1. Introduction

White and Peterson (1996) and Jacobs and Mitchell (1996) found monthly sea surface height (SLH), sea surface temperature (SST), sea level pressure (SLP), and sea ice extent (SIE) anomalies in the Southern Ocean from 1982 to 1995 dominated by a broad interannual signal of a 3- to 7-yr period in zonal wavenumber– frequency spectra, propagating slowly eastward around the Southern Ocean in fixed phase with one another. They found that the four variables were dominated by global zonal wavenumber 2, with individual phases taking about 8 years to circle the globe at ∼45° of longitude per year (i.e., ∼0.08 m s⁻¹). They called this interannual signal the Antarctic circumpolar wave (ACW).

Recently, we refined these results by conducting multitaper-method singular-value-decomposition (MTM-SVD) analysis on monthly SST, SLP, and sea ice concentration (SIC) anomalies extending in latitude from 30° to 90°S (centered on the fall–winter–spring sea ice edge) over the 20 years from 1982 to 2001. The corresponding spectrum of local fractional variance revealed five signals dominating climate variability near the sea ice edge; that is, one southern annular mode (SAM) signal near a 1.0-yr period and four ACW signals near 2.9-, 3.7-, 7.1‐, and 17-yr periods, each characterized by eastward phase propagation. Together, these five signals explained ∼50% of the variance in unfiltered winter SIC anomalies in the Ross and Weddell Seas, with the ACW at a 3.7-yr period dominating neighboring ACW signals by a factor of nearly 2 to 1 from 1983 to 1992.

Thus, the broadscale ACW signal of a 3–7-yr period observed by White and Peterson (1996) and Gloersen and White (2001) was composed of three narrowband ACW signals at 2.9-, 3.7-, and 7.1-yr periods, with the ACW signal at a 3.7-yr period dominating the two neighboring signals from 1983 to 1992. It is characterized by global zonal wavenumber 2 with wave characteristics similar to those observed in the broadscale ACW observed by White and Peterson (1996) over this same epoch. In the present study, we have isolated the dominant ACW signal of the 3.7-yr period from its neighbors by appropriate bandpass filtering (Kaylor 1977). This allows us to investigate its character uncontaminated by interference from its neighbors. We focus on its behavior from 1983 to 1992, allowing us to continue to call it the ACW.

Modeling studies of the ACW (e.g., White et al. 1998; Baines and Cai 2000; White and Chen 2002) determined that its eastward phase propagation depends upon coupling between ocean and atmosphere, not on the eastward advection by the Antarctic Circumpolar Current (ACC). Thus, the ACW does not follow the ACC through the Southern Ocean. Rather, Peterson and White (1998), White and Chen (2002), and White et al. (2002) found the ACW composed of two main tracks; that is, a northern track following the Subtropical Front across the eastern Atlantic, Indian, and western and central Pacific sectors between 30° and 45°S and a southern track along the fall–winter–spring sea ice edge forming around Antarctica each year near 63°S. Both tracks converge at Drake Passage as the ACW propagates from the eastern Pacific sector to the western Atlantic sector.

The main issue in understanding ocean–atmosphere coupling in the ACW is the troposphere response to SST and SIE anomalies. White and Chen (2002) explained this response to SST anomalies in the ACW along the Subtropical Front by diagnosing the troposphere thermal, vorticity, and potential vorticity budgets along this track. They found warm SST anomalies driving upward latent heat flux and greater precipitation, generating upper-level diabatic heating and low-level diabatic cooling in the troposphere. The former is balanced mainly by upward thermal advection and the latter is balanced mainly by poleward thermal advection. This yields anomalous ascending motion throughout the troposphere (i.e., deep convection), accompanied by corresponding low-level convergence and upper-level divergence, and poleward low-level wind anomalies. In the vorticity budget, the anomalous low-level convergence is balanced mainly by the anomalous poleward advection of planetary vorticity, also yielding poleward low-level wind anomalies. In the potential vorticity budget, the anomalous vertical gradient of diabatic heating is balanced at the low level by a combination of anomalous poleward advection of mean vortex tubes and planetary vorticity, both of similar magnitude. This explains why poleward meridional surface wind (MSW) anomalies come to be collocated with warm SST anomalies in the ACW along this track. In the present study, we conduct a similar diagnostic study on the southern track of the ACW along the fall–winter–spring sea ice edge around Antarctica.

