From data fusion to ecosystem response: Evaluating merged precipitation products for multi-threshold meteorological drought monitoring and vegetation impact

Reliable precipitation data is fundamental for accurate drought monitoring and understanding its impacts on terrestrial ecosystems. While multi-source precipitation fusion techniques show promise, a comprehensive evaluation of their performance across varying drought intensities and ecological impacts remains limited. This study evaluated two fused precipitation products—Bayesian Model Averaging Ensemble Precipitation (BMAEP) and Cheng-Kling-Gupta Efficiency Weighted-Ensemble Precipitation (CWEP)—against the Multi-Source Weighted-Ensemble Precipitation (MSWEP) across mainland China (2001–2018). We assessed their capability in monitoring droughts using 3-month Standardized Precipitation Index (SPI3) thresholds (-0.5 to −2.0) and quantified input data contributions. Furthermore, an Extreme Gradient Boosting (XGBoost) model, interpreted via SHapley Additive exPlanations (SHAP), uncovered the response patterns of vegetation (excluding irrigated croplands) to different drought intensities. Results show that fused products, particularly CWEP-2P and BMAEP-2P (both combining the Climate Prediction Center (CPC) and Modern-Era Retrospective Analysis for Research and Applications V2 (MERRA2) datasets), outperformed individual inputs and the benchmark in drought detection. CPC and MERRA2 were identified as the most critical input sources. SHAP analysis revealed that sparse vegetation and rain-fed croplands were highly susceptible to mild and moderate droughts, whereas forests exhibited more profound responses to extreme droughts (SPI ≤ −2.0), likely due to legacy effects. The fused product CWEP-2P effectively captured these ecological response patterns. The findings provide a scientific basis for selecting precipitation inputs and demonstrate the utility of interpretable machine learning in elucidating complex ecosystem responses to climate extremes, offering valuable insights for environmental management.