摘要
以贵州省纳雍县为研究区,基于地理信息系统(GIS)平台,采用Logistic回归模型评价区域滑坡易发性。选取高程、坡度、坡向、地表曲率、岩性、到公路距离、到水系距离和到断层距离8项影响因子,应用Logistic回归模型拟合描述滑坡发生与各项因子关系的回归方程。在此基础上,预测研究区内滑坡易发性的概率值,并将研究区划分为5个易发级别。结合历史滑坡分布,对滑坡易发性分区的合理性进行验证,结果表明滑坡易发性分区图和实际滑坡分布情况吻合较好。此外,应用受试者工作特征曲线对模型结果进行评价,结果显示,曲线下面积AUC值为0.774,Logistic回归模型的预测精度较好。
Based on GIS platform, the landslide susceptibil Province was analyzed by using the logistic regression mo condition parameters including elevation, slope gradient, distances to roads, distance to drainage and distance to fa ity evaluation of Nayong County in Guizhoh del in this paper. A total of eight landslide slope aspect, surface curvature, lithology, ults were considered. The regression equa- tion of the relationship between the landslide and these parameters were analyzed by using the logistic regression model. On this basis, a landslide susceptibility map was produced. The study area was classified into five zones of landslide susceptibility levels. Past records of landslide locations were used to validate the landslide susceptibility map, and it shows a good matching with the identified land- slides. Additionally, the receiver operating characteristics （ROC） curve was drawn as a performance indicator for validation. The area under curve （AUC） value was 77.4%. It reveals that the logistic re- gression model has good prediction accuracy.
出处
《交通科技》
2012年第4期69-72,共4页
Transportation Science & Technology
