1. Introduction
Western boundary currents (WBC) of the oceans’ subtropical gyres are the main oceanic conduit by which excess heat gained near the equator is redistributed poleward (Hogg and Johns 1995; Johns et al. 2011) and subsequently transferred upward to the atmosphere (Ganachaud and Wunsch 2003; Josey et al. 2013; Trenberth and Caron 2001). Syntheses of global observations reveal that WBC regions are experiencing the highest rate of surface warming (Wu et al. 2012; Rouault et al. 2009) and upper-ocean heat gain (Rhein et al. 2013, their Fig. 3.2), possibly as a result of a poleward shift and intensification of these currents (Yang et al. 2016). However, WBCs are swift, narrow, and highly variable (Hogg and Johns 1995), making their changes difficult to observe and predict.
The classic solution to vorticity dynamics gives a prediction for the strength of a WBC: Its transport must balance the interior Sverdrup transport, that is, the zonally integrated meridional transport driven by the wind stress curl across the basin’s interior (Munk 1950; Salmon 1998). However, observational evidence for this balance, which has been tested only at a single latitude within most subtropical gyres, is mixed. Of note is the study of DiNezio et al. (2009), who showed that half of the interannual variability of the Gulf Stream flowing through the Florida Strait near 27°N can be explained, with moderate statistical confidence, by lagged-adjusted wind stress curl anomalies over the basin’s interior near the same latitude. More typically, the influence of bottom topography, the geometry of ocean basins, overturning circulations, nonlinearities, uncertainties related to data sampling, and the time scales of oceanic adjustment all contribute to obscure a simple “Sverdrupian” picture (Wunsch and Roemmich 1985; Schmitz et al. 1992; Hautala et al. 1994; Wunsch 2011; Gray and Riser 2014; Thomas et al. 2014; Roemmich et al. 2016).
The transport of the Agulhas Current (AC), the WBC of the subtropical gyre of the south Indian Ocean, has been observed over 9 months near 32°S (Bryden et al. 2005) and over 34 months near 34°S (Beal et al. 2015). In the mean, it is substantially larger than the zonally integrated Sverdrup transport calculated from climatological winds at these corresponding latitudes (Cásal et al. 2009). This is principally because the AC carries southward a component of the Indian Ocean overturning circulation as well as a transport corresponding to the global thermohaline circulation that enters the Indian Ocean via the Indonesian Seas (Ganachaud and Wunsch 2000; Bryden et al. 2005). However, the increase in mean transport of the AC from 32° to 34°S is consistent with the latitudinal increase of the Sverdrup transport, an observation in support of a time-mean Sverdrupian balance (Cásal et al. 2009; Beal et al. 2015). In terms of multidecadal trends, ocean and climate models and reanalyses point to an intensification of the AC transport in response to an intensification of the westerlies and an increase in wind stress curl over the Indian Ocean (Rouault et al. 2009; Wu et al. 2012; Yang et al. 2016), a result also consistent with Sverdrup dynamics. Yet, in a recent analysis using an observation-based proxy, Beal and Elipot (2016) find that the AC has not strengthened since the 1990s, despite the wind trends. At interannual time scales, model studies have found no clear relationship between AC transport and Sverdrup transport over the interior of the Indian Ocean (Fetter et al. 2007; Biastoch et al. 2009b). This is suggested to be the result of nonlinear dynamics and of shielding from westward-propagating planetary waves by topography.
The goal of this study is to investigate the interannual-to-decadal variability of the AC and the role of atmospheric forcing on this variability. We use the recently derived, 24-yr-long, observation-based proxy of AC transport (Beal and Elipot 2016), together with atmospheric reanalyses, to quantify its linear sensitivity to Sverdrup dynamics and more significantly to more complex patterns of mean sea level pressure and wind stress. We then use the derived interannual sensitivity to conjecture the response of AC transport to long-term atmospheric trends.
2. Data
a. Oceanic data
The oceanic data and methods used to derive the proxy time series of AC transport are described in detail in Beal and Elipot (2016). Briefly, the Agulhas Current Time series (ACT) velocity mooring array was deployed between April 2010 and February 2013 off the east coast of South Africa near 34°S, and under the path of the suite of satellite altimeters, TOPEX/Poseidon and Jason-1 and Jason-2. These coincident data were used to build a series of regression models between in situ transport at each mooring and sea surface slope, thanks to the equivalent barotropic nature of the AC at the location of the array (Elipot and Beal 2015). Two volume transport time series were derived from these regression models, representing a streamwise jet transport Tjet and a geographically fixed boundary layer transport Tbox (Beal et al. 2015). The Tjet was defined as the southwestward transport integrated out to the first maximum in transport per unit distance beyond the half-width of the boundary layer. The Tbox was defined, more straightforwardly, as the net transport integrated out to the mean width of the boundary layer. Boundary layer width was calculated as the mean distance from the coast out to the zero isotach.
Here, Tjet and Tbox are updated with 14 additional months of altimeter data and now span 26 September 1992 to 20 April 2016, with a near-regular 9.92-day interval from the repeat orbital time of the satellites (Figs. 1a,b). The 12 missing estimates out of 870 time steps are filled by linear interpolation. The computer code to obtain the time series is freely available (https://github.com/selipot/actproxy), as are the resulting time series through the ACT website (http://act.rsmas.miami.edu).
Southward (a) Tbox and (b) Tjet transport time series. The thin gray lines are the time series at near 10-day intervals and the thick black lines are the 18-month low-pass-filtered time series. (c) Multitaper spectral estimates for Tbox and Tjet using seven Slepian tapers. The 95% confidence interval is drawn in the lower left.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
While along-track altimeter data from AVISO were used to derive the AC transport proxies above, we use mapped absolute dynamic topography (ADT) from AVISO to investigate the large-scale circulation. Specifically, we use the two-satellite, reference product at ¼-degree horizontal resolution from 1 January 1993 up to April 2016.
To delineate the time-mean boundary of the wind-driven subtropical gyre of the south Indian Ocean we use the CSIRO Atlas of Regional Seas (CARS) hydrographic dataset (Ridgway et al. 2002) (http://www.cmar.csiro.au/cars). We choose the 1650-m2 contour of the vertically integrated dynamic height, 2000 dbar to surface, relative to 2000 dbar. This contour forms an approximately closed loop, passing through the ACT array, turning from southwestward to eastward within the Agulhas retroflection, and passing south of Australia to Tasmania before looping back toward the source region of the Agulhas. We consider this boundary to be representative of a region influenced by wind stress curl over the south Indian basin, and it therefore neglects the Indonesian Throughflow (ITF).
b. Atmospheric data and climate indices
The ECMWF interim reanalysis (ERA-Interim, hereinafter ERA-I; Dee et al. 2011) and NCEP–DOE Atmospheric Model Intercomparison Project reanalysis (NCEP2; Kanamitsu et al. 2002) provide the atmospheric data for our study. NCEP2 is a follow-up to the NCEP–NCAR reanalysis project (Kalnay et al. 1996). From ERA-I we use monthly means of mean sea level pressure (MSLP) and instantaneous turbulent surface stress (i.e., wind stress) derived from daily means at 0.75° by 0.75° spatial resolution. From NCEP2 we use monthly means of MSLP and momentum flux at the surface (i.e., wind stress) at 1.875° by 1.9° spatial resolution. Both reanalysis products span the era of satellite coverage, from January 1979 to present day.
We consider MSLP to diagnose and characterize large-scale climate modes of variability (Allan and Ansell 2006). We consider wind stress (τ) and wind stress curl (∇×τ) as direct mechanical forcings of the ocean. Wind stress curl is calculated from the zonal and meridional components of wind stress using first central differences on a sphere.
To focus on interannual variability, we remove least squares estimates of the linear trends and of the monthly averages representing seasonal cycles, at every grid point for the time period January 1979 to December 2016. This results in detrended and deseasoned time series.
To relate our results to climate phenomena, we consider the 26 climate indices listed by de Viron et al. (2013). These indices are obtained from the Climate Prediction Center of the NOAA/National Weather Service via the Earth System Research Laboratory (ESRL) (http://www.esrl.noaa.gov/psd/data/correlation). Of these, the Antarctic Oscillation (AAO), also known as the southern annular mode (SAM, used hereinafter), is defined as the leading eigenmode of monthly mean 700-hPa geopotential height poleward of 20°S from the NCEP–NCAR reanalysis dataset, covering years 1979–2000. To be consistent with the reanalysis dataset used here, we calculate SAM indices using this same methodology, but taking MSLP from ERA-I and NCEP2 for the time period 1979–2016. Because indices based on reanalyses may exhibit spurious or exaggerated trends due to sparse data at high latitudes, we also consider the observation-based SAM (Marshall 2003) from 1957 to the present, obtained from the British Antarctic Survey (http://www.nerc-bas.ac.uk/icd/gjma/sam.html) via the KNMI Climate Explorer (http://climexp.knmi.nl). Over their respective overlapping periods and at monthly intervals, the SAM indices derived from ERA-I and NCEP2 are both highly correlated with the SAM index from ESRL (>0.97), and with the observation-based SAM (>0.83), after linear detrending. The east central tropical Pacific SST index, Niño-3.4, one of the indices symptomatic of El Niño–Southern Oscillation (ENSO), is also among those considered and obtained from ESRL. We consider, as well, the dipole mode index (DMI; Saji et al. 1999), or Indian Ocean dipole (IOD), representing interannual climate variability over the tropical Indian Ocean region, and the Madden–Julian oscillation (MJO), characterized by an extended empirical orthogonal function (EEOF) analysis applied to pentad velocity potential at 200 hPa. The first EEOF is composed of 10 time-lagged patterns, of which we consider the temporal index associated with the first, centered at 80°E (http://www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_mjo_index/mjo_index.html).
