Role of sensor noise in hyperspectral remote sensing of natural waters: Application to retrieval of phytoplankton pigments

An algorithm is derived to retrieve the concentration of optically active materials, e.g., phytoplankton pigments, etc., from remotely measured spectra of up welled oceanic light. The algorithm takes into account sensor noise in deriving equations for the best linear estimate of concentration mean and residual variance. The algorithm is applied to the problem of phytoplankton concentration retrieval using a modeled hyperspectral sensor based roughly on the LASH imager. The algorithm requires knowing the joint distribution of radiance spectra and concentration. This joint distribution is obtained by simulation using ocean radiance models. It is shown that sensor noise (both shot and dark current) markedly decreases the accuracy of concentration retrieval. However, accuracy is greatly improved if a priori information about observation conditions is known and included in the algorithm. Thus accounting for sensor noise improves retrieval accuracy and affects the choice of observation method.