www.elsevier.com/locate/rse

Remote Sensing of Environm

Global biomass mapping for an improved understanding of the

CO2 balance—the Earth observation mission Carbon-3D

S. Hesea,*, W. Luchtb, C. Schmulliusa, M. Barnsleyc, R. Dubayahd, D. Knorra,

K. Neumanna, T. Riedela, K. Schrftere

aFriedrich-Schiller University Jena, Institute for Geoinformatics, Jena, GermanybPotsdam Institute for Climate Change Impact Research, Potsdam, Germany

cDepartment of Geography, University of Wales, UKdDepartment of Geography, University of Maryland, USA

eJenaOptronik GmbH, Jena, Germany

Received 5 June 2004; received in revised form 17 August 2004; accepted 6 September 2004

Abstract

Understanding global climate change and developing strategies for sustainable use of our environmental resources are major scientific and

political challenges. In response to an announcement of the German Aerospace Center (DLR) for a national Earth observation (EO) mission,

the Friedrich-Schiller University Jena and the JenaOptronik GmbH proposed the EO mission Carbon-3D. The data products of this mission

will for the first time accurately estimate aboveground biomass globally, one of the most important parameters of the carbon cycle.

Simultaneous acquisition of multiangle optical with Light Detection and Ranging (LIDAR) observations is unprecedented. The optical

imager extrapolates the laser-retrieved height profiles to biophysical vegetation maps. This innovative mission will reduce uncertainties about

net effects of deforestation and forest regrowth on atmospheric CO2 concentrations and will also provide key biophysical information for

biosphere models.

D 2004 Elsevier Inc. All rights reserved.

Keywords: Biomass; LIDAR; Carbon; Global change; Biosphere models; Carbon-3D; BRDF; CO2

1. Introduction

In 2003, the Friedrich-Schiller University Jena and the

JenaOptronik proposed the Earth observation mission

bCarbon-3DQ for global biomass mapping combining the

LIDAR system Vegetation Canopy Lidar (VCL) with a

Bidirectional Reflectance Distribution function (BRDF)

imager on one platform.

Carbon-3D improves the knowledge about spatiotempo-

ral patterns and magnitudes of major carbon fluxes between

land, atmosphere, and oceans, and allows quantifying

aboveground stocks. Because land use, land use change,

0034-4257/$ - see front matter D 2004 Elsevier Inc. All rights reserved.

doi:10.1016/j.rse.2004.09.006

* Corresponding author. Tel.: +49 3641 948873; fax: +49 3641

948882.

E-mail address: soeren.hese@uni-jena.de (S. Hese).

and forestry activities, as well as vegetation response to

enhanced levels of atmospheric CO2, are the major

influences on greenhouse gas emissions, quantifying carbon

stocks and changes is critical (Cihlar et al., 2002; Harris &

Battrick, 2001). Aboveground biomass stocks are also a key

parameter in assessing the economic, conservation, and

biofuel potential of land surfaces. The provision of a sensor

that measures these stocks and their change in space and

time is therefore paramount. Global data sets like bGlobalTree CoverQ (Fig. 1) exhibit high uncertainties for forest

cover estimates and amount of carbon stored in different

types of forest.

The mission is also relevant for contributing key data on

the environmental consequences of nonclimatic global

change resulting from continued global population growth,

economic globalisation, and expanding land use.

ent 94 (2005) 94–104

Fig. 1. Global Tree Cover in percent (white: 0%; dark green: 100%) retrieved from NOAA AVHHR (from DeFries et al., 2000). Uncertainties exist for global

forest cover estimates (between 30 and 60 Mio km2) and carbon stored in different types of forests (stock ranges from 100 to 400 tC/ha with biome). (For

interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104 95

The technical innovation of Carbon-3D is the simulta-

neous operation of a LIDAR (NASA’s Vegetation Canopy

Lidar—VCL) with a multiangle imager providing BRDF

information (Bidirectional Reflectance Distribution Func-

tion). Carbon-3D’s strength is the combined information on

both the fine-scale vertical structure of the canopy (through

waveform analysis of the vertical laser profile) and

biophysical properties of the surface targets (through multi-

angular optical observation of vegetation targets with three-

dimensional spatial structure). The BRDF information

allows extrapolation of the point measurements to complete

spatial coverage.

Large-footprint LIDAR remote sensing is a breakthrough

technology for the estimation of important forest structure

characteristics. The waveform information (Fig. 2) enables

direct determination of vegetation height, the vertical

structure of intercepted surfaces, and the subcanopy top-

Fig. 2. Direct retrieval of vertical forest structure by LIDA

ography. A critical issue and key requirement for accurate

parameter assessment is the ability to identify precisely the

top of the canopy, as well as the ground reference level.

Accurate retrieval of vertical forest parameters is essential

as other biophysical forest characteristics, such as biomass,

stem diameter, and basal area, are modelled on the base of

these measurements. Aboveground biomass is commonly

modelled using the height information by performing

regression analyses or applying allometric height–biomass

relations (Drake et al., 2003). Traditional predictive models

require information on stem diameter to estimate biomass

and volume. This parameter is a function of tree height and

thus could be inferred on the basis of LIDAR data.

Recent studies have proven the high benefit of large-

footprint scanning airborne LIDAR; depending on the

respective study area, total aboveground biomass as

estimated from field data could be predicted from airborne

R remote sensing (from Dubayah & Drake, 2000).

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–10496

LIDAR-derived metrics with R2 values of up to 0.96 on

biomass levels of 1300 t/ha, which is far exceeding the

capabilities of radar (Drake et al., 2002b; Lefsky et al.,

1999a).

One major limitation of stand-alone LIDAR systems is

their capability to retrieve locally sparse information on

vertical forest structure only. For spatially consistent biomass

assessment, information on horizontal biophysical forest

parameters, e.g., phenology and forest type, is essential. To

gain such information simultaneously, the proposed Carbon-

3D mission will be equipped with the multiangle imager in

the range of VIS/NIR/SWIR. Driven mainly by the needs of

NASA’s MISR and MODIS mission, considerable work was

done regarding the development of BRDF model inversion

algorithms for vegetation, surface, and climate parameter

retrieval (Knyazikhin et al., 1998; Lucht et al., 2000). With

the high spatial resolution of Carbon-3D, it will be possible

to exploit more complex BRDF models, i.e., physically

based approaches, for precise parameter retrieval. In

addition, the use of Look-Up Table (LUT) techniques is

expected to enable the derivation of the geophysical data

products with a high accuracy (Barnsley et al., 2000).

