estimate the value at 0.94 ¿¿m, while the N/W method takes the ratio of a 0.03 
¿¿m wide band to a 0.07 nm wide band centered on 0.94pm to estimate the 
strength of atmospheric absorption. By analysis of the surface data we find 
that the 11 relatively broad bands describe this region fairly well: The CIBR 
yields a root mean square error at 0.94 of 1.3%, the N/W an error of .6%, and 
the expansion in basis functions an error of 1.2%. Thus we conclude that the 
procedure used here yields a reasonable estimate of surface reflectance in 
regions of moderate water vapor absorption. For measurement of the water 
vapor effect we add a spectral channel at 1.12-1.15 pm. Figure 1 illustrates 
the first basis function derived by using 11 surface basis functions, then 
selecting the interval 1.12-1.15pm as the observation band. 
Figure 1. Basis function describing atmospheric attenuation of AVIRIS data as 
compared to surface spectra. Spectral band is 1.12-1.15pm, while peaks at 
0.94, 1.38 and 1.95 are due to spectral correlations. The regions of strong 
absorption are unsatisfactory for water vapor estimation because they are also 
highly variable (also due to water) in the ground spectra. 
6. CONCLUSION 
From examination of approximately 3000 laboratory and field spectra and 28 
AVIRIS scenes it appears that approximately 20-25 measurements are adequate to 
define the spectral variability of most natural surfaces, excepting minerals. 
Thus improvement of the Thematic Mapper, in terms of spectral bands, is 
recommended. It also appears possible to simplify treatment of atmospheric 
effects for comparing remotely sensed spectra with a spectral library. 
7. REFERENCES 
7.1 Journal References 
Carriere, V., and J.E. Conel, 1993, Recovery of Atmospheric Water Vapor 
Total Column Abundance from Imaging Spectrometer Data Around 940 nm - 
Sensitivity Analysis and Application to Airborne Visible/Infrared 
Spectromenter (AVIRIS) Data, Remote Sens, of Envir., 44:179-204. 
355