A midlatitude oceanic frontal zone is a confluent region of warm and cool ocean currents, characterized by a strong meridional gradient in both sea surface temperature (SST) and surface air temperature (SAT). While recent observational and modeling studies indicate potential impacts of midlatitude oceanic fronts on the extratropical climatological circulation, including storm tracks and eddy-driven westerlies, their impacts on the atmospheric-dominant low-frequency variability (i.e., the annular mode) still remain to be understood. This study explores possible impacts of midlatitude oceanic frontal zones on annular mode signatures in the wintertime Southern Hemisphere (SH). To mimic the SH, sets of idealized aquaplanet experiments are conducted with zonally symmetric distributions of SST prescribed globally at the lower boundary of an atmospheric general circulation model. By systematically changing the latitude of frontal gradient in the SST profile, the experiments reveal that the characteristics of the wintertime annular mode exhibit strong sensitivity to the position of the SST front if situated at midlatitude or subpolar latitude. The annular mode may be interpreted as a manifestation of wobble of the extratropical tropospheric circulation between two “dynamical regimes”—one under the strong influence of SST gradient and the other under the strong control of atmospheric internal dynamics unrelated to the lower-boundary condition. In fact, this interpretation offers insight into the observed interbasin differences in the wintertime signature of the southern annular mode (SAM) that are embedded in the zonally symmetric anomalies. The findings suggest a possible reinterpretation of the climatological-mean state observed in the wintertime SH as the superposition of those two dynamical regimes.
The extratropical tropospheric circulation fluctuates with various temporal and spatial scales. The dominant mode of variability on intraseasonal through interannual scales is manifested as meridional shifts of the midlatitude westerlies with a high degree of zonal symmetry, especially in the Southern Hemisphere (SH) (e.g., Kuroda and Kodera 1998; Thompson and Wallace 2000). This pattern is often referred to as the annular mode, which influences climatic conditions over extensive regions in the extratropics (Limpasuvan and Hartmann 1999; Thompson and Wallace 1998, 2000). Annular modes in the individual hemispheres are referred to as the southern annular mode (SAM) and the northern annular mode (NAM).
While NAM anomalies are maintained by feedback forcing from anomalous activity of synoptic-scale eddies and planetary waves (Limpasuvan and Hartmann 1999, 2000; Kimoto et al. 2001; Lorenz and Hartmann 2003; Watanabe and Jin 2004), SAM anomalies are maintained mainly through feedback forcing generated with anomalous activity of synoptic-scale eddies (Lorenz and Hartmann 2001). The SAM-associated anomalous convergence of eddy momentum flux near the tropopause arises from modulations in meridional propagation and structure of synoptic-scale eddies (Yu and Hartmann 1993; Hartmann and Lo 1998; Shiogama et al. 2004).
The positive phase of the wintertime SAM represents a double-jet structure with distinct separation between the midlatitude polar-front jet (PFJ) and subtropical jet (STJ), while its negative phase represents a merged single-jet structure with the dominant STJ in the upper troposphere (Yoden et al. 1987; Yu and Hartmann 1993; Hartmann and Lo 1998). There exists an interpretation of the wintertime SAM variability as a “regime shift” between two different quasi-equilibrium states of the tropospheric circulation (e.g., Yoden et al. 1987). In fact, Itoh et al. (1999) suggested the presence of dual attractors in the phase space based on the leading modes of variability of the SH tropospheric westerlies as representative of “dynamical regimes,” in arguing their correspondence to the double- and single-jet structures. Nevertheless, the above arguments on the regime-like characteristics of the observed wintertime SAM are rather phenomenological, and in-depth investigation of the underlying dynamics is required.
Eichelberger and Hartmann (2007) found that the meridional structure of the tropospheric annular mode is sensitive to the climatological-mean background state of the westerlies. Nevertheless, not much attention has been paid to the potential importance of sea surface temperature (SST) distribution as the background state of the annular modes, despite the fact that the climatological-mean storm track (Chang et al. 2002) and PFJ tend to form along oceanic frontal zones with sharp SST gradients in each of the hemispheres (Nakamura and Shimpo 2004; Nakamura et al. 2004). In each of the ocean basins, a surface baroclinic zone with sharp surface air temperature (SAT) gradient forms along an oceanic frontal zone. As an extension of the conceptual models of the tropospheric general circulation by Palmén and Newton (1969) and by Lee and Kim (2003), Nakamura et al. (2004) postulated a new model that emphasizes the necessity of taking a midlatitude oceanic frontal zone into account to understand how the climatological latitudes of a storm track and PFJ are determined. The new conceptual model describes a fundamental picture of the tropospheric general circulation that would be realized even without landmasses. The tropospheric circulation observed in the SH may roughly be interpreted in this simple framework with slight modifications for individual ocean basins.
In the present study, potential importance of midlatitude SST gradient across the SH oceanic frontal zones in the climatological-mean tropospheric circulation and its dominant variability is assessed with implications for the observed SAM variability. The assessment is through aquaplanet AGCM experiments where all the landmass and sea ice are removed from the model lower boundary and zonally uniform SST is prescribed. This idealized setting suppresses planetary waves forced by land–sea thermal contrasts and topography, while retaining synoptic-scale eddies and an eddy-driven PFJ in addition to the thermally driven STJ. Although any quantitative assessment of the model reproduction of the observed atmospheric circulation is of course impossible under this idealized setting, the experiment nevertheless allows us to extract the essential features of the atmospheric general circulation and an active role of the ocean, if any, in the atmospheric circulation and its variability.
