Relative contributions of local heat storage and ocean heat transport to cold‐season Arctic Ocean surface energy fluxes in CMIP6 models

The Arctic near‐surface air temperature increases most strongly during the cold season, and ocean heat storage has often been cited as a crucial component in linking the ice‐albedo radiative feedback, which is active in summer, and near‐surface air temperature increase in winter, when the lapse rate feedback contributes to Arctic warming. Here, we first estimate how much local heat storage and ocean heat transport contribute to net surface energy fluxes on a seasonal scale in CMIP6 models. We then compare contributions in a base state under weak anthropogenic forcing to a near‐present‐day state in which significant Arctic amplification is simulated. Our analysis indicates that, in a few regions, ocean heat transport plays a larger role for cold‐season net surface energy fluxes compared with local heat storage. Analyzing differences between past and near‐present‐day conditions suggests that the lapse rate feedback, which mainly acts during the cold season in warm water inflow regions, may be more strongly influenced than previously thought by increased ocean heat transport from lower latitudes.


INTRODUCTION
In an early global climate model study, Manabe and Wetherald (1975) found that surface air temperature at high latitudes responds more strongly to greenhouse gas warming compared with the global mean surface temperature. Since then, Arctic amplification (Holland and Bitz, 2003;Serreze and Francis, 2006;Pithan and Mauritsen, 2014;Previdi et al., 2021) has clearly emerged in the surface air temperature record (England et al., 2021;Chylek et al., 2022). During recent decades, Arctic surface air temperature has increased several times faster than the global mean temperature (Chylek et al., 2022), and Arctic amplification has by now been recognized as a robust feature in climate model simulations and observations (England et al., 2021). Evidence of Arctic amplification on Quaternary time scales has also emerged (Miller et al., 2010;Park et al., 2019). Arctic amplification has often been attributed either to local feedbacks or to a combination of heat transport and local feedbacks, often with an emphasis on local feedbacks. In particular, the absence of efficient mixing of air at the surface and the sea ice retreat are understood to contribute to Arctic amplification (e.g. Manabe and Wetherald, 1975;Pithan and Mauritsen, 2014;Boeke et al., 2021).
Effects of ocean heat transport on Arctic sea ice (see e.g. review by  have been investigated among others by Tsubouchi et al. (2020), , and Decuypère et al. (2022), and several studies suggested that ocean heat transport shapes polar climate change (Goosse et al., 2018). Several studies (Koenigk and Brodeau, 2013;van der Linden et al., 2019;Beer et al., 2020) have also highlighted the role of ocean heat transport for Arctic amplification. The overall effect of ocean heat transport on Arctic amplification was, however, found to be small by Pithan and Mauritsen (2014) in a study that focused on radiative feedbacks. Instead, Pithan and Mauritsen (2014) point to the ocean as an important heat reservoir. The role of the Arctic Ocean as an important heat reservoir has recently also been highlighted in other studies on Arctic amplification (Dai et al., 2019;Boeke et al., 2021;Chemke et al., 2021;Chung et al., 2021;Dai, 2021;Jenkins and Dai, 2021;Liang et al., 2022). In particular, the observation that Arctic amplification is strongest during the cold season whereas the sea-ice albedo feedback is active during summer points to ocean heat storage as a potentially important contributor to Arctic amplification. Several studies suggested that ocean heat release in winter helps to explain why Arctic amplification and the lapse rate feedback are strongest in winter (Pithan and Mauritsen, 2014;Boeke et al., 2021;Chung et al., 2021;Jenkins and Dai, 2021). The Arctic lapse rate feedback strongly depends on surface temperature changes (Boeke et al., 2021;Jenkins and Dai, 2021) and arguably acts even when only the surface air temperature changes while the tropospheric temperature above the air near the surface remains constant (Salzmann, 2017).
Although it is widely understood that the surface albedo feedback (due to sea ice changes) and the lapse rate feedback (due to a lack of efficient vertical mixing in the atmosphere) are important for Arctic amplification, and there is evidence that the lapse rate feedback is affected by ocean heat release in winter (Boeke et al., 2021;Chung et al., 2021;Jenkins and Dai, 2021), when solar absorption at the surface ceases, and that ocean heat transport is correlated with sea ice fraction (Årthun et al., 2012), other aspects are less well understood. In particular, it is still unclear how much ocean heat transport contributes to ocean heat release in winter compared with ocean heat storage. On a more fundamental level, it is still unclear, whether changes in ocean heat transport can be represented in terms of a closed feedback loop or should be classified as important drivers of polar climate change that cannot be expressed within a feedback framework (Goosse et al., 2018). If the main role of the ocean in climate change were an increase of local seasonal heat storage and release, a local feedback framework (e.g. Pithan and Mauritsen, 2014;Chung et al., 2021) would seem appropriate as far as the lapse rate feedback is concerned.
Although ocean heat release in winter is thought to play an important role for Arctic amplification, and although numerous previous studies have investigated heat transport to the Arctic, the relative importance of heat storage and heat transport for the cold-season surface energy release in warm water inflow regions has not been widely addressed. Consequently, it is still unclear whether the cold-season lapse rate feedback is indeed mainly related to local ocean heat storage (Pithan and Mauritsen, 2014;Boeke et al., 2021;Chung et al., 2021;Jenkins and Dai, 2021) and how strongly ocean heat transport from lower latitudes modulates this important feedback, as suggested by the geographical pattern of the lapse rate feedback (figure 1g of Boeke et al., 2021), which tends to be strongest in warm water inflow regions, which are most strongly affected by changes in ocean heat release (Metzner et al., 2020). Here, we investigate the relative contributions of ocean heat storage and ocean heat transport to winter surface energy fluxes based on several coupled climate models that participated in the Coupled Model Intercomparison Project Phase 6 (CMIP6, Eyring et al., 2016). The first goal is to understand whether local heat storage or ocean heat transport is the main contributor to net positive upward surface energy fluxes during Arctic winter in these models, or whether both contribute equally. The second goal is to understand how these relative contributions change in a climate warming scenario and thus to help clarify whether ocean heat transport can be considered an important driver of Arctic climate change or whether a local feedback framework is more appropriate.
For historical simulations, several coupled climate models that participated in CMIP6 simulate Arctic amplification that is consistent with observation and reanalysis datasets (Davy et al., 2018;Ye and Messori, 2021;Chylek et al., 2022), although the ensemble spread tends to be large even for individual ensemble runs from the same model, and single-model ensemble averages tend to differ between models (e.g. Ye and Messori, 2021). As expected in the presence of internal variability, depending on the analysis period, the multi-model ensemble mean either agrees with or deviates from the observations due to internal variability (Chylek et al., 2022). For a hypothetical perfect model, this is expected because internal variability affects observations but is averaged out when averaging over a sufficient number of model runs. Regarding heat transport to the Arctic, Madonna and Sandø (2021) found that CMIP6 models yielded results that are closer to observations compared with models that took part in the Coupled Model Intercomparison Project Phase 5.
In this study, we first analyze monthly mean climate model output data to estimate the relative contributions of ocean heat storage and ocean heat transport to net surface energy fluxes on a seasonal scale. We then analyze changes of these contributions, comparing contributions in a base state under weak anthropogenic forcing to a state in which significant Arctic amplification is simulated.

