Received:2022/04/22

Revised:2022/06/18

Published:2024/05/15

Susceptibility evaluation of geological hazards in alpine canyon regions: A case study of Nujiang Prefecture

FENG Xianjie1,  LI Yimin2,3,  DENG Xuanlun2,  ZHAO Juanzhen1,  YANG Yiming2

1. Institute of International Rivers and Eco-Security, Yunnan University, Kunming 650500, Yunnan, China; 2. School of Earth Sciences, Yunnan University, Kunming 650500, Yunnan, China; 3. Research Center of Domestic High-Resolution Satellite Remote Sensing Geological Engineering, Kunming 650500, Yunnan, China)

Abstract: Geological hazards frequently occur in alpine canyon regions. Objectives This study aims  to investigate the spatial distribution of geological hazard susceptibility. Methods Nujiang Prefecture was selected as the study area. By integrating geological conditions, meteorological and hydrological factors, vegetation cover, and other variables, 12 low-collinearity evaluation factors, including elevation, slope, aspect, curvature, and relief were identified to construct a regional geological hazard susceptibility evaluation index system. Geological hazard susceptibility was assessed at the grid-cell level using three models: the information value (IV) model, the information value-back propagation neural network (IV-BPNN) coupled model, and the information value-support vector machine (IV-SVM) coupled model. Results (1) The susceptibility results were validated using actual geological hazard points, showing good spatial distribution consistency between the hazard points and the three susceptibility results. (2) The susceptibility index is categorized into four levels: low, moderate, high, and extremely high susceptibility areas. The corresponding area proportions of high and extremely high susceptibility classes for the IV model, IV-BPNN model, and IV-SVM model are 37.12%, 32.36%, and 23.08%, respectively. High and extremely high susceptibility areas exhibit a linear distribution, mainly concentrated along river systems such as the Nujiang, Lancang, and Dulong Rivers, near roads, and in areas with active geological structures. (3) The areas under the curve (AUC) of the receiver operating characteristic (ROC) curves for the IV model, IV-BPNN model, and IV-SVM model are 0.884, 0.889, and 0.901, respectively. Conclusions All three geological hazard susceptibility evaluation models demonstrate high prediction accuracy, among which the IV-SVM model shows the highest accuracy and reliable zoning results, offering valuable guidance for local governments in formulating geological hazard prevention and control measures.

Key words: alpine and canyon areas; geological disasters; information model; BP neural network; support vector machine; susceptibility evaluation; Nujiang Prefecture