mixture. That means clearly that grass became dominant at the end of the growing period, in agreement with ground 
observations. The mixture model is used for the bush/grassland. The shape of the curves indicate that the herb layer 
was rather made of forbs at the beginning of the season, then became an equitable mixture of grass and forb at the end 
(Fig. 6d). Since the tendency of the floristic composition from ground observations is realistic, the model is supposed 
to be validated. 
Site scale - Before to drive regional studies, it is relevant to hold estimates of the vegetation processes at the site scale. 
This is achieved through modeling. 
For the grassland (Fig. 5b), no significative difference of biomass exists between the plot and the site at the end of the 
growing season. Therefore, assuming that the floristic composition of the plot and the site were identical, the PAR 
interception for the site is the same than plot measurements. For the millet crop (Fig. 5c), the density of the clumps is 
higher on the plot. It is then needed to consider again the cylinders radiative transfer model to calculate PAR daily 
interception at the site scale. The millet crop started developing in August, reaching a maximum of green healthy 
leaves at mid-September, then decreasing within a month due to the senescence of the leaves. For the bushland (Fig. 
5d), the site differs largely from the plot by the density and size of the bushes, and by the presence of a patchy herb 
layer (Zomia glochidiata) covering 73% of the soil background. We calculate the PAR interception first for the bushes 
using the cylinders model, then for the (herb layer + woody layer) using the mixture model. The total PAR 
interception for site is then estimated by weighting each model-derived interception by their respective canopy 
fraction, assuming that the density and size of the bushes do not depend on the herb location. Modeling is the way to 
separate the PAR interception of the bushes and grass, plus the interception of the leaves and wood of the bushes. The 
herb layer began to intercept at the end of July, reaching a maximum 1st week of September. Then, the PAR 
interception decreases due to the fall of the leaves. Note that the leaves of the bushes begin to grow at the end of June, 
that is earlier than the herb. For the bush/grassland (Fig. 5a) where the mixture model is applied, are displayed the 
total canopy PAR interception, the interception by the bushes only, and the interception by the herb layer. As on the 
bushland site, the herb layer is the dominant factor of the bush/grassland interception at the end of the growing season. 
The herb layer began to grow at the end of July, almost one month later than the leaves of the bushes. The maxima 
appearing on the curves of Figures 5 correspond to cloudy days simulations which reflects the sensitivity of the model 
to the fraction of diffuse radiation. In addition, Licor LAI-2000 measurements were made along transects on the four 
sites. Calculated daily intercepted PAR from Licor are in good agreement with PAR interception estimated from 
modeling (Figures 5). The relation between NDVI and PAR interception is quite linear during the growing period 
(Figure 6). Airborne radiometric data can therefore be used to map PAR interception at regional scale. 
5. CONCLUSION 
A large and unique collection of radiative and biological data has been gathered over representative Sahelian 
vegetation canopies. This data set has been used to validate radiative transfer models for continuous and discontinuous 
canopies, and to derive biophysical parameters to be used in primary production models such as LAI (through the use 
of empirical models) and conversion efficiency from dry matter measurements. The relative constancy of climatic 
efficiency throughout the rainy season is a matter of interest for an accurate assessment of the incoming PAR radiation 
at the ground level. Since the Sahelian vegetation is poorly covering and the soil background is highly reflecting, 
interception and absorption are very close in the PAR range. The interception exhibits a large variability in time and 
space and appears as an interesting tool to study the dynamic of vegetation. Comparatively, the range of variation of 
the hemispherical reflectance, and also NDVI, is substancially reduced, due to the strong contamination by the soil 
background. Once validated, the radiative transfer models for continuous and discontinuous canopies permit to 
calculate interception of the various HAPEX-Sahel sites where biological data are available. The relation between 
NDVI and PAR interception is linear during the growth. Further, our interception measurements, in relation to ground 
and airborne remote sensing data, will be used to map primary production at a regional scale. 
6. REFERENCES 
Asrar G., M. Fuchs, E.T. Kanemasu, and J.L. Hatfield, 1984, 'Estimating absorbed photosynthetic radiation and leaf 
area index from spectral reflectance in wheat Agronomy Journal, 76, 300-308. 
Baret F. and G. Guyot, 1991, 'Potentials and limits of vegetation indices for LAI and APAR assessment', Remote 
Sensing of Environment , 46, 1-25. 
Begue A., 1992, 'Modeling hemispherical and directional radiative fluxes in regular-clumped canopies', Remote 
Sensing of Environment, 40, 219-230. 
Begue A N.P. Hanan, and S.D. Prince, 1994, 'Radiative transfer in shrub savanna sites in Niger - Preliminary results 
from HAPEX-Sahel: II. PAR interception of the woody layer 1 , Agricultural and Forest Meteorology, in press. 
Begue A., 1993, 'Leaf Area Index, daily intercepted PAR and spectral vegetation indices: A sensitivity analysis for 
regular-clumped canopies', Remote Sensing of Environment, 46, 1-25.