Figure 1 shows the spatial distributions of the SH extratropical SLP ERAs related to the Niño3 or Niño4 variations for the 79–98 and 99–10 periods and in the four seasons. Significant SLP anomalies related to the Niño3 or Niño4 changes are observed in all of the four seasons, while the coupling between ENSO and SLP at high latitudes (south of 60°S) are stronger in spring and summer, which is consistent with previous studies (e.g., Fogt and Bromwich, 2006). In spring, the SLP ERA fields related to Niño3 and to Niño4 in 79–98 are similar (Figs 1d , h), and they are both characterized by two negative anomaly centers in the western Atlantic sector and the western Pacific sector, and a positive anomaly center in the eastern Pacific sector. This pattern resembles the distribution of a negative phase of PSA2. The coefficient of correlation between Niño3 and the PSA2 index is –0.59, and the coefficient of correlation between Niño4 and the PSA2 index is –0.53 (Table 1). These correlations are both significant at the confidence level of 95%. In the period 99–10 (Figs 1l , p), however, the SLP ERA field bears no resemblance to PSA2. Instead, the SLP ERA field shows a dipole structure with a positive anomaly center in the Amundsen Sea and Ross Sea, and a negative anomaly center east of New Zealand. Such structure shows similarity to the pattern of a positive phase of SAM in the Pacific sector (Zhang et al., 2018), and the correlation coefficients between Niño3 and SAM and between Niño4 and SAM are –0.33 and –0.42, respectively, which are respectively significant at the confidence level of 95% and 90%. These results support the finding from Yu et al. (2015) that there is strong relationship between SAM and ENSO when ENSO was transitioned into a CP type in the 1990s. On the other hand, a significant relationship between PSA1 and Niño3 (Table 1) in 79–98 does not appear in 99–10, which is consistent with the results from Yeo and Kim (2015) that the Rossby wave response to anomalous Equatorial heating/cooling is weaker in the CP ENSO period.

In the summer of 1979–1998 (Figs 1a ,e), the positive anomaly center in the eastern Pacific sector is still noticeable, while the negative centers in the Atlantic sector and the western Pacific sector are much weaker compared to those in spring. The SLP ERA fields exhibit a PSA-like pattern in the Pacific sector, with alternative signs from the mid to high latitudes depicting a wave train. The PSA1 index is neither significantly correlated with Niño3 nor Niño4 (Table 1), while the correlation coefficients between Niño3 and PSA2 and between Niño4 and PSA2 still reach –0.55 and –0.57 respectively, both of which are significant at the 95% confidence level. The coefficients of correlations between Niño3 and SAM and between Niño4 and SAM are –0.29 and –0.24 respectively, which are also respectively significant at the 95% and the 90% confidence level. For the summer of 99–10 (Figs 1i , m), a PSA-like pattern nearly disappears in the SLP ERA field, and neither the correlation between Niño3 and PSA2 nor between Niño4 and PSA2 is significant. The decadal variability in the ENSO-PSA2 relationship over the two periods are similar to that in spring. On the other hand, the correlation between Niño3 and SAM is –0.31 and is significant at the 90% confidence level, while that between Niño4 and SAM is not statistically significant. Ciasto et al. (2015) found that during the austral warm season, the EP SST variability is correlated with SAM and the CP SST variability exhibits no clear association with SAM. Our results indicate that in summer significant relationships exist both between SAM and the EP SST variability and between SAM and the CP SST variability during the EP ENSO period, while the association of SAM with the CP SST variability is broken down during the CP ENSO period.

Significant SLP ERAs associated with Niño4 and Niño3 are also observed in the autumn and winter of 79–98 (Figs 1f , g) and of 99–10 (Figs 1g , k), respectively. The ERA patterns show resemblance to those in spring, while the positive anomalies are mostly confined south of 60°S in the Amundsen Sea and eastern Ross Sea. For these two seasons, correlations between Niño4 and PSA2 during 79–98 and between Niño3 and PSA2 during 99–10 are both significant at the 90% confidence level, and for both periods the correlation coefficients for autumn are higher than those for winter. For autumn and winter, no significant correlations are found either between Niño3 and SAM or between Niño4 and SAM for either of the two periods. While PSA1 is also one of the dominant modes for the SH extratropical climate variability, we see in Fig. 1 that prominent PSA1-like signatures only appear in the SLP ERA field in the summer of 79–98, while statistically significant correlation between ENSO and PSA1 is only found in the winter season of 79–98 between Niño4 and PSA1 and the spring season of 79–98 between Niño3 and PSA1, with low correlation coefficients below 0.3. Therefore the results from this study suggest that PSA2 plays a more important role in the linkage of the SH subtropical climate variability to ENSO.

