972 
with a more or less pronounced annual periodicity. Ecosystems dynamics refers to longer time scales, and also 
has a spatial dimension. Its understanding (and consequently its modeling) is difficult, in part because the 
responses of an ecosystem to some evironmental changes have time scales longer than the mechanisms which 
have induced them. However, the response of the ecosystems dynamics and the feedback to climate are directly 
linked to the processes which drive the exchanges at the canopy surface, and which deter mine the annual and 
interamial dynamics in the state of the canopy. 
In that paper, we define both the exchanges and the seasonal and interanmial dynamics by the 
term of “ecosystem functioning”. The improvement of its understanding is required to have a better idea of the 
ecosystem dynamics, especially in a context of Global Change. Probably, the long term series of satellite data 
(since the 1970’s for Landsat images) are not long enough to monitor ecosystem dynamics, apart from some 
specific sensitive locations where some important ecological changes appear on short time scales. But they can 
directly help in the understanding of ecosystem functioning, especially by the continous monitoring of seasonal 
and interannual dynamics observed from the regional to the global scale. 
The aim of this paper is to review some works dealing with satellite data, and particularly time 
series, and ecosystem models. In a first part, examples of studies where the coupling of satellite data with 
functional models have allowed a better understanding of either the state, or some mechanisms of natural 
vegetation are reviewed. In a second part, some precise strategies for coupling are discussed in case of crops, 
with crop production models. We are rather confident in the outputs of such models, and consequently they can 
be used as a good tool to test the various methodologies. The conclusions allow us to propose in tire third part an 
approach that we call: “assimilation of satellite data”. This approach seems to be very interesting when we have 
radiometric data about ecosystems canopies (in time and in space), but when some information or some 
knowledge about tire functioning are missing. 
2 - NATURAL VEGETATION MODELING - LINKING WITH SATELLITE DATA. 
We do not pretend here to be exhaustive. Representative examples describe various means of combining 
functional models and satellite data. 
2.1 Modeling the Montana coniferous forests functioning. 
The studies on the Montana coniferous forests have given interesting works: the FOREST-BGC model 
(BioGeoChemical Cycles. Running and Coughlan, 1988). simulates photosynthesis, growth and maintenance 
respirations, évapotranspiration, with mechanistic relations based on biological and physiological rules at a daily 
time step. Nitrogen cycle is described at a yearly time step. The ecosystem, a mature coniferous forest in that 
case, is represented by a “Big-Leaf’ of constant Leaf Area Index (LAI) all along the year. Because the LAI is an 
“expensive” input variable, a NOAA/AVHRR image has been used to derive it over a whole region, using one 
empirical relationship, in order to map the regional forest évapotranspiration and photosynthesis (Running et al., 
1989). The main interest of such a use of satellite data is the spatialization of important variables in order to run 
the model in a diagnostic mode over large areas, but it does not produce any explanation for the observed spatial 
variations of the LAI. To be able to understand it would be an interesting step in the study of the terrestrial 
ecosystems. 
2 2 Equilibrium ecological models. 
Grieg and Running (1977) proposed a model for the estimation of the LAI in mature coniferous forests where the 
soil-water-availability is an important facta 1 (northern Rocky Mountains). They assume that the LAI is a function 
of soil and climate. This ecological concept of climate-soil-LAI hydrologic equilibrium of forests was validated 
by Nemani and Running (1989) by comparison with LAI derived from satellite observations. At a local scale, 
LAI values estimated by the ecological model are well related to the vegetation indices derived from TM 
measurements. The correlation is also coherent at the regional scale, with the use of NOAA/AVHRR data. The 
quality of the results is due to the fact that: a) the stands were homogeneous and usually with complete canopy 
cover, thus minimizing the contribution of understory and soil on spectral reflectance, b) the measured LAI vary 
between 1.6 and 4. so almost always under the values where the visible and near infrared re flectance s begin to 
“saturate”, dearly, the use of remote sensing information can provide understanding of ecological processes, and 
die authors suggested that the knowledge of the spatial variations of climate (especially in that case the 
precipitations), associated with the satellite mapped LAL could help in the estimation of soil hydrologic 
properties at various scales. Another similar example is given by Woodward and Smith (1993), who have used