Abstract

The Southern Ocean (SO) is the region of the World Ocean bordering on Antarctica over which significant exchanges between the atmosphere, the ocean, and the sea ice take place. Here, the strong and nearly unhindered eastward flow of the Antarctic Circumpolar Current plays an important role in mean global climate as it transmits climate anomalies around the hemisphere. Features of interannual variability have been observed to propagate eastward around the SO with the circumpolar flow in the form of a system of coupled anomalies, known as the Antarctic circumpolar wave (ACW). In the present study, the 142-yr series of the Twentieth Century Reanalysis, version 2, dataset (850-hPa geopotential height, sea level pressure, sea surface temperature, surface meridional wind, and surface air temperature) spanning from 1871 to 2012 is used to investigate the presence and variability of ACWs. This examination shows, for the first time, the presence of the ACW before the mid-1950s and interdecadal changes in its characteristics. Modifications in the strength and speed of the circumpolar wave are shown to be linked with large-scale climate changes. Complex empirical orthogonal function analyses confirm that the ACW becomes apparent when the tropical El Niño–Southern Oscillation (ENSO) signal gives rise to the Pacific–South American (PSA) pattern and is a consequence of the constructive combination of the PSA and the subantarctic zonal wavenumber 3. Correlation analyses are also performed to quantify the role played by ENSO teleconnections for the appearance of the ACW, and the impact on the presence of ACWs of three super–El Niño events is investigated.

1. Introduction

The Antarctic circumpolar wave (ACW) is a large-scale oceanic and atmospheric pattern of covarying anomalies that propagates eastward across the Southern Ocean (SO) on subdecadal time scales (White and Peterson 1996, hereafter WP96; Peterson and White 1998; White and Simmonds 2006). This interactive phenomenon, first detected in sea level pressure (SLP) and meridional wind stress data in the period 1985–94 by WP96, originates from lower latitudes in the western subtropical Pacific Ocean (Peterson and White 1998) and manifests as an eastward phase-locked propagating signal in sea surface temperature (SST), precipitable water, and sea ice extent anomalies. The ACW takes about 8–10 yr to encircle the SO and shows a circumpolar wave-2 anomaly pattern, an average speed of 6–8 cm s−1, and a dominant periodicity of 4–5 yr per wave couplet (WP96; Peterson and White 1998). Jacobs and Mitchell (1996) found that sea surface height (SSH) also displays variability consistent with the ACW. Spatially, the ACW is strongest from the southwest Pacific through to the South Atlantic sectors of the subantarctic, where it may be impacted by El Niño–Southern Oscillation (ENSO) teleconnections (Peterson and White 1998). Each wave couplet has internally consistent anomalies of meridional wind, temperature, upper-ocean salinity, and sea ice conditions. For the area east of the low/west of the high, these anomalies take the form of northerlies, warm air, oceanic downwelling and reduced salinity, and sea ice retreat. Over the area west of the low/east of the high, the anomalies assume the opposite phase (i.e., southerlies, cold air, upwelling and increased salinity, and sea ice advance). A possible effect on the sea ice thickness is not discussed owing to limitations in accuracy as well as in temporal and spatial coverage of the available information (e.g., Aulicino et al. 2014; Holland et al. 2014).

The ACW has been studied in several oceanic and/or atmospheric global circulation models, and some of these also focused on atmospheric planetary wave dynamics coupled to ocean models (Qiu and Jin 1997; White et al. 1998; Christoph et al. 1998; Motoi et al. 1998; Cai et al. 1999; Haarsma et al. 2000). In these works several patterns in the atmosphere and ocean circulation associated with the ACW have been identified, the most common being zonal wavenumber 2 and zonal wavenumber 3 (ZW3) and the Pacific–South American (PSA) pattern (Mo and White 1985, hereafter MW; Karoly 1989; Irving and Simmonds 2016). Other studies focused on the local ocean–atmosphere coupled instability mechanisms responsible for generation of the ACW mode and on the important role that the Antarctic Circumpolar Current (ACC) plays in advecting the SST anomalies eastward. The observed low-frequency variability in the SST signal also has been considered as a possible energy source available to provide broadband forcing for interactive ACW modes (Baines and Cai 2000).

The generating mechanisms of the ACW, its characteristics, local effects, persistence, and presence have also been investigated in a number of studies concerned with the variability of the SO on different time scales (White and Cherry 1999; White 2000; White and Cayan 2000; Carril and Navarra 2001; Gloersen and White 2001; White and Chen 2002; White et al. 2002; Connolley 2003; Marsland et al. 2003; ,Fischer et al. 2004). Cai and Baines (2001) showed that the observed ACW may be seen as the sum of wave-2 and ZW3 patterns during the two last decades of the twentieth century. In particular, they demonstrated that the wave-2 component of the ACW is forced by the PSA pattern of SLP anomalies in the southern high latitudes, atmospherically teleconnected to the tropical standing mode of ENSO, while the ZW3 pattern takes a more important role when the intensity of ENSO decreases. They also argued that during transition between different ENSO phases (e.g., when the remote forcing becomes weak or absent) the covarying SST and SLP anomalies generating the PSA propagate eastward into the eastern Atlantic, Indian, and western Pacific sectors of the SO in the form of an ACW. White et al. (2002) found the ACW in the eastern Pacific and western Atlantic sectors to be in damped resonance with remote forcing stemming from the slow eastward phase propagation of covarying SST and SLP anomalies in the global ENSO wave across the Pacific warm pool (White and Cayan 2000). White et al. (2004) found five individual signals dominating climate variability near the sea ice edge with different periodicities, four of these showing ACW characteristics with a 3.7-yr period signal leading with a wave-2 pattern from 1983 to 1992. On the other hand, Venegas (2003) found that most of ACW’s variance comes from a linear combination of a quasi-standing ZW3 pattern with a periodicity of about 3 yr (ACW-3) and a propagating wave-2 pattern with a periodicity of 5 yr (ACW-2). She argued that the ACW-3 exhibits coupled air–sea interactions, and the ACW-2 represents the southern response to the tropical ENSO activity. She also pointed out that the constructive or destructive interference of the two modes with different periodicity and wavelengths determines the observed irregular fluctuations of the ACW.

Interdecadal changes observed in the ACW variability and its modifications in strength and speed have been investigated in connection with large-scale climate changes. Bian and Lin (2012) noted that tropical and subtropical ocean warming and ozone depletion tend to weaken the ACW. Their analysis supports the hypothesis that ENSO can play a role in setting the periods of ACW oscillations. In turn these oscillations can be superimposed upon those induced by the southern annular mode (SAM) at different times and places, leading to amplification or weakening of the ACW. On the other hand, White and Annis (2004) proposed that changes in the appearance of the ACW could have been determined by El Niño variability during the last half of the twentieth century. They observed that before 1977 the ACW expanded equatorward into a warmer subtropical South Pacific Ocean, while after 1977 the ACW receded from a colder subtropical South Pacific Ocean with an equatorward expansion into a warmer subtropical south Indian Ocean. This behavior seems consistent with changes in circulation observed during the mid-1970s over the Southern Hemisphere (SH), namely, the positive trend of the SAM linked to the ozone depletion (Polvani et al. 2011), the concomitant poleward shift of the Hadley and Ferrel cells (Thompson and Wallace 2000), the regime shift of the westerlies (Bian and Lin 2012), and the abrupt changes in the atmosphere–ocean system in the mid- and high latitudes during 1973–78 (Xiao and Li 2007). Recently, Giarolla and Matano (2013), through the investigation of SSH, SST, and wind stress curl data from 1993 to 2010, found two eastward-propagating low-frequency wavelike modes compatible with an eastward-propagating ACW-like pattern. A first wave was identified between 1997 and 2007, and a second was found to start in 2000. Both waves completed a cycle around the globe in about 10 yr, having a periodicity consistent with the ACW period identified by WP96. Because of the long ACW time scales, predictability of regional anomalies is potentially possible several seasons ahead; however, the temporal variability in strength of the pattern between recent decades somewhat limits its prognostic value.

This paper investigates the ACW presence and variability from 1871 to 2012. Our attention is directed to the circumpolar domain (56°–66°S) where the phenomenon exhibits the greatest magnitude (WP96) and where the eastward flow of the ACC and its fronts are the major oceanographic features (e.g., Orsi et al. 1995; Cotroneo et al. 2013). One of our objectives is to ascertain, using this long record, if ACWs occurred when the constructive combination of the ENSO and ZW3 patterns takes place at mid- and high latitudes of the SH, as argued by Cai and Baines (2001), Venegas (2003), and others. The importance of ENSO in modulating or forcing the ACW anomalies around Antarctica has been discussed in the literature. Many studies argued that ENSO plays a leading role in forcing the ACWs in the western subtropical South Pacific Ocean and considered the propagating component of the PSA pattern in the southern Pacific, Atlantic, and western Indian sectors as the essential part of the ACW. On the other hand, some authors found the ACW pattern around Antarctica even when there is no ENSO teleconnection in the SO, which would suggest that the ACW is an independent mode of variability at the midlatitudes, not tropically forced. Our analysis will clarify whether the ACWs are related to the presence of ENSO teleconnections and if these propagating waves are robustly connected with super–El Niño events (Hong et al. 2014). The work is divided as follows. Data sources and methods of analysis are described in section 2. The climate indices used in the paper are discussed in section 3. The SO variability is examined in section 4. The circumpolar ACW variability is described in section 5, and conclusions are presented in section 6.

