copernicus global land operations vegetation and energy · copernicus global land operations –...

Copernicus Global Land Operations – Lot 1

Date Issued: 21.10.2019

Issue: I3.22

Copernicus Global Land Operations

”Vegetation and Energy” “CGLOPS-1”

Framework Service Contract N° 199494 (JRC)

PRODUCT USER MANUAL

DRY MATTER PRODUCTIVITY (DMP)

GROSS DRY MATTER PRODUCTIVITY (GDMP)

COLLECTION 1KM

VERSION 2

Issue I3.22

Organisation name of lead contractor for this deliverable: VITO

Book Captain: Bruno Smets

Contributing Authors: Else Swinnen

Roel Van Hoolst



Issue: I3.22

Document-No. CGLOPS1_PUM_DMP1km-V2 © C-GLOPS1 consortium

Issue: I3.22 Date: 21.10.2019 Page: 2 of 47

Dissemination Level PU Public X

PP Restricted to other programme participants (including the Commission Services)

RE Restricted to a group specified by the consortium (including the Commission Services)

CO Confidential, only for members of the consortium (including the Commission Services)



Issue: I3.22



Document Release Sheet

Book captain: Bruno Smets

Sign Date 21.10.2019

Approval Roselyne Lacaze

Sign Date 30.10.2019

Endorsement: Michael Cherlet Sign Date

Distribution: Public



Issue: I3.22



Change Record

Issue/Rev Date Page(s) Description of Change Release

28.03.2013 All geoland2 document, Issue I1.00, 10.12.2010 I1.00

I1.00 09.02.2015 Adaptation to PROBA-V products I2.00

I2.00 15.01.2016 41 Annex 1 added after external review I2.10

I2.10 20.12.2016 All Updated to Version 2 of DMP products I3.00

I3.00 05.07.2017 34-42 Updated after external review I3.10

I3.10 20.01.2018 22-23 Split DMP and GDMP in two products I3.20

I3.20 08.02.2018 27-32 Update QFlag and netCDF attributes tables I3.21

I3.21 21.10.2019 46-47 Update the Users Requirements chapter to be

compliant with the Applicable Documents I3.22



Issue: I3.22



TABLE OF CONTENTS

Executive Summary .................................................................................................................. 11

1 Background ....................................................................................................................... 12

1.1 Scope and Objectives............................................................................................................. 12

1.2 Content of the document....................................................................................................... 12

1.3 Related documents ............................................................................................................... 12

1.3.1 Applicable document .................................................................................................................................. 12

1.3.2 Input ............................................................................................................................................................ 12

1.3.3 External reference ...................................................................................................................................... 13

2 Algorithm .......................................................................................................................... 14

2.1 Overview .............................................................................................................................. 14

2.2 Retrieval Methodology .......................................................................................................... 14

2.2.1 Background ................................................................................................................................................. 14

2.2.2 History of the product ................................................................................................................................. 15

2.2.3 Inputs .......................................................................................................................................................... 16

2.2.4 Processing steps .......................................................................................................................................... 17

2.2.5 Near Real Time retrieval of the DMP 1km Version 2 .................................................................................. 20

2.3 Limitations ............................................................................................................................ 21

3 Product Description ........................................................................................................... 23

3.1 File Format ............................................................................................................................ 23

3.2 Product Content .................................................................................................................... 25

3.2.1 Data Files ..................................................................................................................................................... 25

3.2.2 QFLAG image ............................................................................................................................................... 26

3.2.3 File and Layer attributes ............................................................................................................................. 27

3.2.4 ‘quicklook’ image ........................................................................................................................................ 31

3.3 Product Characteristics .......................................................................................................... 32

3.3.1 Projection and grid information ................................................................................................................. 32

3.3.2 Spatial information ..................................................................................................................................... 32

3.3.3 Temporal information ................................................................................................................................. 32

3.4 Data Policies ......................................................................................................................... 32

3.5 Contacts ................................................................................................................................ 34

4 Quality Assessment ........................................................................................................... 35



Issue: I3.22



5 Product Usage ................................................................................................................... 37

5.1 Typical use ............................................................................................................................ 37

5.2 Advanced usage: how to analyze the data .............................................................................. 38

5.3 Exploitation in empirical regressive modelling ........................................................................ 42

6 References ........................................................................................................................ 44

Annex: Review of Users Requirements ...................................................................................... 46



Issue: I3.22



List of Figures

Figure 1: The component fluxes and processes in ecosystem productivity. GPP: Gross Primary

Production, NPP: Net Primary Production, NEP: Net Ecosystem Production, NBP: Net Biome

Production (Valentini, 2003) ................................................................................................... 15

Figure 2: Global land cover map (ESA CCI epoch 2010) derived from ENVISAT MERIS and SPOT-

VGT data. .............................................................................................................................. 17

Figure 3: Global overview of the optimized Light Use Efficiency (LUE) values, calibrated with

FLUXNET data, and assigned to the CCI land covers. ........................................................... 18

Figure 4: Process flow of GDMP/DMP. Based on meteo data a daily DMPmax is estimated. At the

end of each dekad, a Mean Value Composite of these DMPmax images is calculated. At the

same time, a fAPAR product is generated. The final DMP10 product is retrieved by the simple

multiplication of the latter two images..................................................................................... 20

Figure 5: Chronograph showing the several periods considered and the associated branches (B,

C-, C+) used to process the data. .......................................................................................... 21

Figure 6: Regions of DMP/GDMP product. Red lines indicate the boundaries on availability of

meteorological data. ............................................................................................................... 25

Figure 7 : Color coding for quicklook images ................................................................................. 32

Figure 8 : Quicklook of DMP for Sudan, 2nd dekad of August 2006. Red = low DMP, green = high

DMP ...................................................................................................................................... 37

Figure 9 : Western Africa, last dekad of July 2006. Absolute difference in DMP with respect to long

term average. Green = better situation, Red = worse ............................................................. 38

Figure 10 : Sahel, seasonal dry matter productivity as calculated by the Biogenerator software of

Action Contre la Faim (http://www.actioncontrelafaim.org/en). Yellow = low cumulative DMP,

Dark green = high cumulative DMP. Map obtained from Ham et al. (2012). ........................... 39

Figure 11 : profiles for DMP for winter cereals in Belgium derived with 6 different un-mixing

methods: Regional mean, Hard classification, CNDVI threshold = 20%, CNDVI threshold =

50%, Linear un-mixing and Neural network ............................................................................ 40

Figure 12 : Regional unmixed means of DMP. Example for the Horn of Africa. ............................. 41

Figure 13: Online dashboard of the time series viewer as provided by Copernicus Global Land

Service (http://viewer.globalland.vgt.vito.be/tsviewer/). .......................................................... 42

Figure 14: Temporal trends of predicted yield calibrated by leave-one-out cross validation

technique ............................................................................................................................... 43



Issue: I3.22



List of Tables

Table 1: Individual terms in the Monteith variant used for Global Land service GDMP/DMP. All

terms are expressed on a daily basis. .................................................................................... 17

Table 2: Definition of the continental tiles distributed via Eumetcast.............................................. 25

Table 3: Range of values and scaling factors of GDMP and DMP. ................................................ 26

Table 4: Quality Flag indicator for GDMP/DMP Version 2. ............................................................ 26

Table 5: Description of netCDF file attributes ................................................................................ 28

Table 6: Description of netCDF layer attributes ............................................................................. 29

Table 7: Description of netCDF coordinate variable attributes ....................................................... 30

Table 8: Description of netCDF grid mapping variable attributes ................................................... 30

Table 9: Description of netCDF attributes for time coordinate variable .......................................... 31

Table 10 : QuickLook - value signification ..................................................................................... 31



