Strengthened Relationship between the Antarctic Oscillation and ENSO After the Mid-1990s during Austral Spring

The Antarctic Oscillation (AAO), the principle mode of circulation in the Southern Hemisphere (SH), is primarily characterized by a large-scale seesaw oscillation of air masses between the Antarctic and midlatitudes (Gong and Wang, 1999; Thompson and Wallace, 2000). The AAO appears year-round and features a strongly barotropic structure (Rogers and van Loon, 1982; Kidson, 1988).

It has been concluded that the AAO exerts a substantial impact on climate in both the Northern Hemisphere (NH) and SH. For example, effects have been reported on Antarctic surface air temperature and sea-ice variability (Kwok and Comiso, 2002; Liu et al., 2004); dust weather frequency in North China (Fan and Wang, 2004); cold-air anomalies and cyclone activities in East Asia (Fan, 2006; Yue and Wang, 2008); hurricane frequency over the Atlantic (Fan, 2009); and precipitation over regions including China (Xue et al., 2004; Wang and Fan, 2005), Australia (Meneghini et al., 2007), western South Africa (Reason and Rouault, 2005), and North America (Sun, 2010). Specifically, (Wang and Fan, 2007) detected that the circulation related to the negative phase of the AAO (-AAO) may provide a favorable environment for typhoon genesis over the western North Pacific. Studies have also shown that the AAO is a useful predictor of summer precipitation in East Asia, in addition to ENSO (Gao et al., 2003; Fan, 2006; Sun et al., 2009). Moreover, (Silvestri and Vera, 2003) proposed that the AAO modulates the influence of ENSO on precipitation in South America during austral spring. (Sun, 2010) discussed the decadal strengthened connection between the boreal spring AAO and the West African summer monsoon after the mid-1980s and attributed such unstable relationships to the modulation of ENSO events. Thus, it is of importance to pay more attention to the interaction between the AAO and ENSO, given the climate decadal variability.

Recent studies have revealed the interhemispherical link of the atmospheric circulation. Specifically, (Lin, 2009) found that the western North Pacific summer monsoon is associated with circulation anomalies in the high-latitude South Pacific and South America through a Rossby wave train. This connection is also presented by (Ding et al., 2016). (Ding et al., 2016) also revealed that global SST strongly accounts for the interannual variability of rotated EOF2 and EOF3 modes of seasonal mean atmospheric circulation during austral winter in the SH. ENSO has been considered the most striking interannual climate variation over the Pacific. Previous studies have suggested that the AAO and ENSO are not two independent physical processes (Ribera and Mann, 2003; Carvalho et al., 2005; Ciasto and Thompson, 2008; Liu and Xue, 2010). For example, (Karoly, 1989) noted that, during an El Niño event in austral summer, there are stable and zonally symmetric anomalies of the SH circulation with increased height at low and high latitudes, and decreased height at midlatitudes——a finding later demonstrated by results using CAM2.0 (Zhou and Yu, 2004). (L'Heureux and Thompson, 2006) revealed that both oscillations are strongly related to the zonal-mean zonal flow in the southern high latitudes. (Ding et al., 2012) found significant correlation between the AAO and tropical central Pacific SST during austral winter, and with tropical eastern Pacific SST during austral summer.

However, (Fogt and Bromwich, 2006) documented a strong decadal change in the ENSO-AAO relationship, and that a weak ENSO response in a small region of the South Pacific during the 1980s was followed by a strong teleconnection across nearly the entire South Pacific and the Amundsen-Bellingshausen Seas in the 1990s. (Li et al., 2015) found an enhanced relationship between ENSO and the AAO in January that occurred in the late 1990s. Meanwhile, (Yu et al., 2015) documented that ENSO changed from the eastern to the central Pacific type from the mid-l990s, which has increased the influence of ENSO on the AAO through both an eddy-mean flow interaction mechanism in the troposphere and a stratospheric pathway mechanism.

