Integrating snowpack mechanical properties into snow avalanche susceptibility mapping in continental dry–cold mountain regions

Snow avalanches are among the most destructive natural disasters in mountainous regions, and the evolution of snow properties controls their initiation and release. However, few studies have incorporated snowpack mechanical properties (SMP) into avalanche susceptibility mapping, limiting representation of snow instability processes. To address this gap, this study developed an avalanche susceptibility mapping framework that integrated snow shear strength and snow density, two key SMP, obtained from physically based WRF model simulations in the western Tianshan Mountains of China, where dry–cold avalanches occur frequently. Four machine learning models (SVM, RF, XGBoost, and CatBoost) were used to evaluate their influence. Results show that adding SMP improves model performance by 0.3–4% in AUC and refines susceptibility zoning, with low- and moderate-risk areas shifting by 2–4% while high-risk zones remain stable. Among all models, CatBoost achieves the best performance with AUC = 0.972. The SHAP-based interpretability analysis identifies the leading conditions influencing avalanche occurrence, with elevation (<2100 m), slope (>25°), snow shear strength (<490 Pa), and NDVI (<0.69) emerging as the most influential variables. Low shear strength combined with steep slopes reduces snowpack stability and promotes weak-layer failure. Integrating SMP enhances predictive accuracy and physical interpretability.