Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

Large dam operations can profoundly reshape downstream river morphology and disrupt river-lake interactions, leading to significant hydrological and ecological consequences. This study investigates long-term hydromorphological changes in the Jingjiang Reach of the Yangtze River following the operation of the Three Gorges Reservoir (TGR) since 2003. Based on daily discharge and water level data from three key hydrological stations (Zhicheng, Shashi, and Jianli), power-law stage-discharge relationships were developed to characterize channel evolution. Effective base level analysis indicated negligible change at Zhicheng but substantial riverbed incision at Shashi and Jianli, with mean lowering rates of 0.271 and 0.256 m/year, respectively. By 2024, cumulative channel storage had increased by approximately 2.45 billion m3, equivalent to 11 % of the average volume of Dongting Lake. Under the original hydrological regime, the Yangtze River flow was distributed into Dongting Lake through three main distributary inlets-Songzikou, Taipingkou, and Ouchikou. Results show a marked decline after TGR impoundment, particularly at Taipingkou and Ouchikou, where linear trends suggest potential inflow cessation by the mid-21st century. Extrapolated trends indicate that channel storage may reach 7.4 billion m3 by 2071, coinciding with a transition toward a "one-in, three-out" regime. This evolving hydrodynamic pattern is expected to weaken Dongting Lake's flood regulation capacity and exacerbate hydrological and ecological fragmentation. These findings highlight the necessity of integrated morphological monitoring and adaptive water management to mitigate the cumulative impacts of large-scale reservoir regulation on downstream river-lake systems.

Abstract During July–August 2022, Pakistan (PKT) experienced catastrophic flooding while the Yangtze River Basin (YRB) endured unprecedented heatwaves. While previous studies have examined the physical teleconnections, there remains a critical gap in quantifying the role of anthropogenic forcing in shaping such trans‐regional concurrent extremes. Here, we bridge this gap by combining probabilistic and storyline attribution frameworks to assess both historical and future risks of 2022‐like events. We find that the 2022 event represents a warming‐amplified analogue of the 2010 event, driven by a westward extension of the Western Pacific Subtropical High (WPSH) and an eastward shift of the South Asian High (SAH). Moisture and heat budget diagnosis reveal that dynamically horizontal moisture transport dominated the 2022 PKT precipitation, while surface cloud‐radiative forcing drove the YRB heatwave. Using complex network analysis, we uncover intensified cross‐regional linkages under SSP2‐4.5, SSP3‐7.0, and SSP5‐8.5 scenarios. Crucially, our bivariate probabilistic attribution indicates that anthropogenic forcing accounts for nearly 100% of the likelihood of the 2022 event. Projections show that, by 2071–2100, the probability of such events could rise by 57–326 times, relative to a baseline probability of 0.0015 in historical simulations. Further, storyline attribution demonstrates that anthropogenic thermodynamics and circulation dynamics contributed approximately 60% and 40% to the 2022 event, with nearly half of the dynamic effect attributable to anthropogenic forcing. These results offer a quantitative perspective on the rising risk of concurrent Pluvial Pakistan–Hot Yangtze events under climate change, offering valuable insights for regional climate resilience and adaptation planning.

