Effects of Mid‐Latitude Oceanic Fronts on the Middle Atmosphere Through Upward Propagating Atmospheric Waves

The impact of mid‐latitude oceanic frontal zones with sharp meridional sea‐surface temperature (SST) gradients on the middle atmosphere circulation during austral winter is investigated by comparing two idealized experiments with a high‐top gravity wave (GW) permitting general circulation model. Control run is performed with realistic frontal SST gradients, which are artificially smoothed in no‐front run. The control run simulates active baroclinic waves and GW generation around the mid‐latitude SST front, with GWs propagating into the stratosphere and mesosphere. In the no‐front run, by contrast, baroclinic‐wave activity is significantly suppressed, and GWs with smaller amplitude are excited and then dissipated at higher altitudes in the mesosphere. Westward wave forcing in the winter hemisphere was more pronounced in the control run up to ∼0.03 hPa, resulting in a more realistic reproduction of the middle atmospheric polar vortex. The results demonstrate the importance of realistic mid‐latitude ocean conditions for simulating the middle atmosphere circulation.


Introduction
Climatological-mean mid-latitude sea-surface temperature (SST) field is characterized by "oceanic frontal zones" where meridional SST gradients are pronounced due to confluent warm and cool ocean currents.Along the frontal zones, near-surface baroclinicity remains high through its efficient restoration via vigorous air-sea heat exchanges (Hotta & Nakamura, 2011;Sampe et al., 2010).In combination with moisture supply from warm currents, the mid-latitude SST fronts are favorable for recurrent baroclinic development of synoptic-scale cyclones and anticyclones, thus acting to anchor major storm-tracks (e.g., Nakamura et al., 2004).Nakamura et al. (2008) and Sampe et al. (2010Sampe et al. ( , 2013) ) investigated the importance of midlatitude SST fronts in the tropospheric circulation by comparing two idealized experiments with an "aqua-planet" configuration of an Atmospheric General Circulation Model (AGCM).One of them was conducted with SST profile with realistic frontal gradients, while the other with artificially smoothed SST profile that severely suppressed mid-latitude frontal SST gradients.In the former experiment a storm-track with vigorous synoptic-scale eddies is anchored around the SST fronts and an eddydriven polar-front jet (PFJ) is collocated.Conversely, their AGCM experiment without frontal SST gradients results in notable weakening in climatological-mean mid-latitude eddy activity and PFJ, as well as its annular variability.Minobe et al. (2008) examined the Gulf Stream's influence on the troposphere, based on operational analyses, satellite observations and AGCM simulations.They found that within the marine boundary layer, atmospheric pressure adjustment to frontal SST gradients leads to the time-mean surface wind convergence and thereby the formation of a well-defined precipitation band along the Gulf Stream.Within this precipitation band, upward motion and cloud systems are found to extend into the upper troposphere, as verified through comparison between two AGCM experiments with and without realistic frontal SST gradients across the Gulf Stream.Other studies have also suggested fundamental importance of SST frontal zones in the Southern Oceans in the formation of the Southern Hemisphere (SH) storm-track and eddy-driven PFJ as well as their dominant variability as the SH annular mode (e.g., Ogawa et al., 2016) and related troposphere-stratosphere coupled variability that induced ozone-hole influence on the troposphere (Ogawa et al., 2015).
As mentioned above, most of the previous studies have concentrated on the impact of the mid-latitude SST fronts on the troposphere (e.g., Seo et al., 2023).As synoptic-scale baroclinic waves and precipitation are modulated by the SST fronts, atmospheric gravity waves (GWs) should also be affected, because they are among the main midlatitude sources of GWs (e.g., Hertzog et al., 2008;Jia et al., 2014;Sato et al., 2009).Since GWs can transport both eastward and westward momentum from the troposphere into the stratosphere/mesosphere (i.e., the middle atmosphere) to drive large-scale circulation there, SST fronts can therefore potentially exert a significant remote influence on the middle atmosphere.
The purpose of this study is to investigate whether mid-latitude SST fronts can exert significant influence not only in the troposphere but also in the middle atmosphere through modulating upward propagating waves.To address this issue, idealized experiments were conducted with a "high-top" AGCM with fine vertical resolution, following experimental designs similar to those employed by Nakamura et al. (2008).In this letter, simulated climatological-mean fields for June-July-August (JJA) are analyzed within the context of the austral winter.

