter Across MS In time for each individual MS

Biomass gain Biomass loss

0% - independent data, country specific. Co-variation may be given by the use of default or common BEFs, WD and C content

0 – 100 %, MS estimates could be dependent because of the use of same parameters. Different NFI data and models (which may give systematic errors)

Disturbances 0-100 % within some groups of MS

0-100 %, if same default parameters used in time

SOM organic & SOM mineral soils and DOM

100 %, if Tier 1 or 2 is used for reporting by the country

0 %, if Tier 3 is used for reporting by the country 100 %, if Tier 1 or 2 is used for reporting by the country

Activity data 0 % 0-100 % depending on the reporting datasets

GHG inventories provide annual estimates of GHG emissions/CO2 removal, but

underlying data is not collected annually, which makes the estimates to not be totally independent over the years, while actual data used to not reflect the real emissions/removal in that year, thus real trend is also uncertain, unknown. Trend is dominated by the sinks/sources changing level across time (i.e. between base year and 2008), with highest influence given by strongly increasing or decreasing rates of the parts. MS do not report different relative uncertainty for different years (i.e. often recalculate previous years estimates with most updated methods, in NFI datasets), thus any change of relative uncertainty in time is practically only given by change of sink/source shares in the total EU sink (i.e. improvements are regularly implemented across all time series). Uncertainty of the base year does not receive any particular assessment effort, thus an acceptable level of the trend uncertainty (i.e. 1 % compared to base year) is likely not achievable, especially if the estimation methodologies differ in time [2]. Under independent estimates the uncertainty in the trend is 4.42 %, with a range between -36 to +71%. If assume maximal disturbance occurrence, it negligibly increases. Both in case of time correlated variables and full correlation across MS estimates the relative uncertainty remains at same level, but with much narrower range: -11 to +10 %. Uncertainty in the trend is influenced by same factors: C stock changes in Biomass gain and loss and SOM mineral soils. Actually, accounting for correlations and asymmetrical distribution functions does not necessarily lead to a significant increase in uncertainty in total GHG emissions [8], due to the fact that there is practically no different uncertainty estimated for the base year and latest year. Uncertainty in the trends on C stock changes in pools is not changing either, showing same narrowing of the uncertainty intervals.

Currently incomplete GHG inventories (missing pools, sources) and lacking uncertainty quantification prevent an appropriate assessment of the developments/improvement needs at EU level, but obvious candidates are DOM, SOM and disturbance. In fact there is a good question: is there any added value on reporting uncertainty on C stock change on pools, compared to IPCC recommended one


17

on AD and EF on LU subcategory? One shortage of reporting on AD and EF is the whole complexity reflected in the “single value” of the EF and not in the “total emissions”. On the other side, hidden data on disturbances remain critical in the sense of emissions amount and uncertainty, improvement potential. In order to not change the current CRF tables (i.e. in CRF add a new column under “loss”) NIRs should provide detailed tables on emissions from forest fires and/or other disturbances. Actually, forestry resources assessments are traditionally well supported by statistics, which seems not readily available (i.e. forest yield tables provide absolute figures), but NFIs continually improve [19],[20]. This paper approach on C stock changes in MS’s pools could act as verification tool of approaches based on land use. Current confidence of the uncertainty level is only supported reciprocally by individual MS assessments and other Annex I Parties, with future improvement needs requested by the coming commitments (i.e. verification methods). Actually, simulations based approach add value compared to Tier 1 in terms of defining probability (of the expected value, thresholds); EU perspective on MS contributing uncertainties and changes; risks analysis (i.e. related to disturbances); probability to maintain current sink over some time horizon; highlight methodological improvements needs. Also, independent uncertainty analysis (by the MS GHG assessments) would help clarify how / if it LULUCF entirely or some components (i.e. pools) to be in(ex)cluded in emission trading schemes [7].

Overall, the EU-15 case helps to understand the challenges of assessing the uncertainty, through bottom up aggregation of net removal/emissions. Uncertainty in EU 15 5A sink is in the same order of magnitude as obtained by Tier 1 analysis (reported in EU GHG NIR 2010), although somewhat less trend uncertainty was found by Monte Carlo. The exercise of reducing uncertainty in reporting may involve reporting based on growing stock, which does not involve the area as AD. Aggregation from MS up has the particularity to not add uncertainty, except if there is any co-variation. Fact is that EU sink is equal to MS sums, but the uncertain amount decreases with aggregation, under reciprocal compensation.

Aggregation of several uncertain estimates has particular importance when one has to early assess the probability to reach a certain fix target or to prove the compliance with it (like in KP). Currently, the political issue of uncertainty in the GHG estimations is dealt with only by considering is as a key driver for further improvements, but estimates reliability remain under concern. This negotiated provision seems so far apparently meaningless, as far as the GHG inventory improvements have driven by the datasets availability, while uncertainty was not stated yet as a major factor for it (e.g. only recently under KP reporting few MS reports that N2O emission from drainage methodologies are subject of improvement under high uncertainty).

Regarding the simulation protocol it seems that minimizing the number of calculations performed ensures that the overall uncertainty is more accurately estimated by the model [21]. Also depending on the model used, structural uncertainty may occur such as in case of performing sensitivity analysis on long chains of calculations, when @Risk is sampling inappropriate inputs for specified outputs when correlations are set, thus nonsensical results may arise.


18

4. Conclusions

EU 15 MS report 5A sink as the most uncertain contributor to national estimates. While at EU level the 5A sink is estimated by summing up of MS estimates, the assessment of the uncertainty has eventually to consider covariance given by non independent estimates (e.g. default factors), which may double it compared to independent inputs. Nevertheless, MS estimates of 5A sink are rather independent, under country specific datasets and methodologies. Level uncertainty at EU’s 5A1 sink varies 34-37 % upon the share of the contributing pools and their uncertainty. Under current level of uncertainty the confidence in 5A1 sink trend is relatively good, around 4,42 % with large interval under independent estimates. Time correlated estimates across time and MS apparently increase the confidence in estimated trend value, still independent uncertainty must be quantified in different moments in time upon underlying estimation circumstances (datasets, etc).

The uncertainty computed here is estimated only for CO2 in 5A1 without any gap filling of not estimated C stock changes in pools of some MS, but further improvement of a method for verification could be set up as to ensure full anchorage into the reality and real emission reduction. Analysis based on C stock change in pools could be taken as an example of verification of MS submissions on uncertainty in the future, relaying especially on non-UNFCCC reporting data. There is also a strong need to report disturbances, both as annual estimates and uncertainty, which would allow understanding of the risk related to annual sink performance.

Reference

[1] Nilsson S., Shvidenko A., Jonas M., McCallum I., Thomson A. and Balzter H. (2007):

Uncertainties of a regional terrestrial biota full carbon account: a system analysis, Water, Air, & Soil Pollution: Focus, Volume 7, Numbers 4-5 / September, 2007.

[2] Winiwarter W., Rypdal K. (2001): Assessing the uncertainty associated with national greenhouse gas emission inventories: a case study for Austria, Atmospheric Environment 35 (2001) 5425–5440.

[3] Jonas, M., White T., Marland G., Lieberman D., Nahorski Z. and Nilsson S.(2010): Dealing with uncertainty in GHG inventories. How to go about it? In: Coping with Uncertainty. Robust Solutions, K. Marti, Y. Ermoliev and M. Makowski (eds.) Berlin:Springer, 229-245.

[4] Heath L.S. and Smith J.E. (2000): An assessment of uncertainty in forest carbon budget projections, Environmental Science & Policy 3 (2000) 73-82.

[5] Gillenwater M., Sussman F. and Cohen J.(2004): Practical applications of uncertainty analysis for National greenhouse gas inventories. GHG Uncertainty Workshop - Warsaw. pp. 14.

[6] Frey C.H. and Burmaster D.E. (1999): Methods for Characterizing Variability and Uncertainty: Comparison of Bootstrap Simulation and Likelihood-Based Approaches, Risk Analysis, Volume 19, Issue 1, p 109-130.


19

[7] Rypdal K. and Winiwarter W. (2001): Uncertainties in greenhouse gas emission inventories evaluation, comparability and implications, Environmental Science & Policy 4 (2001) 107–116.

[8] Ramırez A., de Keizer C., Van der Sluijs J.P., Olivier J. and Brandes L. (2008): Monte Carlo analysis of uncertainties in the Netherlands greenhouse gas emission inventory for 1990–2004, Atmospheric Environment 42 (2008) 8263–8272.

[9] European Environmental Agency (2010): Annual European Union greenhouse gas inventory 1990 – 2008 and inventory report 2010.

[10] Monni S. (2005): Estimation of Country Contributions to the Climate Change. Viewpoints of Radiative Forcing and Uncertainty of Emissions, Dissertation for the degree of Doctor of Science in Technology, VTT Publications 577.

[11] Intergovernmental Panel on Climate Change (2003): Good Practice Guidance for Land Use, land-Use Change and Forestry. Penman, J., Gytarsky, M., Hiraishi, T., Kruger, D., Pipatti, R., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. and Wagner, F. (Eds). Intergovernmental Panel on Climate Change (IPCC), IPCC/IGES, Hayama, Japan.

[12] Blujdea V., Bird D.N. and Robledo C. (2009): Consistency and comparability of estimation and accounting of removal by sinks in afforestation/reforestation activities, Mitig Adapt Strateg Glob Change (2010) 15:1–18.

[13] Morgan M. G. and Henrion M.(1990): Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis, Cambridge University Press, New York, 1990.

[14] Cullen A.C. and Frey H.C.(1999): Probabilistic techniques in Exposure Assessment, A handbook for dealing with variability and uncertainty in models and inputs, New York Plenum press.

[15] Intergovernmental Panel on Climate Change (2000): IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories, Intergovernmental Panel on Climate Change (IPCC), IPCC/IGES, Hayama, Japan

[16] Grassi G.(2010): JRC LULUCF tool (personal communications).

[17] Isukapalli S.S.(1999): Uncertainty Analysis of Transport-Transformation Models -A dissertation submitted to the Graduate School-New Brunswick Rutgers, The State University of New Jersey, available at: http://www.ccl.rutgers.edu/~ssi/thesis/thesis-node1.html (last visited July 2010).

[18] Lehtonen A., Cienciala E., Tatarinov F. and Makipaa R.(2007): Uncertainty estimation of biomass expansion factors for Norway spruce in the Czech Republic. Annals of Forest Science 64:133-140.

[19] Tomppo E., Nilsson M., Rosengren M., Aalto P., and Kennedy P.(2002): Simultaneous use of Landsat-TM and IRS-1c WiFS data in estimating large area tree stem volume and aboveground biomass. Remote Sensing of Environment, 82, 156−171.


20

[20] Van den Wyngaert I., Brus D., Walvoort D. and Nabuurs G.-J.(2010): Eurogrid Aggregation of Plot Data of Forest Monitoring Schemes, in Harmonized Methods for Assessing Carbon Sequestration in European Forests Eds E. Cienciala, G. Seufert, V. Blujdea, G. Grassi, Z. Exnerová, JRC Scientific and Technical Report (in press).

[21] United Kingdom (2010) UK Greenhouse Gas Inventory 1990 to 2008: Annual Report for submission under the Framework Convention on Climate Change, AEA Technology plc.


21

Uncertainties in national inventory emissions of methane from

landfills: A UK case study

Keith A. Brown, Joanna MacCarthy, John D. Watterson, Jenny Thomas

AEA, Gemini Building, Fermi Avenue, Didcot, Oxon OX11 0RA, United Kingdom [email protected]

Abstract

Landfill gas is the largest source of man-made methane emissions in the UK, although contributions from this source have fallen significantly since 1990. This has come about through improvements in landfill management practices implemented as a result of the Landfill Directive, and also through the reduction in the amount of biodegradable waste being disposed of in landfills. An important source of information on methane production in landfills has been the data provided by operators of landfill gas energy schemes, where the gas is used as a fuel, predominantly for generating electricity. However, there is much greater uncertainty about the amount of gas that is collected for flaring (there is no obligation on site operators to provide flare stack duty to the competent authorities) and the extent to which the remaining methane may be oxidised in the surface layers of the landfill before reaching the atmosphere. The paper will review these and other principal sources of uncertainty in this aspect of the UK’s emission inventory and will discuss the results of recent meteorological and statistical studies of landfills to reduce this uncertainty.

Introduction

Methane is the second most important greenhouse gas after carbon dioxide and its emission from solid waste disposal sites is one of the largest global sources of anthropogenic methane emissions. According to the UK greenhouse gas inventory (GHGI), methane currently contributes about 7.7 per cent (or 48.5 Mt CO2 equivalent) of the country’s total annual emission of approximately 628.2 Mt CO2 equivalent [1]. Waste disposed to land (i.e. in landfills) is the largest source of UK methane emissions, accounting for about 41 per cent of methane emissions, just ahead of the next largest source, agriculture, with about 38 per cent. The UK’s legally binding target as part of the EU burden sharing agreement, under the Kyoto Protocol (KP), is to reduce emissions in the basket of six greenhouse gases to 12.5% below 1990 levels over the first commitment period 2008-2012. In order to monitor progress and to inform the need for corrective actions, Parties to the United Nations Framework Convention on Climate Change (UNFCCC) and the KP are obliged to maintain national inventories of greenhouse gas emissions and to report the findings annually using a standardised format to the, using methodologies developed by the IPCC and adopted by the UNFCCC.

Methane is formed as organic materials decompose under the oxygen limited (anaerobic) conditions of waste disposal sites. Freshly deposited wastes are decomposed by a wide range of microorganisms that convert complex biopolymers such as carbohydrates and proteins into smaller molecules, some of which are oxidised to carbon dioxide, in the process using up the oxygen entrained from the atmosphere. In large deposits of waste, this aerobic decomposition step occurs fast enough to consume the oxygen entrained in the waste faster than it can be replenished from the atmosphere. Under the resulting oxygen-free conditions, specialised anaerobic methanogenic bacteria convert the organic products of the initial decomposition processes into a mixture of


22

methane and carbon dioxide. Water is also formed that gives rise to the characteristic dark coloured liquor (leachate) produced in waste disposal sites and which can pollute surface and ground water if not properly controlled. The processes of methane formation in waste disposal sites are essentially similar to those that form methane in the guts of ruminant, termites and in waterlogged soils.

The UK’s approach to reporting methane emissions from landfills uses a model based on a methodology developed for this purpose by the Intergovernmental Panel on Climate Change (IPCC), as described below. Countries need to ensure that inventory reports are as accurate as possible and are prepared on a consistent basis and where national emission and activity data are used, that these are based on rigorous science. The UNFCCC annually reviews national inventories to ensure that they comply with these requirements.

This paper provides an overview of some of the uncertainties in estimating methane emissions from landfills and compares emissions calculated by the GHG inventory methodology with two alternative methodologies.

Estimating methane emissions from landfills

Calculating methane emissions on a national basis poses a different set of problems to estimating emissions from an individual landfill or landfill cell, where techniques such as flux boxes and long-path spectrographic techniques have been successfully applied (for example, see [2] and [3]). The difficulties stem from the very large range of variability of methane fluxes from landfills, both spatially (fluxes from similar areas differ markedly within a few metres) and temporally (fluxes vary – and can even change sign – diurnally, seasonally and with changes in barometric pressure) so that emission forecasting based on measurement alone would require an extensive site-based measurement programme and hence would be extremely resource-intensive. The IPCC does not recommend that countries use direct measurements to inform their inventory estimates, but rather to use the measurements to deduce national level parameter values that can then be used in a modelling-based approach.

Estimating national methane emissions from landfills - IPCC modelling approach

The IPCC has developed Guidelines for countries to help them choose the most appropriate method for reporting their greenhouse gas emissions, including methane from landfills. Emissions are related to both a set of activity factors and emission factors. Activity factors are in this case the quantity of the various sorts of organic waste that are disposed of in landfills each year, whilst emission factors determine how much of the degradable organic carbon is actually converted to methane and released to the atmosphere. The Guidelines allow countries to choose a method most appropriate to the level of detailed information they have on both activity and emission factors.

The UK GHGI is based on Tier 3 methodology described in the 2006 edition of the IPCC Guidelines [4], using the IPCC First Order Decay (FOD) based methodology with country-specific parameters and activity data. Equation (1), below, defines the overall approach for calculating methane emission from landfill as the amount of methane generated in the waste, minus the amount of methane recovered (for flaring or other combustion process), correcting for the amount of remaining methane that is oxidised to carbon dioxide. This is represented by equation 3.1 in the 2006 Guidelines:

CH4 emissions = [ Σx CH4 generatedx, T – RT ] (1-OXT), (1)


23

where: CH4 emissions = methane emitted in year T, Gg; T = inventory year; x = waste category or type of material; RT = recovered CH4 in year T, Gg; OXT = oxidation factor in year T (fraction).

Note that only the methane remaining after subtraction of methane recovered is available for oxidation. The FOD methodology is described in detail in the IPCC Guidelines. Key variables used are:

• The mass of dissimilable1 degradable organic carbon (DDOC) that is sent to SWDs each year;

• The fraction of DDOC that is converted to methane, as opposed to carbon dioxide;

• The first-order decay rate constant (k) of DDOC (units: year-1). Typically waste is considered in several different categories with different characteristic decay (i.e. k) rates. The total methane generation is then calculated by summing the methane generated in each category;

• The Methane Correction Factor (MCF). This is used to adjust the methane generated for the type of SWDs. In dumpsites and informal disposal facilities where waste is deposited without compaction a large proportion decays aerobically, or may be burnt on-site, before undergoing anaerobic decomposition after burial in later additions of waste. The IPCC Guidelines proposes the use of MCF ranging from 0.4 for shallow unmanaged dumpsites to 1.0 for modern landfills with compacted waste. All UK landfills are now considered to have a MCF equal to 1.0.

Details of the parameters and variables used are described in the GHGI report [1], along with current estimates of emissions since 1990, which are shown In Figure 1. The results indicate that methane emission from landfills has decreased to about 40% of the levels in 1990. This reduction is largely attributed to the improvement in landfill management and extensive implementation of gas collection, and to a reduction in the amount of biodegradable wastes being landfilled.

Sources of uncertainty

As shown in equation 1, above, there are three sources of uncertainty in the calculation of methane emitted from landfills: that associated with the methane generated; the amount recovered and the fraction oxidised. In addition, there are further uncertainties associated with the parameters used to calculate methane generation, such as rates of decomposition and DDOC levels. Overall, the level of uncertainty in waste composition and quantities going to landfill is believed to have reduced since 1990 with the availability of better quality data, particularly for municipal waste, although scope for further improvement remains, particularly in relation to commercial and industrial wastes. The GHGI reports and overall level of uncertainty in landfill gas emission factor of 48%, based on Monte Carlo simulation [5]. In the following sections we present a qualitative discussion of uncertainties in methane recovery and oxidation.

1 I.e. organic carbon mineralised to methane and carbon dioxide.


24

0

500

1,000

1,500

2,000

2,500

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Methane emissions, Gg/year

Year

Figure 1. Methane emissions (Gg/year) from Solid Waste Disposal Sites in UK [1]

Methane recovery

Landfill gas is actively pumped from modern landfills via gas wells sunk into the waste. The collected gas is then either used or flared. Gas recovery is required to prevent uncontrolled escape of gas that may result in odour nuisance and even fires and explosions, as well as for mitigating greenhouse gas impacts. Although there are many potential uses for the methane in the landfill gas, such as use as a heating or vehicle fuel, or, following removal of contaminants, for injection into the natural gas mains, the predominant use is for on-site electricity generation. In addition, some gas will be flared to prevent its migration to neighbouring property.

There is a high level of confidence in the amount of methane used for power generation. This is because electricity from landfill gas is eligible for financial support mechanisms to encourage renewable energy resources. To receive the financial benefits, landfill gas energy operators have to report the energy generated to government, and hence it is easy to calculate the equivalent amount of methane recovered, given that the conversion efficiency is well-characterised.

There is much less confidence in the amount of methane flared. This is because landfill site operators (at least in the UK) have no obligation to report volumes of methane flared. Information on this factor is obtained from operator surveys and from sales data from flare suppliers (where they are willing to provide it), but this usually does not allow a distinction to be made between new flares that are additional to existing plant, or purchased as replacements for old equipment. Furthermore, data on flare capacity only provides an indication of the maximum amount of methane that could be flared. Given that many flares are used to provide back-up for when gas engines are out of commission for servicing, as well as those used to combust methane produced in excess of engine capacity, the degree of uncertainty is high.

A further complication in estimating recovery rates comes from the time-phasing of methane generation. Typically in moist temperate climates, modern landfills begin


25

forming methane within a year of closure of the operational area or cell, as aerobic decay processes use up oxygen entrained with the waste and the anaerobic conditions needed for methanogenesis become established. There follows a period of between about 3 and 12 years when methane generation peaks, followed by a steady decline, but with some residual methane production occurring decades after waste emplacements. A typical time course is illustrated in Figure 2.

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49

Years after waste landfilled

A

B C

Figure 2. Typical time pattern of methane generation in landfills (for illustration only)

The balance between alternative ways of controlling emissions varies over

the landfill’s life. During the period when maximum amounts of methane are being produced, actively pumped gas collection networks are in use (shown by phase B in the figure), mostly feeding the landfill gas to engines for electricity generation (during phase A), but increasingly supplying flares as methane formation declines.

As well as methane generation varying over time, the composition of the landfill gas also changes. Methane and carbon dioxide are produced in roughly equimolar amounts during methanogenesis, but as a result of the lower aqueous solubility of methane, the concentrations of methane in the landfill gas during peak generation is about 55-60% methane by volume. This declines steadily as methane formation tails off. When the methane concentrations falls below about 25-30% by volume, the gas can no longer be burnt economically in engines. When the methane concentration has declined below about 17% by volume, it can no longer be flared without the use of a pilot fuel. Before this stage is reached, most landfill sites will have reduced the active gas extraction system and some will rely increasingly on biofilter covers for removing the remaining methane through microbial oxidation (shown by phase C in the figure), as further discussed in the next section.

The mix of emission control achieved through the use of gas engines and flares attached to active gas extraction networks will therefore change during the life of a landfill. Site operators will make maximum use of gas engines during the 9 or 10 years of maximum methane generation, with flares for backup and for dealing with excess gas.


26

As gas production declines over time, gas engine capacity is increasingly replaced with flaring and a reduction in active extraction rates. Finally when risks from local impacts of landfill gas escape are judged by the environmental regulator to be minimal, active gas extraction is turned off.

The UK GHGI model allocates waste to four types of landfill, characterised in terms of gas collection efficiency, reflecting the higher standards of gas control, site design and operation introduced to meet the needs of the 1993 Landfill Directive. The model currently allocates all biodegradable landfilled waste in the UK to modern landfills with efficient gas collection systems, although waste landfilled in previous years is allocated to sites having no or much lower levels of gas control. The collection efficiency adopted for modern EU-compliant landfills is 75% of methane formed over the life of the site. The value reflects discussions with industry and includes much higher rates of collection (often >90%) claimed by site operators during the period of peak methane formation (i.e. during phase A, in Figure 2). This high rate is consistent with a published study of methane collection rates [6]. Taking account of emissions from waste already landfilled in old sites with little or no gas extraction reduces the average collection efficiency to about 70% in 2007. About 32% of this is used for power generation and the remaining 38% is attributed to flaring, although as discussed above, the uncertainties in this latter figure are large. Given that overall about 70% of methane formed during the life of UK landfills is recovered, the remaining 30% is available for emission into the atmosphere, after taking account of microbial oxidation in the surface layers of the landfills or in the restoration cover material.

Methane oxidation

As indicated in equation (1), the FOD model applies the methane oxidation fraction to the methane remaining after taking account of recovery. The IPCC Guidelines provide a default value of zero for methane oxidation in shallow dumpsites, with MCF of <1, but allows the use of a factor of 0.1 (i.e. equivalent to 10% of the non-recovered methane) in managed landfills, unless the Party can substantiate the use of a higher figure. Examination of the 2009 NIR submissions to the UN FCCC by 17 western European countries2 revealed that all except one adopted the 0.1 factor for methane oxidation (the exception being Greece which used the zero default applicable to dumpsites).

Methane oxidising bacteria (methanotrophs) are widely distributed in soils, using the oxidation of methane to carbon dioxide, with oxygen as the terminal electron acceptor, as a source of energy. In addition to a supply of methane, they also require aerobic conditions, as well as appropriate levels of moisture, absence of inhibitory agents and non-extreme temperatures and pH. Whilst the presence of methanotrophs can be shown in most soils and their activity demonstrated in laboratory experiments, it has been much harder to establish their role in reducing methane emissions under landfill conditions. Furthermore, their potential importance is expected to change during the life of a landfill.

The extent to which methane oxidation may reduce methane emissions depends on the methanotrophs having sufficient time in which to oxidise the methane. Conditions expected to maximise the opportunity for this interaction would be low flow rates of landfill gas through the porous medium in the landfill surface. High flow rates of landfill 2 The countries were Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Norway, Portugal, Spain, Sweden, Switzerland and United Kingdom.


27

gas which may occur through cracks and fissures in the landfill cap will not allow sufficient opportunity for methanotrophs living in surface moisture on soil particles to access the substrate. In addition, high gas flow rates will tend to force out air from the surface layer and so deprive the methanotrophs of oxygen, which is a limiting factor for methane oxidation. Therefore in the period of maximum methane formation with efficient gas collection (as in phase B in Figure 2), oxidation of the relatively small amount of methane not removed by the extraction system may proceed reasonably efficiently. However, if the gas extraction system is not effective, high rates of mass flow of landfill gas may displace oxygen from the surface layer, so reducing oxidation rates in the bulk of the surface. If the surface is cracked, much of the methane would be expected to escape without exposure to methanotrophs.

In older landfills after the peak period of methane formation is passed (as in phase C in Figure 2), there may be much greater potential for methanogens to control emissions. The mass flow of gas out of the waste will be lower, so reducing the displacement of oxygen from the surface layers.

There appears to be reasonable evidence to indicate that the 0.1 factor for methane oxidation stated by the IPCC may be excessively conservative. Recent studies have indicated that this mechanism can consume considerably more than this, reporting oxidation levels of 14-25% [7], 57-98% [8], 25-46% [9], 4-50% [4], and >96% [10], respectively. However, sustained high rates of oxidation appear to depend on temperature between 5-40oC and moisture content between field capacity and wilting point3 [11].

Thus it appears that there is considerable uncertainty in the rate of methane oxidation, although the importance of this mechanism to reduce emissions during periods of maximum methane formation without effective gas control appears to be limited. Further work is needed to determine the scope for reducing uncertainty in this area.

Comparison of GHGI emissions with other studies

In this section, we compare the GHGI emissions with two other different assessments. In the first of these other studies, undertaken by Jacobs Engineering [12] for the Environment Agency (the UK’s principal regulator for landfills), database of the key factors influencing methane emissions has been constructed from available information for 23,000 landfill sites in England and Wales. 65% of the sites were judged to have no methane emission potential while 20% of the sites have low emission potentials. The high and very high emissions categories account for 2.2% of the sites and 53% of the annual total emissions. Methane generation rates were estimated using the Environment Agency’s GasSym model, developed for site assessment studies. The annual methane emissions of landfills in England and Wales in 2007 were estimated to be 732 ± 253 (95% CI) Gg per year. Scaling up to the whole of the UK by population (England and Wales account for ~88% of the UK population) gives an equivalent emission rate of 831 ± 287 Gg per year. The GHGI emission for 2007 (963 Gg) is well within these confidence limits (compare the Landfills line with Jacobs point (+/- 95% CI) shown in Figure 3.

3 “Field capacity” is the amount of water held by a soil when all the water that can drain away because of gravity has done so; “wilting point” is soil moisture content below which plants wilt.


28

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Methane emissions, Gg/year

YearOther sectors Landfills

NAME using ECMWF met Inventory Uncertainties

NAME2 Estimate Jacobs landfill estimate

Figure 3. Methane emission estimates comparison

Figure 3 also shows the total anthropogenic emissions of methane from landfills and other man-made sources (“Other sectors” bars) reported in the UK GHGI [1], which demonstrates the overall reduction in methane emissions achieved since 1990, along with reductions from landfills. The 95% confidence intervals on the total inventory methane emissions are shown by the vertical bars for years 1990 and 2008 only.

In order to provide some verification of the UK GHGI a high-quality observation station at Mace Head on the west coast of Ireland has been established (see Annex 11 of the GHGI report [1]. The station reports high-frequency concentrations of the key greenhouse gases. A Lagrangian dispersion model NAME (Numerical Atmospheric dispersion Modelling Environment driven by 3D synoptic meteorology is used to generate so called air-history maps. The air-history maps represent the recent 10-day history of the air before it arrives at the observing station, Mace Head, and estimate the dilution in concentration that surface sources would undergo during this transport. These maps have been generated for each 3-hour period from 1990 and enable the observations made at Mace Head to be sorted into those that represent Northern Hemisphere baseline air masses and those that represent regionally-polluted air masses arriving from Europe. From the sorted data an estimate of the time-varying Northern Hemisphere mid-latitude baseline concentration of key greenhouse gases, including methane, is made. The results of the NAME modelling (+/- 95% CI) are compared with the GHGI results for total anthropogenic emissions of methane from all source, in Figure 3, after eliminating the impact of potential natural sources of methane from peat bogs in the vicinity of the measuring station.

The GHGI trend is monotonically downwards whereas the NAME estimates show no clear trend, although the agreement from 2001 onwards is reasonably good. It must be remembered however that the GHGI totals only include anthropogenic emissions whereas


29

the NAME estimates are total emissions combining both anthropogenic and biogenic releases, however biogenic emissions in North Western Europe are thought to be low. Furthermore, the NAME methodology requires uniformity of emissions in time and space, and assumption that is at least questionable for methane emissions. Further work is underway to determine the cause of the differences in trends between the NAME and the GHGI methodology. The UNFCCC Expert Review Team has closely examined the UK’s estimates of landfill methane emissions reported in the GHGI and no adjustments were required.

For 2006 the NAME-inversion method has been applied using data from 11 stations, including Mace Head, across Europe as part of the FP6 European project NitroEurope. The agreement between the Mace Head only results and the GHGI estimates are excellent for this year (see “NAME 2 estimate” point in the graph).

Conclusions

Emissions of methane from UK landfills in 2007 as reported in the GHGI are in good agreement with those obtained from a statistical assessment of emissions from a large number of landfills in England and Wales. However, there still remain areas of uncertainty, particularly in the quantification of methane recovered and flared and the extent of methane oxidation in the surface layers. Overall, emissions from landfills are estimated to have fallen to about 40% of 1990 levels as a result of improvements in landfill gas management and a decrease in the amount of biodegradable waste sent to landfill, as a consequence of improvements brought about through the 1993 Landfill Directive.

Emissions of total anthropogenic methane reported in the GHGI are in good agreement with total (anthropogenic and biogenic sources) calculated from atmospheric concentrations and dispersion modelling over the past few years, assuming biogenic emissions to be a small component of the total. However, it is not yet clear why the measurement and dispersion modelling does not reflect the downward trend in emissions reported in the inventory since 1990, of which the major decrease is associated with reductions in methane emissions from landfill. Further investigation will be needed to explain these differences. Disclaimer and acknowledgement

The views expressed are those of the authors and are not necessarily those of UK government departments, AEA or any other body. The authors gratefully acknowledge financial support from the UK government (Department of Energy and Climate Change). AEA developed and operates the UK’s greenhouse inventory, part of the National Atmospheric Emissions Inventory, on behalf of the government since its inception in the 1980s. The greenhouse gas inventory is the mechanism through which UK reports its emissions in compliance with its obligations under the United Nations Framework Convention on Climate Change and the European Union Monitoring Mechanism.

Reference

[1] UK Greenhouse Gas Inventory, 1990 to 2008. Annual Report for Submission under the Framework Convention on Climate Change. ISBN 0-9554823-9-9, April 2010 http://unfccc.int/national_reports/annex_i_ghg_inventories/national_inventories_submissions/items/5270.php


30

[2] Ullas Hegde et al, Methane and carbon dioxide emissions from Shan-Chu-Ku landfill site in northern Taiwan. Chemosphere, 52 (2003), 1275-1285.

[3] Ito, A. et al, The long-term evolutions and the regional characteristics of atmospheric methane concentrations in Nagoya, 1983]1997. The Science of the Total Environment 263 (2000) 27-45.

[4] Intergovernmental Panel on Climate Change (IPCC) Guidelines for National Greenhouse Gas Inventories (2006). Chapter 3 - Solid Waste Disposal. http://www.ipcc-nggip.iges.or.jp/. The 2006 Guidelines have yet to be adopted by Parties for reporting to the UNFCCC but can be used as scientific information in conjunction with the previous Guidance in Revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories Reference Manual Volume 3. http://www.ipcc-nggip.iges.or.jp/public/gl/invs6e.html and Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories 2000. Chapter 5 – Waste. http://www.ipcc-nggip.iges.or.jp/public/gp/english/5_Waste.pdf.

[5] Brown, KA, et al (1999) Methane Emissions from UK Landfills, AEA Technology, AEAT-5217, Culham.

[6] Spokas, K. et al. Methane mass-balance at three landfill sites: What is the efficiency of capture by gas collection systems? Waste Management 26 (2006) 516-525.

[7] Abichou, T. et al, Methane flux and oxidation at two types of intermediate landfill covers. Waste Management 26 (2006) 1305-1312.

[8] Berger, J. et al, Methane oxidation in a landfill cover with capillary barrier. Waste Management 25 (2005) 369-373.

[9] Einola, J. et al Methane oxidation at a surface-sealed boreal landfill. Waste Management 29 (2009) 2105-2120.

[10] Einola, J. et al Methane oxidation in a boreal climate in an experimental landfill cover composed from mechanically-biologically treated waste. Science of the Total Environment 407 (2008) 67-83.

[11] Spokas, K and Bogner, J. Limits and dynamics of methane oxidation in landfill cover soils. Waste Management (2010) in press.

[12] Jacobs Engineering. Final project Report to the Environment Agency – Categories of Landfills for Methane Emissions. February 2010 in press.


31

Comparison between energy and emission data reported under the ETS and energy balance and greenhouse gas

inventory of Flanders

Dieter Cuypers1, Tom Dauwe1, Kristien Aernouts1, Ils Moorkens1, Johan Brouwers2

1 VITO – Transition, Energy and Environment Boeretang 200, Mol, B-2400, Belgium

[email protected]

2 Flemish Environment Agency Van Benedenlaan 34, Mechelen, B-2800, Belgium

Abstract

Belgium is a federal state where environmental issues fall under the responsibility of the regions. In Flanders, data from the ETS are not used in the GHG inventories. In this paper, we compared the energy and emissions included in the ETS with the total energy consumption and CO2 emissions for the different industrial sectors. During the first period of the ETS (2005-2007), the average share of CO2 emissions was 45 % of total CO2 emissions in Flanders. In 2008 the share even rose to 50%, due to an expanded scope of the ETS in the second trading period. We also examined the differences between the data reported under the ETS and the data used in energy balances and GHG inventories. This analysis shows differences in the calculated emissions in some industry sectors due to different approaches and definitions. Also, in sectors where derived fuels are used, ETS data and energy balances are not easily comparable. We conclude that the involvement of different authorities for GHG inventories and ETS necessitates consultation and coordination among all actors. Further harmonization between the two data sets is needed in light of the division between ETS and non-ETS policies in the EU Climate and Energy Package.

Keywords: Industry, energy, CO2 emissions, ETS, energy balance, non-ETS

1. Introduction

The European Union has played an important and active role in the international climate change debate. To reduce greenhouse gas emissions, numerous policies and measures have been installed at EU, member state, regional and local level. The flagship of the EU climate policy is the EU emission trading scheme (ETS, Directive 2003/87/EC). The ETS is the largest cap-and-trade system with approximately 12.000 installations in 2008 in 27 member states and 3 non-EU countries (i.e. Norway, Iceland and Liechtenstein). It covers almost all large point-source pollution sources of CO2. In the first (2005-2007) and second (2008-2012) trading period of the ETS, installations were allocated emission allowances by the member states to offset their greenhouse gas emissions. The possibility to trade emission allowances will guarantee that emission reductions are achieved at the lowest cost. The types of activities that fall under the ETS are listed in Directive 2003/87/EC. Member states however have an option to either include (opt-in) or exclude (opt-out) certain activities. During the first trading period it became clear that member states interpreted the activity “combustion installations with


32

a thermal input exceeding 20 MW” differently. For the second trading period, the European Commission clarified which activities (in the so-called aggregation rule) fall under the ETS.

The clear division between ETS and non-ETS emissions in current and especially future climate policies (i.e. in the climate and energy package) is also apparent in the EU energy policy. The Energy Services Directive (2006/32/EC) sets a non-mandatory objective to member states to improve energy efficiency with 1% annually between 2008-2016. This directive only applies to the non-ETS sector. The division between ETS and non-ETS does not necessarily correspond to the classic division in sectors and subsectors (e.g. households, industry and services). Therefore there is a need for reliable statistics on ETS and non-ETS energy consumption and related CO2 emissions for monitoring and reporting purposes. In this study we have compared the energy consumption and energetic CO2 emissions of the energy and industrial sector with the energy consumption and CO2 emissions of the ETS installation in Flanders, Belgium in 2005-2008. For reasons of confidentiality information at installation level are aggregated to subsector level.

Belgium is a federal country, where competences are divided between the federal and the regional level. For most climate and energy related competences, regional governments (Flanders, Walloon region and Brussels Capital) are for a large extent responsible for implementing policies and monitoring and reporting obligations. In Flanders, ETS installations have to monitor, report and verify their emissions to the Department of Environment, Nature and Energy (Dept. LNE), whereas the energy consumption and greenhouse gas emissions are monitored by the Flemish Environment Agency.

2. Methodology

We were interested in a few comparisons related to the differences between the ETS sector and the installations and sectors not covered by the ETS Directive. More specifically, we wanted to investigate :

• The differences between ETS and non-ETS in sectors and subsectors that are subject to the ETS Directive (for installations that meet the criteria for inclusion under the ETS); and

• The differences between ETS and non-ETS in the complete sector (which includes all sectors and subsectors, irrespective of whether installations are included under the ETS).

Both comparisons were made for:

• Energy consumption; • Energetic emissions of CO2; • Process related CO2 emissions; and • The sum of all CO2 emissions.

2.1. Energy consumption in Flanders

For energy statistics the Flanders uses the energy statistics manual developed by the IEA [1] with some specific adaptations because of the limited availability of regional


33

energy statistics. As energy import data are only available at the Federal Belgian level for fossil fuels Flemish energy import is deducted from the Gross Inland Consumption, production and distribution losses and international bunker fuels. For biomass, derived fuels (i.e. byproducts from industrial processes which can be reused as fuel) and heat net import is considered to be equal to 0. With the increasing importance of imported biomass and for consistency reasons this assumption will be adjusted in the future.

The representation of the energy statistics derived is called the energy balance. An energy balance represents in a consistent way all streams and stocks per fuel product (energy carrier) in energy units. All stocks, production, imports, exports and consumption are compiled from existing statistics and inquiries and reporting by companies and sector organisations. The energy balance data allow for a division and comparison of energy consumption between the different sectors of an economy. The format of the energy balance for Flanders follows the revised IPCC Guidelines from 1996 [1, 2]. According to these guidelines the energy consumption of autoproducers of electricity is reported in the sector these autoproducers belong to whereas in the past they were treated as a specific category in the Energy sector.

To compare ETS installation energy consumption with the energy consumption of the sectors they belong to, we first extracted the concerned sectors from the energy balance: the Energy sector and its subsectors Natural Gas, Electricity & Heat Production, and Petroleum Refineries; the Industry sector and its subsectors Metal, Chemical, Food, Paper, Textile, and Other Industries; and Commerce & Services. All of the extracted sectors and subsectors could be affected by the ETS Directive, other sectors were not extracted. The sector terminology used is the Flanders Environmental Reporting (MIRA) sector division. To be able to compare energy consumption data, autoproducer energy consumption had to be subtracted from the (sub)sectors and counted towards the considered subsector in the Energy sector as this is the way emissions from energy consumption are treated in the ETS. Per (sub)sector the energy consumption per energy carrier was compared to the energy consumption of the ETS installations belonging to that sector. Some specific energy carriers had to be transferred from one sector to another when sectors use each other’s derived fuels.

Companies with ETS installations are required to report their energy consumption annually as imposed by the Directives 2003/87/EC and 2007/589/EC, based on a verified monitoring plan. These are the ETS data which are freely available on the Dept. LNE website [3].

Energy balance and ETS reporting is done annually which allows for a comparison for the years 2005-2008 as ETS reporting started in 2005. For the comparisons only fossil and derived fuels are considered relevant for this study as they are the ones regulated by a cap-and-trade mechanism. Comparisons made with total energy consumption are comparisons made with total energy consumption of fossil and derived fuels; no renewable or nuclear energy is considered.

2.2. Greenhouse Gas Emissions in Flanders

The energy balance statistics are used by the Flemish Environment Agency to construct an Emission inventory by multiplying energy consumption by emission factors per fuel to calculate energetic emissions. Additional data from inquiries and reporting


34

from companies and sector federations are used for verification. Process emissions are also included in this inventory and are calculated from the mandatory emission reporting by companies and inquiries.

The comparison is done in the same way as for energy consumption. The extracted (sub)sector energetic emissions were calculated from the multiplication of the energy consumption per energy carrier and the emission factor for each energy carrier.

3. Results

3.1. Energy consumption

As the energy consumption of fossil fuels and derived fuels constitute the major source of CO2 emissions by the evaluated sectors, an analysis on these energy consumption data can be seen as a first step for a consistent analysis of the CO2 emissions later on. In Table 1 the results are given for each sector and subsector with the relative part of the ETS installations per (sub)sector expressed in %.

Table 1. Energy Consumption (PJ) per sector for the period 2005-2008 split-up between ETS and non-ETS installations, % expressed as ETS part of total energy consumption of sector

Energy use (PJ)

SECTOR or subsector ETS

non-ETS % ETS

non-ETS % ETS

non-ETS % ETS

non-ETS %

ENERGY 291,5 12,9 96% 280,1 17,9 94% 293,9 11,9 96% 280,3 14,695%

Natural Gas 0,0 2,0 0% 0,0 2,4 0% 0,0 1,9 0% 1,2 0,7 63%Electricity &

Heat 225,9 10,9 95% 211,3 15,6 93% 221,4 9,9 96% 206,4 13,9 94%Petroleum Refineries 65,6 - 100% 68,8 - 100% 72,5 - 100% 72,6 - 100%

INDUSTRY 102,4 187,7 35% 120,1 168,2 42% 111,4 158,2 41% 180,4 87,4 67%

Metal 31,0 71,3 30% 32,7 68,5 32% 31,0 58,1 35% 33,8 50,3 40%Chemical 37,5 85,8 30% 44,7 78,4 36% 44,7 74,6 37% 111,1 10,6 91%

Food 12,5 11,2 53% 12,6 9,3 57% 12,9 8,2 61% 12,9 9,6 58%Textile 1,9 4,7 28% 2,2 4,4 33% 2,0 4,0 33% 1,2 3,3 27%Paper 3,2 1,6 67% 3,4 0,9 79% 3,1 1,2 72% 2,9 1,3 69%Other 16,3 13,0 56% 24,5 6,6 79% 17,8 12,1 60% 18,5 12,4 60%

COMMERCE & SERVICES 0,0 33,2 0% 0,0 31,6 0% 0,0 30,9 0% 0,4 37,8 1%

Phase I2005 2006 2008

Phase 22007

The share of ETS installations in the Energy sector in Flanders stays quite stable around 95% over both phases of the ETS. In the subsectors there is, however, an increased coverage of the Natural Gas sector from 0% in the first trading period to 63% in the second trading period. The minor share of this subsector towards the Energy sector explains the non-effect on the Energy sector ETS share as a whole.


35

The installations in the Natural Gas subsector are compressor stations. The increased share of ETS is due to a former opt-out of these installations in the first trading period.

The other two subsectors, Electricity & Heat and Petroleum Refineries stayed stable around 95% and 100% (complete coverage) respectively. The Petroleum Refineries sector proved to be a difficult exercise because of different methodologies used between the ETS reporting and data from the Energy Balance. This subsector was known to be covered completely by the ETS, though, ETS report data didn’t add up to 100% of energy balance data.

In the Flemish Industry sector the effect of the clarification on the interpretation of the 20 MW rule for the second phase clearly has its effect on the coverage of the sector by the ETS. In the first phase the ETS share lies around 39% increasing to 67% in the second phase (Table 1). This increase is mostly due to the increased coverage of the Chemical Industry subsector (infra) by the ETS.

In the Metal Industry subsector there is a slight increase of the ETS share from 2005 to 2008 (Table 1). The increase during the first phase is due to the lowered energy consumption of the non-ETS installations while the ETS installations energy consumption stays quite constant around 31,5 PJ. The decreased energy consumption of the non-ETS installations continues in 2008. This, combined with an increased energy consumption for the ETS sector results in a 40% coverage. Although only one more installation got covered by the ETS in 2008 while 6 were transferred to the non-ETS sector because of the 3 MW de-minimis rule this can be explained by the fact that one major installation influences the Flemish Metal ETS sector because it is responsible for 2/3 of the energy consumption in the subsector.

The Chemical Industry subsector is clearly affected by the clarification on the interpretation of the 20 MW rule for the second phase (Figure 1). While the total energy consumption of fossil and derived fuels remains quite constant throughout the periods, the share of the ETS rises slightly towards the end of the first phase to 37% to increase to 91% in 2008.

Figure 1. Evolution of the energy consumption (PJ) and share (%) of the ETS installations’ energy consumption in the Chemical Industry in Flanders

2005 2006 2007 2008

TOTAL (PJ) 123,3 123,1 119,3 121,6

non-ETS (PJ) 85,8 78,4 74,6 10,6

ETS (PJ) 37,5 44,7 44,7 111,1

share ETS 30% 36% 37% 91%

30% 36% 37%

91%

0

20

40

60

80

100

120

140

Ene

rgy

Co

nsum

ptio

n (P

J)


36

The Food, Textile, Paper and Other Industries represent only a minor share of all ETS installations in Flanders, the Energy sector and Metal and Chemical Industry sector make up about 90% of all energy consumption of the Flemish ETS sector (Figure 2).

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

2005 2006 2007 2008

Commerce & Services

Other

Paper

Textile

Food

Chemical

Metal

Petroleum Refineries

Electricity & Heat

Natural Gas

Figure 2. ETS energy consumption repartitioning among sectors and subsectors subject to the ETS Directive for the period 2005-2008

The Food Industry ETS share fluctuates slightly due to fluctuations in the non-ETS

installations’ energy consumption. Although four companies enter the ETS sector in 2008 this does not alter the energy consumption of the ETS sector. In the Textile Industry the ETS share falls from 33% in 2007 to 21% in 2008 due to the 3 MW de-minimis rule transferring four companies back to the non-ETS sector in the second phase. In the Paper Industry there is a small decrease in the ETS share due to the 3 MW de-minimis rule applying to one installation. The Other Industries, regrouping all industries which do not resort under any other major subsector, show fluctuations in the first phase which are bigger than the increase of the share from 63% in 2007 to 75% in 2008, which is due to the transfer of 6 installations to the ETS sector in the second phase.

The Commerce & Services sector constitutes a very small share of the total ETS in Flanders (Figure 2). Because of the transfer of three former opt-outs and one other installation the share of ETS energy consumption of the sector increases to 1%.

3.2. CO2 emissions

3.2.1. Energetic emissions

For energetic emissions the same comparison was made as for energy consumption. For emission factors (EF) the IPCC Guidance 1996 and default EF were used, other EF or data were used for: 1) Electricity: reported emissions from power generating facilities; 2) Petroleum Refineries: reported emissions from the companies’ reports; 3) Metal Industry: EF from ETS reports for specific fuels from big emitters; and 4) for derived fuels (most common in the Chemical Industry subsector) specific EF were used when available from company reporting. The results of this exercise are shown in Table 2. It is clear that the same explanations as given for energy consumption per (sub)sector apply here. In general evolutions and fluctuations have the same reasons because the first and second phase the


37

ETS Directive coverage is based on energetic thresholds and CO2 emissions only. By comparing Table 1 and Table 2, even while using company specific EF and reported emissions, inconsistencies appear.

Comparing energy consumption shares and energetic emission shares From a comparison it is clear that the biggest differences lie in the Metal Industry

sector; the ETS share of energy consumption lie between 45% and 49% lower than the share calculated for energetic emissions. The same anomalies were found in the Petroleum Refineries sector. However, this is not reflected in the tables as this subsector is known to be covered completely by the ETS and was treated as a 100% sector. Energetic data from ETS reporting were considerably lower than the total calculated from the energy balance while energetic emissions were somewhat higher than those reported for ETS. This can be explained by the fact that in these sectors a lot of emissions are treated as energetic emissions when reported for the ETS while for the energy balance they were treated as non energetic processes. In fact, although it renders this comparison difficult, the final result should be the same; no difference is made between energetic or process emissions to calculate ETS emissions.

Other inconsistencies are due to the use of: different calorific values; unknown calorific values for heterogeneous and uncommon (e.g. one-time use) fuel types; the use of different EF. Two major eventual differences remain to be solved for the Food (in 2006) and the Other Industries (in 2007) sectors.

Table 2. Energetic emissions (kton CO2) per sector for the period 2005-2008 split-up between ETS and non-ETS installations, % expressed as ETS part of total energetic emissions of sector

Energetic emissions (kton CO2)

SECTOR or subsector ETS

non-ETS % ETS

non-ETS % ETS

non-ETS % ETS

non-ETS %

ENERGY 21.625 884 96% 20.591 1.281 94% 20.508 1.375 94% 19.511 1.165 94%8.224Natural Gas 0 114 0% 0 132 0% 0 108 0% 68 41 63%

Electricity & Heat 18.381 900 95% 17.338 1.215 93% 17.185 1.335 93% 15.971 1.249 93%

Petroleum Refineries 3.244 0 100% 3.253 0 100% 3.323 0 100% 3.471 0 100%

INDUSTRY 7.685 7.072 52% 8.224 6.564 56% 7.854 6.168 56% 11.569 2.183 84%

Metal 3.039 848 78% 3.351 774 81% 3.016 742 80% 3.137 415 88%Chemical 2.487 4.391 36% 2.574 4.143 38% 2.432 4.043 38% 6.156 206 97%

Food 831 673 55% 810 564 59% 1.031 288 78% 852 524 62%Textile 104 271 28% 102 275 27% 95 249 28% 78 179 30%Paper 217 110 66% 236 70 77% 205 81 72% 211 73 74%Other 1.006 780 56% 1.152 738 61% 1.075 765 58% 1.135 786 59%

COMMERCE & SERVICES 0 1.985 0% 0 1.861 0% 0 1.814 0% 22 1.943 1%

Phase I Phase 22005 2006 2007 2008


38

3.2.2. Process emissions

Process emissions occur in Flanders only in some of the (sub)sectors studied. Table 3 gives an overview of these process emissions for these (sub)sectors.

It is clear from a first look at Table 3 that a lot of emissions are not reported in the emission inventory from the Flemish Environmental Agency. The process emissions in the Electricity & Heat sector are limestone emissions and are only reported via the ETS reporting. For the Petroleum Refineries the emission inventory does not report the energetic and process emissions separately as they are not calculated from the energy balance and process emission reporting, but directly from the company reports, which do not distinguish between both. For the Metal Industry the same figure is reported in the ETS as for the emission inventory. This is because the ETS reporting for this sector is already integrated in the emission inventory (a positive consequence of this study). The emissions reported for the Paper sector are from bio sludge which were reported by one company as process emissions while they were treated as energetic emissions for the emission inventory.

Table 3. Process emissions (kton CO2) per sector for the period 2005-2008 reported for ETS installations and for the emission inventory

n a: data not available

3.2.3. Total CO2 emissions

The evaluation of the coverage of the ETS of all CO2-emissions in Flanders, including all process and energetic emissions from other (sub)sectors reported in the emission inventory is visualised in Figure 3.

The clarification on the scope of the ETS increases the ETS share of all CO2 emissions in Flanders from 45% in the first phase to 50% in the second phase.

4. Discussion and conclusion

It can be clear from the above that a simple split-up between ETS and non-ETS is not that straightforward, especially when factoring in control mechanisms through the use

kton CO2

sector ETS Total ETS Total ETS Total ETS Total

Electricity & Heat 11 n a 13 n a 13 n a 18 n a

Petroleum refineries 1.213 n a 1.399 n a 1.373 n a 1.248 n aMetal 2.365 2.365 2.390 2.390 2.154 2.154 2.151 2.151

Chemical 6 2.460 7 2.091 8 2.099 8 1.577

Paper 12 n a 16 n a 10 n a 26 n aOther 227 230 230 236 231 237 236 237

2006 2007 20082005


39

of energy balances and the consecutive split-up between process and energetic emissions. Though these control mechanisms are necessary to provide a consistent analysis.

Figure 3. Evolution of the CO2 emissions (kton CO2) and share (%) of the ETS installations’ CO2 emissions of total CO2 emissions in Flanders

Problems encountered during this analysis are due to different typologies and methodologies used among the different reports.

First, although all CO2 emissions in the ETS are treated equal, the differentiation between process or energetic emissions is a difficulty for good comparisons.

Second, the use of generalised and specific emission factors and calorific values for derived fuels among companies but also between reports and inventories can create considerable differences when large fuel quantities are consumed.

Third, but of lesser importance, is the terminology of fuels: which ‘bio-like’ fuels are biofuels or which compartments of these fuels are considered bio and which not?

Fourth, different methodologies between ETS reporting and the energy balance and emission inventory for treating autoconsumers render comparisons difficult and can cause misunderstandings when evaluating sector wide data. The same applies to some fuels which are interchanged between sectors and are attributed to either one of the sectors involved depending on the framework.

To provide for a clear split-up between the ETS and non-ETS sector these inconsistencies have to be cleared out. On the other hand the comparisons provided us with data from ETS reports that, up to now, were not included in the energy balance or emission inventory. Unreported data in one inventory can be added to complement the other if the different scope is taken into account. At the same time double counting is to be avoided and this is only possible through the use of the same terminology.

We conclude that the involvement of different authorities for GHG inventories and ETS necessitates consultation and coordination among all actors. Further harmonization between the two data sets is needed in light of the division between ETS and non-ETS policies in the EU Climate and Energy Package.

Acknowledgements

The research described was ordered and financed by the Flemish Environment Agency in the framework of Flanders Environmental Reporting (MIRA).

2005 2006 2007 2008

TOTAL (kton CO2) 75.466 73.163 70.962 70.677

non ETS (kton CO2) 41.880 40.171 38.697 35.674

ETS (kton CO2) 33.586 32.992 32.265 35.003

share ETS 45% 45% 45% 50%

45% 45% 45%50%

0

20.000

40.000

60.000

80.000

CO2-emissions (kton CO2)


40

References

[1] IEA. (2004): Energy Statistics Manual. Paris, France. Available at : http://www.iea.org/textbase/nppdf/free/2004/statistics_manual.pdf

[2] IPCC. (1997): Greenhouse gas inventory reporting instructions. In: IPCC 1996 Revised Guidelines for National Greenhouse Gas Inventories, Vol.1. Bonn, Germany. Available at: http://www.ipcc-nggip.iges.or.jp/public/gl/invs4.html

[3] http://www.lne.be


41

Climate Change in the Gulf Countries – Situation and Reactions

Dhari Al-Ajmi

Kuwait Institute for Scientific Research P. O. BOX 24885, 13109-SAFAT-KUWAIT

[email protected]

Abstract

The State of Kuwait, the Kingdom of Saudi Arabia and the Kingdom of Bahrain ratified the UN Framework Convention on Climate Change (UNFCCC). This convention aims to stabilize the greenhouse gases (GHG) concentrations in the atmosphere at a level that would prevent significant potential changes to the global climate. One effective option that has been adopted by various developed countries to achieve this objective is the stabilization of greenhouse gas emissions by the 2000 at their 1990 levels. Being a signatory to the UNFCC, the three Gulf Cooperation Council (GCC) countries have agreed to develop and submit their First National Communication report to the Secretariat of the Conference. This paper represents the major findings of these reports.

Keywords: Greenhouse gases, global climate

1. Introduction

The inventory of anthropogenic emissions and removals by sinks of greenhouse gases has been prepared by the three GCC countries according to the 1996 Guidelines of the Intergovernmental Panel for Climate Change (IPCC, 1997). The greenhouse gases covered in this inventory included the direct greenhouse gases; namely, carbon dioxide (CO2) methane (CH4) and nitrous oxide (N2O).

The process of developing National GHG Inventory passed through various steps as proposed in the 1996 IPCC guidelines including: Identification of the types of data to be collected from each emission source category and sub-sectors, preparation of a list of government ministries and other governmental, semi-governmental, and private organizations, Development of questionnaires or forms to collect the required information, Collection of inventory data and information. Tabulation of the collected data in 1996 IPCC prescribed format, Estimation of the greenhouse gas emissions/sinks based on methodologies recommended by the 1996 IPCC Guidelines; and, Development of the report and summary of total anthropogenic emissions of greenhouse gases and their removals by sinks.

Input data for the inventory were collected for the period covering 1990 through 1996. However, greenhouse gas emissions were estimated for the base year for Kuwait, Saudi Arabia and Bahrain are 1994, 1994 and 1990 respectively as stipulated in the IPCC guidelines. In addition to the questionnaires, various other sources of information were consulted. The emission factors were adopted from the 1996 IPCC Guidelines. Additionally, where more accurate (than the default emission factors suggested in IPCC Guidelines) country-specific information was available, new emission factors were


42

developed and adopted for this study. The major sectors considered were: Energy, Industrial processes, Agriculture, Land-use change and forestry and Waste.

2. Contributions of Major Sectoral Activities to GHG emissions

In Kuwait the total GHGs contribution from all sectors were calculated for the year 1994 (Kuwait, 2006) and outlined in Table 1. It is evident that CO2 emission from energy sector accounts for approximately (99.6%) of total CO2 emissions. The main sources of CO2 emissions were found to be the energy production sector (i.e. power and oil industries). The transportation and household sectors contributed a share of (3.68%) and (1.65%) respectively. The CO2 emissions from cement industry was aggregated as negligibly small (0.4%) from the industrial sector. Unlike other countries, the main contributor of GHGs emissions particularly the CO2 emission in Kuwait is from energy sector.

Table 1. Summary of 1994 GHGS (Gg) Emissions for all Source Sectors in Kuwait

No. Source Sector CO2 CH4 N2O

1 Energy 160343.48 (99.6%) 873.69 (97.07%) 1.044 (81%)

2

Industry processes

654.92 (0.4%)

0

0

3

Agriculture

0

7.97 (0.8%)

0

4

Waste

0

18.78 (2 %)

0.244 (19 %)

The overall methane emission was calculated in 900.44 Gg. The energy sector

contributed about 873.69 Gg with a share of (97.07%) and the contribution of agricultural sector in CH4 emission was estimated to be 7.97 Gg with a share of (0.8%). The methane emission from waste treatment was estimated to be 18.78 Gg with a share of (2%) of the total CH4 emission. From the agriculture sector, CH4 emissions from sheep accounts for the maximum contribution of methane among all livestocks. The total CH4 emission from agricultural and waste treatment was estimated to be only 27 Gg, which shares about (3%) of the total CH4 emission.

In Kuwait, the N2O emissions estimated for all sectors were found negligible compared to CO2 emission. The contribution of N2O emissions from energy sector was about (81%) and that from the waste treatment was (19%).

CO2 emissions in Saudi Arabia in 1990 were 140,948 Gg and CO2 sinks were 15,240 Gg (Saudi Arabia, 2005). As shown in Table 2, the energy sector contributed (90%) of the total CO2 emissions, followed by the industrial processes sector (8%) and the agriculture sector (2%). The major source categories contributing to these CO2


43

emissions (contributions ≥ 2% of the total emissions) were electricity generation (26%), road transport (25%), desalination (15%), petroleum refining (10%), cement production (5%), cement industry (3%), petrochemical industry (3%), aviation (3%) and iron and steel production (2%).

Table 2. Summary of 1990 GHGS (Gg) Emissions for all Source

Sectors in Saudi Arabia


1 Energy 127,385 (90.4%) 87.6 (11%) 0.9 (2.7%)

2

Industrial Processes

10,881 (7.7%)

9.6 (1%)

–

3

Agriculture

2,692 (1.9%)

88 (11%)

30.7 (90.8%)

4

Land-use change and forestry

-15,240 (14.5)%

–

–

5

Waste

–

602 (77%)

2.2 (6.5%)

In Saudi Arabia, CH4 emissions were 787 Gg. The waste sector contributed (77%)

of the total CH4 emissions followed by agriculture (11%), the energy (11%) and the industrial processes sector (1%). N2O emissions were 33.8 Gg. The agriculture sector was the major contributor with (91%) followed by the waste (6%) and the energy (3%) sectors.

In Bahrain the GHGs emissions as outlined in Table 3 (Bahrain, 2005) shows that the main GHG gas emitted is carbon dioxide 16,483 Gg, which constitutes (99%) of the total, followed by methane 140 Gg. N2O was emitted in a very small amount. The main GHG-emitting sector is energy, which account for about (71%) of total emissions on a carbon dioxide equivalent basis. Waste, industry, and transport account for (13%), (10%) and (7%) respectively.

In summary, Table 4 shows that significant amount of total CO2 emissions were released into the atmosphere from all the four sectors in all the three GCC countries.

3. Mitigation Measures

Many GCC countries have expanded road networks at the national and regional levels, and have further developed and upgraded their public transit systems. Policy approaches in the region have focused mainly on switching to fuel types that are less polluting such as unleaded fuel and natural gas. An effort is being made to introduce more sustainable fuels in power generation. Unleaded gasoline has been introduced to


44

the GCC countries, and has been the only fuel produced in Bahrain since July 2000 (UNEP 2003a). Steps have been taken in Saudi Arabia to replace gasoline with natural gas, and the first natural gas powered car came on the roads in March 2001 (Green Gulf Report, 2006).

Table 3. Summary of 1994 GHGs (Gg) Emissions for all Source Se Bahrain


1 Energy 14,633 26.50 0.04

2

Industrial Processes

1,850

2.1

0

3

Agriculture

-

0.84

0.09

4

Waste

-

110.8

0.0

Table 4. The total GHGs (Gg) Emissions for the three GCC countries

The oil and gas industry has also undertaken measures to show environmental

responsibility. Saudi Aramco’s Master Gas System, which significantly reduced the need for flaring, recovers more than 3500 tons of elemental sulfur per day from gas produced in association with crude oil. Also, in July 1999, the Saudi Consolidated Electric Company announced that all service and repair workshops in the city of Jubail , including some at Jubail industrial city itself, would be relocated to a new site outside residential areas and far from the city zones in order to protect the population from pollution.

Bahrain National Gas Company (BANAGAS) has taken measures to curb the effects of air pollution. Measurement of gas emissions produced by gas turbines and driers is

GHGs

KUWAIT

( base year,1994)

SAUDI ARABIA (base year 1990)

BAHRAIN ( base year,

1994)

Total

CO2

160,998.4

125,718

16483

303,199.4

CH4

900.44

787.2

140.24

1,8267.88

N2O

1.288

33.8

0.13

35.22


45

carried out regularly. Also the gas company has eliminated the usage of halons, which have ozone-depleting potential.

Moreover, the GCC countries have sought to unify all regulations, laws, and legislation dealing with various aspects of the environment. These represent the minimum requirements for the enhancement of national legislation. Kuwait, Oman, Qatar, Saudi Arabia and the United Arab Emirates have acceded to the Kyoto Protocol for climate change. The Clean Development Mechanism (CDM) projects under the Protocol are gaining acceptance in the region and are being pursued with special emphasis on carbon sinks.

4. Conclusions

The GCC Countries economy will undoubtedly be impacted by Annex I countries of Kyoto Protocol climate change response measures since these actions will be implemented as policy measures to reduce primarily CO2 emissions. This will reduce oil demand in these countries and directly reduce GCC countries crude oil exports which comprise a large portion of its GDP. However, the GCC countries can adapt to these response measures by diversifying its economy sufficiently away from its crude oil export sales. In order for Annex I countries to implement these response measures, they must attend to the needs of developing countries as stipulated in climate change convention and the Kyoto Protocol.

GCC countries will require assistance from the Annex I countries to diversify its economy in order to adapt to potential climate change related energy policies. However, this will require a joint effort between all Annex I countries and GCC in order to implement solutions for GCC economic diversification. This can be achieved by leveraging the GCC’s potential assets (abundant and low cost energy resource as well as a large youth population) and providing investments as well as implementing technological know-how of Annex I countries.

References

1. Bahrain, 2005. “Bahrain’s Initial Communications to the United Nations Framework Convention on Climate Change, Volume 1: Main Summary Report”. Kingdom of Bahrain: General commission for Protection of Marine Resources, Environment & Wildlife.

2. Saudi Arabia, 2005. “First National Communication of Saudi Arabia: Submitted to United Nations Framework Convention on Climate Change” (UNFCC). Kingdom of Saudi Arabia: Presidency of Meteorology and Environment, Ministry of Defence and Aviation.

3. Kuwait, 2006. “The First Kuwait National Communication under the United Nations Framework Convention on Climate Change”. Kuwait Institute for Scientific Research.

4. Green Gulf Report, 2006. Gulf Research Center, Energy and Resource Institute. UAE.

5. IPCC (1977a). Greenhouse Gas Inventory Reporting Instructions. Revised IPCC Guidelines for National Greenhouse Gas Inventories. Volume 1, Ed. J.T. Houghton, L.G. Meira Filho, B. Lim, K. Treanton, I. Mamaty, Y. Bonduki, D.J.Griggs, and B.A. Callander, Intergovernmental Panel on Climate Change, WGI Technical Support Unit, London, United Kingdom.


46

6. IPCC (1997b). Greenhouse Gas Inventory Workbook, Revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories. Volume 2. Ed. J.T. Houghton, L.G. Meira Filho, B. Lim, K. Treanton, I. Mamaty, Y. Bonduki, D.J. Griggs, and B.A. Callander. Intergovernmental Panel on Climate Change, WGI Technical Support Unit, London, United Kingdom.

7. IPCC (1977c). Greenhouse Gas Inventory Reference Manual. Revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories. Volume 3, Ed. J.T. Houghton, L.G. Meira Filho, B. Lim, K. Treanton, I. Mamaty, Y. Bonduki, D.J. Griggs, and B. A. Callander. Intergovernmental Panel on Climate Change, WGI Technical Support Unit, London, United Kingdom.

8. UNEP 2003a. Global Environment Outlook 2003. Produced by the UNEP GEO Team Nairobi, Kenya: Division of Early Warning and Assessment. United Nations Environment Programme.


47

Addressing uncertainties of GhG emission abatement in Ukraine

Olga Diukanova1, Igor Liashenko2

1Fondazione Eni Enrico Mattei, Isola di San Giorgio Maggiore, I-30124 Venice, Italy [email protected]

2Taras Shevchenko National University of Kyiv, Faculty of Cybernetics, Dept. of Mathematical Informatics, Glushkov boul. 2, build 6, Kyiv 03680, Ukraine

Abstract

The issue of greenhouse gas abatement in Ukraine is associated with multiple uncertainties. First, emission levels depend on economic activity and energy efficiency which in turn are sensitive to climate policy adopted in the country. Second, the impact of future abatement policies on national economy is ex ante unknown. Third, no decision regarding the instrument of emission reduction has yet been made by the Ukrainian government. In order to assist national policymakers in adoption of a feasible and effective GhG reduction policy, a range of alternative post-2012 carbon dioxide reduction targets were evaluated. The methodology is based on a forward-looking dynamic multi-sectoral computable general equilibrium model of a Ramsey-Cass-Koopmans type. The model employs domestic emission trading as an instrument of CO2 reduction in the country. Monte Carlo simulations were performed in order to assess sensitivity of the model results towards the values of key exogenous parameters.

Keywords: uncertainty, post-2012 emission policies, computable general equilibrium model.

1. Introduction

The inadequate economic structure, inefficient functioning of energy sector and lax environmental legislation cause numerous environmental problems that substantially affect Ukraine’s economic performance and health of its population. Large share of national economy is represented by resource-oriented industries that operate outdated industrial infrastructure difficult to modify without significant investments. Ukraine was ranked as one of the most energy- and carbon- intensive countries in the world in terms of intensity per capita and per unit of GDP, IEA (2009). At the same time Ukrainian economy is largely dependent on imported energy, State Statistics Committee of Ukraine (2009). Ukrainian energy sector a source of 70% of domestic greenhouse gases (GHGs) emissions, 78% of which are CO2, MEP of Ukraine (2009). Energy sector is a major contributor to harmful local and transboundary air pollution, EMEP (2008). Due to its low energy efficiency, Ukraine has huge potential for low-cost emission abatement, World Bank (2003). However, GHG emissions have never been regulated in the country and government is unwilling to start their abatement. On COP-15 in Copenhagen Ukrainian government has announced commitment to reduce its GhG emissions by 20 % by 2020 and by 50% by 2050 taking 1990 as a baseline year, UNFCCC (2009). This pledge was ranked by the international community as inadequate, since the emission abatement commitment by 2020 is above business-as-usual projections (Ecofys, 2009).

2. Literature review

There are a number of studies focusing on evaluation of GHG reduction policies and ancillary benefits associated with them that employ CGE approach F. Bosello et. al. (2005), (2000), S. Dessus et. al. (2003) and others.


48

3. Policy scenaria

In order to prevent catastrophic consequences of climate change, the global average temperature should not exceed 2°C above the pre-industrial level. It requires long-term stabilization of global GHGs at a level less than 550 ppm by 2100. This target is equivalent to CO2 stabilisation at a level of 450 ppm and reduction of other greenhouse gases at similar rates. To stabilize CO2 at 450 ppm, global carbon dioxide emissions have to be reduced approximately by 30% in 2050 compared to 1990 levels, IPCC (2007), Ecofys (2006), Criqui et al. (2003). This study develops and evaluates a range of alternative post-2012 carbon abatement policies for Ukraine according to the global CO2 stabilization target at 450 ppm. Departing from the UNFCCC principles that define equitable distribution of emission reduction commitments among the countries (United Nations,1992), the study considers the following CO2 reduction scenarios for Ukraine:

• ET90 scenario reflects the official pledge of Ukrainian government that was proclaimed at the 7th session of AWG-KP meeting in Bonn, UNFCCC (2009) to reduce national CO2 emissions by 20% in 2020 and by 50% in 2050 comparing with the level of 1990.

• ET06 scenario considers 20% reduction of CO2 by 2020 and 50% reduction by 2050 below the 2006 level. It reflects the estimated target for Ukraine according to the NEAA (2009a and 2009b), ERC (2009) and IIASA (2008a) studies. These estimates were based on a combination of different criteria such as ability to pay, historical and per capita emissions, potential and cost of emission reduction.

• Hist scenario reflects Historical responsibility principle that bases allocations upon each country’s historical contribution to damages caused by climate change. Over the 1850-2008 period Ukrainian energy-related CO2 emissions were about 2% of the global energy-related CO2, CAIT (2009). Based on CAIT data it was estimated that historical responsibility principle allows Ukraine for 22% increase in CO2 by 2050 comparing with 2007 level.

• MAC scenario is based on equalization of marginal emission abatement costs among the nations in order to achieve a common stabilizaton target (the Capability Principle). Global marginal abatement cost of meeting a 450 ppm CO2 stabilization target was approximated according to Van Vuuren et.al. (2006) estimates as equal to 66 Euro/t CO2

in 2050. Simulations with the model developed in this study have shown that equalization of Ukrainian MAC with the global one by 2050 allows for 14% increase in national CO2 from the level of 2007.

• pCap scenario foresees equalization of per capita emissions between the countries by 2050 according to the common 450 ppm CO2 stabilization target (the Egalitarian principle). Based on statistical data of CO2 per capita in different countries, EIA (2009a) it was estimated that this target would require 77% reduction of Ukrainian CO2 by 2050 vs. 2007.

These scenaria were compared with a baseline trajectory under the absence of CO2 reduction (BAU scenario).

CO2 emission limit. Since Ukraine has a surplus of “emission rights” equal to 1350 mln. t of CO2-equivalent during the first commitment period of the Kyoto Protocol, emission reduction targets specified by the model scenaria were applied starting from 2012.


49

4. The methodology

To quantify the economy-wide costs of post-2012 carbon reduction policies and ancillary benefits associated with them, a dynamic forward-looking general equilibrium model of Ramsey-Cass-Koopmans type was developed in this study. Since most of the dynamic CGEs are built on the premise of a steady-state economic growth over the model horizon, application of this framework would lead to overestimation of economic impacts from CO2 abatement in Ukraine. Therefore a CGE developed in this study features a long–term demographic crisis in the country along with economic crisis of 2008-2009. The model was formulated as a mixed complimentarily problem (MCP) and solved by GAMS/MPSGE software with PATH solver, Ferris (2000) and Rutherford (1999).

A single representative household is endowed with labor and capital. Its disposable income consists of factor income and governmental transfers. Households’ decision regarding consumption and savings depends on real interest rates. Capital accumulation function equates the current capital stock to the depreciated stock inherited from the previous period and is augmented by the gross investment.

Sectors’ output is produced by the capital, labor and intermediate inputs. Domestic and imported varieties of the same good form Armington aggregate. According to the Armington assumption (Armington, 1969), which is employed in the model, there is imperfect substitution between imported and domestic varieties of the same good and imperfect transformation between domestic consumption and export. The Armington aggregate enters sectors’ production, household and government consumption and investment. Produced output is divided into domestic consumption and export. Ukraine is treated as a small, open economy relative to the international market. It was assumed in the model that the balance of payments surplus/deficit is fixed and the exchange rate is flexible.

Government collects income from taxes and auctioned emission permits and uses this income to finance public consumption and pay transfers to households. It is endowed with

carbon emission permits, which it distributes among the sectors in accordance with selected allocation scheme. Although governmental revenue-neutrality is maintained in the model, the proceeds from permits’ auctioning are redistributed as a lump-sum payment to households.

Although most of the dynamic CGE models employ the assumption of steady-state economic growth, it is incompatible with the evolution of Ukrainian economy. The recent economic crisis led to a sharp economic slowdown in Ukraine in 2008-2009. The country faces a strong demographic crisis with 0.68% annual rate of population decline. These are the reasons why necessary adjustments were made to reflect the unsteady economic growth in the country. First, a time-dependent growth rate was introduced into the model. It was calibrated to the observed and forecasted GDP growth rates, Ministry of Economy of Ukraine (2009b). Since there are no national forecasts beyond 2020, an assumption of 3% annual growth in 2021-2050 was employed. Second, two components were defined for the investment sector: investments aimed at accumulation of productive capital and investments intended at non-productive use in final consumption. Equations of capital accumulation, investments and final demand were adjusted to accommodate these changes. Energy productivity was modelled via the autonomous energy efficiency improvement (aeei) factor which defines a 1% annual increase in energy efficiency. Following McKinsey (2009) a 2.3% increase in annual labor productivity was assumed. It alleviates to a certain extent the negative impact of demographic crisis.


50

The CGE model for Ukraine has 16 production sectors: metallurgy, electricity, coke production, transport, chemical industry, coal mining, mineral production, building and construction, agriculture, production of hydrocarbons, petroleum refinement, non-energy extraction, pulp and paper production, mashinery, food processing and the aggregate of remaining least energy-intensive sectors (ROI). All sectors were assumed to operate under the cost optimization production technology.

Substitution possibilities among various inputs were reflected through the nested production functions.Sectors’ production functions have five levels of nesting. The first level reflects substitution between the aggregates of non-energy inputs and the energy-value-added. Second level shows substitution between the energy and value-added aggregates, and among the non energy inputs. On the third level labor substitutes with capital and electricity substitutes with the aggregate of fuels (petroleum, coal and hydrocarbons). Fourth level features substitution between petroleum and fossil fuels aggregate. Fifths level defines substitution between coal and hydrocarbons. Levels of nesting were defined by the constant elasticity of substitution (CES) functions.

Since coal is a main energy resource in coal mining and coke production, hydrocarbons are main, necessary resource in petroleum refinement and production of hydrocarbons, and petroleum is a main resource for transport sector, for the clarity of analysis an additional top-level of nesting was employed in these sectors’ production functions to define complementarity between the sectors’ main energy-resources and the rest of inputs.

Household and governmental preferences were captured via production function with 3 levels of nesting. First level represents substitution possibilities between the energy aggregate and composite of non-energy goods. Second level reflects substitution between: (a) electricity and the fuels composite, and (b) among the non-energy goods. Third level combines different types of fuels. Levels of nesting were reflected by CES functions. CO2 emissions from fossil fuels were linked to the consumption of coal, hydrocarbons and petroleum by different sectors. Emission coefficients were estimated according to the data of the Ukrainian GHG Inventory, MEP of Ukraine (2009) and EIA (2009a).

The model employs domestic emission trading as a main instrument for CO2 reduction in Ukraine. In order to assure some adjustment period for Ukrainian industries to operate under the carbon constraint, the model starts with 100% free allocation of emission permits based on 2007 output. The amount of distributed for free permits is gradually decreased in order to arrive to 100% auctioning in 2050. According to the output-based allocation scheme employed in the model, each sector allocation is determined by the share of its output in the total production. Since this type of allocation is tied to the level of production, it performs as an implicit subsidy to output. The sector-specific subsidization rate was determined as ratio of permits` value that were allocated to the sector to its revenue.

