USE OF HYPERSPECTRAL (AVIRIS) DATA TO CLASSIFY AND MAP MIXED PHYTOPLANKTON BLOOMS IN COASTAL WATERS

摘要

Hyperspectral (AVIRIS) image data of Florida Bay were classified using image-derived spectra extracted from pixels that corresponded to ground-truth sampling sites. The spectra were incorporated into a spectral library and used as supervised classifiers with the Spectral Angle Mapper algorithm. Spectra used to build the spectral library were extracted from a subset (first 60 of 224 total AVIRIS bands) to generate spectral library endmembers with a wavelength range from 365-930 nm. This wavelength range corresponds to that in which algal pigment signatures are found. Algal accessory pigments were identified and quantified by HPLC analysis of filtered water samples acquired at the ground-truth sights during the AVIRIS flight. Using this approach we were able to discriminate, classify, and map Florida Bay phytoplankton according to bloom composition (diatoms, cyanobacteria, and green microalgae) as well as relative concentrations of bloom components. Mapped images included those both with and without ground-truth sample sites.