Search

The El Niño–Southern Oscillation (ENSO) provides most of the global seasonal climate forecast skill1–3, yet, quantifying the sources of skilful predictions is a long-standing challenge4–7. Different sources of predictability affect ENSO evolution, leading to distinct global effects. Artificial intelligence forecasts offer promising advancements but linking their skill to specific physical processes is not yet possible8–10, limiting our understanding of the dynamics underpinning the advancements. Here we show that an extended nonlinear recharge oscillator (XRO) model shows skilful ENSO forecasts at lead times up to 16–18 months, better than global climate models and comparable to the most skilful artificial intelligence forecasts. The XRO parsimoniously incorporates the core ENSO dynamics and ENSO’s seasonally modulated interactions with other modes of variability in the global oceans. The intrinsic enhancement of ENSO’s long-range forecast skill is traceable to the initial conditions of other climate modes by means of their memory and interactions with ENSO and is quantifiable in terms of these modes’ contributions to ENSO amplitude. Reforecasts using the XRO trained on climate model output show that reduced biases in both model ENSO dynamics and in climate mode interactions can lead to more skilful ENSO forecasts. The XRO framework’s holistic treatment of ENSO’s global multi-timescale interactions highlights promising targets for improving ENSO simulations and forecasts.

Retrospective tropical Indo‐Pacific forecasts for 1961–2015 are made using 28 models from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) plus four models from the North American Multi‐Model Ensemble (NMME), using a model‐analog technique. Forecast ensembles are extracted from preexisting model simulations, by finding those states that initially best match an observed anomaly and tracking their subsequent evolution, requiring no additional model integrations. Model‐analog forecasts from the 10 “best” CMIP5 models have skill for sea surface temperature and precipitation comparable to that of both the NMME model‐analog forecast ensemble and (since 1982) traditional assimilation‐initialized NMME hindcasts. The El Niño–Southern Oscillation (ENSO) forecast skill has no trend over the 55‐year period, and its decadal variations appear largely random, although the skill does improve during epochs of increased ENSO activity. Including the CMIP5‐projected effects of external radiative forcings improves the tropical sea surface temperature skill of the model‐analog forecasts but not within the ENSO region. Seasonal forecasts are made by starting a climate model from an initial estimate of the latest global three‐dimensional ocean, atmosphere, and land conditions and then using supercomputers to run the model's equations forward in time. These extensive calculations are only feasible at a few national operational centers and large research institutions. However, many similar models are also used for long simulations of the Earth's preindustrial climate, which are made freely available for climate change studies. We investigated whether seasonal forecasts might be drawn from the information already existing within these simulations, instead of by making new model computations. Within each simulation, we determine the best matches, or “model‐analogs,” to current observed tropical Indo‐Pacific ocean surface conditions. How these analogs evolve over the next several months within each long simulation is then its seasonal forecast. We find that this much less expensive model‐analog technique is as skillful as the more traditional forecasting method. We then employed it to make forecasts during 1961–2015, using 28 existing climate simulations, significantly expanding previous forecast efforts. This study suggests that with little additional effort, sufficiently realistic and long existing climate model simulations can provide the basis for skillful seasonal forecasts. That is, anyone can be a climate forecaster. Seasonal tropical Indo‐Pacific hindcasts for 1961‐2015 are made from 28 different CMIP5 models, with some as skillful as operational CGCMsENSO hindcast skill has no trend, but it is higher in strong‐ENSO epochs (1980s and 1990s) and lower in weak‐ENSO epochs (1970s and 2000s)Including the effects of projected external radiative forcings improves SST hindcast skill but not within the ENSO region