5. Discussion of Results

Model results prove that Ukrainian official pledge regarding its post-2012 CO2 reduction policy reflected by ET90 scenario does not imply any CO2 abatement until 2028 (Figure 1). For 2008-2028 period the trajectory of ET90 scenario coincides with the baseline trajectory (BAU scenario) and the later does not foresee any emission reduction. Indeed, the ET90 target allows for increase in industry energy-related CO2 by 49% in 2020 and by 24% in 2050 from the level of 2007, generating “hot air” in 2013-2027.


51

Defining “hot air” as difference between the allowed emissions according to the imposed constraint and actual emissions, the EU90 scenario generates 1777 Mt CO2 of “hot air” during the period 2013-2027, as shown on Figure 2. The annual amount of “hot air” gradually decreases from 268 MT in 2013 to 12 MT in 2027.

-100

-50

0

50

100

150

200

pCAP ET06 MAC Hist ET90 BAU

%

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050

Figure 1. Energy-related industry CO2 emissions, % vs. 2007

-50

0

50

100

150

200

250

300

ET90

M t

CO

2

2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026

2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040

2041 2042 2043 2044 2045 2046 2047 2048 2049 2050

Figure 2. Excess of the AAUs

Although the post-2012 policies to some degree slow down production growth comparing with the baseline trajectory, at the same time they promote significant structural changes in favour of least energy–intensive sectors (Figures 3-5). The extent of these structural changes is proportional to the stringency of CO2 abatement target. However, all scenaria show similar pace of growth in cumulative production unitl 2020.

The group of the least energy intensive sectors consists was represented by ROI aggregate of the least energy-intensive sectors. Rest of the sectors defined in the model belong to the group of the most energy-intensive industries.

Cost of CO2 reduction was calculated in terms of a marginal abatement cost (Figure 6). Egalitarian principle reflected by scenario pCAP is especially unfavourable for Ukraine since its per capita CO2 emissions are among the world highest.


52

-50

0

50

100

150

200

250


%

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 Figure 3. Cumulative industry output, % vs. 2007

-100

-50

0

50

100

150

200

250

300


%

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 Figure 4. Output of the most energy intensive sectors, % vs. 2007

-50

0

50

100

150

200

250


%

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 Figure 5. Output of the least energy intensive sectors, % vs. 2007


53

Country’s welfare was measured in terms of GDP (Figure 7). Model results show that all post-2012 CO2 abatement policies allow to preserve the projected GDP growth until 2020. Although strict environmental policies reflected by pCAP and ET60 scenaria require sharp CO2 abatement below the level of 2007, they ensure increase in the national GDP above 2007 level by 100% and 140% correspondingly in 2050. Mild policies (MAC, Hist and ET90) allow for about 200% increase in GDP by 2050 relative to the level of 2007.

Model results indicate significant energy saving reflected in terms of domestic energy consumption comparing with the baseline trajectory (Figure 8).

0

100

200

300

400

500

600

700

800

pCAP ET06 MAC Hist ET90

Eur

o/t

CO

2

2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026

2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040

2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 Figure 6. Marginal CO2 abatement cost, Euro/t CO2

-50

0

50

100

150

200

250


%

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 Figure 7. GDP, % vs. 2007

6. Conclusions

Application of a CGE framework to evaluate the alternative post-2012 CO2 reduction targets for Ukraine allow to conclude that country could benefit from CO2 reduction. CO2 abatement brings structural changes towards the least energy intensive industries with high share of value added and reduction of energy consumption. The magnitude of these effects depends on a stringency of CO2 abatement target and the assumed values of the future economic growth rate, energy and labor efficiency as well as elasticities of substitution. Irrespective to the selected target, all scenaria show similar pace of growth in cumulative output and GDP until


54

2020. Therefore even the most even stringent pCap and ET06 scenaria can be considered as policy measure in 2013-2020. Such targets may be adopted for a longer term if significant technological improvements in energy saving are achieved in the country. Otherwise scenaria MAC and Hist that ensure both economic growth and CO2 abatement can be selected.

-100

-50

0

50

100

150

200

250


%

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 Figure 8. Domestic energy consumption, % vs. 2007

References

1. Amann M., I. Bertok, et. (2008a). GAINS - Potentials and costs for greenhouse gas mitigation in Annex I countries. International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria http://gains.iiasa.ac.at/

2. Armington P.(1969), A Theory of Demand for Producers Distinguished by Place of Production, IMF Staff Papers 16.

3. Barro R. J. and Sala-i-Martin X. (2004). Economic Growth. MIT Press, Cambridge MA.

4. Bosello F., Roson R., Tol R. S.J. (2006). Economy-wide estimates of the implications of climate change: Human health, Journal of Ecological Economics 58, 579– 591.

5. CAIT (2010). Climate Analysis Indicators Tool (CAIT) Version 7.0. (Washington, DC: World Resources Institute, 2010). http://cait.wri.org./cait.php?page=compcoun

6. Criqui P., Kitous A., et al. (2003). Greenhouse gas reduction pathways in the UNFCCC Process up to 2025 - Technical Report. No. B4-3040/2001/325703/MAR/E.1 for the DG Environment. Grenoble, France: CNRS-IEPE.

7. Dessus S., O’Connor D. (2003) Climate Policy without Tears: CGE-Based Ancillary Benefits Estimates for Chile”, Environmental and Resource Economics 25: 287–317, 2003.

8. Ecofys (2006).Greenhouse Gas Stabilization Targets: What are the Near-term Implications?(2006) http://www.ccap.org/docs/resources/67/Lee_HoehneNearterm_Implications_of_Stabilization_Targets.pdf

9. Ecofys (2009). Climate action tracker. Detailed information on individual country pledges for greenhouse gas emission reduction http://www.climateactiontracker.org/country.php?id=365

10. EIA (Energy Information Administration, 2009a). International Energy Statistics. CO2 emissions.


55

http://tonto.eia.doe.gov/cfapps/ipdbproject/iedindex3.cfm?tid=5&pid=5&aid=8&cid=&syid=1980&eyid=2008&unit=MMTCD&products=5

11. EIA (Energy Information Administration, 2009b). International Petroleum (Oil) Prices and Crude Oil Import Costs http://www.eia.doe.gov/emeu/international/oilprice.html

12. International Natural Gas Prices http://www.eia.doe.gov/emeu/international/gasprice.html

13. ERC, (Energy Research Centre, 2009). Analysis of possible quantified emission reduction commitments by individual Annex I Parties.-H.Winkler, A.Marquardt and T. Letete.

14. Ferris M.C., Munson T.S. (2000). GAMS/PATH User Guide. www.gams.com/docs/pdf/path.pdf

15. Höhne N., Galleguillos C., Blok K., Harnisch J., Phylipsen D. (2003). Evolution of commitments under the UNFCCC: Involving newly industrialized economies and developing countries, Environmental Research of the Federal Ministry of the Environment, Nature Conservation and Nuclear Safety. Research Report 20141255 http://www.chem.uu.nl/nws/www/publica/Publicaties2003/e2003-155.pdf

16. IEA (International Energy Agency, 2009). Key World Energy Statistics 2009 http://www.iea.org/co2highlights/co2highlights.pdf

17. IIASA (International Institute for Applied Systems Analysis, 2008a). Comparison of GHG Mitigation Efforts for Annex 1 Parties. http://gains.iiasa.ac.at/gains/Annex1.html

18. IPCC (2007). Fourth Assessment Report http://www1.ipcc.ch/ipccreports/assessments-reports.htm

19. McKinsey (2009). Reviving Ukraine’s Economic growth http://www.mckinsey.com/aboutus/Ukraine_Economic_Growth_ENG.pdf

20. Ministry of Economy of Ukraine (2009a). Strategy of Ukraine's innovative development in 2010- 2020 under the challenges of globalization http://kno.rada.gov.ua/komosviti/control/uk/doccatalog/list?currDir=48718

21. Ministry of Economy of Ukraine (2009b). Main indicators of economic and social development in Ukraine (in Ukrainian)- http://www.me.gov.ua/control/uk/publish/category/main?cat_id=78198

22. MEP (Ministry of Environmental Protection of Ukraine, 2006). Ukraine’s Report On Demonstrable Progress Under The Kyoto Protocol http://unfccc.int/resource/docs/dpr/ukr1.pdf

23. MEP (Ministry of Environmental Protection of Ukraine, 2008). National Report on State of the Environment in Ukraine in 2007 http://www.menr.gov.ua/cgibin/go?node=nac%20dop%20p%20nps

24. MEP (Ministry of Environmental Protection of Ukraine,2009). National Inventory of anthropogenic emissions in Ukraine for 1990-2007http://unfccc.int/national_reports/annex_i_ghg_inventories/national_inventories_submissions/items/4771.php

25. NEAA (Netherlands Environmental Assessment Agency, 2009a). Exploring comparable post–2012 reduction efforts for Annex I countries. http://www.pbl.nl/en/publications/2009/Exploring-comparable-post-2012-reduction-efforts-for-Annex-I-countries.html


56

26. NEAA (Netherlands Environmental Assessment Agency, 2009b). Pledges and actions –A scenario analysis of mitigation costs and carbon market impacts for developed and developing countries. http://www.rivm.nl/bibliotheek/rapporten/500102032.pdf

27. Rutherford T.F. (1999). Applied General Equilibrium Modeling with MPSGE as a GAMS Subsystem: An Overview of the Modeling Framework and Syntax. Computational Economics.– Vol. 14.– Number 1-2

28. State Statistics Committee of Ukraine (2009). Material and energy resources statistics. http://www.ukrstat.gov.ua/

29. UNFCCC (2009). Information relating to possible quantified emissions limitation and reduction objectives as submitted by Parties. http://unfccc.int/resource/docs/2009/awglca7/eng/misc06a01.pdf

30. United Nations (1992). United Nations Framework Convention on Climate Change: UN Document A:AC.237/18.2,3

31. United Nations (2007). Environmental Performance Reviews: Ukraine -Second Review

www.unece.org/env/epr/epr_studies/Ukraine%20II.pdf

32. Van Vuuren D.P., den Elzen M.G.J.(2006). Stabilising greenhouse gas concentrations at low levels: an assessment of options and costs Netherlands Environmental Assessment Agency. MNP Report 500114002/2006.


57

A model for robust emission trading under uncertainties

T. Ermolieva, Y. Ermoliev, M. Jonas, G. Fischer, M. Makowski, F. Wagner, W. Winiwarter

International Institute for Applied Systems Analysis Schlossplatz 1, A-2361 Laxenburg, Austria

[email protected]

Abstract

The international emission trading (IET) scheme was devised to lower the cost of achieving sets of greenhouse gas emission reductions for different countries: emissions are reduced where it is cheapest and emission certificates are then traded to meet the nominal targets in each country. However, carbon markets, like other commodity markets, are volatile. They react to stochastic “disequilibrium” spot prices, which may be affected by speculations and bubbles. The underlying, actual cost of GHG mitigation, i.e. the marginal costs of abatement technologies is only of secondary importance. The market-based emission trading, therefore, does not necessarily minimize abatement costs and achieve emission reduction goals. Although in Copenhagen little of progress has been made towards increasing emission reduction goals and reaching binding agreements, it is likely that emission trading schemes will continue to be one of the essential economic mechanisms for emissions regulations also in post-Kyoto period, both at the national as well at the international level. While the EU has already implemented a carbon trading scheme several years ago, other developed countries such as US and Australia are ready to adopt the cap-and-trade emission trading system. The paper discusses the following key questions: Under which conditions is carbon trading environmentally safe and cost-effective in the long-term, if considered in the context of a stochastic market? How the knowledge about uncertainties may affect portfolios of technological and trade policies or structure of the market, e.g., if knowledge of uncertainty may turn buyer into seller? How uncertainties characteristics may affect market prices and change the market structure? We introduce a basic stochastic trading model allowing us to analyze the robustness of economic mechanisms for emission reduction under multiple natural and human related uncertainties. We illustrate functioning of the robust market with numerical results involving such countries as US, Australia, Canada, Japan, EU27, Russia, Ukraine, etc.

Keywords: Emissions trading, market-based, uncertainties, robust economic mechanisms, detectability, environmental safety, cost-efficiency, stochastic equilibrium

1. Introduction

The aim of this paper is to analyze an integrated approach for joint treatment of natural and human related uncertainties and bilateral carbon emission trading. Proposed stochastic model allows to introduce explicitly the detectability of emissions to control the safety of Kyoto targets for robustness of trading schemes. The introduced scheme takes long-term perspectives on emission trading by using rational expectations.

The dynamics of this scheme is driven by specific decentralized stochastic optimization procedure with different endogenous disequilibrium prices between mutually beneficial bilateral trades, but finally the system converges to cost-effective and environmentally safe global equilibrium. The safety constraints work as a discounting mechanism that discounts the reported emissions to detectable levels. This, in turn, provides incentives for participants to reduce uncertainties before trading.


58

The explicit introduction of emission detectability and safety constraints or(and) increasing returns of new abatement technologies may easily produce the duality gap preventing price-based trading to be environmentally safe and cost-effective scheme. The role of proposed computerized multiagent trading system is central for dealing with long-term perspectives, irreversibility and “lock-in” equilibriums. This system can be viewed as a device for decentralized collective regulation of trades.

2. Emission trading

The public property of large scale pollution makes it impossible to organize complete environmental markets with private demand for and private supply of pollution control. The monetary evaluation of environmental damage and therefore the benefits of cleaning up makes little sense. Yet, the idea of carbon trading is becoming increasingly popular for global climate change control. At the same time, the existence of various exogenous and endogenous inherent uncertainties raises serious concerns regarding the ability of carbon trading markets to fulfill the main purpose of the climate change control without creating world-wide irreversible socio-economic and environmental disruptions. Definitely that interests of profit oriented markets may contradict the main concerns of the environmental safety.

The Kyoto Protocol was established in 1997 under the United Nations Framework Convention on Climate Change proposed in 1992. For each country taking part, the Protocol specifies a target emission level not to be exceeded in this period. However, it also allows for emission trading. Although in Copenhagen very little of progress has been made towards increasing or specifying new emission reduction goals and reaching binding agreements, it is likely that emission trading schemes will continue to be one of the essential economic mechanisms for emissions regulations also in post-Kyoto period, both at the national as well at the international level. While the EU has already implemented a carbon trading scheme several years ago, other developed countries such as US and Australia are ready to adopt the cap-and-trade emission trading system. Emission trading means that each party (country, company) of the Protocol has the possibility to exceed their prescribed emission level given that another party carries out an equivalent emission reduction such that the aggregate emission level remains constant. It is assumed that parties with high emission reduction costs will buy emissions from parties with low emission reduction costs within prescribed Kyoto targets. In other words we can think of parties engaged in a bilateral emission exchange process driven by cost minimizing and environmentally safe sequential decisions without the need for a market.

In contrast, carbon trading markets are more similar to share markets. Parties hold a number of permits to emit a specific amount of emissions. The total amount of permits cannot exceed a limit or a cap on the amount of carbon (or other pollutant) that can be emitted. Parties that need to increase their emissions beyond the cap must buy permits on the market according to prevailing market price. Thus, if in the bilateral emission trading scheme the exchange of emission is driven by abatement costs and safety constraints, in the carbon trading markets the exchange of emissions is driven by prevailing market prices, which are usually modeled by exogenous random processes with potential bubbles induced by speculators. Such price signals may have nothing in common with minimization of abatement costs and environmental safety constraints of participants unless they have rather specific endogenous character defined by the dual model generating dual cost-effective and environmentally safe prices. In fact, the design and


59

even the existence of robust cost-effective and environmentally safe emission trading schemes can be derived from the availability of robust decentralized optimization procedures for underlying emission control model.

There are two principle approaches to control pollution: centralized “command-and-control” methods and decentralized market simulating schemes. If the central agency is fully informed about emissions and abatement cost functions of all parties, the primal problem of finding emission levels that meet given environmental standards in a cost-effective way is a straightforward task. It allows easily to deal with nonconvex cost functions typically encountered in long-term evaluations involving new technologies with increasing returns. The convexity of the model can also be destroyed by explicit introduction the detectability of emissions and environmental safety constraints.

If the central agency is not fully informed, then the primal model has to be solved in a decentralized manner. The bilateral emission trading scheme corresponds to a decentralized solution of the primal model. An alternative to this scheme may be a price-based carbon emission market simulating a decentralized solution of the dual problem. Unfortunately, the cost-effectiveness and environmental efficiency of this solution critically depend on the proper modeling of price processes, in particular, proper treatment of various uncertainties.

3. Emissions uncertainties

There is a number of scientific challenges related to the types and origins of uncertainties (Ermoliev et al., 2000, Godal et al., 2003, Liberman et al., 2007, Winiwarter 2001, 2007) affecting the design of pollution control schemes.

First of all, emissions of greenhouse gases (GHGs) are not directly observable. On the basis of specific emission factors, emissions can be estimated with information on GHG-emitting activities. These activities are assessed by a national agency in each Party and the inferred emission levels are reported to the Convention Secretariat according to specific guidelines developed by the Intergovernmental Panel on Climate Change (IPCC). The accuracy of estimated emissions depends on the quality of the monitoring system in each specific country and on the accuracy of the emission factors used (Ryptal and Zhang 2000). Conversion factors can also be manipulated, as many of them do not apply globally. As emissions of GHGs cannot be observed perfectly, parties can underreport emissions either on purpose or because of uncertainty. Current estimates of the overall uncertainty of the greenhouse gases inventories produced by the countries so far show that for most of them this uncertainty is in the same order of magnitude as, if not greater than, their agreed emission reductions.

The Good Practice Guidance report of the IPCC recommends recalculating historic emissions, whenever inventory methods change or are refined, when new source categories are included, or when errors are identified and need to be corrected. Emission estimates are revised from year to year, also for past periods, due to monitoring and increasing knowledge about emission sources. Figures 1 and 2 show how much may differ the estimates of emissions in different years, for Austria and EU27.

The Figure shows that the more one goes back in time the greater the observed difference is between initial and recalculated estimates of emissions. This can be explained as gradual correction of the mistake of previous calculations, i.e., by an increase in knowledge, in particular, investments in monitoring. Our trading model exploits these phenomenon.


60

Figure 1. EU-27’s CO2 emissions estimated initially and recalculated in 2005. (Source: Hamal, IIASA IR (forthcoming), NIR (1999-2007)).

Figure 2. Austrian’s emissions estimated in 2000 and recalculated in 2001, 2002, 2003 and 2005 (Hamal IIASA-IR, Austrian NIR (2002-2007)).

Apart from monitoring and measuring emissions, another essential source of uncertainties in emission trading is the uncertainties of the abatement costs. The parties have no incentives to reveal information on real costs. If a permit buyer reveals his abatement cost function, the seller can use this information when bargaining on a permit price such that the buyer is worse off than he otherwise would be. Hence, parties have incentives to keep this information private and the specific costs of emission reductions may remain unknown. Besides, abatement costs may vary according to unknown in advance market conditions including “bubbles” created by speculators. They are also subject to both industry wide and firm specific shocks, e.g., uncertainties and shocks connected with new technologies.

Thus, the ways uncertainty is quantified, represented and thereafter accounted for in decision-making have not yet been adequately reflected, neither in national inventory reports nor in the procedures for the design and approval of robust Kyoto flexible mechanisms (including emission trading schemes).


61

Figure 3. Probability density distribution of total emissions for Austria, 2006. [The distribution is most strongly influenced by the lognormal distribution of the uncertainty in N2O emissions. This is a proof of the importance on assumptions taken on N2O emissions on the overall uncertainty of a national GHG inventory.]

The paper shows critical importance of a proper uncertainty representation for emission trading which may have skewed probability distributions. In particular it shows that the use of uncertainty interval representation can leave out of consideration an essential mass of potential emissions and, hence, may be critically misleading.

4. Stochastic model for robust emission trading under uncertainties

Explicit representation of uncertainties allows to develop a stochastic emission trading model without explicit introduction of appropriate risk measures to control the safety of emission reduction targets (e.g., post-Kyoto pledge targets). This type of safety constraints is typical for pollution control, financial applications, stability regulations in the insurance industry and catastrophic risks management (Ermolieva et. al., 2005). In a sense, these constraints work as a probabilistic discounting mechanism which discounts the reported emissions to detectable levels overshooting uncertainty within a specified safety levels, i.e., portion of detectable emission changes.

Since the concept of safety constraints discounts emission changes to detectable levels, this provides incentives to reduce uncertainty before trading. This significantly affects the trade equilibrium state, i.e., this state cannot be achieved without explicit introduction of the safety constraints. In contrast to “black-and-white” uncertainties characterized by intervals, the proposed stochastic model aims to reduce underestimating and overestimating costs by using additional information on likelihoods of uncertainties which can be characterized by precise or/and imprecise probabilities and sets of potential scenarios.

The proposed sequential bilateral trading schemes are based on stochastic decomposition methods. The trade at each step takes place towards minimization of safety-adjusted costs of meeting parties. This generates disequilibrium random prices which are endogenously driven towards the cost-effective and environmentally safe equilibrium price. The proof of the convergence with probability 1 can be used for


62

the feasibility analyses of alternative schemes. The model shows why other trading schemes, especially based on exogenous price processes may not guarantee the cost-effectiveness and environmental safety. In particular, inappropriate introduction of the detectability of emission changes and safety regulations may easily destroy the duality of emission trading model with dramatic consequences on the ability of price-based trading schemes to achieve cost-effective and environmentally safe solutions.

The main concern is also the irreversibility of investments and other decisions, which may follow emission trades. In order to avoid this, we develop a computerized Multi-Agent Trading System (MATS) for stable and fair functioning of proposed robust emission trading scheme. The paper also analyzes path-dependencies of trading schemes relying on instantaneous observable markets situations. To analyze robust emission trading numerically, we use data on the costs of emissions reduction from the GAINS model for the countries and group of countries Australia, Canada, EU27, Japan, Norway, Russia, USA. The cost curves are displayed in Figure 6.

-100

-50

0

50

100

150

200

250

0 10 20 30 40

Australia

Canada

EU27

Japan

Norway

Russian_Federation

Switzerland

Ukraine

US

Figure 6. Cost curves for emissions reduction as percent of pledge targets, USD per tC

Table 1 shows reported emissions levels in 1990 and 2009. Baseline projected emissions level in 2020 is derived with GAINS model, and the pledge emission reduction targets are set equal to low pledge level according to (Wagner and Amman 2009). The data for emissions uncertainties and costs of reducing uncertainties is compiled from IPCC, (Nahorski et al (2007, 2010), Nilsson et al. (2000), Obersteiner (2000), Goddal et. al. (2003), Winiwarter (2001, 2007)). We have employed unreported emission levels in year 2020 as a percentage of the pledge targets. Table 2 shows that marginal cost between countries on reducing reported emissions are equal, in accordance with the outlined procedure. The marginal costs are about 22.5 USD per tC. Tables 3 describes some results of trading under uncertainties. We see that marginal cost between and within each country on reducing reported emissions and on investing in monitoring are equal, also in accordance with the outline procedure. The marginal costs are 84.2 USD per tC, which is almost 4 times higher than in the case without uncertainties. This partially is explained by the need for uncertainties reduction.


63

Table 1. Baseline and pledge targets

Emissions

1990 Emissions

2009 Baseline

2020 Pledges Unc.

(% targets)

USA 6,135.20 7,017.30 422.3 5,898.50 15

Australia 416.2 536.1 573.1 470.4 10

Canada 592.3 720.6 765.5 576.5 15

EU27 5,564.00 5,129.60 4,670.60 4,451.20 15

Japan 1,272.10 1,340.10 1,199.10 1,154.40 15

Russia 3,326.40 2,190.20 2,481.00 2,661.10 25

Table 2. Trades without uncertainty

BAU20 Target20 Emiss Trades Marg.cost emiss

USA 6969 5899 6461.5 563.1 22.5

Australia 573 470 412.8 -57.6 22.5

Canada 766 577 652.1 75.6 22.5

EU27 4671 4451 4493.7 42.8 22.5

Japan 1199 1154 1298.7 144.3 22.5

Russia 2481 2661 1892.7 -768.2 22.5

Table 3. Trades with uncertainty

BAU20 Target20 Emiss Trades Marg.cost emiss

Marg.cost uncert

USA 6969 5899 6461.5 61.1 84.2 84.2

Australia 573 470 412.8 -37.4 84.2 84.2

Canada 766 577 652.1 124.1 84.2 84.2

EU27 4671 4451 4493.7 54.1 84.2 84.2

Japan 1199 1154 1298.7 196.1 84.2 84.2

Russia 2481 2661 1892.7 -398.0 84.2 84.2

References

Energy Business Review (2006): Volatility the only certainty in EU carbon market. Available at: http://www.energy-business-review.com/article_feature.asp?guid= FD09D7CA-3EFC-4229-BA86-1D968025DF5B)

Ermolieva, T., Y. Ermoliev (2005): Catastrophic Risk Management: Flood and Seismic Risks Case Studies. In: Applications of Stochastic Programming [S. W. Wallace and W. T. Ziemba (eds.)]. MPS-SIAM Series on Optimization, Philadelphia, PA, USA.

Ermoliev, Y., M. Michalevich and A. Nentjes (2000): Markets for tradeable emission and ambient permits: A dynamic approach. Environ. Res. Econ. 15, 39–56.

Ermoliev, Y., R. Wets (eds.). 1988. Numerical techniques of stochastic optimization. Computational Mathematics, Berlin, Springer Verlag.

Ermolieva T, Ermoliev Y, Fischer G, Jonas M, Makowski M (2010): Cost effective and environmentally safe emission trading under uncertainty. In: Coping with Uncertainty: Robust Solutions, K. Marti, Y. Ermoliev, M. Makowski eds), Springer-Verlag, Heidelberg, Germany, pp. 79-99.


64

Ermolieva, T., Ermoliev, Y., Fischer, G., Jonas, M., Makowski, M., Wagner, F. (2010): Carbon emission trading and carbon taxes under uncertainties. Climatic Change, DOI: 10.1007/s10584-010-9910-x.

Godal, O., Y. Ermoliev, G. Klassen and M. Obersteiner (2003): Carbon trading with imperfectly observable emissions. Environ. Res. Econ. 25, 151–169.

Hamal (2010): Reporting GHG Emmissions: Change in Uncertainty and Its Relevance for Detection of Emission Changes (forthcoming).

Hudz, H., M. Jonas, T. Ermolieva, R. Bun, Y. Ermoliev and S. Nilsson (2003): Verification times underlying the Kyoto Protocol: Consideration of risk. Background data for IR-02-066, International Institute for Applied Systems Analysis, Laxenburg, Austria. Available at: http://www.iiasa.ac.at/Research/FOR/vt_concept.html.

Lieberman, D., M. Jonas, Z. Nahorski and S. Nilsson (2007): Accounting for Climate Change: Uncertainty in Greenhouse Gas Inventories―Verification, Compliance, and Trading. Springer, Berlin, Germany. Available (as of 3 September 2007) at: http://www.springer.com/west/home/environment/air?SGWID =4-199-22-173719021-0 (cf.

http://www.ibspan.waw.pl/GHGUncert2004/schedule.htm for initial short-paper contributions).

Nahorski & Horabik (2010): Compliance and emission trading rules for asymmetric emission uncertainty estimates; Climatic Change, DOI: 10.1007/s10584-010-9916-4

Nahorski, Z., J. Horabik and M. Jonas (2007): Compliance and emissions trading under the Kyoto Protocol: Rules for uncertain inventories. In D. Lieberman, M. Jonas, Z. Nahorski, S. Nilson, eds., Accounting for Climate Change: Uncertainty in Greenhouse Gas Inventories – Verification, Compliance, and Trading, Springer Verlag, 2007, pp. 119-138. Short paper available at: http://www.ibspan.waw.pl/GHGUncert2004/schedule.htm.

Pickl et al (2010): The impact of uncertain emission trading markets on interactive resource planning processes and international emission trading experiments; Climatic Change, DOI: 10.1007/s10584-010-9912-8

Rypdal, K. and L-C. Zhang (2000): Uncertainties in the Norwegian Greenhouse Gas Emission Inventory’. Reports 2000/13, Statistics Norway, Oslo, Norway.

Rypdal, K and W. Winiwarter (2000): Uncertainties in greenhouse gas emission inventories evaluation, comparability and implications. Environmental Science and Policy, pp. 107-116.

Wagner, F., Amann, M. (2009): GAINS contribution to ETMA request #2B –v1, International Institute for Applied Systems Analysis (IIASA).

Winiwarter, W. and K. Rypdal (2001): Assessing the uncertainty associated with national greenhouse gas emission inventories: a case study for Austria. Atmospheric Environment, pp. 5425-5440.

Winiwarter, W. (2007): National Greenhouse Gas Inventories: Understanding Uncertainties versus Potential for Improving Reliability. In D. Lieberman, M. Jonas, Z. Nahorski, S. Nilson, eds., Accounting for Climate Change: Uncertainty in Greenhouse Gas Inventories – Verification, Compliance, and Trading, Springer Verlag, 2007, pp. 23-30.


65

Uncertainty and variability in corporate GHG inventor ies and reporting

Pedro Faria

Carbon Disclosure Project, 40 Bowling Green Lane, London, EC1R 0NE, United Kingdom [email protected]

Abstract

There are a growing number of initiatives, voluntary or regulatory driven, requiring companies to report their GHG inventory, usually in accordance with the GHG Protocol [1]. There is also an increasing concern with the figures provided, usually reported with no explicit mention to their uncertainty. This paper looks at uncertainty assessment and variability in corporate GHG inventory and reporting.

Keywords: uncertainty, variability, corporate GHG inventory, GHG Protocol

1. Introduction

Corporations play a central role in addressing the challenges of climate change: they shape and drive the economy, often presenting budgets (and GHG Emissions) that are far bigger than states and countries1; they have effective economic and political power and are considered, likewise, as part of the problem as of the solution to stop global climate change.

A corporate Greenhouse Gas (GHG) Inventory is one of the business responses to the challenges posed by climate change. It is a representation of a corporations (or company) GHG sources, sinks, emissions and removals, usually constructed to get a fair estimate of its GHG emissions and are usually made in accordance with the GHG Protocol [2] or ISO 14064 [3]. The GHG Protocol was launched in 2001 and is the de-facto standard for corporate inventories.

Inventories, or parts of it, can be used by multiple stakeholders for multiple purposes. Businesses can use them for internal purposes such as managing GHG risks and opportunities, for establishing baselines used for recognition of early voluntary action, identifying reduction opportunities, public reporting, etc [2]. Uncertainty about the information reported is an important parameter. Companies need to conduct assessment reports also for optimization of operations, maintenance policy and ultimately, to choose between different mitigation measures. In particular, when operating facilities within emissions trading schemes, uncertainty assessment of the figures of CO2 emitted are important, in particular in the presence of high CO2 prices, as even small uncertainties can millions of Euros. Finally, governments are already requiring uncertainty assessments in order for companies to demonstrate compliance with regulatory schemes.

However, despite the generalized use of a common protocol, there is an increasing concern with the figures provided by corporations, especially fears of lack of comparability due to variability of parameters, models and approaches, and the fact that this variability is seldom translated into an uncertainty assessment.

1 As an example see [1] where the fossil fuel emissions of RAO UESR are compared to the fossil fuel emissions of the United Kingdom.


66

This paper analyses the role of uncertainty assessment and variability in corporate GHG inventory. We distinguish between variability, related to changes in the properties of variables with certain dimensions (time, space, approach), from uncertainty attached to the property of the variable, related with lack of knowledge, measurement errors, etc. We also make a distinction between several types of uncertainty: regulatory, model, parameter and measurement.

Whenever adequate, data from the Carbon Disclosure Project (CDP) database2 will be presented to illustrate specific points and contextualize. The Carbon Disclosure Project is an independent not-for-profit organization holding the largest database of primary corporate climate change information in the world.

2. Uncertainty and variability in GHG corporate inventories

With the increasing demand and interest for GHG emissions data, there has been a considerable increase in the amount of work put in to the organization and analysis of country inventories, as well as their uncertainty [4] [5]. Interest about uncertainty of inventory data is growing for corporate GHG emission inventories.

One of the questions of the CDP data set is about uncertainty estimates on the total gross direct emissions (Scope 1) and the % response per uncertainty range reported is presented in Table 1. We will consider as an acceptable uncertainty threshold 10% and as a good one 5%. These values are aligned with the values used in the European Trading Scheme [6] for ad hoc approaches to monitoring. Outstanding in the answers to CDP is the big percentage of answers that responded a remarkable uncertainty figure for an inventory, less than 2%. A good level of uncertainty was reported by 30% and an acceptable level by 20%. Overall less than 16% have reported uncertainty figures higher than 10%, being that 8% reported other things such as “no uncertainty” and “unknown”.

Table 1. Estimates of the uncertainty range of GHG direct emissions reported to CDP

Uncertainty range % responses Cumulative % Less than or equal to 2% 27% 27% More than 2% but less than or equal to 5% 30% 57% More than 5% but less than or equal to 10% 20% 77% More than 10% but less than or equal to 20% 9% 85% More than 20% but less than or equal to 30% 3% 89% More than 30% but less than or equal to 40% 1% 89% More than 40% but less than or equal to 50% 0% 90% More than 50% but less than or equal to 60% 1% 90% More than 60% but less than or equal to 70% 0% 91% More than 70% but less than or equal to 80% 0% 91% More than 80% but less than or equal to 90% 0% 91% More than 90% but less than or equal to 100% 1% 92% Other 8% 100 %

2 Investor 2010 programme data set, cut at the date of 14th of July, which comprises 1817 different inventories from companies. To establish the representativeness of this data set, it should be mentioned that it is composed of 1817 companies headquartered in 51 countries and representing 154 sub-industry codes according to the Global Industry Classification Standard.


67

Considering the overall degree of knowledge available in the corporate GHG community about uncertainty, it is considered that these figures reflect an optimistic approach to uncertainty that also reflects the recent engagement of many companies into GHG accounting. Uncertainty analysis is not an easy subject and it will often require the use of external experts.

In close alignment with the GHGP [7] in this work we will distinguish between the following types of uncertainty

• Regulatory or scenario, this is uncertainty due to choice, because there are multiple methodological approaches and ways of organizing and presenting information due to. This type of uncertainty generates variability and lack of comparability, which can be translated as uncertainty about the final figures.

• Model uncertainty, this is uncertainty that is caused by simplifications of aspects that cannot be easily modeled, such as non-linear effects, lack of spacial or temporal discretization, etc, or to the proper application of estimation methodologies.

• Parameter and measurement uncertainty, related with the uncertainty in quantities defining a theoretical model that relates different variables and can be related with inaccuracies of approximations, reliability of data, instruments used, sampling, operator, environmental conditions, etc.

In order to produce an inventory and calculate a total emission figure corporations have at least three basic steps they have to follow. The first step is the definition of their organizational boundary (or company hierarchy), according to three different approaches: equity share; financial control and operational control [2]. This will define the structure of what will be monitored from the level of a parent company and sub-companies down to the individual facilities and sources.

The second step, is to define the operational boundary, this is, if it will measure direct emissions from within the boundary (Scope 1) or/and also indirect emissions (Scope 2 and 3 that, by definition, are outside the boundary). Scope 2 emissions are indirect emissions due to energy consumption within the boundary. Scope 3 emissions are all other indirect emissions and we will not address them in here.

The third step is to apply the individual calculation models or methodologies. Although the GHG Protocol provides tools to help companies calculate emissions, the exact methodologies are not prescribed and consequently can vary. The calculation methodologies can be applied at several different levels of the corporate structure mentioned above in step 1. Because corporations are often regulated at facility and source level, it will be advantageous to calculate the emissions at this level of data granularity. Also, if corporations have to report for multiple purposes, it will also be convenient to have the data used to calculate emissions in granular form, to allow for it. If not, regulatory variability can introduce inconsistency and uncertainty (error) because values might have to be “reused” or “re-calculated” using ad hoc transformations or methodologies. In the following section we will try to review the three steps described above and distinguish the several sources of uncertainty, as defined here.

2.1. Corporations organizational structure

The three approaches defined in [2] are in practice, besides boundary setting rules, consolidation rules for the GHG emissions data collected in the different parts of the organizational boundary. The relationship between the several entities within


68

the organizational boundary will be expressed, for consolidation purposes, either in terms of a percentage (percentage of shares) or as one/zero value (control/no control).

For a given time boundary, the following model would be used to calculate the emissions:

∑=

∗=m

ni,nni tfT

1, (1)

where Ti is the total emission of scope i,n represents a organisation within the structure and m the number of organizations within the structure, fn is a number between 0 and 1, representing either the control (or lack of) of that organization or the % of share ownership and tn,i are the sub total emissions of scope i. The expression is generally valid for each organizational boundary approach, but the parameters will vary for each approach. Although simple, one can consider regulatory, model and parameter and measurement uncertainty (m, fn, tn,i).

If the exact number of organizations within the boundary is not known, or if the relationships between the several organizations is particularly complex and intricate there might be uncertainty in m. Or you can know m exactly but lack means to know tn,i – as when you own an asset but do not have an effective mean to know the emissions that derive from its use by another entity. When deciding between the three different organizational boundary approaches, something that could be characterized as “regulatory uncertainty”, there will be variability in m and in fn. It is clear that not all companies report on the same boundaries (see Table 2, based on CDP 2010 data). Almost half (48%) of the companies follow a financial control approach and 38% for operational control, but this change according to industry sector, e.g. for Oil&Gas companies these numbers switch. More important a reasonable percentage (11%) report boundaries that can not be categorized as any of the three so, effectively creating issues with comparability and uncertainty about total figures reported.

Table 2. Approaches to organizational boundaries data reported

All sectors Oil & Gas Sector n.º of companies % n.º of companies %

Equity share 63 4% 3 4% Financial control 813 48% 29 40% Operational control 642 38% 39 54% Other 187 11% 1 1% Total 1705 100% 72 100%

Likewise fn will vary if the exact percentage of ownership is uncertain or misrepresented which can happen especially in certain markets where there is not the same level of transparency and scrutiny on corporate governance. Sometimes it might also not be linear to decide who controls a company if special shares exist (e.g. “golden” shares). Finally, the total emissions of scope i for an organization n, will have a probability distribution function associated to it that can be the result of regulatory, modelling or parameter and measurement uncertainties.

Furthermore, we can consider that the model needs to be applied consistently so that its values do not deviate from the expected true value for a given boundary approach, e.g. if financial control approach is being used, one should not consolidate emissions from assets that are operationally controlled. This can be considered more a discretization error than an uncertainty. To avoid this type of discretization errors, certain organizations


69

in the corporate GHG accounting space like the Carbon Disclosure Project3 and the Carbon Disclosure Standards Board are considering that the model should be extended to be represented as

∑=

∗=m

ni,nn,jj,i tfT

1, (2)

where j simply represents separate line items of a financial report that binds a certain value for fj,n according to traditional financing accounting rules. This is, the consolidation rule, that translates in to fn, should be as per consolidated audited financial statement. An example of a separate line item would be “Parent and subsidiaries under control of the parent”, “Joint ventures”, “Associates”, etc.

A common problem to both models is that of time boundary inconsistency, this is, adding figures that do not concern to the same time boundary, something that happens in practice due to the different requirements existent in different countries for reporting. This problem occurs also at a more disaggregate level, for instance, when calculating scope 2 emissions, and it originates in lack of time discretization of the input variables/parameters.

We consider that the full extent of the impact of the consolidation rules to the overall uncertainty of an emission figure is a topic that needs further research. It is a subject that will cross two very different traditions, one related with accounting and law and another related with engineering and science. As a ton of CO2 emitted is actually a physical quantity, other aspects of uncertainty assessment that are common in the accounting and financial tradition, such as fair value of assets and liabilities, depreciation, etc, are not relevant for GHG accounting4.

For simplification purposes, and while some analysis is conducted to evaluate regulatory, model and parameter uncertainty, that the only reasonable way uncertainty can be addressed for Tj,i is by considering the uncertainty associated with tn,i, the variable of the model. If this is the case, the final uncertainty can be calculated in a simple way, by the known expression of the uncertainty of a sum of independent variables

( ) ( ) ( )[ ]i

mn i,ni,nn

i T

tu*t*fTU

∑ == 12

(3)

or for the extended boundary model

( ) ( ) ( )[ ]i,j

mn i,ni,nn,j

i,j T

tu*t*fTU

∑ == 12

. (4)

2.2 Calculation models and methodologies

One specific calculation model can be applied in several different ways, originating what is often expressed as a calculation methodology. The recently introduced US EPA rule [7] requires reporting of greenhouse gas (GHG) emissions from large sources and suppliers in the United States lists more than 100 different calculation methodologies available for scope 1 emissions. Even the most common sources, e.g. stationary

3 See “Draft Framework for responding to the CDP Information Request” available at https://www.cdproject.net/en-US/Respond/Documents/CDP_draft_framework.pdf 4 Although for financial accounting the fact that there is a new asset/liability named “t CO2” creates specific challenges that are not yet fully addressed.


70

combustion, have several different calculation methodologies and there is no agreement on which one should be preferred.

For combustion processes there are two basic approaches to calculate emissions, one based on inputs and another on the outputs. Ackerman and Sundquist [4] have compared two US Power-Plant CO2 data sets, one that is based partly or entirely on monitoring of stack gases (output model) and another that is based on fuel consumption (input model) and concluded that although total emissions at country level differed only slightly (3.5% for electricity and heat exports and 2,3% for electricity generation only) absolute differences between individual facilities could be much larger (16,9% for electricity and heat exports and 25,3% for electricity generation only).

This illustrates the need to validate for each source or process type the different models used to calculate emissions, at least at the level of the facility. This is particularly relevant in cap-and-trade systems, where any considerable deviations can undermine the trust in the system and an uncertainty figure can be monetized.

Also it illustrates that two models that are equivalent at one scale might not be at another: the two models seem to produce equivalent results at national level, but do not agree, at least with the necessary accuracy, at the facility level. For corporate GHG inventories should any of these two methodologies be ruled out? This also can only be answered with further research. There is no explanation for the deviations observed at facility level. Corporate inventories can be very extensive and cover emissions that can rival with country economies. As such, even if one method could be ruled out as non-suitable for monitoring at facility level, it could still be acceptable for monitoring at corporate level if the scale of the inventory is big enough such as the case reported in [1] for RAO UESR GHG inventory (based on the input model). Because corporate GHG inventories vary widely in scale (Figure 1) it will be difficult to rule out a method.

Figure 1. Frequency distribution of reported direct emissions. On the horizontal axis is direct emissions (t CO2-e) and on the vertical axis, n.º of companies

It is difficult to propose a way to quantify this type of uncertainty into an uncertainty assessment. At best, this could be considered as a difference in calibration, so the values for one model could be corrected in relation to the numbers of the other. At present the proposal is to acknowledge this type of uncertainty, but actually disregard a quantification of its effects in the uncertainty assessment.

2.3 Global warming potentials

Global Warming Potentials (GWP) are usually considered in GHG corporate accounting as “uncertainty-less” parameters. Nonetheless, GWPs do have uncertainty


71

associated with their values, something that has been characterized as “scientific uncertainty” [6]. In this section it is not the intention to analyze the specific uncertainty of GWP but rather to ascertain the implications of variability due to choice of different GWP’s. This can be investigated by looking at the data that has been reported to the CDP. An early cut of CDP 2010 investor program data shows that the vast majority (75%) of the GWP reported are based on 100 year GWP, 4% reported non-100 year GWP and 21% referenced other sources and is not possible to determine which GWP is used. The 100 year GWP percentage can still increase if we consider that it is expected that many of the other sources will in fact use 100 year GWP. Furthermore, of the 4% that have used non-100 year GWP, it was also detected some reporting mistakes (wrong reference, but in practice using a 100 year GWP), which would further increase this percentage. This was expected as it is generally considered good practice to use 100 year GWP.

However, when it comes to what reference to select for the 100 year GWP, responses to CDP are no longer so homogeneous: 31% select Assessment Report 4 (AR4), 5% Third Assessment Report (TAR) and 43% the Second Assessment Report (SAR) and 22% report other references which were not tracked. The GHG Protocol and ISO 14062-1 have no rule on to how to choose GWP’s and thus, in practice, this introduces variability. It is curious that companies are using different values for GWP when SAR has been widely adopted by most regulators, as can be observed in Table 3.

Table 3. GWP adopted in main standards and regulations related to GHG accounting and reporting (* PAS2050 has the explicit criteria that the most recent GWP has published by IPCC should be used)

GHG protocol No criteria

ISO 14064 No criteria

EPA Mandatory GHG Reporting (CFR 40) SAR-100

EUETS Directive SAR-100

The Climate Register SAR-100

Regional Greenhouse Gas Initiative Model Rule TAR-100

Australia National Greenhouse and Energy Reporting Guidelines SAR-100

BillantCarbonne SAR-100

PAS2050 4AR – 100 *

The Climate Leaders (EPA) SAR-100

Is it possible to get an estimate of what is the relevance GWP variability and how much it can represent in a total figure? The significance of this variability depends on the relative amount of non-CO2 gases to the amount of CO2 that make the inventory as well as of the variability of 100 year GWP reported in SAR, TAR and AR4 presented in Table 2 for some of the most commonly reported gases.

Overall, for the CDP sample, the amount of CO2-e coming from non-CO2 gases was found to be very small: of the companies that disaggregate their Scope 1 emissions per gas, 98% are direct CO2 emissions. Only a very small percentage of the emissions come from non-CO2 related gases. The most significant GHG after CO2 is methane, accounting for 1,9%. Using the values in Table 4 recalculation of emissions for 10 companies (from diverse sectors) using the different SAR, TAR and 4AR GWP’s was done. On average the total scope 1 would vary 1% due to variation of GWP for methane, nitrous oxide and


72

sulphur hexafluoride, with a maximum value observed of 7%. Based on this limited analyses it seems that GWP is an important variable to control when doing comparison on a company basis – in particular when the amount of non-CO2 gases is relatively high. However, it might not be relevant in analyzes of aggregate emissions figures as the amount of non-CO2 GHG might become to small.

Table 4. 100 year GWP’s for four common gases for SAR, TAR and AR4 and the % variability in relation to SAR

SAR TAR AR4

% ∆ (TAR/SAR)

% ∆ (AR4/SAR)

Carbon dioxide 1 1 1 - -

Methane 21 23 25 9,5% 19,0%

Nitrous oxide 310 296 298 -4,5% -3,9%

Sulphur hexafluoride 23.900 22.200 22.800 -7,1% -4,6%

References

[1] Dudek D., Golub A., Safonov G., Saparov M. (2002): Emission inventory at company level: lessons from Russia. Mitigation and Adaptation Strategies for Global Change 7: 155–172, 2002.

[2] WBCSD, WRI (2004): Greenhouse Gas Protocol – A Corporate Accounting and Reporting Standard, Revised Edition

[3] ISO 14062-1:2006, Greenhouse gases — Part 1: Specification with guidance at the.organization level for quantification and reporting of greenhouse gas emissions and removals

[4] Ackerman K.A., Sundquist E.T. (2008): Comparison of two U.S. Power-Plant Carbon Dioxide Emissions Data Sets. Environmental Science and Technology, Vol. 42, n.º 15

[5] Marland G., Brenkert A., Olivier J. (1999): CO2 from fossil fuel burning: a comparison of ORNL and EDGAR estimates of national emissions. Environmental Science & Policy, n.º 2, pp. 265-273

[6] EC (2007): Commission Decision of 18 July 2007 establishing guidelines for the monitoring and reporting of greenhouse gas emissions pursuant to Directive 2003/87/EC of the European Parliament and of the Council. 2007/589/EC

[7] GHGP (2003): GHG Protocol guidance on uncertainty assessment in GHG inventories and calculating statistical parameter uncertainty, available at http://www.ghgprotocol.org/calculation-tools/all-tools


73

Uncertainty of BAU emissions in LULUCF sector: Sensitivity analysis of the Global Forest Model

Mykola Gusti1,2

1 Lviv Polytechnic National University, 12 Bandery str., Lviv, 79013, Ukraine 2 International Institute for Applied Systems Analysis (IIASA),

Schlossplatz 1, Laxenburg, A-2361, Austria [email protected]

Introduction

Effectiveness of measures on reduction of greenhouse gas emissions from deforestation (or enhancement of sink in case of afforestation or reforestation) is measured against emissions generated under business-as-usual (BAU) scenario. As the BAU scenario is defined (definition of BAU scenario for a socio-ecological system is not considered in this study) the next problem is estimation of BAU emissions. Some countries use their own models, some countries do not have resources to build such models. In order to obtain coherent estimations across countries fine-scale global or regional models should be used. Global Forest Model (G4M) developed at the International Institute for Applied Systems Analysis is a geographically explicit landuse change and forestry model that is used for such estimates.

Uncertainty of the BAU emissions depends on uncertainty of the deforestation and afforestation/refforestation rates. The study is devoted to analysis of G4M parameters and sensitivity of the model to the parameter changes (only deforestation and afforestation/reforestation).

Global Forest Model

Global Forest Model (G4M, successor of DIMA model) developed in the International Institute for Applied Systems Analysis (IIASA) is used for estimation of BAU emissions in landuse, landuse change and forestry (LULUCF) sector as well as assessment of measures on emission reductions in the sector. G4M is a geographically explicit agent based model. The model version considered in the study operates on a regular 0.5x0.5 degree grid. G4M models deforestation, afforestation, and forest management aimed at satisfying exogenous wood demand. Forest in each cell is modelled with a forest of age structure averaged for a respective country, or a normal forest if the age structure information is absent or forest is unmanaged. Landuse change and forest management decisions are made for each grid cell. The landuse change decisions are made by comparing the net present value of forestry (sustainable production of wood during multiple rotation periods) and the net present value of agriculture. Deforestation occurs, if the net present value of agriculture together with benefits from selling wood after the clear-cut of the forest is greater than the net present value of forestry. Afforestation/reforestation occurs, if the environmental conditions are suitable for forestry and the net present value of forestry is greater than the net present value of agriculture. Deforestation and afforestation rates are functions of population density, gross domestic product (GDP) and agriculture suitability. Forest management decisions include use of variable rotation time, variable stocking degree, and thinning–no-thinning options. G4M and its applications are described in [1-3].


74

What do we model?

G4M forest area in 2000 is initialized using GLC2000 [4]. The vegetation classes used (forest classes 1-8, and two mosaic classes 9 and 17 partially) cover areas with open and closed forest with canopy >10% and tree height >3m. In general it is consistent with the forest definition used by the Food and Agriculture Organization (FAO)1.

G4M simulates deforestation (consistent with the IPCC definition: “the direct human-induced conversion of forested land to non-forested land“ [5]), afforestation and reforestation. Afforestation and reforestation are consistent with FAO definition2 [7] and the IPCC definition3, if the modelling is started in 1990). Afforestation and reforestation cannot be distinguished in the model inside one cell.

Landuse-change data

FAO Forest Resource Assessment (FRA) is a major data source on global and national levels. FRA provides global estimates for deforestation and afforestation, and national estimates for net forest area change since 1980s. The assessments are based on data provided by countries, besides remote sensing surveys of the tropics were used in FRA 2000 and FRA 2005 [8]. The recent FRA (FRA 2005) [6] provides the data for 1990, 2000 and 2005.

FRA 2010 providing recent landuse change data should be published soon. FRA 2010 uses remote sensing for estimating deforestation globally. We expect the estimates will be provided only globally and by ecozone (not for individual countries or spatially).

Deforestation and afforestation rates can be estimated from the country reports under the United Nations Framework Convention on Climate Change (UNFCCC) or the Kyoto Protocol, using such categories as Forest Land (FL), Other Land (L) and FL converted to L and L converted to FL. The land-use change data obtained from the reports are not as consistent among countries as the FRA data because countries report for different periods, also a limited number of countries provides regular reports (mostly Annex-I countries). However these data can be considered as more accurate than the FRA 2005 data because they are more recent and are updated regularly.

Grainger (2008) [9] analysed uncertainty of the forest cover data provided in the FRA series. In particular, he paid attention to substantial corrections of natural forest area that were published in the FRA series (FRA 1990, FRA 2000 and FRA 2005).

1 Land spanning more than 0.5 hectares with trees higher than 5 metres and a canopy cover of

more than10% or trees able to reach these thresholds in situ [6] 2 Afforestation is the establishment of forest plantations on land that, until then, was not classified

as forest. Implies a transformation from non-forest to forest. [7] Reforestation is the Establishment of forest plantations on temporarily unstocked lands that are considered as forests [6]

3 Afforestation is the direct human-induced conversion of land that has not been forested for a period of at least 50 years to forested land through planting, seeding and/or the human-induced promotion of natural seed sources.

Reforestation is the direct human-induced conversion of non-forested land to forested land through planting, seeding and/or the human-induced promotion of natural seed sources, on land that was forested but that has been converted to non-forested land. For the first commitment period, reforestation activities will be limited to reforestation occurring on those lands that did not contain forest on 31 December 1989. [5]


75

In many cases the corrections cannot be considered as improvements but alternative estimates. The net forest change for 2000-2005 published in FRA 2005 and obtained from the UNFCCC/Kyoto reports by the Annex-I countries differ by up to 557% (Table 1). The data in the Annex-I countries’ reports are subject to substantial corrections as well. Many countries report the same deforestation and afforestation rates for many years that do not look plausible.

Table 1. Net forest area change in 2000-2005 derived from the country reports under the UNFCCC or the Kyoto Protocol4, and the FRA 2005 data for the Annex-I countries.

Country

UNFCCC/Kyoto FAO

Net forest area

change

UNFCCC-FAO difference

Af-forestation,

ha/year

De-forestation,

ha/year

Net forest area

change, ha/year

ha/year %

Australia 80,339 354,258 -273,919 -193,000 -80,919 42

Austria 11,125 5,229 5,896 5,000 896 18

Belarus 1,274 2,341 -1,067 0 -1,067

Belgium 12,940 233 12,707 50,000 -37,293 -75

Bulgaria 59,147 0 59,147 9,000 50,147 557

Canada 7,796 34,187 -26,391 0 -26,391

Croatia 3,782 0 3,782 1,000 2,782 278

Czech Republic 1,951 699 1,252 2,000 -748 -37

Denmark 1,988 495 1,492 3,000 -1,508 -50

Estonia 11,698 4,548 7,151 8,000 -849 -11

Finland 8,137 11,733 -3,595 5,000 -8,595 -172

France 96,025 38,056 57,969 41,000 16,969 41

Germany 19,880 7,703 12,177 0 12,177

Greece 1,848 172 1,677 30,000 -28,323 -94

Hungary 8,293 511 7,782 14,000 -6,218 -44

Iceland 2,033 0 2,033 2,000 33 2

Ireland 17,120 75 17,045 12,000 5,045 42

Italy 77,068 682 76,386 106,000 -29,614 -28

Japan 1,511 12,094 -10,582 -2,000 -8,582 429

Latvia 3,064 0 3,064 11,000 -7,936 -72

Liechtenstein 6 0 6 0 6 -

Lithuania 7,053 0 7,053 16,000 -8,947 -56

4 Grassi G. and Pilli R. Joint Research Center, Italy, 2010, personal communication.


76

Luxembourg 468 376 91 0 91

Monaco 0 0 0 0 0 0

Netherlands 2,559 1,992 567 1,000 -433 -43

New Zealand 41,072 1,217 39,856 17,000 22,856 134

Norway 34,603 13,700 20,903 17,000 3,903 23

Poland 16,067 621 15,445 27,000 -11,555 -43

Portugal 32,459 63,658 -31,199 40,000 -71,199 -178

Romania 2,274 4,800 -2,526 1,000 -3,526 -353

Russian Federation 310,047 33,367 276,680 -96,000 372,680 -388

Slovakia 705 323 383 2,000 -1,617 -81

Slovenia 972 397 574 5,000 -4,426 -89

Spain 59,306 540 58,766 296,000 -237,234 -80

Sweden 14,664 9,934 4,730 11,000 -6,270 -57

Switzerland 2,053 1,637 415 4,000 -3,585 -90

Turkey 47,648 0 47,648 25,000 22,648 91

Ukraine 81,483 0 81,483 13,000 68,483 527

United Kingdom 15,396 1,046 14,350 10,000 4,350 43

United States of America

664,934 0 664,934 159,000 505,934 318

Hansen et al. (2010) [10] recently released a global map of gross forest cover loss for

2000-2005 with the resolution 18.5x18.5m km. The researchers used combination of MODIS (Moderate Resolution Imaging Spectroradiometer) and ETM+ (Landsat Enhanced Thematic Mapper Plus). Forest is defined as 25% or greater canopy closure at 30x30m pixel for trees higher than 5m. The map shows the gross forest cover loss without distinction of causes for the loss (i.e., logging, fire, windfall, etc.). Only in the tropics (except the Congo Basin) the gross forest cover loss during the observation period (2000-2005) is mostly attributable to conversion of forest land to agriculture land, i.e. deforestation [10]. Fine scale spatial data on deforestation for longer period is available also for the Amazon [11].

There are a few issues that make it difficult to use the spatially explicit data for validation of G4M performance on national or spatial scales: 1) The forest definition is narrower than the FAO or UNFCCC definitions; 2) The data contain information on land cover but not landuse. The data can be used for calibration or validation of the model only in some regions (like Amazon) where the landcover change is close to landuse change.

Model sensitivity to the input data

In order to find the model parameters which influence the results most of all we run the model for each parameter changed ±10% around the default value one by one. The results are presented in Table 2.


77

G4M is most sensitive to GDP, combined wood density and carbon content, net primary production (NPP), share of NPP stored in wood, and population density.

GDP in the base year is initialised on national scale using statistical data and can be treated as well known. Downscaling is done taking into account differences in urban and rural areas [12] with unknown uncertainty. GDP for the cells vary between the IPCC SRES scenarios by up to 160% in 2050.

Combined wood density and carbon content (ftimber) depends on tree species, tree age and environmental conditions. In the EU countries national average for ftimber ranges from 3.5 to 4.8 m3/tC (-13% and 20% around default value 4 m3/tC). We suppose that the spatial variability is higher. Detailed data (including spatial data) for estimation ftimber are not available in many countries.

Table 2. G4M sensitivity to variation in the input data

Parameter varied ±10%

Parameter description

deforRate change, % afforRate change, %

2000 2025 2050 2000 2025 2050

popdence_p10

Population density

-10.5 -3.3 10.3 -41.8 -1.2 31.1

popdence_m10 0.1 -23.1 -26.6 46.5 -3.1 -13.5

agrosuit_p10

Agriculture suitability

5.3 3.6 2.5 1.2 0.4 -0.3

agrosuit_m10 -6.0 -4.0 -2.5 -1.1 -0.5 0.4

discount_p10

Discount rate

0.3 0.8 0.7 -0.2 -0.4 -0.3

discount_m10 -0.4 -1.2 -1.0 0.4 0.5 0.4

NPP_p10

Net primary production

17.5 15.3 7.2 -5.0 -2.6 1.3

NPP_m10 -17.0 -9.8 -7.7 4.1 0.6 0.0

gdp_p10

Gross domestic product

-21.2 15.9 47.9 51.9 -12.6 -16.9

gdp_m10 19.9 35.8 60.2 -59.9 -70.7 -55.6

ftimber_p10 Combined wood density and carbon content

19.3 17.7 8.1 -5.0 -2.4 1.8

ftimber_m10 -19.5 -12.1 -9.0 4.1 0.4 -0.2

plantingcosts_p10

Planting costs

5.6 3.2 0.5 -2.1 0.1 0.6

plantingcosts_m10 -5.3 -3.3 -0.7 2.9 0.0 -0.2

fcuptake_p10 Share of NPP stored in wood

17.5 15.3 7.2 -5.0 -2.6 1.3

fcuptake_m10 -17.0 -9.8 -7.7 4.1 0.6 0.0

biomass_p10

Aboveground biomass

2.6 2.7 1.4 0.0 0.1 0.5

biomass_m10 -2.3 -2.6 -1.8 0.0 -0.2 -0.5

harvloos_p10

Harvest losses

2.9 2.4 1.5 -1.7 -0.7 -0.9

harvloos_m10 -3.3 -2.5 -1.4 2.3 0.6 0.9

lprice_m10

Agriculture land price

-1.7 -1.6 -3.6 3.2 1.6 2.7

lprice_p10 1.3 1.7 2.3 -1.2 -1.6 -2.4

wprice_m10

Stumpage wood price

7.6 4.9 2.9 -3.6 -1.8 -2.2

wprice_p10 -6.4 -4.0 -5.1 5.3 1.4 2.4


78

We consider three spatial datasets of NPP, which were used in G4M – average of 17 model outputs from the NPP model intercomparison [13], result of TsuBiMo model [14] and remote sensing derived NPP [15]. Average difference between the NPP estimates for the cells is about 50% for [13] vs. [14]; 105% for [13] vs. [15] and 518% for [14] vs. [15]. The share of NPP stored in wood (fcuptake) determines the rate of carbon accumulation in forest. In the model a constant value is used.

We suppose the uncertainty of population mapping is fairly good in the base year but the projections (SRES scenarios) differ by up to 1000% on average for the cells.

Model performance on different scales

We cannot estimate performance of the model on global scale because of no data available yet. We expect the FAO Forest Resource Assessment 2010 will be published soon that will allow us to validate the model (afforestation and deforestation rates) on global level and partially on national level (net forest area change).

Landuse change data derived from the countries’ reports to the UNFCCC can be used for model validation on national level. The data in the reports have two year lag therefore we expect 2010 data to be published in 2012. Taking into account the differences in the net forest area change values presented in the FAO FRA and derived from countries’ reports to the UNFCCC either FAO or UNFCCC data must be used both for calibration and validation of the model.

As shown in Soares-Filho et al. (2006) [16] for the Amazon distance to previously deforested land and distance to roads are the strongest predictors of deforestation. The parameters are not used in the G4M version considered in the study. Projection of road development is not available globally thus it is not feasible to include distance to roads as a parameter of a global model. One of the solutions is limiting spatial resolution of the model to the scale where the roads are not “visible” (about 0.5-1 deg) and population density can be a proxy for road network density.

Summary

The object of modelling (deforestation and afforestation/reforestation) is estimated with a high uncertainty on global, national and grid scales that makes difficult the calibration and validation of the model. Reliable data on deforestation and afforestation that are consistent among countries do not exist. At the time the paper has been written the data (i.e., FRA 2010 and/or countries’ reports to the UNFCCC containing 2010 data) required to validate the model on global and national scales for the period 2005-2010 were not available. We expect FRA 2010 to be published at the end of 2010 or early 2011.

Spatially explicit data on forest area change that can be used as a proxy for deforestation are available for some regions, e.g., Amazon. The data can be used for improving the model performance on the grid scale in the tropics.

The model is most sensitive to the following input data: GDP, combined wood density and carbon content, NPP, share of NPP stored in wood, and population density. Historical data on population density can be estimated with sufficient confidence. The GDP, combined wood density and carbon content can be estimated fairly well for the past on national scale using statistical data but downscaling to the grid scale is quite uncertain. NPP is very uncertain on the grid level. Projection of all the parameters is very


79

uncertain on global, national and spatial levels. Taking into account the parameter uncertainties and the model sensitivities to the parameters we can conclude that projected deforestation and afforestation/reforestation are very uncertain as well especially on the grid scale.

The next step in the model uncertainty analysis will be Monte-Carlo simulation of the parameter uncertainties to study interaction of the uncertainties.

References

1. Kindermann G., Obersteiner M., Sohngen B., Sathaye J., Andrasko K., Rametsteiner E., Schlamadinger B., Wunder S. and Beach R. Global cost estimates of reducing carbon emissions through avoided deforestation. Proceedings of the National Academy of Sciences of the U.S.A. (PNAS), 2008, vol. 105, N 30, pp. 10302–10307. http://www.pnas.org/cgi/doi/10.1073/pnas.0710616105

2. Kindermann G., Obersteiner M., Rametsteiner E. and McCallcum I. Predicting the Deforestation–Trend under Different Carbon–Prices. Carbon Balance and Management, 2006, pp. 1:15; doi:10.1186/1750-0680-1-15

3. Gusti M., Havlik P., Obersteiner M. Technical Description of the IIASA Model Cluster. The Eliasch Review; Office of Climate Change, UK, 2008

4. The Global Land Cover Map for the Year 2000. GLC2000 database. European Commission Joint Research Centre. http://www-gvm.jrc.it/glc2000

5. Annex to decision 16/CMP.1, Land use, land-use change and forestry. http://www.redd-monitor.org/wordpress/wp-content/uploads/2009/09/Kyoto_COP001_016.pdf

6. Global Forest Resources Assessment 2005. Food and Agriculture Organization, Rome, Italy, 2006.

7. Global Forest Resources Assessment 2000. Food and Agriculture Organization, Rome, Italy, 2001.

8. Submission by the Food and Agriculture Organization of the United Nations REDUCING EMISSIONS FROM DEFORESTATION IN DEVELOPING COUNTRIES http://www.fao.org/forestry/11262-1-0.pdf

9. Grainger A. Difficulties in tracking the long-term global trend in tropical forest area. Proceedings of the National Academy of Sciences of the U.S.A. (PNAS), 2008, vol. 105 no. 2, pp. 818-823 http://www.pnas.org/content/105/2/818.full.

10. Hansen M., Stehman S., Potapov P. Quantification of global gross forest cover loss. Proceedings of the National Academy of Sciences of the U.S.A. (PNAS), 2010, vol. 107 no. 19 8650-8655 http://www.pnas.org/content/107/19/8650.full

11. Deforestation estimates in the Brazilian Amazon. INPE (Instituto Nacional de Pesquisas Espaciais). Sao Jose dos Campos, 2003 http://www.obt.inpe.br/prodes/

12. Grübler A., O'Neill B., Riahi K., Chirkov V., Goujon A., Kolp P., Prommer I., Scherbov S., Slento E. Regional, national, and spatially explicit scenarios of demographic and economic change based on SRES. Technological Forecasting and Social Change, 2007, vol. 74, Issue 7, pp. 980-1029


80

13. Cramer, W., D.W. Kicklighter, A. Bondeau, B. Moore III, G. Churkina, B. Nemry, A. Ruimy, A.L. Schloss and the Participants of the Potsdam NPP Model Intercomparison Comparing global models of terrestrial net primary productivity (NPP): Overview and key results. Global Change Biology, 1999, vol. 5(S1), pp.1–15.

14. Alexandrov G.A., Yamagata Y., Oikawa T. Towards a Model for Projecting Net Ecosystem Production of the World Forests. Ecological Modelling, 1999, vol. 123, pp.183-191. http://dx.doi.org/10.1016/S0304-3800(99)00128-3

15. Running S., Nemani R., Glassy J., Thornton P. MODIS daily photosynthesis (PSN) and annual net primary production (NPP) product (MOD17). Algorithm Theoretical Basis Document. Version 3.0, 29 April 1999

16. Soares-Filho B., Nepstad D., Curran L., Cerqueira G., Garcia R., Ramos C., Voll E., McDonald A., Lefebvre P., Schlesinger P. Modelling conservation in the Amazon basin. Nature, 2006, vol. 440, pp. 520-523. doi:10.1038/nature04389


81

Spatial cadastres of GHG emissions: Accounting for uncertainty

Khrystyna Hamal, Rostyslav Bun, Nestor Shpak, Olena Yaremchyshyn

Lviv Polytechnic National University str. S.Bandery 12, Lviv, 79013, Ukraine

[email protected]

Abstract

An approach for spatial greenhouse gas emissions inventory based on official fuel statistics and digital maps in Energy and other sectors is presented. Using the developed geo-information technology for spatial GHG emissions analysis the territorial distribution of emissions in the level of elementary plots 2 km x 2 km for the territory of Western Ukraine is obtained. Uncertainty of inventory results is calculated using the Monte-Carlo approach and results of sensitivity analysis are described.

Keywords: spatial GHG emissions inventory, uncertainty, geo-information system.

1. Introduction

An integral part of greenhouse gas inventory process is uncertainty estimation. The problem of high quality uncertainty estimates of greenhouse gas emissions’ inventory results is extremely important for implementation of mechanisms under the Kyoto Protocol (such as Emissions Trading, the Clean Development Mechanism and Joint Implementation) and in the processes of establishing new environment protection treaties.

International agreements towards reduction of greenhouse gas emissions deal with emission and absorption estimates on the country scales and therefore uncertainty estimates of country’s total emissions is of interest. On the other hand it is desirable for governmental bodies of every country to have a tool, which would enable to analyse the separate constituents of many-sided processes of greenhouse gas emissions and absorptions and to obtain the integrated information on the actual spatial distribution of GHG sources and sinks, and thus to find the optimum ways of solving a number of economic or environment protection problems (Bun et al., 2007). Besides referring corresponding emissions to the places where they actually occur gives possibilities to largely improve the inventory process and to reduce uncertainty of overall inventory results.

This article discusses bottom-up inventory analysis. The approaches to modelling geo-referenced cadastres of emissions in Energy and other sectors are described as well as methods of uncertainty reduction using the knowledge of spatial GHG emissions distribution.

2. Spatial greenhouse gas inventory

The results of spatial GHG emissions modelling are data on emission values for a certain time period which additionally contain the information on geographical coordinates of a certain territory under investigation. For climatic models and for the analysis of territorial distribution of total emissions it is desirably to have emission estimates on the level of elementary plots with equal rather small area. The grid cell’s size depends on the inventory purpose and total size of territory under investigation.


82

For example, cutting the territory into cells 30 km x 30 km is reasonable for the territory of the whole country but such cells wouldn’t properly reflect the characteristics of emissions distribution in case of inventory for one city or administrative region.

The spatial GHG emission inventory for a certain territory consists of carrying out inventory for each grid cell in turn using the “bottom-up” approach and further summing up the inventory results for all activity subsectors. GHG emission level from a certain economical activity in one grid cell is in turn a sum of emissions from all emission sources, which are fully or partially located within its borders. In order to build spatial cadastre of certain gas emissions it is necessary to calculate territorially distributed specific emissions of this gas. Such specific emission values are calculated using parameters and data that define the emission process for the selected activity and also take into account geographical location of emission sources. This is, for each anthropogenic activity the specific GHG emission is a function of activity intensity parameters in certain territory and time period, appropriate emission coefficients, geographical coordinates of the territory under investigation and time.

2.1. Point, line and area emission sources

According to the internationally approved GHG emission inventory methodology the Energy sector or any other sector consists of a number of subsectors that in turn may be further disaggregated into separate emission source groups (2006 IPCC). Within a separate cell dissimilar emission sources are located – big and small in size, mobile and stationary etc.

To conclude spatial analysis it is suitable to categorise all emission sources into three groups – line, area and large point sources of emissions (Figure 1). GHG emissions modelling approaches for each of them differ significantly.

Large emission sources with significant emissions and relatively small area belong to large point sources. As an example, power stations, big industrial objects, refinery plants belong to this group. In case that the inventory is carried out for administrative regions, units or the country as a whole these emission sources among others are introduced with points.

Large point sources are necessary to be exactly located in space and the corresponding emissions that occur during their functioning are to be directly located to the point in space using geographical coor-dinates of location of the corres-ponding source (plant). This approach demand availability on the level of individual plant of the information about activity data (amount of fuel used in technological process, amount of industrial production sold etc.) and additional parameters for GHG emissions inventory

Figure 1. Example of classification of GHG emission

sources in separate cell into three types of sources: large point sources, line and area sources


83

that influence emission coefficients (for example, age and productivity of equipment on a certain plant, chemical characteristics of fuel used, detailed information on technological process, efficiency of emission control equipment etc.).

To the line emission sources belong sources of GHG emission to the atmosphere which are represented in the form of lines in spatial coordinate system. Transport roads and railways, oil and gas pipelines are further treated as line emission sources. Spatial modelling of emissions for line objects is conducted by cutting the line emission source into fragments using the grid cell overlapping the road or pipeline network. Later for each fragment of line emission source the corresponding emissions are calculated taking into account a number of parameters which define the level of these emissions (for example, road category, to which belong the road segment which is analysed; day or annual traffic capacity; distance from settlements for road segments; presence of railway stations for railways etc.)

To area sources belong emission sources where emissions occur from the surface that occupies a certain area. Agricultural fields, forests, oceans, seas etc. are examples of area emission/absorption sources. To area sources it is also reasonable to include territories where a big number of small area or line emission sources are concentrated. For example, in this paper area sources also encompass urban road transport network (because of high density of roads and streets), households, territories where agricultural and building works are conducted, small enterprises and plants, small boiler plants, territories where coal, oil or gas are extracted etc.

Spatial approach in GHG inventory enables to take into account the specificity of economic activity for separate, rather small territories, methods and technology of fossil fuel combustion in different economic sectors, technological specificity of extraction and refinement of primary fuels, availability and efficiency of cleaning installations etc. Therefore comparing to “traditional” GHG emission inventory (basing on aggregated data on the level of the whole country) the spatially referenced GHG emission inventory may have significant impact on the accuracy of the total emission estimates (Bun et al., 2007).

2.2. Greenhouse gas emissions inventory from mobile sources

Emissions from all transport types refer to GHG emissions from mobile sources. During fuel combustion in transport the direct acting GHG emissions occur - carbon dioxide, methane, nitrous oxide etc. that increase greenhouse gas effect.

In road transport sector the sources of emissions are the automobiles that are functioning on the roads. As it is practically impossible to investigate emissions from separate car, it is reasonable to interpret roads and highways where automobiles are operating as GHG emission sources. According to the classification methodology presented in previous section these sources belong to line emission sources. Urban road network is treated as area source because of very high density, and only main urban roads are separately treated as line sources.

Statistical information referring use of fossil fuels in road transport sector is available in the level of administrative units and cities in yearbooks of fuel statistics (Transport 2009). Separate parameters that will be further used in the model of spatial GHG emission inventory are available from statistical issues that contain transport statistics and summarising yearbooks. In Ukraine depending on the administrative province the input data for greenhouse gas inventory are available either in the level of administrative regions and cities or at the level of the province as a whole.


84

In general the level of GHG emissions in a certain grid cell depends on the amount of fuel consumed by transport within its borders. That is, before the spatial GHG emission inventory from road transport it is necessary to disaggregate the amount of fossil fuel used by transport to concrete emission sources and further multiply fuel quantity with corresponding emission factors to obtain emission estimates for the certain GHG. All the fuel used within urban road network in a region is disaggregated directly to the territories of cities and suburban areas around cities. The suburban territories of three levels are built around administrative borders of each city – the first has width of half radius of city area, the second – one radius, and the third – one and a half part of city area radius. For big cities the corresponding information on fuel consumption on transport sector is gathered and this data is directly located to the territory, which occupy the city and suburban areas around it. For small cities desegregation of fuel used in transport is made proportionally to population density. The rest of fuel used in transport sector in a region is disaggregated by automobile roads (segments) of a region according to the developed algorithms (including main roads within settlements) taking into account the length and width of each road segment, its capacity and current state. The above approach of fuel disaggregation to the level of elementary plots foresees that the part of fuel that was bought in a settlement for transport purposes is used (burnt) within its borders (for the needs of internal urban transport), large part of fuel is used on automobile roads in suburban territories that are located within a certain distance from the administrative borders of settlement, and the rest fuel is used outside the settlements and located to the road segments according to the road maps.

Emissions for each source type (area and line sources) are calculated using bottom-up approach. The quantity of fuel used of a certain type (diesel, gasoline etc.) is multiplied with the corresponding emission factor, that differs for various automobile operation modes (cold and hot emissions), as well as for different automobile types and control systems. Additionally the age distribution of vehicles is taken into account as well as average speed of vehicles on different road segments (using digital maps of road network and their capacity) and within cities. The average speed of vehicles for a certain road segment is established according to the road type (urban street, rural, highway) by overlapping map of roads with settlements’ map.

Similarly emissions from off-road transport are calculated. Emissions from agricultural machinery and mobile equipment for building purposes belong to this sector. Using digital maps of agricultural fields the statistical data on fuel used for agricultural works are desegregated to these areas proportionally to the area size and ratio of agricultural production harvested on a certain area. Using the similar approach emissions are calculated for construction territories as an area sources. Using digital maps of railways, stations, and appropriate statistics of fuel consumption emissions are calculated for railway transport and allocated to the railway segments as line sources of GHG emissions.

2.3. Greenhouse gas emissions inventory from stationary sources

Emissions from stationary emission sources in energy sector and other sectors contain emissions from processes of heat and power production, oil refinery, heating of residential buildings, industry as well as fugitive emissions from oil, gas and coal extraction processes (IPCC 2006). Common feature for all these sources is that emissions should be directly located to the place they occur.


85

General approach to spatial emissions modeling in sector of heat and power production as well as oil refinery processes is that all sources should be classified to two types – large point sources and small territorially dispersed sources. For each distinguished large point source the information has to be collected on fuel consumption, technology of fuel treatment, implemented emission control systems, age of equipment, chemical characteristics of fuel used etc. Basing on this information GHG emissions are estimated and geocoded to the elementary cell using the address of a certain plant (power stations, big boiler plants, refineries etc.). Total amounts of fuel combusted on small, dispersed sources (small power stations, boiler plants) are located to settlements’ areas (area emission sources) where these sources are located proportionally to consumers presence or heat production.

In residential sector the emission sources are households. These kinds of sources belong to small and territorially dispersed. This is, in spatial emissions models for this sector sources are represented by territories of settlements and are classified as area sources. For most cities there is accurate statistics on fuel usage in residential sector and these data are directly located to the city territory. The rest fuel is distributed by settlements basing on fuel type, settlement type, population density, parameters of average fuel usage of certain type in rural and urban territories etc.

For the sector of oil, gas and coal extraction the digital maps are built of fuel mines with data on amount of fuel extracted in a certain year, method of extraction and additional information. Using the “national” emission coefficients and the information on mine characteristics the emissions are calculated separately for each mine and geocoded to the appropriate cells.