Recent studies assessed the potential importance of the latitude of an oceanic front for the transient eddy activity and midlatitude westerlies through the aquaplanet experiments (e.g., Brayshaw et al. 2008; Chen et al. 2010; Deremble et al. 2012). However, their modifications on the prescribed tropical and/or subtropical SST directly affect the Hadley cell and associated STJ and thus indirectly the midlatitude storm track and eddy-driven PFJ (Lee and Kim 2003). Although their studies imply the significance of midlatitude oceanic frontal zones in the extratropical tropospheric circulation, they fail to separate it from the impacts of the tropical SST. The particular deficiency was first overcome by Nakamura et al. (2008), who used the SST profile with frontal SST gradient observed over the south Indian Ocean for their model experiment and modified the gradient without changing tropical SST. They revealed the particular importance of the oceanic frontal zones for energizing transient eddies along the SST front and thus anchoring and reinforcing the eddy-driven PFJ at the poleward flank of the front as observed. They pointed out the importance of the meridional contrast in the turbulent sensible heat supply from the ocean for the recurrent development of the transient eddies through maintaining the near-surface baroclinicity efficiently, which has been verified through a numerical study by Hotta and Nakamura (2011). Nakamura et al. (2008) also pointed out that the frontal SST gradient could significantly affect the amplitude and meridional structure of the annular mode, which was further discussed by Sampe et al. (2013). These previous studies suggest the necessity of considering a midlatitude oceanic frontal zone to understand not only the climatological-mean state of PFJ but also the annular mode variability. Another aquaplanet experiment by Ogawa et al. (2012) revealed the sensitivity of the climatological-mean PFJ and lower-tropospheric storm track to the latitude of an SST front. It is therefore suggested that considering the latitudinal position of an oceanic front may be necessary to understand the annular mode dynamics.
The aim of this study is to investigate the dependence of the annular mode characteristics on the latitude of an SST front, in order to deepen the understanding of the observed SAM signatures in relation to the underlying SST gradients. As in the previous studies, we utilize aquaplanet AGCM experiments to address this issue, which enables us to extract the essential features of the storm track and PFJ by suppressing planetary waves. The specific experimental design is described in section 2. In section 3 results of our model experiments are presented, and implications of those results for the observed SAM are discussed in section 4. A new interpretation of the wintertime climatological mean state in SH is proposed in section 5. Conclusions and implications are presented in section 6.
2. Experimental design and simulated climatological-mean state
Numerical experiments in this study are conducted with the AGCM for Earth Simulator (AFES; Ohfuchi et al. 2004; Enomoto et al. 2008; Kuwano-Yoshida et al. 2010). The model has 56 levels up to 0.09 hPa. Its horizontal spectral resolution for our experiments is triangular truncation at wavenumber 79 (T79), corresponding approximately to 150-km grid intervals both longitudinally and latitudinally. Although not sufficient for fully resolving mesoscale features of the SST distribution prescribed as the model lower-boundary condition, this resolution is still enough to incorporate frontal SST gradient across an oceanic frontal zone, which is important for realistic representation of both the climatological-mean state and annular mode variability in our aquaplanet experiments (Nakamura et al. 2008; Ogawa et al. 2012; Sampe et al. 2010, 2013).
The lower boundary of the AGCM for our experiments is set as the fully global ocean with different latitudinal profiles of zonally uniform SST (Fig. 1a). As in the previous studies (Nakamura et al. 2008; Sampe et al. 2010, 2013; Ogawa et al. 2012), one of these SST profiles was taken from the climatological monthly SST with 1° resolution based on the OISST data (Reynolds et al. 2007), blended satellite and in situ observations. The profile over the south Indian Ocean (60°–80°E) for austral winter (June–August) was assigned to the model SH (black line in Fig. 1a) and the corresponding austral summertime profile (December–February) to the model NH. With this SST profile characterized by its frontal gradient at 45° latitude in both hemispheres, the AGCM was integrated for 120 months under fixed insolation to its solstice condition after 6 months of spinup. For sensitivity experiments, the intensity of frontal SST gradient was kept the same but the frontal latitude was shifted from 30° to 55° at 5° intervals (colored lines in Fig. 1), as in Ogawa et al. (2012). In any of those modified profiles, SST equatorward of 25° in the two hemispheres was kept unchanged (Fig. 1b), so as not to affect the tropical convection directly that drives the Hadley cell. The SST profiles are thus designed to highlight the dependence of the wintertime annular mode variability on the latitudinal position of the extratropical oceanic frontal zone. For deepening our understanding, another experiment is performed by using a “nonfront” (NF) SST profile (Nakamura et al. 2008; Sampe et al. 2010, 2013), as indicated with dashed lines in Fig. 1. In each of the experiments with the frontal zone, SST gradient poleward of the frontal zone is kept the same as in the NF experiment. Our idealized experiment cannot reproduce the observed atmospheric circulation in the SH, but it can still capture the essential dynamics of the extratropical circulations.
As shown in Fig. 2, the climatological-mean states of storm tracks and westerlies in the winter hemisphere simulated in our experiments are almost identical to their counterparts in Ogawa et al. (2012). In the lower troposphere (Fig. 2a), the axial latitude of climatological-mean westerlies shows certain sensitivity to the latitude of the SST front. The lower-tropospheric westerly axis tends to shift poleward as the frontal latitude increases with a hint of dual westerly axes, while the sensitivity is lower to the subpolar SST front. The sensitivity is even more ambiguous in the upper troposphere, characterized by the nearly constant latitude of the PFJ around 45° (Fig. 2b). In what follows, this sensitivity for the climatological-mean states is interpreted from a viewpoint of annular mode characteristics.
