Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
A distinct class of El Niño events with extreme magnitude (termed “super El Niño” events in this study) is identified after removing decadal variation. These events occurred in 1972/1973, 1982/1983, and 1997/1998. They are distinguished not only by their size but also by associated features such as a Southern Hemispheric transverse circulation that is not similarly robust in other El Niño events. This transverse circulation is characterized by a low-level equatorward flow, which spins off from a high sea-level-pressure anomaly around Australia and then merges into the deep convection anomalies over the central Pacific, and by an upper-level southward divergent flow, which branches off from the convection center and connects to the subsidence of the Australia high. It is suggested that this transverse cell, peaking in boreal summer, serves as an effective booster during the developing stage of a super El Niño by intensifying tropical Pacific low-level westerly winds.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Many attempts have been made to classify El Niño events according to frequency [e.g., Rasmusson et al., 1990], magnitude [e.g., Wolter and Timlin, 1998; Stephens et al., 2007], onset time [Xu and Chan, 2001; Horii and Hanawa, 2004], or zonal structure of sea surface temperature (SST) [Ashok et al., 2007; Kao and Yu, 2009; Kug et al., 2009]. Not intending to add new classification, we single out a set of El Niño events in this study based on a threshold of magnitude. Figure 1 shows that among 18 identified El Niño events during the instrumental era of 1952–2010, only three events (1972/1973, 1982/1983, 1997/1998) with Niño-3 SST anomalies attain values substantially above the rest by one standard deviation. We refer to these extreme events as “super El Niño” and the rest as “regular El Niño”; the partitioning is further supported by performing a K-mean clustering analysis as part of the supplementary material.
Despite a limited number of super El Niño events that have been recorded in the modern period, they account for a disproportionate share of the known features of El Niño due to their large magnitude and their uniform pattern. Therefore, the grouping of super El Niño provides an attractive starting point to study El Niño. This new class of super El Niño events raises several questions. Do super El Niño events possess enough common features to merit a special class, or are they merely a stochastic picking from normality? What are the factors that cause this runaway growth of El Niño? Why do they appear so rarely yet seemingly following a decadal periodicity? In this article, we will answer at least some of these questions.
Various hypotheses and theories have been proposed to explain the underlying causes of the extraordinary magnitude of some El Niño events. They include internal nonlinearities that lead to bursting growth of some El Niño [Timmermann et al., 2003], state-dependent (multiplicative) noise [Jin et al., 2007; Levine and Jin, 2010] such as strong westerly wind events under an El Niño favorable condition [Lengaigne et al., 2004], and nonlinear response of convection to SST [Okumura and Deser, 2010; Choi et al., 2013]. However, the attentions thus far have been mostly focused on the processes within the tropical Pacific, while forces outside the tropics have been overlooked, which could play an indispensable role in pushing El Niño's growth over the hurdle. In this paper, we present evidence for a tight relationship that exists between super El Niño and certain Southern Hemispheric anomalous circulation during El Niño developing stage. We hypothesize that these southern hemispheric features act to boost El Niño's growth and turn it into a super El Niño. In other words, it requires the right amount of push through the positive feedback between equatorial Pacific and southern hemispheric winter subtropical circulation during El Niño developing phase in early boreal summer and autumn for an El Niño to grow into a super El Niño.
The rest of this paper is organized as follows. Section 'Data' describes data used in this study. Section 'Transverse Circulation' focuses on the transverse circulation, highlighting the unique features of super El Niño events with respect to regular El Niño events. To illustrate the structural difference between super and regular El Niño events, the Taylor diagram for each rescaled field is analyzed. In section 'Conclusion and Discussion', results are summarized, and implications of singling out the super El Niño are briefly discussed.