The ACW along the fall–winter–spring sea ice edge has anomalous retracted SIE, warm SST, poleward MSW, and reduced SIC propagating eastward together around more than three-quarters of the Southern Ocean from 30° eastward to 110°E (Gloersen and White 2001). The memory of the ACW in the sea ice pack from one austral winter to the next is conveyed by the upper-ocean diabatic heat storage associated with SST anomalies in the ACW. Yet, this raised the question as to whether SIE and SIC anomalies along the sea ice edge act as passive tracers for the ACW or whether they participate in a coupled interaction between the ocean and atmosphere. Deser et al. (2000) found SIE anomalies actively participating in the coupling by producing sensible-plus-latent heat flux anomalies with magnitudes on the order of the climatological estimates themselves, much larger than those associated with SST anomalies over the open ocean. They found this to have a significant impact on the troposphere circulation near the sea ice edge in the North Atlantic Ocean. In the present study, we find SIE anomalies in the ACW signal along the fall–winter–spring sea ice pack around Antarctica producing upward latent heat flux anomalies, but with magnitudes similar to those driven by SST anomalies along the northern track of the ACW (White and Chen 2002). However, we find it having much greater impact on tropospheric circulation at the sea ice edge, generating deep convection and meridional surface wind responses 2–3 times those along the Subtropical Front.

To understand this, we seek the dominant thermodynamic balances that govern the ACW along the fall– winter–spring sea ice edge by diagnosing its tropospheric thermal and potential vorticity budgets. We utilize the National Centers for Environmental Prediction– National Center for Atmospheric Research (NCEP– NCAR) global atmosphere reanalysis (Kistler et al. 2001), following the methodology of White and Chen (2002). We conduct this diagnostic study for the 10 years from 1983 to 1992 when the 3.7-yr period ACW signal was robust. We find anomalous SIE-induced latent heat flux driving mid- to upper-level diabatic heating and low-level cooling, the former balanced mainly by upward thermal advection and the latter balanced by poleward thermal advection as in White and Chen (2002) but in different proportion. Even so, this balance explains the collocation of anomalous poleward MSW and deep convection with retracted SIE anomalies. It also reveals a new finding; that is, the anomalous SIE-driven deep convection along the fall–winter–spring sea ice edge instigates anomalous Ferrell cells and zonal (Walker-like) cells. The anomalous zonal cells teleconnect opposite phases of the ACW signal along the sea ice edge, while the Ferrell cells teleconnect the various phases of the ACW along the sea ice edge equatorward across the Southern Ocean to opposite phases in the ACW along the Subtropical Front.

2. Data and methods

In this study we analyze 13 oceanic and tropospheric variables, and their various horizontal and vertical derivatives, from three sources. We analyze monthly SST, SLP, latent heat flux (Q_E), air temperature (T), wind (V), divergent wind (V_D), pressure velocity (ω), and the horizontal eddy heat flux divergence (div 〈V″T″〉, where V″ and T″ represent 6-hourly deviations about the monthly mean, and where angle brackets represent the monthly mean operator) from the NCEP–NCAR atmosphere reanalysis dataset (Kistler et al. 2001). Over the Southern Ocean the NCEP–NCAR reanalysis model is driven by the Reynolds SST dataset from the Advanced Very High Resolution Radiometer (AVHRR) satellite sensors after 1982 (Reynolds and Marsico 1993) and the Goddard Space Flight Center (GSFC) SIC dataset from SSM/I satellite sensors after 1979 (Gloersen et al. 1992), with the Television Infrared Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) troposphere temperature and water vapor profile dataset assimilated daily into the model after 1979 (Rao et al. 1990). We also analyze the NCEP Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) reanalysis dataset (Xie and Arkin 1997). Over the Southern Ocean, the precipitation (PCP) estimates from CMAP derive mainly from the NCEP– NCAR reanalysis model, depending on the assimilation of TOVS data after 1979.

