Flash droughts threaten global managed forests Authors: Jianzhuang Pang, Hang Xu, Yang Xu, Yifan Zhang, Xiaoyun Wu, Kexin Li, Zhiqiang Zhang Published: 2025 ABSTRACT AbstractFlash droughts, characterized by rapid onset and increasing frequency, pose significant threats to ecosystem stability and function. However, there remains no global consensus regarding forest responses to flash droughts. Here, using a reconstructed global high spatiotemporal resolution Standardized Precipitation-Evapotranspiration Index dataset and an interpretable machine learning framework, we find that global forests have experienced increasingly rapid, intense, and prolonged flash droughts over the past four decades. Managed forests are more prone to browning from flash droughts than intact forests due to their limited capacity to acclimate to rapid drought stress driven by extreme heat. Notably, our meta-analysis confirms that current forest management practices, designed to maximize ecosystem services, exacerbate the vulnerability of managed forests to flash droughts globally. Our findings highlight the escalating risks posed by increasingly frequent and prolonged flash droughts to managed forests, underscoring the urgent need to integrate resistance and resilience to extreme climatic events into forest management strategies.Subject terms: Natural hazards, Environmental impact Introduction Flash droughts, defined as droughts rapidly escalating into water stress within weeks1, are increasing in frequency and intensity with rising global temperatures2. Rapid shifts in water availability induced by flash droughts increase forest vulnerability, as trees typically struggle to adjust physiological processes to such abrupt changes3. Additionally, flash droughts are associated with compound hot-dry conditions4, further intensified by land-atmosphere feedback, thereby exacerbating stress in forests5 and potentially causing significant adverse impacts6. Indeed, the global response of forests to flash droughts remains a topic of ongoing debate. For instance, a recent study found rapid declines in the Normalized Difference Vegetation Index (NDVI) of forests during flash droughts in Northeast China7. In contrast, research reported that forests exhibit strong resistance to flash droughts, experiencing NDVI declines only after prolonged water stress8. Furthermore, studies from the Americas have documented a counterintuitive increase in NDVI during flash droughts in humid regions9,10. Therefore, understanding global forest responses to flash droughts is essential for identifying forests at potential risk and informing appropriate management strategies. Accurate assessment of flash droughts impacts on global forests requires indicators capable of monitoring drought intensification at sub-weekly intervals1. Soil moisture is a key indicator for identifying flash droughts due to its direct relation to water availability11. However, existing soil moisture products suffer limitations due to short observational records and shallow detection depths constrained by sensor technology. This limitation is particularly problematic in forested areas, where high vegetation optical depth obstructs accurate measurements of deep soil moisture12, thereby introducing substantial uncertainty in flash drought identification. Although the evaporative stress ratio, which quantifies forest evapotranspiration responses relative to anomalous evaporative demand, is recognized as an indirect indicator of water limitation13, the high uncertainty of evapotranspiration measurements from remote sensing and land surface models14 hinders the accurate identification and prediction of flash droughts. By contrast, precipitation and air temperature data provide long-term, higher-density records with relatively high accuracy and predictive reliability, highlighting their potential for early flash drought warning15. Consequently, the Standardized Precipitation Evapotranspiration Index (SPEI), integrating water availability and evaporative demand, is a widely adopted metric for identifying flash droughts and demonstrates strong consistency with soil moisture and evapotranspiration dynamics16. However, most SPEI-based flash drought studies typically focus on local scales16,17. Research is further constrained by the coarse spatiotemporal resolution (i.e., monthly or 0.25°) of available SPEI products at a global scale18. Thus, understanding global forest responses to flash droughts requires developing a long-term, high spatiotemporal resolution SPEI dataset for identifying and characterizing flash droughts precisely. Forest responses to flash droughts arise from the interactive effects among multiple regulating factors. Many studies confirm that meteorological anomalies during flash droughts and regional background climate significantly mediate forest responses7,19. Specifically, meteorological