摘要

Genetic algorithm (GA) is an effective approach in selecting the best factors without considering all possible combinations in landslide susceptibility mapping (LSM). The approach experienced a local optimal solution for hazard mapping. In this study, we propose a novel genetic algorithm (NGA) for solving the problems of optimal precision in selecting conditioning factors based on the crossover and mutation. In the southwestern part of China, including Wenchuan, Ludshan, and Ludian areas, the findings of this study confirm the applicability of NGA, which has a strong robustness compared to GA obviously. Results indicated that the highest area under curve (AUC) of GA is 93.47, 83.45, and 82.21% in Wenchuan, Lushan, and Ludian, respectively. Cumulative error of the precision (Delta R) is 3.19, 10.48, and 6.05%, and error of the highest precision (Delta P) is 0.01, 0.03, and 0.12% for Wenchuan, Lushan, and Ludian, respectively. Compared to the GA, the highest accuracy of NGA is 93.48% (Wenchuan), 83.48% (Lushan), and 82.28% (Ludian). It also revealed that Delta R is 0.77, 1.26, and 1.82%, and Delta P is 0.00, 0.04, and 0.05% for Wenchuan, Lushan, and Ludian, respectively. By comparing with GA, the novel approach of NGA has stronger robustness and higher accuracy on selecting the optimal conditioning factors of landslide. Additionally, the relationship of landslide occurrence with controlling factors was assessed in every study area. According to the results, lithology, distance to roads, elevation, and slope were regarded as the most effective factors for shallow translational landslides. These factors implied that internal structure and composition of rock, anthropogenic activity, and topography factors posed the main impacts on landslide occurrence. Finally, we implemented landslide susceptibility assessment in three study areas. Results showed that high landslide susceptibility was in the east and northeastern parts of Wenchuan; central region northward of Lushan; and southwest, central region, and west of Ludian.

  • 出版日期2017-5
  • 单位成都理工大学; 地质灾害防治与地质环境保护国家重点实验室