Global distribution, trends, and drivers of flash drought occurrence

Flash droughts are rapid-onset drought events that can devastate agriculture, ecosystems, and economies due to swift land-surface desiccation. Despite growing interest, predicting these sub-seasonal events remains challenging. Prior research has advanced detection and monitoring using soil moisture, evaporative demand, evapotranspiration (ET) and potential evapotranspiration (PET)-based evaporative stress metrics, and impact-based approaches, with documented cases across the U.S., Brazil, southern Africa, Spain, western Russia, and Australia. The key question motivating this study is: Which global regions are most susceptible to flash drought occurrence? The study aims to quantify the global spatial distribution and seasonality of flash droughts, assess trends in occurrence from 1980–2015, and diagnose the relative roles of precipitation deficits and enhanced evaporative demand in driving rapid drought development, thereby informing risk assessment and predictability efforts.

The paper synthesizes multiple strands of flash drought research: (a) detection/monitoring using soil moisture percentiles and in situ observations; (b) atmospheric evaporative demand indices (e.g., EDDI) and PET variability; (c) evaporative stress metrics derived from ET/PET ratios such as the Evaporative Stress Index (ESI) and standardized evaporative stress ratio (SESR), which provide early indicators and correspond with impact datasets (e.g., USDM); and (d) impact-based assessments of agricultural and ecological stress. Regional studies have identified rapid drought intensification events across North America, South America, Africa, Europe, and Australia. Mechanistic work highlights land–atmosphere coupling, blocking highs, and compound heat-drought-wildfire linkages. A prior hemispheric-scale analysis using soil moisture identified hotspots in the Sahel and India and a mid-latitude band across Eurasia, broadly consistent with the present study, though spatial details over North America differ. The review underscores gaps in quantifying global patterns, seasonal timing, long-term trends, and the relative contributions of precipitation shortfalls versus elevated evaporative demand.

Data: Four global reanalysis datasets were used for 1980–2015: MERRA, MERRA-2, ERA-Interim, and ERA5. Daily ET and PET were obtained; PET was computed using the FAO Penman–Monteith formulation. The evaporative stress ratio (ESR=ET/PET) was computed daily, aggregated to pentads, and standardized to produce SESR at each grid cell and pentad. The pentad-to-pentad change in SESR (ASESR) was also computed, detrended, and standardized. Flash drought identification: A comprehensive method emphasizing both impact and rapid intensification was applied using SESR/ASESR with four criteria: (1) minimum duration of six pentads (≈30 days); (2) final SESR below the 20th percentile; (3a) most ASESR values at or below the 40th percentile with (3b) no more than one ASESR above the 40th percentile after a worsening step; and (4) the mean change in SESR over the event below the 25th percentile. Percentiles were computed locally per grid cell and pentad across years. This distinguishes rapid drought development from short dry spells and ensures both speed (flash) and impact (drought) components. Compositing and spatial analysis: Each reanalysis’ flash drought frequency map (percent of years with at least one event) was bilinearly interpolated to 0.5°×0.5° and averaged to create a composite frequency map. Arid/hyper-arid (aridity index AI=P/PET<0.2) and cold/low evaporative demand regions (average growing-season PET<1 mm d−1) were masked to focus on regions susceptible to rapid desiccation and likely vegetative impacts. Fifteen regional domains (hotspots and key agricultural areas) were defined for temporal analyses. Seasonality: For each domain, flash droughts were binned by initiation month. For tropical/subtropical regions (30°S–30°N), all months were considered; for mid-latitudes (30–60°), analysis focused on growing seasons (Mar–Oct NH; Sep–Apr SH). Monthly occurrence was expressed as each month’s share of the annual total. Trends: Annual flash drought spatial coverage within each domain was computed as the fraction of grid points experiencing at least one event per year. The Mann–Kendall test assessed monotonic trends; p<0.1 indicated statistical significance. Trends from individual reanalyses were also examined for consistency. Drivers: For each event and grid cell within the domains, accumulated precipitation and PET were computed over the flash development period. The Standardized Precipitation Index (SPI) and standardized PET anomalies were calculated for that same window using 1980–2015 climatology. Events were classified by thresholds: SPI≤−1 (large precipitation deficit), PET≥+1 (large positive evaporative demand anomaly), both, or either. Regional frequencies were aggregated and compared across reanalyses. Validation of impact: Soil moisture percentiles at event end (from reanalyses) were summarized to assess land-surface desiccation associated with SESR-identified flash droughts. Gridded land cover, PET climatology, precipitation climatology and variability, and teleconnection indices (e.g., Niño 3.4) were used to interpret spatial patterns, seasonality, and interannual relationships. Code and derived datasets were archived; data sources included MODIS land cover and the reanalyses noted above.

