Independent component analysis: A tool for algal bloom detection

Abstract

Independent component analysis decomposition of hyperspectral signals has characteristics ideal for detecting harmful algal blooms in coastal waters. In this proceeding, independent component analysis is used to process images taken by the HYPSO-1 satellite at two different locations: Lake Erie and the Salish Sea. For each location, a set of components are produced and compared to two common water quality indices. The results demonstrate that some of the spatial-spectral features uncovered by independent component analysis resemble common water quality indices but also highlight the need for improved interpretation of the recovered features.