Model Description and Experimental Design
The model employed is based on the atmospheric component of the MIROC version 3.2 (Hasumi & Emori, 2004), referred to herein as the MIROC-AGCM.The model has a horizontal resolution of T106 spectral truncation (1.125°× 1.125°) with 168 vertical levels (L168) with the model top at 0.00059 hPa (∼100 km).The vertical level spacing is about 550 m from ∼300 hPa up to 0.003 hPa (see Figure 1d of Kawatani et al., 2019), which should assist with adequate representation of the mean flow interaction with vertically propagating waves.Above 0.0018 hPa, the model includes artificial damping in a sponge layer.Unlike in Nakamura et al. (2008), realistic topography is given as the lower-boundary condition.Topographic GW parameterization based on McFarlane (1987) is employed, but non-orographic GW parameterization is not included in the model.
Although a realistic land-sea distribution and topography are included as well as including seasonal cycle, our AGCM simulations were conducted in a rather idealized manner.Specifically, a zonally uniform SST distribution was prescribed.The meridional SST profiles were obtained by averaging SST over the South Indian Ocean (60°-80°E) for each calendar month based on monthly mean OISST (Raynolds et al., 2007) averaged for 2003-2010.Subsequently, zonally uniform monthly SST fields based on these profiles were prescribed as the lower-boundary condition for the model SH.For the model Northern Hemisphere (NH), monthly SST fields prescribed were taken from those for the corresponding season in the SH (e.g., SST for NH January was taken from SST in SH July).The black line in Supplementary Fig. S1 illustrates the latitudinal profile of the June-July-August (JJA) mean SST in the control (CTL) experiment.In both hemispheres, the SST profile exhibits sharp gradients at 45°latitude.
In the "no-front (NF)" experiment, SST poleward of the mid-latitude front was artificially raised to eliminate the frontal gradients (red line in Supplementary Fig. S1).No modifications were added to tropical/subtropical SST, as it could sensitively influence the intensity of the subtropical jet and Hadley cell.These method and concept align with those in Nakamura et al. (2008).One potential drawback in the NF run is the artificially raised SSTs poleward of ∼45°S and 45°N, which acts to increase precipitation and thus favors cyclone development.This effect is nevertheless overwhelmed by the dominant impact of the lack of the mid-latitude SST fronts on baroclinic-wave activity (Nakamura et al., 2008;Ogawa et al., 2012Ogawa et al., , 2016;;Sampe et al., 2010Sampe et al., , 2013)).Since no non-orographic GW parameterization was employed in the present study, modulations in middle atmospheric circulations in the presence of frontal SST gradients are attributable primarily to resolved waves in the model.
To mitigate the impact of sea ice between the two experiments with different SSTs at high latitudes, sea ice is not included in the model boundary condition, and SST is fixed at 273K where sea ice was originally present.The model was integrated for 150 years in both experiments.Such extended simulations are deemed necessary to capture statistically significant differences in the middle atmosphere, where interannual variability is substantially large.Other model configurations follow those outlined in Kawatani et al. (2019).In this initial paper the focus is on the winter circulation in the SH.The effects of mid-latitude SSTs should be more pronounced in the SH given the broader ocean than in the NH, and the tropospheric coupling to the middle atmosphere is generally believed to be strongest in winter.Differences discussed in this paper are defined as deviations in the 150-year JJA-mean fields in the CTL run from those in the NF run, with statistical confidence areas at 95%, determined by Student's t-test.

