Application of Projection Pursuit Model to Landslide Risk Classification Assessment

It's a top priority to estimate risk of landslides for landslide disaster prevention and disaster reduction. In this paper, taking Xiaojiang ravine, a frequent debris flow occurring area as the study area, a classification assessment system of landslide risk is built by selecting 3 evaluation indexes including slope, cohesive force and internal friction angle based on the geographic information system support. Moreover, the projection pursuit technique is applied to evaluate landslide risk classification. According to linear projection of index factors of evaluation samples, the projection pursuit model to classification evaluation of landslide risk is built and the classification map of landslide risk zoning is drawn by the optimal projection direction of the projection characteristic value. The results show that the area ratio of extremely high risk zone, high risk zone, medium risk zone, low risk zone, extremely low risk zone in the study area is 14.28∶9.41∶69.12∶7.00∶0.19. A total of 60 landslides were verified in the study area according to the established evaluation index system of level 5. The high and extremly high risk zones cover 23.69% of the study area, but the actual number of soil landslides is 45, and the number of soil landslides in those risk zones is 75.00% of the total. The number of soil landslides in medium risk level and above risk zones is 96.67% of the total. The statistical results of landslide risk point density in the classification map of landslide risk zoning show that the density of landslides' volume increases rapidly with the increase of risk level. The evaluation results are relatively more consistent with actual situation and more reasonable compared with the measured results. The model proves to be a new perspective for landslide risk prediction.