Shuiqing Yin

Vegetation growth is increasingly threatened by both extreme climate events and anthropogenic disturbances. Although considerable attention has been given to detecting large‐scale vegetation anomalies, comprehensive analyses of the spatiotemporal dynamics of negative extreme anomalies in vegetation growth (NEGs) and their underlying drivers remain limited. Here, we treat NEGs as event‐scale vegetation declines from onset to termination, enabling coherent tracking of vegetation decline events and their associated drivers. Applied to global leaf area index (LAI) data from 2001 to 2024, we find increasing tendencies in both NEG frequency (0.07% year ⁻¹ , p = 0.13) and affected area (394,471 km ² year ⁻¹ , p = 0.04), reaching 39% of the global land surface by 2024. Interannual variations in NEG frequency and affected area show positive associations with annual mean CO ₂ anomalies, which may suggest a potential influence on the stability of the terrestrial carbon sink. Coincidence analysis further indicates that drought‐related events dominate globally: single‐driver drought events, together with drought‐related compound events involving temperature extremes and wildfires, account for 58.5%–59.8% of all NEGs. By the end of the 21st century, up to 50.8% of the global land surface is projected to experience NEGs annually under the high‐emission scenario, approximately 1.6 times the simulated historical level. These findings highlight the urgent need for climate mitigation to reduce the intensification of compound climate extremes and underscore the importance of understanding how these events may destabilize ecosystems and amplify climate feedbacks in a warming world.

Obtaining precise erosion measurements requires costly fieldwork, making it infeasible to directly survey large domains such as a province or river basin. To extend fieldwork results across such extensive domains, we propose a novel spatial prediction method that treats local erosion distributions as objects in the Wasserstein space. These distributions are mapped into square-integrable trajectories and represented via basis expansion, forming a multivariate random field that captures spatial dependence. By applying local regression and Kriging in this representation, our approach flexibly models and predicts erosion distributions at arbitrary locations. This framework improves prediction for functionals of the distribution, such as the mean and exceedance probabilities. Simulation studies demonstrate that the proposed method outperforms a misspecified parametric alternative and existing Fréchet regression approaches. We illustrate the approach with a detailed erosion analysis in Shaanxi Province, China, where local measurements from surveyed watersheds are extended to predict erosion distributions across the entire province using covariates such as land use and elevation.

This study develops a high‐resolution rainfall regionalization for mainland China based on sub‐daily storm characteristics, addressing key limitations of previous classifications that relied on daily or coarser precipitation data and limited rainfall metrics. Using hourly records from 2275 meteorological stations between 1971 and 2020, over 2.31 million rainfall events (≥ 10 mm) were identified and categorised into five event types (A–E) based on event rainfall depth (total rainfall accumulation, mm), event duration (total hours), and peak rainfall intensity (maximum hourly intensity). Stations were then clustered into six groups and aggregated into eight rainfall zones, integrating both event‐type composition and topographic context. Pairwise correlation analysis reveals that 99.1% and 99.6% of stations exhibit significant positive correlations between event depth and duration, and between event depth and peak intensity, respectively—indicating that higher rainfall depths generally result from either longer durations or stronger intensities. In contrast, 92.8% of stations show a significant negative correlation between duration and peak intensity, suggesting a typical trade‐off whereby shorter events tend to be more intense. Spatially, mean event rainfall depth decreases from southeast to northwest. Mean durations display a “sandwich‐like” pattern across central and eastern China, with the longest events (> 21 h) occurring in the Qinba Mountains and Weihe Plain, Middle‐Lower Yangtze Plain and Northern Southeast Hills, Yunnan‐Guizhou Plateau, and southeastern Tibetan Plateau. Shorter durations dominate both southern coastal and northern inland regions. Peak intensity shows an inverse pattern, with the highest values (> 12 mm·h ⁻¹ ) along the Southern Southeast Hills and North China Plain and elevated values (10–12 mm·h ⁻¹ ) in the western Sichuan Basin. The station classification demonstrates that stations with similar event‐scale rainfall characteristics—despite wide geographic and climatic differences—can be grouped into the same station types, emphasising the role of common physical drivers such as moisture transport and topography. The use of event‐type composition as a key classification metric proves effective in distinguishing station‐level rainfall regimes. Moreover, the resulting rainfall zones align closely with major weather systems and physiographic boundaries, affirming the physical‐geographic basis of the regionalization. This study offers a transferable, event‐based framework for hydroclimatic classification, providing valuable insights for water resource management, flood risk assessment, and climate adaptation planning.