Evaluating Vascular Plant Composition and Species Richness on Horn Island, Mississippi, Using Passive and Active Remote Sensing in Conjunction with Ground Based Measurements

Barrier island vegetation is subjected to chronic abiotic stressors combined with periodic storm events that favor species adapted to harsh environments. These islands are the first landforms to be affected by changes in coastal subsidence and sea-level rise. Evaluating changes in vegetation is important for understanding the impact of global climate change on coastal environments. This study assesses vegetation composition and plant species richness on Horn Island, Mississippi using ground data in conjunction with remotely sensed spectral and LIDAR data. The goals of this research are to: 1) classify and map vegetation composition on Horn Island using hyperspectral and LIDAR data, 2) evaluate changes in vegetation composition through comparison with a vegetation study and classification map from 1979, 3) determine the extent to which vascular plant species richness might be estimated using remotely sensed spectral reflectance indices and spatial variability within these indices, and 4) utilize the vertical distribution of airborne multiple-return LIDAR data to evaluate vascular plant species richness. The vegetation composition of habitat-types on Horn Island can be identified by indicator species that are consistent both over time and among other barrier islands. Additionally, combining airborne hyperspectral and LIDAR data improved the overall classification accuracy of habitats. Although only broad comparisons in vegetation changes could be made between this study and previous maps, these changes were linked with geomorphological changes. In simple linear regressions, various reflectance- and LIDAR-indices correlated significantly (p \u3c 0.05) with richness when habitat-types were considered individually. Regressions of richness with indices derived from within-transect means or spatial variability in reflectance, reflectance band ratios, as well as vertical distribution descriptors and height percentiles from LIDAR data produced estimation errors of 0.4-2.5 species per transect. Best-fit indices from hyperspectral data indicate spectral bands in the near- and mid-infrared spectra are most important in the estimation of plant species richness while LIDAR indices indicate the importance of vegetation height and structural complexity in estimating plant species richness. The capability of utilizing remotely sensed data to classify vegetation composition and estimate species richness provides a promising means of assessing and monitoring vegetation on barrier islands