📄 earth science

Understanding ENSO asymmetry with a reduced stochastic model: deterministic nonlinearity and multiplicative noise

This study employs a reduced stochastic model derived from daily observations to demonstrate that nonlinear deterministic drift, rather than state-dependent noise, is the primary mechanism driving ENSO asymmetry, a finding that also explains intermodel differences in CMIP6 simulations through variations in effective SST damping.

Original authors: Cristian Martinez-Villalobos, Boris Dewitte, Soong-Ki Kim, Malte F. Stuecker

Published 2026-10-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Cristian Martinez-Villalobos, Boris Dewitte, Soong-Ki Kim, Malte F. Stuecker

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The El Niño-Southern Oscillation, or ENSO, is a massive, natural climate pattern that swings back and forth across the tropical Pacific Ocean. It is the engine behind the world's most significant year-to-year weather variations, shifting rainfall patterns, temperatures, and storm tracks from the Americas to Asia. At the heart of this system lies a simple but powerful feedback loop: when the surface water in the eastern Pacific warms up, it weakens the steady trade winds that usually blow from east to west. Those weakening winds, in turn, allow the warm water to spread even further east, creating a self-reinforcing cycle that can grow into a major El Niño event. Conversely, when the water cools, the winds strengthen, pushing warm water away and deepening the cold. While this cycle is well understood, a stubborn mystery has lingered for decades: the swings are not equal. The warm El Niño events tend to be far more intense and dramatic than the cold La Niña events. The distribution of these events is lopsided, with the ocean capable of heating up to extreme degrees but rarely cooling down with the same ferocity. Scientists have long debated what drives this asymmetry. Is it the chaotic, unpredictable bursts of wind that hit the ocean, or is it something built into the fundamental physics of how the ocean and atmosphere interact?

A team of researchers has now peeled back the layers of this mystery by looking at the system with a much sharper, faster lens than ever before. Instead of studying the ocean and atmosphere in monthly snapshots, which smooths over the rapid, chaotic changes, they analyzed daily data from 1982 to 2019. They built a simplified mathematical model that tracks just two key variables: the temperature of the sea surface in the eastern Pacific and the strength of the winds in the central Pacific. By feeding this model with real-world daily observations, they could separate the predictable, deterministic forces of the climate system from the random, noisy jolts of atmospheric weather. Their goal was to see which of these two ingredients—the steady rules of the system or the random noise—was responsible for the lopsided nature of El Niño.

The researchers discovered that the answer lies primarily in the steady rules, not the random noise. When they stripped the system down to its core, they found that the deterministic part of the model—the part that describes how the ocean and wind naturally influence each other—already contained the seeds of the asymmetry. In their daily-resolution model, the physics of the system naturally favors the growth of warm events over cold ones. Specifically, the way the wind responds to warm water and the way the water responds to the wind creates a situation where warm, westerly wind conditions amplify each other more effectively than their cold, easterly counterparts. This built-in bias means that even if the random wind bursts hitting the ocean were perfectly symmetrical, the system would still produce stronger El Niños than La Niñas. The researchers showed that this deterministic nonlinearity is the dominant driver of the skewness, generating the observed lopsided distribution on its own.

This finding challenges a popular theory that had gained traction in previous studies using monthly data. In those coarser models, the rapid, chaotic wind bursts were often treated as the main culprit, with scientists suggesting that the randomness of the atmosphere itself was what made El Niños so extreme. However, by resolving the data to a daily scale, this new study reveals that the "randomness" was actually masking a deeper, deterministic truth. The daily analysis showed that the asymmetry is encoded in the very structure of the coupled ocean-atmosphere system. The random atmospheric noise does play a role, but it is a secondary one. It acts more like a modifier, slightly increasing the chances of the most extreme, record-breaking warm events, but it is not the primary reason the system is lopsided in the first place. The core imbalance is a feature of the deterministic drift, the underlying current of the system that pushes warm events to grow faster and stronger.

To confirm that this mechanism holds up beyond their specific model, the team applied their findings to a suite of twenty-three different climate models used in the latest global climate projections. They looked at how these complex computer simulations handled the daily interactions between sea temperature and wind. They found that the models which produced a more realistic, lopsided distribution of El Niño events were the ones where the effective damping of the sea surface temperature behaved in a specific, non-linear way. In other words, the models that got the asymmetry right were the ones where the "brakes" on the system worked differently depending on whether the ocean was warming or cooling. This connection suggests that the ability of a climate model to reproduce the real-world lopsidedness of El Niño depends heavily on how well it captures these subtle, non-linear feedbacks in the daily interaction between the ocean and the atmosphere.

The study does not claim to have solved every aspect of ENSO, as the real world involves many other factors like ocean waves and interactions with other climate patterns that were not included in this minimal model. However, it provides a clear, mechanistic explanation for a long-standing puzzle. By moving from a monthly view to a daily one, the researchers were able to see that the ocean's tendency to favor extreme warmth is not a fluke of random weather, but a fundamental property of the system's dynamics. The asymmetry is built into the engine itself. This insight offers a new way to evaluate climate models, suggesting that to predict future changes in El Niño intensity, scientists must focus on how well their models represent these daily, non-linear interactions, rather than just trying to tune the random noise. The lopsided nature of El Niño is not a mystery of chaos, but a signature of the system's own internal logic.