Flash droughts, characterized by their rapid onset and intensification, can evolve into long-term agricultural droughts, thereby amplifying adverse impacts on water resources, agriculture, and ecosystems. However, the propagation from short-term flash droughts to long-term agricultural droughts remains limited understood, particularly across different flash drought types. Here we developed an integrated framework that combined convergent cross mapping (CCM), the random forest model, and the copula-based Bayesian approach to investigate the propagation pathways and underlying mechanisms. We applied this framework to analyze the propagation of meteorological, soil, and evaporative flash droughts into agricultural droughts in the Middle and Lower Reaches of the Yangtze River Basin (MLRYRB) from 2000 to 2022. Our results revealed strong causal relationships between flash droughts and agricultural droughts, with an average propagation time of 36.8-48.8 days. Meteorological flash droughts showed the shortest propagation time, while evaporative flash droughts exhibited the longest. Soil flash droughts demonstrated the highest propagation frequency, rate, and sensitivity to agricultural droughts, while evaporative flash droughts showed the lowest translation rates to agricultural droughts. We further found that flash drought severity strongly influenced the propagation of all flash drought types, particularly soil flash droughts, with a threshold value of 11.2 +/- 2.3. Additionally, precipitation and vapor pressure deficit (VPD) emerged as the most critical factor for meteorological and evaporative flash drought propagation, with threshold values of 14.3 +/- 7.6 mm and 7.8 +/- 2.3 hPa, respectively. These findings can advance our understanding of flash drought dynamics and mechanisms, offering important insights for effective drought mitigation.

While accurate assessment of drought evolution is a prerequisite for effective early warning and mitigation, conventional methodologies often struggle to characterise the intricate propagation pathways and feedback mechanisms inherent to the Yangtze River Basin (YRB). To address these deficiencies, this study establishes a causal inference framework for drought cascades by integrating convergent cross mapping (CCM), propensity score matching (PSM) and logistic regression. This approach facilitates the quantitative disentanglement of causal interactions among meteorological (MD), agricultural (AD) and hydrological droughts (HD), while effectively isolating confounding environmental variables. Our findings revealed that, in addition to the traditional MD-AD-HD propagation sequence, typically characterised by a 1-month lag, there existed significant, previously overlooked feedback loops (AD-MD, HD-MD and HD-AD) lacking strong lag signals, with a 0-month lag accounting for 63% of the YRB. Notably, the transformation risk from AD to MD was the highest (60%), whereas that from MD to HD was the lowest (20%), indicating that drought feedback mechanisms exert a consistently stronger influence than conventional downward propagation. Furthermore, precipitation (Pre) and vapour pressure deficit (VPD) were identified as the primary mitigating and intensifying drivers of these cascades (both propagation and feedback), respectively, with a unit increase in Pre reducing the drought cascade risk to as low as 0.18 times and a unit increase in VPD amplifying it to more than 3.0 times. The intensity of these interactions is further amplified in regions with lower altitudes and high soil clay content. Overall, this research offers a novel perspective on the bidirectional coupling of hydro-meteorological extremes, providing a robust scientific framework for nuanced drought risk management.

Compound precipitation and wind extremes (CPWEs) are destructive multivariate events whose risks are amplified under climate change. It is urgent but challenging to assess intensity risk and project future exposure due to complex social-meteorological factor interaction and uncertainty of GCM outputs. To deal with these issues, a novel CPWE risk assessment framework was established containing Bayesian Model Averaging (BMA) based GCM outputs ensemble, Copula-based CPWE selection, and intensity assessment. A comprehensive exposure index was built to evaluate future population and economic exposure. Key findings include: (1) The BMA ensemble outperforms individual GCMs, yielding higher accuracy and robustness in simulating precipitation and windspeed; (2) The intensity of future CPWEs is projected to increase, characterized by enhanced variability and spatial heterogeneity. A significantly larger proportion of regions exhibit upward trends under both SSP245 and SSP585 scenarios, particularly under the latter scenario; (3) High exposure risk persists in the North China Plain and southeastern coast, with most regions experiencing increased risk relative to the historical period. Risk peaks around mid-century, indicating a critical period for climate-socioeconomic tipping points. The results provide critical insights into the spatiotemporal patterns of future CPWEs, supporting the development of effective early warning systems and climate resilience strategies.