3. Methods
To focus on interannual variability, we filter out periods shorter than 18 months by applying a sixth-order Butterworth filter to all time series, run forward and backward in time to avoid phase distortion. We choose 18 months as a time scale that clearly corresponds to spectral gaps in the power spectra of the AC transport proxies Tjet and Tbox, separating interannual variance from variance at annual and shorter time scales (Fig. 1c). Low-pass filtering artificially increases correlation coefficient estimates by making time series more serially auto-correlated. Thus, unless noted otherwise, p values of correlation coefficient estimates are calculated using an effective number of degrees of freedom corresponding to the length of the time series divided by the estimated decorrelation time scale of the interannual Tjet time series, 5.8 months (Emery and Thomson 2001). Statistics are deemed significant when significance is found at the 95% level.
We apply the method of DelSole and Yang (2011) to estimate Southern Hemisphere regression patterns of atmospheric field variables with the AC jet transport. This method consists of regressing a predictand (herein transport) against a subset of principal component time series (PCs) from an eigenmode analysis of the predictors (herein atmospheric variables), rather than against predictors directly, as with a point-by-point regression. From this regression exercise, we obtain a predicted time series for the predictand, called the canonical time series, in addition to spatial regression patterns. The squared correlation coefficient between the original predictand and the canonical time series is called the squared canonical correlation ρ2 and quantifies the amount of variance of the predictand that can be explained through the regression model. Since, by construction, the PCs are uncorrelated, ρ2 is merely the sum of the squared correlation coefficients between the predictands and the PCs. The method of DelSole and Yang (2011) has a number of advantages compared to point-by-point regression methods: (i) it provides a way to formally test for the significance of the resulting regression pattern as a whole, (ii) it avoids the pitfalls of multiple hypothesis testing (Shaffer 1995), (iii) it circumvents the issue of spatial correlation between predictors, which is taken into account by the preliminary eigenmode decomposition, and (iv) it quantifies how much of the variance of the predictand can be explained, avoiding the issue of implicitly and unrealistically explaining 100% of the variance when the field variables have more spatial grid points than time steps. Finally, an added and useful aspect of the method is that it allows us to directly relate the resulting atmospheric patterns to climate mode variability by correlating the selected PCs to climate indices.
Once the number of PCs is chosen, the resulting significant regression pattern is then obtained by multiplying the regression coefficients for the selected PCs by their corresponding eigenvectors (the EOF patterns). The total significant regression pattern can be decomposed into a number of distinct contributions from the individual EOF patterns or any combinations of these. In this way, one can argue that a single term of this sum can be associated with a known climate phenomena if the corresponding PC time series is strongly correlated with a known climate index.
Since wind stress and MSLP are covarying, we regress the Agulhas transport proxies against the PCs arising from a combined EOF analysis of MSLP and wind stress, rather than against the PCs of wind stress and the PCs of MSLP separately. Combined EOF analysis provides a way to capture the covariance between variables of different physical nature by decomposing their correlation matrix into eigenvectors (Kutzbach 1967); thus all atmospheric time series are normalized before computing the PCs. The individual components of the eigenvectors are then rescaled by their corresponding standard deviations in order to obtain the regression patterns.
We conduct the above analyses using both ERA-I and NCEP2 datasets for robustness (Fig. 2). In both cases, we use six EOFs to build the total regression patterns, because the CVSS statistics [Eq. (1)] for both products increase monotonically up to local maxima at 6 PCs (Fig. 2c). Considering the error bars for CVSS, one could argue that only five PCs should be chosen for ERA-I and only two for NCEP2. However, a choice of fewer than six PCs for ERA-I drops the squared canonical correlation ρ2 below significance (Fig. 2d). For NCEP2 using two PCs, ρ2 = 0.12 and is significant. However, ρ2 also corresponds to the fraction of variance of Tjet that can be linearly explained by the regression patterns, and is lower than what can be obtained with either ERA-I or NCEP2 with six PCs (0.28 and 0.29, respectively). In addition, the fraction of covariance of the atmospheric field variables explained by only two EOFs for NCEP2 is 30% (34% for ERA-I), compared to 56% (60% for ERA-I) for six PCs (Fig. 2b). Finally, we also find that the combined EOF analyses for ERA-I and NCEP2 are consistent up to the sixth EOF, as their respective PCs correlate pairwise strongly (correlation >0.74; not shown). Beyond the sixth EOF, we find that their eigenmodes are different and explain less than 5% of the fraction of covariance, so their influence on the large-scale atmospheric circulation patterns should be small. In conclusion, considering the overall consistency of the statistics between the two reanalyses products up to, and including, the 6th PC and corresponding EOF, we choose their first six eigenmodes to ultimately derive regression patterns.
Statistics used for field significance of regression patterns for ERA-I (blue) and NCEP2 (red): (a) Correlation coefficient ρ between Tjet and individual PC time series from the Combined EOF analysis of MSLP and wind stress. Filled symbols indicate correlation coefficients that are deemed significant. (b) Amount of variance explained by the cumulative sum of combined EOFs. (c) CVSS as a function of the number of PC time series included to calculate regression patterns. Error bars correspond to one standard deviation. (d) Cumulative squared canonical correlation coefficients with Tjet. The dashed curve corresponds to the 95% confidence level.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
4. Relevance of Agulhas jet transport to the south Indian gyre circulation
Beal et al. (2015) defined two measures of AC transport: a streamwise jet transport Tjet and a net boundary layer transport Tbox. The Tjet follows the current’s southwestward flow as it meanders over time, while Tbox is fixed geographically. These measures and their resultant proxies are highly correlated but exhibit distinct probability density functions, as well as distinct power spectral density functions (Beal et al. 2015; Beal and Elipot 2016) (Fig. 1). For instance, Tbox is biased positive because it includes northeastward flow and misses offshore southwestward flow during meander events, while Tjet may be biased negative because it disregards local recirculations. In addition, Tjet has less power at the annual and lower frequencies compared to Tbox (Fig. 1c).
Considering their differences, we ask which of the two transports is the most representative of the larger-scale dynamics throughout the Agulhas system and the south Indian Ocean gyre circulation as a whole. It is this quantity we would like to investigate in terms of its response to atmospheric forcing. As an attempt to answer this question, we relate the two transports to the variability of the surface circulation, as inferred from ADT in the vicinity of the Agulhas system. We compute point-by-point regression maps between interannual transports and ADT (Fig. 3). We do not apply here the regression method using eigenmodes (DelSole and Yang 2011), as we do later for atmospheric variables, because we find no improvement in the results.
Point-by-point regression maps at interannual time scales for the period January 1993–May 2016 between absolute dynamic topography and (a) southwest Agulhas boundary transport (SW Tbox) and (b) southwest Agulhas jet transport (SW Tjet). Regression coefficients are shown for one standard deviation of the respective transports, 6.8 Sv (1 Sv ≡ 106 m3 s−1) for Tbox and 5.5 Sv for Tjet. Gray lines are contours of mean ADT for the same period at 10-cm interval. Selected contours are labeled. The ACT array is drawn with a green line.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
At the ACT array, an increase of one standard deviation in southward AC transport, both for Tbox and Tjet, corresponds to a SLA pattern that transitions from negative to positive offshore, leading to an increase in sea surface slope, and this is consistent with an expected acceleration of the local surface flow to the southwest (Fig. 3). More broadly, across the region, the SLA related to Tjet shows that the same negative–positive pattern persists throughout the Agulhas system, from the source region at the southern tip of Madagascar to the second meander of the Agulhas Return Current (36°E, 40°S). This pattern shows that an intensification of Tjet is representative of stronger flows up and downstream of the ACT array, throughout the southwest region of the gyre. In contrast, the SLA field related to Tbox points to a more localized response of the circulation. The negative–positive pattern suggests that variance in Tbox is related to a tight recirculation between the Agulhas and its Return Current, centered upstream of the retroflection at the offshore end of the ACT line. Notably, variance in both transports is related to perturbations in the meandering Agulhas Return Current, which creates strong SLA signals owing to the sharpness of its front.
Looking more closely at the detailed structure of regressed SLA across the AC along the ACT line (not shown), we find for Tbox that there is a weak negative slope from the coast out to a minimum at 135 km offshore, which acts to slow the inshore core of the current, even while the remainder of the current intensifies and net transport increases. This SLA pattern is consistent with a broadening of the boundary layer, as shown by Beal and Elipot (2016). For Tjet the SLA has a second minimum at the offshore end of the ACT line and this pattern corresponds to a narrowing of the jet as it intensifies.
