When Climate Modeling Becomes Climate Mythmaking: “Sea-Level Rise Causes More Cold Snaps”

The modern climate discourse is full of surprises, but every so often a publication appears that manages to outdo even the most ambitious attempts at connecting loosely related climate variables. The recent Nature Communications paper “Intensification of extreme cold events in East Asia in response to global mean sea-level rise” is one such specimen. It attempts to link global mean sea-level rise (GMSL) — especially in the comically tiny range of 15 to 30 centimeters — to an increase in extreme cold outbreaks in East Asia.

Abstract

Today, the global mean sea level (GMSL) stands ~ 20 cm higher than at the beginning of the last century, and the rate of sea-level rise has been accelerating in recent decades. Even a slight, globally uniform sea-level rise can notably impact atmospheric and oceanic circulations at climatic and potentially synoptic scales. However, the extent to which sea-level rise will influence extreme weather remains largely unknown. Here, we focus on East Asia and conduct climate model experiments to investigate the effects of GMSL rise on winter cold extremes. Our experiments demonstrate that GMSL rise promotes stronger and more frequent extreme cold events, and this influence is expected to strengthen significantly in the coming century. This effect is attributed to weakened mid-high latitude westerly winds and increased occurrence of blocking events over Eurasia. Our study presents evidence that GMSL rise can modify synoptic systems and intensify extreme events, suggesting that both coastal and inland countries are exposed to threats arising from GMSL rise.

https://www.nature.com/articles/s41467-025-63727-1

To reach this conclusion, the authors construct a world that doesn’t exist, run simulations for thousands of virtual years, and then present the resulting patterns as evidence for a new climate threat. It reads less like a scientific study and more like a technocratic fable: Once upon a time, a uniform global ocean rose an identical number of centimeters everywhere, and then the atmosphere obediently rearranged itself to produce additional cold spells.

The paper admits at the outset that “the extent to which sea-level rise will influence extreme weather remains largely unknown” — a commendably honest statement. Unfortunately, what follows is an attempt not to illuminate that unknown, but to populate it with model-generated certainties dressed in statistical regalia.

The premise rests on the assumption that “even a slight, globally uniform sea-level rise can notably impact atmospheric and oceanic circulations at climatic and potentially synoptic scales.” That line alone deserves an award for imaginative framing. A globally uniform sea-level rise exists in climate models, but nowhere on Earth. The oceans are not a swimming pool; they are a dynamic, sloshing fluid influenced by gravity anomalies, tectonics, heat transport, winds, freshwater fluxes, and basin geometry. Treating them as a flattened bathtub surface is the kind of abstraction that may simplify a model but certainly does not simplify reality.

Yet the authors treat this artificial uplift as a physical input to the climate system, not as a testing gimmick.

And then the real fun begins.

The 2200-Year Model Voyage

The paper reveals that each simulation is run for 2200 model years:

“All experiments were run for 2200 model years, with analyses focusing on the last 200 years of the model output.”

One can appreciate the computational commitment, but the physical justification is less clear. Running a climate model for millennia under fixed conditions is an easy way to induce artificial equilibrium states or oceanic warm pools that have no analog in observed history. The authors themselves candidly acknowledge this:

“Our coupled sea-level experiments have been conducted over a span of 2200 years, a duration sufficient to induce substantial warming in the North Pacific.”

This is equivalent to admitting that the model has been allowed to drift into a condition that Earth has not experienced — and then treating the drift as a feature rather than a flaw. When a long simulation produces substantial warming in a specific region simply because it was allowed to run long enough, that warming is not a discovery; it’s a numerical artifact.

Yet this artifact becomes the scaffolding for the claim that sea-level rise “intensifies extreme cold events.”

The Uniform Sea Level Assumption: A Modeler’s Fantasy

The foundational design choice is laid out without irony:

“Here, GMSL rise is represented by a globally uniform uplift of the ocean reference surface—an idealized but scientifically justified simplification.”

Idealized? Yes.
Scientifically justified? Only if the goal is to make the model easier to manipulate, not to reflect the physical characteristics of ocean basins.

