Parallel Comparison of the 1982/83, 1997/98 and 2015/16 Super El Niños and Their Effects on the Extratropical Stratosphere

El Niño-Southern Oscillation (ENSO) is a coupled ocean-atmosphere phenomenon that recurrently occurs every two to seven years. It not only dominates tropical climate variations but also exerts a broad influence on the extratropics of both hemispheres. Due to the complexity and diversity of ENSO, researchers have classified ENSO into several types to study its underlying mechanism and impacts. For example, based on its magnitude, ENSO can be classified into moderate and strong events (Wolter and Timlin, 1998; Stephens et al., 2007; Rao and Ren, 2016b, c); based on its location and zonal structure, it can be classified into Eastern Pacific and Central Pacific events (Ashok et al., 2007; Kao and Yu, 2009; Kug et al., 2009; Rao and Ren, 2014); based on its seasonal evolution, it can be classified into early- and late-onset events (Xu and Chan, 2001; Horii and Hanawa, 2004); and finally, based on its period and frequency, it can be classified into quasi-biennial and quasi-quadrennial events (Rasmusson et al., 1990; Ren et al., 2017).

Accompanying the initiation and development of ENSO, the tropical Walker and Hadley circulations deviate from their normal states and exert a substantial influence on the tropical Indian and Atlantic oceans through the so-called "atmospheric bridge" process. Previous studies have proved the existence of significant lead-lag correlations between the SST anomalies over the tropical eastern Pacific in winter and those over the tropical Indian Ocean (TIO; e.g., Nicholson, 1997; Klein et al., 1999; Ding and Li, 2012) and tropical Atlantic (e.g., Enfield and Mayer, 1997; Klein et al., 1999; Alexander et al., 2002) throughout winter and the subsequent spring. Specifically, as ENSO develops, the remote response in the tropical Indian and Atlantic oceans produces anomalous surface winds and heat fluxes into the ocean, which can then induce SST anomalies in those regions (Alexander et al., 2002).

The remote impacts of ENSO on the extratropics are mainly realized via several teleconnections. In the mid-to-upper troposphere, for example, the Pacific-North America (PNA) pattern can link the anomalous latent heating induced by the ENSO SST anomalies over the tropical eastern Pacific to the atmospheric variability in the extratropics (Jin and Hoskins, 1995; Newman and Sardeshmukh, 1998; Annamalai et al., 2007). Actually, it has been reported that the PNA pattern is also modulated by the anomalous diabatic heating over other tropical oceans, particularly over the Indian Ocean (Barsugli and Sardeshmukh, 2002; Annamalai et al., 2007; Rao and Ren, 2016a). Both linear model experiments forced with idealized tropical diabatic heating profiles (e.g., Newman and Sardeshmukh, 1998; Annamalai et al., 2007) and general circulation model experiments forced with idealized SST anomalies (Kumar and Hoerling, 1998; Farrara et al., 2000; Barsugli and Sardeshmukh, 2002; Spencer et al., 2004; Rao and Ren, 2016a) show that the tropospheric circulation response pattern induced by the imposed warm forcing over the Indian Ocean tends to destructively interfere with that induced by the warm forcing over the tropical eastern Pacific. As (Barsugli and Sardeshmukh, 2002) suggested, a nodal line exists near 110^°E for the sensitivity of the PNA response to the tropical warm forcing, across which the polarity of the PNA response seems to reverse. Therefore, the warm SST anomalies over the eastern equatorial Pacific during El Niño compete with the El Niño-related warm SST anomalies over the TIO in modulating the PNA pattern.

Since the PNA pattern is a key bridge linking the tropical SST forcing to extratropical stratospheric variabilities, ENSO is known to be an important factor of influence on the interannual variation of the stratospheric polar vortex in the winter season (e.g., Garfinkel and Hartmann, 2007; Wei et al., 2007; Ren et al., 2012; Xie et al., 2012; Rao and Ren, 2016a, b, c). By modulating the planetary wavenumber-1 from the troposphere to the stratosphere, El Niño can lead to an anomalously weaker and warmer polar vortex, which in turn is coupled with the circulation and climate anomalies in the troposphere (Ren and Cai, 2006; Ren and Hu, 2014; Yu et al., 2014, 2015; Rao et al., 2015; Cai et al., 2016). Stratospheric anomalies have been recognized as important indicators for circulation and climate anomalies in the troposphere in various respects (Ren and Hu, 2014; Yu et al., 2014, 2015; Hu et al., 2015; Rao et al., 2015).