We conclude that Tjet better represents interannual variability of the large-scale circulation of the region, with patterns of ADT consistent with not only a local measure of the strength of the Agulhas jet at the ACT array but also with the strength of the circulation far up- and downstream. Therefore, for the remainder of this paper, we focus on Tjet for our analyses of the sensitivity of the current to atmospheric forcing.
5. Sensitivity of the Agulhas jet to atmospheric interannual forcing
a. Response to interior Sverdrup transport near 34°S
We first test whether simple Sverdrup dynamics can explain interannual variability of AC transport. We calculate interior Sverdrup transports from both ERA-I and NCEP2 by zonally integrating the wind stress curl over the entire Indian Ocean sector, averaged between 33° and 36°S, the range of latitudes spanned by the ACT array. We find null correlations between these Sverdrup transport time series and the AC jet transport Tjet. The lack of correlation could be due to the oceanic adjustment time scale, expected to be as long as 4 to 5 years, the time required for Rossby waves to traverse the entire Indian Ocean at the latitude of ACT at speeds between 4 and 5 km day−1 (Chelton and Schlax 1996). Yet, a cross-spectral analysis typically able to account for lags also fails to reveal any significant coherence between time-varying jet and Sverdrup transports. The same analysis conducted with boundary layer transport Tbox also returns null results. These null results from in situ observations corroborate conclusions from numerical experiments: the Agulhas Current system appears to be shielded from wind-forced oceanic variability to the east of 45°E on seasonal and longer time scales, perhaps by the Mozambique and Madagascar Ridges (Matano et al. 1999; Biastoch et al. 2009b; Fetter et al. 2007). At the same time, Rouault et al. (2009) report, using a model forced by realistic winds, that the AC transport exhibits a linear trend since the 1980s consistent with a linear trend of the Sverdrup transport calculated from Indian Ocean winds between 21° and 33°S. We will investigate linear trends further in section 7.
b. Response to large-scale atmospheric patterns
Beyond the complex geometry of the Indian Ocean and the nonlinearities of the system, another explanation for the lack of linear relationship between Tjet and the Sverdrup transport near 34°S could be that Tjet is influenced by winds at other latitudes, for instance through coastally trapped wave processes carrying signals up- and downstream. This idea is supported by the fact that our transport proxy is representative of the variability of the entire WBC, and not only representative of the variability of the current near a single latitude, as we demonstrated previously (section 4 and Fig. 3). As such, one might reasonably hypothesize that Tjet is driven by atmospheric forcing acting over the entire subtropical latitudinal band and beyond. A drawback of this hypothesis—if we can verify it—is that the mechanisms by which basinwide atmospheric anomalies drive the AC will be more difficult to identify than if a direct response to Sverdrup transport is demonstrated.
To test this hypothesis, we extend our investigation of atmospheric forcing of the AC by calculating regression patterns of Southern Hemisphere MSLP and τ against southward Tjet, following the eigenmode method of DelSole and Yang (2011). We also consider wind stress curl derived from the τ regression patterns. We show and discuss the results for ERA-I only (Fig. 4), because the two products produce very similar results and ERA-I has a higher spatial resolution and leads to slightly better statistics for our study. Here, we emphasize that the total regression patterns obtained by the method of DelSole and Yang (2011) are significant as a whole, thereby avoiding the statistical pitfalls of point-by-point significance testing (see section 3). The resulting atmospheric patterns represent natural modes of variance that can have remote centers of action, in the same way that ENSO has related wind anomalies over the Southern Ocean (Mo and Paegle 2001), representing teleconnections that need not directly impact the AC.
Results of regression analysis with Tjet. Pattern for (a) MSLP, (b) zonal wind stress τx, and (c) resulting wind stress curl (∇×τ). The pattern for meridional wind stress is not shown. In (a), the time-mean MSLP for 1993–2016 is contoured from 980 to 1025 mbar at 5-mbar intervals. In (b) and (c), zero contours of mean wind stress curl are drawn in gray, and zero contours of mean zonal wind stress are drawn in black. In (a), (b), and (c) the boundary of the barotropic south Indian subtropical gyre referenced to 2000 dbar is drawn in dark green. (d) Interannual Tjet time series and corresponding canonical time series.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
The regression pattern for MSLP (Fig. 4a) shows that one standard deviation of southward increase of Tjet corresponds to a deepening of the polar trough around Antarctica, to a strengthening of the anticyclonic high of the south Indian Ocean, and to a weakening of the anticyclonic highs of the South Atlantic and Pacific Oceans. At the same time, the trade winds are weakened south of the equator over the west Pacific Ocean and Atlantic Ocean but strengthened over the Indian Ocean (Fig. 4b). Over the Southern Ocean, the westerlies are strengthened for an increased Agulhas, most prominently in the Indian Ocean sector and between Antarctica and the southern boundary of the time-mean barotropic subtropical gyre. East of the date line, the westerlies are weakened between 40° and 60°S but strengthened poleward of 60°S.
The ∇×τ anomalies (Fig. 4c) induced by the total regression pattern of τ are mostly due to zonal wind stress anomalies away from the coasts and are composed of loosely organized zonal bands within each oceanic basins. Positive ∇×τ anomalies are found in the Indian Ocean between approximately 20°S and the equator—the region of the equatorial gyre and South Equatorial Current. Over the subtropical gyre, positive anomalies are found in the southwest, and negative anomalies in the north, off Madagascar. Strikingly, very strong ∇×τ anomalies are found within the subpolar frontal zone of the south Indian Ocean (Kazmin 2017), between the zero ∇×τ contour to the south and the southern boundary of the Indian Ocean subtropical gyre to the north. Here, the southern boundary of the gyre is defined by a contour of vertically integrated dynamic height relative to 2000 dbar (Ridgway and Dunn 2007) and does not coincide at all longitudes with the zero ∇×τ contour because bottom torque forcing of the Antarctic Circumpolar Current (ACC) displaces the subtropical front northward (De Boer et al. 2013). Fetter et al. (2007) also found wind stress curl anomalies, centered slightly to the southeast of this region, that are linearly associated with interannual variability of the AC.
Next we examine zonally averaged regression patterns over the Indian Ocean sector. Although we have seen that the forcing patterns of the AC are not purely zonal, particularly in the western subtropics (Fig. 4), we nevertheless might expect WBC systems to be influenced by an integral of forcings applied to the east of their locations (Anderson and Gill 1975; Anderson and Killworth 1977). Also, numerical studies that have modeled the response of the Agulhas Current system to changing winds have principally imposed zonally symmetric perturbations over the Indian Ocean or the Southern Ocean (e.g., Durgadoo et al. 2013; Loveday et al. 2014), and so our zonal analysis provides a useful comparison.
We note that the Agulhas jet transport is sensitive to MSLP and τ forcings of relatively weak magnitude compared to the full extent of the observed variability of these fields (Figs. 5d–f). Hence, to obtain a clear impression, we consider the forcing magnitude required to increase Tjet southward from minus one standard deviation below its interannual time mean to one standard deviation above, that is, the zonal averages of twice the atmospheric anomalies from the regression (Fig. 5).
Twice the regression patterns zonally averaged for the Indian Ocean sector (28° to 120°E) for (a) MSLP, (b) τx, and (c) induced ∇×τ. Each panel shows the regressions from the total pattern, and from the separate contributions of EOF1, EOF2, and EOFs 3 to 6. (d)–(f) The results of adding the total zonal anomalies shown in (a), (b), and (c), respectively, to the mean zonal values for 1993–2016. Red shadings indicate positive changes and blue shadings negative changes. Gray shadings show plus and minus one standard deviation of the zonal averages around their mean values. In (e), the black solid lines indicate the latitude of the time-mean maximum τx while the red dashed line indicates the latitude of the maximum resulting from the added anomalies. In (f) similar lines depict the latitudes of the zero wind stress curl.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
For increased Agulhas transport the atmospheric subtropical high is increased and the polar trough is deepened (Figs. 5a,d), implying a strengthening of both the subtropical and subpolar circulations. The meridional pattern of wind stress anomalies reveals more complexity (Figs. 5b,e), with the largest signals corresponding to a narrowing and intensification of the westerlies over the Indian Ocean sector, shaped by a decrease in τx around 38°S and a large increase with a maximum at 52°S. The westerlies jet is primarily strengthened rather than shifted meridionally, since the anomaly peaks at the latitude of the time-mean maximum of the westerlies (Fig. 5e). This corroborates findings from a numerical wind perturbation experiment, although focused on Agulhas leakage response (Durgadoo et al. 2013), but is different from late twentieth-century trend patterns, which suggest a poleward shift in the westerlies (Biastoch et al. 2009a). The increase in the westerlies leads to a positive wind stress curl anomaly, and hence an implied northward Sverdrup transport, between the latitude of ACT and the maximum westerlies at 52°S, and a negative anomaly, implying southward Sverdrup transport, south of here (Figs. 5c,f). Note that the ACT array is located very close to the climatological mean maximum of anticyclonic wind curl. This pattern leads, overall, to a slight expansion southward of the subtropical gyre, as delineated by displacement of the zero ∇×τ line by 0.5° within the subtropical region (Fig. 5f).
