ENSO Atmospheric Teleconnections and Their Response to Greenhouse Gas Forcing

1 Introduction

The El Niño and Southern Oscillation (ENSO) is the most prominent mode of interannual climate variability on Earth. It is generated through coupled interactions between the ocean and atmosphere in the tropical Pacific (Bjerknes, 1969) and alternates between anomalously warm (El Niño) and cold (La Niña) sea surface temperature (SST) conditions. Figure 1 displays a schematic diagram showing the tropical weather and oceanic condition along the equator during normal, El Niño and La Niña conditions. During El Niño, the surface air pressure over the tropical western Pacific becomes higher than in the tropical eastern tropical Pacific compared to a normal year, resulting in weaker trade winds and surface zonal currents. The prevailing rain pattern also shifts further east than normal. Weaker trade winds act to reduce ocean upwelling intensity, leading to a deeper thermocline and warmer-than-normal SSTs in the eastern and central tropical Pacific. During La Niña, in contrast, the trade winds in the tropical Pacific basin are stronger than normal, leading to the increase of the upwelling intensity off South America and along the equator and hence the lower than normal SSTs. The center of the prevailing rain pattern also shifts farther west than normal during La Niña.

ENSO has enormous impacts on natural and human systems including agriculture, forestry, public health, the hydrological cycle, the global carbon cycle, marine and terrestrial ecosystems, and fisheries (McPhaden et al., 2006). The impacts on regions outside the tropical Pacific arise through atmospheric teleconnections (Bjerknes, 1969; Diaz et al., 2001; Horel & Wallace, 1981; McPhaden et al., 2006; Stan et al., 2017; Trenberth et al., 1998) and oceanic teleconnections (England & Huang, 2005; Feng et al., 2013; Frischknecht et al., 2015; Gordon & Fine, 1996; Hazeleger et al., 2004; McPhaden et al., 2006; Vergara et al., 2017). The term “teleconnection” refers to a statistically significant remote response, either concurrent with or time lagged, from a forcing region, which in the case of ENSO is the equatorial Pacific.

ENSO oceanic teleconnections include, for example, heat extremes off the west coast of Australia via the Indonesian Throughflow (Feng et al., 2013). Equatorial wind anomalies associated with ENSO also generate oceanic Kelvin waves along the west coast of the Americas (Clarke & van Gorder, 1994), affecting coastal marine ecosystems and fisheries (Gutierrez et al., 2008; Meyers et al., 1998). In addition, the anomalous poleward oceanic heat transport following El Niño and La Niña events can affect the heat budget of the high-latitude oceans and hence the tropical cyclone activity (Gordon & Fine, 1996; Jin et al., 2014). ENSO atmospheric teleconnections have a strong and broad impact on global climate and weather events, and it is on these teleconnections that we focus this review.

During the mature phase of El Niño, the heat source in the tropical western Pacific expands eastward (Sikka, 1980; Wang, 2002). This eastward movement weakens the Pacific Walker circulation that in turn weakens the lower-tropospheric equatorial easterly trade winds. The reverse occurs during the mature phase of La Niña. Through the tropical atmospheric bridge via modulation of the Walker circulation (i.e., the near-equatorial zonal atmospheric overturning circulation), ENSO atmospheric teleconnections induce changes in cloud cover and evaporation in remote ocean basins such as the South China Sea, the Indian Ocean, and the tropical North Atlantic, with a time lag of 3 to 6 months (Klein et al., 1999; Wang, 2002). For example, during a developing period of El Niño, together with the eastward migration of the prevailing atmospheric convection over the tropical Pacific, the Walker circulation moves to the east, resulting in anomalous atmospheric subsidence over the Indian Ocean region and consequent suppression of convection there. This leads to warming in the Indian Ocean through intensified solar radiation as well as a weakening land-sea contrast in summer, resulting in less Indian summer monsoon rainfall. The opposite happens during La Niña (Webster et al., 1998). During boreal springs after the peak of El Niño events, the northeast trade winds usually weaken and the SST increases over the northern tropical Atlantic Ocean, which is associated with an atmospheric wave train generated in the central tropical Pacific (Huang, 2004).

ENSO atmospheric teleconnections also influence weather and climate conditions over other parts of the globe (Alexander et al., 2002; Chiodi & Harrison, 2015; Deser, 2000; Lau, 1997; Trenberth et al., 1998). Changes in deep convection due to ENSO leads to the changes in atmospheric heating, low-level convergence, and upper-level divergence in the equatorial Pacific (Figure 2; see also Hoskins & Karoly, 1981). The anomalous horizontal component of upper-level atmospheric vorticity forces large-scale atmospheric Rossby waves that propagate into the extratropics (Hoskins & Karoly, 1981; Trenberth et al., 1998) to excite a “Pacific-North American (PNA)” pattern (Horel & Wallace, 1981) (Figures 2a and 2b) and “Pacific-South American (PSA)” pattern (Karoly, 1989) (Figures 2e and 2f) in the Northern and Southern Hemispheres, respectively. These teleconnections influence weather and climate, including extreme events, around the globe.