2.4. Geoinformation system

The geoinformation system is developed for practical implementation of algorithms of geo-spatial GHG emissions inventory, automatic building of corresponding digital maps, visual analysis of obtained results and uncertainty analysis.

The system uses tables of input data and input maps (digital maps of settlements, mines, roadways etc.) and according to the developed algorithms builds geo-referenced databases of inventory results. Each record in this database corresponds to a certain grid cell and contains the information about emission source types within the cell’s borders, the structure of emissions in cell by gas, fuel type and economic sector. The results may be visualised with the help of digital maps and various thematic layers, using which helps to extensively estimate the situation, localise the territories with the highest emission rates, investigate the structure of emissions and make effective decisions on emissions reduction. Figure 2 and Figure 3 show examples of thematic maps with inventory results for western region of Ukraine (102 550 km2 ).

3. Uncertainty analysis

Uncertainty in GHG emissions represents the lack of knowledge about the true value of emissions at a certain area. Uncertainties resulting from the assessment of GHG emissions largely depend on the method used, quality of input data, uncertainties from expert judgements etc. (2006 IPCC ).

Increasing of knowledge about the investigated process can help do decrease uncertainties. Therefore comparing to traditional inventory (on the scale of the whole country) the spatially referenced emissions have additional parameter – geographical


86

coordinates of their location and thus this provides new ways of analysis and reduction of uncertainty of total GHG emission estimates. Adding of new independent information about GHG emission processes for separate emission sources or groups of sources to the information about overall emissions leads to decrease of uncertainty of total results (Winawater 2007).

Figure 2. Specific emissions of СО2 from gasoline combustion by road vehicles in juridical

property (kg/km2, 2009)

Figure 3. Prism-map of specific direct acting GHG emissions, summarized by all subsectors of

Energy sector (2009, 4 km x 4 km; СО2-eqv., kg/km2, because of incompatibly high emission rates

at Burshtyn and Dobrotvir Power Plants the scale of power 0.4 is used)


87

3.1. Verification of spatial emissions disaggregation

Analysis of uncertainty level for separate elementary cells is practically impossible because of lack of all necessary input data and insufficient knowledge about the emission process on the level of relatively small grid cells. Such analysis is justifiable only for separate cells where large point emission sources are located. Verification of spatial emission modeling results on the level of separate grid cells may be carried out by comparing with alternative data sources, such as remote sensing maps, direct point measurements of GHG emissions concentrations etc. For the territory of Ukraine there are no such alternative emissions estimates on the level of relatively small grid cells. But verification of calculated emissions by cells also is possible using two alternative input data in emissions calculations. For example, in transport sector the GHG emission estimates were made using as main input data quantity of fuel used by fuel type, and distance travelled data. Figure 4 presents difference map of these two estimates for CO2, CH4, N2O emissions in Lviv region where each cell’s value is calculated as a relative distinction of emission estimates built for this cell using alternative data sources.

Figure 4. Relative difference between estimates of specific GHG emissions (СО2-eqv.) in road

transport sector calculated using fuel statistics and vehicle mileage parameter in 2007 (%)

With the help of such difference maps it is easy to reveal and localize errors in spatial emission modeling. Figure 4 shows that on level of elementary plots results obtained using two approaches (emission estimates calculated using fuel statistics and vehicle mileage statistics) are generally well agreed (difference do not exceed 24%) with the exception of one region (Starosambirsk region), for which statistical data on fuel used in transport sector do not correspond to vehicle miles travelled statistics in 2007.

3.2. Uncertainty evaluation and sensitivity analysis

Total uncertainty level of emission modeling results depends on uncertainties of all input parameters of emission model. These uncertainties are combined into total


88

uncertainty estimate of inventory results using appropriate statistical tools (IPCC 2006). For such analysis it is important to have independent uncertainty ranges for emission coefficients, statistical data and other parameters of inventory process.

In calculations mainly “national” uncertainty ranges for statistical data, emission coefficients and net calorific values were used, but in case of “national” data lack the default IPCC uncertainty ranges were implemented. The Monte Carlo approach was used for emissions and uncertainty ranges modeling on the scale of 2 km x 2 km cells. An assumption was made that all input parameters are statistically independent. Table 1 contains modeling results of emissions and uncertainty ranges by GHGs and economic sectors for Western Ukraine in 2008.

Table1. Uncertainty estimates of GHG emissions by economic sectors (Western Ukraine, 2008)

Sector Emissions Uncert-ty

СО2 СН4 N2O Total emissions, СО2 eqv.

Heat & power production,

refinery

E (kg) 20027,0 0,287 0,286 20122,1

U(%) -8.99..+9.45 -31,7.+.46,74 -68,37..+160.3 -9,02..+9,48

Industry E (kg) 1731,5 0,067 0,011 1736,2

U(%) -6,61..+6.85 -42,33.+.66,34 -58,97..+123.3 -6,62..+6,85

Road transport

E (kg) 5191,5 1,867 0,332 5332,8 U(%) -5,65..+5.71 -38,33.+.61,64 -42,97..+69.3 -5,72..+5,81

Railway and off-road transport

E (кг) 730,5 0,123 0,315 830,5

U(%) -12,35..+12.36 -34,72..+51,64 -58,87..+112.9 -14,53..+17,91

Residential E (kg) 8135,1 2,281 0,039 8195,0

U(%) -11,44..+11.92 -48,92..+81,64 -62,87..+128.9 -11,47..+11,94

Oil, gas, coal extraction

E (kg) 34,02 43,431 0,0012 946,43

U(%) -78,94..+81.72 -45,82..+48,54 -92,87..+328.9 -44,47..+46,94

Uncertainty ranges of total GHG emissions from Enersy sector in Western Ukraine

territory are as follows: • for СО2: - 5.76%..+ 6.02%; • for СН4: - 41.45%..+ 44.28%; • for N2O: - 36,93%..+60.12%; • for total emissions taking into account Global Warming Potential factor: -

5.74%..+ 5.97%. The highest uncertainty of total emissions is in processes of coal, gas and oil

extraction as well as in transport (with the exception of road transport) and residential sectors (table 1). Relatively high uncertainties refer to emissions in sector “heat and power production” (mainly because solid fuel domination).

The sensitivity of uncertainty of total emissions to change of uncertainties of input parameters – statistical data on economic activity, calorific values and emission coefficients is investigated. Figure 5 graphically shows results of analysis – sensitivity graphics of uncertainties of emissions estimates to improvement of accuracy of input parameters on P percent.

Results show that relative uncertainty of emission estimates for CO2 and total emissions in CO2-eqv. largely depend on uncertainty of statistical data and uncertainty of fuels’ calorific values. Uncertainty of total emissions stays almost unchangeable with


89

the change of uncertainty of N2O emission coefficients and is hardly correlated to improvement of knowledge about CH4 emissions processes. For example, with the reduction of uncertainty of CH4 emission coefficients by half the uncertainty of total inventory results is almost unchangeable but on the other hand uncertainty of overall CH4 emissions changes from 44% to 24%. Similar situation is on case of twice reduction of uncertainty of N2O emission coefficients – overall uncertainty of inventory results for all direct acting GHGs in CO2-eqv. doesn’t change, but there is considerable reduction of N2O emissions’ uncertainty from 60% to 35% (the upper bounds of 95% confidence intervals).

This is, in spite of large uncertainties of CH4 and N2O emission coefficients the improvement of their accuracy has no significant impact on uncertainty reduction of total emissions in CO2-eqv. The most efficient way for uncertainty reduction is improvement of accuracy of statistical data on fuel statistics and calorific values.

a) b)

c) d)

Figure 5. Dependence of total uncertainty of emission estimates to changes of uncertainty (on P %) of input parameters of inventory (Monte-Carlo approach):

а) СО2; b) СН4; c) N2O; d) total emissions

Reduction of uncertainty of statistical data by half leads to reduction of uncertainty of total emissions in Energy sector from 5.84% to 4.87%, the same reduction of calorific values’ uncertainty leads to reduction of overall uncertainty to 4.75%. Moreover, only improvement of accuracy of physical and chemical characteristics of coal that is used as fuel in power plants of Western Ukraine has significant impact on reduction of uncertainty of overall direct acting GHG emission cadastre.


90

Conclusions

The spatial analysis of GHG emissions allows receiving the important information about the actual location of anthropogenic sources of emissions at the regional level. Spatial integration of a distributed inventory on all elementary plots yields a generalized result of traditional inventory. This provides very useful possibilities for analyzing of separate constituents of uncertainty of inventory results and helps to find the most efficient ways for uncertainty reduction. The geo-information technology of GHG spatial inventory at the regional level is an effective tool for support of decision making on the actual problems of ecology and the environmental protection.

References

[1] Bun R., Hamal Kh., Gusti M., Bun A. Spatial GHG inventory on regional level: Accounting for uncertainty, Proc. of the 2nd Intern. Workshop on Uncertainty in Greenhouse Gas Inventories, Laxenburg, Austria, 2007, pp. 27–32.

[2] 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas Inventories Programme, Eggleston H.S., Buendia L., Miwa K., Ngara T. and Tanabe K. (eds), 2006.

[3] Transport Statistics of Lviv Region: Statistical Yearbook, Lviv, Main Statistical Agency of Lviv Region, 2009, 196 p.

[4] Winiwarter W. National greenhouse gas inventories: understanding uncertainties versus potential for improving reliability, / W. Winiwarter // Water Air Soil Pollution: Focus, 2007, v. 7, p. 443-450.


91

Improving resolution of spatial inventory with a statistical inference approach

Joanna Horabik, Zbigniew Nahorski

Systems Research Institute, Polish Academy of Sciences str. Newelska 6, 01-447 Warsaw, Poland

[email protected] [email protected]

Abstract

This paper presents a way of applying the so-called Chow and Lin method for disaggregation of spatial data. An example is used to illustrate the potential usefulness of the proposed technique. In the context of greenhouse gases, this extension might be particularly adequate to improve spatial resolution of inventories in the situation, when highly spatially resolved activity data are not readily available.

Keywords: Disaggregation methods; spatial inventory of emissions; statistical modelling

1. Introduction

Development of spatially distributed inventories of greenhouse gases (GHG) crucially depends on availability of highly spatially resolved activity data. For instance, in Poland activity data relevant to GHG emissions at present is available for 17 country regions, with no more accurate spatial resolution. Information of higher spatial resolution can be obtained for some land use and line emission sources which are related to GHG emissions. These are, however, only proxy data about activities.

We propose to apply spatial statistics methods to produce higher resolution activity data, taking advantage of more detailed land use information. Our approach extends the method proposed in [2] to disaggregate time series based on a related, higher frequency series. We extend this methodology to the case of spatially correlated data.

To model spatial dependence we use the conditional autoregressive (CAR) structure as a relevant statistical tool. Compare also [4] for another application of the CAR structure to model spatial GHG inventory.

The format of the article is as follows. Section 2 describes the disaggregation model, along with an estimation and prediction of the value of interest in a fine grid. The data example, which we present in Section 3, is not on GHG emissions, but on ammonia (NH3) emissions. Nevertheless, we believe the approach proposed can be of interest to GHG inventory preparers as well. Concluding remarks are given in Section 4.

2. The disaggregation framework

This section presents the statistical approach to the issue of spatial disaggregation. We specify the model and provide details on its estimation and prediction in a fine grid. It should be noted, however, that the approach requires knowledge on a variable of interest in a coarse grid, and on some related variables in a fine grid. Moreover, we need to make assumption about a residual covariance structure. Here we apply the conditional


92

autoregressive model, but other approaches (e.g. a geostatistical model) might be potentially used as well.

2.1. The model

We begin with the model specification in a fine grid. Let Yi denote a random variable associated with a missing value of interest yi defined at each cell i for i=1,...,n of a fine grid. Assume that random variables Yi follow Gaussian distribution with the mean µi and

variance 2Yσ

( ).,~| 2Yiii GauY σµµ (1)

Next we assume that given the values µi and 2Yσ , the random variables Yi are

independent, thus the joint distribution of Y=(Y1,..., Yn)T conditional on the mean process

( )Tnµµ ,...,1=µ is Gaussian

( )nYnGau IY 2,~| σµµ (2)

where nI is an identity nn× matrix.

Our approach to modelling the meaniµ expresses an assumption that available covariates explain part of the spatial pattern, and the remaining part is captured through a spatial clustering. For this, we make use of the conditional autoregressive model, which is given through specification of the full conditional distribution functions for i=1,...,n

( ) njiww

wGau

ij i

Tjj

i

ijTiijji ,...,1,,,~|

2

, =

−+ ∑

≠ ++≠

τµρµµ βxβx (3)

where ijw are the adjacency weights ( 1=ijw if j is a neighbour of i and 0 otherwise, also

0=iiw ); ∑=+ j iji ww is the number of neighbours of area i; ix is a vector containing 1

for the intercept 0β and k explanatory covariates of area i; ( )Tkβββ ,...,, 10β = is a vector

of regression coefficients and 2τ is a variance parameter. Given (3), the joint probability distribution of the process µ is as follows, see [1, 3]

( )( ),12 −W-Dβ, ρτX~µ nGau (4)

where X is the matrix with vectors Tix

=

nkn1

1k11

xx1

xx1

L

MOMM

L

X ;

D is an nn× diagonal matrix with +iw on the diagonal; and W is an nn× matrix with

adjacency weights ijw . Equivalently we can write (4) as

( )N0,~εε,Xβµ nGau+= , (5)

where ( ) 1−= WDN ρτ -2 . The model for a coarse grid (aggregated) observed data is obtained by multiplication

of (5) with an nN × aggregation matrix C consisting of 0’s and 1’s, indicating which cells have to be aggregated together


93

( )TNGau CNC0,~CεCεCXβCµ += (6)

where N is a number of observations in a coarse grid. We treat the random variable

Cµλ = as the mean process for random variables ( )TNZZ ,...,1=Z associated with

observations ( )TNzz ,...,1=z of the aggregated model

( )NZNGau IλZ 2,~| σλ . (7)

Thus, random variables NiZi ,...,1, = are conditionally independent

( )2,~| Ziii GauZ σλλ (8)

where [ ]ii λ=λ .

2.2. Estimation and prediction

With the maximum likelihood (ML) method we can estimate parameters 22 ,, τσ Zβ and ρ . First, from (6) and (7) we derive the joint unconditional distribution of Z

( )TNGau CNCMCXβZ +,~ , (9)

where NZ IM 2σ= , see e.g. [6]. Next, we formulate the log likelihood associated with (9)

( ) ( )

( ) ( ) ( )CXβzCXβz

CNCMβ

--

T

1

22

2

1

2log2

log2

1,,,

−+−

−+−=

TT

Z

NL

CNCM

πρτσ

where ⋅ denotes the determinant. With fixed 22 ,τσ Z and ρ , the above log likelihood is

maximised for

( ) ( ) ( ) ( ) ( ) zCNCMCXCXCNCMCXβ11122 ,,

−−−+

+= TTTT

Z ρτσ ,

which substituted back into the function ( )ρτσ ,,, 22ZL β provides the profile log

likelihood

( ) ( )

( ) ( ) ( ) ( )( )

( ) ( ) ( ) ( ) .

2log2

log2

1,,

111

1

111

22

+

+×

+×

+

+

−+−=

−−−

−

−−−

zCNCMCXCXCNCMCXCXz

CNCM

zCNCMCXCXCNCMCXCXz

CNCM

TTTT

T

TTTTT

Z

NL

-

-2

1-

T πρτσ

Further maximisation of ( )ρτσ ,, 22ZL is performed numerically, including the checks on

to ensure that the matrix WD ρ- is non-singular, see [1]. To obtain the standard errors of estimated parameters, one needs to derive the Fisher

information matrix. The asymptotic variance-covariance matrix of the ML estimators is


94

obtained by inverting the expectation of the negative of the second derivatives (the Hessian) of the log likelihood function, and the expectation is evaluated at the ML estimates. In other words, we use the expected Fisher information matrix to obtain the standard errors of parameters. Calculation of the Hessian with respect to the regression coefficients is relatively straightforward, but it becomes more burdensome for the covariance parameters. Due to a limited space of the article, detailed calculations of the explicit formulas for the expected Fisher information matrix will be provided elsewhere; here we report resulting standard errors of parameters.

Model estimation, however, does not provide missing values in a fine grid, which is our primary interest. This task needs to be accomplished with the prediction procedure. To derive the predictor, basically, we follow the steps presented in [4], accommodating adjustments to the present model.

Consider a random variable Y0 associated with missing emission values in a fine grid, and let 0µ denote its mean value. Assume that the distribution of 00 | µY is of

the form (1) and the distribution of µ|0µ is of the form (3). The predictor of the observation Y0, which is optimal in terms of the minimum mean squared error, is given by ( )z|0YE . In deriving this predictor we will make use of the conditional distribution of zµ | , which from (5), (7) and (9) is

( )WWVzµ ,Gau~| n , (10)

where ( ) 1−+= 1-1-T NCMCW and XβNzMCV -1-1T += . To develop the predictor

( )z|0YE we will make use of the following property of the conditional expected value:

( )000 | µYEY = and ( )µ|00 µµ E= . Thus,

( ) ( )[ ] ( ) ( )[ ]

( )

.|

|||||| 00000

∑ ∑

∑

+=

−+=

===

++

+

j jj

0

0jTj

0

0jT0

j

Tjj

0

0jT0

w

w

w

w-

w

wE

EEEYEEYE

z

zµzzz

µρρ

µρ

µµµ

Eβxβx

βxβx (11)

To calculate the rightmost expectation in the last equality of (11) we use the expression (10). Denoting the jth element of the vector WV with l j, we get the predictor in the form

( ) ( )∑ −ρ+=+j

Tjj

0

0jT0 l

w

w|YE βxβxz0 .

3. Data example

We illustrate the proposed procedure using real data set on gridded inventory of NH3 (ammonia) emissions from fertilization (in tonnes per year) reported in a northern region of Poland (the Pomorskie voivodship). Inventory grid cells are of regular 5km× 5km size,

and the whole of cadastral survey compiles n=800 cells, denoted ( )Tnyy ,...,1=y , see Figure 1. In addition, we have available CORINE land cover map for this region to be used as explanatory information. Specifically, for each grid cell we calculate area of these land use classes, which can be related to ammonia emissions. The following CORINE classes were considered (the CORINE class numbers are given in brackets):


95

Non-irrigated arable land (211), denoted ( )Tnxx 1,1,1 ,...,=1x

Fruit tree and berry plantations (222), denoted ( )Tnxx 2,2,1 ,...,=2x

Pastures (231), denoted ( )Tnxx 3,3,1 ,...,=3x

Complex cultivation patterns (242), denoted ( )Tnxx 4,4,1 ,...,=4x

Principally agriculture, with natural vegetation (243), denoted ( )Tnxx 5,5,1 ,...,=5x .

In what follows we will examine models with all the above classes (set 1), and compare the results with models including only Non-irrigated arable land and Complex cultivation patterns (set 2). Secondly, we compare a linear regression with independent (iid) errors vs. spatially correlated errors modelled by the CAR process. We consider the following models:

Model CAR1: - CAR errors, set 1 of covariates Model LM1: - iid errors, set 1 of covariates Model CAR2: - CAR errors, set 2 of covariates Model LM2: - iid errors, set 2 of covariates.

To examine the performance of the disaggregation procedure, first we aggregate emissions into 10km× 10km (coarse) grid cells, then we fit the model and predict ammonia emissions for a 5km× 5km (fine) grid. Finally, we check these results with original inventory emissions of a 5km× 5km (fine) grid. Thus, our simulation study tests the case of a quadruple disaggregation.

Figure 1. Ammonia emissions (in tonnes per year) in a 5km× 5km grid

Maximum likelihood estimates (denoted by Est.) and standard errors (denoted by

Std.Err.) of the parameters for each model are displayed in Table 1. In this table, we can observe that for all the models, the ML estimates of the regression coefficients are similar. From the ratio of regression coefficients and its respective standard errors


96

(i.e. the t-test statistic), we can roughly conclude that all the considered land use classes are statistically significant; in fact, in each case respective p-values proved to be less than 0.05 (not shown). Next let us turn our attention to the error part of the models.

Significantly lower values of 2Zσ estimates under both the CAR models compared with

their linear regression counterparts, indicate that greater variability is explained by the models with spatially correlated errors than the corresponding models with independent errors. As expected, among the spatially correlated models, both variance

parameters 2Zσ and τ2 are higher for CAR2 than for CAR1 model with five land use

classes as explanatory variables. Furthermore, we have parameter ρ, that reflects strength of spatial correlation (note that ρ=0 corresponds to a model with independent errors), see also [1] for more details. A value of parameter ρ is higher for CAR2 model, which illustrates that in models of limited explanatory power, the importance of spatial correlation becomes more pronounced.

Table 1. Maximum likelihood estimates

Results of the four models are also summarized using the Akaike criterion (AIC).

The idea of AIC is to favour a model with a good fit and to penalize for a number of parameters; models with smaller AIC are preferred to models with larger AIC. Table 2 displays AIC for each model, additionally we report the negative log likelihood (-L). Naturally, models with set 1 of covariates provide much better results than the models with another set. Among these respective sets, models with spatial structure considerably improve results obtained with models of independent errors. Note, that this improvement is higher for models with set 2 of covariates (797.6-742.8=54.8) than for models with set 1 of covariates (686.1-642.7=43.4).

Table 2. Model comparison

Model -L AIC CAR1 LM1 CAR2 LM2

312.3 336.1 365.4 394.8

642.7 686.1 742.8 797.6

CAR1 LM1 CAR2 LM2 Est. Std.Err. Est. Std.Err. Est. Std.Err. Est. Std.Err.

β0

β1

β2

β3

β4

β5

2Zσ

τ2

ρ

1.84e-02

1.12e-07

2.54e-07

9.63e-08

1.16e-07

1.26e-07

0.335

0.535

0.948

7.13e-02

4.14e-09

1.94e-07

1.29e-08

2.12e-08

1.37e-08

0.073

0.082

0.001

5.51e-02

1.06e-07

4.62e-07

1.04e-07

1.17e-07

1.31e-07

1.161

-

-

5.59e-02

3.61e-09

1.97e-07

1.32e-08

1.79e-08

1.21e-08

0.109

-

-

0.376

1.08e-07

-

-

1.22e-07

-

0.522

0.807

0.972

9.27e-02

5.27e-09

-

-

2.89e-08

-

0.112

0.122

1.74e-04

0.452

9.58e-08

-

-

1.60e-07

-

1.95

-

-

5.45e-02

4.43e-09

-

-

2.22e-08

-

0.184

-

-


97

Ammonia emissions predicted in a 5km× 5km grid ( )*iy are displayed in Figure 2.

Differences between the four models are negligible, although visual comparison with the original emissions in Figure 1 suggests that both the models based on set 1 of covariates (CAR1, LM1) provide slightly better results. Since the mapped values are binned into 9 classes, therefore some features might have been masked on maps in

Figure 2. To remedy this, in Figure 3 we present the model residuals ( )*iii yyd −= .

Now the difference in prediction results between the models is evident - the best results are obtained for CAR1 model and the worst for LM2 model.

Figure 2. Predicted ammonia emissions (in tonnes per year) in a fine grid

Figure 4 presents the scatterplot of predicted values *iy versus the observations yi for

each model. The straight line has slope 1, thus if the predicted values are close to the original data, points are close to the straight line. This setting, once again, illustrates much better explanatory power of models based on all the land use classes (set 1 of covariates). Moreover, it also shows importance of models with spatial CAR structure. In case of models CAR2 and LM2, introduction of spatial dependence evidently


98

improved accuracy of prediction. In case of models CAR1 and LM1, the applied spatial structure allowed to considerably limit a number of highly overestimated predictions.

Figure 3. Residuals from predicted values

The prediction error, i.e. residuals di are further analysed in Table 3. Namely, we calculate the mean squared error

( )∑ −=i ii yy

nmse

2*1,

which should be as low as possible. The mean squared error indicates how well a model predicts data. Moreover, we report in Table 3 also the minimum and maximum values of

di, and the sample correlation cofficient r between the predicted *iy and observed yi values. In terms of both the mean squared error and the coefficient r, the best model is CAR1 and the poorest model is LM2, following the previous assessments. However, we note that the remaining two models changed their ranks, that is, CAR2 model has lower mse and higher coefficient r than the linear model based on covariates set 1, i.e. LM1 model.


99

Table 3. Analysis of residuals ( )*iii yyd −=

Model mse min(di) max(di) r CAR1 LM1 CAR2 LM2

0.102 0.188 0.158 0.291

-2.144 -2.562 -1.917 -2.498

0.989 0.433 1.362 1.765

0.937 0.882 0.901 0.808

Figure 4. Predicted values (y*) vs. observed (y)

4. Concluding remarks

The major objective of this study was to demonstrate how a variable of interest (e.g. emissions) available in a coarse grid plus information on some related covariates available in a finer grid can be combined together to provide the variable of interest in a finer grid, and therefore to improve its spatial resolution. We proposed a relevant disaggregation model and illustrated the approach on a real data set. The idea stems from the method of Chow and Lin, originally designed for time series data. We applied it to


100

spatially correlated data, and spatial dependence was modelled with the conditionally autoregressive structure, introduced into a linear model as a random effect.

Clearly, performance of the proposed framework depends on at least three factors: (i) explanatory power of covariates available in a fine grid, (ii) strength of a spatial dependence within the predicted value, and (iii) extent of disaggregation. Only the first of these factors was evaluated in our case study, the remaining two were kept constant. The results indicate that, to some extent, inclusion of a spatial dependence structure can compensate for less adequate covariate information.

We note that in our simulation study we used original data in a fine grid to assess quality of resulting predictions. For the purpose of future applications, where original emissions in a fine grid are not known, the proposed disaggregation framework should be completed with a measure of prediction error.

Furthermore, future work will be devoted to application of the method to a case of greenhouse gas inventories.

Acknowledgements

Authors gratefully acknowledge provision of data for this case study from Ekometria – Biuro Studiów i Pomiarów Proekologicznych in Gdańsk, Poland.

References

[1] Banerjee S., Carlin B.P. and Gelfand A.E. (2004): Hierarchical modeling and analysis for spatial data. Chapman & Hall/CRC.

[2] Chow G.C. and Lin A. (1971): Best linear unbiased interpolation, distribution, and extrapolation of time series by related series. The Review of Economics and Statistics Vol. 53, No. 4, pp. 372-375.

[3] Cressie N.A.C. (1993): Statistics for spatial data. John Wiley & Sons, New York.

[4] Horabik J. and Nahorski Z. (2010): A statistical model for spatial inventory data: a case study of N2O emissions in municipalities of southern Norway. Climatic Change. doi: 10.1007/s10584-010-9913-7.

[5] Kaiser M.S., Daniels M.J., Furakawa K. and Dixon P. (2002): Analysis of particulate matter air pollution using Markov random field models of spatial dependence. Environmetrics Vol. 13, pp. 615-628.

[6] Lindley DV and Smith AFM (1972): Bayes estimates for the linear model. Journal of the Royal Statistical Society Series B, Vol. 34, pp. 1-41.


101

Compliance for uncertain inventories: Yet another look?

Olgierd Hryniewicz1, Zbigniew Nahorski1, Joanna Horabik1, Matthias Jonas2

1Systems Research Institute, Polish Academy of Sciences, Newelska 6, 01-447 Warsaw, Poland [email protected]

2International Institute for Applied Systems Analysis, Schlossplatz 1, A-2361 Laxenburg, Austria

Abstract

The paper deals with the problem of compliance under uncertainty, treating it from the view of comparison of uncertain alternatives. We emphasize inadequacy of using nominal emission estimates, as those obtained from emission inventories, when they are subject to high uncertainty. Meanwhile, there exists plenty of techniques of dealing with comparison of such uncertain estimates. Several of them are presented in the paper. Probabilistic and fuzzy approaches are considered and compared. Many of them can be adapted to check fulfillment of obligations on the basis of knowledge of uncertain emission estimates.

Keywords: greenhouse gases inventories, compliance, uncertain alternatives

1. Introduction

A handful of solutions have been proposed to cope with the problem of commitment verification for emission obligations in case of uncertain inventories, see [9]. Many of them pointed to methodological incompetence in using nominal (crisp) values in clearing pollutant emission targets. In practice, only highly inexact knowledge on emission values is available, as is the case of greenhouse gases, see e.g. [8, 10, 11].

To give an example, let us consider verification of a single emission inventory x against a given limit L , i.e. Lx ≤ . A distribution of an inventory uncertainty )(xµ may be asymmetric. See an example in Figure 1 depicting a histogram of emission inventory uncertainty for Austria, obtained by the Monte Carlo method, [16]. Let us suppose that an emission target for Austria is the one marked in Figure 1. The question of interest is whether this party fulfills its commitment, or not. Ignoring uncertainty, the answer is yes, as the nominal value of inventory lies below the target. However, although the nominal value is just below the target, it is more likely that the actual emission may be above the target, because most of the possible emissions (probability mass) is placed to the right of the target value. Can we then responsibly accept the answer yes?

Another example with simplified uncertainty distributions is depicted in Figure 2, with axis placed ad hoc. Let us consider two parties with known triangular distributions of emission uncertainties. The nominal inventories (the top values of the distributions densities) of both parties are very close to each other. Ignoring uncertainty, the party A will be considered compliant (fulfilling the commitment), while the party B will be considered noncompliant. However, confidence in the inventory value of the party B is high, while the confidence in the inventory value of the party A is much lower. Therefore, which party is more credible? Should the party A be considered compliant, while the party B should not?


102

Figure 1. An example of an asymmetric uncertainty distribution of a national inventory: compliant or noncompliant?

Figure 2. Which inventory is smaller, A or B?

It seems reasonable to require that decision on fulfillment of obligations should be fair among parties, in the way, that ordering of inventories should make it possible to decide which inventory outperforms others. From the above examples we can see that, when dealing with uncertain values of possibly asymmetric distributions, neither decisions on fulfillment of obligations nor comparison of inventories treated as certain values, need to be consistent with a common sense understanding of the uncertainty distributions.

For the greenhouse gases, reduction of inventory is often defined in percents, i.e. bc xx ρ≤ , where cx is an emission inventory in the compliance period, bx is

an emission inventory in the basic year (at the beginning of the reduction period), and ρ is a required fraction of emission reduction. Here, again, the task is to compare uncertain inventories in the compliance year, cx , with inventories reduced from the basic year,

bxρ , and to decide whether the former is lower than the latter.

In Section 2 we present different probabilistic- and fuzzy-rooted techniques, which can deal with the uncertain inventories. Section 3 sketches further works.

2. Compliance under uncertainty

2.1. Probabilistic approach

Comparison of uncertain random values has been already considered in various fields. The problem of selection from risky projects has a long history in such areas as finance, R&D projects, IT projects, [5], and several methods have been proposed to compare such


103

projects. The methods can be divided into groups. All the methods presented below are adopted to the considered problem of emission inventories.

In the sequel two uncertain inventories, A and B of Figure 2, will help us to illustrate the described techniques. The question to be answered is as follows: inventory of which party can be considered "`smaller"'. The answer can be either used for ordering inventories of two parties, or checking if cx is smaller than bxρ , i.e. verifying fulfillment

of reduction expressed in percents. The most elementary technique is based on the mean value and the variance (MV).

The smaller the mean value and the variance, the better the inventory is. In the case presented in Figure 2, the respective values are depicted in Table 1. Although the nominal value of the inventory A is smaller than that of B, the mean value of A is greater than the mean value of B. The same is true for the standard deviations. Even this simple criterion shows, that an inventory of the party B should be considered smaller than that of the party A. This is contrary to the result for nominal values, which ignores uncertainty.

The mean value and the variance may be, however, of limited use, since their comparison in pairs may lead to contradictory evaluations of inventories. In these cases a notion of the semivariance can be applied (MSV), which, in our case, is defined as

dxxKxsKS )()(= 22 µ−∫∞

, (1)

where K is a chosen value and )(xµ is the distribution density function of an inventory. In our case K can be conveniently chosen as a given target, and this value is used in

the examples presented below. The smaller the value of 2Ss , the better the inventory is.

In the example considered earlier 22 > SBSA ss , see Table 1. Thus, according to this

criterion, an inventory B is smaller than A. A large group of techniques uses the notion of critical probability (CP), proposed

already in 1952 [15]. Most of the methods in this group require knowledge of the related probability distributions. The measure used to compare inventories is the probability of surpassing the target K

dxxcrpK

)(= µ∫∞

. (2)

A smaller value of crp indicates better inventory. According to Table 1, again, an inventory of the party B is evaluated as the smaller one.

Table 1: Criteria values for comparison of inventories A and B

Method Criterion value for A

Criterion value for B

Inventory chosen

MV 4=Am

3

116=Aσ

1=Bm

3

2=Bσ

B

MSV 13.45=2SAs 0.35=2

SBs B

CP 9

8=Acrp

8

7=Bcrp B

risk 10.6=critAc 2.1=critBc B


104

In other related methods, as the Baumol's risk measure and the value at risk (VaR), the probability of inventory x to be above a critical value critx is fixed, and then

the value critx is calculated. Without going into details, an inventory is smaller when critx

is smaller. In our example, fixing probability to 0.1, the inventory B is chosen as the smaller one.

A technique similar in spirit has been proposed to ensure a reliable compliance. It is called undershooting, [2, 3, 12, 13, 14], and is illustrated in Figure 3. In this approach, it is required that only a small enough α -th part of an inventory distribution may lie above a target. This idea, then used for ordering inventories, becomes equivalent to the CP technique. However, in the undershooting technique, comparison of two inventories, cx and bxρ , is replaced by checking whether the difference bc xx ρ− is not

greater than zero.

Figure 3. Illustration of compliance in the undershooting approach

In the stochastic dominance technique an inventory A is smaller than B, if their cumulative probability functions (cpfs) satisfy )()( xFxF BA ≤ for all x , and the condition is strict for at least one x . It is obvious that not all inventories can be decisively compared this way, see example for inventories A and B above in Figure 4. Although cpf of the party B is greater for most values of x , it is lower than cpf of the party A for a small range of low value arguments. This possible lack of an answer yes or not is not convenient for comparison of inventories. However, some modifications have been proposed to extend the set of inventories which can be compared.

Figure 4. Stochastic dominance criterion for comparison of inventories A and B


105

In the almost stochastic dominance (ASD)1 the inventory B is smaller than A, if the area between both cpfs for )(<)( xFxF AB is small enough (ε times smaller,

usually with 0.5<<0 ε ) part of the whole area between pdfs, dxxFxF ABx|)()(| −∫ .

It can be seen by inspection that this condition is satisfied in our example. Thus, also this technique indicates inventory B as the smaller one.

In conclusion, decision about obligation fulfillment, which is based on deterministic (nominal value) comparison of an inventory with a target, violates a common sense understanding of comparison and ordering of uncertain values. The deterministic approach also contradicts the already existing scientific knowledge on ordering uncertain projects.

Not all techniques for project comparison can be useful for comparing inventories. Some techniques do not always provide an unequivocal answer, other become complicated, and therefore not convenient in practical use. The critical probability group seems to be particularly suitable, if only a distribution of an inventory uncertainty is known.

2.2. Fuzzy set (possibility) approach

A fuzzy set is a generalization of a usual set. A usual set can be defined by its characteristic function, taking either value 0 or 1. A fuzzy set is characterized by an analogous membership function, which takes values from the interval [0,1]. It means, that a point fully belongs to the set for the value of the membership function equal to 1, it does not belong for the value equal to 0, and only "`partly belongs"' for the intermediate values. A fuzzy set describes imprecision and can be compared to the probability density function. However, it is differently normalized, meaning that its highest value is limited by 1. Also the algebra rules are different.

Fuzzy set models of uncertainty can be considered as a competitive approach to the probabilistic one, described above. A few arguments can be given in favor of this approach. First, the probabilistic approach is intrinsically related to the frequency of variable appearance, while it is hardly possible to have frequent inventories at the same year. Second, in the fuzzy set approach determination of the distribution is much more flexible. The distributions can be freely shaped and do not need to follow any known probabilistic distributions to be practically useful. For example, they can be estimates given by experts. In practice, uncertainty of emission inventories often have an expert-quantified character, even if the Monte Carlo simulation is used to estimate its distribution. Third, the algebra in the fuzzy set approach is simpler, in the sense that for complicated problems more often it is possible to get a final analytic solution using the fuzzy approach than using the probabilistic one, see e. g. [13, 12].

The fuzzy sets have been used in the undershooting technique [12] to calculate the difference bc xx ρ− . But their role was only instrumental there, as the rest of

the technique was close to the idea used in probabilistic CP technique. Similar to the probabilistic case, the unique method for the verification of

the inequality AB xx ≤ for fuzzy sets does not exist. However, possibility and necessity measures of Dubois and Prade [1] can be used for this purpose, see also [6]. To use it, the NSD index may be calculated for the relation AB xx f . 1This is the first order ASD. For the second order ASD see [5].


106

The Necessity of Strict Dominance ( NSD) index for two fuzzy sets A and B is defined as

( )( ) ( ){ } ( )BAPossyx

ABNessNSD

BAxyyx

±−−≤

1=,minsup1=

==

;,

µµf

(3)

where ( )xAµ and ( )yBµ are the membership functions of A and B , respectively.

The value of the NSD index represents a measure of necessity that the set B dominates the set A .

The Possibility of Dominance ( PD ) index for two fuzzy sets A and B is defined as ( ) ( ) ( ){ }yxBAPossPD BA

yxyx

µµ ,minsup==;, ≥

± . (4)

PD is the measure of a possibility that the set A is not dominated by the set B .

Figure 5: Illustration of definition of PD and γ−1=NSD indices.

This approach does not prioritize the fuzzy sets itself, like earlier techniques.

It answers the question of the degree of possibility or necessity of dominance of a chosen set by another one.

In our case, for each inventory there is only one point in the, so called, core of the fuzzy set. The core of the fuzzy set is the set of points for which the membership function is equal to 1. Moreover, the membership function decreases (or at least it does not increase) starting from the core point towards both sides. In this case, the NSD index is nonzero only if the core point of the set A is to the left of the core point of the set B. And it is equal to

PDNSD −1= , (5) where PD is the value of the membership functions at the point where they cut, like in Figure 5. Let us note that inventories are now normalized to have the maximal value of the membership function equal to 1. In this example 0.74=PD and 0.26=NSD . Thus, necessity that the set B strictly dominate the set A is not too high in this example, while possibility that A is not dominated by B is quite high. Similarly as in the undershooting technique, some critical values should be set for making decision.

3. Conclusions and further works

Almost all of the techniques presented in this paper could be used for comparing uncertain inventories. Only a few of them can not be used, because in some situations they do not provide any answer. Other approaches use critical values, where an answer is


107

decisive only when a critical value is achieved, and indecisive if it is not. For example, symmetrically to the compliance requirement in the undershooting technique, it would be reasonable to say, that the party is noncompliant, if there is only some (small) likelihood β that the real emission does not achieve the target, see Figure 6.

In this approach the question would arise, how to treat those parties, which are neither compliant nor noncompliant for accepted α and β . One solution, assumed so far tacitly in the undershooting approach, is to treat all parties, which do not satisfy the upper (α ) condition as noncompliant. But it is actually rather fair to say that no decision can be taken for those parties, which do not satisfy neither upper nor lower condition.

The answer proposed for such cases in [7, 4] is to wait until one or another exceedance occur. Rough methods to estimate when this may take place were also designed. It as called verification time and is based on a linear or quadratic prognosis of future emission trajectory. This approach requires far reaching changes in the system of clearing the commitments, as it does not use one common clearing time for all parties. The present regulation focuses on verifying reduction of emissions at the same time for all parties. Another solution for the indecisive parties, accepted by IPPC for the noncompliant parties and using nominal values, is to shift the decision to the next clearing time with some penalty for not fulfilling the compliance condition in the previous period.

Figure 6. Illustration of noncompliance as the symmetry to the undershooting idea

Partly due to a limited space, this paper focuses on a presentation of the methods. Further results are anticipated as its continuation. From a practical point of view, simulation of different techniques for real cases would be interesting. From a theoretical perspective, introduction of uncertainty in definition of uncertainty distribution could give greater credibility to the method, even if it is confined only to assessing some accuracy estimates.

Acknowledgements

Partial financial support for O. Hryniewicz, Z. Nahorski, and J. Horabik from the Polish State Scientific Research Committee within the grant N N519 316735 is gratefully acknowledged.

References

[1] Dubois D., Prade H. (1983) Ranking fuzzy numbers in the setting of possibility theory. Information Sciences, 30:183-224.


108

[2] Gillenwater M, Sussman F, Cohen J (2007) Practical policy applications of uncertainty analysis for national greenhouse gas inventories. Water, Air & Soil Pollution: Focus, 7(4-5):451-474.

[3] Godal O., Ermolev Y., Klaassen G., Obersteiner M. (2003) Carbon trading with imperfectly observable emissions. Environmental and Resource Economics, 25:151-169.

[4] Gusti M., Jęda W. (2002) Carbon management: new dimension of future carbon research. IR-02-006. IIASA, Austria. http://www.iiasa.ac.at/Publications/Documents/IR-02-006.pdf

[5] Graves S. B., Ringuest J. L. (2009) Probabilistic dominance criteria for comparing uncertain alternatives: A tutorial. Omega, 37:346-257.

[6] Hryniewicz O., Nahorski Z. (2008) Verification of Kyoto Protocol - a fuzzy approach. In: L. Magdalena, M. Ojeda-Aciego, J. L. Verdegay (Eds.) Proc. IPMU'08. Torremolinos, Spain, 729-734. http://www.gimac.uma.es/ipmu08/proceedings/papers/096-HryniewiczNahorski.pdf

[7] Jonas M., Nilsson S., Obersteiner M., Gluck M., Ermoliev Y.M. (1999) Verification Times Underlying the Kyoto Protocol: Global Benchmark Calculations. IR-99-062. IIASA, Austria. http://www.iiasa.ac.at/Publications/Documents/IR-99-062.pdf

[8] Jonas M., Nilsson S. (2007) Prior to economic treatment of emissions and their uncertainties under the Kyoto Protocol: Scientific uncertainties that must be kept in mind. Water, Air, and Soil Pollution: Focus, 7:495-511.

[9] Jonas M., Gusti M., Jęda W., Nahorski Z., Nilsson S. (2010) Comparison of preparatory signal analysis. Climatic Change. DOI: 10.100//s10584-010-9914-6.

[10] Jonas M., Marland G., Winiwarter W., White T., Nahorski Z., Bun R., Nilsson S. (2010) Benefits of dealing with uncertainty in greenhouse gas inventories: introduction. Climatic Change. DOI: 10.100//s10584-010-9922-6.

[11] Lieberman D., Jonas M., Winiwarter W., Nahorski Z., Nilsson S. (2007) Accounting for climate change: Introduction. Water, Air, and Soil Pollution: Focus, 7:421-424.

[12] Nahorski Z., Horabik J. (2010) Compliance and emission trading: rules for asymmetric emission uncertainty estimates. Climatic Change. DOI: 10.100//s10584-010-9916-4.

[13] Nahorski Z., Horabik J., Jonas M. (2007), Compliance and emission trading under the Kyoto Protocol: Rules for uncertain inventories. Water, Air & Soil Pollution: Focus, 7(4-5):539-558.

[14] Nahorski Z., Jęda W., Jonas M. (2003) Coping with uncertainty in verification of the Kyoto obligations. In: Studziński J., Drelichowski L., Hryniewicz O. (Eds.) Zastosowania informatyki i analizy systemowej w zarządzaniu. SRI PAS, 305--317.

[15] Roy A.D. (1952) Safety first and the holding of assets. Econometrica, 20:431-449.

[16] Winiwarter W., Muik B. (2007) Statistical dependences in input data of national GHG emission inventories: effects on the overall GHG uncertainty and related policy issues. Presentation at 2nd Intl. Workshop on Uncertainty in Greenhouse Gas Inventories, 27-28 September 2007, IIASA, Laxenburg, Austria. http://www.ibspan.waw.pl/ghg2007/Presentation/Winiwarter.pdf


109

Determination of the uncertainties of the German emission inventories

for particulate matter (PM 10 & PM 2.5) and aerosol precursors (SO2, NOx, NH3 & NMVOC) using Monte-Carlo analysis

Wolfram Joerss1

1 Institute for Futures Studies and Technology Assessment (IZT) Schopenhauerstr. 26, 14129 Berlin, Germany

[email protected]

Abstract

This paper presents the application of a Monte-Carlo simulation for assessing the uncertainties of German 2005 emissions of particulate matter (PM10 & PM2.5) and aerosol precursors (SO2, NOx, NH3 & NMVOC carried out in the PAREST research project. For the uncertainty analysis German Federal Environment Agency’s emission inventory was amended and integrated with a model on the disaggregation of energy balance data. A series of algorithms were developed in order to make efficient and pragmatic use of available literature and expert judgement data for uncertainties of emission model input data. Looking at the results, the inventories for PM10 (95%-confidence interval: -16%/+23%), PM2.5 (-15%/+19%) and NOx (-10%/+23%) appear most uncertain, while the inventories for SO2 (-9%/+9%), NMVOC (-10%/+12%) and NH3 (-13%/+13%) show a higher reliability. The source categories adding the most relevant contributions to overall uncertainty vary across the pollutants and comprise agriculture, mobile machinery in agriculture and forestry, construction sites, small businesses /carpentries, cigarette smoke and fireworks, road traffic, solvent use and stationary combustion.

Keywords: PAREST, particulate matter, aerosol precursors, emission inventories Monte-Carlo simulation, PM10, PM2.5, SO2, NOx, NMVOC, NH3

Introduction

In the PAREST research project, (PArticle REduction STrategies, cf. www.parest.de & [1]), funded by the German Federal Environment Agency (UBA), emission scenarios until 2020 were constructed for particulate matter (PM10 und PM2.5) as well as aerosol precursors SO2, NOx, NH3 and NMVOC, both for Germany and Europe. Reduction measures were assessed and finally air quality in Germany was modelled. In this framework, also the uncertainties of the nationally aggregated German 2005 emission estimates for all covered pollutants were assessed.

The objective of the exercise was on one hand to assess the reliability of the emission estimates and to identify those source categories which add most to the inventories’ total uncertainties, which is the topic of this paper, as well. On the other hand, the results on uncertainties were used within the PAREST project to construct sensitivity runs for air quality modelling.

In the operationalisation of uncertainties we concentrate on the aspect of accuracy in contrast to completeness (for a discussion of different aspects of uncertainty in emission inventories cf. [2]).

For the uncertainty analysis, a Monte-Carlo analysis was deemed to be preferable to a “simple” calculation of error propagation, as the uncertainties met for the input data to the considered inventories are often rather high (up to an order of magnitude) and correlated.


110

The Emission Model

The emission inventories used and enhanced in PAREST build upon the German Federal Environment Agency’s emission data base ZSE (Central System Emissions, [3]), including a wide range of sectoral emission estimation models which are used to feed the database with activity rates (AR), emission factors (EF), emissions (EM) and/or other relevant variables for the calculation of emissions, e.g. split factors (SF). For the inventories relevant for PAREST, the ZSE data base features approx. 900 time series of activity rates and approx. 450 to 700 time series of emission factors for each of the pollutants respectively.

Notably, energy related data for stationary use are processed in the separate model BEU (“Balance of Emission Causes”, described in [4]), which is used to disaggregate the official German energy balance into more than 400 segments of fuel use in order to meet a highly differentiated set of emission factor structure distinguishing a variety of fuels, economic sectors, combustion technologies, installation sizes and applicable environmental legislation. Other sophisticated external models are used to generate ZSE input data e.g. for traffic and mobile machinery, agriculture and solvent use. The full data set is characterised in [5].

Operationalising Uncertainties

The mathematical operations in the assessed deterministic emission model (consisting mainly of the ZSE and BEU models/ data bases) can be roughly reduced to three types of equations:

ARi * EFi,x = EMi,x, (1)

where ARi is the activity rate for source category i, EFi,x is the emission factor for pollutant x in source category i, EMi,x is the emission of pollutant x from source category i,

EMi,x + EMj,x = EMx, (2)

where EMi,x is the emission of pollutant x from source category i, EMi,x is the emission of pollutant x from source category j, EMx is the total emission of pollutant x from source categories i and j, and

ARm = ARP * f in conjunction with ARn = ARP * (1-f), (3)

where ARm is the activity rate for source category m, ARn is the activity rate for source category n, ARP is the primary activity rate (sum of source categories m and n), f is the split factor.

Thus, in the uncertainty assessment of the emission model, we need to attribute an uncertainty to all input variables, .i e. in the example of equations (1) to (3) above to the variables ARi, ARj, EFi,x, EFj,x, ARP and f (the factor f is subject to uncertainty as well!).Based on that, we can compute uncertainties of the model results, in the example above that is for EMx, ARm and ARn, and as well for EMi,x and EMj,x.

In line with the IPCC guidelines [6], we use a relative uncertainty, expressed in percent of the mean (or reference) value which are used in the deterministic model: Mathematically / stochastically, these uncertainties are defined through the 95% confidence interval of an assumed probability distribution. In the interpretation of the application to an emission inventory, however, we switch from a statistical concept of probability to a Bayesian concept which defines probability as the subjective degree of belief [7].


111

For the aggregation of uncertainties in an emission model / inventory along equations (1) to (3) above, the IPCC guidelines [6] offer two basic approaches, i.e. error propagation rules and Monte-Carlo simulation. Error propagation is easier to perform. However, it does not produce reliable results for comparably high uncertainties: [6] sets a variation coefficient of 0.3 as a limit which corresponds to +/- 58% for a normal distribution. Furthermore it is restricted to symmetric probability density functions and cannot account for correlations. As all the restrictions are not met in our application, a Monte-Carlo simulation approach was chosen.

In a Monte-Carlo simulation, each input variable of the emission model is represented by a full probability density function, defined by a shape (e.g. normal, lognormal etc.) a mean (the reference of the deterministic model) and 2.5 % and/or 97,5% percentiles (as the borders of the 95% confidence interval). The Monte-Carlo analysis simulates a large number of random experiments. In each run, for every input variable a random value is taken which then feeds the calculation of the model results. For the whole set of model runs (we generally used 10,000 runs, using @Risk 5.5 software package) the predefined probability distribution for each input variable is maintained. Consequently, the model outputs, i.e. usually aggregated emissions or any intermediate calculation step, are distribution functions as well and can be characterised e.g. by mean value and the 95% confidence interval.

For the characterisation of the input variable we have used a set of mathematically well defined distributions, i.e. normal, triangle, uniform, lognormal and inverse lognormal:

normal Triangle uniform lognormal inverse lognormal

Figure 1. Distribution types used for input variables

While normal, triangle and uniform distributions have symmetric probability density functions, the lognormal distributions are asymmetric and feature different percentage values for the upper and lower border of the confidence interval. These upper and lower values cannot be chosen freely, though: A lognormal distribution is characterised by the parameters µ and σ, and each specific lognormal distribution is well defined through its mean value plus only one percentile [8]:

, (4)

where E(X) is the mean of the lognormal distribution, µ and σ are location parameters of the lognormal distribution and

, (5)

where x(p) is p-quantile of the lognormal distribution, u(p) is the p-quantile of the standardised normal distribution (for p = 2,5%, u(p) equals -1,96; for p = 97,5%, u(p) equals 1,96), µ and σ are location parameters of the lognormal distribution.

2,5% 95,0% 2,5%

1,500 2,500

NormalAlt

("mu";2;0,975;2,5)

Minimum −∞

Maximum +∞

Mittelwert 2,0000

Std.Abw. 0,2551

2,5% 95,0% 2,5%

1,224 2,776

Triang(1;2;3)

Minimum 1,0000

Maximum 3,0000

Mittelwert 2,0000

Std.Abw. 0,4082

2,5% 95,0% 2,5%

1,050 2,950

Uniform(1;3)

Minimum 1,0000

Maximum 3,0000

Mittelwert 2,0000

Std.Abw. 0,5774

2,5% 95,0% 2,5%

0,71 4,52

0 1 2 3 4 5 6

Lognorm(2;1)

Minimum 0,0000

Maximum +∞

Mittelwert 2,0000

Modus 1,4311

Std.Abw. 1,0000

0 2 4 6 8

10


112

With a known (or pre-defined) mean value (i.e. the reference value of the deterministic model) and a guess for just one of the borders of the 95% confidence interval, the other border can be computed. Furthermore, it can be deducted from equations (4) and (5) that for a lognormal distribution the 97.5% quantile’s maximum deviation from the mean is approx. 583%. Accordingly, the minimum 2.5% quantile is at approx. -99,7%

This has important implication for the use of expert judgement or literature for feeding the model with uncertainties of input variables, as a quote of e.g. “mean value X, +200 % / -70%, lognormal” does not meet both equations (4) and (5) and thus cannot be translated into the Monte Carlo analysis without further processing. In parallel, an expert estimation like “factor 10” (i.e. +900%) cannot directly be used as a 97.5% quantile of a lognormal distribution.

In order to pragmatically make use of existing data sources for uncertainties, we developed an algorithm to generate a mathematically well defined lognormal distribution:

1. The distribution type “lognormal” and the mean (reference) value are kept unchanged.

2. The expert guess of the lower deviation is interpreted as a 2.5% quantile. A fitting 97.5% quantile is calculated through equations (4) and (5).

3. We compare the thus calculated value with the original expert guess for the upper deviation and take the arithmetic mean (as a maximum, however the above mentioned 583% limit) as the 97.5% quantile for the Monte-Carlo simulation.

4. The 2.5% quantile of the distribution used for the simulation can again be calculated using equations (4) and (5).

This algorithm was chosen, as in the PAREST context, the reference value was NOT be altered. In a project framework which allows for a change or improvement of a “best guess” reference, other choice might deem appropriate.

As mentioned above, correlations between input variables have to be regarded in the uncertainty assessment. To account for that, the emission model was cleared from all redundancies to make sure that each piece of information on the physical world which is used in several parts of the model is transferred into a probability function exactly once during each model run. This applies both to correlation within a given pollutant inventory as well as across pollutants, if e.g. the same activity rates are used for emission calculation of several pollutants. Additionally, PM2.5 emissions and emission factors are calculated using (uncertain) split factors based on PM10 emissions/ emission factors.

Data Sources for Input Parameters

In order to carry out a Monte-Carlo simulation it was then necessary to determine the shapes and quantiles of the probability functions that replace the deterministic input variables. This means in particular the respective types of distribution and 2.5%/ 97.5% quantiles (the means were fixed anyway through the deterministic model). The sources for these parameters were taken over from primary and secondary literature or estimated through expert judgement and varied with the source categories, pollutants and type of values (i.e. emission factors (EF), emissions (EM), activity rates (AR) or split factors (SF). Table 1 gives a rough overview.

It has to be noted, though, that sometimes for missing spots analogies had to be assumed. For that purpose, pragmatic algorithms were used comparable to the one explained for lognormal distributions in section 3 above.


113

Comparable effort was put into split factors, concerning whose uncertainties hardly any literature source was available. In our emission model, split factors (determining one entity to represent x% (x in the range of 0%-100%) of logically larger entity already calculated in the model) are in particular important on one hand for the disaggregation of energy balance activity data (technology splits, fuel splits, full load hours etc. in the BEU model) and on the other hand for the calculation of PM2.5 emissions based on PM10 emissions. While split factors bearing the values of straight 100% or 0% in the deterministic model were kept unchanged as constant factors in the probabilistic model (otherwise, the mean of the probability distribution could not have met the reference value), for all “real” split factors with the value x with 0% < x < 100% algorithms were defined for the conversion into a probability function: First, the distribution type was defined (lognormal if close to 0% or 100%, normal if rather medium. Second, the 2.5% / 97.5% quantiles were defined depending on both the chosen distribution and on relative position of the reference/mean value.

Table 1. Data sources for uncertainty parameters

Source Category

Data source

Research project reports commissioned by the German Federal Environment Agency (UBA) targeted on emissions, emission factors and/or uncertainties

EMEP/EEA air pollutant emission

inventory guidebook — 2009

[9]

own estimates

stationary combustion

Energy balance (AR), Distribution parameters for energy balance (partly) EF SO2, NOx, NMVOC, PM10

EF NH3 distribution parameters for energy balance

road traffic AR, EF NOx, NMVOC & SO2 EF NH3 & PM10 (exhaust); EF PM10 road/ break/ tyre wear

EF PM10 resuspension

other surface traffic

AR, EF NOx & PM10 EF NMVOC, SO2 & NH3

agriculture AR, EF NH3, NOx, & NMVOC EF PM10 solvent use EM handling of bulk materials

AR & EF (UBA expert judgement)

other source categories

partly partly partly

References [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23]

Generally, for the implementation of Monte-Carlo analysis to the emission model, logical correlations within the respective sets of EF, AR and SF were considered, based on a thorough understanding of the modelled emission sources.

A detailed description of the analysis performed is given in [15].

Results

The aggregated results of the uncertainty assessment are summarised in Table 2. A more specific view on the sectoral contributions to the overall uncertainties shows

Figure 2 below, using an adapted version of the SNAP emission reporting format.


114

Table 2. Aggregated Uncertainties in the German emission inventory 2005

Uncertainties in the German emission inventory 2005

PAREST reference scenario, emission calculation according to inland principle

Pollutant 95%- confidence interval

2,5%-quantile 97,5%-quantile

PM10 -16% 23%

PM2.5 -15% 19%

SO2 -9% 9%

NOx -10% 23%

NH3 -13% 13%

NMVOC -10% 12%

Figure 2. German 2005 Emissions and Uncertainties (SNAP format)


115

Looking at the aggregated results (Table 1 above), the inventories for PM10 (95%-confidence interval: -16%/+23%), PM2.5 (-15%/+19%) and NOx (-10%/+23%) appear most uncertain, while the inventories for SO2 (-9%/+9%), NMVOC (-10%/+12%) and NH3 (-13%/+13%) show a higher reliability. For comparison: The official UBA uncertainty data on the German 2007 GHG inventory [24] is +/ 9.7% (for CO2 equivalents of both CO2 and non-CO2 GHG in sum)1.

Looking across the analysed pollutants (Figure 2 above), the source categories adding the most relevant contributions to overall uncertainty vary across the pollutants and comprise

• agriculture (NOx from fertiliser application, NMVOC from manure management, NH3 from animal husbandry and cultivation of land, PM10 from pig fattening),

• mobile machinery in agriculture and forestry (PM10, PM2,5 and NOx), • construction sites (PM10), • small businesses /carpentries (PM10 and PM2.5), • cigarette smoke and fireworks (PM2.5), • road traffic (PM10 and PM2.5 from resuspension of road dust, NOx from heavy

duty vehicles and passenger cars, NMVOC and NH3 from petrol engines), • solvent use (NMVOC) and • stationary combustion (SO2 from coal-fired power plants and oil-fired domestic

furnaces, PM10 and NMVOC from wood firing).

Conclusions and Look Ahead

Within the PAREST project, the German Federal Environment Agency’s (UBA’s) 2005 emission inventories for particulate matter (PM10 & PM2.5) and aerosol precursors (SO2, NOx, NH3, NMVOC) was amended and for the first time successfully fed into a Monte-Carlo analysis for uncertainty assessment. Progress was made in particular by fully integrating into the uncertainty analysis the UBA’s “BEU” model for processing energy balance data into the inventory. Overall uncertainties of the covered pollutants were determined, and rough sectoral analyses were made.

Looking ahead, however, there appear a couple of issues worthwhile to be treated in future work:

• A full key source analysis according to IPCC standards for each of single pollutant inventories.

• A cross-pollutant key source analysis taking into account ambient air PM generation potentials which can be deducted from the air quality modelling performed in PAREST.

• A further validation of the developed probabilistic emission inventory model by performing sensitivity runs on parameter settings and reviewing of key parameters.

• An application of the methodological advance in particular of energy related uncertainties on the German greenhouse gas inventories.

1 However, the respective methodologies to calculate these figures for particulate matter & precursors on one hand (PAREST) and greenhouse gases on the other hand (UBA) are not identical have not yet been comprehensively compared.


116

References

[1] Builtjes P.; Theloke J., Stern R. and Joerss W. (2010): PAREST-Final Report to the German Federal Environment Agency (UBA), FKZ 206 43 200/01 „Strategien zur Verminderung der Feinstaubbelastung“, forthcoming as UBA-TEXTE (ISSN 1862-4804, Dessau-Roßlau).

[2] Aardenne J.A. (2002): „Uncertainties in emission inventories“, Thesis Wageningen University, ISBN 90-5808-641-0

[3] Federal Environment Agency (2007): Zentrales System Emissionen (ZSE), Emission Data Base of the German Federal Environment Agency (UBA), Version 08.06.2007

[4] Joerss W. and Kamburow C. (2006): „Bilanzierung und Modellierung emissionsrelevanter Daten zum Energieverbrauch in stationären Quellen“. WerkstattBericht Nr. 78. Berlin: IZT - Institut für Zukunftsstudien und Technologiebewertung

[5] Joerss W, Kugler U. and Theloke J.(2010): „Emissionen im PAREST-Referenzszenario 2005 – 2020“; Report to the German Federal Environment Agency (UBA), in the framework of the PAREST project FKZ 206 43 200/01 „Strategien zur Verminderung der Feinstaubbelastung“, Berlin: IZT, forthcoming as UBA-TEXTE (ISSN 1862-4804, Dessau-Roßlau)

[6] Eggleston, H. S.; Buendia, L.; Miwa, K., et al. (Ed.). “IPCC Guidelines for National Greenhouse Gas Inventories”. Prepared by the National Greenhouse Gas Inventories Programme. Japan. http://www.ipcc-nggip.iges.or.jp/public/2006gl/index.html

[7] Morgan, M.G., Henrion, M., (1990): Uncertainty. A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge University Press, Cambridge, ISBN 0-521-42744-4

[8] Müller P.H. (1991): Lexikon der Stochastik; Akademie Verlag, Berlin, 1991, 5. Auflage

[9] EMEP/EEA air pollutant emission inventory guidebook — 2009; EEA Technical report No 6/2009, Copenhagen, 2009; ISBN 978-92-9213-034-3; DOI 10.2800/23924; http://www.eea.europa.eu/publications/emep-eea-emission-inventory-guidebook-2009

[10] Dämmgen U. (2008): Expert Judgement on uncertainties of German agricultural emissions, 01. 04.2008.

[11] Dämmgen U., Haenel H.D., Rösemann C., Conrad J., Lüttich M., Döhler H., Eurich-Menden B., Laubach P., Müller-Lindenlauf M. and Osterburg .. „Calculations of Emissions from German Agriculture - National Emission Inventory Report (NIR) 2009 for 2007“ Johann Heinrich von Thünen-Institut Bundesforschungsinstitut für Ländliche Räume, Wald und Fischerei (vTI) Braunschweig, 2009. ISSN 0376-0723, ISBN 978-3-86576-049-4.

[12] Degel, M and Joerss W. (2009): „Aufbereitung von Daten der Emissionserklärungen gemäß 11. BImSchV aus dem Jahre 2004 für die Verwendung bei der UNFCCC- und UNECE-Berichterstattung - Teilbericht Stationäre Verbrennungsmotoren“. UBA-TEXTE Nr. 45/2009, Dessau-Roßlau: Umweltbundesamt, ISSN 1862-4804


117

[13] Harthan R., Graichen J., Graichen V., Repenning R, Ziesing H.J. and Wittke F. (2007): “Dokumentation der Datenqualität von Aktivitätsdaten für die Berichte über Emissionen aus stationären Feuerungen im Rahmen des Nationalen Inventarberichtes und des Monitoring Mechanismus nach RL EG 99/296“; Endbericht des Forschungs- und Entwicklungsvorhabens FKZ 204 41 132 im Auftrag des Umweltbundesamtes, Berlin, 2007

[14] Handke V, Joerss W., Pfitzner R.; Brinkschneider F. and Schollenberger H. (2004): „Das Qualitäts-System-Emissionsinventare –Handbuch“; Forschungsbericht für das Umweltbundesamt FKZ 202 42 266; Berlin: IZT

[15] Joerss W. and Handke V. (2010): „Unsicherheiten der PAREST-Referenz-Emissionsdatenbasis“; Report to the German Federal Environment Agency (UBA), in the framework oft he PAREST project FKZ 206 43 200/01 „Strategien zur Verminderung der Feinstaubbelastung“, Berlin: IZT, forthcoming as UBA-TEXTE (ISSN 1862-4804, Dessau-Roßlau)

[16] Joerss W. (2010): „Emissionen aus Offener Verbrennung in Deutschland“; Report to the German Federal Environment Agency (UBA), in the framework oft he PAREST project FKZ 206 43 200/01 „Strategien zur Verminderung der Feinstaubbelastung“, Berlin: IZT, forthcoming as UBA-TEXTE (ISSN 1862-4804, Dessau-Roßlau)

[17] Kludt R. (2009): Expert Judgement on uncertainties of activity rates and emission factors for particulate matter emissions from handling of bulk materials, Berlin 26.8.2009

[18] Knörr W., Heldstab J., Kasser F. and Keller M. (2009): „Ermittlung der Unsicherheiten der mit den Modellen TREMOD und TREMOD-MM berechneten Luftschadstoffemissionen des landgebundenen Verkehrs in Deutschland“. INFRAS AG Zürich, Schweiz. ifeu-Institut für Energie- und Umweltforschung. Heidelberg. Commissioned by the German Federal Environment Agency (UBA). FKZ 360 16 023. Heidelberg/Zürich/Bern, 20. Juli 2009

[19] Rentz O., Karl U. and Peter H. (2002): Ermittlung und Evaluierung von Emissionsfaktoren für Feuerungsanlagen in Deutschland für die Jahre 1995, 2000 und 2010. UBA-report 299 43 142.

[20] M. Schaap, A.M.M. Manders, E.C.J. Hendriks, J.M. Cnossen, A.J.S. Segers, H.A.C. Denier van der Gon, M. Jozwicka, F. Sauter, G. Velders, J. Mathijssen, P.J.H. Builtjes (2009): Regional Modelling of Particulate Matter for the Netherlands. Technical Report BOP, research carried out for Ministry of Housing, Spatial Planning and the Environment (VROM), Bilthoven, Niederlande

[21] Struschka M., Zuberbühler U., Dreiseidler A., Dreizler D., Baumbach G., Hartmann H., Schmid V. and Link H. (2003): “Ermittlung und Evaluierung der Feinstaubemissionen aus Kleinfeuerungsanlagen im Bereich der Haushalte und Kleinverbraucher sowie Ableitung von geeigneten Maßnahmen zur Emissionsminderung, Forschungsbericht für das Umweltbundesamt; UBA-TEXTE 41/03, ISSN 1862-4804.

[22] M. Struschka, D. Kilgus, M. Springmann, G. Baumbach (2008): "Effiziente Bereitstellung aktueller Emissionsdaten für die Luftreinhaltung"; Universität Stuttgart, Institut für Verfahrenstechnik und Dampfkesselwesen (IVD) UBA-FB 205 42 322.


118

UBA-FB 001217; UBA Texte 44/08. Dessau-Roßlau, November 2008. ISSN 1862-4804.

[23] Theloke, J (2005).: NMVOC-Emissionen aus der Lösemittelanwendung und Möglichkeiten zu ihrer Minderung. Fortschritt-Berichte VDI Reihe 15 Nr. 252. Düsseldorf: VDI-Verlag 2005

[24] Federal Environment Agency (2010): National Inventory Report for the German Greenhouse Gas Inventory 1990 – 2008 Submission under the United Nations Framework Convention on Climate Change and the Kyoto Protocol 2010; Umweltbundesamt, Dessau, Climate Change Nr. 04/2010


119

Dealing with Uncertainty in Greenhouse Gas Inventories in an Emissions Constrained World

Matthias Jonas1, Volker Krey1, Fabian Wagner1, Gregg Marland2, Zbigniew Nahorski3

1 International Institute for Applied Systems Analysis (IIASA) Schlossplatz 1, 2236 Laxenburg, Austria

[email protected] 2 Oak Ridge National Laboratory, Carbon Dioxide Information Analysis Center (CDIAC)

Oak Ridge, TN, USA 3 Polish Academy of Sciences (SRI), Systems Research Institute

Warsaw, Poland

Abstract

The urgent task under the United Nations Framework Convention on Climate Change (UNFCCC) is to agree on a climate treaty that comes into force in 2012, the year in which commitments under the Kyoto Protocol will cease. Leaders of the world’s major industrialized countries have formally agreed in the wake of the 2009 UN climate change conference in Copenhagen that the average global temperature should not increase by more than 2oC from its pre-industrial level. Compliance with this temperature target can be expressed equivalently in terms of limiting cumulative GHG emissions, for example, up to 2050, while considering the risk of exceeding this target. The emission reductions required are substantial: 50–80% below the 1990 level at the global scale, with even greater reductions for industrialized countries.

Although the issue of translating an approved global emissions constraint to the sub-global level and allocating global emission shares to countries is still unsettled, a crucial question arising and still to be answered is: how should we deal with the uncertainty associated with the accounting of emissions for compliance purposes (i.e., with the uncertainty in GHG inventories)? The accounting of emissions, when bottom-up estimates are compared with top-down constraints derived from scenarios, could still force us to accept additional uncertainties due to still existing accounting gaps. In addition, minimizing the risk of exceeding an agreed global average temperature target may demand considerable undershooting of the most uncertain emissions estimates to ensure that the global total of emissions does not exceed the agreed target.

In our study we make use of a per-capita and GDP related metric to translate a global emissions constraint for 2000–2050 – i.e., a constraint that complies with an agreed global average temperature target and a variable risk of exceeding it – to the country level. The purpose of this exercise is to put uncertainties that are associated with accounting emissions for compliance purposes into a wider quantitative context.

Keywords: Greenhouse gas emissions, emission constraints, contraction & convergence, emission equity, uncertainty in emission estimates, sustainable land use

1. Introduction

The urgent task under the United Nations Framework Convention on Climate Change (UNFCCC) is to agree on a climate treaty that comes into force in 2012, the year in which commitments under the Kyoto Protocol will cease [1,2]. Leaders of the world’s major industrialized countries have formally agreed in the wake of the 2009 UN climate change conference in Copenhagen that the average global temperature should not increase by more than 2 oC from its pre-industrial level [3–6]. Compliance with this temperature


120

target can be expressed equivalently in terms of limiting cumulative greenhouse gas (GHG) emissions, for example, up to 2050, while considering the risk of exceeding this target [7,8]. The emission reductions required are substantial: 50–80% below the 1990 level at the global scale, with even greater reductions for industrialized countries [9,4,6,10; see also below].

Global cumulative emission budgets constitute an important first step in translating long-term GHG concentrations targets (e.g., for 2100) to mid-term emission constraints (here in terms of constraints for 2000–2050). However, these need to be translated further, notably

1. to emission targets on the near-term time scale,

2. and to emission targets on the national scale;

so that governments can implement these through ‘cap-and-trade’ schemes or other kinds of tangible policy effort, such as carbon taxes or regulatory programmes. However, the 2 oC temperature target was considered a political delusion already prior to the Copenhagen conference. Even with a big dose of luck, the effort needed to get to 2 oC would be heroic, as the authors of [8] indicate, and probably far beyond what real governments can achieve [11].

In our study we address the two challenges mentioned above, albeit in a 2 oC target context to demonstrate the formidable task ahead. In linking the short with the mid-term, we start from calculating for Annex I countries 1990–2050 per-capita GHG emission paths that comply with a universal per-capita emissions target in 2050 (‘contraction & convergence’).1 What makes our study unique is the combination of conditions that are (and will be) applied: a constraint for cumulative anthropogenic emissions and for emissions from the terrestrial biosphere on a per-capita level by country; a two-sided treatment of uncertainty, top-down and bottom-up;2 and the introduction of the spatial disconnect between biomass production and biomass consumption. However, we will also explore metrics alternative to the contraction & convergence case to distribute emission allowances.

Our study combines expertise that is available at IIASA, SRI and CDIAC with respect to current emissions, compliance with emission targets under uncertainty, emission pledges, and projections of emissions till 2050 and beyond [13–16]. This 1 The list of Annex I countries to the UNFCCC differs from the list of Annex B countries to the Kyoto Protocol. The countries of concern are: Belarus and Turkey (I not B), Croatia (B not I), and Cyprus and Malta (neither I nor B but EU Member States). Officially, only the former EU-15 Member States are a party under both the UNFCCC and the Kyoto Protocol (I and B). However, the EU-27 seeks to comply with the Kyoto Protocol as a whole (8% emission reduction relative to 1990). 2 Uncertainty reported by climate modelers based on the analysis of (prognostic) emission-climate change scenarios is referred to as ‘top-down’; while uncertainty associated with (diagnostic) emission inventories is referred to as ‘bottom-up’. The former expresses, e.g., the risk of exceeding an agreed temperature target dependent on cumulative GHG emissions till a given point in time; while the latter can be related to the risk that true (but unknown) GHG emissions are greater than inventoried and reported emissions at a given point in time. Note that this comparison of bottom-up emission estimates with top-down, scenario-derived constraints builds on the assumption that unaccounted emissions do not exist.


121

combination allows investigating for Annex I countries near-term emission target paths that comply with the 2 oC temperature target at mid-term (2050)

− by following a national per-capita emissions equity concept (work in progress and described below) and, alternatively, a regional emissions intensity concept (work ahead);

− by including bottom-up uncertainties from emission inventories, which must be considered in addition to those reported top-down by climate modelers in grasping the risk of exceeding the agreed temperature target (described below for the per-capita emissions equity concept);

− by including mitigation costs and also quantifying the costs of factoring in bottom-up uncertainty from inventorying emissions (described below for the per-capita emissions equity concept); and

− by tackling deforestation and other land use (LU) mismanagement by treating the LU sector separately from anthropospheric GHGs (described below for the per-capita emissions equity concept).

Section 2 provides a brief overview of our methodological concept and the models that we apply and intend to apply. Section 3 shows the provisional result typical for a data-rich country. Section 4 provides a preliminary discussion and our summary.

2. Methodological Concept and Models Applied

Tab. 1: Overview of Section 2.

Sub-section

Subsection Thematic Focus Underlying Assumption

2.1 Global emission constraints Anthropogenic GHG emissions: annual GHG emissions of the six Kyoto GHGs (in CO2 equivalent) from the sector/source categories listed in Annex A to the Kyoto Protocol [17]

All emissions are accounted for; i.e., no unaccounted emissions exist

2.2 From Global to National: Per-capita Emissions Equity by 2050

2.3 Per-capita Emissions for 1990–2006 2.4 Pledges for 2010–2020 2.5 Alternative to the Per-capita Emissions

Equity Target in 2050 2.6 Considering Bottom-up Uncertainty 2.7 Land Use and Land-use Change

Management till 2050 Emissions from LU: annual CO2 emissions resulting from LU [19]

2.1. Global emission constraints

In their 2009 special report [6] the German Advisory Council on Global Change (WBGU) proposes to base the allocation of emission rights on three principles: the polluter pays principle, the precautionary principle, and the principle of equality.3

3 Under the polluter pays principle, it is the industrialized countries that have a particular responsibility to cut their GHG emissions due to their high cumulative emissions in the past. The precautionary principle acknowledges, in line with the principle of sustainability, that timely action is required to prevent irreversible damage to present and future generations. The principle of equality postulates individuals’ equal rights without distinction, to the benefits of the global commons [6].


122

WBGU developed the idea of determining an upper limit for the tolerable warming (i.e., mean global temperature) and thus for deriving a global CO2 reduction target by means of an ‘inverse approach’ (backward calculation) [12]. WBGU’s budget approach is the further development of this concept and links it to current international climate policy [but see [11] above). The central idea is that to keep atmospheric warming below 2 oC, the total amount of anthropogenic CO2 emitted to the atmosphere must be limited. WBGU proposes adopting a binding upper limit for the total amount of CO2 that can be emitted from fossil fuel sources up to 2050; and allocating the available emissions amount subject to negotiation. WBGU breaks down the global budget into national emission budgets based on an equal per-capita basis. The budget approach contains four parameters that are political (i.e., negotiable) by nature. These are: 1) the start and 2) end year for the total budget period; 3) the cumulative emissions constraint or, equivalently, the probability of exceeding the 2 oC temperature target; and 4) the year of reference for global population. Our choice of the four parameters – 1) 1990, 2) 2050, 3) 10–43%, and 4) 2050 – differs slightly from the options investigated by WBGU.4

The strength of a global cumulative emissions constraint is that it is compelling: no country can escape. If a country, e.g., wants to choose another (e.g., later) start or base year, it must make clear how its emissions for the missing years are balanced in a global context; i.e., how the community of other countries shall take over the country’s emission burden for these years.

Our approach advances WBGU’s approach by including bottom-up uncertainty from emission inventories, mitigation costs including the costs of considering uncertainty, and by treating the LU sector (and its uncertainty) separately. In Section 2.2 we construct 1990–2050 per-capita emissions equity paths for Annex I countries while focusing on their anthropogenic Kyoto GHG emissions (i.e., excluding LU). In Sections 2.3–2.6 we contrast the countries’ 1990–2050 per-capita emissions equity paths with monitored as well as pledged and projected emissions while additionally considering bottom-up uncertainty. In Section 2.7 we expand our view on anthropogenic GHGs and investigate the countries’ emissions resulting from LU.

