SPI-based drought simulation and prediction using ARMA-GARCH model

Drought is one of the most frequent climate-related disasters occurring in North China Plain. The accurate and timely information of drought is vital for crop production and food security. In this study, the monthly precipitation data during 1965–2015 was used to calculate the Standardized Precipitation Index (SPI) with a time scale of 9 months (SPI-9) at five stations in Shandong Province, North China Plain. The GARCH (Generalized Autoregressive Conditional Heteroskedasticity) model was adopted to eliminate the heteroskedasticity (ARCH effect) in the residuals of ARMA (Autoregressive and Moving Average) model, and the two models were combined into a composite model called ARMA-GARCH model. Both ARMA and ARMA-GARCH models were used to simulate SPI-9 drought index, and the results of comparison between two models showed that the ARMA-GARCH model performed better. Furthermore, the two models were used to predict SPI-9, the result showed that the accuracy of the ARMA-GARCH model is much higher than that of the ARMA model; for maintaining the stability of site-to-site correlation, the ARMA-GARCH model also outperformed the ARMA model. The research indicates that the ARMA-GARCH model could be used to more accurately simulate and predict SPI-9 drought index.