The atmosphere interacts with the underlying ocean and, as such, changes in atmospheric circulation are tightly associated with the anomalies in oceanic conditions. The role of SST in changes in the ENSO-AAO relationship has received little attention in previous works. Therefore, the object of this paper is to elucidate the decadal change in the relationship between ENSO and the AAO and the possible mechanisms involved from the perspective of the impact of SST. Here, we focus on austral spring because it is the most active season in terms of AAO fluctuations, when it may magnify upward into the stratosphere (Thompson and Wallace, 2000).

Following this introduction, section 2 introduces the datasets used in the study. Section 3 investigates the strengthening of the ENSO-AAO relationship and associated atmospheric and oceanic climate variabilities. Numerical simulation results are presented in section 4, and brief conclusions are given in section 5.

The monthly atmospheric reanalysis dataset employed in this research comes from ERA-Interim on a 2.5^°× 2.5^° grid for 1979-2014 (Dee et al., 2011). The variables analyzed include SLP, geopotential height, horizontal wind, vertical velocity, air temperature, and mass stream-function. The monthly SST dataset, at a resolution of 2.0^°× 2.0^°, is from ERSST.v3b, for 1854-2014 (Smith et al., 2008). The common time period is set to 1979-2014.

The Niño3.4 index is computed using the SST anomalies in the equatorial eastern Pacific (5^°S-5^°N, 170^°-120^°W) to describe the ENSO cycle (available online at http://www.cpc. ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_ change.shtml). The AAO index is defined as the first principle component of 700-hPa geopotential height anomalies poleward of 20^°S (available online at http://www.cpc.ncep. noaa.gov/\!products/\!precip/\!CWlink/\!daily_ao_index/\!aao/\!aao. shtml). In order to preserve the positive correlation between the Niño3.4 and AAO indices, the AAO index is multiplied by -1 to indicate the negative phase of the AAO (-AAO). Both indices are standardized and detrended prior to analysis. In this study, austral spring indicates the mean of the months of October-November, for the ENSO-AAO relationship in September is weak during the entire period (figure not shown). The Student’s t-test is used to detect the statistical significance. Additionally, the linear trends are eliminated before correlation and regression analysis.

The temporal variations of the -AAO and Niño3.4 indices in austral spring are presented in Fig. 1a. A strong co-variability of -AAO and ENSO can be identified after the mid-1990s (R=0.62; above the 99% confidence interval); whereas, before that, the connection between -AAO and ENSO is weak, with a correlation coefficient of 0.09. We also calculate the 13-year sliding correlation coefficients between the two indices. It is obvious that the -AAO-ENSO relationship varies with time and that significant correlations exist only after the mid-1990s. The 15- and 17-year running correlations show a consistent result (figures not shown). The results confirm that the strengthened -AAO-ENSO relationship since the mid-1990s is robust. Furthermore, (Clem and Fogt, 2013) proposed that the 1988 La Niña/-AAO event induced a decadal change in the ENSO-AAO relationship. Therefore, we further calculate the running correlations between the two indices with 1988 excluded (figure not shown), and the result suggests that the change in the relationship is less marked without 1988, but a strengthening of the relationship still exists after the mid-1990s.

Figure 1.  (a) Time series of austral spring -AAO and ENSO indices for 1979-2014, both of which are normalized and detrended. (b) The 13-year-sliding correlation coefficients between the -AAO and ENSO indices. The dashed straight lines indicate the standard for 95% and 90% significance, estimated by the Student’s t-test.

Figure 2.  The simultaneous (a, b) SLP (units: hPa) and (c, d) SST (contours; units: ^°C) / UV850 (vectors; units: m s^-1) regressions onto the austral spring -AAO during 1979-93 (left panels) and 1994-2014 (right panels). (e-h) As in (a-d) but for the austral spring Niño3.4 index. Vector winds are significant at the 90% confidence level, based on the Student’s t-test. The dark (light) shaded regions represent the 95% (90%) confidence level, respectively, as estimated using the Student’s t-test. Orange (blue) color areas indicate the significant positive (negative) values.

