1085 
g is arbitrary. Taking 
r well within a strip of 
south-north dust load 
vely. The orientation 
le agreement is better 
it load u is by a factor 
lodel. This difference 
le is assumed to be a 
of the satellite image 
Assuming a constant 
I 365.25 /65) = 0.13 
below the vegetation 
1 on larger scales, e.g. 
ieve that the satellite 
)llutant8 of this kind, 
phenomena. 
TOWARDS ASSIMILATION OF SATELLITE MEASUREMENTS 
IN A GENERIC VEGETATION PROCESS MODEL 
L. KERGOAT and G. DEDIEU 
LERTS, CNES-CNRS, 18 Av E.Belin, bpi 2801, F-31055 Toulouse cedex, France 
ABSTRACT 
The current concern about the global carbon cycle and, more generally, interest in ecosystem functioning 
encourage the development of prognostic process-based vegetation models. Satellite measurements provide a 
unique source of observations when regional to global spatial scales are considered. These data can be used in 
different ways. We focus, in this paper, on a control strategy: The model trajectory is constrained by temporal 
series of remote sensing data. We present a generic vegetation model designed for this assimilation scheme. A 
synthetic experiment for a deciduous-type forest simulation shows that phenological behavior and allocation of 
carbon to the leaves can be controlled by visible and near infrared reflectances, and that carbon fluxes 
estimation can thus be improved. Finally, we point out the interests and requirements of this combined use of 
prognostic model and remote sensing measurements, and discuss the application to natural vegetation at the 
global scale. 
KEYWORDS : vegetation model - global - assimilation - phenology - carbon cycle 
n model. PSI General 
PSI General Energy 
alton Highway. Sub- 
Case Study in Arctic 
T Handbook, Volume 
Des- 
m (5S). Users Guide 
Des- 
te signal in the solar 
heric code. Proc. 5th 
5 available the SPOT 
»ted to P. Courrouyan 
e express our appreci- 
itory measurement of 
ing of the manuscript. 
1 - INTRODUCTION 
Assessing the contribution of the terrestrial biosphere to the global carbon cycle has become a major concern, 
in a global change context (direct effect of [C02] increase and possible related climatic changes). At the 
global scale, validation of the carbon fluxes estimations is non trivial: One possibility is to use atmospheric 
CO2 measurements, available from current stations network, to compare modelled net surface fluxes and these 
measurements, through an atmospheric transport model. This requires an accurate spatial and temporal 
description of the surface CO2 sources and sinks, with emphasis on seasonal and interannual variations. 
Several studies have demonstrated that AVHRR times series allow the monitoring of the vegetation, at regional 
to global scales. Fung et al. 1987 used empirical relationships between NDVI time profiles and photosynthesis, 
to distribute, in time, annual NPP estimates. This results in satisfying simulations of atmospheric [CO2] 
seasonal oscillations. Various authors also applied Monteith's model (Monteith 1972) and derived Absorbed 
Photosynthetically Active Radiation ( APAR ), from the global weekly AVHRR1NDVI archive and incident solar 
radiation. APAR is transformed into dry matter through biological efficiency coefficients and results in global 
NPP estimations (Heiman and Keeling 1988, Ruimy et al. 1994). A major advantage of these diagnostic 
studies is that the response of the vegetation to environmental facing is included, to some extent, in the 
satellite signal. However, we also need to develop predictive understanding of the vegetation functioning. This 
means that prognostic process vegetation models are promising tools to address the response of the global 
system to anthropogenic perturbations (Agren et al. 1991, Mellilo et al. 1993). However, as these models are 
still in their infancy and suffer from the scarcity of relevant large scale data sets, it seems interesting to use 
satellite measurements which are a unique source of observations. 
In this paper, we propose a methodology to control a vegetation model with remote sensing data. A 
medium-term objective is to evaluate the impact of satellite measurements assimilation on the estimations of 
vegetation CO2 uptake and release during the 80's and to investigate whether biophysical processes modelling 
can benefit from remotely sensed information (see Fischer et al. this issue, fa a more general discussion). In 
section 2 we briefly describe the model and mention the specific features required for the control by AVHRR 
measurements. The temporal development of the canopy is of particular importance, fa the carbon budget as 
well as for reflectance modelling. In section 3, we present a synthetic experiment which establishes the 
feasibility of this approach. Finally, we discuss (section 4) the possible contributions of satellite dam 
assimilation to global carbon cycle studies and spatialization of biosphere models.