An extreme Atlantic Niño developed in the boreal summer of 2021 with peak‐season sea surface temperature anomalies exceeding 1°C in the eastern equatorial region for the first time since global satellite measurements began in the early 1970s. Here, we show that the development of this outlier event was preconditioned by a series of oceanic Rossby waves that reflected at the South American coast into downwelling equatorial Kelvin waves. In early May, an intense week‐long westerly wind burst (WWB) event, driven by the Madden‐Julian Oscillation (MJO), developed in the western and central equatorial Atlantic and greatly amplified one of the reflected Kelvin waves, directly initiating the 2021 Atlantic Niño. MJO‐driven WWBs are fundamental to the development of El Niño in the Pacific but are a previously unidentified driver for Atlantic Niño. Their importance for the 2021 event suggests that they may serve as a useful predictor/precursor for future Atlantic Niño events. Atlantic Niño is the Atlantic counterpart of El Niño in the Pacific, often referred to as El Niño's little brother. It was previously thought to have only regional influence on rainfall variability in West Africa, but a growing number of studies have shown that Atlantic Niño also plays an important role in the development of El Niño–Southern Oscillation, as well as in the formation of powerful hurricanes near the coast of West Africa. This study investigates the development of an extreme Atlantic Niño in the summer of 2021. Here, we show that the 2021 event was preconditioned by warm waters piled up near the South American coast, and then directly triggered by a westerly wind burst event that drove the warm waters eastward. The westerly wind burst event was driven by a patch of tropical thunderstorms that formed across the Indian Ocean and moved slowly eastward across the Pacific, South America, and the Atlantic, also known as the Madden‐Julian Oscillation. Westerly wind bursts driven by the Madden‐Julian Oscillation are fundamental for the development of El Niño in the Pacific, but a previously unidentified driver for Atlantic Niño, and thus may improve our ability to predict future Atlantic Niño events. The extreme 2021 Atlantic Niño was preconditioned by a series of oceanic Rossby waves reflected into downwelling equatorial Kelvin wavesOne of the Kelvin waves was greatly amplified by an intense week‐long westerly wind burst event, initiating the 2021 Atlantic NiñoThe westerly wind burst was driven by the Madden‐Julian Oscillation, which is a previously unidentified driver for Atlantic Niño The extreme 2021 Atlantic Niño was preconditioned by a series of oceanic Rossby waves reflected into downwelling equatorial Kelvin waves One of the Kelvin waves was greatly amplified by an intense week‐long westerly wind burst event, initiating the 2021 Atlantic Niño The westerly wind burst was driven by the Madden‐Julian Oscillation, which is a previously unidentified driver for Atlantic Niño

The failed influence of the 2015–2016 El Niño on California rainfall has renewed interest in the relationship between El Niño and U.S. rainfall variability. Here we perform statistical data analyses and simple model experiments to show that sufficiently warm and persistent sea surface temperature anomalies (SSTAs) in the far eastern equatorial Pacific are required to excite an anomalous cyclone in the North Pacific that extends to the east across the U.S. West Coast and thus increases rainfall over California. Among the four most frequently recurring El Niño patterns considered in this study, only the persistent El Niño, which is often characterized by the warm SSTAs in the far eastern equatorial Pacific persisting throughout the winter and spring, is linked to such extratropical teleconnection patterns and significantly increased rainfall over the entire state of California. During the last 69 years, only three of the 25 El Niño events (i.e., 1957–1958, 1982–1983, and 1997–1998) are clearly identified as the persistent El Niño. In addition, the monthly rainfall variance explained by El Niño is less than half that caused by internal variability during the 25 El Niño. Therefore, the rarity of persistent El Niño events combined with the large influence of internal variability effectively explains the fragile relationship between El Niño and California rainfall. This study explores why some historical El Niño events significantly increased rainfall in California but some other events did notStatistical data analyses and model experiments show that persistently warm far eastern Pacific excites an optimal stationary wave patternDuring the last 69 years, only three such events have occurred, which explains the fragile relationship between El Niño and California rainfall

