How Precipitation Timing Asymmetry Restructures Vegetation Phenology in Water‐Limited Ecosystems

Abstract A fundamental paradigm in global phenology posits that temperature serves as the universal initiator of vegetation growth. Here we demonstrate that this assumption fails catastrophically in water‐limited ecosystems, where precipitation timing—not amount—inverts the canonical hierarchy of climate controls. Integrating 22 years of Enhanced Vegetation Index data with four multi‐source hourly precipitation products across northwestern China's drylands, we reveal that regional spring onset advances at 2.2 days·decade −1 on average, yet the ultra‐arid Tarim Basin (<50 mm annual precipitation) advances by 3.4 days·decade −1 while the Hexi Corridor delays by 0.9 days·decade −1 —divergent trajectories that thermal‐centric models cannot reconcile. Most strikingly, nighttime precipitation amount surpasses temperature as the primary spring onset driver in the Tarim Basin Desert Area (TBDA), as indicated by the XGBoost–SHAP attribution framework (5.32), accounting for 24.8% of total SHAP importance and constituting a diagnostic signature of hyperaridity. Partial least squares structural equation modeling (PLS‐SEM) further showed that in the hyperarid TBDA, nighttime precipitation exerted a stronger total effect on spring onset than daytime precipitation (−0.219 vs. −0.124), and its positive pathway to soil moisture was stronger than that of daytime precipitation. Machine learning attribution and PLS‐SEM together show that diurnal precipitation timing is an important control on dryland phenology, particularly for spring onset in the hyperarid TBDA. These findings extend the question of dryland phenology beyond how much it rains to when precipitation occurs, highlighting the combined roles of thermal forcing, precipitation amount, and precipitation timing in Earth system model development and ecological forecasting across expanding arid regions.