2.2. From Global to National: Per-capita Emissions Equity by 2050

The notion of a global emissions constraint received momentum in 2009 and was widely discussed prior to the UN climate change conference in Copenhagen. To start with, we make use of the global 2000–2049 cumulative emissions constraint of 1,500 Gt CO2-eq reported by [7]. It limits the probability of exceeding the temperature target of 2 oC at and post 2050 to 10–43%.5 We use 2050 as the year of reference for global population and take the projected value for 2050 from [18]. Considering, in addition, the decade of the nineties, i.e., global emissions of all Kyoto GHGs for 1990–1999 from

4 The four parameters in WBGU’s ‘historical responsibility’ approach are 1) 1990, 2) 2050, 3) 25%, and 4) 1990; while they are 1) 2010, 2) 2050, 3) 33%, and 4) 2010 in its ‘future responsibility’ approach. In the two approaches the probability of exceeding the 2 oC temperature target refers to cumulative emissions constraints for 2000-2049. 5 In 2000 total emissions of CO2, CH4, N2O and high GWP (global warming potential) gases from anthropogenic sources and land use were about 39.4 Gt CO2-eq, with CO2 contributing about 29.9 Gt. Between 2000–2006 the emissions of CO2 alone totaled already 234 Gt, i.e., 16% of the cumulative constraint for 2000–2050.


123

anthropogenic and terrestrial-biospheric sources [19], a global per-capita GHG emissions equity target of 2.9 tCO2-eq can be established for 2050.

This per-capita GHG emissions equity target is universally valid and constitutes the end point of linear emission target paths that countries have to follow, starting from their per-capita emissions in 1990. Initially, the focus is on the countries’ combined anthropogenic emissions (i.e., it is assumed that countries will practice sustainable land use by 2050; see also Section 2.7 below). These per-capita emission target paths serve as a reference in monitoring the countries’ success in approaching the universal emissions equity target.

2.3. Per-capita Emissions for 1990–2006

The data used in this study for anthropogenic GHG emissions are summarized in Table 2 and reflect the situation prior to our data update (ongoing). We can make use of two alternative data sources.

Tab. 2: Overview of the two alternative data sources used for anthropogenic GHGs.

GHG Source Period Resolution CO2, CH4, N2O, HFCs, PFCs, S F6 (Kyoto GHGs)

UNFCCC 1990–2006 by country; annual

CO2 CH4, N2O, high GWP emissionsa)

CDIAC, UN Pop Div EPA, UN Pop Div

1990–2006 1990–2005

by country; annual by country; every 5 years

a) The gases included are the anthropogenic (direct) GHGs – other than CO2 – covered by the UNFCCC: CH4, N2O, and the high global warming potential (GWP) gases including substitutes for ozone-depleting substances and industrial sources of HFCs, PFCs and SF6 [20].

To contrast the countries’ current per-capita emissions with their emission limitation and reductions commitments as listed in Annex B to the Kyoto Protocol [17], we translate the latter to a per-capita basis. Base year and commitment year per-capita emissions allow introducing linear, 1990-to-2010 Kyoto target paths as reference.6

2.4. Pledges for 2010–2020

Pledges of and baseline emissions for Annex I countries are routinely monitored and updated by IIASA’s GAINS (Greenhouse gas – Air pollution Interactions and Synergies model) group. GAINS allows estimating emission reduction potentials and costs for a range of GHGs and air pollutants; and quantifying the resulting impacts on air quality and total GHG emissions considering the physical and economic interactions between different control measures. As a principle, the analysis employs only such input data that are available in the public domain and that appear credible and consistent in an international perspective.

Methodologically, GAINS (i) adopts exogenous projections of future economic activities as a starting point; (ii) develops a corresponding baseline projection of GHG emissions for 2020 with information derived from the national GHG inventories that have been and are reported by Parties to the UNFCCC; (iii) estimates, with a bottom-up approach, for each economic sector in each country the potential emission reductions that could be achieved through application of the available mitigation measures; and (iv)

6 We use the year 2010 as commitment year and temporal average in (net) emissions over the commitment period 2008–2012.


124

quantifies the associated costs required for these measures under the specific national conditions. The approach includes all six GHGs that are included in Annex A to the Kyoto Protocol (i.e., CO2, CH4, N2O, HFCs, PFCs, SF6) and covers all anthropogenic sources that are included in the emission reporting of Annex I countries to the UNFCCC – Energy; Industrial Processes; Agriculture; Waste; and the LULUCF (Land Use, Land-use Change, and Forestry) sector. In addition, the analysis quantifies the implications of GHG mitigation strategies on air pollution [21].

2.5. Alternative to the Per-capita Emissions Equity Target in 2050

MESSAGE is an energy systems model that captures the global energy system (plus other sectors) and allows identifying cost efficient energy supply technology portfolios and energy conservation efforts for 11 world regions until 2100, the time horizon required for analyzing the implications of mitigation efforts on climate. The model’s principal results include technology-specific multi-sector response strategies for achieving specific GHG emissions, concentrations, or radiative forcing targets [16]. In the context of this study, we use MESSAGE to compile alternatives to the normative, equal per-capita emissions target in 2050 mentioned above. One such alternative are regional GHG emission targets for 2050 based on equal marginal mitigation costs from so-called first-best scenarios which assume full integration and efficient carbon markets. The MESSAGE breakdown of GHG emissions by region for a given target will be complemented by similar results from a recently compiled scenario database that includes some 150 scenarios from about 15 energy-economic and integrated assessment models (work ahead).

2.6. Considering Bottom-up Uncertainty

GHG inventories contain uncertainty for a variety of reasons, and until recently, relatively little attention has been devoted to how uncertainty in emissions estimates is dealt with and how it might be reduced. This situation is slowly changing, with uncertainty analysis increasingly being recognized as an important tool for improving inventories of GHG emissions and removals.

We have available six techniques to analyze uncertain emission changes (also called emission signals) that countries agreed to achieve by the end of the Protocol’s first commitment period, 2008–2012. The techniques allow analyzing uncertain emission signals from various points of view, ranging from signal quality (defined adjustments, statistical significance, detectability, etc.) to the way uncertainty is addressed (trend uncertainty or total uncertainty). For most countries under the Kyoto Protocol the agreed emission changes are of the same order of magnitude as the uncertainty that underlies their combined (carbon dioxide equivalent) emission estimates. Any such technique, if implemented, could "make or break" claims of compliance, especially in cases where countries claim fulfilment of their commitments to reduce or limit emissions.

These techniques to grasp bottom-up uncertainty are described in detail in [14] and are applied to past (see Section 2.3) and pledged emissions (see Section 2.4) so far.

2.7. Land Use and Land-use Change Management till 2050

LU and LUC (land-use change) are discussed widely – and controversially, either optimistically while focusing on selected regions and opportunities to reduce emissions, or unenthusiastically while focusing on the global carbon budget and LU mismanagement as an important, socio-economically rooted part of it.


125

To tackle deforestation and other LU mismanagement, (i) we treat the LU sector separately from anthropospheric GHGs; and (ii) express global LU on a per-capita basis by country to ‘personalize’ carbon debts (primary data source: [19] plus UN Pop Div) while accounting for the countries’ spatial disconnect between biomass production and consumption. The possibility to grasp the spatial disconnect between biomass production and biomass consumption became available only recently [22], which allows making the step also for LU from global to national (work ahead).

To start with, our working hypothesis is that deforestation will cease and other land use will become sustainable until 2050. The validity of this assumption will feed back on the countries’ anthropogenic per-capita emission target paths introduced in Section 2.2 above.

3. Results

Figure 1 below shows the provisional result typical for a data-rich country (here: the US; with pledges as made in 2009).

4. Preliminary Discussion and Summary

This study describes work in progress. So far, we explored (and still do) a contraction & convergence approach which is based on a global cumulative emissions constraint and which will lead to a universally valid emissions equity target in 2050. We still need to explore alternatives, such as regional GHG emission targets for 2050 based on equal marginal mitigation costs from so-called first-best scenarios which assume full integration and efficient carbon markets.

Our insights so far are:

− The concept of a cumulative emissions constraint adds complexity and will make international negotiations more difficult.

− The consequences of a cumulative emissions constraint are frequently underestimated: (i) No country can escape such a constraint, e.g., by choosing individual start and/or base years for accounting its emissions; and (ii) the cumulative emissions of a country above and below its reference emissions target path must balance (here by 2050). This leads us to conclude that it appears unrealistic to achieve the 2 oC temperature target in particular.

− Bottom-up uncertainty is still under-explored and, most likely, underestimated: (i) Reducing the risk interval (here: 10–43 %) of exceeding the given temperature target (here: 2 oC by 2050) translates to increased per-capita emission intervals (work ahead); and (ii) the concept of grasping bottom-up uncertainty could also be used to neutralize top-down uncertainty and the risk of exceeding the agreed temperature target (work ahead). It is recalled that our comparison of bottom-up emission estimates with top-down, scenario-derived constraints builds on the assumption that no unaccounted emissions exist.

− We rely on our future already today; in particular,

• we hope that an alternative, market-based approach renders it possible to reach the 2 oC temperature target.

• we hope that deforestation will cease and other land use will become sustainable until 2050.

• we hope that bottom-up and top-down uncertainty decrease.


126

Figure 1: The figure shows that each individual within the United States (US) must reduce on average his or her GHG emissions by 88% between 1990 and 2050. The task is equally large for other industrialized countries (RU: 87%, EU-27: 76%). Achieving such a target limits the risk of global temperatures rising beyond 2°C to 10–43 % and so avoids global warming consequences beyond this ‘threshold’.

The solid grey line indicates a reference pathway that emissions must follow between 1990 and 2050 to achieve a universal per-capita target of 2.9 t CO2-eq which directly follows from [7]. Countries that emit quantities above this line will need to compensate by emitting below the grey line before 2050 to ensure the target is reached. The solid black curve shows the emissions of the six GHGs (without LU) between 1990 and 2006 as reported by the US to the UNFCCC. Underneath, the red line shows what the per-capita emission levels that the US would have committed to had it ratified the Kyoto Protocol. The solid black dot represents the US estimated emissions for 2010 according to IIASA’s GAINS model (http://gains.iiasa.ac.at/index.php/home-page). The dashed blue and orange lines show expected per-capita emissions according to the ‘2020 pledges’ (conservative and optimistic) made by the US. The costs for achieving these pledges by applying known mitigation techniques are also mentioned in the blue and orange-framed boxes (output of GAINS).

The solid green line is an estimate of the emissions from LU on the territory of the US and the thin brown line shows the land use, land-use change and forestry (LULUCF) emissions reported by the US under the UNFCCC. The broken green line is a zero-order estimate of the US’ share of emissions from global LU if a consumption, not a production, approach is followed in accounting emissions.7 The dotted grey line represents the path to lower these emissions to zero by 2050 assuming that emissions from LU achieve sustainability (which remains to be questioned).

The ranges shown numerically in the boxes and graphically by the ‘I’ shape reflect the range of uncertainty in estimating emissions. 7 The US’ consumed share of global LU emissions has been derived via the ratio of its national to global fossil fuel (FF) emissions. This is a zero-order approach motivated by the fact that sparsely populated regions are mainly net producers of biomass while densely populated regions are net consumers, independent of development status (global view; K. Erb, 2009: pers. comm.). The only purpose of this approach is to show the direction in which a ‘consumption accounting’ will pull a ‘production accounting’. Yet, the magnitude is unclear. The next step in quantifying the spatial disconnect between biomass production and biomass consumption will be to test the concept of ‘embodied human appropriation of net primary production’ [22].


127

References

[1] FCCC (2009a): Copenhagen – Background Information. Fact Sheet, UN Framework Convention on Climate Change (FCCC), Bonn, Germany, pp. 3. Available at: http://unfccc.int/files/press/application/pdf/fact_sheet_copenhagen_background_information.pdf.

[2] FCCC (2009b): Why is an International Climate Change Deal so Important? Fact Sheet, UN Framework Convention on Climate Change (FCCC), Bonn, Germany, pp. 2. Available at: http://unfccc.int/files/press/fact_sheets/application/pdf/fact_sheet_ climate_deal.pdf.

[3] FCCC (2009c): Copenhagen Accord. Advance Unedited Version of Decision -/CP.15, UN Framework Convention on Climate Change (FCCC), Bonn, Germany, pp. 6. Available at: http://unfccc.int/files/meetings/cop_15/application/pdf/cop15_cph_auv.pdf.

[4] Schiermeier Q. (2009): G8 climate target questioned. Nature, 460(7253), 16 July, 313, doi: 10.1038/460313a.

[5] WBGU (2009a): Climate Change: Why 2oC? Factsheet 2/2009, German Advisory Council on Global Change (WBGU), Berlin, Germany, ISBN: 3-936191-33-6. Available at: http://www.wbgu.de/wbgu_factsheet_2_en.html.

[6] WBGU (2009b): Solving the Climate Dilemma: The Budget Approach. Special Report, German Advisory Council on Global Change (WBGU), Berlin, Germany, pp. 55, ISBN: 978-3-936191-27-1. Available at: http://www.wbgu.de/wbgu_sn2009_en.pdf.

[7] Meinshausen M., Meinshausen N., Hare W., Raper S.C.B., Frieler K., Knutti R., Frame D.J. and Allen M.R. (2009): Greenhouse-gas emission targets for limiting global warming to 2oC. Nature, 458 (30 April 2009), 1158–1162, doi: 10.1038/nature08017.

[8] Allen M.R., Frame D.J., Huntingford C., Jones C.D., Lowe J.A., Meinshausen M. and Meinshausen N. (2009): Warming caused by cumulative carbon emissions toward the trillionth tonne. Nature, 458 (30 April 2009), 1163–1166, doi: 10.1038/nature08019.

[9] EU (2009): Decision No 406/2009/EC of the European Parliament and of the Council of 23 April 2009 on the Effort of Member States to Reduce Their Greenhouse Gas Emissions to Meet the Community’s Greenhouse Gas reduction Commitments up to 2020. Official Journal of the European Union, L 140, 52(5 June 2009), 136–148, ISSN: 1725-2555. Available at: http://eur-lex.europa.eu/JOHtml.do?uri=OJ:L:2009:140:SOM:EN:HTML.

[10] Jonas M., Marland G., Winiwarter W., White T., Nahorski Z., Bun R. and Nilsson S. (2010); Benefits of dealing with uncertainty in greenhouse gas inventories: introduction. Clim. Change, doi: 10.1007/s10584-010-9922-6.

[11] Victor D.G. (2009): Global warming: why the 2 oC goal is a polticial delusion. Nature, 459 (18 June 2009), 909, doi: 10.1038/459909a.

[12] WBGU (2005): Scenario for the Derivation of Global CO2 Reduction Targets and Implementation Strategies. Statement on the Occasion of the First Conference of the Parties to the Framework Convention on Climate Change in Berlin. Special Report, German Advisory Council on Global Change, Bremerhaven, Germany. Available at: http://www.wbgu.de/wbgu_sn1995_engl.pdf.

[13] Lieberman D., Jonas M., Nahorski Z. and Nilsson S. (eds.) (2007): Accounting for Climate Change. Uncertainty in Greenhouse Gas Inventories – Verification,


128

Compliance, and Trading. Springer, Dordrecht, Netherlands, pp. 159, ISBN: 978-1-4020-5929-2 [Reprint: Water Air Soil Pollut.: Focus, 2007, 7(4–5), ISSN: 1567–7230]. Available at: http://www.springer.com/environment/global+change+-+climate+change/book/978-1-4020-5929-2.

[14] Jonas, M., Gusti M., Jęda W., Nahorski Z. and Nilsson S. (2010): Comparison of preparatory signal analysis techniques for consideration in the (post-) Kyoto policy process. Clim. Change, doi: 10.1007/s10584-010-9914-6.

[15]: IIASA (2010): GAINS: Comparing GHG Mitigation Efforts between Annex I Countries. Atmospheric Pollution & Economic Development Program, International Institute for Applied Systems Analysis, Laxenburg, Austria. Available at: http://gains.iiasa.ac.at/index.php/gains-annex-1.

[16] IIASA (2007): Energy Modeling Framework. Energy Program, International Institute for Applied Systems Analysis, Laxenburg, Austria. Available at: http://www.iiasa.ac.at/Research/ENE/model/.

[17] FCCC (1998): Report of the Conference of the Parties on Its Third Session, Held at Kyoto from 1 to 11 December 1997. Addendum. Part Two: Action Taken by the Conference of the Parties at Its Third Session. Document FCCC/CP/1997/7/Add.1, UN Framework Convention on Climate Change (FCCC), Bonn, Germany. Available at: http://unfccc.int/resource/docs/cop3/07a01.pdf.

[18] IIASA (2010): 2007 Update of Probabilistic World Population Projections. World Population Program, International Institute for Applied Systems Analysis, Laxenburg, Austria. Available at: http://www.iiasa.ac.at/Research/POP/proj07/index.html.

[19] Le Quéré C., Raupach M.R., Canadell J.G., Marland G. et al. C. et al. (2009): Trends in the sources and sinks of carbon dioxide. Nature Geosci., 2, 831–836, doi: 10.1038/ngeo689 (see also http://www.globalcarbonproject.org/carbonbudget/index.htm).

[20] Scheehle E., D. Godwin D. and Ottinger D. (2006): Global Anthropogenic Non-CO2 Greenhouse Gas Emissions: 1990 – 2020. Report 430-R-06-003 (rev.), Environmental Protection Agency, Washington, D.C., United States of America. Available at: http://www.epa.gov/climatechange/economics/ downloads/GlobalAnthroEmissionsReport.pdf.

[21] Amann A., Bertok I., Borken J., Cofala J., Heyes C., Hoglund L., Klimont Z., Purohit P., Rafai P., Schöpp W., Toth G., Wagner W. And Winiwarter W. (2009): GAINS. Potential and Costs for Greenhouse Gas Mitigation in Annex 1 Countries: Initial Results. Report, International Institute for Applied Systems Analysis, Laxenburg, Austria. Available at: http://gains.iiasa.ac.at/gains/reports/AnnexI-results.pdf (see also http://gains.iiasa.ac.at/index.php/reports).

[22] Erb K.-H., Krausmann F., Lucht W. and Haberl H. (2009): Embodied HANPP: Mapping the spatial disconnect between global biomass production and consumption. Ecol. Econ., 69(2), 328–334 doi: 10.1016/j.ecolecon.2009.06.025.


129

The potential impact of CO2 and air temperature increases on krummholz’s transformation into arborescent form in the southern

Siberian Mountains

Vyacheslav I. Kharuk, Maria L. Dvinskaya, Sergey T. Im

V. N. Sukachev Institute of Forest SB RAS, Academgorodok 50-28, Krasnoyarsk, 660036, Russian Federation

[email protected]

Abstract

The vertical transformation of mat and prostrate krummholz forms of larch (Larix sibirica Ledeb) and Siberian pine (Pinus sibirica Du Tour) has been documented for the southern Siberian Mountains. An analysis of the radial growth increments showed that the widespread vertical transformation of krummholz begin in the mid of 1980s. Radial and apical growth increments correlated with increase in air temperature (summer and “cold period”) and CO2. In areas where Siberian pine grew in krummholz form, larch attained a vertical growth form.

Keywords: alpine forest-tundra ecotone, treeline, climate change, CO2 elevation, krummholz, Larix sibirica, Pinus sibirica, Siberian Mountains

Introduction

The climate-driven impact on tree vegetation is expected to be most significant in the zones where climate variables limit tree growth, e.g., in the alpine forest-tundra ecotone [1]. It has been reported that during the last few decades, increased stand density and tree growth increments have been observed in the northern forest-tundra ecotone, which includes alpine and non-alpine ecotones [2, 3, 4, 5]. A number of advancing treeline cases have been reported for the alpine forest-tundra ecotones of the European, American, Siberian and Ural mountains [6, 7, 8, 2, 9, 10, 4, 11]. It has also been observed that species more common in the south have begun to appear in the northern forest communities [12, 13, 14, 15, 16, 17, 18]. Milder climates are promoting changes in tree morphology, i.e. transforming the mat and prostrate krummholz into their vertical form [19, 20, 4, 10].

Besides the warmer ambient air, tree growth is also facilitated by higher levels of CO2. Physiological growth modeling predicts tree growth rises from the combined effect of CO2 and air temperature increases [21]. Based on a series of elevated CO2 level experiments Ainsworth and Long [22] conclude that the trees were responding to the higher CO2 concentrations by growth and above-ground production increase. Norby et al. [23] analyzed the available data on elevated CO2 simulations, concluding that the response of forest net primary productivity (NPP) to elevated CO2 is highly conserved across a broad range of productivity, with a stimulation at the median level of 23±2%.

The most significant changes in temperature is observed in, and predicted for, Siberia [24], but there are still few studies of climate-driven changes for Siberian forests [3, 4, 9, 10].


130

The main goal of this paper is to analyze the climate-induced changes to tree physiognomy in the alpine forest-tundra ecotone of the southern Siberian Mountains.

The specific questions are: 1. How widespread is the phenomenon of mat and krummholz transforming into its

arboreal form? 2. When do trees start turning from krummholz into arborescent form? 3. Was this transformation connected with climate variables and CO2 concentration

in the ambient air? We measured the radial and apical larch tree increments of the alpine forest-tundra

ecotone.

Materials and methods

The studies were conducted at the Sengilen Ridge, which is located in the Altai-Sayan mountain region; this is a transition area between boreal forests and Mongolian steppes (Fig. 1). The Altai-Sayan mountain region is composed of a system of ridges with elevations up to 4500 m that is divided by a dense drainage network. The region has a severe continental climate. In January, the temperature ranges between –32°С and –12°С. The temperature ranges between +9°С and +18°С in July. Precipitation is about 390 mm yr –1 (Fig. 2). The dominant species is larch (Larix sibirica Ledeb), with an admixture of Siberian pine (Pinus sibirica Du Tour). The life span of study area’s larch and Siberian pines is 600–700 yrs and 400–500 yrs, respectively.

Figure 1. The area of investigation (red box); the Sengilen ridge is located in Siberia’s

southern Altai-Sayan Mountains. Insert: a “post-krummholz” larch. Lower part of the crown approximately corresponds to the winter snow level


131

The region of study included the upper forest elevation belt, and the forest-alpine tundra ecotone. The “forest-alpine tundra ecotone” is defined as the transitional area between the regeneration line and the upper limit of closed forests. “Post-krummholz” trees (i.e., trees which attained vertical growth form) were sampled on seven sites, whose elevations ranged between 2400 and 2600 m (Fig. 1). These aboveground studies were conducted in 2007. The trees were sampled at their base (sample size – 13 trees). The mean tree height was 2.6 m; the mean age was 74 yr; the trees grew at a mean elevation of about 2500 m.

The radial growth increments were measured using LINTAB-III equipment with a precision of 0.01 mm. The master chronology [25] was generated on the basis of eight living larch trees. The master chronology covered the years 1900–2006, and had mean correlation and sensitivity values were 0.66 and 0.33, respectively. The COFECHA [26] and TSAP [27] programs were used to detect double-counted and missing rings based on the master chronology.

The apical growth increment measurements covered the distance between the whorls for the entire measurement period: 15–25 yr, depending on tree sample. Trees were sampled at the mean elevation of 2420 m; there were 28 trees per sample (both post-krummholz and “normal” trees), and they had a mean height 5.8 m. Since growth increments of individual trees varied depending on local site conditions, these readings were normalized for comparison purposes. The increment values (i.e. tree ring widths and distances between the whorls) of a given sample were normalized through a summation of their annual totals. These were made equal to 1.0. The increment value of a given year was taken to be equal to its proportion of the total sum.

The data was processed with the sliding window method of exponential approximation; this tool was used to detect temporal trends [28]. The CO2 dynamics of ambient air were acquired with Tans [29]. The temperature and precipitation data originated from Mitchell and Jones [30]. This information source provides the meteorological data averages of half-degree grid cells (or 35x55 km); the mean elevation of this area was about 2300 m. Excel and StatSoft software [28] were used for statistical analysis.

Results

Correlations

The radial increments had a positive correlation to the summer (JJA) and “cold period” (September-May) temperatures (Fig. 3). The correlation value depended on the period of observation: for summer temperatures, this correlation peaked at a value of 0.65 for a period of about 20 years; for “cold period” temperatures, this correlation depended on the length of the study period (Fig. 3). No significant correlations were observed for precipitation, with the exception of “cold period” precipitation (Fig. 3). A high correlation was observed for CO2 concentrations (R = 0.72; analyzed period 1959–2006 yr). The summer temperatures correlated over that period at a value of 0.57. There were also significant correlation between apical growth increment and summer (R=0.55), cold period (R= 0.37) temperatures, and CO2 concentration (R= 0.83).

Multiple correlation analyses showed that only two factors, CO2 and summer temperatures, were significant. The radial increments correlated to CO2 concentrations (R = 0.59; p > 0.001), and to the summer temperature readings (R = 0.36; p > 0.001). Both factors had a regression coefficient of 0.8 (p > 0.001).


132

-4

-3

-2

-1

0

1

2

3

4

1900 1920 1940 1960 1980 2000

Ann

ual m

ean

sum

mer

tem

pera

ture

an

omal

y (°C

)

Year

(a)

-4

-3

-2

-1

0

1

2

3

4

1900 1920 1940 1960 1980 2000

Ann

ual m

ean

win

ter

tem

pera

ture

an

omal

y (°C

)

Year

(b)

-150

-100

-50

0

50

100

150

1900 1920 1940 1960 1980 2000

Ann

ual s

umm

er p

reci

pita

tion

ano

mal

y (m

m)

Year

(c)

-150

-100

-50

0

50

100

150

1900 1920 1940 1960 1980 2000

Ann

ual w

inte

r pr

ecip

itatio

n an

om

aly

(mm

)

Year

(d)

Figure 2. Summer (a, c) and winter (b, d) temperature and precipitation anomalies

for the Sengilen ridge. The solid line shows a 10-yr filter

0

0.1

0.2

0.3

0.4

0.5

0.6

194019451950195519601965197019751980

R

Year

123p>0.05p>0.01

Figure 3. The correlation between radial increments, air temperature and precipitation.

The given data corresponds to the “sliding” period of analysis (e.g., for 1980, this correlation was calculated for the years of 2006 and 1980; for 1960, this period spanned 2006 through 1960, and so on). 1 – summer (JJA); 2 – “cold period” (September – May) temperatures; 3 – “cold period” precipitation


133

Arboreal vs krummholz

A widespread transformation of krummholz into its vertical form was observed (inserts on Figs. 1 and 4). Notably, in areas where larch can support vertical growth (within upper tree limit), Siberian pines also attain a prostrate growth (Fig. 4, insert). “Post-krummholz” radial and apical growth increments strongly increased over the last two decades (Fig. 4). The beginning of radial growth increase was observed after the minimum in the year 1987 (Fig. 4).

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

1920 1940 1960 1980 2000

Incr

emen

ts (r

elat

ive

un

its)

Year

1

2

3

Figure 4. The radial (N = 13, h ≈ 2.6 m, A ≈ 74 yr) and apical (N = 28, h ≈ 5.8 m)

dynamics of “post-krummholz” larch tree growth increment.1, 2 – radial increments (1– individual, 2 – mean values); 3 – apical increments. Insert photo (left): multistem “post-krummholz” larch and prostrate Siberian pine. Apical shoot damage and needle discoloration cuased by winter dissectaion. Insert photo right: bark chlorophyll is facilitating trees’ survival in harsh environment; it was found even in some dissication-damaged apical leaf-less shoots

In order to emphasize temporal trends, we filtered the mean increment and summer temperature data by exponential filter (with each widow equal to 3 yrs; Fig. 5). Fig. 5 data show how the negative incremental growth rate turned positive. (It is necessary to note that the filter deflected the extreme position of year seven; Fig. 5). This coincided (with a shift about 3 yrs) to similar changes in the summer temperature (Fig. 5). The CO2 concentration rates were almost linear for the entire period of observations (Fig. 5).


134

7

8

9

10

11

12

13

14

0

0.01

0.02

0.03

0.04

1940 1950 1960 1970 1980 1990 2000

Mea

n s

um

mer

tem

pera

ture

( C

)

Mea

n in

crem

ent (

rela

tive

un

its)

Year

1

2

3

4

300

320

340

360

380

CO

2(p

pm)

o

Figure 5. Radial increment and summer temperature dynamics: initial data and exponential

smoothing results. The filter shifted the extreme position of year seven

Discussion

Thus two factors, CO2 and summer temperatures, described radial increment variability better than the lone CO2 data (R = 0.8 vs R = 0.72). This may indicate the synergy effect of CO2 and ambient air temperature increases. The high correlation to CO2 concentrations does not necessarily mean a cause-and-effect dependence. But it is known that CO2 is a fertilizer, and its present ambient air concentration is below the saturation level. In other words, CO2 concentrations are still one of the limiting factors of tree growth. Moreover, CO2 concentrations decrease with drops in barometric pressure. This means that CO2 limits tree growth stronger in the highlands than in the lowlands. In our case, at elevations of about 2500 m, CO2 concentration is about 25% lower (based on barometric formula) than in lowlands. This implies that the fertilization effect of CO2 increases will have a more pronounced effect on alpine forest-tundra zones.

Surprisingly, the growth increments’ correlation to summer precipitation was not significant, where as “cold period” precipitation had a positive impact (Fig. 3). One potential explanation involves the highly stochastic pattern of summer rain activity (Fig. 2c). The other possible reason is that summer precipitation does not limit larch growth; which has been known to survive semi-arid levels of precipitation (~250 mm yr –1) [31]. The third factor is rain-induced temperature decreases: even the midsummer months can cover the mountains in snow or hail. Winter (or “cold period”) precipitation displayed a positive trend in decade (Fig. 2d) that facilitated tree growth through higher soil water content.

A widespread transformation of krummholz into its vertical forms was observed in the study area. Krummholz forms have now been observed only at their upper tree limits, and have been evidenced primarily by Siberian pines (Fig. 4). The begin of these transformations were referred to 1987 and coincides with upturns in the summer temperatures readings, which moved from negative to positive trends (Fig. 5). This date approximately coincides to the period when winter temperatures surpassed their 20th


135

century mean value reading (Fig. 2b). Notably, according to Kullman, this coincides with an increase in tree growth and the upward migration of trees to the western end of Eurasia, in Swedish Scandia [32].

CO2 concentration increases were approximately linear for the observed period (Fig. 3). This indicates that temperature is still the main limiting factor of tree growth in the forest-tundra ecotone. The temperature increase is not quite as straightforward: periods of extremely cold winter cause apical increment damage and mortality (Fig. 4). The synergy of low temperatures and winds caused shoot and needle desiccation and decreased vertical growth (Fig. 4, insert). The significance of winter temperatures to the transformation of krummholz into its vertical forms has been reported earlier [32]. Apical shoot mortality caused multiple stem formation that resulted from the lateral branches’ vertical growth. The multiple stem larches and Siberian pine trees were observed in the zones below their present tree lines. We consider this as an evidences of former tree line position.

Currently, at the Sengilen ridge, larch is much less likely to be found in its krummholz form than the Siberian pine. This is because larch’s dense bark and deciduous leaves provide it with much better resistance against the harsh Siberian climate. We cannot compare how larch and Siberian pine respond to climate change in the study area – the canopy’s proportion of Siberian pine is too low. But in the areas where both species were present, we found indications that positive trends in temperature and precipitation had a more favorable effect on the Siberian pines (with respect to growth and regeneration amounts) [10].

Finally, there is another factor that facilitates tree survival and post-krummholz formation in harsh environments: this is involves “bark photosynthesis”. According to our observations, even some leafless shoots were found to have been alive due to the bark chlorophyll which was observed on the shoot cross-sections (Fig. 4, insert). For some species, the bark was responsible for an estimated 10 to 15 percent of the tree’s entire carbon balance [33]. But the estimated upper-tree-line significance of non-leaf photosynthesis requires further study.

Acknowledgments

This research was supported by Siberian Branch Russian Academy of Science Program # 23.3.33.

References

[1] Holtmeier F.-K. and Broll G. (2007): Treeline advance – driving processes and adverse factors. Landscape Online Vol. 1, pp. 1–33.

[2] Kharuk V. I., Ranson K. J., Im S. T. and Dvinskaya M. L. (2006): Forest-tundra larch forests and climatic tends. Russian journal of ecology Vol. 37, No. 5, pp. 291–298.

[3] Shiyatov S. G., Terent’ev M. M., Fomin V. V. and Zimmermann N. E. (2007): Altitudinal and horizontal shifts of the upper boundaries of open and closed forests in the Polar Urals in the 20th century. Russian Journal of Ecology Vol. 4, No. 38, pp. 223–227.


136

[4] Devi N., Hagedorn F., Moiseev P., Bugmann H., Shiyatov S., Mazepa V. and Rigling A. (2008): Expanding forests and changing growth forms of Siberian larch at the Polar Urals treeline during the 20th century. Global Change Biology Vol. 14, No. (7), pp. 1581–1591.

[5] Harsch M., Hulme P., McGlone M., Duncan R. (2009): Are treelines advancing? A global meta-analysis of treeline response to climate warming. Ecology Letters Vol. 12, No. 1, pp. 1040–1049.

[6] Theurillat J. P. and Guisan A. (2001): Potential impact of climate change on vegetation in the European Alps: a review. Climatic Change Vol. 50, pp. 77–109.

[7] Baker B. B. and Moseley R. K. (2007): Advancing treeline and retreating glaciers: implications for conservation in Yunnan, P. R. China. Arctic, Antarctic and Alpine Research Vol. 39, No. 2, pp. 200–209.

[8] Kullman L. (2007): Treeline population monitoring of Pinus sylvestris in the Swedish Scandes, 1973–2005: implications for treeline theory and climate change ecology. Journal of Ecology Vol. 95, pp. 41–52.

[9] Kharuk V. I., Dvinskaya M. L., Im S. T. and Ranson K. J. (2008): Tree vegetation of the forest-tundra ecotone in the Western Sayan Mountains and climatic trends. Russian Journal of Ecology Vol. 39, No. 1, pp. 8–13.

[10] Kharuk V. I., Ranson K. J., Im S. T. and Dvinskaya M. L. (2009): Pinus sibirica and Larix sibirica response to climate change in Southern Siberian alpine forest-tundra ecotone. Scandinavian Journal of Forest Research Vol. 24, No, 2, pp. 130– 39.

[11] Lenoir J., Gegout J. C., Marquet P. A., de Ruffray P. and Brisse H. (2008): A significant upward shift in plant species optimum elevation during the 20th century. Science Vol. 320, No. 5884, pp. 1768–1771.

[12] Luckman B. H. and Kavanagh T. (2000): Impact of climate fluctuations on mountain environments in the Canadian Rockies. Ambio Vol. 29, pp. 371–380.

[13] Payette S., Fortin M. and Gamache I. (2001): The subarctic forest-tundra: the structure of a biome in a changing climate. BioScience Vol. 51, No. 9, pp. 709–718.

[14] Grace J., Berninger F. and Nagy L. (2002): Impacts of climate change on the treeline. Annals of Botany Vol. 90, pp. 537–544.

[15] Klasner F. L. and Fagre D. B. (2002): A half century of change in alpine treeline patterns at Glacier National Park, Montana, U.S.A. Arctic, Antarctic and Alpine Research Vol. 34, pp. 49–56.

[16] Smith W. K., Germino M. J., Hancock T. E. and Johnson D. M. (2003): Another perspective on altitudinal limits of alpine timberlines. Tree Physiology Vol. 23, pp. 1101–1112.

[17] Esper J. and Schweingruber F. H. (2004): Large-scale treeline changes recorded in Siberia. Geophysical Research Letters Vol. 36, No. 6, DOI 10.1029/2003GL019178.

[18] Kharuk V. I., Dvinskaya M. L., Ranson K. J. and Im S. T. (2005): Expansion of Evergreen Conifers to the Larch-Dominated Zone and Climatic Trends. Russian Journal of Ecology Vol. 36, No. 3, pp. 164–170.


137

[19] Holtmeier F. K., (2003): Mountain Timberlines: Ecology, Patchiness, and Dynamics.Kluwer Academic Publishers, Netherlands , 438 pp.

[20] Gamache I. and Payette S. (2004): Height growth response of treeline black spruce to recent climate warming across the forest-tundra of eastern Canada. Journal of Ecology Vol. 92, pp. 835–845.

[21] Matala J., Ojansuu R., Peltola H., Sievänen R. and Kellomäki S. (2005): Introducing effects of temperature and CO2 elevation on tree growth into a statistical growth and yield model. Ecological Modelling Vol. 181, No. 2-3, pp. 173–190.

[22] Ainsworth E. A. and Long S. P. (2005): What have we learned from 15 years of free-air CO2 enrichment (FACE)? A meta-analytic review of the responses of photosynthesis, canopy properties and plant production to rising CO2. New Phytologist Vol. 165, No. 2, pp. 351–71.

[23] Norby R. J., DeLucia E. H., Gielen B., Calfapietra C., Giardina C. P., King J. S., Ledford J., McCarthy H. R., Moore D. J. P., Ceulemans R., De Angelis P., Finzi A. C., Karnosky D. F., Kubiske M. E., Lukac M., Pregitzer K. S., Scarascia-Mugnozza G. E., Schlesinger W. H., and Oren R. (2005): Forest response to elevated CO2 is conserved across a broad range of productivity. Proceedings of the National Academy of Sciences,Vol. 102, pp. 18052–18056.

[24] IPCC (2007): Climate Change 2007: Synthesis Report. Valencia, Spain. (http://www.ipcc.ch/pdf/assessment-report/ar4/syr/ar4_syr_spm.pdf). Last accessed on 14 December 2009.

[25] Fritts H. C. (1991): Reconstruction Large-scale Climatic Patterns from Tree-Ring Data: A Diagnostic Analysis. University of Arizona Press, Tucson-London, pp. 286.

[26] Holmes R. L. (1983): Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin Vol. 43, pp. 69–78.

[27] Rinn F. (1996): TSAP V 3.6 Reference manual: computer program for tree-ring analysis and presentation. Heidelberg, Germany, pp. 263.

[28] StatSoft Inc., (2001): Electronic book of statistics (in Russian). Moscow: StatSoft. (http://www.statsoft.ru/home/textbook/default.htm). Last accessed on 14 December 2009.

[29] Tans P. (2010): NOAA/ESRL. (www.esrl.noaa.gov/gmd/ccgg/trends/). Last accessed on 2 August 2010.

[30] Mitchell T. D. and Jones P. D. (2005): An improved method of constructing a database of monthly climate observations and associated high resolution grids. International journal of climatology Vol. 25, No. 6, pp. 693–712.

[31] Kloeppel B. D., Gower S. T., Trechel I. W. and Kharouk V. I. (1998): Foliar carbon isotope discrimination in Larix species and sympatric evergreen conifers: a global comparison. Oecologia Vol. 114, pp.153–159.

[32] Kullman L. and Kjällgren L. (2006): Holocene pine tree-line evolution in the Swedish Scandes: recent tree-line rise and climate change in a long-term perspective. Boreas Vol. 35, pp. 159–168.


138

[33] Kharouk V. I., Middleton E. M., Spencer S. L., Rock B. N and Willams D. L. (1995): Aspen bark photosynthesis and its significance to remote sensing and carbon budget estimates in the boreal ecosystem. Journal of Water, Air and Soil Pollution Vol. 82, pp. 483–497.


139

Peculiarities of sequestered carbon assessment in urban forests of Kyiv

Ivan Lakyda

National University of Life and Environmental Sciences of Ukraine Geroiv oborony str, 15, Kyiv, Ukraine

[email protected]

Abstract

Peculiarities of sequestered carbon assessment in urban forests of Kyiv are discussed. Within the framework of the investigation we estimated the amount of sequestered carbon and discussed the corresponding uncertainty. The input data from base forest inventory in three forest-park economies of Kyiv city was processed in accordance with the technique for live biomass and sequestered carbon amounts estimation previously developed for Ukrainian forests. It is found that maturity age and age structure has significant impact upon the results of the computation. The reason for this is realization of the specialized software used for this assessment (CARBON programme). Possible ways for improving the software are suggested.

Keywords: Urban forests, maturity age, age structure, carbon assessment.

1. Introduction

Interest to carbon-depositing functions of forest has risen considerably during last decades. These functions of forest ecosystems acquire special significance in pre-urban and urban forests. These forests perform not only stabilizing and protective functions, but also are a renewable resource. Due to rapid development of modern technologies (bioenergy), this feature becomes more and more important. However, currently there is a lack of scientifically based assessments of carbon-depositing function of urban forests. Within the transition of global forestry to the sustainable forest management paradigm, it is very important to investigate ecological, economic and social aspects of forest objects. Forests, which are located within the borders and (or) nearby urban settlements, (i) are commonly used for recreation of urban inhabitants, (ii) holding cultural and curative services and (iii) are used for keeping favorable ecological conditions. These forests are complex and have diverse nature and are crucial for sustaining urban environment stability functions (Federal Agency of Forest Management of the Russian Federation, 2009).

The results of assessment of carbon-sequestering functions of forest stands may be influenced both by uncertainties, systematic errors and crude errors of forest inventory data, as well as by uncertainties, generated by data processing technique. Within the framework of this research, the influence of the latter group is investigated. Every model, system of models or data processing technique is characterized by data range, within which it produces robust results. Reliability problems may occur both when attempting to produce the result for a dataset outside definitional domain of a model and for a dataset structurized differently than a stereotype built into the model by its developers. Aiming at minimization of influence of technique upon reliability of produced results it is worthwhile to include various data processing strategies or scenarios. Robustness of models, systems of models and data processing techniques may be enhanced in this way.


140

With respect to urban forests of Kyiv city, the main aim of this study was to assess and quantify carbon sequestration, to point out some peculiarities and to discuss uncertainty of this assessment. The assessment is based upon generalized forest fund register of growing stock volumes at level of forestry enterprises or other territorial consolidations and calculated with a system of mathematical models which connect main live biomass components with standing stocks of studied forest stands (Lakyda, 1997). Although this approach gives more generalized results, for practical reasons, it was the only possible way, taking into account present state of information support of Ukrainian forestry.

2. Material and method

The source data used for performing this study were materials of forest inventory of Communal enterprises “Darnytsya forest-park economy”, “Svyatoshyn forest-park economy”, and “Koncha-Zaspa forest park economy” of communal association “Kyivzelenbud”. The dataset from 1999 contains information on areas and growing stocks of forest stands divided by age classes, tree species, and site index classes. The initial dataset needed transformation and data consolidation in order to be presented in a form, which was needed for further calculations with specialized software (the CARBON programme). As a result, characteristics of individual parameters of forest fund of the study object were obtained:

1. Distribution of forested land area and stock by groups of forest-forming species (coniferous, hardwood broadleaves, softwood broadleaves);

2. Percentages of stock of main forest-forming species (pine, spruce, oak, beech, birch, aspen, alder) within the groups of forest-forming species (coniferous, hardwood broadleaves, softwood broadleaves);

3. Distribution of stocks of stands within groups of forest-forming species by age groups (young, mid-age, premature, mature and overmature (Normative and reference materials for forest mensuration in Ukraine and Moldavia, 1987);

4. Average site index classes by M.M. Orlov by groups of forest-forming species.

During data processing, maturity age for all the groups of forest-forming species was set to be equal to protective maturity age of the corresponding species (Normative and reference materials for forest mensuration in Ukraine and Moldavia, 1987). This approach was also used for assigning of age classes to age groups. Average site index classes were set by M.M. Orlov’s scale as weighted average values of individual site index classes to corresponding area.

Carbon sequestration in forests is indissolubly connected with their live biomass production. The algorithm for estimation live biomass and sequestered carbon used in this study was developed and implemented by prof. P. Lakyda in the CARBON programme (Lakyda, 1997).

3. Results and Discussion

The results of calculation of live biomass and sequestered carbon amounts are presented in Table 1.


141

Table 1. Amounts of live biomass and sequestered carbon in urban forests of Kyiv (state on 01.01.1999)

Group of forest-

forming species

Fores-ted

land, 103 ha

Standing stock, 106 m3

Live biomass components, Tg

Live biomass density, kg·m-2

Carbon

leav

es (

nee

dle

s)

bra

nch

wo

od

an

d b

ark

stem

wo

od

an

d b

ark

roo

ts

un

der

sto

rey

veg

etat

ion

tota

l

tota

l, T

g

den

sity

, kg·m

-2

TOTAL 31,3 10,5 0,12 0,49 4,73 1,05 0,29 6,68 21,3 3,32 10,6

including:

Coniferous 26,3 0,10 0,39 4,07 0,90 0,28 5,73 21,8 2,85 10,8

Hardwood broadleaves 3,2 0,01 0,08 0,49 0,11 0,01 0,70 22,4 0,35 11,1

Softwood broadleaves

1,8 0,00 0,02 0,17 0,05 0,01 0,24 13,2 0,12 6,6

One of the important features of urban forests of Kyiv with respect to live biomass and sequestered carbon assessment is their protection category, maturity age, age class distribution and age structure, since difference in these parameters may substantially impact the final result. The abovementioned parameters differ significantly between production forests and protective forests. In the studied forests maturity age values are set to be equal to the age of protective maturity (Normative and reference materials for forest mensuration in Ukraine and Moldavia, 1987). Correspondingly, distribution of stands by age groups is made as following (Table 2):

Table 2. Age criteria for distribution of stands by age groups and groups of forest-forming tree species

Groups of species Age groups

Coniferous Hardwood broadleaves


Young stands 1-40 1-40 1-20

Table 2 (continued) Groups of species

Age groups

Coniferous Hardwood broadleaves


Mid-aged stands 41-100 41-110 21-50 Pre-mature stands 101-120 111-130 51-60 Mature and over-mature stands

121 and older 131 and older 61 and older

The grouping described above is applied to initial dataset in order to present it in a form, required for further processing by CARBON programme. This software was developed and calibrated for assessing amounts of live biomass and sequestered carbon mainly in productive forests, where maturity age, age class distribution and age structure differ from those in protective forests. Therefore, it is highly possible that the results of calculations are shifted.


142

In order to be able to answer the question of degree of influence of the mentioned above peculiarities upon results of the assessment, programme code of the CARBON software was analyzed. The analysis proved necessity of software recalibration with the aim to harmonize it with peculiarities of the study object. After recalibration of the software, amounts of live biomass and sequestered carbon were assessed iteratively. Differences between the first (original software) and the second (recalibrated software) assessments are provided in Table 3.

Table 3. Differences between assessments, %

Group of forest-

forming species

Forested land

Standing stock

Live biomass components

Live biomass density

Carbon

leav

es

(nee

dle

s)

bra

nch

wo

od

an

d b

ark

stem

wo

od

an

d b

ark

roo

ts

un

der

sto

rey

veg

etat

ion

tota

l

tota

l

den

sity

TOTAL 0,0 0,0 -9,8 -5,8 3,3 5,1 -4,9 2,0 2,1 2,0 1,4 including:

Coniferous

-9,4 -7,8 6,4 10,9 -0,6 4,8 4,3 4,8 4,9

Hardwood broadleaves

-7,7 -1,9 2,1 6,2 -14,9 1,6 1,8 1,7 1,2


-12,7 -8,3 0,3 -2,8 -16,4 -1,6 -0,9 -1,5 -1,8

Information in table 3 proves that for some live biomass components differences between results produced by original and recalibrated software exceed 15%, which is significant. At the same time, difference in total amount of live biomass and sequestered carbon are considerably smaller.

The abovementioned facts and calculations confirm necessity of recalibration of the software when assessing bioproductivity of protective forests. In order to eliminate the described problems and allow for peculiarities of protective forests, inclusion of several scenarios of data processing into the CARBON programme is recommended.

References

[1] Federal agency of forestry of Russian Federation, 2009. Urban forests. Available at: http://www.rosleshoz.gov.ru/terminology/g/43 - date of review 20.04.2009 (in Russian).

[2] Lakyda, P.I. Productivity of forest stands of Ukraine by components of aboveground live biomass: abstract of Dr. Hab. Thesis (specialty 06.03.02 “Forest inventory and forest mensuration”). Kyiv, 1997 (in Ukrainian).

[3] Normative and reference materials for forest mensuration in Ukraine and Moldavia. Kyiv, 1987, (in Russian).


143

Reducing uncertainties in GHG inventory using modern agricultural

lands monitoring systems

Lyubov Lebed‘

’ Kazakh Research Institute for Ecology and Climate Seifullin pr.,597, Almaty, 050022, Republic of Kazakhstan

[email protected]

Abstract

The effective measures on GHG emissions reduction in Kazakhstan to fulfill the commitments relate also to agricultural sector, first of all livestock breeding and wheat production on precipitation-fed and irrigated land. Simultaneous implementation of measures aiming at increase of carbon sinks by natural ecosystems including forests and pastures may significantly affect the efforts of Kazakhstan in fulfilling the emission reduction commitments and facilitate economic growth stabilization. In this connection reducing uncertainty of the emissions and sinks estimates become of great importance. According to the IPCC Guidelines, 2006, the uncertainty can be reduced by obtaining additional information and estimating GHG emissions using higher Tiers. This report presents preliminary results of estimating primary bioproduction at pastures as the carbon stock pools. Estimates can be made on the basis of empirical ground measures and modern monitoring of pastures using space images and production processes modeling.

Keywords: GHG inventory, uncertainties, pasture ecosystem, modeling production

1.Introduction

Participation of Kazakhstan in the Kyoto Protocol after its ratification in early 2010, promotes emission trading in Kazakhstan and defines the economic policy and measures in emission reduction. According to the data of the Ministry of Environment Protection agricultural sector contributed from 14.4 to 30.8 Tg to the total national GHG emissions to the atmosphere during 3 last years, where as national total amounted to 233.9-245.8 Tg CO2-equivalent [1]. Direct GHG emissions from agriculture 1.5-2.5 times exceed national sinks estimated at 5.9 to 9.2 Tg. According to the expert judgment the GHG emission reduction commitments of Kazakhstan which mainly relate to energy and industry measures may limit annual economic growth to maximum 6% [2]. In this connection, measures aimed at GHG emissions reduction in agriculture and increasing carbon sequestration by forests at the area of 12.27 million hectares and natural pastures at the area on 187 million hectares may to the certain extent decrease risks of the economic growth. According to the recent studies uncertainty of the GHG emissions estimates in agricultural sector amount to 80-95% and thus reducing uncertainty in the emissions estimates is an important requirement of further studies. First of all it is necessary to impartially assess carbon potential of natural ecosystems in Kazakhstan with modern monitoring practices using space and ground data. Also the conditions of natural ecosystems are to be assessed in the reference years with further changes.

2. Possible emission reduction and increase of emission sequestration by managed agricultural lands

Main GHG flows (primarily carbon dioxide) between atmosphere and ecosystems are regulated by photosynthesis and plants respiration, nitrification and denitrification of


144

soil, genesis of methane in overwetted soils and manure storing reservoirs as well as enteric fermentation of livestock. The intensity of these processes is defined by the natural potential of the area and anthropogenic activity (capability to manage ecosystem). According to FAO improvement of the world agriculture management during 25 years may result in a decrease of carbon dioxide concentration in the atmosphere by 10% through carbon sequestration by soil.

Present day strategic solutions for decreasing carbon consumption in economic sectors foresee steady reduction of GHG emissions by 15% in 2020 to 70 % by 2050 [2]. The measures in agricultural sector foresee the increase of carbon sink by major pools along with decrease of direct emissions of nitrogen oxide and methane. Anthropogenic influence on GHG flows in ecosystems is possible through establishing optimal land use and sustainable land management. State Programme on the boosted industrial and innovative development of Kazakhstan for 2010-2014, accepted in May 2010 at the governmental level as well as the Strategic Development Plan by the Ministry of Agriculture include a number of potential possibilities for agricultural sector both to promote economic indices and sustainability of production and to reduce direct GHG emissions and increase carbon sequestration. Among effective measures it should be noted the planned annual expansion of wheat planting areas by 300 thousand hectares which are to be tilled by water-, soil-, and resource-saving practices including minimal or zero soil treatment with chemical steam. The area of land managed that way is planned to reach 73% of the total land under cereals by 2014. Preliminary results show that this measure only may result in reducing the emissions of nitrogen containing compounds by 10% approximately. Nitrogen emissions reduction from agricultural soils will also be stimulated by regulation of the nitrogen fertilizers application and extension of the areas under annual leguminous plants and perennial herbs to partially substitute chemical fertilizer with green manure. Methane emissions will be reduced through the planned increase of a number of large farms for animal production on the basis of livestock-breeding complex, milk farms, poultry factories and fattening grounds with infrastructure for emissions treatment.

Simultaneously, measures to increase carbon sink are planned in forestry through annual reforestration on the area of 40-50 thousand hectares under state programme “Zhasyl El” for 2008-2010 and increasing the percentage of forest at the area of 3714 thousand hectares, including saxaul forests in deserted zone of Kazakhstan. According to preliminary results, successful implementation of the planned measures the GHG sinks may significantly exceed emissions from agricultural lands by 2020. The positive balance is estimated at 5 Tg of CO2-equivalent, which is around 10% of the total expected GHG emissions reduction in Kazakhstan by 2020.

3. Assessment of biological productivity and carbon content in pasture lands

The cumulative amount of carbon in the ground ecosystems is more than 3.5 times higher that in the atmosphere. Thus minor changes of carbon content in the ground ecosystems result in the significant effect on the atmosphere [3]. Also is was revealed that in the total carbon balance on earth the organic carbon content in soil is 3.5 times higher its content in the vegetation biomass.

Present day pastures in Kazakhstan are presented by the groups of small shrub, semi-shrub and bushy vegetation with some grasses. They are formed in the arid conditions with annual precipitation less than 250 mm on the zonal light chestnut, brown and gray-brown soils with low humus content from <1% to 3-3.5%. All these factors define


145

peculiarities in carbon accumulation and allocation in the area and among above and underground pools. Energy resources of the natural ecosystems functioning are defined by their biological productivity, first of all biomass stocks and annual product. Despite the long-term researches of pastures which were conducted regularly in the second half of the last century on geobotanical and agrometeorological stationary stations [4,5 et], on-route geobotanical studies[6] and aerospectrometric shootings of vast areas [7], the data on the cumulative biological productivity of pastures are rather limited. Numerous geobotanical maps by the Kazakhstan Agency of land management reflect fodder stocks, i.e. annual increment of green biomass, which was estimated at 59.9-61.8 million tons of dry matter per year for late 80s-early 90s in Kazakhstan [6]. For the scale estimate of ecosystems’ biological productivity of great importance are maps of total stock of phytomass, deadmass and annual product on the restored vegetation cover by N. Bazilevich et al. for the territory of Northern Eurasia [8]. Basing on these maps T. Gilmanov estimated potential stocka of the net aboveground productivity for pasture ecosystems in Central Asia, including Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan and Uzbekistan [9]. His estimates show that cumulative aboveground product for Central Asia amounts to 465.7 million tons per year for the pasture area of 262 million hectares. Therefore, because pastures’ area in Kazakhstan is about 187 million hectares, the 333 million tons of aboveground product should be attributed to Kazakhstan, which relates to the amount of fodder stock as 5:1. It should be noted that estimates do not account for significant changes in the natural vegetation if compared to the initial, due to intensive long-term livestock grazing and other anthropogenic activities. These impacts resulted in changes in species structure of vegetation cover, reduction of soil fertility due to erosion and finally decrease of biological productivity of 60% pastures. Aral region (Aral and Kazalinsk regions of Kzylorda oblast) can be considered as a striking example of anthropogenic negative impact on the natural vegetation, where initial native vegetation remained on the <50% of area, and the productivity of haylands and pastures decreased 8-10 and 2-3 times, correspondingly [6]. According to maps by N. Bazilevich the stocks of pasture ecosystem phytomass for the restored vegetation cover is estimated at 7-9 tons per hectare for steppes and sub-boreal semi-desert. For the real sub-boreal desert with 80+% of underground biomass with prevalence of living roots the stocks of pasture phytomass is estimated to be <5 tons per hectare [8].

Thus, preliminary studies show a number of discrepant estimates of biological productivity, related energy resources and carbon potential of natural ecosystems in Kazakhstan, and therefore justify urgent need in the objective estimates taking into account natural potential of the area, biodiversity, anthropogenic disturbance and observed climate change.

4. Estimation of bio- and energy potential of the pasture ecosystem basing on the “Pasture” model

During last years more and more importance obtain issues of estimating actual potential of carbon sequestration by vast pasturable lands of Kazakhstan and other Central Asian countries through their regular monitoring. Modern monitoring of pasture lands is based on the complex of methods including empirical ground and landscape and ecological studies using space systems as well as modeling of production processes at pastures and GIS technology.

For simulation of pasture production authors used agrometeorological “Pasture’’ model, which was refined through long-term pasture study in Kazakhstan [10].


146

The model describes agrometeorological conditions of plants growth, biomass accumulation dynamics as photosynthesis production during vegetative period, its preservation and destruction. Generally model can be presented by the following equations:

Bt = Bm · Rt (1) Bm = F (Ph, A,) (2) Ph = F (J, Co) (3) Rt = Rm · F (T) · F (W) (4)

tbt

−=Β′ exp (5)

bt = F (T, W), (6)

where Bt is the biomass (annual increment), t\hectare, Rt is the plant growth function, dimensionless, Bm is the maximum biomass for vegetative period under optimum conditions, t\hectares, Ph is the photosynthesis product for vegetative period, kg СО2\m2 s, A is the factor of leaves amount at plants, dimensionless, J is the photosynthesis active radiation (PAR), J\m2 s, Co is the СО2 concentration in atmosphere, Rm is the growth function under optimum environment conditions, dimensionless, Т is the air temperature, 0 C;, W is the moisture sufficiency for plants, dimensionless, tB′ is the Biomass in destruction, t\hectare, bt is the biomass destruction

factor, dimensionless; Figure 1 presents calculated dynamic of accumulation biomass distribution in time

for several dominants during the vegetative period generated in Balkhash region in sandy desert and foothill clay desert.

Fig. 1 Modeling of season aboveground biomass for plants dominants in dynamic for 2007 year on Balkhash area pasture lands. Map 1:200 000 scale. 1- Ephemery group, 2- Artemisia terrae

albae, 3- Agropyron fragile, 4- Calligonum aphylum, 5-Kochia prostrate, 6- Ceratocarpus utriculosus.

The identification of model’s parameters was carried out on the basis of long-term materials of the agrometeorological and geobotanical field stationary observations received in 1960s-1980s for the conditions of Kazakhstan deserts and seasonal values of the projective green vegetation cover, %, received from Modis digital images.


147

Pasture model estimated pasture bioecological and economical characteristics, including aboveground biomass (annual growth), potential food volumes, load of livestock and pasture ecological capacity.

The «Pasture» model can be used to estimate Primary biological production (PBP) on pastures as the energy characteristics of natural ecosystem expressed in t/hectare, cal/hectares or other units. The values of annual increment of aboveground biomass can be used as the input data. Experts used the method of peaks or sum of maximum annual biomass increment on each vegetative dominant. The Total production amount of РBP comes from aboveground biomass, including Рg – annual living biomass, Рm – long-term dead biomass and Рf – long-term living biomass (woody), as well as belowground biomass: Рr – living roots and Рmr – dead roots.

Figure 2 and Figure 3 present the examples of map fragments of Primary biological production and the Total biological production in Balkhach area (Scales 1: 1 000 000 and 1: 200 000), which were calculated on the basis of pasture Modis images data and the field information received during field study with contact measurements. Such maps, along with soil carbon maps are used as initial data for calculations of accumulated carbon stocks on the pastures and their dynamics during regular monitoring with remote sensing data.

Fig. 2 Simulation of Primary aboveground production (t\ ha) in condition of temporache grazing. Balkhash area pastures . Map 1: 1 000 000 scale.

References

[1] National report on the environmental situation in the Republic of Kazakhstan in 2008 (2009): Almaty, Kazakh Research Institute of Ecology and Climate, pp.223 (ru).

[2] The strategy of development planning sectors of the economy of the Republic of Kazakhstan (2010): UNDP, Astana, pp.21 (ru).


148

Fig. 3 Simulation of Total Biological Production (t\ha) for Balkhash area pasture lands.

Map 1: 200 000 scale (phragment).

[3] Borisenkov E., Kondrat'ev K. (1988): The carbon cycle and climate, L: Gidrometeoizdat, pp. 319 (ru).

[4] Pastures and hay in Kazakhstan (1970): Stationary study for the Southern Balkhash region and lower reaches of the river Chu - Alma-Ata, pp.233 (ru).

[5] Weather and harvest hay and pasture Muyunkumsko-Betpakdala natural complex (1978): A Collection of Issue on Hydrometeorological security of grazing breeding, Alma-Ata, Vol. 69, pp. 161(ru).

[6] Geobotanical works in the lands use manager system of PKazakhstan (2005): Agency for Land Management, Astana, pp.138 (ru).

[7] Lebed L., Korobova E., Turbacheva T. (1989): Calculation of seasonal fodder reserves of natural pasture in Kazakhstan according to aerospektrometrical surveys – Problems of Hydrometeorological security of grazing breeding. Vol. 103, Moscow, pp.107-113 (ru).

[8] Bazylevych N.I. (1993): Biological productivity of ecosystems of Northern Eurasia, Moscow, pp. 320 (ru) .

[9] Gilmanov T. (1996): Ecology of pastures in Central Asia and the modeling of primary productivity-Central Asia, Seminar Assessment of livestock in the region. Tashkent, Published by the University of California, pp. 150-184.

[10] Lebed L. (2008): Possible changes in agriculture under the influence of climate in Kazakhstan- Environmental Probable of Central sia and their economic, Social, and Security Impacts, The NATO Science for Peaceand Security Programme, pp. 149- 162.


149

Abatement, R&D policies, and negative emission

technology in climate mitigation strategies

Derek Lemoine1, Sabine Fuss2, Jana Szolgayova2,3, Michael Obersteiner2

1 Energy and Resources Group, 310 Barrows Hall, University of California, Berkeley, CA 94720-3050

2 International Institute for Systems Analysis, Schlossplatz 1, A-2361 Laxenburg, Austria 3 Department of Applied Mathematics and Statistics, Faculty of Mathematics, Physics and

Informatics, Comenius University, Bratislava, Slovakia [email protected], [email protected], [email protected], [email protected]

Abstract

In this study, a global decision-maker’s cost-minimizing portfolio of technology and emission policies for achieving a year 2100 carbon dioxide (CO2) concentration target is determined. Technological change depends stochastically on abatement and on public funding of research and development (R&D). A simplified analytic model shows the complex dependence of the optimal portfolio on the parameters governing technical change. The full numerical model shows that technology policies complement abatement, whereas negative emission technologies can delay abatement. The type of technology targeted by public R&D depends on the height of the CO2 target, and the level of public R&D funding depends on the effectiveness of abatement in inducing technological change. Under nearly all scenarios and CO2 constraints, the optimal policy portfolio abates 50-100% of emissions by 2050. Announced 2ºC temperature targets require greater-than-announced abatement by 2030 unless policymakers plan for large-scale future use of negative emission technologies or unless policymakers are willing to accept a substantial chance that the temperature targets will be exceeded.

Keywords: uncertainty, technological change, GHG emission targets, policy portfolio, carbon capture and storage (CCS)

1. Introduction

There are several ways we consider for the world to reach CO2 targets: (1) sustained abatement, (2) initial research and development (R&D) followed by higher abatement, and (3) use of negative emission strategies reducing the total abatement needed. Each path requires significant long-term decarbonization and requires different types of near-term policies and technological progress. Early abatement reduces the cost of later abatement, while R&D delays abatement expecting R&D programs to produce technologies making future abatement cheaper. Negative emission technologies enable greater gross CO2 emissions by removing previously emitted CO2 from the atmosphere.

The results show that the optimal portfolio of abatement, R&D, and negative emission technologies almost always includes decarbonization close to 100% by mid-2000, where the use of near-term abatement and public R&D funding depends on the level of the GHG target, on the feasibility of negative emission technologies, and on beliefs about the sensitivity of technological change to public R&D and abatement. However, abatement decisions are not responsive to the availability of R&D options. The derivation of such (optimal) portfolios could serve as benchmark for global climate policy, but also for national and sub-national policies.


150

In the following section we present the current literature on technology and emission policies, which motivates our choices for the model presented in this paper. We model the effects of technological change on the abatement cost curve in two ways. A numerical model is introduced in section 3, while section 4 presents and discusses the results.

2. Literature Review

Various authors ([1];[2]) have pointed to the fact that a portfolio of technology and emission policies is probably part of a welfare-maximizing climate policy. Focusing on emissions, emission policies provide different incentives for innovation and adoption of low-carbon technologies. In general, firms’ expectations of future climate policies would determine their budgets for low-carbon innovation. Many economic models include a version of this demand-pull process of induced technological change (ITC), but there is no established way of representing it ([3]; [4]; [5]). Focusing on technology, climate policy may concentrate funding on an intensive search for cost-effective low-carbon, where they can be oriented towards long-term technologies needed for anticipated deep abatement, or towards shorter-term technologies that reduce CO2 emissions through incremental advances. [6] find that if future full abatement is more likely and R&D programs more risky, R&D into longer-term carbon-free technologies is preferred and vice versa.

Negative emission technologies might be a third policy option allowing greater gross emissions to achieve the same level of net emissions. In this study, we represent large-scale negative emission technologies by air capture technologies ([7]). Two leading examples are facilities that remove CO2 from the air via chemical reactions and biomass-fired electricity generators that use CCS. [8] used an integrated assessment model to explore the implications of air capture for climate strategy finding that the future availability of air capture reduces near-term abatement efforts and reduces atmospheric CO2 concentrations to pre-industrial levels faster than natural removal mechanisms. In this paper we test cases with and without feasible air capture in order to assess its impact on climate policy decisions. In contrast to [8], we explore how concerns about threshold effects from temporarily high CO2 levels might affect planned air capture use.

The existing literature provides some studies analyzing portfolios of technology and emission policies. [9], for example, consider six types of climate policies applied to the U.S. electricity sector. They find that an emission price is the best single policy, but R&D support is an important additional ingredient of the optimal portfolio. However, they focus on the near term only, neglecting the need for longer-term technological breakthroughs, which are part of the focus of our study.

More closely related to our paper, [10] looks at the implications of uncertainty on portfolio selection and finds that abatement and R&D hedge against different kinds of risks emanating from uncertainty about climate damages.

3. Policy Portfolio Selection

Compared to the literature reviewed above, our model allows for complex interactions between abatement and R&D options and offers an extension by explicitly modeling the simultaneous choice between emission reductions and R&D policies and taking into account the interaction between ITC and public R&D funding, where technological change is stochastic. Finally, the option of air capture and of air-capture R&D is considered.


151

We extend the R&D framework of [11] and add ITC. Abatement and public funding of R&D affect the distribution of future technology outcomes. These outcomes in turn affect abatement cost according to whether the new technologies are “carbon-free” technologies or “emission intensity” technologies. We treat technological change as creating new options for cheaper abatement, with technologies diffusing in the same period as the new abatement options are exercised.

For a given fraction µ of emissions abated, carbon-free technologies (e.g. renewable electricity sources) decrease the cost and marginal cost of abatement by a fraction α, producing relatively small savings at low levels of abatement and greater savings at higher levels of abatement.

R&D can also be targeted to emission intensity technologies, e.g. cheaper CCS for coal-fired plants reducing non-abated emissions by a fraction γ. New emission intensity technologies reduce marginal abatement cost for low levels of abatement but raise it for high levels of abatement: because these technologies make it cheaper to obtain low levels of abatement, they also make it incrementally more expensive to achieve the highest levels of abatement.

We modify the probability distributions for technology outcomes from Baker and Adu-Bonnah (2008) to account for ITC and for previous R&D outcomes. α and γ are the actual technology outcomes, and α

H and γH are the best possible technology outcomes (1 > αH ≥ αbar ≥ 0 and 1 > γH ≥ γbar ≥ 0). Let αhat and γhat be the total targeted technology outcomes as determined by ITC and by α

bar and γbar, the targeted outcomes as selected by levels of public R&D funding. The total targeted outcomes are achieved with probability 1-pα and 1-pγ for exogenously specified probabilities. In those cases where they are not achieved, technology attains α

H or γH (maximal success) with probability αhat

and γhat and does not change from the previous level with probability (1 - αhat) and (1 -γhat). We use an analogous representation for air capture R&D, where the R&D outcome φ is the new cost as a fraction of the original cost. Lower φ is therefore better

R&D outcome, giving 1 ≥ φ ≥ φ H > 0. Making the technology targets more ambitious raises the chance of achieving the greatest possible breakthrough while reducing the chance of total failure.

Functions ITCα : µ → [0,αH] and ITCγ : µ → [0,γH] convert abatement into technology targets, and these ITC technology targets then add to the targeted innovation levels selected through direct R&D expenditures, with the total targeted outcomes capped by the maximal targets. The targets determining the probability distributions for α and γ therefore become αhat = min[αbar + ITCα(µ), αH] and γhat = min[γbar + ITCγ(µ), γH]. We assume that (a) abatement and air capture use generate no ITC for air capture technology, and (b) ITC depends only on the current period’s abatement decision and not on expectations of abatement in future periods. In the base case parameterization, the two ITC functions make low levels of abatement contribute only to emission intensity R&D, while high levels of abatement can also contribute to carbon-free R&D.

The model selects the cost-minimizing level of five different climate policies in each of three periods. This optimal portfolio is conditional on the realizations of technology outcomes and is subject to the expected final CO2 concentration being no greater than an exogenous level. We solve the model using stochastic dynamic programming to obtain the optimal action in each period conditional on each possible state of the world. The periods correspond to 2010-2029, 2030-2049, and 2050-2099, roughly matching the near-term, intermediate-term, and long-term periods for which CO2 emission goals


152

are often discussed. These large time steps also roughly match some infrastructure lifespans, leading us to assume that abatement investment does not carry over from period to period except via its effect on technology.

The objective is to select a sequence of abatement policies {µt} t=13, air capture levels

{κt} t=13, carbon-free public R&D targets {αbar t} t=1

3, emission intensity public R&D targets { γbar t} t=1

3 and air capture public R&D targets {α bar t} t=13 so as to minimize costs subject

to the expected year 2100 GHG concentration GHG3 to be no greater than a specified threshold GHG*:

( ) ( ) ( ) ( )[ ] *GHGGHGE.t.s

jhg,f,,cemin bartH

t

bart

Ht

bart

tttttttt

t

},,,,{ tbarbarbar

≤

φ+

γγ+

αα+φκ+γαµµβ∑

=

−

φγακµ =

3

3

1

1203

1 (1)

The decision variables can have one of the levels given in Table 1. µt is the fraction of BAU emissions et that are abated, and κt gives the quantity of air capture. The R&D targets are evenly spaced in the gap between the maximal outcome and the period 1 starting values for R&D outcomes. ( )⋅c is the average cost of abatement, depending on µt and on the outcomes of previous R&D into carbon-free technologies and emission intensity technologies. f(.) is the cost of air capture and depends on κt and the outcome of past air capture R&D. g(.), h(.), and j(.) represent the R&D funding required by the chosen public R&D targets.

Table 1. Decision Variable Values

Decision Variable Possible Values Abatement µt {0,0.25,0.50,0.75,1} Air Capture κt {0,0.10e3,0.25e3,0,5e3,e3}

Public carbon-free R&D αbar t {0, αH/4, αH/2,3αH/4, αH}

Public Emission Intensity R&D γbar t {0, γH/4, γH/2,3γH/4, γH}

Public air capture R&D φ bar t {1,3/4(1- φ H)+ φ H,1/2(1- φ H)+

φ H+1/4(1- φ H)+ φ H, φ H}

The state variable GHGt records the CO2 concentration at the end of period t:

ttttt eafGHGGHG κµ −−+= − )1(**1 , (2)

where af is the airborne fraction, which gives the percentage of CO2 emissions that remain in the atmosphere after a short adjustment period, and it is set as af = 0.45 ([12]). The state variables αt, t, γt and φ t record the technology outcomes from period t and have the following probability distributions, which depend on the previous period’s R&D outcomes, on the previous period’s abatement, and on the previous period’s direct funding of the corresponding type of R&D:

])),(min[(1(]Pr[ 111H

tbarttt ITCp αµααα αα −−− +−== ; (3)

αα αµαα pITC Ht

bartt −=+= −− 1)]),(min(Pr[ 11 ; (4)

])),((min[]Pr[ 11H

tbart

Ht ITCp αµααα αα −− +== ; (5)

])),(min[1(]Pr[ 111H

tbarttt ITCp γµγγγ γγ −−− +−== ; (6)

γγ γµγγ pITC Ht

bartt −=+= −− 1)]),(min(Pr[ 11 ; (7)


153

])),((min[]Pr[ 11H

tbart

Ht ITCp γµγγγ γγ −− +== ; (8)

barttt p 11]Pr[ −− == φφφ φ ; (9)

φφφ pbartt −== − 1]Pr[ 1 ; (10)

)1(]Pr[ 1bart

Ht p −−== φφφ φ . (11)

Framing the CO2 constraint in terms of expected year 2100 CO2 concentrations ignores concerns about the possibility of threshold effects from temporarily overshooting the targeted concentration (e.g., [13]). We represent these concerns in a set of model runs by making the CO2 constraint bind each period’s concentration rather than just the third period's expected final concentration. This change does not affect model runs in which air capture is infeasible because the inability to produce net negative emissions makes the year 2100 constraint apply to each period’s final concentration. However, it could affect model runs with feasible air capture if the technology is used in later periods to make up for temporary overshoots. The exact functional forms and parameterizations used can be found in the appendix along with a list of the experiments in Table 2.

The goal is to find actions that are robust to beliefs about parameters and to determine which parameters are crucial for optimal plans. We model portfolio selection for each parameter scenario under each of three different CO2 constraints. The most stringent CO2 constraint of 390 ppm is just above the starting CO2 concentration and would be exceeded in the first period. The middling constraint of 435 ppm would be exceeded in period 2, and the least stringent constraint of 550 ppm would be exceeded in period 3. In line with insights from [14] and [15], the CO2 constraint is given exogenously based on risk preferences rather than determined endogenously using some distribution on marginal damages. If prior beliefs allow climate models to be incomplete and to share biases, then the 550 ppm target implies a 90% chance of keeping temperature change below 4ºC, the 435 ppm target corresponds to requiring a 95% chance of keeping temperature change below 4ºC, and the 390 ppm target corresponds to requiring a 90% chance of keeping temperature change below 2ºC ([16]).

4. Discussion of Results

Fig. 1 displays the planned emission paths by CO2 constraint, type of options available, by concerns about tipping points, and by parameter scenarios. In the absence of public R&D options, abatement provides the only means of technological change, and in the absence of feasible air capture technologies, abatement provides the only means of meeting the CO2 constraint. The presence of public R&D options does not tend to affect planned abatement under any of the CO2 constraints. The presence of options for negative emission air capture technologies does not affect planned actions under the 550 ppm constraint. With the 435 ppm constraint, making negative emission technologies available increases near-term abatement while decreasing long-term abatement, and for 390 ppm, it decreases near-term abatement by enabling future air capture to offset higher early emissions. The presence of a strict CO2 threshold does not affect emissions under 435 ppm by much, but increases early abatement under 390 ppm. The strict constraint also shifts some air capture use into the first period for 390 ppm, so as to avoid having to abate 100% of BAU emissions in the first period. Some policy choices are not sensitive to climate targets or to parameters’ values: the optimal portfolio almost always abates at


154

least 50% of period 2 BAU emissions and at least 75% of period 3 BAU emissions. Public funding for R&D is almost never above 50% of the maximum and, unless the CO2 constraint is a strict threshold or there is no discounting, air capture is rarely used before period 3 or without previous air capture R&D. A robust course of action therefore plans for deep abatement from 2030-2100, includes public R&D support that is significant but not a substitute for early abatement, and uses air capture only after deep abatement and in conjunction with ongoing deep abatement. Carbon-free public R&D and emission intensity public R&D often substitute for each other, with expectations of future abatement levels driving the choice between the two types of technology forcing (Fig. 2).

Figure 1. Planned emission paths under the three year 2100 CO2 constraints (rows)

for different types of options and CO2 thresholds (columns).

Knowing a few specific parameters provides many of the remaining details about the optimal course of action, regardless of other parameters’ values, one of the most important parameters being the presence of options to undertake air capture use. The air capture option allows the level of period 3 abatement under the two more stringent constraints to be contingent on abatement R&D outcomes and on air capture R&D outcomes. Air capture and emission intensity R&D thus act as complements, both substituting for carbon-free R&D and for abatement. In cases without air capture, one can almost perfectly predict each period’s abatement if one knows this constraint and nothing else. The possibility of air capture tends to reduce the importance of the CO2 constraint for the determination of abatement levels and abatement R&D decisions because air capture can make the more stringent constraints’ abatement goals look more like those needed for less stringent constraints. In a world without air capture, beliefs about climate change and tolerance for climate change risks almost completely determine immediate abatement and R&D decisions, and in a world with air capture, these beliefs and risk tolerance determine whether air capture is a relevant technology.


155

Figure 2. Probability of undertaking a type of action in each parameterization types of options and CO2 thresholds (columns).