We also analyze the monthly GSFC SIC dataset distributed by the National Snow and Ice Data Center (1998). The SIC estimates are processed into SIE estimates (in kilometers) using both 15% and 50% SIC criteria to identify the sea ice edge. Here we form seasonal averages for the fall, winter, and spring sea ice edge. SIE anomalies are positive northward (Gloersen et al. 1992), but we display their negative value so that retracted SIE anomalies have the same sign as warm SST anomalies with which they covary (Gloersen and White 2001). These SIC data were used as boundary conditions for the NCEP–NCAR reanalysis model (Taylor et al. 2000), in which the sea ice edge around Antarctica is defined using a 50% SIC criterion, and requiring 100% SIC (0% SIC) poleward (equatorward) of the sea ice edge with no gradation in between.

We also analyze monthly extratropical cyclone density (ECD) and the extratropical cyclone intensity (ECI) datasets from 1979 to the present, both variables being derived from a state-of-the-art cyclone tracking scheme developed by Simmonds and Keay (2000). Their method was applied to a 6-h NCEP–NCAR global reanalysis dataset (Kistler et al. 2001). The ECD is defined as the number of cyclones (per unit area) per 6-h “analysis,” while the ECI is defined by the Laplacian of sea level pressure at the center of the cyclone, with units of hPa (°lat)⁻².

Monthly/seasonal estimates for each variable were interpolated onto a common 2° latitude × 5° longitude grid determined by White (1995) to be optimal for resolving basin-scale climate variability over the global upper ocean. Monthly/seasonal anomalies were computed about the long-term monthly/seasonal means over the 20-yr record from 1982 through 2001. Since MTM-SVD analysis of monthly SST, SLP, and SIC anomalies in the Southern Ocean from 1982 to 2001 yielded peak local fractional variance for the ACW near the 3.7-yr period, we isolate this signal from weaker neighboring signals at 2.9- and 7.1-yr periods. This is accomplished by bandpass filtering time sequences of monthly anomalies using a filter with a frequency response function that is flat with steep sides and negligible side lobes, setting half-power points at 3- and 6-yr periods (Kaylor 1977). Bandpass filtering is applied over the entire 20-yr record from 1982 through 2001. However, prior to bandpass filtering we applied maximum entropy spectral analysis (Andersen 1974) to extend the record length. This extension allows the half-power point criterion to be preserved in the frequency response function, and it allows over half the variance of the ACW signal near the end points of the record to be faithfully represented (White 2000).

3. Global patterns of SST, SLP, and SIC anomalies in the ACW

The MTM-SVD analysis of monthly SST, SLP, and SIC anomalies centered on the sea ice edge around Antarctica from 1982 to 2001 revealed the broadscale ACW of 3- to 7-yr periods observed earlier by White and Peterson (1996) and Gloersen and White (2001) to be composed of three narrowband ACW signals near the 2.9-, 3.7-, and 7.1-yr periods (see the introduction). The ACW signal near the 2.9-yr period is characterized by global zonal wavenumber-3 variability (e.g., Cai and Baines 2001), while the ACW signals near the 3.7- and 7.1-yr periods are characterized by global zonal wavenumber-2 variability. This analysis found the ACW signal near the 3.7-yr period dominating its neighbors from 1982 to 1993, but thereafter all three signals contributed about equally to the broadscale interannual variability.

Here, we display the phase sequences of the complex singular value decomposition mode for the ACW signal of a 3.7-yr period (Fig. 1) from the joint MTM-SVD analysis of SST, SLP, and SIC anomalies. The evolution of SST, SLP, and SIC variability in this signal over one-half cycle (180° of phase, Fig. 1) differs only subtly from that observed in the broadscale ACW signal (e.g., White and Peterson 1996; Gloersen and White 2001; Cai and Baines 2001). The evolution of each variable is characterized by global zonal wavenumber-2 variability, propagating eastward at ∼45° of longitude per year, with warm SST weights equatorward of the sea ice edge covarying with reduced SIC weights poleward, with high SLP weights displaced ∼90° of phase to the east. Perhaps the most interesting difference between the evolution of this narrowband ACW signal and that of the broadscale ACW signal (Cai and Baines 2001) is the greater ratio of traveling to standing wave mode amplitude observed in each of the phase sequences. For the remainder of this study, we focus attention on this ACW signal of the 3.7-yr period, referring to it simply as the ACW (see section 2, data and methods). We focus on its behavior from 1983 to 1992 when it dominated neighboring signals.

