1091 
isy observations, for 50 
g C m-2 year -1 instead 
dard deviations of the 
improved (in amplitude 
the first guess. Theses 
m statistically finds the 
knowledge-based link 
i when high LAI values 
lei, can easily generate 
;rvations. We therefore 
lization. The first guess 
3/4166 °Cfor2^2 
ults of adjustment to a 
found, but the simplex 
;nt methods: systematic 
highlights the irregular 
o be possible, for this 
ssimiladon can improve 
ions were not addressed 
xuracy of reflectances 
measurements are not 
a prognostic vegetation 
approach in the field of 
questions. The method 
s direct modelling from 
te biophysical variable, 
nodel course and these 
method should lead to 
reflectance models has 
natural vegetation and 
strategies which can be 
ing in the observations, 
devoted to the analysis 
model and 'observation' 
e data set may contain 
at cost functions can be 
it priori weights to the 
rge error in modelling 
nal reflectance. These 
g information. A priori 
it function, the specific 
>n of noise observation 
led surface-atmosphere 
reflectance model and climatologies as inputs. We can also notice that vegetation indices, designed to reduce 
perturbations, can be assimilated as well. 
LAI inversion requires a strong filtering of the radiometric measurements. These filters are based on a priori 
knowledge (for example Maximum Value Composite of NDVI, Holben 1986). In particular, recent methods 
assume a minimum time for canopy decrease and regrowth. Clearly, this a priori knowledge is given by the 
vegetation model, in the 'model to satellite' approach. Besides, it allows to take into account measurements 
that should have a better statistical relevance than a composite method, which select one datum over a 
prescribed time period (sometimes more than one observation could be use, sometimes they all should be 
rejected). However, it requires accurate direct modelling of reflectances, and therefore LAI and optical 
properties temporal profiles. Development of LA/meter measurements and long term biomass sampling 
coupled with temporal series of ground radiometric measurements are expected to result in better 
understanding and modelling of canopy behavior for natural ecosystems. 
Modelling the structure of vegetation canopies and their seasonal variations is a critical issue a) for carbon 
cycle assessment, b) for coupling with satellite data but also c) for climate researches (Pielke et al. 1993). In 
this study, we focused on seasonal vegetation and satellite measurements in the solar spectrum. Therefore, for 
global scale applications, savannahs, steppes and deciduous forests ecosystems are natural candidates for 
vegetation model control by satellite observations. Moreover, the first information which is expected concerns 
phenology modelling. Besides, inclusion of measurements from other sensors or wavelengths is 
straightforward, for example by adding properly weighted terms in the cost function, as soon as accurate 
forward modelling from vegetation to satellite exists. Water in plant and soil, latent heat fluxes, canopy 
biomass, structure and temperature have been related to microwave and thermal infrared measurements (e.g. 
Kerr and Njoku 1993). 
Finally, we would like to outline that the development of vegetation models and assimilation of satellite data 
should proceed in close connection, because the control strategy requires accurate analysis of the model errors, 
to determine the control variables for instance. 
AKNOWLEDGEMENTS 
This work was carried out at the Laboratoire d'Etudes et de Recherches en Télédétection Spatiale (LERTS) as 
part of the European project "The global Carbon Cycle and its perturbation by man and climate. Part H: 
Biosphere", supported by the Environment program of the Commission of the European Communities. Laurent 
Kergoat is supported by CNRS-INSU 
BIBLIOGRAPHY 
Agren, G.I., McMurtrie, R.E., Parton, W.J., Pastor, J. and Shugart, H.H. 1991. State-of-the-art of models 
of production-decomposition linkages in conifer and grassland ecosystems. Ecological Applications 
1(2), 118-138 
Bonan, G.B. 1991. Atmosphere-biosphere exchange of carbon dioxide in boreal forests. J. Geophys. Res. 
96, 7301-7312. 
Bonan, G.B. 1993. Importance of leaf area index and forest type when estimating photosynthesis in boreal 
forests. Remote Sens. Environ. 43, 303-314. 
Fischer, A., Kergoat, L., and Mougin, E., 1994, Understanding the ecosystem dynamic behavior. What is 
the potential for satellite data assimilation into functional models? Proc. of 6th ISPRS Symposium. 
17-24 Jan 1994, Val d'Isere, France 
Fung, I.Y., Tucker, CJ. and Prentice, K.C, 1987. Application of advanced very high resolution radiometer 
vegetation index to study atmosphere-biosphere exchange of CC>2. J. Geophys. Res. 92, 2999-3015. 
Heimann, M. and Keeling, C.D. 1989. A three-dimensional model of atmospheric OO2 transport based on 
observed winds: 2. Model description and simulated tracer results. Geophysical monographs 55 
(American Geophysical Union). 237-275 
Holben, B.N. 1986 Characteristics of maximum-value composite images from temporal AVHRR data. Int. 
J. Remote Sensing, 7(11), 1417-1434 
Hollingsworth, A. 1990. Assimilation of remotely sensed atmospheric data from new satellite systems in 
the 1990's. Proc. of the international symposium assimilation of observations in meteorology and 
oceanography, held in Clermont-Ferrant (France) 9-13 July 1990,WMO, 119-26