The anomalies in the trades are smaller and point to a strengthening and equatorward shift of the tropical winds (Figs. 5e,f). This pattern is shaped by an increase in the easterlies over the equatorial gyre, from the equator to the South Equatorial Current at 15°S, and a slight weakening farther south (Fig. 5b). Overall, perturbations in the trades related to an increase in AC transport lead to a northward expansion of the subtropical gyre, as delineated by displacement of the zero ∇×τ line by 0.75° within the tropical region.
In summary, interannual variability in AC jet transport is associated with patterns of MSLP and wind stress consistent with spinup and -down of the subtropical and subpolar circulations and expansion and contraction of the subtropical gyre. Changes in the winds are most pronounced poleward of 34°S and are associated predominantly with the strength, rather than the position, of the westerlies jet over the Indian Ocean sector, such that the transport of the Agulhas jet is increased with a strengthening of the westerlies jet at its core.
From the regression analysis, the squared canonical correlation is 0.29, which is the square of the correlation coefficient (0.54) between Tjet and the canonical time series 
6. Climate mechanisms
The regression patterns we have found that relate atmospheric anomalies to AC strength are spatially complex and are likely to be the result of several mechanisms of both local and remote origins. To try to tease apart these mechanisms, we examine individually each of the eigenmodes that contribute to the total patterns shown in Fig. 4. We find that the first two PCs exhibit significant and strong (>0.8) correlations with ENSO and the SAM, respectively, while the remaining four do not correlate with common climate indices. Thus, we examine the first two modes independently and then consider the contribution of the sum of the remaining four modes.
a. Indo-Pacific influence
Climate variability of the tropical regions of the Pacific and Indian basins drives the dominant atmospheric variance associated with changes in AC transport (Fig. 6). The first PC is significantly and strongly correlated with both the low-passed Madden–Julian oscillation index at 80°E (MJO) and the Niño-3.4 index at the same level, 0.85. The MJO is a coupled tropical disturbance that propagates eastward across the Indo-Pacific on intraseasonal time scales. The MJO is strongly modified by ENSO at interannual time scales (Zhang 2005), as confirmed here by the low-passed time series of the Niño-3.4 and MJO indices (Fig. 6d). Both phenomena have extratropical influences (e.g., Karoly 1989; Matthews and Meredith 2004; Pohl et al. 2010; Cai et al. 2011) and thus contribute to the overall climate variance of the Southern Hemisphere atmosphere (L’Heureux and Thompson 2006). The tropical Indian Ocean DMI also imprints on PC1, with a correlation of 0.48.
Contribution to the total regression pattern of the first EOF pattern for (a) MSLP and (b) τx. (c) The resulting wind stress curl. In (a), (b), and (c) the boundary of the barotropic south Indian subtropical gyre referenced to 2000 dbar is drawn in dark green. Color scales saturate at plus and minus the 99th percentile of absolute values for each panel. Landmasses are shaded dark gray. In (a) the time-mean MSLP for 1993–2016 is contoured from 980 to 1025 mbar at 5-mbar intervals. In (b) and (c), zero contours of mean wind stress curl are drawn in gray, and zero contours of mean zonal wind stress are drawn in black. In (a), (b), and (c) the boundary of the barotropic south Indian subtropical gyre referenced to 2000 dbar is drawn in dark green. (d) The PC time series (PC1) and the low-passed MJO and Niño-3.4 time series scaled to transport units based on their regressions with Tjet.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
The contribution of EOF1 (Fig. 6) to the total regression patterns (Fig. 4) is evident within the trade winds band in all three oceanic sectors. As an example, in its positive phase, EOF1 is characterized by reduced trade winds in the western tropical Pacific (Fig. 6b), symptomatic of a positive phase of ENSO but strengthened trade winds over the equatorial Indian Ocean, consistent with expectations for a positive phase of the DMI (Gadgil et al. 2004). At higher latitudes, EOF1 captures the extratropical influence of ENSO within the central and eastern Pacific, known as the Pacific–South American mode (e.g., Mo and Paegle 2001). For the positive phase of ENSO, this mode manifests as a strong positive MSLP anomaly upstream of Drake Passage associated with a slowdown there of the westerlies jet and leads to strong positive ∇×τ anomalies to the south of the zero ∇×τ contour and negative anomalies to the north. In contrast, the Indian and Atlantic sectors exhibit an intensification of the northern flank of the westerlies (Figs. 5b and 6b), leading to positive ∇×τ anomalies to the north and negative ∇×τ anomalies to the south of the zero ∇×τ contour (Figs. 5c and 6c).
We can further investigate how these tropical modes, captured in EOF1, directly influence oceanic circulation and AC transport by calculating point-by-point regressions between detrended, low-pass-filtered ADT and the partial canonical time series obtained from PC1, scaled to transport units (Fig. 6d). These regressions are used to illustrate changes in ADT gradient, and hence surface geostrophic velocity, within the AC, as well as changes in gyre strength and position, as depicted by ADT contours, that are associated with an increase southward of Tjet from minus one standard deviation to plus one standard deviation (Fig. 7a). We find a spatially coherent strengthening of the AC jet, from its origin near 28°S all the way downstream to the Agulhas retroflection region. Outside the AC system, there is a strong impact on the equatorial and tropical regions of the Indian Ocean gyre, such that a strengthening of Tjet is associated with a westward contraction of the equatorial gyre or Seychelles Dome (Yokoi et al. 2008), an eastward extension of the subtropical gyre east of Madagascar, and a northward displacement of the Leeuwin Current system. There also appears to be a reduced ITF, which agrees with observations of a weaker, shallower ITF during El Niño (e.g., Schott et al. 2009; Sprintall et al. 2014).
Absolute dynamic topography contours obtained by subtracting (light lines) or adding (heavy lines) to the time mean the results of the regression of ADT against one standard deviation of the partial canonical time series from (a) PC1 and (b) PC2. Contours are drawn from −1.489 to 1.235 m at 0.075-m intervals north of 40°S and at 0.15-m intervals south of 40°S. Color shading in the vicinity of the Agulhas Current show changes of ADT gradient (Δ|∇ADT|) along the ACT array in the offshore direction (black line) related to subtracting then adding the regression results to the time mean ADT. In each panel the boundary of the barotropic south Indian subtropical gyre referenced to 2000 dbar is drawn in black.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
Despite these notable changes in surface circulation, EOF1 can explain only a modest 5% of the interannual variability of Tjet (correlation 0.22). The standard deviation of the canonical time series for this mode is 1.2 Sv, and its maximum peak to peak amplitude is 6 Sv (Fig. 6d). Interestingly, we find a stronger linear link between Tjet and ENSO, with a correlation of 0.34 between Tjet and the Niño-3.4 index. Hence, ENSO alone can explain 11.5% of the variance of the interannual variability of the AC, accounting for a peak-to-peak transport variance of 9.6 Sv (Fig. 5d), with El Niño leading to a strengthening of the jet. Recently, using a coupled climate model with eddy-resolving ocean resolution, ENSO was linked to a modulation of Agulhas leakage at a two-year lag, with the AC acting as a conduit for this modulation (Putrasahan et al. 2016). This link was also found in ADT observations, but was not reproduced in an equivalent coupled model with low ocean resolution. Putrasahan et al. (2016) concluded that anomalously weak trade winds over the Indian Ocean related to El Niño lead to an oceanic teleconnection mediated by Rossby waves and the resolved meridional boundary flow of the AC itself. Our zero-lag regression analysis with ADT cannot evidence propagating signals, but we find that the autocorrelation of the low pass filtered Niño-3.4 index time series, which has increased spectral power between 2 and 7 years (e.g., Rasmusson et al. 1981), is positive near the 2-yr lag. This means that if the AC exhibits a linear link with ENSO with a 2-yr lag, then our analysis, which low-pass filters and detrends the data to focus on interannual time scales, would reveal a positive correlation at zero lag, as we find. Thus, it is possible that our observed interannual correlations between Tjet and Niño-3.4 and PC1 are representative of changes in the AC driven by oceanic teleconnections with tropical Indo-Pacific climate variability.
The IOD mode also contributes to the EOF1 pattern (correlation 0.48), but the AC transport and the DMI are not directly correlated. The IOD has been argued to influence interannual variability in the southward oceanic transport through the Mozambique Channel (Ridderinkhof et al. 2010), one of the source regions for the AC (Biastoch and Krauss 1999; Beal et al. 2006; Elipot and Beal 2015). Overall, our analysis suggests that an aggregate of all climate forcings originating in the tropical Indo-Pacific regions has a significant impact on the interannual variability of the AC. ENSO has a notable influence on the AC, but does not act independently of either the MJO or IOD.
b. Mid- to high-latitude influence
The second atmospheric mode of variance related to variability of the AC represents climate variability predominantly over the Southern Hemisphere subtropical and subpolar latitudes (Fig. 8). PC2 is strongly and significantly correlated with the low-passed and detrended SAM indices derived from ERA-I (0.87), from NCEP2 (0.82), and from observations (0.87) (Fig. 8d). As a result, the contributions of EOF2 (Fig. 8) to the total regression patterns (Fig. 4) are dominant poleward of 20°S, and are generally axisymmetric, although this breaks down a little to the west of Drake Passage (Figs. 8a–c). Close to Antarctica, there is a deepening of the time-mean pressure trough (Fig. 8a), with a related strengthening of the westerlies to the south of their maximum and a weakening to the north (Fig. 8b). This pattern leads to strong positive ∇×τ anomalies in the region of the time-mean zero ∇×τ contour and the ocean subpolar frontal zone, as well as within the southeastern reaches of the subtropical gyre (Fig. 8c). When zonally averaged over the Indian Ocean sector, the contributions of EOF2 are similar to the total regression patterns south of 20°S, but with a poleward shift of 2° to 3° (Figs. 5a–c).