Data provided by the BRDF instrument onboard the

Carbon-3D satellite are also required to extrapolate spatially,

regionalizing the LIDAR point measurement on vegetation

structure between the transects and to regions where no

LIDAR data exist. The spatial extrapolation of forest

structure, as derived from locally sparse LIDAR data using

radiometer measurements, is central to the mission objectives

(see Section 5.1).

2. Strategic positioning of Carbon-3D

The role of vegetation in reducing atmospheric levels of

CO2 has been recognised in a number of international

agreements [e.g., United Nations Framework Convention on

Climate Change (UNFCCC), Kyoto Protocol], which

specifically require countries to quantify their carbon stocks

and changes. Plants store carbon in above- and belowground

biomass, 90% of the aboveground carbon is stored in tree

stems—which are being reduced through natural (diseases,

wildfires, drought/flooding) and anthropogenic impacts

(logging, pollution, human-induced fires), or vice versa

increased through regeneration or promoted growth asso-

ciated with elevated CO2 levels in the atmosphere (Wardle

et al., 2003).

Currently, the magnitude of the terrestrial carbon sink is

considerably reduced by expanding land use. In the 1990s,

terrestrial uptake has been estimated to have been 1.6–4.8

billion tons of carbon per year (GtC/year), a notable fraction

of the 6.3 GtC/year that have been emitted from fossil fuel

burning. However, losses due to land use are estimated to

have amounted to 1.4 to 3.0 GtC/year, leaving the net

uptake of carbon from the atmosphere by the biosphere to

have been about 1.0F0.8 GtC/year (House et al., 2003).

This sink is the result of the combined changes in carbon

content of vegetation, litter, and soils. Changes in the carbon

flux from vegetation into the litter and soil pools can be

estimated from the carbon pools in vegetation. The relative

magnitudes of these fluxes demonstrate the importance of

interactions between atmospheric composition, biospheric

biogeochemistry, and land use for determining the rate of

climate change.

Carbon-3D will be an essential contribution to carbon

cycle investigations by providing crucial and unique data on

vegetation biomass, vegetation productivity, and vegetation

types and structure.Model-based extensions of these data will

also allow estimations of soil carbon stocks and dynamics

with unprecedented spatial detail, and hence the land surface

carbon balance. Particularly, the spatial heterogeneity of

carbon-related vegetation properties, such as the status of

regrowth in intensely managed forests, will be quantified.

These products will fill substantial gaps in current

continental-scale carbon assessments which are specifically

related to problems of spatial heterogeneity and scaling, as

well as reliable quantifications of pool sizes. By observing

biomass around the globe, the proposed mission will build a

much needed bridge between knowledge gained at sites and

the spatially poor resolved information from atmospheric

inversion studies (Janssens et al., 2003). Biogeochemical

process models of vegetation and soil and the associated

carbon and water fluxes will benefit from the upscaling and

validation data that will be available, and it will be possible

to assimilate the observations into these simulations.

Vegetation biomass is the direct result of vegetation

productivity, that is, of the net carbon exchange between the

atmosphere and the photosynthetically active tissues of

plants. Estimates of net primary production will be

improved by not only using estimates of absorbed light, as

do current satellite-based algorithms, but also the observed

vegetation structure in terms of types and age. This will

provide a firm basis for the estimation of autotrophic

respiration, a critical but highly uncertain component of

previous satellite-driven global model estimates of net

primary production.

The implications of changes in biomass for the atmos-

pheric greenhouse gas balance and the future evolution of

climate change are critical. The temporal and spatial

variations in observations of atmospheric CO2 contain a

strong biospheric signal. There is substantial evidence from

model studies that the current role of the terrestrial

biosphere as a net sink for CO2, and therefore a brake on

the build-up of anthropogenic CO2 in the atmosphere, could

be reversed during this century as a result of climate change.

3. Modelling carbon fluxes—scientific and technological

state of the art

An observation system for carbon stocks and fluxes in

vegetation consists of both observations and modelling

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104 97

(Fig. 3). Observations characterise processes, spatial and

temporal patterns of vegetation structure and activity, while

modelling allows to infer from such observations a full

carbon account of the surfaces viewed by computing the

climate-dependent biogeochemical processes that contribute

to the overall surface budget, including the dynamics of

carbon pools not accessible to observation (e.g., in soils;

Nemani et al., 2003; Potter et al., 2003). A side effect of

such modelling is that the magnitude of evapotranspiration,

and resulting effects upon soil moisture can be estimated at

the same time due to the close coupling of the carbon and

water cycles in vegetation (Gerten et al., 2004).

3.1. In situ biomass and carbon determination

At the tree level, biomass cannot be assessed directly by

nondestructive methods but is usually derived from meas-

urements, such as diameter-at-breast-height (dbh), tree

height, and wood density using conversion factors. These

biomass estimation methods are relatively time consuming

and expensive, not uniformly standardised across the world,

and are characterized in many regions by undersampling.

Uncertainty also arises from insufficiently validated allo-

metric equations. Upscaling of inventory data to the

continental scale remains a challenge (Janssens et al.,

2003) and is possible only where the sampling density is

sufficient (Kauppi et al., 1992; Liski et al., 2003; Nabuurs et

al., 2003; Shvidenko & Nilsson, 2002).

Fig. 3. Role of Carbon-3D in the observation an

About 100 flux towers have been installed around the

globe to directly measure net carbon exchanges above

canopies (e.g., Lloyd et al., 2002). These measurements of

net fluxes do not immediately allow a differentiation between

CO2 exchanges through vegetation and CO2 exchanged by

soils. An extrapolation of the obtained data to the continental

scale (Papale & Valentini, 2003) remains uncertain, but they

provide valuable process understanding. Direct measure-

ments of atmospheric CO2 concentrations are undertaken at a

network of stations. Their spatial differences can be inverted

by taking into account wind and emission patterns, as well as

the effect of the oceans to deduce source-sink distributions

for very large regions (Bousquet et al., 1999). The results

suffer from a lack of regional pattern and insufficient

sampling but are very valuable as the only currently available

independent information on large-scale net fluxes.