Although a super El Niño (maximum SST anomaly exceeding 3 K in Niño regions) only occurs once in every one or two decades, its climatic influences are known to be globally much stronger than those of an ordinary El Niño (Zhai et al., 2016). For example, the heavy rainfall and frequent floods in China caused by the latest super El Niño in 2015/16, have yielded more than 360 human deaths and up to RMB 45.51 billion of direct economic losses. Therefore, an accurate understanding of super El Niños and their effects on climate is urgently needed to reduce these huge impacts. One important aspect of their influence is the possible effect of super El Niños on the extratropical stratosphere, which may benefit our understanding and prediction of the climate anomalies in super El Niño winters. Accordingly, the present study examines the responses of the northern winter stratospheric polar vortex to three historic super El Niño events and their relationship with extratropical tropospheric anomalies. A parallel comparison of their SST patterns, seasonal evolutions, and their effects in the extratropics is performed.

The remainder of the paper is organized as follows: In section 2, we introduce the data and methods employed in the study. Section 3 presents a parallel comparison of the SST patterns and their seasonal evolutions for the three super El Niño events. Section 4 compares their effects in the extratropical stratosphere and provides parallel analyses of the influencing processes. A summary and discussion is provided in section 5.

The SST analysis datasets used in this study include NOAA's Extended Reconstructed SST (ERSST) dataset, version 4 (Smith and Reynolds, 2003), the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) dataset (Rayner et al., 2003), and the Japan Meteorological Agency's Centennial In Situ Observation-Based Estimates (COBE) dataset (Ishii et al., 2005). Their horizontal resolutions are 2^°× 2^°, 1^°× 1^°, and 2^°× 2^°, on a global uniform latitude-longitude grid, respectively. In general, the three super ENSO evolutions are quite consistent among the three SST datasets. The climatology in each dataset is calculated based on the period 1981-2010, and the annual cycle of the global SST is obtained. The annual cycle is removed from the original data to compute the SST anomalies, and the long-term trend of the SST anomalies in each calendar month is removed.

To better capture the evolution of the ENSO events, several indices——namely, the Niño3, Niño4, and Niño3.4 indices——are also adopted in this study. We calculate the monthly Niño3 index by averaging the SST anomalies over the Niño3 region (5^°S-5^°N, 150^°-90^°W). Similarly, the monthly Niño4 and Niño3.4 indices are also calculated by averaging the SST anomalies over the Niño4 (5^°S-5^°N, 160^°E-150^°W) and Niño3.4 (5^°S-5^°N, 170^°-120^°W) regions. In fact, Niño3 is the best index to capture the super El Niños, since the maximum SST anomalies are just located in the Niño3 region, where the thermocline is much shallower than that in the Niño4 region.

Figure 1.  SST anomalies (units: K) during the three super El Niños [1982/83 (left), 1997/98 (middle) and 2015/16 (right)] winters (December-February), from (a-c) COBE, (d-f) ERSST, and (g-i) HadISST.

The oceanic analysis used here is provided by the Global Ocean Data Assimilation System (GODAS) at the National Centers for Environmental Prediction (NCEP), which gives an estimate of the oceanic state to initialize the coupled climate forecast system. GODAS is based on a quasi-global configuration of the Geophysical Fluid Dynamics Laboratory's Modular Ocean Model, version 3. The model domain extends from 75^°S to 65^°N, and it has a horizontal resolution of 1^° (longitude) × (1/3)^° (latitude). The model has 40 levels, with a 10-m resolution in the uppermost 220 m. Forced by the momentum flux, heat flux, and freshwater flux from the NCEP-DOE (DOE: Department of Energy) atmospheric Reanalysis-2 (Kanamitsu et al., 2002), GODAS assimilates temperature profiles from expendable bathythermographs, tropical atmosphere-ocean moorings, and Argo profiling floats (Saha et al., 2006). More information about GODAS can be found at http://www.esrl.noaa.gov/psd/data/gridded/data.godas.html.

