Spatiotemporal dynamics and nonlinear associations of snow phenology with environmental factors in the forest-steppe ecotone of the Greater Khingan Mountains

Study region The western slope of the Greater Khingan Mountains, northeastern China, represents a climate-sensitive forest–steppe ecotone where seasonal snow processes strongly influence regional hydrological and ecological dynamics. Study focus Using a 1-km snow phenology dataset derived from Advanced Very High Resolution Radiometer (AVHRR) observations during the 1982–2019 hydrological years, this study investigated the spatiotemporal variations of snow cover duration (SCD), snow cover start date (SCS), and snow cover melt date (SCM). Modified Mann–Kendall trend analysis, Pearson correlation analysis, cross-wavelet analysis, and an annual pixel-based eXtreme Gradient Boosting (XGBoost)–SHapley Additive exPlanations (SHAP) framework were applied to characterize snow phenology changes and quantify the temporal variations in model-based contributions of environmental factors. New hydrological insights for the region Snow cover duration decreased significantly at a rate of −0.42 d/a, mainly associated with delayed snow cover onset (SCS: +0.46 d/a), while SCM showed only a weak decreasing trend (−0.046 d/a), indicating an asymmetric seasonal response of snow phenology. Temperature showed significant statistical associations with SCD and SCM, whereas total precipitation exhibited spatially heterogeneous relationships with snow phenology. The annual XGBoost models achieved moderate predictive performance, with spatial cross-validation R² values of 0.52, 0.40, and 0.56 for SCD, SCS, and SCM, respectively. SHAP analysis revealed that environmental contributions varied among years and snow phenology indicators: precipitation and NDVI frequently contributed substantially to SCD variation, NDVI showed increasing relative contributions for SCS after the mid-2000s, and the contribution structure for SCM varied among years. These findings indicate that snow phenology changes in this transitional cold region are closely associated with temporally varying environmental factors rather than a single dominant factor, providing new insights into seasonal snow storage changes and potential hydrological responses in cold regions.