2. Data and methods

To appreciate the ACW’s complex structure, the data analyses have been conducted over the broad area extending from 20°S to the South Pole. Since the ACW variability shows the greatest magnitudes around 56°S (WP96), particular attention has been paid to the circumpolar belt between 56° and 66°S.

a. Gridded datasets

In this study we use monthly means of SLP, 850-hPa geopotential height (HGT850), surface air temperature (SAT), and meridional surface wind (VWIND) available on a 2° regular grid from the Twentieth Century Reanalysis, version 2 (20CRV2), data provided by the National Oceanic and Atmospheric Administration (NOAA) and the Cooperative Institute for Research in Environmental Sciences (CIRES), spanning 1871–2012. The 20CRV2 utilizes a new version of the National Centers for Environmental Prediction (NCEP) atmosphere–land model with time-evolving interpolated monthly SST and sea ice concentration (SIC) fields from the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) dataset as prescribed boundary conditions, newly compiled surface pressure observations, and the radiative effects of historical time-varying CO2 concentrations, volcanic aerosol, and solar variations (Compo et al. 2011). The reanalysis is performed with the ensemble Kalman filter described in Compo et al. (2011) and based on the method of Whitaker and Hamill (2002) to produce an estimate of the complete state of the atmosphere and the uncertainty in that estimate. The short-term forecast ensemble is generated in parallel from 56 nine-hour integrations of a state-of-the-art atmospheric general circulation model, a 2008 updated experimental version of the atmospheric component of NCEP’s operational Climate Forecast System model (Saha et al. 2006). In this paper the means of the analysis ensemble fields have been used.

Because the dataset is derived using only observations of synoptic surface pressure and the uncertainty is approximately inversely proportional to the density of observations, one could question the value of the 20CRV2 in the SH, where there are very few historical in situ data. Bengtsson et al. (2004) and Kanamitsu and Hwang (2006), using a three-dimensional variational (3DVAR) data assimilation system, such as used in the NCEP–National Center for Atmospheric Research (NCAR), NCEP–Department of Energy (DOE), and 40-yr European Centre for Medium-Range Weather Forecasts Re-Analysis (ERA-40) reanalyses, have suggested that surface pressure observations alone cannot be used to make a good analysis of midtropospheric circulation, even in the relatively data-rich Northern Hemisphere. However, more recent investigations of Compo et al. (2006) have shown that additional newly recovered surface pressure observations is sufficient to generate useful weather maps of the low-tropospheric extratropical circulation back to 1930 over the SH. Moreover, it has also been observed that by using a more advanced data assimilation system, such as four-dimensional variational (4DVAR) and ensemble Kalman filtering, it is feasible to produce high-quality upper-air fields using only surface pressure observations (Bengtsson 1980; Whitaker et al. 2004, 2009; Anderson et al. 2005; Compo et al. 2006), “provided the assimilation system is ‘dynamical’ enough” (Thépaut 2006). Uncertainty estimates are consistent with actual differences between first guess and pressure observations even as the network changes over more than 100 yr (Compo et al. 2011). The beneficial impact of these synoptic observations on the analysis of the tropospheric circulation has been demonstrated for synoptic time scales and even more so when those observations are composited and averaged to produce multiyear means (Compo et al. 2011).

We stress that here we use the monthly means rather than the 6-hourly analyses. The standard error of the mean decreases as 1/(N)1/2, where N is the number of observations, and hence the monthly means have an order of magnitude less variance than the synoptic analyses. The data presented in Fig. 9b of Compo et al. (2011) indicate that the number of monthly observations south of 20°S very rarely falls below 180, a number that suggests monthly analyses of considerable value. We also comment that recent studies (Pena-Ortiz et al. 2013; Freitas et al. 2015) have explicitly pointed to the value of the 20CR data in the SH. As an additional aspect of assessing the quality of the reanalysis we have interpolated the individual SLP monthly means in 20CRV2 to the locations of Faraday/Vernadsky (65°15′S, 64°16′W) and Darwin (12°28′S, 130°49′E) and have correlated these time series with those of the observations at these two locations. For Faraday/Vernadsky we obtain a strong correlation of r = 0.97 (p < 0.01) for the period 1950–2006 and r = 0.96 (p < 0.01) for the period 1876–2012 at Darwin. We also calculate these correlations after removing the seasonal cycle in the time series obtaining a correlation of r = 0.96 (p < 0.01) for Faraday–Vernadsky (1950–2006) and of r = 0.77 (p < 0.01) at Darwin (1876–2012). These approaches of calculating correlations are both important components of the assessment of the reanalysis dataset deviations. We comment that the original correlations above indicate how well 20CRV2 captures the important seasonal cycle, while the second (with the deseasonalized data) reflects how well 20CRV2 captures the monthly anomalies. Hence, we can have considerable confidence in using 20CRV2 for our study, which focuses on low-frequency variations.

Monthly means of SST data provided through a recent version of the Extended Reconstructed SST, version 3b (ERSST.v3b), analysis that is available on a 2° regular grid from International Comprehensive Ocean–Atmosphere Data Set have been used. ERSST.v3b is generated using in situ buoy/ship SST observations and improved statistical methods that allow stable reconstruction using sparse data when satellite data are not used. This monthly analysis begins January 1854, but because of sparse data the SST signal is heavily damped before 1880. Afterward the strength of the signal is more consistent over time. Although ERSST shows key limitations as the large uncertainty affecting the Antarctic Circumpolar Ocean (Smith and Reynolds 2003, 2004; Smith et al. 2008), it is suitable for long-term global and basinwide studies as in our analyses.

A climatological mean over the 142 yr (1871–2012) is calculated for each month. These means are subtracted, by month, from the time series of monthly averages to remove the seasonal cycle in the series.

b. Analysis techniques

In our paper, two different techniques of analysis have been applied, namely the caterpillar singular spectrum analysis (SSA) algorithm and the multitaper method (MTM) of spectral analysis.

The SSA algorithm has been applied for prefiltering time series prior to spectral analysis. The aim of the SSA is to make a decomposition of a time series into the sum of a small number of independent and interpretable components such as a slowly varying trend, periodic or quasi-periodic oscillatory components, and random noise. In our analysis, periodicities as long as four months are captured. For a more detailed description of the method see Vautard et al. (1992), Allen and Smith (1997), Ghil et al. (2002), Hassani (2007), and Myung (2009).

The MTM of spectral analysis provides a powerful method for spectral estimation (Thomson 1982; Percival and Walden 1993) and signal reconstruction (e.g., Park 1992) of a time series, which is believed to exhibit a spectrum containing both continuous and singular components. The aim of the MTM is to compute a set of independent estimates of the power spectrum by premultiplying the data by orthogonal tapers with an optimal trade-off between spectral resolution and the stability or “variance” properties of the spectral estimate. For a more detailed description of the method see Thomson (1982), Park (1992), Percival and Walden (1993), and Mann and Lees (1996).

We also investigate the properties of our data with complex empirical orthogonal function (CEOF) analysis, a procedure suitable for studying propagating waves (Bretherton et al. 1992; Cherry 1996, 1997; Cai and Baines 2001).

3. Southern Hemisphere climate indices

Two different SAM indices are used in this study: 1) the station-based index of Marshall (2003), which spans the period from 1957 to 2012, and 2) the leading principal component (PC) of HGT850 anomalies south of 20°S (Thompson and Wallace 2000). EOF analysis has been applied to all HGT850 monthly anomalies over our domain, allowing us to identify the leading PC for describing the SAM pattern oscillations from 1871 to 2012. This PC explains 17.9% of variance and the correlation between the index of Marshall (2003), and our PC for the period 1957–2012 is r = 0.83 (p < 0.01), highlighting their strong similarity. [For convenience, we have chosen the (arbitrary) sign of the leading EOF (and its PC) in such a way that our SAM signal (Fig. 1a) displays a positive summer and autumn trend as does the SAM index of Marshall (2003)].

Fig. 1.

Patterns of weights of EOF1 of HGT850 deseasonalized monthly anomalies over (a) 1871–2012, (b) 1957–2012, (c) 1979–2012, and (d) 1957–78. They account for 17.9%, 17.1%, 16.3%, and 19.2%, respectively, of the total variance on monthly time scales. Contours are at interval of 0.003 and white-to-red (blue) colors indicate positive (negative) weights.

Fig. 1.

Patterns of weights of EOF1 of HGT850 deseasonalized monthly anomalies over (a) 1871–2012, (b) 1957–2012, (c) 1979–2012, and (d) 1957–78. They account for 17.9%, 17.1%, 16.3%, and 19.2%, respectively, of the total variance on monthly time scales. Contours are at interval of 0.003 and white-to-red (blue) colors indicate positive (negative) weights.

To allow an assessment of the statistical robustness of the first EOF over the period 1871–2012, EOF analyses have been conducted for the HGT850 monthly anomalies (south of 20°S) over two subperiods (i.e., 1957–2012 and 1979–2012) of relevance in our investigation. The first period is coincident with the temporal length of station-based values used by Marshall (2003) to compile the SAM index series, while the second covers the satellite epoch. The leading EOF derived from the period 1957–2012 explains 17.1% of variance (Fig. 1b), while the leading EOF over the period 1979–2012 explains 16.3% of variance (Fig. 1c). The three patterns displayed in Fig. 1 are very similar, albeit with the hint of a stronger feature protruding into the southeast Pacific for the more recent epoch. The two PC time series show correlations of r = 0.86 (p < 0.01) with the Marshall index over each of the corresponding periods, thus highlighting their strong linkage and similarity. In addition, the PC of the first EOF derived from the period 1871–2012 shows a very strong association with the leading PC referred to the period 1957–2012 (r = 0.98 and p < 0.01 over the period 1957–2012) and with that calculated over the period 1979–2012 (r = 0.96 and p < 0.01 over the period 1979–2012). The SAM pattern has shown a change in its structure over time since the leading EOF derived from the period 1979–2012 (Fig. 1c) shows lower pressure values over Antarctica and the presence of a meridional bulging in the Pacific sector, which is not as marked in the first EOF performed over the entire record (Fig. 1a). In addition, the EOF referred to 1979–2012 shows the increase of the high pressure values in the anticyclonic belt at the midlatitudes, where a clear ZW3 pattern can be observed.

