Urmia lake water level fluctuation hydro informatics modeling using support vector machine and conjunction of wavelet and neural network

In this research, the simulation of Urmia Lake water level fluctuation by means of two models was applied. For this, Support Vector Machines (SVM), and Neural Wavelet Network (NWN) models that conjugated both the wavelet function and ANN, developed for simulating the Urmia Lake water level fluctuation. The yearly data of rainfall, temperature and discharge to the Urmia Lake and water level fluctuation were used. Urmia Lake is the biggest and the hyper saline lake in Iran. The outcome of the SVM based models are compared with the NWN. The results of SVM model performs better than NWN and offered a practical solution to the problem of water level fluctuation predictions. Analysis results showed that the optimal situation occurred with use of precipitation, temperature and discharge for all station and water level fluctuations at the lag time of one year (RMSEs) of 0.23, 0.41 m obtained by SVM, NWN, respectively, and SSEs of 0.43, 1.33 and R 2 of 0.97, 0 obtained by SVM, NWN, respectively. The results of SVM model show better accuracy in comparison with the NWN model.