The Asymmetry of the El Niño–Southern Oscillation: Characteristics, Mechanisms, and Implications for a Changing Climate

Abstract

The El Niño–Southern Oscillation (ENSO) is inherently asymmetric, a primary characteristic where its warm phase (El Niño) and cold phase (La Niña) differ in amplitude, spatial pattern, and temporal evolution. This review synthesizes over two decades of research to provide a comprehensive overview of ENSO asymmetry. It systematically examines the observed manifestations, evaluates the competing physical mechanisms, and analyzes the ongoing challenges in climate modeling. The key findings in the literature indicate that this asymmetry is driven by complex interactions of nonlinear processes, where atmospheric mechanisms such as state-dependent westerly wind bursts and threshold responses of deep convection are now considered dominant driving factors, which are subsequently amplified and modulated by oceanic feedback. The main challenge in this field is that most of the current state-of-the-art climate models underestimate ENSO asymmetry, which is related to mean-state bias and brings uncertainty to future predictions. Furthermore, a key finding from recent projection studies is that while the asymmetry in ENSO’s sea surface temperature is expected to weaken in a warmer climate, the asymmetry of its global rainfall impacts may paradoxically be amplified. Future research should focus on balanced improvements in ocean and atmospheric model components, development of new diagnostic tools to clarify the roles of different feedbacks, or establishment of a framework that clearly links asymmetry to the full spectrum of ENSO diversity. By consolidating the current state of knowledge and highlighting key unresolved questions, this work provides an essential roadmap to improve the prediction and projection of Earth’s most far-reaching mode of climate variability.

1. Introduction

The El Niño–Southern Oscillation (ENSO) is the strongest and consequential mode of natural climate variability on interannual timescales [1]. It is characterized by irregular oscillations in sea surface temperatures—the oceanic component known as the El Niño (warm phase) and La Niña (cold phase) cycle—and a corresponding large-scale shift in atmospheric pressure between the eastern and western Pacific, known as the Southern Oscillation. This phenomenon originates from the complex coupling between the ocean and atmosphere in the tropical Pacific region, which radiates globally through the atmospheric teleconnection network and has profound impacts on weather patterns, ecosystems, and society far beyond its equatorial source area [2]. The classic perspective of this cycle is dominated by the positive Bjerknes feedback loop: a slight increase in sea surface temperature (SST) in the eastern equatorial Pacific weakens the easterly trade winds, thereby reducing cold water upwelling from lower-level seawater and leading to further warming. This feedback mechanism drives the occurrence of El Niño events. The oscillation is eventually terminated and reversed by negative feedbacks, such as the discharge of equatorial heat content by oceanic waves, which cause the system to swing into the opposite phase [3,4]. On the contrary, the cooling of the East Pacific SST will enhance trade winds, promote cold water upwelling, and thus strengthen the La Niña state. This fundamental oscillation leads to significant interannual fluctuations in global temperature, precipitation, and frequency of extreme weather events, making it a core element in seasonal-to-interannual climate prediction. A schematic overview of the distinct ocean–atmosphere states during El Niño and La Niña, along with the historical time series of the oscillation, is provided in Figure 1.

While the concept of warm-cold oscillations provides us with a basic framework, it has become very clear that El Niño and La Niña are not simply mirror images. Their differences in amplitude, spatial distribution, temporal evolution, and global impact are collectively referred to as ENSO asymmetry [5]. This asymmetry is not just a subtle statistical difference, but a fundamental characteristic that reveals the dominant role of nonlinear processes in coupled climate systems. Understanding these nonlinearities is essential, as they lead to the diverse “flavors” of ENSO events and the inherent irregularity that makes their prediction a persistent challenge [6]. In recent years, people have increasingly recognized that certain aspects of ENSO asymmetry have been strengthened, which highlights the necessity of understanding its intrinsic dynamics and potential response to climate change. The broad features of the ENSO phenomenon and its global impacts have been documented in several comprehensive reviews [1,7].

The manifestations of this asymmetry are diverse and profound. The most widely documented difference is amplitude, where strong El Niño events typically have larger magnitude than the strongest La Niña events. This feature is often quantified by the positive skewness of the probability distribution of SST anomalies in the Eastern Pacific [8]. The temporal evolution of the two phases also differs markedly, as illustrated by the historical behavior of the ENSO index (Figure 1c). La Niña events often last for many years—a phenomenon now known as “multi-year La Niña”—while El Niño events rarely exhibit this characteristic, typically transitioning rapidly to a cold or neutral state after the peak [9,10]. The increasing frequency of La Niña events in recent years has a significant impact on global prolonged drought and flood patterns [11]. This asymmetry in duration is closely related to the that in phase transition. The transition from El Niño state to La Niña state is a common pathway, while the reverse transition is relatively rare [12].

The spatial and propagational characteristics of ocean anomalies also have fundamental asymmetry. The peak warming during strong El Niño events is often located further east, near the coast of South America, than the peak cooling during La Niña, which is typically centered further west in the central Pacific. Furthermore, a striking interdecadal asymmetry has been observed in the propagation of SST anomalies. While La Niña cooling has consistently maintained a robust westward propagation from the eastern Pacific across decades, the behavior of El Niño has changed significantly [13]. Prior to the climate regime shift of the late 1970s, El Niño warming also tended to propagate westward, consistent with the canonical patterns of that era [14]. In the period since, however, strong El Niño events are now often characterized by a clear eastward propagation from the western Pacific. This asymmetry is not limited to the sea surface, it has a clear three-dimensional structure and exhibits different patterns in the subsurface ocean thermocline [15]. The key is that the remote impacts of the ENSO are also asymmetric. The response of subtropical atmosphere to warm and cold events is often unequal and opposite. For example, the well-known Pacific North America (PNA) teleconnection mode—the main channel through which the ENSO affects North American climate—exhibits asymmetric effects, and the amplification effect in mid-to high-latitude regions is significantly stronger than the initial asymmetry of tropical SST [16]. Similar asymmetry also exists in the stratospheric pathway, where moderate intensity El Niño and strong La Niña events are more efficient in modulating the stratospheric polar vortex than their counterparts [17].

Explaining the physical origins of these asymmetric phenomena has become a core focus of research on the ENSO. The relevant hypothesis suggests that there are a series of nonlinear mechanisms in the ocean and atmosphere. The asymmetry in event duration, for example, is closely linked to asymmetric oceanic recharge/discharge processes. On average, the oceanic discharge of heat following an El Niño is stronger than the subsequent recharge process, and this is further compounded by off-equatorial Rossby waves during La Niña that can inhibit the equatorial recharge process, thus favoring the persistence of the cold state [18]. In addition, other marine processes, such as nonlinear temperature advection (i.e., anomalous ocean currents transporting anomalous temperatures), are considered crucial for generating stronger warming events [8]. In the subsurface, nonlinear dynamic heating has been shown to weaken subsurface cooling after strong El Niño events, thereby weakening subsequent La Niña events and affecting their intensity and structural asymmetry [19]. At the same time, atmospheric mechanisms have gradually become a potential dominant contributing factor. These mechanisms include state-dependent forcing caused by sporadic westerly wind bursts (WWBs), which are more frequent and intense during warm periods and therefore tend to amplify El Niño events [20,21]. In addition, the highly nonlinear sensitivity of deep atmospheric convection to SST, which is much more sensitive to warming above a certain threshold than other conditions, is now considered by many researchers as the main cause of ENSO asymmetry [22,23].

The modern paradigm of ENSO diversity, which distinguishes between Eastern Pacific (EP) and Central Pacific (CP) “flavors” of El Niño, is essentially closely related to this asymmetry. The mechanisms that generate these different types are inherently nonlinear, and the events themselves contribute differently to the overall asymmetry of the ENSO cycle [6,24]. For example, the most extreme El Niño events typically belong to the EP type, while multi-year La Niña events often follow closely behind, highlighting the deep connection between ENSO spatial and temporal asymmetry.

