Landslide risk assessment of high-mountain settlements using Gaussian process classification combined with improved weight-based generalized objective function. (January 2022)

• Record Type: Journal Article
• Authors: Gao, Zemin; Ding, Mingtao; Huang, Tao; Liu, Xingwang; Hao, Zheng; Hu, Xiewen; Chuanjie, Xi
• Is Part Of: International journal of disaster risk reduction. Volume 67(2022)
• Publication Date: 2022-01
• Subjects: The high-mountain landslide -- Risk assessment -- Gaussian process classification -- Improved weight-based generalized objective function -- Southwest China; Emergency management -- Periodicals; Risk management -- Periodicals; Disaster relief -- Periodicals; Hazard mitigation -- Periodicals; 363.34
• DOI: 10.1016/j.ijdrr.2021.102662
• Languages: English
• ISSNs: 2212-4209

Abstract

Abstract: High-mountain landslides in the hydrogeological environment expose the residents and properties of mountain settlements to immediate risks, especially in the southwestern part of the Tibetan Plateau, where severe damage is frequent. The risk assessment of geohazards is provided as an effective disaster prevention and mitigation management measure. To reasonably predict the spatial distribution characteristics of high-mountain landslide risks, the Gaussian process classification (GPC) and the improved weight-based generalized objective function were applied to quantify the hazard, vulnerability, and risk values of Li County, Southwest China. The findings revealed that the hazard, vulnerability, and risk levels in the studied areas were comparatively high, with percentages in the moderate–high ranges of 41.715%, 40.634%, and 57.135%, respectively. The towns of Zagunao, Ganbu, and Taoping are among the regions with a high grade of risk, with risk values of 0.407, 0.347, and 0.335, respectively. These high-risk distribution sites are concentrated predominantly in the municipalities and their fringes, whereas other prefectures are in the medium–below risk level range. Following the examination of the GPC model and integral appraisal performance, this risk assessment framework exhibits good predictability, with an accurate proportion of predictions in the hazardous location interval reaching 99.857% and an ROC-AUC value of 0.97. The proceedings of this study can provide essential materials and references for decision-making for hydrogeological-environmental hazard policy formulation, territorial spatial planning, and other regional-related risk assessment projects. Highlights: Improvement of the weights of the generalized objective function by introducing the entropy and the information method. The integration of machine learning algorithms and improved weight-based generalized objective function. Completing the risk, vulnerability, and hazard mapping of high-mountain landslides under geo-environmental conditions.