The above methods are coordinated in observing net-

works, such as CARBOEUROFLUX, AMERIFLUX,

FLUXNET, and LTER (FAO, 2003), and the Terrestrial

Ecosystem Monitoring Sites (TEMS) managed by the

Global Terrestrial Observing System GTOS.

3.2. Simulation of Biomass and NPP in biosphere models

Carbon fluxes between atmosphere and biosphere can be

estimated using a range of global biosphere models that

simulate the magnitude and geographical distribution of

biomass and NPP (Cramer et al., 1999). These models range

d modelling strategy of the carbon cycle.

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–10498

in complexity from regressions between climatic variables

and one or more estimates of biospheric trace gas fluxes to

quasi-mechanistic models that simulate the biophysical and

ecophysiological processes. Several different classification

systems and descriptions of biosphere models exist in the

literature (e.g., Beringer et al., 2002; Cramer et al., 1999,

2001; Peng, 2000; Prentice et al., 2000). The simplest

differentiation is that into Static (Equilibrium) and Dynamic

Vegetation Models (SVMs and DVMs). SVMs are time-

independent and assume equilibrium conditions in climate

and terrestrial vegetation to simulate the global distribution

of potential vegetation by relating the geographic distribu-

tion of climatic parameters to vegetation. DVMs are time-

dependent process-based models that simulate the carbon

balance of ecosystems under climatically induced changes

to ecosystem structure and composition (Peng, 2000). The

great strength of DVMs is their generality and predictive

capability. By accurately representing the biophysical

processes involved, they allow calculations of the long-

term behaviour of vegetation systems under changing

climate (Cramer et al., 1999; Foley et al., 1996; Peng,

2000; Prentice et al., 2000; Sitch et al., 2003). To achieve

this, they use generalised plant functional types (PFTs) to

represent the state of vegetation in each grid cell and to

simulate plant succession processes like establishment, tree

growth, competition, and mortality.

Fire occurrence is modelled based on natural probabil-

ities as a joint function of vegetation and litter state and

climate. The majority of models addresses the potential

natural state of vegetation, although increasingly human

processes are being incorporated, most notably agriculture.

In mechanistic models, biogeochemical fluxes in the

vegetation–soil system are computed on the basis of soil

type, climate, and atmospheric carbon dioxide concentra-

tion without additional reference to external data.

Exchanges of water and carbon through the stomates of

leaves are simulated as physiologically coupled water and

carbon balances. Hydrological processes taken into consid-

eration include percolation, evaporation, and runoff, inter-

ception by vegetation, snow fall and permafrost. Carbon

assimilation is calculated from numerical simulations of

photosynthesis. The carbon gained is used for plant

respiration, growth, and reproduction following allometric

relationships that ensure functional and structural coher-

ence. Carbon fluxes into the litter and soil are estimated

from vegetation parameters. Computations of climatically

dependent soil decomposition are used to calculate the net

carbon balance of the vegetation–soil system as the

difference between vegetation net uptake and carbon release

from soils and through fire. Vegetation models have been

shown to be able to reproduce a wide range of observed

data from seasonal cycles of atmospheric carbon dioxide

concentrations and growth rates to runoff and soil moisture,

global vegetation patterns, and satellite-observed trends in

vegetation greenness (Gerten et al., 2004; Kucharik et al.,

2000; Lucht et al., 2002; McGuire et al., 2001; Sitch et al.,

2003; Wagner et al., 2003b). Due to their generalised nature

however, they are not fully capable of producing the full

spatial and temporal detail of the real world, nor can human

alterations of the land surface be inherently formulated as

mechanistic processes.

Recently, a new type of regional vegetation models,

which aim at combining process-based descriptions and

available knowledge on individual landscapes in the form of

a multilayer Geo Information System (GIS), has been

proposed (Nilsson et al., 2003; Shvidenko & Nilsson, 2002).

3.3. Biomass determination by remote sensing

Dong et al. (2003) and Myneni et al. (2001) provide

examples of using optical remote sensing data to determine

biomass with high spatial resolution at the continental scale

using NDVI. Their method of correlating seasonally

integrated measures of satellite-observed vegetation green-

ness with ground-based inventory data amounts to a spatial

interpolation and extrapolation of the inventory data. While

pioneering, the work is indirect in that direct space-based

measurements of biomass are not available. The upscaling

used relies on the unknown degree to which the sample of

inventories used is representative of the whole of a

continental area.

Correlations of Synthetic Aperture Radar (SAR) data to

biomass have been proven at low frequencies L- and P-band

(Eriksson et al., 2003; Le Toan, 1992; Ranson et al., 1997;

Rowland et al., 2002) and at C-band using ERS-1/2 repeat-

pass interferometry (Santoro et al., 2002) and combining ERS

tandem interferometric coherence and JERS backscatter

(Balzter et al., 2002; Wagner et al., 2003a). Multipolarimetric

SAR data allow interpretation of the canopy structure up to

200 t/ha aboveground biomass at P-band and 100 t/ha at L-

band (Dobson et al., 1992). Recently developed PolInSAR

techniques, combining polarimetry and interferometry

(Cloude & Papathanassiou, 1998; Papathanassiou & Cloude,

2001) provide better vegetation characterisation and a

sensitivity up to 400 t/ha (Mette et al., 2003). No such

system is foreseen yet for operational service in space, but the

PolInSAR techniques will be tested within the framework of

the bKyoto and Carbon InitiativeQ with Phased Array type L-band Synthetic Aperture Radar (PALSAR) data from the

Advanced Land Observing Satellite (ALOS; Rosenqvist et

al., 2001, 2003). ALOS is planned to be launched in 2005.

In contrast to the frequent attempts of using SAR for

biomass estimations, only a few airborne light detection and

ranging (LIDAR) missions have been developed and

validated: the Laser Vegetation Imaging Sensor (LVIS)

and the Scanning Lidar Imager of Canopies by Echo

Recovery (SLICER). Results of these studies have demon-

strated that large-footprint LIDAR instruments show great

promise in biomass estimation of tropical as well as

temperate forests due to the information about the location

of the intercepting surfaces and subcanopy topography

(Drake et al., 2002a; Lefsky et al., 1999a,b).