Differences Associated With Baroclinicity Due To SST Fronts
First, differences arising from the mid-latitude SST fronts are confirmed.Figures 1a-1c show climatological JJAmean temperature at 925 hPa in the CTL run, and its meridional gradient in each of the CTL and NF runs.In the CTL run, strong meridional gradients are found around ∼45°S associated with SST gradient (Fig. S1b), while the gradients are much smaller in the NF run.These results confirm that baroclinicity in the NF run is significantly weaker than in the CTL.
Figures 1d-1f display mean precipitation in the CTL and NF runs, and differences between them (CTL minus NF runs), respectively.In the CTL run, bands of locally enhanced precipitation form around 43°S on the poleward flank of the SST front.In contrast, such precipitation bands become less obvious in the NF run, indicating sharp SST fronts are essential for the formation of precipitation bands (Minobe et al., 2008).At subpolar latitudes, ) in (d) CTL and (e) NF runs as well as (f) their differences (CTL-NF).(g) The corresponding 250 hPa u'v' (m 2 s 2 ) with 4 ≤ s ≤ 11 in the CTL run (contour) and its difference between CTL and NF runs (colors).(h) As in (g), but for 100 hPa φ'w' (J kg 1 m s 1 ) with s ≥ 12.The differences with statistical confidence ≥95% are shaded in (f)-(h).

Geophysical Research Letters
10.1029/2024GL108262 greater mean precipitation (differences of ∼0.6 mm day 1 ) in the NF run is consistent with warmer SSTs (Fig. S1a) than in the CTL run.
Differences in synoptic-scale baroclinic wave activity are investigated by calculating meridional eddy flux of zonal momentum u'v' and poleward eddy heat flux v'T' with zonal wave number s from 4 to 11 (4 ≤ s ≤ 11), where T′, u′, and v′ are fluctuations in temperature, zonal and meridional wind velocities.Figure 1g shows u'v' at 250 hPa in the CTL run and its differences between the CTL and NF runs.Differences of u'v' (shaded in blue) are negatively larger in the southern flank of the zonally extended maxima of the mean u'v' in the CTL run (as indicated by dashed lines), indicating enhanced poleward flux of westerly momentum.Meridional sections of zonal-mean u'v' and v'T' are presented in Supplementary Fig. S2.The negative u'v' in the CTL and NF runs maximizes around 250 hPa at ∼33°S, while their difference peaks at ∼45°S around the SST front.Distributions of v'T' as a measure of baroclinic eddy growth indicate significantly larger negative values in the CTL run in the lower troposphere, with the maximum differences around 850 hPa also at ∼45°S around the SST front.These findings illustrate that baroclinic wave activity is notably larger in the CTL than in the NF around the SST fronts, consistent with other previous studies (e.g., Nakamura et al., 2008).
Next, distributions of GWs propagating from the troposphere into the stratosphere are examined through the vertical component of energy flux, ϕ'w', with s ≥ 12, where ϕ′ and w′are fluctuations in geopotential height and upward velocity.In mid-latitudes, wave components with s ≥ 12 are mainly attributed to GWs.Note that fluctuations with s ≥ 12 encompass other waves, such as medium-scale traveling waves (Sato et al., 1993).However, GWs are expected to dominate this quantity, as well as the upward flux of zonal momentum and horizontal wind divergences, as demonstrated subsequently.Contours in Figure 1h illustrate mean ϕ'w' at 100 hPa in the CTL run, which generally corresponds with those of u'v' at 250 hPa.Differences in ϕ'w' also resemble those of u'v', extending zonally on the poleward flank of the maximum mean eddy flux in the CTL run.Some peak values of ϕ'w' are found over Tasmania Island, New Zealand and the Andes, where such peaks are not obvious in u'v'.Watanabe et al. (2008) pointed out that GWs are excited vigorously when baroclinic waves hit the Andes, as also seen in this figure .Supplementary Fig. S3 shows longitude-time sections of 100-hPa horizontal wind divergence evaluated only for s ≥ 12, which is considered as being associated with GWs, at 45°S around the SST front in the CTL and NF runs.The period selected is August of the 11th model year, but general characteristics are similar in all other years.In the CTL run, amplitude of GWs tends to be larger in wide longitudinal sectors relative to the NF run.Properties of GWs in these sectors, such as vertical wavelengths and periods, are described in Kawatani et al. (2004).These results once again indicate that GWs are generated more vigorously around baroclinic waves whose activity is enhanced in the CTL and those GWs then propagate into the stratosphere.

