Study on drought events in China based on time-varying nested Archimedean-copula function

Abstract

Abstract Drought forecasting, which can enable contingency actions to be implemented in advance of a drought, plays a significant role in reducing the risks and impacts of drought. In this study, a simulation framework of the occurrence probability of drought events based on a nested copula function and Gibbs sampling is proposed to effectively compensate for the high-dimensional problems and lack of initial data in traditional methods. Precipitation data of 718 meteorological stations from 1961 to 2018 in China was analyzed. The results showed that the occurrence location of drought events was mainly concentrated from 35° to 42° north latitude and 105° to 120° east longitude, with the occurrence time mainly concentrated from September to November. The Archimedean-copula function, constructed based on latitude, longitude, and occurrence time, could precisely determine the spatiotemporal joint probability distribution of drought events (RMSE: 0.01). The optimal time-varying nested Archimedean-copula functions were obtained from February to May (spring), June to September (summer) and October to January (autumn and winter). Compared with the nested Archimedean-copula function, the accuracy of Gibbs sampling and simulation based on the time-varying nested Archimedean-copula function was increased by 84.05% latitude and 69.76% longitude. The results provide an effective means for scientific drought forecasting, and water resource management departments can take preventive measures at an early stage.