1129 
environmental data (latitude, soil type), the fanning practices (sowing date and density) and some plant 
characteristics (in particular the photoperiod sensitivity and the vernalization sensitivity). This growth model 
has been developed for winter wheat crops. In addition, a water budget model is integrated and the needs in 
nitrogen are considered into the AFRCWHEAT2 version used here. The meteorological data are provided by 
the station of Boigneville, for the years 1991 and 1992. 
AFRCWHEAT has been validated over European countries (Weir et al., 1984), so we can assume 
that, on a region like the Beauce, using the meteorological data relative to the year, the photosynthesis is well 
simulated. However, the plant phenology, and thus the production, is still uncertain if we cannot evaluate 
model parameters relative to plant properties, and above all the sowing date which is an important initial 
condition. The leaf area index (LAI) of the vegetation cover given by the functional model allows the linking 
with a canopy radiative transfer model. 
23.2. The radiative transfer models. 
As simulated reflectances are sensitive to the radiative transfer model used, we test 4 different models. 
•The Baret model. This model computes vertical bidirectional reflectances (Baret, 1988) using a 
simple description of the radiative transfer. Its validity field is limited to the monitoring of cereal crops with 
satellite measurements. The radiometric response of a crop depends on various parameters which are taken 
into account in this model. The reflectance is first driven by canopy characteristics : LAI, leaves orientation 
(mean angle) and leaf optical properties (reflectance and transmittance). It also depends on external parameters 
like solar position (zenith angle) and soil optical properties (reflectance). 
•The SAIL canopy model. The SAIL (Scattering by Arbitrarily Inclined Leaves) model (Verhoef, 
1984) is a generalization of the Suits model (Suits, 1972) which computes the reflectance in the sensor 
direction as a function of canopy parameters and of acquisition and solar angles. Here, the leaves orientation is 
represented by a discrete function of the leaf inclination angle distribution (Leaf Inclination Density Function 
or LIDF). It takes into account LAI, optical leaf properties (reflectance and transmittance) and soil reflectance. 
•The EXTRAD model. A model has initially been computed by De Wit to simulate crop 
photosynthesis, then Idso (Idso and De Wit, 1970) used it to simulate diurnal variations of reflectances over a 
com crop. Goudriaan (1977) has extended and improved the Idso-De Wit model with a description of the 
radiative flux profile into a cover (Goel, 1988). Thus, the EXTRAD model (EXTinction of RADiation) has 
been developed to calculate the profile of solar radiation in the canopy. Vegetation cover characteristics are : 
the LAI, diffusion coefficient of leaves and the distribution of leaf inclination angles. The external parameters 
consist in the solar angle and the fraction diffuse sky irradiance. 
•The Nilson-Kuusk model. This analytical model (Nilson and Kuusk, 1989) computes the single and 
the multiple scattering of radiation in both canopy and soil. It takes into account specular reflection of 
radiation on leaves and canopy hot spot. The model assumes that the plant canopy is statistically 
homogeneous. The inputs concern first optical properties : reflectance and transmittance of leaves, refractive 
index of the leaf cuticular wax layer and soil reflectance. Second, it consists in structural parameters : LAI, 
leaf inclination distribution (eccentricity and inclination of elliptical distribution), leaf-hair index and a 
dimensionless leaf size coefficient Finally, model needs geometrical and illumination characteristics : solar 
and view angles and the fraction of direct solar radiation in total irradiance above canopy. 
The optical input parameters of these reflectance models summarized bellow are used for every test 
Leaf reflection coefficient : 0.12 in visible, 0.46 in n.Lr., 
Leaf transmission coefficient : 0.01 in visible, 0.50 in n.ij., 
Soil reflectance : 0.13 in visible, 0.19 in n.Lr.. 
Optical properties of the leaves correspond to values observed in literature (Bouman, 1992 ...) for winter 
cereals. The LAI is modelled daily by AFRCWHEAT2. Due to the great sensitivity of reflectances to the 
distribution of leaf inclination angles, some tests have been carried out to find the most suitable one. 
Concerning soil reflectances, we have taken the mean values obtained for bare soil pixels on SPOT images. 
The view and illumination geometry corresponds to the configuration of SPOT observations. 
2.4 Data assimilation technique 
Various parameters, which have a strong influence on estimated carbon fluxes or production, arc generally not 
well-known at a regional scale : It concerns above all the phenology of the plant, but also some characteristics 
relative to the crop variety. The growth model we used allows a good description of the plant phenology which 
depends on meteorological conditions. According to a sensitivity study, the sowing date is the most pertinent 
parameter : It influences both phenological development and net primary production. Thus, we attempt