In this study, we used the SST data from the Hadley Centre Sea Ice and Sea Surface Temperature data set (HadISST) [Rayner et al., 2003], and the monthly atmospheric variables from the 40 year European Centre for Median-Range Weather Forecasts (ECWMF) Re-Analysis (ERA-40) [Uppala et al., 2005]. Monthly anomalies were obtained as departures from monthly mean climatology for the period of 1958–2001; then, the Kolmogorov-Zurbenko (KZ) filter [Eskridge et al., 1997] was applied to remove high-frequency (less than six months) and longer-than-ENSO scale (8 years) variations. The KZ filter gives iterative moving average that removes high-frequency variation relative to the window size; the method cleanly separates various time scales of meteorological variables and has the same accuracy as the wavelet method. Hence, the band-pass filter (3 month window, two-iteration KZ filter minus 43 month window, two-iteration KZ filter) in particular preserves the peaks within ENSO time scale yet alleviates the abruptness of 1976/1977 climate change. The filtered data were highly correlated (r > 0.9) with those filtered using other commonly employed methods, e.g., a 6 year high-pass filter based on successive application of running means centered at 25 and 37 months [Zhang et al., 1997]. As a result, our conclusions presented here do not depend on our choice of a specific filter, and only analyses of KZ-filtered data are discussed. Regular El Niño events (excluding the three super El Niño cases) are defined as Niño-3 SST anomalies exceeding 0.5 standard deviation for at least 3 months. The time span covered by the ERA-40 data includes nine regular El Niño events (1963/1964, 1965/1966, 1968/1969, 1969/1970, 1976/1977, 1986/1987, 1987/1988, 1991/1992, and 1994/1995). Since the number of super El Niño is limited in the observation, we also used the output of a coupled model, GFDL-ESM2M, to investigate super El Niño with much more (12) events. The GFDL-ESM2M model output was obtained from the Coupled Model Intercomparison Project phase 5 (CMIP5) [Taylor et al., 2011] data archive (http://cmip-pcmdi.llnl.gov/cmip5/index.html). Simulated super El Niño in the pre-industrial experiment for a 500 year simulation period was analyzed and compared to regular super El Niño in the model.
3 Transverse Circulation
During the El Niño developing stage (June–November, or JJASON), unique features can be identified outside the tropics in the super El Niño composite compared with the regular El Niño composite (Figures 2c–2d). Of particular interest are the intensified 10 m westerlies over the western tropical Pacific that are closely associated with a large-scale coherent pattern in sea level pressure (SLP), with high pressure over southern Australia and low pressure over the southern Pacific. These two systems form an SLP dipole in the southern midlatitudes, far from the low latitudes where the typical Southern Oscillation is located. The anomalous high pressure over Australia (hereafter Australian high) spins off strong equatorward flow along the east coast of Australia, which moves across the equator near New Guinea, and veers off to the east from a potential vorticity perspective [e.g., Rodwell and Hoskins, 1995] to accelerate westerly winds in the central Pacific. The veering off is also facilitated by a NW-SE aligned tropical trough anomaly near Melanesia, a result mainly caused by Rossby response to El Niño's anomalous convection, in addition to some regional effects near the Solomon Sea (see numerical experiments in Annamalai et al., 2010). We propose that this extra-tropical forcing from the Southern Hemisphere adds on top of the Bjerknes instability [Bjerknes, 1969; Wyrtki, 1975]. By strengthening the westerly wind in the ENSO source region, the anomalous circulation in the Southern Hemisphere acts as a booster to amplify El Niño's growth, a critical step toward becoming a super El Niño.
Stephens et al.  pointed out that a standing wave pattern in SLP in the Southern Hemisphere, similar to the one in Figure 2c, during strong El Niño developing stage weakens the annual cycle and drives the southwesterly wind toward the equatorial Pacific. They proposed that the South Pacific low-pressure anomaly plays an important role in regulating the strength of trade winds, which leads to changes in El Niño amplitude. In this study, however, we find that in the case of super El Niño events the Southern Hemisphere SLP dipole interacts with the tropics in a more intimate way through a self-sustaining transverse cell that connects the El Niño source region and the Australian high.
