关键词: 滑坡易发性分区;机器学习;混合分类模型;空间分析;延安市;陕西省;

Abstract: For landslide prevention and control, it is of great practical significance to study the appropriate method of landslide susceptibility zoning in the county area. In view of this, based on 717 landslide samples collected from Wuqi County, Yan’an City, Shaanxi Province, the slope, aspect, elevation, plane curvature, profile curvature, average annual rainfall, distance from road, distance from river, rock and soil mass type and NDVI were used as an impact factors, and their corresponding entropy indices were calculated to construct a modeling dataset based on entropy indices. Subsequently, based on the modeling dataset, coupled index of entropy (IOE) and logistic regression tree model (LMT), an IOE-LMT hybrid classification model was established to draw a zonal map of landslide susceptibility in Wuqi County. A variety of statistical metrics, area under the ROC curve (AUROC) and mean absolute error (MAE) were used to evaluate the partition accuracy and the generalization performance of the model. The results showed that the generalization performance of the IOE-LMT model was strong (AUROC=0.942), and the accuracy of the landslide susceptibility zoning was high. Landslide in the study area was prone to happen in the loess gullies, and the landslide susceptibility in the north of the study area was significantly higher than that in the south. The evaluation results are reasonable and reliable, and can provide reference for local landslide prevention and land space planning.

Key words: landslide susceptibility zoning; machine learning; mixed classification model; spatial analysis; Yan 'an City; Shaanxi Province