A multivariate approach for statistical assessments of compound extremes

Global warming may affect the regime of hydroclimatic systems and induce more frequent occurrences of extremes, such as drought, heat wave and flood. Apart from the assessment of each extreme, recent decades have witnessed a surge in the study of compound extremes, i.e., the concurrence of multiple extremes. To aid the understanding of compound extremes, a variety of studies has been conducted to assess the dependence among different variables or extremes. As such, it is important to model multiple contributing variables of compound extremes to characterize the associated risk taking into account the dependence. In this study, a multivariate approach based on the meta-Gaussian model is proposed for the statistical analysis of compound extremes in the trivariate case. The application of the proposed approach is illustrated with the compound drought and hot extreme in the U.S. based on monthly precipitation, soil moisture and temperature from the North American Land Data Assimilation System (NLDAS-2). The likelihood of the occurrence of compound drought and hot extreme is assessed based on the joint distribution, which is shown to be higher in regions with significant land atmosphere interactions. The impact of precipitation and temperature on the occurrence of agricultural drought is also assessed based on the conditional distribution. Overall, results show that the proposed method provides a useful tool for statistical assessments of the compound extreme through constructing the joint and conditional distribution.