Comparing all the SLP ERA fields in Fig.1, it is noted the weakest SLP anomalies occur in the summer of 99–10 related to Niño4, the autumn of 79–98 related to Niño3, the autumn of 99–10 related to Niño4 and the winter of 79–98 related to Niño3, corresponding to the four periods without any significant correlations of ENSO to either of the major climate modes dominating the SH extratropical climate variability, i.e., SAM, PSA1 or PSA2 (Table 1). On the other hand, comparing the SLP ERA fields associated either with Niño3 or with Niño4 between the 79–98 and 99–10 periods, we cannot reach a conclusion that the ERAs induced by Niño3 variations are stronger in 79–98 when ENSO is dominated by the EP type, and that ERAs induced by Niño4 variations are stronger in 99–10 when ENSO is dominated by the CP type. In fact, in the EP ENSO period 79–98, during spring and summer the SLP ERAs induced by Niño3 and Niño4 variations are comparable, while during autumn and winter the SLP ERAs related to Niño4 are even stronger than those related to Niño3. In the CP ENSO period, during spring and winter there are no considerable differences between the SLP ERAs related to Niño3 and the ERAs related to Niño4, while during summer and spring the SLP anomalies related to Niño3 are even stronger than those related to Niño4. These indicate that the SH extratropical atmospheric variations associated with a specific ENSO type are not necessarily strong when this ENSO type is dominant.

Figure 2 shows the ERAs fields of the SH extratropical surface (2 m) wind related to the Niño3 or Niño4 variations for the 79–98 and 99–10 periods. The surface wind ERA distributions show a close correspondence to the SLP ERA distributions, such that a positive (negative) anomaly center in the SLP field is generally accompanied by anticyclonic (cyclonic) wind anomalies. In spring, the wind ERA field is generally characterized by westerly wind anomalies in the mid latitudes and easterly wind anomalies in the high latitudes, while noticeable anticyclonic wind anomalies are observed in the central and eastern Pacific sector corresponding to the positive SLP anomaly center (Fig. 1). It is noted that in the spring of 79–98, while the SLP anomaly field associated with Niño4 (Fig. 1h) is similar to that with Niño3 (Fig. 1d), the wind anomaly field shows noticeable differences. The anticyclonic anomaly structure in the central and eastern Pacific sector associated with Niño4 is weaker compared to those associated with Niño3, and instead it is mainly characterized by northeast wind anomalies on the northern branch of the anticyclone. The northeasterly anomalies, instead of southeasterly anomalies in the same region as in Fig. 1d, are a result of a weakened correlation between ENSO and PSA2 that results in weak SLP anomalies in the western Pacific sector between 40°S and 60°S. The difference between the ERA field associated with Niño3 and that associated with Niño4 in 79–98 also occurs in the summer, autumn and winter seasons in the same period (Figs 1e-g). In 99–10, no significant differences are observed between the wind ERA field associated with Niño3 and that associated with Niño4 in winter and spring, while in autumn the wind ERAs associated with Niño3 are strongest among all of the wind ERA fields, correlated with strong SLP ERAs. This is the period when the ENSO variation is only significantly correlated with PSA2, and the fact that there is no influence from the other major climate modes such as SAM and PSA1, and thus no offset in the climate variable anomalies may result in the strong SLP ERAs in this period. For the surface wind, same as SLP, there is no evidence that the ERAs induced by Niño3 variations are stronger in the EP ENSO dominated period, and that ERAs induced by Niño4 variations are stronger in the CP ENSO dominated period.