Differences of Wave Fluxes and Forcing Attributed to SST Fronts
In the preceding section, distributions of baroclinic wave activity and GW generation are found to be closely related, with significant differences in upward propagating GWs around the latitude of the mid-latitude SST front.In this section, differences in atmospheric waves, primarily GW momentum fluxes, attributed to the SST front are investigated, to address the issue whether the SST front can modulate large-scale circulation in the middle atmosphere.
Figure 2 represents meridional sections of mean upward wave fluxes of zonal momentum ρ 0 u'w' with s ≥ 12 along with the zonal-mean zonal wind in the CTL and NF runs, and the ρ 0 u'w' difference between the two runs, where ρ 0 is the mean-state atmospheric density.In the wintertime SH, negative ρ 0 u'w' extends from the troposphere into the mesosphere in the extratropics, consistent with upward propagation of GWs carrying westward momentum.In the summertime NH, the negative ρ 0 u'w' is seen in the troposphere, while positive ρ 0 u'w' extends vertically in the extratropical middle stratosphere, suggesting GW filtering around the tropopause level.Both the positive and negative ρ 0 u'w' in the middle atmosphere values decay with height, indicating waves breaking and/or dissipation as they propagate upward.Compared to the NF run, negative ρ 0 u'w' is significantly stronger in the CTL run from the upper troposphere up to ∼0.03 hPa and weaker above this altitude in the SH.
The maximum zonal wind speed of the mesospheric polar-night jet (PNJ) around 0.4-0.5 hPa exceeds ∼120 m s 1 in the CTL run and ∼140 m s 1 in the NF runs; these jet intensities are overestimated compared to the observations.The current T106 model resolution cannot represent GWs with s ≥ 107.It is known that with a model horizontal resolution of T213, for example, the maximum PNJ intensity becomes weaker due to stronger resolved wave forcing (Watanabe et al., 2008), although such high-resolution models are currently too expensive for 150-year integrations.In addition, the present model does not employ non-stationary GW parameterization, which may result in the overestimated PNJ intensity in the middle atmosphere due to underestimated total wave forcing.Despite these factors, the maximum intensity of the mesospheric PNJ is nevertheless reduced and thus more realistic in the CTL run than in the NF run.
To estimate resolved zonal wave forcing quantitatively, Eliassen-Palm (EP) flux divergence is calculated (refer to Section 3.5 in Andrews et al., 1987) for all wave components and then compiled for 1 ≤ s < 3, 4 ≤ s ≤ 11 and s ≥ 12. Figures 3a and 3b display meridional sections of EP-flux divergence with all wave components together with zonal-mean zonal wind in the CTL and NF runs, respectively.In the wintertime SH, westward wave forcing represented as negative EP-flux divergence tends to overlap the mean westerly wind.The maximum westward forcing that acts to reduce the westerlies exceeds 100 m s 1 day 1 in the mesosphere.In the summertime NH, large eastward wave forcing dominates in the mesosphere, where mean zonal wind is easterly, and the eastward wave forcing maximizes in the upper mesosphere.Differences in the EP-flux divergence between the two experiments due to all wave components (Figure 3c) show stronger westward wave forcing in the CTL run over most of the SH stratosphere and mesosphere up to ∼0.03 hPa, while the forcing is weaker above ∼0.03hPa.
Figures 3d-3f illustrate the corresponding differences in the EP-flux divergence for the individual zonal wavenumber bands.Large differences with statistical significance in wave forcing with s ≥ 12 are found in the mesosphere in both the NH and SH.Interestingly, the difference in westward wave forcing in the SH mesosphere maximizes at ∼40-50°S nearly above the SST front (Figure 3f).As shown in Supplementary Fig. S3, GWs with smaller amplitude tend to be generated in the NF run, and they can propagate to higher altitudes even into the upper mesosphere, resulting in stronger westward wave forcing above ∼0.03hPa than in the CTL run.Although the differences are not as large as those associated with s ≥ 12, statistically significant differences in wave forcing associated with 1 ≤ s ≤ 3 and 4 ≤ s ≤ 11 are also found in the mesosphere (note color intervals are not linear in order to visualize the full range of differences).The difference due to 1 ≤ s ≤ 3 exhibits a rather complicated striped distribution, while its counterpart due to 4 ≤ s ≤ 11 also hints some meridional structure around 0.1-0.01hPa.Possible factors for these complicated patterns may be attributable to the different shape of the mesospheric mean zonal winds between the two experiments, leading to alterations in the position of wave dissipations and variations in in-situ generation of planetary waves and/or synoptic-scale waves around the upper part of the mesospheric jet (cf.Okui et al., 2021;Sato & Nomoto, 2015;Watanabe et al., 2009).
More detailed analyses are required to explain the causality of these differences, such as the wave generation processes in the troposphere (e.g., separation of GW sources associated with precipitation and/or baroclinic wave activity), GW intermittency (e.g., Ern et al., 2022;Hertzog et al., 2012), and the generation of planetary and baroclinic/barotropic waves in the upper part of the mesospheric jets.These aspects are beyond the scope of the present study and will be investigated in our future work.
Finally, how large-scale circulations and temperature in the middle atmosphere are modulated by the SST fronts is investigated.Figures 3g-3i depict meridional sections of zonal-mean zonal wind, temperature and residual mean

