When the Data Disagrees: Ice Melt, Ocean Currents, and the Predictions That Never Came True

A data-driven examination of common climate claims versus what the measurements actually show

1. Introduction — Why This Paper Exists

Every now and then, a discussion comes along where someone makes a series of confident claims about climate science. Ice is melting at "unprecedented" rates. The Gulf Stream is shutting down. The IPCC has it all figured out. And if you answer their questions with data they didn't expect, you get told you've "lost your point."

This paper exists because claims deserve scrutiny. Not dismissal. Not hand-waving. Scrutiny. With data, citations, and measurements. If a claim holds up to the data, great. If it doesn't, we should say so plainly.

What follows is a systematic examination of several common positions that come up in climate discussions. No names. No personal attacks. Just teh claims versus the measurements. Let the data do the talking.

2. The Claims Under Examination

The following positions are examined in this paper. Each was stated as fact in a public discussion. Each deserves verification.

Each of these will be addressed with peer-reviewed data. Not opinion. Not "the science says so." Actual measurements from actual instruments.

3. IPCC Predictions vs Observed Reality

3.1 The Models Run Too Hot

Let's start with the big one. The IPCC relies on climate models — specifically the CMIP6 ensemble — to project future warming. The question is simple: how well do these models match what we actually measure?

Not well, as it turns out.

McKitrick and Christy [2020] compared 38 CMIP6 models against satellite observations, radiosondes, and reanalysis products over the 1979–2014 period. Their finding was blunt: every single one of the 38 models overpredicted warming. Not most. Not many. All of them. The average difference between models and observations was statistically significant [McKitrick & Christy, 2020].

This isn't a minor discrepancy. The models predict a strong warming signal in the tropical mid-troposphere — the so-called "tropical hotspot" — that should be the fingerprint of greenhouse gas warming. After four decades of looking, it's still not there in the observations. The models say it must exist. The atmosphere says otherwise.

Scafetta [2021, 2025] took this further. He ranked CMIP6 models by how well they reproduce observed surface warming from 1980–2021. The models that came closest to observations were those wiht an Equilibrium Climate Sensitivity (ECS) between 1.8°C and 3.0°C. The IPCC's AR6 assumed an ECS range of 2.5–4.0°C — skewing toward the warm end [Scafetta, 2021].

In plain English: the IPCC's preferred models are too sensitive to CO2. They warm too fast. And they've been doing this for decades.

3.2 The IPCC Can't Even Agree on a Schedule

If the science is so settled, you'd think the body responsible for assessing it could at least agree on when to publish thier next report.

The IPCC completed AR6 in 2023. They're now working on AR7. After four consecutive plenary sessions — Istanbul, Sofia, Hangzhou, and Lima — they still cannot agree on a delivery timeline. Carbon Brief called this "unprecedented in IPCC history" [Carbon Brief, 2025].

The Earth Negotiations Bulletin recorded that IPCC Chair Jim Skea described the outcome of planning in annual increments as "the worst option." Nepal's delegation said the result would "harm the IPCC's legitimacy" [IISD, 2025].

The next attempt to agree on dates is March 2026 at IPCC-64. This is an administrative problem, not a scientific one. But it matters for a simple reason: when a scientific body repeatedly delays its synthesis report, it raises questions about whether the literature has reached conclusions it can confidently assert. If the science were as settled as claimed, agreeing on when to write it down shouldn't take four failed attempts.

3.3 AR6 on Cold Extremes — A Regional Problem

The IPCC's AR6 Chapter 11 stated it is "virtually certain" that the frequency and intensity of cold extremes have decreased on the global scale since 1950 [IPCC, 2021]. "Virtually certain" is IPCC-speak for 99–100% probability.

At the global average level, that may hold. But IPCC regional projections for mid-latitude cold extremes are a different story.

In January 2026, Europe's average temperature was −2.34°C, which is 1.63°C below the 1991–2020 average. It was the coldest January since 2010 [Copernicus, 2026]. Temperatures plunged 12–15°C below seasonal averages across Central and Western Europe. Finland recorded −40°C. The WMO described it as an "extraordinary weather event" with a "generational Arctic Outbreak" [WMO, 2026].

Four consecutive severe Northern Hemisphere winters. Cold records broken year after year across Europe, North America, and Asia. As we discuss in Section 6, the mechanism is well understood — La Niña combined with polar vortex disruptions. But the frequency and severity of these events raises a legitimate question: if global warming is supposed to reduce cold extremes, why are the regional patterns doing the opposite? The IPCC's own models did not predict this pattern of recurring severe mid-latitude cold outbreaks. The "virtually certain" label applies to global averages. But people don't live in global averages. They live in regions. And the regions are getting hammered.