Based on the identified change in the -AAO-ENSO relationship after the mid-1990s, we select two periods [1979-93 (P1) and 1994-2014 (P2)] to investigate the associated atmospheric and oceanic climate variabilities. Figures 2a-d depict the linear regressions of SLP and SST/850-hPa horizontal wind (UV850) upon the -AAO index for the two periods. During P1, the -AAO corresponds to positive SLP anomalies in the Antarctic and negative SLP anomalies over the southern oceans, including the Southwest Indian Ocean, Southwest Pacific and Southwest Atlantic, accompanied by weakened circumpolar westerlies and strengthened westerlies in the midlatitudes. During P2, the -AAO signal is amplified and significant in the tropics. The negative SLP anomalies in the Southwest Pacific expand equatorward and cover most of the tropics-midlatitudes in the eastern-central Pacific, with positive SLP anomalies over the tropical eastern Indian Ocean. The negative SLP anomalies in the South Indian Ocean shift eastward and the anomalous magnitudes weaken. Compared with P1, SLP anomalies associated with -AAO are weaker in the Indian Ocean sector but stronger in the Pacific sector during P2, which implies a stronger projection of external forcing onto AAO during P2 given that the Indian Ocean sector is dominated by internal atmospheric dynamics while the Pacific sector is dominated by tropical SST anomalies (Ding et al., 2012). Consistently, anomalous westerlies prevail over the equatorial Pacific, with easterlies over the equatorial Indian Ocean. That is, the AAO is concurrent with the Southern Oscillation (SO) pattern (Walker and Bliss, 1930), which is not present during P1. The -AAO features a roughly zonal symmetry during both periods and displays the largest variance over the Amundsen Sea (the maximum local explained variance by the AAO is above 0.80 for both periods). From the oceanic perspective, during P1, the performance of -AAO is rather weak and confined to a tiny part of the southern Pacific and Atlantic. During P2, the SST anomalies in the southern oceans become spatially broader. Moreover, significant positive SST anomalies appear in the equatorial eastern-central Pacific, with negative values in the western Pacific, a pattern resembling El Niño, which is consistent with the weakened trade winds (Fig. 2d).

Figures 2e-h illustrate the patterns of linear regressions of SLP and SST/UV850 with the Niño3.4 index for the two periods. During P1, a positive phase of ENSO (i.e., El Niño) corresponds to a negative SO pattern in the tropics, positive SLP anomalies over a small portion of the Southeast Pacific, and negative values over the Weddell Sea, exhibiting a wave-train pattern from the eastern equatorial Pacific to the South Atlantic. However, this wave pattern is very weak poleward of 70^°S, and has no projection on the AAO. Anomalous westerlies occur over the equatorial Pacific and easterlies over the equatorial Indian Ocean. During P2, the southward shift of positive SLP anomalies over the Southeast Pacific moves down over the Antarctic, along with negative SLP anomalies over the midlatitude southern oceans, accompanied by a weakening of circumpolar westerlies, especially over the South Pacific. The El Niño-related circulation regimes agree well with those associated with the -AAO pattern during P2 (Figs. 2b and d). Moreover, the responses of the synchronous SST to ENSO events also show some differences between the two periods. During P1, the El Niño mode features positive SST anomalies in the equatorial eastern-central Pacific and the tropical Indian Ocean, and negative SST anomalies in the tropical western Pacific. During P2, the SST anomalies in the tropical oceans enlarge longitudinally, co-occurring with positive SST anomalies in the South Pacific.

Apart from the regression results, composite analyses are performed to verify the decadal change in the co-variability of -AAO and ENSO since the mid-1990s. The warm (cold) events are tagged based on the criterion that the austral spring Niño3.4 index exceeds 1.0 standard deviation (is less than -1.0 standard deviation), as shown in Table 1. The two-tailed Student’s t-test is used for the significance for the warm-minus-cold ENSO composites.