Understanding the interaction between the tropical Pacific and Atlantic Oceans has challenged the climate community for decades. Typically, boreal summer Atlantic Niño events are followed by vigorous Pacific events of opposite sign around two seasons later. However, incorporating the equatorial Atlantic information to variabilities internal to the Pacific lends no significant additional predictive skill for the subsequent El Niño‐Southern Oscillation (ENSO). Here we resolve this conundrum in a physically consistent frame, in which the nascent onset of a Pacific event rapidly induces an opposite‐signed summer equatorial Atlantic event and the lead correlation of Atlantic over Pacific is a statistical artifact of ENSO's autocorrelation. This Pacific‐to‐Atlantic impact is limited to a short window around late spring due to seasonally‐amplified Atlantic atmosphere‐ocean coupling. This new frame reconciles the discrepancies between the observed and multi‐model simulated inter‐basin relationship, providing a major advance in understanding seasonally‐modulated inter‐basin climate connections as well as their predictability. Previous studies interpreted the observed lead/lag relationship between Atlantic Niño/Niña and Pacific Niño/Niña sea surface temperature anomalies as evidence for a precursory role of the equatorial Atlantic on the development of El Niño–Southern Oscillation (ENSO) events. This study clearly demonstrates that this statistical relationship is not related to Atlantic‐to‐Pacific causality, but can rather be explained by seasonally modulated equatorial Atlantic's response to ENSO. We find that Pacific ENSO events drive equatorial Atlantic events rather than vice versa, and reconcile the apparent discrepancies between the observed and multi‐model simulated tropical Pacific/Atlantic relationship. The lead correlation of the equatorial Atlantic over the Pacific is not related to Atlantic‐to‐Pacific causalityThe tropical Pacific/Atlantic interaction is consistent with the nascent onset of Pacific events driving the equatorial Atlantic eventsThe discrepancies between the observed and multi‐model simulated tropical Pacific/Atlantic relationship can be reconciled in this new frame The lead correlation of the equatorial Atlantic over the Pacific is not related to Atlantic‐to‐Pacific causality The tropical Pacific/Atlantic interaction is consistent with the nascent onset of Pacific events driving the equatorial Atlantic events The discrepancies between the observed and multi‐model simulated tropical Pacific/Atlantic relationship can be reconciled in this new frame

Here we show that the characteristics of the Indian Ocean Dipole (IOD), such as its power spectrum and phase relationship with the El Niño–Southern Oscillation (ENSO), can be succinctly explained by ENSO combination mode (C‐mode) wind and heat flux forcing together with a seasonal modulation of the air/sea coupled Indian Ocean (IO) Bjerknes feedback. This model explains the observed high‐frequency near‐annual IOD variability in terms of deterministic ENSO/annual cycle interactions. ENSO‐independent IOD events can be understood as a seasonally modulated ocean response to white noise atmospheric forcing. Under this new physical null hypothesis framework, IOD predictability is determined by both ENSO predictability and the ENSO signal‐to‐noise ratio. We further emphasize that lead/lag correlations between different climate variables are easily misinterpreted when not accounting properly for the seasonal modulation of the underlying climate phenomena. A physical null hypothesis for the IOD is presentedA simple low‐order model captures the IOD spectrum and ENSO/IOD phase relationshipIOD predictability is determined by ENSO predictability and ENSO signal‐to‐noise ratio

Observations and model simulations of the climate responses to strong explosive low-latitude volcanic eruptions suggest a significant increase in the likelihood of El Niño during the eruption and posteruption years, though model results have been inconclusive and have varied in magnitude and even sign. In this study, we test how this spread of responses depends on the initial phase of El Niño-Southern Oscillation (ENSO) in the eruption year and on the eruption's seasonal timing. We employ the Geophysical Fluid Dynamics Laboratory CM2.1 global coupled general circulation model to investigate the impact of the Pinatubo 1991 eruption, assuming that in 1991 ENSO would otherwise be in central or eastern Pacific El Niño, La Niña, or neutral phases. We obtain statistically significant El Niño responses in a year after the eruption for all cases except La Niña, which shows no response in the eastern equatorial Pacific. The eruption has a weaker impact on eastern Pacific El Niños than on central Pacific El Niños. We find that the ocean dynamical thermostat and (to a lesser extent) wind changes due to land-ocean temperature gradients are the main feedbacks affecting El Niño development after the eruption. The El Niño responses to eruptions occurring in summer are more pronounced than for winter and spring eruptions. That the climate response depends on eruption season and initial ENSO phase may help to reconcile apparent inconsistencies among previous studies. In our simulations the likelihood of El Niño-like responses almost doubles after a Pinatubo-size eruptionThe central Pacific El Niños develop much stronger warming responses to volcanic forcing than the eastern Pacific El NiñosEl Niño responses are highly sensitive to the seasonal timing of a Pinatubo-size eruption