There is something remarkable about taking an inherently heterogeneous real-world process and flattening it into a uniform forcing, then concluding that this manufactured homogeneity creates complex climate responses. It’s a bit like digitally raising the floor of your house by a few centimeters in a video game and concluding that this explains thunderstorms in your backyard.

Later the authors concede that real-world regional sea-level patterns produce “minor and less significant” effects. This is their way of admitting that their modeled effect depends on a scenario that Earth does not produce:

“Regional sea-level variations may also influence winter extreme cold events in East Asia, although the effects appear minor and less significant.”

In other words: the real ocean doesn’t generate the effect they want, but the fictional one does.

The CO2 Paradox: High Sea-Level Rise, Fixed Atmospheric CO₂

The paper freezes atmospheric CO2 concentration at 400 ppm for every simulation, even when sea-level rise is pushed to absurd levels (5, 10, 20 meters):

“In all SL experiments, atmospheric CO2concentration was fixed at 400 ppm (close to current levels) to isolate the impact of GMSL rise.”

This creates a physical impossibility. If sea level rises by 5 to 20 meters — a scale associated with glacial melt over millennia — CO2 would not be holding still at 400 ppm. The entire scenario becomes a detached, ahistorical sandbox, not a representation of anything that could plausibly occur.

The authors even admit that in a world where sea level rises beyond about 2.5 meters, rising CO2 and its warming effects would overwhelm their hypothesized cold-event mechanism:

“The warming caused by high CO2 levels… could potentially offset the effects of sea-level rise.”

Put more plainly: the mechanism only functions in a physically impossible world — large sea-level rise without greenhouse forcing.

Statistical Tricks: When Model Output Becomes “Observed” Data

Extreme cold days (ECDs) are defined using a model-derived threshold:

“We define the temperature threshold by the 10th percentile of the winter daily surface air temperature distribution of the PiControl experiment.”

Using a single model’s internal distribution to define what “extreme cold” means is not inherently problematic. What is problematic is treating 200 years of model output — from a single model — as if they represent 200 independent years of real-world observations.

The authors then apply t-tests grid by grid, highlighting results that achieve a confidence level as low as 90%. In any field outside climate modeling, a 90% confidence level is what you’d use for screening, not for declaring robust risk. But because the model produces thousands of cold events across centuries of output, the statistical machinery can dress even small, model-driven fluctuations in the garb of significance.

The paper is full of statements such as:

“The increase in max persistence can be significant (90% confidence level).”

and

“The solid dots indicate the mean change is significant at a 90% confidence level.”

When you treat synthetic output as empirical data, statistical significance becomes easier to obtain than a parking space at a climate conference.

The Blocking Events and Atmospheric Teleconnections

One of the more creative parts of the study is the claim that raising sea level weakens Eurasian mid-latitude westerlies, strengthening blocking events that funnel cold Arctic air into East Asia. The authors assert:

“As sea-level rise, winter background westerly winds weaken… favoring the development of blocking events.”

This presumed causal chain runs as follows:

1. Uniform sea-level rise warms the North Pacific (due to long-term model drift).
2. That warming triggers Rossby wave anomalies.
3. These anomalies weaken westerlies.
4. Weakened westerlies allow blocking events to persist.
5. Blocking events allow cold Arctic air to spill into East Asia.

This is a marvelous illustration of how, once a model is allowed to run long enough, everything can be connected to everything else. The authors even admit that the North Pacific warming — the very heart of this mechanism — is a product of the extended simulation:

“A duration sufficient to induce substantial warming in the North Pacific.”

In other words, the model generates the mechanism because the model is asked to.

To reinforce the illusion of significance, they provide regressions such as:

“R² = 0.45, p = 0.05” between blocking frequency and cumulative intensity of extreme cold events.

Only in climate modeling can one regress two synthetic variables, both produced by the same simulation, both influenced by the same artificial forcing, and call the correlation “evidence.”