3. Wintertime annular mode simulated in aquaplanet experiments
a. Definition of the model annular mode
To extract low-frequency wind variability associated with the annular mode, a Butterworth filter with a cutoff period of eight days was applied to zonal-mean zonal wind—[U], where the square brackets represent zonal-mean statistics—at 925 hPa, where STJ is very weak (Fig. 2a). This level is thus chosen to highlight the variability of PFJ that accompanies notable anomalies in the surface westerlies. To identify the annular mode signature in the model extratropical troposphere, an empirical orthogonal function (EOF) analysis was applied to the low-pass-filtered [U] anomalies over the entire analysis period (3600 days) within the domain poleward of 20°S in the model winter hemisphere. In this EOF analysis, the 925-hPa [U] anomalies had been weighted by the square root of cosine of latitude. As observed (e.g., Thompson and Wallace 2000), the model annular mode is defined as meridional wobble of the latitude of PFJ, and by convention the positive (negative) phase is defined as the phase that corresponds to a poleward (equatorward) shift of the PFJ axis. The annular mode is extracted from the first EOF (EOF1) for all the experiments except the one with the SST front at 35° latitude, where EOF2 represents the annular mode. Fractions of the total [U] variance explained by EOF1 and EOF2 are shown in Fig. 3a. In the experiment with the SST front at 35°, the variance fractions explained by EOF1 and EOF2 are comparable and they are therefore not separated statistically. In fact, the annular-mode characteristics in the experiment with a subtropical SST front show notable differences from the observed SAM, as discussed later.
b. Sensitivity of annular mode structure and persistence to the SST front latitude
Instead of displaying the obtained EOF patterns themselves, meridional profiles of 925-hPa [U] anomalies regressed on the standardized principal components (PCs) for all the experiments are compiled in Figs. 3d and 3e. The annular mode, which accompanies zonal wind anomalies with a high degree of zonal symmetry (not shown), shows its strong sensitivity to the latitude of the SST front in its meridional structure (Fig. 3d). Two centers of action in the [U] anomalies associated with the annular mode are located poleward and equatorward of the SST front, representing their seesaw relationship with the node situated near the SST front. In the absence of the SST front (i.e., the NF experiments), EOF1 represents meridional fluctuations of the midlatitude westerly axis, as well as the poleward extension or contraction of the midlatitude westerly belt whose climatological axis is at 38° latitude (Fig. 2a). As shown in Fig. 3b, the e-folding time of the PC that represents the annular mode is in the range of 7 through 14 days, in agreement with the observed SAM (approximately 10 days; Thompson and Woodworth 2014). Figure 3b appears to suggest no systematic relationship between the persistence of the model annular mode and the SST front latitude. Once rearranged as in Fig. 3c, however, the simulated results nevertheless show a clear tendency for the persistence of the annular mode to be reduced with the climatological latitude of the eddy-driven westerly jet as long as the SST front is situated at a middle or subpolar latitude (i.e., 40°–55°), which is consistent with the tendency for the SAM signature simulated in global climate models (Barnes and Hartmann 2010). The persistence is reduced further in a simulation with a subtropical SST front or without any front. Another EOF, which is EOF2 except for the experiment with the SST front at 35° latitude (EOF1 in this case), basically represents the strengthening or weakening of PFJ rather than its meridional displacement (Fig. 3e; cf. Fig. 2a). Except in the case of the SST front at 30° latitude, the e-folding time for the particular PC is about a week (Fig. 3b), indicating that this mode of variability is less persistent than the annular mode.
c. Characteristics of the annular mode in its positive phase
The meridional distribution of [U] anomalies associated with the model annular mode (Fig. 3d) suggests the sensitivity of its characteristics to the latitude of the SST front. In this section, typical behaviors of the PFJ and storm track for the positive phase of the annular mode are described. Figure 4a shows the meridional profile of 925-hPa [U] composited for the positive events. As the STJ is confined into the upper troposphere (Nakamura et al. 2004), the results mostly describe the behavior of PFJ. In the positive phase, the latitude of the PFJ axis shows strong sensitivity to the latitude of the SST front, with a systematic poleward displacement relative to the front as observed climatologically over the southern oceans (Nakamura et al. 2004; Nakamura and Shimpo 2004). This poleward displacement means that the near-surface westerly axis is inclined to stay on the cooler side of the SST front. The cooling by the underlying ocean (Fig. 4f) acts to drive the westerlies by pulling air parcels poleward (Hoskins 1991), which counteracts frictional damping. In the positive phase of the annular mode, the latitude of a storm track defined either as the maximum activity of synoptic eddies (extracted as 8-day high-pass-filtered fluctuations) near the surface also shows its high sensitivity to the latitude of the SST front (Figs. 4b,c). In all the experiments, the low-level storm track axis is closer to the SST front compared to the PFJ axis regardless of the definition of the storm track: the maximum wind fluctuations (Fig. 4b) or the maximum eddy heat flux serves as a measure of baroclinic development of eddies (Fig. 4c). As is obvious in Figs. 4a–c, the PFJ and storm track in the experiments with the subtropical SST front (at 30° or 35° latitude) are also located systematically poleward of the SST front, which will be discussed later.
In the positive phase of the annular mode, the upper-tropospheric PFJ is well separated from the STJ (Fig. 4d), as in both the observations and model simulations (e.g., Hartmann and Lo 1998; Nakamura et al. 2008; Sampe et al. 2013). As their low-level counterpart, the upper-level axes of the eddy-driven PFJ (Fig. 4d) and the storm track (not shown) are almost collocated, showing notable sensitivity to the latitudinal shift of the SST front. The sensitivity is consistent with the same sensitivity as found in the convergence maximum of an upper-level poleward flux of westerly momentum associated with transient eddies (Fig. 4e), which is in agreement with both the observations and model simulations (Yu and Hartmann 1993; Hartmann and Lo 1998).