Nearly two decades ago, Diaz et al. (2001) reviewed the principal features of ENSO atmospheric teleconnections in terms of regional patterns of surface temperature, precipitation, and atmospheric circulation. They highlighted the connections between the decadal-to-multidecadal variability in tropical SSTs and ENSO properties along with decadal changes in the ENSO atmospheric teleconnections during the 20th century. In particular, Diaz et al. (2001) highlighted the periods before and after the mid-1970s when both climate states in the Pacific (Graham, 1994; Trenberth & Hurrell, 1994; Zhang et al., 1997) and ENSO characteristics (An & Wang, 2000; An et al., 2006; Fedorov & Philander, 2000; Wang & An, 2002) significantly changed. The shift in the background state over the tropical Pacific during the mid-1970s included increased SST in the eastern and central tropical Pacific (Graham, 1994). In addition, ENSO amplitude as well as its dominant period increased after the late 1970s compared to before the late 1970s (Fedorov & Philander, 2000; Wang & An, 2002). Since the 1990s, however, ENSO properties such as the spatial structure and amplitude of SST anomalies (SSTAs) as well as the ocean mean state have changed again (Capotondi et al., 2015; Jo et al., 2013, 2014, 2015; Kosaka & Xie, 2013; Lyon et al., 2014; McPhaden, 2012; McPhaden et al., 2011; Meehl et al., 2014, 2015; Xiang et al., 2013) (Figure 3). It has been observed that the center of both El Niño and La Niña SST anomalies has shifted to the west from the period of 1979–1997 compared to the period of 1998–2015 (Figures 3a–3d). Furthermore, the amplitude of El Niño events weakened from before the late 1990s to the first 15 years of the 21st century (Figures 3a and 3b). La Niña events, however, have remained relatively unchanged. On the other hand, most ocean basins have warmed since the 1990s except for the eastern Pacific, the high-latitude Southern Ocean, and the eastern North Pacific (Figure 3e). The changes of ENSO properties as well as the ocean mean state are thus different from those that occurred during the 20th century (Diaz et al., 2001).

Substantial progress in understanding ENSO dynamics, diversity, and impacts has been achieved in the past decade (Cai, Santoso, et al., 2015; Capotondi et al., 2015; Marathe et al., 2015; Yeh et al., 2014). Previous publications during this time though have provided only a brief review on impacts associated with ENSO diversity (Capotondi et al., 2015; Yeh et al., 2014) and how ENSO teleconnections associated with ENSO extremes would change under greenhouse warming (Cai, Santoso, et al., 2015). There is still a debate about the processes associated with the nonlinearity of the ENSO atmospheric teleconnections, the diversity of the ENSO atmospheric teleconnections, and the role of regional air-sea coupled processes outside the tropics in modulating the ENSO atmospheric teleconnections. These issues motivate our review of ENSO atmospheric teleconnections (hereafter, ENSO teleconnections) associated with changes in ENSO properties and ocean mean state in the tropical Pacific in recent decades and how they may change under greenhouse warming. In addition, we will show that changes in the ENSO teleconnections significantly control the statistics of extreme weather events. Finally, we will review how climate models simulate the effects of greenhouse gas forcing on ENSO teleconnections in the future.

2 ENSO's Spatial Patterns and Their Teleconnections.

Interest in ENSO teleconnections has undergone a revival in the past decade considering the diversity in ENSO spatial patterns (e.g., Capotondi et al., 2015; Trenberth & Stepaniak, 2001) (Figures 3a–3d and 4a and 4b). The recent focus has been on differences in impacts on remote temperature and precipitation between events in which the maximum anomalous SST is located in the eastern tropical Pacific (namely, Eastern Pacific El Niño) and events with maximum anomalous SST in the central tropical Pacific (namely, El Niño Modoki, Central Pacific El Niño, or Warm pool El Niño) (Ashok et al., 2007; Kao & Yu, 2009; Kug, Jin, et al., 2009) (Figures 4a and 4b). In addition, Central Pacific (CP) El Niños have occurred more frequently than Eastern Pacific (EP) El Niños since the late 1990s (Capotondi et al., 2015; Lee & McPhaden, 2010; Yeh et al., 2015), which might be due to either anthropogenic forcing (Kim & Yu, 2009; Yeh et al., 2009) or natural variability (Giese & Ray, 2011; McPhaden et al., 2011, Newman et al., 2011; Yeh et al., 2011) or both. However, there is still a debate about the robustness of these changes before and after the late 1990s because of the relatively short observational record (Stevenson et al., 2010; Wittenberg, 2009). There is also an argument that the location of El Niño SST maxima varies considerably, ranging as a continuum from the western to the eastern Pacific for the period of 1871–2008 rather than being simply bimodal with maxima in the central and eastern Pacific (Giese & Ray, 2011).

