Modelling daily water level fluctuations of Lake Van (Eastern Turkey) using Artificial Neural Networks

Abstract

In this paper, two different Artificial Neural Network (ANN) techniques, namely the Feed-Forward Neural Networks (FFNNs) and the Radial Basis Function Neural Networks (RBFNNs), are used for the prediction of the daily water level of Lake Van. The water level fluctuations of Lake Van have been changing throughout history due to global warming processes. The estimated daily water levels using ANNs and the corresponding observed values correlate well. The results of the models are compared using Mean Squared Error (MSE) and Determination Coefficient (R-2) statistics. Comparison of the results has shown that the FFNN model performs better than the RBFNN model in the predictions. The forecasting results have indicated that the water level fluctuations of Lake Van have a long-term decreasing trend in the future with the water level reduced by as much as 15 cm compared with the lowest lake level of the 2011 water year.