8 
The effect of water vapor absorption is very significant for the NOAA-AVHRR near 
IR channel. This channel is used together with the visible channel to derive the normalized 
vegetation index (NDVI) used to monitor the dynamics of global vegetation (Tucker et al., 
1985; Justice et al., 1985). The effect of water vapor on the NDVI is large (Holben, 1986; Tanre 
et al., 1992). There is no operational method to eliminate this effect on a global basis. The 
composite technique used in the generation of the global vegetation index reduces the effect 
of water vapor significantly (by 30-50%). Future generation of NOAA-AVHRR sensors (the 
K, L and M series) and the NASA Earth Observing System (EOS) MODIS instrument are 
planned to have narrow channels in the near IR that will eliminate this problem. 
Direct correction of the aerosol effect is based on determination of the aerosol opacity 
(expressed by the optical thickness and the resulting path radiance) from specific pixels in 
the image and applying it to correct the same image (Kaufman and Sendra, 1988; Holben et 
al., 1992). The aerosol path radiance, Lp, and optical thickness, r a , can be determined above 
dark surfaces for which the surface reflectivity can be estimated. Once such surface type is a 
green forest area. Since the aerosol scattering may make it difficult to find forest pixels, the 
forest pixels are determined based on their reflectance in the mid-IR (2.15 or 3.75 pm) where 
the aerosol effect is minimal (the wavelength is much larger than the particle size). 
In indirect correction for the aerosol effect, instead of correcting the measured 
radiances for the aerosol effect, a remote sensing function is defined (such as the NDVI used 
for remote sensing of vegetation) in a new form that minimizes the dependence on the 
aerosol loading. One such function is the Atmospheric Resistant Vegetation Index (ARVI) 
that instead of the red channel used in the NDVI, uses a combination of the red and blue 
channels (Kaufman an Tanre, 1992). This redefinition of the NDVI as ARVI retains a 
similar information about the surface properties while being less sensitive to the aerosol 
effect. 
In this paper we shall discuss the aerosol effect on remote sensing of the surface 
reflectance, techniques for remote sensing of the aerosol from space, and application of the 
direct and indirect correction techniques to remote sensing from the AVHRR and EOS- 
MODIS. 
2. THE EFFECT OF AEROSOL ON REMOTE SENSING 
Aerosol particles scatter and absorb solar radiation. The scattering and absorption 
characteristics depends on the chemical and physical properties of the particle. Several 
chemical compounds have specific absorption characteristics in some parts of the solar 
spectrum, somewhat similar to the absorption bands of atmospheric gases. An example is 
the red Saharan dust that absorbs stronger in the blue part of the spectrum (Patterson, et al., 
1977). Black carbon is responsible for most of the aerosol absorption, for aerosol originated 
from populated areas (aerosol emitted or formed down the wind from industrial and car 
emissions) and aerosol in smoke from biomass burning, which is widely prevailing in the 
tropical latitude. Its absorption is uniformly distributed across the solar spectrum 
proportional to the aerosol scattering, which decreases with the wavelength. The aerosol 
single scattering albedo, co Q , is a measure of the aerosol absorption. It is defined as the ratio 
of aerosol scattering to scattering + absorption. (co 0 =l for complete scattering and co 0 =0 for 
complete absorption.) Fig. 1 shows the dependence of the atmospheric effect on the surface 
reflectance, for several values of the aerosol single scattering albedo and aerosol optical 
thickness (Fraser and Kaufman, 1985). Increase in the surface reflectance or decrease in the 
value of oo 0 reduces the net aerosol effect on the radiance or makes it more negative. For a 
given value of (0 o there is a value of the surface reflectance, called the critical reflectance, 
for which the aerosol effect is close to zero for any value of the aerosol optical thickness. 
Remote sensing of this value was used to determine the aerosol absorption from space 
(Kaufman, 1987). 
0.1 
o.o 
o.o 
0.0 
0.0 
CL 0. 
I 
-j -0.0 
- 0.0 
- 0.0 
- 0.0 
- 0.1 
Fig. 1: The rat 
reflectance for t 
thickness z A at 
angel is 40°, tht 
the atmospheric 
the dotted line, 
positive (p c =0. 
Kaufman, 1985) 
The effect 
2. The abscissa is 
reflectance in ch; 
actual NDVI (wit: 
the difference be 
atmosphere. Not« 
(low NDVI) and 
on the NDVI is b 
Bare soils (small 
values of surface 
because of the c 
conclusion may n 
be completed. Lar 
reflectance. The ] 
unit) for tropical 
channel 2 p 2 = 0.1 
correspond to lo\ 
lava surface. Tho 
radiance and the 
for different types