Susceptibility mapping and zoning of highway landslide disasters in China

• Published in: PloS one Vol. 15; no. 9; p. e0235780
• Main Authors: Yin, Chao, Li, Haoran, Che, Fa, Li, Ying, Hu, Zhinan, Liu, Dong
• Format: Journal Article
• Language: English
• Published: San Francisco Public Library of Science 04.09.2020; Public Library of Science (PLoS)
• Subjects: Biology and Life Sciences; Cartography; China; Computer and Information Sciences; Decision trees; Disaster management; Disasters; Earth Sciences; Earthquakes; Emergency preparedness; Genetic algorithms; Highway planning; Laboratories; Landslides; Landslides & mudslides; Learning algorithms; Location; Machine learning; Mapping; Methods; Mitigation; Optimization; Particle swarm optimization; People and Places; Physical Sciences; Principal components analysis; Research and Analysis Methods; Roads & highways; Seismic engineering; Social Sciences; Statistical methods; Statistics; Support vector machines; Susceptibility; Swarm intelligence; Traffic control; Transportation services; Zoning; China
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0235780

Prominent regional differentiations of highway landslide disasters (HLDs) bring great difficulties in highway planning, designing and disaster mitigation, therefore, a comprehensive understanding of HLDs from the spatial perspective is a basis for reducing damages. Statistical prediction methods and machine learning methods have some defects in landslide susceptibility mapping (LSM), meanwhile, hybrid methods have been developed by combining the statistical prediction methods with machine learning methods in recent years, and some of them were reported to perform better than conventional methods. In view of this, the principal component analysis (PCA) method was used to extract the susceptibility evaluation indexes of HLDs; the particle swarm optimization-support vector machine (PSO-SVM) model and genetic algorithm-support vector machine (GA-SVM) model were implemented to the susceptibility mapping and zoning of HLDs in China. The research results show that the accumulative contribution rate of the four principal components is 92.050%; evaluation results of the PSO-SVM model are better than those of the GA-SVM model; micro dangerous areas, moderate dangerous areas, severe dangerous areas and extreme dangerous areas account for 24.24%, 19.49%, 36.53% and 19.74% of the total areas of China; among the 1543 disaster points in the HLDs inventory, there are 134, 182, 421 and 806 located in the above areas respectively.