Issue: I3.22



List of Acronyms

AD Applicable Document

AFI Area Fraction Image

AR Autotrophic Respiration

ATBD Algorithm Theoretical Basis Document

AVHRR Advanced Very High Resolution Radiometer

CCI Climate Change Initiative, an ESA program

CEOS Committee on Earth Observation System

CGLS Copernicus Global Land Service

CNDVI Cumulative NDVI

CNES Centre National d’Etudes Spatiale

CSV Comma Separated Values

CTIV Centre de Traitement des Images VEGETATION

DM Dry Matter

DMP Dry Matter Productivity

DN Digital number

Efficiency

EBF Evergreen Broadleaf Forest

EC European Commission

ECMWF European Centre for Medium-Range Weather Forecasts

ECV Essential Climate Variable

ENVISAT Environment satellite from ESA, contains MERIS instrument

ERA-Interim A global atmospheric reanalysis dataset from ECMWF

ESA European Space Agency

fAPAR Fraction of absorbed photosynthetically active radiation

FP7 Framework Programme 7

GAUL Global Administrative Unit Layers

GCOS The Global Climate Observing System

GDMP Gross Dry Matter Productivity

GIO GMES Initial Operations

GLC Global Land Cover

GLIMPSE Global image processing software

GPP Gross Primary Productivity

Ha Hectares

HDF Hierarchical Data Format

ISLSCP International Satellite Land-Surface Climatology Project Initiative

JRC Joint Research Centre

LPJ-DGVM Lund-Potsdam-Jena Dynamic Global Vegetation Model

LPV Land Product Validation group of CEOS

LUE Light use efficiency



Issue: I3.22



MARS Monitoring Agricultural ResourceS

MARSOP MARS-operations

MERIS Medium Resolution Imaging Spectrometer

MODIS Moderate Resolution Imaging Spectroradiometer

NBP Net Biome productivity

NDVI Normalized Difference Vegetation Index

NEE Net-Ecosystem exchange

NEP Net Ecosystem productivity

NetCDF Scientific data Network Common Format

NGO Non Governmental Organization

NPP Net Primary Productivity

NRT Near Real Time

PAR Photosynthetically active radiation

PROBA-V Vegetation instrument on PROBA platform, Belgian mission

PUM Product User Manual

R Radiation

RMSE Root Mean Square Error

RS Remote Sensing

SPOT Système Pour l’Observation de la Terre

T Temperature

UN-LCCS United Nations Land Cover Classification System

VGT VEGETATION sensor onboard SPOT4/5

VITO Flemish Institute for Technological Research

VR Validation Report

WGS World Geodetic System

XML Extensible Markup Language

XSL Extensible Style sheet Language



Issue: I3.22



EXECUTIVE SUMMARY

The Copernicus Global Land Service (CGLS) is earmarked as a component of the Land Monitoring

service to operate “a multi-purpose service component” that provides a series of bio-geophysical

products on the status and evolution of land surface at global scale. Production and delivery of the

parameters take place in a timely manner and are complemented by the constitution of long term

time series.

Since January 2013, the Copernicus Global Land Service is continuously providing Essential

Variables like the Leaf Area Index (LAI), the Fraction of Absorbed Photosynthetically Active

Radiation absorbed by the vegetation (FAPAR), the surface albedo, the Land Surface

Temperature, the soil moisture, the burnt areas, the areas of water bodies, and additional

vegetation indices, which are generated every hour, every day or every 10 days on a reliable and

automatic basis from Earth Observation satellite data.

This document describes the CGLS Dry Matter Productivity (DMP) Collection 1km Version 2

product, derived from SPOT/VEGETATION and PROBA-V. The product provides Gross Dry Matter

Productivity (GDMP) and Dry Matter Productivity (DMP) information, calculated from the CGLS

FAPAR Collection 1km Version 2. The product is generated globally and made available to the

user every 10 days, with a maximum of 3 days lag in Near Real Time (NRT), followed by

consolidations in the course of the next 6 dekads.



Issue: I3.22



1 BACKGROUND

1.1 SCOPE AND OBJECTIVES

The Product User Manual (PUM) is the primary document that users have to read before handling

the products. It gives an overview of the product characteristics, in terms of algorithm, technical

characteristics, and main validation results.

The product is derived from the CGLS Fraction of Absorbed Photosynthetically Active Radiation

(FAPAR) Collection 1km Version 2 [CGLOPS1_ATBD_FAPAR1km-V2], and hence follows the

delivery scheme of this dataset [CGLOPS1_PUM_FAPAR1km-V2].

1.2 CONTENT OF THE DOCUMENT

This document is structured as follows:

Chapter 2 presents a description of the algorithm

Chapter 3 describes the technical characteristics of the product

Chapter 4 summarizes the main results of the quality assessment exercise

Chapter 5 gives examples of the product usage

The users’ requirements are recalled in Annex.

1.3 RELATED DOCUMENTS

1.3.1 Applicable document

AD1: Annex I – Technical Specifications JRC/IPR/2015/H.5/0026/OC to Contract Notice 2015/S

151-277962 of 7th August 2015

AD2: Appendix 1 – Copernicus Global land Component Product and Service Detailed Technical

requirements to Technical Annex to Contract Notice 2015/S 151-277962 of 7th August 2015

AD3: GIO Copernicus Global Land – Technical User Group – Service Specification and Product

Requirements Proposal – SPB-GIO-3017-TUG-SS-004 – Issue I1.0 – 26 May 2015.

1.3.2 Input

Document ID Descriptor

CGLOPS1_ATBD_DMP1km-V2 Algorithm Theoretical Basis Document of the Dry Matter

Productivity Collection 1km Version 2

CGLOPS1_ATBD_FAPAR1km- Algorithm Theoretical Basis Document of the LAI, FAPAR,



Issue: I3.22



V2 FCover Collection 1km Version 2 products

CGLOPS1_QAR_DMP1km-V2 Quality Assessment Report of the Dry Matter Productivity

Collection 1km Version 2 products

CGLOPS1_QAR_FAPAR1km-

PROBAV-V2

Quality Assessment Report of the PROBA-V FAPAR,

Collection 1km Version 2 product

GIOGL1_QAR_FAPAR1km-VGT-

V2

Quality Assessment Report of the SPOT/VGT FAPAR

Collection 1km Version 2 product

CGLOPS1_PUM_FAPAR1km-V2 Product User Manual of the LAI, FAPAR, Fcover Collection

1km Version 2 products

GIOGL1_VR_DMP1km-V1 Validation Report of the DMP Collection 1km Version1

products.

1.3.3 External reference

SPOT-VGT http://www.spot-vegetation.com/index.html

PROBA-V http://proba-v.vgt.vito.be/

http://www.spot-vegetation.com/index.html

http://proba-v.vgt.vito.be/



Issue: I3.22



2 ALGORITHM

2.1 OVERVIEW

DMP, or Dry Matter Productivity, gives an indication of the dry matter biomass increase (growth

rate) of the vegetation and is directly related to the well-known NPP (Net Primary Productivity), but

customized for agro-statistics and expressed in kilograms of dry matter (kg DM) per hectare per

day.

DMP images, in this case derived from SPOT-VGT (until December 2013) and PROBA-V (from

January 2014) imagery combined with (modelled) meteorological data from ECMWF, are made

available at 1km resolution and are updated every 10 days.

2.2 RETRIEVAL METHODOLOGY

2.2.1 Background

The productivity of an ecosystem is a fundamental ecological variable. It measures the energy

input to the biosphere and terrestrial carbon dioxide assimilation. Monitoring ecosystem carbon

fluxes is a key issue in mitigating global change and widely used in climatology studies. In general,

ecosystem productivity is a good indicator for the condition of the land surface area and status of a

wide range of ecological processes (Gower et al. 2001). It is a unique integrator of climatic,

ecological, geochemical and human influences (Running et al. 2009) and a suitable indicator for

agronomic analyses.

Ecosystem productivity is a complex and dynamic system characteristic influenced by many

factors. Theoretically it can be expressed in the form of carbon fluxes. Figure 1 gives a schematic

representation of the different component fluxes and processes. Gross primary production (GPP) is

the production of organic compounds from atmospheric carbon through the process of

photosynthesis. Net primary production (NPP), is the difference between gross primary production

(GPP) and autotrophic respiration. Net ecosystem production (NEP) is the net amount of primary

production after the costs of autotrophic respiration by plants and heterotrophic decomposition of

soil organic matter. Net biome production (NBP) is the amount of carbon that remains in the

vegetation after anthropogenic removals and losses by disturbances.