Compared to the climatological-mean state shown by Ogawa et al. (2012), the latitudinal association among the storm track, PFJ, and SST front is more evident in the positive phase of the annular mode, especially when the SST front is located at a middle or subpolar latitude. In other words, the association in the climatological-mean state is dominated by a contribution from the positive phase of the annular mode.
d. Characteristics of the annular mode in its negative phase
Figure 5 shows the composited statistics for the negative events of the annular mode. In sharp contrast to the positive phase (Fig. 4a), the PFJ axis in the negative phase stays around 38° latitude for all the experiments (Fig. 5a), showing little sensitivity to a latitudinal shift of the SST front. A similar insensitivity of the PFJ latitude to SST front latitude has already been discussed by Ogawa et al. (2012), but only for a subpolar SST front (cf. Fig. 2a). Referring to its latitudinal correspondence to the mean state simulated in the NF experiment by Nakamura et al. (2008) and Sampe et al. (2010), Ogawa et al. (2012) argued that this latitudinal insensitivity of the climatological PFJ axis might be due to the dominance of atmospheric “internal” dynamics unrelated to the lower-boundary condition (Robinson 2006), through which the eddy-driven PFJ can maintain itself by organizing baroclinically developing eddies and thereby enhancing the feedback forcing. It is evident in Fig. 5a that the PFJ axis in the negative phase of the annular mode simulated in any of our experiments with frontal SST gradient well corresponds to the climatological-mean PFJ axis in the NF experiment, as indicated with a white dashed line in the figure. In the negative phase, oceanic cooling with downward surface sensible heat flux on the cooler side of the SST front is greatly reduced (not shown). Unlike in the positive phase, the latitudinal collocation is overall diminished between the SST front and eddy-driven jet. The low-level storm track axis marked with the peak variance of meridional wind fluctuations (Fig. 5b) tends to reside around the climatological-mean storm track in the NF experiment at 40° latitude, despite the low-level eddy heat flux that still maximizes in the vicinity of the SST front (Fig. 5c), which is likely to include a substantial contribution from the thermodynamic response to enhanced near-surface temperature gradient across the SST front. Unlike in the positive phase of the annular mode, the dynamic response as an impact of the front is presumably suppressed in the negative phase, largely because of the absence of an upper-level westerly jet that acts as a waveguide for upper-tropospheric disturbances. Thus the coupling of upper-level and near-surface PV anomalies important for their baroclinic growth is unlikely to occur and thereby the dynamical response around the SST front. It is thus suggested that in the negative phase of the annular mode the latitudes of the storm track and eddy-driven PFJ tend to be determined by atmospheric internal dynamics, at least for the experiments with a midlatitude or subpolar SST front.
In the negative phase of the annular mode, the 300-hPa [U] profile in any of the experiments (Fig. 5d) is characterized by a prominent STJ, while PFJ almost diminishes in the midlatitudes. Regardless of the frontal latitude, the upper-level storm track resides around 40° latitude (Fig. 5f), corresponding to its climatological-mean position in the NF experiment (white line). The associated eddy momentum flux convergence (Fig. 5e) maintains the midlatitude westerlies in agreement with the observations and model experiments (Yu and Hartmann 1993; Hartmann and Lo 1998).
For the experiments with subtropical SST fronts (at 30° or 35° latitude), the PFJ axis is closer to the SST front in the negative phase of the annular mode, by definition, than in its positive phase, making a sharp contrast to the situation in the other experiments. In fact, the storm-track activity with the subtropical SST front (Figs. 5b,c) tends to be stronger in the negative phase than in the positive phase (Figs. 4b,c), suggestive of the stronger influence of the surface baroclinicity along the SST front in the negative phase. The annular mode signature for the subtropical SST front is different from the counterpart for the subpolar or midlatitude SST front, as discussed more in detail later in this section.
e. Regime-like characteristics of the model annular mode realized under frontal SST gradient
As demonstrated in the preceding subsections, the sensitivity of the annular mode characteristics to the latitude of SST front depends strongly on its phase. Latitudes of the PFJ and storm track show their strong dependence on the latitude of SST front in the positive phase, while such dependence becomes ambiguous in the negative phase. This tendency suggests that the annular mode variability may be a manifestation of wobble of the extratropical tropospheric circulation between two regimes—one under the strong control of atmospheric internal dynamics and the other under the strong influence of frontal SST gradients. The latter is characterized by the dynamical response to SST gradients and thus the association between the storm track, eddy-driven PFJ and SST front. In this subsection, we discuss the regime-like characteristics of the annular mode in the experiments with extratropical SST fronts.
Figure 6 shows the probability density of the 925-hPa PFJ latitude as a function of the SST front latitude. For the experiments with midlatitude or subpolar SST fronts (45°–55° latitude), distinct dual peaks are found with a well-defined probability minimum in between. One of the peaks is located about 5° poleward of the SST front, while the other is located equatorward of the front and around the same latitude as a distinct single peak in the NF experiment. Dual peaks are hinted also for the experiment with the SST front at 40° latitude. Bimodality of the probability density distribution in the physical space can be regarded as a manifestation of two regimes of the zonal mean circulation. Latitudes of these two peaks correspond to those of [U] composited for the individual phases of the annular mode (cf. Fig. 6 with Figs. 4a and 5a). In fact, for each of the experiments, the climatological-mean residence time of the [U] maximum at given latitude (Fig. 7a) tends to maximize around the dual probability peaks (Fig. 6), and these peaks correspond to the maxima in the residence time averaged separately for the positive and negative phases of the annular mode (Figs. 7b,c). For the positive phase of the annular mode (Fig. 7b), the mean residence time of the [U] maximum tends to decrease with its latitude, following the SST front latitude. For the negative phase (Fig. 7c), by contrast, the corresponding residence time is insensitive the front latitude, and so is the latitude of the [U] maximum itself. These results suggest two distinct quasi-equilibrium states of the model extratropical circulation in the presence of an extratropical SST front. Therefore, the variability associated with the model annular mode may be interpreted as wobble between these two quasi-equilibrium states or regimes.
