Anthropogenic and atmospheric variability intensifies flash drought episodes in South Asia

Droughts are complex, multifaceted extremes driven by natural climate variability and anthropogenic influences, with substantial impacts on water, agriculture, energy, ecosystems, and society. Flash droughts (FDs) differ from conventional droughts by their sudden onset and rapid intensification over weeks to months, posing acute challenges for monitoring, early warning, and risk management. Recent severe FDs in several global hotspots have heightened interest, yet South Asia’s FD evolution characteristics and mechanisms remain insufficiently understood. Knowledge gaps include limited study of the rapid FD onset phase, the transition of FD characteristics across consecutive seasons (spring–summer), and the influence of large-scale atmospheric variability (e.g., monsoon processes, land–atmosphere coupling) on FD onset speed and intensity. This study aims to systematically investigate FD evolution in South Asia using a multivariate probabilistic framework, quantify onset speed, frequency, severity, duration, and joint return periods, and assess the roles of atmospheric circulation variability and anthropogenic climate change in altering FD risk across South Asia.

Prior studies have documented FD occurrences and drivers in global hotspots (Australia, South China, Central U.S.) and in Spain, often focusing on meteorological drivers and warm-season events in water-scarce regions. Research has shown that FDs can trigger compounding extremes (heatwaves, wildfires), and that land–atmosphere interactions and monsoon processes influence regional extremes, but their role in FD onset speed and seasonal transitions in South Asia remains underexplored. Recent works introduced multivariate FD indicators, highlighted increased evaporative demand, and showed that coupling between soil moisture and atmospheric aridity can accelerate FD onset. Studies in India suggest monsoon variability and anthropogenic warming may amplify future FD risk. However, robust quantification of FD evolution (onset speed, duration, severity) with joint probability methods and formal attribution to anthropogenic climate change over South Asia has been limited.

Study region and periods: South Asia (SA), with focus on spring–summer transition (March–August) and summer (June–August), 1979–2021.

Datasets:

• Reanalysis: ERA5 at 0.25°×0.25° for hourly root-zone soil moisture (layers 0–7 cm, 7–28 cm, 28–100 cm) and daily precipitation, 2 m air temperature, and evapotranspiration (averaged to consistent epoch, 1979–2021). Monthly ERA5 fields for circulation analysis: 500-hPa geopotential height, 850-hPa winds, 2 m temperature, mean sea level pressure, specific humidity, cloud liquid water, 500-hPa vertical velocity, and SST.
• Models for attribution: 10 CMIP6 coupled GCMs at 0.25°×0.25°, experiments: ALL (historical anthropogenic+natural forcings, 1979–2014, extended with SSP245 2015–2021) and NAT (natural-only forcings, 1979–2021). Soil moisture processed into percentiles to harmonize differing depths across models.

Flash drought (FD) identification and metrics:

• Soil moisture percentile computed per grid and pentad using 1979–2021 climatology.
• FD event defined when 5-day pentad soil moisture declines from >40th to <20th percentile with average drop rate ≥5 percentile points per pentad; termination when soil moisture rebounds to the 20th percentile. Events shorter than 15 days (3 pentads) are excluded.
• Onset speed/intensification rate: average rate of soil moisture reduction from 40th to 20th percentile considering all pentad changes between onset and termination. Spring–summer onset speed also expressed as regional mean change in soil moisture percentile from spring (Mar–May) to summer (Jun–Aug).
• FD features: frequency (percent of pentads with FD in a season), duration (average event length in days), severity (accumulated deficit of soil moisture percentile from 40th threshold until termination).

Trend and anomaly analyses:

• Modified Mann–Kendall trend tests for 1979–2021 on FD frequency, severity, duration.
• Interannual standardized anomalies of precipitation, temperature, soil moisture, and evapotranspiration over FD-affected land areas.

Joint return period (JRP) analysis:

• Marginal distributions for FD duration and severity fitted using candidate distributions (Gamma, GEV, Weibull, Normal, Log-normal, Inverse Gaussian).
• Dependence modeled using copulas (Gaussian, Frank, Gumbel); best fit selected by Akaike Information Criterion; bivariate probability estimated with AND/OR formulations; OR-type return period (T_OR) used to characterize combined severity-duration risk, with JRPs computed for spring–summer and summer separately.

Attribution analysis (Fractional Attributable Risk, FAR):

• Compute change in FD onset speed as spring minus summer soil moisture percentile for ALL and NAT ensembles at each grid cell, 1979–2021.
• FAR = (FD_ALL − FD_NAT) / FD_NAT × 100, interpreted as anthropogenic contribution to increased onset speed risk.
• Significance testing: Student’s t-test for ALL vs NAT differences (5%); country-wise distributions assessed with Wilcoxon rank-sum (median shift) and Kolmogorov–Smirnov (distribution family) tests.

Atmospheric circulation diagnostics:

• Seasonal-mean anomaly composites (relative to 1979–2021 climatology) of 500-hPa height and 850-hPa winds; 2 m temperature, MSLP, specific humidity; 500-hPa vertical velocity; cloud liquid water; and SST to diagnose synoptic-to-regional processes associated with FD evolution.

