National Oceanography Centre, University of Southampton, Southampton, UK
Centre de Recherches sur la Terre et le Climat Georges Lemaître, Earth and Life Institute, Université Catholique de Louvain, Louvain-la-Neuve, Belgium
Corresponding author: S. E. Close, Centre de Recherches sur la Terre et le Climat Georges Lemaître, Earth and Life Institute, Université catholique de Louvain, Place Louis Pasteur 3, BE-1348 Louvain-la-Neuve, Belgium. (email@example.com)
 The time series of high-quality hydrographic measurements of the ACC in Drake Passage is revisited to investigate the extent of baroclinic adjustment in the region on time scales of up to two decades. We find that substantial adjustment of the upper kilometer of northern Drake Passage has occurred on interannual to decadal time scales, driven primarily by tropical Pacific forcing via the poleward propagation of boundary waves. The decadal-scale signal consists of a marked deepening of isopycnals (by ∼200 m between the early 1990s and the late 2000s), and reflects ENSO-forced wind-driven baroclinic changes over an extensive region of the eastern South Pacific.
 The baroclinic response of the Antarctic Circumpolar Current (ACC) to decadal-scale increases in the strength of the westerlies over the Southern Ocean [Thompson et al., 2011, and references therein] is an issue of significance, having the potential to influence the global ocean circulation via changes in stratification [Wolfe and Cessi, 2010], inter-ocean exchange [Rintoul et al., 2001] and meridional overturning [Toggweiler and Samuels, 1995].
 Hydrographic [Böning et al., 2008] and altimetric [Sokolov and Rintoul, 2009] studies suggest that Southern Ocean baroclinicity has remained essentially unperturbed by recent climatic changes in forcing. A possible explanation for this is put forward by Hallberg and Gnanadesikan  and Meredith and Hogg who propose, based on the analysis of eddy-permitting models and altimetric measurements, that the ACC may be approximately ‘eddy-saturated’ on interannual to interdecadal time scales. In this scenario, the tendency of Ekman flows to increase isopycnal slopes when wind forcing is intensified invigorates the eddy field through baroclinic instability, and the action of the energised eddies restores isopycnals to their original state on time scales as short as 2–3 years. Other authors, however, use idealised models to suggest that a key reason for the ACC's apparent insensitivity to climatic changes in forcing is that its baroclinic adjustment involves communication between the Southern Ocean and global ocean circulation over centennial time scales via boundary, equatorial Kelvin and Rossby waves [e.g.,Allison et al., 2011].
 In this article, we investigate the extent of baroclinic adjustment in Drake Passage on time scales of up to two decades. We find substantial adjustment of the upper kilometer of northern Drake Passage on interannual to interdecadal time scales, driven primarily by tropical Pacific forcing via the poleward propagation of boundary waves. The decadal-scale signal consists of a marked deepening of isopycnals (by ∼200 m between the early 1990s and the late 2000s), reflecting ENSO-forced wind-driven baroclinic changes over an extensive region of the eastern South Pacific.
 In situ hydrographic data from the WOCE SR1b repeat section crossing Drake Passage between ∼54°S, 58°W and ∼61°S, 54°W (Figure 1) are used to quantify pressure variability. This section has been occupied quasi-annually since 1993/4 [seeYelland, 2009, Table 1.1]. Argo measurements in the longitude range 50–70°W in the region to the north of the Subantarctic Front (SAF), an area found here to vary relatively coherently, supplement the time series during the post-2000 period. Data from the WOCE P06E repeat hydrographic transect, extending from the Chilean coast into the South Pacific along 32.5°S and occupied in May 1992, September 2003 and January 2010, are also analysed. All data are gridded to a 0.01 kg m−3 neutral density (γn) grid, and data shallower than 120 dbar excluded from the analysis to minimize variability associated with the mixed layer.
 The weekly, delayed-time, updated, merged, gridded sea level anomaly product of AVISO is used to quantify sea level anomaly (SLA). For consistency, only the period after the launch of the ERS-2 satellite (April 1995) is analyzed. High-frequency mesoscale variability is removed by smoothing the data using a 45-day moving average filter. Tide gauge data lying along the western South American coastline are employed, spanning a latitude band extending from the equator to 35°S (more southerly stations are omitted due to poor data coverage over our study period). Yearly means, tides, annual and semi-annual cycles are removed from the data, and the inverse barometer correction applied. Additionally, wind stress and sea level pressure (SLP) data from the European Centre for Medium-range Weather Forecasting ERA-Interim reanalysis and the National Centers for Environmental Prediction-National Center for Atmospheric Research (NCEP-NCAR) reanalysis are utilized, and the Bivariate ENSO time series (BEST) used to characterize ENSO.