(a)–(c) As in Fig. 6, but for the second EOF pattern. (d) The associated PC time series (PC2), the low-passed AAO (i.e., SAM) indices calculated from ERA-I and NCEP2, and the low-passed SAM index of Marshall (2003), all scaled to transport units based on their regressions with Tjet. The linear trends for the climate indices calculated for the time period of the transport proxy, scaled to transport units, are also drawn in (d) with an arbitrary offset starting from the black open circle.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
Looking at the influence of PC2 on the large-scale ocean circulation and on AC surface velocities using the same analysis as for the first mode, we find that when the AC jet transport is increased southward as the ADT gradient perpendicular to the coast is slightly reduced inshore but is increased offshore (Fig. 7b), but this pattern does not extend upstream. Instead, the greatest anomalies appear locally and in the region of the Agulhas Return Current. The ADT contour displacements suggest some surface circulation changes within the tropical regions, with an inward contraction of the Seychelles Dome and a westward contraction of the subtropical gyre east of Madagascar. At the mid- and high latitudes, ADT contours migrate southeastward in the eastern subtropical Indian Ocean for an increased AC, and within the Southern Ocean the contours migrate southward at many longitudes between 30° and 100°E. This last result is consistent with Kim and Orsi (2014), who found that sea surface height contours associated with the subantarctic and polar fronts shifted poleward between 90° and 100°E during positive SAM events, and with Gille (2014), who found that a transport-weighted average latitude of the ACC is negatively correlated with the SAM (i.e., ACC is shifted south for positive SAM phases). Linear relationships between the SAM and the strength of the ACC, although weak, have also been demonstrated, both in a zonally averaged sense (Gille 2014) and as the ACC flows through Drake Passage (Meredith et al. 2004; Koenig et al. 2016).
Our results corroborate previous findings about the ACC, but also suggest that the influence of the SAM on oceanic circulations extends directly to the Indian Ocean subtropical gyre and to the strength of its WBC. This connection from extratropical latitudes to subtropical latitudes may be related to the positive ∇×τ anomalies between about 40° and 55°S (Figs. 5c and 8c), which extend southward the region of positive curl (not shown), and thus the meridional extent of the subtropical gyre, which appears to translate into a strengthening of the AC. A mechanism that could explain these results is that the strength of the AC is modulated on interannual time scales by Rossby waves excited south of 40°S that are traveling not only westward but also northward along the southern boundary of the subtropical gyre, to eventually reach the latitudes of the AC system. Such meridional propagation of Rossby waves (e.g., Glazman and Weichman 2005) could be induced in general by interactions with the mean flow and bottom topography (Killworth and Blundell 2003), and specifically here by wave refraction as the westward propagating waves impinge on the ACC and the Agulhas return Current flowing east-southeast within the Indian Ocean sector (Hughes 1996). For positive phases of PC2, this could explain how ADT contours migrate southeastward within the eastern subtropical Indian Ocean (Fig. 7b).
The latitudinal position of the subtropical front in the sector south of Africa and the position of the westerlies have been linked to variability in Agulhas leakage in the past, although on decadal and paleo time scales (de Ruijter et al. 1999; Biastoch et al. 2009a; Beal et al. 2011), but the relationship between these and AC strength has not been clear. Here, PC2 and its strong correlation with SAM suggest that both the strength and position of the westerlies in the Indian Ocean sector influence Agulhas transport. However, the regression anomalies from this second mode—like the total regression anomalies (Fig. 5e)—illustrate that only the strength of the westerlies is implicated. This subtlety is further supported by our finding that while PC2 is strongly correlated with the SAM, Tjet is not (0.1 with ERA-I SAM index, 0.14 with the NCEP2 SAM index, null with the observation-based SAM). We understand this to imply that ways in which PC2 differs from the SAM might be responsible for obscuring a direct linear relationship between the SAM and Tjet. As an example, extratropical wind stress anomalies for this mode are the strongest within the Indian Ocean sector (Fig. 8) but that is not the case for the SAM [e.g., Thompson and Wallace 2000; see also Fig. 12 (below)].
Similarly to the tropical Indo-Pacific mode EOF1, our subtropical-subpolar mode EOF2 can explain only 5% of the interannual variability of Tjet (correlation 0.23). The partial canonical time series for this mode has a standard deviation of 1.2 Sv and maximum peak-to-peak amplitude of 5.4 Sv (Fig. 8d).
c. Other influences
The remaining four modes contributing to the total regression pattern, EOFs 3 to 6, are analyzed as their sum (Fig. 9) since none of the corresponding PCs correlate individually with common climate indices. The combined pattern is complex and far less zonally uniform than the first two modes. Notably, there is a positive ∇×τ anomaly region to the south and west of Madagascar, directly over the Mozambique Channel and Agulhas Current system (Fig. 9c), which would act to input positive vorticity. This may drive a regional recirculation that intensifies the AC. Zonally, over the Indian Ocean sector as a whole (Fig. 5), this local subtropical wind anomaly is averaged out, so that the combined contribution from EOFs 3 to 6 appears to have a far smaller magnitude over the subtropics than the individual contributions from EOFs 1 and 2. The sum of EOFs 3–6 nevertheless account for 19% of the interannual variance of the AC transport, since the correlation coefficient between southward Tjet and the partial canonical time series for these combined modes is 0.43 (Fig. 9d). The transport variability from these modes exhibits a strong decadal variability with a typical amplitude of 2.3 Sv and peak-to-peak amplitude of 10 Sv.
(a)–(c) As in Fig. 6, but for EOF patterns 3 to 6. (d) The partial canonical time series obtained from PC time series 3 to 6.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
7. Implications for long-term linear trends
Having quantified the interannual atmospheric forcing of the AC jet transport and characterized its components in terms of climate modes, we turn our attention to multidecadal trends. There is evidence that atmospheric wind stress may be undergoing decadal and longer changes, which are expected to have an impact on oceanic circulation (Marshall 2003; Beal et al. 2011; Swart and Fyfe 2012; Rhein et al. 2013; Wu et al. 2012). In particular, Wu et al. (2012) and Yang et al. (2016) use a mixture of century-long ocean and atmospheric reanalyses and CMIP5 models to suggest that shifting and strengthening westerlies, related to a positive trend in the SAM, have led to a long-term intensification of the Southern Hemisphere western boundary currents, including the AC. In contrast, in an earlier study using our satellite altimeter proxy, we found no significant trend in the transport of the AC since the early 1990s (Beal and Elipot 2016). An extension of the Tjet time series by 14 months, from 1993 to 2016, also exhibits an insignificant trend in southward transport of −0.7 ± 2.2 Sv, confirming our earlier result.
How can we reconcile these seemingly contradictory results? We first estimate the linear trends for MSLP, τx, and ∇×τ within the time period of the altimeter proxy, from October 1992 to October 2015, for both reanalysis products (Fig. 10). The linear trend patterns for all three variables are similar between the two reanalysis products, but differ in magnitude. Generally, NCEP2 exhibits larger trends compared to ERA-I (Rhein et al. 2013; Swart and Fyfe 2012). In particular, the negative trend of the Antarctic circumpolar trough and the positive trend of τx south of 40°S are larger in NCEP2, and the weakening of the trade winds in the tropical Indian Ocean is also stronger and more widespread. We focus on the Indian Ocean sector by calculating the linear trends for the zonally averaged atmospheric variables between 28° and 120°E since 1992, and also since 1979, the beginning of the reanalyses (Fig. 11). While we do not quantify the significance of these trends, we note that in all cases their magnitudes are smaller than one standard deviation at each latitude. We find that the linear trends since 1979 and since 1992 are generally consistent for ERA-I but not for NCEP2, especially at high latitudes.
