Southwestern United States drought of the 21st century presages drier conditions into the future

The study addresses why the American Southwest (California, Nevada, Arizona, Utah, New Mexico, Colorado) has experienced unprecedented drought and water stress since the start of the 21st century, and whether these conditions presage a persistently drier future. The authors aim to quantify the separate and joint roles of precipitation (moisture supply) and temperature (moisture demand) using instrumental and paleoclimate records extending to the 16th century, to test if recent trends can be explained by historical variability. They develop a new reconstruction of the Standardized Precipitation-Evaporation Index (SPEI) to evaluate drought from a climatological moisture-balance perspective and compare it with soil-moisture-oriented PMDI reconstructions. They further assess recovery probabilities from the current precipitation deficit, examine atmospheric circulation patterns associated with bidecadal droughts (16th century vs. present), and evaluate future trajectories of aridity using CMIP5/CMIP6 simulations. Given uncertain precipitation projections but robust warming, a key focus is isolating the quantitative impact of temperature on regional drought conditions and water resource recovery potential.

Previous work has documented persistent 21st-century drought conditions in the U.S. Southwest and placed them in a multi-century context using tree-ring and other paleoclimate reconstructions, indicating that recent drought severity rivals or exceeds historical megadroughts (e.g., Cook et al. 2010; Woodhouse et al. 2010; Cook et al. 2015; Ault et al. 2016; Cook et al. 2016; Williams et al. 2020, 2022). Studies have examined precipitation variability and hydroclimatology in California and the West, focusing on moisture delivery mechanics and long returns to normal (Diaz & Wahl 2015; Wahl et al. 2017). Prior findings suggest uncertain future precipitation trends regionally due to the SW’s latitudinal position between projected higher-latitude wetting and subtropical drying, while temperature increases are robust (IPCC AR6). Soil moisture-based drought reconstructions (PDSI/PMDI) and the newer Living Blended Drought Atlas provide long-term spatial context for drought severity. Recent attribution work has highlighted a substantial anthropogenic contribution to ongoing megadrought conditions, especially via warming-driven vapor pressure deficit and evaporative demand increases. These studies provide the foundation for the present work’s separate evaluation of precipitation and temperature drivers, new SPEI reconstruction, probabilistic recovery analysis, and model-based projections.

- Study region and periods: The U.S. Southwest defined as CA, NV, AZ, UT, NM, CO. Precipitation analyzed on water-year basis (Oct–Sep), temperature on calendar-year basis. - Instrumental data: Regional area-weighted series from NOAA NCEI Climate at a Glance for precipitation (WYs) and temperature (CYs). SPEI gridded (0.5°) data from CSIC were aggregated to the region and to WY means. - Paleoclimate reconstructions: Temperature from Wahl & Smerdon (5° grid, extracted cells matching the SW domain); precipitation from Wahl et al. (0.5°, closely matched to SW domain). Reconstructions were centered and scaled to be compatible with instrumental series (temperature zero-anomaly relative to 1904–1980; precipitation centered/scaled to match 1916–1977 calibration period). Expected value (EV) reconstruction time series were used. - New SPEI reconstruction: Regressed regional instrumental SPEI on instrumental temperature and precipitation (monthly SPEI aggregated to WY). High model skill (r=0.89 fitted vs. actual, SE=0.145; predictors nearly independent, r≈−0.11). Validation with reconstructed T and P as predictors against instrumental SPEI yielded r≈0.91 and near-identical behavior (r=0.97 between the two fitted series). 95% confidence intervals computed from multiple regression standard error of prediction; CIs for 21-year lowess smooths obtained by dividing annual CI by sqrt(21). The 2020 SPEI value was imputed from 2018 due to missing data. - PMDI data: Used NOAA’s Living Blended Drought Atlas (combined reconstruction + instrumental) at 0.5° resolution. For extended analysis, PMDI for the SW domain was taken back conservatively to 600 CE (grid-cell coverage >99% up to this time). PMDI smooth CIs were indirectly estimated by scaling SPEI smooth CIs to PMDI scale. - Statistical analyses: 21-year lowess smoothing to characterize bidecadal conditions; rank analyses for precipitation, temperature, and joint ranks. Probability density functions and Monte Carlo resampling: 10,000 random 20-year means for precipitation; 1,000,000 random 40-year means for temperature to evaluate extremeness of current means. Autocorrelation analyses to characterize white-noise (precipitation) vs. persistent (temperature) behavior. Standardized regression coefficients (β weights) of SPEI and PMDI vs. temperature and precipitation computed to quantify changing relative influences. - Recovery time estimation: Applied Wahl et al. method using the whiteness of precipitation to compute the length of time for cumulative precipitation to recover from the current deficit to climatological normal, considering all possible starting years (forward and backward to avoid edge effects), for recovery windows up to 75 years. - Circulation analysis: Compared boreal winter (DJF) sea-level pressure (SLP) composites for the early drought (1571–1590) using a paleoreconstruction (analog data-model assimilation) with modern reanalysis (NCEP/NCAR) for the current drought (2001–2021). Anomalies referenced to 1948–1980; RECON mean and SD matched to INST for amplitude comparability. Significance mapped (t-test) at p≤0.1/0.2. - Climate model projections: CMIP6 SSP585 (primary), plus CMIP5 RCP8.5 and RCP4.5 for comparison. One run per model to avoid ensemble-size bias. Extracted regional CY temperature, WY precipitation, and total column soil moisture (mrso). Soil moisture calibrated to instrumental SPEI (and PMDI) via mean and variance matching over 1901–2019 to yield self-consistent RECON+INST+MODEL time series. Presented medians and 20th/80th percentiles. Emphasized temperature robustness vs. precipitation uncertainty. - Assumptions and calibrations: Precipitation treated as white-noise in recovery analyses and as suggested by models; PMDI uncertainty bands approximated by rescaling SPEI smooth CIs; mean/variance matching used to bridge observational/reconstruction/model datasets.

