Assessing the uncertainty of radiometric properties in coral reef environments

The spectral information within a single pixel from a satellite image of a coral reef environment can theoretically reveal information on primary and/or secondary environmental factors. That which is of most interest within an image to reef managers and remote sensing scientists is coral species or genera identification, and coral spatial distribution and health. Attempts to map these dimensions, however, have slowed due to the inherent uncertainty of interpreting spectral signatures, which are a function of not only single environmental factors, but as well, the dynamism of their interactions within an aquatic system. What remains uncertain in identifying corals within an image is the precision with which they can be differentiated from submerged features with similar reflectance properties. In attempt to address this uncertainty we revisit spectral derivative analysis of signatures collected within the tidal range of Bunaken Island, Indonesia. Through comparisons of five derivative equations our goal is to identify the one that reveals the greatest potential for spectral discrimination of narrowly defined categories. Results indicate that the second and third spectral derivative equations of smoothed data best preserve reflectance change with wavelength and allow for discrimination between features within narrow wavelength ranges.