We present a systematic exploration of differences in the spatiotemporal sea surface temperature (SST) evolution along the equatorial Pacific among observed El Niño events. This inter‐El Niño variability is captured by two leading orthogonal modes, which explain more than 60% of the interevent variance. The first mode illustrates the extent to which warm SST anomalies (SSTAs) in the eastern tropical Pacific (EP) persist into the boreal spring after the peak of El Niño. Our analysis suggests that a strong El Niño event tends to persist into the boreal spring in the EP, whereas a weak El Niño favors a rapid development of cold SSTAs in the EP shortly after its peak. The second mode captures the transition and resurgence of El Niño in the following year. An early‐onset El Niño tends to favor a transition to La Niña, whereas a late‐onset El Niño tends to persist long enough to produce another El Niño event. The spatiotemporal evolution of several El Niño events during 1949–2013 can be efficiently summarized in terms of these two modes, which are not mutually exclusive, but exhibit distinctive coupled atmosphere‐ocean dynamics. A systematic analysis identifies two leading modes of inter‐El Niño variabilityThe two modes describe spring persistence, transition, and resurgence of El NiñoSeveral El Niño events since 1949 can be readily summarized with the two modes

Quantifying the response of atmospheric rivers (ARs) to radiative forcing is challenging due to uncertainties caused by internal climate variability, differences in shared socioeconomic pathways (SSPs), and methods used in AR detection algorithms. In addition, the requirement of medium‐to‐high model resolution and ensemble sizes to explicitly simulate ARs and their statistics can be computationally expensive. In this study, we leverage the unique 50‐km large ensembles generated by a Geophysical Fluid Dynamics Laboratory next‐generation global climate model, Seamless system for Prediction and EArth system Research, to explore the warming response in ARs. Under both moderate and high emissions scenarios, increases in AR‐day frequency emerge from the noise of internal variability by 2060. This signal is robust across different SSPs and time‐independent detection criteria. We further examine an alternative approach proposed by Thompson et al. (2015), showing that unforced AR variability can be approximated by a first‐order autoregressive process. The confidence intervals of the projected response can be analytically derived with a single ensemble member. An “Atmospheric River” (AR) is a weather phenomenon characterized by strong, narrow moisture transport that brings heavy rainfall to land. They serve as a critical water resource but also can cause damaging flash floods and high winds. Thus, knowing how AR activity will change in the future climate can help us to mitigate potential AR‐related disasters and promote effective water resource management. However, this task is challenging due to uncertainty in how fast the climate will warm and in how much the noise of natural climate variability can obscure the signal from global warming. In addition, several different definitions of AR exist, which raises questions about the sensitivity of AR changes to AR definition. In this study, we use a next‐generation global climate model to evaluate the influence of these uncertainties and to determine the time when humanity will notice a discernible change in AR activity. We find that the response to global warming can be robustly identified by 2060 across all explored methods of computation. We further examine a less expensive approach, which enables us to quantify the uncertainty in warming signals and estimate the time of signal emergence in a single realization of nature. Increases in atmospheric river (AR) frequency emerge from the noise on internal variability by 2060 in simulations with a global climate modelA computationally efficient method is developed for determining the time of emergence (ToE) of AR responses to global warmingThe AR ToE is not sensitive to the criteria of AR detection algorithm or shared Socioeconomic pathways Increases in atmospheric river (AR) frequency emerge from the noise on internal variability by 2060 in simulations with a global climate model A computationally efficient method is developed for determining the time of emergence (ToE) of AR responses to global warming The AR ToE is not sensitive to the criteria of AR detection algorithm or shared Socioeconomic pathways