Figure 3.  Warm-minus-cold ENSO composites of austral spring (a, b) SLP (units: hPa) and (c, d) Z300 (units: 10 m) during 1979-93 (left panels) and 1994-2014 (right panels). The dark (light) shaded regions represent the 95% (90%) confidence level, as estimated using the Student’s t-test. Orange (blue) color areas indicate the significant positive (negative) values.

Figure 4.  Warm-minus-cold ENSO composites of austral spring (a, b) 1000-hPa air temperature (T1000; units: K), (c, d) T300 (units: K), and (e, f) U200 (units: m s^-1) during 1979-93 (left panels) and 1994-2014 (right panels). The dark (light) shaded regions represent the 95% (90%) confidence level, as estimated using the Student’s t-test. Orange (blue) color areas indicate the significant positive (negative) values.

Figure 3 illustrates the warm-minus-cold ENSO composites of austral spring SLP and 300-hPa geopotential height (Z300) for the two periods. During P1, an El Niño event is characterized by scattered atmospheric anomalies restricted within the tropics-midlatitudes, with no significant area over the southern high-latitudes. This is expected, and verifies the fact that no significant AAO pattern co-occurs with ENSO events. During P2, a -AAO pattern projects very well onto the El Niño spatial signature, with apparent positive SLP anomalies in the Antarctic and negative anomalies over the midlatitude southern oceans (Fig. 3c), which largely mirrors the regression map of SLP onto the Niño3.4 index (Fig. 2f). Furthermore, the -AAO is characterized by an equivalent barotropic component (Fig. 3d).

The features of the thermal structure associated with ENSO are presented in Fig. 4. During P1, El Niño features near-surface warming anomalies over the equatorial eastern Pacific. In the high troposphere, two warming centers occupy each side of the equatorial Pacific, with cooling anomalies over the Southwest Pacific, which enhances the thermal gradient between the tropics and midlatitudes (Figs. 4a and c). Thus, a strengthened subtropical westerly jet appears, with anomalous easterlies over the equatorial Pacific in the upper troposphere (Fig. 4e). During P2, the near-surface warming anomalies over the equatorial eastern Pacific enlarge longitudinally (the significant area is larger). Meanwhile, ENSO has an enhanced influence on air temperature in the southern high-latitudes, with a near-surface warming center over the Ross Sea. In the upper troposphere, two warming centers over the tropics are strongly prominent. The cooling anomalies over the Southwest Pacific are elongated over the midlatitudes and warming anomalies occur over the Antarctic. The thermal rearrangement increases the tropics-midlatitudes thermal gradient but decreases the Antarctic-midlatitudes thermal gradient, thus leading to the enhanced (weakened) subtropical (polar) jet (Fig. 4f). Actually, the configuration of 200-hPa zonal wind (U200) related to El Niño for P2 exhibits a meridional wave-train pattern across the southern Pacific basin, which transports ENSO-related signs southward and then impacts southern high-latitude circulations. Early studies suggested that ENSO may modulate the zonal wind anomalies over the southern extratropical regions via the eddy momentum flux (Chen et al., 1996; L'Heureux and Thompson, 2006).

To highlight the atmospheric circulation responses, we also examine the warm-minus-cold ENSO composites of vertical motion and zonal-mean mass stream-function in austral spring for the two periods (Fig. 5). During P1, in El Niño cases, an anomalous sinking movement dominates 5^°-20^°S and a tiny portion of the southern high-latitudes, with ascending motion over the equator. The existence of southern Hadley circulation is consistently detected. During P2, significant upward motion co-occurs over the equator and midlatitudes with downward movement over the low latitudes and subpolar regions in SH. Additionally, the deep southern Hadley circulation expands poleward. Notably, both the southern Ferrel and polar cells are also prominent and broaden poleward (the area surrounded by the zero line is larger). This suggests a strong coupling between the polar and tropical latitudes of the SH. ENSO may have an enhanced association with the atmospheric anomalies over the high- and midlatitudes in the SH via the poleward extent of the southern meridional circulation (Figs. 2-4).