Under global climate change, drought frequency and severity in Central Asia (CA) have risen sharply, threatening ecological security. Despite extensive studies on drought evolution, a quantitative framework for revealing the joint mechanisms and compound risks of multiple drought types remains lacking. Therefore, this study analysed droughts in CA from 1982 to 2022 by integrating multiple indicators to characterise meteorological (Standardised Precipitation Evapotranspiration Index, SPEI), agricultural (Palmer Drought Severity Index, PDSI) and hydrological droughts (i.e., Gravity Recovery and Climate Experiment [GRACE]-Drought Severity Index, GRACE-DSI). A Vine Copula model was subsequently employed to construct multidimensional dependence structures among key drought characteristics. The main findings were as follows: (1) meteorological droughts were predominantly short-term (e.g., 3 month), constituting approximately 96.7% of events, whereas hydrological and agricultural droughts exhibited substantial proportions of medium- to long-term events (e.g., larger than 6 months), at 35.5% and 68.1% respectively, indicating their stronger cumulative effects and recovery lags; (2) significant time-lagged couplings occurred among drought types, with high joint probabilities concentrated in the Tianshan Mountains and central arid core. Agricultural droughts exhibited joint probabilities above 0.8 at 3-6 month scales, while extending the timescale to 12 months substantially strengthened synchronisation across all drought categories, highlighting the importance of incorporating longer timescales in drought early warning systems; and (3) driven by increased duration, severity and intensity, the joint return periods of meteorological and hydrological droughts generally ranged between 3 and 10 months, whereas those of agricultural droughts exceeded 8 months even at short timescales. The findings can provide valuable insights into the multidimensional drought couplings in CA.

City clusters, being concentrations of socio-economic activities and resource-environment pressures, confront significant challenges in achieving sustainable development due to spatial heterogeneity in factors such as resource allocation, environmental governance, and economic growth. Although nexus thinking has gained widespread recognition in the sustainable management, a critical gap exists in quantitatively assessing how these spatial disparities influence the sustainable development of city clusters within an integrated framework. To bridge this gap, this study develops a novel assessment framework based on society-water-energy-environment (SWEE) nexus. By incorporating spatial heterogeneity analysis, it overcomes the limitations of traditional isolated assessments. Applied to the City Cluster in the Middle Reaches of the Yangtze River (CCMRYR) from 2010 to 2022, the study quantifies a coupling coordination degree as a Sustainable Development Index (SDI). The results revealed a fluctuating upward trend in the SDI of the cities situated in the CCMRYR, the average value of which increasing from 0.610 in 2010 to 0.729 in 2022. There existed a high level of sustainable development in the capital cities of Wuhan, Changsha and Nanchang. However, these cities had limited capacity to spearhead the sustainable development of the entire city cluster. Factor detection based on the geo-detector identified the dominant influencing factors for spatial differentiation of the SDI shifting from daily municipal wastewater treatment capacity (explanatory power: 0.745) in 2010 to urbanization rate (explanatory power: 0.513) in 2022. Crucially, interaction detection confirmed that the combined influence of factor pairs was consistently stronger than their individual effects, highlighting the necessity of synergistic policy-making. These findings provide quantitative evidence for formulating targeted policies based on the nexus concept to address spatial development imbalances within city clusters, which is crucial for promoting the sustainable development of city clusters.

Global vegetation dynamics profoundly change the terrestrial water cycling processes, especially with cascading effects on hydrological drought evolution. Over the past three decades, China has experienced extensive vege-tation greening. However, it remains poorly understood how much of a role vegetation changes play in hy-drological drought. In this study, we employ a process-based distributed hydrological model integrated with vegetation dynamics, and design two scenarios using the observed Leaf Area Index (LAI) data (noted as S1) and detrended LAI data (noted as S2) to examine the effect of vegetation changes on hydrological drought across China. The results show that hydrological drought occurs frequently across China, with an average drought frequency of 31.89% and a significant drying trend (average Standardized Runoff Index (SRI) trend: 0.133 decade-1, p <0.05). Vegetation change significantly influences drought categories and drought trends, while it has a slight effect on drought frequency over the whole of China. Specifically, vegetation change accelerates the drying trend by 25.47%, with the national average drying trend declining from 0.106 decade-1 in S1 to 0.133 decade-1 in S2. Spatially, vegetation increase (or decrease) amplifies drying (or wetting) trends in 61.04% (or 21.45%) of China, particularly in semi-humid regions. These findings highlight that vegetation greening alters surface water-energy balances, thereby intensifying regional hydrological drought in most areas in China. This study provides valuable insights for drought risk assessment and water resource management in ecosystems undergoing greening.