El Niño and La Niña comprise the dominant mode of tropical climate variability: the El Niño and Southern Oscillation (ENSO) phenomenon. ENSO variations influence climate, ecosystems, and societies around the globe. It is, therefore, of great interest to understand the character of past and future ENSO variations. In this brief review, we explore our current understanding of these issues. The amplitude and character of ENSO have been observed to exhibit substantial variations on timescales of decades to centuries; many of these changes over the past millennium resemble those that arise from internally generated climate variations in an unforced climate model. ENSO activity and characteristics have been found to depend on the state of the tropical Pacific climate system, which is expected to change in the 21st century in response to changes in radiative forcing (including increased greenhouse gases) and internal climate variability. However, the extent and character of the response of ENSO to increased in greenhouse gases are still a topic of considerable research, and given the results published to date, we cannot yet rule out possibilities of an increase, decrease, or no change in ENSO activity arising from increases in CO2. Yet we are fairly confident that ENSO variations will continue to occur and influence global climate in the coming decades and centuries. Changes in continental climate, however, could alter the remote impacts of El Niño and La Niña. Copyright © 2010 John Wiley & Sons, Ltd.

Large ensembles of model simulations require considerable resources, and thus defining an appropriate ensemble size for a particular application is an important experimental design criterion. We estimate the ensemble size (N) needed to assess a model’s ability to capture observed El Niño‐Southern Oscillation (ENSO) behavior by utilizing the recently developed International CLIVAR ENSO Metrics Package. Using the larger ensembles available from CMIP6 and the US CLIVAR Large Ensemble Working Group, we find that larger ensembles are needed to robustly capture baseline ENSO characteristics (N> 50) and physical processes (N> 50) than the background climatology (N≥ 12) and remote ENSO teleconnections (N≥ 6). While these results vary somewhat across metrics and models, our study quantifies how larger ensembles are required to robustly evaluate simulated ENSO behavior, thereby providing some guidance for the design of model ensembles. To account for uncertainties arising from the chaotic nature of the climate system, Earth system models are often used to generate a large number of simulations under slightly different initial conditions. These large ensembles enable the consistency between models and observations to be addressed while accounting for the internal variability in the climate system. Creating a set of ensemble simulations requires substantial resources, and so in this study we diagnose what ensemble size is sufficient to robustly represent the simulated behavior of the El Niño/Southern Oscillation (ENSO), one of the most important modes of variability affecting climate worldwide. To estimate the ensemble size required to characterize the ENSO simulation, ensemble members of CMIP6 and Large Ensemble models are analyzedA broad range in the relative performance of models exists with internal variability influencing the robustness of some ENSO characteristicsThe required ensemble size depends on metric, duration of observational record, and model; the size can be a small as 6 or greater than 50 To estimate the ensemble size required to characterize the ENSO simulation, ensemble members of CMIP6 and Large Ensemble models are analyzed A broad range in the relative performance of models exists with internal variability influencing the robustness of some ENSO characteristics The required ensemble size depends on metric, duration of observational record, and model; the size can be a small as 6 or greater than 50

Atmospheric rivers (ARs) exert significant socioeconomic impacts in western North America, where 30%of the annual precipitation is determined by ARs that occur in less than 15%of wintertime. ARs are thus beneficial to water supply but can produce extreme precipitation hazards when making landfall. While most prevailing research has focused on the subseasonal (≤5 weeks) prediction of ARs, only limited efforts have been made for AR forecasts on multiseasonal timescales (≥3 months) that are crucial for water resource management and disaster preparedness. Through the analysis of reanalysis data and retrospective predictions from a new seasonal‐to‐decadal forecast system, this research shows the existing potential of multiseasonal AR frequency forecasts with predictive skills 9 months in advance. Additional analysis explores the dominant predictability sources and challenges for multiseasonal AR prediction. Atmospheric rivers (ARs), narrow corridors of intense moisture transport and heavy precipitation, are an important water resource but also a cause of flooding‐related disasters for western North America. Consequently, predictions of AR frequency several seasons in advance potentially would be of great value, but such operational forecasts are currently lacking due to the challenges in simulating such intense, small‐scale weather phenomena and their predictability sources on seasonal timescales. In this study, we examine the forecast skill of AR frequency on seasonal‐to‐multiseasonal timescales (≥3 months) in a new generation seasonal‐to‐decadal prediction system developed at the Geophysical Fluid Dynamics Laboratory. We find that AR frequency can be skillfully forecast at least 9 months in advance over certain regions of the west coast of North America, such as California and Alaska, while the forecasts are only reliable for the first season in other regions. This regional variability can be further explained by the large‐scale climate variability pattern that is responsible for much of the skill, which is strongly modulated by slowly varying sea surface temperature (SST) variations. A prototype probabilistic seasonal AR forecast product is proposed. Atmospheric river frequency can be skillfully forecast at least 9 months in advanceThe skills are significant over certain regions of western North America such as Alaska and CaliforniaEl Nino‐Southern Oscillation and Interdecadal Pacific Oscillation are important predictability sources Atmospheric river frequency can be skillfully forecast at least 9 months in advance The skills are significant over certain regions of western North America such as Alaska and California El Nino‐Southern Oscillation and Interdecadal Pacific Oscillation are important predictability sources