4. Zonal phase relationships among SST, SIE, ECD, ECI, and SLP anomalies in the ACW

Now, we construct time–longitude diagrams of interannual SST, −SIE, ECD, ECI, and SLP anomalies along the mean fall–winter–spring sea ice edge (defined by the 15% SIC criterion, Fig. 2a) over the 10 yr from 1983 to 1992. Each time–longitude diagram extends zonally around the globe from 30° to 30°E (Fig. 2b). Here, we see all five variables composed of two zonal wavelengths, propagating eastward at ∼45° of latitude per year over more than three quarters of the globe eastward from 30° to 110°E longitude. The zonal phase relationship among the five variables can be discerned qualitatively by referencing them against the sloping black line and quantitatively by computing zonal-lag cross-correlations among the five variables (Fig. 2c). These zonal-lag cross-correlations find warm SST anomalies aligned with retracted SIE anomalies (both positive): with low SLP anomalies displaced ∼90° of phase to the west. They also find low SLP anomalies collocated with greater ECD and ECI, the interpretation of which is given in section 9.

5. Zonal phase relationships among SIE, Q_E, PCP, D₈₅₀, and D₂₀₀ anomalies in the ACW

To test the Deser et al. (2000) hypothesis, we display time–longitude diagrams of interannual −SIE, Q_E, PCP, divergence at 850 hPa (D₈₅₀), divergence at 200 hPa (D₂₀₀) anomalies along the fall–winter–spring sea ice edge (defined by the 50% SIC criterion, Fig. 3a) over the 10-yr record. Each time–longitude diagram extends zonally around the globe from 30° to 30°E (Fig. 3b). Here we find the ACW signal propagating eastward from 30° to 110°E in each of the five variables, though Q_E and PCP anomalies display a significant standing mode characterized by nodes in the Amundsen and Bellingshausen Seas. The zonal phase relationship among the five variables can be discerned qualitatively by referencing them against the sloping black line and quantitatively by computing zonal-lag cross-correlations among the five variables (Fig. 3c).

The zonal-lag cross-correlations between positive −SIE anomalies (retracted toward Antarctica) and negative D₈₅₀ anomalies (low-level convergence), and positive D₂₀₀ anomalies (upper-level divergence) find the influence from SIE anomalies in the ACW penetrating from the sea surface to the top of the troposphere, associated with deep convection. Retracted SIE anomalies are also collocated with greater upward Q_E and higher PCP, as expected from Deser et al. (2000). This also fulfills the necessary condition for SIE anomalies to drive the overlying troposphere (White and Chen 2002). The magnitudes of anomalous SIE-induced Q_E and PCP, ranging over ±4 W m⁻² and ±10 mm month⁻¹, respectively, are nominally the same as those induced by SST anomalies in the ACW along the subtropical front (White and Chen 2002).

6. Diagnosis of the tropospheric thermal budget of the ACW

We diagnose the thermal budget of the troposphere for the ACW along the sea ice edge; that is,

View Expanded

where T is the temperature, V the wind velocity, θ the potential temperature, and ω the pressure velocity in the p direction; div 〈V″T″〉 is the eddy thermal flux divergence in units of kelvins per second, Q^′_D/(ρ_AC_PA) is the diabatic heating in units of kelvins per second, C_PA is the specific heat of air, and ρ_A is the density of air decreasing upward; ( ) and ( )′ indicate the long-term monthly mean and interannual anomaly, respectively. We estimate the anomalous diabatic heating [Q^′_D/(ρ_AC_PA)] on the right-hand side of Eq. (6.1) as the residual of the sum of terms on the left-hand side.

We follow the covariance methodology of White and Chen (2002), computing the zonal-lag regression coefficients between −SIE anomalies and the anomalous components in Eq. (6.1) as a function of pressure level in the troposphere from 900 to 200 hPa (terms a through e) in Fig. 4 over the time–distance diagram that follows the track of the ACW along the sea ice edge around Antarctica for the 10 yr from 1983 to 1992. In this case, the usual regression coefficients are multiplied by the standard deviation of the SIE anomalies to yield temperature tendency units of Kelvins per second.