In the context of global warming, comprehending the long-term trends of multiple drought types and their characteristics is crucial for mitigating and adapting to drought risks. However, studies on long-term trends rarely incorporate linear and non-linear features, a gap that requires more attention. Therefore, this study aimed to address this gap by exploring the linear and non-linear trends of multiple drought types (i.e., meteorological, agricultural and hydrological drought) and their characteristics (i.e., duration, frequency, intensity and severity) from 1940 to 2023 in the Yangtze River Basin (YRB). The main conclusions were as follows: (1) agricultural drought was most likely to occur, followed by meteorological and hydrological droughts among three droughts in the YRB, especially in Wujiang River Basin and Yibin-Yichang Reach. Meteorological drought showed the highest drought frequency (DF) and intensity (DI), and hydrological drought showed the highest duration (DD) and severity (DS) and (2) meteorological, agricultural and hydrological droughts were projected to intensify in 56.94%, 74.62% and 73.31% of the YRB, respectively. The trends exhibited a decrease in meteorological DD and DS, and an increase in agricultural DF and the hydrological drought characteristics in most of the YRB and (3) drought indices and their characteristics exhibited a negative correlation, while a positive correlation among the drought indices or characteristics (with a maximum kappa value of 0.85). This study provides a refined understanding of drought evolution and critical information for region-specific water resources management and climate adaptation strategies.

Study region: This study focuses on mainland China, where regions are classified into arid, sub-arid, sub-humid and humid types based on the aridity index (AI) thresholds. Study focus: This study quantified the combined effects of climate change and vegetation restoration (1982-2020) on Terrestrial Water Storage Anomaly (TWSA) using partial least squares structural equation modeling (PLS-SEM). We analyzed spatiotemporal trends and partial correlation relationships of TWSA and its related climate and vegetation variables, and evaluated direct and indirect pathways influencing TWSA across different climate zones. New hydrological insights for the region: (1) TWSA showed an overall decline of-0.267 cm/a, with notable decreases in southeastern Tibet, North China, and the Ili River Basin, whereas increases occurred in South China, the Songhua River Basin, and northern Tibet. (2) Since 2000, accelerated vegetation greening exerted heterogeneous impacts on TWSA. In arid/ sub-arid regions, initial vegetation expansion improved water retention, but exceeding local water carrying capacity ultimately led to net water loss. In humid/sub-humid regions, greening promoted water conservation, benefiting TWSA in humid regions and mitigating its decline in sub-humid regions. (3) Climate change influenced TWSA directly or indirectly through vegetation change. Precipitation was the primary positive driver, with its effect amplified by vegetation dynamics in humid/sub-humid regions but dampened in arid/sub-arid regions. Rising temperatures exerted a negative indirect effect on TWSA, with amplified effects in arid/sub-arid regions post-2000 but diminishing impacts in humid/sub-humid zones.