Since the GODAS data are forced by the NCEP-DOE reanalysis, we also use this reanalysis dataset in our case study to extract the atmosphere circulation anomalies during the super ENSOs. In this way, the possibility of a mismatch between oceanic and atmospheric conditions is greatly diminished. We also use the NCEP-NCAR (NCAR: National Center for Atmospheric Research) Reanalysis-1 (Kalnay et al., 1996), and find that the result shows little sensitivity to the choice of reanalysis dataset. In the NCEP-DOE reanalysis data, the 3D atmospheric variables, such as air temperature, winds, and geopotential height, as well as several surface variables, such as mean sea level pressure and precipitation, have a post-processed horizontal resolution of 2.5^°× 2.5^°, extending from the North to the South Pole. In the vertical direction, this operational model uses a sigma coordinate, and outputs are post-processed at 17 standard pressure levels for 3D variables from 1000-10 hPa by the NCEP-DOE releasers. More information about the NCEP-DOE reanalysis can be found at http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis2.html.

Figure 1 shows the winter SST anomaly patterns during the mature phase of the three super El Niños, from different analyses. As the super El Niños peak in the boreal winter, the maximum SST anomalies mainly appear in the Niño3 region and the cold tongue area. This is especially true for the first two super El Niños (Figs. 1a, d and g; Figs. 1b, e and h). However, the 2015/16 super El Niño has some differences from the 1982/83 and 1997/98 events, in that the anomalous SST center is just situated in the Niño3 area and shows little eastward extension to the cold tongue region (Figs. 1c, f and i). In addition, basin-wide warm SST anomalies over the TIO during the mature ENSO phase are clearly seen, while the SST anomalies over the tropical Atlantic are relatively weaker. The general picture obtained from the three SST datasets is quite consistent, even though ERSST has a coarser resolution than the other two datasets.

Figure 2.  As in Fig. 1, but for the seasonal evolutions of equatorial SST anomalies (units: K) from the leading summer [July(-1)] to the second following summer [July(2)], relative to the mature phase of the three super El Niños. The white plus sign in each panel denotes the timing and longitude of the maximum SST anomaly. The horizontal lines mark the concurrent winter relative to the mature phase of the super El Niños (January of year 1).

The seasonal evolution of the equatorial SST anomalies from the summer of year(-1) to the summer of year(2) is shown in Fig. 2 for the three events. Note that we use the numbers -1, 0, 1 and 2, in parentheses, to denote the previous, concurrent, decaying, and following years, respectively, relative to the mature-phase year of the super El Niño, which peaks in December(0)-January(1)-February(1). The growth of the super El Niños is quite asymmetric between the development and decay phases. Also, the warm SST anomalies are formed around 18 months before the mature phase, in July(-1), although the eastern equatorial Pacific is dominated by cooling anomalies in the 1982/83 and 1997/98 cases (Figs. 2a, d and g; Figs. 2b, e and h). The warm SST anomalies remain in the west equatorial Pacific until the ENSO developing spring [April(0)], and then the warm SST anomalies propagate eastwards gradually and develop into a mature super El Niño in the following cold season. Compared with the 1982/83 case, which matures in December(0) (Figs. 2a, d and g), the 1997/98 super El Niño seems to peak one month earlier, in November(0) (Figs. 2b, e and h). However, the super El Niño then decays rather rapidly; from early summer in year(1), the SST anomalies change in sign from positive to negative, and develop towards the negative phase. The COBE, ERSST and HadISST datasets capture quite consistent features for the historical super El Niños in 1982/83 and 1997/98. The 2015/16 super El Niño also ceases in the early summer (Figs. 2c, f and i), and is expected to develop into a mature La Niña in the following winter if it grows in a similar way as the 1982/83 and 1997/98 cases. There are also several differences between the latest super El Niño and the 1982/83 or 1997/98 cases, in that the warm SST develops in the whole equatorial Pacific from the summer of year(-1), and the warm center is located a little farther westward in the peak winter, which is quite consistent with Figs. 1c, f and i.