Table 1

BRDF instrument description

Sensor/FOV Along-track capability/F3.58�F28Directional sequence Between 3 and 7 angles

(desirable; F508, F258, 08)Spatial resolution Nadir: b25 m; off-nadir: ~50 m

Image swath Swath: 50–100 km

Spectral bands Three to five channels

(blue, red, NIR, SWIR 1.6 and 2.2 Am)

Radiometric resolution 12 bits

SNR 200:1

Data rate 8 Mbps compressed

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104 99

Due to the correlation between light absorption and net

carbon uptake by vegetation, a range of diagnostic methods

exists for converting optical satellite observations into

estimates of net primary production (e.g., Nemani et al.,

2003; Potter et al., 2003; Veroustraete et al., 2002). The

relationship between these quantities is considerably mod-

erated by effects of temperature and soil moisture, limiting

the accuracy of the assessment. The use of satellite data,

however, ensures fine spatial and temporal detail. Derivation

of biomass from these estimates requires use of a full

biogeochemical process model.

Design lifetime 3 years

4. Description of the mission

Carbon-3D is designed to ensure the overall science-

driven goal: global acquisition of a combined, synergistic

BRDF-LIDAR data set to continuously map height profiles

and reflectance to retrieve biomass (Fig. 4).

Required ground data will be obtained from terrestrial

monitoring sites linked to worldwide observation networks

and associated regional projects. The scientific project

leader is coordinating ESA’s Landcover Implementation

Office starting in 2004 and will be able to establish close

links to global landcover validation activities and the CEOS

landcover validation team. If Carbon-3D will be selected for

the phase A and B study in 2004 and 2005, the system will

be launched in 2009.

4.1. The Carbon-3D BRDF sensor

The BRDF imager is a multispectral pushbroom imager

that is sampling the Bidirectional Reflectance Distribution

Function of the target scenery (Table 1). The mission design

lifetime will be 2 years with an orbit of 390–410 km (with

Fig. 4. Carbon-3D BRDF imager and V

monthly reboost to 410 km using monopropellant hydra-

zine). The spatial resolution at nadir will be b25 m to match

the VCL resolution, allowing for approximately 50 m spatial

resolution at the extreme off-nadir angles for BRDF angular

acquisitions. The BRDF Imager will consist of two identical

Three-Mirror-Anastigmat (TMA) imaging subsystems, a

nadir imager and an off-nadir imager.

4.2. The Carbon-3D VCL sensor

The Vegetation Canopy LIDAR (VCL) instrument

consists of three near-infrared laser beams (Table 2).

The instrument is designed to operate at an orbital

altitude of 400 kmF10%. That does require orbital

maintenance every few weeks, but higher orbits have a

significant impact on the quality of the received laser signal.

Laser altimetry is the only space-based remote sensing

technique capable of measuring tree heights in closed

canopies. Waveform analyses based on extensive airborne

and spaceborne laser altimetry have revealed the need for

footprint sizes of about one to two canopy diameters. This

CL data acquisition configuration.

Table 2

VCL instrument description

Lasers 3 Nd:YAG diode-pumped pulsed

lasers at 1064 nm

Laser pulses 242 pps (land), 10 mJ per pulse

(EOL; 15 M, BOL)

Telescope 0.9 m f/1 parabolic mirror with 20 mrad

total FOV and 0.3 mrad IFOV

Waveform digitisation 250 Mega samples/s

Resolution 25-m (60 Arad) footprint diameter,

400-km altitude

Track spacing and swath 4 km (three tracks with 4 km spacing),

swath: 8 km

Elevation accuracy ~1 m in low slope terrain

Veg. height accuracy ~1 m limited by 100:1 pulse detection

dynamic range

Design lifetime 1 year, goal: 2 years

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104100

guarantees a resulting reflection from the vertical top of

canopies within the sampled area, as well as sufficient intra-

and intertree gaps required to image the underlying ground.

Sufficient laser output energy and dynamic range in the

receiver are required to detect small (~1%) returns from the

canopy top and, even in dense canopies, the underlying

ground. Smaller footprints underestimate true canopy height

(reduced probability of sampling the top of the canopy).

Conversely, with larger footprints, similar to those of the

ICESat mission (Ice, Cloud, and land Elevation Satellite)

with the Geoscience Laser Altimeter System (GLAS;

Zwally et al., 2002), the fraction of the total return

contributed by the canopy top is greatly reduced, making

height measurements inaccurate, especially in mature forests

with great height variability (ICESat’s primary scientific

objective is to measure changes in elevation of the Green-

Fig. 5. Simulated coverage for the VCL instrument (for its NASAVCL mission o

assuming 50% loss of data due to cloud obscuration. As can be seen, even under

observations to reliably estimate canopy height at 8�8 km resolution (from Dub

land and Antarctic ice sheets as part of NASA’s Earth

Observing System of satellites).

4.3. Carbon-3D in case of degradation

The proposed BRDF instrument design is generally

conceived to minimize technical risks. In the case of

degradation of one spectral band, the core products of the

Carbon-3D mission could be assessed with reliable accu-

racy. Additionally, data provided by other sensors, such as

MODIS could be used to minimize information loss,

whereby errors due to different acquisition times and spatial

resolution have to be considered. Concerning the retrieval of

LAI, the degradation of the red or NIR channel would be

critical. The channel at 2.2 Am is of special interest for the

atmospheric correction, and in case of its breakdown, data

provided by other missions have to be utilized. Furthermore,

the deterioration of the spatial resolution of the optical

instrument would be conceivable. As a result, land cover

information and vegetation parameters derived by the

optical instrument would be less precise (the BRDF instru-

ment onboard of Carbon-3D has a high resolution—25 m at

nadir and 50 m at off-nadir). Carbon-3D is designed for

vegetation parameter retrieval at global scales. For such

applications, a coarser spatial resolution would be admis-

sible, especially when considering that the LIDAR infor-

mation has to be extrapolated over large areas.

