Circulation dampened heat extremes intensification over the Midwest USA and amplified over Western Europe

Global near-surface air temperature increased by about 1.1 °C since pre-industrial due to anthropogenic forcing, intensifying heat extremes worldwide. Yet, regional disparities exist: the Midwest USA (MUS) shows a weak decreasing trend in the intensity of daytime heat extremes since 1951 despite moderate summer mean warming (a “warming hole”), whereas Western Europe (WEU) exhibits a remarkable >3 °C increase in heat extreme intensity since 1951, outpacing global-mean change. The reasons for these opposite regional trends remain unclear. Internal variability can strongly modulate regional warming rates, especially on shorter time scales that characterize heat extremes, and external forcings and local feedbacks can alter atmospheric circulation states. This study investigates why observed heat extreme intensity trends are unusually low over the MUS and unusually high over WEU and explores implications for the future. Using observations of circulation and temperature together with Earth System Model large ensembles, the authors disentangle circulation-induced and thermodynamic (forced) contributions to observed trends via a dynamical adjustment approach.

Prior work attributes the MUS warming hole and cooling of heat extremes to large-scale ocean–atmosphere patterns, increased irrigation and cropland intensification enhancing evapotranspiration-driven cooling, regional reforestation, and aerosol changes. Considerable uncertainty remains regarding the roles of anthropogenic forcings versus unforced internal variability over the MUS and North America. In WEU, identified as a heatwave hotspot, rapid warming and intensifying extremes have been linked to declining aerosol forcing and cloud cover, changes in atmospheric circulation including increased frequency and persistence of midlatitude jet stream states and blocking anticyclones, and land–atmosphere feedbacks associated with soil moisture deficits. Internal variability, including Atlantic multidecadal variations, can substantially alter regional warming rates and circulation. These mechanisms broadly influence regional extremes by modifying circulation regimes, teleconnections, and land–atmosphere interactions.

Data: Daily maximum temperature (TX) from multiple observational/reanalysis sources were used: CPC (1979–2021), EOBS (1951–2020; Europe only), GHCNDEX (1951–2021), HADEX3 (1951–2018), and ERA5 (1951–2021). Geopotential height at 500 hPa (Z500) from ERA5 served as the circulation proxy. Heat extreme indices include TXx (annual maximum daily TX), Tx5d (annual hottest 5-day mean of TX), Tx15d (annual hottest 15-day mean of TX), and Tn5d (annual hottest five nights; mean of daily minimum temperature). Regional domains: MUS following Melillo et al. (2014); WEU defined as −10° to +15°E, 36–60°N. Circulation domains: MUS Z500 over 116–60°W, 16–68°N; WEU Z500 over 30°W–35°E, 16–80°N. Trends were computed with Theil–Sen’s slope estimator and expressed both per decade and per °C of global-mean surface temperature (GMST) change. Modeling: Initial-condition large ensemble from CESM2 (89 members) and CMIP6 multi-model ensembles were analyzed under historical forcings (1951–2014) and SSP3-7.0 (2015–2021). A 2070-year CESM2 pre-industrial control simulation was used to train the statistical model. Dynamical adjustment: Z500 was detrended by subtracting the daily global-mean Z500 from each grid cell prior to averaging over extreme days to remove forced thermal expansion. A regularized ridge regression f was trained (k-fold cross-validation) on pre-industrial CESM2 to predict heat extremes from detrended Z500: T = f(Z500). Applying f to observed/reanalysis Z500 yields the circulation-driven (dynamic) component of heat extremes; the residual is interpreted as the thermodynamic (externally forced and other non-circulation) component, acknowledging that residuals can also include local feedbacks and any circulation effects not captured. Uncertainties from internal variability were assessed using spread across ensemble members. Spatial maps and regional averages were analyzed; comparisons to model ranges used the 5th–95th percentile distributions.

