Earth's global average near-surface air temperature has increased by 1.1 °C since the pre-industrial period due to anthropogenic forcings, leading to an increase in heat extremes regionally and globally. However, some regions show disproportionate changes; for example, the Midwest USA (MUS) has experienced a weak decreasing trend in heat extreme intensity since 1951, while Western Europe (WEU) has seen a remarkable increase of over 3 °C since 1951. The MUS exhibits a weak cooling trend in heat extremes despite moderate warming of daily average summer temperatures, a phenomenon sometimes referred to as a "warming hole." Several studies attribute this to large-scale ocean-atmospheric patterns, increased irrigation, regional reforestation, and aerosol changes. However, the extent to which these trends are due to anthropogenic forcings or unforced internal variability remains unclear. In contrast, WEU is a heatwave hotspot with a faster warming rate than many other parts of the world in the last four decades. This disproportionate warming has been attributed to a decline in aerosol forcing and cloud cover, changes in atmospheric circulation states (including increased frequency and persistence of the midlatitude jet stream), and local factors like soil moisture deficit and land-atmosphere feedback mechanisms. Unforced internal variability in the climate system can also significantly alter regional warming rates by affecting atmospheric circulation states. This study aims to understand the contrasting trends in heat extremes intensity over the MUS and WEU and their implications for the future, using observations, and Earth System Model (ESM) large ensemble simulations to examine how circulation-induced changes explain these trends. This understanding is crucial for constraining future changes in heat extremes and designing adaptation and resilience plans.

Existing literature highlights contrasting trends in heat extremes between the Midwest USA and Western Europe. The Midwest's "warming hole," characterized by a cooling trend in heat extremes despite overall warming, has been linked to various factors such as large-scale atmospheric patterns (Partridge et al., 2018), increased irrigation (Mueller et al., 2016; Nikiel & Eltahir, 2019), regional reforestation (Ellenburg et al., 2016), and aerosol changes (Meehl et al., 2012). However, the relative contributions of these factors and the role of internal variability remain debated (Banerjee et al., 2017; Fischer et al., 2021). In contrast, Western Europe's significant increase in heatwave intensity has been associated with changes in atmospheric circulation (Rousi et al., 2022), decreased aerosol forcing and cloud cover (Dong et al., 2017, 2022), and land-atmosphere feedback mechanisms (Sousa et al., 2020; Stegehuis et al., 2021). Studies emphasize the influence of internal variability on regional warming trends (Deser et al., 2012, 2014) and its stronger manifestation on shorter timescales (Fischer et al., 2013). The role of basin-scale changes in the Atlantic multidecadal variations on WEU summer climate has also been highlighted (Della-Marta et al., 2007; Sutton & Dong, 2012). These studies provide a backdrop for understanding the complexities of regional climate change and the need to disentangle the effects of different forcing mechanisms.

This study uses daily maximum temperature (TX) data from multiple observational sources (CPC, EOBS, GHCNDEX, HADEX3) and ERA5 reanalysis (1951-2021) to analyze heat extreme trends. The analysis focuses primarily on the 5-day maximum of TX (Tx5d), but also considers the 1-day maximum (TXx) and the hottest 15-day period (Tx15d) to ensure robustness. Regional averages are calculated using area-weighted averages over defined regions for the MUS and WEU. Geopotential height at 500 hPa (Z500) from ERA5 reanalysis is used as an atmospheric circulation proxy. Initial-condition large ensembles from CESM2 and a multi-model ensemble (MME) from CMIP6 are utilized to compare observed trends with model simulations under historical and SSP3-7.0 scenario forcings. The Theil-Sen slope estimator is used to calculate trends. A dynamic adjustment approach is employed to quantify the contributions of circulation- and thermodynamic-induced changes to observed trends. This involves using regularized ridge regression to establish the relationship between heat extremes and detrended Z500 (removing forced components to isolate the dynamic component). The model is trained on 2070-year pre-industrial simulations from CESM2 to avoid biases from forced warming. The dynamic component represents circulation-induced changes, while the residual component represents thermodynamic-induced changes (including external forcings and local feedbacks). This methodology allows for the separation of the influence of atmospheric circulation changes on heat extremes from other contributing factors.

Observational analysis reveals a remarkable increase in Tx5d intensity across WEU and a weak decreasing trend over much of the MUS during 1979-2021. The MUS shows a slight cooling trend in Tx5d, contrasting with warming in Tn5d (warmest nights) and daily average summer temperatures. This cooling is more pronounced in shorter-duration heat extremes (TXx). In contrast, WEU experiences a significant warming in Tx5d (>3 °C since 1979), consistent across different datasets and time scales. The observed trends in both regions deviate significantly from the 95% range of both CESM2 LE and CMIP6 MME simulations. Dynamic adjustment analysis reveals that atmospheric circulation significantly dampened Tx5d trends in the MUS (~0.2 °C/decade cooling, equivalent to ~1°C cooling per °C of global warming). This circulation-induced cooling offsets the weak thermodynamic-induced warming, resulting in the overall cooling trend. In WEU, circulation amplified warming (~0.2 °C/decade), contributing to about one-third of the total observed Tx5d trend. Thermodynamic-induced changes account for the remaining two-thirds. The observed circulation-induced trends are outside the range of climate model simulations for both regions, with MUS trends at the very low end and WEU trends at the very high end. Even after accounting for circulation-induced effects, the model simulations still show discrepancies with observed thermodynamic-induced trends, particularly in the MUS.

The study's findings highlight the crucial role of atmospheric circulation in shaping regional trends of heat extremes. In the MUS, circulation-induced cooling counteracts thermodynamic warming, leading to an overall weak cooling trend in Tx5d. This suggests that factors such as land use changes, irrigation, and aerosol forcing, which primarily affect short-duration heat extremes, might play a more significant role than previously thought. In WEU, circulation amplifies warming, exacerbating the effects of thermodynamic warming. This emphasizes the importance of considering both thermodynamic and dynamic factors when assessing regional heat extreme trends. The discrepancies between observed and simulated trends, even after accounting for circulation effects, point towards limitations in climate models' representation of regional processes and internal variability, underscoring the need for improved model parameterizations. The inability of models to fully capture the observed trends, even after accounting for circulation, suggests the influence of factors not fully represented in the models.

This study demonstrates that atmospheric circulation plays a crucial role in modulating heat extreme trends, damping them in the MUS and amplifying them in WEU. The observed circulation-induced trends in both regions fall outside the range of climate model simulations, indicating potential limitations in current models' representation of regional processes and internal variability. Future research should focus on improving model representation of regional factors, such as land use changes and irrigation in the MUS, and further investigating the mechanisms driving the observed circulation changes in both regions to better constrain future projections and inform adaptation strategies.

The study acknowledges several limitations. The dynamic adjustment method may not fully capture all circulation-induced trends, potentially underestimating or overestimating the influence of internal variability. The method's reliance on residual components to estimate thermodynamic trends may lead to misinterpretations if other unexplained components exist. The focus on two specific regions limits the generalizability of the findings, and some processes, like irrigation and land use, are not precisely quantified in the model. Furthermore, disentangling the forced and unforced components of circulation changes remains challenging.