Hyperspectral classification approaches for intertidal macroalgae habitat mapping: a case study in Heligoland

作者：Natascha Oppelt · 发表于：Optical Engineering · 年份：2012 · DOI：10.1117/1.oe.51.11.111703 · 被引用次数：42 · 研究领域：Marine and coastal plant biology、Marine Biology and Ecology Research、Isotope Analysis in Ecology

Analysis of coastal marine algae communities enables to adequately estimate the state of \ncoastal marine environment and provides evidence for environmental changes. \nHyperspectral remote sensing provides a tool for mapping macroalgal habitats if the algal \ncommunities are spectrally resolvable. We compared the performance of three classification \napproaches to determine the distribution of macroalgae communities in the rocky intertidal \nzone of Heligoland (Germany) using airborne hyperspectral (AISAeagle) data. The classification \nresults of two supervised approaches (maximum likelihood classifier and spectral angle \nmapping) are compared with an approach combining k-Means classification of derivative \nmeasures. We identified regions of different slopes between main pigment absorption \nfeatures of macroalgae and classified the resulting slope bands. The maximum likelihood \nclassifier gained best results (Cohan’s kappa = 0.81), but the new approach turned out as \ntime effective possibility to identify the dominating macroalgae species with sufficient \naccuracy (Cohan’s kappa = 0.77), even in the heterogeneous and patchy coverage of the \nstudy area.