The Civilian Access Control Zone (CACZ), south of the Demilitarized Zone (DMZ) separating North and South Korea, has functioned as a unique bio-reserve owing to restrictions on human use. However, it is now increasingly threatened by damaged land and slope failures. In this study, a machine-learning-based method was used to assess slope stability by introducing the random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and logistic regression (LR) approaches. These classification models were trained and evaluated on 393 slope stability cases from 2009 to 2019 to assess slope stability in the northern area of the Civilian Control Line, South Korea. For comparison, the performance of these classification models was measured by considering the accuracy, Cohen’s kappa, F1-score, recall rate, precision, and area under the ROC curve (AUC). Furthermore, 14 influencing factors (slope, vegetation, structure conditions, etc.) were considered to explore feature importance. The evaluation and comparison of the results showed that the performance of all classifier models was satisfactory for assessing the stability of the slope, the ability of LR was validated (accuracy = 0.847; AUC = 0.838), and XGBoost proved to be the most efficient method for predicting slope stability (accuracy = 0.903; AUC = 0.900). Among the 14 influencing factors, the external condition was the most important. The proposed supervised learning method offers a promising method for assessing slope status, may be beneficial for government agencies in early-stage risk mitigation, and provides a database for efficient restoration management.

Erosion control dams (ECDs) in forested mountain streams are key pieces of infrastructure for disaster risk reduction, but they have also been reported to disturb the hydro-geomorphic balance of stream ecosystems. Open-type ECDs have increasingly been used to mitigate the loss of longitudinal connectivity caused by closed-type ECDs; however, quantitative comparisons between the two types remain scarce. We evaluated the presence of fish (Rhynchocypris oxycephalus) and amphibian (Glandirana emeljanovi) using logistic regression analysis in a forested mountain stream on Mt. Oseo, Republic of Korea, where closed- and open-type ECDs were installed in sequence. The best-fit models indicated different habitat preferences: R. oxycephalus responded to hydro-geomorphic parameters, whereas G. emeljanovi responded to riparian forest vegetation and soil parameters as well. Although six parameters varied significantly between the sections divided by the two types of ECDs, all observed values fell within the ranges previously reported for mountain streams in natural conditions. Bayesian Network Analysis further identified crown density, stream-bed slope, and number of step pools as high-level parameters that can be regulated through erosion control works, including engineering and forestry techniques. These findings link the presence of both species to stream conditions and are translated into an ecosystem management framework, filling a knowledge gap where comparable studies are scarce.

Rapid urban expansion into mountainous terrain, together with increasing exposure to extreme weather, has increased the need for transparent and interpretable slope-hazard assessment in forested and residential mountain environments. However, relatively few studies have integrated field-survey information with physically based simulation outputs for interpretable multi-class classification within an established slope-hazard grading framework. This pilot study developed a simulation-informed machine-learning framework for classifying composite slope-hazard grades (Grades I–III) using 71 landslide-prone and 52 debris-flow-prone sites in Sejong City, Republic of Korea. For the landslide dataset, selected field-survey indicators were integrated with SoilWorks-derived slope-stability information, whereas for the debris-flow dataset, selected field-survey indicators were integrated with FLO-2D-derived simulation information. Four machine-learning algorithms—Random Forest (RF), Logistic Regression (LR), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM)—were evaluated under severe class imbalance. RF showed the strongest overall performance in the original holdout evaluation, achieving an accuracy of 0.857 for both datasets and weighted F1-scores of 0.849 for the landslide dataset and 0.836 for the debris-flow dataset. Repeated stratified train–test analyses further showed that RF retained the highest mean overall performance, although the debris-flow dataset was more sensitive to sample allocation. Gini importance, Shapley Additive Explanations (SHAP), and targeted ablation analysis indicated that simulation-derived variables provided strong discriminatory information within the established grading framework. The proposed framework provides a structured and interpretable approach for integrating field-based assessment indicators with physically based simulation outputs and may provide supplementary support for classifying and prioritizing sites requiring further assessment. The findings also provide a basis for future investigation of reduced-input screening approaches, although the present study does not establish that reduced field surveys or simulation outputs alone are sufficient for formal grade assignment. Accordingly, the framework should be interpreted as a supplementary classification approach within the existing assessment system rather than as an independent predictor of future landslide or debris-flow occurrence.