Evidence of anthropogenic impacts on global drought frequency, duration, and intensity

The study investigates whether and how anthropogenic climate forcing has altered key characteristics of meteorological droughts globally, addressing a gap in detection and attribution literature that has largely focused on temperature and precipitation extremes rather than drought features. Droughts have wide-ranging ecological, agricultural, energy, and hazard-cascading impacts, and observations have shown increasing trends in drought frequency, duration, and severity in several regions since the mid-20th century. With projections indicating further increases in drought occurrence and severity across many continents, quantifying anthropogenic influences on drought frequency, maximum duration, and intensity is critical. The authors aim to detect and attribute changes in these drought features between the late 19th and late 20th centuries and to distinguish the roles of greenhouse gases versus anthropogenic aerosols, including how altered atmospheric evaporative demand affects net water availability.

Prior studies have attributed changes in temperature and precipitation extremes to human influence, often using historical simulations with and without anthropogenic forcings to estimate fractions of attributable risk. Observational records (e.g., GPCC and CRU) indicate positive trends in meteorological drought characteristics and land area under drought. While hydrological drought in the western U.S. has been linked to anthropogenic warming through snow-related processes, meteorological drought attribution has been less explicit. Tree-ring-based drought atlases and modeling studies (e.g., Marvel et al.) suggest increasing human influence on global soil moisture drought, with aerosol emissions implicated in mid-20th-century regional drying (e.g., the Sahel) and weakened monsoons (e.g., South Asia). Multivariate fingerprinting (Bonfils et al.) shows joint human-driven changes in temperature, rainfall, and aridity. Overall, literature highlights complex, regionally varying drought responses to combined greenhouse gases and aerosols, motivating a comprehensive, feature-based attribution of meteorological drought.

Data: Monthly CMIP6 simulations (nine models, three realizations each) for four scenarios: historical (all major anthropogenic and natural forcings), historical natural-only (natural forcings only), GHG-only (greenhouse gases only), and AER-only (anthropogenic aerosols only). Model outputs were regridded to a common 1° grid using nearest-neighbor interpolation, with land areas between 60°S and 60°N analyzed.

Drought indices and definitions: A non-parametric standardized precipitation index (SPI) was generated following Farahmand and AghaKouchak (2015). For each pixel, 6-month precipitation totals (rolling window) were ranked against the historical natural-only climatology for the corresponding month to obtain empirical probabilities, which were then transformed to a standard normal variate to yield SPI values. Drought events were defined as 6-month SPI dips below −1.5. Event features extracted included: frequency (count of events), maximum event duration, and maximum event intensity (magnitude of the minimum SPI during an event). Features were computed for two periods (1851–1900 and 1956–2005), and shifts were calculated as differences between late 20th and late 19th centuries.

Attribution metrics: Probability Ratio (PR) maps quantified the likelihood of drought occurrences in forced scenarios relative to natural-only conditions during 1956–2005. PR > 1 indicates increased drought likelihood under the forced scenario. PR was computed for historical, GHG-only, and AER-only SPI-based droughts.

Dependence analysis: Using IPCC AR5 regions, non-consecutive drought events (SPI < −1.5) from 1850–2005 were used to construct bivariate kernel density estimates (KDEs) of event duration versus median intensity. Differences between historical and natural-only joint distributions were tested with the 2D, two-sample Kolmogorov–Smirnov test (Fasano–Franceschini), with most regions significant at alpha = 0.05 except the Tibetan Plateau.

Net water availability: To account for atmospheric evaporative demand, standardized precipitation-evapotranspiration index (SPEI) was computed analogously to SPI using precipitation minus potential evapotranspiration (PET), enabling assessment of drought features and PR under historical, GHG-only, and AER-only scenarios. PET formulation follows established SPEI methodologies; potential evapotranspiration responds strongly to temperature and thus to GHG forcing.

Statistical treatment and significance: Spatially aggregated distributions (PDFs) of shifts across land pixels (60°S–60°N) were compared between historical and natural-only scenarios for SPI and SPEI features using two-sample t-tests. Stippling on maps denotes pixels with statistically significant increases in the respective feature per methods described. Analyses minimized short-term internal variability by contrasting multi-decade periods and focused on differences across experiments to reduce bias impacts.

