This study presents a landslide susceptibility assessment (LSA) for South Korea using stacking ensemble machine learning. The research aimed to construct a high-resolution (30 m) landslide susceptibility map across the entire country by integrating diverse natural factors, including past climate data. A total of 30 input variables were used, including 9 from past climate model data (MK-PRISM), 12 topographical factors, and 9 environmental factors. Sixteen machine learning algorithms were tested, and a stacking ensemble was applied to the four algorithms with the highest area under the curve (AUC). A separate assessment model was also developed for landslide events affecting areas larger than 1 ha. The highest-performing classifier was CatBoost, achieving an AUC of approximately 0.89 for both assessments. The most influential input variables were distance from roads, daily maximum precipitation, digital elevation model, and soil depth. CatBoost, lightGBM, XGBoost, and Random Forest had the highest AUC in descending order for all landslide events, while for large events, the order was CatBoost, XGBoost, Extra Tree, and lightGBM. The stacking ensemble enabled the creation of two landslide susceptibility maps. The study highlights the importance of using a stacking ensemble approach to improve the performance of landslide susceptibility models. The results demonstrate that CatBoost is the most effective algorithm for this task. The study also emphasizes the need for high-resolution input data and the importance of considering long-term historical data in LSA. The findings provide a statistical method for constructing a high-resolution landslide susceptibility map for South Korea, which can be used to assess landslide risks and inform disaster prevention strategies. The study also discusses the limitations of the approach, including the challenges of accurately determining landslide locations and the need for further research to improve the model's accuracy.This study presents a landslide susceptibility assessment (LSA) for South Korea using stacking ensemble machine learning. The research aimed to construct a high-resolution (30 m) landslide susceptibility map across the entire country by integrating diverse natural factors, including past climate data. A total of 30 input variables were used, including 9 from past climate model data (MK-PRISM), 12 topographical factors, and 9 environmental factors. Sixteen machine learning algorithms were tested, and a stacking ensemble was applied to the four algorithms with the highest area under the curve (AUC). A separate assessment model was also developed for landslide events affecting areas larger than 1 ha. The highest-performing classifier was CatBoost, achieving an AUC of approximately 0.89 for both assessments. The most influential input variables were distance from roads, daily maximum precipitation, digital elevation model, and soil depth. CatBoost, lightGBM, XGBoost, and Random Forest had the highest AUC in descending order for all landslide events, while for large events, the order was CatBoost, XGBoost, Extra Tree, and lightGBM. The stacking ensemble enabled the creation of two landslide susceptibility maps. The study highlights the importance of using a stacking ensemble approach to improve the performance of landslide susceptibility models. The results demonstrate that CatBoost is the most effective algorithm for this task. The study also emphasizes the need for high-resolution input data and the importance of considering long-term historical data in LSA. The findings provide a statistical method for constructing a high-resolution landslide susceptibility map for South Korea, which can be used to assess landslide risks and inform disaster prevention strategies. The study also discusses the limitations of the approach, including the challenges of accurately determining landslide locations and the need for further research to improve the model's accuracy.