The most used ecosystem productivity variables in earth observation are GPP and NPP. Accurate

estimations of NEP and especially NBP with ecosystem models are currently hampered by high

uncertainties in the model results (Luyssaert, et al., 2010).



Issue: I3.22



Figure 1: The component fluxes and processes in ecosystem productivity. GPP: Gross Primary

Production, NPP: Net Primary Production, NEP: Net Ecosystem Production, NBP: Net Biome

Production (Valentini, 2003)

Relation between Net Primary Production (NPP) and Dry Matter Productivity (DMP)

DMP, or Dry Matter Productivity, represents the overall growth rate or dry biomass increase of the

vegetation, expressed in kilograms of dry matter per hectare per day (kgDM/ha/day). DMP is

directly related to NPP (Net Primary Productivity, in gC/m²/day), but its units are customized for

agro-statistical purposes.

1 kgDM/ha/day = 1000 gDM/ha/day = 0.1 gDM/m2/day (1)

According to Atjay et al. (1979), the efficiency of the conversion between carbon and dry matter is

on the average 0.45 gC/gDM. So NPP and DMP only differ by a constant. In practice, to scale

DMP to NPP following calculation should be done:

NPP [gC/m2/day] = DMP [kgDM/ha/day] * 0.45 * 0.1 (2)

Similar to the relation between NPP and DMP also the agronomic equivalent of the GPP (Gross

Primary Productivity) can be defined as GDMP (Gross Dry Matter Productivity). Scaling from NPP

to GDMP or oppositely is similar to the scaling described above. The main difference between the

NPP/DMP and GPP/GDMP is the inclusion of the autotrophic respiration.

2.2.2 History of the product

A simplified Monteith (1972) approach, using remote sensing data to provide the fraction of

absorbed PAR-radiation (fAPAR), was first implemented by Veroustraete (1994) and provided NPP

estimates for European forests. This basic algorithm was consequently applied for entire Europe in

the frame of the MARSOP project to produce DMP-estimates on a regular basis since around



Issue: I3.22



2000, with unaltered parameterization. It was also implemented in the GLIMPSE software (Eerens

et al., 2014). A parallel development, the C-Fix model focusing on NPP (Veroustraete et al., 2002,

Veroustraete et al., 2004, Verstraeten et al., 2006), tried to improve the basic algorithm with the

implementation of a water stress factor, a temperature compartment model and biome specific

LUE values. Despite reasonable success over Europe (Verstraeten et al., 2010), this model has

not been implemented in an operational context.

Over the years different versions of the DMP product arose which were distributed via separate

channels to the user community. Since 2004, VITO produces global DMP-images derived from the

10-daily syntheses of SPOT-VEGETATION on behalf of the FOODSEC unit of JRC-MARS. Since

2007, part of this information was further distributed to users in Africa and South-America via the

GeoNetCast service by another VITO-project called DevCoCast (or its predecessor VGT4Africa).

In the summer of 2009, it was decided within the MARS-group to launch a new DMP version, using

improved inputs (meteorological data and fAPAR). The DMP Version 1, produced in the

Copernicus Global Land Service (CGLS), is using the same meteo input data, algorithm and

constants as for this latter MARSOP DMP. The MARSOP3 project ended in 2014 but the

Copernicus Global Land Service, uses the same meteorology input data, algorithm and constants

as for this latter MARSOP DMP. This version of the GL DMP is referred to as version 1. Recently,

an improvement action in Copernicus Global Land Service, led to the DMP Version 2. This latter

version has removed unreliable low fAPAR values through gap filling and uses biome specific light

use efficiencies, and is described in detail in CGLOPS1_ATBD_DMP1km-V2.

2.2.3 Inputs

Meteodata: Up till 2014, the global meteo data were delivered by MeteoConsult in the frame of

MARSOP3, in the form of daily CSV-files, providing for each “grid-cell” (at 0.25° resolution) the

values of all standard meteorological variables. Basically, all daily data are “operational forecasts

for the next 24 hours” derived from ECMWF (ERA-Interim for the years 1989-2008). Radiation, Tmin

and Tmax are derived from these dataset for the DMP calculations. The Copernicus Global Land

Service adopted the retrieval of the global meteorological data from MeteoConsult in dito format.

fAPAR: fAPAR is estimated from the SPOT-VGT (until December 2013) and PROBA-V (since

January 2014) data. The currently used fAPAR is the CGLS fAPAR Collection 1km Version 2,

produced every 10 days at 1km resolution. It is described in the ATBD

[CGLOPS1_ATBD_FAPAR1km-V2].

The fAPAR Version 2 benefits of three major improvements compared to the fAPAR used in DMP

Version 1. They are:

Improvement of the smoothness of the product

Less missing data which means a higher continuity in the time series

Higher accuracy when validated with in-situ measurements.



Issue: I3.22



Land Cover information: The DMP v2 uses biome specific Light Use Efficiency (LUE) values. The

information on the global distribution of land cover types comes from the ESA CCI Land Cover

Map (epoch 2010), which was derived from ENVISAT-MERIS and SPOT-VGT imagery of the

period 2008-2010. The land cover classes are based on the UN Land Cover Classification System

(LCCS). The global map is shown in Figure 2.

Figure 2: Global land cover map (ESA CCI epoch 2010) derived from ENVISAT MERIS and SPOT-VGT

data.

2.2.4 Processing steps

The GDMP and DMP are currently computed with the following Monteith variant (see Table 1):

GDMP = R.c.fAPAR.LUEc.T.CO2[.RES] (3)

DMP = R.c.fAPAR.LUEc.T.CO2.AR[.RES] (4)

Table 1: Individual terms in the Monteith variant used for Global Land service GDMP/DMP. All terms

are expressed on a daily basis.

TERM MEANING VALUE UNIT

GDMP Gross Dry Matter Productivity 0 - 640 kgDM/ha/day

DMP Dry Matter Productivity 0 – 320 kgDM/ha/day

R Total shortwave incoming radiation (0.2 – 3.0µm) 0 – 320 GJT/ha/day

c Fraction of PAR (0.4 – 0.7µm) in total shortwave 0.48 JP/JT

fAPAR PAR-fraction absorbed (PA) by green vegetation 0.0 ... 1.0 JPA/JP

LUEc Light use efficiency (DM=Dry Matter) at optimum Biome-specific kgDM/GJPA

T Normalized temperature effect 0.0 ... 1.0 -

CO2 Normalized CO2 fertilization effect 0.0 ... 1.0 -

AR Fraction kept after autotrophic respiration 0.5 -

RES Fraction kept after omitted effects (drought, pests...) 1.0 -



Issue: I3.22



Some remarks to Table 1:

Incoming solar (shortwave) radiation R is mostly reported in terms of kJT/m²/day with variations

between 0 and 32 000. This corresponds with 320 GJT/ha/day (1 hectare is 10 000m², and 1 GJ

is 1 000 000 kJ).

The fraction c of PAR (Photosynthetically Active Radiation, 0.4-0.7 µm) within the total

shortwave (0.2 – 3.0µm) range is set to a global constant of c =0.48 according to McCree

(1972).

fAPAR is derived from SPOT/VGT (until 2013) and PROBA-V (from 2014) as 10-daily

composites with the Version 2 method [CGLOPS1_ATBD_FAPAR1km-V2].

LUE is biome-specific and estimated by calibrating the GDMP with flux tower GPP

measurements. Figure 3 shows the global map when these LUE values are assigned to the CCI

land covers. The LUE values holds for optimal conditions.

Figure 3: Global overview of the optimized Light Use Efficiency (LUE) values, calibrated with

FLUXNET data, and assigned to the CCI land covers.

The behavior of T is simulated with the model of Wang (1996).

The CO2 fertilization effect, CO2, is defined as the increase in carbon assimilation due to CO2

levels above the atmospheric background level (or reference level). This effect is influenced by

the temperature. In version 1, the yearly dynamics of the atmospheric concentration was fixed at

355.61 ppmv. In version 2, the atmospheric CO2 concentration is implemented as a linear

function of the calendar year as:

(5)

The factor AR represents the autotrophic respiration component. To avoid any uncertainty

through modelling this factor, as done in Version 1, a constant factor of 0.5 is used in Version 2.



