For effective and reliable management of water resources, the accurate forecasting of precipitation patterns is highly essential. The precipitation process is a complex hydrological component that is influenced by several hydro-meteorological variables, and it is varied from one region to another. In this study, the performance of coupled and standalone data intelligence models including wavelet artificial neural network (WANN), wavelet gene expressing programming (WGEP), artificial neural networks (ANN), and gene expressing programming (GEP) is undertaken for precipitation forecasting at four stations located in Iran (i.e., Ardabil, Khalkhal, Meshginshahr, and Parsabad). In this regard, monthly precipitation data are utilized from 1997 to 2016. The developed forecasting models are constructed using correlated lag time information. Two statistical performance metrics: coefficient of determination (R2) and root mean square errors (RMSE), are calculated for forecasting accuracy inspection. The obtained results indicated the capacity of the WANN coupled model as a superior forecasting model compared with the other data intelligence models. This was observed at the four investigated meteorological stations. However, the potential of the discrete wavelet transform has also demonstrated an enhancement in the predictability performance of the GEP model. This is indicating the capability of the pre-processing data time series as a prior stage for the forecasting process. Further, the results evidence the capability of the WANN coupled model in capturing the peak values of the precipitation. Overall, the applied coupled data intelligence model provided a significant contribution to the precipitation forecasting at this region contributing to the base knowledge of water resources engineering.
