Modeling of Daily Water Level Fluctuations of the Urmia Lake Using Extreme Learning Machine Model

In recent decades, the water level of the Urmia Lake has decreased due to over utilization of surface and groundwater resources, prevention of surface water resources discharges to the Urmia Lake and also climate changes, which caused water and environmental crisis in the region. Therefore, modeling Lake level fluctuations is essential for its water resources planning and management. The aim of the study is to forecast the Urmia Lake water level fluctuations for one, three and seven days ahead using extreme learning machine (ELM). Also, the artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) models used to compare the ability of ELM. For this purpose, the daily water level (Dec 1965- Nov 2012) of the lake was used. To develop the models, the current day and six days water level lags as the inputs used for forecasting the one [h(t)], three [h(t)] and seven [h(t)] days ahead water levels. The data sets were divided into two subsets training/validation (85%) and testing (15%) and after modeling, the performance of the models were evaluated based on coefficient of determination (R2), root mean square errors (RMSE) and Nash–Sutcliffe coefficient (NSC). The results showed that the ELM model for one-step-ahead water level modeling with R2 = 0.9995, RMSE = 0.0151 m and NSC = 0.9995 respectively, outperformed in comparison with ANN and ANFIS models. Also, it was observed that ELM models learned faster than the other models during model development trials whereas the ANFIS models had the highest computation time.