The zonal-lag regression finds retracted SIE anomalies associated with greater Q_E and higher PCP (Fig. 4g). We find anomalous SIE-induced latent heat flux and precipitation associated with diabatic heating at mid to upper level and diabatic cooling at low level (Figs. 4e and 4g). This profile of diabatic heating increases upward from a maximum negative estimate of ∼−0.8 × 10⁻⁶ K s⁻¹ near 900 hPa to a positive maximum estimate of ∼1.6 × 10⁻⁶ K s⁻¹ near 300 hPa. It is balanced by the sum of the horizontal and vertical thermal advection (terms a, b, and c), with the eddy thermal flux divergence negligible (d). The mean advection of anomalous temperature and the anomalous advection of mean temperature tend to cancel each other (terms a and b), but the latter dominates the former throughout the column, yielding a net warming tendency ranging from ∼1.4 × 10⁻⁶ at the low level to ∼0.8 × 10⁻⁶ K s⁻¹ at the upper level. At the low level (upper level) this warming tendency overwhelms (is overwhelmed by) the cooling tendency from the vertical thermal advection, increasing upward from ∼0.8 × 10⁻⁶ at the low level to ∼2.2 × 10⁻⁶ K s⁻¹ at the upper level (c). Thus, the anomalous SIE-induced Q_E directed upward into the troposphere appears to drive mid- to upper-level diabatic heating (presumably through higher precipitation) in the relative absence of horizontal eddy flux divergence (d). The resulting diabatic heating profile is balanced mainly by vertical thermal advection at the upper level and net horizontal thermal advection at the low level, the latter dominated by anomalous poleward advection of mean temperature.

The association between retracted SIE anomalies, greater upward Q_E, higher PCP, and higher mid- to upper-level diabatic heating suggests that retracted SIE anomalies drive deep diabatic heating in the troposphere. Indeed, the vertical integral of the mass-weighted diabatic heating profile directly over the retracted SIE anomaly is ∼0.7 W m⁻², with low-level diabatic cooling tending to compensate mid- to upper-level diabatic heating in the mass-weighted integral. This estimate is ∼1/2 the anomalous Q_E at the sea surface of ∼1.3 W m⁻² but is equivalent to the net heat flux of ∼0.7 W m⁻² into the column generated by the higher PCP of ∼0.6 mm month⁻¹. This integral balance indicates that anomalous SIE-induced Q_E and PCP are sufficient to account for the net diabatic heating occurring in the tropospheric column.

The center of low-level diabatic cooling is located ∼45° of phase to the east of retracted SIE anomalies near 900 hPa, achieving a maximum value of ∼−0.8 × 10⁻⁶ K s⁻¹ in Fig. 4e. It is balanced principally by the warming tendency from the anomalous poleward advection of mean temperature (Fig. 4b). Even farther to the east near 90° of phase, this low-level diabatic cooling is associated with the anomalous downward latent heat flux and represents an intrinsic feedback from atmosphere to ocean. Since the latter is displaced ∼90° of phase to the east of the retracted SIE anomalies, it yields an anomalous warming tendency in the upper ocean that contributes to some or all of the eastward phase propagation of the ACW along the sea ice edge (White et al. 1998).

If vertical thermal advection was the only term on the left-hand side of Eq. (6.1) balancing the diabatic heating on the right-hand side, then the profile of the pressure velocity would resemble that of diabatic heating with ascent at the mid- to upper-level and descent at the low level (Fig. 4). Yet, we observe anomalous SIE-induced ascending motion of uniform sign throughout the column, maximum at mid level (f), approximately collocated with anomalous SIE-induced upward Q_E (g). Thus, ascending motion derives from the vertical heat advection (c) required to balance the warming tendency from diabatic heating (e) and from net horizontal thermal advection (terms a and b).