Despite significant progress, accurately characterizing these nonlinear processes remains a major challenge for state-of-the-art climate models. The successive generations of the Coupled Model Intercomparison Project (CMIP) have shown persistent biases, with most models systematically underestimating the observational intensity of ENSO asymmetry [25,26]. This persistent model deficiency is often associated with other long-standing model errors, such as the “cold tongue bias” in the tropical Pacific mean state [27,28] and the inability to correctly simulate key feedback processes [29,30]. The gap between observations and models not only limits the understanding of fundamental mechanisms, but also introduces significant uncertainty into predicting how the ENSO and its global impacts evolve in a warming world [31]. Some studies predict that ENSO asymmetry may weaken in the future as the background conditions warm [32,33], but this issue remains highly controversial.

While the foundational reviews cited above provide excellent overviews of the general ENSO phenomenon, the specific topic of ENSO asymmetry has grown into a major subfield of its own, with a rapid increase in dedicated research over the past two decades. To our knowledge, a comprehensive synthesis focused squarely on the characteristics, mechanisms, and implications of this asymmetry has been lacking. To address this gap, the primary objectives of this review are threefold: (1) to systematically catalog the observed characteristics of ENSO asymmetry, from its signature in the tropical Pacific to its global teleconnections; (2) to critically evaluate the current understanding of the competing physical mechanisms that generate this asymmetry; and (3) to assess the persistent challenges in simulating ENSO asymmetry in state-of-the-art climate models and the implications for future climate projections. To achieve these objectives, this review is structured as follows. Section 2 systematically organized the observational characteristics of ENSO asymmetry, from characteristic signals in tropical Pacific SST to its asymmetric teleconnection. Section 3, the core of this study, delves into the physical mechanisms and provides a detailed overview of the competing and complementary ocean and atmospheric driving factors that are believed to generate this asymmetry. Section 4 evaluates the ongoing challenges faced by climate models in simulating these features, linking common model biases to the misidentification of key nonlinear processes. Section 5 explores how the modern ENSO diversity paradigm is intrinsically related to its asymmetric properties. Finally, Section 6 consolidates the current understanding of how ENSO asymmetry could evolve under anthropogenic warming, summarized in Section 7 by highlighting key unresolved questions and future research directions.

This article is a narrative review aimed at systematically reviewing and critically evaluating the progress in research on ENSO asymmetry over the past two decades. To ensure a comprehensive and representative examination of the field, we conducted a structured literature search through major scientific databases, primarily Web of Science and Scopus. The core search keywords include “ENSO asymmetry”, “El Niño diversity”, “nonlinear ENSO”, “multi-year La Niña” and “ENSO teleconnection asymmetry”, and the search scope is further expanded by consulting the reference list of basic literature. A broad search for articles with “ENSO” in the title from 2000–2025 reveals a vast and growing body of international research, with major contributions from numerous countries, particularly the United States and China (Figure 2). Our review synthesizes the most influential and foundational papers from this global effort. The main inclusion criteria are the physical mechanisms, observational features, modeling, and future predictions directly related to ENSO asymmetry, with a focus on integrating the most influential and highly cited papers to construct a coherent scientific narrative. Although this review aims to comprehensively elucidate the core physical sciences, the in-depth analysis of ENSO asymmetric biogeochemical or socio-economic impacts goes beyond the scope of research, which is a key limitation. Comprehensive information will be presented by theme in subsequent chapters.

2. The Manifestations of ENSO Asymmetry

The ENSO is essentially asymmetric. Although early conceptual models viewed this phenomenon as a linear periodic oscillation, decades of observational evidence have revealed a much more complex and detailed picture. El Niño and La Niña are far from perfect opposites. They differ systematically in amplitude, spatial structure, temporal evolution, and have profound impacts on the global climate system. These asymmetries are not insignificant statistical anomalies, but rather surface manifestations of profound nonlinearities inherent to the coupled ocean–atmosphere dynamics of the tropical Pacific [5,34]. Cataloging these diverse manifestations is a crucial first step toward understanding the physical mechanisms that drive them. This section will summarize the core observational characteristics that define ENSO asymmetry, providing an empirical basis for the subsequent content of this review.

2.1. Amplitude Asymmetry

As introduced, the most well-known and fundamental feature of ENSO asymmetry is the difference in intensity between its warm and cold phases. Observation records consistently indicate that the strongest El Niño event produce significantly greater SST anomalies than the strongest La Niña event in the eastern equatorial Pacific [35,36]. This difference is statistically quantified as the positive skewness of the probability distribution of SST anomalies in key monitoring areas, such as Niño-3 region [37]. While the mean and variance of SST anomalies provide a first-order description of the ENSO, skewness reveals a key third-moment characteristic: the tail of extreme positive (warm) events is longer than that of negative (cold) events. This positive skewness is most significant in the far-eastern Pacific and gradually weakens towards the west, highlighting its connection with the dynamics of the cold tongue in the East Pacific.

This amplitude asymmetry exhibits robust characteristics across different observation datasets and time periods, despite the existence of decadal variability in its intensity [5]. The significance of this skewed distribution is profound: it constitutes the main evidence that the ENSO cannot be fully described by linear systems driven solely by random weather noise. A purely linear system, by definition, should produce a symmetric Gaussian anomaly distribution [38]. The observed positive skewness is therefore a clear indication of underlying nonlinear physical processes, which preferentially amplify the warm phase of oscillations [34,35]. The fundamental asymmetry of SST amplitude extends to other key variables in the coupled system. For example, the atmospheric response is not symmetrical. The zonal wind-stress response to warm SST anomalies was observed to be stronger than that of cold SST anomalies of the same magnitude [39]. Similarly, precipitation response is highly asymmetric. In the typically colder eastern Pacific, positive SST anomalies can easily trigger deep convection, while negative SST anomalies in this region have limited impact on the already dry conditions [40]. This fundamental asymmetry in magnitude lays the foundation for other asymmetric behaviors of the ENSO, from its spatial expression to its global teleconnections.

2.2. Spatial and Propagational Asymmetry

In addition to simple amplitude differences, the spatial distribution of warm and cold events also has fundamental asymmetry. The geographical distribution of SST anomalies during El Niño events does not simply invert that of La Niña. The key difference between the two lies in their zonal location and meridional extent. Strong El Niño events typically exhibit the largest SST anomalies in the far eastern Pacific region, and are often concentrated along the coasts of South America. In contrast, the strongest cooling during intense La Niña episodes tends to peak farther westward, in the central Pacific [36]. This spatial difference is a major component of ENSO diversity, where the EP El Niño event differs in mechanism and structure from the CP El Niño event. This asymmetry arises from the fact that the strongest El Niño events almost always belong to the EP type, while strong La Niña events often exhibit a CP type structure [36,41].

This asymmetry extends into the subsurface ocean, forming a unique three-dimensional structure. The depth and zonal extent of thermocline anomalies are asymmetric between different stages. During a strong El Niño event, the thermocline significantly deepens and covers a broad region, while during a strong La Niña event, the corresponding thermocline becomes shallower, usually with a smaller amplitude and centered further west [15]. The sum of composite warm and cold subsurface temperature anomalies, highlighting systematic differences that cannot be captured by linear analysis. This subsurface asymmetry is crucial as it directly affects the key oceanic feedback intensity that controls the evolution of ENSO events. Asymmetry is also apparent in sea level anomalies, with the sea level fall in the western tropical Pacific during El Niño being significantly stronger than the sea level rise during La Niña [42].

One of the most significant manifestations of asymmetry is the zonal propagation of SST anomalies. Prior to the climate regime shift of the late 1970s, both El Niño and La Niña events were characterized by westward propagation. However, there is a clear asymmetry in the subsequent period: El Niño events, especially strong El Niño events, often exhibit warm anomalies spreading eastward from the western or central Pacific. In sharp contrast, La Niña events continue to show a stable and strong trend of cold anomalies spreading westward from the eastern Pacific [13]. This dynamical divergence is a fundamental break in the mirror-image paradigm, which cannot be explained by simple linear oscillator theory. It points to phase-dependent processes that either promote or suppress the propagation of anomalies in specific directions. Subsequent research has confirmed that this propagation asymmetry is a true characteristic of the climate system, rather than an artifact of changes in the observational network [43].