Geophysical Research Letters
10.1029/2024GL108262 meridional circulations expressed by stream function in the CTL run (contours) and the differences between the CTL and NF runs (colors).
In the lower and upper troposphere, the westerlies between 40°S and 60°S are significantly stronger in the CTL run than in the NF run (Figure 3g), as an indication of the intensified eddy-driven PFJ as the influence of the SST front, consistent with Nakamura et al. (2008).Meanwhile, the westerly subtropical jet around 30°S is weaker only by ∼2 m s 1 in the CTL run, although this difference is negligible compared with the jet intensity (exceeding 50 m s 1 ).This is due to our experimental design in which SST gradient is unchanged in the tropics and subtropics in order not to affect the Hadley circulation substantially, as in Nakamura et al. (2008).In the SH middle atmosphere, by contrast, zonal-mean westerlies between 40°S and 60°S are overall weaker in the CTL run, due to stronger westward wave forcing as shown in Figure 3c.

10.1029/2024GL108262
Corresponding to the zonal wind differences, zonal-mean temperatures in the CTL run are significantly warmer in the high-latitude SH from the mid-stratosphere into the lower mesosphere (∼30 hPa through ∼0.05 hPa) and cooler above (Figure 3h).It is noteworthy that in our CTL run, though idealized (given the zonally homogeneous SSTs), the presence of the mid-latitude SST front leads to a warming of the stratospheric winter pole and thereby reducing a cold pole bias existing in the NF run and in most AGCM simulations.
Consistent with the wave forcing differences, the stratospheric Brewer-Dobson circulation and mesospheric residual circulation in the SH are significantly stronger in the CTL run than in the NF runup to ∼0.03 hPa (Figure 3i), in association with the stronger westward wave forcing driving poleward motions.Above ∼0.03 hPa, in contrast, mesospheric residual circulations in the SH are weaker, corresponding to the weaker westward wave forcing.These results demonstrate that the mid-latitude SST front can affect not only the troposphere but also the middle atmosphere circulations through enhancing upward propagating atmospheric waves.

