Identifying climate and topographic factors for Korean Fir (Abies koreana) of Mt. Hallasan using machine learning under projected climate change scenarios

The Korean fir ( Abies koreana ), an endemic species that exists in isolated populations on certain high-elevation mountains in South Korea, has been experiencing a steady decline in recent years. This study aimed to predict future suitable habitats of Korean fir on Mt. Hallasan, which marks the southernmost limit of the species’ distribution, and to identify the key environmental factors influencing its presence and spatial distribution. A total of 36 variables, including bioclimatic, topographic, and soil characteristics, were utilized to predict species occurrence and density at a spatial resolution of 100 meters. The analytical methods employed comprised logistic or lasso regression, Random Forest, Gradient Boosting, XGBoost, LightGBM, and an ensemble modeling approach. Furthermore, a hurdle model was constructed using these variables. The relative importance of each predictor and its influence on the response variables were evaluated through Shapley Additive exPlanations (SHAP) analysis. Among the models tested, XGBoost showed the best performance for predicting species occurrence (AUC>0.9), while LightGBM was most effective for modeling density (R 2 =0.69). Projections under future climate scenarios, specifically SSP3-7.0 and more severe pathways, more than half of the suitable habitat and over 90% of the population will be lost by the end of the 21st century, with faster declines occurring in areas of higher population density. Both the presence and density models were primarily influenced by the Snow Average , a derived variable, preceding the start of growth, showing a marked increase once a critical threshold of 290 mm, which is the insulation standard, was surpassed. The second key variable, bio03 , was associated with values below this critical threshold. Other important predictors in the presence model included bio15 and Positive Openness , while the density model was more strongly affected by topographic variables. The research highlights that successful conservation and management of Korean fir populations depend on minimizing drought stress before the growth season and enhancing the physical stability of their habitats. Achieving this necessitates ongoing habitat monitoring and thorough observation to identify critical thresholds for essential environmental factors, particularly microtopographic characteristics and snow cover patterns.