Issue: I3.22



The factor RES (residual) is only added in equation (3) to emphasize the fact that some

potentially important factors, such as the effect of droughts, nutrient deficiencies, pests and plant

diseases, are clearly omitted here. As a consequence, the product might better be called

“potential” DMP. On the other hand, it might be argued that the adverse effects of diseases and

shortages of water or nutrients are manifested (sooner or later) via the RS-derived fAPAR.

Given the simple elaboration of the epsilons (and dropping the factor RES), equation (3) and

equation (4) can be rewritten as:

GDMP = R.c.fAPAR.LUEc.T.CO2 = fAPAR.LUEc.R.(T,CO2) = fAPAR.LUEc.GDMPmax (5)

DMP = GDMP.AR = GDMP . 0.5 (6)

with (T,CO2)=c.T.CO2 and LUEc is the land cover specific LUE. This formulation better highlights

the fact that (within the limits of the described model) GDMP is only determined by four basic

factors: fAPAR, radiation (R), temperature (T) and CO2. At the same time, equation (5) provides a

practical method which allows bypassing the differences in temporal and spatial resolution

between the inputs. In practice, the meteorological inputs (R, T) are provided on a daily basis and

at a very low resolution, here 0.25° x 0.25°. On the contrary, fAPAR is provided on a dekadal

basis, having pixels with a size of around 1 km². The ESA CCI land cover map is also applied at 1

km resolution to include the Light Use Efficiency (LUE or LUEc) factor. Then, the final GDMP/DMP-

product has this 1km resolution and dekadal frequency.

So, in practice, we use the procedure (Eerens et al., 2004) illustrated in Figure 4 and can be

formulated as follows (the subscripts 1 and 10 indicate daily and dekadal products, Nd is the

number of days in each dekad)

GDMP10 = fAPAR10 . LUEc . GDMPmax,10 with GDMPmax,10 = {GDMPmax,1} /Nd (8)

DMP10 = GDMP10 . AR with AR = 0.5 (9)



Issue: I3.22



Figure 4: Process flow of GDMP/DMP. Based on meteo data a daily DMPmax is estimated. At the end

of each dekad, a Mean Value Composite of these DMPmax images is calculated. At the same time, a

fAPAR product is generated. The final DMP10 product is retrieved by the simple multiplication of the

latter two images.

2.2.5 Near Real Time retrieval of the DMP 1km Version 2

The production of the DMPv2 is managed similarly to the production of FAPAR Collection 1km

Version 2. A brief description is given below but the reader is also referred to the product user

manual of this fAPAR [CGLOPS1_ATBD_FAPAR1km-V2]. A distinction is made between the past-

time series production and the near real time products.

The past-time series are defined as past observations where, for a given dekad, the ‘n’

dekads before and after are available. ‘n’ is the number of dekads required for

convergence, i.e. the length (in dekad) of the convergence period (see Figure 5). ‘n’ is fixed

to 6.

The real time products are defined as the most recent observations, for which no full

convergence period (n=6 dekads after) is available. These images are referred to as RT0

(most recent dekad) till RT5 (fifth most recent dekad). RT6 is than the “final” and

consolidated version. This temporal window of six dekads shifts each time a new dekad is

processed. This results in six successive updates of the fAPAR that converge (from RT0 till

RT6) towards the past time series values. The DMP follows this approach. In practice, the



Issue: I3.22



product is rather stable from RT2 onwards hence only RT0, RT1, RT2 and RT6 are

delivered to the user.

Note that for the first past dekads (from d1 to d1+n), no past data is available (Figure 5). Here,

Branch C is run in reverse mode (it is called C- as opposed to the forward mode for real time

estimation called C+).

Figure 5: Chronograph showing the several periods considered and the associated branches (B, C-,

C+) used to process the data.

In summary, the most recent NRT derived DMP images are temporary files, which are updated 3

times before they are labelled as “stable and final” version.

2.3 LIMITATIONS

Following remarks should be taken into consideration when interpreting the DMP products:

The biome-specific LUE’s are calibrated against flux tower Eddy Covariance (EC) data. These

flux data are considered as ground truth measurements of GPP. But in fact, this GPP is a

combination of Net Ecosystem Exchange (NEE) measurements and a modelled respiration factor.

Additionally, there are other reasons to question the relation between flux measurements and

modelled GPP estimates. The spatial representativeness of the tower, for example, is strongly

dependent on local conditions like the towers footprint, wind direction, etc. Thus, caution should be

made when comparing 1 km pixel values with tower observations. Each tower also has its specific

environmental conditions like soil properties, precipitation, elevation,… This makes difficult to

generalize tower specific parameterization into broad categories like land cover.

Convergence PeriodConsolidated PeriodHistoric Period

dx+1dx dx+n

BC-

Past Projection

C+

d1 d1+n dx-n



Issue: I3.22



The biome-specific parametrization requires accurate information on the land cover type. At

global scale, a land cover map will always contain a significant fraction of falsely classified pixels.

Some potentially important factors, such as drought stress, nutrient deficiencies, pests and plant

diseases, are omitted in the DMP product. As a consequence, the product might better be called

“potential” DMP. On the other hand, it might be argued that the adverse effects of drought,

diseases and shortages of nutrients are manifested (sooner or later) via the RS-derived fAPAR.

The DMP algorithm does not include a water stress factor to account for short term drought

stresses. In drought sensitive vegetation types, this can lead to an overestimation of the actual

plants productivity. The temperature dependency factor uses generalized values derived from the

parameterization on European forests (Veroustraete et al., 2002) and these parameterization may

not hold for other biome types and agro-ecological zones. Especially the difference between C3

and C4 vegetation types is pronounced. C3 plants (like the forests used in the current calibration)

will have their optimal temperate around 20-25 degrees whereas C4 plants will reach their optimum

around 30-35 degrees (Massad et al., 2007; Yamori et al., 2013).

The autotrophic respiration is calculated as a simple fraction of GPP and is therefore assumed to

have the same ecophysiological behaviour. In other models (e.g. used for MODIS product

retrieval), the autotrophic respiration is considered as an independent component.



Issue: I3.22



3 PRODUCT DESCRIPTION

The Dry Matter Productivity Collection 1km Version 2 (DMPv2) products follow the naming

standard:

c_gls_<product>[-<RTx>]_<YYYYMMDDhhmm>_<AREA>_<SENSOR>_V<Major.Minor.Run>

where

<product> is either DMP (Dry Matter Productivity) or GDMP (Gross Dry Matter Productivity)

<RTx> is an optional parameter. It is uses whenever a real-time product is provided:

RT0: Near Real Time product

RT1 or RT2: Consolidated Real Time product (in convergence period), where the

number equals the number of times the RT0 product was successively updated

RT6: Final consolidated Real Time product

For OFFLINE reprocessing, this parameter is omitted.

<YYYYMMDDhhmm> gives the temporal location of the file. YYYY, MM, DD, hh, and mm

denote the year, the month, the day, the hour, and the minutes, respectively.

<AREA> gives the spatial coverage of the file. In this case, <AREA> is GLOBE, the name

used for the full globe.

<SENSOR> gives the name of the sensor family used to retrieve the product, so VGT

referencing SPOT-VEGETATION, and PROBAV for PROBA-V.

<Major.Minor.Run> gives the version number of the product. “Major” increases when the

algorithm is updated. “Minor” increases when bugs are fixed or when processing lines are

updated (metadata, colour quicklook, etc...). “Run” increases whenever a new processing

run (with the same major and minor version) is performed without a change in the algorithm

(e.g. due to a change in input data). This version refers to Major = 2.

3.1 FILE FORMAT

The DMP and GDMP Collection 1km Version 2 products are delivered in compressed Network

Common Data Form version 4 (netCDF4) files with global coverage and metadata attributes

compliant with version 1.6 of the Climate and Forecast (CF) conventions.



Issue: I3.22



The multi-layer GDMP netCDF files contain the following layers:

o GDMP: Gross Dry Matter Productivity value

o QFLAG: Quality Flag indicator

The multi-layer DMP netCDF files contain the following layers:

o DMP: Dry Matter Productivity value

o QFLAG: Quality Flag indicator

Separately from the data files, the following files are available:

An xml file containing the metadata conforms to INSPIRE2.1.