Figure 3 shows the ERAs fields of the SH extratropical SAT related to the Niño3 or Niño4 variations for the 79–98 and 99–10 periods. The SAT anomalies are largely determined by the surface wind anomalies. For example, in the regions dominated by anticyclonic wind anomalies, such as in the central and eastern Pacific sector in most subplots of Fig. 3, the SAT anomaly field is characterized by a negative center near the Antarctic Peninsula and a positive center in the central Pacific sector, which are respectively created by the southward advection of warm air from the lower latitudes and northward advection of cold air from the Antarctic continent by the anomalous anticyclone. In the spring of 79–98, the period when a strong relation exists between PSA2 and ENSO (Table 1), SAT exhibits the strongest ERAs associated with both of Niño3 or Niño4 in the high latitudes, and particularly in the eastern Pacific sector and the western Atlantic sector as for the SLP ERAs. Associated with the prominent negative SLP anomaly center in the western Atlantic sector in the spring of 79–98 (Figs 1d, h) and the related anomalous cyclonic winds (Figs 2d , h), which appear as a result of the strong coupling between PSA2 and ENSO, the SAT anomaly field in the Atlantic sector exhibits a dipole structure with strong negative SAT anomalies east of the Antarctic Peninsula and noticeable positive SAT anomalies in the eastern Atlantic sector. Such anomalies patterns are found to exert significant impacts on the anomaly field of the mixed layer temperature and sea ice concentration, as will be discussed below. Strong SAT anomalies also occur in the winter of 79–98 associated with Niño4 (Fig. 3g), the only season in which significant correlation between ENSO and PSA1 exists (Table 1). Another notable feature in the SAT ERA field is that in the autumn of 99–10 (Figs 3j , n), the anomaly patters are both dominated by three positive centers in the mid latitudes, indicating a wavenumber-3 structure. For the anomaly field associated with Niño3 (Fig. 3j), this structure can be explained by the close relation between Niño3 and PSA2 in this season as shown in Table 1, while for the anomaly field associated with Niño4, there are no significant correlations of Niño4 either with PSA2, PSA1 or SAM. The SLP and surface wind ERAs are also quite weak in this season, implying that the SAT anomalies are not resulted from the wind anomalies but instead from other processes. It is noted that significant anomalies associated with Niño4 are found for the mixed layer temperature (MLT) in the autumn of 99–10 (Fig. 4n) and are characterized by a wavenumber-3 structure in the mid and high latitudes, with three positive anomaly centers located exactly in the same regions as the SAT positive anomaly centers, seemingly capable of explaining the SAT anomaly pattern through a oceanic feedback to the atmosphere. However, the surface heat flux anomalies at the three positive SAT anomaly centers are positive (Fig. 5n), meaning that the net heat fluxes are from the atmosphere into the ocean and that it is the SAT anomalies modulating the MLT anomalies rather than the other way round. Therefore the SAT anomalies are created by anomalies of other atmospheric processes, and probably by cloudiness that is not included in this study.

On the other hand, significant positive SAT anomalies are observed in the autumn of 1979–1998 associated with Niño3 in the central Pacific sector (Fig. 3b), while the SLP and surface wind anomalies are quite weak and insignificant (Figs 1b, 2b). The strong positive anomalies in the MLT field (Fig. 4b) together with the weak, but significant negative anomalies in the total surface heat flux field (Fig. 5b) in this region suggest that the positive SAT anomalies are modulated by the positive MLT anomalies by the ocean-to-atmosphere heat fluxes. It is unknown that if the MLT anomalies are directly affected by the ENSO variations via an ocean teleconnection or affected by atmospheric variables like clouds or precipitation via an atmosphere teleconnection and then atmosphere-ocean interaction. As shown in Table 1, in the autumn of 79–98 there is no significant correlation of the Niño3 index to either of the dominant SH climate variation indices, i.e. SAM, PSA1 or PSA2. Therefore, it is very likely that the MLT anomalies are resulted from ocean teleconnection, the mechanism behind which demands further analysis that is beyond the scope of this study.

As mentioned above, in most cases the MLT ERA field shows a resemblance to the SAT ERA field by the atmosphere-ocean interactions, such that positive (negative) anomaly centers of SAT are normally followed by positive (negative) anomaly centers of MLT by atmosphere-ocean heat fluxes. A positive anomaly center in the central Pacific sector is a persistent feature in the MLT ERA field in both periods and in all seasons resembling the SAT anomaly pattern. A major difference from the SAT anomaly field is that the positive MLT anomaly centers in the central Pacific sector are strongest in summer and autumn, and weakest in spring when the SAT anomalies are quite strong. The decoupling between the MLT and SAT anomalies in spring are associated with the surface wind anomalies, which have a strong northward component in the eastern Pacific sector (the eastern branch of the anomalous cyclone) that transports cold water near Antarctica to the central Pacific area by Ekman flows, which are directed leftward of the anomalous winds. Therefore, the MLT anomalies are co-influenced by the SAT anomalies via air-sea heat fluxes and by surface wind anomalies via advective oceanic heat fluxes.

