352 
2. Soils, principally from the United States (Stoner, et al., 1980). These 
564 spectra represent soil samples obtained by the U. S. Department of 
Agriculture's Soil Conservation Service, plus a few from foreign countries. 
Laboratory spectra were measured with soils at field capacity, which tends to 
reduce spectral variability. The measurement range was 0.55-2.32 fim. 
3. Selected vegetation species. These and subsequent spectra were obtained 
for the interval 0.4-2.5 jam. The 597 laboratory spectra represent four 
species (Pinus coulteri - coulter pine, Ceanothus crassifolius Torr. - 
ceanothus, Salvia mellifera Greene. - black sage, and Adenostoma fasciculatum 
Hook & Arn. - chamise) in a variety of modes - leaves, stems, flowers, fruits, 
needles, etc, under varying conditions of fresh/drying/dead. The spectra were 
originally obtained in order to assess potential moisture status determination 
methods (Cohen, 1991a, 1991b, 1991c) 
4. Spectral diversity (Satterwhite and Henley, 1991). This collection of 285 
samples is the best single spectral survey for a variety of surfaces, 
including trees, shrubs, grasses, soils, and rocks. Both field and laboratory 
spectra are included, with emphasis on spectra from arid and semiarid 
environments. The collection is too varied to be summarized here. 
5. Vegetation. These 24 laboratory spectra, of high quality but 
undocumented, represent mainly leaves of trees and bushes: sassafras, locust, 
red oak, yew, poplar, etc. (D. Krohn, private communication, 1991). 
6. Water. Two spectra for sea water and lake water are included in the 5S 
spectral code (Tanre et al., 1990). 
7. Sedimentary rocks. These 160 laboratory spectra from sedimentary rocks in 
the western U. S. have been documented for potential geologic applications 
(Lang, et al., 1990) Many of the identified absorption bands fall in the 
regions which we exclude due to atmospheric water vapor absorption. 
Table 1 presents 19 recommended bands for characterizing the seven spectral 
collections, where water vapor regions are excluded, as are the regions 
0.4-0.55/im and 2.31-2.50 /im, based on the limitations of the LARS data. 
Table 1. Recommended bands for the seven spectral collections (¿¿m) 
These 19 spectral bands describe the ensemble very well, in the sense that E 
is reduced to a value of 0.004%. When the full range 0.40 - 2.5 n m is 
considered we must eliminate data sets 1 and 2 (LARS spectra), which do not 
describe the short and long wavelength regions. For the full spectral domain 
we find that 22 spectral bands are sufficient to describe the ensemble, with a 
residual of E = 0.003%. Thus we conclude that approximately 20 spectral 
intervals are sufficient to describe data sets 1-7, representing commonly 
observed materials such as soils, vegetation and rocks, at least to the extent 
that the available data sets are representative. The addition of minerals and 
man made materials (not described here) introduces considerably more localized 
spectral variability, and many more bands are required. Whether most mineral 
types occur frequently at large enough scale to be seen in satellite data is 
unknown, but may be examined through the study of AVIRIS data. 
0.55-0.60 
0.61-0.65 
0 . 66 - 0.68 
0.69-0.71 
0.72-0.75 
0.76-0.79 1.20-1.28 2.03-2.09 
0.80-0.93 1.29-1.34 2.10-2.16 
0.94-1.02 1.48-1.55 2.17-2.22 
1.03-1.13 1.56-1.70 2.23-2.31 
1.14-1.19 1.71-1.80