A quicklook in a coloured GeoTIFF format, sub-sampled to 25% in both horizontal and

vertical directions from the DMP or GDMP layer.

For a more user-friendly viewing of the XML metadata, an XSL transformation file can be

downloaded on https://land.copernicus.eu/global/sites/cgls.vito.be/files/xml/c_gls_ProductSet.xsll.

This file should be placed in the same folder as the XML file.

The products are delivered as one single global image, covering 180°W to 180°E longitude and

80°N to 60°S latitude. The globe is divided in 36 tiles in longitude and 18 in latitude according to

Figure 6. Only the green tiles, covering land, are generated and hence integrated into the global

image. The meteorological input is bound from 75°N to 50°S, hence the green tiles outside this

window are filled with empty values (Table 3). In winter season, the northern hemisphere has less

observation, hence the number of available tiles could be less.

Note that continental tiles, as defined as in Table 2 for AFRI and SOAM are distributed over

EUMETCast. The distribution is limited to the real-time mode, hence products with file name

convention c_gls_DMP-RT0_* or c_gls_GDMP-RT0_*.

https://land.copernicus.eu/global/sites/cgls.vito.be/files/xml/c_gls_ProductSet.xsl

http://land.copernicus.eu/global/sites/default/files/xml/c_gls_ProductSet.xsl

http://land.copernicus.eu/global/sites/default/files/xml/c_gls_ProductSet.xsl



Issue: I3.22



Figure 6: Regions of DMP/GDMP product. Red lines indicate the boundaries on availability of

meteorological data.

Short name Continent

AFRI Africa: 30°W – 60°E, 40°N – 40°S

SOAM South America: 110°W – 30°W, 20°N – 60°S

Table 2: Definition of the continental tiles distributed via Eumetcast.

3.2 PRODUCT CONTENT

3.2.1 Data Files

Table 3 gives the digital numbers associated with the physical DMP and GDMP range and no data

values. Note that no uncertainties are associated to DMP and GDMP values. The physical values

are retrieved by:

PhyVal = DN * scale_factor + add_offset

where the scale_factor and add_offset are given in the table below. This rescaling formula does

not apply to the binary numbers representing special flag values (Table 3).



Issue: I3.22



Table 3: Range of values and scaling factors of GDMP and DMP.

GDMP

(kg Dry Matter / ha / day)

DMP

(kg Dry Matter / ha / day)

Minimum value 0 0

Maximum value 655.34 327.67

Maximum DN value 32767 32767

Missing value -1 -1

Specific value for sea -2 -2

Scale_factor 0.02 0.01

Add_offset 0 0

The physical DMP (GDMP) values range between 0 and 327.67 (655.34) kg of dry matter per

hectare per day. The lowest values are found over non vegetated areas like deserts, urban areas

or outside the growing season in cold climate zones. The highest values indicate high productive

pixels, often found over forested areas in the peak of the growing season or in tropical zones.

3.2.2 QFLAG image

In addition to the quality flags in the negative range of the data image (Table 3), the quality of the

GDMP/DMP product is also provided in the bitwise Quality Flag (QFLAG) layer. This layer is

derived from the FAPAR Version 2 [CGLOPS1_PUM_FAPAR1km-V2] quality indicator as

referenced in the last column of the table.

Table 4: Quality Flag indicator for GDMP/DMP Version 2.

Bit Qualitative indicator Meaning Mapping to FAPAR

1 FAPAR missing or invalid 0 = Available FAPAR value

1 = Missing FAPAR value

QFLAG Bit 8

2 Meteorological input missing or

invalid

0 = Available Meteo value

1 = Missing Meteo value

3 FAPAR correction for high

latitude (lat > 55°N, SZA>70°)

0 = No high latitude

1 = High Latitude

QFLAG Bit 10

4 Evergreen Broadleaf Forest

(EBF)

0 = no EBF

1 = Evergreen Broadleaf Forest

QFLAG Bit 11



Issue: I3.22



(EBF)

5 Bare Soil

0 = no Bare Soil

1 = Bare Soil

QFLAG Bit 12

6 Not used -

7 FAPAR climatology 0 = FAPAR not filled

1 = FAPAR filled with climatology

QFLAG Bit 3 and

QFLAG Bit 13

8 FAPAR gap filling by

interpolation

0 = not interpolated

1 = interpolated

QFLAG Bit 3 and

QFLAG Bit 14

The user is referred to the fAPAR product [CGLOPS1_ATBD_FAPAR1km-V2] to investigate

details of the fAPAR characteristics [CGLOPS1_PUM_FAPAR1km-V2] and quality

[CGLOPS1_QAR_FAPAR1km-PROBAV-V2] if required.

3.2.3 File and Layer attributes

The netCDF file contains a number of attributes

at the file-level, see Table 5

at the level of the variables (layers), see Table 6

at the level of the standard coordinate (dimension) variables for latitude (‘lat’) and longitude

(‘lon’), holding one value per row or column respectively: see Table 7

at the level of the grid mapping (spatial reference system) variable (named ‘crs’): see Table

8.

At the level of time coordinate, see Table 9



Issue: I3.22



Table 5: Description of netCDF file attributes

Attribute name Description Data

Type Example

Conventions Version of the CF-Conventions used String CF-1.6

title A description of the contents of the file String 10-daily Dry Matter Productivity 1KM:

GLOBE 2014-12-10T00:00:00Z

institution The name of the institution that

produced the product String VITO NV

source The method of production of the

original data String Derived from EO satellite imagery

history

A global attribute for an audit trail. One

line, including date in ISO-8601

format, for each invocation of a

program that has modified the dataset.

String 2018-02-07 Processing line DMP2

1KM

references

Published or web based references

that describe the data or methods

used to produce it.

String https://land.copernicus.eu/global/prod

ucts/dmp

archive_facility Specifies the name of the institution

that archives the product String VITO

product_version Version of the product (VM.m.r) String V2.0.1

time_coverage_start Start date and time of the total

coverage of the data for the product. String 2014-02-01T00:00:00Z

time_coverage_end End date and time of the total

coverage of the data for the product. String 2014-02-10T23:59:59Z

platform Name(s) of the orbiting platform(s) String Proba-V

sensor Name(s) of the sensor(s) used String VEGETATION

identifier Unique identifier for the product String

urn:cgls:global:dmp_v2_1km:DMP_20

1412100000_GLOBE_PROBAV_V2.0

.1

parent_identifier

Identifier of the product collection (time

series) for the product in Copernicus

Global Land Service metadata

catalogue.

String urn:cgls:global:dmp_v2_1km

long_name Extended product name String Dry matter productivity

orbit_type Orbit type of the orbiting platform(s) String LEO

processing_level Product processing level String L3

processing_mode

Processing mode used when

generating the product (Near-Real

Time, Consolidated or Reprocessing)

String Offline

copyright

Text to be used by users when

referring to the data source of this

product in publications (copyright

notice)

String Copernicus Service information 2018

https://land.copernicus.eu/global/products/dmp




Issue: I3.22



Table 6: Description of netCDF layer attributes

Attribute Description Data Type

Examples for DMP layer

CLASS Dataset type String DATA

long_name

A descriptive name that indicates a variable’s

content. This name is not standardized.

Required when a standard name is not

available.

String Dry matter productivity

1km

scale_factor

Multiplication factor for the variable’s contents that

must be applied in order to obtain the real values.

Omit in case the scale is 1.

Float 0.01

add_offset

Number to be added to the variable’s contents (after

applying scale_factor) that must be applied in order to

obtain the real values. Omit for offset 0.

Float 0.0

units

Units of a the variable’s content, taken from

Unidata’s udunits library.

Empty or omitted for dimensionless variables.

Fractions should be indicated by scale_factor.

String kg / hectare / day

valid_range

Smallest and largest values for the variable.

Missing data is to be represented by one or

several values outside of this range.

Same as

data variable 0, 32767

_FillValue

Single value used to represent missing or

undefined data and to pre-fill memory space in

case a non-written part of data is read back.

