List of Figures

Figure 19.1 Relative spectral variation as a function of spatial scale and level of biological organization, from biochemicals to global biosphere. The spatial scales typical of direct measurements and satellite sensors are indicated. Current satellites are the Advanced Very High Resolution Radiometer (AVHRR), the Landsat series Multispectral Scanner (MSS), the Thematic Mapper (TM), the European Systeme Probatoire d'Observation de la Terre (SPOT), and the European Radar Satellite (ERS-1). The minimum pixel resolution is shown at the tip of the arrows and the field-of-view of the sensor is shown by the line length.


Figure 19.2 Typical chronosequence of secondary pine-hardwood forests succession in the eastern United States. The upper tree profile depicts community dynamics and canopy properties expressed as percentage of maximum values, illustrating changes in ecosystem structure, canopy gaps, LAI, and biomass. The typical range of maximum LAI is 6-8 (Nehmeth, 1971), of pine density is 1,000-30,000 tress/ha, of hardwood density is 5000-6000 trees/ha, of above ground pine biomass is 20-40 kg/m2, and of above ground hardwood biomass is 18-30 kg/m2 (Peet and Christensen, 1987). The lower tree profile illustrates differences in the canopy location of maximum signal derived from optical and radar sensors.


Figure 19.3 Three reflectance end members used to model spectral variation from an AVIRIS scene covering part of Owens Valley, California, a typical mixed pixel spectrum, and a residual spectrum. Mixtures of these end members—49% vegetation (foliage from semiarid shrub species), 19% shade, 30% granitic (gray) soil, and 0% weathered (tan) soil (not shown)—provide a best fit to the measured pixel. Image mixtures are calculated from the multispectral variation on a pixel-by-pixel basis, using a simple linear calibration (Smith et al., 1990 a,b). The residual spectrum represents the remaining pixel variation unaccounted for by the model.


Figure 19.4 Contour plot of (A) positive (higher reflectance) and (B) negative (lower reflectance) residuals across the 400-to 2500-nm spectrum, extracted from 12 separate locations (means of 9 pixels) on the valley floor. The maximum deviation from the model was <± 0.3 and the mean <± 0.2. Biogeochemical properties represented by the residual spectra may be identified by comparison with spectra of known materials through spectral matching routines.

List of Plates

Plate 3 (A) AVIRIS vegetation end member front the Sierra Nevada bajada (left edge of image) and the floor of Owens Valley, California, including the city of independence (right edge of image), at junction of Independence Creek (upper center, extending from the left to right) and the toe of the alluvial fan, obtained July, 1989. (B) Thematic Mapper (TM) vegetation end member from May 1985, and (C) TM, December 1982. Note that A is a higher resolution image and covers only the central area within the black frame in B. The Sierra Nevada bajada is missing from the left side and the Owens River is missing from the right side of the images. Images are color-density sliced into low (0-20%, gray), intermediate (21-30%, yellow), and high (>30% green) vegetation cover classes. The vegetation end member image overlays the shade end member (shown as gray tones), which adds some information about topography to the display. 

 
Plate 4 Vegetation end member concentrations calculated from three spectral regions. The visible wavelengths (473-643 nm, A) have limited spectral contrast, making mixtures of vegetation, soils, and shade difficult to separate. The regions in the near-infrared (783-877 nm, B) and the shortwave infrared (1286-2375 nm, C) have greater contrast between vegetation and background materials, and increased spectral resolution of vegetation. The relative abundance of each end member corresponds directly to the image brightness. Note that these images only include a small area around Independence, California, seen in the upper right in Plate 3A. 

 
Plate 5 Fraction images of the four reflectance end members derived from 171 AVIRIS bands. Atmospheric water vapor bands and bands of low signal/noise were excluded from the analysis. Images from left to right show the gray soil (A), shade (B), tan soil (C), and vegetation (D) end members. The relative abundance of each end member corresponds directly to the image brightness; fractions sum to unity. 

 
Plate 6 Composite false-color images formed from three end members or residuals. Images derived from analyses of satellite data may be recombined into new composite images displaying properties not visualized-directly in the original data. (A) Composite image showing interactions among the visible (blue), near-infrared (green), and shortwave infrared (red) vegetation fractions. The spatially distinct areas differentially contribute to the composite vegetation end member from the three spectral regions. (B) Composite image of the three end members: tan soil (red), vegetation (green), and gray soil (blue). The hue varies with the magnitude of the numerical value; colors depend on the relative proportions of the end members in each pixel. (C) Composite image of residuals at 525-nm (blue), 809-nm (green), and 1100-nm (red) regions. The high residuals are not random but show clear wavelength-specific spatial associations suggesting biogeochemical differences in surface conditions. 

 
Plate 7 Residual images (difference between calibrated reflectance and estimated mixture spectrum) show areas where mixtures of the four end members do not fit the measured spectral variation at specific wavelengths (shown from left to right are 574, 986, 1254, and 1333 nary).