2.3. Temporal Asymmetry

The lifecycle of ENSO events reveals significant spatiotemporal asymmetry in their duration and transition from one stage to another. Over a century of observational records have shown that La Niña events have a stronger tendency to persist compared to El Niño events [9]. Although most El Niño events develop, peak, and subside within a year, a notable fraction of La Niña events persist for two to three consecutive years. These “multi-year” La Niña events do not occur randomly, but are systematic features of ENSO cycles. Since the 1970s, the multi-year La Niña events have occurred more frequently, with five out of six events since 1998 lasting for two to three years. Understanding their dynamics has become an urgent research priority, as they are associated with some of the most severe and long-lasting drought and flood events on a global scale [16].

The asymmetry of this duration is inherently related to the asymmetry of the phase sequence. The progression from mature El Niño into the following year’s La Niña event is a very common evolutionary path, representing the classic “discharge” stage in recharge-discharge oscillator models [44]. The strong oceanic heat release in the equatorial Pacific often “overshoot” to a neutral state after major El Niño events, directly triggering La Niña events. However, the reverse path is much rarer. Mature La Niña events are unlikely to be immediately replaced by El Niño events, but show a greater propensity to terminate in a neutral state or continue until the following year [39]. The unidirectional preference in this phase transition process is a clear manifestation of nonlinear phenomena. This indicates that the oceanic and atmospheric feedback mechanisms that control event termination operate differently and with varying efficiencies during the warm and cold phases. The mechanism that leads to this sequence asymmetry is believed to involve the asymmetric wind-stress sensitivity to SST anomalies, which generates stronger and faster negative feedback during the regression of El Niño evolution [39].

2.4. Asymmetric Teleconnections and Global Impacts

The asymmetry of the ENSO is not confined to the tropical Pacific, but can spread globally, resulting in asymmetric remote impacts that are often amplified relative to the initial forcing. During ENSO events, atmospheric Rossby waves emanated from tropical regions form the basis of teleconnection, and these waves themselves also exhibit asymmetry. For example, the response of winter circulation in the Northern Hemisphere to EP La Niña events is significantly different from the response to EP El Niño events [41]. During extreme El Niño events, a strong PNA pattern is formed. During the EP La Niña event, the response of PNA type significantly weakened, but a unique and significant teleconnection pattern appears on the Atlantic-Eurasian region, which is completely absent during the El Niño event. This indicates that different stages of the ENSO can activate fundamentally different atmospheric pathways, a feature that cannot be captured by linear analysis. These teleconnection nonlinear features have been recognized in early foundational research [45].

The asymmetry of atmospheric pathways has a profound impact on regional climate anomalies. For example, the well-documented influence of the ENSO on rainfall in Australia is highly asymmetric. The teleconnection pathway of El Niño on the drying effect on southeastern Australia is mainly an indirect pathway, mediated by the Indian Ocean Dipole (IOD). In contrast, the path of La Niña wetting effect is more direct, originating from the Pacific via the Pacific South America (PSA) pattern [46]. This path difference gives rise to differences in the spatial distribution and intensity of precipitation anomalies. In addition, the rainfall response itself has nonlinear characteristics. The strength of La Niña episodes exerts a more pronounced influence on extreme rainfall intensity across eastern Australia than the intensity of El Niño events [47].

Furthermore, the strength of the teleconnection can be asymmetric even when the pathway is the same. In the stratosphere, the influence of the ENSO exhibits both nonlinear and asymmetric characteristics. Observational analysis shows that moderate intensity El Niño events and strong La Niña events are more effective in warming and cooling the polar stratosphere in the northern hemisphere during winter than their strong El Niño and moderate La Niña events [17]. This complex relationship indicates that the extratropical response is not simply proportional to the intensity of tropical SST anomalies, but rather depends on its sign and intensity in a nonlinear fashion. The rich tapestry of asymmetric manifestations, from the amplitude of SST anomalies in the tropical Pacific to the intensity of polar vortex in the stratosphere, highlights the complexity of ENSO phenomena and provides a rigorous observational benchmark that any complete ENSO theory or realistic model of the ENSO must be able to explain.

3. Physical Mechanisms Driving ENSO Asymmetry

3.1. The Central Role of Nonlinearity

The existence of ENSO asymmetry is a clear signature of the nonlinear properties of coupled ocean–atmosphere systems. While linear conceptual models, such as classical delay oscillators, have played an important role in explaining the fundamental oscillation behavior of the ENSO, these models are inherently symmetrical and cannot explain the significant differences between El Niño and La Niña events alone [5]. Deviation from this symmetrical oscillation behavior—manifested as amplitude differences, temporal irregularities, and skewed statistical distributions—directly points to the influence of physical processes that do not respond linearly to the sign or magnitude of SST anomalies.

Therefore, understanding ENSO asymmetry requires identifying and quantifying key nonlinear processes within the tropical Pacific climate system. Research over the past two decades has shown that these nonlinear features are not limited to a single component, but are present in the ocean, atmosphere, and most critical coupling feedback mechanisms. The main mechanisms can be broadly categorized into three areas: (1) the nonlinear response of the atmosphere to SST forcing, such as the threshold behavior of deep convection and the state dependence of random wind events; (2) nonlinear ocean dynamics, including the temperature advection by anomalous ocean currents (nonlinear temperature advection) and asymmetric feedbacks involving the thermocline; (3) the modulation of these nonlinear processes by the background climatological mean state of the tropical Pacific. Although early theories often emphasized oceanic advection [8], recent evidence suggests that atmospheric processes may be the primary source of asymmetry, subsequently shaped and amplified by oceanic dynamics. This section will systematically review the evidence supporting these distinct but interconnected mechanisms.

3.2. Atmospheric Drivers of Asymmetry

The response of the atmosphere to underlying SST anomalies is far from linear. This nonlinearity is the primary source of ENSO asymmetry, which introduces asymmetry in wind, heat flux, and precipitation patterns that force the ocean. Three key atmospheric mechanisms have been identified as major contributors: state-dependent forcing from WWBs, threshold-like response of deep convection to SST, and asymmetric cloud-radiation feedback.

Westerly wind bursts are short-lived, high-frequency relaxations or reversals of the easterly trade winds in the western and central equatorial Pacific. They are a form of stochastic “noise” in the climate system that can serve as a powerful trigger or amplifier for El Niño events [21]. There is fundamental asymmetry in their nature: they are westerly anomalies. Therefore, they can generate downwelling ocean Kelvin waves propagating eastward, deepening the thermocline and promoting surface warming, thereby initiating or strengthening El Niño events. However, they cannot produce the opposite initiate to trigger or amplify La Niña events.

The impact of WWBs is further amplified by their state-dependent nature. Observation and modeling studies consistently indicate that the frequency and amplitude of WWBs are not random, but modulated by the background SST state [20]. When the western Pacific warm pool anomalously warms and expands eastward, the likelihood of significant WWBs increases, which is a characteristic of developing or existing El Niño events. This forms a positive feedback loop: initial warming makes WWBs more likely to occur, and these WWBs further drive warming. Levine et al. [20] demonstrated through a conceptual model that this state-dependent noise forcing is sufficient to generate realistic El Niño-La Niña amplitude asymmetry, leading to stronger and more extreme El Niño events. This mechanism also provides a pathway for ENSO diversity, as the specific time, location, and intensity of WWBs can lead to different event evolutions, contributing to the distinction between EP and CP El Niño events [21]. The asymmetric forcing of WWBs not only contributes to amplitude asymmetry, but also affects time asymmetry, as the strong oceanic response they trigger can lead to rapid discharge of warm pools and subsequent rapid transition to La Niña conditions. The incorporation of realistic WWB parameterization schemes has been proven to be crucial for improving the simulation of ENSO asymmetry and diversity in climate models [48].

While WWBs represent the stochastic components, perhaps the most fundamental nonlinearity in the atmosphere is the deterministic response of deep convection to the underlying SST. Deep atmospheric convection does not increase linearly with SST. On the contrary, when the water temperature is cooler than approximately 27.5 °C, deep convection is strongly suppressed, and once this threshold is crossed, deep convection will increase dramatically and nonlinearly [22]. This threshold behavior is the primary driver for ENSO asymmetry.