CMIP6 data
We analyzed data from several global climate models that participated in CMIP6 (CMIP6, Eyring et al., 2016). In the CMIP6 historical simulations, emissions of anthropogenic greenhouse gases and anthropogenic aerosol precursors are prescribed based on estimates of past emissions for the years 1850 to 2014. Effects of volcanic aerosols and solar variability are also taken into account. In the ssp245 scenario, anthropogenic emissions for the years 2014 to 2100 are assumed to follow a medium pathway, which sets this scenario apart from other high-and low-emission scenarios. Eyring et al. (2016) provide an overview of the CMIP6 model experiment set-up. Table 1 provides an overview of the model runs that were used in this study. We selected the models for which output data for at least three realizations for the ssp245 scenario including the corresponding historical runs were available. When computing a multimodel mean, we first combined the realizations for each model and then averaged, so that each model result is weighted identically independent of the number of realizations. We did not, however, combine results from similar models prior to computing the multimodel mean. For one of the models (UKESM1-0-LL; see Table 1), results were provided by two different groups. These results were combined on the grounds that both groups ran the same model.

Computation of monthly mean surface energy fluxes
The local net upward energy flux at the surface F net was computed as where R = LW emi − SW abs − LW abs is the net radiative upward flux at the surface, LHF is the upward latent heat flux, and SHF is the upward sensible heat flux. LW emi is emitted terrestrial (long-wave) radiation, SW abs is absorbed solar (short-wave) radiation, and LW abs is absorbed terrestrial radiation at the surface. SW abs represents the difference between incoming and outgoing (reflected) solar radiation at the surface.