Several studies have been conducted for the VCL

instrument that explore the effects of degraded or lost

lasers. A minimum science mission of one laser operating

for 1 year would provide enough canopy data to provide an

order of magnitude increase in our knowledge of global tree

heights and thus of biomass. The simulated coverage for

rbital parameters, similar to those of Carbon-3D) for cells of 8�8 km, and

severe degradation, a majority of the Earth’s surface has sufficient LIDAR

ayah et al., 2000).

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104 101

VCL data without fusion with multispectral data is shown in

Fig. 5. With fusion of BRDF data, a degraded mission of

one laser for 1 year provides far better information on spatial

variability.

Table 3

Potentials of VCL and the Multiangle imager for deriving vegetation

parameters (partly from Dubayah et al., 2000)

LIDAR—vertical BRDF—horizontal LIDAR and

BRDF—3D

! Vegetation height ! Surface and

spectral reflectance

! Three-dimensional

structure of land cover

! Vertical distributionof intercepted

surfaces

! Surface radiance ! Improved land cover

and vegetation-type

mapping

! Subcanopytopography

! Surface emissivity

! Biomass ! Albedo! Crown volume ! Land cover

mapping

! Improved vegetation

parameter retrieval

(e.g., LAI)! Stem diameter ! Vegetation fraction

! Basal area ! Phenology! Forest age ! Chlorophyll content ! Derivation of life

form diversity

! Density of

large trees

! LAI, fAPAR, NPP

5. Carbon-3D vegetation parameters and data products

5.1. Carbon-3D LIDAR data extrapolation

Developing ways of using optical data (i.e., the multi-

angle imager) for spatial extrapolation of forest structure as

derived from locally sparse LIDAR data is central to the

mission objectives.

The challenge is similar to problems of spatial extrap-

olation routinely met in other ecological mappings based on

sampling. Extrapolating detailed measurements at a

restricted number of sites requires an understanding of the

spatial structure of the relevant system over larger areas.

Monitoring plots in forestry are routinely extrapolated to

much larger areas of forest. In agricultural monitoring, yield

models for selected sites are extrapolated to the province or

country level using Geographic Information Systems (GIS).

Estimates of the greenhouse gas exchange of large parts of

continents are based on the measurements made by a small

number of flux towers that are scaled to the respective cover

fractions of the larger area.

In the case of lidar-based samples of vegetation

biophysical properties, their extrapolation to larger areas

requires a mapping of the spatial distribution of vegetation

coverage and type. Classifying pixels of similar spectral and

angular characteristics into polygons to which LIDAR-based

vegetation properties can be scaled is currently the most

feasible and accessible approach, suggesting a combination

of a sampling sensor (the LIDAR) with an extrapolation

sensor (the optical sensor). Spectral differences allow

differentiation of polygons of different vegetation types,

while angular information provides some information on the

three-dimensional structure of this vegetation (open or a

closed canopy). Multiangular, multispectral data also pro-

vide an optimal constraint on retrievals of leaf area index.

The resultant combination of continuous parameters and

categorical polygons can then be associated with the unique

information retrieved along the LIDAR scan path.

It is important to realize that in such a scheme, the optical

sensor provides a means of extrapolating the LIDAR data,

not a means of substituting the information content derived

from it. Myneni et al. (2001) use optical data to extrapolate

forest inventory data collected on the ground to the whole of

the northern hemisphere. Veroustraete et al. (2002) use

optical data to extrapolate flux exchange measurements.

Regionalization via data fusion is based on the assumption

that there are relationships among vegetation structures and

spatially continuous optical data. One strategy is to develop

better canopy models that use sparse LIDAR data as model

initializations, and then use multispectral imagery to map the

model-derived structures (Dubayah et al., 2000). Alterna-

tively, straightforward empirical modelling procedures could

be used for data integration purposes (Dubayah et al., 2000).

Hudak et al. (2002) tested five aspatial and spatial

approaches to combine LIDAR data with limited coverage

in horizontal plane with Landsat data to predict canopy

height.

Regression, kriging, cokriging, as well as kriging and

cokriging of regression residuals, were used. They conclude

(at that time in the context of the upcoming missions ICESat

and VCL) that an integrated modelling technique is the most

efficient means to assess vegetation parameters at locations

unsampled by the LIDAR.

A combined system of LIDAR and an optical sensor

sampling the whole globe and extrapolated to complete

coverage with matching temporal and spatial characteristics

will provide a mapping opportunity for large parts of the

earth where ground information from forestry and agricul-

ture is not available nor reliable or consistent.

5.2. Data products

Carbon-3D will provide level 1 (raw data), level 2

(postprocessed data: radiance, reflectance, vertical distribu-

tion of interception), and level 3 data (empirical or semi-

empirical science algorithms that build the products;

Table 3). The combination of the BRDF and the LIDAR

information will lead to new and innovative data products

[three-dimensional structure of landcover with global cover-

age, improved landcover and vegetation type mapping,

vegetation parameter retrieval (LAI), and life form and

physiognomic diversity analysis (compare with Table 3)].

Level 4 products (NPP, forest age) will be the output from

the modelling science community.

The following list of applications are anticipated for the

Carbon-3D products: global 3-D canopy structure maps,

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104102

improved parameterisation of Global Circulation Models

(GCM), improved parameterisation of biogeochemical

vegetation models (DGVMS), improved forest parameters,

and deforestation information for forest inventories, fire

susceptibility, fuel accumulation, biodiversity, desertifica-

tion, shrub encroachment research, detection of habitat

features associated with rare or endangered species, and the

control of subsidised land use.

6. Relevance to international science-orientated

programmes

Since the early 1990s, international organisations have

been working towards the establishment of systematic, long-

term observation systems. To facilitate progress in the

challenge of obtaining and disseminating global carbon

cycle observations, space agencies and international

research programmes have established a coordination

mechanism: the Integrated Observing Strategy Partnership

(IGOS-P). IGOS-P established a Terrestrial Carbon Obser-

vation theme (TCO) under guidance of GTOS (Harris &

Battrick, 2001). Carbon-3D directly serves the IGOS

requirements and supplies the missing global biomass map

as a solid prerequisite to carbon accounting.

