Sea water chlorophyll-a estimation using hyperspectral images and supervised Artificial Neural Network

Abstract

The use of satellite hyperspectral images has improved the extraction of information compared to multispectral images. Although designed as a technical demonstration for land applications, Hyperion satellite hyperspectral images are used to estimate sea water parameters in the coastal area. A combination of turbid river inputs, as well as the open sea flushing, determines the quality of the sea water in the coastal area and the status of its environment. In addition, the existence of different source of pollution adds to the complexity of the coastal sea water analysis. The field campaigns to retrieve sea water parameters provided by the past completed projects were coincident with acquisition of the Hyperion image covering the pilot area. A robust method based on a supervised Feed-Forward Back-Propagation Artificial Neural Network (ANN-BP) algorithm is applied to retrieve the concentration of chlorophyll-a from hyperspectral image. In addition, Hyperion images are used to show the variation of chlorophyll-a during two different periods of time. The variation is due to many manmade environmental disasters such as oil spill and continuous discharge of chemical and solid wastes. The research proves that the new method based on ANN has improved the mathematical regression methods to a coefficient of determination almost equal 1 compared to about 0.4 for the methods not based on ANN-BP.