On the possibility of mapping living corals to the species level based on their optical signatures

Abstract

Several methods (Spectral Angle Mapper, first derivatives, spectral clustering) were used to test the possibility of discriminating live corals at the species level by using reflectance spectra obtained by close range underwater remote sensing. The Spectral Angle Mapper used with an in situ spectral library of 73 coral colonies recognised all corals correctly in an artificial remote sensing image created from the same coral spectra. The results did not change when “grey” noise was added to each pixel in the artificial image or when the artificial image was made from reflectance spectra modelled through a 5-m water column. The results showed that the within-species variability in optical properties of a set of four coral species, for which at least three different specimens were measured, was as high as between-species variability. Thus, it is highly unlikely that hard corals can be discriminated at a species level by means of remote sensing. Cluster analysis showed that it is difficult to recognise corals also at the genus level ( Acropora vs. non-Acropora) based on their reflectance spectra. However, visual inspection of the coral spectra and clustering analysis suggest that it is possible to separate three main optical types of corals: green, blue, and brown corals. This may form the basis for an optical classification of corals that is more suited for remote sensing than traditional taxonomy.