This research models plant species distributions across the arid White Mountains in eastern California. We estimated environmental and spectral envelopes of major plant species using multivariate analyses of vegetation composition in 500+ ground control points relative to geologic, topographic, and hyperspectral (AVIRIS) data at those sites. We then predicted species distributions across 830 km2, and validated the statistical models.

The White Mountains, rising to 4000+ meters along the western boundary of the Basin and Range province, provide classic examples of the effects of geology and climate on vegetation. Their well-mapped mosaic of dolomite, limestone, argillite/phyllite/quartzite, granitoid, and volcanic rocks affects chemical, thermal, and physical properties of soils. Local climate is driven by regional and local topography. As elevation increases, precipitation increases and temperature decreases. Aspect, slope, and surrounding topography determine potential insolation, so that south-facing slopes are warmer and north-facing slopes cooler at a given elevation. Topographic position (ridge, slope, canyon, or meadow) and slope angle affect sediment accumulation and soil depth. These factors form complex environmental gradients that have profound effects on plant distributions. Spectral data provide snapshots of existing vegetation structure across broad areas. This study compares different combinations of hyperspectral, geologic, and topographic variables to describe the environmental and spectral envelopes of species distributions using Canonical Correspondence Analysis (CCA). CCA models species "envelopes" in multidimensional environmental space, which can then be projected across entire landscapes. Different methods of incorporating AVIRIS data include Minimum Noise Fraction (MNF) bands, raw AVIRIS bands, and band ratios. Two-thirds of the ground control sites were used for calibration of CCA models, and one-third were reserved for evaluation using an independent measure of fit (kappa). This investigation examines how to best extract the most useful information for mapping arid vegetation from the enormous volumes of data incorporated within 224 narrow bands of AVIRIS imagery.