anomalies impact tree transpiration and photosynthesis by directly altering evaporative demand and water availability4. In addition, background climate influences the direction and magnitude of forest responses to meteorological anomalies, thereby regulating how forests respond to flash droughts. Moreover, forest responses vary according to flash drought characteristics. Increased drought intensity and duration generally exacerbate NDVI declines20. Fundamentally, the insufficient quantification of flash drought characteristics—such as onset rate, peak stress, and stress duration—continues to impede insights into how these attributes mediate forest responses. Studies quantifying the impact of regulating factors on forest responses21, usually focus on specific regions or isolated factors, inadequately addressing the spatial variability and their complex interactive effects. These limitations—insufficient quantification of flash drought characteristics and narrow regional or isolated-factor approaches—have led to inconsistent identification of key influencing factors across studies19,22. Consequently, a global reassessment of various regulating factors and their interactive effects on forest responses to flash drought is essential for accurately identifying forests at potential risk. Managed forests cover over 2 billion hectares (~54% of the global forest area), reflecting worldwide efforts to reduce forest loss and provide multiple ecosystem services, such as soil and water conservation, climate change mitigation, and socio-economic benefits23. However, recent studies raise concerns regarding the sustainability of these services, primarily because managed forests exhibit limited acclimatization to extreme climates24,25. Forest management practices (e.g., planting, logging, and fertilization) may compromise forest resilience and acclimatization capacity to water stress by altering characteristics such as age structure, canopy architecture, stand density, and root systems26. For instance, a recent study reported significantly greater reductions in basal area increments in Central Europe managed forests compared to intact forests under slowly developing drought conditions27. However, whether managed forests and intact forests differ in response to flash droughts remains unclear. Furthermore, the influence of forest management practices on forest responses to flash droughts under varying environmental conditions remains a critical knowledge gap. This gap may hinder forest managers and natural resource policymakers in addressing future natural disturbances, threatening the sustainability of forests and their associated benefits. In this study, we quantify forest responses to flash droughts and analyze the effects arising from multiple regulating factors over the period 1979–2018. First, we identify global flash droughts and analyze the spatiotemporal patterns of flash drought characteristics using a global high spatiotemporal resolution SPEI dataset (gSPEI-HR, 0.1°, daily). This dataset, achieving unprecedented spatiotemporal resolution, is derived from standardized 5-day (i.e., pentad) moving-average water deficits (precipitation minus potential evapotranspiration) based on the land component of the fifth generation of European ReAnalysis dataset. Forest response is quantified using standardized NDVI anomalies (ΔNDVI) during flash droughts, obtained from satellite-based reanalysis datasets. Subsequently, an interpretable machine learning framework is applied to identify key factors that regulate the impacts of flash droughts on global forests across various background climates, meteorological anomalies, forest traits, and flash drought characteristics. Finally, we address the knowledge gap regarding the complex influence of forest management by examining differences in forest responses under management practices and exploring interactive effects between forest management and other regulating factors. We find that current forest management practices inadvertently exacerbate the global vulnerability of managed forests to flash droughts, amplifying their susceptibility to widespread browning under changing climate. Results and discussion Spatiotemporal patterns of flash drought characteristics from 1982 to 2022 Over the past four decades, global forests have experienced flash droughts (Fig. 1), characterized by accelerating onset, prolonged stress duration, and increasing intensities (Supplementary Fig. 1). These flash drought characteristics show a significant spatial heterogeneity across latitudes. Low-latitude forests experience flash droughts more frequently, with faster onset (onset rate = −0.89 pentad−1), longer duration (stress duration = 4.82 pentads), and greater intensity (peak stress = −1.89) compared to forests at other latitudes. Moderate frequency flash droughts observed