The higher sensitivity of PFJ and storm track to the SST front in the positive phase of the annular mode than in the negative phase implies a stronger anchoring effect by the front on them. Previous studies (e.g., Sampe et al. 2010; Hotta and Nakamura 2011) demonstrated the importance of the meridional contrast in surface turbulent sensible heat flux (SHF) across an SST front in maintaining the surface baroclinicity and thereby allowing the recurrent development of synoptic eddies along the SST front (referred to as “oceanic baroclinic adjustment” in these studies). Figure 8 shows the difference between the composited gradients of zonal-mean SHF between the two phases of the annular mode (positive minus negative). For each of the experiments, the SHF gradient across the SST front is indeed stronger in the positive phase than in the negative phase, which suggests that more efficient oceanic baroclinic adjustment is operative in the positive phase.
Figure 9 shows latitudinal profiles of probability density of the latitude of the near-surface [U] maximum, based on daily sampling for the four experiments with SST fronts at 55°, 50°, 45°, and 40° latitude, as well as for the NF experiment. Three different profiles of the probability density plotted in each of the panels in Fig. 9 are based on separate samplings over the positive and negative events of the annular mode and over the entire experimental period. In the experiments with the SST front at 55°, 50°, and 45° latitude, the probability based on the full-period sampling (black line) exhibits dual peaks, as consistent with Fig. 6. Obviously, the PFJ axis in the positive phase of the annular mode (red line) tends to be located in the vicinity of the poleward peak, which corresponds to the period when the extratropical atmospheric circulation shows strong sensitivity to the SST front. In contrast, the [U] axis in the negative phase of the annular mode (blue line) tends to be located at the equatorward peak located around 38° latitude, which corresponds to the distinct single peak of the probability of the [U] maximum over the entire period for the NF experiment (black line in Fig. 9e). The dual-peak feature in the probability density is weaker in the experiment with SST front at 40° latitude (Fig. 9d), but still a weak peak is found at 38° separately from the dominant peak at 45° latitude. These results support the notion that the annular mode simulated with a midlatitude or subpolar SST front represents vacillation between the two quasi-equilibrium states, where the effectiveness of the oceanic baroclinic adjustment along the SST front is essentially different.
Figure 10 shows meridional sections of typical quasi-equilibrium states of the zonal-mean storm track activity and PFJ as realized during extreme events of the positive and negative phases of the annular mode in the presence of the SST front at 55° latitude. In a regime represented by the positive phase of the annular mode (Fig. 10a), a well-defined PFJ forms poleward of the SST front throughout the depth of the troposphere accompanied by strong storm-track activity around the front, marked by enhanced upward Eliassen–Palm (EP) flux (Andrews et al. 1987) as an indicator of baroclinic growth of eddies. In contrast, eddy activity is not well organized as a storm track in the other regime represented by the negative phase of the annular mode (Fig. 10b), and the eddy-driven PFJ is weaker and located at around 40° latitude. This situation is similar to the climatological-mean state in the absence of SST front (Fig. 10c), indicative of the dominance of atmospheric internal dynamics for the maintenance of the PFJ in this regime.
As suggested from Fig. 6, the regime-like behavior of the model annular mode almost diminishes in the presence of a subtropical SST front, marking a sharp contrast to the case of a midlatitude or subpolar SST front. For the subtropical SST front at 30° or 35° latitude, the probability density of the latitude of the 925-hPa [U] maximum shows a single peak as in the case of the NF experiment (Fig. 6). Figures 11a and 11b show the same probability density as in Fig. 9, but for the experiments of SST front at 30° and 35° latitude, respectively. The annular mode is found to represent a wobble of the near-surface PFJ axis around its climatological-mean position, which is similar to the case of the NF experiment (Fig. 9e). The descending branch of the Hadley cell formed under the STJ acts to retard baroclinic eddy growth and associated driving of the surface westerlies below the STJ (Nakamura and Shimpo 2004). Under this influence of the Hadley cell, the storm track is unlikely to form in the vicinity of a subtropical SST front, and near-surface [U] axis tends to be situated around the latitude at which the axis forms in the NF experiment. The residence time of the [U] axis is thus particularly long in the presence of the subtropical SST fronts (Fig. 7).
4. Implications for the observed wintertime southern annular mode
a. Definition of the southern annular mode
In this section, we argue that some of the characteristics of the wintertime annular mode revealed from our aquaplanet experiments are applicable to the observed wintertime SAM. In the SH winter, midlatitude oceanic fronts are located at different latitudes between the basins on the basis of the OISST data (dark green dots in Fig. 12). Specifically, the SST fronts in the South Atlantic and south Indian Ocean are located around 45°S, while the front is located around 55°S in the South Pacific. In the following, monthly mean atmospheric statistics derived from a global reanalysis dataset (JRA-55; Kobayashi et al. 2015) for the 55 years from 1958 to 2012 are used, in addition to monthly mean sensible heat flux based on the OAFlux dataset (Yu et al. 2008).