- Global hotspots and frequencies (1980–2015): Highest composite flash drought occurrence (30–40% of years) in Brazil, the Sahel, the Great Rift Valley, and India. Additional notable regions (20–30%): central Mexico, Indochinese Peninsula, northern Australia. Mid-latitude local hotspots (10–20%): central United States, Iberian Peninsula, Asia Minor, southwestern Russia, northeastern China. Agreement between reanalyses was strong in several tropical/subtropical hotspots (coefficients of variation <0.3) and lower in many mid-latitude regions (0.3–0.6), especially the central U.S. - Seasonality: Mid-latitude Northern Hemisphere regions generally peak in May–July (central U.S., southwestern Russia, northeastern China), with exceptions: Iberian Peninsula shows a bimodal peak (June and September), Asia Minor increases through the season. Southern Hemisphere mid-latitudes show differing patterns: Argentina exhibits variable monthly occurrence; southeastern Australia peaks near the end of the austral growing season. In the tropics/subtropics, Northern Hemisphere regions (Mexico, Sahel, India, Indochinese Peninsula) peak during boreal growing season; Southern Hemisphere regions (Brazil, Great Rift Valley, northern Australia) peak during the austral growing season. - Trends in spatial coverage (Mann–Kendall, p<0.1): Significant increases in six regions (central United States, Iberian Peninsula, Asia Minor, Brazil, Sahel, southeastern Australia); significant decreases in three regions (India, Great Rift Valley, northern Australia). Magnitudes vary: modest (~4%) increases in central U.S. and Iberian Peninsula; large changes (14–26%) in Asia Minor, India, Sahel, Great Rift Valley, northern Australia. Minimal changes (<2%) in Mexico, Argentina, Indochinese Peninsula. Most regions showed consistent trend signs across reanalyses; at least three reanalyses had significant trends in four of the nine significant regions. - Drivers of rapid intensification: Across all domains, large precipitation deficits (SPI≤−1) occurred in 33% of events, large positive PET anomalies (≥+1) in 31%, and either anomaly in 44% of events; both occurred concurrently in ~20%. Regional contrasts: European domains (Iberian Peninsula, Asia Minor, western Russia) were more PET-driven; domains in the Americas (central U.S., central Mexico, Amazon, Brazil, Argentina) were more precipitation-deficit-driven. Concurrent SPI and PET anomalies were more common in tropics/subtropics (20–34%) than most mid-latitudes (5–16%, except central U.S.). - Soil moisture impacts: End-of-event soil moisture fell to the 20th percentile or lower in 11/15 regions and to the 25th percentile or lower in some portion of all 15 regions, confirming substantial land-surface desiccation associated with SESR-identified flash droughts. - Contextual drivers and consistency: Hotspots align with regions of strong land–atmosphere coupling (central U.S., Sahel, India). Anticyclones/blocking highs contribute to simultaneous rainfall suppression and elevated evaporative demand (e.g., 2012 central U.S., 2010 western Russia). Higher climatological PET in the tropics/subtropics (often >5 mm d−1) likely contributes to higher hotspot frequencies relative to mid-latitudes (3–5 mm d−1). High interannual rainfall variability in the tropics also elevates risk. ENSO (Niño 3.4) correlates with annual flash drought extent in several domains (Amazon, Argentina, Indochinese Peninsula, Great Rift Valley). Regions with increasing PET trends (Iberian Peninsula, Brazil, Sahel) also show increasing flash drought coverage.