2.2.2
Multiyear averaged annual cycle Because we are interested in the annual cycle of surface energy fluxes, we computed multiyear mean annual cycles of surface energy fluxes from monthly mean data starting with March instead of January and ending with February instead of December. This reduces the length of a given multiyear time series by 12 months. The idea behind this is to reduce potential effects from mismatches at the beginning and end of the time series. When starting the averaging in December, there are always two Northern Hemisphere winter months at the beginning of the time series that do not correspond to any spring or summer months included in the time series. On the other hand, only one winter month at the end of a conventional multiyear time series corresponds to one full summer and fall. A mismatch between the winter months at the beginning of a time series and the winter month at the end, which may potentially influence multiyear averages, can then arise because of internal variability and/or long-term trends. Our decision to start time series analysis with March and end with February does not completely avoid the potential problem of a mismatch. But it is expected to reduce the potential impacts of remaining mismatches.

Decomposition of surface energy fluxes
The multiyear mean monthly mean net upward surface energy flux can be either positive or negative. A positive upward flux indicates that, in the monthly mean, energy is transferred from the surface toward the atmosphere. A negative upward flux indicates that the surface receives energy from above.
Based on the sign of the monthly mean net upward surface energy flux computed from the CMIP6 model output, we artificially decompose the multiyear mean monthly mean net upward energy flux into contributions from months in which, on average, the surface emits energy toward the atmosphere (here, F pos ) and months in which, on average, the surface receives energy (here, F neg ): where F pos (F neg ) is the monthly mean net upward energy flux at the surface during a month with a monthly mean positive (negative) net upward energy flux at the surface.
We first computed the time mean annual cycle for each model run as described in Section 2.2.2. Then, we averaged the annual cycles from the different realizations to obtain the ensemble mean annual cycle for each model. The ensemble mean annual cycle was then decomposed into positive and negative contributions. In order to account for the variation of the number of days in a month, we weighted the monthly mean surface energy fluxes by the number of days in the specific month divided by one-twelfth of the number of days in the specific year. Multimodel means were computed by averaging over the results from this analysis.

2.2.4
Estimating contributions of local heat storage and other processes In order to estimate the contribution of processes other than local heat storage to the seasonal surface energy flux from multiyear data containing complete 12-month cycles, we compute the ratio where |F x | denotes the absolute of F x . The choice of the denominator in Equation (2) ensures that the ratios that will be shown in the maps (i.e., at grid scale) vary between −1 and +1. This denominator equals twice the seasonal ocean heat uptake plus ocean heat transport. For r = 0, the annual mean local net upward surface energy flux is zero and local heat storage dominates. For r = −1 or r = 1, one can assume that local heat storage plays only a minor role. The dominant process for r = −1, r = 0, and r = 1 One caveat of this interpretation is that r is computed from monthly mean data. It should therefore be regarded as a first-order estimate. On the one hand, since we are interested in the seasonal cycle, some filtering of high-frequency variability seems desirable. On the other hand, using high-frequency output data and an integral-preserving filter for smoothing would potentially lead to more exact results. However, in addition to limited data availability for daily data, this would still leave open the question of what exactly the filter time-scale should be, and it seems unlikely that the additional effort would result in fundamentally different results.
In order to determine in which direction r changes, we compute the difference: where r 2 is r for March 2000 to February 2050 and r 1 is r for March 1850 to February 1900. For computing r 2 , we use ssp245 runs (starting with the year 2015) together with the corresponding historical runs (up to the year 2014). By corresponding run, we mean the run that provides the initial conditions; that is, the "parent" run. For computing r 1 , we take into account only the historical runs that served as a starting point for the ssp245 runs. For each ssp245 run, we combine the ssp245 run and the corresponding historical run into one time series. We then analyze time slices from these combined time series. For regions with positive r, a positive d implies an increased contribution of ocean heat transport to winter surface upward energy fluxes relative to local heat storage.

Definition of subregions
We define Arctic Ocean subregions based on a mask that was originally devised for the "Sea Ice Back To 1850" dataset (Walsh et al., 2017). The regions are shown in Figure 1. Figure 2 shows multiyear mean net upward surface energy fluxes from the CMIP6 models listed in Table 1 for historical model runs. Positive net upward surface energy fluxes indicate that, on average, the surface emits energy and acts to heat the atmosphere. In the absence of a significant heat source inside the Arctic Ocean, positive net upward surface energy fluxes over a multiyear period can only be sustained by ocean heat transport. Another potential candidate would be the freezing of liquid water, during which heat is released. Melting sea ice, on the other hand, requires energy. Negative upward (i.e., positive downward) surface energy fluxes indicate that, on average, energy is taken up by the ocean or sea ice at the surface.