Consider the cold eastern equatorial Pacific in terms of climatology. During strong El Niño events, SST can rise by 3–4 °C, pushing vast areas of the ocean surface well above the convective threshold. This triggered a large-scale nonlinear increase in atmospheric heating and precipitation, resulting in a strong westerly response that further amplifies the warming through the Bjerknes feedback. In contrast, during a strong La Niña event, the decrease in SST is similar, but due to the water temperature being below the convective threshold, the reduction in convection is relatively small and linear. Therefore, the response of the atmosphere to the −3 °C SST anomaly is much weaker than that to the +3 °C anomaly.

Recent studies suggest that this nonlinear convective feedback is the main mechanism leading to the genesis of ENSO asymmetry. Geng et al. [22] concluded that nonlinear atmospheric convection feedback alone can explain the observed amplitude and evolution asymmetries by employing a recharge oscillator model modified with various nonlinearities. In a more recent study, Srinivas et al. [23] used a series of atmospheric and oceanic general circulation models to isolate the linear and nonlinear components of wind stress response. Their results confirm that this nonlinearity (mainly governed by the SST-driven response of deep atmospheric convection) explains the majority of the positive ENSO skewness in the Eastern Pacific. This mechanism also naturally explains the spatial asymmetry of precipitation, as the center of anomalous convection shifts much farther east during El Niño than the center of suppressed convection shifts during La Niña [49].

The third, more subtle atmospheric mechanism involves asymmetric feedback from clouds and radiation. The changes in cloud cover associated with El Niño and La Niña are not symmetric, resulting in asymmetric modulation of surface radiation budget. Chen et al. [50] studied this process using observational data and CMIP5 simulations. They found that the response of shortwave and longwave cloud-radiative forcing to changes in SST is weaker during the cold La Niña phase than during the warm El Niño phase.

During El Niño, the eastward movement of deep convection is associated with a decrease in the climatological stratus cloud decks over the eastern Pacific, resulting in a significant increase in incoming solar radiation (positive feedback, amplification effect). During La Niña, the enhancement of these stratus decks is weak, resulting in weak negative feedback. The asymmetric feedback in this cloud-radiation-SST interaction contributes to overall asymmetry, although it is generally considered a secondary effect relative to the direct nonlinear effects of wind stress forcing and convective heating. The misrepresentation of these low-level cloud feedbacks remains a persistent source of bias in many climate models, further complicating the accurate simulation of ENSO asymmetry.

3.3. Oceanic Drivers of Asymmetry

Although atmospheric nonlinearity may be the main source of ENSO asymmetry, the dynamical and thermodynamical responses of the ocean are also crucial in shaping the magnitude, structure, and evolution of ENSO asymmetry. Ocean processes can amplify initial atmospheric asymmetry, introduce their own nonlinearity, and determine the spatiotemporal structure of ENSO cycles, such as the persistence of La Niña.

One of the earliest and most widely studied mechanisms of ENSO asymmetry is nonlinear temperature advection (NTA). This process refers to the advection of the anomalous flow field on the average temperature gradient, and more importantly, the advection of the anomalous flow field on the temperature anomaly (the u’T’ term). Previous studies suggest that NTA is stronger during El Niño, thus contributing to its larger amplitude. Duan et al. [8] proved that the NTA, particularly nonlinear zonal advection, critically amplifies El Niño magnitude, while having a relatively small impact on La Niña. This asymmetry arises from the systematic positive product of zonal ocean current anomalies and zonal temperature anomalies during the warm phase, thereby strengthening the warming process.

Recent research has extended this concept to the subsurface. Hayashi and Jin [19] emphasized the crucial role of subsurface nonlinear dynamical heating (NDH). They pointed out that after a strong El Niño peak, sharp positive NDH anomalies will develop along the thermocline, which will weaken the linear advective subsurface cooling. This process significantly weakens the intensity of the subsurface cold signal as they reach the eastern Pacific, thereby weakening the subsequent La Niña event. This mechanism directly leads to amplitude asymmetry (strong El Niño events followed by a weaker than expected La Niña events), as well as the structural asymmetry between the phases. Many climate models have failed to capture this strong NDH phenomenon, a key reason for their poor simulation of ENSO asymmetry [51].

The core dynamical feedbacks of the ENSO cycle, thermocline feedback and zonal advective feedback, are themselves asymmetric. Thermocline feedback links changes in SST to upwelling of water in the thermocline, which is modulated by ocean wave dynamics. The zonal advective feedback associates SST changes with anomalous zonal advection of the average east–west temperature gradient. Guan et al. [52] quantified these feedbacks and found that the positive oceanic feedback responsible for event growth was significantly stronger during El Niño than during La Niña.

This asymmetry is partly related to the mean state of the ocean. The mean thermocline is deeper in the west Pacific and shallower in the east. When El Niño occurs, the eastern thermocline deepens further, but its sensitivity to wind stress is more robust. When La Niña happens, the eastern thermocline becomes shallower, and when it becomes very shallow, the response to wind stress forcing may become more efficient to lead to a stronger feedback. This phenomenon of “shallow thermocline” contributes to the persistence of La Niña events. The persistence of subsurface temperature anomalies from one season to the next is modulated by the average seasonal period of the mixed layer depth and contributes to the asymmetry of event evolution [53]. In addition, as shown by Dewitte et al. [54], the long-term enhancement of thermocline feedback in the western-central Pacific thermocline after the 1976 climate change contributed to the observed changes in ENSO characteristics.

Other ocean processes also contribute to asymmetry. Tropical Instability Waves (TIWs) are westward-propagating waves originating from the instability of equatorial ocean current systems. They cause significant meridional heat mixing and are the main source of heat for the equatorial cold tongue. The activity of TIWs is highly dependent on the system state: active under La Niña and neutral conditions, but strongly suppressed during El Niño [55]. This leads to asymmetric negative feedback. During the La Niña event, strong TIWs transports warm water towards the equator, suppressing cold anomalies. During El Niño events, the absence of TIWs means that this damping mechanism is shut down. This asymmetric damping mechanism results in a larger amplitude of El Niño events than La Niña events [30].

Salinity changes also play an asymmetric role. During El Niño, an increase in rainfall in the central and eastern Pacific creates a fresh, buoyant surface layer, which may strengthen stratification and retain warm water near the surface. During La Niña, a decrease in rainfall has the opposite effect. Guan et al. [56] pointed out that these asymmetric salinity anomalies further amplify the temperature asymmetry between El Niño and La Niña by affecting vertical mixing and stratification. Beyond this modulating role, some studies have also suggested that salinity variability can play an active part in initiating events, with positive salinity anomalies in the upper-central Pacific potentially acting as a trigger for the development of a La Niña event [57].

3.4. The Influence of the Mean State

The various nonlinear atmospheric and oceanic mechanisms mentioned above do not operate in isolation. Their efficiencies largely depend on the background climatological mean state of the tropical Pacific. The mean east–west SST gradient, the depth and sharpness of the thermocline, and the size of the western Pacific warm pool collectively form the basis for the development of ENSO events. Therefore, the average state deviation simulated by climate models is the main reason why they cannot truly reproduce ENSO asymmetry.

A large number of studies have linked the common “cold tongue bias” in climate models—the non-existent westward expansion of the equatorial cold tongue in many models—to the weakening and excessive symmetry of the ENSO. Sun et al. [27,28] pointed out that the smaller and warmer warm pools in the CMIP3 and CMIP5 models, combined with this cold tongue bias, push the deep convection region excessively westward. The westward movement of this convection weakens the nonlinear air–sea interaction, limiting the westerly anomaly to the western Pacific, thereby preventing the development of strong asymmetric El Niño events with characteristics of the eastern Pacific. The latest study by Bayr et al. [58] confirms that this bias still exists in the CMIP6 model and is a key reason for its underestimation of strong EP El Niño events and their associated amplitude asymmetry.

