Abstract: This paper aims to construct a landslide susceptibility assessment system that integrates spatial heterogeneity characteristics, focusing on Lengshuijiang City, a typical resource-based city in Hunan Province, in response to the limitations of existing landslide susceptibility assessment methods in dealing with spatial heterogeneity and nonlinear relationships. Through field investigations and historical records, this paper obtains data on 325 landslide events from May 2015 to July 2024, and establishes a landslide susceptibility evaluation index system consisting of 12 factors, covering multi-dimensional variables such as terrain, vegetation coverage, distance-related factors, and lithology. In the data preprocessing stage, operations such as grid division, data projection transformation, missing value imputation, and standardization are carried out. The study adopts the Spatial Constrained Multivariate Clustering （SCMC） method to analyze the spatial distribution patterns of landslide events, and uses the Generalized Additive Mixed Model （GAMM） combined with “deviance explained” to evaluate the importance of variables. Meanwhile, GIS technology and the natural breaks method are utilized to achieve the visualization and classification of landslide susceptibility. The results show that the GAMM model considering spatial random effects outperforms the model without considering spatial random effects in terms of indicators such as AIC, BIC, pseudo R², and log-likelihood, and performs more excellently in identifying landslides in high-risk areas. The study reveals that variables such as profile curvature, distance to roads, terrain wetness index, and distance to mining areas are of extremely high importance in the assessment of landslide susceptibility. In addition, the research results indicate that the landslide-prone areas in Lengshuijiang City exhibit significant clustering, and the model considering spatial effects can more accurately reflect this pattern, effectively avoiding misjudgments in low-risk areas. The assessment system constructed in this study has important application value for the prevention and control of landslide disasters in Lengshuijiang City and similar resource-based cities.