 Mean pressure fields on isopycnal surfaces are defined as a function of dynamic height by fitting cubic splines to all available in situ data. Time series of the pressure anomaly are calculated by subtraction of these fields from observations, as shown in Figure 1b for the isopycnal range 27.1 ≤ γn ≤ 27.23 kg m−3 between the SAF (defined using the dynamic height criterion of Naveira Garabato et al. ) and 55°S (see Figures 1a and 1c), the region showing the strongest baroclinic variability [Close, 2011]. Substantial baroclinic changes (equivalent to ∼50–100 dbar) are evident on interannual time scales, accompanied by a decadal-scale deepening of isopycnals by ∼200 m over 1995–2010 that is most obvious after 2005. This tendency characterizes the upper kilometer of the water column (γn < 27.45 kg m−3) of northern Drake Passage, and is unchanged by the usage of either solely ship- or Argo-based data.
 In situ pressure anomaly and seasonally averaged SLA are strongly correlated in an area in northeastern Drake Passage spanning the location at which the in situ data are obtained (Figure 1a: 52–54°S, 56–62°W, ), suggesting that a downward (upward) displacement of isopycnals occurs in conjunction with a greater abundance of lighter (denser) density classes, yielding a net steric increase (decrease) in SLA. A proxy record for isopycnal pressure anomaly is hence defined using SLA data from this area. The agreement between the records suggests that, on seasonal and longer time scales, stratification changes in the hydrographic data are primarily indicative of baroclinic variability, rather than higher-frequency mesoscale activity.
 To investigate the origin of this variability, SLA in the proxy area is correlated with SLA elsewhere at varying lag (Figure 2). The signature of ENSO is apparent in the highly correlated area in the equatorial Pacific at lags of 24 and 16 months. For shorter lags, a band of positive correlation extends diagonally away from the western South American coastline. This distribution is qualitatively consistent with the westward propagation of baroclinic Rossby waves generated at the South Pacific's eastern edge by the passage of boundary waves in the manner described by Allison et al. , amongst others. The boundary wave propagation speed of O(0.1) m s−1 implied by the correlation analysis of Figure 2is used in conjunction with idealized solutions of the quasi-geostrophic potential vorticity conservation equation [Gill, 1982] and the Global Atlas of First-Baroclinic Rossby Radius of Deformation [Chelton et al., 1998] to predict the theoretical position of a westward-propagating first-mode baroclinic Rossby wave packet triggered by the passage of a boundary wave initiated at the time ofFigure 2a (see Figures 2b–2d). The approximate correspondence between the theoretical positions and the band of high correlation supports the notion that baroclinic variability in northern Drake Passage primarily reflects the passage of boundary waves generated by tropical Pacific forcing.
 The theoretical propagation speed of first-mode baroclinic Kelvin waves is of order 0.5–1 m s−1 [Enfield and Allen, 1980]. (The propagation speed of O(0.1) m s−1 implied by Figure 2is significantly smaller than the theoretical first-mode baroclinic Kelvin wave speed of 0.5–1 m s−1. This point was noted by Johnson , who suggested that the difference indicates higher-mode contributions to the boundary waves.) Such waves would travel the length of the western South American coastline in 1–2 months, a period comparable to the temporal resolution of the altimeter data. Tide gauge data, available hourly, are hence used to study propagation along the continental shelf.Figure 3b illustrates the key features of the entire record. Boundary wave phase speeds of up to ∼1–2 m s−1are implied, consistent with first-mode baroclinic Kelvin waves.
 The Hovmöller diagram in Figure 3bspans the strong ENSO event of 1997/8 (December–June), during which a largely continuous southward propagation of little-modified boundary wave signals emanating from the equator may be seen. In contrast, the more quiescent period that follows (after June 1998) is characterized by a series of positive SLA features that originate at ∼17°S (most evident in the June–August and October–December periods) and appear to be superimposed on a lower-frequency signal originating at the equator. SLA south of ∼17°S thus exhibits variability that may be decoupled from that further north (particularly during periods of weak ENSO activity), implying that forcing localised to this vicinity may modify boundary wave signals originating at the equator before they reach Drake Passage.
 The baroclinic anatomy of the boundary waves is evident in the P06E section data. Nodal structures, separating positive and negative pressure anomalies in the vertical, are observed in the 2003 and 2010 transects (Figure 4), consistent with a first-mode baroclinic Kelvin wave structure. As discussed bySuginohara  and Hughes and Meredith , boundary waves generated at an equatorial source resemble baroclinic Kelvin waves in low-latitude regions of strong stratification and evolve toward a barotropic continental wave structure as they propagate poleward into areas of weaker stratification. As the initially horizontal node associated with the first-mode baroclinic Kelvin wave mode tilts and transforms into the vertical node associated with the barotropic wave structure, pressure anomalies that are initially separated in the vertical become laterally separated. This suggests that the hydrographic observations at 32°S are consistent with those in Drake Passage: the positive pressure anomalies shown inFigure 4would be shifted upward and poleward (off-shelf) according to the preceding theory, leading to a deepening of isopycnals in northern Drake Passage, as observed inFigure 1b.
