Quantification of biomass and leaf-area-index in a Charleston, SC estuary using low-altitude AVIRIS imagery

Estuaries are highly dynamic ecosystems that need to be better managed to mitigate current and future damage. New management policies are often derived from studies that examine the biophysical variables that are associated with an estuary. In certain instances, remote sensing may provide a more timely, cost-effective method of data collection. Until recently, the examination of vegetation presence, health, and biomass has been determined with broad spectral band sensors. The purpose of this research was to evaluate the relationship between specific biophysical variables and high spatial resolution hyperspectral data and to determine their relationship to ecologically significant measurements of plant populations. This study had three objectives: (1) to examine the statistical relationship between total aboveground biomass and LAI of smooth cordgrass (Spartina alterniflora), (2) to examine the correlation between the low-altitude Airborne Visible Infrared Imaging Spectrometer (AVIRIS) data and the in situ biophysical variables, (3) to determine whether or not unique spectral signatures for Spartina alterniflora can be derived from the AVIRIS data. The prediction of total aboveground dry biomass and LAI from AVIRIS imagery was performed using a variety of methods. These methods included statistical analysis (vegetation indices) and spectral mixture tuned matched filtering. The vegetation indices used were Simple Ratio, normalized difference vegetation index (NDVI), soil adjusted vegetation index (SAVI), and Infrared Index. The spectral analysis involved: computing the minimum noise fraction (MNF), identifying pixel purity, visualizing the data, and performing the mixture tuned matched filtering. A relationship between the in situ biophysical variables and the low-altitude AVIRIS data has been established through the statistical and the spectral analyses. Biomass maps were derived from the regression relationships of the vegetation indices and biomass. The amount of acreage covered by various biomass ranges was calculated from these maps. The mixture tuned matched filtering generated a unique spectral signature for the high marsh area. Based on the results in this study a combination of precise in situ spectral signatures and in situ biophysical variables would provide a base map for the study area. The future of hyperspectral data as a tool for resource managers appears promising.