Drought Simulation using Two CEEMD-GPR and GPR-GARCH Integrated Models (Case Study: Northwest of Iran)

Drought is one of the most important natural disasters affecting agriculture section and water resources. Droughts often occur in arid and semi-arid regions. Therefore, drought forecasting is necessary and plays an important role in the planning and management of water resources. So far, numerous drought prediction methods have been proposed in the literature, including time series models, regression models, probabilistic models, machine learning models, physical models, and a host of hybrid models. Although all of these methods have shown promising results in terms of improving accuracy of drought forecasts, the impact of climate change on droughts has highlighted the need for more advanced methods for predicting this event. Engle (1982)  proposed the ARCH model which can depict the variance of the time series and eliminate the heteroskedasticity caused by the constant time series variance. The GARCH model was further developed based on the ARCH model, the advantage of which is that it can use a simpler form to represent a high-order ARCH model. On the other hand, in recent years, the Meta model approaches have been applied in investigating the hydraulic and hydrologic complex phenomena. Hybrid models involving signal decomposition have also been found to be effective in improving prediction accuracy of time series prediction methods (Amirat et al., 2018). Complementary Ensemble Empirical Mode Decomposition analysis is one of the widely-used signal decomposition methods for hydrological time series prediction. Decomposition of time series reduces the difficulty of forecasting, thereby improving forecasting accuracy. Due to the complexity of the drought phenomenon and the effect of various parameters on its prediction, in this study, the capability of GPR as a kernel-based approach and also integrated CEEMS-GPR and GPR-GARCH models were assessed for drought modeling based on six-month SPI index for the three cities of Tabriz, Urmia, and Ardabil in Iran during the period 1978-2017. In fact, this study attempts to create a novel method by combining the CEEMD and GARCH models with the GPR to enhance the estimation accuracy of the six- month SPI drought index.