Artificial neural network (ANN)-based simultaneous inversion of optically active ocean-colour variables from IRS-P4 OCM data

Abstract

An artificial neural network (ANN) based procedure is developed to estimate concentrations of Chlorophyll-a, Suspended Particulate Matter (SPM) and absorption due to chromophoric dissolved organic matter (CDOM) in the seawater from satellite detected normalized water-leaving radiance (nLw) of the IRS-P4 Ocean Colour Monitor (OCM) satellite data. An ocean colour reflectance model was used to generate surface spectral reflectance corresponding to first five bands of IRS-P4 OCM sensor, using three optically active oceanic water constituents as inputs. ANN model making use of reflectance in five visible bands was trained, tested and validated to invert the spectral reflectance for the simultaneous estimation of three optically active constituents. The retrieved chlorophyll-a concentrations from ANN based algorithm were compared with the corresponding chlorophyll-a distribution obtained by global empirical algorithms e.g. Ocean Chlorophyll-4 (OC4) algorithm. ANN derived chlorophyll-a estimates were found to have reduced the over estimation in coastal waters often observed with the use of band ratio algorithms.