Figure 3.  Equatorial ocean potential temperature anomalies (units: K) from GODAS in the (a-c) previous winter, (d-f) leading summer, (g-i) concurrent winter, and (j-l) following summer, relative to the mature phase of the three super El Niños [1982/83 (left), 1997/98 (middle) and 2015/16 (right)]. The vertical lines mark the western (150^°W) and eastern (90^°W) boundaries of the Niño3 region.

To further certify the differences between the 2015/16 super El Niño and the two 20th century cases, we show in Fig. 3 the equatorial ocean potential temperature anomalies from GODAS during the ENSO previous winter, developing summer, mature winter, and decaying summer. Different from the two former cases, which show obvious eastward propagation (leftmost and middle columns in Fig. 3), the 2015/16 super El Niño shows little zonal propagation (rightmost column in Fig. 3). Specifically, the warm potential temperature anomalies are formed, especially at the depths of 50-250 m in the western equatorial Pacific, from the previous winter in the former two cases (Figs. 3a and b), and propagate eastwards gradually into the Niño3 region, with the maximum center at a depth of ∼100 m, in the developing summer (Figs. 3d and e). In the mature phase, the potential temperature anomalies peak at the eastern boundary of the Niño3 region, and large cooling anomalies (approximately -4 K) form in the western equatorial Pacific (Figs. 3g and h). The cold anomalies seem to follow the eastward march of the warm anomalies, and are transported into the Niño3 region in the decaying summer, which might be responsible for the sudden end to the super El Niño and the initiation of a follow-up La Niña (Figs. 3j and k).

Different from the 1982/83 and 1997/98 El Niños, the 2015/16 super El Niño shows large potential temperature anomalies remaining stationary near the Niño3 region (Figs. 3c, f and i), although the potential temperature anomalies strengthen gradually. We can see that the cooling anomalies accompany the formation of warm anomalies in the Niño3 region and remain in the western equatorial Pacific at depths shallower than 250 m (Fig. 3c). However, the cold anomalies extend rapidly into the Niño3 region in the decaying ENSO summer (Fig. 3l). Thus, we also expect to see a mature follow-up La Niña, based on this oceanic mechanism.

In fact, a super El Niño cannot impact the atmospheric circulation directly, especially in the upper troposphere. The anomalous release of latent heating induced by convective activity provides an energy source for the circulation adjustment. According to this route, Fig. 4 presents the tropical rainfall anomalies and the circulation anomalies in the upper troposphere during different stages of the three super El Niños. During the previous winter, quite large circulation anomalies appear in the mid-high latitudes, and meanwhile rainfall anomalies are quite scattered and distributed irregularly, inconsistent among the three super El Niños (Figs. 4a-c). The dipole height anomalies along the dateline in the meridional direction during the previous winter, somewhat like a North Pacific Oscillation (NPO) distribution, are very consistent among the three super El Niños (Figs. 4a-c). The anomalous westerly south of the negative lobe of the NPO-like height anomalies may be favorable for the initiation of an El Niño event, via the Bjerknes feedback mechanism, if it can touch the near-surface (nearly barotropic for the negative lobe; not shown). The initiation of an ENSO event via extratropical forcing has been noted previously by (Vimont et al., 2001), who proposed that the NPO anomalies can remain in the ocean and initiate an ENSO event via the so-called footprint mechanism.

Figure 4.  As in Fig. 3, but for precipitation (color shading; units: mm d^-1) and 200-hPa geopotential height [contours (solid, positive; bold, zero; dashed, negative); units: m; interval: 25 m] anomalies.

Since warm SST anomalies continue growing in JJA(0) (JJA: June-July-August), and the forcing is still not strong, the atmospheric circulation anomalies in the developing summer are small and inconsistent among the three super El Niños (Figs. 4d-f). Positive rainfall anomalies are consistent among the three super events and are organized in the central and eastern equatorial Pacific (Figs. 4d-f).