However, the relationship between the mean state and the ENSO is bidirectional. Although the mean state can regulate the ENSO, the asymmetry of ENSO events can rectify the mean state itself in the long term. Due to the fact that warm events are stronger than cold events, the long-term average of ENSO-related anomalies is not zero, resulting in a net “residual” effect that can alter the mean thermocline structure and SST gradient [59]. This creates the dilemma of whether the chicken or the egg gives birth to the chicken when diagnosing causal relationships. Liang et al. [60] successfully defined this rectification effect using a low-order model, indicating that it originates from a nonlinear advection and may result in a time-averaged state significantly higher than the equilibrium state. Recently, Huang et al. [61] proposed a method based on Box–Cox normalization, defining a “normalized mean state” to eliminate this ENSO rectification effect, providing a clearer framework for studying the interaction between background climate and ENSO variability. This work emphasizes that a realistic mean state structure, especially the realistic cold tongue and deep convection pattern, is a key prerequisite for simulating the complete spectrum of the ENSO’s nonlinear behavior and its fundamental asymmetry.

In short, the asymmetry of the ENSO is not caused by a single mechanism, but by the complex interaction of multiple nonlinear processes. Although the response of the atmosphere to SST, particularly deep convection, appears to be the primary driver for initial asymmetry, ocean dynamics are crucial for amplifying this signal and shaping the unique spatiotemporal characteristics of El Niño and La Niña. All of these processes are essentially modulated by the average state of climatology, highlighting the close relationship between the deviation of the model in the average state and its persistent difficulty in simulating the asymmetry of the ENSO in reality.

4. The Challenge of Simulating ENSO Asymmetry in Climate Models

The accurate simulation of the ENSO and its diverse behavioral patterns remains one of the most significant challenges in the field of climate modeling. Although modern state-of-the-art models have made great progress in capturing the fundamental periods and spatial patterns of the ENSO, they have always struggled to cope with its nonlinear characteristics. The asymmetry between El Niño and La Niña, as a direct manifestation of these nonlinear features, has been proven to be a particularly stubborn deficiency and persists in multiple generations of climate models. This section will evaluate this long-standing challenge, trace the historical background of biases from CMIP3 and CMIP5 to the current CMIP6 model, link these biases to underlying physical processes, and explore potential pathways toward improvement.

4.1. A Historical Perspective on Model Biases (CMIP3 & CMIP5)

The underestimation of ENSO asymmetry is not a new issue. The early evaluation of models participating in the CMIP3 revealed a common issue: most models simulate the ENSO as too symmetrical compared to observed results. Sun et al. [27] conducted a comprehensive analysis of 19 CMIP3 models and found that there were common biases in the simulated ENSO statistics, characterized as a significant weak asymmetry between the warm and cold phases. This deficiency is linked to systematic biases in the climatological mean state, notably a western Pacific warm pool that was smaller yet warmer than in reality, which sets the stage for unrealistic air–sea interaction environments.

This issue became the focus of analysis in the subsequent CMIP5 collection. Zhang and Sun [25] evaluated 14 CMIP5 models and confirmed that a persistent shortcoming is the continued underestimation of ENSO asymmetry. They quantified this problem by computing the skewness of SST anomalies in the Niño-3 region—a statistical metric of asymmetry [62]. They found that nearly every model was unable to reproduce the positive skewness seen in observations, and some models even simulated negative skewness. Their analysis further reveals that this weak surface asymmetry is mainly due to underestimation of warm SST anomalies during El Niño events, while the simulated La Niña anomaly is closer to the observed value. By comparing the fully coupled runs with the corresponding atmosphere-only (AMIP) experiment, they demonstrated that this bias is at least partially due to the atmosphere, specifically due to insufficient precipitation response to the observed warm SST anomalies in the eastern Pacific. This provides a key insight: the inability to simulate asymmetry is not only an ocean problem, but also deeply rooted in atmospheric components and their coupling with the ocean, a conclusion supported by analysis of nonlinear wind stress coupling [39].

4.2. Current Status in CMIP6 Models

With the improvement of model resolution, physical processes, and the inclusion of more comprehensive Earth system components, people had hoped that the CMIP6 model would make significant progress in simulating ENSO asymmetry. However, recent evaluations indicate that despite progress in certain areas, fundamental problems still exist. Zhao and Sun [26] provided the clearest picture to date for the comprehensive analysis of 19 CMIP6 models. On the surface, the key finding is that most CMIP6 models still systematically underestimate ENSO asymmetry. The average SST skewness of the Niño-3 region in the model set is still significantly weaker than the observed values, and some models continue to exhibit skewness close to zero or even negative values. There are significant differences between the models, but the central tendency of the model ensemble is clear: significant asymmetry in the observed world is almost absent in the simulated world.

However, a key new insight of the CMIP6 generation is the distinction between surface and subsurface representation. Zhao and Sun [26] found that the simulation of ENSO asymmetry is more realistic in the subsurface than at the surface. Although the positive asymmetry of the subsurface temperature in the eastern Pacific is still underestimated, it agrees more closely with observations than the surface representation does. This indicates that while the model begins to capture certain nonlinear oceanic dynamics occurring in the thermocline, these improvements are not effectively transmitted to the surface. This disconnect indicates that there are still persistent biases in the thermodynamic processes controlling the surface mixing layer or in the coupled air–sea feedback connecting the ocean and atmosphere.

While the models face difficulties in terms of amplitude asymmetry, they show mixed success with other features. For example, Zhao and Sun [26] found that although CMIP6 models underestimated the strength of seasonal phase locking for El Niño and La Niña, they did capture the correct ensemble average peak months, despite significant differences between models. The models also perform better in simulating the phase locking of El Niño than that of La Niña. This indicates that although the nonlinear feedback that generates strong amplitude asymmetry has not been accurately captured, some basic seasonal dynamic characteristics have been reflected by the model. However, simulating asymmetric teleconnection remains a major challenge. Sengupta et al. [63] evaluated 50 CMIP6 models and found that although the models performed reasonably in simulating the basic nature of teleconnection, they performed significantly worse in capturing the asymmetry of observed precipitation responses. The asymmetry bias in the model was found to be related to the inability to accurately simulate the skewness of both local precipitation and Niño-3.4 SST distributions.

4.3. Linking Model Biases to Physical Processes

The persistent existence of weak ENSO asymmetry in the model is not a random error, but is systematically related to other well-documented models’ biases. Scientific literature has now established a causal chain that links the error in simulating the mean state with atmospheric feedback defects, which in turn prevent the development of strongly asymmetric ENSO events. The main culprit is the equatorial “cold tongue bias”. As shown by Bayr et al. [58] in their recent CMIP6 model, most models produce a cold tongue that stretches excessively westward into the Pacific warm pool. This biased mean state has a profound impact on asymmetry. It hinders the southward migration of the Intertropical Convergence Zone (ITCZ) during the development of El Niño, a character of observation during strong EP El Niño events. Due to anchoring deep atmospheric convection too far west, the model is unable to trigger the powerful nonlinear Bjerknes feedback in the eastern Pacific, which is an essential condition for generating extreme asymmetric warming.

This failure to correctly position convection directly leads to a weak and westward-displaced westerly wind response during El Niño. As Zhang et al. [15] first pointed out in the CCSM model family, insufficient asymmetry in deep convection can give rise to insufficient asymmetry in zonal wind stress. This discovery was confirmed by Liu et al.’s [29] multi-model analysis, which attributed the deterioration of ENSO asymmetry in high-resolution atmospheric models to the “degradation of nonlinear atmospheric feedback”. The systematic bias of the zonal ocean current with extreme westward shift in the CMIP6 model also contributes to this problem, resulting in an overestimated zonal advective feedback and the formation of an excessive central-Pacific warming pattern [64,65].

The resulting weak and misplaced wind forcing then fails to drive a sufficiently strong oceanic response. The subsurface NDH mechanism, which is crucial for generating asymmetry [19], is also too weak in most models. In a comprehensive analysis, Hayashi et al. [51] pointed out that the ENSO asymmetry simulated in the CMIP model is directly and linearly proportional to the efficiency of the simulated subsurface NDH. Most models lie far below the observed values, confirming that their inability to generate strong NDH is the key reason for the inability to generate observed asymmetry. This highlights a chain of error propagation: biased mean states lead to flawed atmospheric responses, which in turn weaken the nonlinear response of the ocean, ultimately resulting in an overly symmetric ENSO. This issue is further complicated by the significant differences in the asymmetry simulated by the CMIP5 models, ranging from significant negative values to significant positive values, highlighting the difficulty of simulating observed positive asymmetry [66]. Consequently, the models are difficult to simulate the full range of asymmetric behaviors, from the significant evolutionary differences between EP and CP El Niño events [67] to the complex transitional patterns between phases [12,24].

