Get Complete Project Material File(s) Now! »
Characterisation of ENSO diversity
Although the names and definitions of these El Niño types are slightly diﬀerent, the Warm Pool, Central Pacific, Dateline and Modoki El Niño events are often considered to describe the same type of events, in which the warmest SST anomalies occur in the western-to-central Pacific. By analysing the evolution of several El Niño events in the E-C phase space between May(0) (of their developing year) and January(1) (of their mature phase), Takahashi et al. (2011) confirmed that Modoki, CP and moderate El Niños can be grouped in the same category, distinct from extreme warm events (see their Figure 1). It should be noted that the canonical El Niño, referring to as the El Niño composite of Rasmusson and Carpenter (1982), is often compared to EP El Niño events in literature (Kug et al., 2009; Kim et al., 2009). However, it shares more features with moderate El Niños than with strong EP El Niños, despite an initial warming in the eastern Pacific that the CP events do not have (Takahashi et al., 2011; Dewitte and Takahashi, 2017). For instance, in the E-C phase space, the canonical El Niño event of 1957-58 (Rasmusson and Carpenter, 1982) is close to Modoki or CP El Niño events (Takahashi et al., 2011).
Commonly, the typical Eastern Pacific (EP) El Niño develops strongly in the (far) eastern Pacific (Niño-1+2 region, see Figure 1.4a) during boreal spring. Although sometimes the developing SST anomalies do not have propagating characteristics and are thus considered as quasi-stationary (Kug et al., 2009), it generally extends westward during summer (Kao and Yu, 2009), meeting a second warming zone that develops in the central Pacific following local westerly wind events (Wang and Weisberg, 2000). The peak of positive SST anomalies is reached in boreal winter, before retreating in an almost identically opposite way that restores the normal cold waters of the eastern Pacific Cold Tongue during the following spring. Typically, the Niño regions experience this decay in April of the second year except in the Niño-1+2 region, where the decay may occur later, in July (Kao and Yu, 2009). It should be noted however that some diﬀerences appear in the development of strong El Niño events: during the 1982-83 events for instance, no initial warming in the extreme east has been observed but the SST anomalies herein were amplified in the spring after the mature phase of the event.
The Central Pacific (CP) El Niño has a diﬀerent structure with positive SST anoma-lies confined in the western-to-central Pacific (Kao and Yu, 2009). Positive SST anomalies often begin during the summer months near the international dateline (180° longitude). The event spreads meridionally with extension to the eastern subtropics, but fails to propagate zonally to the extent that the EP El Niño does. The mature phase of the event, with maxi-mum SST anomalies amplitudes, is also often reached in winter and the event retreats during spring of the second year. Furthermore, CP (EP) El Niño events are characterised by weaker (stronger) SST anomalies and have a shorter (longer) duration, around 9 (15) months.
Lai et al. (2015) analysed the spatial evolution over the equatorial Pacific of all observed El Niño events over 1980-2013. They concluded that almost all events start with a warming (+0.5 ◦C) in the central Pacific. They also showed that all CP El Niños and their hybrid El Niños decay with a warming in the central Pacific. Only EP El Niño, which correspond in their study to the two strong El Niño events of 1982-83 and 1997-98, end with a warming in the eastern Pacific. Thus, the initiation phase would be similar between El Niño events, which would then diﬀer in their direction of propagation (see paragraph 188.8.131.52).
Some studies suggested that other classifications, such as moderate versus extreme El Niño events could be more relevant. Among the diﬀerent SST anomalies metrics presented above, the “TD” classification also integrates information on the strength of the event. It leads to a categorisation between strong and moderate El Niño events, based on the underlying Bjerknes feedback (Takahashi and Dewitte, 2016) as well as between moderate CP and moderate EP El Niño events, based on the variability in the far eastern Pacific related to an air-sea interaction mode (Dewitte and Takahashi, 2017).