Arctic net upward surface energy fluxes in CMIP6 models
Negative multiyear mean net upward surface energy fluxes in Figure 2 Figure 1. Based on CMIP6 historical simulations the southern Labrador Sea, much of Hudson Bay, and in many of the models in parts of the Beaufort Sea and the East Siberian Sea as well. Long-term average net transfer of energy via the atmosphere to the Arctic Ocean is found mainly at higher latitudes outside the main warm water inflow in the marginal seas.
The net surface energy fluxes in the central Arctic, the Beaufort Sea, the Chukchi Sea, and the East Siberian Sea in Figure 2 are generally small. Especially for the central Arctic, this is expected due to sea ice cover persisting throughout the year in the historical runs. Most models compute slightly positive net upward surface energy fluxes over the central Arctic which may, for example, result from heat conduction through the sea ice, from leads, which are usually parametrized in global climate models, or, depending on the region, also from intermittent sea-ice-free conditions, in combination with ocean heat transport. Only the two Centre National de Recherches Météorologiques (CNRM) models (Figure 2e,f) show regions of slightly negative net upward surface energy fluxes extending into the central Arctic.
At depth, inflow of warm Atlantic water affects the entire Arctic Ocean (e.g. Carmack et al., 2015). At the surface, the main signatures of warm Atlantic water are found in the Barents Sea, where warm water subsides and then further circulates across almost the entire Arctic Ocean. In most of the Greenland Sea and in southern parts of the Labrador Sea, where cold water from the Arctic Ocean flows southward and subsides below warmer Atlantic water, on average, the ocean acts to heat the atmosphere as well. This is evidenced by positive net upward surface energy fluxes in Figure 2. Warm water from the Pacific enters from the Bering Sea into the Chukchi Sea, resulting in average positive net upward surface energy fluxes in the Bering Sea and the Chukchi Sea.

3.2
Contributions of ocean heat storage and ocean heat transport to surface energy fluxes on a seasonal scale in CMIP6 models by region

3.2.1
Example: Barents Sea in the CanESM5 model We now decompose the monthly mean data into data from months with positive and months with negative net upward surface energy fluxes. An example for the  Figure 3b. If we assume that the seasonal components must balance in the multiyear mean, in the absence of a heat source inside the ocean, this multiyear average net upward surface energy flux must be supplied by ocean heat transport. Figure 3b therefore suggests that the contribution from ocean heat transport to the sum of the monthly mean positive fluxes is larger than the contribution from local ocean heat uptake during summer. In this particular example for the Barents Sea, ocean heat storage appears to account for less than half of the positive net upward surface energy flux from September to March.