4.4. Pathways to Improvement: The Role of Resolution and Parameterization

Addressing these persistent biases requires a multi-faceted approach that focuses on both model resolution and underlying physical parameterization schemes. Although improving model resolution is often seen as a panacea, recent research has revealed a more complex and sometimes counterintuitive picture. Liu et al. [29] conducted a critical analysis of this issue. They found that improving ocean resolution is usually beneficial. It leads to better-resolved eddy-driven heat transport, which helps alleviate cold tongue bias and improve mean SST and precipitation patterns. This in turn allows for clearer reproduction of ENSO spatial patterns. However, they found that simply increasing atmospheric resolution may have negative effects, leading to a deterioration of ENSO asymmetry. This is because without a corresponding increase in ocean resolution, higher atmospheric resolution can lead to the weakening of the nonlinear atmospheric feedback that is essential for building asymmetry. This indicates that the resolution of the ocean and atmosphere needs to be simultaneously improved to achieve balance.

In addition to resolution, improving the physical parameterization scheme that represents the unresolved process is crucial. The study by Imada and Kimoto [30] is an excellent case study. They studied the effects of TIWs, which remains typically unresolved in medium-resolution climate models and whose variability is known to be asymmetric [55]. By introducing a new parameterization scheme to describe the asymmetric heat transport associated with TIWs, they successfully generated a significant asymmetric negative feedback effect on the ENSO, which helps explain the observed stronger amplitude El Niño phenomenon. This indicates that targeted improvements in the characterization of key physical processes, including not only TIWs but also salinity effects [56], can yield substantial gains. Indeed, the choice of key physical parameterizations, such as for atmospheric convection, has been shown to be a primary factor in determining the skill of ENSO prediction, and the fidelity of these schemes is often evaluated using error measurement metrics like SST skewness [68]. The ultimate goal is to improve the simulation of fundamental nonlinearities in air–sea interactions [28], which are not only crucial for ENSO asymmetry but also have a critical impact on its decadal modulation [69,70], as well as their teleconnection to regions such as Australia [71] and the Indian Ocean [46]. If this goal is not achieved, it will weaken the reliability of future ENSO predictions, both for the ENSO itself [32,33] and its regional impacts, such as precipitation changes [49].

5. Asymmetry in ENSO Diversity

The traditional view of the ENSO as a simple periodic oscillation between a single warm and cold state has been replaced by a more detailed paradigm of ENSO complexity [7]. This modern framework acknowledges that ENSO events exhibit significant diversity in spatial patterns, temporal evolution, amplitude, and underlying physical mechanisms. This diversity is not random noise, but is inherently related to the fundamental asymmetry of the system. The nonlinear process that prevents La Niña from becoming an El Niño mirror is the same process that allows for the existence of different event types. This section explores this deep interplay, first analyzing the inherent asymmetry in the El Niño binary opposition between the Eastern Pacific and Central Pacific, then focusing on the phenomenon of multi-year La Niña as an extreme manifestation of temporal asymmetry, and finally discussing how decadal-scale climate variability modulates ENSO diversity and its asymmetry.

5.1. Asymmetry Within EP and CP El Niño Diversity

The primary axis of ENSO spatial diversity is the distinction between El Niño events in the Eastern Pacific and the Central Pacific. EP El Niño is often considered a typical type, characterized by the strongest SST anomalies in the coastal waters of South America, extending westward along the equator. In contrast, the SST anomaly peak of CP El Niño (also known as the “El Niño Modoki”) occurs in the equatorial central Pacific [24]. This spatial diversity is a direct result of different physical mechanisms that initiate and sustain warm events, and these mechanisms themselves have inherent asymmetry.

The key source of this diversity is the nature of atmospheric forcing, particularly from WWBs. Chen et al. [21] proposed a unified perspective, in which symmetric and typical ENSO cycles serve as baseline oscillations, while the asymmetry, irregularity, and diversity of events are mainly driven by the stochastic and state-dependent nature of WWBs. The location and intensity of these wind bursts are crucial and lead to different event evolutions [72]. For instance, the characteristics of westerly wind anomalies in the early months of the year can act as a precursor, with stronger and more eastward-extended westerlies favoring the development of an EP El Niño, while weaker, westward-confined westerlies are more conducive to a CP type event [73].

The response of the atmosphere to different El Niño types is highly asymmetric. One dominant feature is the anomalous Western Pacific Anticyclone (WNPAC), which serves as a key bridge in the ENSO teleconnection [74]. Compared with CP El Niño, WNPAC during EP El Niño is significantly stronger and more well-defined [67]. The asymmetry of WNPAC is directly attributed to the different locations of anomalous atmospheric heating. Wang et al. [75] investigated the underlying cause of this asymmetric circulation and found that it arises from a complex interplay between linear and nonlinear moist enthalpy advection over the WNP. Their analysis revealed that while the nonlinear term tends to induce a negative precipitation anomaly and an anticyclones’ response during both El Niño and La Niña winters, it is the combination with the much stronger linear moist enthalpy advection during El Niño that leads to the overall powerful WNPAC. This in turn leads to asymmetric impacts on regional climate. For example, the amplitude of winter extratropical teleconnection modes (such as the Pacific North America pattern) during the EP La Niña period is weaker than that during the EP El Niño period, which is a direct consequence of weaker tropical forcing during the cold phase [41]. This asymmetry likewise carries over to the influence of extreme events. Distinct “flavors” of El Niño exert diverse, asymmetric impacts on Western North Pacific tropical cyclone (TC) activity [76]. The asymmetric influence on precipitation over Southern China are further modulated by intraseasonal oscillations, which have different responses to EP and CP events [77]. It is important to note that while the EP/CP dichotomy is a useful framework, it represents a simplification of a broader ENSO continuum, which also includes other variations such as coastal and uncoupled events that contribute to the full spectrum of regional climate impacts [78].

5.2. The Phenomenon of Multi-Year La Niña

Perhaps the most significant manifestation of temporal asymmetry is that La Niña events tend to persist for two or sometimes three consecutive years, which is extremely rare in El Niño events. In recent years, the increasing frequency of such multi-year events has brought profound economic and social impacts, often leading to sustained and widespread droughts [11].

Recent studies have shown that not all multi-year La Niñas are the same. Wang et al. [11] classified them into two categories based on their precursors: “Super El Niño-to-multiyear La Niña” (SE2ML) and “CP El Niño-to-multiyear La Niña” (CPE2ML). This confirms that the early El Niño type is a key factor determining the subsequent evolution of La Niña.

The dynamics of these multi-year events are a subject of intense research, with recent studies highlighting a complex array of contributing factors beyond the tropical Pacific. The persistence of cold conditions can be influenced by specific atmospheric triggers, such as the historical southeasterly wind anomalies observed in March 2022 [79]. Inter-basin interactions are also crucial, with anomalous conditions in the Indian and Atlantic Oceans potentially energizing and sustaining La Niña events in the absence of strong oceanic preconditioning in the Pacific [80]. Furthermore, changes in the background mean state, such as a strengthened zonal SST contrast, have been shown to alter the physical processes governing La Niña’s growth and maintenance [81]. Both oceanic dynamics [82] and specific thermodynamic processes [83] have been identified as key to prolonging these events. In a novel linkage, it has even been proposed that remote forcing from intense Australian wildfires, through aerosol-cloud interactions, could contribute to tropical Pacific cooling and help sustain La Niña conditions [84].

The dynamic characteristics of these multi-year events also differ. Park et al. [85] identified a “mid-latitude leading double-dip La Niña” in which the connection with the Pacific Meridional Mode (PMM) is crucial for maintaining the cold event into its second year. A unique recent oceanic process characterized by rapid double reversal of zonal ocean circulation has also been identified as an important mechanism in multi-year La Niña events [86]. The transition from a multi-year La Niña event to a strong El Niño event, although uncommon, has become more likely in the context of transient greenhouse warming, posing new challenges to predictability [87].

