Estimating Coral Reef Benthic Coverage of Communities from Airborne Hyperspectral Remote Sensing Data: Multiple Discriminant Function Analysis and Linear Spectral Unmixing
A staged approach for the application of linear spectral unmixing techniques to airborne hyperspectral remote sensing data of reef communities of the Al Wajh Barrier, Red Sea, is presented. Quantification of the percentage composition of four different reef components (live coral, dead coral, macroalgae and carbonate sand) contained within the ground sampling distance associated with an individual pixel is demonstrated. In the first stage, multiple discriminant function analysis is applied to spectra collected in situ to define an optimal subset combination of derivative and raw image wavebands for discriminating reef benthos. In the second phase, unmixing is applied to a similarly reduced subset of pre-processed image data to accurately determine the relative abundance of the reef benthos (R2 > 0.7 for all four components). The result of a phased approach is an increased signal- to-noise ratio for solution of the linear functions and reduction of processing burdens associated with image unmixing.
Introduction
Standardized survey protocols are of value to coastal managers because they permit assessment of regional biological diversity and ecological status, allow comparison of status within and between ecoregions and facilitate the detection of change in coastal ecosystem status against established baselines (English et al. 1997). Remotely sensed imagery provides an attractive tool for quantitative and systematic monitoring of coral reef health at a broad synoptic scale (e.g. the typical swath width of a remotely sensed image ranges from 500 m to 100 km). Since the 1970s, sensing instruments have increasingly been operated from satellite or airborne platforms to acquire consistent imagery over reefs that would otherwise be expensive or a logistical challenge to survey.
Seafloor coverage data are typically derived from imagery through the application of a supervised image classification that assigns each pixel to the class to which it appears most spectrally similar. More recently, hyperspectral sensors have become available, which sample many narrow sections of the electromagnetic spectrum to provide a contiguous coverage of radiance measurements across (Mather 2004)…
