974 
3 - CROP PRODUCTION MODELING - COUPLING WITH RADIOMETRIC DATA. 
3.1 Examples of 3 different strategies of coupling. 
Crop production models are functional models, built to simulate the processes related to the growth and the 
canopy development along the growing season, and which determine the final production. A very large range of 
crop functional models exists, the simple ones use less than 10 parameters, while some very detailed models, 
where water and nitrogen cycles in the soil are described, use mere than 50 parameters. They all have in common 
some general characteristics. They generally work at a daily time step, and LAI is a key variable involved in an 
iterative process: daily as simil ation is a function of the external conditions (solar radiation, temperature, 
humidity, water and nutrient supply), and of the state of tire cover (phenological stage. LAI). During the growing 
period, a part of the daily assimilated dry matter is allocated to the leaves, updating the LAI value. A 
phenological submodel, mainl y driven by temperatures, and also by daylength for winter cereals, gives the timing 
of emergence, of the various stages during growth, and of senescence. It controls the values of the allocation 
coefficient to the various plants components. As in case of natural vegetation, the temporal description of the 
photosynthetically active canopy predicted by the model is of particular importance for the final yield. Some 
authors have proposed to couple these models with remote sensing information for several years, and the 
different strategies which have been used are good examples of advantages and limits of this coupling. Del 6 colle 
et al. (1992), made a synthetic review of 3 major approaches: facing, readjustment and correcting term. The 
third ore (Faivre et al., 1991) does not interfere with the model, only an the major output, the yield in that case. 
Consequently, there is no improvement in the descriptiai of the processes in the model. Even if the spatial pattern 
of the crops behavior are retrieved, the result does not lead to a spatialization of the modeling. So, we will 
comment oily the farcing and the readjustment strategies, by giving four different examples of recent a current 
work, ranging iron the facing of remotely sensed LAI to the assimilation of satellite reflectances into a crop 
functional model. 
The models used fa these examples are GRAMI (Maas, 1992) fa sorghum, SUCROS fa sugarbeets (Spitters et 
al. 1989), and AFRCWHEAT fa winter wheat (Weir et al. 1984). 
GRAMI is a very simple model which simulates the daily changes in three state variables: green 
LAI, above ground dry matter, and the stage of development A classical Beer function gives the Absorbed 
Photosynthetically Active Radiation (APAR) from the LAI Similarly to die Monteith (1977) approach, the daily 
production of dry matter is then computed from APAR through a conversion efficiency, changing with the 
phenological stage. The increment in the LAI depending on this daily production is also a function of the 
phenological stages. The drawback of the simplicity of this model is the extrem sensitivity to the initialization of 
LAI at emergence. 
SUCROS describes the main processes of the assimilation of dry maitw (photosynthesis and 
respiration) in a more mechanistic way. Biophysidogical relations are used to compute the photosynthesis, three 
times a day, and at three levels within the cover. As in GRAMI, the canopy is still represented by a “Big-Leaf’, 
characterized by a LAI value and a radiation extinction coefficient 
AFRCWHEAT is a much more complex model which takes into account the sensitivity to the 
photoperiod and vernalization effects in the phenological submodel. The simulati on of the cover structure is very 
important shoots and leaves are discrete individuals, their pop ulations change, enrh leaf has an increasing green 
surface, turning yellow when secenescence occurs. Such a description has appeared to be useful fa a better 
simulation of the winter wheat production in European agricultural regions. In fact, the assumption that the 
canopy can be represented by a “Big Leaf’ with a global LAI seems accurate enough fa the simula tion of the 
exchanges of mass and energy, but is not sufficient to give accurate es tima tion of grain production, which also 
depends on the number of shoots at an the sis (when all shoots are assumed to be able to carry grains). The 
simulation of this number needs a precise modeling of file tillering mechanisms, depending on the farming 
practices (sowing date, sowing density, choice of the variety). 
Although these three examples have very different levels of complexity, they all show how difficult and 
important a good description of the canopy is. Classical relations between LAI and radiometric measurements in 
the short wavelengths were used to couple the three models described above with radiative Ham in three different 
ways, which are schematically represented on figure 1 : 
3.1.1 Forcing of LAI derived from satellite data in the production model Del 6 colle and Guirif (1988) reversed 
LAI values from SPOT/HRV data over wheat fields in Camaigue (South of France). Due to the low temporal 
frequency and the irregularity of file LAI retrieved along the wheat growing season, a statistic al model (Baret,