Technical Summary: Understanding ENSO Asymmetry with a Reduced Stochastic Model

Problem Statement
The El Niño-Southern Oscillation (ENSO) exhibits a pronounced asymmetry, characterized by stronger and more spatially extensive warm events (El Niño) compared to cold events (La Niña), resulting in positively skewed sea surface temperature (SST) distributions in the eastern Pacific. While this non-Gaussian behavior is well-documented, the dynamical origin remains debated. Previous studies have proposed explanations involving nonlinear deterministic feedbacks, state-dependent atmospheric noise (multiplicative noise), or a combination of both. A key challenge in disentangling these contributions lies in the multiscale nature of ENSO: its evolution is continuously influenced by high-frequency atmospheric variability (e.g., wind bursts) that acts as stochastic forcing. Traditional analyses often rely on monthly-mean data, which coarse-grains these fast processes, potentially obscuring the distinction between resolved deterministic drift and unresolved stochastic forcing. This study addresses the question of how nonlinear deterministic dynamics and stochastic forcing contribute to ENSO asymmetry within a unified, interpretable framework estimated at daily resolution.

Methodology
The authors develop a minimal stochastic framework using daily observational data (1982–2019) from NOAA OISST and NCEP-NCAR Reanalysis 1. The core of the methodology involves:

Two-Variable Stochastic Model: A coupled system is constructed using the Niño 3 SST index ( $T$ ) and the Niño 4 zonal wind index ( $U$ ). The system is modeled by stochastic differential equations (SDEs) containing deterministic drift terms and stochastic forcing residuals.
- The model includes both linear and nonlinear formulations. The nonlinear version incorporates quadratic and cubic terms in the SST drift and a state-dependent wind-SST interaction term ( $\gamma_T UT$ ).
- Stochastic forcing is parameterized as correlated Gaussian white noise acting on both variables, estimated directly from the residuals of the fitted system.
Adiabatic Elimination: Leveraging the clear timescale separation between the fast atmospheric wind variable ( $U$ , decorrelation $\sim$ 13 days) and the slow oceanic SST variable ( $T$ , decorrelation $\sim$ 353 days), the wind variable is adiabatically eliminated. This reduction yields an effective one-dimensional stochastic equation for the Niño 3 SST index.
- Crucially, this reduction process naturally generates Correlated Additive-Multiplicative (CAM) noise in the reduced SST equation, rather than imposing it ad hoc. The multiplicative component arises from the interaction between SST and wind anomalies.
- The elimination also modifies the effective deterministic drift, introducing state-dependent damping terms.
Analytical Stationary Distributions: The reduced one-dimensional equations are analyzed using the Fokker-Planck equation to derive analytical solutions for the stationary probability density functions (PDFs). Four model configurations are compared:
- Linear + Additive noise
- Linear + CAM noise
- Nonlinear + Additive noise
- Nonlinear + CAM noise
CMIP6 Application: The derived analytical framework, specifically an approximation for skewness based on the effective drift coefficients and variance, is applied to historical simulations from 23 CMIP6 models to assess intermodel differences in ENSO asymmetry.

Key Results

Dominance of Deterministic Nonlinearity: The analytical comparison of the four model configurations reveals that nonlinear deterministic drift is the leading contributor to ENSO skewness. The "Nonlinear + Additive" model reproduces the observed positive skewness and the enhanced warm tail of the Niño 3 distribution with high fidelity. In contrast, the "Linear + CAM" model (linear drift with multiplicative noise) produces only modest departures from Gaussianity and fails to capture the observed asymmetry.
Secondary Role of State-Dependent Noise: While state-dependent (multiplicative) noise does affect the distribution, its primary role is to modify the probabilities of extreme events (slightly increasing the likelihood of extreme warm events and decreasing extreme cold events) rather than generating the fundamental skewness of the distribution.
Mechanistic Origin of CAM Noise: The study demonstrates that the CAM noise structure often used phenomenologically in conceptual ENSO models emerges naturally from the adiabatic elimination of the fast wind variable in a coupled SST-wind system. This provides a mechanistic link between the resolved coupled dynamics and the state-dependent noise terms.
Phase-Space Structure: Deterministic phase-space analysis shows that nonlinearities create an asymmetry in the flow geometry where warm-westerly states (positive $T$ , positive $U$ ) experience reduced effective damping or enhanced growth compared to their cold-easterly counterparts. This structural asymmetry is encoded in the drift term before stochastic forcing is applied.
CMIP6 Intermodel Differences: When applied to CMIP6 models, the analytical skewness approximation (derived from effective drift coefficients and variance) correlates positively with the skewness diagnosed directly from model output ( $r \approx 0.74$ ). This suggests that intermodel differences in ENSO asymmetry are, to leading order, related to differences in effective SST damping and SST variance across models.

Significance and Claims
The paper claims to provide a "mechanistic interpretation" of ENSO statistics by isolating the roles of nonlinear dynamics and stochastic forcing within a daily-resolution framework. Its primary significance lies in:

Reframing the Asymmetry Debate: It argues that within a daily-resolved coupled system, the asymmetry is primarily encoded in the deterministic drift (effective damping) rather than being solely a product of stochastic forcing. This contrasts with some monthly-mean perspectives where asymmetry is often attributed to state-dependent noise.
Unifying Deterministic and Stochastic Views: The study posits that deterministic nonlinearities and multiplicative noise should not always be viewed as distinct physical ingredients. Instead, they may reflect different reduced descriptions of the same underlying coupled variability, depending on the temporal resolution (daily vs. monthly) at which the system is analyzed.
Diagnostic Tool for Climate Models: The reduced framework offers a tractable basis for interpreting non-Gaussian climate variability. By linking skewness to effective drift coefficients and variance, it provides a process-oriented metric to evaluate why CMIP6 models underestimate observed ENSO asymmetry, suggesting that these models may have deficiencies in their effective damping structures or variance balances.

The authors maintain a modest scope, acknowledging that the model is a minimal representation that omits processes like thermocline variability, spatial SST structure, and extratropical precursors. However, they assert that this reduction is sufficient to isolate the specific contributions of nonlinear drift and stochastic forcing to the Niño 3 SST asymmetry.

More like this