In addition to a possible more physical approach to the phenomenon, the diﬀerentiation of strong or extreme El Niño events is of prime importance because of the greater and dramatic socio-economics global impacts (Philander, 1990; McPhaden et al., 2006). Another approach for identifying extreme El Niño has been proposed by Cai et al. (2014) and focuses on precipitation anomalies rather than SST anomalies. Characterised by an important warming in the eastern Pacific with SST exceeding 28 ◦C, the two observed strong El Niño events of 1982-83 and 1997-98 induced precipitation anomalies in the eastern equatorial Pacific, normally dry and cold (Philander, 1983; McPhaden, 1999). Indeed, during strong El Niño events, the ITCZ moves from its climatological location (8°N) to the eastern equatorial Pacific (Rasmusson and Carpenter, 1982; Lengaigne and Vecchi, 2010) due to the weakening of the zonal and meridional SST gradients. It leads to atmospheric convection and exceptional rainfall (> 5 mm/day) in the eastern Pacific. Cai et al. (2014) showed that the precipitation in the eastern Pacific increases non-linearly with the increase of SST anomalies in Niño-3 and the decrease in the eastern Pacific meridional SST gradient. Cai et al. (2014, 2015a, 2017) defined thus an extreme El Niño event when the boreal winter (DJF) rainfall in the Niño-3 region is beyond an arbitrary threshold of 5 mm/day.
More recently, an index based on the longitude of the deep convection in the equatorial Pacific basin has been proposed by Williams and Patricola (2018). Based on the same idea as the precipitation index of Cai et al. (2014), it allows in addition to discriminate the eﬀect of the increase in column water vapour due to warming in order to focus on the rain-related ENSO eﬀect. The proposed ENSO Longitude Index (ELI) contains longitude information of where deep convection occurs along the equatorial Pacific. Surprisingly they proposed only one index to characterise ENSO diversity, arguing that it accounts for the non-linear, threshold-like response of deep convection to SST. The index is not weighted by how much SST exceeds the threshold, it is not an indication of total magnitude of the event. However, depending on the resulting longitude value, the index makes it possible to determine the displacement of the deep convection associated with El Niño events and so, indirectly, the qualitative importance of the warming in the eastern Pacific where the relationship rainfall-SST is non-linear (see paragraph 184.108.40.206). One of the main advantage of the index is that it is robust to changes in the SST background state since the convection threshold used for the calculation of the index is defined by comparison with the tropics-wide average SST.
Finally, Dewitte and Takahashi (2017) showed that CP and Modoki El Niño events may be apprehended as moderate CP events while canonical El Niño events as moderate EP events. This classification is close to that proposed by Chen et al. (2015a) who defined three distinct SST warm patterns: the extremely strong El Niño event, with large warming near the South American coast, the CP El Niño, a weak warm event centred near the dateline and the canonical El Niño event with moderate warming in the central-eastern Pacific. We will come back in more details on this classification in terms of amplitude (extreme versus moderate) closely related to the classification in terms of SST anomalies patterns (EP versus CP) in the chapter 3.
The diﬀerent classifications agree that the observed El Niño events of 1983-83, 1997-98 and to a lesser extent 2015-16 are strong or extreme events (Takahashi et al., 2011; Santoso et al., 2017; Cai et al., 2017). However, even between extremes events with the same magnitude of SST anomalies across the central and eastern equatorial Pacific (L’Heureux et al., 2017) (see Niño-3.4 index in Fig. 1.5), the temporal and spatial evolutions of the tropical Pacific show important discrepancy (Abellán et al., 2017b). The anomalous warming of the 1997-98 El Niño develops in the eastern tropical Pacific in late boreal spring – early summer and then extends to the central Pacific the next winter. The anomalous warming of the recent El Niño event is less intense in the eastern Pacific and extends further west of the dateline. The associated precipitation field is even more contrasted with a spatial extension far eastward and a maximum of precipitation in the Niño-3.4 region in winter for the 1997-98 event while the displacement of the precipitation centre remains confined to the central Pacific with a maximum of precipitation located in the Niño-4 region in winter for the 2015-16 event (see spatial patterns in Fig. 1.5).
