Abstract Changes in seismic properties can help assess slope damage, which is influenced by external factors such as temperature, relative humidity, rainfall, air pressure, and local seismicity. However, the contributions of these factors, and how the dominant factor modulates transient and seasonal seismic velocity changes (), remain unclear across different slope types. We conduct 2 years of in situ monitoring of daily on two representative slopes in southwest China: the Pubugou (PBG) rock slope and the Jinpingzi (JPZ) clastic‐soil slope. We apply an eXtreme Gradient Boosting (XGBoost) regressor and SHapley Additive exPlanation (SHAP) to quantify each factor's marginal contribution. The XGBoost model outperforms other common models in predicting . SHAP analysis highlights temperature as the most influential factor while revealing a contrasting pattern, a linear correlation on the PBG slope versus a nonlinear relationship on the JPZ slope. This discrepancy is primarily attributed to the inherent material properties across slope types.
Li et al. (Wed,) studied this question.