Value must be outside of valid_range.

Same as

data variable -1

missing_value

Single value used to represent missing or

undefined data, for applications following older

versions of the standards.

Value must be outside of valid_range.

Same as

data variable -1

grid_mapping Reference to the grid mapping variable String Crs

flag_values Provides a list of the flag values. Used in

conjunction with flag_meanings.

Same as

data variable -2

flag_meanings Descriptive phrases for each flag value,

separated by blanks

String (blank

separated

list)

Sea



Issue: I3.22



Table 7: Description of netCDF coordinate variable attributes


Example

CLASS Dataset type String DIMENSION_SCALE

DIMENSION_

LABELS Label used in netCDF4 library String Lon

NAME Short name String Lon

standard_name

A standardized name that references a

description of a variable’s content in CF-

Convention’s standard names table. Note

that each standard_name has corresponding

unit (from Unidata’s udunits).

String longitude

long_name


content. This name is not standardized.

Required when a standard name is not

available.

String longitude

units Units of a the variable’s content, taken from

Unidata’s udunits library. String degrees_east

axis Identifies latitude, longitude, vertical, or

time axes. String X

_CoordinateAxisType Label used in GDAL library String Lon

Table 8: Description of netCDF grid mapping variable attributes


Example

GeoTransform

Six coefficients for the affine transformation

from pixel/line space to coordinate space,

as defined in GDAL’s GeoTransform

String

-180.0000000000

0.0089285714 0.0

80.0000000000 0.0

-0.0089285714

_CoordinateAxisTypes Label used in GDAL library

String,

blank

separated

GeoX GeoY

_CoordinateTransform

Type Type of transformation String Projection

grid_mapping_name Name used to identify the grid mapping String latitude_longitude

inverse_flattening

Used to specify the inverse flattening (1/f)

of the ellipsoidal figure

associated with the geodetic datum and

used to approximate the

shape of the Earth

Float 298.257223563

long_name A descriptive name that indicates a

variable’s content. String

coordinate

reference system

longitude_of_prime_m

eridian

Specifies the longitude, with respect to

Greenwich, of the prime meridian

associated with the geodetic datum

Float 0.0

http://www.gdal.org/gdal_datamodel.html



Issue: I3.22



semi_major_axis

Specifies the length, in metres, of the semi-

major axis of the

ellipsoidal figure associated with the

geodetic datum and used to

approximate the shape of the Earth

Float 6378137.0

spatial_ref Spatial reference system in OGC’s Well-

Known Text (WKT) format String

GEOGCS["WGS

84",DATUM["WGS

_1984",…

AUTHORITY["EPS

G","4326"]]

Table 9: Description of netCDF attributes for time coordinate variable


Example

CLASS Dataset type String DIMENSION_SCALE

NAME Short name String time

long_name


content. This name is not standardized. Required

when a standard name is not available.

String Time

units Units of a the variable’s content, taken from

Unidata’sudunits library. String

Days since

1970-01-01 00:00:00

axis Identifies latitude, longitude, vertical, or

time axes. String T

calendar Specific calendar assigned to the time variable,

taken from the Unidata’sudunits library. String standard

3.2.4 ‘quicklook’ image

The quicklook is a GEOTIFF file and represents either DMP or GDMP layer. The spatial resolution

is sub-sampled, using nearest neighbour re-sampling, to 25% in both directions.

The quicklook is provided in an unsigned character (8 bits) with a scale factor of 1.28 and an offset

of 0 (see Table 10).

Table 10 : QuickLook - value signification

Digital Value Description Decode

0 No DMP n.a.

1 to 255 Unit: kg Dry Matter / ha / day Scaling factor: 1.28

Offset: 0

The physical values can be retrieved from the digital numbers using the following formula:

physical value = (digital number * Sc) + Off, where Sc and Off are the scaling

factor and offset parameters from the “Decode” column in the table above.



Issue: I3.22



The quicklook is provided with an embedded color table (Figure 7).

Figure 7 : Color coding for quicklook images

3.3 PRODUCT CHARACTERISTICS

3.3.1 Projection and grid information

The product is displayed in a regular latitude/longitude grid (plate carrée) with the ellipsoïd WGS

1984 (Terrestrial radius=6378km). The resolution of the grid is 1/112°.

The reference is the centre of the pixel. It means that the longitude of the upper left corner of the

pixel is (pixel_longitude – angular_resolution/2.). The product is provided at global scale. However,

through a custom order, the user is able to select either region according to his/her needs.

3.3.2 Spatial information

As shown in Figure 6, the DMP/GDMP product is provided from longitude 180°W to 180°E and

latitude 80°N to 60°S.

3.3.3 Temporal information

The temporal information “YYYYMMDDhhmm” in the filename corresponds to the end date of the

dekad, just like the input FAPAR product.

For example, the file c_gls_DMP _201308310000_GLOBE_VGT_V2.0.1.nc corresponds to the

dekad from 2013-08-21 to 2013-08-31.

The temporal information is detailed in the netCDF attributes time_coverage_start and

time_coverage_end as well as the XMLmetadata fields "EX_TemporalExtent : beginPosition" and

"EX_TemporalExtent : endPosition", denoted as “YYYY-MM-DDTHH:MM::SS”, giving the

beginning and the end of the 10-days time period, respectively. The startPosition of the 10-days

time period is always set to the 01st, 11th or 21th day of the month.

3.4 DATA POLICIES

All users of the Global Land service products benefit from the free and open access policy as

defined in the European Union’s Copernicus regulation (N° 377/2014 of 3 April 2014) and

Commission Delegated Regulation (N° 1159/2013), available on the Copernicus programme’s web



Issue: I3.22



site, http://www.copernicus.eu/library/detail/248). Products from legacy R&D projects are also

provided with free and open action.

This includes the following use, in so far that is lawful:

a) reproduction;

b) distribution;

c) communication to the public;

d) adaptation, modification and combination with other data and information;

e) any combination of points (a) to (d).

EU law allows for specific limitations of access and use in the rare cases of security concerns,

protection of third party rights or risk of service disruption.

By using Sentinel Data or Service Information the user acknowledges that these conditions

are applicable to him/her and that the user renounces to any claims for damages against

the European Union and the providers of the said Data and Information. The scope of this

waiver encompasses any dispute, including contracts and torts claims that might be filed in

court, in arbitration or in any other form of dispute settlement.

Where the user communicates to the public on or distributes the original DMP/GDMP products,

he/she is obliged to refer to the data source with (at least) the following statement (included as the

copyright attribute of the netCDF file):

Copernicus Service information [Year]

with [Year]: year of publication

Where the user has adapted or modified the products, the statement should be:

Contains modified Copernicus Service information [Year]

For complete acknowledgement and credits, the following statement can be used:

"The products were generated by the Global Land Service of Copernicus, the Earth Observation

programme of the European Commission. The research leading to the current version of the

product has received funding from various European Commission Research and Technical

Development programs. The product is based on SPOT-VEGETATION 1km data (copyright CNES

and distribution by VITO).”

Or using the following citation for the PROBA-V products

"The products were generated by the Global Land Service of Copernicus, the Earth Observation

program of the European Commission. The research leading to the current version of the product

has received funding from various European Commission Research and Technical Development

programs. The product is based on PROBA-V data (© ESA).”

http://www.copernicus.eu/library/detail/248



Issue: I3.22



The user accepts to inform the Copernicus Global Land Service about the outcome of the use of

the above-mentioned products and to send a copy of any publications that use these products to

the support contact below.

3.5 CONTACTS

The DMP/GDMP products are available through the Copernicus Global Land Service website at

the address:


Scientific, Production and Distribution support contact:

https://land.copernicus.eu/global/contactpage

Accountable contact: European Commission Directorate-General Joint Research Center

Email address: [email protected]


mailto:[email protected]



Issue: I3.22



4 QUALITY ASSESSMENT

An extensive evaluation of the CGLS DMP Collection 1km Version 2 is done and described in the

validation report [CGLOPS1_QAR_DMP1km-V2]. The CGLS DMP product is validated against

various sources of reference data. The idea was that using multiple sources of reference

information would stimulate the “convergence of evidence”. The validation is following the

guidelines, protocols and metrics defined by the Land Product Validation (LPV) group of the

Committee on Earth Observation Satellite (CEOS). Adjusted to the specificities of the CGLS DMP,

the validation is done on three levels: spatial consistency analysis, global statistical analysis and

temporal consistency analysis. This analysis compares the CGLS product with DMP/GDMP

estimates from models or satellite remote sensing products. Separately, an accuracy assessment

is done by comparing the CGLS product with in situ measurements (FLUXNET).

