Tuominen, J. and Lipping, T., 2014. Feasibility of benthic cover-type mapping in turbid waters near estuaries using hyperspectral remote sensing.

In this study the feasibility of benthic cover-type mapping in turbid waters near estuaries using hyperspectral remote sensing is explored. The majority of research papers addressing remote sensing of benthic cover types concentrate on optically clear, shallow coastal and reef waters. In order to study the effects of water depth and water quality on benthic cover-type classification, a synthetic data set was generated. Synthetic spectra were calculated using in situ optical measurements of water quality and spectra of different bottom types. The data set was classified using the spectral correlation mapper (SCM) and the Euclidean distance (ED) classifier. A simple water-column correction method was also tested. The overall mapping accuracy of SCM classification without water-column correction reached 47.8% when the depth range 0.0–3.0 m was studied. The mapping accuracy increased to 66.0% using the ED classifier with water-column correction. When water quality was changed from least turbid to most turbid, the overall accuracy decreased to 50.4%.