• Anthropogenic forcing has significantly increased meteorological drought frequency, maximum duration, and maximum intensity (SPI-based) across large parts of the Americas, Mediterranean/southern Europe, western and southern Africa, and eastern Asia when comparing 1956–2005 to 1851–1900. Historical natural-only simulations showed no coherent regional increases and a slight global shift toward wetter conditions.
• Spatially aggregated land distributions (60°S–60°N) display stronger upper tails and greater variability for historical versus natural-only conditions for all three SPI features, indicating widespread anthropogenic-driven increases.
• Probability Ratios (PR) for SPI drought occurrences (threshold −1.5, 6-month): historical PR > 1 in many regions, with statistically significant increases in parts of southern/eastern Europe, northern/western Africa, India, Central and South America. GHG-only forcing elevates drought likelihood in southern Europe, northern and southern Africa, Central America, and parts of South America, aligning with projected 21st-century precipitation changes. AER-only forcing produces distinct patterns, with significant roles in South and East Asia, Central America, the Sahel, and sub-Arctic margins, consistent with aerosol-driven monsoon weakening and Sahel drying reported in prior studies.
• Joint behavior: Bivariate distributions of event duration and median intensity (SPI) differ significantly between historical and natural-only in most IPCC regions (except the Tibetan Plateau). Regions like Central America/Mexico and the Mediterranean show concurrent shifts toward longer and more intense events, with notable increases in the upper tail of intensity.
• Net water availability (SPEI): Historical and GHG-only forcings yield even larger increases in drought probability ratios globally, especially in tropical and subtropical bands, indicating strong PET-driven drying under GHG influence. AER-only generally reduces SPEI-based drought PR in these bands, but greenhouse gas effects dominate globally. SPEI-based shifts in frequency, maximum duration, and maximum intensity mirror SPI but with stronger anthropogenic signals, highlighting enhanced evaporative demand under warming.

The findings directly address the research question by demonstrating that anthropogenic forcing has measurably intensified multiple interrelated drought characteristics globally. Historical simulations with human influence show significant increases in drought frequency, duration, and intensity relative to natural-only variability, confirming that observed late 20th-century drying patterns in several regions are unlikely to be due to internal variability alone. Separating forcings clarifies mechanisms: greenhouse gases drive widespread increases in drought occurrence through thermodynamic (increased atmospheric water vapor, intensification of moisture transport, wet-get-wetter/dry-get-drier tendencies), and dynamic responses (poleward Hadley cell expansion), while anthropogenic aerosols exert strong regional effects, particularly weakening Northern Hemisphere monsoons and contributing to Sahel drying. Incorporating evaporative demand via SPEI reveals that greenhouse gases substantially increase drought likelihood beyond precipitation-driven metrics, underscoring the role of rising temperatures and PET in reducing net water availability. These results are relevant for risk assessment and water resource planning, indicating regions of heightened sensitivity to anthropogenic climate change and suggesting that GHG-driven drying signals will persist and dominate over aerosol influences in the 21st century.

This study provides comprehensive, model-based evidence that anthropogenic forcing has increased the frequency, maximum duration, and maximum intensity of meteorological droughts across many regions worldwide. By leveraging CMIP6 experiments and standardized indices (SPI, SPEI), the work disentangles the relative roles of greenhouse gases and aerosols, showing greenhouse gases as the dominant driver of increased drought likelihood and reduced net water availability due to enhanced atmospheric demand. The bivariate analysis further indicates significant shifts in the joint distribution of drought duration and intensity across most regions. Future research should extend attribution to other drought types (hydrological, agricultural), refine representation of aerosol–cloud–radiation interactions and vegetation responses, evaluate sensitivity to different drought indices and PET formulations, and reduce model biases (e.g., SST and ITCZ biases) to improve regional fidelity and risk projections.

• Model biases: CMIP6 models retain SST biases affecting ENSO simulation and teleconnections, and a persistent double-ITCZ, contributing to precipitation biases and potential regional drought simulation errors.
• Aerosol processes: Limited understanding and representation of aerosol feedbacks may hinder accurate replication of real-world aerosol-driven drought changes.
• Land–atmosphere and vegetation responses: Models may underrepresent vegetation and land-surface responses to elevated CO2 and warming, which can affect regional hydroclimate.
• Bias correction and methodological sensitivities: Detection of trends can be sensitive to model uncertainty, bias correction choices, and drought metric selection.
• Scope: Focus is on meteorological drought (SPI/SPEI) and two multi-decade periods; other drought types may respond differently to forcings. While differences between experiments and multi-decade contrasts reduce some biases and variability, uncertainties remain in regional attribution and magnitude estimates.