The Use of Hyperspectral Remote Sensing to Assess Vascular Plant Species Richness on Horn Island, Mississippi

Relationships of plant species richness with spectral indices derived from airborne hyperspectral image data were evaluated for several habitat-types on Horn Island, Mississippi, northern Gulf of Mexico. A 126-band hyperspectral data cube of Horn Island acquired by the HyMap imaging system covered the 450-2500 nm spectrum at a 3 m Ground Sample Distance (GSD). Reflectance spectra were extracted from 5-11 HyMap pixels representing each of 95,15-m vegetation line transects that were established randomly on the island. In simple regressions, no index related to richness when data from all habitat-types were combined. a number of reflectance and spectral derivative indices related significantly (p \u3c= 0.05) with richness However, when habitat-types were considered separately. Only those indices which passed a fidelity test based on the consistency of index response to changes in plant and environmental moisture were selected as potentially reliable indicators of richness. Transect coefficient of variation (CV) for R-1056/R-966 and R-920/R-834 related negatively with richness in meadows and transition zones, respectively. The CV of R-951/R-1100 and R-904 related negatively with woodland richness. Negative regression slopes, field observations and spectral mixtures indicated that richness in these habitats declined with increased bare soil exposure. In marsh habitat, positive relationships of richness with mean R-618/R-2475 or the CV of R-514/R-2459 were explained by the increasing presence and patchy distribution of broadleaved vegetation in the progression from wetter, low-richness sites to slightly-elevated, higher-richness sites. Present results combined with those of a previous grassland study suggested that remotely-sensed indicators of soil exposure may be generally useful in the assessment of plant species richness in mesic habitats. (c) 2008 Elsevier Inc. All rights reserved