Figure 1.5 – Time-longitude sections of the equatorial band (2°S-2°N) SST anomalies (◦C) for (left) the 1997-98 El Niño event and (middle) the 2015-16 El Niño event (HadISST v1 dataset). The con-tours represent the evolution of the precipitation (mm/day) with intervals every 4 mm/d (GPCP v2.3 dataset). On the right figure is shown the time-evolution of (solid lines) the Niño-3.4 index (SSTA averaged over 5°S-5°N, 190-240°E, black box of the left figures) and (dashed lines) the E-index during both events (1997-98 in red, 2015-16 in blue).
When using the E-index defined by Takahashi et al. (2011), diﬀerences between both events are significant (see E-index in Fig. 1.5): the maximal amplitude is of 2 ◦C greater for the 1997-98 event than for the 2015-16. The warm period of the 1997-98 El Niño, when the E-index is greater than 1.5 ◦C, lasts twice as long, from June 1997 to June 1998 when the 2015-16 event warm period lasts from July 2015 to January 2016. Finally, while the 1997-98 event presents a marked peak in December 1997, the 2015-16 event reaches a SST anomalies peak not as large but that persists more than the 1997-98 event. These diﬀerences of evolution of the E-index are reflected in the spatial pattern of the two events, with a significant warming in the extreme east of the Pacific Ocean (Niño-1+2 region) for the 1997-98 event (L’Heureux et al., 2017), the E-index explaining most of the variability of the eastern Pacific (see paragraph 3.2.2) when the 2015-2016 event had an important warming in the Niño-4 region in addition to the exceptional warming in Niño-3.4 region (Xue and Kumar, 2017; L’Heureux et al., 2017; Abellán et al., 2017b). Even if these two events are characterised as strong events, their temporal and spatial evolution is drastically diﬀerent.
Williams and Patricola (2018) pointed out that the ONI index cannot eﬀectively capture the diﬀerences between the 1997-98 and 2015-16 El Niño events because the variation between them in the phase space PC1-PC2 is orthogonal to the variability explained by ONI (their figure 3a). Santoso et al. (2017) raised also the diﬃculty of defining strong El Niño since extreme El Niño, defined by Cai et al. (2014) from a threshold of precipitation in the eastern part of the tropical Pacific, does not necessarily propagate eastward as extreme El Niño events defined by Santoso et al. (2013) or McPhaden and Zhang (2009) do.
Thus, the El Niño classification into strong events may also depend on the metrics used. Similarly, the classification between CP and EP El Niños is dependent on the metrics used, as shown by Singh et al. (2011) and Jeong and Ahn (2017) in observations, or by Stevenson et al. (2017) using a large ensemble Coupled General Circulation Model (CGCM). These studies inter-compared diﬀerent methods using several datasets. Singh et al. (2011) compared the “AS”, “KY”, “KU” methods among others to their Sea Surface Salinity (SSS) clustering. Jeong and Ahn (2017) compared “AS”, “KY”, “KU”, “RJ” and “TD” methods. Stevenson et al. (2017) used the “KU”, “KY” and Yeh et al. (2009a)’s methods.
The use of these diﬀerent indices leads to diﬀerent classifications of El Niño events type, which may question the validity of the results that estimate the projected evolution of El Niño statistics in a warmer climate. The use of simplified indices, useful for studying complex climate phenomena, is above all a tool for accessing the diversity of ENSO in a qualitative way. It must be kept in mind that compositing El Niño events via these metrics can mask significant diﬀerences between events of the same class.
Knowing that ENSO diversity causes diﬀerent global impacts (see paragraph 1.1.3 and Larkin and Harrison (2005a); Ashok et al. (2007) among others) and may contribute diﬀer-ently to climate variability (see paragraph 1.3.2 and Lee and McPhaden (2010); Choi et al. (2012); Johnson (2013) among others), the characterisation of El Niño flavours by adapted metrics is crucial to understand their associated dynamics, from the atmospheric stochastic forcing to ENSO feedbacks (see section 1.2 of the chapter). The characterisation of the ENSO diversity is also of prime importance to estimate changes in frequency of occurrence of El Niño due to internal variability (see section 1.3) and due to global warming (see section 1.4).
