A Neural Network Model for Estimating Sea Surface Chlorophyll and Sediments from Thematic Mapper Imagery

Abstract

Two important parameters used for monitoring coastal water quality are the concentrations of chlorophyll and suspended sediment in surface waters. Ocean color remote sensing provides a convenient method of determining these concentrations from upwelling radiances. In the open ocean, it is not difficult to derive empirical algorithms relating the received radiances to surface concentrations of chlorophyll. In turbid coastal waters, however, this is much more difficult due to the presence of high concentrations of suspended sediments and dissolved organic material, which overwhelm the spectral signal of chlorophyll. Neural networks have been proven successful in modeling a variety of geophysical transfer functions. Here, a neural network is employed to model the transfer function between the chlorophyll and sediment concentrations and the satellite-received radiances. It was found that a neural network with two hidden nodes, using the three visible Landsat Thematic Mapper bands as inputs, was able to model the transfer function to a much higher accuracy than multiple regression analysis. The RMS errors for the neural network were <10%, while the errors for regression analysis were >25%.