parameters such as vegetation cover fraction f v , LAI and canopy height which are essential parameters for a
coupling with visible reflectance models or active and passive microwave models, are also simulated. The
model runs on a daily step basis. Being developped for regional applications, the model simulates processes and
structural parameters which, at that scale, should be taken as average, characterizing what can be considered as
“the most likely average" situation. A direct application of the present model, at a local scale, would therefore
be meaningless.
The model validation represents an essential step before performing the coupling with satellite data. This is
done by comparing model simulations to measurements of net primary production representative of the scale
considered i.e. at least, 1 x 1 km 2 . To this end, the validation is carried out independently on two typical
sahelian regions, namely the Ferlo region (Senegal) and the Gourma region (Mali) where standing biomass
measurements over large zones have been systematically collected for more than ten years (figure 1).
Figure 1: Locations of the test sites. Ferlo (Senegal) and Gourma (Mali)
In both regions, estimation of the standing biomass has been performed by harvest at the end of the rainy
period. A more complete description of measurement procedures can be found in Tucker et al (1983) for the
Ferlo and in Justice and Hiemaux (1986) for the Gourma. Only biomass measurements made on sandy soils are
retained for the model-measurement comparison, in order to consider homogeneous regions from this point of
view. For the Ferlo, biomass measurements taken into consideration are representative of a region of about
10x10 km 2 whereas for the Gourma, the available measurements correspond to homogeneous areas of only lxl
km 2 . On these two regions, average values of sand content of soils reach about 90%. Maximum tree cover is
about 20% for the Ferlo (Le Houdrou, 1989) and 10% in the Gourma (Hiemaux, 1986) for the different sites
under study, with an average value of about 5%. Trees are therefore assumed to have negligible effects on the
water budget. Meteorological data were obtained from 36 and 15 stations distributed in the Ferlo region and the
Gourma region respectively. For the Gourma, only three experimental sites are considered (sites 1, 18 and 31,
for location see Justice and Hiemaux, 1986).
The comparison of the model simulations with aboveground biomass measurements was carried out for the
period 1976-1987 for the Ferlo and 1985-1992 for the Gourma. Figure 2 shows the comparison between
simulated aboveground biomass and measurements for the Ferlo and the Gourma for the period 1976-1987 and
1985-1992 respectively. This data set, including 127 measurements representative at a regional scale, offers an
unique opportunity to validate the present model. Here, a very good agreement is found with r=0.95 indicating
that the STEP model is able to provide a reliable regional estimation of the aboveground biomass and related
processes under different climatic conditions.
1113