Further international research with respect to the global

carbon cycle is organised through the ESA Living Planet

Programme, including the Earth Observation Envelope

Programme (EOEP) and the Earth Watch Programme

(EWP), the Orbiting Carbon Observatory (OCO) mission,

and the Global Monitoring of Environment and Security

(GMES), a joint initiative of the European Commission and

the European Space Agency.

Numerous international and national political conven-

tions and multilateral agreements have been adopted

targeting global climate change [Kyoto Protocol—United

Nations Framework Convention on Climate Change

(FCCC); UNFCCC, 2000].

7. Summary

Carbon-3D will be of importance for vegetation and

carbon cycle studies, as it provides the first instrument

capable of retrieving accurate biomass information from

regional to global scales. The presented proposal for an

Earth observation mission should also stimulate the research

in that field. More work has to be done to integrate LIDAR

and radiometer-derived data sets.

Remote sensing only serves to diagnostically analyse the

current state of the vegetation. It can neither analyse the

actual flux of carbon through the system nor predict the

future development and changes of vegetation patterns.

Therefore, prognostic vegetation models are necessary for a

prediction of future sources and sinks of carbon (Lucht et

al., 2001).

The multiangle imager provides BRDF data, delivering

more comprehensive land surface information in terms of its

spectral, directional, spatial, and temporal characteristics

than data acquired from monodirectional observations

(Asner, 2000; Barnsley et al., 1997; Diner et al., 1999;

Verstraete et al., 1996). The determination of the chemical

and physical structure of land surfaces improves biophysical

modelling. Second, these data products improve existing

vegetation parameters, such as LAI, fAPAR, and NPP

(Knyazikhin et al., 1998; Roberts, 2001). Furthermore, a

reliable computation of albedo is granted improving climate

modelling. These data can only be achieved from a

multiangle instrument as on-board Carbon-3D.

Acknowledgement

Gratefully acknowledged is the work of all members of

the Carbon-3D team which responded to the announcement

of the German Aerospace Center (DLR) in 2003 for a

national Earth observation mission.

References

Asner, G. (2000). Contribution of multi-view angle remote sensing to land

surface and biogeochemical research. Remote Sensing Reviews, 18,

137–162.

Balzter, H., Talmon, E., Wagner, W., Gaveau, D., Plummer, S., Yu, J. J., et

al. (2002). Accuracy assessment of a large-scale forest cover map of

central Siberia from synthetic aperture radar. Canadian Journal of

Remote Sensing, 28(6), 719–737.

Barnsley, M. J., Allison, D., & Lewis, P. (1997). On the information content

of multiple view angle (MVA) images. International Journal of Remote

Sensing, 18, 1937–1960.

Barnsley, M. J., Lewis, P., O’Dwyer, S., Cutter, M., Disney, M. I., Hobson, P.

D., et al. (2000). Potentials of CHRIS-PROBA for mapping vegetation

characteristics from space. Remote Sensing Reviews, 19, 171–189.

Beringer, J., Mcllwaine, S., Lynch, A. H., Chapin, F. S., & Bonan, G. B.

(2002). The use of a reduced form model to assess the sensitivity of a

land surface model to biotic surface parameters. Climate Dynamics, 19,

455–466.

Bousquet, P., Peylin, P., Ciais, P., Le Quere, C., Friedlingstein, P., & Tans,

P. P. (1999). Regional changes in carbon dioxide fluxes of land and

ocean since 1980. Science, 290, 1342–1346.

Cihlar, J. R., Denning, S., Ahern, F., Arino, O., Belward, A., Bretherton, F.,

et al. (2002). Initiative to quantify terrestrial carbon sources and sinks.

Eos. Transactions, vol. 83(1),1, 6–7.

Cloude, S. R. & Papathanassiou, K. P. (1998). Polarimetric SAR

interferometry. IEEE Transactions on Geoscience and Remote Sensing,

36(5), 1551–1565.

Cramer, W., Bondeau, A., Woodward, F. I., Prentice, I. C., Betts, R. A.,

Brokvin, V., et al. (2001). Global response of terrestrial ecosystem

structure and function to CO2 and climate change: Results from

six dynamic Vegetation Models. Global Change Biology, vol. 7

(pp. 357–373).

Cramer, W., Kicklighter, D. W., Bondeau, A., Moore, III, B., Churkina, G.,

Nemry, B., et al. (1999). Comparing global models of terrestrial net

primary productivity (NPP): Overview and key results. Global Change

Biology, 5(S1), 1–15.

Diner, D. J., Asner, G. P., Davies, R., Knyazikhin, Y., Muller, J. -P., Nolin,

A. W., et al. (1999). New directions in earth observing: Scientific

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104 103

applications of multiangle remote sensing. Bulletin of the American

Meteorological Society, 80, 2209–2228.

DeFries, R., Hansen, M., Townshend, J. R. G., Janetos, A. C., &

Loveland, T. R. (2000). A new global 1 km data set of percent tree

cover derived from remote sensing. Global Change Biology, 6,

247–254.

Dobson, M. C., Ulaby, F. T., Le Toan, T., Beaudoin, A., Kasischke, E. S., &

Christensen, N. (1992). Dependence of radar backscatter on coniferous

forest biomass. IEEE Transactions on Geoscience and Remote Sensing,

vol. 30(2), 412–415.

Dong, J., Kaufmann, R. K., Myneni, R. B., Tucker, C. J., Kauppi, P. E.,

Liski, J., et al. (2003). Remote sensing estimates of boreal and

temperate forest woody biomass: Carbon pools, sources and sinks.

Remote Sensing of Environment, vol. 84 (pp. 393–410).

Drake, J. B., Dubayah, R. O., Clark, D. B., Knox, R. G., Blair, J. B.,

Hofton, M. A., et al. (2002a). Estimation of tropical forest structural

characteristics using large-footprint lidar. Remote Sensing of Environ-

ment, 79, 305–319.

Drake, J. B., Dubayah, R. O., Knox, R. G., Clark, D. B., & Blair, J. B.

(2002b). Sensitivity of large-footprint lidar to canopy structure and

biomass in a neotropical rainforest. Remote Sensing of Environment, 81,

378–392.

