Ocean-forcing of cool season precipitation drives ongoing and future decadal drought in southwestern North America

The study addresses why the southwestern United States has experienced a prolonged megadrought since around 1998/99 and whether it is driven by natural variability, anthropogenic forcing, or both. The context is a rapidly growing population and critical agricultural water use facing unprecedented stress from reduced precipitation and warming-enhanced evaporative demand. Prior research links wet conditions in the late 20th century to a warm tropical Pacific (positive PDO) and the subsequent dryness to a cool tropical Pacific (negative PDO), with additional influence from a warm-phase Atlantic Multidecadal Oscillation (AMO). Climate models project future reductions in winter and spring precipitation in parts of the West due to radiative forcing, but coupled models have challenges reproducing realistic Pacific and Atlantic decadal variability and tropical Pacific SST trends. Therefore, the purpose is to isolate the roles of observed ocean-driven decadal variability (PDO/AMO) and forced change on cool season precipitation and summer soil moisture in the Southwest, and to assess best-, middle-, and worst-case hydroclimate outcomes through the early 2040s.

The paper situates its work within several strands of literature: (1) historical hydroclimate variability and megadroughts in the western U.S., including tree-ring reconstructions and recent attributions of warming to increased drought risk; (2) the influence of decadal ocean variability, particularly the Pacific Decadal Oscillation (PDO) and Atlantic Multidecadal Oscillation (AMO), on North American precipitation; (3) climate model projections of future drying in Mexico and the U.S. Southwest, especially spring drying driven by changes in moisture advection; and (4) limitations of coupled climate models in simulating realistic Pacific and Atlantic decadal variability and tropical Pacific SST trend patterns, including the observed strengthening of the equatorial Pacific zonal SST gradient and a lack of warming in the cold tongue that is outside the range of many models. The literature underscores that tropical Pacific variability exerts strong control on cool season precipitation teleconnections, that Atlantic variability can modulate North American hydroclimate, and that internal variability and forced change interact to drive decadal drought risk.

Design: The authors use an atmosphere–land model forced by prescribed sea surface temperatures (SSTs) to separate SST-forced variability from internal atmospheric noise and to explore near-term futures with realistic decadal ocean variability. Model: NCAR Community Atmosphere Model version 6, low-resolution configuration (CAM6-LR; 2° x 2°, 32 levels), run with '-chem none' (no prognostic aerosols), coupled to the Community Land Model (CLM) for soil moisture calculations. Standard CMIP6 forcings (trace gases, solar, stratospheric aerosols, land use/land cover) are applied to 2014, then SSP370 thereafter; non-stratospheric aerosols are climatological (1995–2005) with a bulk aerosol scheme. Observational/reanalysis data: NOAA CPC Unified precipitation, ERA5 SST, land surface air temperature, precipitation, humidity and 200 hPa geopotential heights; soil moisture from ERA5 (top 1 m) and NLDAS‑2 Noah. Period: Jan 1979–Aug 2021. Historical simulations: CAM6‑LR is forced by NCAR “blended” SSTs (HadISST1.1 + NOAA OIv2). Sixteen ensemble members start Jan 1, 1979, following a one‑year spin-up with perturbed initial conditions. Historical analyses define a late 20th-century pluvial (Dec–May 1979–1998) and a 21st-century megadrought (Jun 1998–Aug 2021). Future projections: Boundary‑forced large ensembles in which SSTs combine (a) empirically derived PDO/AMO/ENSO variability and (b) a radiatively forced SST trend, added to a 1979–2018 observed monthly SST climatology. Two grand ensembles ("All") are constructed using two different forced SST trends: (1) CMIP6-based: monthly multimodel mean SST change from 40 CMIP6 models under SSP370; (2) Hadley-based: extrapolated observed trend via regression of monthly HadISST (1900–2020) onto total GHG forcing and projected forward using SSP370-consistent forcing from MAGICC. Members also include SSP370 radiative and LU/LC forcings consistent with CMIP6, initialized as continuations of historical members. PDO/AMO/ENSO construction: A cyclostationary Linear Inverse Model (CSLIM) is trained on HadISST SST PCs (1958–2017), with the secular trend removed using the least-damped eigenmode of a stationary LIM. CSLIM generates 100 realizations of 60-year SST histories (6000 years total). The dataset is divided into overlapping 21-year chunks, from which strong PDO+ and PDO− segments (and neutral ones) are selected based on the North Pacific EOF-based PDO index, ensuring similar temporal evolution, inclusion of realistic ENSO variability, and minimal 21-year natural trends to avoid double-counting trends. AMO states are defined by North Atlantic (0–60°N) area-mean SST selection for AMO+ and AMO−, and neutral near zero. Combinations assembled: PDO+AMO+, PDO−AMO−, PDO+AMO−, PDO−AMO+, and neutral-neutral (PDOnAMOn). Each grand ensemble totals 80 simulations spanning these combinations. Region and seasons: Southwest North America defined as 25°N–40°N, 125°W–100°W, land only. Cool season defined as Dec–May (DJFMAM); warm/summer season as Jun–Aug (JJA). Future evaluation period: DJFMAM 2031–2041 and JJA 2032–2041, typically expressed relative to 1979–2021 (DJFMAM) and 1979–2020 (JJA) climatologies. Diagnostics and attribution: The ensemble mean isolates the SST-forced atmospheric response. The separate influences of PDO and AMO are isolated using linear combinations across sub-ensembles: PDO effect estimated as [(PDO−AMO+ + PDO−AMO−)/2] − [(PDO+AMO+ + PDO+AMO−)/2]; AMO effect as [(PDO+AMO+ + PDO−AMO+)/2] − [(PDO+AMO− + PDO−AMO−)/2]. Correlations, box-and-whisker distributions, and spatial composites are used to assess model–observation agreement and scenario differences. Soil moisture controls are examined via cross-member relationships among DJFMAM precipitation change, JJA soil moisture change, and JJA surface air temperature change. Comparisons: Historical model results are compared with NOAA‑CPC (precip), ERA5 (multiple fields), and NLDAS‑2 (soil moisture). Future scenarios are evaluated for best case (PDO+AMO−) and worst case (PDO−AMO+), under both CMIP6-like and Hadley-like forced SST trends, and contrasted against neutral-neutral and grand-ensemble means to isolate forced responses.

