Impact of tissue, canopy, and landscape factors on the hyperspectral reflectance variability of arid ecosystems

摘要

Changes in vegetation distribution and condition commonly occur in arid ecosystems clue to land use and climate variability. Most arid land remote sensing efforts have focused on detecting vegetation change using spectral: indices, such as the normalized vegetation index, with limited success. Less attention has focused on using the continuous shortwave spectrum (0.4 mu m to 2.5 mu m) for studying vegetation in arid environments. Using field measurements and a photon transport model, we quantified the absolute and relative importance of tissue, canopy, and landscape factors that drive pixel-level shortwave reflectance variation along a land-cover gradient in the Chihuahuan Desert, New Mexico. Green foliage, wood standing litter, and bare soil had distinctive spectral properties, often via specific, narrow absorption features and through overall differences in the shape of their shortwave spect ra. While the amount of each plant material varied significantly along the land-cover gradient, foliar optical properties remained relatively stable, supporting the hypothesis that resource variation (e.g., water and nutrients) is more strongly resolved at the scale of whole plant canopies (e.g., via allocation and production) than at the leaf level. Significant variation in vegetation type and condition along the gradient resulted in only subtle changes in pixel-level reflectance variability, which could be determined in, high spectral resolution Airborne Visible and Infrared Imaging Spectrometer data. Most important, the relative impact of tissue, canopy and landscape factors on pixel-level reflectance shifted with plant composition and phenology. Mie compared the ability to resolve these vegetation and soil factors using Airborne Visible and Infrared Imaging Spectrometer; Moderate Resolution Imaging Spectrometer, and Landsat Thematic Mapper optical channels and found that feu; factors could be accounted for unless most of the spectral range was adequately sampled. (C) Elsevier Science Inc., 2000. [References: 55]