The SAM signature in austral winter (June–August) has been extracted through an EOF analysis that was applied to monthly mean anomalies of zonally varying 925-hPa U poleward of 20°S from which the remote influence of El Niño–Southern Oscillation (ENSO) had been statistically removed. This removal was based on the regression coefficient of the U anomalies (e.g., Honda et al. 2001) evaluated locally with the Southern Oscillation index (SOI) available from the Japan Meteorological Agency (http://www.data.jma.go.jp/gmd/cpd/db/elnino/index/soi.html). The removal of the ENSO influence was possible throughout the analysis period by using monthly mean data (e.g., Renwick and Revell 1999).
b. Consistency of the SAM characteristics with the aquaplanet experiment
Monthly anomalies of 925-hPa U as a wintertime SAM signature defined in the preceding subsection are shown in Fig. 12, as a linear regression map against the first PC time series. From a hemispheric viewpoint, the wind anomalies exhibit a high degree of zonal symmetry, representing a seesaw in U values between the subtropical and subpolar latitudes. Nevertheless, the positions and strength of the centers of action of the seesaw differ from one ocean basin to another. While this moderate zonal asymmetry in the wintertime annular mode has already been pointed out previously (e.g., Codron 2007), no study has discussed its relation to the underlying oceanic conditions. In fact, Fig. 12 indicates that the nodal latitude of the SAM-related U variability tends to follow the latitude of SST front, which is consistent with the results in our aquaplanet experiments.
Furthermore, the regime-like characteristics of the annular mode simulated in our aquaplanet experiments are also found in the observed U composited separately for the strong positive and negative events of the SAM. As shown in Figs. 13a,c, the latitude of 925-hPa U maximum composited for the positive SAM events averaged over the Pacific is displaced poleward by approximately 10° relative to its counterpart for the Atlantic and Indian Oceans, which appears to be in agreement with our aquaplanet experiment. The latitudinal difference of the PFJ axis seems to be in accordance with the corresponding latitudinal difference in the aquaplanet experiments. For the negative SAM events (Figs. 13b,c), by contrast, the corresponding interbasin latitudinal differences in the 925-hPa U maximum are diminished, which is again in agreement with the aquaplanet experiments (Fig. 5). As shown in Fig. 14a, probability for the latitude of the observed wintertime U maximum averaged for the South Pacific sector (170°–120°W) exhibits broad peaks in the middle and subpolar latitudes, hinting at a regime-like behavior of the low-frequency U variability in this sector. Essentially the same argument can be made for the latitude of the lower-tropospheric storm track, if applied to the composites for the positive and negative events (not shown). In fact, the OAFlux data reveal that the cross-frontal gradient of upward sensible heat flux averaged within each ocean basin tends to be stronger in the positive phase (not shown), which suggests that the oceanic baroclinic adjustment acts more efficiently in the positive phase than in the negative phase. Therefore, the modest zonal asymmetry of the wintertime SAM structure may be understood as the interbasin differences in the SAM signature yielded by the corresponding differences in the SST front latitude.
It is noteworthy that the dual-peak feature by the probability for the PFJ latitude is diminished in the south Indian Ocean sector (50°–100°E; Fig. 14b) with an SST front observed at 45°S, seemingly inconsistent with the aquaplanet experiment with the SST front prescribed at the same latitude (Fig. 14d). It can be interpreted, however, from a viewpoint of the distance between the SST front and the particular latitude at which the PFJ is driven by atmospheric internal dynamics in the negative phase of the annular mode (Fig. 5). In fact, this particular latitude differs by about 4° between the aquaplanet experiments (38°S) and the observations (42°S). Taken this latitudinal difference into account, it seems more reasonable to compare the aquaplanet experiments with the SST front at 50° and 40° latitude to the observed U variability in the South Pacific and south Indian Ocean, respectively. In fact, the probability for the latitude of [U] axis in the experiment with SST front at 50° latitude (Fig. 14e) also suggests a regime-like behavior as in the case with SST front at 55° latitude (Fig. 14c), whereas such a behavior is ambiguous with the SST front at 40° latitude (Fig. 14f), as actually observed in the south Indian Ocean (Fig. 14b). The latitudinal excursion of the PFJ axis between the positive and negative phases of the SAM (as indicated by arrows along the abscissa in Fig. 14) is greater over the South Pacific than over the south Indian Ocean, in agreement with the corresponding aquaplanet experiments (Figs. 13c,d). Again, the observed difference is in better agreement with the comparison between the experiments with the SST fronts at 50° and 40° latitude (Figs. 14e,f) than with the fronts at 55° and 45° latitude (Figs. 14c,d), although such a subtle difference is less clear for the low-level storm track latitude (not shown). For the regime-like behavior of the wintertime annular mode, it is therefore important for the SST front to be situated poleward, by at least several degrees in latitude, of the particular PFJ latitude determined by atmospheric internal dynamics, as confirmed consistently in both aquaplanet experiments and observed facts (see arrows in Figs. 14a,b,e,f).
5. Interpretation of the climatological-mean state
Regime-like characteristics of the wintertime annular mode discussed in the preceding sections can leave some imprints in the climatological statistics. Figures 15a–d show the climatological-mean profiles of 925-hPa [U] (black line) and the composited profiles for the strong positive and negative events (red and blue lines, respectively) of the model annular mode in the experiments with the SST fronts located at 45°, 55°, 40°, and 50° latitude, respectively. Each of the climatological-mean profiles exhibits a hint of dual peaks that reflect the two quasi-equilibrium states or regimes represented by the two phases of the annular mode, and the climatological-mean state can therefore be interpreted as the superposition of the two regimes corresponding to the two phases of the annular mode. The latitude of the equatorward peak in the climatological-mean state, although rather ambiguous for the SST front at 40° latitude (Fig. 15c), is almost the same among those experiments, corresponding to the negative phase of the annular mode represented in Fig. 5a. In contrast, the latitude of the poleward peak differs, corresponding to the latitudinal difference of the SST front, which can be understood as an imprint of the positive phase represented in Fig. 4a.