- Historical attribution: The 21st-century megadrought in the Southwest is characterized by a cool-season (Dec–May) precipitation decline linked to a shift toward cooler central–eastern tropical Pacific SSTs (La Niña–like), with associated upper-tropospheric ridging from the North Pacific across southern North America. The atmosphere-only model forced by observed SSTs reproduces the observed decadal precipitation reduction and circulation anomalies. The correlation between observed and ensemble-mean cool-season precipitation is 0.66, indicating about one-third of interannual variance is SST-forced. - Non-overlapping decadal means: Modeled cross-ensemble spreads of multidecadal average cool-season precipitation for 1979–1998 vs. 1998–2021 do not overlap, evidencing a statistically significant SST-driven decadal shift; observed values lie near the outer edges of the modeled distributions. - Soil moisture linkage: Spring and summer soil moisture declined across the region in the 21st century. Interannual-to-decadal JJA soil moisture variations track DJFMAM precipitation anomalies strongly; no clear additional offset toward drier soils independent of precipitation emerges, despite warming. NLDAS‑2 and model agree on the decadal shift (ERA5 shows a larger dry shift). - Best vs. worst near-term futures: Future cool-season precipitation (2032–2041 relative to 1979–2021) depends strongly on PDO/AMO phases. Worst-case (PDO−AMO+) yields broad drying over the Southwest; best-case (PDO+AMO−) yields potential abatement but not uniform wetting due to persistent North Pacific highs and circulation trends. The worst-minus-best contrast shows about 0.3 mm/day precipitation difference over the Southwest, a large fraction of the model’s mean cool-season precipitation (~1 mm/day). - Forced SST trend dependence: The Hadley-based forced SST trend (continuation of observed lack of equatorial Pacific cold tongue warming and stronger zonal gradient) produces stronger drying teleconnections (North Pacific high and reduced equatorial Pacific precipitation) than the CMIP6 multimodel mean trend (more El Niño–like), implying greater radiatively forced cool-season precipitation reduction if observed Pacific trends persist. - Ensemble separations: For cool-season precipitation, the driest (Hadley trend, PDO−AMO+) and wettest (CMIP6 trend, PDO+AMO−) scenario ensemble spreads do not overlap. Summer precipitation ensembles overlap, but JJA soil moisture maintains the dry/wet separation inherited from DJFMAM precipitation; spreads again do not overlap. - Probabilities of drying: In grand ensembles, 75% (CMIP6 trend) and 75–95% (Hadley trend) of members show JJA soil moisture drier than the historical average, indicating robust summer soil drying due to both reduced precipitation (especially under Hadley trend) and common warming-driven increases in evaporative demand. - No return to late-20th-century wetness: Even under best-case decadal variability, the model indicates a high probability that cool-season precipitation and summer soil moisture will not return to the late 20th-century pluvial levels by the 2030s–2040s.

