A multiple-drought cascading framework based on causal inference

Abstract

Comprehensively evaluating evolution of drought has an important guiding significance to drought early warning systems and disaster control. Drought is a major natural disaster. Although there are multiple methods including statistics to study drought propagation at present, the evolution of droughts is a complex phenomenon, existing physical and data-driven methods demonstrate unsatisfactory estimation performance. To address this issue, a new type of multiple-drought cascading framework was proposed. The peculiarity of this framework is that based on the principle of causal inference, a generalized linear regression model was established to quantify cascading effect of multiple droughts in disaster chain systems and explore influences of complicated factors outside the system on the cascading effect. The proposed framework was evaluated based on meteorological, soil and groundwater drought conditions of the Yanhe River basin, China. The results prove that study from the perspective of cascading disasters can intuitively characterize disasters of different types of drought signal propagation in the disaster chain system. The results also revealed a top-to-down drought long-chain cascading relationship in moisture transfer in the atmosphere, pedosphere, and lithosphere and discovered that climate change and vegetation restoration are the primary influencing factors of different links. In the framework, a novel model was constructed to quantify multiple-drought cascading effect considering influences of complex factors to facilitate the enhancement of policy planning and can help reduce negative influences of multiple cascading disasters. This framework is also expected to be applied in other areas to enrich the understanding of mechanisms underlying multiple-drought cascading disasters.