148 
Figure 4 shows this new index applied to the mineral (clay, sand) and peat soils. (Pozzolana and 
pebbles soils have been left out because they are quite unlikely to be found as a vegetation background). For 
the clayey and sandy soils group, the slope of the soil line is close to 1 (Fig. 2), therefore this new index is very 
similar to the "normal" TSAVI for the mineral soils. A differentiation is expected between the two groups of 
soils, indeed the Simplified TSAVI provides a separate curve for each soil group, with the two curves 
converging at high values of canopy cover (Fig. 4). At lower levels of vegetation cover, the index gives 
satisfactory results when the soil group is known. Therefore we do not need to know too much detail about 
the soil, only whether it is predominantly peat or mineral soil, which could be obtained easely from a soil map 
of the area. 
At the bottom of the figure 3, the standard deviation of this index has been estimated for each group 
of soils. In this context, this index performs better than the other indices previously tested. Assuming that the 
type of soil can be differentiated into either group, a sensitivity analysis has been done to provide quantitative 
information concerning the accuracy with which the cover parameter can be retrieved from the vegetation index 
values. Because of the complexity of inverting the SOILSPECT-SAIL model, the sensitivity analysis has been 
done by fitting the average values for each of the two groups of soils (Figure 4) and then inverting the two 
curves. The range in cover estimations has then been calculated by considering, for each LAI, the difference 
between the maximum and the minimum S-TSAVI values for all soils. For each soil group, the mean error 
is 0.05 of a unit of canopy cover [l-exp(-K*LAI)], and the maximum deviation from the fitted mean is about 
half this value, i.e. roughly 3% error in cover, which is a very good precision for vegetation monitoring, 
including yield prediction. 
4. The dual-angle approach 
It has been demonstrated that the Simplified TSAVI provides a satisfactory vegetation index, that is 
insensitive enough to the effects of the soil background. From this index it is now possible to retrieve, in a 
classical way, the biophysical parameters such as LAI, biomass, PAR and productivity. The next step is to 
investigate how the two viewing angles: 0 and 55' of the same scene may provide more information about the 
canopy. 
Data from at least two different directional reflectances have been proved useful for monitoring 
atmospheric effects, but have not yet been exploited to the same extent in vegetation studies. Directional effects 
of reflectances are difficult to use because they are due to a combination of many factors, such as soil, hot 
spot, specular reflection, sun position, etc. Vegetation indices reduce most of these effects. As for TSAVI, the 
Simplified TSAVI index is very stable for changes in relative azimuth view angle (when assuming an 
azimuthally uniform leaf angle distribution of the canopy in the SAIL model). Even the hot-spot dos not affect 
the TSAVI. Variations with the solar zenith angle are also very small. However, calculations show that we may 
expect a more systematic difference between the two viewing angles. 
Figure 5a presents the difference between the index values calculated for the forward view angle and 
for the nadir view, as a function of the canopy cover and for the two groups of soils. For each LAI, the five 
points correspond to different mean leaf inclination angles (from aplanophile canopy: 25’, to an erectophile 
canopy: 65"). For a given LAI the variations with vegetation cover are not greater than 0.1, and most often 
around 0.05, which is of the same order as the precision with which the plant cover can be retrieved from the 
index. In this point of view it would be difficult to extract any supplementary information. 
However, if we consider that for an intermediate amount of vegetation (LAI < 3) the relationship 
between plant cover and LAI is linear (Sellers et al., 1992), and therefore that a LAI in this range can be 
retrieved from the nadir view vegetation index with the same precision as above, then it may be possible to 
distinguish between an erectophile and a planophile canopy. Figure 5b shows the directional variations of the 
vegetation index as a function of the leaf inclination angle for representative values of LAI. For small LAI 
(LAI<0.1), or for high vegetation cover (LAI = 8 ), the differences are small, but for an intermediate cover, 
the difference is up to 0.1 of the index value between an erectophile and a planophile canopy, and therefore 
can be useful for determination of the canopy architecture. A possible application of this consideration might 
be the detection of stress or disease in vegetation. For example, it has been shown that a hydric stress on crop 
can be detected at an early stage by the leaf inclination effect (Rondeaux and Herman, 1991). This could be 
done by using the two view angles.