Drought forecasting with regionalization of climate variables and generalized linear model

Abstract

Spring drought forecasting is essential in South Korea for managing water resources reliably and cultivating agricultural products efficiently, as seasonal rainfall difference often drives water shortage during spring. In the current study, a novel scheme for spring drought forecasting was suggested by extensively searching appropriate predictors from the global climate variable: here mean sea level pressure (MSLP) of the winter season due to its time lag for forecasting. The target series was estimated with the median of the spring precipitation series of the weather stations over South Korea, called the accumulated spring precipitation (ASP). A number of points of the MSLP data were detected as significant cross-correlation with the ASP and also the points were regionally grouped. Therefore, the regionalization for the high correlation points was performed, resulting in three regions, such as Arctic Ocean (R1), South Pacific (R2), and South Africa (R3). The R1 and R2 regions are located at the places where climate indices have been developed such as Arctic Oscillation and North Atlantic Oscillation for R1 and the indicator of El-Nino and Southern Oscillation for R2. The generalized linear model (GLM) was adopted in ASP drought forecasting with the driven three regionalized indices as the predictors of the ASP. The result indicates that the regionalized indices can produce a good performance in forecasting the ASP. The forecasting result can be employed as a good tool for managing water resources and planning better cultivation in agriculture industries.