- Current drought extremity: At the ~21-year scale, the ongoing drought is the most intense in the SW since at least 600 CE (PMDI) and since 1571 in both SPEI and PMDI reconstructions. - Temperature anomaly is non-random: The mean 1981–2020 SW temperature is ≈11 standard deviations above the Monte Carlo distribution of random 40-year means, implying p≈0 that it is a random draw from the last 500 years’ variability. - Precipitation behavior: The precipitation time series behaves as white noise with no significant autocorrelation beyond lag 0; current dryness is a rare but random realization of the long-term precipitation process. - Joint temperature–precipitation ranks: Since 2000, 5 of the 10 highest warm/dry joint-ranking years occur, highlighting the dominant role of recent warming. - Shift in relative influences: Standardized regression analyses indicate a reduction of precipitation’s influence relative to temperature by about one-third during the current drought: 34% reduction from SPEI (instrumental) and 37% from PMDI (RECON+INST), consistent with other studies’ estimates (≈33–46%). - Recovery probabilities for precipitation deficit: Estimated chances of recovery times based on historical precipitation alone are very low: ≤10 years: 2.9%; ≤15 years: 6.0%; ≤20 years: 8.4%; ≤30 years (to mid-century): 13.6%. Overall, full recovery by mid-century is unlikely; the probability within 10–15 years is about 5%. - Future projections: CMIP6/CMIP5 simulations show robust continued warming with slight or uncertain changes in precipitation. Temperature-driven increases in evaporative demand desiccate soils, yielding strongly negative expected values for SPEI/PMDI through the 21st century; 20th percentile outcomes are extremely negative and far outside the historical range. - Circulation comparisons: Wintertime (DJF) SLP ridging over the NE Pacific associated with the late-1500s drought was stronger than during the current drought, implying that even stronger circulation-driven drying remains possible; coupled with warming, such periods could produce aridity exceeding current conditions. - Water resources: Given long recovery times, persistent warming, and current demand levels, major regional reservoirs (e.g., Powell, Mead) have a very low chance of regaining full capacity in the foreseeable future.

By separating precipitation (supply) and temperature (demand) and reconstructing a meteorological drought index (SPEI) alongside PMDI, the study demonstrates that the unprecedented severity of the current bidecadal drought is driven not only by rare precipitation deficits but, critically, by anthropogenic warming that enhances evaporative demand. The precipitation series’ white-noise nature implies that sequences long enough to erase current deficits are infrequent, while warming further suppresses recovery by diminishing the effectiveness of typical moisture deliveries. Probabilistic analyses place recent temperature well outside the historical distribution, and β-weight estimates quantify a roughly one-third reduction in precipitation’s influence relative to temperature during the current drought. Climate model projections reinforce the expectation of continuing aridification via robust warming and declining soil moisture, even without strong changes in mean precipitation. Circulation comparisons suggest that more intense NE Pacific ridging than currently observed is within the historical envelope; combined with continued warming, such conditions could produce multi-decadal aridity surpassing today’s drought. These findings directly answer the research question: the 21st-century drought presages a persistently drier regime in the SW with low probabilities of rapid recovery, underscoring significant risks to water resources, ecosystems, and socio-economic systems.

The study provides a comprehensive, empirically grounded assessment that the 21st-century SW drought is the most intense ~21-year period in at least 1,400 years and that recent warming is incompatible with random historical variability. Temperature increases have reduced the relative role of precipitation in determining drought severity by roughly one-third, and recovery from current deficits is unlikely on decadal to multi-decadal horizons based on precipitation statistics alone. Together with model projections indicating continued warming and soil moisture decline, the prospects for full reservoir recovery under current demand are very low. The work integrates instrumental records, paleoreconstructions, circulation analyses, and climate model outputs to produce a self-consistent view of past, present, and likely future aridity in the SW. Potential future research directions include: improving regional precipitation projections and variability representation; refining drought metrics that integrate atmospheric demand, snowpack changes, and runoff timing; assessing the combined impacts of extreme circulation regimes and warming on multi-decadal drought risk; and exploring scenario-based implications of water demand management for reservoir recovery probabilities.

- Precipitation modeling and variability: Regional precipitation projections are uncertain and model amplitudes can differ from observations and reconstructions; results rely on the robustness of temperature projections and assume precipitation retains a white-noise character. - Reconstruction scope and uncertainty: Analyses use expected-value (EV) reconstructions; PMDI confidence intervals for smooths were indirectly estimated by scaling from SPEI, introducing approximation. Early PMDI coverage declines before 600 CE; 600 CE was chosen conservatively to ensure high grid coverage. - Statistical assumptions and sample size: Recovery analysis assumes precipitation whiteness; deviations could affect recovery probabilities. SLP composite differences are based on small samples (n=20 per period), limiting spatial extent of statistical significance. - Calibration and bridging: Mean/variance matching was used to calibrate model soil moisture to observed SPEI/PMDI, which may not capture all process differences. The 2020 instrumental SPEI value was imputed from 2018 due to missing data. - β-weight estimation constraints: For the SPEI reconstruction, constant regression coefficients over the reconstruction period mean β-weight differences reflect changing variances, potentially overestimating shifts; PMDI-based β-weights provide an independent check.