Remote sensing water quality inversion using sparse representation: Chlorophyll-a retrieval from Sentinel-2 MSI data

Abstract

Remote sensing has been widely applied to estimate the water quality parameters using empirical approaches. Numerous algorithms have been developed for water quality retrieval. However, the identification of key features for inversion and the reduction of uncertainty with limited field measurements are crucial aspects. In this study, satellite-based sparse representation is developed as an optimization model to simultaneously determine the major spectral features, and estimate water quality maps under noisy environment. Result shows that the blue–green ratio is the important feature for estimation of chlorophyll-a (Chl-a) concentration, whereas the NIR-red algorithm is the better one with in retrieving Chl-a in a high concentration case. The Chl-a map is estimated by using main spectral features of Sentinel 2 MSI data constrained with observations (correlations between observations and estimations: over 0.9). The rapid mapping of Chl-a in inland water allows us to assess spatial distribution of the water quality. This study provides reliable and interpretable information for policymakers to implement effective water quality management practices.