Water Column Radiative Transfer Compensations for Coral Reef Remote Sensing
Eric
J
Hochberg, University of Hawaii, hochberg@hawaii.edu
(Presenting)
Marlin
J
Atkinson, University of Hawaii, mja@hawaii.edu
Deterministic remote sensing classification of coral reef bottom-types requires compensation for the combined effects of variable water depths and optical properties. One approach is implicit compensation, where a classifier is trained using sea surface remote sensing reflectance spectra, and imagery is classified directly without removing water column effects. Model results suggest that, for relatively clear water at least, coral, algae and sand are largely distinguishable from each other to depths of ~20 m using this approach. We are currently collecting field data to construct the spectral library necessary for further investigation and application. Another approach is to explicitly remove water column effects from remote sensing imagery, then apply a classifier built using known bottom-type reflectance spectra. We are developing an empirical technique based on simple decorrelation of visible wavebands from known water depth. Decorrelated wavebands are corrected to absolute reflectance, then the imagery is classified based on a library of reflectance spectra from coral reefs around the world. This investigation is ongoing.