729 
View Considerations 
The view angle effects on vegetation indices are demonstrated in Figure 6 using the simulated data with US standard 
atmosphere and Continental aerosol models at 23 km visibility after Rayleigh correction. None of the vegetation indices 
considered are perfectly symmetric about the nadir. In general, the forward scattering directions (positive view angles) 
resulted in higher vegetation index values than the backscattering (negative view angle). Also, the maximum values of 
all indire occurred at large view angles, not at the nadir as assumed by maximum value compositing algorithm (Holben, 
1986). The PVI appeared to vary litde with sensor view angle. This could be advantageous for compositing, but on the 
other hand this also indicates that the PVI is not sensitive to vegetation changes because off-nadir viewing sensors do 
see more vegetation than nadir viewing sensors. Consequendy, the choice of vegetation indices with respect to view 
angle responses depends the purpose of using vegetation indices. 
View Angle (degree) 
View Angle (degree) 
Figure 6. Bidirectional effects on vegetation indices under varying conditions, (a) ground data, (b) 
with simulated atmosphere and (c and d) with Rayleigh correction. 
CONCLUDING REMARKS 
All vegetation indices examined in this study were shown to be primarily sensitive to vegetation changes although their 
dynamic ranges were quite different. The vegetation sensitivities of most vegetation indices are dependent upon the 
vegetation densities of the targets. At low densities (LAI <1.0), the NDVI was shown to be the most sensitive index 
to vegetation variations, while at high vegetation densities (LAI>1.0) the MSAVI was shown to be most sensitive one. 
With the use of blue spectral band in ARVLSARVI, and ASVI, it appeared that vegetation sensitivities were increased. 
The GEMI, ARVL and NDVI were very sensitive to soil background variations. This is most likely due to 
the assumption that all isolines converge at the origin in red-NIR plane (NDVI, and ARVI), or due to the non-linear 
combinations of spectral bands (GEMI). The SAVI reduced soil noises, but the use of the constant L (0.5) buffered the 
sensitivity to vegetation variations, as the L value of 0.5 is usually much larger than the red reflectances. The MSAVI, 
with a variable self-adjusting L function, increased the vegetation sensitivities while resulting in a minimal soil 
background effects. However, both MSAVI and SAVE are quite sensitive to atmospheric conditions and, therefore, would 
be a good vegetation indicator for only ground- or low altitude-based remote sensing measurements, or for space 
remote sensing measurements with atmospheric corrections. 
The NDVI, SAVI, and MSAVI were also sensitive to atmospheric conditions. The inclusion of the blue band 
in ARVI and ASVI reduced the sensitivity to atmospheric variations significantly. But the use of blue band worsened 
tbe soil background noise in the ARVI, especially at low vegetation densities. The use of the blue band in SARVI