For illustrative purposes, in the following discussion we consider the existence of two extremes of the El Niño continuum associated in a statistical sense with distinct SST anomaly patterns represented by the 1997/1998 EP and the 2009/2010 CP El Niño events (Takahashi et al., 2011). ENSO teleconnections are sensitive to the longitude where atmospheric deep convection is triggered (empirically where SST > 27.5 °C) (Barsugli & Sardeshmukh, 2002). These longitudes are different for EP El Niño and CP El Niño (Figures 4c and 4d), leading to different ENSO teleconnections. For example, the PNA-like Rossby waves propagating into the extratropics in the Northern Hemisphere with wave numbers of 3–4 during winter are dominant when an EP El Niño occurs (as in 1997/1998; Figure 5a). In contrast, the 500 hPa geopotential anomalies in response to CP El Niño SST forcing (as in 2009/2010) are associated with an increase in the geopotential height over the Arctic and a decrease in the surface pressure over the North Pacific and North Atlantic Ocean during winter (Figure 5b), somewhat similar to the spatial pattern of a negative phase of Arctic Oscillation (AO)-like circulation (Thompson & Wallace, 1998). In other words, the center of action for ENSO teleconnections during EP El Niño events is shifted to the east over the Eurasian continent, North America, and the North Atlantic. Conversely, during CP El Niño events, a westward shift is observed in the North Pacific. Further analysis indicates that the ENSO teleconnections during EP El Niño events and CP El Niño events are different in the Northern Hemisphere from the developing phase to the decaying phase (not shown). In particular, the location of their seasonal centers of action is shifted in the zonal and meridional direction in the high latitudes including the Arctic.

ENSO teleconnections in response to EP El Niño and CP El Niño result in different remote impacts in precipitation or temperature or both around the globe (Cai & van Rensch, 2012; Garfinkel et al., 2013; Graf & Zanchettin, 2012; Kumar et al., 2006; Mo, 2010; Preethi et al., 2015; Rodrigues et al., 2011; Taschetto & England, 2009; Wang & Hendon, 2007; Weng et al., 2007; Yu & Zou, 2013). Consequently, this diversity affects the predictability of ENSO teleconnections as well as the characteristics of extreme weather events in response to ENSO as illustrated in global land surface air temperature anomalies (Figure 6) and precipitation anomalies (Figure 7) during the 1997/1998 EP El Niño and the 2009/2010 CP El Niño winters. The spatial pattern of global surface air temperature and precipitation anomalies in response to the EP El Niño was quite different from that in response to the CP El Niño. For example, a dipole-like structure of surface air temperature anomalies over eastern North America is a dominant feature for CP El Niño, which is not the case for EP El Niño (Figures 6a and 6b). In addition, during the EP El Niño, central and eastern Europe exhibited positive surface air temperature anomalies during boreal winter, while the CP El Niño was associated with an anomalously cold winter in eastern Europe (Ashok et al., 2007; Graf & Zanchettin, 2012). It is also evident that surface air temperature anomalies in Antarctica were mirror images of one another for the EP and CP El Niños (Figures 6a and 6b). Likewise, precipitation anomalies in Europe and in Antarctic were mirror images of one another for the EP and CP El Niños (Figures 7a and 7b). Furthermore, it is found that CP El Niños are more effective in reducing the precipitation anomalies over India during the winter peak than EP El Niños, which is in contrast to eastern Australia and Mexico where the CP El Niño causes much precipitation than EP El Niño (Figures 7a and 7b).

A number of studies since the mid-2000s have focused on EP and CP El Niño teleconnections. Some studies compared differences in precipitation and temperature anomalies in response to a single EP El Niño and CP El Niño event (e.g., Figures 4-6) (Ashok et al., 2009; Barnard et al., 2011; Kumar et al., 2006; McPhaden, 2004). Others studies analyzed observed differences in teleconnection composites based on a number of EP El Niño and CP El Niño events (Ashok et al., 2007; Garfinkel et al., 2013; Graf & Zanchettin, 2012; Taschetto & England, 2009; Weng et al., 2007; Yeh et al., 2009; Yu & Zou, 2013). The interpretation of these results requires caution because of the limited number of El Niño events in the record. Thus, there are uncertainties in the precise nature of many regional ENSO precipitation and temperature teleconnections. Nevertheless, the leading ENSO teleconnection features during EP El Niño and CP El Niño, derived from available observations, provide clues on how and why EP El Niño and CP El Niño influence remote areas differently. The following summarizes the precipitation and temperature response to EP and CP El Niños for each continent based on the published literature. It should be noted, however, that these responses represent the El Niño signal against a background of considerable atmospheric noise. Correlations are relatively low suggesting that El Niño teleconnections account for only a fraction of the total variance in many parts of the globe.

3 ENSO Teleconnections Impacted by Ocean Basins Outside the Tropical Pacific.

There is a clear influence of ENSO on SSTAs outside the tropical Pacific Ocean (Deser et al., 2010). However, the slowly evolving state of the ocean outside the tropical Pacific is able to influence not only ENSO dynamics, variability, and diversity (Dommenget & Yu, 2017), but also ENSO teleconnections through local air-sea interactions particularly in the North Pacific, the North Atlantic, and the Indian Ocean. Thus, ocean state changes outside the tropical Pacific can lead to decadal modulation of ENSO teleconnections. The mean state of most ocean basins has warmed after the late 1990s except the eastern Pacific, the high-latitude Southern Ocean and the eastern North Pacific (Figure 3e). These mean state changes are the result of natural variability, anthropogenic forcing, or their combined influence. Because there are few studies examining the influence of SSTAs outside the tropical Pacific on the teleconnections resulting from EP El Niño and CP El Niño events, the following subsection mainly describes how these remote oceans impact ENSO teleconnections without making a distinction between EP El Niño and CP El Niños.