The spatial continuity analyses showed that the DMP Version2 has much less missing values than

its previous version and reaches a global product completeness of >90% throughout the year. The

spatial consistency analysis showed that the CGLS DMP deviates from global reference data

(remote sensing: MODIS and modelled: ISLSCP-II) mainly in the boreal regions where the CGLS is

lower and the tropics, where the CGLS is higher. The global statistical analysis showed that the

GDMP matches quite well the one from MODIS, expect for broadleaved evergreen forests and

cropland, where the CGLS product is higher and in boreal regions where the CGLS is lower. When

comparing temporal profiles with Bigfoot and FLUXNET, it was shown that the CGLS GDMP

captures quite well the temporal behaviour of the reference datasets. Expect for cropland where

the growing season of the in-situ product is steeper and more productive. The temporal

smoothness is comparable to that of MODIS. The accuracy assessment showed that there is quite

some scattering when compared to FLUXNET, but the values are centred mostly around the 1:1

line.

From the analyses above, these are the most important findings of the report:

The DMP Version2 has much less missing values than its previous version and reaches a

global product completeness of >90% throughout the year.

Compared to the DMP Version1, this DMP Version2 is more in line with reference products,

especially when looking at the order of magnitude of the estimates.

A global comparison with MODIS and modelled ISLSCP-II data showed that the CGLS

DMP Version2 fairly well agrees with the reference products, expect for boreal regions

where the CGLS has lower and the tropics where the CGLS has higher values.

The order of magnitude of the DMP and GDMP of broadleaved evergreen (tropical) forests

is significantly higher than that of MODIS. But compared to in-situ measurements of

FLUXNET and Bigfoot, the CGLS product matches quite well, at least for GDMP. There is

an indication that for DMP, the CGLS values are overestimating the actual situation.



Issue: I3.22



For cropland, the CGLS product is also higher than its MODIS equivalent. But compared to

FLUXNET and Bigfoot, it still underestimates the measurements. The seasonality of these

reference products are not very well captured by the CGLS product. In many cases, the

FLUXNET season is much steeper and reaches higher productivities than that of the CGLS

product.

The temporal behaviour of PROBA-V products compared to FLUXNET data showed similar

agreements as for SPOT-VGT.

The comparison of PROBA-V and SPOT-VGT on the limited overlapping period (January-

March 2014) showed RMSE values up till 12 kg DM/ha/day for the vegetation active

regions in the southern hemisphere. For the northern latitudes, this period is off season

hence no clear conclusions can be made.

The difference between the NRT product RT0 and the consolidated RT6 (after two months)

is acceptable (< 10 kg DM/ha/day). For operational monitoring in the context of early

warning, the RT0 product is stable enough. But for quantitative multiannual temporal

analysis such as trend analysis, the RT6 product should be preferred.



Issue: I3.22



5 PRODUCT USAGE

The examples shown below are from DMP Collection 1km Version 1 product. Similar usage can be

done with Version 2 product.

5.1 TYPICAL USE

The DMP products are provided in a single image and are commonly visualized through

specialized software or via quicklooks. Note that, in order to retrieve actual DMP from the images,

the digital numbers have to be divided by 100 (Table 3). Missing values due to clouds or snow/ice

and sea are indicated with special flags (see also Table 3).

DMP values are classified into a number of classes and colour-coded for easy interpretation.

Country, province or district boundaries and a lat/lon grid are superposed on the map. An example

of such a quicklook is provided below in Figure 8.

These visual representations of the Dry Matter Productivity product can then be inspected to

identify zones of high or low productivity and as such they are useful for general evaluations of the

ecosystem productivity of an area.

Figure 8 : Quicklook of DMP for Sudan, 2nd

dekad of August 2006.

Red = low DMP, green = high DMP



Issue: I3.22



5.2 ADVANCED USAGE: HOW TO ANALYZE THE DATA

For dry matter productivity assessment over the growing season, an analysis needs to be made of

a multi-temporal set of those images so that the progress of the growing season and changing

growth rates can be assessed (e.g. by looking at the differences of consecutive images in the

same season, or by comparing to previous years or seasons or long term averages). Alternatively,

the Dry Matter Productivity images (kg/ha/day) can be cumulated over time from the start of the

season in order to estimate the final dry matter (kg/ha) which had been produced by the vegetation

over time. When properly calibrated against official yield statistics (post-processed) DMP products

can even be used for crop yield estimation and forecasting.

Possible post-processing operations:

1. Calculation of differences (in absolute values or percentages):

Differencing consecutive images in one season makes it possible to assess the progress

(evolution) of dry matter productivity and growth rates during the season.

Calculation of the differences between current DMP images and images of previous years

or seasons or long-term averages, allows the detection of anomalies in vegetation growth,

e.g. Figure 9.

Figure 9 : Western Africa, last dekad of July 2006. Absolute difference in DMP with respect to long

term average. Green = better situation, Red = worse

2. Cumulating DMP over time, from the start of the season onwards, provides estimates of final dry matter production over time, an example in Figure 10.



Issue: I3.22



Figure 10 : Sahel, seasonal dry matter productivity as calculated by the Biogenerator software of

Action Contre la Faim (http://www.actioncontrelafaim.org/en). Yellow = low cumulative DMP, Dark

green = high cumulative DMP. Map obtained from Ham et al. (2012).

3. Calculation of land cover or crop specific regional means statistics::

Computation of DMP values per administrative (agro-statistical) unit or other regions

(preserved areas, forests) and per land-cover class / crop type further enhances the product for

the purpose of dry matter productivity monitoring in specific areas and/or for specific vegetation

classes or crop types.

Several techniques can be used for un-mixing, ranging from the calculation of simple regional

means (if no land use information is available), regional means per class or crop type (if per

pixel the class/crop type is known), weighted regional means (if per pixel the area fractions of

the classes/crops are known) or using more complex techniques as matrix inversion (linear un-

mixing) or nonlinear estimation with neural networks. The method to be used mainly depends

on the quality of the available information on crop type and location.

After un-mixing the results are stored in a database. In this way, the image data become

compatible with the other agro-statistical information (official areas/yields, model-outputs,…) in

a spatial and thematic sense. The agro-statistical region (or district) replaces the pixel as

spatial unit, and the class/crop becomes the thematic unit. The procedure also involves an

important data reduction.

By generating graphs from the database the seasonal evolution of the dry matter productivity

for a certain region and class/crop can easily be monitored and, in the same way, comparisons

can be made with previous years or with the long term average.

In the graph below (Figure 11), DMP profiles are shown for winter cereals in Belgium. Un-

mixing was done according to 6 different methods (Verbeiren, 2006).



Issue: I3.22



Figure 11 : profiles for DMP for winter cereals in Belgium derived with 6 different un-mixing

methods: Regional mean, Hard classification, CNDVI threshold = 20%, CNDVI threshold = 50%,

Linear un-mixing and Neural network

In the next Figure 12 is explained how regional land cover specific statistics can be derived for the

Horn of Africa. Weighted DMP means per region and class are computed based on the FAO-GAUL

(Global Administrative Unit Layer) regions map and on Area Fraction Images (AFI, one per class)

derived from Africover and GLC2000 classifications. AFI’s contain per pixel the proportion (0-

100%) of the area (1km²=100ha) covered by the concerned class (example: class 10, pastoral

dry). In the example a threshold of 50% is specified to withdraw pixels from the computation if their

area fraction for a given class is below this threshold. The resulting graph shows the DMP

evolution of pastoral dry areas in the Bakool region in Somalia in 2005, 2006 and 2007.



Issue: I3.22



Figure 12 : Regional unmixed means of DMP. Example for the Horn of Africa.

