Deep Neural Networks for Water Quality Mapping

Water is essential to the ecosystem, it has a significant role. Many factors influence water quality, including industrialization, pollution, and climate change. In this study, three coastal water quality parameters were mapped using Deep Neural Networks (DNN) combining in-situ measurements and Landsat-8 data. Indeed, three parameters measurements of Chlorophyll-a (Chl-a), Suspended Solids (SS), and Turbidity concentration and their relevant multispectral Landsat reflectances were used to train Back Propagation (BP) DNN. This paper focuses on the choice of the best model architecture which gives the optimum number of hidden layers and nods number's achieving the minimum of Root Mean Square Error (RMSE). When applying to Algerian coasts, the DNN result showed a correlation coefficient (R) of 97.19
^%
with a 97.39 % of accuracy.