To depict the transverse cell, anomalous upper-level atmospheric fields and vertical motion in the upper-middle troposphere are shown in Figure 2a. During super El Niño years, anomalous divergent winds generated from enhanced deep convection over the central equatorial Pacific move southwestward and then converge over the Maritime Continent and the Australian high (Figure S2). The upper-level atmosphere features a unique pattern of anomalous Rossby wave source (RWS) (see Sardeshmukh and Hoskins, 1988 for the derivation of anomalous RWS). Anomalous divergent winds advect large vorticity southward and result in positive RWS anomalies (, shown by purple contour) around 25°S by crossing the maximum absolute vorticity gradient latitude that is intense in the austral winter. Moreover, corresponding to the upper-level convergence over southeastern Australia, descending motion in the troposphere induces vortex stretching in the upper troposphere and produces a pronounced patch of negative RWS anomalies (, green contour) over southeastern Australia and the central South Pacific. This subsidence over southeastern Australia results in vortex shrinking, which drives additional equatorward flow in the lower troposphere. A low-level high-pressure anomaly stands out to its left to maintain the Sverdrup balance (Figure S3) [Hoskins and Karoly, 1981]. The low-level southerly winds and the Australian high are thus integrated in a transverse cell to facilitate El Niño's growth.
Since the southerly winds along northeastern Australia directly interact with El Niño by affecting the strength of zonal wind stress over the western and central equatorial Pacific, a Southern Hemisphere booster (SHB) index is defined utilizing the 850 hPa meridional wind averaged over 10°S–30°S and 140°E–170°E normalized by the mean over the period of January 1958 to December 2001. Figures 2e and f show the temporal evolution of the transverse cell comparing with the normalized Niño-3 index. The SHB index leads the Niño-3 index by about 3 months during the El Niño onset/developing stage (Figure 2e); the index strengthens rapidly from April, peaks in August, and weakens after October. This feature is only found in the super El Niño group. In the class of super El Niño, the prior existence of SHB indicates that the Southern Hemispheric anomalous circulation is not passively forced by El Niño. In contrast, such a lead could not be detected in regular El Niño (Figure 2f), and the magnitude of the transverse cell is either too weak or absent altogether. The lack of SHB in regular El Niño is consistent with its sluggish growth and small amplitude. More details on the SHB index and Niño-3 index for individual El Niño events are provided in Figures S4–S6, which display temporal evolutions similar to the composite. Figure 2g contrasts the SHB index in JJASON(0) against the Niño-3 index in D(0)J(1) for the 12 El Niño events. The scatter diagram indicates that the strength of the transverse cell and the magnitude of El Niño are positively correlated (r = 0.87). The magnitudes of both indices for super El Niño events are proportionally larger than those for regular El Niño events. It appears that only El Niño events with the SHB index larger than two standard deviations during summer–autumn can grow into a super El Niño in the following winter.
One may suspect that regular El Niño events have a similar Southern Hemispheric structure but cannot be identified visually due to its weak strength. To examine this possibility, Taylor diagrams [Taylor, 2001] are used for analyzing pattern similarity for the associated spatial fields, which are rescaled by the November–January mean Niño-3 SST anomaly for each El Niño case to mitigate the amplitude difference between the super and regular El Niño events to a large extent. The rescaled super El Niño composite is used as the reference. The SLP (50°S–15°N, 90°E–90°W), vertical pressure velocity averaged over 300–700 hPa (35°S–15°N, 90°E–120°W), 10 m meridional winds (30°S–10°N, 120°E–180°), and the 200 hPa RWS vortex stretching term (45°S–15°S, 70°E–90°W) are chosen as the four representative fields of the transverse cell to quantify spatial similarity of each “normalized” El Niño case.
As shown in Figures 3a–3b, the three super El Niño events are clustered together and strikingly distinct from the rest. Their standard deviations are close to one, indicating that the magnitude of the spatial pattern is also close to the reference point (their composite). In contrast, regular El Niño events spread out in the Taylor diagrams due to the lack of resemblance to super El Niño. The large spatial standard deviation of regular El Niño events results from the lack of coherence in the transverse cell.
Figure 4 presents a schematic diagram demonstrating the extraordinary amplification of super El Niño events through a Southern Hemispheric transverse cell during the developing stage. Although the growth of super El Niño is mainly caused by the Bjerknes feedback in the equatorial region, the transverse cell further strengthens the low-level tropical westerly winds with the aid of the Australian high to booster super El Niño's growth during the austral winter.
