| Regional precipitation-induced landslide is one of the major geologic hazards in our country, which features large areas of damage, simultaneity, sudden breakout and serious consequences. For quite some time, close attentions have been paid to the forecast and early warning of single landslide in our country while the research of regional early warning proceeds quite slowly. It is essential to carry out the research into the forecast and early warning of precipitation-induced landslide and establish a scientific, complete and practical early warning and forecasting approach at the beginning of the national campaign launched for the geologic hazard weather forecast and early warning. Yucheng District of Ya'an City, the so-called rain city of Sichuan province is chosen as the research target in the paper. Based on the rich practical data, the paper makes the in-depth research into various factors affecting the occurrence of landslide, conducts an overall analysis on the relationship between the time-space distribution of landslide and rainfall, and explores the method to determine the threshold of precipitation inducing the regional landslide. It also preliminarily establishes the precipitation-induced landslide forecast and early warning system. Based on the inverse modeling carried out in the research area in 2004, all approaches are practical, feasible and can provide scientific basis for the active disaster mitigation and prevention. The key researches, approaches and conclusions are as follows: (1)The paper proposes the CF logic regression model to address the consolidation of different data layers during the landslide critical evaluation through GIS-based statistical analysis model and the determination of data layer superposition weight. By combining the certainty coefficient (CF) and logic regression model (Binary Logistic) and with the assistance of SPSS statistical analysis software, the factor selection, heterogeneity data consolidation and the data layer superposition weight are solved and the precision and efficiency of landslide susceptivity zoning are improved. Sensitivity analysis on the landslide impact factor in the research area is carried out by using the landslide certainty coefficient; Key factors affecting the landslide in the research area are determined by the consolidated inspection on the certainty coefficient. Logic statistical analysis predictive model is established, and through homogeneous meshing, the space quantitative evaluation on landslide susceptivity is realized. Research shows that landslide is mostly likely to happen at the petrofabric with interlacing softness in Yucheng district, dip slope of 10°～40°facing the east and southwest as well as the forest with elevation between 600～1000 m. Key factors affecting the landslide risk include the aspect of slope, elevation, petrofabric, type of slope and inclination. Based on the calculated cell landslide probability of occurrence, the landslide risk in Yucheng district is divided into 5 grades, namely extremely unstable, unstable, latent unstable, basically stable and stable. Inspection is carried out on the new landslide sites of Yucheng district. The result shows that the extremely unstable area and unstable area contain 86% of the verified landslide, indicating that the evaluation result is in reasonable agreement with the actual landslide. (2) Logic regression model is introduced into the precipitation-induced landslide early warning and forecast, based on which, the precipitation-induced regional landslide threshold expression and the probability of rainfall's impact on the landslide is established. Quantitative investigation has been conducted on the expressed rainfall precipitation for the rainfall threshold inducing the regional landslide based on the logical regression coefficient. Expression considering both the rainfall intensity and rainfall threshold during the rainfall is established. Research indicates that the occurrence of landslide in Yucheng district is relevant to the rainfall during the 4 days, i.e.
|
PDF Full Text Request