3.1 North Pacific Ocean.

There is a slow decadal evolution of SSTAs in the Pacific referred to as Pacific Decadal Oscillation (PDO) (Mantua et al., 1997) or the Interdecadal Pacific Oscillation (IPO) (Power et al., 1999). The PDO is defined as the leading empirical orthogonal function (EOF) mode of SST in the North Pacific north of 20°N, and it is the most dominant mode of SST variability in the North Pacific on decadal timescales (Newman et al., 2016). A positive PDO is characterized by cool temperatures in the western and central North Pacific and by anomalously warm temperatures to the east, north, and south of these cool temperatures. Opposite conditions occur during the negative phase of PDO. The PDO results from a combination of different physical processes, including local North Pacific atmosphere-ocean interactions involving reemergence of SST anomalies (Alexander et al., 1999), ENSO-forced variability from the tropics, and oceanic zonal advection of SST anomalies in the Kuroshio-Oyashio Extension. These processes operate on different timescales to collectively generate PDO-like SST anomaly patterns (Newman et al., 2016, Schneider & Cornuelle, 2005). The IPO is an alternative definition of the PDO but with an equatorial focus. As such it features a decadal timescale “ENSO-like” SST fluctuation in the Pacific, though less equatorially confined and relatively more prominent in the extratropics, especially the North Pacific, compared to ENSO (Han et al., 2014; Meehl & Hu, 2006; Power et al., 1999). A positive phase of IPO is characterized by a warm tropical Pacific and a cool extratropical Pacific and vice versa for a negative phase of IPO.

The slow evolution of SSTAs in the Pacific associated with the PDO/IPO can modulate ENSO teleconnections by interacting with the atmospheric circulation anomalies in the midlatitudes, or by changing the location of atmospheric deep convection in the tropical Pacific (Gershunov & Barnett, 1998; Hong et al., 2014). This modulation process can weaken or reinforce ENSO teleconnections on decadal timescales (Gershunov & Barnett, 1998; Hu & Huang, 2009; Kim et al., 2014; Yu & Zou, 2013). For example, during the positive phase of the PDO, or equivalently the IPO (Han et al., 2014), climate anomalies during boreal winter over North America and East Asia associated with El Niño tend to be strong and more spatially coherent (Fuentes-Franco et al., 2016; Hu & Huang, 2009; Kim et al., 2014; Kumar et al., 2016; Pavia et al., 2006). When the IPO is in a negative phase, on the other hand, the impact of La Niña on Australian precipitation is enhanced (Power et al., 1999). La Niña during a negative phase of the IPO modifies the location of the South Pacific Convergence Zone (SPCZ), which is responsible for delivering more rain-bearing cloud bands across eastern Australia (Folland et al., 2002). Usually, ENSO teleconnections around the Pacific basin are intensified when the ENSO and the PDO/IPO are in phase, and weaker when ENSO and the PDO/IPO are out-of-phase.

The phase of the PDO/IPO changed from positive to negative in the late 1990s associated with a recent decade long “hiatus” in observed globally averaged surface temperature. This period showed little increase or even a slightly negative trend of global mean surface temperature during the early 2000s (Hong et al., 2014; Kosaka & Xie, 2013; Meehl et al., 2011) though there is controversy about this issue depending on different definitions of “hiatus” and different data sets used (Medhaug et al., 2017; Risbey & Lewandowsky, 2017). This phase shift was characterized by a cooling in the eastern tropical Pacific (see Figure 3e) and a shift of the atmospheric convection to the west in the tropical Pacific during ENSO events (Jo et al., 2015; Kim et al., 2014; Xiang et al., 2013) (Figure 8). These changes result in changes in ENSO teleconnections to midlatitudes (see Figure 5). Figures 8a and 8b display the composite of anomalous outgoing longwave radiation (OLR) when El Niño occurred before and after the late 1990s, respectively. The center of negative OLR anomalies after the late 1990s is shifted to the west by about 15–20 in longitude compared to that before the late 1990s. The zonal distribution of OLR anomalies, averaged between 2.5°N and 2.5°S, before and after the late 1990s supports this notion (Figure 8c). Similarly, for La Niña, the center of positive OLR anomalies also shifted to the west by about 10° longitude from the late 1990s to after the late 1990s (Figures 8d–8f). It is known that the OLR is a proxy for atmospheric heating, anomalies of which force changes in atmospheric circulation that affect the middle-to-high latitudes (Chiodi & Harrison, 2015). These changes in OLR lead to altered forcing of the large-scale atmospheric Rossby waves by changing convergence patterns in the lower and upper atmosphere (Hoskins & Karoly, 1981), modulating the atmospheric teleconnections into the middle-to-high latitudes. For example, the anomalous Aleutian low pressure associated with El Niño events became deeper and its center shifted southwestward after the late 1990s primarily due to the westward shift in the location of the tropical convective heating around the dateline (see Figures 8a and 8b) that occurred after the late 1990s (Jo et al., 2015).

