运用3S技术和数理统计模型开展滑坡灾害敏感性分析评价仍然是当前区域滑坡研究的热点，评价单元、评价因子和评价模型的科学合理选取、确定和构建是滑坡敏感性分析评价的关键。以陇中黄土高原典型生态脆弱区的甘肃省天水市为研究区，基于证据权法（WOE）模型，选取地形因子、地质因子、人类活动因子、生态因子等4类13个孕灾因子，采用2种评价单元（栅格单元、斜坡单元）和2种数据分辨率（30 m、10 m），创建4个不同的滑坡孕灾因子数据集（GU30、SU30、GU10、SU10），并结合研究区详细的滑坡编目数据，开展研究区滑坡孕灾因子敏感性定量分析，并进行敏感性分区评价和结果对比分析验证。孕灾因子敏感性分析结果表明，地层岩性、高程、地形起伏度、地表粗糙度和坡度等5个孕灾因子是区内滑坡发育的关键控制因子，即地质因子和地形因子控制和决定了区内滑坡的形成发育，地层岩性的整体敏感性值最高；敏感性分区评价结果表明，按滑坡点、滑坡面积和ROC曲线分别进行统计分析验证，基于斜坡单元（slope-unit，SU）分区的极高区滑坡占比高于基于栅格单元（grid-unit，GU）分区的极高区滑坡占比，基于斜坡单元10 m分辨率的滑坡敏感性极高区滑坡占比最高，基于斜坡单元10 m分辨率的滑坡敏感性评价结果的AUC值最大，其精度和准确度最高。因此，在滑坡敏感性评价中采用斜坡单元和高分辨率数据具有更高的评价精度效果。研究区滑坡敏感性分区结果与区内实际滑坡发育分布吻合，该分区结果也可为研究区滑坡灾害防治和国土空间规划利用提供技术借鉴。

Analysis and assessment of landslide susceptibility using 3S technology and mathematical statistical model is still the current hot topics. It is very important for selection of unit and factor and model to the landslide susceptibility assessment reasonably. Tianshui City is selected as the study area which is in the typical regions of Longzhong ecologically vulnerable area. Based on the weights-of-evidence approach （WOE） and two kinds of assessment units （grid-unit， GU， slope-unit， SU） and also two resolutions （30 m， 10 m）， a total of four different data sets （GU30， SU30， GU10， SU10） is established. Then quantitative analysis and contrast assessment division of the loess landslide susceptibility in the study area though selected four kinds of thirteen factors. The results show that stratum lithology， altitude， relief amplitude， roughness and slope are the key control factors of landslide development in the area. The geological factors and topographic factors control and determine the formation and development of the landslides in study area. The susceptibility value of stratum lithology is the highest value. Statistical analysis of susceptibility division is carried out according to the point and area of landslide respectively. The landslide ratio shows that the division based on the slope-unit is higher than that of the grid- unit. Landslide susceptibility area in highest division based on the slope-unit and 10m high resolution is accounted for the highest. Therefore， it has higher precision in landslide susceptibility analysis and assessment that using the slope-unit and high resolution data. The landslide susceptibility result is consistent with the actual landslide development in the study area. The division results can also be used for the landslide prevention and control and national spatial planning in the study area.