726 
RESULTS 
Vegetation, Soil, and Atmosphere Sensitivities 
As the ARVI and SARVI were meant to be used with data after the Rayleigh scattering correction, we applied the data 
before the corrections to NDVI, SAVI, PVI, MSAVI, and GEMI, and to all other indices after the Rayleigh correction. 
The dynamic ranges of the NDVI, SAVI, PVI, and MSAVI, are shown in Figure 2a. while their sensitivities to the 
vegetation density are plotted in Figure 2b with ground data (before atmosphere simulation). The lower limits of all 
vegetation indices considered are similar (0.0 - 0.1). The dy nami c ranges of these indices are 0.82, 0.72, 0.40, and 0.81 
for NDVI, SAVI, PVI, and MSAVI respectively. The sensitivity to vegetation changes of these indices (Figure 2b) 
differed, depending on the vegetation densities. In general, when vegetation density is low (LAI < 1.0), the NDVI is 
most sensitive to vegetation variations and SAVI and MSAVI are the second. When LAI is greater than 1.0, MSAVI 
become the most sensitive index to vegetation, while the SAVI was the second, and the NDVI became the lowest when 
LAI > 2.5 due to its saturation at high vegetation densities (Qi et al. 1994). 
The sensitivities of these vegetation indices to soil background variations are plotted in Figure 2c and those 
to atmospheric visibility are given in Figure 2d. The soil sensitivity (Fig. 2c) can be divided into four zones: LAI < 
0.2, 0.2 <LAI<0.4, 0.4< LAI < 1.2, and LAI > 1.2. In the first zone, the NDVI is the most sensitive index to soil 
variations, while the PVI is the least sensitive one. In the second zone, the sensitivity of NDVI remained the highest, 
but the MSAVI and SAVI dropped to their lowest levels. In the third zone, the NDVI became less sensitive to soils than 
the PVI, but still more than MSAVI and SAVI. When the LAI is greater than 1.2, NDVI became the lowest of all 
indices considered. Virtually, NDVI is independent of soil background at this time. This is due most likely to its 
saturation to vegetation changes as seen in Figures 2a and 2b. The atmospheric sensitivity (Fig. 2d) of PVI was highest 
and the NDVI was lowest, while the MSAVI and SAVI fall in between. 
Figure 2. Dynamic ranges (a) of NDVI, PVI, SAVI, and MSAVI indices and their sensitivities to 
vegetation (b), soils (c), and atmosphere(d) using simulated data. 
In Figure 3, all eight vegetation indices are compared using the data simulated with US standard atmosphere 
and continental models, but corrected for Rayleigh scattering. Notice the differences in the lower limits of these 
vegetation indices (Fig. 3a). The use of blue band in ARVI, SARVI, and ASVI resulted in their values being about -0.2 
for bare soils, while that for GEMI was still 0.3. The use of the blue band in ASVI resulted a greater vegetation 
dynamic responses than that of SARVI. The sensitivity to vegetation variations of ARVI, SARVI, GEMI. and ASVI 
are lower than NDVI, SAVI, PVI, and MSAVI at LAI less than about 0.4 (Fig. 3b). At high vegetation densities, the 
use of blue band resulted in high vegetation sensitivities. However, the sensitivities of all vegetation indices, except 
for NDVI at LAI less than 0.4, are similar due to the atmospheric effects. In general, the MSAVI and the ASVI were 
more sensitive to vegetation variations than other indices, especially when vegetation density was high (Fig. 3b).