Appendix: Functional Forms and Parameterizations

The average cost in the base case of abating fraction µt of BAU emissions et given R&D outcomes αt and γt is:

( ) ( ) ( ) ( )

µα−

µ=γαµ 00100 ,,c,,,zc

zmin,,c tt

t

tttt

where [ ]0),1/()(max ttttz γγµ −−≡ .The average abatement cost in the low-cost

parameterization is denoted as ( )ttt ,,d γαµ . Zero abatement is free and the normalization

is ( ) 100001 =,,c . Two marginal cost representations are developed by assuming that the carbon prices

reported by [17] represent the marginal cost of abatement, that abatement of 25% has a marginal cost of $20/tCO2, that abatement of 50% makes marginal costs either quintuple (base case) or triple (low-cost case) to $100/tCO2 or $60/tCO2, that higher levels of abatement follow the same geometric progression, and that the marginal cost of abating a given fraction of contemporary emissions is unaffected by previous abatement. The further assumption that marginal costs increase linearly between the discretized points allows us to identify the average cost at each possible level of abatement:


156

Base case: ( ) ( ) ( ) ( ) 10000128007504800504200250 ==== ,,c,,,.c,.,,.c,.,,.c

Low-cost: ( ) ( ) ( ) ( ) 2700112007500600504200250 ==== ,,d,,,.d,.,,.d,.,,.d

All other cost functions in this model are expressed in terms of ( )⋅c . For values of zt that fall between µ’s discretization, we define the abatement cost functions by treating average cost as piecewise linear between the discretized points.

For any given upper limits for R&D targets and outcomes, we assume the funding it takes to aim for the chosen public target to be independent of the level of the target but it does depend on the percentage of the maximum target that it represents. Due to the lack of previous empirical work, we also treat the cost of reaching a percentage of the maximal level of R&D as being a fraction y of the base case cost for abating the same percentage of period 1 emissions:

100 e*,,c*ygHHgH

+αα

αα=

αα

=

HhH

gyhγγ

γγ

*

( ) 11

100

1

1e**,,c*yj

HHjφ−φ−

φ−φ−=φ

The cost of emission intensity R&D is assumed to be a fraction yh of the cost of carbon-free R&D represented in terms of average abatement costs.

ITC is defined as leading a percentage abatement to produce R&D targets that are the same percentage of their maximal levels.

))(,0max()( HITC αυµµ αα −=

))(,0max()( HITC γυµµ γγ −=

By including the parameter ν we can vary the effectiveness of ITC across scenarios and make ITC more effective within a given scenario for emission intensity technologies than for carbon-free technologies.

Air capture has constant marginal costs: ( ) ( )00,,xc,f κφ=φκ . x = 0.75 corresponds to air capture cost of $115/tCO2, which is near the low end of current estimates, and x = 1 corresponds to air capture cost of $415/tCO2, which is above many current estimates (e.g., [18]; [19]).

Initially, the CO2 concentration is GHG0 = 385 ppm. BAU emissions come from scenario A2r in the International Institute for Applied System Analysis (IIASA) GGI Scenario Database (see also Riahi et al., 2007). Summing over each period’s years yields et in Gt CO2: e1 = 750, e2 = 1150 and e3 = 4500. The BAU path produces CO2 concentrations of 428 ppm in 2030, 493 ppm in 2050, and 749 ppm in 2100.

References

[1] Jaffe, A. B., R. G. Newell, and R. N. Stavins (2005): A tale of two market failures: Technology and environmental policy. Ecological Economics Vol. 54, No. 2-3, pp. 164-174.

[2] Bennear, L. and R. Stavins (2007): Second-best theory and the use of multiple policy instruments. Environmental and Resource Economics Vol. 37, No. 1, pp. 111-129.


157

Table 2. Listing of Experiments

Scenario Parameter Values BAU Values Base Case - -

1) Cheap abatement ( )⋅d ( )⋅c

2) Cheap R&D yg=yj=0.25 yg=yj=0.5 3) Cheap em. Intensity

R&D yh=0.5 yh=1

4) Cheap abatement, R&D and air capture

( )⋅d ,x=0.75, yg= yj=0.25

( )⋅c , ( )⋅d ,x=1, yg= yj=0.5

5) Limited R&D scope α

H=γH=0.25, φ H =0.75

αH=γH=0.75, φ H =0.25

6) Greater R&D scope α

H=γH=0.95, φ H =0.05

αH=γH=0.75, φ H =0.25

7) Limited R&D control pα=pγ= φp =0.75 pα=pγ= φp =0.25

8) High discounting β=0.9 β=0.95 9) No discounting β=1 β=0.95 10) Perfect ITC να= νγ=0 να= 0.5,νγ=0.25 11) Better ITC for both

technologies να= 0.25,νγ=0 να= 0.5,νγ=0.25

12) Better ITC for intensity technology

νγ=0 νγ=0.25

13) No ITC να= νγ=100 να= 0.5,νγ=0.25 14) Cheap air capture x=0.75 x=1

[3] Sue Wing, I. (2006): Representing induced technological change in models for

climate policy analysis. Energy Economics Vol. 28, No. 5-6, pp. 539-562.

[4] Gillingham, K., R. G. Newell, and W. A. Pizer (2008): Modeling endogenous technological change for climate policy analysis. Energy Economics Vol. 30, No. 6, pp. 2734-2753.

[5] Popp, D., R. G. Newell, and A. B. Jaffe (2009): Energy, the environment, and technological change. National Bureau of Economic Research Working Paper Series No. 14832.

[6] Baker, E. and K. Adu-Bonnah (2008): Investment in risky R&D programs in the face of climate uncertainty. Energy Economics Vol. 30, No. 2, pp. 465-486.

[7] Keith, D. W. (2009): Why capture CO2 from the atmosphere? Science Vol. 325, No. 5948, pp. 1654-1655.

[8] Keith, D. W., M. Ha-Duong, and J. K. Stolaroff (2006): Climate strategy with CO2 capture from the air. Climatic Change Vol. 74, No. 1-3, pp. 17-45.

[9] Fischer, C. and R. G. Newell (2008): Environmental and technology policies for climate mitigation. Journal of Environmental Economics and Management Vol. 55, No. 2, pp. 142-162.


158

[10] Baker, E. (2009): Optimal policy under uncertainty and learning about climate change: A stochastic dominance approach. Journal of Public Economic Theory Vol. 11, No. 5, pp. 721-747.

[11] Baker, E., L. Clarke, and E. Shittu (2008): Technical change and the marginal cost of abatement. Energy Economics Vol. 30 No. 6, pp. 2799-2816.

[12] Denman, K., G. Brasseur, A. Chidthaisong, P. Ciais, P. M. Cox, R. E. Dickinson, D. Hauglustaine, C. Heinze, E. Holland, D. Jacob, U. Lohmann, S. Ramachandran, P. da Silva Dias, S. Wofsy, and X. Zhang (2007): Couplings between changes in the climate system and biogeochemistry. In S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. Averyt, M. Tignor, and H. Miller, eds., Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 500-587. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press.

[13] Lenton, T. M., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf, and H. J. Schellnhuber (2008): Tipping elements in the Earth's climate system. Proceedings of the National Academy of Sciences Vol. 105, No. 6, pp. 1786-1793.

[14] Weitzman, M. L. (2009): On modeling and interpreting the economics of catastrophic climate change. Review of Economics and Statistics Vol. 91, No. 1, pp. 1-19.

[15] Ackerman, F., S. DeCanio, R. Howarth, and K. Sheeran (2009): Limitations of integrated assessment models of climate change. Climatic Change Vol. 95, No. 3-4, pp. 297-315.

[16] Lemoine, D. M. (2010): Climate sensitivity distributions depend on the possibility that models share biases. Journal of Climate (in press).

[17] Hoogwijk, M., D. Vuuren, S. Boeters, K. Blok, E. Blomen, T. Barker, J. Chateau, A. Grübler, T. Masui, G. Nabuurs, A. Novikova, K. Riahi, S. R. du Can, J. Sathaye, S. Scrieciu, D. Urge- Vorsatz, and J. Vliet (2008): Sectoral emission mitigation potentials: Comparing bottom-up and top-down approaches, Ecofys.

[18] Rhodes, J. S. and D. W. Keith (2005): Engineering economic analysis of biomass IGCC with carbon capture and storage. Biomass and Bioenergy Vol. 29, No. 6, pp. 440-450.

[19] Riahi, K., A. Grübler, and N. Nakicenovic (2007): Scenarios of long-term socio-economic and environmental development under climate stabilization. Technological Forecasting and Social Change Vol. 74, No. 7, pp. 887-935.


159

Uncertainties of results of GHG inventories: Europe 2020

Myroslava Lesiv, Andriy Bun, Mykola Medykovsky

Lviv Polytechnic National University str. Bandera 12, Lviv, 79013, Ukraine

[email protected]

Abstract

The main two factors of changes in uncertainty in relative terms are “knowledge increasing” and structural changes in GHG emissions. In this article the influence of structural changes in GHG emissions on total uncertainty is analyzed considering EU’ “20-20-20” targets. The projections of uncertainty levels till 2020 under scenarios of future energy pathways have been made.

Keywords: uncertainty, GHG emissions, Energy sector

1. Introduction

Uncertainty in GHG emissions is the value, which refers to a lack of confidence in the components of a cadastre as a result of any casual factors, such as uncertainty of the sources of emissions, absence of transparency of the inventory process etc. It is very important to increase knowledge on uncertainty and reasons of its change to reduce uncertainties in GHG inventories and to set future emissions targets. Unfortunately, there is not enough experience in assessing and compiling uncertainty and changes in uncertainty in relative terms.

The first-ever estimates on change of uncertainty in the past were provided in IR [5]. The author calculated the combined related uncertainties for EU-15, which consider accuracy and precision. The achieved results for the period 1990-2005 show the uncertainty decrease approximately 4.24% each year. This change is expected to be more significant for LULUCF sector and for other than carbon dioxide gases which are more uncertain. The decrease in the past is caused by the knowledge increasing and structural changes in fossil fuel consumption. The 95% relative uncertainty change was caused only by knowledge increase and only 5% by structural changes in fossil fuels consumption.

The aim of this article is to analyse the influence of structural changes in emissions on total uncertainty, making projection of future uncertainty levels under scenarios of structural changes in GHG emissions considering new measures – EU’ “20-20-20” targets.

2. EU’ "20-20-20" climate/energy targets

In order to limit the global average temperature increase to not more than 2 degrees C above preindustrial levels [6], developed countries as a group should reduce their emissions to 30% below 1990 levels by 2020. The EU has set the example by committing to a 20% reduction in its emissions compared to 1990 levels by 2020, irrespective of whether or not an international agreement is concluded.

The European commission adopted the Climate Action and Renewable Energy package on January 23, 2008. It was a part of implementing the Integrated Energy and Climate Change package of 2007. The Member States agreed to:


160

- cut the GHG emissions by at least 20% of 1990 level (30% if other developed countries commit to comparable cuts);

- cut the energy consumption by 20% of projected 2020 levels – by improving energy efficiency;

- increase the use of renewables (wind, solar, biomass, etc) to 20% of total final energy consumptions.

The EU 27 is making good progress towards its 2020 emission reduction target of – 20 % [12].

3. Mathematical background

Projections of future uncertainty levels under scenarios of structural changes in GHG emissions have been made according to the next scheme (Figure 1):

Figure 1. Scheme for making projections of uncertainties of GHG emissions in Energy sector

In the study emission data and uncertainties are taken from National Inventory Reports [7] by countries, and Annual European Inventory Report to the UNFCC [8]:

- GHG emissions in CO2 – equivalent by sectors; - CO2 emissions of fossil fuels consumption; - uncertainties by gases and sectors. For calculating GHG emissions in Energy sector the uncertain quantities of emission

are used such as GHG emissions in CO2-equivalent from different sources (with the exception of CO2 emissions from fossil fuel consumption), and CO2 emissions of fossil fuel consumption, which according to the IPCC Guidelines [2] are calculated as:

COFCCNCVADE ⋅⋅⋅= , (1) where E is the carbon dioxide emission, AD is the activity data in physical units, NCV is the net calorific value (energy per physical units), CC is the carbon content (mass of carbon per unit of energy on a net calorific value basis) in kg/GJ (kilogramme/gigajoule), COF is the carbon oxidation factor.

For calculating uncertainties, the approach 1 is applied, which is described in IPCC Guidelines [2]. Here the expression is used, which combines the uncertainties (addition and subtraction):


161

( ) ( ) ( )k

kktotal x...xx

xU...xUxUU

+++⋅++⋅+⋅

=21

2222

211 , (2)

where totalU is the percentage uncertainty in the sum of the quantities (half of the 95-percent confidence interval divided by the total (i.e., mean) and expressed as a percentage); this term ‘uncertainty’ is thus based upon the 95-percent confidence interval, ix and iU are the uncertain quantities, and the percentage uncertainties associated with them, respectively.

The greenhouse gas inventory is principally the sum of products of emission factors, activity data, and other estimation parameters. Therefore, this expression (2) can be used repeatedly to estimate the uncertainty of the total inventory.

Achieved results are described below.

4. Projection of total uncertainty levels

Using the methodology described in previous section the projections of future levels of uncertainty of GHG emissions in CO2 equivalent were made for the period 2005-2020 considering EU’s “20-20-20” targets. The projections refer to EU-15, because data required is available only for EU-15. One of the main assumptions is that energy demand is constant during investigated period.

The Figure 2 displays calculated projections of emission uncertainties for EU-15 countries with assumptions:

cut GHG emissions in all sectors according to EU’s “20-20-20” targets, i.e. 21% – sectors covered by ETS – Emission Trading System, 10% – sectors non-covered by ETS (as result, uncertainty will increase by 0.2% per year; this value is very small compared to two previous cases, due to reduction of GHG emissions not only in energy sector but also in all others).

Figure 2. Projection of uncertainty level of GHG emissions in CO2–equivalent

considering new targets for EU-15 States

Increase in uncertainty is not a positive effect, even if the values are small. However, these changes in uncertainty can be balanced if taken into account another driver of change – “learning process”.

5. Change in uncertainty considering new measures in Energy

Energy sector is the most important sector with about 80% of total EU emissions. GHG mitigation measures are set mainly in this sector.


162

Key assumptions that were used to make projections of uncertainty level in CO2 emissions from fossil fuel combustion are described below.

GHG emissions including CO2 emissions from renewable energy sources are either low or zero. Increasing the share of renewable energy sources in the EU fuel mix will therefore result in significantly lower GHG emissions. The additional renewable energy deployment needed to achieve the 20% target will reduce annual CO2 emissions in a range of 600-900Mt in 2020 [9].

One of the important renewable energy sources is biomass [13]. Amounts of biomass used as fuels are included in the national energy consumption reports but the corresponding CO2 emissions are not included in the national total emissions as it is assumed that the biomass is a product of sustainable manner. If the biomass is harvested at an unsustainable rate the amount of carbon dioxide emissions are accounted for as a loss of biomass stocks in the Land-Use, Land-Use Change Forestry sector [8].

Projections were made only for EU-27. All data required are available from National Inventory Reports [7], and European GHG Inventory Report [8]. Total uncertainties by gas for EU were calculated considering reported data of 26 Member States, except Malta, because this country does not report uncertainties.

All projections from the previous section are based on the assumption that energy consumption will be constant. Nevertheless, considering changes in energy demand we can receive prognoses that are more realistic. Two scenarios described in Second Strategic Review [10] to examine European future energy are used to make projections of uncertainty levels:

- Baseline – is an energy demand which is projected according to current trends and policies [11];

- New Energy Policy – is the case of taking action to achieve agreed EU targets on climate change mitigation, mainly a reduction of 20% in GHG emissions compared to 1990, along with the share for renewables in the final energy demand by 2020, and to bring about a substantial improvement in energy efficiency.

Except from the policy assumptions, all the other assumptions (technology, economic structure, demographic development, etc.) remain unchanged between the Baseline case and the New Energy case. Both scenarios start from common projections, notably on economic growth (2.2 % on average up to 2020). The Baseline includes current trends and policies as implemented in the Member States up to the end of 2006 (more information in [11]). The New Energy Policy scenario assumes vigorous implementation of new policies to make substantial progress on energy efficiency for reaching other energy and climate targets. The 20% RES (Renewable Energy Sources) and GHG targets are assumptions for the New Energy Policy.

Both the Baseline and New Energy Policy cases depend on a moderate or high oil price environment. The moderate price environment means an oil price of 61$ (2005)/barrel in 2020. The high price environment would have an oil price of 100 $/barrel in 2020.

The results are shown in the Figure 3. In case of Baseline scenario:

moderate oil price – uncertainty will decrease by 0.13% per year; because of using more gas in fossil fuels combustion that has low uncertainty;

high oil price – uncertainty will decrease by 0.19% per year; because of changes in fossil fuels combustion: less oil and gas, more solids, that has bigger uncertainty, but very small amount to compare with other fuels.


163

Figure 3. Changes in uncertainty of CO2 emissions of fossil fuels combustion

consider future energy demand for the period 2005-2020

In case of New Energy Policy scenario: moderate oil price – uncertainty will increase by 0.5% per year; because of

combination of reductions in combustion of all fuels, using more renewables; high oil price – uncertainty will increase by 0.27% per year; because of

reductions in combustion of all fuels, but bigger than in case with moderate price. So, implementation of New Energy Policy can cause the increase of uncertainty of

GHG emissions due to structural changes in fossil fuel combustion, but the percentage of change is very small and can be balanced by another factor of uncertainty change – “knowledge increase”.

6. Conclusions Uncertainty of GHG emissions changes over time because of two factors: structural

changes in GHG emissions and “knowledge increasing”. Projections of total relative uncertainty for EU-27, which were made considering

EU’s “20-20-20” targets for the period 2005-2020, show that under new measures the level of uncertainty will increase by 0.2% per year. This change is very small.

Made projections of future uncertainty levels in Energy sector confirm that percentage of change in uncertainty due to structural changes in GHG emissions is negligible; it varies from 0.27% to 0.5% per year. Even if these values are too small, they show the increase of relative uncertainty levels. If take into account another factor of change – knowledge increasing, these values can be balanced but the achieved reduction of uncertainty level will not be so effective. Besides, experiments were made for CO2 emissions excluding LULUCF sector which is the most stable and does not change significantly in time. Increase of renewables in energy consumption can cause the increase of GHG emissions in other sectors, and it would lead the increase of total uncertainty level.

The knowledge increasing with time is very important to better estimate uncertainties. Structural changes in GHG emissions cause small percentage of relative uncertainty change, so we should put more effort on knowledge increasing.

References

[1] Revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories, Vol. 1: Greenhouse Gas Inventory Reporting Instructions; Vol. 2: Greenhouse Gas Inventory Workbook; Vol. 3: Greenhouse Gas Inventory Reference Manual. Intergovernmental


164

Panel on Climate Change (IPCC) Working Group I (WG I) Technical Support Unit, IPCC/OECD/IEA, Bracknell, United Kingdom, 1997. Available at: http://www.ipcc-nggip.iges.or.jp/public/gl/invs1.htm.

[2] 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas Inventories Programme, Eggleston H.S., Buendia L., Miwa K., Ngara T. and Tanabe K. (eds), 2006.

[3] Jonas M., Nilsson S., Bun R. et al. Preparatory Signal Detection for Annex I Countries under the Kyoto Protocol―A Lesson for the Post-Kyoto Policy Process. Interim Report IR-04-024, International Institute for Applied Systems Analysis, Laxenburg, Austria, 2004, 91 p. Available at: http://www.iiasa.ac.at/ Publications/Documents/IR-04-024.pdf

[4] Bucki R. Mathematical modelling of allocation processes as an effective tool to support decision making / R. Bucki // Information and Telecommunication Systems. – Bielsko-Biała : Polish Information Processing Society, 2008. – N. 17. – P. 7-13.

[5] Hamal K. Reporting greenhouse gas emissions: Change in uncertainty and its relevance for the detection of emission changes. Interim report. Laxenburg, Austria: International Institute for Applied Systems Analysis, 2009 (forthcoming).

[6] European Commission. EU action against climate change – Leading global action to 2020 and beyond (2009). Directorate-General for the Environment Information centre (BU-9 0/11) B-1049 Brussels. Available at: http://ec.europa.eu/environment/climat/pdf/brochures/post 2012 en.pdf.

[7] National Inventory Reports (2003-2007) under the UNFCCC Treaty. Available at: http://unfccc.int/national_reports/annex_i_ghg_inventories/national_ inventories_submissions/items/4303.php

[8] Annual European Community Greenhouse Gas Inventory 1990–2007 and Inventory Report 2009. Technical Report No. 4, European Environment Agency (EEA), Copenhagen, Denmark, 2009. Available at: http://www.eea.europa. eu/publications/european-community-greenhouse-gas-inventory-2009.

[9] COM (2006) 848 final. Renewable Energy Road Map – Renewable energies in the 21st century – building a more sustainable future. Brussels, 10.01.2007. Available at: http://ec.europa.eu/energy/energy_policy/doc/03 renewable energy roadmap en.pdf.

[10] SEC 2871. Second Strategic Review – An EU security and solidarity action plan – Europe’s current and future energy position – Demand – resources – investments. Brussels, 2008. Volume 1. Available at: http://ec.europa.eu/energy/ strategies/2008/2008_11_ser2_en.htm.

[11] European Communities (2008). European Energy and Transport – Trends to 2030 – Update 2007. Prepared by the Institute of Communication and Computer Systems of the National Technical University of Athens (ICCS-NTUA), E3M-Lab, Greece. Available at: http://ec.europa.eu/dgs/energy_transport/figures/ trends_2030_update_2007/energy_transport_.

[12] EEA. Greenhouse Gas Emission Trends and Projections in Europe 2009. Report No. 9, European Environment Agency (EEA), Copenhagen, Denmark, 2009. Available at: http://www.eea.europa.eu/publications/eea_report_2009_9.

[13] Zanchi G., Pena N., Bird N. The upfront carbon debt of bioenergy. Graz, Austria: Joanneum Research, May 2010.


165

Reducing the uncertainty of methane recovered (R) in greenhouse gas inventories from waste sector and of adjustment factor (AF) in landfill

gas projects under the clean development mechanism

George Magalhães, Francisco do Espirito Santo Filho, João Wagner Alves, Matheus Kelson, Roberta Moraes

School of Arts, Sciences and Humanities. University of Sao Paulo Av. Arlindo Betio, 1000. 03828-000. Sao Paulo – Brazil

[email protected]

Abstract

This paper aims to offer a contribution to better understanding about the methane emissions from passive systems in landfills, helping to reduce associated uncertainties for the definition of the values of Recovered methane (R) in greenhouse gas (GHG) emission inventories from waste sector in Brazil and to contribute for better estimating of the parameter Adjustment factor (AF) in landfill gas (LFG) destruction projects under the Clean Development Mechanism (CDM). In Brazilian National GHG inventory, the R value adopted was 0. In other words, the amount of methane destructed by passive system at Brazilian landfills is insignificant. Furthermore, the Brazilian landfills gas destruction CDM projects commonly adopted an AF value as 0.20 (20%). This value was adopted by more than 50% of the Brazilian LFG destruction CDM project activities registered at the UNFCCC by year 2008. Based on a sample with 154 Brazilian municipal solid waste landfills and on IPCC and UNFCCC methodologies, it was estimated the R and AF to these landfills. The results of paper show that R, in the National GHG inventory of waste sector, and AF, in CDM landfills projects, probably were underestimated and overestimated, respectively. Default values to R and AF are proposed in the conclusions.

Keywords: Greenhouse gas, CDM, methane, biogas, landfills

1. Introduction

Methane (CH4) content in LFG can be recovered and destroyed by flaring or, as a more convenient alternative, be used as energy source (fuel) for electricity or heat generation, which can meet the energy demand of landfills, wastewater treatment plants, composting plants and all of other anaerobic treatments applied to solid waste or wastewater. Methane is an important GHG and, by recovering or destroying methane, the total GHG emissions are reduced.

Estimating the amount of recovered methane in landfills is important in national greenhouse gas inventories because CH4 emissions from the municipal solid waste (MSW) disposal represents a significant part of emissions in most of the developing countries where landfilling is the common practice in terms of MSW management.

In case of LFG capture and destruction initiatives registered as project activities under the CDM or under other GHG emission reduction trade schemes, the amount of methane which is recovered in the landfill by passive systems before the project implementation is considered as part of baseline emissions. In most cases, when data of methane recovered before project implementation is not available, is usual to estimate


166

the average share of methane recovered based on common practice adopted by similar landfills in the country or region and also by taking into account applicable national or regional legislation dealing with the issue.

In Brazil, while so far there is no regional or national legislation which requires or recommends LFG to be captured and partially destroyed (burnt) due to safety or environmental reasons. The recovery of methane generated by anaerobic decomposing of solid waste at landfills has not been a common practice in the country. So, in many of the Brazilian landfills which have a passive system to collect and flare LFG, such devices are normally designed in a very rudimentary manner and/or LFG is captured and burnt in a very unsystematic (not controlled or monitored) and ineffective manner. As a result, the collection and destruction efficiencies for such flares are very low.

It is important to highlight that, in Brazil, there are no landfills with active system (forced) to collect and destroy methane that does not enjoy the benefits of CDM. All Brazilian landfills with high methane collection and destruction efficiency are implemented under the CDM.

In passive LFG burn systems in Brazil, normally there is no control about the LFG collection efficiency, LFG flaring efficiency or control over the fraction of time the chimneys are actually lit. Moreover, there is not enough number of LFG extracting wells installed in the landfill area that would ensure a reasonable and safe efficiency for the chimneys flaring LFG when compared against the total amount of LFG generated in these sites.

According to IPCC methodologies [1], [2], [3] to national GHG emission inventories from waste sector, recovered methane must always to be discounted from the estimated amount of generated methane when estimating net GHG emissions from landfills.

It is important to reduce the associated uncertainties of GHG inventories, mainly of the national inventory, thus generating GHG emissions reports which are quantitatively close to the reality of Brazilian emissions, it is crucial to offer a better basement and methodological approaches to GHG emission inventories developers, policy makers, decision makers, planners and stakeholders.

1.1 Definition of Methane recovered (R) and Adjustment factor (AF)

R is defined as the recovered and destroyed methane at landfills. This data belongs to national inventory on GHG emissions from solid waste management. The Intergovernmental Panel on Climate Change - IPCC provides methodology to estimate national methane emissions from solid waste management and R is the amount that is discounted for being destroyed.

AF is an Adjustment factor parameter related to the estimation of destroyed methane in the baseline scenario for LFG capture and destruction. To estimate AF in the CDM projects could be used the UNFCCC ACM0001 [4] or AMS-III.G baseline methodologies. The definition of AF is influences directly the baseline emissions for such CDM project activities, thus it influences quantitatively the amount of Certified Emission Reductions (CERs) to be claimed by the CDM projects, since this is estimate from the baseline.

The Adjustment factor (AF) and methane Recovered (R) are functions of the fraction of methane destroyed by passive LFG venting systems. In Brazil, at Reference report on


167

methane emissions from solid waste disposal, related with National Communication, R is currently assumed as zero [5]. Commonly, AF has been historically set as 0.20 (20.0%) in LFG CDM projects in Brazil comprising LFG collection and destruction [4]. Some observations about the common practice in terms of LFG passive venting and flaring in landfills in Brazil suggests that adopted values for R and AF are underestimated and overestimated, respectively, therefore affecting the correctness accounting in GHG inventories and CDM project activities, thus increasing the uncertainty in theses estimations.

1.2 GHG inventories

The Brazilian Reference report on methane emissions from solid waste management indicates that controlled and efficient LFG collection and flaring are not a common procedure in landfills or dump sites in operation around the country. Most common solutions are passive systems with uncontrolled open burning of LFG where the system is most of the time kept unlit. The level of control and efficiency of LFG which is actually flared is thus very low.

After approval of Kyoto Protocol, its ratification by Brazilian Government and the creation of the Carbon Markets, Brazil registered its first projects with methane capture and destruction, where in some cases it includes energetic recovery. Nevertheless, Brazil has 5,564 municipalities [6], and the total of projects is less than 30 [5] which represents only a small fraction of the landfills and dump sites in the country.

The National Communication is part of Brazilian Government commitments at United Nations Framework Convention on Climate Change - UNFCCC. In this document[5] it is acknowledged that only the recovered methane that is collected by forced LFG extracting systems and flared in enclosed high efficiency flares by registered CDM projects have their environmental contribution on mitigation of climate change properly accounted.

2. Methodology

The IPCC provides a methodology to estimate CH4 emissions in National GHG inventories from waste sector [2]. The estimation of CH4 generated at SWDS, based on a First order decay (FOD) modelling, comprise data about the amount of degradable organic carbon disposed at solid waste disposal site (SWDS), the anaerobic conditions of SWDS, the climate in the region of SWDS, the oxidation factor of SWDS and on a normalization factor. According to this methodology the recovered methane must always to be discounted from the estimated amount of generated methane when estimating net GHG emissions from landfills.

Therefore, to estimate R it was used the equation below,

)()()()()()( .).( xixPRxixxBLx QMDQQMDR +−= , (1)

where MDBL is the methane destroyed in baseline at the year x, Q is the amount of methane generated at the year x, Qi is the amount of methane generated by LFG recovery projects at year x, MDPR is the methane destroyed by the LFG recovery projects at year x, i is the projects that recovery and destroy methane and x is the year of inventory.


168

For CDM projects, as established by the most recent versions of applicable CDM baseline methodologies (ACM0001 [4] and AMS-III.G) the amount of methane that would have been destroyed/combusted during the year in the absence of the project shall be discount from the total LFG captured due to the project as part of the emission reduction determination.

According to ACM0001, in the absence of regulatory, contractual requirements or historic data, the AF shall be estimated by a suitable procedure chosen by the project developer. As required by CDM rules and procedures, all calculations must be a third part (auditing company during the validation and/or verification phases for the proposed CDM project activities).

Therefore, according to the UNFCCC ACM0001 methodology, to estimate AF to the LFG recovery project activities it was used the equation below,

PRBL MDMDAF /= , (2)

where MDBL is the methane destroyed at the baseline of the project and MDPR is the methane destroyed by the LFG recovery project.

The MDBL can be estimated from the equation

MDBL = Ps . Ca . Ftc . CEOF , (3)

where Ps is the fraction of LFG vented through the passive system, in the landfill in question, Ca is the fraction of chimneys available for flaring, Ftc is the fraction of time chimneys are actually lit and CEOF is the combustion efficiency of an open flame.

To estimate the AF from the equation (2), it is necessary to estimate MDPR. This parameter is estimated from the product between Collection efficiency of the project (CEPR) and Flare efficiency (FE).

So, the formula to estimate AF can be written as follow:

)./()...( FECECEFtcCaPsAF PROF= , (4)

Complementarily, the LFG vented through the passive system (Ps) can be estimated by the equation

Ps = (W / We) . CEPS , (5)

where W is the number of wells installed, We is the number of wells expected and CEPS is the collection efficiency of passive system.

Finally, the estimation of the number of Wells expected (We) can be calculated by the equation

We = A / Wi2 , (6)

where A is the available area of the landfill used to disposal MSW and Wi2 is the density of the grid of wells. This density is equal the square of the distance between wells (Wi).

3. Data

3.1 Landfill, landfill area, wells and wells expected

This study was performed on based on available data published by Brazilian Ministry of Cities [7], Brazilian Ministry of Environment [8] and data provided by


169

landfill managers and other contributors. For this reason, for better information accuracy, it would be necessary a field research or data gathering study at all landfills.

According to TNO [9], World Bank [10] and US-EPA [11], and adopting a conservative approach, the collection efficiency used in this paper is 0.75. These references are detailed on section 3.6. To reach an optimized LFG collection efficiency, Brazilian experts on MSW disposal recommend the use a 30x30m grid of LFG extracting and flaring wells. By adopting a conservative approach, it was considered that all Brazilian landfills follow the suggestion of Brazilian experts and have a 30x30m grid of wells. The authors acknowledge emphasize this does not reflect the reality in landfills in Brazil. The adopted approach is, thereby, conservative.

Bellow are presented for the selected Brazilian landfills the data about landfill location, landfill area (only the available area to solid waste disposal), operation type, number of wells and distance between wells. The number of wells expected (We) for each landfill was estimated by applying the Equation (6).

In some cases, data for some of landfills listed above could not be up to dated. Similarly, some landfills could currently have implemented forced LFG collection and destruction or utilization projects under the CDM.

3.2 Collection efficiency of passive system – CEPS

In the absence of consolidated references about CEPS and adopting a conservative approach, it was assumed that the CEPS is equal the CEPR used for active (forced) LFG systems. The authors acknowledge emphasize this does not reflect the reality for existent passive LFG capture and venting/flaring systems commonly used in landfills around the country. The adopted approach is, thus, conservative. So, the CEPS adopted in this study were 0.75.

3.3 Chimneys available for flaring – Ca

The National System of Sanitation Data (SNIS, in Portuguese) annually issue results of a permanent data collection in landfills. This data collection includes a question about “existing biogas venting” (yes = “sim” or no = “não”).

Considering the SNIS results, more than 50% of landfills must not have methane collection, so that, there are not chimneys available for flaring in these landfills is 0.

Furthermore, alignment problems, high level of leachate, covering material, soil consolidation, closure and other factors negatively influences on fraction of wells available for flare LFG in these landfills.

Because there are no data about it, a conservative approach was adopted, this paper considers the country average Ca as 0.50. The authors acknowledge emphasize this may not reflect the reality for existent passive LFG capture and venting/flaring systems commonly used in landfills around the country. The adopted approach is thus conservative.

3.4 Fraction of time the chimneys are actually lit – Ftc

Rainfall, wind, LFG pressure and operational characteristics are the main reason to wells stop to flare methane. Also, there are no data about this. So, adopting a conservative approach, this paper estimates Ftc as 0.50. The authors acknowledge emphasize may not reflect the reality for existent passive LFG capture and venting/flaring systems commonly used in landfills around the country. The adopted approach is thus conservative.


170

Table 1. Landfill location, operation, area (A), number of wells (W), distance

between wells (Wi) and number of wells expected (We)

Landfill location Operation A W Wi We

[m2] [wells] [m] [wells] Americana Municipal 650,000 20 30 178

Belo Horizonte* Municipal 59,600 110 30 722 Betim Municipal 60,000 24 30 66

Blumenau Municipal 101,400 14 30 67 Caieiras* Private 95,000 37 30 113 Camaçari n.a. 106,420 50 30 106

Carapicuíba Municipal 600,000 0 30 118 Contagem Municipal 200,000 24 30 667

Cuiabá Municipal 314,000 13 30 222 Curitiba Private 45,000 200 30 349

Duque de Caxias* Private 148,000 49 30 1,556 Embu Municipal 270,000 4 30 50

Goiânia Private 100,000 44 30 300 Gravataí Municipal 240,000 44 30 111 Guarujá Private 128,000 70 30 267

Itaquaquecetuba* Private 600,000 30 30 142 Jaboatão dos Guararapes Municipal 1,000,000 16 30 667

João Pessoa* Municipal 300,000 6 30 1,111 Joinville Private 600,000 30 30 333

Natal Mixed 200,000 5 30 667 Niterói Municipal 170,000 34 30 222 Osasco Private 98,000 101 30 189 Palmas Municipal 705,000 30 30 109

Paulínia* Private 100,000 59 30 783 Ribeirão das Neves Municipal 1,400,000 3 30 111

Salvador* Private 650,000 55 30 164 Santos Municipal 47,268 19 30 53

São Francisco do Conde Municipal 35,533 10 30 39 São Leopoldo Private 40,000 9 30 44

São Paulo – Bandeirantes* Private 1,500,000 400 30 1,667 São Paulo – São João* Private 800,000 125 30 889

Serra* Private 155,025 28 30 172 Valinhos Private 190,000 50 30 211 Vera Cruz Municipal 270,000 6 30 300 Vitória* Private 1,172,000 8 30 1,302

* – landfill currently with CDM project.

n.a. – Data not available.

Sources: [7][4].


171

3.5 Combustion efficiency of an open flame – CEOF

The combustion efficiency of open flame, for the time of which the chimney is actually lit, varies with the type of flare. So, in accordance with experts’ judgment, a conservative value of CEOF is 0.50.

3.6 Collection efficiency of the project – CEPR

The collection efficiency of active (forced) LFG extracting systems is much higher than the collection efficiency of passive systems. Based on international references, this paper considered CEPR as 0.75.

The references advise that “It is necessary to apply an efficiency or collection factor to the above-noted LFG generation estimates to assign a recoverable quantity of LFG. Essentially, this factor recognizes that not all of the LFG that is being generated can be collected. There are economic tradeoffs between cost of collection and efficiency of collection. A well designed gas collection system can typically collect 75 percent of more of the total quantity of gas that is generated. The level of certainty and inherent financial risk can be effectively managed assuming that the appropriate input parameters are assigned to the modeling and there is a sound understanding of the specific waste management system for the site.” [10].

And, the “[…]Collection systems operated for energy recovery may be more efficient than those where the collected gas is not put productive use because each cubic foot of gas will have a monetary value to the owner;/operator. In addition, newer systems may be more efficient than the average system in operation today. Nevertheless, there continues to be economic limits on the tightness of well spacing and other factors that are difficult or impossible to control. Therefore, a reasonable assumption for a newer collection system operated for energy recovery is 75 to 85 percent collection efficiency. Multiplying the total landfill gas generation estimated by methods A or B by 75 to 85 percent should yield a reasonable estimate of landfill gas available for energy recovery.” [11]

And complementarily, “At several landfills gas was recovered. […] In some cases, the efficiency is very low. This is related to the capacity of the utilization. If this is limited, there is no need for the recovery of larger amounts of landfill gas. When gas recovery is optimal, differential efficiencies are up to 50-75%. When a top liner system is applied, or when a landfill is capped with clay, efficiencies are up to almost 100%.” [9]

These references were taken into account for defining CEPR as 0.75.

3.7 Flare efficiency (FE)

The Flare efficiency – FE for the active (forced) LFG extracting and flaring system using enclosed flares is set as 0.99. To assure this efficiency the temperature must be above 850oC and under 1.200oC. This data could be checked at UNFCCC CDM projects.

4. Results

4.1 LFG vented through passive system - Ps

Based on Equation 5, it was estimated Ps to the landfills of the sample that have available data (35 landfills). The other 119 landfills of the sample have Ps as 0. Ps data are presented at Table 2.


172

Table 2. LFG vented through passive system – Ps

City Ps City Ps

Americana 0.0844 Joinville 0.0675 Belo Horizonte 0.1142 Natal 0.0056

Betim 0.2718 Niterói 0.1148 Blumenau 0.1575 Osasco 0.4010 Caieiras 0.2463 Palmas 0.2066

Camaçari 0.3553 Paulínia 0.0565 Carapicuíba 0.0000 Ribeirão das Neves 0.0203 Contagem 0.0270 Salvador 0.2508

Cuiabá 0.0439 Santos 0.2713 Curitiba 0.4299 São Francisco do Conde 0.1900

Duque de Caxias 0.0236 São Leopoldo 0.1519 Embu 0.0600 São Paulo – Bandeirantes 0.1800

Goiânia 0.1100 São Paulo – São João 0.1055 Gravataí 0.2970 Serra 0.1219 Guarujá 0.1969 Valinhos 0.1776

Itaquaquecetuba 0.1582 Vera Cruz 0.0150 Jaboatão dos Guararapes 0.0180 Vitória 0.0046

João Pessoa 0.0041

Average Ps 0.1411

4.2 Methane destroyed at baseline – MDBL

Observing a sample with 226 Brazilian landfills, 119 do not have wells installed [7]. It means that there is a great possibility to 52.65% of Brazilian landfills have zero as MDBL.

According to the Equation 3 and using data from Table 1 and Table 2 was estimated the MDBL to the Brazilian landfills that have available data.

Thus, the Brazilian weighted average MDBL, including the 119 landfills that it is zero, is presented in Table 3.

In Table 3 is also presented the Sample average MDBL. This average does not consider the landfills that do not have wells installed (119 landfills). It is notorious that the Sample average MDBL is a value higher than Weighted average MDBL, because all landfills that have zero as MDBL was only considered in the Weighted average MDBL.

It is reinforced that it would be necessary a field research or data gathering study at all landfills for a better information accuracy.

4.3 Adjustment factor – AF

With the Brazilian weighted average MDBL presented at Table 3, and MDPR as 0.7425, the weighted average AF was estimated to each landfill with available data and it is presented in Table 4. To other 119 landfills that does not have methane collection and destruction system, the AF is considered 0.


173

Table 3. Methane destroyed in baseline – MDBL

City MDBL City MDBL

Americana 0.0105 Joinville 0.0084 Belo Horizonte 0.0143 Natal 0.0007

Betim 0.0340 Niterói 0.0143 Blumenau 0.0197 Osasco 0.0501 Caieiras 0.0308 Palmas 0.0258 Camaçari 0.0444 Paulínia 0.0071

Carapicuíba 0.0000 Ribeirão das Neves 0.0025 Contagem 0.0034 Salvador 0.0314

Cuiabá 0.0055 Santos 0.0339 Curitiba 0.0537 São Francisco do Conde 0.0237

Duque de Caxias 0.0030 São Leopoldo 0.0190 Embu 0.0075 São Paulo – Bandeirantes 0.0225

Goiânia 0.0138 São Paulo - São João 0.0132 Gravataí 0.0371 Serra 0.0152 Guarujá 0.0246 Valinhos 0.0222

Itaquaquecetuba 0.0198 Vera Cruz 0.0019 Jaboatão dos Guararapes 0.0023 Vitória 0.0006

João Pessoa 0.0005 Landfills where MDBL = 0 119 Total of landfills 154 Sample average MDBL 0.0176 Weighted average MDBL 0.0040

The AF results are presented, to landfills of the sample that do not have CDM

projects installed, in the hypothetical case of they become CDM projects with the implementation of active (forced) LFG extracting and flaring systems, using enclosed high efficiency flares.

In Table 4 are also presented the AF actually adopted by some of the currently registered CDM projects (AFPR) for the mere purpose of comparison.

5. Conclusions

Analyzing the estimated results to the methane destroyed in baseline (via passive systems) to Brazilian sample of landfills, we conclude is possible that the Brazilian National Communication, when consider R as zero had underestimated its data.

The authors believe that adopting the approach presented in this paper is a good way to have a lower uncertainty, even minimally, in Brazilian Reference report on methane emissions from solid waste disposal and other greenhouse gas inventories from waste sector, part of National Communication. Thus, is recommended, if there is no available data, to adopt a R value close to 0.0040. It is important to mention that this R value represents the fraction of methane generated that was recovered. To estimate the tonnes of methane recovered, the R must be multiplied by the amount of methane generated.


174

Table 4. Estimated MDBL, MDPR and AF to Brazilian landfills

City MDBL MDPR AF AFPR City MDBL MDPR AF AFPR

a b c=a/b a b c=a/b

Americana 0.0105 0.7425 0.0142 n.p. Joinville 0.0084 0.7425 0.0114 n.p.

Belo Horizonte 0.0143 0.7425 0.0192 n.p. Natal 0.0007 0.7425 0.0009 n.p.

Betim 0.0340 0.7425 0.0458 n.p. Niterói 0.0143 0.7425 0.0193 n.p.

Blumenau 0.0197 0.7425 0.0265 n.p. Osasco 0.0501 0.7425 0.0675 n.p.

Caieiras 0.0308 0.7425 0.0415 0.20 Palmas 0.0258 0.7425 0.0348 n.p.

Camaçari 0.0444 0.7425 0.0598 n.p. Paulínia 0.0071 0.7425 0.0095 0.20

Carapicuíba 0.0000 0.7425 0.0000 n.p. Ribeirão das

Neves 0.0025 0.7425 0.0034 n.p.

Contagem 0.0034 0.7425 0.0045 n.p. Salvador 0.0314 0.7425 0.0422 n.ap.

Cuiabá 0.0055 0.7425 0.0074 n.p. Santos 0.0339 0.7425 0.0457 0.20

Curitiba 0.0537 0.7425 0.0724 n.p. São

Francisco do Conde

0.0237 0.7425 0.0320 n.p.

Duque de Caxias

0.0030 0.7425 0.0040 0.05 São

Leopoldo 0.0190 0.7425 0.0256 n.p.

Embu 0.0075 0.7425 0.0101 n.p. São Paulo -

Bandeirantes 0.0225 0.7425 0.0303 0.20

Goiânia 0.0138 0.7425 0.0185 n.p. São Paulo - São João

0.0132 0.7425 0.0178 0.20

Gravataí 0.0371 0.7425 0.0500 n.p. Serra 0.0152 0.7425 0.0205 n.p.

Guarujá 0.0246 0.7425 0.0331 n.p. Valinhos 0.0222 0.7425 0.0299 n.p.

Itaquaquecetuba 0.0198 0.7425 0.0266 n.p. Vera Cruz 0.0019 0.7425 0.0025 n.p.

Jaboatão dos Guararapes

0.0023 0.7425 0.0030 n.p. Vitória 0.0006 0.7425 0.0008 n.p.

João Pessoa 0.0005 0.7425 0.0007 0.10

Brazilian sample average MDBL, MDPR and AF 0.0176 0.7425 0.0238

Brazilian weighted average MDBL, MDPR and AF 0.0040 0.7425 0.0054

n.p.: no LFG extracting and destruction CDM project activity is registered and implemented at landfill.

n.ap.: not applicable to this landfill.

The CDM is one important initiative to disseminate LFG recovery high efficiency

practice at Brazilian landfills. It is important to highlight that all of landfills with high methane collection and destruction efficiency (active system) are activities implemented under the CDM.


175

UNFCCC has registered 142 LFG projects in the world at the moment (July, 15th 2010), of which more than 30% had set the value of the parameter AF as zero. (UNFCCC, 2010)

It is possible that designers of the first CDM projects in Brazil comprising LFG capture and destruction have adopted an overestimated and conservative approach for determining the value of the parameter AF. Indeed, AF was set as 0.20 (20%) in more than 50% of the Brazilian LFG capture and destruction CDM project activities registered at the UNFCCC by year 2008. Moreover, for about a quarter of the registered CDM projects, AF set was 0.10 (10%).

Considering the weighted average AF estimated for the sample of Brazilian landfills as 0.0054 (as shown in Table 4), it is possible that designers of the first CDM projects in Brazil, comprising LFG capture and destruction, have adopted an overestimated approach for determining the value of the parameter AF.

To reduce uncertainty of the value of AF in new CDM project activities comprising LFG collection and destruction (using ACM0001 or AMS-III.G CDM baseline methodologies) to be proposed in the future, the results of this paper should be considered.

References

[1] IPCC – International Panel on Climate Change. (1996): Guidelines for national greenhouse gas inventories. United Kingdom, IPCC.

[2] IPCC – International Panel on Climate Change. (2000): Good practice guidance and uncertainty management in national greenhouse gas inventories. Japan, IPCC.

[3] IPCC – International Panel on Climate Change. (2006): 2006 IPCC Guidelines for greenhouse gas inventories. Japan, IPCC.

[4] UNFCCC – United Nations Framework Convention on Climate Change. (2010): Approved consolidated methodology – ACM0001: Consolidated baseline and monitoring methodology for landfill gas projects activities. Available at <http://cdm.unfccc.int/methodologies/PAmethodologies/approved.html>. Accessed on March, 2010.

[5] BRAZIL. Ministry of Science and Technology. (2006): Methane emissions from the solid waste disposal and treatment – Reference Report. Brasília, Ministry of Science and Technology. 86 p.

[6] IBGE – Brazilian Institute of Geographic and Statistics. (2010): Brazil in synthesis. Available at <http://www.ibge.gov.br/brasil_em_sintese/default.htm>. Accessed on: March, 2010.

[7] SNIS –National System of Sanitation Data. (2008): Management diagnostic of municipal solid waste - 2007. Brazilian Ministry of Cities, Database.

[8] BRAZIL. Ministry of Environment. (2001): Research of potential electrical generation in big cities in Brasil – Estudo do potencial de geração de energia renovável proveniente de aterros sanitários nas regiões metropolitanas e grandes cidades do Brasil. Brasília, Brazilian Ministry of Environment.

[9] TNO - Institute of Environmental and Energy Technology. (1995): Landfill gas formation, recovery and emissions. Contract 3555220/1910. Netherlands, TNO. 91 p.


176

[10] WORLD BANK. (2003): Handbook for the preparation of landfill gas to energy projects in Latin America and Caribbean. Draft, Ref. 19399(6). p 78 and a curve at figure 8.3.

[11] US-EPA. (1996): Turning a Liability into an Asset: A Landfill gas-to-energy project development handbook. EPA 430-B-96-0004. p 2-8.


177

The U.S. NRC report on monitoring and verification of national greenhouse gas emissions inventories

Gregg Marland

Environmental Sciences Division Oak Ridge National Laboratory

Oak Ridge, TN 37830-6335 USA [email protected]

Abstract

In March, 2010, the U.S. National Research Council (NRC) released a report entitled “Verifying Greenhouse Gas Emissions: Methods to Support International Climate Agreements”. This report looks at the state of national greenhouse gas emissions inventories and asks what is necessary to decrease the uncertainty in emissions estimates and to provide independent monitoring and verification of emissions estimates in the case of an international agreement to establish emissions limitations.

The protocols of the UN Framework Convention on Climate Change provide a monitoring process that focuses on procedural and reporting issues but does not have access to data that would provide independent verification of self-reported, national emissions estimates. The NRC report explores ways to improve both self-reported estimates and other nation’s ability to verify them. The report concludes that accounting for CO2 emissions would be improved if all countries had the capacity for full bottom-up and top-down accounting of all sources and sinks. Additional atmospheric monitoring from ground-based stations and space-based measurements could permit independent information for verification or falsification of some components of emissions inventories. Whereas useful data for verification of CO2 emissions could be achieved within a few years, estimates of emissions of other greenhouse gases will remain uncertain in the near term. But focus on CO2 would enable independent checks (with <10% uncertainty) on fossil-fuel combustion and deforestation, which are responsible for three-fourths of UNFCCC greenhouse gas emissions. This paper presents a synopsis of the context and content of the NRC report by one of its authors.

Introduction

The United Nations Framework Convention on Climate Change (UNFCCC) requires that countries listed in its Annex I submit annually “national inventories of anthropogenic emissions by sources and removal by sinks of all greenhouse gases not controlled by the Montreal Protocol” (Article 4). Since 2003 these inventories have been subject to an annual technical review process. The Kyoto Protocol requires that those countries listed in its Annex B (essentially the same list) submit an annual inventory of greenhouse gas emissions and requires that expert review teams provide, annually, a “thorough and comprehensive technical assessment of all aspects of the implementation by a Party of this Protocol” (Article 8). These expert review teams are able to review processes, procedures, data sources, emissions factors, etc., but they do not have access to independent information that would allow them to check the reliability of self-reported, national emissions inventories. Since the time of drafting of the UNFCCC there have been questions about the importance of independent confirmation of emissions


178

inventories but there simply is not data on which to base such analyses, and many observers have held the opinion that “current reporting and review requirements are sufficiently rigorous to provide a reliable basis to assess implementation of Parties’ quantitative emissions targets” (Breidenich and Bodansky, 2009; see also U.S. Government Accountability Office, 2010).

Nonetheless, confidence in an international treaty would be greatly enhanced by the ability of the international community to provide some independent verification of emissions inventories, or to at least be able to provide falsification of some components of an emissions inventory. As former U.S. president Ronald Reagan was frequently quoted: “Trust but verify”. A recent report from the U.S. National Research Council (NRC) has examined the prospects for improving verification of national greenhouse gas emissions inventories. We examine here the context of the NRC report and summarize its perspective and conclusions.

The focal questions are basically twofold. What is the uncertainty in national emissions inventories? What are the prospects for reducing this uncertainty and for being able to provide independent estimates with sufficiently small uncertainty to verify self-reported, national estimates?

The issue of verifying emissions estimates was elevated by the Bali Action Plan, produced at the 13th Conference of the Parties to the UNFCCC in December, 2007. The Bali Action Plan calls for addressing “Enhanced national/international action on mitigation of climate change, including, inter alia, consideration of: (i) Measurable, reportable, and verifiable nationally appropriate mitigation commitments or actions, including quantified emission limitation and reduction objectives, by all developed country Parties.”

The topic of verification was very pointedly broached by current U.S. President Barack Obama in a speech at the UNFCCC 15th Conference of the Parties (COP15) In Copenhagen, in December, 2009. President Obama said “we must have a mechanism to review whether we are keeping our commitments, and exchange this information in a transparent way. These measures need not be intrusive, or infringe on national sovereignty. They must, however, ensure that an accord is credible, and that we’re living up to our obligations. Without such accountability, any agreement would be empty words on a page.” The Copenhagen Accord that emerged from COP15 follows up: “Delivery of (emissions) reductions …by developed countries will be measured, reported and verified in accordance with existing and any further guidelines adopted by the Conference of the Parties, and will ensure that accounting of such targets…is rigorous, robust and transparent”. The Copenhagen Accord moves toward comprehensive global inventories in that it would add to current annual reporting of emissions by Annex I countries the additional requirement that non-Annex I countries report emissions inventories every 2 years.

Subsequent to Copenhagen the European Commission (2010) has supported building a “robust and transparent emissions and performance accounting framework” and acknowledged that “among the most difficult negotiations in Copenhagen were those on monitoring, reporting and verification (MRV). Transparency is key to ensure mutual trust and demonstrate the effectiveness and adequacy of targets and actions”. Zhang (2010) notes that “At Copenhagen, China eventually compromised to agree to open emission data to international consultation and analysis”.


179

We note that Alexander Zahar (2010) has reviewed early work of the expert review teams to see the extent to which they are able to discover problems in national inventories or if they reveal widespread efforts to distort emissions inventories in favor of the reporting countries. Zahar examined reviews of national emissions reported for 1990 – the baselines against which national commitments will be compared under the Kyoto Protocol. What Zahar observed is that of 37 Annex I countries reviewed, 34 emissions inventories were revised as a result of the review process. For 15 countries these revisions were very small (less than 0.4% for the total of all greenhouse gases). For the remaining countries, 1990 emissions were revised downward for 15 countries (with a maximum revision of 6.2%) and upward in 4 counties (with a maximum revision of 3.8%). In summary, the expert review teams generally found only small needed corrections and did not reveal widespread effort to overestimate the baseline values.

Within this context, the U.S. National Research Council initiated a study in December, 2008, and in March, 2010, released a report entitled: “Verifying Greenhouse Gas Emissions: Methods to Support International Climate Agreements”.

The NRC Report

The NRC report was prepared by a committee of 13 experts that was convened “to assess current capabilities for estimating and verifying greenhouse gas emissions and to identify ways to improve the capabilities” (page 1). Although the NRC report concluded, like others before it, that developed countries can estimate fossil-fuel-based CO2 emissions accurately enough to support monitoring of a climate treaty, it simultaneously concluded that the same is not true for other greenhouse gases and that current systems are not adequate to support monitoring if developing countries were to take on quantitative commitments. On the other hand, currently available methods are not sufficiently accurate to provide independent checks on any of these self-reported emissions estimates. The NRC addressed all long-lived greenhouse gases that are commonly included in international agreements but spent most of its time on the most important and most tractable greenhouse gas (CO2), which is responsible for nearly ¾ of anthropogenic greenhouse gas emissions (in terms of global warming potential) covered by the UNFCCC.

The committee suggested that strategic investments could improve national reporting and yield useful capability for independent verification within 5 years. It suggested that CO2 emissions from fossil fuels could be estimated by each country, and checked using independent information, with less than 10% uncertainty (2 sigma). “However, self-reported estimates of N2O, CH4, CFC, HFC, PFC, and SF6 emissions will continue to be relatively uncertain and we will have only a limited ability to check them with independent information.” Because of the dominant role of CO2, useful treaty support and verification may not require accurate measurement of all greenhouse gas emissions. The report suggested that it is a realistic near-term goal to reduce uncertainties of fossil-fuel CO2 emissions to 10% and to provide checks (especially for large, high-emitting countries) using independent methods that are equally accurate. But fundamental research is needed before other gases can be estimated at the national level with reasonable accuracy using independent methods. Useful simplifications are that almost 90% of global greenhouse gas emissions are in the energy and AFOLU (agriculture, forestry, and land use) sectors and 80% of emissions are from ¼ of the countries on Earth.


180

While recognizing shortcomings, the NRC report acknowledged that UNFCCC procedures “have been, and will likely continue to be, the primary means for monitoring and verifying greenhouse gas emissions under a new international climate treaty.” The principal shortcomings identified are that detailed emissions reports are not required of developing countries, independent data with which to check self-reported data are minimal, and the uncertainty in some emissions estimates is greater than the expected emissions reductions. It is difficult to compare inventories with physical measurements because the inventories do not provide complete accounting of all sources and sinks of greenhouse gases, the inventories do not have spatial and temporal resolution, and final inventory values are often not available for as much as 2 years after they occur. Especially in the land-use component there is a fundamental mismatch between the national inventories and what might be obtained from atmospheric measurements.

Verifying national inventories with independent data generally will rely on physical measurements of the atmosphere (from ground-based stations, towers, aircraft, balloons, spacecraft) and modeling that relates these measurements to the sources and sinks of greenhouse gases. In principle tracer-transport models could provide independent estimates of greenhouse gas emissions by country but the uncertainties are very large for a variety of reasons. The anthropogenic emissions of greenhouse gases are small compared to the large atmospheric background, modeling requires accurate representation of flows and mixing within the atmosphere, the most important anthropogenic greenhouse gas also have large natural sources and sinks, and our measurements in the atmosphere are from too few times and too few places.

The NRC report examined 3 categories of methods for estimating greenhouse gas emissions: national inventories, atmospheric and oceanic measurements and models, and land-use measurements and models. The report also had 3 broad categories of recommendations: strengthen national inventories; improve “the ability to independently and remotely estimate national, annual fossil-fuel CO2 emissions and to monitor emissions trends”; and develop better capability to estimate and check emissions and sinks related to land use. By focusing on fossil-fuel use and land use (in the later case including CH4 and N2O), the recommendations cover over 90% of global greenhouse gas emissions.

To strengthen national inventories the NRC committee recommended extending “regular, rigorous reporting and review” to developing countries, extending top-tier methods to the most important sources in each country, and implementing inventories at finer spatial and temporal resolution. More complete inventories globally and better spatial and temporal resolution would facilitate comparison with estimates derived from atmospheric measurements and modeling. The report recognized that external funding and training would often be required to strengthen developing country institutions and systems to build and retain sustained expertise.

To provide independent estimates of fossil-fuel CO2 emissions would require additional measurements in the atmosphere and improved tracer-transport modeling. The NRC committee suggested surface atmospheric monitoring stations near cities and large point sources, CO2 sensing satellites, and extensive measurements of 14C. They recommended a replacement for the Orbiting Carbon Observatory that failed on launch in February, 2009, national and international cooperation to extend the atmospheric sampling network to focus on large point and regional sources and


181

underrepresented regions globally, and extending the capability of measuring 14C. Estimates in the report suggest that 57% of U.S. CO2 emissions from fossil fuels come from 1.7% of the land area and that focusing on these more measureable sources could confirm trends in emissions changes. Expanded capability for measuring 14C would aid in separating fossil fuel and biospheric fluxes.

To develop better capability to estimate and check emissions and sinks related to land use is motivated by recognition that the agriculture, forestry, and other land-use sector is the second largest source of greenhouse gases (30%) but the largest source of uncertainty – with uncertainty at the national level often in the range of 50-100%. To support estimates and provide independent insight on sources and sinks related to land use, the NRC committee recommended a standing group to produce, from satellite imagery, at least every 2 years, global maps of land-use and land cover change. It also recommended a focused, interagency group to comprehensively review current information and to design a research program to improve methods for estimating CO2, CH4, and N2O emissions from agriculture, forestry, and land-use.

The NRC report concluded that improvements in emissions inventories and useful initiatives toward independent verification of the fulfillment of international commitments could be achieved within a 5-year time frame, given sufficient financial resources. Targeted, fundamental research would improve monitoring and verification of all greenhouse gas emissions. The results would yield substantial scientific gains in addition to improving confidence in international agreements.

References cited:

Breidenich, C., and D. Bodansky, 2009. Measurement, reporting and verification in a post-2012 climate agreement. A report prepared for the Pew Center on Global Climate Change, Arlington, Virginia, USA., 32 pp.

European Commission, 2010. International climate policy post-Copenhagen: Acting now to reinvigorate global action on climate change; Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, COM (2010) 86 final, 9 March, available at http://ec.europa.eu/environment/climat/pdf/com_2010_86.pdf.

National Research Council, 2010. Verifying Greenhouse Gas Emissions: Methods to support international climate agreements, The National Academies Press, Washington D.C., U.S.A.,

Obama, B., 2009. Remarks by the President at the Morning Plenary Session of the United Nations Climate Change Conference, December 18, 2009, The White House, Office of the Press Secretary, available at http://www.whitehouse.gov/the-press-office/remarks-president-morning-plenary-session-united-nations-climate-change-conference.

UNFCCC, 2008. Report of the Conference of the Parties on its thirteenth session, held in Bali from 3-15 December, 2007, FCCC/CP/2007/6/Add.1, 14 March 2008, available at unfccc.int/resources/2007/cop13/eng/06a01.pdf#page:3


182

U.S. GAO, 2010. The quality, comparability, and review of emissions inventories vary between developed and developing nations, U.S. Government Accountability Office GAO-10-818, 47 pp.

Zahar, A., 2010. Does self-interest skew state reporting of greenhouse gas emissions? A preliminary analysis based on the first verified emsiions estimates under the Kyoto Protocol. in press, Climate Law v.1, no.2.

Zhang, Z. X., 2010. Copenhagen and beyond: Reflections on China’s stance and responses, transcript of a speech presented at the International workshop on Climate change policies, Presidency of Complutense University, Madrid, Spain, February 18-19, 2010, version of March, 2010.


183

Multi-agent approach to simulation of the greenhouse gases emission permits market

Zbigniew Nahorski1, Jarosław Stańczak1, Piotr Pałka2

1Systems Research Institute, Polish Academy of Sciences, Newelska 6, 01-447 Warsaw, Poland [email protected]

2Warsaw University of Technology, Faculty of Electronics and Information Technology, Institute of Control and Computation Engineering, Nowowiejska 15/19, 00-665, Warsaw, Poland

Abstract

The paper presents the problem of simulation of the greenhouse gases emission permits market. The method of simulating the market is based on a specialized evolutionary algorithm (EA) or on a multi-agent system. Negotiation of prices between the parties with possibility to investigate influence of purchase/sale prices on the market are introduced. The trading parties can choose among two options: (i) a bilateral negotiation, and (ii) a tender; but only the first option is presented in this version of the paper. They have also several strategies to choose among.

Keywords: greenhouse gases, emission trading, market simulation, multi-agent systems, evolutionary programming

1. Introduction

The market behavior is usually analyzed either using a static optimization model or a game-theoretic approach. In both cases it is assumed that an appropriate information is available to the parties acting on the market. Recently, agent-based models are often used to investigate the dynamics of the market behavior using a simulation approach. The parties are represented by intelligent programming agents who negotiate and trade the goods according to given market rules and the information available to the agents. This paper uses this approach, focusing on the greenhouse gases emission permits market.

Thus, the method proposed in this paper does not assume an ideal market. Neither the equilibrium prices are assumed in the trading, nor the dynamic trading without taking prices into account, like in [3, 5], are considered. A more sophisticated, dynamic market model is introduced, with possibility of price negotiation and the influence of real prices on the agent behavior (similar assumptions can be found in [2]). The number of transactions between the start of the market and equilibrium is not known in advance. Only transactions profitable for both participants are accepted during simulations. Each transaction that is profitable for both parties moves the market toward equilibrium.

Transaction can be simulated using a specialized, multi-criteria evolutionary algorithm or a multi-agent platform for multicommodity exchange [11]. This gives a possibility to trace activities of agents participating in the market and to build more realistic model of conducted transactions. In particular, intelligent agents are considered and trading strategies of the agents are analyzed and optimized. Each agent minimizes its own objective function, which is the costs including spendings on emission reduction plus expenditures for the permits. This approach takes into account the purchase/sale price of permits, which considerably influences the profitability of transactions and


184

the decision to buy/sell permits, i.e. whether it is better to reduce emissions or to buy permits. Obtained so far results of simulations are close to the equilibrium point calculated for the problem where the common costs for all trading agents are minimized.

In this workshop version of the paper we focus on the bilateral trade. Other trading mechanisms, like tender or auctions will be considered in more details in further versions. The uncertainty in prices is taken care of in the paper and the method to solve it is the trade negotiations. We leave for future consideration emission uncertainty and focus here more on the simulation methods. The emission uncertainty can be, however, easily taken into account using the effective permits, as proposed in [15] and used already in simulation of trade in [19].

The contents of the paper is as follows. Section 2 motivates and presents the rules of the considered market. In Section 3 the multi-agent system used for simulating the market is described. Selection of trading strategies by using an evolutionary algorithm is addressed in Section 4. Section 5 contains results of simulation, and Section 6 concludes.

2. Market

Let us consider a market with N parties trading the emission permits. Each party has been pre-allocated nK permits, Nn ,1,= K , usually called "the Kyoto targets".

At the compliance time the party must not emit more than the number of permits it possesses. However, it may freely sell or buy permits to achieve the target. We denote by

nx the emission of the n th party and by ny the traded permits. Emission must be

nonnegative, 0≥nx , but number of traded permits ny may be positive, when bought, or

negative, when sold. To achieve reduction, the following condition at the starting time has to be satisfied

n

N

n

sn

N

nKx ∑∑

1=1=> .

It is common to refer to percent emission reduction, i.e. to present the reduction as snnn xK )(1= δ− , where nδ100 is the percent of reduced emission. In this paper we use,

however, the absolute reductions. The central planner perceives the use of emission trading system as minimization of

the common cost function

),,(=),(=)( 11=

Nnn

N

nKKKKcKF K∑ , (1)

when )( nn Kc is the cost born by the n th party to reduce the emission to the level nK ,

and subject to

01=

= KKn

N

n∑ ,

where 0K is the total allowable emission of market participants. Using the Lagrange method, the solution is

01=

=;,1,=),(= KKNnKc n

N

nnn ∑′−λ K . (2)


185

Thus, at the optimal emissions the marginal costs of all parties are equal and λ is an optimal price.

The n th party is looking to minimize its cost of reducing the emission and buying/selling the permits ny , i.e. to minimize the function

nnnnnn yxcyxf λ+)(=),( (3) subject to

nnn yKx +≤ .

Above, 0≥λ is the price of one unit of permits. For known functions )( nn xc for all n , the problem (1) may be solved and

the optimal price λ can be found. This would enable the parties to solve the problem (3). However, the functions )( nn xc are not known to the central planner. Moreover, the trade is continuing in time, as some parties negotiate the transactions earlier and some later, not knowing the optimal price λ . Just, they have to cope with the uncertainty in earning/loosing money in the trade.

In [5], an approach to simulate the bilateral exchange of permits between parties has been proposed, called the dynamic bilateral trading. Its idea is that two parties meet randomly and sell/buy permits, if it is feasible for both parties. This happens when the marginal costs of the parties differ, i.e. the first condition in (2) is not satisfied. It is a simple observation that each such transaction makes the cost function (1) smaller, see [3]. As the function (1) is constrained from below by 0 , and the sequence of the values of the function (1) is decreasing with each transaction, then it converges, although not necessary to the global minimum of (1). In the original paper [4] it is assumed that the number of exchanged permits goes to zero (albeit not too quickly) and then it is proved that the sequence converges to the global minimum with probability 1. Thus, the approach actually uses a stochastic optimization method. That approach does not take into account the prices of the permits. Neither it is considered, if the parties really need to buy or sell the permits.

A modification of this problem, in which the prices are allowed for, has been presented in [19]. As before, only feasible transactions are considered. But the price π of the transaction is drawn at random from the interval constrained by the marginal costs of the trading parties. The number of trading permits is drawn randomly as well. It is considered that the parties look during the trading at the financial result in each transaction, viz. the difference between the gain from selling the permits and costs of reducing the emissions. This can be written as

)]()([= 11 −− −+−π tnn

tn

tnn

tn

tn

tn xcyxcyg .

Let us consider a transaction between parties n and m . Assume that the marginal costs

of both parties differ and that it holds )(=>)(= 11 −− ′−λ′−λ tmm

tm

tnn

tn xcxc . Then, the m th

party can sell permits to the n th one (the transaction is feasible). Denoting by 0>tnmy

the number of permits traded, and following the argumentation from [5], the sum of the results of both parties in the transaction is

)]()([)()(== 1111 tnm

tnm

tnm

tnn

tnm

tnn

tm

tn

tnm yxcyxcxcxcggg −++−++ −−−−

and therefore, it is the difference of the reduction costs of both parties before and after transaction. It can be also written as


186

−−+−+−−−−−

tnm

tnm

tnm

tnm

tnm

tnn

tnm

tnnt

nmtnm

y

yxcxc

y

xcyxcyg

)()()()(=

1111

and for small tnmy approximately as

0>)(=))()(( 11 tm

tn

tnm

tmm

tnn

tnm

tnm yxcxcyg λ−λ′+′−≈ −− .

Thus, any feasible transaction increases the value of the common cost function (1), at least for small enough number of traded permits.

Denoting now by ),,,,(= 21

111

iTNNi yyyyy KK all lots of emissions traded until

the i th iteration of the algorithm, see explanation of iterations below, the following problem is considered in [19]

tnnm

n

iT

tii gmaxyG )(

1==)( ∑ (4)

subject to

0=1=1=

tn

N

n

iT

ty∑∑ ;

maxtn yy ≤≤0 ;

nntn

tn

tn Kxyxx =,= 01 +− ;

)(0

)(1

)()( =,= nmnmtn

tnm

tnm Kxyxx −− ;

iTt ,1,= K ;

iTn

iT

tn

iTn yKx ∑+≤

1=;

iTn

iT

tnm

iTnm yKx ∑+≤

1=)()( ;

Nnnnm|Nnm ,1,=},)(,{1,)( KK ≠∈ ; Ii ,1,= K .

Above, )(nm is the party bound by transaction to the n th party. Let us notice that

for each t only one transaction is chosen; the one, for which tnnmg )( is the greatest.

It means that all values of tny and tnπ except the winning ones are lost. The variable

tnnmg )( depends on the random choice of t

ny and tnπ . Thus, the choice of a winner is

a stochastic one, but those cases where the difference is greater, have greater probability to win.

A party can choose one of two strategies to find a good partner for winning the above competition. One is to choose a partner randomly, like in [5]. The second strategy is a tender. A randomly drawn number of permits is offered for sale and that

feasible partner is chosen, for whom the value tnnmg )( is the greatest, for randomly drawn

prices. If it is feasible, the number of permits and the price are drawn randomly and no


187

further screening is done. These strategies are drawn randomly with probabilities assigned each time by an evolutionary algorithm, as described in more details in the sequel. The iterations i are connected with consecutive loops of the evolutionary algorithm.

Let us also notice that the algorithm does not start from the emissions snx but nK . This is connected with the fact that actually the algorithm does not simulate the real trade but maximizes iteratively the common function (1) and is, up to some details, a subversion of a more general algorithm from [5]. Both algorithms should therefore end with the same solution, if it is unique. However, in the algorithm from [5], the costs of optimal solution for each party is calculated using the equilibrium price λ . In [19], the costs are sums of incomes from all transactions.

In this paper we follow a general idea of the above algorithm. However, only local goal functions are considered and more advanced methods of negotiating prices are applied. Each party, or agent in the multi-agent system nomenclature, maximizes only his

own income tng .

A party initiating a consecutive transaction is drawn randomly. The number of traded permits is drawn randomly, as before. However, in this paper, in the bilateral trade the price of a possible transaction is negotiated with the prospective partner. There are several negotiation strategies to be chosen by a party and the probability of the choice evolves according to the evolutionary algorithm. One of the simpler is the linear bidding when each negotiating party steps off a constant sum from its price in consequent biddings until both parties meet.

The formal definition of the market is as follows. The randomly drawn n th party

enters the trade and or randomly drawn tny , tries to maximize its quality function

tn

tn

gmaxπ

(5)

by negotiating transactions with chosen randomly parties, among nmNm ≠∈ },,{1,K

nntn

tn

tn Kxyxx =,= 01 +− ,

NnyKx n

t

ntn ,1,=,

1=K

τ

τ∑+≤ ,

Tt ,1,= K . A limitation on the number of traded permits is also introduced.

3. Multi-agent system

As set forth in [20], multi-agent system is a system composed of two or more autonomous software agents communicating with each other and striving for their own purposes. Such a system should achieve some overarching objectives and should operate in accordance with intentions of the system designer. Nevertheless, the system does not implement these objectives directly, but through individual objectives of each of the agents and their interactions. Multi-agent systems are considered as a tool for simulation of market processes in [18, 20].

Each agent represents single party, which is guided by its own interests. The individual party comes to interact with others, motivated by the desire to achieve


188

certain gains from the exchange of permits, i.e. to reduce the costs including trading plus expenditures for the permits. The overarching goals of the system can be identified with the objectives of the central planner, i.e. to minimize the common cost -- the sum of costs to reduce the emission limits. However, these goals will not always be met.

The implementation of the multi-agent platform for the multi-commodity exchange [17] is used to simulate the permission trading. In this implementation we use the standards which enable agents to co-exist and interact. Such a standards (agents communicative language) are defined by FIPA organization [8]. Moreover, to meet the needs of the wide understood multi-commodity exchange, we use the M3 model [13] to describe offers. The platform is implemented in Java language, using the JADE (Java Agent DEvelopment) and JAXB (Java Architecture for XML Binding) frameworks. More detailed description of the design and implementation of the multi-agent platform for multi-commodity exchange can be found in [17].

A particular agent, to be able to perform trading processes, needs to collect some specific information about other agents, and particularly about the current trade process. We can certainly imagine a case in which each agent broadcasts the messages about initiating of some new trade to all other agents. However, in large systems, such a solution would involve a massive communication effort. Moreover, agents would obtain a huge amount of unnecessary information.

As often there is no need for the central entity to manage the trade process in the multi-agent environment, there is a clear need there to share information. For this purpose we define a special agent called Morris Column agent [17]. The task of such agent is to offer some public location, where other agents may ``hang a notice'' to report certain information about trade processes provided by them. In addition, the Morris Column agent should provide a functionality of searching and removing the information.

We assume existence of a number of negotiating agents and of only one Morris Column agent. In this paper we consider mainly the bilateral trading mechanism. Each agent can choose to be a passive negotiator, i.e. it submits to the Morris Column its willingness to negotiate and waits for the negotiation partner, or an agent can choose to be an active negotiator, i.e. it searches for a negotiating partner by querying the Morris Column. We assume that agents choose to be passive or active randomly. Agents negotiate in randomly established pairs and conduct bilateral contract depending on the expected profit.

Also the tender mechanism is considered, although no comprehensive simulations were performed yet. For it, we assume that one agent plays a role of an operator, and the rest of agents plays a bidder roles. The selection of the operator from a number of agents in distributed environment is a problem. A solution can be to apply the bully election algorithm [14] for selecting the operator from a group of agents. Thus, the operator is chosen randomly, rest of agents submits bids for selling or buying a number of permits, at the price specified by the bidder. The operator chooses the most profitable bid, taking into account its preferences. Afterwards, the new operator is chosen randomly and the process is repeated iteratively.

4. Evolutionary method applied in computer simulations

The evolutionary-based method is one of two approaches described in this paper and used to simulate the greenhouse gases permits market. The method presented is not using an usual basic evolutionary algorithm but rather a specialized one with autonomic agents with their own optimized quality functions (multi-criteria approach).


189

Evolutionary algorithms (EA) are based on the phenomenon of natural evolution. Like in the nature, they operate on a population of individuals (also agents, members or solutions), which are reproduced, modified, evaluated and the best of which are selected in a sequence of generations. Individuals are encoded solutions of the problem to be solved. They are modified by evolutionary operators at each step of the algorithm, emulating random changes (mutation) and recombination of parent genes in the genome of natural organisms. The best ones are selected as members of the next generation using some selection method that mimic the natural selection, to promote the best individuals. A basic evolutionary algorithm works in the manner shown in Algorithm 1, but this simple scheme requires many problem specific improvements to work efficiently. Two types of algorithms are discussed in the sequel.

Evolutionary algorithm

1. Initialization of the population of solutions. 2. Reproduction and modification of solutions using specialized

operators. 3. Evaluation of the solutions obtained. 4. Selection of individuals for the next generation. 5. If a stop condition not satisfied, go to step 2.

Algorithm 1. Evolutionary algorithm. In this approach, an individual of the EA population contains information about all the countries participating in the market. Thus, it forms a complete solution of the problem. It is possible to create as many individuals as necessary and obtain this way many basically different solutions. However, this basic scheme has an important shortcoming. All participants of the market are valued using one criterion or quality function and it is impossible to develop their individual abilities, strategies and gains.

Agent-based evolutionary algorithm can be treated as a specialized, developed version of the evolutionary algorithm. The scheme of its operation is presented in Algorithm 2. In this method each trading party constitutes a separate agent and the population of solutions in the basic evolutionary algorithm is narrowed to the number of parties participating in the trade. Each member of the population has its own quality function and tries to optimize it, contrary to the standard EA method, where there is one function value for all market participants. This approach allows parties to optimize their actions independently and obtain results for their actions, instead only averaging the results of all parties. Although the equilibrium point obtained in both methods is almost the same, the results obtained by parties may be quite different.

Agent-based evolutionary algorithm

1. Initialization of agents. 2. Modification of agent states using specialized

operators. 3. Evaluation of new agent states obtained. 4. If a stop condition not satisfied, go to step 2.


190

Algorithm 2. Agent-based evolutionary algorithm. Now, each individual in the agent evolutionary method contains information to be used in market simulation. The information needed to describe all actions of agent/population members is as follows and is encoded as a vector of eight numbers:

1. the marginal cost associated with a given number of permits possessed by the country (shadow price);

2. the real current price of a permit for sale/purchase; 3. the real current value of a permit for sale/purchase; 4. the current number of units for sale/purchase; 5. the net number of units sold/purchased ; 6. the current emissions level; 7. the previous emissions level (before the present transaction); 8. the present and previous values of the objective function.

