Introduction

The North Atlantic Warming Hole (NAWH) is a region south of Greenland where the sea surface temperature (SST) has been cooling over the instrumental era. While the global ocean surface has warmed by about 0.55 \(^\circ\)C over the last century, unequivocally driven by anthropogenic radiative forcing (IPCC AR6 SPM), the NAWH region has cooled by about 0.75 \(^\circ\)C since 18701. This North Atlantic cooling trend is also a feature of historical simulations and future projections of coupled models participating in the Coupled Model Intercomparison Project phase six (CMIP6)2 and earlier generations of climate models, suggesting that the NAWH is a robust response to external forcing such as greenhouse gas emissions.

The NAWH has garnered significant research interest in terms of mechanisms of its formation and its potential impacts. Previously proposed mechanisms for NAWH development fall under two categories: ones that argue that the cooling trend is primarily ocean driven, and those with a role for both the ocean and the atmosphere (either coupled or contributing independently). There has been extensive evidence to support a buoyancy-driven argument via the slowing down of the Atlantic Meridional Overturning Circulation (AMOC)1,3,4,5,6,7, which results in local cooling as less warm water is transported northward. Indeed, the SST fingerprint of an AMOC decline includes a cooling subpolar North Atlantic3,8; however, it is difficult to establish long-term trends in AMOC strength due to a short observational record and lack of measurements spanning different latitudes9. While a recent study by Fan et al. supports a weakening AMOC contributing to the NAWH’s formation, it also cautions overestimating AMOC’s role and suggests air–sea interaction and changes in the subpolar gyre playing more dominant roles10. Keil et al. also suggests that the slowing of AMOC interacts with a shift of the subpolar gyre to impact high-latitude ocean heat transport. This idea is augmented by Gervais et al. and Keil et al., suggesting a role for local deep mixing. In addition, Keil et al. and Hu and Fedorov found that the NAWH was present even in model simulations that did not have a coupled ocean, suggesting that the atmosphere may play a role in the formation of the NAWH via shortwave cloud feedback, Rossby wave connections to the Indian Ocean, and an increase in local storminess as possible sources of causation11,12,13. Ghosh et al.14 provides evidence for an alternative hypothesis for NAWH intensification involving the strengthening of the subpolar gyre (SPG) through a density driven mechanism. That study argues that salinity reduction at the eastern side of the SPG, which may be driven by a weakening AMOC9,15,43,17, causes an intensification in the SPG which in turn intensifies the NAWH. While the experimentation and result of Ghosh et al. is quite different than our current study, the observed increased westerlies over the NAWH are consistent. Ghosh et al. attributes the wind-driven changes to internal variability; however, our study provides another possible source of the increased westerlies that may bolster in combination or replace the attribution to internal variability altogether.

Many previous studies seek to understand how midlatitude SST anomalies, such as the NAWH, can affect atmospheric circulation, as well as the sensitivity to the anomaly’s location18. The response to such usually falls under two categories, one consisting of a baroclinic linear direct response19,20, and another consisting of an equivalent barotropic nonlinear response18,21,19,20,21,22,26. The nonlinear barotropic response is dominated by the transient eddy forced response via an enhanced SST gradient27,25,29. It has been found that the transient eddy forced response appears similar to the dominate internal variability present25 which in the North Atlantic is the NAO.

On a hemispheric scale, the NAWH may modulate the strength of the North Atlantic Oscillation (NAO) which may impact precipitation over Europe1,27,30,28,29,33. Previous studies have also suggested a teleconnection between the NAWH and the tropical Indian Ocean, with ramifications for the Asian-Australian winter monsoon season1,12,34. Diagnosing the local atmospheric response of the NAWH is a first step toward understanding its maintenance and broader climatic implications.

