为探究流域内物源的空间分布状态对泥石流易发性的影响，采用数学统计原理最近邻指数对物源的空间聚类进行量化，并基于2 243条小流域为评价单元，以纵比降、面积-高程积分、地形湿度指数、地震峰值加速度、岩性坚硬度为孕灾指标，物源的聚集度指标、连通性指数、物源储量等物源指标为核心，依托LightGBM模型探究金沙江上游石鼓-岗托河段的泥石流易发性。研究过程分别计算物源因子的指标体系与不包含物源因子的指标体系。两种结果均表明：较高、高易发区主要集中在奔子栏-巴塘河段。通过ROC（receiver operating characteristic curve）曲线分析可得，加入物源特征指标后得出AUC（area under the curve）值与不含物源特征的AUC值相比提升了6%，表明在加入物源指标后，模型呈现出良好表现，预测精度较高；也证明了物源特征指标对于泥石流发生概率的关联性很大。

To explore the impact of the spatial distribution of material sources within a watershed on the susceptibility to debris flows. The nearest neighbor index was adopted, based on the principles of mathematical statistics, to quantify the spatial clustering of material sources. Using 2243 small watersheds as evaluation units, the longitudinal gradient, area-elevation integral, topographic wetness index, peak ground acceleration of earthquakes, and rock hardness were taken as disaster-prone indicators, and the aggregation index of material sources, connectivity index, and material reserves were taken as the core material source indicators. The LightGBM model was relied on to investigate the susceptibility to debris flows in the Shigu-Gangtuo section of the upper reaches of the Jinsha River. The research process calculated the index system of material source factors and the index system without material source factors. Both results indicate that the high and very high susceptibility areas are mainly concentrated in the Benzilan-Batang section. The receiver operating characteristic curve(ROC) curve analysis shows that after incorporating the material source characteristic indicators, the area under the curve(AUC) value increases by 6% compared to the AUC value without material source characteristics, indicating that the model performs well and has high predictive accuracy after the inclusion of material source indicators. It also proves that the material source characteristic indicators are highly correlated with the probability of debris flow occurrence.