Multimodel spread
Before analyzing multimodel means in the next section, we investigate the spread between different models in this section. In Figure 4, we focus on the Barents Sea, the Beaufort Sea, the central Arctic, and the Chukchi Sea. Similar plots for the remaining regions are shown in Supporting Information Figures S1 and S2.
In the Barents Sea (Figure 4a), ocean heat transport contributes more to the sum of the positive net upward surface energy fluxes than seasonal ocean heat storage in all the model simulations except the Goddard Institute for Space Studies (GISS) model family and KACE-1-0-G. On the whole, the results for net upward surface energy fluxes in the Barents Sea are qualitatively similar in the various model simulations, except that the net fluxes in the GISS model family and to some extent also KACE-1-0-G are smaller. The large multiyear net upward surface energy fluxes in the majority of the models confirm the importance of warm water inflow in the Barents Sea in the model simulations.
The Beaufort Sea is characterized by very small net surface energy fluxes and a moderate annual cycle of net surface energy fluxes in all the models (Figure 4b). Throughout most of the historical simulations, the Beaufort Gyre is characterized by particularly thick sea ice, which limits surface energy fluxes. Monthly mean positive and negative net upward surface energy fluxes almost balance each other in all the models. On the whole, the results are qualitatively similar across the models.
In the central Arctic (Figure 4c), the amplitude of the seasonal cycle of the net upward surface energy fluxes is again much smaller than in the Barents Sea (Figure 4a). Net surface energy fluxes per unit area are also smaller in magnitude than in the Barents Sea, which is in line with a larger annual mean sea-ice cover. Ocean heat transport from lower latitudes tends to play a lesser role in the central Arctic relative to local heat storage. Yet, in most models, ocean heat transport plays a non-negligible role for the decomposed positive net upward surface energy fluxes even in the central Arctic. An exception are the two CNRM models that showed negative net upward surface In these models, regional compensation results in small area-average net upward surface energy fluxes over the central Arctic. Figure 4d for the Chukchi Sea shows again non-negligible long-term net surface energy fluxes, which are consistent with inflow of warmer water, in this case especially from the Pacific through the Bering Strait. However, unlike in the Barents Sea, heat storage contributes more to the monthly mean positive net upward surface energy fluxes compared with ocean heat transport. Though the Arctic Ocean is comparatively well connected with the Atlantic, the Bering Strait is relatively shallow and relatively narrow.  Figure 5, one finds large contributions from ocean heat transport in each of them. The Barents Sea and the Greenland Sea stand out in that these are the only two regions in which the net upward surface energy flux is larger than the absolute of the sum of the monthly mean negative upward surface energy fluxes. This speaks to the importance of ocean heat transport for supporting the large positive upward surface energy fluxes in these regions. As one moves along the main import pathways of warm water from the Atlantic and the Pacific, ocean heat transport becomes less important relative to seasonal heat storage.
Our simple method to estimate the contribution of ocean heat transport to seasonal net upward surface energy fluxes yields values above 20% in the majority of the marginal seas. The Beaufort Sea and the East Siberian Sea are exceptions. Especially the Beaufort Sea away from the coast is characterized by robust sea-ice coverage. The Beaufort Sea and the East Siberian Sea are both outside the main inflow of warm Atlantic water. No notable net enhancement of simulated annual mean surface energy fluxes due to the influence of ocean heat transport is found in the Canadian Archipelago and Hudson Bay.

Ratio of non-local to local contributions
In order to provide some rough visual guidance to the qualitative importance of non-local versus local contributions to seasonal net surface energy fluxes, we computed the ratio r of net upward surface energy flux to the sum over the monthly mean positive and the absolute of the monthly mean negative net upward surface energy fluxes as defined in Equation (2). The pattern of r in Figure 6 is largely in line with the discussion of the results from the preceding analysis of regional net upward surface energy fluxes.
Warm water inflow regions show fairly large positive r, indicating a prominent contribution of ocean heat transport to the monthly mean net upward surface energy fluxes that occur in the colder part of the year. Negative r implies that net energy flux via the atmosphere to the ocean and sea ice plays a role. Small r means that local heat storage is large compared with the net upward surface energy flux. The main finding in Figure 6 is the large regions in which r is not overly small.
For much of the Arctic, r is positive, because the net surface energy flux in the numerator is positive, indicating that in the multiyear mean the energy is transferred from the ocean and sea ice toward the atmosphere. In warm water inflow regions, the contribution of ocean heat transport to monthly mean positive net upward surface energy fluxes can be of similar magnitude to the seasonal heat storage, as indicated by the preceding regional analysis and reflected in the darker red colors in Figure 6. In these regions, r would be even larger if only positive net upward surface energy fluxes were included in the denominator in Equation (2). Instead, the denominator contains the net flux and twice the (balanced) contribution due to seasonal heat storage, as defined by the sums of the absolute values of the positive and the negative monthly mean net upward surface energy fluxes. Figure 6 also shows limited regional compensation occurring within the Baffin Bay, Labrador Sea, and Gulf of St Lawrence (BLL) region. This indicates that the preceding analysis of regional net upward surface energy fluxes for this particular region would have benefited from a region mask in which the BLL region is further subdivided into individual regions. The main disadvantage of r is that a large part of the information in the previous sections is lost when focusing exclusively on this ratio.

Comparison between the past and near-present-day
Comparing differences of the ratio r between the past and near-present-day conditions allows us to qualitatively analyze whether seasonal heat storage has become more or less important relative to non-local processes in a warmer climate with significant Arctic amplification. Figure 7 shows the difference d between r for March 2000 to February 2050 and r for March 1850 to February 1900, before Arctic amplification started to emerge. Red colors are found in regions that are strongly influenced by ocean heat transport and in which annually averaged positive net upward surface energy fluxes are typical (cf. Figure 2). This indicates that in a near-present-day climate, annual mean net upward surface energy fluxes, which are driven by ocean heat transport, take a more prominent role relative to seasonal heat storage in these regions. Blue colors indicate that the net upward surface energy fluxes have decreased and/or the seasonal heat storage contributes more strongly to the denominator. In either case, the contribution of ocean heat transport to the net surface fluxes has decreased compared with either seasonal heat storage and/or energy transfer via the atmosphere to the ocean.
Blue colors in the central Arctic Figure 7 are largely compatible with increased downward surface energy fluxes during summer and increased seasonal ocean heat storage. However, Figure 7 suggests that, for many of the models, ocean heat transport plays an increasingly important role in parts of the marginal seas.

