Water Quality Monitoring in Large Reservoirs Using Remote Sensing and Neural Networks

Abstract

Abstract-- Water quality monitoring in lakes and reservoirs using water samples and laboratorial analysis is expensive and time consuming. The use of artificial neural networks to predict water quality using satellite images shows great potential to make this process faster and at lower costs. This article discusses an indirect method to estimate the concentration of pigments (chlorophyll-a), an optically active parameter in water quality. A model based on artificial neural networks, using radial base functions architecture, was developed to predict Tucurui's Reservoir chlorophyll-a concentrations. As input to the neural networks spectral information from Landsat imagery was used, while pigment concentration were used as output information. To train and validate the model we used data from the years 1987, 1988, 1995, 1999, 2000 and 2004. The tested model showed a correlation coefficient of 0.92 for the estimation of pigment (chlorophyll-a) concentrations, indicating its applicability to predict this water quality parameter.