Southern Hemisphere linear trend patterns for time period October 1992–October 2015 (in respective units per decade) for (top) ERA-I and (bottom) NCEP2 for (left) MSLP, (center) zonal wind stress, and (right) wind stress curl. Note the different scales between ERA-I and NCEP2. In (a) and (d) the mean MSLP for 1993–2016 is contoured from 980 to 1025 mbar at 5-mbar intervals. In (b), (c), (e), and (f) zero contours of mean wind stress curl are drawn in gray, and zero contours of mean zonal wind stress are drawn in black. In all panels the boundary of the barotropic south Indian subtropical gyre referenced to 2000 dbar is drawn in dark green. Color scales saturates at plus and minus the 99th percentile of absolute values of each variable. Landmasses are shaded dark gray.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
Absolute linear change from trend estimates averaged over the Indian Ocean sector (28°–120°E) as a function of latitude for (top) ERA-I and (bottom) NCEP2 for MSLP, zonal stress τx, and wind stress curl ∇×τ, for time periods January 1979–January 2016 and October 1992–October 2015. Red (blue) shaded areas indicates positive (negative) changes. The latitude of the ACT array is indicated by a horizontal dashed line. For τx, horizontal solid lines indicate the latitude of the maximum for the time mean (black), beginning (blue), and the end (red) of the time periods, as implied by the linear change from the trends. For ∇×τ, the lines indicate the latitude of the zero for the time mean, beginning, and end of the time periods. The gray shadings show plus and minus one standard deviation of the zonal averages around their mean values.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
Point-by-point regressions patterns with the SAM index for time period October 1992–April 2016 for (top) ERA-I and (bottom) NCEP2 for (left) MSLP, (center) zonal wind stress, and (right) wind stress curl. In (a) and (d) the mean MSLP for 1993–2016 is contoured from 980 to 1025 mbar at 5-mbar intervals. In (b), (c), (e), and (f) zero contours of mean wind stress curl are drawn in gray, and zero contours of mean zonal wind stress are drawn in black. In all panels the boundary of the barotropic south Indian subtropical gyre referenced to 2000 dbar is drawn in dark green. Color scales saturates at the 99th percentile of absolute values of each variable. Landmasses are shaded dark gray.
Citation: Journal of Climate 31, 8; 10.1175/JCLI-D-17-0597.1
The trends from ERA-I represent a southward displacement and strengthening of the westerlies within the Indian Ocean sector (Fig. 11, top). The zonal wind stress maximum has migrated south by 3.75° and increased by 16% since 1979, and by 0.75° and 4.5% since 1992. In contrast, NCEP2 (Fig. 11, bottom) only exhibits a strengthening over the longer period, and no meridional displacement for either periods. The weaker westerlies trend since the early 1990s could be explained by ozone recovery (Son et al. 2008). Despite their differences, the trends of the westerlies in both products lead to a southward displacement of the zonally averaged position of zero ∇×τ by 2.4°–3° since 1979 (0.8°–1° since 1992). This comes about as a result of a narrowing of the equatorward flank of the westerly jet in both ERA-I and NCEP2 cases, in addition to the shift for ERA-I. North of the ACT array, within the tropics and subtropics, the trends within the Indian Ocean sector are generally weaker. Moreover, the two atmospheric products have dissimilar trends, notably with NCEP2 showing weakened trade winds and a southward displacement of the tropical zero ∇×τ, whereas ERA-I exhibits negligible trends.
Comparing the atmospheric trend patterns (Figs. 10 and 11) to the sensitivity patterns for the AC (Figs. 4 and 5) it is clear they are dissimilar. The most compelling difference at the hemispheric scale is in the patterns of MSLP, which have opposite signs. For ∇×τ, we saw that the Agulhas appears to respond to expansions both north and south of the region of positive ∇×τ in the Indian Ocean (Fig. 5f), which is not evident in the trend patterns from either ERA-I or NCEP2. Also, the region of positive curl to the south and west of Madagascar, which we argued earlier could have a strong influence on the Agulhas as part of the regression pattern explaining 19% of its variance, is replaced by negative curl in the trends. Nevertheless, both the Agulhas sensitivity and linear trends are characterized by an intensification of the westerlies, although an increase of Tjet is not associated with a poleward migration of the westerlies, as seen in ERA-I trends.
Overall, the trend patterns, while being uncertain because they differ between products, are unlike the Agulhas forcing patterns. Assuming that the AC responds to wind forcing similarly at interannual time scales and longer, it therefore seems unlikely that these atmospheric changes can force a significant trend in the Agulhas. Nevertheless, we can model how the Agulhas may change in response to the atmospheric trend patterns by noting their similarities to the SAM regression patterns (Fig. 12), which we know exhibits a positive trend (Marshall 2003). To quantify this explicitly, we calculate the trends of the SAM indices during the time period of the proxy. We find that all three indices have positive trends, which we scale by the regression coefficients between these indices and Tjet (Fig. 8d). In this way, we estimate that the observation-based SAM index implies an insignificant strengthening of AC transport (southward Tjet), and that the reanalyses suggest small but significant strengthening trends of 0.24 ± 0.19 and 0.42 ± 0.25 Sv decade−1 for ERA-I and NCEP2 SAM, respectively. The reanalyses may overestimate the trends (Marshall 2003); nevertheless the trends are still very small and not inconsistent with the Agulhas proxy, which has an insignificant trend of −0.7 ± 2.2 Sv decade−1, encompassing all the SAM-based estimates and their uncertainties. In conclusion, there may be a small strengthening transport trend due to the positive trend in SAM since 1992, but it is an order of magnitude less than the interannual-decadal variability of the Agulhas (5.4 Sv), and therefore currently indiscernible above the uncertainties.
8. Summary and conclusions
We are able to linearly relate 29% of the interannual variance of the AC jet transport to Southern Hemispheric patterns of atmospheric forcing. These patterns are complex, but consistent with a spinup and -down of the subtropical and subpolar circulations and expansion and contraction of the subtropical gyre. Changes in the winds related to the AC appear strongest poleward of the ACT array and are associated with the strength of the westerlies jet over the Indian Ocean sector but not its latitudinal position. The explained variability of the AC jet transport has a standard deviation of 2.9 Sv, a dominant fraction of its total standard deviation of 5.4 Sv. We cannot identify a linear link between Agulhas transport and the interior Sverdrup transport of the Indian Ocean at the latitude band of the ACT array.
We examined the atmospheric regression patterns that drive the AC jet transport using the eigenmode method of DelSole and Yang (2011), allowing us to further decompose the resulting regression patterns into contributions that are clearly associated with common modes of climate variability, namely, ENSO and the SAM. Our study follows a number of other studies that have reported a linear influence of large-scale modes of climate variance on oceanic transport, including the Gulf Stream and the ACC (e.g., Matthews and Meredith 2004; Meredith et al. 2004; DiNezio et al. 2009; Koenig et al. 2016). The first two modes of atmospheric variance related to AC transport can be described as a tropical Indo-Pacific mode, dominated by ENSO, and a subtropical-subpolar mode, dominated by SAM. We find that ENSO alone can explain 11.5% of the interannual variance of the AC, while, surprisingly, the influence of the SAM is insignificant, despite strong regression signals evident in the strength of the westerlies related to Agulhas transport. Concretely, we have been able to quantify the sensitivity of the AC jet transport to ENSO (typical response of 1–2 Sv) and to a SAM-related mode (typical response of 1.2 Sv). Despite a lack of significant correlation, the sensitivity of the AC jet transport solely to the SAM index can be quantified but is weaker, no more than 1 Sv on interannual time scales.
The remaining four modes of atmospheric variance are not correlated with common climate modes, but together account for 19% of the interannual variability of AC transport. Their combined atmospheric patterns are complex, but we conjecture that the presence of positive wind curl directly over the Agulhas Current system and its sources, to the south and west of Madagascar, plays an important role in driving the strong decadal variability, of amplitude 10 Sv, related to these modes.
Some caution over these results is necessary, as our altimeter proxy of AC jet transport captures only 55% of the total variance of the transport as measured in situ (Beal et al. 2015; Beal and Elipot 2016), while the atmospheric reanalyses may suffer from gaps and inconsistencies. Nevertheless, observation-based analyses relating oceanic transport variability to atmospheric circulation and coupled modes are necessary to provide a benchmark for models and ultimately improve predictability.
The response of western boundary currents to atmospheric circulation changes could have significant impact on future climate. Studies based on reanalyses and climate models (Wu et al. 2012; Yang et al. 2016) suggest a strengthening of the AC in response to the positive SAM trend. We find that the AC transport proxy does not have a significant trend, as shown previously (Beal and Elipot 2016), and that its atmospheric regression patterns are unlike atmospheric trend patterns. Further, we use the positive trend in SAM and its relation to AC transport to infer a trend of less than 0.5 Sv decade−1 in relation to shifting and intensification of the westerlies, an order of magnitude less than the interannual variability of the AC and undetectable given observational uncertainties. To conclude, we find it unlikely that there has been a significant intensification of the AC over the past 24 years, given the null trend in the jet proxy and the weak sensitivity of the AC transport to the SAM. The disagreement between our results and those based on reanalyses and climate models may be due to the inability of the latter to properly resolve western boundary currents.
Acknowledgments
NCEP reanalysis 2 data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, from their website (at http://www.esrl.noaa.gov/psd/). The DMI was downloaded from NOAA (http://stateoftheocean.osmc.noaa.gov/sur/ind/dmi.php). The climate indices were downloaded from the sources listed in de Viron et al. (2013), including the NOAA/Earth System Research Laboratory (https://www.esrl.noaa.gov). The altimetry data used are produced by SSALTO/Duacs and distributed by AVISO (http://www.aviso.oceanobs.com/duacs/). This work was supported by the U.S. National Science Foundation through the ACT project, Award OCE-0850891, and the ASCA project, Award OCE-1459543.
REFERENCES
Allan, R., and T. Ansell, 2006: A new globally complete monthly historical gridded mean sea level pressure dataset (HadSLP2): 1850–2004. J. Climate, 19, 5816–5842, https://doi.org/10.1175/JCLI3937.1.