The strongest impacts of ENSO on the atmosphere mainly appear in the concurrent mature winter and the decaying season (Figs. 4g-i). As the SST anomalies reach maxima (Figs. 1 and 2), an increase in rainfall and enhanced convection is observed in the central and eastern equatorial Pacific, which corresponds to a release of anomalous latent heating (Figs. 4g-i). Although the SST anomalies are centered in the Niño3 region, the maximum positive rainfall anomalies are located a little farther west. This pattern is mainly caused by the nonlinear SST-rainfall relationship, in which SST anomalies in the warm pool region more readily induce a rainfall response, while the same SST anomalies in the cold tongue region are less efficient (Hoerling et al., 1997; Rao and Ren, 2016b, c). With the diabatic heating released in the central equatorial Pacific, a couple of anomalous high centers are excited on both sides of the equator (Gill, 1980). In the Northern Hemisphere, the tropical high center is accompanied by an anomalous low center in the North Pacific, another anomalous high center over continental Canada (Figs. 4g-i), and another anomalous low center in the southeastern United States although weak in Fig. 4i, as different lobes of the PNA teleconnection. The PNA pattern is an important ENSO teleconnection in the troposphere, which connects the tropical diabatic heating with the extratropical circulation. The PNA pattern is mainly projected to the zonal planetary waves, modulating the extratropical stratospheric circulation. The upper troposphere height anomalies have relatively large inter-event variation, which may be associated with the SST pattern differences. In the 1982/83 and 2015/16 mature winters, the basin-wide warm SST anomalies appear in the TIO (Figs. 1a, d and g; Figs. 1c, f and i), together with scattered positive rainfall (or latent heating) anomalies in that region (Figs. 4g and i). However, basin-wide warm SST anomalies do not form in the TIO in the 1997/98 mature winter (Figs. 4b, e and h), and the TIO is mainly covered with negative rainfall (latent heating) anomalies (Fig. 4h). The TIO latent heating is also associated with the PNA anomalies, but its sign is opposite to that induced by the tropical Pacific forcing (see section 1). It is obvious that the 1997/98 PNA response is strongest among the three super El Niño events, partially due to the different behaviors of the TIO rainfall.

In the decaying summer, the upper tropospheric circulation anomalies are quite weak, except that the South Asian high seems to be intensified, since positive latent heating (rainfall) appears over the TIO in this season (Figs. 4j-l).

The stratospheric variation mainly lies in the northern winter and is characterized by changes in the intensity and position of the stratospheric polar vortex. It has been shown that the stratospheric polar vortex is anomalously weaker and warmer during warm ENSO mature winters, associated with an increase in the extratropical planetary wave propagating upwards into the stratosphere (Garfinkel and Hartmann, 2007; Rao and Ren, 2016a, b, c; Ren et al., 2017). Figure 5 shows the seasonal evolution of the polar cap temperature and circumpolar westerly anomalies from 1000 to 10 hPa. It is indeed true that the polar night jet during some winter months of the mature super El Niños is weakened, and the polar cap is anomalously warmer, but the negative zonal wind and positive temperature anomalies appear in different calendar months. During the 1982/83 and 2015/16 cases, positive zonal wind anomalies in the circumpolar region, and negative temperature anomalies in the Arctic, appear in early winter and persist from October(0) to late-January(1) (Figs. 5a and c). From early-February(1), negative zonal wind and positive temperature anomalies begin to develop in the upper stratosphere and propagate downwards gradually into the upper troposphere. However, the warm temperature anomalies in the Arctic and the anomalous easterly in the circumpolar region appear much earlier during the 1997/98 super El Niño than during the other two cases (Fig. 5b). The Artic is mainly controlled by warm anomalies throughout the 1997/98 winter, together with circumpolar easterly anomalies.

Figure 5.  Pressure-time evolution of polar cap (60^°-90^°N) temperature anomalies (color shading; units: K) and circumpolar (65^°-75^°N) westerly anomalies (contours; units: m; interval: 1 m s^-1), for the (a) 1982/83, (b) 1997/98 and (c) 2015/16 super El Niños. The vertical lines mark the concurrent winter relative to the mature phase of the super El Niños (January of year 1).