We examine links between tropical Pacific mean state biases and El Niño/Southern Oscillation forecast skill, using model‐analog hindcasts of sea surface temperature (SST; 1961–2015) and precipitation (1979–2015) at leads of 0–12 months, generated by 28 different models from the fifth phase of the Coupled Model Intercomparison Project (CMIP5). Model‐analog forecast skill has been demonstrated to match or even exceed traditional assimilation‐initialized forecast skill in a given model. Models with the most realistic mean states and interannual variability for SST, precipitation, and 10‐m zonal winds in the equatorial Pacific also generate the most skillful precipitation forecasts in the central equatorial Pacific and the best SST forecasts at 6‐month or longer leads. These results show direct links between model climatological biases and seasonal forecast errors, demonstrating that model‐analog hindcast skill—that is, how well a model can capture the observed evolution of tropical Pacific anomalies—is an informative El Niño/Southern Oscillation metric for climate simulations. Research institutions around the world have developed computer models to conduct simulations of Earth's climate to support the Intergovernmental Panel on Climate Change Assessment. Therefore, it is necessary to assess aspects of these models using historical observations. Here, we develop a new model metric that involves making forecasts over several seasons, using existing long simulations of Earth's preindustrial climate conducted by many research institutions and provided to the community for use in climate studies. Within each simulation, we find the best matches, or “model‐analogs,” to historical observations of tropical Indo‐Pacific Ocean surface temperatures and sea level. The evolutions of these analogs over the next several months within the simulation are then used as forecasts. Long‐term accuracy of the forecasts indicates how well a model reproduces the time evolution of the observations. Precipitation forecast accuracy (regardless of lead) and sea surface temperature forecast skill (at 6‐month or longer leads) emerge as key identifiers of the models that best simulate tropical Indo‐Pacific climate. Skill of monthly ENSO hindcasts (1961–2015) of 28 CMIP5 models, determined from model‐analogs, is related to tropical Pacific model biasesModels with poorer precipitation hindcast skill generally have larger cold and dry biases in the equatorial cold tongue regionModel‐analog ENSO hindcast skill may be a useful metric for CMIP6 models, given its correspondence with CGCM mean and variance errors

In this Review, the middle initial of author Kim M. Cobb was omitted. The original Review has been corrected online.

Strong El Niño events are followed by massive summer monsoon flooding over the Yangtze River basin (YRB), home to about a third of the population in China. Although the El Niño–Southern Oscillation (ENSO) provides the main source of seasonal climate predictability for many parts of the Earth, the mechanisms of its connection to the East Asian monsoon remain largely elusive. For instance, the traditional Niño3.4 ENSO index only captures precipitation anomalies over East Asia in boreal winter but not during the summer. Here we show that there exists a robust year‐round and predictable relationship between ENSO and the Asian monsoon. This connection is revealed by combining equatorial (Niño3.4) and off‐equatorial Pacific sea surface temperature anomalies (Niño‐A index) into a new metric that captures ENSO's various aspects, such as its interaction with the annual cycle and its different flavors. This extended view of ENSO complexity improves predictability of YRB summer flooding events. ENSO's complex temporal and spatial diversities can be well captured when combining equatorial and off‐equatorial SSTAs to a new metricThis metric captures a robust year‐round ENSO/monsoon relationship and the Yangtze River basin summer flooding eventsThe ENSO complexity captured by the metric can be predicted several months in advance, thereby providing monsoon predictability