This asymmetry in duration and transition is closely related to the Spring Predictability Barrier (SPB), a well-known phenomenon in the Northern Hemisphere where ENSO forecast skill significantly declines during spring. Chen et al. [88] pointed out that SPB itself also has asymmetry: spring SST anomalies in El Niño years exhibit markedly greater persistence than those in La Niña years, making El Niño more predictable in the short term. This is attributed to stronger zonal advection feedback during El Niño, which helps maintain warm anomalies against SPB. The role of inter-basin coupling is also crucial. For example, the Indian Ocean can weaken the ENSO SPB and its impact is stronger during La Niña events, leading to asymmetry in predictability [6].

5.3. Modulation by External and Decadal Variability

The asymmetry and diversity characteristics of the ENSO are not stationary and unchanging, but are modulated by other mode of climate variability on decadal and longer timescale. This low-frequency modulation adds another layer of complexity, causing the dominant features and the strength of asymmetry to vary over time.

The most significant modulation factor among them is the Interdecadal Pacific Oscillation (IPO), also known as the Pacific Decadal Oscillation (PDO). The observational analysis by Lin and Zheng [89] shows that during the positive PDO phase, El Niño events occur markedly more often than La Niña events, and this frequency asymmetry is consistent in most CMIP5 models. Okumura et al. [90] clearly revealed this modulation mechanism through a comprehensive analysis of a 1300-year model simulation, indicating that IPO primarily affects the frequency and duration of El Niño events. At the same time, the second leading mode of Tropical Pacific Decadal Variability (TPDV) regulates the amplitude and pattern asymmetry of the two phases. This modulation extends to teleconnection. Dong et al. [70] demonstrated that IPO asymmetrically modulates the remote effects of the ENSO, with warm IPO/ENSO phases being larger than cold phases in terms of atmospheric response. The combined effects of the ENSO and PDO on the stratosphere are also highly asymmetric [91]. This extends to regional climate impacts, where the combined phases of the ENSO and the PDO have been shown to asymmetrically influence dry and wet conditions in the U.S. Great Plains [92]. The decadal variability of this asymmetry is itself a result of changes in the underlying background state, which can alter the efficiency of nonlinear processes [69].

Other ocean basins also play a crucial role. The Atlantic Multi-decadal Oscillation (AMO) has been shown to have asymmetric effects on the ENSO, with its negative phase (cold Atlantic) prioritizing the strengthening of El Niño events by enhancing air–sea coupling in the Pacific [93]. There is also an asymmetric relationship between the IOD and the ENSO. Fan et al. [94] found that ENSO events that emerge early are linked to the later formation of strong IOD events, while Weller et al. [95] pointed out that exaggerated La Niña effects distort the climate impact of the IOD in climate models.

The PMM is another key external source of influence. Fan and Huang [96] pointed out that PMMs, as modulators rather than triggers of ENSO events, have a stronger enhancing effect on El Niño than on La Niña. The asymmetry of the PMM effects exacerbates the overall complexity of ENSO transition processes [97]. The fidelity of climate model simulations of these nonlinear interactions has profound implications for future predictions. Karamperidou et al. [31] established a direct link between a model’s ability to simulate ENSO nonlinearities and its projection of tropical Pacific warming, emphasizing that a deep understanding of the interactions between ENSO asymmetry, diversity, and decadal variability is a key requirement for achieving reliable climate predictions.

6. ENSO Asymmetry in a Changing Climate

As the global climate warms due to anthropogenic greenhouse gas emissions, a key issue facing the climate science community is how the dominant modes of natural variability will respond. Given the central role of the ENSO in driving global climate anomalies, its future behavior is crucial. The key aspect of its future trajectory lies in the evolution of asymmetry. Due to the fact that ENSO asymmetry is a product of nonlinear feedback, which is sensitive to background climate conditions, it is expected to undergo changes. However, predicting these changes is full of uncertainty, which directly stems from the model bias discussed in the previous section. This section synthesizes the current understanding of the possible evolution of ENSO asymmetry, with a focus on the expected changes in its amplitude and structure, the physical mechanisms driving these changes, and the ultimate impact on future teleconnection and regional climate impacts.

6.1. Projected Changes in ENSO Asymmetry and Amplitude

The question of how the basic characteristics of the ENSO will change in a warmer world has been a fiercely debated topic for many years, with little consensus on early research and model generations [98,99]. Some studies, especially in the CMIP5 era, predict that extreme El Niño and La Niña events are projected to occur more frequently [100,101,102], which would imply a potential intensification of the ENSO’s asymmetric extremes. However, recent analyses and models that simulate key nonlinear processes more accurately are presenting a different picture.

More and more evidence suggests that ENSO amplitude asymmetry may weaken in a warmer climate in the future. Geng and Cai [33] provided a convincing prediction based on the CMIP6 model that although ENSO asymmetry may only show slight changes in the 21st century, its asymmetry will significantly weaken after 2100 under sustained high emission scenarios (Figure 3). This means that the amplitude discrepancy between strong El Niño and strong La Niña events will decrease, leading to a symmetrical ENSO state. This expected weakening is not simply a statistical phenomenon, but is closely related to a fundamental shift in the dynamics of the tropical Pacific. This result is consistent with Ham’s [32] earlier study using the CMIP5 model, which found a significant reduction in Niño3 skewness in the RCP4.5 scenario and directly attributed it to changes in atmospheric feedback.

This projected trend of weakening asymmetry is part of a broader trend of overall weakening of the ENSO itself in the context of long-term, equilibrated warming. Tuckman’s [103] idealized simulation shows that as climate warms, the Walker circulation weakens, leading to a decrease in the extreme nature of overall ENSO events. As the extreme nature of the event decreases, its behavior tends to be linear, naturally reducing asymmetry. This conclusion is supported by comprehensive GCM studies, which have shown a significant decrease in ENSO amplitude under long-term warming scenarios [104,105]. Although there are still significant differences in predictions between models [106,107], the models that best capture the nonlinear characteristics of the ENSO often exhibit a weakening responses [108].

6.2. The Role of a Warming Mean State

The projected weakening of ENSO asymmetry is tightly linked to changes in the mean state of tropical Pacific climatology. The prevailing theory suggests that under the warming of greenhouse effect, the equatorial Pacific will evolve into an average state with more “El Niño characteristics”, characterized by more rapid warming in the eastern equatorial Pacific than in the west [109]. This leads to a weakening of the zonal SST gradient and a weakening of the Walker circulation [110]. This altered background state has a profound impact on the generation of asymmetric by nonlinear feedback.

As described by Ham [32], the nonlinear relationship between SST and precipitation is a key driving factor for asymmetry. In a warmer average state in the future, when the eastern Pacific approaches the convective threshold, additional warming during El Niño events will trigger milder and more linear atmospheric responses. At the same time, the background conditions become wetter, allowing for a more significant reduction in response to cold SST anomalies during La Niña events [33]. This dual effect—weakened atmospheric response under El Niño events and enhanced response under La Niña events—collectively diminishes the nonlinear characteristics of the system.

The oceanic mechanism also plays a crucial role. Kohyama et al. [108] argue that in warmer climates, the enhanced thermal stratification of the upper ocean makes the thermocline “stiffer” and less sensitive to wind perturbations. This enhanced stability suppresses the large-scale vertical thermocline displacement and associated nonlinear dynamic heating required for extreme, asymmetric El Niño events [19]. This is consistent with the theoretical understanding that ENSO behavior is highly sensitive to the structure of the thermocline [111]. The change in mean state is not independent of the ENSO itself. The nonlinear rectification process, where the asymmetry of ENSO events leaves a permanent imprint in the average state, is a key feedback mechanism [59,60,112]. This complex interaction is further modulated by long-term changes in ocean heat uptake, especially in the Southern Ocean, which may affect tropical climate on a decadal timescale [113,114]. The paleo-climate records provide a long-term background and evidence from the warm period of the Pliocene suggests that a significantly warmer world may exhibit a “permanent El Niño-like” state with heavily dampened variability [115,116].