The composite results are highly consistent with the regression analyses. The above results certify the strengthening of the linkage between -AAO and ENSO after the mid-1990s. However, a relevant question is: what accounts for the decadal change of the -AAO-ENSO relationship? To answer this, we perform EOF analysis on the regional SST (70^°S-30^°N, 120^°E-60^°W) in austral spring during the two periods. The SST EOF1 features a dominant El Niño mode in the tropical Pacific for both periods (Figs. 6a and b). Next, we examine the ENSO-related SST anomalies by regressing SST onto the normalized time series corresponding to the EOF1 mode (SST_PC1) for the two periods, respectively (Figs. 6c and d). During P1, the SST_PC1 mode is characterized by a positive SST anomaly in the equatorial central-eastern Pacific and a negative anomaly in the western Pacific (i.e., El Niño). During P2, the ENSO-related SST anomalies (at 30^°-40^°S) become broader, more poleward and eastward (farther east of 120^°W during P2 compared with P1). This may produce a more zonally symmetric meridional thermal gradient between the mid and high latitudes during P2, and then force zonal wind anomalies in the subtropical and polar regions (as shown in Fig. 4), which would affect the response in the overall AAO. Moreover, significant positive SST anomalies occupy the South Pacific, which implies that ENSO has an intensified linkage with SST anomalies in the South Pacific.

As shown in Figs. 2c and d, -AAO is tightly correlated with SST in the southern oceans for both periods, especially the South Pacific SST. ENSO has a noticeably tight connection with the South Pacific SST during P2 rather than during P1 (Figs. 2g and h). A South Pacific SST (SST_SP) index is defined as the normalized area-averaged SST in the South Pacific [(40^°-74^°S, 80^°W-180^°); rectangle in Fig. 6d]. Actually, the SST anomalies in the South Pacific have an enhanced association with ENSO after the mid-1990s. Figure 7a depicts the lead-lag correlations of the austral spring [ON(0), i.e., October-November of the current year] SST_SP index with the previous Niño3.4 index during P2. The lead-lag correlations are faint during P1, so the results are not shown. During P2, previous ENSO events have an influence on the South Pacific SST. The significant correlation of ENSO with ON(0) SST_SP occurs from May-June of the current year [MJ (0)] when ENSO leads SST_SP by about 4-5 months. Early studies showed that ENSO can excite the Pacific-South America teleconnection pattern, which then propagates into the South Pacific to induce the SST anomalies (Li et al., 2013). Additionally, the South Pacific SST leads the ON(0) AAO by about 2 months, and the synchronous correlation peaks (Fig. 7b). We also examine the lead-lag correlations of AAO with the ON(0) SST_SP index (figure not shown), and the result suggests that the correlations are not significant until October. The above results imply that previous SST anomalies in the South Pacific exert an impact on the AAO. (Ding et al., 2015) also revealed positive feedback from extratropical SH SST anomalies to the AAO, based on two ensembles of numerical experiments. This finding means that, apart from the fast teleconnection from the tropics to the southern high-latitudes via the atmosphere (Yu et al., 2015), there is another, slow, oceanic pathway. That is, ENSO may induce SST anomalies in the South Pacific, which may then affect the AAO. The strengthened ENSO-SST_SP relationship is one possible cause for the intensified relationship between ENSO and the AAO. The SST anomalies in the South Pacific persist well (figure not shown). The anomalous atmospheric circulations related to the austral spring SST_SP index are investigated in the following section.

