Comparison of intelligent models to predict water level fluctuations of Zarival Lake using groundwater level

In recent decades, drought and lack of water resources management has caused many lakes and wetlands to be in critical conditions. Surface water level prediction is an important and complex hydrological process but it is required for better management and improvement of their ecosystem. In this research, four soft-computing techniques including wavelet artificial neural network (WANN), artificial neural network (ANN), adaptive-neuro-fuzzy inference system (ANFIS) and gene expression programming (GEP) were used to predict 2-month water level fluctuations of Zarivar Lake. The predicted water levels in each technique were compared with observed data and statistical indicators, RMSE, MAE and R2 were used to evaluate the performance of each method. The results proved that WANN performed considerably better and its prediction was more accurate. After WANN, the accuracy of ANFIS, GEO and ANN, respectively, were better and closer to observed data. The selected technique in this research can be recommended to predict the water levels in lakes and wetlands with enough accuracy.