Forest structure and biomass estimates derived from red edge and water absorption geometry using AVIRIS data

Geometric analysis of spectral signatures collected from known forest sites using narrow spectral bands in the red-edge and five water absorption regions are the basis for developing ratios to estimate forest biomass and canopy structure. The study area used to test the methodology was located within Congaree Swamp National Monument in central South Carolina. This is one of the oldest stands of hardwood bottomland forest preserved in the United States. Low altitude-high spatial resolution AVIRIS data collected at 3,800-meter altitude, that produced a 3.8 meter spatial resolution, were associated with USGS-NPS National Vegetation Inventory data in order to develop narrow band spectral analysis. Ratios were developed based on geometric properties of spectral signatures associated with biomass and canopy categories identified during spectral analysis and used as input data for a series of high dimensional unsupervised classifications. The ratios were developed from the four best performing spectral intervals: the red edge and three water absorption regions. Classification results were compared to test sites for accuracy assessment. The ratios produced from selected spectral bands in the regions between 707 and 754 nm, 918 and 994 nm, 1448 and 1480nm, and 1963 and 1993 nm showed significant spectral discrimination of vegetation features associated with the three categories of biomass and canopy.; The research indicates that the limited areal coverage provided by ground truth data prevented a good correlation between image processing results and vegetation inventory data. Nonetheless, results of spectral analyses showed that biomass and canopy attributes could be identified using band ratios based on geometric properties of electromagnetic responses in the spectral regions selected. It is recommended that further research should be based on more extensive ground truth in order to evaluate forest feature detection ability of the ratios proposed. This methodology can be applied to other forest areas with minor adjustments.