Vulnerability assessment of disaster chains: A case study of rainstorm–landslide disaster chains in the Greater Bay Area

Abstract

Considering the triggering and superposition effects of disaster chains and the different characteristics of each dimension of vulnerability assessment, a disaster chain vulnerability assessment model was established and constructed from three aspects, namely, the sensitivity of the disaster-prone environment, the exposure of disaster-bearing bodies, and the adaptability of the disaster chain. The sensitivity was calculated based on a convolutional neural network (CNN) and a coupling model of a parameter optimal geographical detector and analytic hierarchy process (OPGD-AHP). The exposure degree was evaluated via the sequential relationships-TOPSIS method. The adaptability of the proposed method was assessed according to the entropy weighting-TOPSIS approach. In addition, ArcGIS technology was used for vulnerability assessment and vulnerability zonation. The rainstorm–landslide disaster chain in the Guangdong–Hong Kong–Macao Greater Bay Area (GBA) was employed as an example for empirical analysis. The results revealed that the rainstorm–landslide disaster chain in the GBA exhibited the following distribution characteristics: high- and secondary high-vulnerability areas (the vulnerability index was between 0.776 and 1.000) in the northwestern, northeastern, and southeastern regions and low- and secondary low-vulnerability areas (the vulnerability index was less than 0.726) in the central, south-central, and eastern regions. With gradual transmission, the disaster chain has the phenomenon of increasing loss and decreasing probability, that is, the phenomenon of "small probability, enormous risk". There were not only overlapping relationships between single disaster types and high-vulnerability areas in the disaster chain but also triggering and synergistic effects. There is also a clear synergistic effect among the low-vulnerability areas. The results of the disaster chain assessment are more consistent with the actual situation than the results of single-disaster models are.