三峡库区万州区滑坡灾害易发性评价研究

 滑坡灾害易发性研究在滑坡灾害风险管理与城市规划等方面具有非常重要的现实意义。以往的研究中，鲜有对指标因子状态划分作有关深入分析和讨论的。鉴于此，以滑坡灾害频发的三峡库区万州区为研究对象：首先，选取影响滑坡发生的7个致灾因子(地层岩性、地质构造、水系分布、坡度、坡向、坡体结构及土地利用)作为滑坡易发性的评价指标，依据各指标条件下滑坡累计发生频率曲线斜率的变化，并结合滑坡面积比和分级面积比曲线对指标因子的状态进行分级；其次，根据全区655个历史滑坡数据，分别运用信息量模型和逻辑回归模型建立各自的滑坡易发性评价体系；再则，采用快速聚类法(K-means cluster)对以上2种方法所得到的易发性结果进行分级，并基于GIS平台，得到全区滑坡易发性区划图；最后，从模型结果、精度、适用条件等方面对2个模型进行讨论和比较，研究结果表明：信息量模型和逻辑回归模型的预测精度分别为73.0%和54.9%，前者预测能力要优于后者。

Evaluation of landslide susceptibility for Wanzhou district of Three Gorges Reservoir

Evaluation of landslide susceptibility for Wanzhou district of Evaluation of susceptibility of landslide hazard plays an important role in landslide hazard risk management and urban planning. In previous studies，few scholars had done any in-depth analysis and discussion on the state division of index factors. Wanzhou district of the Three Gorges Reservoir region where landslide disasters take place frequently is the focus of this study. Seven influence factors including stratum lithology，geological structure，water distribution，gradient，direction and structure of slopes，and land utilization were chosen to be the evaluation indexes. The state of each index was graded based on the variation of gradients of the cumulative frequency curve of landslide，the landslide area ratio curve and the grading area ratio curve. The information value model and the logistic regression model were used to build the susceptibility evaluation systems based on the data of 655 landslides in the history of the district. The susceptibility results of the two methods above were graded by adopting the K-means cluster analysis. A zoned map of landslide susceptibility for the whole district was obtained based on GIS platform. The two models were compared with respect to the modeling results，the accuracy，the application condition etc. The results showed that the prediction accuracies of the information value model and the logistic regression model reached 73.0% and 54.9% respectively，indicating that the information value model had better performance than the logistic regression model.