ie NOAA-AVHRR near 
derive the normalized 
getation (Tucker et al., 
;e (Holben, 1986; Tanre 
on a global basis. The 
Index reduces the effect 
A-AVHRR sensors (the 
VIODIS instrument are 
his problem. 
a of the aerosol opacity 
from specific pixels in 
Sendra, 1988; Holben et 
n be determined above 
e such surface type is a 
3 find forest pixels, the 
(2.15 or 3.75 pm) where 
; particle size). 
rrecting the measured 
'such as the NDVI used 
he dependence on the 
:getation Index (ARVI) 
on of the red and blue 
»VI as ARVI retains a 
sensitive to the aerosol 
sensing of the surface 
and application of the 
:he AVHRR and EOS- 
ttering and absorption 
)f the particle. Several 
ome parts of the solar 
: gases. An example is 
ctrum (Patterson, et al., 
, for aerosol originated 
rom industrial and car 
idely prevailing in the 
s the solar spectrum 
avelength. The aerosol 
t is defined as the ratio 
scattering and co 0 =0 for 
ric effect on the surface 
do and aerosol optical 
ance or decrease in the 
it more negative. For a 
the critical reflectance, 
rosol optical thickness, 
absorption from space 
Fig. 1: The radiance of the earth-atmosphere system (in reflectance units) minus the surface 
reflectance for nadir observation as a function of the surface reflectance. The total aerosol optical 
thickness x A and the single scattering albedo co A are indicated for each line. The solar zenith 
angel is 40°, the wavelength is 0.61 pm. Power law size distribution was used with v-3. Note that 
the atmospheric effect is zero for an empty atmosphere, (L - p=0) and the aerosol effect is zero for 
the dotted line. For surface reflectance under a given critical value (p c ) the aerosol effect is 
positive (p c =0.25 for co A -0.96) and above this value the effect is negative, (after Fraser and 
Kaufman, 1985). 
The effect of aerosol on the NDVI derived from the AVHRR data is examined in Fig 
2. The abscissa is the reflectance in AVHRR channel 1 (0.64 |im) and the y-axis is the 
reflectance in channel 2 (0.83 pm). Solid lines, labeled NDVI, represent iso-lines of the 
actual NDVI (with no atmospheric effect) and dashed lines, labeled 5, represent iso-lines of 
the difference between the actual NDVI and the NDVI* measured at the top of the 
atmosphere. Note that the two oval areas on the figures roughly correspond to bare soils 
(low NDVI) and highly vegetated surfaces (high NDVI). Simulation of the effect of aerosol 
on the NDVI is based on an aerosol optical thickness corresponding to a visibility of 23 km. 
Bare soils (small NDVI) show small perturbation to the aerosol effect. It corresponds to 
values of surface reflectances which are less sensitive to the presence of a scattering layer 
because of the compensation between atmospheric reflectance and transmission. That 
conclusion may not be valid for very turbid atmospheres since the compensation may not 
be completed. Larger differences are observed for high NDVIs because of the low channel 1 
reflectance. The perturbations introduced by aerosol scattering may reach 0.10 (in NDVI 
unit) for tropical or boreal forests where the reflectance in channel 1 is pi=0.02 and in 
channel 2 p2=0.20. Note that the largest differences which were found for low NDVIs 
correspond to low reflectances in both bands. Such conditions are representative water or 
lava surface. Those results confirm the importance of atmospheric effects on the detected 
radiance and the NDVI, and point out the relative importance of atmospheric parameters 
for different types of surfaces. 
9