Anderson, D. L. T., and A. E. Gill, 1975: Spin-up of a stratified ocean, with applications to upwelling. Deep-Sea Res., 22, 583–596, https://doi.org/10.1016/0011-7471(75)90046-7.
Anderson, D. L. T., and P. D. Killworth, 1977: Spin-up of a stratified ocean, with topography. Deep-Sea Res., 24, 709–732, https://doi.org/10.1016/0146-6291(77)90495-7.
Beal, L. M., and S. Elipot, 2016: Broadening not strengthening of the Agulhas Current since the early 1990s. Nature, 540, 570–573, https://doi.org/10.1038/nature19853.
Beal, L. M., T. K. Chereskin, Y. D. Lenn, and S. Elipot, 2006: The sources and mixing characteristics of the Agulhas Current. J. Phys. Oceanogr., 36, 2060–2074, https://doi.org/10.1175/JPO2964.1.
Beal, L. M., and Coauthors, 2011: On the role of the Agulhas system in ocean circulation and climate. Nature, 472, 429–436, https://doi.org/10.1038/nature09983.
Beal, L. M., S. Elipot, A. Houk, and G. Leber, 2015: Capturing the transport variability of a western boundary jet: Results from the Agulhas Current time-series experiment (ACT). J. Phys. Oceanogr., 45, 1302–1324, https://doi.org/10.1175/JPO-D-14-0119.1.
Biastoch, A., and W. Krauss, 1999: The role of mesoscale eddies in the source regions of the Agulhas Current. J. Phys. Oceanogr., 29, 2303–2317, https://doi.org/10.1175/1520-0485(1999)029<2303:TROMEI>2.0.CO;2.
Biastoch, A., L. M. Beal, J. R. E. Lutjeharms, and T. G. D. Casal, 2009a: Variability and coherence of the Agulhas Undercurrent in a high-resolution ocean general circulation model. J. Phys. Oceanogr., 39, 2417–2435, https://doi.org/10.1175/2009JPO4184.1.
Biastoch, A., C. W. Böning, F. U. Schwarzkopf, and J. R. E. Lutjeharms, 2009b: Increase in Agulhas leakage due to poleward shift of Southern Hemisphere westerlies. Nature, 462, 495–498, https://doi.org/10.1038/nature08519.
Bryden, H. L., L. M. Beal, and L. M. Duncan, 2005: Structure and transport of the Agulhas Current and its temporal variability. J. Oceanogr., 61, 479–492, https://doi.org/10.1007/s10872-005-0057-8.
Cai, W., A. Sullivan, and T. Cowan, 2011: Interactions of ENSO, the IOD, and the SAM in CMIP3 models. J. Climate, 24, 1688–1704, https://doi.org/10.1175/2010JCLI3744.1.
Cásal, T. G. D., L. M. Beal, R. Lumpkin, and W. E. Johns, 2009: Structure and downstream evolution of the Agulhas Current system during a quasi-synoptic survey in February-March 2003. J. Geophys. Res., 114, C03001, https://doi.org/10.1029/2008JC004954.
Chelton, D. B., and M. G. Schlax, 1996: Global observations of oceanic Rossby waves. Science, 272, 234–238, https://doi.org/10.1126/science.272.5259.234.
De Boer, A. M., R. M. Graham, M. D. Thomas, and K. E. Kohfeld, 2013: The control of the Southern Hemisphere westerlies on the position of the subtropical front. J. Geophys. Res. Oceans, 118, 5669–5675, https://doi.org/10.1002/jgrc.20407.
de Ruijter, W. P. M., J. R. E. Lutjeharms, and P. J. van Leeuwen, 1999: Generation and evolution of natal pulses: Solitary meanders in the Agulhas Current. J. Phys. Oceanogr., 29, 3043–3055, https://doi.org/10.1175/1520-0485(1999)029<3043:GAEONP>2.0.CO;2.
de Ruijter, W. P. M., H. M. van Aken, E. J. Beier, J. R. Lutjeharms, R. P. Matano, and M. W. Schouten, 2004: Eddies and dipoles around South Madagascar: Formation, pathways and large-scale impact. Deep-Sea Res. I, 51, 383–400, https://doi.org/10.1016/j.dsr.2003.10.011.
Dee, D., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
DelSole, T., and X. Yang, 2011: Field significance of regression patterns. J. Climate, 24, 5094–5107, https://doi.org/10.1175/2011JCLI4105.1.
Dencausse, G., M. Arhan, and S. Speich, 2010: Spatio-temporal characteristics of the Agulhas Current retroflection. Deep-Sea Res. I, 57, 1392–1405, https://doi.org/10.1016/j.dsr.2010.07.004.
de Viron, O., J. Dickey, and M. Ghil, 2013: Global modes of climate variability. Geophys. Res. Lett., 40, 1832–1837, https://doi.org/10.1002/grl.50386.
DiNezio, P., L. Gramer, W. Johns, C. Meinen, and M. Baringer, 2009: Observed interannual variability of the Florida Current: Wind forcing and the North Atlantic Oscillation. J. Phys. Oceanogr., 39, 721–736, https://doi.org/10.1175/2008JPO4001.1.
Durgadoo, J. V., B. R. Loveday, C. J. C. Reason, P. Penven, and A. Biastoch, 2013: Agulhas leakage predominantly responds to the Southern Hemisphere westerlies. J. Phys. Oceanogr., 43, 2113–2131, https://doi.org/10.1175/JPO-D-13-047.1.
Elipot, S., and L. M. Beal, 2015: Characteristics, energetics, and origins of Agulhas Current meanders and their limited influence on ring shedding. J. Phys. Oceanogr., 45, 2294–2314, https://doi.org/10.1175/JPO-D-14-0254.1.
Emery, W. J., and R. E. Thomson, 2001: Data Analysis Methods in Physical Oceanography. 2nd ed. Elsevier, 638 pp.
Fetter, A., J. R. E. Lutjeharms, and R. P. Matano, 2007: Atmospheric driving forces for the Agulhas Current in the subtropics. Geophys. Res. Lett., 34, L15605, https://doi.org/10.1029/2007GL030200.
Gadgil, S., P. N. Vinayachandran, P. A. Francis, and S. Gadgil, 2004: Extremes of the Indian summer monsoon rainfall, ENSO and equatorial Indian Ocean oscillation. Geophys. Res. Lett., 31, L12213, https://doi.org/10.1029/2004GL019733.
Ganachaud, A., and C. Wunsch, 2000: Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408, 453–457, https://doi.org/10.1038/35044048.
Ganachaud, A., and C. Wunsch, 2003: Large-scale ocean heat and freshwater transports during the World Ocean Circulation Experiment. J. Climate, 16, 696–705, https://doi.org/10.1175/1520-0442(2003)016<0696:LSOHAF>2.0.CO;2.
Gille, S. T., 2014: Meridional displacement of the Antarctic Circumpolar Current. Philos. Trans. Roy. Soc., 372A, 20130273, https://doi.org/10.1098/rsta.2013.0273.
Glazman, R. E., and P. B. Weichman, 2005: Meridional component of oceanic Rossby wave propagation. Dyn. Atmos. Oceans, 38, 173–193, https://doi.org/10.1016/j.dynatmoce.2004.11.002.
Gray, A. R., and S. C. Riser, 2014: A global analysis of Sverdrup balance using absolute geostrophic velocities from Argo. J. Phys. Oceanogr., 44, 1213–1229, https://doi.org/10.1175/JPO-D-12-0206.1.
Hautala, S. L., D. H. Roemmich, and W. J. Schmilz, 1994: Is the North Pacific in Sverdrup balance along 24°N? J. Geophys. Res., 99, 16 041–16 052, https://doi.org/10.1029/94JC01084.
Hogg, N. G., and W. E. Johns, 1995: Western boundary currents. Rev. Geophys., 33, 1311–1334, https://doi.org/10.1029/95RG00491.
Hughes, C. W., 1996: The Antarctic Circumpolar Current as a waveguide for Rossby waves. J. Phys. Oceanogr., 26, 1375–1387, https://doi.org/10.1175/1520-0485(1996)026<1375:TACCAA>2.0.CO;2.
Johns, W., and Coauthors, 2011: Continuous, array-based estimates of Atlantic Ocean heat transport at 26.5N. J. Climate, 24, 2429–2449, https://doi.org/10.1175/2010JCLI3997.1.
Josey, S. A., S. Gulev, and L. Yu, 2013: Exchanges through the ocean surface. Ocean Circulation and Climate: A 21st Century Perspective, G. Siedler et al., Eds., International Geophysics Series, Vol. 103, Academic Press, 115–140.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. Hnilo, M. Fiorino, and G. Potter, 2002: NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1643, https://doi.org/10.1175/BAMS-83-11-1631.
Karoly, D. J., 1989: Southern Hemisphere circulation features associated with El Niño-Southern Oscillation events. J. Climate, 2, 1239–1252, https://doi.org/10.1175/1520-0442(1989)002<1239:SHCFAW>2.0.CO;2.
Kazmin, A. S., 2017: Variability of the climatic oceanic frontal zones and its connection with the large-scale atmospheric forcing. Prog. Oceanogr., 154, 38–48, https://doi.org/10.1016/j.pocean.2017.04.012.