While other studies have proposed potential causes and remote impacts of the NAWH and its relation to anthropogenic warming, we explore the atmosphere’s fundamental response to the NAWH, including how potential coupled feedbacks may allow the NAWH to persist throughout the observational record. Previously, Karnauskas et al. investigated the fundamental response to the NAWH in an older and lower resolution atmospheric general circulation model (AGCM). Karnauskas et al. found a negative phase NAO-like tendency (Fig. 1a) response over the North Atlantic as well as a negative feedback mechanism locally over the NAWH. This result directly contradicts the findings of Gervais et al. and He et al. which identify a positive NAO-like response to the NAWH with a higher resolution model hierarchy35,36. The objective of this study is to characterize and diagnose the regional atmospheric response to the NAWH in an updated, higher-resolution model, including how it may depend upon the NAWH’s latitude. We run idealized simulations using a state-of-the-art AGCM, differing only in terms of whether the NAWH is present in the boundary (SST) forcing, and where that forcing is located. As this approach is similar to that of Karnauskas et al., but with a different and significantly more modern model with a higher resolution, we then take the opportunity to explore the extent to which results may be sensitive to model resolution.

Fig. 1
figure 1

Direct DJF surface response comparison between Karnauskas et al. using the (a) ECHAM4.6 model (2.8\(^\circ\)) and the (b) CESM2 model (1\(^\circ\)). Also included are (c) the ERA5 observed DJF in units per 83 years with (b) the positive SST trends omitted. Sea level pressure (contoured every 0.25 mb, zero omitted, negative dashed) and surface wind vector anomalies are shown (> 0.5 m/s) and scaled by 5, as well as the SST forcing (\(^\circ\)C) used in each experiment (colors).

Full size image

Results

Simulated atmospheric response to the NAWH

A pair of high and low sea level pressure (SLP) anomalies over the North Atlantic of ± 1.5 mb develops in response to the NAWH surface forcing (Figs. 1b, 2a,c), which is also evident aloft (Fig. 2e). These signals are statistically significant at the 95% confidence level (Supplemental Fig. 5). Accompanying these changes in the pressure field are positive zonal wind anomalies-approximately 1.5 m/s near the surface (Figs. 1b, 2c) and up to 3 m/s aloft (Figs. 2e, 3a,c). In response to the NAWH, the jet stream becomes extended; the contour of 25 m/s reaches further to the east by \(\sim\)8\(^\circ\) (Fig. 2e). This further supports the Gervais et al. result that the vertically (Fig. 3a) and horizontally propagating baroclinic eddy energy extends the North Atlantic jet eastward (Gervais et al. 2019). The maximum surface westerly anomaly occurs near the southern edge of the surface low increasing by \(\sim\)1.5 m/s (Fig. 2c).

Fig. 2
figure 2

(a,b) SST forcing (contoured every 0.2 \(^\circ\)C, zero omitted, negative dashed). Response of zonal wind speed (m/s) to the NAWH and the NAWH-S (c,d) at the surface and (e,f) at the level of the jet stream, calculated relative to CTRL. Thick contours define a threshold of zonal wind speed \(\le\) 25 m/s representing the location of the North Atlantic jet stream (CTRL-black, NAWH-white). Response of (c,d) sea level pressure (contoured every 0.25 mb, zero omitted, negative dashed) and (e) geopotential height (contoured every 5 m, zero omitted, yellow solid and negative green), (f) as in (e) but contoured every 3 m.

Full size image
Fig. 3
figure 3

(a,b) Vertical cross-section of the DJF response of zonal wind (m/s), and the position of the control jet steam (contoured every 5 m/s, zero omitted). SST forcing along 320\(^\circ\)W associated with (c) NAWH and (d) the NAWH-S, indicating the location of the prescribed SST anomalies relative to the location of the zonal wind features shown above.

Full size image

Interestingly, the response of SLP to NAWH forcing projects strongly onto the positive phase of the NAO (Supplemental Fig. 2)31,29,30,34. We note that this response is nearly the opposite of Karnauskas et al., who found a negative NAO-like response to the NAWH in a different model (Fig. 1a), but consistent with Gervais et al. The ERA5 trend result is similar to the model response, with a low SLP anomaly to the north, a high SLP anomaly to the south, and consistent with geostrophic balance, accelerated westerlies directly over the NAWH (Fig. 1b,c). The main difference is a subtle tilt of the SLP anomalies. Because the real world as represented in ERA5 data has both positive and negative SST anomalies present in the North Atlantic basin, as well as a plethora of other influences on the regional climate that are subject to natural decadal variability, such a close match between observed trends and our simulated response was not anticipated. The physical mechanisms and potential feedbacks associated with the emergent features in our CESM2 experiment are described in Gervais et al. and He et al. and will be outlined in the following sections, with additional insights arising from consideration of model resolution, vertical stability, and proximity of the NAWH to the jet stream.

