A Novel Survival Analysis Model for Quantifying Time-Lagged and Nonlinear Effects of Meteorological Conditions on Snow Phenology

Snow phenology is a crucial indicator that captures the dynamic changes in snow cover, which play a significant role in shaping hydrological processes and influencing ecosystem functioning. Recent climate change has affected the temporal dynamics of snow accumulation and ablation processes, particularly on the Tibetan Plateau. However, accurately quantifying how meteorological conditions influence snow phenology, especially the nonlinear and time-lagged effects, remains challenging. To address this challenge, we present a novel survival analysis model, a state-of-the-art approach used in medical research, to examine the complex effects of meteorological conditions on snow onset date (SOD) and snow end date (SED). Rigorous validation demonstrates that incorporating nonlinear and time-lagged relationships enhances both the accuracy and interpretability of the model, providing deeper insights into snow cover dynamics on the Tibetan Plateau. Specifically, our findings indicate that meteorological factors show an average delay of 11-12 days for SOD and SED across the Tibetan Plateau. A 1 °C increase in temperature or a 1 W/m² increase in shortwave radiation reduces the probability of SOD by 10.7% and 1.7%, while increasing the probability of SED by 8.2% and 0.6%. Conversely, a 1 mm increase in precipitation or a 1 m/s decrease in wind speed increases the probability of SOD by 11.2% and 25.0%, and decreases the probability of SED by 12.5% and 17.6%, respectively. In addition to enhancing the quantitative understanding of how various meteorological factors influence snow phenology, the proposed model presents a promising approach for forecasting snow cover dynamics under future climate change scenarios.