It is critical to note, however, that this projected “El Niño-like” change in the mean state is in direct conflict with observed trends over the past several decades, which have shown a strengthening of the Walker circulation and an increased east–west SST gradient, a “La Niña-like” trend. This divergence between model projections and historical observations represents a major unresolved question in climate science. As concluded by Lee et al. [117], this discrepancy may reflect either an error in the models’ forced response to greenhouse gases, an underestimation of multi-decadal internal variability that is currently masking the forced signal, or a combination of both. Resolving this dichotomy is essential for building confidence in future ENSO projections.

6.3. Implications for Future Teleconnections and Regional Climate Impacts

The potential weakening of the ENSO’s intrinsic asymmetry has important but complex implications for global impact. A key finding is that the asymmetric changes in SST anomalies in the tropical Pacific do not necessarily translate into a simple weakening of remote teleconnections. In fact, for certain regions and variables, the situation may be exactly the opposite.

Huang and Chen [49] studied the future trends of precipitation anomalies caused by the ENSO and found that the asymmetry of precipitation response is projected to further intensify under the background of climate warming. This is because in a warmer and moister environment, the response of the atmosphere to any given circulation anomaly will be amplified. Even if the forcing of underlying SST becomes more symmetric, the resulting shifts in convection may still lead to stronger and more spatially distributed precipitation anomalies. This indicates that the regional characteristics influenced by ENSO asymmetry may become more significant.

This conclusion is strongly supported by recent comprehensive research by Lieber et al. [118]. Their analysis of the CMIP6 model predicts that more regions worldwide will experience an amplification effect between ENSO historical teleconnection and mean and extreme temperatures and precipitation, rather than a weakening effect. The response to extreme events is very similar to the mean response, which means that regions where ENSO impacts are expected to increase may also face amplification effects of similar intensity of related extreme events. This amplification effect is not uniformly distributed and depends on the specific teleconnection path. For example, the “atmospheric bridge” linking the tropics and extratropics [119,120] may be altered, leading to a shift in extreme rainfall patterns in regions such as Australia [121].

The future direction of the ENSO’s relationship with other climate mode is also a key factor. For example, the asymmetric interaction between the ENSO and IOD is likely to be modulated by mean state changes [95,122]. In addition, transition dynamics, such as the rare but impactful shift from multi-year La Niña to strong El Niño, are projected to become more likely, posing new challenges for prediction and adaptation [87].

Ultimately, the credibility of these future predictions depends on the ability of climate models to accurately simulate underlying nonlinear processes. Karamperidou et al. [31] established a clear correlation between the model’s ability to simulate ENSO nonlinear processes and its projected warming patterns. This underscores a core challenge: significant biases like the equatorial cold tongue bias persist [58,77], they not only degrade the simulation of current ENSO asymmetry, but also create significant uncertainty on the predictions of future ENSO events.

7. Synthesis and Future Directions

The El Niño–Southern Oscillation is essentially asymmetric. This asymmetry is not a slight deviation from the ideal state of symmetry, but rather a core feature reflecting the profound nonlinear characteristics of the coupled ocean–atmosphere system. In the past two decades, the scientific community has shifted from simply recording this asymmetry to analyzing its underlying causes and consequences. This review integrates a broad body of research to construct a complex but internally consistent overall picture.

The manifestations of asymmetry have been well documented in multiple fields. At the amplitude level, the positive skewness of tropical Pacific SST anomalies (driven by larger and more extreme El Niño events rather than La Niña events) is a fundamental observational result [7]. On the temporal scale level, this asymmetry is manifested by the tendency of La Niña phenomena to persist for many years, as well as the more common and rapid transition from El Niño to La Niña [11]. On the spatial scale level, the trend of strong El Niño events spreading eastward since the late 1970s is in sharp contrast to the characteristic of La Niña events continuing to spread westward, which has distinct subsurface features. The key is that these asymmetries in tropical regions will be projected onto the global climate, driving asymmetric teleconnection and having profound impacts on regional weather and extreme events.

A consensus has emerged that these asymmetries stem from a series of nonlinear physical processes. Although ocean dynamics processes such as nonlinear temperature advection and subsurface heating are undoubtedly crucial [19], recent evidence increasingly points to atmospheric processes as the primary genesis of asymmetry. The state dependence of stochastic westerly wind bursts, which tend to amplify warm events [20], and the sharp nonlinear threshold response of deep atmospheric convection to sea surface warming [22], are now considered the dominant driving factors. These atmospheric nonlinear effects generate asymmetric forcing, which are then shaped, amplified, and modulated through oceanic feedback and background climatological mean states. The modern ENSO “flavors” paradigm, such as the EP/CP dichotomy, is now understood not as a separate phenomenon, but rather an inherent component of the nonlinear asymmetric system [21].

Despite the significant progress, there are still several fundamental questions remain at the forefront in ENSO research, which highlight the remaining gaps in the understanding. The main challenge lies in quantifying the hierarchy of physical driving factors. Although a series of nonlinear mechanisms have been identified, their relative importance remains under active discussion. Is the origin of asymmetry mainly attributed to the deterministic nonlinearity of atmospheric convection [22], or is it dominated by the stochastic and state-dependent nature of wind bursts [20]? In addition, how these atmospheric driving factors interact with oceanic processes, as subsurface heating [51], to generate the observed asymmetry is not fully understood. The uncertainty in this mechanism is closely related to the persistent and confusing disconnect between the understanding of physical processes and the performance of state-of-the-art climate models. Why do most CMIP6 models systematically underestimate ENSO asymmetry despite improvements in resolution and physical processes [26]? Thoroughly clarifying the complex network of interacting model biases—from the prevalent cold tongue bias [58] to flawed convective parameterization schemes—remains a major challenge. The recent surge in multi-year La Niña events has also raised urgent questions about their causes and predictability: is this trend a manifestation of internal decadal variability or a forced response to anthropogenic warming [11]? Ultimately, these uncertainties converge into key questions about the future evolution trajectory of the ENSO. Some studies project that asymmetry will weaken [32,33], while other studies emphasize that the impact may be exacerbated. Due to the inability of the model to fully capture the nonlinear characteristics of the ENSO, the lack of consensus in the model still poses a significant obstacle to confidence in future projections [31]. Indeed, it is notable that many of the fundamental challenges highlighted in reviews nearly two decades ago—such as the precise roles of different feedbacks, the irregularity of the cycle, and the limits of long-term predictability—remain at the forefront of the field today [123]. This underscores the profound complexity of the ENSO system and the continued need for foundational research into its dynamics in a changing climate.

Addressing these unresolved questions requires collaborative and multi-pronged efforts through the use of new technologies, innovative diagnostics, and a renewed focus on the fundamental physics of coupled systems. In terms of modeling, the future development path requires not only incremental increases in resolution, but also the adoption of a balanced strategy to ensure coordinated development of oceanic and atmospheric components, in order to avoid degrading nonlinear feedbacks [29]. More importantly, it is crucial to make targeted improvements to parameterization schemes for key physical processes. For instance, the choice of atmospheric convection parameterization has been shown to be a crucial factor in determining a model’s ENSO prediction skill [68], and continued focus on these schemes, as well as those for cloud-radiative feedback and ocean mixing, is essential. In addition to model development, the scientific research community also needs to adopt more advanced diagnostic and methodological methods. By using a novel framework to separate background states from ENSO rectification effects [60,61,124] and applying machine learning to build powerful “ENSO emulators” [125], it is expected to diagnose complex behaviors and identify sources of model errors. Finally, future research must continue to integrate knowledge across the full ENSO spectrum. A comprehensive understanding requires going beyond a single indicator and developing a clear framework that links asymmetry with ENSO event diversity. Continuous optimization through conceptual models to incorporate state-dependent noise and other nonlinear factors will provide a theoretical basis for these efforts [126]. By better understanding the dynamic mechanisms that generate different types of events, this study can deepen the understanding of the complexity of asymmetry itself, thereby gaining deeper insight into predicting the most influential climate variability patterns on Earth.