All three super El Niño winters witness a warmer stratospheric polar vortex, although the timing of the warming is different among the three cases. In fact, the interannual variability of the stratospheric circulation is also controlled by other factors, including the thermal state in the TIO (Fig. 6a), the solar cycle (Fig. 6b; ftp://ftp.ngdc.noaa.gov/STP/space-weather/solar-data/solar-features/), and the stratospheric Quasi-Biennial Oscillation (QBO) in the tropics (Figs. 6c and d). Thus, we need to filter out contributions from these factors and check the residual anomalies through multivariate (partial) regression analysis. We do not use tropical Indian SST indices, since the SST anomalies in that region are highly correlated with ENSO (Ren et al., 2012, 2016; Rao and Ren, 2016a). Instead, the TIO rainfall is a good indicator for the diabatic heating anomalies associated with the TIO thermal state (Fig. 6a). The possible effects of the solar cycle (Fig. 6b) on the stratosphere have been documented widely (Kodera and Kuroda, 2002; Camp and Tung, 2007). As suggested in (Wallace et al., 1993), the QBO can be well represented by the two EOF modes of the tropical equatorial zonal wind (Figs. 6c and d).

Figure 6.  Monthly time series of (a) the area mean precipitation over the tropical Indian Ocean basin (30^°S-30^°N, 40^°-110^°E), (b) the solar cycle index [F10.7, the solar radio flux at 10.7 cm (2800 MHz)], and (c) the two QBO modes (QBO1 in blue and QBO2 in red) that are obtained from the leading and second EOFs of the equatorial (5^°S-5^°N) zonal wind anomalies in 100-10 hPa. The spatial patterns of QBO1 and QBO2 are shown in (d).

Figure 7.  Multivariate regression coefficients against (a, b) QBO1, (c, d) QBO2, (e, f) the solar cycle index, and (g, h) the IOB (Indian Ocean basin) precipitation index, of the circumpolar zonal wind (left; 65^°-75^°N; color shading; units: m s^-1) and polar cap temperature (right; 60^°-90^°N; color shading; units: K) anomalies in each calendar month from the leading summer [July(0)] to the following summer [June(1)]. Contours mark the regression coefficients at the 90% (gray) and 95% (black) confidence levels. Poisitive (negative) regression coefficients that are significant at 90% and 95% confidence levels are marked by solid (dashed) lines. The vertical lines mark the midwinter (January of year 1).

Different from the high correlation between the TIO SSTs and ENSO, the sign of the TIO precipitation index is more uncertain, even under the three super El Niños. Specifically, positive precipitation anomalies appear in the TIO basin during the 1982/83 winter, while negative precipitation anomalies dominate during the 1997/98 and 2015/16 winters (Fig. 6a). The three super El Niños also take place under different phases of solar activity (Fig. 6b), although the solar interference within the stratospheric response to ENSO is not particularly large. It is clear that the 1982/83 El Niño appears after a minimum solar flux, and the 1997/98 event appears after a maximum solar flux, while the 2015/16 event appears just before a future maximum solar flux. It can be easily identified from Figs. 6c and d that the 1982/83 and 2015/16 winters are just in the westerly phase of the QBO, while the 1997/98 winter is in a QBO transition phase from westerly to easterly. A westerly phase of the QBO favors a cold and strong stratospheric polar vortex, which will cancel part of the warm ENSO signal from a linear assumption. It is expected that the TIO, solar cycle and QBO signal components can be linearly superposed, and a partial regression computation will decompose them, although the nonlinear reaction between any two or all of those factors may exist, which is usually hard to precisely assess.

Partial regression patterns of the circumpolar zonal wind and polar cap temperature anomalies for each factor are displayed in Fig. 7. It is obvious from Figs. 7a-d that the QBOs have quite large impacts on the Arctic by adjusting the meridional position (latitude) of the critical line (U=0), across which planetary waves cannot propagate (Holton and Tan, 1980). Although the climate system obtains extra energy during the solar maximum period (Figs. 7e and f), the circumpolar westerly is somewhat intensified (1 m s^-1), associated with a positive Northern Annular Mode-like response in the early winter (Matthes et al., 2006). The impacts of the TIO diabatic heating on the northern extratropical stratosphere have been reported by (Rao and Ren, 2016a) via numerical experiments, and are verified again here through observational diagnosis (Figs. 7g and h). They reported that a warm TIO is favorable for an intense and cold stratospheric polar vortex response. The unstable relationship between super El Niño and the stratospheric polar vortex might be affected by those factors.