• Hotspot regions and seasonality: FD frequency is elevated in subtropical and semi-arid areas—central India, western Pakistan, and eastern Afghanistan—particularly during the spring–summer growing season (1979–2021). Simultaneous FD onsets are evident over India and Pakistan, with highest frequencies in summer.
• Trends: South-central India, western Pakistan, and eastern Afghanistan exhibit statistically significant increasing trends (p<0.05) in FD severity; some regions (Nepal, Bhutan, Bangladesh, Sri Lanka) show declining trends for certain metrics. FD frequency intensified notably during 2010–2020 across many countries.
• Durations: Mean FD durations are about 20–80 days in south-central India and 20–60 days in southern Pakistan; about 15–25 days in northern Afghanistan and 20–50 days in central Bangladesh, reflecting substantial drought intensification potential.
• Areal extent and hydroclimate anomalies: Spring–summer shows a larger median FD-affected area (~24% of SA land) than summer (~17%). FD-affected years feature warm temperature anomalies, precipitation deficits, limited soil moisture, and elevated evapotranspiration across SA; temperature and precipitation anomalies are key in rapid soil moisture declines.
• Joint return periods (JRP): Many summer FD events have JRPs <5 years and tend to be less severe and shorter than spring–summer events. Spring–summer experiences more severe and longer FDs, with the most severe events reaching 50–100-year JRPs. (Reported examples include events lasting several weeks to ~2 months with higher severity in spring–summer, and ~1 month with lower severity in summer.)
• Anthropogenic attribution: Anthropogenic climate change significantly amplifies FD onset speed in spring–summer. Median FAR values by country: Afghanistan ~60%, Pakistan ~80%, India ~90%; Nepal ~45%, Sri Lanka ~55%, Bangladesh ~65%. Differences between ALL and NAT are significant in many grid cells (5% level), with positive shifts in medians and distributions confirmed by WR and KS tests.
• Physical mechanisms: Spring–summer is characterized by a dipolar pressure pattern (positive geopotential height anomalies over western SA, negative over eastern SA), anomalous wind/pressure gradients, inhibited convection, and moisture transport disruptions, leading to precipitation deficits. In summer, low-pressure systems migrate westward with associated vertical velocity anomalies. Enhanced air temperature gradients in the Arabian Sea (linked to SST anomalies) and anticyclonic tendencies over continental SA are associated with reduced precipitation and warmer conditions that favor rapid soil moisture decline.

The study demonstrates that South Asian flash droughts preferentially occur and intensify during the spring–summer transition when land–atmosphere coupling and large-scale circulation anomalies most strongly suppress moisture transport and convection. The multivariate and bivariate analyses reveal that FD risk depends jointly on severity and duration, with spring–summer events exhibiting longer durations and higher severities than summer events. These seasonal differences align with atmospheric diagnostics showing geopotential height dipoles, anomalous wind patterns, and SST-driven temperature gradients that reduce precipitation and enhance evaporative demand.

Attribution analysis indicates a substantial anthropogenic contribution to the increased onset speed of FDs, particularly in Afghanistan, Pakistan, and India, implying that continued warming will further elevate FD risks. The results address the study’s goals by quantifying onset speed, frequency, severity, duration, and JRPs and by elucidating the atmospheric mechanisms and human influence that shape FD evolution. These findings are relevant for water resources, agriculture, and energy sectors, emphasizing the need for improved early warning, seasonal outlooks, and adaptation strategies tailored to hotspot regions and seasonal transitions.

This work provides a comprehensive multivariate characterization of flash droughts in South Asia, quantifying their onset speed, frequency, severity, duration, areal extent, and joint return periods, and linking their evolution to large-scale atmospheric variability and anthropogenic climate change. Key contributions include: (i) identification of spring–summer as the critical season for severe and long-duration FDs in central India, western Pakistan, and eastern Afghanistan; (ii) quantification of JRPs that underscore greater compound risk in spring–summer than summer; (iii) diagnosis of synoptic-to-regional atmospheric patterns that inhibit moisture transport and convection; and (iv) formal attribution demonstrating substantial anthropogenic intensification of FD onset speed across multiple countries.

Future research should: (a) incorporate vertical soil moisture profile variability and dynamics to capture depth-dependent processes; (b) advance coupled land–atmosphere modeling to better represent feedbacks during seasonal transitions; (c) refine attribution with larger ensembles and event-based frameworks; (d) integrate agricultural impacts and socio-economic exposure; and (e) develop and evaluate early warning systems leveraging multivariate indicators and subseasonal-to-seasonal predictions for hotspot regions.

• Soil moisture representation: The analysis emphasizes the temporal dynamics of the mean state of the vertical soil moisture profile; it does not fully account for variability across the vertical column, which may affect FD detection and interpretation.
• Seasonal framing: Attribution and onset speed comparisons focus on inter-scenario differences between spring and summer seasons at given locations, potentially omitting other transitional windows or intra-seasonal variability.
• Model and data uncertainties: Results depend on ERA5 reanalysis and a subset of CMIP6 models; uncertainties in hydro-meteorological fields, SST patterns, and land-surface processes may affect circulation diagnostics and attribution outcomes.
• Event definition thresholds: FD identification relies on percentile thresholds (40th to 20th) and a minimum duration (≥15 days), which, while standard, may exclude certain rapid but shorter-lived events or be sensitive to percentile calibration.
• Spatial heterogeneity: FAR spatial variability may reflect regional differences in water–energy limitations and model spread; localized processes (e.g., irrigation, land use) are not explicitly resolved.