AUTHOR=Zhou Xiaohong , Xing Yin 

TITLE=Spatial consistency assessment and landslide susceptibility prediction optimization

JOURNAL=Frontiers in Earth Science

VOLUME=Volume 13 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2025.1702688

DOI=10.3389/feart.2025.1702688

ISSN=2296-6463

ABSTRACT=Currently, although various landslide susceptibility models can achieve high prediction accuracy, their results have significant differences in spatial distribution, resulting in high prediction uncertainty, which poses a challenge to optimizing assessment methods applicable to such complex geological hazards. In order to reduce uncertainty, this study proposes a machine learning ensemble modeling method that combines spatial consistency analysis. Taking Ruijin City in Jiangxi Province as the research area, based on the selection of 12 influencing factors and hyperparameter optimization, three algorithms including XGBOOST, Random Forest (RF), and Support Vector Machine (SVM) were used to generate landslide susceptibility maps. All models performed well, with AUC values ranging from 0.84 to 0.93. However, spatial consistency analysis shows that the spatial correlation between maps between models is only 0.78 to 0.84, indicating that although the prediction accuracy is high, there is still significant spatial heterogeneity and uncertainty. Therefore, a logistic regression (LR) fusion model based on historical landslides was constructed. Use the compilation results as the dependent variable and the results of the three models as the independent variables. The results indicate that XGBOOST contributes the most, followed by RF and SVM. By integrating the three prediction results, a comprehensive vulnerability map was finally obtained, which was superior to the single model in terms of spatial consistency (correlation coefficient 0.87–0.91) and prediction accuracy (AUC = 0.95). This research framework effectively reduces the uncertainty of landslide prediction and improves the reliability and accuracy of evaluation results.