A comparison of spectral macroalgae taxa separability methods using an extensive spectral library

Abstract

Remote sensing is one the most promising approaches to coastal area cartography, including mapping algae forests. After discrimination of algal communities from other benthic habitats, next step is species discrimination (from other algae). Spectral signature provides the most complete remote description to characterize any algae. In this work spectral signatures are studied from the point of view of taxa separability to assess the potential use of remote sensors to map seaweed in coastal waters. Three approaches were tested: Red-Green-Brown colorimetry (sRGB), optimal spectral boundary separation based on True Skill Statistics (TSS-OB), and pigment absorbance band detection by Derivative Spectroscopy (DS). An extensive spectral library of 36 algal species present in the Atlantic Galician coast (NW of Spain) is used to test and validate these methods. The results show that the three broad taxa of red, green and brown algae can be separated by all three methods (Cohen's kappa of 0.697, 0.891 and 0.910, respectively). The TSS-OB and the DS approaches provide almost perfect classification (despite some anomalous specimens), with DS being slightly better. The sRGB approach, useful for in situ photographic classification, also provides good results.