After the partial regression signal is filtered out from the original circulation anomalies for each factor one by one, we display the temporal evolution of the extratropical circulation anomalies in Fig. 8. Although the interferences of the QBO, solar cycle, and TIO diabatic heating are not strictly linear, warm temperature anomalies in the stratospheric polar cap region and the anomalous easterly in the circumpolar region does indeed become larger from Fig. 5 to Fig. 8. The cold anomalies and westerly anomalies appearing in the early winter of the 1982/93 and 2015/16 super El Niños also become smaller in the residual data (Figs. 8a and c) than that in the original data (Figs. 5a and c). In short, the super El Niños have strong impacts on the extratropical stratosphere in the cold season if the interferences from other controlling factors are filtered out, although it is impossible to completely exclude them due to their nonlinear interactions.

Figure 8.  As in Fig. 5, but with the reconstructed monthly fields (products of the multivariate regression coefficients shown in Fig. 7 and the respective time series all removed from the original data fields).

Ren et al. (2012, 2016) indicated that ENSO not only has a large influence on the stratosphere in the concurrent winter, but also in the following winter. The delayed impact of ENSO on the stratosphere is mainly achieved via the delayed midlatitude stratospheric warming in the decaying warm season of ENSO, and the following transportation into the Arctic by the active wavenumber-2 in the subsequent winter. We do not present the dynamic analysis again, but give the distributional pattern of the zonal-mean temperature and zonal wind anomalies in Fig. 9. Note that the three consecutive winter months selected for the 1997/98 super El Niño are different from those selected for the other two events, since the 1997/98 event matures relatively earlier than the other two super El Niños (see the white plus sign in Fig. 2). As the maximum SST forcing is reached in different months for the three events, we choose the February-March-April mean for the 1982/83 and 2015/16 events, but the December-January-February mean for the 1997/98 case. It can be seen that, as the ENSO forcing reaches a maximum in the concurrent winter, the subtropical jet (∼ 25^°N; 200 hPa) is enhanced via thermal wind theorem, since the meridional temperature gradient is enlarged (Figs. 9a-c). However, in the extratropics, the maximum anomalies are in the stratosphere. The circulation patterns are quite consistent among the three super El Niños, although the circulation patterns are given during different months for the three super El Niños.

Figure 9.  Pressure-latitude distribution of zonal mean temperature (color shading; units: K) and zonal wind (contours; units: m; interval: 1 m s^-1) anomalies in the (a-c) concurrent winter and (d-e) following winter, relative to the mature phases of the 1982/83 (left), 1997/98 (middle) and 2015/16 (right) super El Niños. Note that the months selected to represent winter are slightly different for the three super El Niños.

In the next winter, relative to the ENSO mature phase for the 1982/83 and 2015/16 cases (Figs. 9d and e), a similar anomaly pattern reappears in the extratropical stratosphere, except that the maximum temperature anomalies lie at a lower level (below 50 hPa) than in the concurrent winter (above 10 hPa). The delayed response of the northern winter stratospheric polar vortex to ENSO is connected to the midlatitude troposphere-stratosphere coupling process, as reported by Ren et al. (2012, 2016). In the decaying summer of ENSO, the extratropical troposphere is covered with anomalously cold and persistent signals in the upper troposphere, which are coupled by anomalously warm and persisting signals in the midlatitude stratosphere. The combination of climatological planetary waves since the decaying autumn of ENSO and the warm temperature anomalies in the midlatitude stratosphere favors the poleward transport of the eddy heat flux, and therefore an anomalously warm and weak stratospheric polar vortex in the following winter. But what will happen in the winter of 2016/17 (the next winter relative to the 2015/16 ENSO mature phase)? Will the stratospheric polar vortex be weaker and warmer, just as in Figs. 9d and e? From just two historical records, this is hard to predict, because the stratosphere is facing unprecedented changes——the easterly phase of the QBO, initiated at the beginning of 2016 (Osprey et al., 2016), is interrupted by an abrupt westerly above 50 hPa (Fig. 10).