Finally in connection with these EOF patterns we remark that they all have the period 1979–2012 in common, and hence it may not be surprising they exhibit similar structures. To quantify this influence we have also performed an EOF analysis for the nonoverlapping periods 1957–78 (Fig. 1d). It is reassuring to note that the resulting EOF1 (which explains 19.2% of the variance) bears great similarity to the three presented above, albeit with some interesting phase shifts.

For estimating the SH response to the tropical ENSO events, monthly anomalies of the Southern Oscillation index (SOI) have been utilized. The SOI is defined as the normalized atmospheric surface pressure difference between Tahiti and Darwin (Ropelewski and Jones 1987) [see Allan et al. (1991) and Können et al. (1998) for details of the early pressure sources].

The ZW3 oscillating pattern is quantified here by a minor modification of the index of MW. Specifically, a monthly index for the ZW3 oscillating pattern at mid- and high latitudes through the SLP monthly mean anomalies has been constructed from 1871 to 2012. One-point correlations are calculated between monthly anomalies at a base point of 50°S, 96°E and anomalies over grid points of the southern domain south of 20°S. Our base point is shifted one degree east compared to the MW base point, it being the closest grid point to their selected location. Two maxima of positive correlation are found at 58°S, 150°W and 38°S, 4°W (not shown). These points represent the second and the third centers of the ZW3 pattern. The second center is at the same location of MW, while the third center is shifted of one degree east with respect to their selected point, being, as above, our closest SLP grid point. The normalized time series of monthly SLP anomalies (SLP*) are determined at these three points and then the ZW3 index is constructed from the following:

 
formula

A positive index denotes anomalously high SLP over the three maxima and vice versa. The correlation coefficients at the second and the third centers away from the base point are smaller than those presented by MW. This is understandable because we have not performed a seasonal analysis and also the number of samples utilized is much larger than that used by MW (we used monthly means over 142 yr, a total of 1704 samples, while they used only 8 yr, 40 samples of winter data). The correlation coefficient in the centers located in the Pacific and Atlantic sectors reach maximum values of r = 0.37 and r = 0.43, respectively (p < 0.01). This indicates that, assuming one independent event every 2 months, the ZW3 pattern in our correlation maps is robust when a number of samples significantly larger than that of MW is used. The ZW3 index reflects the variability on time scales longer than 2 months, and, although it is not derived on seasonal scale as done by MW, it exhibits a stable ZW3 pattern during all calendar months. Some authors have commented that the MW index of ZW3 has the shortcoming of being sensitive to zonal phase shifts. This is because the stationary nature of the MW approach cannot fully capture the seasonal phase shifts (approximately 15° of longitude on average) experienced by ZW3 (van Loon and Rogers 1984; MW) or the occurrence of patterns whose phase does not approximately coincide with the location of the three analysis points. For this reason, new approaches to diagnosing SH planetary wave activity have been developed (Irving and Simmonds 2015). However, the MW approach is suitable for the investigations undertaken here.

4. Spectral analysis of SLP and SST circumpolar anomalies

Four indices for monthly and standardized SLP, SST, SAT, and VWIND anomalies, zonally and meridionally (56°–66°S) averaged, have been constructed to collectively represent the circumpolar variability in our 142 yr of record (Figs. 2a–d). The decrease in pressure over the satellite era is consistent with that shown by Simmonds (2015) for the ERA-Interim data. The accumulated (in time) circumpolar indices, containing the cumulative sum of the elements of each index, have been calculated (Fig. 2e) in order to more clearly reveal the low-frequency variations. These cumulative indices are a very insightful way of showing changes and trends of the original indices. All cumulative series show significant variations in sign, which are consistent with the climate changes starting in the middle of last century. The abrupt shift shown by cumulative time series, starting around the mid-1970s, is particularly noteworthy since it indicates that in this decade a general change in the atmospheric and oceanic circulation took place. From this decade until present, the circumpolar SLP and VWIND cumulative anomalies exhibit negative tendencies, which are consistent with the warming showed by the circumpolar SST and SAT cumulative anomalies (Fig. 2e), although a weak negative tendency in SST cumulative anomalies is discernible starting from the early 2000s. There are compelling reasons to believe that the negative tendency shown by the VWIND cumulative anomalies (indicating, on average, increased poleward flow of warm air) is consistent with the increasing amplitude of the planetary waves in the SH observed by D’Agostino and Lionello (2017) in the last century. They found the change in wave amplitude to correspond with an intensified thermal gradient between the mid- and high latitudes and the strengthening of the baroclinicity and Ferrel cell activity. To quantify the connection, we have calculated the correlation coefficient between our cumulative VWIND anomaly series and the cumulative anomalies of the planetary wave index (PWI) used by D’Agostino and Lionello (2017) to describe the behavior of the baroclinic activity between 40° and 70°S. We found a value of r = −0.96 (p < 0.01), which highlights their strong correspondence.

Fig. 2.

Time series of (a) SLP, (b) SST, (c) SAT, and (d) VWIND standardized monthly anomalies zonally (0°–360°) and meridionally (56°–66°S) averaged. The black line stands for low-pass-filtered anomalies retaining variability longer than 2 yr. (e) The accumulated circumpolar SLP (blue), SST (red), SAT (green), and VWIND (black) indices of standardized monthly anomalies averaged between 56° and 66°S. The time series span the 142-yr period from 1871 to 2012.

Fig. 2.

Time series of (a) SLP, (b) SST, (c) SAT, and (d) VWIND standardized monthly anomalies zonally (0°–360°) and meridionally (56°–66°S) averaged. The black line stands for low-pass-filtered anomalies retaining variability longer than 2 yr. (e) The accumulated circumpolar SLP (blue), SST (red), SAT (green), and VWIND (black) indices of standardized monthly anomalies averaged between 56° and 66°S. The time series span the 142-yr period from 1871 to 2012.

The spectral analysis conducted upon the Antarctic circumpolar domain reveals remarkable features related to the ACW. The MTM coherence spectrum analysis has been performed combining the circumpolar SLP and SST anomalies, with each field prefiltered through the SSA algorithm. The magnitude-squared coherence Cxy represents a function of frequency with values between 0 and 1, and it denotes the fraction of a common variance between two-time series through a linear relation. The frequency resolution of the MTM coherence spectrum varies between ±1/(NΔt) and ±p/(NΔt), where N is the length of the time series, Δt is the sampling interval, and p is the bandwidth parameter of the tapers. In our analysis N = 1704 samples (142 yr × 12 months), Δt = 1/12 yr, and p = 2; hence the resolution of the spectrum is very high and varies between ± 0.007 and ± 0.014 cycles per year (cpy), which allows the separation of the different interannual components of interest here.

The joint circumpolar SLP and SST anomalies reveal coherent energy on the interannual time scale (Fig. 3), with periods ranging from around 4 to 6 yr (frequencies from 0.16 to 0.25 cpy). These periods are consistent with the scales of appearance of ACW. Two peaks, statistically significantly different from zero at the 99% level, are detected at 0.19 cpy (period of 5.2 yr) and 0.25 cpy (period of 4 yr), with correlation coefficients at these frequencies of r = 0.98 and r = 0.91, respectively. These peaks can be linked to the modulation of different signals, even though the breadth of the peaks suggests frequency modulation of the same signal during the 142 yr analyzed. The reconstructed spatial patterns at the two peaks (not shown) are not identical, which points to the value of considering these separately. The high-frequency resolution of the spectrum provides in detail the frequencies at which the interaction between the signals is realized, also giving useful insights into their combination.

Fig. 3.

MTM coherence spectrum between the time series of circumpolar monthly SLP and SST anomalies. The 142-yr monthly time series have been prefiltered through the SSA algorithm prior the calculation of the spectrum. The spectrum has been tested for the presence of coherent “harmonic” signals, which are significant as measured by the Thomson variance ratio test for periodic signals (F test) and through adaptive estimation procedure. Yellow, green, black, and red dashed–dotted lines are the 50%, 90%, 95%, and 99% significance levels of F test, respectively, and the resolution of the spectrum varies between ± 0.007 and ± 0.01 cpy.

Fig. 3.

MTM coherence spectrum between the time series of circumpolar monthly SLP and SST anomalies. The 142-yr monthly time series have been prefiltered through the SSA algorithm prior the calculation of the spectrum. The spectrum has been tested for the presence of coherent “harmonic” signals, which are significant as measured by the Thomson variance ratio test for periodic signals (F test) and through adaptive estimation procedure. Yellow, green, black, and red dashed–dotted lines are the 50%, 90%, 95%, and 99% significance levels of F test, respectively, and the resolution of the spectrum varies between ± 0.007 and ± 0.01 cpy.

