Remote sensing derived vegetation indices are an efficient metric for mapping vegetation in coastal areas. Spatially explicit analyses are used to address questions of ecosystem structure and related processes and as a tool for informed management and change detection in coastal habitats. High resolution airborne imagery is available, providing increased ability to compare and classify vegetation based on variable spectral reflectance features. We addressed the following questions in this study: Which vegetation indices best reveal spatial patterns of plant composition and abundance, are such indices useful for cross site comparisons of vegetation abundance and distribution, and are within site patterns and across site differences relatable to hydrology, climate, and geomorphic settings. Using a University of Nebraska aircraft with an AISA instrument, 1 or 1.5 m resolution, hyperspectral images were obtained between 2002 and 2008 at National Estuarine Research Reserves (NERRs) in Florida, South Carolina, Mississippi, Delaware, Maryland, Georgia, and Texas. To date, analyses were performed for four locations (SC, MS, GA, and TX). Non salt marsh features were masked using wavelength specific reflectance differences or manually using image interpretation and a digitizing tablet.  ENVI software was used to mosaic flight lines, calculate vegetation indices, and produce classified images and related statistics.

Results/Conclusions

We compared salt marsh images with a suite of indices, including NDVI, SAVI, VARI_green, and PRI. For TX, MS, GA, and SC, the mean, median, and standard deviation values for NDVI were (0.381, 0.396, 0.104); (0.415, 0.405, 0.114); (0.355, 0.373, 0.160); and (0.443, 0.447, 0.108) respectively. Similar inter-site differences were found with SAVI, whereas PRI revealed different patterns perhaps related to physiology. When NDVI values were stratified into three classes (<0.35, 0.35-0.50, and >0.50), more significant inter-site variations were noted. Overall, TX had the lowest proportion of pixels in the >0.50 class (6.5%), indicating a lower amount of high biomass stands, while SC had the highest (28.4% of pixels). A substantial portion of GA marsh pixels (44.1%) were below 0.35, corresponding to extensive exposed mud flats, salt pans, and low biomass high marsh. GA also had the largest coefficient of variation for NDVI (45.1%). Another difference was noted in the degree of high biomass vegetation along waterways, with striking occurrences of tall Spartina in GA, moderate occurrences in SC, and much lower abundances in the low tidal amplitude MS and TX sites. This approach allows rapid assessments of cover and abundance.