The water surface elevation (WSE) of rivers serves as fundamental data for various hydrological research and applications. The recently launched Surface Water and Ocean Topography (SWOT) satellite offers a revolutionary altimetry approach by providing wide-swath elevation mapping using a SAR Interferometer (InSAR) operating at Ka-band. While SWOT provides unprecedented spatio-temporal coverage of WSE, it has not been systematically compared with reference water stage databases. Currently, due to difficulties in accessing recent and globally homogenous gauge station records, established WSE derived from radar altimetry (RA) missions is the most suitable dataset to perform global validation of WSE. This study presents the first global-scale intercomparison of the two altimetry systems, the wide-swath InSAR technique used for the first time by SWOT and the classical along-track RA using the SAR technique, and identifies several representative factors influencing their consistency. SWOT WSE are compared with virtual stations derived from Sentinel-3 and Sentinel-6 missions, across five different node quality categories ("good", "suspect", "degraded", "bad" and a combined "all" group without "bad" data). The analysis further examines the potential influences from river width, river ice, backscattering coefficients (sigma0), and dark water fraction in modulating data consistency. The root mean square error (and correlation coefficient) between WSE from SWOT and RA in "good" and "suspect" data are 0.80 m (0.85) and 1.62 m (0.78), respectively, while those for "degraded" and "bad" data rise significantly to 8.80 m (0.60) and 16.91 m (0.50). The combined "all" category yields an overall RMSE (CC) of 5.15 m (0.65). For rivers wider than 160 m, SWOT measurements with "good" and "suspect" quality demonstrate notably improved consistency with RA compared to narrower rivers. Under frozen conditions, the reduced consistency between SWOT and RA is most evident in the "degraded" and "bad" quality data, with average reductions in CC of 0.17 and 0.21, respectively. In addition, radar backscatter strongly impacts the quality of SWOT-based WSE, as both extremely low values (dark water) and very high values (specular ringing) can lead to unrealistic estimates. Overall, this study offers important insights into the global performance of SWOT-based WSE estimation and informs the future refinement and application of SWOT data in hydrological research.

The mismatch between natural carbon sequestration supply and social carbon demand (i.e., anthropogenic CO2 emissions), referred to as CSSD, poses a critical challenge to achieving carbon neutrality. Here, the carbon sequestration supply (CSS) and demand (CSD) of the Yangtze River Basin in China, which encompasses multiple administrative divisions and water resource zones, were estimated using the Carnegie–Ames–Stanford approach model and energy-related carbon emission accounting from 2001 to 2021, respectively. Then, the supply–demand index (SDI) and the 4-quadrant model were applied to analyze the mismatch of CSSD at different spatial scales. Additionally, the optimal-parameter-based geographical detector and partial correlation analysis were used to reveal the main drivers of CSSD and their interaction effects. The results indicated the following: (a) Both CSS and CSD exhibited increasing trends and marked spatial heterogeneity. High CSD values were mainly situated in the urban agglomeration zones, where CSS values were relatively low, highlighting the spatial mismatch of CSSD. (b) Carbon deficit zones (SDI < 0) were concentrated in urban agglomerations, and the proportion of mismatched zones increased with increasing spatial scale but decreased over time. Notably, these spatial patterns were more accurately identified at the prefecture-level city scale than at the water resource zone level. (c) Climatic factors positively affected CSS, whereas socioeconomic factors influenced CSS and CSD in contrasting directions, aggravating mismatch. Among them, socioeconomic factors and land use/land cover were the primary driving factors (maximum q-statistic = 0.60). Interactions between climatic and socioeconomic factors exhibited synergistic effects. These findings will support effective management toward carbon neutrality.

Study region Yangtze River Basin of China Study focus This study examines the spatiotemporal patterns, seasonality, periodic variability, and hydroclimatic contributions of drought–flood abrupt alternation (DFAA) in the Yangtze River Basin during 1961–2023. Drought-to-flood (DF) and flood-to-drought (FD) events are identified using a standardized DFAA index (DFAAI) constructed from consecutive monthly standardized precipitation evapotranspiration index (SPEI) values. Trend, hotspot, spectral, and SHapley Additive exPlanations (SHAP) analyses assess their frequency, intensity, evolution, and associated predictors. New hydrological insights for the region DFAA frequency and intensity exhibit distinct spatial patterns, indicating that the recurrence and severity of rapid water-balance reversals are not spatially uniform across the Yangtze River Basin. High-frequency areas occur mainly in the middle–lower basin and lake regions, exceeding 10 events per decade, whereas the strongest events cluster along the Sichuan Basin margin. Basin-mean linear trends are weak, but DFAA variability strengthened after 2000. DF events span June–October, while FD events concentrate in July–September and show stronger synchronization among sub-basins, reflecting different monsoon-season pathways of moisture recovery and depletion. DF and FD frequencies show recurrent 3–5-year variability, whereas intensity contains longer-period components. SHAP results indicate that rapid precipitation and humidity changes are most consistently associated with the initiation of water-balance reversals, while temperature and radiation conditions contribute more strongly to their intensity, especially for FD. Large-scale circulation indices mainly modulate the background climate state.

