Investigating the effects of climate variables on river flow estimation using single and hybrid-wavelet methods of soft computing

Climate change and its impact on the status of water resources can endanger the various aspects of life and human life on Earth. In this study, single and hybrid-wavelet artificial neural network, Adaptive neuro-fuzzy inference system and gene expression programming were used modeling flow parameter. For this purpose, monthly climatic data with 21-year (1996-2016) statistical period of flow, temperature and precipitation of Tapik station in the Nazluchay River of West Azerbaijan province has been used. In this study, the effects of delayed flow parameters, precipitation, temperature and periodic effect (monthly coefficient) in models have been investigated. The results show the superior performance of wavelet hybrid models compared to single models of soft computing and the positive effect of applying periodic effects on river flow modeling. Also, wavelet transformation by analyzing data and separating the noise enables the ability to upgrade the performance of hybrid models as compared to single models. For the optimal model (i.e. hybrid wavelet-gene expression programing model, the values of correlation coefficient and root mean square error indices were obtained as 0.98 (maximum) and 326.2 (m3/s) (minimum), respectively.