7. Diagnosis of the tropospheric potential vorticity budget in the ACW

The importance of the net horizontal thermal advection in the thermal budget [Eq. (6.1)] of the ACW along the sea ice edge suggests that the net horizontal vortex tube advection is important in the potential vorticity budget (White and Chen 2002); that is,

View Expanded

Here, the hydrostatic approximation (∂p/∂z = −ρ_Ag) has been assumed and the horizontal eddy flux divergence has been neglected, with N² representing the buoyancy frequency [(g/θ)∂θ/∂z] (Peixoto and Oort 1992, p. 49). Normally, derivation of the potential vorticity equation assumes the thermal wind balance to hold throughout the column [i.e., ∂V′/∂z = (g/Tf₀)kx∇T′ and ∂V/∂z = (g/Tf₀)kx∇T] (Peixoto and Oort 1992, p. 156) on global space scales and interannual time scales (Pedlosky 1987). Under this approximation, the third term in brackets on the left-hand side of Eq. (7.1) goes to zero and ∇T′ in the first term and ∇T in the second term can be expressed in terms of ∂V′/∂z and ∂V/∂z, respectively. However, we find the thermal wind approximation at the low level to be less accurate than at the mid- and upper levels, requiring all terms in Eq. (7.1) to be evaluated. This yields a residual estimate for Q^′_D/ (ρ_AC_PA) on the right-hand side of Eq. (7.1) consistent with the residual estimate for Q^′_D/(ρ_AC_PA) on the right-hand side of Eq. (6.1). So, Eq. (7.1) allows us to examine the relative importance of the absolute vorticity advection that is, V · ∇ζ′ + V′ · (∇ζ + ∇f), and the net vortex tube advection, that is,

View Expanded

in response to the vertical gradient of diabatic heating that is, ∂/∂z[Q^′_D/(ρ_AC_PA)(f₀g/(N²T))].

We examine vertical sections of zonal-lag regression coefficients between components in Eq. (7.1) and −SIE variability directly underneath (Fig. 5) in a manner consistent with Fig. 4. We find the individual terms of the vortex tube advection (terms b, e, and f in Fig. 5) on the left-hand side of Eq. (7.1) to be larger than the individual terms of the absolute vorticity advection (terms a, c, and d), with maximum values in the former (latter) of ∼27 × 10⁻¹² s⁻² (∼16 × 10⁻¹² s⁻²), with anomalous advection of mean relative vorticity negligible everywhere (d). Yet, the mean advection of anomalous vortex tubes (b) and anomalous advection of mean vortex tubes (e) are of the opposite sign and tend to cancel. Even so, the latter dominates the former, yielding a sum that is negative maximum (∼−11 × 10⁻¹² s⁻²) at the low level near 700 hPa and negative minimum (∼−6 × 10⁻¹² s⁻²) at the mid- to upper level near 400 hPa. Closer to the sea surface near 900 hPa, the net geostrophic horizontal vortex tube advection (b and e) is offset by the ageostrophic horizontal vortex tube advection of similar magnitude (f). Near 700 hPa, the negative anomalous advection of planetary vorticity (c) is offset by the positive mean advection of anomalous relative vorticity (a). Thus, the negative maximum vertical gradient of diabatic heating of ∼−11 × 10⁻¹² s⁻² near 700 hPa (g) is balanced principally by the net vortex tube advection (b, e, and f). This indicates that the anomalous SIE-induced low-level circulation derives mainly from consideration of the thermal balance, not the vorticity balance.

At the mid- to upper level near 350 hPa, the vertical gradient of the diabatic heating goes to zero (g). Above that, near 200 hPa, the positive diabatic heating gradient is balanced principally by mean advection of anomalous relative vorticity (a). At this level, the net vortex tube advection terms on the left-hand side of [Eq. (7.1) (terms b, e, and f in Fig. 5)] mostly cancel one another, while the mean advection of anomalous relative vorticity of ∼16 × 10⁻¹² s⁻² dominates the anomalous meridional advection of planetary vorticity by ∼−7 × 10⁻¹² s⁻². Thus, the anomalous SIE-induced upper-level circulation derives mainly from consideration of the vorticity balance.