Periodicities found in the SLP–SST coherence spectrum (Fig. 3), even though these quantities are zonally and meridionally averaged, strongly suggest associations with the presence of SAM, ENSO, and ZW3 patterns. Our hypothesis is that around these frequencies the superposition of the SAM and the SH ENSO response is apparent together with the contribution of the ZW3 variability in the background. We propose that these climate modes of variability cooperate in generating the ACWs and that their combination acts to modulate the interdecadal variations observed in the ACWs around the SO on subdecadal scales. To highlight these features, in Fig. 4 we show the power spectra for the SAM index, the SOI, and the ZW3 index. The SAM (Fig. 4a) exhibits the strongest contribution at around the 2.9-yr period (0.34 cpy), but significant energy (peak significant at the 95% level) is also found at around the 4-yr period (0.25 cpy). For this latter period, a peak in SLP–SST anomalies coherency spectrum was also found (Fig. 3). The SOI and the ZW3 index show significant coherent energy (>95% significance level) at around the 4.5-yr period (0.22 cpy) and 4.7-yr period (0.21 cpy), respectively, indicating an almost coincident frequency modulation of the signals (Figs. 4b,c).

Fig. 4.

MTM power spectra for standardized monthly anomalies of (a) the SAM index, (b) the SOI, and (c) the ZW3 index based on the 142–yr period from 1871 to 2012. Spectra have been tested for the presence of harmonic signals, which are significant as measured by the Thomson variance ratio test for periodic signals (F test) and through adaptive estimation procedure with robust red noise estimator. The null hypothesis is determined in the frequency interannual subrange from 0 to 0.8 cpy. The choice k = 3 tapers with bandwidth parameter p = 2 returns a very high resolution of the spectra varying between ± 0.007 and ± 0.01 cpy. Yellow, green, black, and red dashed–dotted lines are the 50%, 90%, 95%, and 99% significance levels of F test, respectively.

Fig. 4.

MTM power spectra for standardized monthly anomalies of (a) the SAM index, (b) the SOI, and (c) the ZW3 index based on the 142–yr period from 1871 to 2012. Spectra have been tested for the presence of harmonic signals, which are significant as measured by the Thomson variance ratio test for periodic signals (F test) and through adaptive estimation procedure with robust red noise estimator. The null hypothesis is determined in the frequency interannual subrange from 0 to 0.8 cpy. The choice k = 3 tapers with bandwidth parameter p = 2 returns a very high resolution of the spectra varying between ± 0.007 and ± 0.01 cpy. Yellow, green, black, and red dashed–dotted lines are the 50%, 90%, 95%, and 99% significance levels of F test, respectively.

We finally comment that some of the periodicities highlighted here have had a little exposure in the literature. For example, Venegas (2003) found a 5-yr period signal in the local fractional variance (LFV) spectrum for the joint SLP and SST anomalies, arguing this signal was the tropical ENSO response over the SO. Through a similar analysis on monthly SLP, SST, and SIC anomalies, White et al. (2004) found a 3.7-yr period signal dominating the LFV spectrum, with a wave-2 pattern from 1982 to 2001.

5. SLP and SST circumpolar variability

Since the significant interannual peaks detected in the coherence SLP–SST spectrum (Fig. 3) lie between the periodicities from 4 to 6 yr (0.16–0.25 cpy), the SLP and SST signals have been filtered between these two frequencies through the application of a bandpass Lanczos filter with the aim of isolating the interannual variability of interest. Figure 5 shows the time–longitude diagrams of the SLP (Fig. 5a) and SST (Fig. 5c) anomalies meridionally averaged between 56° and 66°S. It can be seen that from 1871 to 1943 there is no clear sign of ACWs in the SLP or SST anomalies, except for weak anomalies, which seem to propagate eastward during 1911–17 and 1920–31. More obvious ACW structures can be seen from about 1944 to 1971, although some phases of the waves are partially interrupted for some periods. From 1972 to about 1979, there is little evidence of eastward-propagating waves and this was also observed by Mélice et al. (2005). They found little evidence of the ACWs in the 1970s through the investigation of two SST records measured at Marion Island station in the subantarctic Indian Ocean and Gough Island station in the South Atlantic Ocean. Similarly, Simmonds (2003), using the NCEP SLP reanalysis (starting 1958) and SLP analyses of the World Meteorological Centre (starting 1972), also found no eastward-propagating features before 1981, arguing that a change in regime from standing to eastward-propagating waves took place in the early 1980s. Consistent with these findings, Bian and Lin (2012) found no strong indication of ACWs from 1973 to 1980–81 in filtered near-surface atmospheric temperature. We note that since 1980, the ACWs identified by WP96 and others are clearly visible until about 1991 (Figs. 5a,c).

Fig. 5.

Time–longitude diagrams of the interannual SLP anomalies (hPa) meridionally averaged between 56° and 66°S and bandpass filtered in time for periods ranging (a) from 4 to 6 yr and (b) from 3 to 7 yr over the 142 yr from 1871 to 2012. (c),(d) As in (a),(b), but for the interannual SST anomalies (°C). White–to–red (blue) colors indicate positive (negative) anomalies.

Fig. 5.

Time–longitude diagrams of the interannual SLP anomalies (hPa) meridionally averaged between 56° and 66°S and bandpass filtered in time for periods ranging (a) from 4 to 6 yr and (b) from 3 to 7 yr over the 142 yr from 1871 to 2012. (c),(d) As in (a),(b), but for the interannual SST anomalies (°C). White–to–red (blue) colors indicate positive (negative) anomalies.

During 1944–71, approximately five ACWs encircled the globe, and on average each ACW required about 5.6 yr (336 months/5 ACWs) to propagate around the entire SO. This result is similar to that of Bian and Lin (2012), which identified about six ACWs in the interannual near-surface air temperature during 1951–73, with an average period for each ACW of about 3.8 yr. However, distinct from Bian and Lin (2012), who analyzed bandpass-filtered anomalies retaining variability between 3 and 5 yr, our time–longitude diagrams have been calculated retaining variability between 4 and 6 yr, thus removing triennial signals in the fields. This feature also reflects the periodicity observed for each ACW, which in our case, is longer. From 1979 to 2012, only two ACWs made complete traverses of the SO with an average period of 6.5 yr (156 months/2 ACWs, during 1979–91), and, in this case, our result differs from that of Bian and Lin (2012), who detected about six ACWs from 1982 to 2010, with an average period of about 4.8 yr. This suggests that the ACWs after 1979 were longer than those observed prior to 1971. On the other hand, after 1979 the ACWs were stronger, with more pronounced amplitudes and greater continuity compared to those in previous decades. To cast light on the interpretation of the results presented above, we show the time–longitude diagrams for bandpass-filtered SLP (Fig. 5b) and SST (Fig. 5d) anomalies retaining variability between 3 and 7 yr, as done by WP96. Comparing Fig. 5a (Fig. 5c) and Fig. 5b (Fig. 5d) it can be noticed that the filtering process does not remove a considerable amount of variance and that, although there is a little change in the variability observed, the ACW cycles appear to be consistent with those described in the previous analyses.

However, after 1991, there is little hint of an ACW, except perhaps for the SST anomalies whose cycles seem to be longer (Figs. 5c,d). Therefore, not only have the periodicities of the ACWs changed during past decades, but a consistent modification in their behavior has been observed since the mid-1970s, when the circulation exhibited a change. These results are compatible with those of White and Annis (2004) and of Bian and Lin (2012). The former observed the relationship between the interdecadal variations in the ACW and the regime shift in the circumpolar westerly between 50° and 60°S, while the latter showed the multidecadal changes in the way El Niño evolved before and after 1977. Our results differ from those of Connolley (2003), which suggested that ACWs only existed during 1985–94, with no clear eastward-propagating phases beyond that period. This difference is probably due to the different filter window used for his study (3–7-yr bandpass filter) and to his choice of the latitudinal domain (56°S). We also point out that his analysis finished at 1999.

To explore further these issues, CEOF analyses have been individually performed for the interannual SLP and SST anomalies in the longitude–latitude domain of 56°–66°S. Since the aim is to detect the interannual variability, the SLP and SST anomalies have been filtered with a Lanczos bandpass filter, retaining variabilities between 4 and 6 yr. Before the analysis, the anomalies have been multiplied by the square root of the cosine of the latitude to allow the appropriate area weighting. The first three CEOFs of the SLP anomalies (hereinafter CEOFSLP1, CEOFSLP2, and CEOFSLP3) explained 45.2%, 25.9%, and 8.3% of the variance, respectively, while the first three CEOFs of the SST anomalies (CEOFSST1, CEOFSST2, and CEOFSST3) explained 37.1%, 22.7%, and 10.9% of variance.

Time series for the atmospheric HGT850 at each SH grid location (0°–90°S) have been regressed onto each of the three CEOFs for the SLP and SST anomalies, and results are shown in Fig. 6. The CEOF analysis returns real and imaginary parts, with the real part being the input data and the imaginary part representing signals that are related with the real part but are 90° out of phase (in time) with the real part. The standing oscillation of a mode is represented by the real part and the temporal coefficients are often similar to a mode of the standard EOF analysis. The additional information of the imaginary part provides the temporal and spatial phase relationships between the standing oscillation and its subsequent features. Together they provide insights into the evolution and propagation of the mode.

Fig. 6.