To modify solutions, the following specialized operators were used: 1. bilateral sale -- two randomly chosen countries conduct price negotiations1 and if

they agree, the solution is modified; 2. tender -- the country considered offers a number of permits for sale, other

countries submit offers to buy2, the best submitted offer is chosen by the calling party and then the states of the winner and seller are modified; as before, this mechanism has been used in a limited scale.

The genetic operators are executed randomly during simulations, although agents can conduct ranking of profits gained using the operators and try to take part in more profitable ones more frequently.

5. Simulation results

The simulation was carried out on the case study described in [9]. The case study includes five parties: USA, EU, Japan, CANZ (Canada, Australia, New Zealand) and EEFSU (East European and Former Soviet Union). We assume that these parties comply with the Kyoto Protocol regulations, so they should reduce CO2 equivalent emissions.

The parties have been specified emission limits nK , BAU (business as usual) emissions

nx0, , and the functions of cost reduction nc (index n defines the party). As set forth in

[9], the cost of reduction is a square function of the size of the reduced emissions

≥−

nn

nnnnnn

xxfor

xxforxxac

0,

0,2

0,

0

<)(= (6)

The marginal cost of the emission permit (shadow price) np is the derivative of

the function nc

≥−

nn

nnnnnn xxformin

xxforxxap

0,

0,0, <)(2= (7)

1In the bilateral trade, parties randomly bid their start prices based on their earlier experiences and try to converge with a constant step to an acceptable price. The numbers of traded units are chosen randomly from a given interval. 2Several strategies of price bidding by potential buyers can be applied.


191

The variable nx stands for the current emission. In the beginning of the simulation,

we assume that each party starts from its Kyoto targets: nn Kx =0 . The "min " means the minimum price value, and it was introduced to the calculations for more realistic conditions for the prices calculation. The data for the case study are presented in Table 1.

Table 1: Data for the case study -- parameters of function of cost reduction.

Party BAU emission Cost function parameter Kyoto limit nx0, [MtC/y] na [MUSD/(MtC/y) 2 ] nK [MtC/y]

USA 1 820.3 0.2755 1251 EU 1 038.0 0.9065 860 Japan 350.0 2.4665 258 CANZ 312.7 1.1080 215 EEFSU 898.6 0.7845 1314

The simulations based on the specialized evolutionary algorithm (EA) are presented

in Table 2 and those on a multi-agent platform for the multi-commodity exchange in Table 3.

Table 2: Results of simulation using evolutionary method, bilateral trade.

Final Last Corresponding No. of traded Permission Reduction Party emission transaction marginal price permissions cost cost price (shadow price) [MtC/y] [USD/tC] [USD/tC] [MtC/y] [MUSD/y] [MUSD/y] USA 1 556.7 143.58 145.24 305.7 48 472.41 19 147.46 EU 960.2 129.52 141.05 100.2 14 025.68 5 487.95 Japan 321.2 123.81 142.07 63.2 10 539.82 2 049.17 CANZ 249.0 142.30 141.16 34.0 2 001.36 4 497.47 EEFSU 811.9 141.68 137.60 -503.1 -75 039.30 6 040.79

Table 3: Results of simulation using multi-agent system, bilateral trade.

Final Last Corresponding No. of traded Permission Reduction Party emission transaction marginal price permissions cost cost

price (shadow price) [MtC/y] [USD/tC] [USD/tC] [MtC/y] [MUSD/y] [MUSD/y] USA 1 559.83 142.09 143.52 308.83 59 794.97 18 692.00 EU 959.45 142.47 142.41 99.45 20 837.18 5 593.85 Japan 321.62 138.89 140.00 63.62 13 736.11 1 987.87 CANZ 248.44 141.91 142.40 33.44 3 768.45 4 575.90 EEFSU 808.66 141.72 141.12 -505.34 -98 136.71 6 346.66

As can be noticed, the results obtained using two presented methods are quite

similar. Values of final emissions, numbers of traded permissions, final shadow prices and the reduction costs are almost identical, bigger differences can be seen in values of permission costs. This fact is probably caused by different mechanisms of reaching contracts in two considered methods. The multi-agent platform is in general closer to real market, thus this results can be more realistic. Moreover, the results obtained using both


192

methods are very similar to the results obtained for the centralized market for the greenhouse gases permits [1].

There are also several more factors that can influence the costs of reached contracts. The most important one is the strategy of setting price for bids in negotiations. This problem arises especially in the case of EEFSU. This party has the level of emission lower than the Kyoto target and its reduction costs are equal to 0 for the most of transactions. Of course, this party does not want to sell permits for free but it is difficult to set acceptable price level for transactions of this party -- theoretically any positive price is profitable.

Figure 1: Trajectory of consecutive contracts unit prices during single simulation,

using multi-agent platform and bilateral negotiations.

6. Conclusions

Multi-agent or evolutionary methods are often applied in market simulations, mainly due to the fact that such methods are able to deal with complicated systems with many interactions between their elements and participants. Economic systems can be quite easily modeled, simulated and controlled using these kind of artificial intelligence (AI) systems. The multi-agent and evolutionary based approaches presented in this paper seem to be a suitable tool for analyzing economic phenomena, especially for dynamic market models with elements of uncertainty in prices (negotiated prices, tenders).

This paper concentrates on presenting possibility of using some AI tools for simulating the trade. Simulation results presented in this paper are preliminary ones. More elaborated results will be presented during the Workshop and published in next publications.

Acknowledgments

Partial financial support from the Polish State Scientific Research Committee within the grants N N514 044438 and N N516 375736 is gratefully acknowledged.


193

References

[1] Bartoszczuk P.,Horabik J. (2007) Tradable Permit Systems: Considering Uncertainty in Emission Estimates, Water Air Soil Pollution: Focus, 7:573-579.

[2] Bonatti M., Ermoliev Y., Gaivoronski A. (1998) Modeling of multi-agent market systems in the presence of uncertainty: The case of information economy. Robotics and Autonomous Systems, 24(3-4):93-113.

[3] Ermolieva T., Ermoliev Y., Fischer G., Jonas M., Makowski M. (2010) Cost effective and environmentally safe emission trading under uncertainty. In: K. Marti, Y. Ermoliev, M. Makowski (Eds.) Coping with Uncertainty. LNEMS 663. Springer Verlag, 79-99.

[4] Ermoliev Y., Klaasen G., Nentjes A. (1996) Adaptive cost-effective ambient charges under incomplete information. Journal of Environmental Economics and Management, 31:37-48.

[5] Ermoliev Y., Michalevich M., Nentjes A. (2000) Markets for tradeable emission and ambient permits: A dynamic approach. Environmental and Research Economics, 15:39-56.

[6] EBR (2006) Volatility the only certainty in EU carbon market. Energy Business Review. http://www.energy-business-review.com/article_feature.asp?guideFD09D7CA-3EFC-4229-BA86-ID968025DF5B.

[7] EEA (2006) Application of the emission trading directive by EU member states. Tech. Rep. 2. European Environment Agency, Denmark. http://reports.eea.europa.eu/technical_report_2006_2/en /technicalreport_2_2006.pdf.

[8] Foundation for Intelligent Physical Agents. http://www.fipa.org/.

[9] Horabik J. (2005) On the costs of reducing GHG emissions and its underlying uncertainties in the context of carbon trading. Report no. RB/34/2005, IBS PAN.

[10] IETA (2005) Emission trading master agreement for the EU scheme. In: Version 2.1 as of 13 June. International Emission Trading Association (IETA), Geneva, Switzerland. http://www.ieta.org/ieta/www/pages/getfile.php?docID=1001.

[11] Kaleta M., Pałka P., Toczyłowski E., Traczyk T. (2009) Electronic Trading on Electricity Markets within a Multi-Agent Framework. LNAI 5796 Computational Collective Intelligence. Semantic Web, Social Networks and Multiagent Systems. Springer Verlag, 788-799.

[12] Lopes F., Woolridge M., Novais A. Q. (2008) Negotiation among autonomous computational agents: principles, analysis and challenges. Artificial Intelligence Review, 29:1-44.

[13] M 3 -- Multicommodity Market Data Model. http://www.openm3.org/.

[14] Mamun Q.E.K., Masum S.M., Mustafa M.A.R. (2004) Modified bully algorithm for electing coordinator in distributed systems. WSEAS Transactions on Computers. 3(4):948-953.

[15] Nahorski Z., Horabik J., Jonas M. (2007), Compliance and emission trading under the Kyoto Protocol: Rules for uncertain inventories. Water, Air & Soil Pollution: Focus, 7(4-5):539-558.


194

[16] Ó Broin P., O'Riordan C. (2007) An evolutionary approach to deception in multi-agent systems. Artificial Intelligence Review, 27:257-271.

[17] Pałka P., Całka M., Kaleta M., Toczyłowski E., Traczyk T. (2010) Design and Java implementation of the multi-agent platform for multi-commodity exchange. III Krajowa Konferencja Naukowa Technologie Przetwarzania Danych, Wydawnictwa Naukowo-Techniczne, Warsaw, 184-196.

[18] Shoham Y., Leyton-Brown K. (2008) Multiagent Systems Algorithmic, Game-Theoretic, and Logical Foundations Cambridge University Press.

[19] Stańczak J., Bartoszczuk P.(2010) CO2 emission trading model with trading prices. In: T. White, M. Jonas, Z. Nahorski, S. Nilsson (Eds.) Benefits of dealing with uncertainty in greenhouse gas inventories. Special issue of Climatic Change, DOI: 10.1007/s10584-010-9905-7.

[20] Woolridge M. (2001) Introduction to multiagent systems. John Wiley & Sons.

[21] Yeh Ch.-H. (2007) The role of intelligence in time series properties. Computational Economics, 30:95-123.


195

Accounting for uncertainties and time preference in economic analysis of tackling climate change through forestry

Maria Nijnik1 and Guillaume Pajot2

1Socio-Economic Research Group, The Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen, AB15 8QH, United Kingdom

[email protected] 2Gretha, Université Montesquieu-Bordeaux 4 avenue Léon Duguit 33 608 Pessac, France

Abstract

The paper discusses the development of economic techniques for dealing with uncertainties in cost-benefit analysis (CBA) via the examples of planting trees to mitigate climate change. In consideration of uncertainty, time preference and intergenerational equity, the traditional CBA framework has been challenged with regard to the question of discounting/non-discounting of carbon uptake benefits, and also because it usually uses a constant and positive discount rate. We investigate the influence of various discounting protocols on a CBA of carbon sequestration in forestry. In this paper, the idea of declining discount rates is also considered. A numerical example dealing with tree plantations in the United Kingdom (UK) is provided. We show that discounting protocols have a considerable influence on the results of economic analysis, and therefore, on the decision-making processes related to climate change mitigation (CCM) strategies. The paper concludes with a few climate policy implications of dealing with uncertainties, and a brief discussion of what different discount rates imply.

1. Introduction

An important social function of forests that receives attention in current research is their climate change mitigation role [1]. There are several ways via which forestry could contribute to CCM: tree plantations, increasing carbon density (e.g. low thinning or long rotations); increasing carbon storage in soils; increasing carbon storage in wood products [1]; and forestry-based renewable energy scenarios. Numerous studies over the last decade have addressed the cost-effectiveness of forestry as a carbon sink [2]; [3]; [4]; [5]; [6]; [7]; [8]; [9]; [10]. Van Kooten et al. [11] carried out a meta-analysis of 68 studies, with a total of 1047 observations worldwide. Baseline estimates of the costs of sequestering carbon through forest conservation (based on analysis of 981 estimates from 55 studies of the costs of creating carbon offsets using forestry) range from US$46.62–US$260.29/tC ($12.71–$70.99/tCO2).

1 Tree planting and agroforestry activities increase costs by more than 200%. The studies identified substantial variability in marginal costs in different countries and in different settings. Although such variation is partly due to the diversity of methods and assumptions used, it indicates that terrestrial carbon sequestration (CS) involves a great deal of uncertainty.

The debates around uncertainties and time preference in economic analysis of policy alternatives related to natural resource use have a long history. In forestry, many effects are long-lived, and growing forests provide some of their benefits far into the future. Manifold uncertainties, therefore, relate to future demand and supply of forest resources 1 The conversion factor from CO2e to C is 3.67 (i.e. 44/12).


196

and ecosystem services; to their stock, flow and reproduction. Difficulties in estimating the future benefits of carbon sink are associated with uncertainties about the dynamics of carbon. In establishing tree plantations for CS, in addition to uncertainties associated with ecological, future technological, economic, environmental, and social aspects of CCM through forestry, there is a great deal of uncertainty associated with property rights, institutional settings and managerial aspects in forestry. Changes in policies, markets and social norms also contribute to uncertainties. Cost-benefit analysis of planting trees to mitigate climate change, therefore, runs into uncertainties, and the extent to which mitigation strategies can be justified on efficiency grounds largely depends on the discount rate employed in the analysis.

At the level of society, discounting has two basic sources. The first source represents social time preference: we have to discount the future because of diminishing marginal utility of consumption. The second source is the productivity of capital or its social opportunity cost; meaning that if certain resources are invested instead of being consumed, then these resources would provide in future a higher level of consumption than if they were consumed now [12]. With perfectly functioning markets, the social rate of time preference would equal the social opportunity cost of capital, since there is no logic in undertaking the investment unless future benefits offset the social rate of time preference. In the actual world, however, the situation is quite different and a choice has to be made. There are many justifications to discounting. In addition to time preference or as a measure of alternative investments, a positive inflation rate and uncertainty over future earnings could be mentioned.

In forest economics, justification of discounting above all relies on the uncertainty of upcoming events and their outcomes. Since the economy is affected by random shocks, the uncertainty about the growth of income induces people to invest more for the future. This precautionary effect provides an argument for reducing the discount rate [13]. The concept of intergenerational equity, Clark's concern about the risk of depletion of natural resources [14] and Ramsey’s [15] idea of ethical indefensibility provide arguments in support of non-discounting (or low discount rates) in forestry. However, financial returns in the forestry sector are commonly low, whereas investment risks are high due to long rotation ages of timber, possible forest fires and often insecure tenure rights etc. Therefore, in consideration of decision-making on long-term investments under risks and uncertainties, in order to secure income generation sooner rather than later, support is usually given to the use of a positive (even rather high) value of discount rates in private forestry [16].

Particular difficulties arise with the choice of discount rates in climate change related CBA. Firstly, this is because marginal damages from atmospheric carbon are not certain over time. There is also uncertainty about the benefits of carbon control strategies for future generations. Hanley and Spash [17] argue that even if the preferences of upcoming generations for CCM strategies were to be taken into account, it would be impossible to accurately reflect these, because future preferences are unknown. Uncertainty in climate change economics is exceptionally high. There is no distinct way to adjust the discount rate for risk in the present value of uncertain future benefits and losses.

On the one hand, there is a probability that marginal damages will not worsen in the long run. Such a probability might increase over time, provided there is continuous technological advance. We may also count on the increasing stock of knowledge; the development of human (and man-made) capital that enables future generations to solve problems which cannot be solved today. On the other hand, however, even if


197

the most considerable damages/losses caused by climate change are imminent in the far future, discounting at a positive rate assigns them insignificant present values (PV), ultimately advocating little immediate action for their alleviation.

Nordhaus [18] argues that the efficient degree of control of carbon concentrations would be deemed zero in the case of high costs, low damages and high discounting, and vice versa. Often, the costs of mitigation strategies are high and if we believe that future technology will enable carbon emission reductions at almost no cost, an infinite discount rate could be suggested. The opposite assumption results in the reverse conclusions. The more rapidly CO2 concentrations are projected to be increasing in time, the less future carbon benefits should be discounted. Overall, largely because of uncertainties, an important question that arises while employing CBA of planting trees to mitigate climate change is that of carbon discounting. We investigate the influence of various discount settings on the economics of carbon sequestration in forestry.

The paper first considers the establishment of forest plantations, for which storage, energy and wood product CCM scenarios are investigated. The storage option presumes planting and growing of trees for a period of 40 years without considering future use of wood and land after timber harvesting [3]. However, more often than not, trees are cut after a certain time horizon. The additional timber to be received from the forest would enlarge the supply of wood for the industry. Moreover carbon stored in wood products is an addition to the total carbon sink. The energy policy option suggests using extra wood received from the plantations as a substitute for fossil fuels. This policy also contributes to carbon sinks provided by the forests and to the reduction of carbon dioxide concentrations in the atmosphere. In this research, we firstly exercise different discount rates, examining their impacts on policy decisions under the above mentioned scenarios by taking the case of Ukraine as an example. Then, as the traditional framework of CBA has been challenged [19] [20], and also because it uses a constant and positive discount rate [21] [22] which reduces the weight of future benefits/costs, we estimate the costs of carbon sequestration strategies under various discounting protocols, including the declining discount setting. More recently, concerns over intergenerational equity have raised scope for using declining discount rates [23], [24], [25], [26], [27]. This takes the case of a tree plantation in the UK as an example. We discuss the implications of these findings, and conclude that the choice of a discounting protocol exerts considerable influence on decision making processes.

2. Influence of different discount settings on the results of economic analysis of carbon sequestration policy implications for Ukraine

Prospects for the creation of forest plantations to sequester carbon in Ukraine were elaborated. Fast-growing tree-species across forestry zones (on land defined as suitable for afforestation) were considered for planting: hybrid poplar in the Wooded Steppe zone; mixed species scenario (30% hybrid poplar, 20% alder and other hardwoods and 20% pine and other softwoods) in the Polissja zone; pine in the Steppe and in the Crimea; and spruce in the Carpathian Mountains (for more information, see [5], [21] and [22]). A discount setting of 4% for costs was employed in the research for Ukraine.

The assorted tree species across the forestry zones possess particular functional forms of tree growth. Growth functions were estimated by using the State Committee of Forestry data. Benefits of carbon uptake were simulated in physical terms of carbon sequestration that depends on tree growth of the chosen species. Then, the total


198

discounted carbon per ha for the above ground biomass across the analysed forestry zones was computed as follows [28]:

( ) ( )rt

t rtrs

e

etVdsesV

Cg −

−−⋅

−

−ηφ

=∫

1

0, (1)

where: t – is rotation age; s – is growing age; V – is volume of stem wood, m3 ; r – is rate of growth.

The first term in parentheses is for the discounted carbon that accumulates during the growth of trees, and the second term stands for the carbon released to another account during timber harvesting. By dividing by 1-e-rt, we obtain the sum of the infinite series of "return" that accrues every t years of timber rotation. While considering a storage option, only the first term in parentheses is pertinent.

η (=1.51 – for coniferous and =1.69 – for softwood deciduous stands, as poplar and mixed species) is a coefficient that translates bole biomass into total above ground biomass;

φ (=0.32, for coniferous, and = 0.34, for softwood deciduous stands) is a coefficient that converts tree growth into carbon [29].

Carbon sequestered by the root pool was estimated in m3 per ha, according to the following relationship [28]:

R(G) = 1.4319 G 0.639, (2)

where: R – is root biomass, m3; G – is above ground biomass, m3. For spruce, it is the constant 0.2317 that relates root biomass to above ground

biomass, while for pine, the understory biomass is related to the growing stock [29], as:

R (V) = 0.146 (V)-0.519 (3)

The total discounted Cr per ha for the root account was computed, as in Van Kooten and Bulte, [28]:

( )∫−•

φ=t

rsr dseGRC

0

, (4)

where: s – is growing age; r – is rate of growth. The economic analysis of establishing forest plantations was conducted under three

discount settings for carbon removals. The first presumes no discounting of physical carbon. By not discounting carbon uptake benefits, we suggest that the value of marginal carbon damages in the future will increase at the real rate of discount. This assumption implies that all the carbon sequestered is valued equally, no matter when it is captured. This would suggest that if the afforestation costs are equal for different stands (e.g. for poplar that is fast-growing, with initially high rates of CS, but for which growth soon decelerates, and for spruce that grows much more slowly but for longer, and at the age of 100 years accumulates up to 500 tonnes of carbon per hectare), the planting of spruce would be preferred.

However, from an economic perspective, and given uncertainty and risk in long-term projections, the setting of a 0% discount rate for carbon benefits is a very specific


199

assumption. Therefore, to justify the efficiency of carbon mitigation policies economically, positive discount rates of 2% and 4% for carbon uptake have been employed in this study. Discounting of carbon uptake benefits at a social rate of discount (4%) presumes that marginal damages from emissions are constant over time. This is a reasonable assumption since we have no evidence to predict any other scenario. Moreover, as a 4% discount rate was applied to the discounting of costs in this study, it is logical to use this rate also for the discounting of benefits. Under these assumptions, cumulative carbon uptake was simulated. The discounted carbon per ha, across forestry zones, for the storage policy option, under different discount rates assigned for carbon uptake benefits, is shown in Table 1.

Table 1. Simulated carbon sequestered per ha, under the storage scenario

Carbon (t C ha-1) Polissja Wooded Steppe Steppe Carpathians Crimea

0% 203.2 236.3 37.3 178.6 99.6

2% 91.29 106.19 16.76 80.26 44.75

4% 41.0 47.7 7.5 36.0 20.1

The highest estimates of carbon uptake over 40 years are for hybrid poplar in the Wooded Steppe, and for the mixtures that include poplar in the Polissja zone. The resulting estimates largely depend on the discount rates employed in the analysis. The figures produced for Ukraine are comparable with the corresponding estimations provided for Canada [28], the Netherlands [30], [31] or Finland [32]. But while the carbon uptake estimates are comparable with those elsewhere in the world, the costs per tonne of carbon removed in Ukraine are comparatively low (Table 2).

Table 2. Present value of carbon sequestration costs, € per tonne C*

Forestry zones € per tonne at the discount rates of

0% 2% 4%

Polissja 5.8 7.1 8.7

Wooded Steppe 4.6 5.9 7.2

Steppe 78.5 120.0 173.3

Carpathians 8.7 12.7 17.9

Crimea 16.2 15.6 32.0

The Ukraine 9.5 18.1

*C is in permanent tonnes

The average for the country costs of sequestering carbon are 9.5 € per tonne of carbon, if carbon is not discounted and 18 € per tonne when both the costs and benefits are discounted at the same discount rate of 4%. Thus, if/when the timing of CS is important, the forests that capture carbon quickly, such as poplar stands in the Wooded Steppe, might have a substantial advantage e.g. over the pine stands in the Steppe.

The wood product scenario stands for the option of cutting trees when their growth decelerates so that carbon is stored in wood products. Wood is used in construction and in the production of various goods. In this case, wood is a sink of carbon, with the duration


200

of the sink equivalent to the life of the goods. There are different views on the duration of life of wood products. In the present study, we assume that wood products release carbon back to the air after 40 years of storage [28]. Both in the wood-for-coal substitution and in bio-energy scenarios, the rotation ages of poplar and mixed species are 20 years, until the growth of the trees slows down. Coniferous trees are considered for harvesting when they reach 40 years of age. Thus, in the Steppe zone planted with pine, the age of harvesting is 40 years, and in a period of 80 years, two harvests take place. At the time of the second harvest, wood products made from the timber harvested 40 years ago release carbon back to the atmosphere. These were basic assumptions while computing carbon saving estimates.

Regarding a fossil fuel substitution (i.e. bio-energy) policy scenario, we assume that after harvesting, the wood is burned to generate energy and to replace an energy equivalent amount of fossil fuel. Only the above ground carbon fixed in wood is suitable for generating energy. When wood is used for energy production, carbon stored in wood biomass is released as CO2 upon burning. Thus, using timber for energy is by itself a carbon neutral process. The net gain here is the amount of carbon dioxide that would have been released by burning fossil fuel, if not replacing it with the burning of woody biomass. The results of the study are manifold (and are presented in-depth [5] and [21]), and among the other outputs and suggestions they indicate that under wood-for-fuel substitution and wood products sink policy options, the important factors that influence the results are, again, the discount rate employed in the analysis (Table 3).

Table 3. Costs of CS over 80 years per forestry zone, alternative scenarios

Scenario Discount rate for carbon Polissja Wooded Steppe Steppe

Energy 0% 16.5 16.2 153.4

4% 405.3 397.1 3764.4

Product 0% 33.0 32.4 215.8

4% 69.1 67.7 1035.9

The time horizon considered in the models is also a very important factor. While carbon uptake in a storage option has just one time effect, the fuel and product substitution strategies are repeatable. Thus, these scenarios are more sustainable means of carbon management, in the long run. An extension of the period under investigation provides more useful outcomes, because a continuous process could be shown. Because the effects for avoidance of carbon release through the replacement of non-timber materials are repeatable, social benefits under a wood product sink scenario in the long run are expected to be considerably higher than under the strategy of carbon fixation alone.

3. Effects of the use of declining discount rates on the decision-making process of tackling climate change: an example of afforestation in the United Kingdom

In the following section, we examine the economics of carbon sequestration, uncertainty and discounting, by considering a case of afforestation in the UK. We employ a bottom-up approach, i.e. the methodology suggested by Stavins and Richard [4], and in


201

the example presented above, we assess the costs and benefits of forestry over one rotation. The net present value (NPV) of one rotation by Fisher was calculated as:

rTpvecTMaxF −+−=)( , (5)

where: p is the timber price;

c is the plantation cost;

v is the timber volume;

T is a temporal variable;

r is the discount rate.

The cost per tonne of CS, i.e. the cost-effectiveness, was then estimated by dividing the opportunity costs of land conversion to forests by the number of tonnes of carbon sequestered:

C

NPVNPVCE gFarForestry

∆−

= min

, (6)

where: ForestryNPV is the NPV of forestry;

gminFarNPV is the NPV of farming;

C∆ is the carbon stock gain over one rotation; CE is the cost effectiveness of carbon sequestration.

We assume that the opportunity cost of land (NPV farming) is zero. This should not compromise the outcomes, which are related to the impact of various discounting techniques on the costs of CS. We begin the analysis with a Sitka Spruce plantation growing in the UK. In line with Moran et al. [9], we assume that the private discount rate is 7%, and that the social discount rate is 3.5%. For the declining discount rate protocol, the HM Treasury guidelines [33] were followed (shown in Figure 1).

Figure 1: The long term discount rate based on HM Treasury Guidelines, [33]

At the beginning of a project, we can simulate and anticipate what the CS rate of

a plantation will be. However, in the real world, forestry appears to be a risky activity. Fires, storms, pests and diseases can be a threat to the ability of a forest to sequester carbon. Also, different forests exhibit different sequestration patterns, as was shown


202

earlier in this paper. There are slow growing species that generate distant benefits in the future (as spruce in Ukraine’s Carpathians, for instance). There are fast growing species that generate immediate carbon sequestration benefits (e.g. hybrid poplar, as shown in our previous example from the Wooded Steppe, in Ukraine). Early sequestration does not have the same effect as late sequestration, because often CS can be seen as a potential means to “buy” time. A plantation would offset the effect of human activities while new technologies with a low carbon intensity, are developed. In this case, it would be relevant to encourage quick sequestration rates plantations rather than slow rates ones.

To take this dimension into account, it is possible to discount carbon benefits (as shown in the previous section). This methodology makes possible comparisons between forestry/land use related projects that exhibit different sequestration patterns; e.g. fast growing versus slow growing plantations; or projects aiming at sequestering carbon in the biomass versus projects aiming at sequestering carbon in soils. Moreover, financial costs are compared to carbon storage benefits on an identical basis, Songhen and Brown [34]; Nijnik [21].

1 2

0 0

( ) ( )T T

rt rtB s t e dt s t e dt− −′ ′= −∫ ∫ , (7)

where B describes the benefits of the project;

1T is the rotation length implemented in the first project and

2T the rotation length implemented in the second project. In our next example, which considers afforestation in the United Kingdom, CS rates

were derived from Bateman and Lovet [35]; details are also given in Nijnik et. al [36]. Yield tables from the UK’s Forestry Commission (FC) were used to analyze timber volumes and estimate the costs and benefits of forestry operations. Timber prices were taken from the FC’s website. Results showing the influence of various discounting protocols for Sitka spruce plantations are given below (please note that negative values indicate a negative cost, i.e. a positive return per tonne of carbon sequestered).

Table 4 Influence of the discount rates applied on the carbon stocks and on the costs of CS

Rotation 49 yrs

NPV Forestry (£/ha)

Carbon stocks (tC/ha)

Discounted carbon stocks (tC/ha)

Costs of carbon sequestration with carbon stocks not discounted (£/tC)

Costs of carbon sequestration with discounted carbon stocks (£/tC)

declining discount rate

constant discount rate (3.5%)

constant discount rate (7%)

£432

£99

-£947

131.93

131.93

131.93

64.94

62.23

35.57

-£3

-£1

£7

-£7

-£2

£27

Declining discount rates generate the highest NPV in the case of a 49 year rotation.

This also generates the highest carbon stocks, and so the lowest costs of CS. It actually even gives a positive return per tonne of carbon sequestered. The private discount rate of 7% shows that the NPV of forestry is negative and does not pass the CBA. This obviously raises the costs of CS. This is even more obvious when carbon stocks are discounted; carbon stocks appear to be relatively low, which then leads to an increase in


203

CS costs. It is worth noting that the declining discount rate protocol does not affect discounted carbon stocks too much; it shows that with such rates, marginal increases discounted are negligible, so discounted carbon stocks are almost equal in the declining discount rate and social discount rate cases. However, the NPVs are considerably different; which has a significant impact on CS costs.

4. Discussion

The immediate consequence of positive discounting is that future benefits (costs) are worth less than present benefits (costs). This consideration was raised long before climate change was considered as a problem. In 1928, Ramsey presented discounting as an ethically indefensible practice [15]. In 1942, Cyriancy-Wantrup recommended the use of zero or negative discount rates when health, education or national defense services are to be valued [19]. In 1948, Harrod defined discounting as a polite expression for rapacity. His argument was based on ethical grounds and environmental issues. Effectively, it appears that the consequences of discounting are dramatic when the economic valuation process deals with environmental issues occurring in the distant future [20].

In contrast, some authors argued that high discount rates were not necessarily a bad thing for the environment, for example Price [23]; high discount rates reduce the number of viable projects. Given that some projects might be a threat for the environment, the use of high discount rates can be considered as a positive thing. Another positive thing about high discount rates is that investments are directed towards the most efficient projects, so that revenues earned will give future generations the means to adapt to a degraded environment.

Uncertainties related to tackling climate change have raised the importance of investigating the discounting issues further. There have been recent advances in the theory of discounting. The discounting of carbon uptake benefits has been an issue of wide scientific debates [3; 5]. Acknowledging the will to give more weight to future costs and benefits, some have advocated the use of declining discount rates. From a theoretical perspective, their use has been justified by Groom et. al [24]. Some other theoretical developments [25; 26] show that declining discount rates are necessary to achieve intergenerational equity. Several justifications to the use of declining discount rates in CBA were given [27]:

- Time preference: experiments show that people, in their everyday lives, tend to use higher discount rates for present trade-offs than for more distant trade-offs;

- Pessimism about the future justifies the hyperbolic discounting; the discount rate being formally a function of consumption growth: if consumption is expected to fall, or if there is a probability of this, then we should use declining discount rates;

- Uncertainty also justifies declining discount rates. In the face of uncertain futures with several likely discount rates, the average of discount factors corresponding to likely discount rates is called the certainty equivalent discount factor; it is then possible, working backwards, to find the certainty equivalent discount rate. This certainty equivalent discount rate appears to be declining.

This has been recognised by official governments as the UK Government now recommends the use of declining discount rates for economic appraisal [33]. A report published for the French Government in 2005 also recommends the use of declining discount rates when investments spread costs and benefits over more than 30 years [37].


204

The choice of discounting protocol has a dramatic impact on the economic acceptability of the afforestation project. Declining discount rates improve considerably the profitability of forestry, and so reduce the costs of CS. The immediate consequence of a declining discount rate in CBA is that more projects will pass the CBA test. Environmental strategies generating long term benefits will also be favoured, which would be particularly relevant in the case of climate change as it would generate costs over centuries.

Declining discount rates tend to favour long term projects. However [38], declining discount rates give more weight to future benefits, compared to current benefits. Declining discount rates might emphasize that the longer term strategy is less costly than the short term strategy. However, there are situations when short term benefits could be preferred to long term benefits. For example, climate change is characterized by thresholds beyond which irreversible effects appear. If the threshold is close, we might prefer a short term strategy sequestering carbon quickly or drastically reducing emissions. The CBA shows only when carbon stocks changes are discounted.

5. Conclusions

Under the traditional CBA framework, it is common to use a positive and constant discount rate. However, this practice has been challenged. This particularly concerns the discounting of carbon uptake benefits. If the value of benefits from net emissions reduction was known, it would be easier to discount their expected flow, in the same way as we discount the costs. However, the benefits of carbon uptake are uncertain, and because of time preference, the majority of economists argue that future carbon reductions decrease in value rapidly. At discount rates of 5-10%, the PV of any amount of carbon sequestered in some 50 years from now sharply approaches zero. Declining discount rates have been proposed as a way to deal with uncertainty, and also achieve intergenerational equity in CBA. The estimation protocol also includes a provision for discounting carbon sequestration. This would allow consideration of uncertainty on carbon sequestration itself. In this paper, we have investigated the consequences of various discounting protocols on the costs of CS in forestry and have shown that the influence of choice of discounting rate on the decision making process is considerable.

To conclude, in the present paper, some innovative insights on accounting for uncertainties and time preference in the economic analysis of tackling climate change through forestry have been put forward and discussed. However these questions have not been deliberated in all their depth, as this piece of research does not discuss the theoretical or ethical rationales for various discounting protocols. The current paper introduces the problem as a new challenge for the future. It shows the consequences of using different discount rates in CBA on the decision-making process and as such is a valuable piece of research for policy makers.

References

[1] IPCC (2007): Climate Change 2007: Mitigation. Metz B., Davidson O. R., Bosch P.R., Dave R., Meyer L. A. (eds), Cambridge University Press, Cambridge.

[2] Newell R. G. and Stavins R. N. (2000): Climate Change and Forests Sinks: Factors Affecting the Costs of Carbon Sequestration. Journal of Environmental Economics and Management 40 (3), pp. 211-235.


205

[3] Van Kooten G.C. (2004): Climate Change Economics. Edward Elgar, Cheltenham. [4] Stavins R. N. and Richards K. (2005): The costs of US forest-based carbon

sequestration. Report for the Pew Center on Global Climate Change. Arlington, VA. Available at: http://www.pewclimate. org/docUploads/Sequest_Final.pdf

[5] Nijnik M. (2005): Economics of climate change mitigation forest policy scenarios for Ukraine. Climate Policy 4, pp. 319-336.

[6] Enkvist P.A., Naucler T. and Rosander J. (2007): A cost curve for greenhouse gas reduction. The McKinsey Quarterly 1, pp. 35-45.

[7] Nijnik M. and Bizikova L. (2008): Responding to the Kyoto Protocol through forestry: a comparison of opportunities for several countries in Europe. Forest Policy and Economics 10, pp. 257-269.

[8] Brainard J., Bateman I. J. and Lovett A. A. (2009): The social value of carbon sequestered in Great Britain’s woodlands. Ecological Economics 68, pp. 1257–1267.

[9] Moran D., MacLeod M., Wall E., Eory V., Pajot G., Matthews R., McVittie A., Barnes A., Rees B., Moxey A., Williams A. and Smith P. (2008): UK marginal abatement cost curves for the agriculture and land use, land use change and forestry sectors out to 2022, with qualitative analysis of options to 2050. Final Report to the Committee on Climate Change, № RMP4950, pp. 152.

[10] ADAS (2009): Analysis of policy instruments for reducing Greenhouse Gas Emissions from Agriculture, Forestry and Land Management – Forestry Options. Report to DEFRA. SFF0602.

[11] Van Kooten G.C., Eagle A.J., Manley, J. and Smolak, T. (2004): How costly are carbon offsets? A meta-analysis of carbon forest sinks. Environmental Science and Policy 7, pp. 239-251.

[12] Pearce D. and Turner R. (1990): Economics of Natural Resources and the Environment. Hertfordshire.

[13] Gollier C. (2001): The Economics of Risk and Time. Massachusetts Institute of Technology, USA.

[14] Clark C. (1973): The Economics of Over-exploitation. Science 181, pp. 630-634. [15] Ramsey, F. P. (1928): A Mathematical Theory of Saving. Economic Journal, 38, pp.

543-559. [16] Samuelson, P. (1961), The Evaluation of Social Income: Capital Formation and

Wealth. In: Lutz and Hague (eds.) The Theory of Capital. New York, USA.

[17]Hanley N. and Spash C. (1998): Cost-Benefit Analysis and the Environment. Edward Elgar Publishers, Cheltenham, UK and Northampton, MA USA.

[18] Nordhaus W. (1991): To Slow or not to Slow: the Economics of the Greenhouse Effect. Economic Journal 101, pp. 920-938.

[19] Cyriacy-Wantrup S. V. (1942): Private Enterprise and Conservation. Journal of Farm Economy 24, pp. 12-21.

[20] Harrod R. F. (1948): Towards the Dynamic Economy. St Martin's Press, London, UK.

[21] Nijnik M. (2002): Dealing with Uncertainties and Time Preference in Cost-Benefit Analysis of Planting Tree to Mitigate Climate Change. In: van Ierland E.C., Weikard H.P. and Wesseler J. (eds) Risk and uncertainty in environmental and resource economics, Wageningen University Pubs., The Netherlands, pp. 143-153.


206

[22] Nijnik M. (2010): Carbon capture and storage in forests. Issues in Environmental Science and Technology of the Royal Society of Chemistry, Cambridge 29, pp. 203-238.

[23] Price C. (2005): How Sustainable is Discounting? Economics, Sustainability and Natural Resources Kant and Berry (eds.), volume 1, Chapter 6.

[24] Groom B., Hepburn C., Kondouri P. and Pearce D. (2005): Discounting the Future: the Long and Short of it. Environment and Resource Economics, 31.

[25] Chichilnsky G. (1997): What is Sustainable Development? Land Economics, 73, pp. 467-491.

[26] Li C. and Lofgren K.-G. (2000): Renewable Resources and Economic Sustainability: A Dynamic Analysis with Heterogeneous Time Preferences. Journal of Environmental Economics and Management, 40, pp. 236-250.

[27] Hepburn C. J. and Koundouri P. (2007): Recent Advances in Discounting: Implications for Forest Economics. Journal of Forest Economics, volume 13, Issue 2-3, pp. 169-189.

[28] Van Kooten G. and Bulte E. (2000): The Economics of Nature: Managing Biological Assets. Oxford, UK: Blackwell Publishers.

[29] Lakida P., Nilsson S. and Shvidenko A. (1995): Estimation of Forest Phytomass for Selected Countries of the Former European USSR. WP-95-79, IIASA, Laxenburg, Austria.

[30] Slangen L., Van Kooten G. and Van Rie P.F. (1997): Economics of Timber Plantations on CO2 Emissions in the Netherlands. Tijdschrift van de Landbouw 12 (4), pp. 318-33.

[31] Nabuurs G., Dolman A., Verkaik E. et.al. (1999): Resolving Issues on Terrestrial Biospheric Sinks in the Kyoto Protocol. Dutch National Research Programme on Global Air Pollution and Climate Change. Report no: 410 200 030, pp. 100.

[32] Pussinen A., Karjalajnen T., Kellomaki S. and Makipaa R. (1997): Potential Contribution of the Forest Sector to Carbon Sequestration in Finland. Biomass and Bioenergy 13 (6), pp. 377-387.

[33] HM Treasury (2003): The Green Book: Appraisal and Evaluation in Central Government: London,

[34] Sohngen B. and Brown S. (2008): The cost and quantity of carbon sequestration by extending forest rotation age. Available at: http://aede.osu.edu/people/sohngen.1/forests/Cost%20of%20Aging_v2.pdf

[35] Bateman I. J. and Lovett A. A. (2000): Estimating and valuing the carbon sequestered in softwood and hardwood trees, timber products and forest soils in Wales. Journal of Environmental Management 60, pp. 301-323.

[36] Nijnik M., Pajot G., Moffat A. and Slee B. (2009): Analysing opportunities of British forests to contribute to climate change mitigation. European Association of Environmental and Research Economics, Amsterdam. Available at: http://www.eaere2009.org/Programme/Detailed

[37] Lebegue D. (2005): Révision du taux d’actualisation des investissements publics. Rapport pour le Commissariat General du Plan.

[38] Valatin G. (2009) Forestry Carbon: Valuation, Discounting and Risk Management, Discussion Paper, Forest Research.


207

Uncertainty of site-specific FOD for the national inventory of methane emission

Sang-hyup Oh, Gwisuk Heo, Jin-Chun Woo 1

1 Korea Research Institute of Standard and Science(KRISS), 1 Doryong-dong Yusong-gu Daejeon (305-600) Korea

[email protected]

Abstract

We have developed a standard procedure and continued to modify it for the preparation of uncertainty of a first order decay constant (FOD) of methane which would be useful to use the calculation of site-specific emission in the category of solid waste land-fill site. As an example, we applied this standard procedure to a FOD value obtained from Daejeon land-fill site in Korea. Since a numerical approach was adopted for the calculation of FOD with many simultaneous model equations, we applied matrix technique for the propagation of uncertainties instead of the Mote-Carlo simulation. Since there were no data on the seasonal variation of methane emission at that time, it was not included. It was also found that the treatment of correlation between uncertainties was very important in the combing measurement uncertainties. As a result, FOD value of Daejeon land-fill site was 0.163 with standard uncertainty of 0.013(8.3 %) and expanded uncertainty was 0.026(16 %).

Keywords: Frst order decay constant (FOD), solid waste site, uncertainty, standard procedure, correlation, uncertainty propagation

1. Introduction

The reliability of GHG emmision inventory data is one of the pre-requisites for the good policy decision-making in the national and international reduction of GHG. In order to provide the more reliable GHG emmision inventory data, 2006 quideline of IPCC recommanded to incorporate with high level of tier 2 or 3 instead of tier 1. In the category of solid waste disposal sites(SWDS), national or site-speccific FODs are essential data for the application to high level of tier 2 or 3. In Korea, several research institutes have devote themseves to the development of national or site-speccific FODs which are usful to apply national or site-specific GHG inventoy.

In the sense of reliability of the FOD as a research product, well-designed quality control and assurance(QC/AC) in the developing step and review processes by third party are needed. For this type of application, we developed a standard procedure for the calculation of uncertainty of the FOD value which would be useful to the development projects. For the completeness of the procedure, until now, we have continued to modify the procedure after the many application to real situation of data. In the procedure, choice of uncertainty components and quantification were designed by 2006 quideline of IPCC and uncertainty propagations were followed with the idea of 2006 quideline of IPCC, Guide to the Eexpression of Uncertainty in Measurement(GUM) and DIN 13157. Correlation factors were also included for the propagation steps of correlated uncertainty components. In this paper, we explained the procedure for uncertainty calculation with an example of the mathane emission data obtained from Daejeon Lanf-fill site in Korea.


208

2. Measurement data and calculation of FOD

Actual methane emission quantities from Daejeon land-fill site in Korea were obtained and used for the evaluation of the developed procedure. Land-fill site, experimental methods and results were described in the section.

2.1 Measurement of emission

Total area of Daejeon land-fill site is about 271,616 m2 and they were measured in Jun. of 2009. At the 3 positions at around 15 red circles as shown in Figure 1, emission gas was sampled as the representation of ground surface. And also emission gas was sampled at 3 positions(red double circle) for elimination holes .

Figure 1. Sampling positions at Daejeon Land-fill site

Dynamic sampling method was incorporated using plastic chamber and bags as shown in Figure 2. Gas temperature and pressure were also measured directly at the sampling site. After sampling, bags were moved to analysis Lab. and determined the individual gas concentration by Gas-Chromatography(GC) with one point calibration method with a reference gas.

The surface flux at each point was measured for 3 times and calculated using equation 1. Since dynamic sampling method was adopted, so the flow rate of sweeping gas of nitrogen was maintained at 5.0 L/min.

310−×××××=std

chamber

chamber

std

chamber

sweep

gas

wii P

P

T

T

A

Q

V

MCE , (1)

where Ei , flux of CH4 at sampling point i , Ci, CH4 Conc. of sampling point i,

stdstd

samplei C

A

AC ×= (2)


209

where Asample, response of sample from GC,

Astd, response of standard gas from GC,

Cstd, Conc. of reference standard gas,

Achamber=0.1256 m2, surface area of chamber,

Mw= 16.01 g/mol , molecular weight of CH4,

Tstd, standard state Temp.,

Vgas, 22.4 L/mol (gas volume at STP),

Tchamber, Temp. of sampling condition (u =±0.9%),

QSWEEP , 5 L/min, input gas flow (u =±5%),

Pstd, standard pressure,

Pchamber, pressure at sampling time (u =negligible).

Figure 2. Dynamic gas sampling chamber used

The measurement results of concentration and flux at each sampling position was summarized at Table 1. Excluding the extreme values of S11-2, average flux was 55.2 mg/m3min. Total emission quantity from the land-fill surface area in 2009 was 7873 ton/yr calculated using equation 3.

WoSURFACE AEE ××= )0005256.0( , (3)

where Eo ; average flux of CH4 emission ( 44 points) ,

n

EE

n

ii∑

== 10 , (4)

where Eo = average flux of CH4 emission ,

Ei = Flux of CH4 at sampling point i,

n = number of the measurement point (44 point),

Aw; land-fill area (271313 m2 , u=±5%).


210

Table 1. Methane concentration and flux at each sampling position

Sample No.

Temp.

℃

Press. mbar

Conc. µg/L

Flux mg/m2min

Sample No.

Temp.

℃

Pressure mbar

Conc. µg/L

Flux mg/m2min

S1-1 31.2 1008.5 1,865 74.2 S8-3 26.4 1011.3 726 28.9

S1-2 30.3 1008.5 776 30.9 S9-1 31.2 1010.1 2,447 97.4

S1-3 30.7 1008.5 396 15.8 S9-2 31.8 1010.1 4,326 172

S2-1 30.5 1009.7 286 11.4 S9-3 32.1 1010.1 675 26.9

S2-2 31.6 1009.7 1,310 52.2 S10-1 31.4 1008.4 4,692 187

S2-3 32 1009.7 3,756 150 S10-2 30 1008.4 4,658 185

S3-1 30.3 1010.6 1,119 44.6 S10-3 29.4 1008.4 1,385 55.1

S3-2 29.7 1010.6 509 20.3 S11-1 29.9 1008.4 3,823 152

S3-3 30.8 1010.6 460 18.3 S11-2* 30.5 1008.4 63,223 2,517

S4-1 26.9 1011.3 2,294 91.3 S11-3 30.3 1008.4 3,791 151

S4-2 27.2 1011.3 493 19.6 S12-1 29.4 1008.4 139 5.54

S4-3 28.7 1011.3 2,385 95.0 S12-2 29.5 1008.4 168 6.68

S5-1 31.1 1010.6 0 0 S12-3 27.2 1008.4 35.0 1.39

S5-2 30.2 1010.6 5.07 0.20 S13-1 30.5 1007.6 1,592 63.4

S5-3 31.6 1010.6 3.38 0.13 S13-2 29 1007.6 2,529 101

S6-1 28.5 1010.6 1,185 47.2 S13-3 29.7 1007.6 1,341 53.4

S6-2 28.1 1010.6 432 17.2 S14-1 31.5 1007.6 90.9 3.62

S6-3 30.2 1010.6 420 16.7 S14-2 31.6 1007.6 214 8.52

S7-1 26.5 1010.6 543 21.6 S14-3 30.9 1007.6 155 6.19

S7-2 26.1 1010.6 1,075 42.8 S15-1 32.6 1007.6 1,612 64.2

S7-3 29.7 1010.6 947 37.7 S15-2 31.8 1007.6 2,246 89.4

S8-1 27 1011.3 1,270 50.6 S15-3 29.9 1007.6 1,354 53.9

S8-2 26.6 1011.3 1,480 58.9

* outlier

For 3 elimination holes, flow rate and concentration were measured. Total methane emission from 3 elimination holes in the year 2009 was 2480 ton/yr calculated using equation 5.

536.31)( 333222111 ×××+××+××= TWTWTWTWTWTWTWTWTWTW VACVACVACE ,(5)

where ETW ; emission flux from 3 emission holes in 2009 (ton/yr), CTW1~3 ; Conc. of emission hole i , Atw1~3; area of emission hole i, Vtw1~3 ; velocity of emission hole i.

In order to correct the oxidation and recovery, corrected total emission quantity was obtained using equation 6 and constants provided from 2006 IPCC guideline. Calculated total corrected emission quantity from Daejeon land-fill area in the year 2009 was 14378 ton/yr

)1()1(2009

2009 ROX

QQ

TR −×−

=, (6)


211

where, QR2009; corrected quantity after oxidation and recovery correction (ton/yr),

Q 2009 ; measured quantity of CH4 emission in 2009(ton/yr),

where TWSURFACE EEQ +=2009 ,

OXT ; oxidation ratio (0.1, u=±10%),

R ; recovery ratio (0.2, u=±10%).

2.2 Activity data of waste

Total mass of decomposable degradable organic carbon at Daejeon land-fill site was calculated with equation 7 and activity data obtained from the local government which manages the Daejeon site. The type of waste and total mass of decomposable degradable organic carbon were summarized in the Table 2

MCFDOCDOCWDDOCN

iifiiTTm ×

××= ∑=1

,,, , (7)

where DDOCm,T ; mass of decomposable degradable organic carbon at the disposal time in the year, T,

WT, i ; amount of waste of type i in the year, T, DOCi ; ratio of organic carbon of type i , DOCf,i ; fraction of DOC of type i that can decompose (fraction), MCF =1; CH4 correction factor for aerobic decomposition in the year, T (fraction), i = 1 – N ; type of waste.

Table 2. Waste deposited from 1996 to 2009 at Daejeon land-fill site(ton)

Year Rubber Others Wood Fibers Sludge Food Paper DDOC m,T

1996 600 3500 7254 0 1000 600 3000 15954

1997 902 5,218 1,816 0 1,232 20,933 13,908 43948

1998 765 4,091 1,599 0 1,547 15,998 12,498 36,497

1999 625 4,246 1,199 0 1,142 12,263 10,371 29,846

2000 275 4,914 2,156 0 1,205 10,461 11,261 30,273

2001 260 6,707 1,378 0 11 11,259 12,193 31,807

2002 265 5,135 1,293 0 1,404 969 2,968 12,034

2003 284 6,447 1,335 0 721 5,904 12,968 27,660

2004 132 4,831 336 0 27 5,576 13,627 24,531

2005 66 2,815 344 0 53 0 4,460 7,738

2006 69 4,362 319 0 22 212 4,806 9,790

2007 74 5,500 306 0 124 732 4,130 10,865

2008 74 5,500 306 0 124 732 4,130 10,865

2009 74 5,500 306 0 124 732 4,130 10,865


212

2.3 Site-specific FOD(k)

The site-specific FOD at Daejeon land-fill site was calculated using the equation 8, a measurement result of total emission(QR2009) and activity data of total mass of decomposable degradable organic carbon(from DDOC m1996 to DDOC m2009)

19961996

19961996

199619971997

19971997

200720082008

20082008

200820092009

20092009

20092009

1

1

1

1

1216

,m,ma

k,ma,md

k,ma,m,ma

k,ma,md

k,ma,m,ma

k,ma,md

k,ma,m,ma

k,ma,md

,mdR

DDOCDDOC

)e(DDOCDDOC

)eDDOC(DDOCDDOC

)e(DDOCDDOC

.

.

)eDDOC(DDOCDDOC

)e(DDOCDDOC

)eDDOC(DDOCDDOC

)e(DDOCDDOC

/FDDOCQ

=−⋅=

⋅+=

−⋅=

⋅+=

−⋅=

⋅+=

−⋅=

⋅⋅=

−

−

−

−

−

−

−

(8)

where DDOCma,T ; DDOCm accumulated in the site at the end of year T,

DDOCmd,T ; DDOCm deposited into the site in year T,

F ; fraction of CH4, by volume, in generated landfill gas (fraction),

16/12 ; molecular weight ratio CH4/C (ratio).

Actually, Newton’s numerical method was applied to obtain the solution. But exact solution was not obtained because of over-estmated large value of total corrected emission quantity in the year 2009(QR2009) or unknown source of uncertainty. Therefore FOD value with minimum calculation error was taken and the value of FOD(k) of methane emission in Deajeon land-fill site was 0.163.

3. Uncertainty of site-specific FOD(k)

The key concepts of uncertainty in the IPCC 2006 guideline are almost same as that of GUM widely used for the measurement fields. But the techniques for the choice of uncertainty components and quantification of uncertainty are well described in IPCC 2006 guideline while the definition and terminology are well desgined in GUM. Therefore, in the procedure developed, the technique for the choice of uncertainty components and quantification of uncertainty were mainly followed to IPCC 2006 guideline while the definition and terminology were used by rules of GUM. The main flow chart for the evaluation of uncertainty is summarized in Figure. 3.


213

Figure 3. Flow chart of uncertainty evaluation

3.1 Uncertainty components

In order to find the components of measurement uncertainty, all the measurement and calculation steps, and input variables described by model equation 1 – 8 were verified. Moreover, the lack of completeness of model equations was seriously verified with following view points of probable incompleteness of model equation:

1) Repeatability, 2) Uncertainty of reference gas used, 3) Uncertainty due to instrumental drift during the analysis, 4) Uncertainty due to non-linearity of instrument, 5) Uncertainty due to matrix effect and interference, 6) Homogeneity and stablity, 7) Effect of humidity and sampling flow rate.

In order to find the components of uncertainty of activity data and constants provided by 2006 IPCC guideline, all the calculation steps, and input variables described by model equation 1 – 8 were verified and examined with expert opinions, data from reference and 2006 IPCC guideline.

3.2 Adjustment of model equation

After the careful review process on the uncertainty components, equation 1-8 were modified if the models had lack of completeness. So dummy factors with unit value were added to the equation 2 and 4 for furher treatment of qantification and propagation of uncertainty such as equation 9 and 10

chamberlin-waterstdstd

samplei f fC

A

AC ×××= , (9)

where f chamber=1; factor due to in-homogeneity in chamber (u =±5%), flin-water=1; factor due to non-linearity and water-effect (u =± 0.01%).

Searching uncertainty components & setting up model Equation

Quantifying uncertainty components and PDF

Combined standard uncertainty by uncertainty propagation

Extended uncertainty at the level of confidence, 95 %


214

seasonohom

n

ii

o ffn

EE ⋅⋅=

∑=1 , (10)

where f homo=1; factor of data dispersion (u =±0.5),

f season=1; factor of seasonal variation effect.

3.3 Quantifying and propagating uncertainties

The standard uncertainties of all the variables in equation 1-7 and 9-10 were quantified by A and B type evaluation after setting-up and adjusting the model equation. If there were no data for the quantification of uncertainty in measurement process, additional experiments were performed. Since there were no data on the seasonal variation of methane emission at that time, it was not quantified. And, except uncertainty of fhomo, PDFs of all uncertainty components in measurement process were assumed as Gaussian or t distribution. The uncertainty of activity data and constants were obtained with expert opinions, data from reference and 2006 IPCC guideline. And PDFs of many of uncertainty components in activity data were assumed as square distribution. After the quantification, equation 11 was used for the propagating uncertainties

( ) ( ) ( ) ijjij

N

j i

N

ic rxuxu

x

f

x

fyu

11

2

∂∂

∂∂= ∑∑

==

, (11)

where ( )Nxxxxfy ⋅⋅⋅= 3,2,1, .

As shown in equation 11, correlation factors were incorporated in the propagation step. In the actual application of equation 10, each of the uncertainty of Ei was treated as if it was correlated(r ij =1) one another since the original data of each uncertainty was same. If PDF of an uncertainty component is symmetrical and there were no correlation effect, equation 12 and 13 were used for the propagating uncertainties

( ) ( ){ }( )∑

=

⋅⋅⋅⋅⋅⋅−⋅⋅⋅±⋅⋅⋅

=N

i Ni

Nii

x,x,x,x,xf

x,xux,,x,x,xf)yu

1 321

32122

,

,

(12)

( ) ( ){ }( )∑

=

⋅⋅⋅⋅⋅⋅−⋅⋅⋅±⋅⋅⋅

=N

i Ni

Nii

x,x,x,x,xf

x,xUx,,x,x,xf)yU

1 321

32122

,

,

(13)

After quantification and propagation, results and uncertainty of input quantity were summarized as shown in Table 3-7.

Level of standard uncertainty of output variables, Ei and ETW were about 3.5% and 18 %. And standard uncertainty of output variables, ESURFACE and QR2009 were 10% and 8.9 %. For the propagation of uncertainty of equation 8, equation 13 and 14 was used. If PDF of input uncertainty was symmetrical and there were correlation effect, the following procedure was used with matrix calculation:


215

Table 3. Standard uncertainty for equation 9

No Input Unit Standard uncertainty Symbol value

1 Cstd 503.5 ppm 1.25 2 Astd 1308.5 m2 13.6 3 As1-1 7592.8 m2 20.8 4 Mw 16 g/mol 0 5 Vgas 22.4 L/mol 0 6 Qsweep 5.0 L/min 0 7 Achamber 0.126 m2 0 8 Tstd 273 K 0 9 Ts1-1 304 K 2.73 10 Pstd 1013 mbar 0 11 Ps1 1008.5 mbar 0 12 flin-water 1 - 0.011 13 fchamber 1 - 0.028

Output E1-1 74.2 ton/yr 2.54 (3.43%) U ton/yr 5.02 (6.76%)

Table 4. Standard uncertainty for equation 3 and 10

No INPUT Standard uncertainty

No Input Standard uncertainty Symbol Value

(ton/year) Symbol Value

(ton/year) 1 E1-1 74.2 2.54 24 E8-3 28.9 0.9 2 E1-2 30.9 1.1 25 E9-1 97.4 3.3 3 E1-3 15.8 0.55 26 E9-2 172 6.1 4 E2-1 11.4 0.40 27 E9-3 26.9 1.08 5 E2-2 52.2 1.84 28 E10-1 187 6.6 6 E2-3 150 5.20 29 E10-2 185 6.3 7 E3-1 44.6 1.55 30 E10-3 55.1 1.9 8 E3-2 20.3 0.71 31 E11-1 152 5.3 9 E3-3 18.3 0.64 32 E11-2 10 E4-1 91.3 3.15 33 E11-3 151 5.1 11 E4-2 19.6 0.68 34 E12-1 5.54 0.2 12 E4-3 95.0 3.26 35 E12-2 6.68 0.2 13 E5-1 0 0 36 E12-3 1.39 0.05 14 E5-2 0.20 0.0073 37 E13-1 63.4 2.2 15 E5-3 0.13 0.0046 38 E13-2 101 3.6 16 E6-1 47.2 1.64 39 E13-3 53.4 1.8 17 E6-2 17.2 0.60 40 E14-1 3.62 0.1 18 E6-3 16.7 0.59 41 E14-2 8.52 0.3 19 E7-1 21.6 0.76 42 E14-3 6.19 0.2 20 E7-2 42.8 1.51 43 E15-1 64.2 2.2 21 E7-3 37.7 1.31 44 E15-2 89.4 3.1 22 E8-1 50.6 1.74 45 E15-3 53.9 1.9 23 E8-2 58.9 2.07 46 A

W 271313 13565.6

Output ESURFACE

7873 791 (10.0%) 47 fhomo

1 0.080 U 1582 (20.1%)


216


No Input Unit Standard uncertainty Symbol Value

1 CTW1

227 g/m3min 11.7

2 CTW2

219 g/m3min 1.4

3 CTW3

54.7 g/m3min 2

4 ATW1-3

0.785 m2 0.0092

5 VTW1-3

0.20 m/sec 0.035

6 fIR

1 - 0.029

Output ETW

2479 ton/yr 440 (18%)

U ton/yr 880 (36%)


No Input Unit Standard uncertainty Symbol Value

1 ESURFACE 7873 ton/yr 791

2 ETW 2479 ton/yr 440 Output Q2009 10325 ton/yr 905(8.7 %)

3 OXT 0.1 - 0.0058 4 R 0.2 - 0.012

Output QR2009 14378 ton/yr 1279(8.9%) U ton/yr 2557(18 %)

Table 7. Standard uncertainty for DDOC m,T

Year DDOC m,T

Standard uncertainty

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

15954 43948 36,497 29,846 30,273 31,807 12,034 27,660 24,531 7,738 9,790 10,865 10,865 10,865

2233 6161 5109 4178 4238 4453 1684 3872 3434 1083 1370 1521 1521 1521


217

1) Matrix expression of nF equation; such as 0 FFn 2, 1, == TFFF )( KK

, 2) Matrix expression of input and output parameters; such as )...( 1

TmXX=X ,

and )..( 1T

nY...Y=Y where input and output parameter’s notes are m and n,

3) Calculating of Fx and Fy 4) Calculating Sensitivity matrix; x-1y F-FQ =

,

a)

∂∂= yx,

k

jx F │

X

F where j= 1, …… nF and,

b)

∂∂= y,x

k

jy

Y

F F │ where k=1,……….n or m,

4) Calculating Sensitivity matrix; x-1y F-FQ =

,

5) Calculating uncertainty variance matrix of output parameters; TQQUU xy =

where yU and xU are uncertainty variance matrix of output and input quantities .

As a result, FOD value of Daejeon land-fill site was 0.163 with standard uncertainty of 0.013(8.3 %) and expanded uncertainty was 0.026(16 %).

4. Conclusion

We have developed a standard procedure and continued to modify it for the preparation of uncertainty of FOD of methane which would be used for the calculation of national or site-specific GHG emission inventory in the category of solid waste land-fill site. And, as an example, we it to a FOD value obtained from Daejeon land-fill site in Korea. From the application, some tips for proper treatment of uncertainty were found and it is as follows:

1) FOD value of Daejeon land-fill site was 0.163 for methane emission with standard uncertainty of 0.013(8.3 %) and expanded uncertainty was 0.026(16 %).

2) Total methane emission quantity obtained from the study was over-estimated.

3) The treatment of correlation between uncertainties for propagation was very important

4) Since there was no data for seasonal variation of emission, future work is needed.

References

[1] Guide to the Expression of Uncertainty in Measurements; ISO, Geneva, Switzerland, 1993.

[2] 2006 IPCC Guidelines for National Greenhouse Gas Inventories; Eggleston H.S., Buendia L., Miwa K., Ngara T. & Tanbe K.Eds., IGES, Japan.

[3] Quantifying uncertainty in analytical measurement EURACHEM Guide; EURACHEM, Teddington, UK, 1995.

[4] The Fitness for Purpose of Analytical Methods; A Laboratory Guide to Method Validation and Related Topics; Eurachem, http://www.eurachem.org/guides/valid.pdf, 1988.


218

[5] DIN 1319-4(1999-02); Grundlagen der Meßtechnik

[6] W. H. Press, S. A. Teukolsky, W. T. Vetterling, B.P. Flannery: Numerical Recipes in FORTRAN - The Art of Scientific Computing. 2. Auflage, Cambrige University Press(Cambridge) 1992


219

Application of the IPCC uncertainty methods to EDGAR 4.1 global greenhouse gas inventories

Jos G.J. Olivier1, John A. van Aardenne2*, Suvi Monni2, Ulrike M. Döring2, Jeroen A.H.W. Peters1 and Greet Janssens-Maenhout2

1 Netherlands Environmental Assessment Agency (PBL) P.O. Box 303, NL-3720 AH Bilthoven, The Netherlands

[email protected] 2 Joint Research Centre, Institute for Environment and Sustainability (JRC-IES)

Climate Change Unit, TP290, 2749, I-21027, Ispra (Va), Italy * Now at European Environment Agency (EEA), Kongens Nytorv 6,

1050 Copenhagen K, Denmark

Introduction JRC and PBL have compiled a comprehensive EDGAR v4.1 global emissions

dataset for the period 1970-2005 for the ‘six’ greenhouse gases included in the Kyoto Protocol (CO2, CH4, N2O, HFCs, PFCs and SF6), which was constructed using consistently the 2006 IPCC methodology and combining activity data (international statistics) from publicly available sources and for the first time - to the extent possible - emission factors as recommended by the IPCC 2006 guidelines for GHG emission inventories (Figure 1). This dataset, that covers all countries, provides independent estimates for all anthropogenic sources from 1970 onwards that are consistent over time and comparable between countries. Where appropriate emission abatement or recovery was taken into account, based on data reported by Annex I countries under the UN Climate Convention or based on other publicly available data sources. The resulting emissions of all gases identified in the Kyoto Protocol are reported using the 1996 IPCC source category classification for ease of recognition of the scope of each category and to allow for easy comparison with national greenhouse gas inventories reported by Annex I countries.

Thus we provide full and up-to-date inventories per country, also for developing countries that go beyond the mostly very aggregated UNFCCC reports of the developing countries. Of the 220 UN nations in 2005 only 43 industrialised countries (‘Annex I’) annually report their national GHG emissions in large detail from 1990 up to (presently) 2008, while most developing countries (‘non-Annex I’) for the UN Climate Convention (UNFCCC) and the Kyoto Protocol only report a summary table with emissions for one or more years (many only for 1994) (UNFCCC, 2005). More information on methods, data sources and differences with previous data is provided in the documentation available at the EDGAR 4 website: http://edgar.jrc.ec.europa.eu . Moreover, the time series back in time to 1970 provides for the UNFCCC trends a historic perspective. As part of our objective to contribute to more reliable inventories by providing a reference emissions database for emission scenarios, inventory comparisons and for atmospheric modellers, we strive to transparently and publicly document all data sources used (Olivier et al., 2010) and assumptions made where data was missing, in particular for assumptions made on the shares of technologies where relevant.


220

0

5

10

15

20

25

30

35

40

45

50

1970 1975 1980 1985 1990 1995 2000 2005

Forest fires CO2

Postburn CO2 and N2O

HFCs, PFCs and SF6

Other CO2, CH4 and N2O

Waste CH4

Agriculture N2O

Agriculture CH4

Industrial processes

Fossil fuel: fugitive andflaring

CO2 and non-CO2 fromfossil fuel combustion

Figure 1. Trend in global greenhouse gas emissions 1970-2005 (unit: Pg CO2-eq.) (source: EDGAR 4.1)

Uncertainty in global and national greenhouse gas inventories

We present our estimate the global inventories of the main greenhouse gases and their trend by major source and region and the methods used to estimate the uncertainty in total regional and total global emissions and representative estimates at country level. Since the uncertainty estimates start with the data used at country level, we have aggregated sources and countries to regions where significant correlation of activity data or emission factor uncertainty exists between source categories or between countries, e.g. when using regional or global default emission factors (Olivier and Peters, 2002).

While using IPCC methodology and default emission factors whenever possible, this also allows us to use the default uncertainty estimates provided in the 2006 guidelines in most cases. Many Annex I countries may apply higher tier methods than was done for EDGAR 4.1 and may also apply country-specific emission factors rather than IPCC default values, that should in most cases result in lower uncertainties.

Uncertainty estimates are made for different reasons. In scientific inventories such as EDGAR and in official national GHG inventories. In scientific inventories, it is good scientific practice to assess and report on the uncertainty of the results as an expression of the overall quality of the resulting emissions as judged by the compilers. A preliminary estimate of uncertainties in global emissions of CH4 sources in EDGAR 3.2 based on IPCC default values appeared to be comparable with uncertainties estimated by global budget studies (Olivier, 2002). This is useful information for atmospheric modellers that require uncertainty estimates for all parameters in their model of which emissions are an important one, so the uncertainty in emissions is part of the overall uncertainty assessment of the model application. On the other hand, for official national greenhouse gas inventories uncertainty estimates are made just as a means for prioritising inventory improvement activities. Since the focus of these inventories lies in reporting emission inventories according to the guidelines, estimating uncertainties is often not given a high priority and IPCC default uncertainty values are applied. Knowing these different approaches to uncertainty estimates is pivotal information for using and interpreting these


221

different types of emissions inventories by the Earth System and Atmospheric Modelling communities.

Besides application in comparisons to other greenhouse gas inventories, emission uncertainty estimates are also important information for atmospheric modellers when estimating emissions (‘inferred emissions’) from measured atmospheric concentrations by so-called inverse modelling. Here a priori emissions are required with uncertainty estimates for each major sources and region to restrict the model to areas where emissions are believed to be most uncertain. Also uncertainty estimates for both emission datasets are required to assess their comparability. Inverse modelling of global or regional emissions has been done for several gases now, such as CH4 and HFC-134a (Villani et al., 2010). Recently more results on recent trends in F-gas emissions such as HFC-23 (Montzka et al., 2010), CF4, C2F6, F3F8 (Muhle et al., 2010) and SF6 (Levin et al., 2010; Rigby et al., 2010) have been published. The methods we applied in estimate uncertainties in total global emissions of our scientific inventory may also be used for combining official emission inventories reported by countries to the UNFCCC, e.g. for use in atmospheric models for verification purposes.

Comparison with official Annex I inventories

Apart from reporting the estimated uncertainty per source category, we also document the tier level of the methods used to compile the EDGAR 4.1 inventories. Therefore it is of interest to compare per category the difference between reported national emissions as well as reported uncertainty estimates for them and EDGAR estimates of emissions and their uncertainty. In Figure 2 we show comparisons for selected Annex I countries of emissions reported to the UNFCCC and EDGAR 4.1 estimates, for national total emissions (without uncertainty). In Figures 3 and 4 we compare emissions of major source sectors of CH4 and N2O for the same countries. Through this comparison we can assess whether or not the IPCC default methods and/or default emission factors show a significant bias for application by industrialised countries or that the uncertainty in the reported emissions is so large that no robust conclusion can be drawn. Except for some notable sources in particular countries most source estimates seem to agree reasonable, taking into account the uncertainties that often resemble the (default IPCC) uncertainty in the emission factors used. The notable exceptions have to be investigated further to determine the causes of the large differences: inconsistent activity data of national and international statistics, the use of very different country-specific emission factors due to country-specific circumstances, the use of high tier or country-specific methodology, a judgement error in selecting the emission factors or a calculation error.

If they would show a bias, this would warrant the use of asymmetrical uncertainty ranges when using lower tier IPCC methods or default factors. Moreover the comparison provides insight on the net gain of using higher tier methodology and allows identifying those regions or sectors where application of higher tier methodology would be most beneficial.

Since the uncertainty estimates start for the data used at country level, we have aggregated sources and countries to regions where significant correlation of activity data or emission factor uncertainty exists between source categories or between countries, e.g. when using regional or global default emission factors.

Areas where higher tier methods or country-specific emission factors instead of default IPCC factors will increase the inventory quality are:


222

Total CH4 emissions in 2005 (excl. LULUCF)

-

5,000

10,000

15,000

20,000

25,000

30,000

Australia France Germany Italy Japan Netherlands RussianFederation

UnitedKingdom

UnitedStates

Gg

UNFCCC

EDGAR4.1

Total N2O emissions in 2005 (excl. LULUCF)

-

200

400

600

800

1,000

1,200

Australia France Germany Italy Japan NetherlandsRussianFederation

UnitedKingdom

UnitedStates

Gg

UNFCCC

EDGAR4.1

Figure 2. Comparison of national total CH4 (a) and N2O (b) emissions

in 2005 between EDGAR 4.1 and UNFCCC for selected countries (without LULUCF) (unit: Gg)

(a) CO2 emission factors for fuel combustion (1A). Natural gas, coal, petrol and

diesel in road transport are often used and in large amounts and therefore cover a large fraction of national GHG emissions. It is known that carbon contents of gas and coal can vary significantly, depending on where it is produced. Also Annex I reporting of petrol emission factors shows a considerable spread in values and a tendency to depart from the IPCC default values (see examples provided in Table 1). As we can see, determining a country-specific value for these fuels may improve the accuracy in this part of the inventory. In particular for natural gas and for diesel in road transport the IPCC defaults, although still within the estimated uncertainties, seem to be somewhat biased to the low side (by 4 and 2.5%). For coal this conclusion cannot be drawn from the table since the values reported by Annex I countries refer to total “solid fuel”, which may include not only coal, but also coal-derived gases such as coke oven gas and blast furnace gas as well as brown coal.


223

(b) CH4 emission factors for animals (4A) and rice production (4C) may be improved compared to (region-specific) default values by using higher tier methods to determine these values. This is particularly relevant if the productivity (e.g. meat or milk production per animal) changes significantly over time or when the national circumstances result in different values of parameters that have been used to calculate regional default IPCC emission factor values in the 2006 IPCC guidelines1.

(c) CH4 from landfills and wastewater (6A and 6B). More up-to-date country-specific information or estimates, such as of the amounts of MSW generated and the fraction landfilled, the waste composition and the Degradable Organic Carbon fraction, and their change over time, will improve the accuracy of the emission estimates.

(d) CO2 from large-scale biomass burning and deforestation and sinks from biomass growth (5) The uncertainty of this category could be reduced by using more detailed information. However due to the limited accuracy of the key parameters for the emissions and sinks calculation due to the variability in biomass types, their spatial distribution and the inherently limited knowledge of the extent of logging, burning and other forest degradation, will in general prevent making a quite accurate estimate of emissions and sinks. However, in case this source category is one of the largest key categories, more capacity to perform a more detailed assessment of changes over time will improve the emissions/sink estimates, albeit still rather uncertain.

-

1,000

2,000

3,000

4,000

5,000

6,000

Aus

tral ia

-UN

Aus

tral ia

-ED

GA

RFr

ance

-UN

Fran

ce-E

DG

AR

Ger

man

y-U

NG

erm

any-

EDG

AR

Ital

y-U

NIt

aly-

ED

GA

R

Japa

n-U

NJa

pan-

EDG

AR

Net

herl

ands

-UN

Net

herl

ands

-ED

GA

RR

ussi

a-U

N/5

Rus

sia-

ED

GA

R/5

UK

-UN

UK

-ED

GA

RU

nite

d St

ates

-UN

/5U

nite

d St

ates

-ED

GA

R/5

Gg

Sectoral CH4 emissions in 2005

6B

6A

4A

1B2

1B1

Figure 3. Comparison of sectoral CH4 emissions in 2005 between EDGAR 4.1 and

UNFCCC data for selected countries: 1B1 – coal mining, 1B2 oil and gas, 4A – animals, 6A – landfills, 6B - wastewater (unit: Gg) (Russia and USA *0.2)

1 Note that the uncertainty of indirect N2O emissions from agriculture cannot be reduced due to

the largely inherent uncertainty of this source category


224

-

50

100

150

200

250

300

Aus

tralia

-UN

Aus

tralia

-ED

GA

RFr

ance

-UN

Fran

ce-E

DG

AR

Ger

man

y-U

NG

erm

any-

EDG

AR

Ital

y-U

NIt

aly-

ED

GA

R

Japa

n-U

NJa

pan-

EDG

AR

Net

herl

ands

-UN

Net

herl

ands

-ED

GA

RR

ussi

a-U

N/2

Rus

sia-

ED

GA

R/2

UK

-UN

UK

-ED

GA

RU

nite

d St

ates

-UN

/4U

nite

d St

ates

-ED

GA

R/4

Sectoral N2O emissions in 2005

6B

4D3

4D2

4D1oth

4D11

4B

2B

1A3b

Figure 4. Comparison of sectoral N2O emissions in 2005 between EDGAR 4.1 and UNFCCC data for selected countries: 1A3b – road transport, 2B –chemical industry, 4B – animal waste (stables), 4D11 – synthetic fertilisers, 4D1-other – other direct soil emissions, 4D2 – pacture, range, 4D3 –

indirect N2O, 6B - wastewtaer (unit: Gg) (Russia *0.5 and USA *0.25)

Table 1. Variability in CO2 factors from fuel combustion reported by Annex I countries and comparison with IPCC default values in the 2006 guidelines. Unit: kg/GJ (LHV). Uncertainties expressed as 2 standard deviations (SD). (source: UNFCCC, 2009)

Fuel type Sector IPCC default EF

Unc. [%]

Unc. (low)

Unc. (high)

Average EF

reported

Stand. dev.

reported

Unc. (low)

Unc. (high)

Diffe-rerence

Diff- erence

coal residential sector 98.3 3.3 94.6 101.0 96.6 6.6% 83.9 109.3 -1.7 -1.7%

coal power generation a) b)

94.6 5.4 89.5 99.7 99.0 8.1% 82.9 115.1 4.4 4.7%

coal industry a) c) 94.6 7.2 87.3 101.0 99.5 22.9% 53.9 145.1 4.9 5.2%

natural gas

all sectors 56.1 3.6 54.3 58.3 58.4 19.0% 36.2 80.6 2.3 4.1%

petrol road transport 69.3 4.0 67.5 73.0 71.0 2.6% 67.3 74.7 1.7 2.5%

diesel road transport 74.1 1.5 72.6 74.8 73.5 0.8% 72.3 74.7 -0.6 -0.8%

a) Less reliable for hard coal, since coal-derived gases such as coke oven gas and blast furnace gas as well as brown coal can be included here (Annex I countries refer to “solid fuel”). This is much less so for the residential sector.

b) For IPCC default value for other bituminous coal was used.

c) For IPCC default value for coking coal was used.

Table 1 also provides another way to look at the uncertainty in using IPCC default emission factors, e.g. for CO2 from fuel combustion, is by assessing the spread in the values of country-specific emission factors and comparing the average of the country-specific values with the IPCC default value. This could only be done for a number of sectors and fuels for which the UNFCCC data refer to rather homogeneous fuel types.


