Nitrate inversion based on remote sensing in the Pearl River Estuary, China

Abstract

Against a background of ongoing economic and industrial development, the impact of human activities on the offshore environment has become increasingly significant in recent years. In many regions, contamination of the marine environment necessitates the need for dynamic, real-time and wide-coverage water quality monitoring and satellite remote sensing, such that multi-temporal and spatial scale observation of various oceanic parameters are realized. In this paper, using salinity, in situ spectral data, nitrate and other water quality data from five sampling campaigns in the Pearl River Estuary and adjacent waters, studies on the inversion of nitrate concentrations using an artificial neural network were performed. Two inversion models, based on input of the measured salinity and field spectral data and just the field spectral data, were investigated. After comparing the accuracy of the two models, it was considered that the nitrate model with salinity and field spectral data as input performed best. Finally, the selected model was applied to MODIS data to achieve inversion of the nitrate concentrations in the estuary waters and obtain the temporal distribution of nitrate concentrations as a means to monitor and evaluate water quality in the estuary.