Spatial patterns generated by regressing the unnormalized and unfiltered 850-hPa geopotential height upon the temporal coefficients of the normalized (a)–(c) real part of CEOFSLP1, CEOFSLP2, and CEOFSLP3, respectively; (d)–(f) as in (a)–(c), but for the imaginary part; (g)–(l) as in (a)– (f), but for CEOFSST1, CEOFSST2, and CEOFSST3 of interannual SLP and SST anomalies (filtered between 4 and 6 yr) over the 142 yr from 1871 to 2012 . Regressions upon the real and imaginary parts are marked with an “.r and .i”, respectively, in the panel titles. White-to-red (blue) colors indicate positive (negative) anomalies, with contours at interval of 0.8 m per unit of the CEOF temporal coefficient for SLP field and 0.6 m per unit of the CEOF temporal coefficient for SST field. The atmospheric circulation pattern of the imaginary part is 90° out of phase (in time) with the pattern corresponding to the real part of each CEOF.

Fig. 6.

Spatial patterns generated by regressing the unnormalized and unfiltered 850-hPa geopotential height upon the temporal coefficients of the normalized (a)–(c) real part of CEOFSLP1, CEOFSLP2, and CEOFSLP3, respectively; (d)–(f) as in (a)–(c), but for the imaginary part; (g)–(l) as in (a)– (f), but for CEOFSST1, CEOFSST2, and CEOFSST3 of interannual SLP and SST anomalies (filtered between 4 and 6 yr) over the 142 yr from 1871 to 2012 . Regressions upon the real and imaginary parts are marked with an “.r and .i”, respectively, in the panel titles. White-to-red (blue) colors indicate positive (negative) anomalies, with contours at interval of 0.8 m per unit of the CEOF temporal coefficient for SLP field and 0.6 m per unit of the CEOF temporal coefficient for SST field. The atmospheric circulation pattern of the imaginary part is 90° out of phase (in time) with the pattern corresponding to the real part of each CEOF.

The anomaly pattern in HGT850 associated with the real part of CEOFSLP1 (Fig. 6a) is the dominant EOF mode of atmospheric variability discussed by Thompson and Wallace (2000) as the SAM (Simmonds and King 2004). This pattern exhibits a zonally quasi-uniform phase over the high southern latitudes, with an opposite sign in the midlatitudes. Although it is mostly zonally symmetric in character, the SAM shows a strong low pressure center in the central Pacific sector and two centers of opposite anomalies located in the Indian sector and in the western South Pacific near New Zealand, with the first being the strongest. The HGT850 regression pattern upon the imaginary part of the CEOFSLP1 (Fig. 6d) shows again the SAM with two significant differences. The first is the slight eastward shift of the center of negative regression maximum in the Pacific sector, which is associated with a general decrease of regression values over the whole of Antarctica. This feature is accompanied by a second center of negative regressions in the Indian sector, and these, in concert, resemble a wave-2 pattern. The second is the eastward shift of the center of positive regression maximum in the midlatitudes, from New Zealand to the central Pacific, which resembles the SO response to the tropical ENSO variability. The influences of the ENSO teleconnection can be discerned in the regression pattern, since the negative regressions over northern Australia and Indonesia are of the same sign as those in the central South Pacific Ocean. This pattern is very similar to that observed by Venegas (2003) for the ACW-2 signal, but she argued that the essence of the signal was to be found in the atmospheric anomalies arriving in the western Pacific from low latitudes, being probably a high-latitude manifestation of the influence of the ENSO phenomenon. The similarity between our regression pattern and the ACW-2 signal is due to the fact that the SAM and ENSO components are superimposed on the interannual scale, and their characteristics appear mixed. The SAM variability is also observed in the anomaly regression patterns for the real and the imaginary part of the CEOFSST2 (Figs. 6h,k). In this case, the anomaly pattern for the imaginary part of the CEOFSST2 (Fig. 6k) more clearly displays the ENSO character, which is consistent with a dipolar structure between the southern Pacific and the Atlantic sectors.

The anomaly pattern in HGT850 regressed upon the real part of the CEOFSLP2 (Fig. 6b) shows a discernible PSA pattern, and it is consistent with the Rossby wave response to equatorial anomalous heating (Hoskins and Karoly 1981). The pattern is similar to that found by Cai and Baines (2001) in the 500-hPa height regression pattern upon the real part of CEOFSST1, in which the SST anomalies have been filtered with a bandpass filter for period ranging from 3 to 7 yr over the 17 yr from 1981 to 1997. In our case, the ENSO mode is seen in the second CEOF for the SLP anomalies, and hence it plays a secondary role compared to the SAM observed in the first CEOF. Interesting features of variability can be observed in the HGT850 regression pattern on the imaginary part of the CEOFSLP2 (Fig. 6e): first, the development of the negative regression center in the Indian sector around 60°E, 45°–50°S, which confers a ZW3 pattern to the pressure variability, and second, the eastward movement of the positive regression center in the Pacific sector that drastically reduces its amplitude when it reaches the eastern Pacific. However, the ZW3 pattern seems to represent the lagged response (90° out of phase) of variability after which the tropical forcing becomes weak or absent and hence when the PSA pattern starts to propagate eastward. Moreover, it is important to note that a negative ZW3 variability pattern develops after a warm ENSO event (positive PSA pattern). Similar results are found in the regression patterns associated to the real and imaginary part of the CEOFSST1, according to patterns described by Cai and Baines (2001).

The HGT850 regression pattern upon the real part of the CEOFSLP3 (Fig. 6c) exhibits a clearer and distinctive ZW3 variability similar to the regression pattern on the imaginary part of the CEOFSLP2 (Fig. 6e). In this case the three centers in the midlatitudes are stronger and shifted somewhat eastward and the anomalies over Antarctica have a sign opposite to those in the midlatitudes. This feature was also noticed by MW. The anomaly pattern when HGT850 is regressed onto the imaginary part of the CEOFSLP3 (Fig. 6f) emphasizes the evolution of the ZW3 variability. After a negative ZW3 oscillation observed in the anomaly pattern in phase 0°, with the development of three negative regression centers in the Atlantic, Indian, and Pacific sectors, three positive centers develop near the Antarctic edge and migrate toward the midlatitudes in phase 90°. The anomaly patterns in the HGT850 regressed on the real and imaginary part of the CEOFSST3 (Figs. 6i,l) display similarity with the SAM and ZW3 variability.

Correlation analysis performed among the first three CEOFs of SLP anomalies and the climate indices has shown that the real part of the CEOFSLP1 is strongly related with the SAM index (r = 0.77) and negatively linked to the ZW3 index (r = −0.60). These results indicate that the SAM and ZW3 patterns exhibit the same sign of anomaly around Antarctica when oscillations of their respective indices are in opposition. The real part of the CEOFSLP2 is correlated with the SOI (r = −0.62) and more weakly with the SAM index (r = −0.30), while the real part of the CEOFSLP3 is linked to the ZW3 index (r = −0.35). Correlation analysis applied to the first three CEOFs of SST anomalies has shown that the real part of CEOFSST1 represents the ENSO component (r = −0.66 with SOI) in which modest influences of the ZW3 pattern are apparent (r = 0.35 with ZW3 index). By contrast, the real part of CEOFSST2 is linked to the SAM index (r = 0.40) as well as the real part of CEOFSST3 (r = −0.18) (all the correlation coefficients mentioned above are significant at the 99% confidence level). Results indicate that, on the interannual scale, the SAM is the leading mode in the SLP variability, and its strong relationship with the ZW3 pattern indicates that these two modes are intrinsically linked. Nevertheless, the ZW3 pattern seems to play a role in the background, since the SAM superimposes itself on it, thus masking the three anomaly centers. The ENSO component, on the other hand, significantly affects the SST variability more than the SAM does. Further, influences of the ZW3 pattern are less significant since they marginally affect the SST anomalies in combination with the ENSO component apparent in the CEOFSST1.

The correlation between CEOFSLP2 and CEOFSST1 is r = 0.66 (p < 0.01), implying that the two modes show the same variability, which is driven by ENSO. Similarly, the correlation between the CEOFSLP1 and the CEOFSST2 is r = 0.30 (p < 0.01), indicating a considerable degree of covariability of them. The cross-correlation analysis reveals that the CEOFSLP2 leads the CEOFSST1 by about 2–3 months, when the maximum correlation coefficient of r = 0.69 (p < 0.01) is found. This lag is compatible with the SST response to the PSA pattern in the SLP anomalies. The CEOFSLP1 leads the CEOFSST2 by 2–3 months (r = 0.32; p < 0.01), thus indicating the effects of the SAM in the modulation of the circumpolar SST field. The fact that the SST anomalies respond to the SAM and ENSO modes at a coincident lag indicates that the two patterns are superimposed on the interannual scale. Moreover, since the SST response to the PSA variability is stronger than that induced by the SAM, it is reasonable to conclude that the ACWs in the SST anomalies are principally modulated by the ENSO teleconnections (Cai and Baines 2001).

The spectral analysis performed on the real part of the first three CEOFs for the SLP and SST anomalies confirms the fact that the SAM and ENSO patterns are superimposed. Actually, the CEOFSLP1 and the CEOFSST1 reveal the leading frequency components at the 4.7-yr period, both significant at the 99% confidence level. Similarly, the CEOFSLP2 and the CEOFSST2 exhibit highly significant energy at the 4.9-yr period. On the other hand, the CEOFSLP3 reveals dominant energy at the 4.6-yr period as well as the CEOFSST3 does at the 4.9-yr period. These results suggest that the SLP variability shows a ZW3 pattern as a background upon which the SAM and ENSO components are superimposed. Simultaneously, the SST variability reveals the leading signature of ENSO on which the influences of the SAM pattern are superimposed. The influences of the ZW3 variability on the SST anomalies are marginal and less important than that associated with the SAM and ENSO. This indicates that the ACWs largely come from the combination between the ENSO and ZW3 patterns in the SLP anomalies, with the ENSO as the principal modulator of the SST field. Irving and Simmonds (2015) found the phase of ZW3 to be greatly influenced by the state of the SAM but not so much by ENSO. This indicates that the state of SAM could largely affect the fate of ENSO–ZW3 combination, highlighting the complexity of these associations.