A “wicked” proble
As mentioned by Christina Karamperidou at the recent ENSO conference in Guayaquil, Ecuador in 20182, ENSO is a “wicked” problem. A wicked problem is defined by its diﬃculty or impossibility to solve because of incomplete or contradictory knowledge and changing requirements that are often diﬃcult to recognise3. The main feature of a wicked problem is that it has no definitive formulation, neither stopping rule, nor ultimate test of a solution. There is always more than one solution for a wicked problem, with the appropriateness of the solution depending greatly on the individual perspective of the designer of the solution. Taking into account the elements presented previously, it appears that ENSO diversity is diﬃcult to define and can be assessed in many diﬀerent ways, whether one want to diagnose the event’s intensity or its spatial pattern for instance. This issue raises several questions.
How to define ENSO diversity? Which variable(s) should we use? Temperature or precipitation or more at once? Should the flavours be defined by their intensity or their spatial pattern? Should we analysed the climate variables at a fixed time, commonly during the mature phase of the events, or characterise their temporal evolution? If, conversely, the descriptive approach through climate variables is not the one chosen, should we rely on physical processes? If so, which ones? As long as one does not better understand the mechanisms that govern the diﬀerent flavours of ENSO, it is diﬃcult to consider putting forward a particular physical process. There is (for now) no definitive formulation for ENSO diversity and the many metrics proposed over the last twenty years reflect this ambiguity.
An important underlying issue of the diversity definition is how we should approach ENSO diversity? As a continuum representation of an asymmetric phenomenon or as two independent modes? The underlying issue is to determine whether the two El Niño flavours involve diﬀerent physics or can emerge fro the same dynamics.
The continuum view comes from the apparent continuous distribution of the longitudes of the ENSO SST anomalies peaks: El Niño and La Niña events have a large longitudinal distribution, and their anomaly centres intertwine. In this framework, weak or moderate El Niño and extreme La Niña have similar patterns with a broad range of longitudes while extreme El Niño are outliers with intense warming in the eastern Pacific. Johnson (2013), using observations over 1950-2011, suggested a framework representing ENSO as a continuum that can be explained by 9 statistically distinguishable patterns. In a CGCM capturing realistically ENSO diversity, the GFDL CM2.1 (see paragraph 220.127.116.11 for a description of the model), the peaks of the El Niño events shift eastward as they strengthen (Fig. 1.6a(b), reprinted from Capotondi et al. (2015)), a result also observed in the reanalysis dataset SODA v2.2.4 (see Figure 9a of Giese and Ray (2011) and paragraph 2.1.1 for a presentation of the reanalysis dataset). However, in the reanalysis dataset over the period 1871-2008, the central longitude of the ENSO warming is indistinguishable from a Gaussian distribution centred near 140°W (Giese and Ray, 2011) when a weak bi-modality appears in the pre-industrial (PI) control simulation of the CGCM (Fig. 1.6aa). Note that the centre of warming used by Giese and Ray (2011) is determined via the CHI index (see paragraph 18.104.22.168), which considers the warming over an area at least equal to the area of the Niño-3.4 region. The centre is calculated for each month of the event. In contrast, Capotondi et al. (2015) used maximum SST anomalies (5°S-5°N) in DJF to determine the associated longitude in the model. Although these two methods are not comparable, El Niño warming centres appear to occupy a wide range of longitudes. Fedorov et al. (2015) confirmed that CP and EP El Niños could be viewed as a continuum especially when focusing on the average warming magnitude and the anomaly heat centre rather than on the maximum SST anomalies and the exact longitude of this maximum. Characterising El Niños as strictly central or eastern events appears contradictory with this continuity in the distribution of longitudes (Giese and Ray, 2011; Johnson, 2013; Capotondi et al., 