CONCLUSION
We analyzed the relative contributions of seasonal heat storage and ocean heat transport to net upward surface energy fluxes in the Arctic Ocean in CMIP6 models. Our analysis was based on a simple decomposition of net upward surface energy fluxes into contributions from months with positive and months with negative net upward surface energy fluxes and on the argument that, in the absence of significant local heat sources in the Arctic Ocean, long-term annual mean net upward energy fluxes can only be explained by ocean heat transport. Our analysis suggests that, in the CMIP6 models, on average, ocean heat transport contributes more strongly to seasonal positive net upward surface energy fluxes than seasonal heat storage in the Greenland Sea and the Barents Sea. Ocean heat transport was found to play an important role in other regions that are directly influenced by warm water inflow from higher latitudes as well, most notably the Bering Sea and the BLL region. Based on our simple method, we estimate the contribution of ocean heat transport to seasonal net upward surface energy fluxes to be above 20% in the majority of the marginal seas. A non-negligible contribution of ocean heat transport to seasonal net upward surface energy fluxes was also found in the central Arctic. In the East Siberian Sea, the Canadian Archipelago, Hudson Bay, and the Beaufort Sea the contribution of ocean heat transport to seasonal net upward surface energy fluxes was found to be very small in the multimodel average.
We also analyzed changes of the contributions of seasonal heat storage and ocean heat transport to net upward surface energy fluxes in the Arctic Ocean in CMIP6 models in a near-present-day climate compared with a reference period in which Arctic amplification had not yet emerged. We found a decreasing contribution of ocean heat transport to the net upward surface energy fluxes in the central Arctic, which is in line with increased downward surface energy fluxes during summer and increased seasonal ocean heat storage. However, in parts of the marginal seas, ocean heat transport plays an increasingly important role relative to seasonal ocean heat storage for positive net upward surface energy fluxes during the colder seasons.
As noted in Section 1, several previous studies have pointed out that ocean heat release in winter plays an important role for Arctic amplification and specifically also for the lapse rate feedback. A number of studies have, however, attributed this increased ocean heat release in winter mainly to increased ocean heat storage, whereas other studies have also highlighted the role of ocean heat transport. Our goal, therefore, was to compare the contributions of ocean heat storage and ocean heat transport to surface energy fluxes in winter in CMIP6 models, in part because such models are frequently being used in studies of Arctic amplification.
Our analysis indicates that not only increased ocean heat storage constitutes a quantitatively relevant contribution to increased surface warming in winter. In some warm water inflow regions, ocean heat transport is more important for winter surface energy fluxes than ocean heat storage, and for some regions winter ocean heat transport increases even more than ocean heat storage. Based on this outcome, together with the previous findings that ocean heat release in winter plays an important role for Arctic amplification (Dai et al., 2019;Boeke et al., 2021;Chemke et al., 2021;Chung et al., 2021;Dai, 2021;Jenkins and Dai, 2021;Liang et al., 2022), affects the lapse rate feedback (Pithan and Mauritsen, 2014;Boeke et al., 2021;Jenkins and Dai, 2021), and that the lapse rate feedback is especially strong in warm water inflow regions during the cold season (Boeke et al., 2021), we conclude that both increased heat storage and ocean heat transport contribute to the lapse rate feedback and Arctic amplification. This finding is of direct relevance to answering the question as to whether changes in ocean heat transport can be represented in terms of a closed feedback loop or should be classified as important drivers of polar climate change that cannot be expressed within a feedback framework (Goosse et al., 2018). However, because atmospheric heat transport tends to increase whenever ocean heat transport decreases, one can still find Arctic amplification in the absence of ocean heat transport; for example, in coupled models in which an atmosphere model is coupled to a slab-ocean model (Chemke et al., 2021). Polar amplification in models occurs even when placing a continent with a flat surface in the southern high latitudes, preventing not only ocean heat transport but also ocean heat storage (Salzmann, 2017;Hahn et al., 2020;Singh and Polvani, 2020).