225

Changes from 1996 to 2006 IPCC guidelines for GHG inventories

Please note that the emission factors used in EDGAR 4.1 are based on the 2006 IPCC guidelines, which may differ from the Revised 1996 IPCC guidelines. For many sources the changes are small, but for some, they can be significant. For CO2 emissions differences are due to the following reasons: • national energy statistics used may differ slightly due to updates included in more

recent releases, which may not be included in the data submitted to the IEA. For EDGAR 4.1 the release of 2007 was used (IEA/OECD, 2007);

• for the UNFCCC, if countries do not have country-specific emission factors, they will use the default CO2 emission factors from the Revised 1996 IPCC Guidelines, which differ slightly due to different default oxidation factors (coal updated value +2%, oil products +1%, natural gas +0.5%) and due to updated defaults for carbon content for some fuels of which the quality may vary considerably (mainly refinery gas, updated value -7%, coke oven gas -7%, blast furnace gas +7%, coke -1%);

• for CO2 from non-energy use or use of fuels as chemical feedstock countries may apply either higher tier methods using more country-specific information or calculate CO2 emissions from carbon released in fossil fuel use labelled in the sectoral energy balance as ‘non-energy use’ or ‘chemical feedstock’ using default fractions stored provided in the CO2 Reference Approach chapter. For EDGAR 4.1, default emission factors and methods from the 2006 IPCC Guidelines were applied, which may give rise to considerable differences compared to the 1996 guidelines. For indirect N2O emissions from atmospheric deposition of NH3 and NOx emissions

from agriculture as reported in EDGAR 4.1 are substantially lower than those presently reported by most Annex I countries due to two markedly lower emission factors compared to the values recommended in the 1996 IPCC Guidelines and the IPCC Good Practice Guidance (IPCC, 1997, 2000): • the default IPCC emission factor (“EF1”) for direct soil emissions of N2O from

the use of synthetic fertilisers, manure used as fertiliser and from crop residues left in the field has been reduced by 20%;

• the default emission factor (“EF5”) for indirect N2O emissions from nitrogen leaching and run-off been reduced by 70%. Thus our EDGAR 4.1 emissions can in some cases also be an indicator of how much

emissions may change if countries use IPCC default emission factors and change them to the defaults in the 2006 IPCC guidelines.

Conclusions

EDGAR inventories are of interest for both Annex I and non-Annex I countries. For the first group they provide a measure to see the impact of using higher tier methodologies and more country-specific emission factors and technology information versus using IPCC standard methodology readily applicable by using widely available statistics as activity data and default emission factors. In other words, how the uncertainty in their national emissions estimate has improved compared to the less detailed default estimate. For the latter group of developing countries they provide an estimate of recent trend and level of national greenhouse gas emissions and assist in identifying the largest sources.

Using uncertainty estimates based on IPCC default uncertainty values seems at first sight a rather crude method. However, since the uncertainty in the various sources differs so widely, the results are likely to provide a fair estimate of the uncertainty in total


226

emissions per gas at national, regional and global level. The difference of EDGAR and official greenhouse gas emissions of Annex I countries also indicates the applicability of the tier 1 IPCC methodology and default emission factors to developing countries (within the uncertainty estimates).

References IEA/OECD (2007, 2009). Energy Balances of OECD and Non-OECD Countries. On-line

data service. Internet: http://data.iea.org IPCC (1997). Revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories.

IPCC/OECD/ IEA, Paris. IPCC (2000). Good Practice Guidance and Uncertainty Management in National Greenhouse

Gas Inventories, IPCC-TSU NGGIP, Japan. IPCC (2006). 2006 IPCC Guidelines for National Greenhouse Gas Inventories. Eggleston, S.,

Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (eds.). IPCC-TSU NGGIP, IGES, Japan. Internet: http://www.ipcc-nggip.iges.or.jp/public/2006gl/index.html

Levin, I. et al. (2010). The global SF6 source inferred from long-term high precision atmospheric measurements and its comparison with emission inventories. Atmos. Chem. Phys., 10, 2655–2662, 2010.

Monzka, S.A. et al. (2010. Recent increases in global HFC-23 emissions. Geophys. Res. Lett., 37, L02808, doi:10.1029/2009GL041195.

Muhle, J. et al. (2010) Perfluorocarbons in the global atmosphere: tetrafluoromethane, hexafluoroethane, and octafluoropropane. Atmos. Chem. Phys., 10, 5145–5164, 2010.

Olivier (2002) On the Quality of Global Emission Inventories. Approaches, Methodologies, Input Data and Uncertainties. Thesis Utrecht University. Utrecht, Utrecht University. ISBN 90-393-3103-0. Internet: http://www.library.uu.nl/digiarchief/dip/diss/2002-1025-131210/inhoud.htm

Olivier, J.G.J. and J.A.H.W. Peters (2002) Uncertainties in global, regional and national emission inventories. In: Van Ham, J., A.P.M. Baede, R. Guicherit and J.F.G.M. Williams-Jacobse (eds.):‘Non-CO2 greenhouse gases: scientific understanding, control options and policy aspects. Proceedings of the Third International Symposium, Maastricht, Netherlands, 21-23 January 2002’, pp. 525-540. Millpress Science Publishers, Rotterdam. ISBN 90-77017-70-4.

Olivier, J.G.J., G. Janssens-Maenhout and J.A. van Aardenne (2010). Part III: Greenhouse gas emissions: 1. Shares and trends in greenhouse gas emissions; 2. Sources and Methods; Total greenhouse gas emissions. In: "CO2 emissions from fuel combustion, 2010 Edition”, pp. III.1-III.49. International Energy Agency (IEA), Paris (also available on CD ROM).

Rigby, M. et al. (2010) History of atmospheric SF6 from 1973 to 2008. Atmospheric Chemistry and Physics Discussions, Volume 10, Issue 5, 2010, pp.13519-13555

UNFCCC (2005). Sixth compilation and synthesis of initial national communications from Par-ties not included in Annex I to the Convention. Note by the secretariat. FCCC/SBI/2005/18, 25 October 2005. Internet: http://unfccc.int/resource/docs/2005/sbi/eng/18.pdf

UNFCCC (2009). Sources and availability of GHG data for non-Annex I Parties. Internet: http://unfccc.int/ghg_data/ghg_data_unfccc/data_sources/items/3816.php. Retrieved 31 July 2009

Villani, M.G. et al. (2010) Inverse modeling of European CH4 emissions: sensitivity to the observational network. Atmos. Chem. Phys., 10, 1249–1267, 2010.


227

Reducing uncertainties on carbon emissions from tropical deforestation: Brazil Amazon study case

Jean P. Ometto, Ana Paula Dutra Aguiar, Carlos A. Nobre

Earth System Science Centre (CCST) National Institute for Space Research (INPE)

Av. dos Astronautas, 1758 12227-010. São Jose dos Campos, SP, Brazil

[email protected]

Abstract

In the last year, Brazil announced the voluntary commitment to reduce its greenhouse gas emissions from 36.1% to 38.9% by 2020 and, to this end, such a commitment requires cutting down 80% of the deforestation in the Amazon rainforest until that year. Much of the uncertainty on the role of forests for carbon emissions is due to the lack of reliable deforestation data. The Brazilian’s National Institute for Space Research (INPE) carries on, since 1988, annual surveys of deforestation in the Amazon, an area of about 5 million km2. The existence of historic data on deforestation allows us to constrain the contribution of land cover change to greenhouse gases emissions since 40 years.

Our work estimate the carbon emission rates for the Brazilian Amazon, combining annual maps of new clearings and spatial information on biomass distribution for different vegetation types. The model also incorporates the temporal dynamics related to the deforestation process and its intraregional heterogeneity, including the percentage emitted by successive burning along the years and biological decay, the percentage of biomass used as timber, contribution of bellow ground emissions by root decay, and secondary vegetation growth.

Introduction

The increase rate of global green house gases concentration in the atmosphere has no parallel in the last one million year of the Earth history. From 1958 to 2004 the mean global CO2 emissions increased at the rate of approximately 1.3% per year, whilst the last 5 years (2004 to 2009) changes in emissions observed an increasing rate at the order of 3% per year (Global Carbon Project, www.globalcarbonproject.org). The IPCC AR4 [1] states that the tropical deforestation accounts to 10-20% of the global CO2, with the uncertainly on this range mostly due to biomass estimates and regional heterogeneity. The pattern of biomass burning emission follows different trajectory in recent years when compared to fossil fuel emissions, which shows an increasing curve. Because of that, and due to recent reduction in deforestation rates worldwide, the relative contribution of the land use component in the global carbon budget is decreasing [2]. In spite of this, the standing forests are major carbon stocks and play an important role in the climate, carbon sequestration, biodiversity, and local resources for indigenous communities [3].

The Amazon region in South America, as the largest continuous area of remaining rainforest in the World, has a critical role in the global carbon budget. The Brazilian Amazon alone contains more carbon stored in its biodiversity than the amount of global human-induced CO2 emissions of an entire decade [4]. A major question posed to the fate


228

of this ecosystem is how the carbon balance of tropical forests responds to rapid, on-going changes in climate and atmospheric composition [5], [6], [7]. [8] showed that the Eastern Amazon forest takes 70 years to recover the nitrogen dynamic back to its original state, after deforestation for agricultural purpose, with direct implications to the carbon balance in the ecosystem. Other important points to take in consideration are the deforestation rate, and biomass burning, in this region for the past 30 years. The use of well developed remote sensing techniques (INPE, Brazil) to monitor and calculate the extent of tropical deforestation has considerably improved our ability to estimate rates of this process in these regions, what is certainly going to become an indispensable auditing tool for implementation of mitigation strategies.

This paper brings the results of a mathematical model proposed in order to systematize the calculation of annual green house gases emission from deforestation in the Brazilian Amazon. The auspice of this effort was construct under the voluntary compromise of emission reduction that the Brazilian Government has committed up to 2020 (see Figure 1), once land use change and deforestation are the main sources of GHG in the Brazilian emissions profile. According to [9] and the Brazilian Ministry of Science and Technology (MCT), the contribution of the Amazon deforestation to the country total emissions reaches 55%, summing to 75% when considering agriculture and cattle ranching.

The model here described considers, first, the biomass distribution according to vegetation maps overlapping with the deforestation-monitoring maps, produced annually by the National Institute for Space Research (PRODES System, [10]); the dynamic of the deforestation process, as well as its interregional peculiarities. Other components of the model logic are: the speed and frequency that trees are cut and burned; the percentage of timber extracted; equations of organic matter decay in soils; root decomposition; and emissions factors for different gases species (CH4, N2O, CO2) which are associated to fire intensity and quality of organic matter. The secondary vegetation re-growth and its dynamic are also considered.

Land-use activities—whether converting natural landscapes for human use or changing management practices on human-dominated lands—have transformed a large proportion of the planet’s land surface. By clearing tropical forests, practicing subsistence agriculture, intensifying farmland production, or expanding urban centres, human actions are changing the landscape in pervasive ways [11]. In Brazil the National Institute for Space Research (INPE) has a historical series of satellite mapping of deforestation in the 5 million km2 of the Brazilian Amazon region, identifying 16% of forest loss until 2007 [10], [12], [13]. The annual data, produced since 1988, show a decreasing rate from 2004 (Figure 1), however the total area deforested in the 2005-2009 periods is still important, reaching 65,000 km2.

Most of the degradation processes are concentrated in the southern and eastern parts of Amazônia (Figure 2). Central and other less accessible areas are still relatively well protected, opening great possibilities for conservation initiatives and carbon fluxes monitoring activities. Recent plans from the Brazilian Government for paving roads and developing infrastructure, with increasing presence of highly capitalized agribusiness companies, can pose a high threat to central areas in a very near future. In fact, an increase of 25–40% in Amazon deforestation due to projected road paving could counterbalance nearly half of the reductions in C emissions that would be achieved with the Kyoto Protocol [4]. Typical Amazon forests contain, on average, around 250 tons of biomass per hectare, thus deforestation in Amazônia alone, can release from 500 to 900 Mton of CO2 annually, 2–4% of world emissions, according to [14], [15], [16]. [17]


229

observed that depending if the regrowth dynamic of the secondary vegetation is considered the total carbon emission might vary significantly. Complimentary, [6], analysing data produced for the Amazon region, showed a broad range of values of emissions estimates, depending on the region, vegetation, or estimate method considered.

Figure 1. Deforestation rate and reduction simulation (Brazilian Ministry of Science

and Technology, 2009)

Figure 2. Brazilian Amazon map indicating (i) non forested areas (pink); (ii) deforestation after 1997 through 2004 (yellow and orange); (iii) forest; clouds (light blue).


230

Aside to the Amazon, the central savanna in Brazil, named Cerrado, represents 23% of the land surface, spreading across 2,031,990 km² of the central Brazilian Plateau. The most extensive woodland/savanna region in South America and second major Brazilian biome, after the Amazon (Figure 2), the Cerrado is also the only hotspot that consists largely of savanna, woodland/savanna and dry forest ecosystems. Within the region, there is a mosaic of different vegetation types, including tree and scrub savanna, grassland with scattered trees, and occasional patches of a dry and closed canopy forest. The biome, however, has been subjected to rapid rates of land conversion to agriculture and pasture, with important consequences to local change in microclimate and carbon fluxes. Moderate Resolution Imaging Spectroradiometer (MODIS) composite vegetation index has being used in the region to analyse the seasonal patterns of photosynthetic vegetation activity and examine the potential separation of Cerrado formations. The Cerrado formations exhibited a high seasonality contrast with a pronounced dry season from June through August and wet season from November to March. The converted agricultural areas had a higher contrast than the native Cerrado, and the forest formation had the lowest seasonal contrast, showing a potential use of remote imagery to monitor and quantify the carbon stocks and dynamic in the region. In spite of emissions data from deforestation in Cerrado not being presented in this paper, the structure of the model can be adjusted to different regions; land use and land cover characteristics.

Results

The metodology was based on previous work by [14], [18], [19], [20], [21], however including new components looking to incorporate a better representation of the deforestation process itself. Inter-regional differences were represented by biomass and deforestation maps [10], [22], the structure of the agriculture production and the economical development frontier.

The modelled results indicate a reduction of emission in recent years (2007-2008) from a mean value at the range of 700-800 Mton CO2 per year (considering the period from 1999 to 2008) to 500-550 Mton CO2 per year. The dynamic of the secondary vegetation need to be better understood and regionalized. According to our calculation the contribution of the re-growth of deforested vegetation does not contribute significantly to the emissions reduction, overall because, according to [23], the residence time of the secondary vegetation is relatively short (circa of 5 years), after which the vegetation is cut and burned again. Figure 3 presents the amount of emissions by: (i) model results (red line) and (ii) the emissions without the dynamic processes represented by the model (blue line); as well as, (iii) the differences among carbon sequestration by the secondary vegetation (yellow line) and (iv) the emissions due to the deforestation of this vegetation (green line).

The model predicted, over the deforestation reduction target Brazil has proposed, 600 to 650 Mton of CO2 of avoided emissions until 2020. However on considering a non-managed re-growth of secondary vegetation these numbers can get up to 740 Mton of CO2.

The model results highlight the importance of regional differences on estimating emissions from tropical forests deforestation, as well as including the secondary forest dynamic as potential to recuperate regional ecosystem, but also to mitigate historical emissions from deforestation. The dynamic of the deforestation process itself is also


231

critical to be considered in emissions models, once depending on the activity replacing the original forest the clear cut deforestation might be concluded in few months or years. Information’s of this nature are critical to mitigation mechanisms as REDD, for instance. Considering the global balance published by [2], our results place deforestation in the Brazilian Amazon contributing to 1.1 to 1.9% of the total global carbon emissions.

Figure 3. Emission Model results

References

[1] Intergovernmental Panel on Climate Change (2007). www.ipcc.ch/publications_and_data/publications_and_data_reports.htm

[2] Le Queré, C. Et al. (2009). Trends in the sources and sinks of carbon dioxide. Nature Geoscience, DOI: 10.1038/ngeo689

[3] Laurance, W. F., H. L. Vasconcelos, and T. E. Lovejoy. 2000. Forest loss and fragmentation in the Amazon: implications for wildlife conservation. Oryx 34:39–45

[4] Carvalho, J.A.; Costa, F.S.; Veras, C.A.G.; Sandberg, D.V.; Alvarado, E.C.; Gielow, R.; Serra, A.M.; Santos, J.C., Biomass Fire Consumption and Carbon Release Rates of Rainforest-clearing Experiments Conducted in Northern Mato Grosso, Brazil, J. Geophysical Research, 106(D16), 17877-17887, 2001.

[5] Gash JHC, Huntingford C, Marengo JA, Betts RA, Cox PM, Fisch G, Fu R, Gandu AW, Harris PP, Machado LAT, von Randow C, Silva Dias MA (2004) Amazonian climate: results and future research, Theor. Appl. Climatol. 78, 187-193, LBA Special Issue

[6] Ometto JPHB, Nobre AD, Rocha HR, Artaxo P, Martinelli LA. Amazonia and the Modern Carbon Cycle: Lessons Learned (2005) Oecologia. DOI: 10.1007/S00442-005-0034-3

[7] Gloor M, Phillips, O. L., Lloyd, J. J. et al. (2009). Does the disturbance hypothesis explain the biomass increase in basin-wide Amazon forest plot data? Global Change Biology, 15: 2418-2430

[8] Davidson E.A., C. J. R. de Carvalho, A.M. Figueira, F.Y. Ishida, J.P.H.B. Ometto, G.B. Nardoto, R.T. Sabá, S.N. Hayashi, E.C. Leal, I.C.G. Vieira and L.A. Martinelli.


232

(2007). Recuperation of nitrogen cycling in Amazonian forests following agricultural abandonment. Nature, v.447, doi:10.1038/nature05900

[9] Cerri et al (2009). Brazilian Greenhouse Gas Emissions: The Importance Of Agriculture And Livestock. Sci. Agric. (Piracicaba, Braz.), v.66, n.6, p.831-843, November/December 2009

[10] Instituto Nacional de Pesquisas Espaciais (INPE) (2009), PRODES: Assessment of Deforestation in Brazilian Amazonia, Natl. Inst. for Space Res, São Jose´ dos Campos, Brazil. www.inpe.br/prodes

[11] Foley, J A, Defries, R, Asner, G P, et al (2005). Global consequences of land use change. Science. 390: 570-574

[12] Almeida, A.; Stone, T. ; Vieira, Ima Célia Guimarães ; Davidson, E. (2009). Non-Frontier deforestation in the Eastern Amazonia. Earth Interactions.

[13] Fearnside PM (2000). Global warming and tropical land use change: Greenhouse gas emissions from biomass burning, decomposition, and soils in forest conversion, shifting cultivation, and secondary vegetation. Climatic Change 46: 115-158.

[14] Loaire et al (2009), Boosted carbon emissions from Amazon deforestation. GEOPHYSICAL RESEARCH LETTERS, VOL. 36, L14810, doi:10.1029/2009GL037526, 2009

[15] Potter, C., Klooster, S., Genovese, V. (2009). Carbon emissions from deforestation in the Brazilian Amazon Region. Biogeosciences, 6: 2369-2381

[16] Santilli, M., Moutinho, P., Schwartzman, S., et al. (2005) Tropical deforestation and the kyoto protocol. Climatic Change, 71: 267–276

[17] Hirsch AI, Little WS, Houghton RA et al. (2004) The net carbon flux due to deforestation and forest re-growth in the Brazilian Amazon: analysis using a process-based model. Global Change Biology, 10, 908–924.

[18] Houghton, R. A., et al. (2000), Annual fluxes of carbon from deforestation and regrowth in the Brazilian Amazon, Nature, 403(6767), 301 – 304, doi:10.1038/35002062.

[19] Houghton, R. A. (2003), Why are estimates of the terrestrial carbon balance so different?, Global Change Biol., 9(4), 500–509, doi:10.1046/j.1365-2486.2003.00620.x.

[20] Houghton, R. A. (2005), Aboveground forest biomass and the global carbon balance, Global Change Biol., 11(6), 945 – 958, doi:10.1111/ j.1365-2486.2005.00955.x.

[21] Ramankutty, N., et al. (2007), Challenges to estimating carbon emissions from tropical deforestation, Global Change Biol., 13(1), 51 – 66, doi:10.1111/j.1365-2486.2006.01272.x.

[22] Saatchi, S. S., et al. (2007), Distribution of aboveground live biomass in the Amazon basin, Global Change Biol., 13(4), 816– 837.

[23] Almeida, C. (2009). Estimativa da área e do tempo de permanência da vegetação secundária na Amazônia legal por meio de imagens Landsat/TM .INPE, 2009. 130p. (INPE-15651-TDI/1429)


233

Application of spatio-temporal emission-factors (STEFs) for carbon footprinting of Indian coastal zones

J.S. Pandey, R. Kumar, S.R. Wate, T. Chakrabarti

National Environmental Engineering Research Institute (NEERI) [Council of Scientific and Industrial Research (CSIR)]

NAGPUR – 440020, India

[email protected]

Abstract

It is increasingly felt today that scientists should move beyond the boundaries of their own disciplines. Especially when it concerns issues and impacts related with global environmental and climatic perturbations. Knowledge of ecological, socio-political, and several economic policies is essential when one requires to grow a sound understanding of ecosystems. Extrapolating the GHG-emissions measured at selected parts and selected intervals of time would require development of appropriate region-specific Spatio-Temporal Emission-Factors (STEF). The present paper discusses and illustrates the issues connected with STEF, which help in quantifying region-specific and process-specific Ecological and Carbon Footprints (EF & CF). These are the issues on which we should focus our immediate attention and continuously work on their future refinement. Need for spatial as well as temporal extrapolations of STEFs to global scales are also discussed In particular, the article looks into the issues related to methane emission potential (MEP) of many important Indian Coastal Regions. In the anoxic sediments of wetlands and mangroves, methanogenesis occurs in the presence of high concentrations of organic material resulting into methane-emissions. These emissions may differ significantly from one geographical region to another and depend, inter alia, on various bio-physical and bio-chemical parameters. In addition, there may be diurnal, seasonal and annual variations too. Moreover, the main scientific debate at the moment is centered around the uncertainties associated with estimations of EF & CF. Development of STEFs will not only reduce the spatial uncertainties but also the uncertainties associated with diurnal, seasonal and annual variations and would ultimately help in evolving site-specific, region-specific and ecosystem-specific environmental management plans aimed at combating the climate regulated environmental crisis.

Key Words : Spatio-Temporal Emission-Factors (STEF), Ecological Footprint (EF), Carbon Footprint (CF), Methane-Emission, Coastal Zones

Introduction

Any developmental activity (technology, industry or infrastructure) undergoes a feasibility study or a cost-benefit analysis before its adoption. Most of the feasibility studies have an inadvertent economic bias. But at the same time one should not lose sight of the inherent ecological costs in the process. Impact of global warming and climate change are the ecological costs, one is likely to entail. Often times, while the technology development brings about a local or regional economic benefit, the ecological costs (environmental impacts) to be paid are at the global scales. Thus, it should be only after a thorough ecological analysis that a technology should be adopted or an infrastructure developed or an industry installed. For doing this kind of cost-benefit analysis, recent


234

techniques like Ecological Foot-printing (EF) (Rees and Wackernagel, 1999) and Carbon Foot-printing (CF) (en.wikipedia.org/wiki/Carbon_footprint) etc. should be taken recourse to (Pandey et al. 2009; Pandey et al. 2001a; Pandey et al. 2010).

However, these techniques as also the conventional Environmental Impact Assessment (EIA)-Process have many inherent limitations including those of data availability, appropriate model selection and their (models’) application. This is mainly because of inherent data limitations. As far as environmental as well as climate models are concerned, scientific literature is replete with them. Be it air quality, water quality, soil quality or for depicting meteorological, biological, ecological and socio-economic processes, a large number of models have been developed. While developing all encompassing and comprehensive models and applying them may itself be a very difficult task, the more daunting task is to obtain reliable data sets for all the parameters included in these models. Therefore, often times we have to rely on many approximations and inherent assumptions. This inevitably introduces errors and uncertainty in the analysis. In short, either the data-available is not in a systematic pattern or it does not have the desired/required resolution. Under the circumstances, there is a strong need for developing such models that are based on realistic and practically available data-sets and are more appropriate (Pandey et al., 2010).

Coastal environment plays a very significant role in any nations economy and infrastructure development. It is a huge reservoir of physico-chemical and biological resources. Indian coastline measures up to approximately 7500 kms. They contain flora and fauna, store nutrients in sediments and maintain the ecological food-chains. In association, they provide several food items, ecological goods and services (mangroves, fish, oil, gas, minerals, tidal energy, recreation, transportation, ports, industries etc.). They also serve as a habitat for genetically, ecologically and economically valuable organisms. However, these habitats are continuously and consistently being threatened because of ever rising air, water and land pollution. The major activities that are responsible for coastal pollution are discharge and disposal of untreated domestic and industrial wastes in the form of gaseous emissions, liquid effluents and sludge, discharges of coolant waters, harbour activities such as dredging, cargo handling, dumping of ship wastes, spilling of cargo’s chemicals and aquaculture.

The present exercise is an attempt to look into the environmental issues connected with ecologically vulnerable coastal zones and estimate their CF initially in terms of methane emissions by applying appropriate models. Many STEFs have earlier been developed by us (Pandey et al., 2010) on the basis of global data available. These models interlink BOD with methane emissions, carbon footprint (CF) with ecological footprint (EF), waste generated etc. Development of these kinds of models results in reducing not only the spatial uncertainties but also the uncertainties associated with diurnal, seasonal and annual variations and would ultimately help in evolving site-specific, region-specific and ecosystem-specific environmental management plans aimed at combating the climate regulated environmental crisis. The present modeling exercise estimates the methane emission potentials (MEP) of various ecologically sensitive Indian coastal regions.

This will be an important contribution towards designing an Integrated Coastal Zone Management Plan as they (Coastal Zone Management Plans) have the most important objective in terms of conservation, preservation and judicious utilization of coastal resources on a sustainable basis so as to uplift the socio-economic conditions of the local


235

people without causing adverse impact on the environment. Determination of waste allocation capacity (WAC) is an important step towards integrated coastal zone management.

Methane Emissions

Recently, methane has gained a special importance because of its global warming potential being as high as 23.5 times (more than carbon dioxide). Side by side, there are many controversies and uncertainties related to methane emissions from hydroelectric dams (Fearnside, 2004; Rosa, 2004; Pandey et al., 2007). Atmospheric concentration of methane has almost doubled after industrial revolution. In industrial countries, 15% of total GHG contribution comes as methane emission (http://ec.europa.eu/environment/ climate/campaign/pdf/gasesen.pdf) and it is expected to contribute to 18% of the total expected global warming over the next 50 years. Along with the warming effect, methane also participates in troposphere ozone formation, which amplifies methane's direct infrared absorption by approximately 70% (Milich, 1999).

Methane is also emitted by the water bodies, which remain stagnant for long time. This includes waste water. 10% of total methane emissions come from industrial and municipal waste water (http://grida.no/climate/ipcc_tar/wg3/120.htm). Anaerobic bacteria are the main factors behind methane emission. Apart from ‘Temperature’ (T), ‘Biochemical Oxygen Demand’ (BOD) and ‘Chemical Oxygen Demand’ (COD) are other two important water-quality parameters on the basis of which methane emission from a particular water-body can be determined. Municipal and industrial wastewater having higher BOD or COD values emit more methane under the similar climatic conditions. Organic fractions present in the municipal wastewaters are degraded to produce methane (Fadel and Masood, 2001; Mishra et al. 2008a,b; Tembhare et al. 2008; IPCC, 2006; http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol5.html).

Methodology for the Present Estimates

We made a comparison between methane emissions from wastewater in various developed and developing countries (USEPA, 2005) and observed that over the years there has been continuous increase in methane emission from wastewater in most of the countries. For instance, in America, methane emission has increased from 24.85 million MT (CO2-eq.) (in 1990) to 35.21 million MT (CO2-eq.) (in 2005). During the same time in India, this increase was from 56.90 million MT (CO2-eq.) to 73.25 million MT (CO2-eq.) respectively (http:// www.epa.gov/methane/ pdfs/global_emissions.pdf). As a first step in our analysis we ranked various countries according to their relative contributions in terms of per capita methane emissions. Subsequently, we correlated these values (Figure 1) with per capita BOD emissions from the select countries (from the information available) for investigating the correlation between Per Capita BOD Per day and Total Methane Emission (Gg/year) (Pandey et al. 2010).

Data from some important coastal regions of India (Paradip, Vishakhapatnam, Tuticorin, Kochi and Mangalore) was selected (http://infochangeindia.org / Coastal-Night-mares / Coastal-cities-need-to-clean-up-their-act.html) for the present analysis (Table 1). Based on their population and sewage discharges BOD-loading rates (g per capita per day) has been estimated with the assumptions of 530 mg BOD per litre of


236

sewage water (http://www.unep.or.jp / itec / Publications / Tech Publications / TechPUb – 15 / 3-2 Asiawest / 2-1.asp). Subsequently, annual methane emission potentials (MEP) have been estimated (Figure 1) on the basis of correlations depicted in Pandey et a. (2010).

Table 1

S. No.

Location BOD g / (c * d) [g per capita per day]

Methane Emissions Gg/(year)

1. Paradip 46.6 75.839 2. Visakhapattnam 27.1 38.715 3. Tuticorin 23.19 31.273 4. Cochin 31.8 47.662 5. Mangalore 29.68 43.626

Gross Methane Emissions (Gg/y)

3442

62

85y = 34.179e0.016x

R2 = 0.486

y = 1.9035x - 12.869R2 = 0.5987

y = 0.1045x2 - 10.579x + 317.57R2 = 0.9334

0

20

40

60

80

100

120

140

160

180

200

0 20 40 60 80 100

BOD (g per person per day)

Met

hane

Em

issi

ons

(Gg/

y)

Figure 1

Results and Discussion

Three kinds of trend-lines have been obtained for the gross methane estimation, linear, exponential and polynomial (Figure 1). Although the polynomial model shown in our earlier paper (Pandey et al., 2010) has the highest correlation (R2 = 0.9334), it sometimes gives misleading (unacceptable) results. For example, it can give higher methane emissions for lower values of BOD. Hence, in the present analysis, only the linear model has been used for estimating methane emitting potentials (MEPs) of the selected coastal regions.

These estimates (Figures 2 through 6) need to be validated with many site-specific measures. The greatest uncertainty in these kinds of extrapolation (estimation) exercises creeps in from the fact that all the micro-sites under the study (given) area are supposed to have the similar emission-trends, which may not be the case in reality. Moreover, methane emissions depict only a portion of the gross CF. For estimating total CF one has to also include emissions of other GHGs. The assumption of 530 mg BOD per lire of sewage water may also introduce some uncertainty in the final estimates as it overlooks site-specific details of BOD-generation. Other source of uncertainty can be related to the population estimates used for the analysis.


237

6.60.75

46.6

75.839

0

10

20

30

40

50

60

70

80

PARADIP

Sewage (MLD)Population (Lakhs)BOD [g /(c*d)]Methane (Gg/Year)

Figure 2

68

13

27.1

38.715

0

10

20

30

40

50

60

70

VISAKHAPATTNAM


Figure 3

17.5

4

23.19

31.273

0

5

10

15

20

25

30

35

TUTICORIN


Figure 4


238

36

6

31.8

47.662

0

5

10

15

20

25

30

35

40

45

50

COCHIN


Figure 5

28

5

29.68

43.626

0

5

10

15

20

25

30

35

40

45

MANGALORE


Figure 6

Conclusion

This article presents an important step (way forward) in the direction of appropriate Climate Change Research needed at the moment, especially those pertaining to the Integrated Coastal Zone Management Plans. However, these researches need to be extended and pursued further if we really want to strike a balance between ecology and economy. Future exercises are needed, which should aim at the dynamics of Ecological Footprints (Pandey et al., 2001a); analysis of Environmental Risks by way of developing models which deal with the issues like Temporal Risk Gradients (TRG) (Pandey et al., 2001b) and Ecological Economics of Natural Resources (Pandey et al., 2004). Similarly, apart from studying the models which represent methane emissions, there is also a need to quantify total Carbon Footprint (CF) (http : // www.up.ethz.ch / education / term_paper / termpaper_hs07 / Farrer_rev_termpaper_hs07.pdf). These exercises will be extremely useful from the Coastal Zone Management and Vulnerability point of view, especially when it pertains to quantification of issues like Waste Allocation Capacity (WAC) etc.

Disclaimer

The present paper has been brought out on the basis of a systematic data analysis and appropriate synthesis of the information available from various sources.


239

Subsequently, this information has been used for appropriate quantifications. The likely sources of uncertainty in the estimates have also been discussed. The views expressed are those of the authors’ and the Institution (NEERI) may or may not share the same views.

Acknowledgement

Apart from the staff members who have directly helped in the preparation of this manuscript, the authors are also grateful to all those who have indirectly contributed in various efforts including that of collecting information, analysis and interpretation of results.

References

Fadel ME, Massoud M (2001) Methane emissions from wastewater management Environmental Pollution 114:177-185.

Fearnside, P.M. 2004. Greenhouse Gas Emissions from Hydroelectric Dams : Controversies Provide a Springboard for Rethinking a Supposedly ‘Clean’ Energy Source : An Editorial Comment. Climate Change 66 : 1-8.

IPCC. 2006. IPCC In : Guidelines for National Greenhouse Gas Inventories. Available at http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol5.html

Milich L (1999) The role of methane in global warming: where might mitigation strategies be focused? Global Environmental Change 9:179-201

Mishra AP, Pandey JS, Wate SR (2008a) Municipal wastewater, methane emissions and global perspective. Souvenir : International Conference on Water Crisis - Challenges and Opportunities, 28-29 February, NEERI, Nagpur, India, p ww – 24.

Mishra AP, Tembhare M.W., Pandey JS, Kumar R, Wate SR (2008b) Carbon footprint : Where India stands in global Scenario. Souvenir : International Conference on Recent trends in Environmental Impact Assessment (RTEIA – 2008), NEERI, Nagpur, India, November 23-25, p 34

Pandey, J.S., Khan, S., Joseph, V. and Singh, R.N. 2001a. Development of a Dynamic and Predictive Model for Ecological Footprinting (EF). Journal of Environmental Systems 28 (4) : 279-291.

Pandey, J.S., Khan, S. and Khanna, P. 2001b. Modeling and Quantification of Temporal Risk Gradients (TRG) for Traffic Zones of Delhi City in India. Journal of Environmental Systems Vol. 28(1) : 55-69.

Pandey, J.S., Joseph, V. and Kaul, S.N. 2004. A Zone-wise Ecological-Economic Analysis of Indian Wetlands.Environmental Monitoring and Assessment 98 : 261-273.

Pandey, J.S., Wate, S.R. and Devotta, S. 2007. Development of emission factors for GHGs and associated uncertainties. PROCEEDINGS : 2nd International Workshop on Uncertainty in Greenhouse Gas Inventories. International Institute for Applied Systems Analysis, A-2361 Laxenburg, Austria, 27-28 September, 2007.

Pandey, J.S. 2009. Inter-disciplinarity of Issues Connected with Climate Change, Food Security and Energy Alternatives. The International JOURNAL of CLIMATE Change Impacts and Responses, Volume1 (In Press)


240

Pandey, J.S., Kumar, Rakesh and Wate, S.R. 2009. Ecological and Carbon Footprints (EF & CF), Climate Change and Associated Issues

(http://aaarabstracts.com/specialty/viewabstract.php? Paper= 40)

Pandey, J.S., Kumar, R., Wate, S.R. and Chakrabarti, T. 2010. Methane Emissions from Wastewater, Wetlands, Mangroves and Hydroelectric Dams : Developing Appropriate Emission Factors for Region-specific GHGs (ERG). Asia-Pacific Business Review, Vol. VI (1) : 29-41.

Rees, W.E. and Wackernagel, M. 1999. Monetary Analysis : Turning a Blind Eye on Sustainability. Ecological Economics 29 : 47-52.

Rosa, L.P., Dos Santos, M.A., Matvienko, B., Dos Santos, E.O. and Sikar, E. 2004. Greenhouse Gas Emissions from Hydroelectric Reservoirs in Tropical Regions. Climate Change 66: 9-21.

Tembhare M, Mishra AP, Pandey JS, Wate SR (2008) Methane emission from wastewater treatment : A comparative analysis. International Conference on Water Crisis-Challenges and Opportunities, 28-29 February, NEERI, Nagpur, India, p ww 21.


241

Changes in European Air Emissions 1970 – 2010: decomposition of determining factors

Peter Rafaj1, Markus Amann1, Henning Wuester2

1 International Institute for Applied Systems Analysis (IIASA), A-2361, Laxenburg, Austria 2 United Nations Framework Convention on Climate Change (UNFCCC), Bonn, Germany

[email protected]

Abstract

This paper presents an in-depth analysis of the factors that have caused the observed evolution of SO2, NOx and CO2 emissions in Europe within the time period 1970 to 2010. Historic energy balances are used together with population and economic growth data to quantify impacts of major determinants contributing to changes in emission levels: energy intensity, conversion efficiency, fuel mix, and pollution control. Time series of historic emission levels are distinguished for countries in West Europe and East Europe showing differences in the importance of emission driving forces. The declining trend in SO2 emissions in West Europe results by three-quarters from combined reduced energy intensity and improved fuel mix, while for reducing NOx emissions dedicated end-of-pipe abatement measures play a dominant role. An increase in air emissions in East Europe until the mid 90’s is associated with the growth of energy intensive industries that overweight the positive impact of better fuel quality and fuel mix. Continuous decrease in energy intensity and higher conversion efficiency are the main determinants that keep the growth in European CO2 emissions at moderate rates.

Keywords: Air pollution, CO2, decomposition analysis, historic emissions, international policies.

1. Introduction

Environmental policies have shown to be effective in reducing air pollution over the last two decades. Sulphur emissions, for instance, have reduced dramatically since the seventies. While much of the emission reduction is attributed to targeted abatement policy, some of it seems simply the result of structural changes in the economy, for instance reducing the size of heavy industry, or changes in energy consumption and fuel mix. Measures taken to reduce emissions and structural changes are likely to be affected by a country’s economic development and wealth. Abatement policies may, however, be primarily a reaction to perceived environmental problems and only to a lesser extent depend on economic parameters. National measures to reduce emissions may target national problems, but, in the context of transboundary pollution, they may also be influenced by measures in other countries and international policies.

The adoption of the Gothenburg Protocol to Abate Acidification, Eutrophication and Ground-level Ozone (UNECE, 1999) and the work for the related National Emission Ceilings Directive of the European Union (EC, 2001; EC, 2005) provides a clear perspective for European air pollution policies. On this basis it is possible to examine changes in emissions to gain a better understanding on their determinants. The Gothenburg Protocol sets emission ceilings for sulphur, nitrogen oxides (NOx), ammonia and volatile organic compounds (VOCs) for 2010 by relating emission changes to 1990 levels. This study investigates changes in SO2 and NOx emissions extending the period covered by the Protocol backwards to 1970 and up to 2010. In the perspective of recent climate change debate and adoption of Europe's Climate and energy package


242

(EC, 2008), it is also relevant to examine factors that drive changes of European CO2 emissions. The availability of underlying data starting from 1970 until present allows the investigation of long-term trends.

A debate about the main determinants of environmental improvements has been led in the economic literature under the heading “environmental Kuznets curve” (EKC). The Kuznets curve refers to the inverted U-shaped relationship between environmental quality and per-capita income (Kuznets, 1955). Most of the literature refers to SO2; few studies investigated other emission species. The analyses, however, does not take sufficiently account for all specific aspects of the environmental problems examined. A refined analysis of the determinants of changes in air emissions and their environmental effects may help to understand to what extent and through which mechanism the level of economic development impacts environmental quality and the role that other determinants play.

The objective of this study is to identify the main factors responsible for reductions of air emissions and related environmental improvements. The analysis contributes to the evaluation of the relative importance of the main driving forces for atmospheric protection policies: welfare (as suggested by EKC), awareness of the harmful environmental impacts of pollution (as favoured by the natural sciences), or technological and structural factors. The paper also draws conclusions about policy measures and other developments that may influence air emissions in the future. A better understanding of whether and why there is a decoupling of pollution from economic growth in Europe once certain level of economic wealth has been achieved provides insights that could be useful for other regions (e.g. Asia) and other pollutants (e.g. black and organic carbon).

2. Methodology and data sources

The protocols under the Convention on Long range Transboundary Air Pollution (CLRTAP) or United Nations Framework Convention on Climate Change (UNFCCC) provide a policy basis for targeted multi-pollutant abatement of emissions in most European countries. In order to determine how much these international policy strategies impact emissions, it is necessary to estimate what reductions actually result from dedicated mitigation measures. To do this, it is useful to distinguish between the consequence of changes in the energy structure, overall economic changes and technological advances. Analyzing the differences between regions and countries may help to understand why emission reduction rates differ. If international policy efforts were important, they would tend to reduce the differences between countries, at least those that become Party to the protocols and with respect to the importance of control measures.

To quantify the relation between targeted abatement measures and autonomous emission changes, an analysis of the factors leading to emission changes is performed. In a first step, the relationship between these factors is clarified. Equations derived are then applied to the underlying databases in order to quantify the importance of the different factors in two larger European regions.

2.1. Determinants of emission changes

In a compact notation, the emission changes relative to a selected base year can be defined as

( ) X*eff*ENE

EMIS**

GDP

ENE*GDPEMIS ∆−

∆η∆

∆=∆ 1 .


243

The following four factors can be distinguished: 1. Energy intensity effect that is due to change in the energy (ENE) requirements for

a unit of gross domestic product (GDP). Changes in energy intensity determine the overall energy consumption and drive the resulting emission levels (EMIS). Time evolution of differences in energy intensity across countries reflects variations in socio-economic structure as well as behavioural patterns.

2. Efficiency of the energy system, which comprises changes in the conversion efficiency (η) at different level of the energy flow, i.e., improved efficiency of converting primary fuels into electricity, efficiency of combustion of final energy carriers in the industry, transport or household sectors, and finally the efficiency of end-use devices such as vehicles or light bulbs. Efficiency improvements are either mandated by regulations or they are motivated through fuel availability and prices.

3. Fuel mix composition affects substantially the emission intensity and involves temporal inter-fossil fuel switching and a change in non-fossil fraction in the energy supply. Also substitution of traditional fuels with electricity and heat belongs to this mitigation component. Fuel switches are introduced through environmental regulations, cost minimization and convenience, but they are conditional to the accessibility of energy grids and infrastructure.

4. Emission control measures reduce exhausts emitted per unit of energy consumed through end-of-pipe measures, as well as though improved fuel quality due to, for example, lowering sulphur content of coal or heating oil. Changes in the emission factors over time can be in addition influenced by the modified import patterns and by exploration of resources with different characteristics. The resulting emission coefficient depends on the removal efficiency (eff) of an abatement measure adopted at a specific application rate (X).

The factors determining trends in emission levels are listed above in a particular order, which follows the sequence of their implementation in respective sectors. The decomposition allows identifying the factors that are most influential for emission developments. First, the upper end of emissions is calculated for the hypothetical case in absence of any emission reduction. This emission path results from the growth in overall energy consumption at the constant energy intensity of GDP and unchanged fuel mix, i.e., it represents the energy consumption effect due to GDP growth. In the following step, changes to the energy intensity and efficiency improvement are accounted in the emission balance while keeping the fuel mix of the base year fixed across the whole computation period. Next, impact of the fuel substitution is calculated by unfixing the share of fuels, but the emission factors are kept constant over time. Finally, contribution of control measures to the emission reductions is derived from time dependent emission factors and through comparison to the reported historic emission data.

2.2. Data used

This analysis makes use of combined information from different statistics, databases and models for calculating emission changes within the 40-year period staring from 1970 to 2010 with a 5-years step. Geographical coverage of calculations distinguishes two groups of counties under examination: a) West Europe (WEU) comprising 15 countries with the European Union (EU) membership before 2004 together with Switzerland and Norway; b) East Europe (EEU) including 12 new members that joined EU after 2004, Balkan countries, Turkey, Belorussia, Ukraine and Moldova. Remaining European countries, i.e., Iceland and Russian Federation were not included in the analysis reported herein.


244

SO2, NOx and CO2 emissions originate almost entirely from combustion of energy carriers, therefore the emission estimates are primarily based on statistical energy data and fuel balances for combinations of 9 fuel categories utilized in 5 sectors. For most of the countries, energy statistics from International Energy Agency (IEA 2009a, 2009b) have been used for the years 1970 to 2005. Energy consumption in the year 2010 is based on the projections developed for the Baseline scenario for revision of the National Emission Ceiling Directive and implemented within the GAINS model (Capros et al., 2008).

The same data sources were used for extracting information on autonomous factors that contribute to the emission reduction, such as GDP, energy intensity and population growth. On the other hand, the available databases do not provide sufficient material that would allow a full quantification of the mitigation role of efficiency improvements. Therefore, two factors defined in Section 2.1. – energy intensity and efficiency improvements – are aggregated within one term that covers both processes.

The respective emission factors for the base year 1970 and consecutive years have been extracted from databases of RAINS and GAINS models (Klaassen et al., 2004), and if necessary, these coefficients were adjusted such that resulting emissions match the official national estimates reported to European Monitoring and Evaluation Programme (EMEP) under CLRTAP (EMEP, 2009). CO2 emissions have been calculated on a more aggregated level using the total primary energy supply provided by IEA energy balances (IEA 2009a). The resulting CO2 emissions have been made consistent with the historic estimates reported by IEA (2009c) for years 1970 to 2005.

3. Results of data analysis

This section presents the overall trends in SO2, NOx and CO2 emissions and their determinants over a 40-year period, during which international cooperation to combat the environmental effects of air pollutants has intensified and reached a peak with the adoption in 1999 of the Gothenburg Protocol to Abate Acidification, Eutrophication and Ground-level Ozone (UNECE, 1999). This period shows unprecedented historic levels of the air emissions in Europe, but at the same time also in the reductions achieved through control of these pollutants and the environmental effects that result from pollution (Schöpp et al., 2003).

The emissions of SO2 culminated in WEU around the year 1970, NOx emissions achieved their highest levels ten years later. Simultaneously, continuous growth in GDP has been observed over the whole time period. This development indicates that most of the countries in WEU were able to decouple the economic growth from an increase in air pollution in all relevant sectors, although the timing has been sector-specific and differences occurred for two emission species. For CO2 emissions, such a decoupling effect has not yet been observed.

During the period covered by this study, important political changes took place in Europe. Some of these led to substantial changes to the economic structure as well as to significant alterations of the energy system, particularly in East European counties. Because of differences in historical evolution of emission trends and underlying drivers, results for WEU and EEU are presented separately.

3.1. Developments affecting emissions

Per capita emissions of SO2, NOx and CO2 as a function of income are depicted in Figure 1. In WEU, emissions of SO2 and NOx tend to decline with the growing income,


245

although the NOx function peaks at income levels much higher than observed for SO2. Based on statistical data, per capita CO2 appears to be rather inelastic to the income and until 2010 remains relatively stable. Collapse of centrally planned economy in EEU resulted in a rapid welfare reduction after 1990 associated with phasing out of inefficient industries and energy supplies (Cofala, 1994). Subsequently, per capita emissions declined remarkably in a short period of time. Recent adoption of air pollution policies and improved energy efficiency keeps the downward trajectory of the SO2 and NOx emissions in EEU. On the other hand, per capita CO2 emissions show a growing tendency linked with the fast growing GDP of transition economies.

1

10

100

5,000 10,000 15,000 20,000 25,000 30,000

Income (US$ 2000/cap)

Per

-cap

ita e

mis

sion

s lo

g

(

t/cap

)

(kg

/cap

)

EEU WEU

CO 2

NO x

SO 2

1970 → → 2010

Figure 1. Emissions per capita versus per capita income in Europe

There has been a different development observed for WEU and EEU concerning changes in energy intensity, expressed as energy use per unit of GDP produced. In WEU, energy intensity gradually decreased by nearly 30% in present as compared to the base year 1970. In some WEU countries, the oil price shocks affected the level and structure of energy consumption in the mid 1970s and the early 1980s. In many western European countries, energy growth slowed down around 1990. Some of these economies, managed to decouple energy growth from economic growth. As it is shown in Figure 2, until 1990s the energy intensity remained stable or even increased in the EEU region. The strong decline after 1990 is due to the recession in the countries in central and eastern Europe.

Admittedly, the continuous economic and energy structure changes during the period examined had an impact on the emission intensity, defined as an amount of air pollutants released per unit of energy consumed. The structural changes following the oil price shocks, especially the second one, and those initiated during the transition period in central and eastern Europe are most significant. In economic terms, the period sees a relative decline of industrial activity and an increase in transport. Some of these and


246

the other restructuring effects are more pronounced at a country level than at the overall European scale. In addition, changes in these emission intensities reflect modifications in the composition within the specific economic sector, for instance, shifts from manufacturing to services, changes in the ratio between passenger and freight transport and finally improvements in conversion efficiency.

CO2

-100%

-75%

-50%

-25%

0%

-50% -25% 0%Change in energy intensity

WEU

EEU

NOx

-100%

-75%

-50%

-25%

0%

-50% -25% 0%

Change in energy intensity

SO2

-100%

-75%

-50%

-25%

0%

-50% -25% 0%

Change in energy intensity

Cha

nge

in e

mis

sion

inte

nsity

1970 1970 1970

2010

2010

2010

Figure 2. Evolution of changes in energy and emission intensity relative

to the base year (Index: 1970 = 0%)

3.2. Decomposing changes in SO2 emissions

The decomposition approach discussed in Section 2.1 allows for quantification of emission reductions determined trough the drivers responsible for time evolution of emission profiles. To simplify the picture, in a first step, the three main effects discussed above are aggregated for graphic representation. Figure 3 shows total SO2 emissions in the WEU and EEU regions between 1970 and 2005, and, as projected, up to 2010. The dark grey area (“Remaining emissions”) shows the developments in real emissions. The upper limit of the area graph (the marked line) shows the hypothetical SO2 emission level that would have occurred in the absence of any mitigation component and represents the GDP growth effect. Thereafter, the graph shows the three factors of emission reductions: changes in energy intensity and efficiency, changes in the energy structure, and control measures.

In WEU, SO2 emissions have declining trend over the computation period. Changes in the fuel mix, combined with the reduced energy intensity, offset the continued growth in energy consumption. Once control measures come into the picture, emissions decrease even more rapidly. By 2010, the SO2 emission reductions that can be attributed to control measures and to energy intensity improvements are of similar magnitude, while the fuel mix changes become the main emission abatement element. As it is shown in Figure 3, increase in energy intensity in EEU (stripped area) overweighs the effects of fuel switches and better fuel quality, resulting in the moderate sulphur emission growth until mid 80’s. After 1990, the decline in total energy consumption brings emissions down, and this process is accelerated through efficiency improvements, coal substitution with natural gas and through adoption of pollution control legislation in most of the EEU countries.


247

WEU

0

20

40

60

1970

1975

1980

1985

1990

1995

2000

2005

2010

Mt SO2/yr

EEU

0

10

20

30

40

1970

1975

1980

1985

1990

1995

2000

2005

2010

Mt SO2/yr

Increased through worsening of energyintensity

Avoided through energy intensityimprovement of GDP

Avoided through changes in fuel mix

Avoided through end-of-pipe measures

Remaining emissions

Hypothetical uncontrolled emissionsfor constant energy intensity and fuelmix

Figure 3. Determinants of SO2 emission reductions compared to 1970

3.3. Decomposing changes in NOx emissions

The growth in NOx shows a different pattern when compared to that in SO2 emissions in both groups of counties under examination. While SO2 in WEU declines by some 40% between 1970 and 1990, NOx increases by 27% in that period. In EEU, the increase in NOx from fuel combustion by more than 65% is reported for the years 1970 to 1990. This increase in European nitrogen oxides occurs mainly due to the growth in energy consumption, but in contrast to SO2, it results from the growth in energy use not only in stationary emission sources but also in transport. After 1985, control measures are phased in, gradually bringing emissions down. This is the general picture that emerges from analysis of the main factors determining emission changes between 1970 and 2010 shown in Figure 4. Structural change only seems to play a minor role.

WEU

0

10

20

30

1970

1975

1980

1985

1990

1995

2000

2005

2010

Mt NOx/yr

EEU

0

3

5

8

10

1970

1975

1980

1985

1990

1995

2000

2005

2010

Mt NOx/yr

Increased through worsening of energyintensity and fuel mix



Avoided through end-of-pipe measures

Remaining emissions

Hypothetical uncontrolled emissionsfor constant energy intensity and fuelmix

Figure 4. Determinants of NOx emission reductions compared to 1970

The analysis of drivers behind the NOx emission changes is more difficult, than that for SO2. Different factors tend to overlap and the same factors change the way they impact emissions over time. Also the choice of the reference year is very important. While structural change reduces the emission growth up to 1990, it increases emissions (or more precisely: it diminishes the impact of control measures) for the period after 1990. This trend is more pronounced in the EEU region (stripped area).


248

3.4. Decomposing changes in CO2 emissions

Compared to 1970, CO2 emissions from fossil fuel combustion have been gradually increasing in WEU, reaching levels in 2005 by 10% higher than in the base year. Energy crisis in 1970’s followed by the oil glut resulted in temporary demand reductions in the middle of 1980’s (Salameh, 2004). Changes in economy structure, improved energy productivity, as well as energy saving measures constitute the main source of CO2 emission cuts over the time period 1970 to 2010. High oil prices reinforced the introduction of alternative and less carbon-intensive fuels into the energy markets, and until early 1990’s fuel switches had similar decarbonisation impact as the drop in energy intensity. By 2010, changes in fuel mix contribute by around 30% to overall CO2 reductions relative to the hypothetical level driven by GDP and energy consumption.

In EEU, CO2 emissions experienced a growth by 2.1% per year up to 1990. Consumption of coal, oil and gas has increased relative to the 1970 levels, as well as the energy intensity of the economy had increased by 1990. Inefficient use of fossil fuels offset the CO2-reducing effect of growing nuclear and hydropower supply capacities. Transition of EEU counties towards market oriented economy since 1990’s resulted first to attenuated market distortion of fuel prices, and simultaneously to a rapid drop in energy use stimulated through the conversion of industrial sector. The recent economy recovery is associated with an increase of CO2 emissions in the last decade (see Figure 5).

WEU

0

2000

4000

6000

8000

1970

1975

1980

1985

1990

1995

2000

2005

2010

Mt CO2/yr

EEU

0

1000

2000

3000

1970

1975

1980

1985

1990

1995

2000

2005

2010

Mt CO2/yr

Increased through worsening of energyintensity



Remaining emissions

Hypothetical emissions for constantenergy intensity and fuel mix

Figure 5. Determinants of CO2 emission changes compared to 1970

4. Summary and conclusion

The main objective of this study was to identify the principle factors responsible for reductions in European air emissions. For SO2, by the year 2010, structural change has been a dominant factor for emission reductions. Additional emission reductions are due to reduced energy intensity of some sectors. By the end of the computation period in 2010, about 25 percent of total reduction was due to targeted abatement end-of-pipe measures. It might be expected that in coming decades the share of emission reductions, which is due to control measures, will rise due to a penetration of measures through the stock of existing capital and due to more countries applying more advanced control measures.

For NOx, structural change on the whole is less important. Emission-reducing structural change in manufacturing industry and the power plant sector is outweighed by emission-increasing structural change in the transport sector. The application of control


249

measures is the most important factor explaining emission reductions. Reductions in energy intensity and fuel saving measures have been identified as the decisive factor for mitigation of CO2 emissions during the time period under analysis. Changes in fuel mix contribute to about one third to overall CO2 abatement. In the near future, the adoption of carbon capture might play a role in carbon mitigation as a factor belonging to the group of end-of-pipe measures.

One of the motivations of the study was to contribute to the understanding of the relationship between emissions and economic growth. The environmental Kuznets curve hypothesis suggests that there is an inverted U-shaped relationship, and emissions first increase with economic growth and subsequently decrease, once a certain level of wealth has been passed. Much debate remains about the reasons for such a relationship, if it exists. Two of the four factors identified in this study are clearly determined by economic parameters. Efficiency gains, energy demand and structural change out of heavily polluting industries are related to economic growth. Studies show that they follow an inverted U shaped Kuznets curve relationship. Also technological change will be driven by economic factors, but it is less clear whether and how pollution intensities of individual sectors are related to economic growth. It may, however, be sufficient that two of the main factors responsible for emission changes follow the Kuznets curve, for such a relationship to be observed between emission data (at least for some countries and some pollutants) and economic growth.

There is little evidence that control measures are directly linked to economic growth, but a formal analysis of this relationship, or indeed the alternative hypothesis, that emission controls are related to some environmental factor, such as, for instance, deposition in excess of critical loads, ecosystem sensitivity or the transboundary nature of pollution, remains to be done.

The results of this study may allow a different perspective of future emission scenarios. Emission reductions, as forecasted, are mainly driven by control measures. It should be examined whether structural change cannot be expected to continue to play the role that it had in the past. At least, for sectors responsible for NOx emissions (in particular transport) there seems to be wide room for emission-reducing structural change. This could imply that emission reduction possibilities are much larger and cheaper than presently projected. One should, however, avoid any over-optimism. As there seems no automatic mechanism to ensure that an environmental Kuznets curve relationship observed in the past will hold in the future, there is no guarantee, that structural change will continue the role it played in the past. As structural change, energy consumption and technology respond to many driving forces other than environmental pressures, one cannot exclude the possibility that developments go in the opposite direction, i.e., that a period of emission-reducing pressures, is followed by a period of emission increasing changes.

References

[1] Capros, P., L. Mantzos, V. Papandreou and N. Tasios (2008): European Energy and Transport Trends to 2030 — Update 2007. European Commission Directorate-General for Energy and Transport, Brussels, Belgium.

[2] Cofala J. (1994): Energy reform in Central and Eastern Europe. Energy Policy. Volume 22, Issue 6, June 1994, Pages 486-498.

[3] EC (2001): Directive 2001/81/EC on national emissions ceilings for certain atmospheric pollutants. Commission of the European Communities, Brussels, Belgium.


250

[4] EC (2005): Communication from the Commission to the Council and the European Parliament on a Thematic Strategy on Air Pollution. SEC(2005) 1132. Commission of the European Communities, Brussels, Belgium.

[5] EC (2008): 20 20 by 2020. Europe's climate change opportunity. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. COM(2008) 30 final. Commission of the European Communities. Brussels, Belgium.

[6] EMEP (2009): Transboundary acidification, eutrophication and ground level ozone in Europe in 2007. Status Report 1/09, Joint MSC-W & CCC & CEIP Report, Norwegian Meteorological Institute, Oslo, Norway.

[7] IEA (2009a): ‘Energy Balances of OECD and Non-OECD Countries - 2009 Edition. Database on CD. Energy Statistics Division - IEA Publications, International Energy Agency. Paris, France.

[8] IEA (2009b): ‘Energy Statistics of OECD and Non-OECD Countries - 2009 Edition. Database on CD. Energy Statistics Division - IEA Publications, International Energy Agency. Paris, France.

[9] IEA (2009c): CO2 Emissions from Fuel Combustion (2009 Edition), International Energy Agency, Paris, France.

[10] Klaassen, G., Amann, M., Berglund, C., Cofala, J., Höglund-Isaksson, L., Heyes, C., Mechler, R., Tohka, A., Schöpp, W. and Winiwarter, W. (2004). The Extension of the RAINS Model to Greenhouse Gases. IR-04-015. International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria.

[11] Kuznets, Simon (1955): Economic Growth and Income Inequality, American Economic Review, 45(1), 1-28.

[12] Salameh, Mamdouh G. (2004): Oil Crises, Historical Perspective, in: Encyclopedia of Energy, Pages 633-648, Elsevier, Oxford, UK.

[13] Schöpp W., Maximilian Posch, Sophia Mylona, and Matti Johansson (2003): Long-term Development of Acid Deposition (1880–2030) in Sensitive Freshwater Regions in Europe. Hydrology and Earth System Sciences, Volume 7, Issue 4, 2003, pp.436-446.

[14] UNECE (1999): Protocol to the 1979 Convention on Long-range Transboundary Air Pollution to Abate Acidification, Eutrophication, and Ground-level Ozone, United Nations Economic Commission for Europe, Geneva, Switzerland.


251

The government mechanisms of environmental protection

Rodolfo Rubén Salassa Boix

Universidad Rovira i Virgili de Tarragona, Spain [email protected]

Abstract

During the past fifty years, climatic change has become one of the most relevant and urgent issues at the international level because it affects not only every country on Earth, but it also has a profound effect on future generations. Additionally, the consequences of these phenomena are practically irreversible. The concern about the effects of global warming has mobilized the nations of the world to adopt various measures to stop, or at least reduce to a minimum, this devastating climatic process. We call these measures “governmental mechanisms of environmental protection”. Climatic change has given rise to a great debate: Which of these government measures is the most adequate to reach the desired goals?

Personally, I believe that it is necessary to study each individual contamination case to determine which the best instruments combination for the damage mitigation is. At first, it is nearly impossible to establish an absolute superiority of one mechanism of protection over another. In fact, it is most probable that the better mechanism tending to mitigate the effects of global warming is a mixture of the variables of the different mechanisms.

1. Concept and categories

Concern over ecological deterioration is growing. The law has not ignored this reality and has implemented a number of mechanisms of protection in the past few years in an effort to ameliorate this global problem.

At first the law used a subjective environmental defense, that is, focused on conflicts between individuals. With time the phenomena of publicatio has emerged; that is to say, the increase use of government intervention to mitigate environmental damage.

We could say that government measures of environmental protection (hereinafter “GOMEPs”) are those legal tools that can be adopted by government administrations, within their public functions, to preserve the environment by preventing potential damage, by promoting its protection, creating incentives not to damage it, or sanctioning those who fail to follow ecological protection norms.

Numerous classifications of GOMEPs have been devised, all of them intended on systemizing measures of environmental protections that can be adopted by the governments. It is well known that in the law there is no single way to categorize the legal concepts, and GOMEPs are no exception given that there are as many classifications as there are scholars in the matter.

The alternatives are as follows: 1) Total indifference by the public administrations, thus leaving the environmental problem and its solutions in the hands of private initiative or 2) Active participation of the governments which can result in different levels of action. The delicate environmental situation of the Planet has made it practically impossible to find an absolute absence of ecological governmental regulation. Assuming


252

governmental intervention, we will classify the GOMEPs according to the degree of autonomy they give to actual or potential contaminating subjects. Thus we find instruments of environmental protection that are restrictive, permissive, or voluntary enforcement.

1.1. Restrictive instruments

Restrictive instruments [1], also called direct or administrative instruments, are included among practices of command and control of public administrations. They are all those measures that exercise direct action over contaminators through a series of rules the violation of which generate sanctions over the violators.

This process is adaptable to the authoritarian model ruling relations between public controllers and controlled private entities. This model is evident through a public policy of applying sanctions with the purpose of overseeing business activities [2]. The intent is the maximum restriction of private freedom of entities by establishing mandatory obligations and limitations that must be strictly fulfilled and observed by them [3].

The narrow margin of deviation from the rules allowed to contaminators given by this instrument is due to the fact that, in contrast to permissive GOMEPs, restrictive GOMEPs are intended to prohibit socially intolerable conduct. Because the instrument is oriented to insure that contamination does not exceed a certain limit the ideal situation would be that no one would exceed the limit. That is the reason for draconian sanctions against those who fails to comply. On the contrary, the regulation of tolerable social conduct requires the implementation of permissive mechanisms given that they are more efficient from an economic point of view [4].

Considering their purpose, restrictive instruments can be classified as preventive repressive, and disclosure measures.

1.1.1. Preventive measures

They are those that, based on the principle of prevention, look to prevent ecological damage. Among preventive measures, the most important are:

- Establishment of levels of contamination. The purpose is to fix maximum (acceptable) levels of pollution. Contamination cannot be completely eradicated because it has been and is one of the principal consequences of economic and technological advances. But that fact cannot result in the uncontrolled utilization of the environment. It is imperative to find equilibrium between economic and technical development and protection of the environment.

- Regulation of content and characteristics of prime materials so that they won’t include potentially contaminant elements. This is accomplished by developing and use of specific production processes.

- Implementation of specific technological standards and imposition of certain technological levels. Along maximum limits of allowable contamination there is the establishment of minimum standards of environment-friendly technology. These measures tend to establish a standard of environmental quality.


253

- Mandatory insurance to cover damage to the environment. The requirement of mandatory insurance assumes the creation of a market where the risk of contamination is shifted from the contaminating entity to the insurer. In this manner, the potential costs of environmental degradation are covered in advance.

1.1.2. Repressive measures

In order for preventive measures to accomplish their goal it is necessary that vigilance and supervision by public regulatory entities be supported and complemented by an adequate repressive mechanism. Repressive measures act, in the first place to punish violators and, to a lesser degree, to prevent undesirable conduct.

1.1.3. Disclosure measures

All the GOMEPs, not only those that are restrictive in nature, depend for their implementation on an enormous flow of information which is beyond the collection capability of the regulating public entities by themselves. Therefore, the collaboration by actual and potential contaminators is essential.

It is at this point that disclosure measures emerge. These measures force private entities to provide the regulatory agencies with all the information that may be useful in controlling their contaminant activity. In and of themselves these measures will not prevent environmental contamination, but their mission is to facilitate the application of the rest of the GOMEPs.

1.2. Permissive instruments

Permissive GOMEPs allow private entities an ample margin of action when is time to plan their entrepreneurial strategy. They allow them to choose the strategy that best suits their economic interests [5].

Permissive GOMEPs encompass all the measures that influence cost and benefits of the different alternatives among which the economic agents can choose, with the focus on incentives to orient their actions in a direction favorable to the environment [6]. Permissive instruments can rise through positive, negative and mixed measures.

1.2.1. Positive measures

They aim to stimulate the contaminant entities to avoid, or at least diminish, environmental damage caused by their activity. This is done through financial or economic help.

Considering the financial activity of the government (income and expenses), this help can be divided in two groups. First, we have measures that influence public expenditures given that they necessarily involve the transfer of recourses from the controlling agencies to those they control (government subsidies). Second, we find measures that affect government income. The incentive is maintained but, in contrast to the subsidies, here the government releases and destines part of its income to promote pro-environment activities. These types of instruments materialize through a series of fiscal benefits [7].


254

- Tax exemptions. Environmental tax exemptions relieve the tax payer of the obligation to pay contributions they owe. The exemptions generally involve property and income taxes. While it is true that this measure results in not treating all taxpayers equally, it can be defended as incentive to activities that favor the environment.

- Tax deductions. In the environmental protection sphere, tax deductions aim to benefit contaminators by allowing them to reduce their taxable base by the amounts spent in measures that indirectly favor the environment. These measures can include acquisition or updating of technology, machinery, or installations that protect the environment.

- Reduction of tax rates. This advantage is observed in relation to the tax burden affecting consumption. The clearest case is with reductions in the value added tax (VAT). This is because a reduction in the tax rate rates applied to the taxable base reduces the final price of pro environment products to be paid by consumers. This is one way to influence consumers to choose pro environment products over products harmful to the environment.

- Favorable amortization/Depreciation rules. Amortization is an accounting tool by which the value of material investments lost through a period of time is reduced from income. There a number of different systems of amortization, and they will not be discussed herein, but it is clear that a system that allows a greater annual allowance for amortization/depreciation of property used to protect the environment constitutes a fiscal benefit and an incentive to invest in said property.

1.2.2. Negative measures

In contrast to the measures discussed above, the purpose of negative measures is to discourage environmental damage through surcharges or increases in the price for using environmental resources.

- Ecotax. This type of tax has the same general characteristics as other taxes – they are collected as income by the government within its power to collect tributes. However, to this characteristic it must be added the fact that it is a permissive instrument of environmental protection. Ecotaxes influence the conduct of private entities by increasing the prices of goods and services that tend to hurt the environment.

The application of ecotaxes constitutes a direct way to exert a price for the utilization of environmental goods forcing tax payers to assume the costs of that use. In this manner ecotaxes not only modify the behavior of private entities, but also immediately yields resources to be destined to the protection of the environment.

To qualify as “ecotax” the primary function of the tax must be the preservation of the environment by discouraging contaminant activities (a negative measure).

1.2.3. Mixed measures

They are measures that combine measures that encourage certain conduct and measures that discourage other actions by private entities. For that reason they cannot be labeled neither positive nor negative.

- Emissions trading markets. The first application of this innovative tool is in the area of atmospheric emissions of carbon dioxide. First, the total quantity of acceptable emissions of “green house” gases is determined to set a quota. Second, emission rights


255

(hereinafter ER) are assigned to green house gas producing entities until the quota is reached. Entities that emit less than their allowance can sell their unused ERs to other entities that emit more than their allowance of ERs.

This is a mixed measure because it combines positive and negative measures. On one hand reduction of emissions below authorized levels are encouraged. On the other hand, contamination above authorized levels are discouraged because those who exceed their particular ER are forced to pay for additional ERs or pay heavy fines.

- Price intervention. This scheme influences prices of certain products for the purpose of either encouraging or discouraging their use. The desired result is to favor environment friendly products and to discourage products harmful to the environment.

1.3. Voluntary compliance instruments

These measures allow maximum freedom and flexibility to regulated entities because they are void of coercive elements. The benefits to the contaminators are twofold. One benefit is the goodwill and positive publicity generated by voluntarily adopting measures that protect the environment. Another benefit is the increase in popularity of their products because they have a limited impact on the environment. Among these measures we find the following:

- Certificate of quality. Consist of issuing certificates to those products that comply with certain requirements of environmental protection quality. Through independent evaluators these certificates state that the process of production of the product has produced a minimum environmental impact. These labels allow the final consumer to easily identify products least harmful to the environment. The economic benefit to the producer is a direct advertisement of its product and, in theory, an increase in its demand.

- Environmental audit. This is an ecological control to which private entities voluntarily submit for the purpose of improving their image over their competitors. The objective of these audits is to evaluate, at a particular point in time, the entity’s impact on the environment.

- Ecological certificate. Consists in the issuing of a document to producing entities whereby an independent third party certifies that the entity complies with a series of pre established ecological protection parameters. This mechanism is similar to the certificate of quality but here the certification is for the whole entity and not just a particular product.

2. Selecting an adequate mechanism of environmental control

Beginning in the 1970s there is a marked rise in the use of mechanisms of environmental control. This in turn opened a debate regarding which type of instrument, restrictive or permissive, was more efficient. At first restrictive instruments were used by governments. Today that tendency is reverting and the majority of experts understand that regulatory instruments have reached their limit of efficiency and therefore must be replaced.

A considerable portion of the authors affirms that restrictive GOMEPs have significant disadvantages over permissive measures [8]:


256

1) The high costs of vigilance and control necessary to be effective; 2) The limited capacity to adapt to changing situations/conditions; 3) The meager incentive to develop less contaminant technologies, or to use existing

technologies. The great advantage offered by permissive mechanisms is the lowering of costs of

their implementation and enforcement. Supervision is no longer minucios and continuous, but instead is oriented to determining whether the controlled entity has opted for the conduct that entitles it to receive the incentive [9].

Consequently, ecotaxes (or alternatively the systems of negotiable emission permits), and subsidies are the more suitable instruments for imposing negative and positive externalities, given that they provide the correct incentives to reach the optimum assignment of recourses [10].

Beyond the discussion of comparisons between restrictive and permissive mechanisms, there is a debate regarding which permissive GOMEP is the most effective. In favor of the mechanisms that rely on price it is said that the application of ecotaxes constitute a simple and direct method of assigning a price to the utilization of environmental assets. This forces the using entities to assume the costs of using them. This in turn results in changing the behavior of the entities as well as generating government income to be used in environment protection and restoration. These end results justify the utilization of ecotaxes [11].

On the other hand, some authors point out that trading emission markets hold a slight advantage over ecotaxes because trading emission markets reduce uncertainty, and the desirable level of contamination is reached more assuredly [12]. This is so because in order to predict with accuracy the impact of ecotaxes one must know the marginal costs and benefits to the regulated entities. If this information is deficient trading emissions markets allow more flexibility to the regulated entities.

We also find authors who, assuming an eclectic posture, maintain that emission right markets as well as adequate taxing systems have the capability of reaching a set of ecological standards at a minimum cost. Therefore, they understand that the use of each system must be strictly considered as alternatives [13].

3. Conclusion

Beyond the discussions herein it has been found that repressive measures as well as permissive measures have advantages and disadvantages [14]. Consequently, it is wise to use the benefits of each to counter, to the degree possible, their disadvantages.

Moving towards a more eclectic posture, it is impossible to establish a priori an absolute superiority of one GOMEP over another that would allow for the outright dismissal of the latter [15].

When is time to choose the most adequate instrument, each type of contamination must be separately analyzed. It is highly likely that choosing the best control instrument will require a mixture of different instruments [16]. Consequently, each particular case must be studied to determine the best possible combination of instruments capable of mitigating a specific contamination.


257

In support of the last conclusion above, Stains states “there is no simple answer, a political panacea. Inevitably, there must be a case by case analysis” [17].

Reference

[1] Mc Loughlin, J. y Bellinger, E. G. (1993) : Environmental pollution control. An introduction to principles and practice of administration, International environmental law and policy series, Graham & Trotman/Martines Nijhoff, p. 63 y ss.; Revesz, R. L.; Sands, P. and Stewart, R. B. (2000): Environmental law, the economy and sustainable development. The United States, the European Union and the International Community, Cambridge University Press, Cambridge, p. 174 y ss.

[2] Del Brío González, J. Á. (2001): Medio ambiente y empresa: de la confrontación a la oportunidad, 1° Edition, Civitas Ediciones, Madrid, p. 81.

[3] Castillo i Solsona, M., Enciso i Rodríguez, J. P. y Farré i Perdiguer, M. (2002): Fiscalitat autonòmica i medi ambient, Atelier, Generalitat de Catalunya (Institut d’Estudis Autonòmics), Barcelona, p. 26.

[4] Magadán Díaz, M. y Rivas García, J. (2004): Fiscalidad y medio ambiente en España, 2° Edition, Septem Ediciones, Oviedo, p. 11/12. Rosembuj, T. (1988): Elementos de derecho tributario”, PPU, Barcelona, p. 44.

[5] Yábar Sterling, A. (2002): La protección fiscal del medio ambiente. Aspectos económicos y jurídicos, Marcial Pons, Madrid, p. 128.

[6] OCDE (1991), Politique de l’environnement comment appliquer les instruments économiques, Paris, 1991.

[7] Vaquera García, A. (1999): Fiscalidad y medio ambiente, Lex Nova, Valladolid, pp. 57-58.

[8] Cuerdo Mir, M. y Ramos Gorostiza, L. (2000): Economía y naturaleza, Síntesis, Madrid. Alonso Oroza, S. y otros (1999): Política ambiental y desarrollo sostenible, Instituto de Ecología y Mercado, Madrid, p. 109.

[9] Lozano Cutanda, B. (2007): Derecho ambiental administrativo, 8° Edition, Dykinson, Madrid, p. 104.

[10] Buñuel González, M. (1999): El Uso de Instrumentos Económicos en la Política del Medio Ambiente, C.E.S. Colección de Estudios, Madrid, p. 91.

[11] Franco Sala, L. (1995): Política Económica del Medio Ambiente, Cedecs, Barcelona.

[12] Lozano Cutanda, B. (2007): Derecho ambiental administrativo, 8° Edition, Dykinson, Madrid, p. 444.

[13] Yábar Sterling, A. (2002): La protección fiscal del medio ambiente. Aspectos económicos y jurídicos, Marcial Pons, Madrid, p. 138 y ss.; y Buñuel González, M. (1999): El Uso de Instrumentos Económicos en la Política del Medio Ambiente, C.E.S. Colección de Estudios, Madrid, p. 122.

[14] Pichot, F. y Rapado, J. R. (1994): La fiscalidad y el medio ambiente. Políticas complementarias, Versión española, Mundi-Prensa, Madrid, pp. 26-27.

[15] Ruesga, S. M. y Durán, G. (2000): Empresa y medio ambiente, Pirámides, Madrid, p. 111.


258

[16] Iranzo, J. E. (2000): Medio ambiente y mercado en España, Instituto de estudios económicos, Madrid, p. 128.

[17] Stavins, R. N. (1993): “Transaction costs and the performance of markets for pollution control, Discussion paper Nº QE93-16, Resources for the future, p. 15.


259

Uncertainties of forest carbon accounting with an international

application of the carbon budget model for Canadian forest sector (CBM-CFS): South Korean case

Kwang-IL Tak1, Hyeon-Kyu Won2, Kyeong-hak Lee2 and Man-Yong Shin1

1Faculty of Forest Environmental System, Kookmin University, 861-1 Jeongneung-dong, Seongbuk-gu, Seoul, Korea 136-702

[email protected] 2Climate Change Centre, Korea Forest Research Institute, Seoul, Korea

57 Hoegui-ro, Dongdaemun-gu, Seoul, Korea 130-712

Abstract

When a foreign developed forest carbon model is applied to a country, the country could face an added uncertainty in addition to the uncertainties inherent to the model and carbon accounting itself. A recent test-run application of the model to a Korean forest revealed a soil carbon stock much lower than the values supported by scientific researches. One of the reasons is suspected to be the historical litter raking, which has not been inventoried by any governmental statistics and has remained as an uncertainty for forest carbon accounting. At the same time, the model has no capability with itself of estimating the impact of the litter raking on soil carbon stock directly. As a means to reduce the uncertainty, an indirect simulation with disturbance matrix by CBM-CFS3 was run and calculated the impact of the litter raking to be reflected in the estimation of soil carbon of current forest ecosystem.