Simulated atmospheric response to the NAWH-S

Had the NAWH developed at a location considerably southward (Fig. 2b) from its observed location (Fig. 2a), would the atmospheric response be shifted as well? Surprisingly, a southward-shifted NAWH (NAWH-S) produces a pair of SLP anomalies in almost the exact same location over the North Atlantic region as the original NAWH experiment, but with an amplitude about 25% weaker (Fig. 2d). Similarly, the pair of upper-level pressure anomalies are evident in the same configuration, yet slightly weaker (Fig. 2f). In accordance with geostrophic balance, similar positive zonal wind anomalies arise from the NAWH-S in the same position as the NAWH response, but again, with a smaller magnitude-reduced by \(\sim\)50% (Figs. 2d,f, 3b,d). In response to the NAWH-S, the jet stream is extended via the enhanced westerlies, but the 25 m/s contour extends only \(\sim\)5\(^\circ\) further to the east (Fig. 2f). Note that the NAWH-S elicits a similar pressure and zonal wind response in the region, but the local response directly above the NAWH-S is different, consisting of anomalous easterlies, about − 0.5 m/s (Fig. 3b,d). It appears from this sensitivity experiment that the structure and dynamics of the NAO constrain the response of the basin-scale atmosphere to persistent SST forcing resembling the NAWH, even if its position is varied by thousands of kilometers.

Feedbacks and role of resolution

The NAWH’s consistent presence over time raises the question of whether local mechanisms may be reinforcing the signal. A positive feedback mechanism by way of the transient eddy forced response as described in Gervais et al. provides a framework for understanding the implications of the response in the CESM2 experiments presented in this study, which have a considerably higher resolution than those of Karnauskas et al. The transient eddy forced response involves enhanced westerlies leading to the development of small-scale transient eddies in the midlatitude jet stream that can further strengthen the westerlies, and so on. This nonlinear barotropic response can be seen qualitatively via cross section with evidence of a downward mixing of westerly momentum from the level of the jet stream (Fig. 3a,c). Because the presence of the NAWH-S still enhances the westerlies at the height of the jet stream via the regional response, a weaker transient eddy forced response is still seen that propagates down to the surface likely through vertical momentum mixing (Fig. 3b). Once the jet stream and surface westerlies become enhanced over the NAWH, the increase in wind speed will amplify sensible and latent heat flux to further the cooling effect36.

Karnauskas et al. used an identical experimental design and similarly constructed albeit much older AGCM (the Max Planck Institute ECHAM4.6 model37) and with significantly lower resolution (2.8\(^\circ\)). Likely due to the lower resolution, that simulation did not adequately resolve the small-scale transient eddy mean feedback response dominating the response in the 1\(^\circ\) CESM2 model. As in Karnauskas et al., without consideration of eddies, changes in vertical stability would indeed have dominated the response (Fig. 4). As there is colder SST in the NAWH region, there is a more stable boundary layer with less implied vertical mixing and downward entrainment of westward momentum from the westerlies aloft38,39. A deceleration in the surface westerlies through this process yields a pair of opposing SLP anomalies and a decrease in latent heat flux ultimately weakening the NAWH-hence the sign of the feedback is opposite (negative) in lower resolution models such as that used in Karnauskas et al.

Fig. 4
figure 4

Layer-mean virtual temperature (DJF) in the lower, middle, and upper troposphere over (a) the observed NAWH (55\(^\circ\)N, 320\(^\circ\)W) and (b) the NAWH-S (35\(^\circ\)N, 320\(^\circ\)W) relative to the control experiment. The profiles relative to control are calculated by NAWH-CTRL (black), NAWH-S—CTRL (blue), and those from ECHAM as in Karnauskas et al. (red).