2015). Moreover, the spatial patterns of CP and EP events are often compared to the first two modes of the tropical Pacific variability obtained from an EOF analysis (Ashok et al., 2007). The second EOF mode, whose pattern may resemble CP El Niño spatial pattern (see paragraph 22.214.171.124 and Fig. 2.8), may rather be considered as a “modulator” for describing inter-event diﬀerences and requires a substantial projection onto the first EOF mode to appear. Indeed, an El Niño event may be considered as an appropriately weighted superposition of the two modes. Cai et al. (2015a) argued that this feature of ENSO involving both EOF modes represents a continuum. Another argument is that the existence of two modes of ENSO variability would imply the existence of diﬀerent deterministic processes that favour one mode over the other. Lai et al. (2015) argued that observed El Niño events (1980-2013) form a continuum with CP and EP El Niño events at each end. The “location” of an El Niño event on this continuum depends on the respective value of two parameters, the zonal wind anomalies and the recharged state of the western Pacific (see paragraph 126.96.36.199). However, Newman et al. (2011) showed that the central and eastern Pacific variability associated with the two regimes may arise from natural random variation, as may the multidecadal variations in ENSO flavours characteristics (see paragraph 1.3.2). Note that they described ENSO flavours as diﬀerent combinations of two orthogonal spatial patterns, precursors to CP and EP El Niño events and excited by stochastic atmospheric forcing, which can be assimilated to a continuum of mixed CP and EP patterns.
Conversely, numerous studies argued that EP and CP El Niños are distinct regimes be-cause of the distinct seasonal evolution of their SST anomalies patterns, their distinct meteo-rological impacts, their distinct oceanic and atmospheric processes or their frequency change after the climate regime shift of 1977 (Ashok et al. (2007); Kao and Yu (2009); Kug et al.
EL NIÑO-SOUTHERN OSCILLATION
(2010); Ren and Jin (2011) among many others). Ashok et al. (2007) raised first the question of whether the Modoki El Niño, with a pattern of warming in the central Pacific flanked by colder SST anomalies on both sides along the equator, is a part or not of El Niño evolu-tion. They suggested that the two regimes are actually two major orthogonal modes of the ocean-atmosphere coupled system in the tropical Pacific. A comparative analysis of various linear and non-linear approaches and composite analysis techniques has been conducted on observed and reanalysed climate datasets to further investigate whether these regimes are two distinct phenomena (Marathe et al., 2015). In particular, they analysed the seasonal evolution of Modoki and canonical El Niño events and suggested that due to their distinct evolution in summer season, their persistent warming in central Pacific and their distinct teleconnections, Modoki El Niño events should be classified as distinct events. They called for considering the Modoki El Niño as a separate phenomenon, at least for the purpose of applications and impact studies. With a diﬀerent approach, Ren and Jin (2011) showed that the two types of El Niño, named WP and CT El Niño in their study, may be considered as two independent modes of ENSO although the data is too short to be statistically significant. Indeed, the two modes correspond to diﬀerent zonal propagation of ENSO SST anomalies and the climate shift in 1976-77 led to ENSO regime changes with the increase in CP ENSO regime.
Table of contents :
1.1 El Niño-Southern Oscillation
1.2 ENSO’s theories
1.3 Multi-scale interactions
1.4 Global warming
2 Data and methods
2.1 Data description
2.2 Methodological tools
2.3 Heat budget computation and validation
3 Characterising ENSO diversity
3.2 Robustness of definitions
3.3 Correspondence of definitions
3.4 Discussion: A coupled system in transition
Synthesis and Perspectives
4 Low-frequency modulation of ENSO diversity
4.2 Low-frequency modulation of the tropical Pacific
4.3 Relationship between ENSO and mean state
Synthesis and Perspectives
5 ENSO diversity and global warming
5.2 Article submitted to Climate Dynamics
5.3 Supplementary materials
Synthesis and Perspectives
Conclusions et Perspectives