【摘要】 宣威市地形地貌复杂，岩质较弱，易发生滑坡。本文以宣威市为例，选取地形地貌、气候水文和地质构造等12个指标，采用证据权模型(WOE),得到各指标分级状态的证据权重；再通过WOE模型与CatBoost算法耦合，得到研究区滑坡灾害易发性评价指数；最后选取SHAP对滑坡影响因子进行全局和局部解释。研究结果显示：(1)采用WOE和WOE-CatBoost耦合模型AUC值分别0.845 5和0.984 6,滑坡极高易发区的滑坡密度分别为0.181 0个/km~2和0.297 0个/km~2;(2)高程、地层岩性、NDVI和降雨是影响滑坡发生的主要因素。得到结论WOE-CatBoost模型可以用于滑坡易发性的预测，且滑坡极高、高易发区主要分布于研究区的东南部和东北部，低易发区主要分布在研究区的中部和西南部。更多还原

【Abstract】 Xuanwei City is characterized by complex topography and geomorphology, weak lithology, and frequent landslide occurrences. This study selects 12 evaluation factors, including topography, climatic hydrology, and geological structure in Xuanwei City, and applies the Weight of Evidence(WOE) model to obtain the evidential weights of each factor’s graded states. A coupled WOE-CatBoost model is subsequently developed to derive landslide susceptibility indices for the study area. SHAP is further employed for global and local interpretability analyses of landslide-influencing factors. Results demonstrate:(1) The WOE and WOE-CatBoost models achieve AUC values of 0.845 5 and 0.984 6, respectively, with landslide densities in extremely high susceptibility zones reaching 0.181 0/km~2 and 0.297 0/km~2;(2) Elevation, lithology, NDVI, and rainfall are identified as primary controlling factors for landslide occurrence. The findings confirm that the WOE-CatBoost model effectively predicts landslide susceptibility, with extremely high and high susceptibility zones predominantly distributed in southeastern and northeastern regions, while low susceptibility areas are concentrated in central and southwestern regions.更多还原