Full size image

Examination of vertical stability (by way of \(\Delta \bar{T}_v\) profiles) from all experiments at two points over the NAWH and NAWH-S allows a direct comparison of stability when both proximity to the jet stream and model resolution are factors (Fig. 4). The observed NAWH (55\(^\circ\)N, 320\(^\circ\)W) is in close proximity to the jet stream, while the NAWH-S (35\(^\circ\)N, 320\(^\circ\)W) is not directly influenced by the jet. The \(\Delta \bar{T}_v\) profiles over the observed NAWH from low- and high-resolution models have an opposite vertical stratification response over the cold SST forcing explaining the opposite wind response between the two studies (Fig. 4). When comparing the low resolution \(\Delta \bar{T}_v\) profile (red) over the observed NAWH with the high resolution NAWH-S (blue) \(\Delta \bar{T}_v\) profile over the southern point, the results appear similar and indicate a more stable boundary layer as expected (Fig. 4). This suggests that when there is no jet stream to interact with, the atmospheric response may not be resolution dependent; however, when the jet stream does directly interact with the NAWH, higher resolution is required to capture the transient eddy forced response.

Discussion

The results of our NAWH experiment are consistent with Gervais et al. and He et al., while the NAWH-S experiment validates that even an increased distance from the jet stream is insufficient to change the regime response to the one simulated by the lower resolution model. We have also reconciled the differences with the results of Karnauskas et al. by considering stability and the role of resolution in the delicate balance between mechanisms. However, for the NAWH to be sustained by the transient eddy feedback, it must be located where the surface wind response is one that reinforces the cooling by way of increased wind speed and turbulent heat fluxes. If a cold SST anomaly were to develop to the south (within 20\(^\circ\)), the jet stream would still be enhanced, but the local wind response over that feature would be anomalous easterlies. The enhanced jet stream may still produce the transient eddy response to such an SST anomaly in the subtropics and, in principle, may even form a new warming hole beneath the midlatitude jet stream, while over time, such a subtropical anomaly would not be aided by a local positive feedback and would likely dissipate.

To address the caveat of our experiments being run at different resolutions as well as with different models, we include low- and high-resolution experiments with CESM1 (Supplemental Fig. 3). The results indicate that there may be model dependencies at lower resolutions, but not at higher resolution, as the higher-resolution CESM1 and the CESM2 results are very similar.

Conclusions

This paper examines the regional boreal wintertime atmospheric response to the NAWH as observed since 1870. The response of the surface atmospheric circulation projects strongly onto the positive phase of the NAO (Supplemental Fig. 2), and its relative insensitivity to location within \(\sim\)20\(^\circ\) latitude underscores the importance of the midlatitude dynamics constraining the NAO. The agreement between our surface response to those of He et al. and Gervais et al., coupled with a positive feedback being a likely mechanism for a persistent NAWH, also underscores the necessity of higher resolution models in this region to correctly balance competing mechanisms for air–sea coupling.

While the regional atmospheric response in the North Atlantic is not critically dependent on the NAWH’s location (at least to within 20\(^\circ\) latitude), the persistence of a warming hole and its maintenance through interaction with the atmosphere is dependent on its proximity to the midlatitude jet stream. Our experimental design, consistent with global climate model simulations of future global warming scenarios, suggests that it is possible for the NAWH to persist longer than might be expected based solely upon the initiating mechanism (whether it be an AMOC decline or otherwise), and in turn modulate the midlatitude circulation across the North Atlantic and downstream across the Northern Hemisphere midlatitudes in the 21st century. While these processes may be occurring in fully coupled models, quantifying their contributions, and identifying the role of resolution may be challenging due to the lack of controlling for other factors that govern jet stream changes, boundary layer stability, and NAO-like patterns. This calls for a better understanding of both the future response of the NAWH to global warming and future slowdown of the AMOC, as well as a spotlight on the climate impacts associated with the NAWH for the Eurasian continent.