3.3 Indian Ocean

The Indian Ocean is home to some of the warmest water in the world's ocean, exceeding 28°C (Tao et al., 2015). Therefore, deep convection is active over this warm water and is a key source for atmospheric teleconnections into both hemispheres (Xie et al., 2016). The atmospheric teleconnection patterns due to the two major modes of Indian Ocean SST variability, the Indian Ocean Basin Mode (IOBM), and Indian Ocean Dipole (IOD) (Saji et al., 1999) have different geographical locations in terms of their center of actions compared to ENSO teleconnenctions (Cai et al., 2011). The IOBM can be defined as the first EOF of Indian Ocean SST anomalies, and the IOD is defined as the difference of the area mean SST anomalies in the western Indian Ocean (10°S–10°N, 50–70°E) and the eastern Indian Ocean (10°S–0°, 90–110°E) (Saji et al., 1999). While a positive (negative) phase of IOBM is characterized by a basin-wide warming (cooling) in the Indian Ocean, a positive (negative) phase of the IOD indicates an anomalous warm (cold) SST in the western Indian Ocean and an anomalous cold (warm) SST in the eastern Indian Ocean with a dipole-like structure in the zonal direction. The driving mechanisms for the IOBM and the IOD are different (Saji et al., 1999, Webster et al., 1999), though both exhibit a strong seasonality. That is, the peak intensity of the IOBM is observed in boreal winter, while the IOD peaks in boreal fall (An, 2004), leading a phase shift between IOBM and IOD (An, 2004; Ha et al., 2016).

ENSO impacts are modulated by the IOD in the Northern and Southern extratropics during boreal winter (Saji & Yamagata, 2003; Tao et al., 2015). For example, wave trains like the PNA and the PSA are prominent during El Niño in the boreal winter when accompanied by a positive IOD in the previous fall (Figures 9a and 9b). In contrast, atmospheric teleconnections across the Arctic and the Antarctic are observed in the Northern and Southern Hemispheres, respectively, in the case of El Niño accompanied by a negative IOD in the previous boreal fall (Figures 9c and 9d). However, the regions where a high level of statistical significance of atmospheric teleconnections are reached are very limited in the midlatitudes of both hemispheres (Figures 9c and 9d).

In addition, a positive IOD is associated with development of the East Indian Ocean wave train, resulting in an equivalent barotropic ridge over southern Australia, leading to less rainfall than usual over much of southern Australia during austral winter (Cai et al., 2011). Furthermore, the number of times El Niño occurs with a positive IOD is larger than the number of times when El Niño occurs with a negative IOD. In contrast, the number of frequency of La Niñas occurring with a negative IOD is larger than for La Niñas with a positive IOD (Table 1). There are clear asymmetries in the spatial pattern of rainfall in the tropical Pacific between an El Niño concurrent with a positive IOD (Figure 10a), and a La Niña concurrent with a negative IOD (Figure 10b) (Cai et al., 2011, 2012; Tao et al., 2015), resulting in an asymmetry in impact, including greater precipitation anomalies in the central tropical Pacific in El Niño events concurrent with a positive IOD (Figure 10c).

Table 1. The Year of El Niño and La Niña Events During Boreal Winter Accompanying With a Positive and Negative IOD in Previous Fall (September-October-November), Respectively

	El Niño during boreal winter	La Niña during boreal winter
Positive IOD in previous fall	1982/1983, 1986/1987, 1987/1988, 1991/1992, 1994/1995, 1997/1998, 2002/2003, 2004/2005, 2006/2007, 2015/2016	1983/1984, 1985/1986, 2007/2008, 2008/2009, 2011/2012
Negative IOD in previous fall	2009/2010, 2014/2015	1984/1985, 1988/1989, 1995/1996, 1996/1997, 1998/1999, 1999/2000, 2000/2001, 2005/2006, 2010/2011

• Note. El Niño (La Niña) refer to the years when the NIÑO3.4 SST index during boreal winter (December-January-February, DJF) is greater (less than) than 0.5°C (−0.5°C) in amplitude. The DJF NIÑO3.4 SST index is defined by time series of DJF mean SST anomaly averaged over the NIÑO3.4 region (170°W–120°W, 5°N–5°S). The IOD index is defined as the difference in SST anomaly between the tropical western Indian Ocean (50°E–70°E, 10°S–10°N) and the tropical southeastern Indian Ocean (90°E–110°E, 10°S to equator) during boreal fall (Saji et al., 1999). The seasonal mean anomaly is defined as seasonal mean deviations from a climatological (1979–2015) seasonal mean and a liner trend is removed. The SST data set is taken from the ERSST.v4. (Liu et al., 2014).

The South Asian High, which forms near the tropopause over the Indochinese peninsula before the Bay of Bengal monsoon onset in the late spring, is stronger than normal when El Niño occurs, leading to less precipitation in the following summer over East Asia (Wei et al., 2014). However, South Asian High is stronger when a positive IOBM in previous winter is concurrent with El Niño, while not as clear when not accompanied by an IOBM (Yang et al., 2007). Following an El Niño event in the previous winter, for example, a positive IOBM persists through boreal summer and leads to an increase of precipitation in the Indian Ocean. This leads to stronger than usual South Asian High through a Matsuno-Gill-type response (Gill, 1980; Matsuno, 1966), which is characterized by the quasi-steady response via the equatorial Rossby and Kelvin waves to a given tropical atmospheric heating in the Indian Ocean (Yang et al., 2007). Subsequently, this causes less precipitation over East Asia in summer as suggested in Sun et al. (2011). Similarly, ENSO characteristics including evolution, spatial pattern, and amplitude can be altered by interbasin interactions involving the IOBM, the IOD, and ENSO (Ha et al., 2016; Okumura et al., 2011), which modifies ENSO teleconnections. The pathway of ENSO teleconnections from the tropical western Pacific to East Asia can be altered by the state of IOBM, with a positive phase exciting a low-level anticyclonic circulation over the subtropical Northwest Pacific. This leads to changes in the centers of action in the Pacific-Japan-like atmospheric teleconnections forced by convective forcing related to ENSO, affecting precipitation variability in East Asia during summer by changing the transport of moisture (Ding et al., 2011).

