00^ 
Figure 6 : Temporal variation of original and 
filled signal (-o- measured visible, modeled 
visible, -x-measured near infrared, -*- modeled 
near infrared, using Shibayama and Wiegand's 
model, (from Cabol and Dedieu, 1994). 
3.4 On atmospheric 
One should emphas 
assessment of associ; 
of the methods descr 
be corrected relative 
variable in space an 
detailed review of a 
importance, affecting 
depending on the an 
atmospheric water vt 
obtained from Gener 
the lack of data over 
of aerosols is also c 
magnitude as the rel 
episodes can mask c< 
are often unavailable 
measurements of aer< 
(Holben et al., 1994), 
Figure 7: Time series of observed and 
modeled TOA reflectances in the visible 
(a) and near infrared (b) and of derived 
NDVI (c). Crosses represent the 
observations and open circles the 
observations normalized by the 
bidirectional function, (from Ba et al., 
1994). 
4 - APPLICATION 
TIME SERIES 
Ba et al. (1994) have 
modelling, but whic 
Atmosphere reflectar 
and directional TOA 
are adjusted against 
basic assumption im] 
uncoupled. Rfl is the 
reflectances correcte 
G(0j,9 v ,<}>) is a 2-p 
coupled surface-atmc 
Ba et al. (IS 
during the FIFE expt 
observed reflectance 
the visible and near i 
day fluctuations in th 
by the bidirectional fi 
in the visible, and 0.0 
(Table 1). This may ] 
method. In the near i 
detectable feature in l 
summer and winter 
reflectance minimum 
resulting NDVI, com] 
exhibits a relative mi 
the "raw" NDVI dab 
rainfall data (not sho\ 
28