In the spring of 79–98 that is characterized by significant correlation between PSA2 and ENSO, similar to the SAT anomaly pattern, a dipole structure appears in the MLT anomaly field in the Atlantic sector, with significant negative anomalies near the Antarctic Peninsula and positive anomalies in the eastern Atlantic sector. This dipole structure is absent in the spring of 99–10. Such decadal variation reflects the variation in the surface heat flux anomaly field in spring (Figs 5d-h): large-scale negative and positive heat flux anomalies exist in the western and eastern Atlantic sectors respectively in 79–98, which nearly disappear in 99–10. Significant decadal variations in the MLT anomaly field also exist in autumn when the negative anomalies in the western Antarctic Peninsula region and in the western Pacific sector in 79–98 disappear in 99–10, and in winter when the MLT anomaly field turns into a more annular structure in 99–10. While there are no significant correlations between SAM and Niño3 or between SAM and Niño4 in the winter of 99–10 (Table 1), the wind anomaly fields do exhibit an annular structure in this period (Figs 2k, o), which are responsible for the annular patterns in the MLT anomaly fields by advective heat fluxes associated with wind-driven Ekman transports. Yeo and Kim (2015) observed a prevailing of warm SST anomalies in the Pacific sector east of New Zealand in spring and summer when CP ENSO dominates, which is associated with a westward shift of positive SLP anomalies in the Pacific. Such phenomenon is not observed in the MLT anomaly fields in Fig. 3, though weak negative MLT anomalies east of New Zealand in 79–98 become obscure in 99–10.

The MLD ERA field shows a good correspondence to the surface heat flux and MLT ERA fields with an anti-correlation, consistent with the results from Sallee et al. (2010) and Zhang et al. (2018) that the SH extratropical MLD anomalies are majorly controlled by the surface heat flux anomalies. Occasionally, surface wind also plays a significant role in the MLD anomalies. In the spring of 79–98, the heat flux ERA fields associated with both Niño3 and Niño4 show prominent signatures of PSA2 and are dominated by a zonal wavenumber-3 structure in the mid and high latitudes (Figs 5d,h). The MLD anomaly field is featured by a positive center in the western Atlantic sector and a negative center in the eastern Pacific sector, collocated with the negative and positive heat flux anomaly centers respectively. The absence of significant MLD anomalies in the western Pacific sector where significant heat flux anomalies exist is a result of weak, insignificant wind anomalies. The dipole structure in the MLD anomaly field disappears in the autumn of 99–10 (Figs 6l , p) when a close ENSO-PSA2 relation no longer exists. Significant decadal variations in the ERA field of MLD are also found for the autumn season, when strong negative MLD anomalies south of 40°S in the eastern Pacific sector and positive anomalies across the Pacific sector between 40°S and 20°S associated with strong negative heat flux anomalies become hardly detectable in 99–10. The latter is associated with the disappearance of the elongated surface heat flux anomalies in the mid latitudes of the Pacific Ocean in 99–10, while the former is a result of disappeared positive MLT anomalies in the eastern Pacific sector in conjunction with nil easterly wind anomalies (which will not weaken the prevailing westerlies and thus will not result in shallowed MLD) in this area.

It is also noted that while the heat flux anomalies in Fig. 5j are strongest among all of the heat flux ERA fields, which results from the strongest surface wind anomalies existing in the same period (Fig. 2j), the respective MLD anomalies (Fig. 6j) as well as the MLT anomalies (Fig. 4j) are quite weak. Besides winds and surface heat fluxes, advection and vertical mixing are the other two factors that play a significant role in the MLD variation (Sen Gupta and England, 2006). It is very likely that the strong wind anomalies have led to strong anomalous horizonal advection of heat that could cancel out the heat anomalies brought by the atmosphere-ocean heat exchange. One possible evidence for such hypothesis is that in the central Pacific sector, which is characterized by strong positive heat flux anomalies (Fig. 5j) but weak MLD and MLT anomalies (Figs 6j and 7j). This could result from the fact that the strong anomalous anticyclonic wind in this area induces westward Ekman transport on its eastern branch and eastward Ekman transport on its western branch, carrying waters from the negative MLT anomaly centers in the eastern and western Pacific sector to the positive MLT anomaly center in central Pacific and thus resulting ultimately in weak MLT and MLD anomalies in this region.