1131 
These results allow us to think that the spherical distribution gives the best agreement. Moreover, this 
configuration is commonly used in many studies. Subsequently, we will employ this type of distribution. 
However, we can observe that although the agreement with spherical distribution is good in May when the 
cover is well-developed, uniform distribution is as correct as the spherical one in April. This can be explained 
by the evolution of the leaf inclination during the plant growth and also its dependence to the variety. But the 
main reason seems to be the fact that we assume the cover is an homogeneous medium which totally covers 
the ground whatever the obtained LAI. The hypothesis is valid when LAI equals 7 or 8 like in May. On the 
other hand, in April, leaf density is low (2 or 3) and the crop is in a clump form. In this case, as the cover 
fraction is low, the direct use of reflectance model is not satisfactory. 
3.12. Reflectance time profiles : test of several radiative transfer models 
The most represented configuration for winter cereals concerns the winter wheat variety 'Soissons', sowed on 
the Julian day 288 (15 October). It represents 14% of the surveyed fields surface. We can simulate the 
reflectances corresponding to this configuration using successively AFRCWHEAT2 to model the LAI and then 
a radiative transfer model to obtain reflectances in visible and n.Lr.. The cover fraction is here taken into 
account ; the values are coming from ground measurements (SCEES, 1985). Fust, the LAI given by the crop 
model has been corrected with this coefficient. Then, for each wavelength, we compute the daily reflectance as 
follow : 
P ' = T * p + (1 - X ) p soil (2) 
where p and p' are the cover reflectances respectively obtained without and with considering the cover 
fraction T , p so n is the soil reflectance. 
We display simulations obtained at Nadir in the principal plan with a daily time step (Fig. 2). The observed 
values are represented for each acquisition date. 
Figure 2 : Time seasonal profile of observed and modelled reflectances. Nadir simulation in visible (lowest 
curves) and n.i.r. (highest curves) with different models : Baret (-), SAIL (..), EXTRAD (—), Nilson-Kuusk (-.). 
Observations (*), with error bars which are 2 times standard deviation in height. 
Figure 2 displays the behaviour of simulated reflectances over the year, and observed values for the SPOT 
acquisition days. There is a great sensitivity of the reflectances with the radiative transfer model used, in 
particular in n.i.r.. However, observations are included in this variation range. 
Concerning directional effects, view zenith angles relative to SPOT data are small. Thus, in this case there is 
not a great difference between Nadir simulation and directional modelling (Moulin and Fischer, 1993). In 
consequence, we will test the 4 reflectance models to assimilate remotely sensed data even though some of 
them cannot take into account view geometry. 
We have shown that simulation can reproduce observation over a wheat cover by the means of a 
functional model linked to a radiative transfer model. In the second step, observations are used to adjust the 
sowing date and simulate carbon fluxes using the growth model.