Conclusion
Recent studies have demonstrated that the mid-latitude SST frontal zones influence precipitation, synoptic-scale baroclinic-wave activity and eddy-driven PFJ in the troposphere.Those baroclinic waves are known to serve as a significant source of atmospheric GWs that propagate upward into the middle atmosphere.The transport of mean zonal momentum by GWs are known to play crucial roles in driving middle atmospheric circulation (and so is sometimes referred as a "wave-driven circulation").Given this framework, the present study aims to investigate whether the mid-latitude SST fronts can exert a remote impact on the middle atmosphere through modulating upward-propagating atmospheric waves.
The present study used the high-top MIROC AGCM with fine vertical resolution to conduct the following two experiments: one employing zonally uniform SST with a meridional gradient based on the observed South Indian Ocean climatology (CTL run), and the other featuring an artificially suppressed SST front (NF run), following the configuration by Nakamura et al. (2008).Realistic topography was incorporated in both experiments.To specifically examine differences in wave forcing attributed to resolved waves, which can potentially be modulated by the presence or absence of the SST fronts, non-orographic GW parameterization was not included.The model integrations were conducted for 150 years, with the climatological-mean annual cycle of ozone, SST and other factors.The focus of this paper is on climatological JJA-mean fields, corresponding to the SH winter.
Consistent with recent studies, distinct precipitation bands are seen only in the CTL run near the SST frontal zone, and baroclinic-wave activity is more prominent in the CTL compared to the NF run, with its maximum difference found around the SST front.GWs are found to be generated in association with synoptic-scale baroclinic waves and propagating into the stratosphere, as illustrated in Kawatani et al. (2004).Zonal wave forcing of the mean flow in most of the middle atmosphere is stronger in the CTL than in the NF run.Specifically, the largest differences in the wave forcing are associated with GWs, defined by wave components with s ≥ 12. Wave momentum fluxes and amplitudes of GWs are substantially larger in the CTL than in the NF run.Westward wave forcing is thus significantly larger in the wintertime SH up to ∼0.03 hPa in the CTL run but smaller above, because GWs with smaller amplitudes generated in the NF run tend to propagate to higher altitudes in the upper mesosphere.Zonal-mean zonal wind, temperature and mean residual circulations are also significantly different between the CTL and NF runs, and the zonal wind and temperature distributions appear to be more realistic in the experiment with the mid-latitude SST frontal zones.
Although the idealized experiments with T106 resolution conducted in the present study can explicitly resolve GWs only within the limited spectral domain, they have successfully demonstrated the potential impact of the mid-latitude SST front on the middle atmosphere for the first time.The present experiments may underestimate the importance of the oceanic fronts, as the model can only capture the GW effects that are explicitly resolved.In the real atmosphere there is evidence suggesting that even smaller scale convectively generated GWs may play a role in the response to frontal SST gradients (Kang et al., 2017;Plougonven et al., 2015).GWs generated around the SST front may exert an even more pronounced impact on the middle atmosphere than the results obtained from the present simulations.
To verify this, it would be beneficial to use high-top climate models with much higher horizontal resolution that can resolve more detailed structures of SST fronts and cover wider wave spectral domains.Additionally, higher vertical resolution should be considered to better capture the generation and dissipation of GWs (e.g., Watanabe Geophysical Research Letters 10. 1029/2024GL108262 et al., 2015)).Utilizing convection-permitting models would also be advantageous for investigating GW generation, given their capacity to simulate increased GW momentum fluxes at mid-to-high latitudes around the oceanic front zones (Stephan et al., 2019).
Climate models often exhibit a systematic bias of delayed springtime breakdown of the SH polar vortex, possibly due to insufficient wave forcing.McLandress et al. (2012) demonstrated that the cold-pole bias is significantly reduced and the vortex breaks down earlier in AGCM simulations with extra GW forcing near 60°S.This extra GW forcing near 60°S mimics the forcings caused by laterally propagating GWs so as to focus onto the westerly jet core (Sato et al., 2009(Sato et al., , 2012) ) and by orographic GWs above small islands in the SH (Alexander et al., 2009;Alexander & Grimsdell, 2013) that are not resolved nor parameterized in most AGCMs, have effects on largescale circulations in the middle atmosphere.The present study suggests that including realistic mid-latitude SST fronts in high-resolution AGCMs should have impacts on improving simulation of the large-scale circulation in the middle atmosphere.It is also suggested that climate models with sufficiently high resolutions in their atmospheric and oceanic components can lead to realistic reproduction of mid-latitude SST fronts and thereby more realistic representation of storm-tracks and GW generation to substantially reduce model biases in the middle atmosphere.

Figure 1 .
Figure1.(a) Climatological JJA-mean temperature (K) at 925 hPa in the CTL run, and its meridional gradient in (b) CTL and (c) NF runs.The climatological JJA-mean precipitation (mm day 1 ) in (d) CTL and (e) NF runs as well as (f) their differences (CTL-NF).(g) The corresponding 250 hPa u'v' (m 2 s 2 ) with 4 ≤ s ≤ 11 in the CTL run (contour) and its difference between CTL and NF runs (colors).(h) As in (g), but for 100 hPa φ'w' (J kg 1 m s 1 ) with s ≥ 12.The differences with statistical confidence ≥95% are shaded in (f)-(h).

Figure 2 .
Figure 2. Meridional sections of climatological JJA-mean ρ 0 u'w' (m Pa) with s ≥ 12 in the (a) CTL and (b) NF runs, as well as (c) their differences.Contours represent zonal-mean zonal wind (m s 1 ) in the (a) and (c) CTL and (b) NF runs.The difference with statistical confidence ≥95% is shaded in (c).