Keywords: forest carbon accounting models, CBM-CFS3, litter raking, South Korea

1. Introduction

South Korea is not among the annex I countries of Kyoto Protocol and not yet subject to the voluntary carbon emission reduction responsibility by the protocol. Nevertheless, South Korea has ratified the protocol and is responsible to report its national greenhouse gas emission regularly. South Korea also made it clear of its commitment to participate in the international efforts for climate change mitigation. Korea has been building a national greenhouse gas inventory system. National forest carbon accounting system is one of important components of the national GHG inventory system and is under construction by Korea Forest Research Institute (Lee, 2008). On the other hand, South Korea is expected to build a model not only for international reporting to UNFCCC but also for international negotiations related to greenhouse gas emission reduction which are expected to happen in near future, since South Korea is the world’s 11th biggest greenhouse gas emission country and the international pressure for reducing the national emission has been mounting over recent years.

Forest carbon accounting is an important part of national Green House Gas (GHG) accounting in Korea. Forest ecosystem can be used as a means for the country’s efforts to reduce greenhouse gas emission through human intervention in forest ecosystem process. Korea Forest Research Institute, a forest carbon reporting agency, is responsible to develop internationally transparent and accountable forest carbon accounting system to undertake analyses for forest policy development at national level as well as for international reporting.


260

There are numerous forest carbon models around the world. Models can be group into two categories. One is based upon the tree growth driven by simulating photosynthesis such as BIOME-BGC (Running and Gower, 1991) and 3-PG (Landberg and Waring, 1997).

The other is based upon the growth driven by forest inventory information, i.e. yield curves. Examples of this type of model are CBM-CFS3, FORECAST and CO2FIX. Each type of model offers its own advantages. Photosynthesis driven models require wide range of data at time steps ranging from hourly to monthly, whereas yield curve driven models can take advantage of existing inventory data and they are very well suited for human and natural disturbances (Kurz et al., 2009). Some selected models are listed as in Table 1.

Table 1. Some selected forest carbon accounting models

Model Name Developed and Primary Use

Key Features

CBM-CFS3 Canada Estimates carbon stock and carbon change over time for a forest estate.

FORECAST British Columbia

Estimates carbon stock and carbon change over time for a forest ecosystem.

CO2FIX The Netherlands

Estimates carbon stock at the ecosystem and landscape level. Has biomass, soil, products, bioenergy, financial and carbon accounting modules.

CARBINE United Kingdom

Estimates carbon stock and carbon change over time for a forest estate. UM based.

BioMetre Austria Estimates carbon stock and carbon change over time for a forest estate.

COLE USA Estimates carbon stock and carbon change over time for a forest estate.

HARVCARB USA Tracks the fate of wood products by type (e.g. lumber, panels, and paper) and carbon pools (e.g. in-service, used for energy, stored in landfills and waste emissions) over time sequences.

TimberCAM Australia Tracks the fate of carbon stored in trees through their harvesting, conversion to wood products, utilization and disposal

Source: Bull (2008)

These models are designed to report at different reporting levels. For example, CBM-CFS3 is designed to reporting at national level but it is suited for reporting at the operational level as well. Similarly, at the operational level, FORECAST and CO2FIX have also been designed and could work at the operational level for either Kyoto standards, Voluntary market standards or NGO standards, where they exist (Bull, 2008).

Considering the existence of a number of readily available models and the amount of time and efforts required to develop its own model being extremely costly, countries


261

like South Korea could be tempted to use other country developed model instead of developing its own.

But such international application is often considered to create a problem of additional uncertainty besides ones inherent to the model itself. Many forest carbon models are developed based upon the dynamics of local forest ecosystem. More specifically, tree species, soil productivity, growth rate, and land use history of the country where a model is developed are often a lot different from a country where the model is to be applied.

Same uncertainties issues emerged when CBM-CFS3 is applied to a forest in Korea. In this paper the authors try to discuss about how these uncertainties could be addressed in the international applications of a forest carbon model.

2. CBM-CFS3 Model

Carbon Budget Model of the Canadian Forest Sector (CBM-CFS3) is a stand and landscape level modeling framework used to simulate the dynamics of all forest carbon stocks. It is designed for Canadian forests but it complies with the carbon estimation methods outlined in the guidelines of the Intergovernmental Panel on Climate Change (IPCC).

CBM-CFS3 accounts for carbon stocks and stock changes in biomass and dead organic matter pools. It was developed initially to assess carbon dynamics at the national and landscape level, and has been developed to explore carbon dynamics at operational level. The model can be used to assess past carbon stock changes using information on management actions and natural disturbances that have occurred, or to evaluate future changes resulting from scenarios of management actions and natural disturbances. The known uncertainties of the model are as follows (Bull, 2008):

• peat-land carbon dynamics • climate change impacts on forest growth • the impacts of climate change on disturbance regimes • insect disturbances that cause reductions in growth rates • it does not represent inter-annual variability in tree growth or productivity

brought about by local climate or weather (e.g. growing season length, moisture availability).

CBM-CFS3 has been applied in several countries including Russia, Mexico, United States, China and Spain (CFS Carbon Accounting Team, 2009). South Korea is one of the newest countries to apply CBM-CFS3 to its forest. South Korea chose this model to test its suitability for national and international reporting as well as carbon stock management at operational level.

3. CBM-CFS3’s application to Korean forest carbon accounting

The forest in Korea CBM-CFS3 has been applied is 33,732 ha of national forest. It is managed by Korea Forest Service’s Regional Office located in Hongcheon, Gangwon Province. The forest is typical South Korean productive coniferous dominant forest mixed with oak species. Age class of the forest ranges from 10 to 100 years.

Over 40 year age class accounts for over 80% of the forest. Growing stock volume of the forest is over 4.2 million cubic meters, or 123 cubic meters per hectare, 20% higher


262

than national average (Korea Forest Service, 2008). This forest carries several features to deserve one of Korea’s role model forests and became the first Forest Stewardship Council (FSC) certified forest in Korea in 2006.

The reason this forest has been selected for this study is manifold. It is actively managed forest and role model management forest, which includes playing same role model for carbon stock management. The forest inventory data is relatively well documented and readily accessible.

4. Initial Application of CBM-CFS3

CBM-CFS3 is originally designed for Canadian forests and the default parameters are developed based upon Canadian forest ecosystem dynamics. Nevertheless, the model is flexible enough to apply to different ecosystems with changes in parameters. The model can accommodate new sets of parameters from a different ecosystem dynamics specific to a country that the model is applied. Among those parameters are climate data, tree growth rate, and dead organic matter turnover rate (Kurz, et., al., 2009).

At the initial application of CBM-CFS3, all these parameters were changed to satisfy local forest ecosystem dynamics in Korea. Despite these parameter changes, while biomass carbon stock is within the range of acceptable variation, the results of running CBM-CFS3 for the study area in Korea produced a significantly higher carbon stock specifically in soil and dead organic matter (DOM) pools than those from comparable studies on Korean forests as in Table 2.

Table 2. Biomass and Soil Carbon from Initial Run of CBM-CFS3

Unit: Mg C ha-1

*Above ground biomass only. Source: Lee et., al (2009), Park et., al. (2001), Park (2003)

5. Uncertainties specific to Korean forests

Low soil carbon value of Korean forests is suspected to be related to Korea’s land-use history. Korean forest has been heavily exploited for 36 years during the Japanese colonization from 1910 to 1945 and the remaining forests were destroyed again during the Korean War between 1950 and 1953. After the war was ended the majority of the forest in Korea became subject to severe soil erosion (Tak et al, 2007). In addition, before briquette coal and other fossil fuel was introduced to farming villages, forest floor had no time to collect litter or dead organic matter because of litter raking by farming activities. Litter was raked in the forest to supply domestic fuel and to produce compost for farming. Not only litter but also most of dead wood was also removed. So forest floor of Korea remained near barren until the litter raking practice was finally stopped in the mid 1970s. But there is no way of estimating the impact of litter raking from the current condition of the forest through CBM-CFS3 model. Unfortunately, the amount of litter raked and the extensity of litter raking have not been inventoried by Korea Forest Service. So the impacts of litter raking on forest soil carbon remain as a huge uncertainty when CBM-CFS3 is applied for forest carbon accounting in Korea.

Soil Carbon Biomass Carbon CBM-CFS3 Run Results 156.7 196.2 Comparable Korean Study Data

46.8 – 75.9 91.3 – 279.9*


263

6. Coping with the uncertainties

Although the CBM-CFS3 does not have a mechanism in it to deal with litter raking practices for its carbon accounting, the model is flexible enough to handle such uncertainty problems by a way of going around the problem. To solve this problem, litter raking is assumed as a disturbance which would bring significant impact on carbon dynamics of a forest ecosystem.

CBM-CFS3 calculates carbon stocks of various pools of forest ecosystem based upon ecosystem dynamics and to project future carbon stock changes based upon assumptions of management and disturbances. A disturbance matrix is built to simulate the litter raking impact on forest carbon dynamics. Disturbance matrix is an input and output table to show how much carbon from a source is transferred to which sink and by how much. The 25 rows and 30 column of disturbance matrix are shown in Table 3.

Table 3. Rows and columns of disturbance matrix

Row (source) Description

Column (sink) Description

1 Softwood merchantable stemwood 1 Softwood merchantable stemwood 2 Softwood foliage 2 Softwood foliage 3 Softwood others 3 Softwood others 4 Softwood submerchantable 4 Softwood submerchantable 5 Softwood coarse roots 5 Softwood coarse roots 6 Softwood fine roots 6 Softwood fine roots 7 Hardwood merchantable stem 7 Hardwood merchantable stem 8 Hardwood foliage 8 Hardwood foliage 9 Hardwood other 9 Hardwood other 10 Hardwood submerchantable 10 Hardwood submerchantable 11 Hardwood coarse roots 11 Hardwood coarse roots 12 Hardwood fine roots 12 Hardwood fine roots 13 Very fast soil pool 13 Very fast soil pool 14 Fast soil pool 14 Fast soil pool 15 Medium soil pool 15 Medium soil pool 16 Slow soil pool 16 Slow soil pool 17 Peat pool 17 Peat soil pool 1 Softwood merchantable 18 CO2 2 Softwood foliage 19 CH4 3 Softwood others 20 CO 4 Softwood sub-merch 21 Forest product sector 5 Softwood coarse roots 1 Softwood merchantable 6 Softwood fine roots 2 Softwood foliage 7 Hardwood merch 3 Softwood others 8 Hardwood foliage 4 Softwood sub-merch 9 Hardwood other 5 Softwood coarse roots 10 Hardwood submerch 6 Softwood fine roots 11 Hardwood coarse roots 7 Hardwood merch 12 hardwood fine roots 8 Hardwood foliage 13 Above Ground Very Fast soil C 9 Hardwood other 14 Below Ground Very Fast soil C 10 Hardwood submerch 15 Above Ground Fast soil C 11 Hardwood coarse roots 16 Below Ground Fast soil C 12 hardwood fine roots 17 Medium Soil C 13 Above Ground Very Fast soil C 18 Above Ground slow soil C 14 Below Ground Very Fast soil C 19 Below Ground Slow soil C 15 Above Ground Fast soil C 20 Softwood Stem Snag 16 Below Ground Fast soil C


264

21 Softwood Branch Snag 17 Medium Soil C 22 Hardwood Stem Snag 18 Above Ground slow soil C 23 Hardwood Branch Snag 19 Below Ground Slow soil C 24 Black C 20 Softwood Stem Snag 25 peat 21 Softwood Branch Snag

22 Hardwood Stem Snag

23 Hardwood Branch Snag

24 Black C

25 Peat

26 CO2

27 CH4

28 CO

29 NO2

30 products

Source: Kurz et., al. (2009)

A value is assigned to each coordinate of 25 row and 30 column carbon sink and source table to mimic litter raking practice. Litter or deadwood is removed out of the forest is considered as emission. The disturbance matrix built for litter raking is shown as in the table 4.

Table 4. Disturbance matrix for litter raking

Row (Source)

Column (Sink) Proportion

1 1 1 2 2 1 3 3 1 4 4 1 5 5 0.8 5 30 0.2 6 6 0.5 6 30 0.5 7 7 1 8 8 1 9 9 1 10 10 1 11 11 0.8 11 30 0.2 12 12 0.5 12 30 0.5 13 30 1 14 30 1 15 30 1 16 16 0.5 16 30 0.5 17 30 1 18 18 0.1 18 30 0.9 19 19 0.5 19 30 0.5 20 20 0.2 20 30 0.8 21 15 0.2 21 30 0.8 22 22 0.2 22 30 0.8 23 23 0.2 23 30 0.8 24 24 1 25 25 1

Source: Kurz et., al. (2009)


265

In order to quantify the magnitude of the impact of uninterrupted litter raking on soil carbon, two CBM-CFS3 simulations were run based upon 10 years of litter raking assumptions, 10% and 50% of raking respectively. The simulation results are presented in the Figure 1 and Figure 2.

Figure 1. 10 year simulation with litter raking 10% of the area

Figure1 shows that litter raking of 10% of the study area is not enough to bring down

the soil carbon stock value to a realistic level. Figure2 shows a value close to more realistic soil carbon stock value. These two graphs show the magnitude of the impact of litter raking practiced for 10 years on soil carbon stock. Present soil carbon stock value can be more realistically estimated with the calculated impacts of litter raking on soil carbon dynamics.

Figure 2. 10 year simulation with litter raking of 50% of the area

7. Conclusion

CBM-CFS3 is a forest carbon accounting system developed by Canadian Forest Service for Canadian forests. As an early carbon model CMB-CFS3 could be applied to other countries which have not developed its own model yet. When a model is applied internationally, one can easily face with an uncertainty like the impact of litter raking on soil carbon in South Korea because of its historical practice over an extended period of time. Although the model is not designed to calculate such impact specific to certain country directly, it has capacity to deal with such problem by indirect method within


266

the model. Using disturbance matrix and simulation of the model, such impact of litter raking on soil carbon can be estimated and be reflected on calculating more realistic value of present soil carbon stock.

Acknowledgement

This research including two month visit to the carbon accounting team of Pacific Forestry Centre in Victoria, Canada has been funded by Korea Forest Research Institute. Authors thank Canadian Forest Service’s Pacific Forestry Centre to let the author to visit and research at the carbon accounting team. W.A. Kurz provided expert advices on how to address the uncertainties issues related to the international application. E.T. Neilsson gave endless technical advice to run CBM-CFS3 model. Authors thank both of them tremendously.

References

[1] Lee, K.H, Kim, S.H. and Son, S.M. (2008): Development of forest carbon accounting system in Korea. Korea Forest Research Institute.

[2] Running, S.W. and Gower, S.T. (1991): FOREST-BGC, A general model of forest ecosystem processes for regional applications. II. Dynamic carbon allocation and nitrogen budgtets. Tree Physiol. 9, 147-160.

[3] Landberg, J.J., Waring R.H., 1997. A generalized model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. Forest Ecol. Manage. 95, 209-228

[4] Kurz, W.A., C.C. Dymond, T.M. White, G. Stinson, C.H. Shaw, G.J. Rampley, C. Smyth, B.N. Simpson, E.T. Neilson, J.A. Trofymow, J. Metsaranta, M. J. Apps. (2009): CBM-CFS3: A model of carbon-dynamics in forestry and land-use change implementing IPCC standards. Ecological Modelling Vol. 220, No. 4, pp480-504.

[5] Bull, Gary. (2008): Forest carbon institutions, markets, management and models: status report for British Columbia, Canada

[6] CFS Carbon Accounting Team. (2009): The CBM-CFS3: Technology transfer and international uses. CBM-CFS3 training workshop, July 9, Victoria.

[7] Korea Forest Service. 2008. 10 year management plan. North Regional Office. [8] Lee, S.K., Y.H. Son, N.J. Noh, S.J. Heo, T.K.Yoon, A.R. Lee, S.A. Razak, and

W.K. Lee. (2009). Carbon storage of natural pine and oak pure and mixed forests in Hoengseong, Kwangwon. J. of Korean Forest Society. Vol. 98, No.6, pp 772-779.

[9] Park, K.S. and S.W. Lee. (2001). A study on carbon production dynamics of natural cork oak dominant forest ecosystem in Gongju, Pohang and Yangyang areas in Korea. J. of Korean Forest Society. Vol. 90, No.6, pp 692-698.

[10] Park, I.H., D.Y. Kim, M.J. Lee, H.O. Jin, Y.H. Son, Y.K. Seo. (2003): Carbon production of mongolian oak and cork oak stands of Chuncheon area in Korea. J. of Korean Forest Society. Vol. 92, No.1, pp 52-57.

[11] Tak, K., Y. Chun, P. Wood. 2007. The South Korean Forest Dilemma. International Forestry Review. Vol. 9, No.1, pp 548-557


267

Uncertainties implied in the country specific baselines caused by

different approaches applied for recalculating the NMVOC emissions into CO2 equivalents

Jochen Theloke1, Folke Dettling2

1 Institute for Energy Economics and the Rational Use of Energy Universitaet Stuttgart, Heßbrühlstrasse 49a, 70565 Stuttgart, Germany

[email protected] 2 German Federal Environmental Agency

Wörlitzer Platz 1, 06844 Dessau-Roßlau, Germany [email protected]

Abstract

In this presentation will be described a comparison of different country specific approaches for recalculating Non-Methane Volatile Organic Compounds (NMVOC) emissions from solvent use into CO2 equivalents and the specific implications onto the baseline emissions which are the basis for the Kyoto and post Kyoto reductions targets. Additional we will also ask, if this recalculation make principal a sense, if we taking into account the atmospheric fate of NMVOC`s and their contribution to the greenhouse gas effect.

Keywords: NMVOC, Solvent use, CO2 recalculation, baseline, reporting obligations, Ozone

1. Introduction

All countries are obliged to recalculate the emissions of non-methane volatile organic compounds (NMVOC) into CO2 equivalents in the greenhouse gas (GHG) emission data sets officially reported to UNFCCC. NMVOCs are a mix of several hundred species which very different numbers of carbon. The quantitative speciation of this substance class is mainly unknown. In principal NMVOC need to be reported as an indirect GHG, because ozone is a strong GHG and NOX as well as NMVOC are precursors of this short lived species. NMVOC emissions are caused mainly by solvent use and on-road as well as off-road activities. A minor share of NMVOC comes from several combustions processes as well as mining and extraction of fuels.

2. Methodology for recalculation CO2 from solvent use related NMVOC emissions

Solvent use is one main activity for NMVOC emissions and covers a wide range of processes and products. Significant quantities of solvents are used both for industrial applications (mainly coatings, cleaning solvents, and printing solvents), but also for non-industrial applications (mainly aerosols, decorative paints, and consumer products). In principle all chemicals that can be classified as NMVOC must be included in the analysis, which implies that it is essential to have an explicit definition of NMVOC. The definition of NMVOC is, however, not consistent. In the EMEP-guidelines for calculation and reporting of emissions, NMVOC is defined as ”all hydrocarbons and hydrocarbons where hydrogen atoms are partly or fully replaced by other atoms, e.g. S, N, O, halogens, which are volatile under


268

ambient air conditions, excluding CO, CO2, CH4, CFCs and halons”. The amount of chemicals that fulfil these criteria is large.

Because there exist no reference methodology for this recalculation the applied CO2 equivalents differ significantly from country to country. This has implications to the country specific baselines.

Neither the revised 1996 IPCC Guidelines for National Greenhouse Gas Inventories nor the chapter 5 of the recent 2006 IPCC Guidelines include any specific method. In Volume 1, Chapter 7 of the 2006 IPCC Guidelines is an in principal not applicable methodology for the solvent use sector described: Calculating CO2 inputs to the atmosphere from emissions of carbon containing compounds from NMVOC: Inputs CO2 = Emissions NMVOC * C * 44/12

Where C is the fraction carbon in NMVOC by mass (default = 0.6)

The default factor is derived from fossil fuels and can not applied to solvents. This methodology lets at least a large space for interpretation. So it is not surprising, that no more than half of the parties report estimates in this sector calculated as indirect CO2 -Emissions. In doing so the CO2-to-NMVOC ratio varies from 0 to 3.8 in the different estimates. This is revealing of the need for a reference method.

Table 1 shows the large differences of correlations between reported NMVOC emissions and recalculated CO2 emissions for different countries. The differences are not understandable. We expect country specific differences because the sectoral structure of solvent differs large from country to country, but we do not understand the magnitude of large differences between countries. In table 6 are shown the country specific officially reported NMVOC emissions form the solvent use sector for 2008, recalculated CO2 emissions as reported by the countries, the CO2 emissions recalculated on base of the officially IPCC methodology, the country specific relationship between reported (recalculated) CO2 emissions and the IPCC based calculation of emissions. In the last column are shown the country specific CO2 to NMVOC correlations as reported by the countries.

3. Implication to the country specific baselines

Because there exists no reference methodology for the recalculation of NMVOC emissions from solvent use to CO2 emissions the countries apply country specific estimations which differ extremely from country to country. This has implications to the country specific baselines. If a country estimate higher C-content for the solvent used in the base year as in the reporting year is the result a contribution to mitigation due to different assumptions about the country specific C-content of applied solvents in different years and vice versa.

4. Conclusion

All countries are obliged to recalculate the emissions of non-methane volatile organic compounds (NMVOC) into CO2 equivalents in the greenhouse gas (GHG) emission data sets officially reported to UNFCCC. NMVOCs are a mix of several hundred species which very different numbers of carbon. The quantitative speciation of this substance class is mainly unknown. In principal NMVOC need to be reported as an indirect GHG, because ozone is a strong GHG and NOX as well as NMVOC are precursors of this short lived species. The contribution of NMVOC in form of CO2 to the GHG effect is negligible. NMVOC emissions are caused mainly by solvent use and on-road as well as off-road activities. A minor share of NMVOC comes from several combustions processes as well as mining and extraction of fuels.


269

Table 1. Correlations of reported NMVOC and CO2 emissions on base of the country

submissions for 20081

Liechtenstein 0,2 0,8 0,5 72,8% 3,8Luxembourg 4,5 10,8 9,8 9,6% 2,4Latvia 15,8 49,1 34,7 41,4% 3,1Finland 23,5 51,8 51,8 0,0% 2,2Hungary 23,4 65,4 51,6 26,8% 2,8Ireland 27,6 86,0 60,7 41,7% 3,1Denmark 27,5 65,0 60,5 7,5% 2,4Norway 48,9 146,8 107,6 36,4% 3,0Sweden 75,0 169,0 165,0 2,5% 2,3Netherlands 59,1 130,0 130,1 -0,1% 2,2Greece 54,0 160,7 118,8 35,2% 3,0Switzerland 41,4 162,1 91,1 77,8% 3,9Austria 97,1 231,9 213,6 8,5% 2,4Romania 69,2 134,7 152,3 -11,5% 1,9Czech Republic 90,0 282,8 197,9 42,9% 3,1Portugal 76,5 230,6 168,3 37,0% 3,0Spain 465,8 1138,3 1024,8 11,1% 2,4Italy 483,1 1272,2 1062,9 19,7% 2,6France 382,0 1190,6 840,4 41,7% 3,1European Community 3424,1 8134,7 7532,9 8,0% 2,4Belarus 54,0 NA 0,0Belgium 55,2 IE,NA 0,0Bulgaria 15,4 11,4 33,9 -66,3% 0,7Canada IE,NA NA,NE 0,0Croatia 74,4 218,0 163,7 33,2% 2,9Estonia NA,NE NA,NE 0,0Germany 714,1 2142,4 1571,1 36,4% 3,0Iceland 1,9 5,8 4,1 41,7% 3,1Japan 1204,9 NA,NE 0,0Lithuania 29,3 91,2 64,4 41,7% 3,1Monaco 0,0 NE 0,0New Zealand 34,8 NA,NE 0,0Russian Federation 1704,1 NA,NE 0,0Slovakia 34,0 0,1 74,7 -99,9% 0,0Slovenia 12,8 NA,NE,NO 0,0Turkey NA,NE NA,NE 0,0Ukraine 120,2 NA,NE 0,0United Kingdom 386,5 NE 0,0USA 3833,8 NA,NE 0,0Malta 1,5 NA 0,0Poland 198,3 618,0 436,3 41,7% 3,1Cyprus 4,7 3,0 10,3 -70,6% 0,6Australia 155,8 NA 0,0Kazachstan NA,NE NA,NE 0,0

CO2 Calculation on base of the

IPCC methodology [kt]

Correlation between country

specific and IPCC

methodologyPartyCO2 to NMVOC

RelationNMVOC [kt] CO2 [kt]

1http://unfccc.int/national_reports/annex_i_ghg_inventories/national_inventories_submissions/items/5270.php


270

The applied recalculations approaches differ from country to country. The official IPCC default methodology is not applicable to solvent use. It is derived from fossil fuels and does not taken into account the high diversity of the applied solvents.

The described situation has also an impact to the country specific mitigation targets in relation to the base year. The authors recommend changing the reporting obligations due to the several reasons mentioned above and especially under taken into account that NMVOC emissions affect mainly the effects from short lived species.

References

2006 IPCC Guidelines for National Greenhouse Gas Inventories Volume 3 Industrial Processes and Product Use, 2007 (http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol3.html)

2006 IPCC Guidelines for National Greenhouse Gas Inventories Volume 1 General Guidance and Reporting, Chapter 7, 2007 (http://www.ipcc-nggip.iges.or.jp/public/2006gl/vol1.html)


271

Analysis and comparison of uncertainty assessment methodologies for high resolution Greenhouse Gas emission models

Balendra Thiruchittampalam, Jochen Theloke, Melinda Uzbasich, Matthias Kopp, Rainer Friedrich

Institut für Energiewirtschaft und Rationelle Energieanwendung, University of Stuttgart, Heßbrühlstr 49a, 70565 Stuttgart, Germany

[email protected]

Abstract

The aim of this document is to present a first analysis of uncertainty assessments methods, which are suitable to determine the uncertainty of high resolution emission models. The selected methods are the most suitable and relatively most applicable methods to determine the uncertainty of high resolution GHG emission models.

Keywords: Uncertainty assessment, spatial, temporal, high resolution emission model

1. Introduction

The objective of the high resolution fossil fuel emission models is to describe anthropogenic emissions of GHG at high spatial and temporal resolutions. These models can provide in particular the actual state, future trends and possible developments of Greenhouse Gas (GHG) emissions. Such information is essentially required to estimate and develop adequate emission control options, therefore eminent for policy makers and other stakeholders.

In general emissions are available from inventories as annual country totals without further temporal and spatial distinction. The task of high resolution emission models is the allocation of annual national totals to small spatial and temporal scales. The development of such models is going towards very high spatial resolutions like 1km x 1km for the global scale and also high temporal scales e.g. hourly. This process neglects usually the aspects of the uncertainty assessment. The aim of this proposal is to analyze existing methods for uncertainty assessment and to identify the most appropriate methods for the high resolution emission models.

Many methodologies and tools suitable for supporting uncertainty assessment have been developed and reported in the scientific literature [2]. The first step of the proposed procedure will be the identification of the most promising methods like the Monte Carlo analysis, scenario analysis, sensitivity analysis etc. The second step is to analyze the applicability and complexity of the identified methods for high resolution models. This analysis has to be carried for the spatial as well as for the temporal resolution. The results of the proposed investigation will be the comparison of the common uncertainty assessment methodologies for the high resolution models for the spatial and the temporal dimension with references on the application and the software requirements.


272

2. Principal approach of high resolution emission models

2.1 Methodology for the spatial resolution of GHG emissions

As already mentioned, emissions are available from inventories as annual country totals without further temporal and spatial distinction. The emissions reported to the UNFCCC are available in the CRF (IPCC Common Format for Reporting) source categories as national totals. The mission of the spatial distribution of GHG emission inventories is therefore to identify indicators, which are capable of describing the spatial pattern of the emission sources. The spatial patterns are influenced by several attributes like the location and capacities of power plants, traffic volume and position of roads, production figures and location of industrial facilities, etc. Thus, high-resolution emission models categorize each CRF emission source category into point, line and area sources. Basically, the spatial distribution of emission inventories can be distinguished into two main procedures. The first step is to regionalize the national totals to administrative levels (e.g. NUTS3). The second is to distribute this regionalized information to the grid using georefernced proxy data. Georefernced proxy data are information like coordinates or land use data, which describe the spatial location of geographical features. The principal approach of the regionalization and the grid distribution is described in figure 1.

Figure1. Principal approach of the spatial distribution of national GHG emission inventories

The allocations of the different emission sources demand as depict in figure 1 diverse proxy data sets. The distribution of point sources e.g. from energy industries needs information like capacities, emissions and coordinates (power plants, iron- und steelworks, refineries, cement production, airports, etc.). The collection of the proxy data for line sources e.g. road traffic consider data like traffic counts for roads, indicators for ship and train traffic. Area sources like residential combustion are first allocated to administrative units using statistical data like population density, employees divided by branch, animal numbers, etc. the next step is to relate the emissions to corresponding land use categories. Using the described methodology it is possible to analyze and also visualize the results of the spatial allocation for each sector (see figure 2 a. power plants


273

for the year 2005 and 2b. road traffic for the year 2005) and also the overall spatial pattern of the anthropogenic CO2 emissions (see figure 3a. for Europe and 3b. for Germany). Figure 2 shows the emissions of CO2 of large point sources of the energy industry in Europe for the year 2005 (power plants, refineries, etc. about 1200 facilities). The CO2 emissions of point sources and line sources are aggregated to a grid with the resolution 5 min longitude x 5 min latitude.

a)

b) Figure 2. Annual CO2 emission (2 a. power plants and 2b. road traffic)

for the year 2005

2.2 Methodology for the temporal resolution of GHG emission

For further temporal allocation of the annual amounts of the spatial distributed CO2 emissions, sector specific time profiles are required. The aim of the temporal resolution is to identify indicators to describe the temporal patterns of the spatially resolved emission data. The spatially distributed data is available for each CRF sector. As a result, it is necessary to distinguish the information about the temporal patterns according the CRF source categories. The method of the temporal resolution is based on three types of time profiles. The applied time profiles are: monthly, weekly and hourly profiles (see also figure 4). To derive such time profiles lots of sector specific indicator data like load curves for power plants, workings hours for industrial plants, traffic count data for road traffic etc. are necessary. Additionally, for more adequate temporal allocation the knowledge of the influence of the temperature is required.


274

a)

b) Figure 3. Example for the overall spatial resolution of CO2 emissions

(3a. Europe left and 3b. Germany right)[1] The influence of the temperature can be derived from correlating the time segment

with the corresponding value. For example, the derivation of temperature dependent monthly time profiles for the energy supply (power plants) is described in the figures 5 and 6. Figure 5a shows the correlation between the ambient temperature and the fuel consumption in Germany and 5b describes the differences for the European countries. The correlation is based on consumption data from Eurostat and temperature data provided by Haylock, M.R. et. al. (2008) [4]. After deriving country and fuel specific temperature correlation, CO2 specific shares of the emissions divided by fuel type can be used to allocate the resulting temperature dependent monthly time profiles for power plants. Figure 6a illustrates the inverse dependency of the fuel consumption for different fuel types for Germany in 2005 and Figure 6b describes the share of the fuel types for CO2 emissions for the European countries derived from gathered data available at from Gains [5].

The resulting yearly, weekly and hourly time profiles for the European power plants are depict in the figures 6, 7 and 8. Figure 6 shows the monthly time profile for Germany divided by fuel type. Figure 7 shows the weekly profile for European countries derived from the Union for the Co-ordination of Transmission of Electricity (UCTE) respectively European Network of Transmission System Operators for Electricity (ENTSO-E).


275

Figure 4. Approach for the temporal resolution using sector specific time profiles

a)

R2 = 0,6669

0,05

0,06

0,07

0,08

0,09

0,1

0,11

0,12

-5 0 5 10 15 20 25

Temperature in [°C]

Sha

re (

norm

ailz

ed)

b)

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1

AT

BE

BG

CZ

DE

DK

ES

FI

FR

GR

HU

IE

IT

NL

PL

PT

RO

SE

SK

UK

Coefficient of determination (R²)

Figure 5. Coefficient of determination of the correlation for coal consumption in power plants and ambient temperature (5a. for Germany left and 5b. summary

for Europe right) [1]


276

a)

0,06

0,065

0,07

0,075

0,08

0,085

0,09

0,095

0,1

0,105

0,11

1 2 3 4 5 6 7 8 9 10 11 12

Month

Sha

re (n

orm

aliz

ed)

-2

0

2

4

6

8

10

12

14

16

18

20

Tem

pera

ture

in [°

C]

Lignite

Natural Gas

Coal

Heavy fuel (oil)

Temperature

b)

0%

20%

40%

60%

80%

100%

ATBE BG CH CY CZ

DE DK EE ES FIFR GR HR HU IE IT LT LU LV M

T NLNO PL PT

RO SE SISK TR

UK

Natural gas

Heavy fuel (oil)

Hard coal

Brown coal

Figure 6. Fuel consumption in power plants divided by fuel type for the year 2005

(6a. monthly values for Germany, left and 6b. summary of the yearly share for Europe, right) [1]

The gradient for the different countries is similar as the figure 7 describes and shows decrease towards the weekends. On the contrary, figure 8 explains the difference of hourly profiles between working days (e.g. Mondays) and weekends (e.g. Saturdays).

3. Uncertainty of high resolution emission models

3.1 Overall uncertainty of high resolution GHG emission models

Generally, the main reason of performing an uncertainty analysis is to assess the uncertainty in the model output that derives from uncertainty in the inputs [3].

High resolution GHG emission models can be divided into two main parts as described in the previous chapter: spatial and temporal distribution. Additionally,


277

the main input parameter is the sector specific emissions available from the submissions to the UNFCCC. Therefore the overall uncertainty can be described as aggregation of the uncertainty of the emission inventories, the spatial distribution with the necessary input parameters and the temporal resolution with the necessary input parameters.

UEM = UEI + US + UT

with EM: Emission distribution model; EI: Emission inventory; S: Spatial distribution; T: Temporal resolution.

Figure 7. Weekly profiles for European countries derived from UCTE for the year 2006

The overall uncertainty can be derived using the following equation for combining

uncertainties from IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories:

.

According to the IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories, where uncertain quantities are to be combined by addition, the standard deviation of the sum will be the square root of the sum of the squares of the standard deviations of the quantities that are added with the standard deviations all expressed in absolute terms.

Moreover, the IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories indicates that, where uncertain quantities are to be combined by multiplication, the same rule applies except that the standard deviations must all be expressed as fractions of the appropriate mean values:

.


278

a)

Time profiles for Monadys in Europe

0,025

0,03

0,035

0,04

0,045

0,05

0,055

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour

Sha

re [%

]

AT

BA

BE

BG

CH

CZ

DE

DK

ES

FR

GR

HR

HU

IT

ME

MK

NL

PL

PT

RO

RS

SI

SK

UA

b)

Time profiles for Saturdays in Europe

0,03

0,035

0,04

0,045

0,05

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour

Sha

re [

%]

AT

BA

BE

BG

CH

CZ

DE

DK

ES

FR

GR

HR

HU

IT

ME

MK

NL

PL

PT

RO

RS

SI

SK

UA

Figure 8. Difference between the hourly time profiles for power plants in Europe for Mondays and Saturdays[1]

3.2 Uncertainty of emission GHG inventories

The uncertainties analysis of the GHG inventories is usually provided by the countries submitting the yearly inventories to the UNFCCC. Basically, according to the IPCC Good Practice Guidance and Uncertainty Management in National


279

Greenhouse Gas Inventories, there are two methods, which can be applied to identify the uncertainty of emission inventories: The error propagation and the Monte Carlo analysis. As the uncertainty assessment of the inventories is also available from the National Inventory Reports, these can be easily implemented in the overall uncertainty estimation and therefore not described more in detail in this document.

3.3 Uncertainty of spatial GHG distribution

In general, the uncertainty of the spatial distribution is described through the uncertainties of the emission input and the distribution parameters. The spatial distribution follows the equation:

∑=

i

i

i

value

valueEE CRFiCRF *,

with ECRF,i : emission of a specific CRF sector for a specific grid cell or administrative unit;

ECRF : National total emission for a specific CRF sector; value: specific spatial distribution parameter ;

CRF: Sector from the IPCC Common reporting format;

i : Specific grid cell or administrative unit.

The uncertainty for each CRF sector from national emission inventories are available from National Inventory Reports (NIR) and can be integrated in the assessment for the spatial distribution. On the contrary, the uncertainties of the distribution parameters should be described using probability density functions (PDFs). In common, the shape of the probability density function (PDF) cannot be determined empirically due to the lack of suitable data. Therefore, expert judgement is necessary to define the shape of the PDF.

The result of the spatial distribution is principally the sum of products of sectoral emissions and the share of the corresponding distribution parameter. Therefore, the equations to combine the uncertainties from the IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories described in chapter 3.1 can be applied repeatedly to estimate the uncertainty of the overall results from the spatial distribution.

3.4 Uncertainty of temporal GHG resolution

According to the spatial distribution, the uncertainty of the temporal distribution is described through the uncertainties of the emission input and the distribution parameters. The uncertainty of emission input for each grid cell has to be determined in the procedure of the estimation of the uncertainty of the spatial distribution and integrated into the temporal distribution. The temporal resolution follows the equation considering the result of the spatial distribution:

∑=

h

h

h

value

valueEE iCRFhiCRF *,,,


280

with ECRF,i : emission of a specific CRF sector for a specific grid cell or administrative unit

ECRF,i,h : Emission of a specific CRF sector for a specific grid cell and time iteration value: specific temporal distribution parameter CRF: Sector from the IPCC Common reporting format (see Annex 1) h : index for the temporal iteration.

The uncertainty of the spatial distribution is the result of the uncertainty assessment of the spatial part and has to be integrated in the assessment for the temporal distribution. According to the spatial distribution, the uncertainties of the temporal distribution parameters should be described using probability density functions (PDFs). Also in compliance with the spatial distribution, the shape of the probability density function (PDF) cannot be determined empirically due to the lack of appropriate data. Therefore, expert judgement is also necessary to define the shape of the PDF for the temporal resolution.

The result of the temporal resolution is also principally the sum of products of sectoral emissions per grid and the share of the corresponding distribution parameter. Therefore, the same procedure to combine the equations of the uncertainties has to be applied repeatedly to estimate the uncertainty of the overall results from the temporal distribution.

4. Methodologies for uncertainty assessment of high resolution GHG emission models

4.1. Error propagation

Uncertainty assessment using the error propagation is relatively easy to apply and delivers quick results and therefore suitable for analysis to get an overview of the magnitude of the uncertainty of high resolution emission models. However, the error propagation equations are limited to simple correlations and are valid only if the following conditions are met [2]:

(1) the uncertainties have Gaussian (normal) distributions;

(2) the uncertainties for non-linear models are relatively small: the standard deviation divided by the mean value is less than 0.3; and

(3) the uncertainties have no significant covariance.

As a matter of fact, the error propagation should only be used for high resolution emission models, if quick results are necessary, otherwise error propagation are more suitable for screening purposes and preliminary analysis (see also [2]). The advantage of the uncertainty assessment is, that is easy to apply, delivers fast results and needs only little of computational effort.

4.2 Monte Carlo Analysis

Constitutive from the results of the error propagation the uncertainty assessment using the Monte Carlo analysis can be performed. The Monte Carlo analysis is suitable for detailed uncertainty analysis and is used according to IPCC Good Practice Guidance


281

and Uncertainty Management in National Greenhouse Gas Inventories in particularly where uncertainties are large, distribution is non-normal, the algorithms are complex functions and/or there are correlations between some of the activity sets, emissions factors, or both. Its purpose is to trace out the structure of the distributions of the model output and in its simplest form this distribution is mapped by calculating the deterministic results for a large number of random draws from the individual distribution functions of input data and parameters of the model [2].

Therefore, to perform a Monte Carlo analysis it is necessary to define PDFs for all model inputs and distribution parameters. As already mentioned, the shape and all other attributes of the PDF cannot usually be determined empirically due to the lack of suitable data. Expect for few sectors like the road transport, where i.e. traffic count data is available. As a consequence, expert judgement is essential to define the shape and the all other attributes of the PDFs. For parameters, where the expert judgement is not sufficient enough the normal distribution can be applied. Although, the normal distribution is usually used in cases of small uncertainty range and symmetric relative to the mean (see also the IPCC Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories).

The advantage of Monte Carlo analysis is its general applicability and that it does not impose many assumptions on probability distributions and correlations and that it can be linked to any model code [2]. The main disadvantages are the computational effort and the necessity to define PDFs, in particular when a great number of model parameters are involved. The computational effort remains even though, methods have been developed to reduce the number of model runs and also a variety of software solutions are available. The disadvantage of the commercial and also the free available software is that none are applicable directly for the use for high resolution emission models and therefore individual modifications are unavoidable.

4.3 Scenario Analysis

Scenario analysis aims to describe logical and internally consistent sequences of events to explore how the future may, could or should evolve from the past and present [6]. The scenario analysis is in particular for the GHG very interesting, because of the influence of the GHG on the climate. The opportunity to analyze future trends and possible developments of Greenhouse Gas (GHG) emissions is one of the main advantages of high resolution emission models. Such information is essentially required to estimate and develop adequate emission control options, therefore eminent for policy makers and other stakeholders. For instance, the spatial and temporal influence of abatement measures can be investigated or the spatial and temporal impact of worst case and best case scenarios can be studied in detail. Scenarios can ensure that assumptions about future developments are made transparent and documented and are often the only way to deal with the unknown future [2]. A limitation for qualitative scenarios is that the distribution parameters are based in the present respectively in the past. The computational effort is the same as a usual model run and involves no additional software.


282

4.4 Expert Judgment

An expert judement for the uncertainty assessment of high resolution emission models is especially necessary in cases where no sufficient data is available to determine the uncertainty. It is widely used in quantitative risk analysis to quantify uncertainties in cases where there are no or too few direct empirical data available to infer on uncertainty [2].

Expert judgement is essential mainly for two issues: the definition of PDFs for the distribution parameters and for the uncertainty judgement of the geospatial distribution. The geospatial distribution describes i.e. coordinates of power plants or the pattern of specific land use categories, this kind of uncertainties can only be investigated through visualization. The uncertainty of the spatial emission pattern has to be judged by an expert with the necessary experience on the certain area. Expert elicitation typically involves the following steps [2]:

(1) Identify and select experts.

(2) Explain to the expert the nature of the problem and the elicitation procedure. Create awareness of biases in subjective judgements and explore these.

(3) Clearly define the quantity to be assessed and choose a scale and unit familiar to the expert.

(4) Discuss the state of knowledge on the quantity at hand (strengths and weaknesses in available data, knowledge gaps, and qualitative uncertainties).

(5) Elicit extremes of the distribution.

(6) Assess these extremes: could the range be broader than stated?

(7) Further elicit and specify the distribution (shape and percentiles or characterising parameters).

(8) Verify with the expert that the distribution that you constructed from the expert’s responses correctly represents the expert’s beliefs.

(9) Decide whether or not to aggregate the distributions elicited from different experts (this only makes sense if the experts had the same mental models of the quantity for which a distribution was elicited).

The main advantage of expert elicitation is that issue, where no sufficient data is available can be considered and accessed. The limitation of expert judgements is that these are subjective and not easily verified.

4.5 Sensitivity Analysis

The final considered method is the sensitivity analysis. Sensitivity analysis (SA) is the study of how the variation in the output of a model (numerical or otherwise) can be qualitatively or quantitatively apportioned to different sources of variation, and of how the outputs of a given model depend upon the information fed into it [7]. Therefore, the SA can be described as the identification of the sensitivity of the results from a certain model input. In principal, three types of sensitivity analysis can be distinguished [8]:


283

• Screening, which is basically a general investigation of the effects of variation in the inputs but not a quantitative method giving the exact percentage of the total amount of variation that each factor accounts for.

• Local SA, the effect of the variation in each input factor when the others are kept at some constant level.

• Global SA, the effects on the outcomes of interest of variation in the inputs, as all inputs are allowed to vary over their ranges.

Although, the sensitivity analysis is crucial, the application for high resolution emission models requires a huge number of model runs due to the number of input parameters. As a consequence, the computational efforts are high and time intensive. The strength of SA is that it provides insight in the potential influence of all sorts of changes in input and helps discrimination across parameters according to their importance for the accuracy of the outcome [2].

5. Conclusion

The aim of this document is to present a first analysis of uncertainty assessments methods, which are suitable to determine the uncertainty of high resolution emission models. The selected methods are the most suitable and relatively most applicable methods to determine the uncertainty of high resolution GHG emission models. Until now, the results from high resolution models were validated only by expert judgment or in some cases by comparison with measurements (needs additionally a transport and a meteorological model). The disadvantage of expert judgement is that, the result is subjective and also only qualitative. Even though, the expert judgement can’t be neglected, the remaining methods are capable of delivering quantitative values, which wasn’t considered for the high resolution emission models until now. Quantitative results are capable of identifying weaknesses of the high resolution emission model and have the potential to improve the high resolution emission models.

References

[1] B. Thiruchittampalam, R. Köble, J.Theloke, M. Uzbasich, U. Kummer, T. Geftler, S. Wagner, R. Friedrich (2009): Fossil fuel emission modelling: approach and results for Europe. Poster, 8th International Carbon Dioxide Conference - Jena, Sep 13-19 2009.

[2] Jens Christian Refsgaard, Jeroen P. van der Sluijs, Anker Lajer Højberg, Peter A. Vanrolleghem: Uncertainty in the environmental modelling process - A framework and guidance, Environmental Modelling & Software 22 (2007) 1543-1556.

[3] Per-Anders Ekström, Robert Broed (2006): Sensitivity Analysis Methods and a Biosphere Test Case Implemented in EIKOS.

[4] Haylock, M.R. (2008), A European daily high-resolution gridded data set of surfacetemperature and precipitation for 1950–2006, JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, D20119, doi:10.1029/2008JD010201, 2008.

[5] IIASA, 2008: gains.iiasa.ac.at/


284

[6] Van Der Heijden, K., 1996. Scenarios: The Art of Strategic Conversation. John Wiley & Sons, ISBN 0471966398.

[7] Andrea Saltelli, Karen Chan, Marian Scott, Sensitivity Analysis John Wiley & Sons publishers, Probability and Statistics series, 2000.

[8] Van der Sluijs, J.P., Janssen, P.H.M., Petersen, A.C., Kloprogge, P.,Risbey, J.S., Tuinstra, W., Ravetz, J.R., 2004. RIVM/MNP Guidance for Uncertainty Assessment and Communication Tool Catalogue for Uncertainty Assessment. Utrecht University (URL: http://www.nusap.net/sections.php?op¼viewarticle&;artid¼17).


285

The improvement of greenhouse gas inventory as a tool for reduction of emission uncertainties for oil activities in Russia

Nina E. Uvarova

Institute of Global Climate and Ecology str. Glebovskaya, 20-B, Moscow 107258, Russian Federation

[email protected]

Abstract

Reliable national inventory is one of the steps to efficient greenhouse gas emission mitigation. National inventory reliability is defined by uncertainty level. Oil activities are the key category of the national greenhouse gas inventory in Russia. The quantitative greenhouse gas emission uncertainty is included in the National Inventory Report of the Russian Federation1. In general, it corresponds to the IPCC Tier 12. The uncertainty of the estimations mainly depends upon activity data and emission factors used. Uncertainty values both of activity data and emission factors are equal to 5 % and 25% respectively. The overall uncertainty estimates are 21%1. The quality of greenhouse gas inventory for oil activities could be improved. An attempt to improve the level of uncertainty of greenhouse gas emission estimates was made by a shift from production-based approach (IPCC Tier 1) to mass balance approach (IPCC Tier 2). The methodology for mass balance approach is provided the 2006 IPCC Guidelines for National Greenhouse Gas Inventories3. Mass balance approach allows to not only obtain highly accurate assessments, but also allows to cross-check and verify calculations on the step-by-step basis. The parameters were recalculated for the national economy conditions of the Russian Federation. The IPCC Tier 2 has been adapted to specific features of national oil operations. The greenhouse gas inventory was calculated from 1990 to 2008. The uncertainty of the inventory was estimated and compared with that for Tier 1 approach2. The comparison showed that the application of the higher tier results in the lower uncertainties of the estimates.

Keywords: greenhouse gas inventory, uncertainty, accuracy, oil operations

1. Introduction

According the Fourth Assessment Report of the Intergovernmental Panel on Climate Change4 the increased concentration of human-induced greenhouse-gases in the atmosphere leaded to global climate changes such as dramatic temperature increasing. As a result of the entry into force in February 2005 of the Kyoto Protocol, the international community of 40 developed countries is committed to undertake joined efforts to reduce greenhouse gas emissions to the atmosphere. The level of the implementation of the commitments under the Kyoto Protocol is judged through the national inventory reports, which are annually submitted and reviewed by the international groups of Experts. As a Party to the United Nations Framework on Climate Change (UNFCCC) Russian Federation has been obliged to prepare, publish and regularly update national emission inventories of greenhouse gases. Reliable national inventory is one of the steps to efficient greenhouse gas emission mitigation. National inventory reliability is defined by uncertainty level that becomes more important in the case of the key categories of the national greenhouse gas inventory.


286

2. Key category choice

The choice of the key category depends upon its contribution to total greenhouse gas emissions. As it has been reported in the National inventory report of Russian Federation1, greenhouse gas emission from oil activities was estimated as Key by the trend in the National greenhouse gas inventory. At the Fifteenth Conference of the Parties to the UN Framework Convention on Climate Change (UN FCCC, Copenhagen, December 2009) the Russian Federation indicated of its intention to reduce greenhouse gas emissions by 15 to 25% below the 1990 year by 2020 year. The improvement of the reliability of the national greenhouse gas inventory particularly for Key categories such as oil activities promotes for identification of the sources with the higher mitigation potential.

3. Uncertainty as provided by National inventory of Russian Federation

According the Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories the integrated quantitative assessment of the greenhouse gas emission uncertainty from oil activity is quite difficult because of its complex structure2. The quantitative greenhouse gas emission uncertainty is included in the National Inventory Report of the Russian Federation1. It corresponds to Tier 12. The uncertainty mainly depends upon activity data and emission factors used. Uncertainty value of emission factors, used for greenhouse gas emission estimation in the national inventory is found to be about 25%2. The source of the activity data on the operations with oil are provided by the national statistics their accuracy is quite high and lies in the range that has the upper point 5%5. Consequently activity data of the Russian Federation is quite reliable to be used for greenhouse gas emission estimation. To make assessment of greenhouse gas emission uncertainty following equations was used2:

( ) 2/122

21_ σσ +=DEVIATIONSTANDARD , (1)

( ) ( ) ( )n

nnTOTAL xxx

xUxUxUU

+++•++++•

=...

...

21

2222

211 , (2)

where σ1 and σ2 are the standard deviations of the probability density functions of the emissions in year t1 and t2. The 95% confidence limits (this time of the mean or the difference in the means) will be given by plus or minus approximately two standard deviations; UTOTAL is the percentage uncertainty in the sum of the quantities (half the 95% confidence interval divided by the total (i.e. mean) and expressed as a percentage); xi and Ui are the uncertain quantities and the percentage uncertainties associated with them, respectively2.

The overall uncertainty for oil operations was estimated as 21%1. The improvements of the national greenhouse gas emission inventory allow to reduce the emission uncertainty. This could be done by using the mass-balance approach that based on another input data which does not contain IPCC default emission factors. The approach is provided in the 2006 IPCC Guidelines3.

4. Improving the level of uncertainty of greenhouse gas emission estimates

An attempt to improve the level of uncertainty of greenhouse gas emission estimates was made by a shift from production-based approach (IPCC Tier 1) to mass balance


287

approach (IPCC Tier 2) for oil activities excluding storage and transportation operations of the oil, which the lower contribution to the entire emission profile. The methodology for mass balance approach is provided the 2006 IPCC Guidelines for National Greenhouse Gas Inventories3.

6103.42)1()1( −⋅⋅⋅⋅−⋅−⋅⋅= gasgasflaredoilventing yMXCEQGORE , (3)

6103.42)1()1(444

−⋅⋅⋅⋅−⋅⋅−⋅⋅= CHCHflaredoilflaringCH yMFEXCEQGORE , (4)

[ ] 6103.42)1(()1()1(4222

−⋅⋅−⋅⋅+⋅+⋅⋅−⋅⋅−⋅⋅= sootNMVOCNMVOCCHNMVOCCOCOflaredoilflaringCO XyNcyNcyMFEXCEQGORE , (5)

ONflaredoilflaringON EFXCEQGORE22

)1( ⋅⋅−⋅⋅= , (6)

where Eventing - direct amount (Gg/y) of GHG gas (CH4 or CO2) emitted due to venting at the oil production facilities; EСО2, СН4, N2Oflaring - direct amount (Gg/y) of greenhouse gas (CH4 or CO2 or N2O) emitted due to flaring at the oil production facilities; GOR is the average gas-oil ratio (m3/m3) referenced at 15°C and 101.325 kPa; Qoil - total annual oil production (103m3/y); Mgas – molecular weight of the gas of interest (e.g., 16.043 for CH4 and 44.011 for CO2); NCi – number of moles of carbon per mole of compound i (i.e., 1 for CH4, 2 for C2H6, 3 for C3H8, 1 for CO2, 2.1 to 2.7 for the NMVOC fraction in natural gas and 4.6 for the NMVOC fraction of crude oil vapors); yi – mol or volume fraction of the associated gas that is composed of substance i (i.e, CH4, CO2 or NMVOC); CE – gas conservation efficiency factor; Xflared – fraction of the waste gas that is flared rather than vented. With the exception of primary heavy oil wells, usually most gas flared; FE – flaring destruction efficiency (i.e., fraction of the gas that leaves the flare partially or fully burned ). Typically, a value; Xsoot – fraction of the non-CO2 carbon in the input waste gas stream that is converted to soot or particulate matter during flaring. In the absence of any applicable data this value may be assumed to be 0 as a conservative approximation; EFN2O – emission factor for N2O from flaring (Gg/103 m3 of associated gas flared), refer to the IPCC emission factor database (EEDB), manufacturer’s data or other appropriate sources for the value of this factor; 42,3×10-6 – the number of kmol per m3 referenced at 101.325 kPa and 15°C (i.e.42.3×10-6 kmol/m3) times a unit conversion factor of 10-3 Gg/Mg which brings the results of each applicable equation to units of Gg/y (the inverse value of the Molar volume, i.e. 1/Vm).

The activity data analysis showed that it is possible to use the mass-balance approach (IPCC Tier 2) for greenhouse gas emission estimation for oil activities after its adaptation for national conditions of the country. For this purpose the parameters were recalculated.

According the 2006 IPCC Guidelines for National Greenhouse Gas Inventories3 the constant 42,3×10-6 kmol/m3 - the number of kmol per m3(i.e.42.3×10-6) times a unit conversion factor of 10-3 Gg/Mg which brings the results of each applicable equation to units of Gg/y - is referenced at 101.325 kPa and 15°C. In the Russian Federation as the standard conditions are 101.325 kPa and 20°C. So the constant was recalculated using the Molar Volume (estimated for normal conditions – temperature is equal to 0°C, pressure is equal to 101.325 kPa) and the Mendeleev-Clapeyron equation for ideal gases. After recalculations the new constant was obtained and its value is equal to 41,57656×10-6 kmol/m3.

The other activity data contains the information on associated oil gas. In this case it is possible to avoid additional uncertainty making calculation with those of associated oil gas. For this reason for formula

OILQGOR• (part of common equations (3), (4), (5), (6))


288

equal CE

Q GASOILUSED __ was derived. The fraction of the waste gas that is flared rather than

vented (Xflared) was estimated using additional activity data for oil gas activity by formula

( )CECE

QQ

GASOILUSED

FLARED

−• 1__

.

The data on chemical content of the oils produced in Western Siberia are used in our calculations because of their dominance in total oil production in the country. The greenhouse gas inventory was calculated from 1990 to 2008 years.

4. Comparing the estimated uncertainty with one for Tier 1 approach

Thus mass balance approach allows not only to obtain highly accurate assessments, but it also allows to cross-check and verify calculations at each estimation stage. Finally it is also possible to compare value of greenhouse gas emission from oil activity obtained by mass-balance approach with initial activity data.

The uncertainty of the greenhouse gas emissions calculated with the use of mass-balance approach was estimated as recommended by the IPCC. The value obtained was 18 %, which is lower that obtained through the calculations with the use of production- based approach. The further reduction of the uncertainty can be achieved by improvement of accuracy of the parameters such as chemical composition of the associated oil gas and parameters of flaring such as Xsoot (fraction of the non-CO2 carbon in the input waste gas stream that is converted to soot or particulate matter during flaring).

References

[1] NIR 2009. The National Report on the Inventory of Emissions by Sources and Removals by Sinks of the Greenhouse Gases of the Russian Federation. Federal Service for Hydrometeorology and Environmental Monitoring, Moscow. [In Russian]. Available at http://unfccc.int/national_reports/annex_i_ghg_inventories/national_inventories_submissions/items/4771.php

[2] IPCC, 2000. Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories. IPCC National Greenhouse Gas Inventories Programme. IGES/OECD/IEA. 2000.

[3] IPCC, 2006. 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas Inventories Programme, Eggleston H.S., Buendia L., Miwa K., Ngara T., and Tanabe K. (Eds.). Vol. 2 Energy, IPCC/IGES.

[4] IPCC, 2007. Climate Change 2007 – The Physical Science Basis Contribution of Working Group I to the Fourth Assessment Report of the IPCC. Geneva, - 104 pp.

[5] Statistical yearbook of Russia. 2009. Rosstat. Moscow, - 795 pp.


289

Using a fuzzy inference system for the map overlay problem

Jörg Verstraete 1,2

1 Systems Research Institute, Polish Academy of Sciences ul. Newelska 6, Warsaw, 01-447, Warsaw

[email protected] 2 Department for Telecommunications and Information processing, Ghent University

Sint Pietersnieuwstraat 41, Gent, 9000, Belgium [email protected]

Abstract

In this contribution, the problem of ill aligned spatial data is considered. The problem is commonly known as the map overlay problem, and occurs when data from different grids are combined. When the different grids don't line up properly, determining what portion of the data associated with a tile in one grid is relevant for the data associated with a partially overlapping tile in the latter grid becomes a problem. As it is in general not possible to derive an exact value, we opted for an approach that results in fuzzy sets. Rather than process the data itself, a seemingly intelligent system to make a decision on which portions could be relevant, was developed. The presented approach makes use of a fuzzy inference system, a system built up of a number of if-then rules containing fuzzy predicates. These rules are used to evaluate set of input values to yield one or more output values. The input values can be fuzzy sets themselves, output values are always fuzzy sets (in our case fuzzy numbers) which are then defuzzified.

Keywords: Map overlay problem, fuzzy inference system

1. Introduction

When dealing with gridded data, it can be necessary to combine data from different sources: one source could supply data of emissions of specific gasses, whereas another source could supply land use information; the combination of both is needed to derive a link between the data. In our case, one grid concerns emission data, while a finer grid contains covariate data, e.g. information of land use or population data that is known to have a relation with the emissions represented in the former grid. In literature, there have been a number of approaches, ranging from simple aerial weighting to spatial smoothing and various regression methods to solve this problem [1]. In general however, it can be concluded that no exact solution is possible: the gridded data itself is usually an approximation, interpreting it to match a different grid will only increase the uncertainty or the imprecision. Unlike the aforementioned methods that describe algorithms to manipulate the data to better match the other grid, we considered a different approach, using fuzzy set theory and a fuzzy inference system.

The more detailed workings of fuzzy sets and the inference system will be explained in the next sections. The concept of the approach is that we derived rules that describe how data of one grid should be redistributed over the second grid. These rules were mainly derived from specific example cases; once implemented, the inference system applies these rules on the real data and redistributes the values accordingly. The accuracy of the result depends largely on the number of rules considered, the fuzzy sets used to represent the data and on how well the rules reflect the desired behaviour.


290

In the next section, we will introduce the fuzzy inference system. For this, a brief introduction in fuzzy set theory is required, with some explanation on fuzzy numbers and representation of linguistic terms. After this, the concept of the fuzzy rulebase and its workings can be explained. The subsequent section will elaborate on the application of the inference system in the context of the map overlay problem. First, a simple example will be used to derive the rules and explain the concept. A more advanced example will then illustrate the feasibility of this approach so far. The conclusion will summarize the findings.

2. Fuzzy inference system

2.1. Introduction to fuzzy set theory

Fuzzy set theory was introduced by Zadeh in [2] as an extension of classical set theory. In a fuzzy set, the elements are assigned a membership grade in the range [0,1]. These membership grades can have different interpretations [3]: a veristic interpretation implies that all the elements belong to some extent to the set, with the membership grade indicating the extent; whereas a possibilistic interpretation implies there is doubt on which elements belong, now the membership grade is expressing the possibility that an element belongs to the set. Last, it is also possible for the membership grades to represent degrees of truth. In [3] it was shown that all other interpretations can be traced back to

one of these three. The formal definition of a fuzzy set A~

in a universe U and its membership function

Aµ~ is given in (1)

( )( ){ }UxxµpAA

∈= |,~

~

)(

]1,0[:

~

~

xµx

Uµ

A

A

a

→

(1)

Various operations on fuzzy sets are possible: intersection and union are defined by means of functions that work on the membership grades, called respectively t-norms and t-conorms. Any function that satisfies these criteria is a t-norm, respectively t-conorm.

T-norm T-conorm

( ) ( )xyTyxT ,, = ( ) ( )xySyxS ,, =

( ) ( )dcTbaT ,, ≤ if ca ≤ and db ≤ ( ) ( )dcSbaS ,, ≤ if ca ≤ and db ≤

( )( ) ( )( )cbaTTcbTaT ,,,, = ( )( ) ( )( )cbaSScbSaS ,,,, =

( ) aaT =1, ( ) aaS =0,

Commonly used t-norms and t-conorms are the Zadeh-min-max norms, which use minimum as the intersection and the maximum as the union (other examples are limited sum and product, Lukasiewicz, ...) [4].

Fuzzy sets can be defined over any domain, but of particular interest here are fuzzy sets over the numerical domain, called fuzzy numbers [5]: the membership function represents uncertainty about a numeric value. The fuzzy set must be convex and normalized (some authors also claim the support must be bounded, but this property is not strictly necessary). Using Zadeh’s extension principle [1], it is possible to define mathematical operators on such fuzzy numbers (addition, multiplication, etc.).


291

Fuzzy sets can also be used to represent linguistic terms, such as high, low; this allows one to determine which numbers are considered high in a given context. Linguistic modifiers also exist and are usually a function that alters the membership function for the term it is associated with, allowing for an interpretation of the words like very and somewhat. It is necessary to make a distinction between an inclusive and an exclusive interpretation: are values that match very high still considered to be high? In real world, people could say about a person: “he is not tall, he is very tall”, which is an exclusive interpretation: “very tall” does not imply “tall”.

The main difficulty when using fuzzy sets is the definition of the membership functions: why are the fuzzy sets and membership grades chosen as they are, and on what information is this choice based.

2.2. Fuzzy rule base

In the fuzzy inference system, a rulebase using fuzzy premises and conclusions are used. The rulebase is comprised a set of rules that are of the form “if x is A, then y is B”. Here “x is A” is the premise and “y is B” is the conclusion; x and y are values, with x the input value and y the output value. Both are commonly represented by fuzzy sets, even though x can be a crisp value. In the rule, A and B are labels, such as “high” or “low”, also represented by fuzzy sets as described above. It is also possible to combine premises using logical operators (and, or, xor) to yield more complex rules. The “is” in the premise of the rules is a fuzzy match, implying that a value can (and most likely will) match multiple premises: a value 80 can match both “high” and “very high” albeit to a different extent. For any input (fuzzy or crisp), the process of matching the value will yield a fuzzy value indicating how well the input matches. The “is” in the conclusion is a basic assignment. It is important to note that x and y can be from totally different domains, a classic example from fuzzy control is “if temperature is high, then cooling fan speed is high”.

2.3. Interpreting the output

Typical is that all the rules are evaluated and that more than one rule can match: a value x can be classified as high to some extent and at the same time as low to much lesser extent. As multiple rules can match, y can be assigned multiple values by different rules: all these values are aggregated using a fuzzy aggregator to one single fuzzy value. For each rule, the extent to which the premise matches impacts the function that is assigned to y. While the output of the inference system is a fuzzy set, in practise the output will be used to make a decision and as such needs to a crisp value. To derive a crisp value (defuzzification), different operators exist. The centroid calculation is the most commonly used; it returns the centre of the area under the membership function.

2.4. Example

Consider the simple example of a fan controller, with 3 temperature distinctions (low, normal and high). The fuzzy sets used to indicate these distinctions are shown on Figure 1. Similarly shaped fuzzy sets are used to indicate a low, normal or high fan speed. On Figure 2, the rulebase used to link the temperatures with the a speed for the fan is shown; with only a single input, the rulebase is very straightforward.

If the given input temperature matches one membership function, the outputted value of the y is exactly the function that matches. For temperatures that match multiple rules, the value of y is calculated from the output values of all the matching rules, as illustrated on Figures 3 and 4.


292

Figure 1. The fuzzy sets used to represent

low, normal and high

Figure 2. The rulebase used for the example

of the temperature and fan

Figure 3. An input of 10 (left) and the resulting

output (right) of 13 after defuzzification

Figure 4. An input 70 (left) and the resulting

output (right) of 64 after defuzzificatio

On figures 3 and 4 it can be seen that the output is a fuzzy set, which needs to be defuzzified. There are several methods to define the defuzzification, and choosing a different method will lead to different – but very similar – results.

2. Application of the inference system

2.1. Conceptual example

2.1.1. Description

To illustrate the workings of the fuzzy inference system for the map overlay problem; first a simple conceptual example will be considered. The example consists of a grid comprised of two square grid tiles that holds emission data (em1 and em2) and a grid built up of three grid squares that holds covariate data (cov1, cov2 and cov3); illustrated on Figure 5. Both grids cover the same area, so the different tiles don’t line up properly; cov2 is split into cov2a and cov2b. While all covi and emi are known, the question is how the emission values can be distributed over the grid with covariates. This problem is equivalent to correctly distributing the covariate values of a tile over its different portions: knowing how the cov2 tile should be split is sufficient to derive an appropriate distribution of the related emission. In this simple example, the calculation can be done very easily; but this example will it allows us to derive the rules and verify results.

Figure 5. The grids used in the conceptual example

IF temp=low THEN fanspeed=low IF temp=normal THEN fanspeed=normal IF temp=high THEN fanspeed=high


293

2.1.2. Deriving the rules

In order to derive the rules for this simple example, we first consider a number of extreme cases as shown on Table 1. For ease of interpretation; all the values (both for covariates and emissions) are in the range 0-100. The first 5 rows show the known data; the rows cov2a and cov2b show how cov2 should be distributed based on the known data.

Table 1. Examples for the conceptual dataset

em1 100 100 0 100

em2 100 0 100 100

cov1 100 0 100 0

cov2 100 100 100 100

cov3 100 100 0 0

cov2a 50 normal 100 high 0 low 50 normal

cov2b 50 normal 0 low 100 high 50 normal

In the rulebase, values are compared against predefined fuzzy sets, not against each

other. To derive the rules, first assume that the covariates are equal: cov1= cov3. If the emission em1=em2, then it is obvious that cov2 should be evenly split over both cov2a and cov2b. If em1<em2, it implies that cov2 contributes more to em2 than to em1; as a result cov2b>cov2a. To make a rule that represents this case, we need to define the rule as:

for every value of A (big, small, ...). The output value clearly depends on the difference between em1 and em2: the greater this difference is (em1=very small and em2=very big), the smaller the value of cov2a should be. This yields a number of additional rules. An analogue reasoning holds when em1>em2.

Next, assume the emissions are equal: em1=em2. If cov1<cov3, then it implies that, as emissions are equal, cov2 contributes more to em1 than to em2; so cov2a>cov2b; the greater the difference between cov1 and cov3, the more this should be reflected in the output. Consequently, we obtain the rule:

This is again for every value of A, and again the greater the difference between cov1 and cov2; the more cov2a should differ from cov2b. A similar reasoning holds when cov1>cov3.

In general, neither the emissions nor the covariates will be equal. This implies that rules for those cases must be defined as well. In the current example, we considered the impact of changes to either emissions and covariates to be similar. To define the rules, we considered three predefined fuzzy sets for the emissions (representations for low, normal and high), three possible values for the covariates and nine possible values for the outputted percentage; all the fuzzy sets are shown on Figure 6. The fuzzy sets for the emissions were chosen as triangular fuzzy sets, whereas the sets for covariates and percentages are bell-shaped.

IF cov1=small AND cov2=big AND em1=A AND em2=A AND THEN cov2a=small

IF em1=small AND em2=big AND cov1=A AND cov3=A THEN cov2a=small


294

(a)

(b)

(c)

Figure 6. The fuzzy sets used to represent low, normal and high emissions (a); low, normal and high covariates (b) and the fuzzy sets used to determine the outputted percentages (c). For each variable, every function has a name mfi, starting from mf0 for the leftmost function.

Below are some examples of the rules are shown using the fuzzy sets – the whole rulebase consists of 80 rules:

2.1.3. Examples

Using the above rulebase, we can verify some of the examples. The outputted number represents which percentage of the covariate of cov2 is said to relate to cov2a.

Table 2. Verification of the example dataset

em1 100 100 0 100 em2 100 0 100 100 cov1 100 0 100 0 cov2 100 100 100 100 cov3 100 100 0 0 desired cov2a

50 100 0 50

fuzzy result

defuzzified 50 95.78 4.22 50

2.1.4. Remarks

Due to the fact that all the conditions are fuzzy, some results appear less optimal than we could envision them; this is mainly the case in the extreme values. Simply adding rules for the cases where one of the emissions or covariates that play a part in determining the portion is equal to 0 will not really help, as this does not prevent the other rules from matching. For a more optimal performance, testing for zero values and then applying

if (em_a == mf0 & em_b == mf0 & cov_a == mf0 & cov_b == mf0 ) -> cov_percentage = mf4;









295

a more customized rulebase could yield better results for those situations. For values other than these extreme cases, the outputs are nicely in between. For contradictory inputs (e.g. high emission but low covariate on one side), the results may appear a bit awkward, but this is a result of the inconsistent input.

2.2. Advanced example

2.2.1. Description

The simple example served as a means of explaining the concept. A more complicated example will be considered now. The previous example is scaled up somewhat: we now consider a 2x2 grid representing emissions and perfectly overlapping 3x3 grid containing covariates, as shown in figure 7. In this example, there basically are 3 different cases to be considered: covariate squares covered by one emission square (cov11,cov13,cov31,cov33), squares covered by two emission squares (cov12, cov21, cov23,cov32) and squares covered by 4 emission squares (cov22). As the circumstances are quite different, each of these three cases will require a different approach.

Figure 7. The grids used in the advanced example

2.2.2. Deriving the rule bases

The concept is similar as before: the emission and covariates for known and related squares is used in the premise of the rulebase. Due to the larger nature of the example, it is impossible to consider all the possible combinations of emissions and covariates like before (this would yield 3^12 cases). Various options exist to limit the number of rules. As a simple approach, we opted to consider the ratios between emissions and covariates. As a reference to determine which ratios are high and which are low the following value is used

∑

∑

==

===

2,1,0,2,1,0

1,0,1,0

cov

em

lkkl

jiij

R

Values greater than this ratio are considered to be high, values lower considered to be low.

For the first cases, the completely covered tiles cov11, cov13, cov31 and cov33, there is no need for the fuzzy inference system, as the covariate is known and needs not to be split.


296

For the second case, the tiles covered by 2 emission squares (cov12,cov21, cov23, cov32) we need to determine which portions are relevant; we will use cov12 as the example (the other three are similar); and determine the value for the portion of cov12 covered by em11. Values for the relevant ratios are needed; the neighbouring tiles that are completely covered by emission tiles are considered to determine the ratios; we will consider 2 ratios for cov12. The first ratio R1 will be defined such that it has a proportional relation to cov12a, whereas R2 will be defined to have an inverse proportional relation to cov12a. As possible definitions for R1, we have:

312111

21111 covcovcov

ememR

+++= or

3332312111

2221111 covcovcovcovcov

emememR

++++++=

The choices for R2 are similar

332313

22122 covcovcov

ememR

+++= or

3132332313


emememR

++++++=

Initial tests have shown that using either definition does not make for a big difference in the end result. Note that in the above definitions only make use of the covij that are fully covered by the emission squares considered. It is possible to also include the covij that are partly covered by the considered emission squares definitions for R1 could use the partially covered covij as well, yielding

22312111

21111 covcovcovcov

ememR

++++= or

23223332312111

2221111 covcovcovcovcovcovcov

emememR

++++++++=

There could be similar alternative definitions for R2, but this change most likely of little impact in the end result and would complicated things too much for a proof of concept. In the example, we will therefore consider the initial definitions.

The approach for the third situation, determining how cov22 should be split, is quite similar, but now different definitions for R1 and R2 are needed. To determine the portion of cov22 for the part covered by em11, the following formulas will be used:

11

111 cov

emR = or

211211

111 covcovcov

emR

++=

The choice for R2 is similar

3231332313


emememR

++++++= or

21123231332313

2221122 covcovcovcovcovcovcov

emememR

++++++++=

Using the ratios is bound to provide for less accurate results, so to compensate for this, more values have been chosen for both the relations and the covariates: we now consider 5 possible reference values for the relations, and 9 possible values for the outputted percentages, with a similar naming scheme as before. As in the conceptual example, a number of typical, predictable cases with desired results is used to derive the rulebase. For the determination of cov12a, some cases are listed in the table below.


297

em11 100 0 100 100

em12 0 100 100 0

em21 0 0 0 100

em22 0 0 0 0

cov11 100 0 100 100

cov12 100 100 100 100

cov13 0 0 100 0

cov21 0 0 0 100

cov22 0 0 0 0

cov23 0 0 0 0

cov31 0 0 0 0

cov32 0 0 0 0

cov33 0 0 0 0

R

R1 1 inf 1 1

R2 inf inf 1 inf

cov12a 100 0 50 100

Based on this table, an appropriate rulebase similar to the one for the basic example is derived. Below, some of the 25 rules are listed.

The same rulebase can be used to determine the portion of cov22, but of course using

the appropriate definitions for the inputted relations R1 and R2.

2.2.3. Results and remarks

The rulebase exhibits the expected behaviour: the portion of the covariate is estimated correctly; the four example cases listed above yield results similar to the simple example. The examples are more difficult to verify though, as changing the values of the emissions and covariates for the different cases has the side effect of changing the reference ratio R. This in turn impacts the fuzzy sets used to describe high, low and so on. So far, the rulebase has been tested with simple example, but further verification is needed. As before, some very extreme cases (e.g. covariates that are 0) can yield less than optimum results, but such cases could be detected and considered separately beforehand.

3. Optimizations

3.1. Inputs

From the two examples, it obvious that the use of the ratio decreases the accuracy. Using the actual values however would yield a rulebase of unmanageable size. It may however be possible to find better groups to use (e.g. summation of emissions and summation of relevant covariates, or multiple ratios) or devise a different rulebase altogether, and obtain a better result while still keeping a relatively small rulebase.

if(R1 == mf0 & R2 == mf0) -> cov12a = mf4;

if(R1 == mf0 & R2 == mf1) -> cov12a = mf3;

if(R1 == mf0 & R2 == mf2) -> cov12a = mf2;

if(R1 == mf0 & R2 == mf3) -> cov12a = mf1;


298

3.2. Rulebase

In the current models, very intuitive and simple rule were used. These rules should just be seen as a first step in a proof of concept. This allows it to work for many cases but still may cause it to be less successful in other cases. The use of additional technologies (e.g. neural networks) is one approach that could allow the rulebase to be determined automatically based on a much large number of cases, rather than constructed from some intuitive results. Especially in combination with the above optimization, this should yield better results.

3.3. Use more available information

Currently, some information is not used: some covariate tiles that partly covered by an emission tile are not used. The main reason for this is that the whole point is trying to determine how to split them, but of course this may be too much of a simplification for general cases. The fuzzy inference system however allows for a more fuzzy input, which makes it possible for us to derive a representative fuzzy value for these tiles; a partially covered covariate tile could be counted as contributing it surface area (as an approximation). At present, it is not clear yet how this fuzzy value should be determined, but it will be important: the risk is that introduction more fuzzy data at the inputs could make the output value too fuzzy to be truly useful.

4. Conclusion

In this contribution, we presented a novel approach to consider the map overlay problem. To determine how data should be distributed between ill aligned grids, a fuzzy inference system is used. The methodology is still in quite early development, but is showing promising results. While the current examples are simple and quite well-behaved, more complex forms of ill-alignment should be possible as the method does not use the geometry itself. Future work first concerns employing a better methodology to determine and refine the rulebase and the input, and then scaling up the methodology to larger and more complex examples. Lastly, realistic examples need to be considered to verify the results in more real world situations.

References

[1] Gotway C.A., Young L.J. (2002): Combining incompatible spatial data; Journal of the American Statistical Association, June 2002, Vol. 97 No. 458, pp. 632-648.

[2] Zadeh L.A. (1965): Fuzzy Sets; Information and Control, 1 3 (1965); pp. 338–353.

[3] Dubois D., Prade H. (1997): The three semantics of fuzzy sets; Fuzzy Sets and Systems 90, pp. 141-150.

[4] Dubois D., Prade H. (2000): Fundamentals of Fuzzy Sets. Kluwer Academic Publishers.

[5] Klir G. J., Yuan B. (1995): Fuzzy sets and fuzzy logic: Theory and applications; New Jersey: Prentice Hall.

[6] Zimmerman H-J. (1999): Practical Applications of Fuzzy Technologies; Kluwer Academic Publishers.