Since the SLP and SST anomalies are filtered between 4 and 6 yr, a complete cycle of evolution is achieved in about 5 yr. This indicates that the time lag between the real part and the imaginary part of each CEOF corresponds to about 7.5 months (because a half cycle is about 2.5 yr). Therefore, from the regression patterns it is possible to estimate the combination between the signals. It is known that influences of warm (cold) ENSO signal, whose essential part is consistent with the simultaneous organization of a positive (negative) PSA pattern, extend over the SO during austral winter. About 7.5 months later, and when the tropical forcing linked to the previous warm (cold) ENSO event has probably started to dissipate, the SO variability exhibits three negative (positive) centers at the midlatitudes as shown in the HGT850 regression pattern upon the imaginary part of the CEOFSLP2 and CEOFSST1 (Figs. 6e,j). This suggests that the propagating component of the positive (negative) PSA pattern, observed in the HGT850 regression pattern upon the real part of the CEOFSLP2 and CEOFSST1, migrates eastward and mixes with a negative (positive) ZW3 pattern that develops in the midlatitudes. The ZW3 pattern, in turn, acts to reinforce the eastward propagation of the SLP and SST anomalies, thus giving the impression of a global ACW circuit. The evolution of the negative (positive) ZW3 pattern, captured in the HGT850 regression patterns on the CEOFSLP3 (Figs. 6c,f), suggests that the subsequent cold (warm) ENSO event overlaps the negative (positive) ZW3 pattern, supporting again the hypothesis of the combination between the signals in the Pacific sector. It is worth noting that the change of the polarity shown by the low-frequency ZW3 pattern, that is, from negative to positive (positive to negative) values, will be in agreement with the eastward-propagating component of the negative (positive) PSA pattern, thus establishing a new ACW cycle. These arguments imply that the conditions for the existence for the ACWs are met by the combination of ENSO and ZW3 components over the SO.

Figure 7 displays Hovmöller (time–longitude) diagrams of the first three reconstructed CEOFs of the interannual SLP and SST anomalies meridionally averaged between 56° and 66°S. These reconstructions are obtained by considering the real part of the standardized PCs. The CEOFSLP1 (Fig. 7a) and the CEOFSST2 (Fig. 7f) exhibit the oscillations of the zonal SAM pattern during the past decades, the strongest of which occur during 1991–2002. The CEOFSLP2 (Fig. 7b) and the CEOFSST1 (Fig. 7e) show the eastward-propagating variability linked to the ENSO oscillations, while the CEOFSLP3 (Fig. 7c) and the CEOFSST3 (Fig. 7g) exhibit a ZW3 pattern. The Hovmöller diagrams of the combination of the ZW3 and ENSO modes (Figs. 7d,h) allow the resumption of the ACWs’ propagation during the past decades. The observed ACW cycles are in agreement with those seen in the Hovmöller diagrams of SLP (Figs. 5a,b) and SST (Figs. 5c,d) previously discussed and are more clearly discernible. This supports the notion that during ACWs periods, the ENSO and ZW3 variability constructively interact. In particular, the ACWs generally appeared in correspondence to the more significant ENSO events and when the ZW3 variability exhibits significant oscillations. Moreover, it can be noticed that the ACWs are seen during periods when the interannual SAM pattern exhibits lower amplitudes or absent oscillations compared to the superimposed ENSO counterpart.

Fig. 7.

Time–longitude diagram of (a) the SLP signals (hPa) reconstructed from the CEOFSLP1, meridionally averaged between 56° and 66°S; (b),(c) as in (a), but from the CEOFSLP2 and CEOFSLP3, respectively; and (d) the sum of (b) and (c). (e),(f),(g) As in (a)–(c), but for the reconstructed SST signals (°C). (h) The sum of (e) and (g). White-to-red (blue) colors indicate positive (negative) anomalies, with contours at interval of 0.04 for (a), (b), and (d) and 0.02 for (c) in interannual SLP field. Contour is 0.07 for (e), (f), and (h), and 0.04 for (g) in interannual SST field. Color scale for (c) and (g) has been enlarged to mark the patterns.

Fig. 7.

Time–longitude diagram of (a) the SLP signals (hPa) reconstructed from the CEOFSLP1, meridionally averaged between 56° and 66°S; (b),(c) as in (a), but from the CEOFSLP2 and CEOFSLP3, respectively; and (d) the sum of (b) and (c). (e),(f),(g) As in (a)–(c), but for the reconstructed SST signals (°C). (h) The sum of (e) and (g). White-to-red (blue) colors indicate positive (negative) anomalies, with contours at interval of 0.04 for (a), (b), and (d) and 0.02 for (c) in interannual SLP field. Contour is 0.07 for (e), (f), and (h), and 0.04 for (g) in interannual SST field. Color scale for (c) and (g) has been enlarged to mark the patterns.

To assess some further quantitative aspects of the ACW, the temporal variations of the reconstructed SLP signal (Fig. 7d) in the central eastern Pacific sector centered at 120°W have been used as an indicator of the ACW cycles evolution (Fig. 8). It can be insightful to define an ACW episode as a period during which the amplitude of the SLP series exceeds the threshold of one standard deviation, and this level can be used to quantify the strength and duration of ACW episodes. It can be noticed that the ACW signal has become more stable with time and its amplitudes have strengthened reaching a maximum from the early 1980s to the early 1990s. The nature of the association between the SOI and ACW episodes can be clearly observed. This is also supported by the strong correlation (r = −0.64; p < 0.01) between the two time series.

Fig. 8.

Temporal variations of the reconstructed SLP signal in the central eastern Pacific sector centered at 120°W, coming from the sum between the real parts of CEOFSLP2 and CEOFSLP3 (blue line), and the SOI (red dashed–dotted line). ACW episodes are realized when the amplitude of the reconstructed SLP series exceeds the threshold of one standard deviation (two dark horizontal lines). Time series span the 142-yr period from 1871 to 2012.

Fig. 8.

Temporal variations of the reconstructed SLP signal in the central eastern Pacific sector centered at 120°W, coming from the sum between the real parts of CEOFSLP2 and CEOFSLP3 (blue line), and the SOI (red dashed–dotted line). ACW episodes are realized when the amplitude of the reconstructed SLP series exceeds the threshold of one standard deviation (two dark horizontal lines). Time series span the 142-yr period from 1871 to 2012.

During the decades analyzed, ACWs have shown variations in character, including speed, the amplitude and global continuity. This is because the relationships between the climate modes have changed over time, and hence the constructive or the destructive interactions between the climate patterns have arisen to various extents. This explains the appearance of only some phases of the waves in certain periods. To extend these results to the whole SO, the CEOF analysis has been performed for bandpass-filtered (4–6 yr) SLP and SST anomalies south of 20°S. Results (not shown) are similar to those discussed above, which is not surprising given that the SAM is the leading pattern on the interannual scale in the SLP anomalies and that ENSO has a similar relationship with interannual SST variability.

6. Conclusions

Interannual SLP and SST signals have been investigated with the aim of diagnosing the presence and structure of ACWs in the past, starting from 1871, and to detect interdecadal changes in the behavior of the phenomenon. In addition, variations in the climate associations across the circumpolar domain of the SO have been connected to changes in strength and speed observed in ACWs during past decades. Our results show that features of interannual variability have been observed to propagate eastward also before the mid-1950s in the form of partial ACW phases. Around the mid-1970s, an abrupt shift in circumpolar fields investigated took place. After 1979 each ACW was found to be stronger with pronounced amplitudes and better continuity compared to those in previous decades. This feature also reflects consistent modifications in periodicities associated to the phenomenon, since irregular ACW cycles have shown longer phases with the advance of decades. ACWs have been found when the constructive combination between the ZW3 and ENSO events is realized at the mid- and high latitudes in the SH. In particular, it has been observed that the positive (negative) ZW3 oscillations are in agreement with the arrival of the warm (cold) ENSO event over the SO during austral winter, thus establishing ideal conditions for the combination.

During extreme El Niño events (1972/73, 1982/83, and 1997/98), Hong et al. (2014) found a very robust SH extensive circulation consisting in a large-scale coherent SLP dipole pattern in the southern midlatitudes. They argued that the extratropical forcing from the SH anomalous circulation, peaking in boreal summer, acts as an effective booster to amplify El Niño’s growth by intensifying tropical Pacific low-level westerly winds. Our results reveal that ACWs are well matched with the South Pacific ENSO teleconnection but not necessarily with the extreme of tropical ENSO events. In particular, while the strongest ACWs of WP96 seem to be consistent with the super–El Niño event of 1982/83 over the tropical Pacific (Hong et al. 2014), the weak ACW phases detected in the early 1970s and during 1997/98 are not. Probably in these years the modulation of the South Pacific teleconnection magnitude could be been strongly affected by SAM phase (Fogt et al. 2011; Cerrone et al. 2017), whose pattern might have contrasted the propagation of ENSO anomalies over the SO, thus resulting in weaker ACW anomalies.