8. Anomalous Ferrell cells and zonal (Walker-like) cells in the ACW

The anomalous SIE-induced ascending motion in the ACW along the sea ice edge in fall–winter–spring (Fig. 4f) are connected to neighboring SIE-induced descending motion via divergent wind anomalies (V_D). To demonstrate this, we display a short animation sequence of the zonal vertical section of divergent wind anomalies in the ACW along the fall–winter–spring sea ice edge (defined by the 50% SIC criterion), extending zonally around the globe from 30° to 30°E each year for 3 yr from 1986 to 1988 (Fig. 6). Also displayed are the corresponding zonal profiles of the anomalous −SIE in the ACW along the sea ice edge, with retracted SIE anomalies positive. Here, we find anomalous ascending (descending) motion occurring nominally over the retracted (expanded) SIE anomalies, forming zonal circulation cells in the troposphere along the sea ice edge that are similar to zonal Walker cells in the Tropics (Bjerknes 1969). Moreover, we find the eastward propagation of SIE anomalies in the ACW over the 3 yr accompanied by eastward propagation of the corresponding zonal (Walker-like) cells, which connect corresponding anomalous ascending and descending motion along the sea ice edge. In these anomalous zonal cells, vertical wind speeds range over ±0.0013 m s⁻¹, while zonal divergent wind speeds range over ±0.40 m s⁻¹. Some of the anomalous zonal cells extend vertically from the low level near 900 hPa into the lower stratosphere above 200 hPa, while others extend from the low to midlevel near 500 hPa.

Next, we display the meridional vertical section of divergent wind anomalies in the ACW along 180° longitude, extending from the Ross Sea to New Zealand for 1986 and 1988 (Fig. 7). These two realizations show covarying warm (cool) SST adjacent to the sea ice edge and inferred retracted (expanded) SIE anomalies along the sea ice edge, associated with cool (warm) SST anomalies at the Subtropical Front near 30°S. In these anomalous Ferrell cells, vertical wind speeds range over ±0.0013 m s⁻¹ while meridional divergent wind speeds range over ±0.55 m s⁻¹. Thus, the ACW along the Subtropical Front near 30°S in the western Pacific sector is teleconnected to the ACW along the sea ice edge near 63°S in the Ross Sea via these anomalous Ferrell cells.

9. Discussion and conclusions

The ACW signal near the 3.7-yr period dominates the SAM signal near the 1.0-yr period and the three other ACW signals at the 2.9-, 7.1-, and 17-yr periods in explaining ∼50% of the variance in winter SIC anomalies in the Ross and Weddell Seas from 1983 to 1992. Here, we examined the troposphere response to SIE anomalies in this 3.7-yr period ACW signal along the fall–winter–spring sea ice edge forming around Antarctica each year. We began by displaying time–longitude diagrams of interannual SST, SIE, ECD, ECI, and SLP anomalies along the sea ice edge. All five variables displayed the familiar characteristics of the ACW eastward from 30° to 110°E (White and Peterson 1996), with low SLP anomalies collocated with increased extratropical cyclone density and intensity in the fall–winter–spring synoptic storm aggregate. Next, we displayed corresponding time–longitude diagrams of interannual SIE, Q_E, PCP, D₈₅₀, and D₂₀₀ anomalies along the sea ice edge. Again, these five variables displayed the standard ACW characteristics from 30° to 110°E, with retracted SIE anomalies collocated with greater upward latent heat flux and higher precipitation, low-level divergence, and upper-level divergence, the latter associated with deep convection. Thus, anomalous circulation throughout the troposphere propagated eastward with this ACW signal along the fall–winter–spring sea ice edge.

To establish the thermodynamics governing the troposphere response to anomalous SIE-induced latent heat flux and precipitation, we diagnosed the anomalous thermal and potential vorticity budgets of this ACW signal. In the thermal budget, we found anomalous SIE-induced latent heat flux and precipitation driving mid-to-upper-level diabatic heating and low-level cooling, similar to that observed along the northern track of the ACW (White and Chen 2002). We found the upper-layer diabatic heating balanced by vertical thermal advection and low-level diabatic cooling balanced by net horizontal thermal advection, together yielding ascending motion throughout the column and poleward wind at the low level. In the potential vorticity budget, we found the vertical gradient of diabatic heating balanced mainly by net horizontal vortex tube advection at the lower level and absolute vorticity advection at the upper level. This revealed a hybrid response of the troposphere circulation to anomalous SIE-induced diabatic heating; that is, with an equivalent barotropic meridional wind response deriving from a thermal balance at the low level and a vorticity balance at the upper level, summarized in a schematic diagram (Fig. 8).

