Comparison of three empirical methods for water depth mapping with case study of Pratas Island

摘要

Statistical methods to map water depth from medium-high resolution multispectral images were easier and more popular than wave spectrum bathymetry or water scattering-based implementation. However, less studies compared the effectiveness of the popular statistical methods for pelagic islands. This study used the Log ratio transform, primary component analysis and independent component analysis methods to retrieve water depth of Pratas Island, using one Landsat 8 Operational Land Imager (OLI) image. Results showed that the Log ratio transformation was not the best method as the proposer suggested. The first primary component and the second independent component are good predictors for absolute water depth ranging from 0 to 20m, while Log Ratio was more sensitive to water depth ranging from 0 to 5m, IC2 was sensitive to water depth between 5 and 10 m.