4 Extreme Events Due To ENSO Teleconnections

This section reviews studies of the influences of ENSO on extreme events without consideration of whether events are of the EP or CP type, since there are few studies comparing the differing influences of EP and CP El Niños on these events. This limitation is mainly due to the limited availability of long-term daily observations over the globe, particularly precipitation data over ocean areas.

ENSO can affect the intensity and occurrence of extreme events such as heat waves, droughts, and floods (King et al., 2016). Using available indices to represent frequency and intensity of extremes, a few of studies have investigated global response patterns of extremes to ENSO over land areas where reasonably long observational records exist (Kenyon & Hegerl, 2008, 2010; Meehl, Tebaldi, et al., 2007). These studies find that temperature and precipitation extremes over many land areas are affected by ENSO teleconnections, showing statistically significant (5%) differences between El Niño and La Niña years and that the response patterns for extremes to ENSO basically follow the mean ENSO teleconnection patterns.

ENSO's influence on temperature extremes occurs globally, but most profoundly around the Pacific Rim and over North America (Arblaster & Alexander, 2012; Kenyon & Hegerl, 2008). Extreme maximum temperatures increase markedly during El Niño over India, Southeast Asia, Australia, and southern Africa while temperatures become cooler over southeastern North America, and vice versa during La Niña. Precipitation extremes, throughout the world, are also affected by ENSO although responses are less spatially coherent than those for temperature extremes (Kenyon & Hegerl, 2010). During boreal winter, El Niño induces statistically significant (5%) increases in precipitation extremes over southwestern North America, Central America, South America and India, while Australia, southern Africa, Southeast Asia, and Canada experience reductions in heavy precipitation (Figures 11a–11d). Extreme rainfall in northern Peru and Ecuador were particularly notable during the extreme El Niño events such as 1982/1983 and 1997/1998 (Pineda et al., 2013). La Niña brings generally opposite responses, and these response patterns of precipitation extremes generally follow the total precipitation responses.

Coupled climate models, which exhibit a wide range of behaviors in the simulation of ENSO events and thus varying teleconnection patterns (Meehl, Stocker, et al., 2007; Langenbrunner & Neelin, 2013), also indicate a close relationship between the responses of seasonal means and seasonal extremes. This is not unexpected, but there are no studies confirming this relationship in the observations. Figure 12 shows spatial correlation coefficients (bars) between the mean precipitation response and the extreme precipitation response to ENSO obtained from each CMIP5 model compared to that from the National Center for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis. Overall, high spatial correlations exist in all the models with multimodel mean values being very similar to the observed (r > 0.7). The high values of correlation indicate consistent teleconnection patterns between mean climate and extremes to ENSO for a given model. Studies of regional extremes related to ENSO teleconnections find a good agreement between the seasonal mean and seasonal extreme responses (Gershunov & Barnett, 1998). For example, drier and hotter conditions in both means and extremes occur over northeastern and southern Australia during El Niño, while wetter and cooler conditions prevail during La Niña (Min et al., 2013). This mean-extreme relationship is found for ocean wave heights over northeastern North Pacific with increases during El Niño and decreases during La Niña (Kumar et al., 2016). An opposite response occurs over the tropical western Pacific and Indian Oceans, in accord with ENSO's general impact on sea level pressure and the associated wind speed. This also supports the prominent impact of ENSO on coastal erosion and flooding across the Pacific (Barnard et al., 2015). Meehl, Tebaldi, et al. (2007) also showed that projections for the patterns of extremes over the U.S. during future El Niño events are expected to change because El Niño teleconnection patterns over the U.S. are projected to shift eastward and northward in a warmer climate. This mean-extreme relation in terms of ENSO teleconnections in both observations and models indicates the likelihood that the same physical mechanisms are at work. The strong physical link between seasonal mean and seasonal extreme response implies that reducing biases in mean responses may improve the simulation of extremes.

There remain large uncertainties, however, in understanding, evaluating, and predicting ENSO-induced changes in extreme events. Challenges include asymmetric nonlinear responses of extremes to El Niño and La Niña, influences of EP/CP El Niños on extremes in terms of amplitude, frequency, and duration, decadal modulation of ENSO impacts on extremes, and identification of global warming signals in extreme responses to ENSO teleconnections. With respect to nonlinear responses, impacts of extreme ENSO events on the strength and pattern of teleconnections need to be studied together with underlying physical mechanisms given that extreme El Niño/La Niña events are expected to occur more frequently after the mid 21st century. This expectation is based on coupled model simulations forced with Representative Concentration Pathways (RCP) 8.5 scenario (Cai, Wang, et al., 2015; Cai et al., 2014), where the RCP8.5 scenario denotes a greenhouse gas concentration trajectory with a radiative forcing of +8.5 W/m² in the year 2100 (IPCC, 2013). Overall, ENSO teleconnections influence both seasonal mean climate on interannual timescales and seasonal weather and climate extremes, with similar spatial patterns. This relationship provides important implications for evaluating the ability of models to simulate weather and climate extremes in the present climate and also for future climate projections (Arblaster & Alexander, 2012; Min et al., 2013).