 The deepening of isopycnals by ∼200 m in 18 years in the upper kilometer of northern Drake Passage is striking, and contrasts with the invariant baroclinic structure on decadal time scales reported by preceding observational studies. The tendency appears to be associated with changes in atmospheric forcing near the western coast of South America around ∼17°S, where meridional wind stress is correlated with SLA in Drake Passage at 15 month lag (r= 0.67), consistent with cross-slope Ekman flow perturbing the upper pycnocline and triggering boundary waves there [Gill and Clarke, 1974]. Meridional wind stress has intensified between 1990 and 2010 (Figure 5), consistent with enhanced upwelling, an increase in near-surface density and thus the near-coast negative pressure anomaly evident aboveγn ∼ 26.7 kg m−3 in Figure 4. The decadal-scale increase in SLA shown inFigure 3a, apparent only south of ∼30°S, further supports this link.
 The evolution of the SLP field (not shown) demonstrates the intensification of meridional wind stress off the Peruvian coast to be linked to an expansion of the South Pacific High (SPH, a permanent anticyclone shown in Figure 5a), in the presence of little concurrent decadal-scale change in the atmospheric system to the east, which is locked to the orography of the Andean Altiplano [Vuille, 1999]. The expansion of the SPH intensifies the zonal SLP gradient at the boundary between the two atmospheric systems, leading to an enhancement of meridional wind stress at the interface.
 The maximum monthly SLP at the centre of the SPH between 1990 and 2010 exhibits a significant correlation with the BEST index (r = −0.54). Folland et al.  show that ENSO modulates the position of the South Pacific Convergence Zone, with northeastward (southwestward) migration, associated with a weaker (stronger) SPH [Trenberth and Shea, 1987], occurring in response to El Niño (La Niña). A robust negative trend, suggesting increased prevalence of La Niña, is evident in the post-1990 BEST index, consistent with a significant role of the ENSO-modulated interdecadal expansion of the SPH in driving the decadal change in the baroclinicity of northern Drake Passage. Further,Dewitte et al.  note that equatorial Kelvin waves triggered in response to El Niño Modoki have reduced amplitude and weaker impact on the Peru upwelling system relative to those triggered by canonical El Niño. The increased occurrence of El Niño Modoki in recent years [Lee and McPhaden, 2010] may thus enhance the oceanic impact of the expansion of the SPH, with intensified, ENSO-modulated wind forcing off the Peruvian coast increasingly overwhelming the weakening boundary signals of equatorial origin.
 The changes in baroclinicity implicate substantial variations in the baroclinic volume transport through Drake Passage. The transport relative to 3000 dbar implied by the variability observed in northern Drake Passage (with the density profile at the southern edge held constant as the mean of all SR1b section repeats) exhibits interannual variability characterised by a standard deviation of 5.4 Sv, a maximum anomaly of 12.6 Sv and a range of 20.1 Sv (the mean total baroclinic volume transport being 106.8 Sv. The equivalent baroclinic volume transport variability associated with the density profile at the southern edge of the passage with the northern profile held constant is substantially smaller, having a standard deviation of 1.7 Sv and a range of 6.6 Sv, concomitant with a relatively slight deepening tendency of 0.32 dbar / yr over the top 1 km over the water column during the study period. Our approximation of an unchanging southern density profile is thus reasonable). The interdecadal increase associated with the deepening of isopycnals in the same period is 1.5 Sv. The baroclinic transport is significantly correlated with the BEST index (r= 0.64) at a lag of 25 months, with transport increasing (decreasing) following El Niño (La Niña) events, suggesting that tropical Pacific forcing plays an important role in regulating inter-ocean exchange through Drake Passage on interannual and longer time scales. As suggested byAllison et al. , the baroclinic structure of the ACC is thus controlled significantly by processes external to the region. These drivers may plausibly confound or mask the ACC's response to decadal-scale variations in the strength of the overlying westerlies.
 The decadal-scale deepening of isopycnals detected in northern Drake Passage reflects baroclinic adjustment over a larger region of the eastern South Pacific. The altimetric data suggest the propagation of the SLA signal into the interior: this is apparent not only as a transient signal (Figure 2), but also as a longer-term expression in the linear trend of SLA over 1995–2010 (Figure 5c). The consequences of decadal changes in tropical Pacific forcing for the South Pacific circulation thus appear to be far-reaching.
 This study would not have been possible without the effort of many people at sea and ashore to collect, calibrate, and process the Drake Passage section measurements. The U.K. Natural Environment Research Council (NERC) funded 17 of the section occupations used in this work through Research Grant GR3/11654 and several core-strategic research programmes at the National Oceanography Centre, Southampton (NOCS) and the British Antarctic Survey. It supported SEC through a PhD studentship based at NOCS, and ACNG through a NERC Advanced Research Fellowship (NE/C517633/1). We thank Elaine McDonagh, Brian King, Peggy Courtois, Eleanor Frajka-Williams and Mike Meredith for helpful discussions.
 The Editor thanks the two anonymous reviewers for their assistance in evaluating this paper.