Figure 8 presents linear regression maps of the synchronous SLP, Z300 and SST/UV850 onto the austral spring SST_SP index for the two periods. During P1, the SST_SP-associated atmospheric anomalies appear to project coherently on a -AAO pattern at both lower and upper layers, with decelerated circumpolar westerlies (Figs. 8a, c and e). Moreover, a negative SST anomaly occurs in a tiny part of the equatorial western Pacific, apart from the positive SST anomaly in the South Pacific (Fig. 8e). During P2, remarkable changes occur. In the lower troposphere, positive SLP anomalies in the Antarctic are amplified, together with positive values in the tropical regions, including the tropical Indian Ocean, Indonesia and Australia. Negative SLP anomalies in the southwestern Pacific spread equatorward, residing in the tropics-midlatitudes of the eastern-central Pacific (Fig. 8b). This implies that the high SST in the South Pacific is marked by the concurrence of negative SO and negative AAO patterns during P2. Meanwhile, the trade winds over the equatorial Pacific weaken (Fig. 8f). In the upper troposphere, positive height anomalies over the Antarctic and negative height anomalies over the South Pacific are spatially broader and quantitatively larger than those during P1, along with positive height anomalies over the tropical Pacific (Fig. 8d). Moreover, the negative SST anomalies in the equatorial western Pacific expand each side of the equator, together with positive values in the equatorial eastern-central Pacific, a mode similar to El Niño, which implies a significant relationship between ENSO and SST in the South Pacific after the mid-1990s.

Figure 5.  Latitude-pressure profiles of warm-minus-cold ENSO composites of austral spring zonal-mean omega (units: 10^-2 Pa s^-1) during (a) 1979-93 and (b) 1994-2014. (c, d) As in (a, b) but for the zonal-mean mass stream-function (units: 10¹⁰ kg s^-1). The dark (light) shaded regions represent the 95% (90%) confidence level, as estimated using the Student’s t-test. Orange (blue) color areas indicate the significant positive (negative) values.

Figure 6.  The EOF1 modes of the regional SST (70^°S-30^°N, 120^°E-60^°W) in austral spring during (a) 1979-93 and (b) 1994-2014. The explained variance of the principle mode is indicated by the percentage. (c, d) Linear regressions of the synchronous SST (units: ^°C) onto the normalized time series of the first principle component of the regional SST (SST_PC1) during (c) 1979-93 and (d) 1994-2014. The dark (light) shaded regions represent the 95% (90%) confidence level, as estimated using the Student’s t-test. Red (purple) color areas in (c, d) indicate the significant positive (negative) values.

Figure 7.  (a) Lead-lag correlations between the austral spring [ON(0)] SST_SP index and the previous Niño3.4 index. (b) Lead-lag correlations between the ON(0) -AAO index and the previous SST_SP index. The dashed straight lines indicate the standard for 90% significance, as estimated by the Student’s t-test.

Figure 8.  Regression maps of austral spring (a, b) SLP (units: hPa), (c, d) Z300 (units: m) and (e, f) SST (contours; units: ^°C)/UV850 (vectors; units: m s^-1) onto the SST_SP index during 1979-93 (left panels) and 1994-2014 (right panels). Vector winds are significant at the 90% confidence level, based on the Student’s t-test. The dark (light) shaded regions represent the 95% (90%) confidence level, as estimated using the Student’s t-test. Orange (blue) color areas indicate the significant positive (negative) values.

In this section, we report the results from conducting numerical experiments to test the reanalysis results. We use CAM4.0 (Gent et al., 2011) to test whether the enhanced ENSO-SST_SP relationship is responsible for the deepened connection between ENSO and -AAO since the mid-1990s. CAM5.0 is the atmospheric component of CESM, with finite volume dynamics and 26 hybrid sigma pressure levels. The "F_2000" component is selected, with prescribed climatological SST and sea ice and an active land model. The CO₂ concentration is set at a constant value of 367.00 ppm during the simulation. We carry out a control experiment with the model's climatological SST as the boundary forcing. Then, perturbation experiments are conducted, wherein perturbation experiment I is similar to the control run but superposed with the difference of the austral spring (September-November) SST in the Niño3.4 region (5^°S-5^°N, 170^°-120^°W) between 1994-2014 and 1979-93; and perturbation experiment II also resembles the control run but is added with the differences of the austral spring SST in the Niño3.4 and the South Pacific (40^°-74^°S, 80^°W-180^°) regions between the two periods. Each run is integrated for 20 years and the results in the last 10 years are analyzed. The average for the last 10 years is equivalent to an ensemble of 10 experiment simulations with changes in the specified SST.