The Self-Organizing Maps: When Patterns Organize Themselves

The paper uses Self-Organizing Maps (SOMs) to classify synoptic patterns. This algorithm takes large numbers of similar circulation states and compresses them into clusters representing typical atmospheric configurations.

SOMs are not inherently problematic — they are useful clustering tools — but they rely completely on the dataset provided. Feed them biased model output, and you get biased clusters.

The authors cluster the model’s output into three synoptic patterns (SOM1, SOM2, SOM3), then note:

“The frequency and max persistence of SOM1 increases in almost all sea-level experiments.”

Since SOM1 is defined as the pattern associated with cold events in East Asia, this is simply restating the model’s behavior: when we force the model with a uniform sea-level rise, the cold-related pattern happens more often. This is not a discovery of nature but an internal property of the artificial system.

The Paper’s Own Disclaimers Tell the Story

Buried in the Discussion section is a string of admissions that, taken together, dismantle the very conclusions the paper claims to offer.

1. Model warming is exaggerated due to long runs: “The warming in this region… may exhibit a slower rate and a smaller magnitude [in reality].”
2. Uniform sea-level rise is unrealistic compared to regional variations: “We used a uniform sea-level rise and did not account for regional differences.”
3. Transient responses are not represented: “They do not account for transient responses.”
4. High sea-level rise scenarios are physically inconsistent: “Atmospheric CO₂ concentrations are expected to far exceed 400 ppm.”
5. Regional sea-level variations produce little effect: “The effects appear minor and less significant.”

One might expect such limitations to temper the conclusions. Instead, the paper presses forward, declaring:

“Our study presents evidence that GMSL rise can modify synoptic systems and intensify extreme events.”

Evidence is a strong word for findings produced entirely within a model world that bears little resemblance to the real one.

The Grand Finale: “Urgent Assessment”

The study concludes with one final crescendo of technocratic alarm:

“An urgent assessment of the global disaster risk stemming from sea-level rise is imperative.”

This is perhaps the most striking line in the entire paper. After constructing a scenario based on a uniform sea-level rise that does not exist, allowing the model to drift for thousands of imaginary years, and generating cold outbreaks via artifacts of that drift, the authors finish with a call for an urgent global disaster assessment.

It’s the academic equivalent of sketching a hypothetical animal in a notebook and then insisting the Department of Agriculture issue emergency guidelines for feeding it.

What This Tells Us About Climate Science Today

This paper is not an outlier; it is an example of a broader pattern where climate models have evolved into self-referential systems. Inputs are simplified for convenience. Outputs are treated as observations. Statistical tests are applied to synthetic data. And elaborate narratives are woven to suggest real-world risks based solely on model-world cause-and-effect.

The danger here is not that such studies exist — creative modeling can be useful — but that they are routinely presented as evidence, not speculation.

Policymakers are then encouraged to act on the basis of what numerical worlds do, not what the real world demonstrates. When models become the source of threats and models become the evidence for those threats, we’ve crossed into mythmaking.

The uncertainty acknowledged at the beginning of the paper — “largely unknown” — is replaced by synthetic certainty at the end: a certainty that reinforces the prevailing climate narrative and supports calls for more research, more funding, and more policy intervention.

Conclusion

In the end, the paper illustrates how climate modeling can drift far from empirical grounding. A uniformly rising global ocean — a physical impossibility — is used to generate a warmed North Pacific — a model artifact — which is then used to claim that sea-level rise intensifies extreme cold events in East Asia.

If this were not enough, the authors decorate the output with t-tests, regressions, and confidence levels pulled from a dataset that exists only inside a computer. The result is presented as a threat requiring urgent assessment.

If this is not p-hacking on steroids, it is at least the next generation of climate numerology: p-hacking through simulation, where the “p” stands not for probability, but for parameterization.

Cold snaps in East Asia are real. Sea-level rise is real. But the bridge constructed in this paper between the two is as artificial as the uniform ocean surface the authors command into existence with their model code.

As always, the important lesson is this: uncertainty is not a flaw in climate science; dismissing uncertainty is.

And in this case, the uncertainty is large enough to sail the entire model ocean through.