Copernicus Global Land Service provides an online dashboard (Figure 14) to view temporal

profiles of regional and land cover specific averages of the DMP.

10-daily DMP images

Image with administrative regions from FAO-GAUL and

area fraction per class (example: class 10, pastoral

dry) as derived from GLC2000/Africover

ID CLASS NAME

1 Cropland (from GLC2000)

2 Grassland (from

GLC2000)

3 Continuous large fields

4 Continuous small fields

5 Clustered large fields

6 Clustered small fields

7 Isolated small fields

8 Irrigated large fields

9 Irrigated small fields

10 Pastoral dry

11 Pastoral wet



Issue: I3.22



Figure 13: Online dashboard of the time series viewer as provided by Copernicus Global Land

Service (http://viewer.globalland.vgt.vito.be/tsviewer/).

5.3 EXPLOITATION IN EMPIRICAL REGRESSIVE MODELLING

The DMP product and in particular the regional and class/crop type specific regional averaged

statistics (see above) of the cumulative DMP or maximum DMP (maximum growth rate) are often

used for crop yield estimation and prediction, as they are considered to be indicative for agricultural

production. Multiple regression and neural networks are the most commonly used techniques to

relate DMP and official yields. Availability of a sufficiently large time series of satellite data and

reliable yield statistics is crucial for obtaining good results.

Figure 14 shows an example of the temporal dynamics of yield statistics versus the CGLS DMP at

1 km derived from SPOT-VEGETATION, an enhanced version of the DMP including a water stress

factor and the classic vegetation indices (fAPAR and NDVI) for the period 1999-2012 (Durgun et

al., 2016)

http://viewer.globalland.vgt.vito.be/tsviewer/



Issue: I3.22



Figure 14: Temporal trends of predicted yield calibrated by leave-one-out cross validation technique

compared with the actual yield for 1999–2012 period per region for Morocco (top) and Belgium (bottom). Examples obtained from Durgun et al. (2016).



Issue: I3.22



6 REFERENCES

Atjay, G.L., Ketner, P. & Duvigneaud, P. (1979). Terrestrial primary production and phytomass. In:

The Global Carbon Cycle [Bolin, B., E.T. Degens, S. Kempe, and P. Ketner (eds.)], John Wiley

& Sons, Chichester, pp. 129-181.

Durgun et al. (2016). Testing the contribution of stress factors to improve wheat and maize yield

estimations derived from remotely-sensed dry matter productivity. Remote Sensing, 8, 170.

doi:10.3390/rs8030170

Eerens H., Piccard I., Royer A. & Orlandi S. (2004). Methodology of the MARS Crop Yield

Forecasting System. Volume 3: Remote Sensing, data processing and analysis, Office for

Official Publications of the European Communities, Luxemburg, 2004 (EU-Monograph EU

21291/EN3).

Eerens, H. et al. (2014). Image time series processing for agriculture monitoring. Environmental

Modelling & Software, 53, pp. 154-162.

Ham, F., Fillol, E. “Chapter 10: Pastoral surveillance system and feed inventory in the Sahel” in FAO. 2012. Conducting national feed assessments, by Michael B. Coughenour & Harinder P.S. Makkar. FAO/ACF Animal Production and Health Manual No. 15. Rome, Italy.

Gower, S. et al. (2001). Net primary production and carbon allocation patterns of boreal forest

ecosystems. Ecological Applications, 11, pp.1395-1411.

Luyssaert, S. et al. (2010). The European carbon balance. Part 3: forests. Global Change Biology,

16, pp. 1429-1450.

Massad, R-S., Tuzet, A. & Bethenod, O. (2007). The effect of temperature on C4-type leaf

photosynthesis parameters. Plant, Cell and Environment, 30, pp. 1191-1204.

Mccree, K. J. (1972). Test of current definitions of photosynthetically active radiation against leaf

photosynthesis data. Agricultural Meteorology, 10, pp. 443-453.

Monteith, J.L. (1972). Solar radiation and productivity in tropical ecosystems. Journal of Applied

Ecology, 19, pp. 747-766.

Running, S. et al. (2009). Next-Generation Terrestrial Carbon Monitoring. Geophysical Monograph

Series, 183, pp.49-69.

Valentini, R. (2003). Fluxes of Carbon, Water and Energy of European Forests, Vol. 163. Berlin:

Springer.

Verbeiren S. (2006). Comparison of different unmixing methods for yield estimation in Belgium and

China, Oral presentation at the ISPRS conference in Stresa, December 2006



Issue: I3.22



Veroustraete F. (1994). On the use of a simple deciduous forest model for the interpretation of

climate change effects at the level of carbon dynamics. Ecological Modelling, 75/76, pp. 221-

237.

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, pp. 376-399.

Veroustraete, F., Sabbe, H., Rasse, D. P., & Bertels, L. (2004). Carbon mass fluxes of forests in

Belgium determined with low resolution optical sensors. International Journal of Remote

Sensing, 25, pp. 769-792.

Verstraeten, W. W., Veroustraete, F., and Feyen, J. (2006). On temperature and water limitation of

net ecosystem productivity: Implementation in the C-Fix model. Ecological Modelling, 199, pp.

4-22.

Verstraeten, W. W., Veroustraete, F., Wagner, W., Van Roey, T., Heyns, W., Verbeiren, S., et al.

(2010). Remotely sensed soil moisture integration in an ecosystem carbon flux model. The

spatial implication. Climatic Change, 103, pp. 117-136.

Yamori, W., Hikosaka, K., Way, D.A. (2013). Temperature response of photosynthesis in C3, C4

and CAM plants: temperature acclimation and temperature adapation. Photosynth Res, 119 (1-

2), pp. 101-117.



Issue: I3.22



ANNEX: REVIEW OF USERS REQUIREMENTS

According to the applicable documents [AD2] and [AD3], the user’s requirements relevant for DMP

are:

Definition:

Dry Matter Productivity is an indicator designed to represent the growth rate of dry biomass.

It is equivalent to the Net Primary Productivity (NPP) but has been adapted to agro-

statistical use by transformation of the units into kilograms of dry matter produced per

hectare and per day (kg DM/ha/day)

Geometric properties:

o The baseline pixel size shall be 1km or 300m.

o The target baseline location accuracy shall be 1/3rd of the at-nadir instantaneous

field of view.

o Pixel co-ordinates shall be given for the centre of pixel.

Geographical coverage:

o Geographic projection: lat-long.

o Geodetical datum: WGS84.

o Pixel size: 1/112° - accuracy: min 10 digits.

o Coordinate position: pixel centre.

o Global window coordinates (40320 columns, 14673 lines):

Upper Left: 180W, 75N.

Bottom Right: 180E, 56S

Time definitions:

o As a baseline, the biophysical parameters are computed by and representative of

dekad, I. E. for ten-day periods (“dekad”) defined as follows: days 1 to 10, days 11

to 20 and days 21 to end of month for each month of the year.

o As a trade-off between timeliness and removal of atmosphere-induced noise in

data, the time integration period may be extended to up to two dekads for output

data that will be asked in addition to or in replacement of the baseline based output

data.

o The output data shall be delivered in a timely manner, i.e. within 3 days after the

end of each dekad.

Accuracy requirements:

o Baseline: wherever applicable the bio-geophysical parameters should meet the

internationally agreed accuracy standards laid down in document "Systematic

Observation Requirements for Satellite-Based Products for Climate". Supplemental



Issue: I3.22



details to the satellite based component of the "Implementation Plan for the Global

Observing System for Climate in Support of the UNFCCC (GCOS#200, 2016)"

o Target: considering data usage by that part of the user community focused on

operational monitoring at (sub-) national scale, accuracy standards may apply not

on averages at global scale, but at a finer geographic resolution and in any event at

least at biome level.

As the DMP is not directly an Essential Climate Variable (ECV), GCOS does not specify requirements. According to [AD3], the users requirements in terms of accuracy for 10-daily composites of

DMP, expressed in kg of DM / ha / day, are:

Threshold: 20

Target: 15 in general but inferior to 10 for crops and up to 20 for grassland

Optimal: 5.