Modelling level change in lakes using neuro-fuzzy and artificial neural networks

SummaryAccurate estimation of level change in lakes and reservoirs in response to climatic variations is an important step for the development of sustainable water usage policies, particularly for complex hydrological systems such as Lake Beysehir, Turkey. In this study, level changes of Lake Beysehir were estimated using adaptive neuro-fuzzy inference system (ANFIS), artificial neural networks (ANN) and a seasonal autoregressive integrated moving average (SARIMA). The ANN and ANFIS models were first trained based on observed data between 1966 and 1984, and then used to predict water level changes over the test period extending from 1985 to 1990. The performances of the different models were evaluated by comparing the corresponding values of mean squared errors (MSE) and decisive coefficients (R2). While all models produced acceptable results, the minimum MSE value (0.0057) and the maximum R2 value (0.7930) were obtained with ANFIS model, followed by the three-layered artificial neural network model (ANN1). The lowest performance was observed with the SARIMA model.