Monthly precipitation prediction improving using the integrated model based on kernel-wavelet and complementary ensemble empirical mode decomposition

Precipitation is one of the most important components of water cycle and plays an important role in assessing the climatic characteristics of each region. Estimates of monthly rainfall are important for various purposes such as flood estimation, drought, irrigation planning, and river basin management. In the present study, the monthly rainfall of Tabriz station was investigated using the intelligent Gaussian Process Regression (GPR) method based on Complementary Ensemble Empirical Mode Decomposition (CEEMD) and Wavelet Transform (WT). In this regard, different models were defined based on teleconnection patterns and climatic elements, and the impact of different input parameters was assessed. The obtained results from the models proved the high capability and efficiency of the applied method in predicting the monthly precipitation. It was observed that in prediction of the monthly precipitation, NAO, Nino3, MEI and climatic elements including mean monthly temperature and relative humidity, as well as precipitation related to the previous months, are effective in prediction and improve the models accuracy. The results showed that time series decomposition based on wavelet transformation led to more accurate outcomes compared to the complementary ensemble empirical mode decomposition. The best evaluation of test series using wavelet transform decomposition was obtained for the state of modeling based on teleconnection patterns and climatic elements with the values of DC=0.889, R=0.961 and RMSE=0.036. Also, based on the sensitivity analysis, Pt-3 was found to be the most effective parameter in modeling.