Figure 10.  Pressure-time evolution of the equatorial (5^°S-5^°N) zonal-mean zonal wind (color shading; units: m s^-1) from January 1981 to June 2016. The vertical lines show the concurrent winter and the following winter, relative to the mature phase of the (a) 1982/1983, (b) 1997/1998 and (c) 2015/16 super El Niños. The horizontal lines mark the tropical tropopause.

Since 1981, three super El Niño events——those of 1982/ 83, 1997/98 and 2015/16——have been observed in three SST analyses (i.e., COBE, ERSST and HadISST). The maximum SST anomalies during the mature phase of these super El Niños are located in the central and eastern equatorial Pacific, consistent among the three SST analyses. The warm SST anomalies appear earliest in the western equatorial Pacific and precede the mature phase of the super El Niños by more than 18 months, i.e., in July(-1). From the previous winter, relative to the mature phase of ENSO, the warm SST anomalies begin to develop gradually in the western equatorial Pacific. From the summer of year(0) to the mature winter, the mixed layer temperature develops gradually and reaches a maximum, in phase with the peak of the SST anomalies in the eastern equatorial Pacific cold tongue region.

The super El Niños show quite a large influence on the atmospheric circulation from the troposphere to the stratosphere. In the previous winter, an NPO-like pattern appears in the northern extratropical troposphere (Figs. 4a-c), consistently among the three super El Niños, which could be viewed as a precursor for the initiation of an El Niño event. In the developing summer, as ENSO is just beginning to develop and the SST forcing is not too large, quite weak and inconsistent height anomalies are seen among the three events. In the upper troposphere, large circulation anomalies mainly occur in the mature winter and decaying summer. During the mature winter, a PNA teleconnection appears in the extratropics, in spite of the subtle differences between the intensities as well as the zonal (longitudinal) position of the anomalous North Pacific low. The PNA teleconnection in the winter of1997/98 is the strongest among the three super El Niños, which may be connected to the weakest cancelling interference among them of the TIO diabatic forcing (Fig. 1; Figs. 4g-i; Fig. 6a; Rao and Ren, 2016a). In the decaying summer, positive height anomalies dominate over the Indian Peninsula and the Indochina Peninsula, indicating an intensification of the South Asian high.

The northern winter stratosphere shows large changes in the polar cap temperature and the circumpolar westerly, if the interferences of the QBO, TIO heating, and solar cycle are linearly filtered out from the circulation data. Associated with the positive PNA response in the three super El Niño winters, positive polar cap temperature anomalies and circumpolar easterly anomalies are also observed in the mature winters of the three super El Niños, although the starting months show some differences. The stratospheric polar vortex in the following winter, relative to the mature phase of the 1982/83 and 1997/98 super events, is also anomalously weaker and warmer. Meanwhile, it is still not clear what will happen in the winter following the mature phase of the 2015/16 super El Niño. It is also not clear whether the unprecedented interruption of the easterly QBO phase by the westerly during summer 2016 is caused by the 2015/16 super El Niño, or if the interruption will impact the evolution of ENSO in the coming months (Osprey et al., 2016).

Given the low sample size of recorded super El Niño events, much uncertainty exists in this study. Therefore, further studies that use model outputs are needed to verify our diagnosis. As indicated by Rao and Ren (2016b, c), strong super El Niños show lower efficiency than moderate ones in modulating the extratropical stratospheric circulation, but the absolute values of the extratropical circulation anomalies are still quite large. It is hypothesized that the zonal shift in the PNA lobe in the North Pacific is also modulated by the TIO thermal state (Figs. 4g-i).

Ren et al. (2016) documented that the 1982/83 and 1997/98 ENSOs belong to the quasi-biennial (QB) type if they are classified according to their Niño3 evolution. If the Niño3 index reverses its sign from positive to negative in the decaying July-September, ENSO can be categorized into the QB type, which usually shows a concurrent impact on the northern winter stratosphere. This may not be completely true for super ENSO events, since the warmer response in the midlatitude stratosphere is quite strong and can persist throughout the decaying warm season. Just as during a moderately strong quasi-quadrennial ENSO event, the warm anomalies in the midlatitude stratosphere developing in the decaying warm season may be transported into the polar region by the active planetary wave in the following winter, relative to the ENSO mature winter.