The findings demonstrate that the recent and ongoing megadrought is primarily driven by ocean-forced decadal variability affecting cool-season precipitation, with the tropical Pacific playing the dominant role and the tropical Atlantic modulating the response. The model’s ability to reproduce the observed decadal shift and circulation patterns when forced by observed SSTs confirms the causal chain: cool tropical Pacific → teleconnected North Pacific/southern U.S. ridging → suppressed cool-season precipitation over the Southwest → reduced spring and summer soil moisture. For the next two decades, decadal variability in the PDO and AMO remains the leading determinant of hydroclimate outcomes: a negative PDO together with a positive AMO sustains megadrought conditions, whereas a positive PDO with a negative AMO offers the greatest relief. Radiatively forced changes add a generally drying tendency, stronger if the equatorial Pacific cold tongue does not warm (Hadley-like trend) than if models’ El Niño–like warming occurs (CMIP6-like trend). Despite this, summer soil moisture continues to be governed mainly by cool-season precipitation anomalies rather than by warming-induced evaporative demand alone; the latter adds a secondary, common drying across scenarios. The results answer the research questions by quantifying the relative roles: decadal ocean variability is dominant for cool-season precipitation and summer soil moisture in the near term, with radiatively forced drying superimposed. However, predictability of PDO/AMO on decadal horizons is limited, constraining actionable forecasts for decision-makers. The study underscores that, even under favorable ocean variability, a full return to late 20th-century wetness is unlikely in the near term.

This work shows that decadal ocean variability—especially the tropical Pacific—has driven the 21st-century decline in cool-season precipitation and consequent spring–summer soil moisture reduction in the U.S. Southwest, and will continue to dominate hydroclimate outcomes through the 2030s–2040s. Worst-case futures arise under PDO−AMO+ conditions and best-case under PDO+AMO−, with radiatively forced drying amplifying impacts if the equatorial Pacific cold tongue continues not to warm. Even with favorable variability, a return to the late 20th-century pluvial levels of winter precipitation and summer soil moisture is unlikely. Methodologically, using an atmosphere–land model forced by empirically derived PDO/AMO/ENSO variability plus alternative forced SST trends provides realistic projections relative to observed climatology. Future research should aim to: improve decadal predictability of PDO/AMO and their interbasin interactions; better constrain the forced trend in tropical Pacific SSTs (especially the cold tongue response); reduce structural uncertainties in teleconnections; and assess water resource implications under a range of management and adaptation strategies given the high likelihood of continued aridity.

- Modeling framework: Atmosphere–land model with prescribed SSTs cannot capture coupled ocean–atmosphere feedbacks; AMO/PDO are imposed rather than emergent, and within sub-ensembles AMO variability is prescribed uniformly. - Forcings and scenarios: Only SSP370 forcing pathway is used; aerosols outside the stratosphere are simplified as climatological; land–atmosphere feedbacks are represented at 2° resolution which may miss regional topographic and convective details. - SST trend uncertainty: Two forced SST trend scenarios (Hadley-regressed and CMIP6 multimodel mean) bracket possibilities but do not encompass full structural uncertainty in tropical Pacific responses (e.g., Walker circulation strength, cold tongue behavior). - Observational products: Soil moisture discrepancies between ERA5 and NLDAS‑2 indicate observational uncertainty; attribution of evaporative-demand impacts is inferred from cross-member relations rather than direct flux partition analyses. - Predictability: Decadal predictability of PDO/AMO is limited, constraining deterministic forecasts of best or worst-case states. - Reference period choices and model biases: Results depend on definitions of pluvial/megadrought periods and on CAM6‑LR mean-state and teleconnection biases common to atmosphere-only models.