Furthermore, Figs. 15b,d lead to an interpretation for the counterintuitive weakness in the climatological sensitivity of the latitude of [U] maximum to the latitude of an SST front if located at subpolar latitude (Ogawa et al. 2012). In fact, the equatorward peak of the climatological-mean [U] in the experiments with subpolar SST fronts (at 50° and 55° latitude) is more representative of the negative phase of the annular mode, whose persistence is higher than its counterpart for the positive phase (Fig. 7). The particular phase represents a regime in which atmospheric internal dynamics is dominant, thus showing no sensitivity to the latitude of an SST front. This interpretation supports the argument by Ogawa et al. (2012) that the climatological weak sensitivity of the latitude of PFJ to subpolar SST fronts may be due to the dominant contribution of atmospheric internal dynamics that is unrelated to the lower boundary condition (Robinson 2006).
The aforementioned interpretation is also applicable to the wintertime climatological-mean state actually observed over the SH. Figures 15e and 15f show the meridional profiles of 925-hPa U in the climatological mean and the composites for the individual phases of the SAM for the south Indian Ocean and the South Pacific, respectively. As pointed out by Nakamura et al. (2004), the wintertime climatological-mean PFJ latitude depends on the latitude of SST front in each of the ocean basins (black lines in Figs. 15e,f). This dependence arises basically from the positive phase of SAM (red lines in Figs. 15e,f), but not from its negative phase (blue lines in Figs. 15e,f), which is in good agreement with the aquaplanet simulations. The climatological-mean U profiles observed over the south Indian Ocean and the South Pacific resemble the results of the experiments with SST fronts located at 40° and 50° latitude, respectively (Figs. 15c,d), rather than those with the SST fronts located at 45° and 55° latitude (Figs. 15a,b). We therefore argue that it is the meridional distance of an SST front from the particular latitude of PFJ determined by atmospheric internal dynamics in the negative phase of the annular mode that is important not only for the annular mode characteristics discussed in the section 4b but also for the meridional structure of the climatological-mean state represented by the superposition of the two regimes that correspond to the individual phases of the annular mode.
6. Discussion and conclusions
In contrast to previous studies on the climatological-mean impacts of a midlatitude oceanic frontal zone on a storm track and eddy-driven PFJ, the present study has demonstrated that the importance of an oceanic frontal zone can strongly influence the characteristics of the tropospheric annular variability in winter and thereby the climatological-mean state. The most significant finding from our experiments is the regime-like behavior of the annular mode when the oceanic front is located in extratropical latitudes as actually observed. The regime represented by the positive phase of the annular mode tends to be maintained through enhanced meridional gradient of upward sensible heat flux at the frontal latitude (Fig. 8), which is important for the recurrent development of baroclinic eddies to form a storm track and thereby the formation of PFJ (Nakamura et al. 2004, 2008). As suggested in the theory of baroclinic instability from a PV viewpoint (Hoskins et al. 1985) and indeed confirmed climatologically in aquaplanet experiments (e.g., Sampe et al. 2010, 2013) and in more realistic model simulations (e.g., Taguchi et al. 2009; Nonaka et al. 2009), a surface baroclinic zone anchored along an oceanic frontal zone is essential for recurrent development of baroclinic eddies through the coupling of PV anomalies both at the surface and near the tropopause, rather than the midtropospheric baroclinicity associated with the vertical shear of STJ. Indeed, this thermodynamic influence of the oceanic frontal zone is evident in the positive phase of the annular mode, which is characterized by the close association of a lower-tropospheric storm track and PFJ with an oceanic frontal zone (Figs. 4a,b). Although the PFJ axis can fluctuate purely as atmospheric self-excited variability (Limpasuvan and Hartmann 2000), the eddy development can be activated once the PFJ axis approaches in the vicinity of an oceanic frontal zone at a middle or subpolar latitude. In this situation, the eddy momentum and heat fluxes act to anchor PFJ over the cool ocean surface slightly poleward of the frontal latitude. The residence time of the PFJ axis around the frontal zone can thus be prolonged to constitute a dynamical regime represented as the positive phase of the annular mode. While a considerable contribution of low-frequency wave disturbances to the maintenance of anomalous low-level baroclinicity associated with the annular mode has been pointed out (Zhang et al. 2012; Nie et al. 2013), the low-frequency wave disturbances also contribute to the polarity change of the wintertime SAM (Shiogama et al. 2005), which is in agreement with aquaplanet experiments (not shown). The wintertime transition events of the annular mode in aquaplanet experiments will be discussed in another paper.
During the negative phase of the annular mode, by contrast, the PFJ axis can stay far away from the frontal zone. Although low-level eddy heat flux is enhanced locally in the vicinity of the oceanic front (Fig. 5c), the storm-track latitude is insensitive to the frontal latitude (Fig. 5b). In this regime, the PFJ and storm track are uncoupled from the oceanic front and the latitude of their joint axes seem to be determined by atmospheric internal dynamics, as indicated by the distinct peak of their residence time around 38° latitude regardless of the latitude of the oceanic front (Fig. 7). To clarify barotropic or baroclinic mechanisms involved in the internal dynamics in this phase, further analysis, as in Nie et al. (2014), will be carried out in our future study. The regime-like behavior tends to be most apparent for a subpolar oceanic front, which acts to anchor the storm track in the positive phase of the annular mode far away from its latitude in the negative phase. A schematic describing the regime-like behavior of the wintertime annular mode is shown in Fig. 16a.
