 This study examined the spatial variations in the relationship between landslide susceptibility and its explanatory factors in Nianshan, China. It compared the performance of geographical random forest, GRF, against random forest, RF models, finding that GRF achieved a higher accuracy with an area under the curve AUC of 0.86 as it considered the spatial heterogeneity among variables. Additionally, the local feature importance derived from GRF allowed researchers to identify the impact of conditioning factors varied across space, providing implications for policy development by local governments to focus on different conditioning factors in specific counties to prevent and mitigate landslides. Furthermore, the study used spatial cross-validation, SEV, to evaluate the model performance, which addressed the over-optimism bias in model error evaluation. This article was authored by Xiaoyang Dai, Yuncheng Zhu, Kai Sun, and others.