Comparison of Different Data-Driven Approaches for Modeling Lake Level Fluctuations: The Case of Manyas and Tuz Lakes (Turkey)

Abstract

Modeling lake level fluctuation is very essential for planning and design of hydraulic structures along the lake coasts. In this study, namely two different adaptive-neuro-fuzzy inference system (ANFIS) including ANFIS with grid partition (ANFIS-GP) and ANFIS with subtractive clustering (ANFIS-SC), and gene expression programming (GEP) were applied to forecast 1-, 2- and 3-month ahead lake-level fluctuations of Manyas and Tuz, Turkey. Comparison of the models indicated that the optimal ANFIS-GP models performed better than the optimal ANFIS-SC and GEP models in forecasting 1- and 3-month ahead lake levels while the ANFIS-SC model showed better accuracy than the other models in 2-month ahead forecasting. The ANFIS-GP model comprising lake level values of current and one previous months successfully forecasted 1-month ahead lake level with root mean square error (RMSE) of 0.251 and coefficient of determination (R2) of 0.872. For the Tuz Lake, the optimal ANFIS-SC models were found to be better than the optimal ANFIS-GP and GEP models for forecasting 1- and 2-month ahead lake levels while the GEP model performed better than the other models in and 3-month ahead lake level forecasting. The ANFIS-SC model comprising lake level values of current and three previous months successfully forecasted 1-month ahead lake level with RMSE of 0.120 and R2 of 0.724. Based on the comparisons, it was found that the GEP, ANFIS-GP and ANFIS-SC models could be successfully employed in forecasting lake level fluctuations.