The effect of wavelet-based denoising in the prediction water level of Urmia lake

Investigation of water level fluctuations in lakes has been received more national and international attention in recent years due to the importance of lake conservation and their significant role as the central water bodies of natural heritance. So, the prediction of wat er levels in Lake Urmia is highly necessarydue to a few meters’ declines of water level(WL)in the last decade and also to prevent possible natural hazards in the future. In this study. Support Vector Machine(SVM) and Artificial Neural Network(ANN) methods were used to predict the water level of Lake Urmia. Then, using the wavelet denoising method,the input data were decoded and compared with the two methods mentioned and compared with noisy models. To simulate and predict the water level of the Urmia water level data during the period 1350 to 1391, using three measuring Vanyar, Simineh Rood, and Zarrineh Rood stations. The results showed theadequacy of both models with a slight advantage over the ANN model with the SVM model. In the model of ANN-Wavelet de-noising and SVM-Wavelet de-noising, the accuracy of the model was reduced compared to the noisy models. The magnitude of this reduction in the support vector machine model ishigher than that of the ANN model.

Keywords:

Support Vector Machine , Artificial Neural Network , Wavelet de-noising , Urmia Lake , Water Level.