5 ENSO Teleconnections Under Global Warming

Changes in ENSO teleconnections under global warning will be largely dependent on two factors, i.e., changes in the mean state both in the tropical Pacific and elsewhere around the globe and changes in ENSO properties such as its spatial pattern and amplitude. These two factors may lead to changes in both the intensity and the location of convective heating associated with ENSO, resulting in the changes in ENSO teleconnections (Cai, Santoso, et al., 2015; Dommenget & Yu, 2017). Furthermore, global warming can intensify the nonlinear tropical precipitation response to El Niño events, which may add more complexity of ENSO teleconnections in a warmer world (Power et al., 2013).

Analyses of Coupled Model Intercomparison Project Phase 3 (CMIP3) and Phase 5 (CMIP5) climate models under greenhouse warming show that models disagree considerably on future changes in ENSO properties (Cai, Santoso, et al., 2015; Collins et al., 2010; Ham & Kug, 2016; IPCC, 2013; Meehl, Stocker, et al., 2007; Zheng et al., 2016), but they generally indicate that warming of the background state on which ENSO events develop is likely to be faster in eastern tropical Pacific than in the western tropical Pacific. Because of the thermodynamic scaling of temperature and precipitation in a warmer climate (Held & Soden, 2006), the strength of the zonally asymmetric atmospheric circulation in the tropics (i.e., the Walker circulation) decreases in future climate experiments according to CMIP3 and CMIP5 climate models. This induces substantial changes to the thermal structure and circulation of the tropical oceans by decreasing the frequency of strong updrafts and increasing the frequency of weak updrafts, resulting in “El Niño-like” conditions in the tropical Pacific mean state (Vecchi & Soden, 2007).

There exists a relationship between the ENSO amplitude and the mean precipitation over the eastern tropical Pacific in CMIP5 climate models, i.e., a wetter mean state with greater precipitation acts to increase ENSO amplitude by strengthening positive coupled feedbacks in future climate (Watanabe et al., 2012). Also, tropical Pacific SST trends are projected to lead to substantial changes in the spatial patterns of ENSO-driven variability in surface temperature and precipitation of the tropical Pacific (Vecchi & Wittenberg, 2010). These changes are likely to lead to changes in ENSO teleconnections (Kug, An, et al., 2009; Meehl & Teng, 2007; Müller & Roeckner, 2008; Schneider et al., 2009; Stevenson, 2012; Zhou et al., 2014) as described next.

There is an expected eastward shift of the PNA teleconnection pattern under global warming (Kug, An, et al., 2009; Meehl, Stocker, et al., 2007; Zhou et al., 2014), resulting from a systematic eastward migration of convection centers associated with both El Niño and La Niña in the CMIP3 climate models (Müller & Roeckner, 2008; Power et al., 2013). In a warmer climate, precipitation anomalies are enhanced and they move eastward over the equatorial Pacific because the enhanced mean SST warming in the eastern basin reduces the barrier to deep convection there. Such an eastward shift in tropical convective anomalies can cause ENSO-forced PNA teleconnections to shift eastward and to intensify under greenhouse warming (Müller & Roeckner, 2008). There is also the possibility that in a warmer climate these teleconnections may contribute to altering the basic state in the midlatitudes, leading to a change in the intensity and the position of midlatitude Rossby waves associated with ENSO convective heating anomalies (Branstator & Selten, 2009). That is, global warming can cause a two-way interaction between the trend of atmospheric basic state and the modes of variability such as PNA. On the other hand, it has been suggested that El Niño teleconnections could weaken due partly to an increase in the static stability of the atmosphere in the midlatitudes (Schneider et al., 2009). In contrast, La Niña teleconnections under global warming are likely to strengthen the Aleutian low, although with a smaller mean eastward shift in the CMIP5 climate models than in the CMIP3 climate models (Stevenson, 2012).

Recently, studies with precipitation-based ENSO indices have suggested that the frequency of extreme El Niño and La Niña events will increase after the mid-21st century under the RCP8.5 scenario (Cai et al., 2014; Cai, Wang, et al., 2015), possibly causing more severe weather and climate variations under a very high CO₂ concentrations above 650 ppm. Furthermore, it has been suggested that extreme El Niño frequency increases linearly with the global mean temperature before 2050 toward a doubling at 1.5°C warming under the RCP2.6 scenario and that an increasing frequency of extreme El Niño events continues for up to a century after global mean temperature has stabilized (Wang et al., 2017). The dominant mode of global precipitation-evaporation variability on interannual timescales is mainly due to ENSO teleconnections on interannual timescales. Therefore, the increase in frequency of extreme ENSO events may cause more severe droughts and floods in a warmer climate, with largest increases in the tropical Pacific and polar regions (Christensen et al., 2013; Seager et al., 2012). It has also been noted that as the specific humidity of the atmosphere increases with warming according to the Clausius-Clapeyron relation, the same circulation patterns as a present could produce more moisture convergence and hence more rainfall; therefore, even in the absence of any changes in ENSO properties, there would be increase in ENSO-related precipitation anomalies (Power et al., 2013; Seager et al., 2012). However, there is also a general weakening of the tropical circulation, and the associated dynamical effect could offset the thermodynamic impact due to an increase in specific humidity (Chadwick et al., 2013; Huang & Xie, 2015).