Drake, J. B., Knox, R. G., Dubayah, R. O., Clark, D. B., Condit, R., Blair,

J. B., et al. (2003). Above-ground biomass estimation in closed

canopy Neotropical forests using lidar remote sensing: Factors

affecting the generality of relationships. Global Ecology and

Biogeography, vol. 12 (pp. 147–159). Dubayah, R. & Drake, J. B.

(2000). Lidar remote sensing for forestry applications. Journal of

Forestry, vol. 98 (pp. 44–46).

Dubayah, R., Knox, R., Hofton, M., Blair, J. B., & Drake, J. B. (2000).

Land surface characterization using LIDAR remote sensing. In M. Hill,

& R. Aspinall (Eds.), Spatial Information for Land Use Management

(pp. 44–46). Singapore7 International Publishers Direct.

Eriksson, L. E. B., Santoro, M., Wiesmann, A., & Schmullius, C. (2003,

July). Multi-temporal JERS repeat-pass coherence for growing stock

volume estimation of Siberian forest. IEEE Transactions on Geoscience

and Remote Sensing, 41(7), 1561–1570.

FAO Food and Agricultural Organization of The United Nation, 2003:

Global terrestrial observing system GTOS-networks. http://www.fao.

org/gtos/tems/network_list.jsp, found on 08-13-03.

Foley, J. A., Prentice, I. C., Ramunkutty, N., Levis, S., Pollard, D., Sitch, S.,

et al. (1996). An integrated biosphere model of land surface processes,

terrestrial carbon balance and vegetation dynamics. Global Biogeo-

chemical Cycles, 10, 603–628.

Gerten, D., Schaphoff, S., Haberlandt, U., Lucht, W., & Sitch, S. (2004).

Terrestrial vegetation an large-scale water balance. Hydrological

evaluation of a dynamic global vegetation model. Journal of Hydrology,

286, 249–270.

Harris, R.A. & Battrick, B. (eds.), (2001), SPECTRA-Surface processes

and ecosystem changes through response analysis. In: ESA SP-1257

(5), ESTEC, Noordwijk, The Netherlands.

House, J. I., Prentice, I. C., Ramankutty, N., Houghton, R. A., & Heimann,

M. (2003). Reconciling apparent inconsistencies in estimates of

terrestrial CO2 sources and sinks. Tellus, 55B, 345–363.

Hudak, A. T., Lefsky, M. A., Cohen, W. B., & Berterretche, M. (2002).

Integration of lidar and Landsat ETM+data for estimating and mapping

forest canopy height. Remote Sensing of Environment, 82, 397–416.

Janssens, I. A., Freibauer, A., Ciais, P., Smith, P., Nabuurs, G. -J., Folberth,

G., et al. (2003). Europe’s terrestrial biosphere absorbs 7 to 12% of

European anthropogenic CO2 emissions. Science, 300, 1538–1542.

Kauppi, P. E., Mielik7inen, K., & Kuusela, K. (1992). Biomass and carbon

budget of European forests 1971–1990. Science, 256, 70–74.

Knyazikhin, Y., Martonchik, J. V., Diner, D. J., Myneni, R. B., Vertraete,

M., Pinty, B., et al. (1998). Estimation of vegetation canopy leaf area

index and fration of absorbed photosynthetically active radiation from

atmosphere-corrected MISR data. Journal of Geophysical Research,

vol. 103 (pp. 32239–32356).

Kucharik, C. J., Foley, J. A., Delire, C., Fisher, V. A., Coe, M. T., Gower, S.

T., et al. (2000). Testing the performance of a dynamic global ecosystem

model: Water balance, carbon balance and vegetation structure. Global

Biogeochemical Cycles, 14, 795–825.

Lefsky, M. A., Cohen, W. B., Acker, S. A., Parker, G. G., Spies, T. A., &

Harding, D. (1999b). Lidar remote sensing of the canopy structure and

biophysical properties of Douglas-Fir Western Hemlock Forests.

Remote Sensing of Environment, 70, 339–361.

Lefsky, M. A., Harding, D., Cohen, W. B., Parker, G. G., & Shugart, H. H.

(1999a). Surface lidar remote sensing of basal area and biomass in

deciduous forests of eastern Maryland, USA. Remote Sensing of

Environment, 67, 83–98.

Le Toan, T. (1992). Relating forest biomass to SAR data. IEEE

Transactions on Geoscience and Remote Sensing, 30(2), 403–411.

Liski, J., Korotkov, A. V., Prins, C. F. L., Karjalainen, T., Victor, D. G., &

Kauppi, P. E. (2003). Increased carbon sink in temperate and boreal

forests. Climatic Change, 61, 89–99.

Lloyd, J., Shibistova, O. B., Zolotoukhine, D., Kolle, O., Arneth, A., Wirth,

C., et al. (2002). Seasonal and annual variations in the photosynthetic

productivity and carbon balance of a central Siberian pine forest. Tellus,

54B, 590–610.

Lucht, W., Bondeau, A., Sitch, S., Cramer, W., Venevsky, S., & Thonicke,

K. (2001, January 8–12). Synergistic use of satellite data and a dynamic

model of the global biospheric carbon cycle. 8th Int. Symp. Phys. Meas.

Signat. in Remote Sensing (pp. 659–668). France7 Aussois.

Lucht, W., Prentice, I. C., Myneni, R. B., Sitch, S., Friedlingstein, P.,

Cramer, W., et al. (2002). Climatic control of the high-latitude vegetation

greening trend and Pinatubo effect. Science, 296, 1687–1689.

Lucht, W., Schaaf, C. B., & Strahler, A. H. (2000). An algorithm for the

retrieval of albedo from space using semiempirical BRDF models. IEEE

Transactions on Geoscience and Remote Sensing, 38(2), 977–998.

McGuire, A. D., Sitch, S., Clein, J. S., Dargaville, R., Esser, G., Foley, J., et

al. (2001). Carbon balance of the terrestrial biosphere in the twentieth

century: Analysis of CO2, climate and land use effects with four

process-based ecosystem models. Global Biogeochemical Cycles,

15(1), 183–206.

