Analysis of Landslide Susceptibility and Accuracy in Different Combination Models

This paper carried out the studies in Bailong river basin,where landslide incidences frequently occurred, to explore the applicability and the rationality of the landslide susceptibility evaluation model.Six indicators of slope, top ographic relief, distance from fault, stratum lithology, watershed gully density and vegetation normalization index were selected assusceptibility evaluation indication. The authors took 2 093 landslide disaster points in the study area as sample data and determined the factor classification status by using the information value, deterministic coefficient value and the mutation law of evidence weight value curve, combined with the frequency ratio curve of the landslide area and the graded area as the criticality of the grading division value. Based on the state grade of each index and correlation analysis, three combination models with logistic regression, of information value method, deterministic coefficient method and evidence weight method were used to evaluate the susceptibility of regional landslide hazard.The three combination models were compared and discussed from the aspects of the results, the applicability and the accuracy.Results show that in the evaluation of regional landslide susceptibility, all the three models are ideal. The prediction accuracy of the combination model of information value method with logistic regressionis better than the other two combination models, with a prediction accuracy of94.6%. The results provide reference for geological disaster prevention and land use in this area.