Figure 4 : Time series of observed surface 
reflectances (---). modelled reflectances (...) and 
corrected reflectances (o) in the visible and near 
infrared, for 2 different test sites, (from Leroy and 
Roujean, 1994). 
Figure 5 : Time series of observed NDV1 (—) and 
corrected NDVI (o) for 2 test sites. Inserted in the 
Valensole plateau figure is a plot of the local rainfall 
(mm) as a function of time, (from Leroy and 
Roujean, 1994). 
of the same surface in the presence of shadows. When 0 S increases, the shadows extension increases and the 
discrepancy between kg and the monthly average of observed reflectances becomes larger. This effect may 
be in some cases larger in the visible than in the near infrared (as for example for the Beauce site in Figure 
4), which again may be expected since the important absorption by plants in the visible enhances shadowing 
effects. 
Table 1 shows that the level of fluctuations 5 remaining after modelling ranges between 1 and 3 
percents of reflectance and is on average of the order of 0.015 in the visible, and 0.02 in the near infrared. 
Thus modelling of bidirectional effects reduces substancially the level of noise-like fluctuations, as can be 
seen by comparing these figures to Figure 4. Analysis shows that the noise reduction factor (standard deviation 
of noise-like fluctuations divided by 8) ranges from 1.3 to 3 depending on the sites (Leroy and Roujean, 1994). 
The time series of observed Normalized Difference Vegetation Indices (NDVI), defined with the 
atmospherically corrected reflectances observed in the visible and near infrared, have been plotted in Figure 5 
for the Beauce plain and Valensole plateau test sites. The corrected vegetation indices, dfined with the kO in 
the visible and near infrared have also been plotted in superimposition on this Figure. As for the time series of 
corrected reflectances, the time series of corrected vegetation indices have a rather smooth aspect and often 
differ from the observed monthly averaged vegetation indices. This is particularly clear for the Beauce site, 
where the corrected NDVI has a steeper slope around Julian day 180, and a lower value in early spring and late 
fall, than the corresponding observed NDVIs. Thus, in this case, the dynamic range of NDVI through the 
annual cycle has been increased by the correction of directional effects. It is, moreover, interesting to note that 
a second maximum of NDVI appears around Julian day 260 on the time series of corrected NDVI of the 
Valensole plateau test site, which could not be perceived by inspection of the original curve of observed 
NDVI. The conjecture that this second maximum corresponds to a real increase of vegetation activity is 
26 
supported by the r 
evidence of strong 
Site 
arid Crau 
humid Crau 
Valensole plateau 
Beauce plain 
Champagne 
Remoulins 
Landes 
Niger 
- 
" 
" 
- 
" 
" 
Kansas prairie 
Table 1 : Value of re 
Roujean (1994); Ref. 
33 Extracting sur 
Surface albedo exb 
with satellite meas 
surface bidirection 
given by Eq. (4) is 
with a set of daily 
3.66°E), from Ma 
(Rahman and Dedi 
vapor content is de 
Niamey airport Nc 
of the continental t 
climatology. Since 
composition is cho 
models of various 
1992; Rahman et i 
Ross, 1981; Verstr; 
The level 
models under test 
irrespectively of th 
in test sites in a ten 
visible reflectance 
reflectances (using 
reflectances. 
An attracti 
can supplement th 
reconstruction of ra 
sensor would see c 
reflectance in the 
reflectances obtaini 
as could be seen by 
and reflectances ac 
the Niger semi-arid