The study addresses the central question of global susceptibility by identifying and comparing multi-reanalysis hotspots, their seasonality, and trends, and by quantifying the relative roles of precipitation deficits and elevated evaporative demand. Results show that both drivers contribute with comparable frequency, implying that flash drought risk cannot be assessed solely from precipitation anomalies; monitoring and prediction must incorporate evaporative demand (temperature, VPD, radiation, wind) alongside precipitation. The spatial clustering of hotspots in regions with strong land–atmosphere coupling suggests positive feedbacks—soil moisture depletion reduces ET, increases surface temperature and VPD, suppresses convection, and accelerates drying—play a pivotal role. Persistent anticyclones/blocking further compound rapid intensification by curtailing rain and increasing PET. Seasonal patterns are consistent with regional hydroclimate regimes: monsoon onset/variability (India, Sahel, northern Australia) and ITCZ migration modulate timing, with delays or weak onset amplifying risk when PET is high. In the Amazon, dry-season vegetation dynamics and enhanced radiation elevate evaporative stress, predisposing to flash drought. Trends indicate that some agriculturally important regions (e.g., central U.S., Iberian Peninsula, Sahel, Brazil, southeastern Australia) have seen increasing flash drought coverage over 1980–2015, aligning in some cases with increasing PET during growing seasons, consistent with warming-driven increases in evaporative demand. Conversely, decreases in India, Great Rift Valley, and northern Australia may reflect changes in precipitation magnitude/timing and large-scale circulations (e.g., South Asian monsoon weakening, Walker circulation changes). Teleconnections such as ENSO modulate interannual variability of flash drought coverage in multiple regions, underscoring the potential for seasonal predictability. Collectively, the findings underscore the global prevalence of rapid drought development outside of arid/cold regions, the centrality of evaporative demand, and the vulnerability of major crop-producing areas, with implications for food security, wildfire risk, water resources, and broader socioeconomic stability.

This study provides a global, multi-reanalysis climatology of flash droughts based on evaporative stress, identifying prominent hotspot regions, their seasonal timing, recent trends in spatial coverage, and the relative contributions of precipitation deficits and elevated evaporative demand. Key contributions include: (1) mapping hotspot regions with strong inter-dataset consensus (e.g., Brazil, Sahel, Great Rift Valley, India) and notable mid-latitude locales; (2) characterizing regional seasonality linked to monsoons, ITCZ shifts, and growing seasons; (3) detecting significant regional trends—both increases and decreases—over 1980–2015; and (4) demonstrating that both precipitation shortfalls and positive PET anomalies are frequent and often concurrent contributors to flash drought development. The results highlight substantial desiccation impacts as reflected in soil moisture percentiles and emphasize risks to agricultural regions. Future work should pursue: improved sub-seasonal predictability that integrates evaporative demand and land–atmosphere coupling; disentangling the roles of teleconnections and decadal variability; extending analyses with longer records and alternative drought metrics; and assessing the cascading socioeconomic impacts to support early warning and resilience planning.

- Metric and data dependence: Flash droughts are identified via SESR/ASESR from reanalyses; results may differ using other variables (e.g., soil moisture-based) or satellite-only products, though major hotspots show cross-method consistency. Reanalyses vary in land-surface schemes and forcings, contributing to inter-dataset variability in some regions (e.g., central U.S.). - Temporal scope: The 1980–2015 period is relatively short and influenced by internal climate variability; detected trends should not be extrapolated into the future. SESR and ASESR were detrended, but residual climate-change effects on thresholds may persist. - Spatial processing and masking: Bilinear interpolation to 0.5° may smooth fine-scale features. Arid/cold regions were masked (AI<0.2 or low PET), so results exclude some environments and emphasize areas with potential vegetative impacts. - Driver diagnostics: SPI and PET anomalies were threshold-based summaries over event windows and may not capture all relevant processes (e.g., soil/vegetation properties, advection, radiation partitioning). - Uncertainties in trends and drivers: Some domains show disagreement among reanalyses regarding trend magnitude or significance; PET and precipitation trends and their attribution remain uncertain in certain regions.