Acknowledgments

This study was performed in the framework of Plankton biodiversity and functioning of the Ross Sea Ecosystems in a changing Southern Ocean (P–Rose) and Multiplatform Observations and Modelling in a sector of the Antarctic Circumpolar Current (MOMA) projects as part of the Italian National Program for Research in Antarctica (PNRA). This research was also made possible by Australian Research Council Grant DP110101388 (Simmonds). Data used in this study are provided by the NOAA/OAR/ESRL/PSD, Boulder, Colorado, from their website (http://www.esrl.noaa.gov/psd/).

REFERENCES

REFERENCES
Allan
,
R. J.
,
N.
Nicholls
,
P. D.
Jones
, and
I. J.
Butterworth
,
1991
:
A further extension of the Tahiti–Darwin SOI, early SOI results and Darwin pressure
.
J. Climate
,
4
,
743
749
, doi:.
Allen
,
M. R.
, and
L. A.
Smith
,
1997
:
Optimal filtering in singular spectrum analysis
.
Phys. Lett. A
,
234
,
419
428
, doi:.
Anderson
,
J. L.
,
B.
Wyman
,
S.
Zhang
, and
T.
Hoar
,
2005
:
Assimilation of surface pressure observations using an ensemble filter in an idealized global atmospheric prediction system
.
J. Atmos. Sci.
,
62
,
2925
2938
, doi:.
Aulicino
,
G.
,
G.
Fusco
,
S.
Kern
, and
G.
Budillon
,
2014
:
Estimation of sea-ice thickness in Ross and Weddell Seas from SSM/I brightness temperatures
.
IEEE Trans. Geosci. Remote Sens.
,
52
,
4122
4140
, doi:.
Baines
,
P. G.
, and
W.
Cai
,
2000
:
Analysis of an interactive instability mechanism for the Antarctic circumpolar wave
.
J. Climate
,
13
,
1831
1844
, doi:.
Bengtsson
,
L.
,
1980
:
On the use of a time sequence of surface pressures in four-dimensional data assimilation
.
Tellus
,
32A
,
189
197
, doi:.
Bengtsson
,
L.
,
K. I.
Hodges
, and
S.
Hagemann
,
2004
:
Sensitivity of the ERA-40 reanalysis to the observing system: Determination of the global atmospheric circulation from reduced observations
.
Tellus
,
56A
,
456
471
, doi:.
Bian
,
L.
, and
X.
Lin
,
2012
:
Interdecadal change in the Antarctic circumpolar wave during 1951–2010
.
Adv. Atmos. Sci.
,
29
,
464
470
, doi:.
Bretherton
,
C. S.
,
C.
Smith
, and
J. M.
Wallace
,
1992
:
An intercomparison of methods for finding coupled patterns in climate data
.
J. Climate
,
5
,
541
560
, doi:.
Cai
,
W.
, and
P. G.
Baines
,
2001
:
Forcing the Antarctic circumpolar wave by El Niño–Southern Oscillation teleconnections
.
J. Geophys. Res.
,
106
,
9019
9038
, doi:.
Cai
,
W.
,
P. G.
Baines
, and
H. B.
Gordon
,
1999
:
Southern mid- to high-latitude variability, a zonal wavenumber-3 pattern, and the Antarctic circumpolar wave in the CSIRO coupled model
.
J. Climate
,
12
,
3087
3104
, doi:.
Carril
,
A. F.
, and
A.
Navarra
,
2001
:
Low-frequency variability of the Antarctic circumpolar wave
.
Geophys. Res. Lett.
,
28
,
4623
4626
, doi:.
Cerrone
,
D.
,
G.
Fusco
,
I.
Simmonds
,
G.
Aulicino
, and
G.
Budillon
,
2017
:
Dominant covarying climate signals over the Southern Ocean and Antarctic sea ice influence over the last three decades
.
J. Climate
,
30
,
3055
3072
, doi:.
Cherry
,
S.
,
1996
:
Singular value decomposition analysis and canonical correlation analysis
.
J. Climate
,
9
,
2003
2009
, doi:.
Cherry
,
S.
,
1997
:
Some comments on singular value decomposition analysis
.
J. Climate
,
10
,
1759
1761
, doi:.
Christoph
,
M.
,
T. P.
Barnett
, and
E.
Roeckner
,
1998
:
The Antarctic circumpolar wave in a coupled ocean–atmosphere GCM
.
J. Climate
,
11
,
1659
1672
, doi:.
Compo
,
G. P.
,
J. S.
Whitaker
, and
P. D.
Sardeshmukh
,
2006
:
Feasibility of a 100-year reanalysis using only surface pressure data
.
Bull. Amer. Meteor. Soc.
,
87
,
175
190
, doi:.
Compo
,
G. P.
, and Coauthors
,
2011
:
The Twentieth Century Reanalysis Project
.
Quart. J. Roy. Meteor. Soc.
,
137
,
1
28
, doi:.
Connolley
,
W. M.
,
2003
:
Long-term variation of the Antarctic circumpolar wave
.
J. Geophys. Res.
,
107
,
8076
, doi:.
Cotroneo
,
Y.
,
G.
Budillon
,
G.
Fusco
, and
G.
Spezie
,
2013
:
Cold core eddies and fronts of the Antarctic Circumpolar Current south of New Zealand from in situ and satellite data
.
J. Geophys. Res. Oceans
,
118
,
2653
2666
, doi:.
D’Agostino
,
R.
, and
P.
Lionello
,
2017
:
Evidence of global warming impact on the evolution of the Hadley Circulation in ECMWF centennial reanalyses
.
Climate Dyn.
,
48
,
3047
3060
, doi:.
Fischer
,
H.
,
F.
Traufetter
,
H.
Oerter
,
R.
Weller
, and
H.
Miller
,
2004
:
Prevalence of the Antarctic Circumpolar Wave over the last two millennia recorded in Dronning Maud Land ice
.
Geophys. Res. Lett.
,
31
,
L08202
, doi:.
Fogt
,
R. L.
,
D. H.
Bromwich
, and
K. M.
Hines
,
2011
:
Understanding the SAM influence on the South Pacific ENSO teleconnection
.
Climate Dyn.
,
36
,
1555
1576
, doi:.
Freitas
,
A. C. V.
,
J. S.
Frederiksen
,
J.
Whelan
,
T. J.
O’Kane
, and
T.
Ambrizzi
,
2015
:
Observed and simulated inter-decadal changes in the structure of Southern Hemisphere large-scale circulation
.
Climate Dyn.
,
45
,
2993
3017
, doi:.
Ghil
,
M.
, and Coauthors
,
2002
:
Advanced spectral methods for climatic time series
.
Rev. Geophys.
,
40
,
1003
, doi:.
Giarolla
,
E.
, and
R. P.
Matano
,
2013
:
The low-frequency variability of the Southern Ocean circulation
.
J. Climate
,
26
,
6081
6091
, doi:.
Gloersen
,
P.
, and
W. B.
White
,
2001
:
Reestablishing the circumpolar wave in sea ice around Antarctica from one winter to the next
.
J. Geophys. Res.
,
106
,
4391
4396
, doi:.
Haarsma
,
R. J.
,
F. M.
Selter
, and
J. D.
Opsteegh
,
2000
:
On the mechanism of the Antarctic circumpolar wave
.
J. Climate
,
13
,
1461
1480
, doi:.
Hassani
,
H.
,
2007
:
Singular spectrum analysis: Methodology and comparison
.
J. Data Sci.
,
5
,
239
257
.
Holland
,
P. R.
,
N.
Bruneau
,
C.
Enright
,
M.
Losch
,
N. T.
Kurtz
, and
R.
Kwok
,
2014
:
Modeled trends in Antarctic sea ice thickness
.
J. Climate
,
27
,
3784
3801
, doi:.
Hong
,
L.-C.
,
LinHo
, and
F.-F.
Jin
,
2014
:
A Southern Hemisphere booster of super El Niño
.
Geophys. Res. Lett.
,
41
,
2142
2149
, doi:.
Hoskins
,
B. J.
, and
D. J.
Karoly
,
1981
:
The steady linear response of a spherical atmosphere to thermal and orographic forcing
.
J. Atmos. Sci.
,
38
,
1179
1196
, doi:.
Irving
,
D.
, and
I.
Simmonds
,
2015
:
A novel approach to diagnosing Southern Hemisphere planetary wave activity and its influence on regional climate variability
.
J. Climate
,
28
,
9041
9057
, doi:.
Irving
,
D.
, and
I.
Simmonds
,
2016
:
A new method for identifying the Pacific–South American pattern and its influence on regional climate variability
.
J. Climate
,
29
,
6109
6125
, doi:.
Jacobs
,
G. A.
, and
J. L.
Mitchell
,
1996
:
Ocean circulation variations associated with the Antarctic circumpolar wave
.
Geophys. Res. Lett.
,
23
,
2947
2950
, doi:.
Kanamitsu
,
M.
, and
S. O.
Hwang
,
2006
:
The role of sea surface temperatures in reanalysis
.
Mon. Wea. Rev.
,
134
,
532
552
, doi:.
Karoly
,
D. J.
,
1989
:
Southern Hemisphere circulation features associated with El Niño–Southern Oscillation events
.
J. Climate
,
2
,
1239
1252
, doi:.
Können
,
G. P.
,
P. D.
Jones
,
M. H.
Kaltofen
, and
R. J.
Allan
,
1998
:
Pre-1866 extensions of the Southern Oscillation index using early Indonesian and Tahitian meteorological readings