Methods

We conduct our experiments with the National Center for Atmospheric Research (NCAR) Community Earth System Model version 2.0 (CESM2) specifically in atmosphere-only mode with the Community Atmosphere Model version 6 (CAM640) at approximately 1\(^\circ\)horizontal resolution. The Hadley Centre Sea Ice and Sea Surface Temperature data set (HadISST41) is used as the present-day climatology for SST. Note that this is the uncoupled version of CESM2; SSTs are prescribed. The first experiment (CTRL) is forced by the monthly SST climatology, averaged from 1995–2005, with all other sources of forcing including CO2 concentration held constant at the year 2000 level. The second experiment (NAWH) is conducted identically, but with the cooling trend defining the NAWH added to modern climatology. In the third experiment (NAWH-S), our application of the cold anomaly defining the NAWH is shifted 20\(^\circ\) southward in order to investigate whether the local and regional atmospheric responses to the warming hole are critically sensitive to its latitude. Experiments are run for 45 years; the first five years are removed to allow the model atmosphere to reach equilibrium. We isolate the atmospheric response to NAWH and the southward shifted NAWH (NAWH-S) by subtracting time-averaged outputs of the control experiment from those of the NAWH experiments, i.e. (NAWH - CTRL) and (NAWH-S - CTRL) respectively. These experiments follow the methodology of Karnauskas et al., allowing for direct comparison (e.g., Figs. 1 and 4). Karnauskas et al. utilized the Max Planck Institute (MPI) ECHAM4.6 model37 containing a Gaussian grid at a T42 spectral horizontal resolution (corresponding to approximately 2.8\(^\circ\) grid resolution) with 19 vertical levels.

As in Karnauskas et al., the NAWH forcing is obtained by calculating the linear SST trend using HadISST (in its original 1\(^\circ\) resolution) from 1870–2019 and removing the basin-wide median trend (0.49 \(^\circ\)C per century, calculated from 55\(^\circ\)S–60\(^\circ\)N). A rectangular domain centered on the prominent cooling trend in the North Atlantic is defined, in which all grid cells with SST trend \(\le\) − 0.25 \(^\circ\)C per century are retained. The anomaly field within the box is then linearly interpolated to the CESM2 model grid (0.9x1.25). As all external radiative forcing is held constant at 2000-year levels, and SST fields are simply repeating climatology, any interannual variability arising in the model solutions are the result of internal atmospheric noise. To ensure that no changes in sea ice boundary conditions occur due to the placement of the SST anomaly field, changes were only applied south of 60\(^\circ\)N accompanied by a liner tapering across 60\(^\circ\)N to smooth any sharp artificial gradients in SST forcing. The NAWH-S forcing is constructed identically to the NAWH, but simply displaced 20\(^\circ\) southward from the observed NAWH position.

Seasonal means are calculated over the 40 years of analyzed output from December through February (DJF) and from June through August (JJA). We focus primarily on DJF due to the robust seasonality of the jet stream and associated midlatitude atmospheric circulation during boreal winter. A response arises in these variables during all seasons, but the response is strongest in the DJF season (also noted by Gervais et al. and Karnauskas et al. To characterize the upper-level atmospheric circulation including the midlatitude jet stream, the 250 mb pressure surface is analyzed.

To approximate atmospheric stability, layer-mean virtual temperature is calculated for the lower (1000–850 mb), middle (850–500 mb), and upper troposphere (500–200 mb) using the hypsometric equation (Supplemental Fig. 1):

$$\begin{aligned} h = z_2 - z_1 = \frac{R \cdot \bar{T}_v}{g} \ln {\left( \frac{p_1}{p_2} \right) } \end{aligned}$$
(1)
$$\begin{aligned} \bar{T}_v = \frac{(z_2 - z_1) \cdot g}{R \cdot \ln {\left( \frac{p_1}{p_2} \right) }} \end{aligned}$$
(2)

where h is the thickness of the layer, z is geopotential height, R is the specific gas constant for dry air, \(\bar{T}_v\) is the layer-mean virtual temperature, g is gravitational acceleration, and p is pressure. Virtual temperatures (Equation 2) are calculated for the DJF season directly over the observed NAWH (55\(^\circ\)N, 320\(^\circ\)W) and over the NAWH-S (35\(^\circ\)N, 320\(^\circ\)W) to consider the role of stability in governing the response to SST forcing throughout all model experiments, including the lower-resolution results of Karnauskas et al.

To contextualize our simulated response(s) of the North Atlantic atmosphere to the NAWH, we use the 0.25\(^\circ\)-resolution ERA5 reanalysis42. ERA5 is produced by European Centre for Medium-Range Weather Forecasts (ECMWF) and assimilates diverse observations from satellites and ground stations along into a numerical weather prediction (NWP) model. Since the cold SST trend forcing our model was over a longer period than the ERA5 reanalysis period, we used the entire length of this observed data set, 1940–2023. Consistent with our model forcing, we examine linear trends in SST, SLP and near-surface winds for the DJF seasonal means.