in high-latitude forests exhibit onset rates comparable to those in low-latitude forests, though with relatively weaker intensity (peak stress = −1.88) and shorter durations (stress duration = 4.52 pentads). Mid-latitude forests experience substantially lower flash drought frequency than both low- and high-latitude forests, featuring slower onset (onset rate = −0.85 pentad−1), shorter duration (stress duration = 4.35 pentads), and lower intensity (peak stress = −1.82). Spatial patterns and latitudinal variations in flash drought characteristics across global forests. a frequency (events per decade), b onset rate (pentad−1), c stress duration (pentads), d peak stress magnitude (dimensionless), and e zonal averages of flash drought characteristics by latitude (yellow: frequency; red: peak stress magnitude; blue: stress duration; green: onset rate). The map outline is based on 1:110 million scale coastline data from Natural Earth (naturalearthdata.com). The Sahelian Belt, exhibiting high sensitivity to precipitation variability28, is identified as the region frequently experiencing precipitation-driven flash droughts (Fig. 1a and Supplementary Fig. 2). Prolonged dry spells (Supplementary Fig. 1c) exacerbate soil moisture depletion in this semi-arid region29, leading to increased sensible heat flux from the drying surface, elevated air temperatures, and potentially further amplifying the severity of flash droughts30. The Amazon Basin experiences the highest severity of flash droughts with onset rate <−1 pentad−1, stress duration >5 pentads, and peak stress <−2 (Fig. 1). Flash droughts in this region are generally induced by the substantial rises in air temperature and the declines in precipitation (52.70%, n = 27903 of 52945; Supplementary Fig. 2), which further exacerbate compound hot-dry conditions through the land-atmosphere feedback31. Furthermore, increases in temperatures and precipitation variability under high-emission scenarios in the future may further intensify compound hot-dry conditions32. In the Central Siberian Plateau, the substantial rise in temperature leads to a marked increase in atmospheric dryness and evaporative demand33, making flash droughts to enter the stress stage (SPEI ≤ −1.28, see “Methods”) rapidly (Fig. 1b). Notably, both the onset rate and intensity are intensifying more rapidly than those observed in low-latitude forests (Supplementary Fig. 1b, d). Although regional discrepancies exist, particularly in ecologically diverse areas such as the Amazon Basin, Southern Africa, and Southeast Asia, the broad spatial patterns of flash drought frequency derived from SPEI are generally consistent with those derived from soil moisture-based indicator (Supplementary Fig. 3a). Such regional differences are consistent with previous studies employing various drought indicators4,11,28,34–37. For example, our SPEI-based identification of flash drought frequency aligns closely with ecological indicator-based results (i.e., standardized evaporative stress rate)28,35,37 (Supplementary Fig. 4), while soil moisture-based indicators show higher flash drought frequencies in these regions4,36. Differences among datasets, spatial resolutions, and study periods may further explain the regional discrepancies between our findings and those of previous studies16,34. Nevertheless, we emphasize that despite these variations in flash drought frequency, the overall spatial patterns of flash drought characteristics remain robust and are largely insensitive to the choice of drought indicators (Supplementary Fig. 3b–d). Impacts of flash droughts on global forests Flash droughts cause browning in 42.73% of global forests (ΔNDVI < 0), primarily occur along the east coast of the United States, the Brazilian highlands, the Sahelian Belt, and southeast Asia (Fig. 2a). The structure (i.e., leaf area index (LAI)) and function (i.e., solar-induced chlorophyll fluorescence (SIF)) of these forests also are impaired due to flash droughts, with consistency relative to NDVI changes being 63.19%and 63.11%, respectively (Supplementary Fig. 5). Interestingly, the spatial patterns of NDVI response during soil moisture-based flash droughts closely resemble those from SPEI-based flash droughts yet exhibit smaller response magnitude (Supplementary Fig. 6). Shallow soil moisture data fail to adequately represent the capacity of deep-rooted trees to utilize deeper water storage during droughts38. Moreover, SPEI integrates rapid moisture depletion identified by precipitation and changes in evaporative demand, thus offering more comprehensive characterization, better predictability, and earlier warnings of forest stress under flash droughts than soil moisture alone37. Impacts of flash drought on global forest. a Spatial distribution of forest response (ΔNDVI) in intact forests (IF; green indicates ΔNDVI > 0, yellow indicates