Killworth, P. D., and J. R. Blundell, 2003: Long extratropical planetary wave propagation in the presence of slowly varying mean flow and bottom topography. Part II: Ray propagation and comparison with observations. J. Phys. Oceanogr., 33, 802–821, https://doi.org/10.1175/1520-0485(2003)33<802:LEPWPI>2.0.CO;2.
Kim, Y. S., and A. H. Orsi, 2014: On the variability of Antarctic Circumpolar Current fronts inferred from 1992–2011 altimetry. J. Phys. Oceanogr., 44, 3054–3071, https://doi.org/10.1175/JPO-D-13-0217.1.
Koenig, Z., C. Provost, Y.-H. Park, R. Ferrari, and N. Sennéchael, 2016: Anatomy of the Antarctic Circumpolar Current volume transports through Drake Passage. J. Geophys. Res. Oceans, 121, 2572–2595, https://doi.org/10.1002/2015JC011436.
Kutzbach, J. E., 1967: Empirical eigenvectors of sea-level pressure, surface temperature and precipitation complexes over North America. J. Appl. Meteor., 6, 791–802, https://doi.org/10.1175/1520-0450(1967)006<0791:EEOSLP>2.0.CO;2.
L’Heureux, M. L., and D. W. J. Thompson, 2006: Observed relationships between the El Niño–Southern Oscillation and the extratropical zonal-mean circulation. J. Climate, 19, 276–287, https://doi.org/10.1175/JCLI3617.1.
Loveday, B. R., J. V. Durgadoo, C. J. Reason, A. Biastoch, and P. Penven, 2014: Decoupling of the Agulhas leakage from the Agulhas Current. J. Phys. Oceanogr., 44, 1776–1797, https://doi.org/10.1175/JPO-D-13-093.1.
Marshall, G. J., 2003: Trends in the southern annular mode from observations and reanalyses. J. Climate, 16, 4134–4143, https://doi.org/10.1175/1520-0442(2003)016<4134:TITSAM>2.0.CO;2.
Matano, R. P., C. G. Simionato, and P. T. Strub, 1999: Modeling the wind-driven variability of the south Indian Ocean. J. Phys. Oceanogr., 29, 217–230, https://doi.org/10.1175/1520-0485(1999)029<0217:MTWDVO>2.0.CO;2.
Matthews, A. J., and M. P. Meredith, 2004: Variability of Antarctic circumpolar transport and the southern annular mode associated with the Madden-Julian oscillation. Geophys. Res. Lett., 31, L24312, https://doi.org/10.1029/2004GL021666.
Meredith, M. P., P. L. Woodworth, C. W. Hughes, and V. Stepanov, 2004: Changes in the ocean transport through Drake Passage during the 1980s and 1990s, forced by changes in the southern annular mode. Geophys. Res. Lett., 31, L21305, https://doi.org/10.1029/2004GL021169.
Mo, K. C., and J. N. Paegle, 2001: The Pacific–South American modes and their downstream effects. Int. J. Climatol., 21, 1211–1229, https://doi.org/10.1002/joc.685.
Munk, W. H., 1950: On the wind-driven ocean circulation. J. Meteor., 7, 80–93, https://doi.org/10.1175/1520-0469(1950)007<0080:OTWDOC>2.0.CO;2.
Pohl, B., N. Fauchereau, C. J. C. Reason, and M. Rouault, 2010: Relationships between the Antarctic Oscillation, the Madden–Julian oscillation, and ENSO, and consequences for rainfall analysis. J. Climate, 23, 238–254, https://doi.org/10.1175/2009JCLI2443.1.
Putrasahan, D., B. P. Kirtman, and L. M. Beal, 2016: Modulation of SST interannual variability in the Agulhas leakage region associated with ENSO. J. Climate, 29, 7089–7102, https://doi.org/10.1175/JCLI-D-15-0172.1.
Rasmusson, E. M., P. A. Arkin, W.-Y. Chen, and J. B. Jalickee, 1981: Biennial variations in surface temperature over the United States as revealed by singular decomposition. Mon. Wea. Rev., 109, 587–598, https://doi.org/10.1175/1520-0493(1981)109<0587:BVISTO>2.0.CO;2.
Rhein, M., and Coauthors, 2013: Observations: Ocean. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 255–317.
Ridderinkhof, H., P. M. Van Der Werf, J. E. Ullgren, H. M. Van Aken, P. J. Van Leeuwen, and W. P. M. De Ruijter, 2010: Seasonal and interannual variability in the Mozambique Channel from moored current observations. J. Geophys. Res., 115, C06010, https://doi.org/10.1029/2009JC005619.
Ridgway, K., and J. Dunn, 2007: Observational evidence for a Southern Hemisphere oceanic supergyre. Geophys. Res. Lett., 34, L13612, https://doi.org/10.1029/2007GL030392.
Ridgway, K., J. Dunn, and J. Wilkin, 2002: Ocean interpolation by four-dimensional weighted least squares application to the waters around Australasia. J. Atmos. Oceanic Technol., 19, 1357–1375, https://doi.org/10.1175/1520-0426(2002)019<1357:OIBFDW>2.0.CO;2.
Roemmich, D., J. Gilson, P. Sutton, and N. Zilberman, 2016: Multidecadal change of the South Pacific Gyre circulation. J. Phys. Oceanogr., 46, 1871–1883, https://doi.org/10.1175/JPO-D-15-0237.1.
Rouault, M., P. Penven, and B. Pohl, 2009: Warming in the Agulhas Current system since the 1980’s. Geophys. Res. Lett., 36, L12602, https://doi.org/10.1029/2009GL037987.
Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360–363, https://doi.org/10.1038/43854.
Salmon, R., 1998: Lectures on Geophysical Fluid Dynamics. Oxford University Press, 400 pp.
Schmitz, W. J., J. D. Thompson, and J. R. Luyten, 1992: The Sverdrup circulation for the Atlantic along 24°N. J. Geophys. Res., 97, 7251–7256, https://doi.org/10.1029/92JC00417.
Schott, F. A., S.-P. Xie, and J. P. McCreary Jr., 2009: Indian Ocean circulation and climate variability. Rev. Geophys., 47, 1–46, https://doi.org/10.1029/2007RG000245.
Shaffer, J. P., 1995: Multiple hypothesis testing. Annu. Rev. Psychol., 46, 561–584, https://doi.org/10.1146/annurev.ps.46.020195.003021.
Son, S.-W., and Coauthors, 2008: The impact of stratospheric ozone recovery on the Southern Hemisphere westerly jet. Science, 320, 1486–1489, https://doi.org/10.1126/science.1155939.
Sprintall, J., A. L. Gordon, A. Koch-Larrouy, T. Lee, J. T. Potemra, K. Pujiana, and S. E. Wijffels, 2014: The Indonesian seas and their role in the coupled ocean–climate system. Nat. Geosci., 7, 487–492, https://doi.org/10.1038/ngeo2188.
Swart, N. C., and J. C. Fyfe, 2012: Observed and simulated changes in the Southern Hemisphere surface westerly wind-stress. Geophys. Res. Lett., 39, L16711, https://doi.org/10.1029/2012GL052810.
Thomas, M. D., A. M. De Boer, H. L. Johnson, and D. P. Stevens, 2014: Spatial and temporal scales of Sverdrup balance. J. Phys. Oceanogr., 44, 2644–2660, https://doi.org/10.1175/JPO-D-13-0192.1.
Thompson, D. W. J., and J. M. Wallace, 2000: Annular modes in the extratropical circulation. Part I: Month-to-month variability. J. Climate, 13, 1000–1016, https://doi.org/10.1175/1520-0442(2000)013<1000:AMITEC>2.0.CO;2.
Trenberth, K. E., and J. M. Caron, 2001: Estimates of meridional atmosphere and ocean heat transports. J. Climate, 14, 3433–3443, https://doi.org/10.1175/1520-0442(2001)014<3433:EOMAAO>2.0.CO;2.
Wu, L., and Coauthors, 2012: Enhanced warming over the global subtropical western boundary currents. Nat. Climate Change, 2, 161–166, https://doi.org/10.1038/nclimate1353.
Wunsch, C., 2011: The decadal mean ocean circulation and Sverdrup balance. J. Mar. Res., 69, 417–434, https://doi.org/10.1357/002224011798765303.
Wunsch, C., and D. Roemmich, 1985: Is the North Atlantic in Sverdrup balance? J. Phys. Oceanogr., 15, 1876–1880, https://doi.org/10.1175/1520-0485(1985)015<1876:ITNAIS>2.0.CO;2.
Yang, H., G. Lohmann, W. Wei, M. Dima, M. Ionita, and J. Liu, 2016: Intensification and poleward shift of subtropical western boundary currents in a warming climate. J. Geophys. Res. Oceans, 121, 4928–4945, https://doi.org/10.1002/2015JC011513.
Yokoi, T., T. Tozuka, and T. Yamagata, 2008: Seasonal variation of the Seychelles Dome. J. Climate, 21, 3740–3754, https://doi.org/10.1175/2008JCLI1957.1.
Zhang, C., 2005: Madden-Julian oscillation. Rev. Geophys., 43, RG2003, https://doi.org/10.1029/2004RG000158.