- Contrasting observed trends (1979–2021): WEU Tx5d warmed by >3 °C (~4 °C per 1 °C GMST); MUS Tx5d shows a slight cooling, with strongest cooling over western/southern MUS. Shorter-duration extremes (TXx) in MUS cooled by ~0.7 °C over four decades (~−1 °C per 1 °C GMST), whereas Tx15d exhibits very weak warming. - Nighttime extremes: WEU Tn5d warmed at ~0.4 °C/decade since 1979; MUS Tn5d weakly warmed. Diurnal range increased in WEU (Tx5d > Tn5d trend) and decreased in MUS (Tx5d cooling with slight Tn5d warming). - Model–observation mismatch in MUS: Observed regional-average Tx5d trend lies outside the 95% range of both CESM2 LE and CMIP6 MME. No CESM2 member simulates MUS cooling; ensemble means show warming of ~3 °C (CESM2) and ~2 °C (CMIP6) since 1979, vs observed ~−0.25 °C. When normalized by GMST, MUS observed Tx5d trends are below the 5th percentile of CMIP6 and below all CESM2 members. - WEU trends vs models: Observed Tx5d trends fall within but at the very high end of CESM2 and CMIP6 ranges; when normalized by GMST, observed WEU Tx5d trends exceed the 95th percentile in both ensembles. Tn5d trends over WEU fall within ensemble ranges. - Circulation contributions: In MUS, circulation induced ~−0.2 °C/decade cooling in Tx5d (~−1 °C per 1 °C GMST), sufficient to offset weak thermodynamic warming, yielding overall cooling. In WEU, circulation added ~+0.2 °C/decade warming (~+1 °C per 1 °C GMST), accounting for about one-third of the observed Tx5d trend; thermodynamic contributions account for about two-thirds (>0.5 °C/decade, ~3 °C per 1 °C GMST), especially strong over central WEU (France, Germany, northern Italy, UK parts). - Consistency across metrics: For MUS, Tx15d shows stronger thermodynamic warming and stronger circulation-induced cooling relative to Tx5d; for Tn5d, circulation reduces trends while thermodynamic warming is larger than for Tx5d. WEU shows similar partitioning, with circulation contributing an even larger fraction for Tn5d (~half). - Internal variability: Large spread in ensemble trends, particularly over MUS (uncertainties about twice those over WEU in CESM2), underscores limited predictability of regional extreme heat trends. Globally, most regions fall within model 95% ranges, yet ~40% (CMIP6) and ~28% (CESM2) of land area fall outside. - Spatial patterns: Circulation-induced cooling in MUS is strongest in the lower/southern MUS; WEU circulation-induced warming peaks over Germany, France, and parts of the UK.

The analyses address why MUS and WEU exhibit opposite trends in heat extreme intensity. Over MUS, circulation changes have dampened or reversed thermodynamic warming in daytime extremes, producing slight cooling despite global warming and moderate mean summer warming. Over WEU, both thermodynamic forcing and circulation changes amplify warming, placing observed trends at the high end of model ranges. The findings imply that unusual circulation trends have strongly modulated recent extreme heat trends in both regions. However, whether these circulation-induced trends reflect externally forced circulation responses or unforced internal variability remains unresolved. This distinction is pivotal for future projections: if externally forced, WEU may continue to see amplified extreme heat trends and MUS trends could remain damped; if internal, these trends may reverse in coming decades, moderating WEU increases and enhancing MUS warming. The partial reconciliation achieved by removing circulation effects indicates that models may misrepresent some forcings (e.g., aerosols, irrigation, land-use/management) or overestimate thermodynamic responses, and that the statistical adjustment may conservatively estimate circulation impacts. Overall, the results stress the need for improved representation and attribution of circulation changes and local forcings to constrain regional projections and guide adaptation planning.

The study shows that atmospheric circulation has been a key modulator of extreme heat trends since 1979, damping or reversing daytime heat extreme trends over the Midwest USA (~−0.2 °C/decade; ~−1 °C per 1 °C GMST) and amplifying trends over Western Europe (~+0.2 °C/decade; ~+1 °C per 1 °C GMST). In WEU, thermodynamic forcing accounts for roughly two-thirds of the observed warming, with circulation adding about one-third, leading to observed trends near or above the upper bound of model ranges. In MUS, both circulation- and thermodynamic-induced trends sit at the low end of model ranges, yielding observed cooling outside model envelopes. These insights highlight the central role of circulation in shaping regional extremes and the need to determine whether observed circulation trends are forced or internal. Future work should pursue process-based attribution of circulation responses (e.g., targeted model experiments, storyline approaches, and refined dynamical adjustment), and improve model representation of aerosols, irrigation, land use, and land–atmosphere feedbacks. Better constraints on forced versus internal components will improve reliability of regional projections and inform adaptation and resilience strategies in both regions.

- Methodological: The thermodynamic component is inferred as a residual after removing circulation effects via statistical dynamical adjustment, potentially misattributing lagged or unresolved circulation influences to thermodynamics. The approach likely provides a conservative estimate of circulation-induced trends, especially for short-duration extremes with strong land–atmosphere feedbacks. Detrending Z500 by global-mean removal may inadvertently remove parts of internally driven circulation signals or over-remove forced components, biasing partitioning. - Modeling/data: Models may inadequately represent key forcings and processes (e.g., aerosols, irrigation, land-use/management, soil moisture feedbacks), and may overestimate thermodynamic warming. Internal variability induces large irreducible uncertainty in regional trends, especially over MUS, limiting predictability. Even after accounting for circulation, observed MUS thermodynamic trends remain at the low end of model ranges, indicating unresolved factors. - Scope: Focus on two regions with unusual trends; although illustrative, they are not the most extreme globally. Some localized regions elsewhere may show stronger decreases but lack spatial coherence.