Central Asia (CA), a region highly vulnerable to drought, is experiencing increasingly severe drought conditions under climate change. However, current understanding of the spatiotemporal dynamics of drought, its propagation mechanisms, and the quantitative contributions of key drivers, particularly snow cover and vegetation, remains limited. Therefore, a three-dimensional (3D) framework was employed to extract meteorological, hydrological, and agricultural drought (MD, HD, and AD) events and their propagation pairs, subsequently integrating the extreme gradient boosting model with the shapely additive explanations to systematically investigate the spatiotemporal evolution of growing-season drought events across CA, propagation characteristics from MD to HD (i.e., MD-HD) and MD to AD (i.e., MD-AD), and their dominant driving factors. The main findings are as follows: (1) MD, HD, and AD events exhibited similar spatiotemporal patterns, characterized by peak severity periods during 1940-1950 and 2010-2023, a southeast-northwest banded distribution of high-severity centroids, and a dominant east-west migration direction (>73 %), (2) analysis of MD-HD and MD-AD propagation revealed a spring-season concentration (>52 %), geographically distinct hotspots (i.e., MD-HD and MD-AD in the northwestern lowlands and the northern agricultural zones, respectively), and a prevalent east-west propagation direction (>67 %), and (3) MD severity exerted the strongest influence on propagation, with elevated LAI facilitating and increased snow depth/snowmelt suppressing this process, further compounded by synergistic effects among MD characteristics. This study provides a scientific basis for the early warning of drought-induced disaster chains in arid and semi-arid regions.

Accurate prediction of flood events is important for flood control and risk management. Machine learning techniques contributed greatly to advances in flood predictions, and existing studies mainly focused on predicting flood resource variables using single or hybrid machine learning techniques. However, class-based flood predictions have rarely been investigated, which can aid in quickly diagnosing comprehensive flood characteristics and proposing targeted management strategies. This study proposed a prediction approach of flood regime metrics and event classes coupling machine learning algorithms with clustering-deduced membership degrees. Five algorithms were adopted for this exploration. Results showed that the class membership degrees accurately determined event classes with class hit rates up to 100%, compared with the four classes clustered from nine regime metrics. The nonlinear algorithms (Multiple Linear Regression, Random Forest, and least squares-Support Vector Machine) outperformed the linear techniques (Multiple Linear Regression and Stepwise Regression) in predicting flood regime metrics. The proposed approach well predicted flood event classes with average class hit rates of 66.0%-85.4% and 47.2%-76.0% in calibration and validation periods, respectively, particularly for the slow and late flood events. The predictive capability of the proposed prediction approach for flood regime metrics and classes was considerably stronger than that of hydrological modeling approach.