Mette, T., Papathanassiou, K. P., Hajnsek, I., & Zimmermann, R. (2003,

January 14–16). Forest biomass estimation using polarimetric SAR

interferometry. Proceedings of POLinSAR 2003. Italy7 Frascati.

Myneni, R. B., Dong, J., Tucker, C. J., Kaufmann, R. K., Kauppi, P. E.,

Liski, J., et al. (2001). A large carbon sink in the woody biomass of

Northern forests. PNAS, 98(26), 14784–14789.

Nabuurs, G. -J., Schelhaas, M. -J., Mohren, G. M. J., & Field, C. B. (2003).

Temporal evolution of European forest sector sink from 1950 to 1999.

Global Change Biology, vol. 9 (pp. 152–160).

Nemani, R. R., Keeling, C. D., Hashimoto, H., Jolly, W. M., Piper, S. C.,

Tucker, C. J., et al. (2003). Climate-driven increases in global

terrestrial net primary production from 1982 to 1999. Science, 300,

1560–1563.

Nilsson, S., Shvidenko, A., McCallum, I., Schmullius, C. (2003 July 21–

25). Quantification of Full Terrestrial Biota Major Greenhouse Gases

Budget at a Regional Scale: A Combination of Modeling Systems,

Geographical Information Systems and Remotely Sensed Data.

Proceedings of the International Geoscience and Remote Sensing

Symposium 2003, (IEEE International Geoscience and Remote Sensing

Symposium), Toulouse, France.

Papale, D., & Valentini, R. (2003). A new assessment of European forests

carbon exchanges by eddy fluxes and artificial neural network

spatialization. Global Change Biology, 9, 525–535.

Papathanassiou, K. P., & Cloude, S. R. (2001). Single-baseline polarimetric

SAR interferometry. IEEE Transactions on Geoscience and Remote

Sensing, 39(11), 2352–2363.

Peng, C. (2000). From static biogeographical model to dynamic global

vegetation model: A global perspective on modelling vegetation

dynamics. Ecological Modelling, 135, 33–54.

Potter, C., Klooster, S., Myneni, R., Genovese, V., Tan, P. -N., & Kumar, V.

(2003). Continental-scale comparisons of terrestrial carbon sink

S. Hese et al. / Remote Sensing of Environment 94 (2005) 94–104104

estimated from satellite data and ecosystem modelling 1982–1998.

Global and Planetary Change, 39, 201–213.

Prentice, C., Heimann, M., & Sitch, S. (2000). The carbon balance of the

terrestrial biosphere: Ecosystem models and atmospheric observations.

Ecological Applications, 10(6), 1553–1573.

Ranson, K. J., Sun, G., Lang, R. H., Chauhan, N. S., Cacciola, R. J., &

Kilic, O. (1997). Mapping of boreal forest biomass from spaceborne

synthetic aperture radar. Journal of Geophysical Research, 102(D24),

29,599–29,610.

Roberts, G. (2001). A review of the application of BRDF models to infer

land cover parameters at regional and global scales. Progress in

Physical Geography, 25(4), 483–511.

Rosenqvist, 2., Ogawa, T., Shimada, M., & Igarashi, T., (2001), The ALOS

Kyoto and Carbon Initiative—An introduction to CEOS, Proceedings

of 15th CEOS Plenary Meeting, Kyoto, Japan, 6–7 November 2001.

CEOS/15/CR/04, 2001.

Rosenqvist, 2., Shimada, M., Igarashi, T., Watanabe, M., Tadono, T., &

Yamamoto, H. (2003, July 21–25). Support to multi-national environ-

mental conventions and terrestrial carbon cycle science by ALOS and

ADEOS-II—the Kyoto and Carbon Initiative. Proceedings of

IGARSS’03, Toulouse, France.

Rowland, C., Balzter, H., Dawson, T., Luckman, A., Skinner, L., &

Patenaude, G. (2002). Biomass estimation of Thetford forest from SAR

data: Potential and limitations. ForestSAT Conference August 2002,

Edinburgh, UK.

Santoro, M., Askne, J., Smith, G., & Fransson, J. E. S. (2002). Stem volume

retrieval in boreal forests with ERS-1/2 interferometry. Remote Sensing

of Environment, 81(1), 19–35.

Shvidenko, A. & Nilsson, S. (2002). Dynamics of Russian Forests and the

Carbon Budget in 1961–1998: An assessment based on long-term forest

inventory data. Climatic Change, 55, 5–37.

Sitch, S., Smith, B., Prentice, I. C., Arneth, A., Bondeau, A., Cramer, W., et

al. (2003). Evaluation of ecosystem dynamics, plant geography and

terrestrial carbon cycling in the LPJ Dynamic Global Vegetation Model.

Global Change Biology, 9, 161–185.

UNFCCC, (2000), A guide to the Climate Change Convention and its

Kyoto Protocol.-Bonn, Germany.

Veroustraete, F., Sabbe, H., & Eerens, H. (2002). Estimation of carbon mass

fluxes over Europe using the C-Fix model and Euroflux data. Remote

Sensing of Environment, 83, 376–399.

Verstraete, M. M., Pinty, B., & Myneni, R. (1996). Potential and limitations

of information extraction on the terrestrial biosphere from satellite

remote sensing. Remote Sensing of Environment, 58, 201–214.

Wagner, W., Luckman, A., Vietmeier, J., Tansey, K., Balzter, H.,

Schmullius, C., et al. (2003a). Large-scale mapping of boreal forest

in Siberia. Remote Sensing of Environment, 85, 125–144.

Wagner, W., Scipal, K., Pathe, C., Gerten, D., Lucht, W., & Rudolf, B.

(2003b). Evaluation of the agreement between the first global remotely

sensed soil moisture data with model and precipitation data. Journal of

Geophysical Research, 108, 4611.

Wardle, D. A., Hfrnberg, G., Zackrisson, O., Kalela-Brundin, M., &

Coomes, D. A. (2003). Long-term effects of wildfire on ecosystem

properties across an island area gradient. Science, 300, 972–975.

Zwally, H. J., Schutz, B., Abdalati, W., Abshire, J., Bentley, C., Brenner,

A., et al. (2002). ICESat’s laser measurements of polar ice, atmosphere,

ocean and land. Journal of Geodynamics, 34, 405–445.