| The productivity, biodiversity, and the ecological functions of estuaries are adversely affected by increasing amounts of landcover change occurring in the areas immediately surrounding them. Net primary productivity (NPP) measurements aid the understanding how landcover change impacts estuarine health. NPP is determined from annual changes in the amount of vegetation present and is commonly measured by leaf area index (LAI) and/or biomass. Remote sensing has been demonstrated as a tool for synoptically predicting LAI and biomass in a way that is both non-intrusive and cost-effective. The capture of these estuarine biophysical parameters by remote sensing has traditionally been based on broad-band multispectral two-band vegetation indices. The use of narrow-band (hyperspectral) imagery has also been demonstrated to capture LAI and biomass measurements within estuarine and other environments. However, the full potential of hyperspectral data to capture these parameters within the estuarine environment has yet to be explored. In this study, an Airborne Imaging Spectrometer for Applications (AISA) hyperspectral dataset of Murrells Inlet, SC was utilized to test hyperspectral prediction methods for LAI and biomass measurements of Spartina alterniflora, the key indicator species for estuaries in the southeastern United States. Hyperspectral data were used to compare the predictive capabilities of: traditional two-channel broad-band vegetation indices, narrow-band versions of traditional two-channel broadband vegetation indices, hyperspectral vegetation indices, and derivatives of hyperspectral reflectance data. The research also attempted to compare hyperspectral data analysis tools (Pixel Purity Index(TM) and Mixture Tuned Matched Filtering(TM)) to the vegetation index approach for mapping Spartina alterniflora LAI and biomass. Hyperspectral vegetation indices were found to be significantly less effective at explaining biomass and LAI variance than traditional broad-band vegetation indices. First and 2 nd order spectral reflectance derivatives were shown to improve variance explanation. Principal components analysis and stepwise multiple regression were also shown to improve variance explanation. Due to a variety of in situ and overhead data collection issues, no Spartina alterniflora endmembers were obtained for examination by hyperspectral data analysis tools. |
