Estimation of various constituents of case 2 waters using neural network algorithms from ocean colour satellite data

An artificial neural network model has been developed to relate the ocean constituents and water-leaving radiances acquired from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS). The network has been trained using Levenberg-Marquardt algorithm on a dataset, obtained through in-situ sampling, containing measured water-leaving radiances and concentrations of water constituents. Chlorophyll and sediment maps were generated for a region in the Arabian Sea and on a comparative study with maps generated using the OC-2 algorithm the neural network model, for chlorophyll estimation, was seen to have better approximation properties. An intuitive study of sediment maps was also conducted. The source codes were generated in MATLAB. The model thus aims to provide a basis for future monitoring and prediction systems in the ocean.