ΔNDVI < 0) and managed forests (MF; blue indicates ΔNDVI > 0, brown indicates ΔNDVI < 0). b Frequency distribution of ΔNDVI responses: vertical lines indicate mean values (brown: all forests, n = 418,791; green: IF, n = 245,651; blue: MF, n = 173,140); the black vertical line represents ΔNDVI = 0. c Heatmap showing average ΔNDVI responses binned by mean annual temperature (MAT) and mean annual precipitation (MAP). ΔNDVI denotes anomalies in the Normalized Difference Vegetation Index during flash droughts relative to concurrent normal conditions. The map outline is based on 1:110 million scale coastline data from Natural Earth (naturalearthdata.com). The variability of forest responses to flash droughts is mediated by multiple interacting regulatory factors. Machine learning based importance analysis identifies forest management (intact and managed forest; 18.37%) as the primary factor influencing spatial variations in ΔNDVI patterns, followed by mean annual temperature (MAT; 12.07%), temperature anomalies (10.30%), and stress duration (9.98%) during flash droughts (Table 1). Managed forests typically exhibit lower NDVI than intact forests and show greater susceptibility to browning (mean ΔNDVI = −0.11; Fig. 2b), a pattern consistently observed across diverse forest types (Supplementary Fig. 7). While the total influence is limited, variations in age and canopy height within specific ranges can significantly affect forest responses. Specifically, forest resilience to flash droughts is enhanced with increasing age (Supplementary Fig. 8a). Expanded root system depth and extent enable forests to access more water to meet rising evaporative demand during flash droughts39. However, intensified competition for resources such as light, water, and nutrients in closed-canopy mature stands may offset the advantage conferred by stronger individual drought resistance40. As canopy height increases, the risk of browning decreases (Supplementary Fig. 8b). Taller trees possess a greater capacity for light acquisition, making them benefit more from increased radiation and favorable temperature conditions during flash droughts41. Nevertheless, additional increases in canopy height elevate vulnerability because the longer water transport distances amplify hydraulic failure risks42. Table 1Importance of factors on forest responses based on the XGBoost modelFactor typeSum. Importance (%)FactorFactor importance (%)Background climate conditions19.10Mean annual temperature12.07Mean annual precipitation7.03Forest traits44.91Forest management18.37Forest age6.99Tree density6.99Canopy height6.89Maximum rooting depth5.67Meteorological anomalies during17.16Temperature anomalies10.30Precipitation anomalies6.86Flash drought characteristics19.37Stress duration9.98Peak stress4.89Onset rate3.96 Importance of factors on forest responses based on the XGBoost model The impacts of flash droughts on forests vary depending on the background climatic conditions. NDVI declines are more common in forests located in regions with higher MAT and lower mean annual precipitation (MAP; Fig. 2c). In regions with higher MAT, forest photosynthesis is limited by increasing temperature, thus resulting in browning during flash droughts (Supplementary Fig. 8g). In regions with lower MAP, forests exhibit increased susceptibility to browning during flash droughts owing to water deficit (Supplementary Fig. 8h). In contrast, forests in colder (MAT < 5 °C) or wetter (MAP > 1500 mm) regions tend to show neutral or beneficial responses to flash droughts, potentially altering managed forests responses to flash droughts. For instance, managed forests in the Eastern European Plain (MAT ≈ 0 °C, MAP > 800 mm) exhibit greening during flash droughts. This greening response occurs because rising temperatures enhance the activity of previously subdued photosynthetic enzymes19, allowing forests to utilize abundant available soil water to sustain photosynthesis. Furthermore, prolonged flash droughts significantly exacerbate forest browning (Table 1 and Supplementary Fig. 8). During the onset stage of flash drought, defined as the period of rapidly intensifying water deficit preceding the stress stage (see “Methods”), forests NDVI increase due to elevated photosynthetic capacity (Supplementary Fig. 5c) associated with higher temperatures and solar radiation4. As flash drought persists, soil moisture declines rapidly due to increased transpiration exceeding replenishment from precipitation (precipitation anomalies <−0.6; Supplementary Fig. 8f). Low soil water available consequently limits photosynthesis, offsetting the initial benefits during the onset stage and resulting in forest browning (stress duration > 5 pentads, Supplementary Fig. 8j). The beneficial effects observed during the onset stage are typically associated only with moderate temperature increases (0.5