Figure 9.  Differences in austral spring (a) SLP (units: hPa), (c) Z300 (units: m) and (e) U200 (units: m s^-1) between perturbation experiment I and the control run (EXP1 minus EXP0). (b, d, f) As in (a, c, e) but for the differences between perturbation experiment II and the control run (EXP2 minus EXP0). Dark (light) shading indicates significant values at the 95% (90%) confidence level, as estimated using the Student’s t-test. Orange (blue) color areas indicate the significant positive (negative) values.

The differences between experiment I and the control run (referred to as EXP1 minus EXP0) reflect the solo role of the SST in the Niño3.4 region, proving the ENSO-associated circulations before the mid-1990s. The modeled atmospheric anomalies are also confined within the tropical and midlatitude regions, as in the observations. Comparatively, the differences between experiment II and the control run (referred to as EXP2 minus EXP0) reflect the joint role of the SST in the Niño3.4 and South Pacific regions, demonstrating the related circulations after the mid-1990s. The simulated SLP exhibits the simultaneous occurrence of negative SO and negative AAO patterns (Fig. 9b). Moreover, a wavenumber-3 pattern is prominent at midlatitudes in the lower and upper layers in both the simulated and observed results, although there is a slight difference in the three centers of action. Meanwhile, the maximum variability can be observed over the West Antarctic in the simulated and observed results, especially the presence of strong positive Z300 anomalies within (60^°-90^°S, 180-80^°W). However, the modeled positive SLP anomalies within (60^°-90^°S, 180-60^°W) are situated farther west compared with the observations. A strengthened subtropical jet and a weakened polar jet exist in the upper troposphere (Fig. 9f). Furthermore, a "- + - +" pattern of zonal wind exists from the equatorial Pacific southward to the Antarctic, along with positive anomalies over northern and southern South America and negative anomalies over central South America. Generally, the model simulations are qualitatively similar to the observed circulation.

However, there are some differences in the intensity and position of the circulation pattern. The discrepancy between the model simulations and observations may be attributable to the bias between the idealized SST used in the simulation and the real SST variability. Besides, the model's capability in simulating the southern extratropical climate is very limited, which generally produces large deviations. Nevertheless, the model simulation can approximately duplicate the major features of the large-scale circulation, and partly the regional circulation features.

This paper uses both reanalysis data and numerical experiments to explore the strengthened relationship between ENSO and -AAO in austral spring after the mid-1990s. Composite and regression analyses show that their correlation is statistically insignificant during 1979-93 (P1), whereas it becomes significant during 1994-2014 (P2).

During P1, ENSO spatial signatures are restricted to the tropics-midlatitudes, whereas the -AAO signals are dominant over the Antarctic-midlatitudes. Thus, the connection between -AAO and ENSO is weak. However, during P2, marked changes occur. The ENSO-associated anomalies expand poleward. The El Niño-related SLP and geopotential height anomalies project on a -AAO pattern. The ENSO-related temperature anomalies reduce the mid-high latitudes thermal gradient but increase the tropics-midlatitudes thermal gradient, leading to an enhanced subtropical jet and a weakened polar jet. Meanwhile, the ENSO-related southern three-cell circulations extend poleward. Additionally, during P2 the ENSO-related SST anomalies (at 30^°-40^°S) become broader, more poleward and eastward. This may produce a more zonally symmetric meridional thermal gradient between the mid and high latitudes, and then force zonal wind anomalies in the subtropical and polar regions. These conditions jointly contribute to the heightened -AAO-ENSO relationship since the mid-1990s. Numerical simulation reveals that the deepened connection between ENSO and SST_SP since the mid-1990s is one cause for the strengthened -AAO-ENSO relationship. However, there is still one interesting question that remains: why does the relationship between the previous ENSO and the South Pacific SST strengthen after the mid-1990s? This issue is of great interest and needs further investigation in future work.