.
J. Climate
,
11
,
2325
2339
, doi:.
Mann
,
M. E.
, and
J.
Lees
,
1996
:
Robust estimation of background noise and signal detection in climatic time series
.
Climatic Change
,
33
,
409
445
, doi:.
Marshall
,
G. J.
,
2003
:
Trends in the southern annular mode from observations and reanalyses
.
J. Climate
,
16
,
4134
4143
, doi:.
Marsland
,
S. J.
,
M.
Latif
, and
S.
Legutke
,
2003
:
Antarctic circumpolar modes in a coupled ocean–atmosphere model
.
Ocean Dyn.
,
53
,
323
331
, doi:.
Mélice
,
J.-L.
,
J. R. E.
Lutjeharms
,
H.
Goosse
,
T.
Fichefet
, and
C. J. C.
Reason
,
2005
:
Evidence for the Antarctic circumpolar wave in the sub-Antarctic during the past 50 years
.
Geophys. Res. Lett.
,
32
,
L14614
, doi:.
Mo
,
K. C.
, and
G. H.
White
,
1985
:
Teleconnections in the Southern Hemisphere
.
Mon. Wea. Rev.
,
113
,
22
37
, doi:.
Motoi
,
T.
,
A.
Kito
, and
H.
Koide
,
1998
:
Antarctic circumpolar wave in a coupled ocean–atmosphere model
.
Ann. Glaciol.
,
27
,
483
487
.
Myung
,
N. K.
,
2009
: Singular spectrum analysis. Ph.D. dissertation, University of California, Los Angeles, 64 pp.
Orsi
,
A. H.
,
T. W.
Whitworth
III
, and
W. D.
Nowlin
Jr.
,
1995
:
On the meridional extent and fronts of the Antarctic Circumpolar Current
.
Deep-Sea Res. I
,
42
,
641
673
, doi:.
Park
,
J.
,
1992
: Envelope estimation for quasi-periodic geophysical signals in noise: A multitaper approach. Statistics in the Environmental and Earth Sciences, A. T. Walden and P. Guttorp, Eds., Edward Arnold, 189–219.
Pena-Ortiz
,
C.
,
D.
Gallego
,
P.
Ribera
,
P.
Ordonez
, and
M.
Del Carmen Alvarez–Castro
,
2013
:
Observed trends in the global jet stream characteristics during the second half of the 20th century
.
J. Geophys. Res. Atmos.
,
118
,
2702
2713
, doi:.
Percival
,
D. B.
, and
A. T.
Walden
,
1993
: Spectral Analysis for Physical Applications. Cambridge University Press, 612 pp.
Peterson
,
R. G.
, and
W. B.
White
,
1998
:
Slow oceanic teleconnections linking the Antarctic circumpolar wave with the tropical El Niño–Southern Oscillation
.
J. Geophys. Res.
,
103
,
24 573
24 583
, doi:.
Polvani
,
L. M.
,
D. W.
Waugh
,
G. J. P.
Correa
, and
S.-W.
Son
,
2011
:
Stratospheric ozone depletion: The main driver of twentieth-century atmospheric circulation changes in the Southern Hemisphere
.
J. Climate
,
24
,
795
812
, doi:.
Qiu
,
B.
, and
F.-F.
Jin
,
1997
:
Antarctic circumpolar waves: An indication of ocean–atmosphere coupling in the extratropics
.
Geophys. Res. Lett.
,
24
,
2585
2588
, doi:.
Ropelewski
,
C. F.
, and
P. D.
Jones
,
1987
:
An extension of the Tahiti–Darwin Southern Oscillation index
.
Mon. Wea. Rev.
,
115
,
2161
2165
, doi:.
Saha
,
S.
, and Coauthors
,
2006
:
The NCEP Climate Forecast System
.
J. Climate
,
19
,
3483
3517
, doi:.
Simmonds
,
I.
,
2003
:
Modes of atmospheric variability over the Southern Ocean
.
J. Geophys. Res.
,
108
,
8078
, doi:.
Simmonds
,
I.
,
2015
:
Comparing and contrasting the behaviour of Arctic and Antarctic sea ice over the 35-year period 1979–2013
.
Ann. Glaciol.
,
56
,
18
28
, doi:.
Simmonds
,
I.
, and
J. C.
King
,
2004
:
Global and hemispheric climate variations affecting the Southern Ocean
.
Antarct. Sci.
,
16
,
401
413
, doi:.
Smith
,
T. M.
, and
R. W.
Reynolds
,
2003
:
Extended reconstruction of global sea surface temperatures based on COADS data (1854–1997)
.
J. Climate
,
16
,
1495
1510
, doi:.
Smith
,
T. M.
, and
R. W.
Reynolds
,
2004
:
Improved extended reconstruction of SST (1854–1997)
.
J. Climate
,
17
,
2466
2477
, doi:.
Smith
,
T. M.
,
R. W.
Reynolds
,
T. C.
Peterson
, and
J.
Lawrimore
,
2008
:
Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006)
.
J. Climate
,
21
,
2283
2296
, doi:.
Thépaut
,
J.-N.
,
2006
: Assimilating only surface pressure observations in 3D and 4DVar. Proc. ECMWF/GEO Workshop on Atmospheric Reanalysis, Reading, United Kingdom, ECMWF, 107–112.
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
, doi:.
Thomson
,
D. J.
,
1982
:
Spectrum estimation and harmonic analysis
.
Proc. IEEE
,
70
,
1055
1096
, doi:.
van Loon
,
H.
, and
J. C.
Rogers
,
1984
:
Interannual variations in the half-yearly cycle of pressure gradients and zonal wind at sea level on the Southern Hemisphere
.
Tellus
,
36A
,
76
86
, doi:.
Vautard
,
R.
,
P.
Yiou
, and
M.
Ghil
,
1992
:
Singular-spectrum analysis: A toolkit for short, noisy chaotic signals
.
Physica D
,
58
,
95
126
, doi:.
Venegas
,
S. A.
,
2003
:
The Antarctic circumpolar wave: A combination of two signals?
J. Climate
,
16
,
2509
2525
, doi:.
Whitaker
,
J. S.
, and
T. M.
Hamill
,
2002
:
Ensemble data assimilation without perturbed observations
.
Mon. Wea. Rev.
,
130
,
1913
1924
, doi:.
Whitaker
,
J. S.
,
G. P.
Compo
, and
T. M.
Hamill
,
2004
:
Reanalysis without radiosondes using ensemble data assimilation
.
Mon. Wea. Rev.
,
132
,
1190
1200
, doi:.
Whitaker
,
J. S.
,
G. P.
Compo
, and
J.-N.
Thépaut
,
2009
:
A comparison of variational and ensemble-based data assimilation systems for reanalysis of sparse observations
.
Mon. Wea. Rev.
,
137
,
1991
1999
, doi:.
White
,
W. B.
,
2000
:
Influence of the Antarctic circumpolar wave on Australian precipitation from 1958 to 1997
.
J. Climate
,
13
,
2125
2141
, doi:.
White
,
W. B.
, and
R. G.
Peterson
,
1996
:
An Antarctic circumpolar wave in surface pressure, wind, temperature and sea-ice extent
.
Nature
,
380
,
699
702
, doi:.
White
,
W. B.
, and
N. J.
Cherry
,
1999
:
Influence of the Antarctic circumpolar wave upon New Zealand temperature and precipitation during autumn–winter
.
J. Climate
,
12
,
960
976
, doi:.
White
,
W. B.
, and
D. R.
Cayan
,
2000
:
A global El Niño–Southern Oscillation wave in surface temperature and pressure and its interdecadal modulation from 1900 to 1997
.
J. Geophys. Res.
,
105
,
11 223
11 242
, doi:.
White
,
W. B.
, and
S.-C.
Chen
,
2002
:
Thermodynamic mechanism responsible for the tropospheric response to SST anomalies in the Antarctic circumpolar wave
.
J. Climate
,
15
,
2577
2596
, doi:.
White
,
W. B.
, and
J.
Annis
,
2004
:
Influence of the Antarctic circumpolar wave on El Niño and its multidecadal changes from 1950 to 2001
.
J. Geophys. Res.
,
109
,
C06019
, doi:.
White
,
W. B.
, and
I.
Simmonds
,
2006
:
Sea surface temperature–induced cyclogenesis in the Antarctic circumpolar wave
.
J. Geophys. Res.
,
111
,
C08011
, doi:.
White
,
W. B.
,
S.-C.
Chen
, and
R. G.
Peterson
,
1998
:
The Antarctic circumpolar wave: A beta effect in ocean–atmosphere coupling over the Southern Ocean
.
J. Phys. Oceanogr.
,
28
,
2345
2361
, doi:.
White
,
W. B.
,
S.-C.
Chen
,
R. J.
Allan
, and
R. C.
Stone
,
2002
:
Positive feedbacks between the Antarctic circumpolar wave and the global El Niño–Southern Oscillation wave
.
J. Geophys. Res.
,
107
,
3165
, doi:.
White
,
W. B.
,
P.
Gloersen
, and
I.
Simmonds
,
2004
:
Tropospheric response in the Antarctic circumpolar wave along the sea ice edge around Antarctica
.
J. Climate
,
17
,
2765
2779
, doi:.
Xiao
,
D.
, and
J.
Li
,
2007
:
Spatial and temporal characteristics of the decadal abrupt changes of global atmosphere-ocean system in 1970s
.
J. Geophys. Res.
,
112
,
D24S22
, doi:.

Footnotes

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).