Compound events (CEs) pose great challenges to disaster risk management as they can amplify risks to globally interconnected socio-economic systems. Identifying the hotspots of CEs and understanding their impact factors are critical for developing targeted climate adaptation strategies. Here, we analyzed 12 types of CEs, the pairwise combinations of seven different hazards (e.g, heatwave, drought and extreme precipitation) and one precondition (i.e., antecedent soil moisture), using global observations and reanalysis datasets of hydrometeorological variables across 520 major river basins during 1980-2019. The hotspots of CE were revealed based on their return periods, and the frequencies and seasonality of CEs in each continent were further explored. Lastly, the impact factors were determined among multiple atmospheric circulation variability modes using the odds ratio (OR) value. The results indicate that CE hotspots are mainly located in basin of Eastern Asia, Eastern North America, Western North America, the Mediterranean, and Northern Australia. Compound drought-heatwave events (D-H), compound antecedent soil moisture-extreme precipitation events (M-P) and spatially compound extreme precipitation events (spat.P) generally occur more frequently than other CEs, accounting for 12.97 similar to 27.05%, 5.22 similar to 22.92% and 15.68 similar to 21.13% of all compound events in the six continents, respectively. Among them, M-P and spat.P show strong seasonality only in Asia and the South-West Pacific. The El Ni & ntilde;o-Southern Oscillation (ENSO), Arctic Oscillation (AO), and North Pacific Pattern (NP) are the important impact factors of most CEs and are statistically associated with the occurrences of CEs, with 42.19 similar to 84.83%, 33.80 similar to 71.14%, and 0.79 similar to 25.35% of CEs occurring during ENSO, AO, and NP anomalies, respectively. These findings reveal the spatial heterogeneity of CEs driven mainly by atmospheric circulation anomalies, which could provide a basis for basin-scale climate risk management.

Drought indices based on probabilistic statistical distributions are widely used in drought assessment, and their calculation generally assumes stationarity in hydro-meteorological variables. However, the nonstationarity induced by climate change and human activities may largely challenge the traditional stationarity-based drought assessments. In this study, we systematically diagnose the nonstationary changes in precipitation, water deficit, runoff, and soil moisture across the Chinese mainland from 1961 to 2019. These hydro-meteorological variables are key inputs for the calculation of meteorological, hydrological, and agricultural drought indices. Additionally, we investigate the impact of nonstationarity in hydro-meteorological variables on drought assessments. We find that 44.7

Flash droughts pose severe risks to vegetation growth and ecosystem stability. Vegetation recovery time following flash drought is a key indicator of ecosystem resilience. However, global patterns and drivers of vegetation recovery across climate zones remain unclear. The Ko & uml;ppen-Geiger classification links background climate and ecological responses, providing a framework for assessing recovery. This study analyzed the spatiotemporal recovery patterns from 2001 to 2023 using gross primary productivity (GPP) and solar-induced chlorophyll fluorescence (SIF) data. The key drivers were examined across different climate zones using SHAP analysis based on the XGBoost model. Dominant factor distributions were clarified using partial correlation analysis. Results indicated: (1) The global mean vegetation recovery time was approximately 55.70 days. Tropical climates showed the shortest recovery time, while arid climates and Mediterranean subtypes exhibited longer recovery time. Recovery time increased in 52.91% of the vegetated area, while 47.09% showed shortening trends. Notably, shortening trends were more prevalent in ecosystems historically exposed to chronic climatic stress (hot-summer temperate, monsoon continental, and arid zones). (2) Precipitation and shortwave radiation were the primary drivers, jointly explaining 44.54%-56.12% of the variance in the recovery time. Temperature ranked third in most regions but contributed less than 10% in tropical climates. Soil moisture dominated in tropical and cold-arid climates, while vapor pressure deficit was critical in continental and hot-arid climates. (3) Partial correlation analysis revealed the spatial distribution of dominant factors and their influences on recovery time. In water-limited regions (arid zones and hot-summer continental climates), increased precipitation facilitated recovery, whereas heightened radiation exacerbated water stress and delayed recovery. Conversely, in tropical climates and cold-summer continental climates, vegetation recovery showed a stronger sensitivity to energy availability: higher radiation promoted recovery, whereas excessive precipitation inhibited it. These findings are critical for predicting ecosystem responses to climate change and for formulating targeted adaptation strategies.