Deconvolution of measured spectra based on principal components analysis and derivative spectroscopy

Submerged coral reef ecosystems are difficult to monitor remotely due to the specific problems of optical attenuation of energy due to the overlying water column, and spectral identification of optically similar substrates. Spectral similarity of submerged coral reef ecosystem features results in low confidence levels when visual interpretation of remotely sensed imagery is used to create thematic maps. For example, both white sand and bleached coral have optically similar spectral responses, so confusion arises when classifying an image. To overcome the problems associated with visual interpretation of submerged coral reef features, spectral reflectance measurements of coral reef features were taken with a high spectral resolution spectroradiometer. Spectral deconvolution methods were used to identify specific wavelength regions where distinction between features could be accomplished. The results of the spectral derivative analysis and principal component analysis performed reveal that discrimination with a hyperspectral instrument is possible in the wavelength regions of 596-602 nm and 656-665nm