Changes in tropical Pacific mean SST and ENSO properties under greenhouse warming will thus likely continue to alter ENSO teleconnections due to changes in mean state of the atmosphere over the globe as evident in the current CMIP3 and CMIP5-type class climate models. However, future ENSO teleconnection changes do not currently show strong intermodel agreement from region to region, which is mostly due to the uncertainty of model physics and thermodynamics in accounting for the sensitivity for the climate variability to global warming. Therefore, it is important to identify which factors affect uncertainty in future model projections.

6 Summary and Discussion

ENSO properties including the spatial pattern and amplitude and the ocean mean state outside the tropics have changed over the past few decades, which might be due to anthropogenic forcing, natural variability, or their combination. This has resulted in changes in ENSO teleconnections, as manifested by changes in the centers of action of associated anomalous atmospheric circulation in the extratropics. These changes in ENSO teleconnections also lead to a different response in precipitation, temperature, or both around the globe. In addition, changes in ENSO teleconnection patterns have affected the statistics of extreme events in temperature and precipitation and the predictability of teleconnection impacts. ENSO teleconnections differ in detail from event to event. They can be also influenced by conditions outside the tropical Pacific Ocean and modified by a longer timescale climate phenomenon. Furthermore, nonlinearities in global ENSO teleconnections strongly complicate the investigation of ENSO-related climate variability. Van Oldenborgh and Burgers (2005) also propose that precipitation variability due to ENSO teleconnections could be explained by random fluctuations on decadal timescales.

ENSO teleconnections are not stationary and their predictability is not high except for certain regions such as Indonesia, eastern Australia, and eastern Africa, thus representing one of the key challenges for weather and climate prediction. For example, the 2015/2016 El Niño event was one of the strongest El Niño in recent history, rivaling the previous two strong El Niño events in 1982/1983 and 1997/1998. However, the temperature and precipitation anomalies due to the 2015/2016 El Niño differed from the expectation associated with previous strong El Niño events (Jong et al., 2016, Nicolas et al., 2017, Kumar & Chen, 2017). For example, the very strong 1982/1983 and 1997/1998 El Niños brought heavy rains to California during the winter season, which was not the case in the 2015/2016 El Niño. On the other hand, unlike during the strong 1982/1983 and 1997/1998 El Niños, the unusual extent and duration of the melting in west Antarctica was likely favored by the strong 2015/2016 El Niño event.

It should be also noted that the statistical significance of the changes in precipitation, temperature, and extreme events due to the changes in ENSO teleconnections is marginal in many regions. There is still a debate about the robustness of the changes in ENSO properties and its teleconnections before and after the late 1990s because of the short observational record (Fedorov & Philander, 2000; McPhaden et al., 2011; Stevenson et al., 2010; Wittenberg, 2009). It has also been pointed out that it may be more appropriate to use atmospheric parameters more directly related to forcing of teleconnections such as OLR, which is a measure of deep atmospheric convection, rather than conventional SST indices to assess the far-field impacts of ENSO (Cai et al., 2014; Chiodi & Harrison, 2013, 2015). Better understanding of how large-scale environmental conditions during ENSO events determine OLR and deep atmospheric convection will lead to improved seasonal precipitation forecasts for many areas of the globe. A further complicating factor is nonlinearity in ENSO teleconnections, which result from a combination of nonlinearities in ENSO SST variability and nonlinearities in the atmospheric response to a given ENSO SST anomaly (Frauen et al., 2014).

To understand ENSO teleconnections under global warming, it is crucial to realistically simulate changes in the mean state of the tropical Pacific Ocean and the changes in ENSO properties including spatial pattern and amplitude. These factors are closely associated with the changes in the intensity and the location of tropical convective forcing, which ultimately determine the characteristics of ENSO teleconnections. In particular, the current CMIP-class climate models still suffer from the double Intertropical Convergence Zone (ITCZ) syndrome, which is characterized by a mean precipitation pattern that is too symmetric near the equator associated with an overly zonal SPCZ (Bellucci et al., 2010; Li & Xie, 2014; Lin, 2007). Such mean state bias affects ENSO mechanisms, variability, and asymmetry (Bellenger et al., 2013, Guilyardi et al., 2004, 2009, Lloyd et al., 2011) in the Pacific and also the nature of ENSO teleconnections in the midlatitudes. Simulation of the mean precipitation pattern in the ITCZ/SPCZ is tightly linked to the ability of models to correctly simulate the pattern and magnitude of mean SST in the tropical Pacific. In addition, the CMIP-class climate models have a persistent warm bias in eastern boundary current systems (Richter, 2015), which has a large influence on regional air-sea interactions and ENSO teleconnections. Therefore, improvements in model realism for these features are likely to lead to improvements in the realism of simulated ENSO teleconnection patterns and in the confidence of future regional responses to greenhouse warming.

Erratum

In the originally published version of this article, the reference and DOI for Schemm et al. were incorrect. It has since been replaced with the correct reference and DOI (https://doi.org/10.1175/JAS-D-16-0072.1) and this version may be considered the authoritative version of record.