Prediction ofWater Sodium Absorption Ratio (SAR) using ANN andWavelet Conjunction Model (Case Study: Rudbar Station of Sefidrud River)

One of the most important factors of sustainable development of watersheds is qualitative and quantitative availability of suitable water resources. In this study, the artificial neural network (ANN), multi - variable linear regression (MLR), genetic programming (GP) and hybrid wavelet ANN (WANN) models were considered for modeling the monthly sodium absorption ratio (SAR) in Sefidrud River - Roudbar Station, and the effect of data preprocessing on model performance was investigated using the discrete wavelet Transform method. For this purpose, observed time series of river discharge and SAR were decomposed into several sub - time series at different scales by discrete wavelet transform. Then these sub - time series were introduced as inputs to the ANN model. The results showed that the hybrid wavelet - neural network model was more suitable for predicting maximum SAR values than the MLR, ANN, and GP models. Furthermore, the hybrid model could simulate the hysteresis phenomenon for SAR modeling rigorously, while multi linear regression method was incapable of detecting it.