1130 
hereafter to use satellite measurements in order to estimate this characteristic. The linkage of crop model with 
radiative model, using a 'first guess' sowing date, provides a temporal profile of reflectances. On the other 
hand, observations represent the reference set of reflectances. We assume that the first guess sowing date is 
uncertain. Then, a mathematical method is used to minimize the merit function by iterations, and thus the 
difference between modelling and observation. 
J(SD)= I(p* -p'"“') 2 0) 
i=l 
/ is the merit function for the sowing date SD, pfbs et pf no< ^ are the reflectances respectively observed and 
modelled for the acquisition day i. 
The final purpose is to process the simulation using the retrieved parameter in order to estimate 
carbon fluxes or production. 
3 - RESULTS AND DISCUSSION 
3.1. Simulation at the field scale 
We confront modelled reflectances with the values observed on SPOT scenes. The sowing date and the plant 
variety are the most important variables concerning the growth and the development of the wheat for 
environmental conditions given. Thus, when these two parameters are well-known, we can assume the 
modelling providing by AFRCWHEAT2 is reliable. 
3.1.1. Confrontation of simulations and observations 
As we have noticed an important sensitivity of predicted reflectances to the inclination of leaves, we perform 
simulations for several cases of angle distribution. For winter wheat, we attempt to obtain the most suitable 
distribution after the confrontation of these values to SPOT observations. 
Statistical results obtained for ground data over test-sites, allow us to know the distribution of sowing 
dates and varieties for winter cereals (wheat and barley). Consequently, we have deduced the different 
combinations of sowing date and variety, represented over the site. Simulations are performed in SPOT 
wavelengths for the main configurations using successively the cover growth model and SAIL radiative model. 
For winter cereals, the inclination of leaves is close to theoretical distributions. Some measurements (Baret, 
1986) have shown that it varies between uniform and erectophile distribution with the variety and the 
development stage. The comparison between modelling results and observation has been carried out for 3 
types of distribution for winter wheat crops. Since variations of reflectances with leaf angles are more 
important in n.i.r. than in visible, we display results relative to nix below. Figure 1 concerns 3 dates of the 
wheat activity period. 
Figure 1 : Modelled versus observed reflectances, in n.i.r., for 3 leaf angle distributions : uniform (mean angle 
is 45°) (+), spherical (mean angle is 573°) (*), erectophile (mean angle is 6324°) (o). The lowest group of 
points is relative to the date 09 April, the highest contains the values of two close dates : the 13 and 18 May. 
Each point represents the result concerning one combination (sowing date and variety).