A preliminary study was conducted on simulations of a state-of-the-art coupled climate model which shows that the Australian high also contributes significantly to the growth of model simulated super El Niño events. Within the framework of the CMIP3 and CMIP5, several recent papers [e.g., Lengaigne and Vecchi, 2010; Santoso et al., 2013; Cai et al., 2014] have laid the foreground for looking into various aspects of super (extreme) El Niño events. Here, we show the results from the output of the GFDL-ESM2M, generally regarded as one of the most reliable models to reproduce El Niño signals (see Supplementary Figures 8 and 11 in Cai et al., 2014). To be consistent with Figure 1, a super El Niño in the model is simply defined by the normalized Niño-3 SST greater than 2.5 standard deviations averaged over November–December–January, and a regular El Niño, by the value greater than 0.5 but less than or equal to 2.5 standard deviations. A total of 12 super and 109 regular El Niño events are found over a 500 year simulation period of the pre-industrial experiment in the GFDL-ESM2M output, and the two El Niño composites are shown in Figure 5.
The super El Niño events in the model feature a strong SHB signal during the developing stage, similar to what is found in the observation. For the regular El Niño events in the model, the SHB signal is much weaker. Moreover, we can infer the effect of SHB in terms of its role in strengthening El Niño in the sense that the SHB tends to peak ahead of the Niño-3 index as found in the observation. There exist discrepancies, however, between model simulation and observations. The Australian high in the model is further to the northeast, so the domain for evaluating the SHB index has to be shifted to 20°S–0° and 150°E–180°. Furthermore, the temporal evolutions of El Niño and its associated SHB in the model are different from those the observation. The simulated super El Niño events have a significantly longer duration due to earlier onset and later termination and have double peaks in August (year 0) and February (year 1). These biased features of simulated El Niño are also reflected in the associated SHB, which is likely due to model biases in simulating climate mean state and annual cycle.
4 Conclusion and Discussion
Evidence is presented to delineate a unique feature shared by three super El Niño events recorded in recent years (1972/1973, 1982/1983, 1997/1998): a distinct transverse cell mainly consists of an anomalous surface high in southern Australia, enhanced convection over the ENSO regime in the central equatorial Pacific, and a pattern of upper-level Rossby wave source in between. The transverse cell is pronounced in the super El Niño development phase during the boreal summer/austral winter. It is featured with the lower-level equatorward wind anomalies along northeast Australia, which can bring southern cold surge thus more westerly wind burst events to strength the El Niño. Thus, this cell is hypothesized to serve as a booster to facilitate the runaway growth of a super El Niño.
After rescaling to mitigate the amplitude difference between the super and regular El Niño events, Taylor diagrams show that the Southern Hemispheric anomalous pattern is indeed a unique feature shared by all three super El Niño events but not by regular El Niños. The output of CMIP5 GFDL-ESM2M supports the results based on the observation, despite some discrepancies between the model and the observation.
The Australian high, a key ingredient of the SHB, intensifies about 2 months before a super El Niño develops. Thus, our study sheds some light on the pre-condition of super El Niño events, which may have implications for their predictability. Moreover, the Australian high is prone to be modulated by other two major sources of low-frequency variability, known as the Antarctic Circumpolar Wave [White and Peterson, 1996] and the Southern Annular Mode [Lau and Nath, 2004]. Thus, occasional supposition of concurring forcing from these different kinds of low-frequency variability may effectively excite the decadal but aperiodic occurrence of super El Niño events, a possibility worthy of further exploration.
We thank two anonymous reviewers for their helpful comments, and we acknowledge the World Climate Research Programme's Working Group on Coupled Modelling, which is responsible for CMIP; we thank the climate modeling group (GFDL-ESM2M) for producing and making the model output available. For CMIP, the U.S. Department of Energy's Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. F.J. is supported by US NSF grant ATM1034798, Department of Energy grant DESC005110, and NOAA grant NA10OAR4310200.