Regime-like behavior of the wintertime SAM has already been pointed out in previous studies but from a rather phenomenological viewpoint (Yoden et al. 1987; Itoh et al. 1999). They argued that the positive phase of SAM represents the double-jet structure with distinct separation between PFJ and STJ while the negative phase represents the merged single-jet structure. The findings of the present study add further significance to their interpretations from a dynamical viewpoint. In fact, the present study emphasizes that the regime-like behavior of the wintertime SAM discussed in these previous studies can be understood in depth from a consideration of the influence of the oceanic front. Our aquaplanet experiments suggest that the particular regime represented by distinct double jet structure associated with the positive phase of SAM corresponds to the regime when the extratropical tropospheric circulation is under the enhanced influence of the oceanic front. One of the two attractors in phase space found by Itoh et al. (1999) based on reanalysis data may be related to the surface temperature gradient determined by the lower-boundary condition, whereas the other is formed by atmospheric internal dynamics unrelated to the lower-boundary condition (Fig. 16a). Therefore, the regime-like behavior of the annular mode as revealed in the bimodality of the probability density function in the zonal wind axis (Figs. 9a–c) tends to be more apparent when the SST front is located farther away from the particular latitude determined by the atmospheric internal dynamics. The marked “split-flow regime” over the wintertime South Pacific as discussed in Bals-Elsholz et al. (2001) is owing probably to the location of the oceanic frontal zone at 55° latitude. On the contrary, the regime-like characteristic is ambiguous in the south Indian Ocean even in winter (Fig. 14b), where the SST front is located at 45° latitude, only about 3° poleward of the particular latitude (42°) that is considered to be determined by the atmospheric internal dynamics (Fig. 15e). This observed situation is consistent with the ambiguous regime-like behavior simulated in a particular aquaplanet experiment with the SST front located at 40° latitude, which is only about 2° poleward of the particular latitude (38°) determined by the internal dynamics in the model. The regime-like behavior is thus almost absent in the situation where the SST front is too close to the particular latitude determined by the atmospheric internal dynamics, which is rather similar situation to our NF experiment. Figure 16b schematically describes the annular mode that could be realized when SST front is located near the particular latitude or absent. The annular mode in either of these situations represents a small wobble of PFJ around the climatological-mean state without regime-like behavior.
Such an oceanic influence on the regime-like behavior of the annular mode has not been found in a previous attempt with aquaplanet experiments by Michel and Rivière (2014), presumably because the tropical SST distribution prescribed in their model is not quite realistic and the midlatitude SST gradient prescribed seems weaker than in the observations. Furthermore, unlike the present study, they did not take into account the seasonality of the SAM signature as observed by Codron (2005, 2007). In fact, the regime-like behavior in our aquaplanet experiment is obvious in winter but less so under the summertime SST profile (not shown), which is consistent with the observations. The weakening of the regime-like character may be related to the absence of climatological STJ in the summer hemisphere, which will be discussed in a future study.
The present study has shown that the wintertime climatological-mean state of the extratropical tropospheric circulation may be understood as a superposition of the distinct regimes that correspond to the two phases of the annular mode both in the observations and our aquaplanet experiments (Fig. 15). We argue that the wintertime extratropical tropospheric circulation and its dominant variability (annular mode) can no longer be interpreted purely as atmospheric internal processes but should be interpreted from a viewpoint of their coupling with the underlying ocean. Although in reality the direct thermodynamic forcing by a midlatitude oceanic front is limited to a narrow latitudinal band within a certain longitudinal extent, its influence can be extended zonally as downstream eddy development along a storm track and also meridionally through eddy propagation and associated momentum transport. It is thus suggested that an appropriate representation of midlatitude oceanic frontal zones in global climate models is required for the realistic reproduction of the wintertime annular mode variability, and thereby the climate reproduction and future projection. In fact, through their AGCM experiment and analysis of outputs of multiple climate models that participate in phases 3 and 5 of the Coupled Model Intercomparison Project (Meehl et al. 2007; Taylor et al. 2012), Ogawa et al. (2015) have shown that the SH midlatitude oceanic fronts can be a critical factor in reproducing the observed climatic trend induced by stratospheric ozone depletion in the late twentieth century, by maintaining the SAM structure and its vertical connection with the stratospheric variability. Furthermore, consideration of the SH oceanic fronts may help interpret other large-scale SH atmospheric variability such as the baroclinic annular mode (BAM; Thompson and Barnes 2014; Thompson and Woodworth 2014). Represented as the pulsing of the eddy kinetic energy, the BAM signature in the SH may be controlled by the surface baroclinic zones along oceanic fronts and will therefore need to be addressed in a future study.
It should be pointed out that the ocean currents that maintain the oceanic fronts in the midlatitude SH are driven by the surface westerlies, which are forced by transient eddies along the storm track. We have nevertheless confirmed that latitudinal profiles of SST gradient in the individual basins of the wintertime Southern Ocean are almost unchanged between the positive and negative phases of SAM (not shown), which justifies our comparison of the observed SAM characteristics with their counterpart in the aquaplanet experiments conducted under the fixed SST profiles. Still, anomalous surface westerlies associated with SAM can in turn impact midlatitude SST (Ciasto and Thompson 2008). Furthermore, the latitudes of oceanic frontal zones may shift in the future, responding to a trend in the surface westerlies that may be induced by the global warming (e.g., Beal et al. 2011). We argue that deeper understanding of the impacts of midlatitude oceanic frontal zones on the observed extratropical circulation should be made from a viewpoint of fully coupled ocean–atmosphere interactions.
The authors thank the three anonymous reviewers for their constructive comments. The use of the AFES was made possible in support of Japan Agency for Marine-Earth Science and Technology (JAMSTEC). This study is supported in part by the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) through a Grant-in-Aid for Scientific Research in Innovative Areas 2205 and Grant-in-Aid for Scientific Research 25287120 and by the Japanese Ministry of Environment through the Environment Research and Technology Development Funds A-1201 and 2-1503, together with the Nordforsk GREENICE project 61841.
Current affiliation: Graduate School of Bioresources, Mie University, Mie, Japan.
This article is included in the Climate Implications of Frontal Scale Air–Sea Interaction Special Collection.