The principal difference between this deep diabatic heating scenario in the ACW along the sea ice edge near 63°S and that along the Subtropical Front between 30° and 45°S (White and Chen 2002) lies in the relative intensity of the anomalous meridional advection of mean temperature at the low level in the thermal budget and the corresponding anomalous meridional advection of mean vortex tubes in the potential vorticity budget. In both budgets along the sea ice edge, these advective components are 2–3 times larger than over the Subtropical Front. Thus, along the sea ice edge the net horizontal advection of vortex tubes at the low level dominates the net advection of absolute vorticity in the potential vorticity budget, while along the Subtropical Front both advection terms are comparable (White and Chen 2002). This difference occurs because of the stronger background temperature gradient in the lower troposphere across the fall–winter–spring sea ice edge than across the Subtropical Front.

An outstanding question concerns the source of anomalous SIE-induced mid-to-upper-level heating and low-level cooling. This diabatic heating profile derives from some combination of release of latent heat through anomalous condensation and the net radiational heating from anomalous cloud fraction (Roads et al. 1998). One likely scenario goes like this; that is, low-level radiational cooling attends upper-level latent heat release when an abnormal number of high towers are generated in individual extratropical cyclones within the synoptic storm aggregate (White and Chen 2002). This scenario now seems more likely since we found low SLP anomalies in the ACW along the sea ice edge associated with increased extratropical cyclone density and intensity in the fall–winter–spring synoptic storm aggregate. This indicates that low SLP anomalies in the ACW are a proxy for increased extratropical cyclone activity. Further support for this hypothesis is given by the zonal phase relationship between greater PCP and low SLP anomalies, the former displaced ∼90° of phase to the east of the latter, similar to that occurring in individual extratropical cyclones (Browning 1990). Thus, it appears that extratropical cyclone activity near the sea ice edge is significantly affected by changes in SIE with amplitudes O(100 km) and wavelength scales O(4500 km). This remains to be tested in an ocean–atmosphere– cryosphere coupled model capable of significant synoptic storm development.

A feedback from atmosphere to ocean was observed in the diagnostics of the thermal budget of the ACW along the sea ice edge. Maximum low-level diabatic cooling was observed displaced ∼45° of phase to the east of retracted SIE anomalies, balanced principally by a warming tendency from the anomalous meridional advection of mean temperature. Farther to the east, this low-level diabatic cooling is associated with anomalous downward latent heat flux, the latter driving an anomalous warming tendency in the upper ocean. Since the latter is displaced ∼90° of phase to the east of retracted SIE anomalies, it contributes to the eastward phase propagation of this ACW signal along the sea ice edge. This phase displacement is greater than that observed in the broadscale ACW along the Subtropical Front where low-level diabatic cooling is associated with anomalous downward latent heat flux ∼45° of phase to the east of warm SST anomalies (White and Chen 2002).

We found anomalous SIE-induced deep convection in the troposphere along the fall–winter–spring sea ice edge driving anomalous meridional Ferrell cells equatorward from the sea ice edge and anomalous zonal cells zonally along the sea ice edge. In zonal vertical sections of anomalous divergent wind, neighboring retracted and expanded SIE anomalies in the ACW signal along the sea ice edge were teleconnected via corresponding zonal tropospheric cells. These are similar to zonal Walker cells teleconnecting warm and cool SST anomalies associated with ENSO in the tropical Indo–Pacific Ocean (Bjerknes 1969). In meridional vertical sections of anomalous divergent wind anomalies in the ACW signal, extending from the Ross Sea to the North Island of New Zealand, warm SST anomalies (and retracted SIE anomalies) at the sea ice edge were found to be teleconnected to cool SST anomalies in the ACW along the Subtropical Front near 30°S. Thus, the ACW signal along the sea ice edge in the Ross and Amundsen Seas were teleconnected to the ACW signal along the Subtropical Front in the western and central Pacific sectors of the Southern Ocean via anomalous Ferrell cells. It remains to determine the three-dimensional character of these Ferrell/Walker cells: that is, how they tie into the global spiral pattern of the covarying SST and SLP anomalies in the ACW over the Southern Ocean (White et al. 1998).