4. "Unprecedented" Ice Melt — What the Data Actually Shows

The claim that ice is melting at an "unprecedented rate" gets thrown around a lot. Rarely with a citation. Let's look at what the actual satellite data says.

4.1 Arctic Sea Ice

The Arctic has lost ice since satellite records began in 1979. That's real. The long-term trend in September minimum extent is a decline of about 12.1% per decade relative to the 1981–2010 average [NSIDC, 2025]. That's roughly 74,000 km² per year.

But the decline is not accelerating. The NSIDC itself notes that while the last 19 years (2007–2025) are the lowest 19 on record, "there has been no significant trend in September minimum extents" during that period [NSIDC, 2025]. The 2012 record low has not been broken in 13 years. The rate of decline was steepest in the 1990s and 2000s, then levelled off after the 2007 step-change.

"Unprecedented" implies acceleration. The long-term trend is real, but the rate of decline has flattened. Those are different things, and the distinction matters.

4.2 Antarctic Sea Ice

Antarctic sea ice tells a more complex story. The 2023 and 2024 maximums were the lowest and second-lowest on record. The 2025 maximum (17.81 million km²) was the third-lowest [NSIDC, 2025].

But by the end of December 2025, Antarctic sea ice extent was roughly in line with the 1981–2010 average [NSIDC, 2025]. It recovered. Not that you'd know from reading the headlines.

The Antarctic picture is complicated by something most people don't talk about: what's happening under the ice. We'll get to that in Section 7.

4.3 The "Unprecedented" Problem

The satellite record goes back to 1979. That's 47 years. In geological terms, that's nothing. We have proxy data showing the Arctic was ice-free during the Holocene Thermal Maximum roughly 6,000–9,000 years ago — when CO2 levels were about 260 ppm, far lower than today [IPCC AR6, Ch. 9].

So "unprecedented in the satellite era" is accurate. "Unprecedented" without qualification is not. And when someone states it as fact without a citation, they're not doing science. They're doing rhetoric.

5. The AMOC Meltwater Hypothesis — Dismantled

This is the claim that meltwater from ice sheets is disrupting the Atlantic Meridional Overturning Circulation (AMOC), which includes the Gulf Stream, and this is causing colder European weather.

It sounds plausible. It makes for good headlines. There's just one problem: the latest data doesn't support it.

5.1 The AMOC Has Not Declined

In January 2025, Foukal, Terhaar, and Vogt published a study in Nature Communications using air-sea heat flux data to reconstruct AMOC behaviour from 1963 to 2017. Their conclusion: the decadal-averaged AMOC at 26.5°N has not weakened over this period [Foukal et al., 2025].

Previous studies that claimed AMOC decline used sea surface temperature (SST) as a proxy. The Woods Hole team showed that SST "doesn't work as well as initially thought" for reconstructing the AMOC. When they used the more physically direct measure — air-sea heat fluxes — the decline vanished [Foukal et al., 2025].

They also tested 24 CMIP6 Earth System Models and found that SST anomalies cannot robustly reconstruct the AMOC. The method that produced the scare headlines was flawed from the start.

5.2 Even If It Were Declining, the Timeline Is Wrong

Even if the AMOC were weakening, the timeline for cooling effects on Europe would be decades to centuries. The AMOC is a massive thermohaline circulation system. It doesn't respond to ice melt on a year-to-year basis. The cold snaps hitting Europe right now are atmospheric events, not oceanic ones.

The IPCC's own AR6 admits to "low confidence in quantitative projections of AMOC decline" because models cannot reproduce past AMOC behaviour [IPCC, 2021]. When your models can't hindcast, your forecasts are worthless.

5.3 The Collapse Predictions

In 2023, Ditlevsen and Ditlevsen published a statistical model in Nature Communications estimating AMOC collapse around 2057 ± 28 years, with a lower confidence bound as early as 2025 [Ditlevsen & Ditlevsen, 2023]. Media headlines ran with "as soon as 2025," which was a tail probability, not the central estimate. To be fair to the authors, thier central prediction hasn't been tested yet. But the lower bound has passed without event, and Foukal et al. [2025] shows no decline at all in the underlying AMOC data. The urgency narrative built on this paper is not supported by current observations.

6. What Is Actually Causing European Cold Winters

If it's not meltwater disrupting the Gulf Stream, then what's driving these severe European winters?

The answer is well understood by meteorologists, even if it doesn't make for sexy headlines: La Niña combined with Polar Vortex disruptions.

6.1 La Niña

A La Niña event emerged in December 2024. La Niña leads to colder-than-normal temperatures in Western Europe through well-documented teleconnection patterns. Butler et al. [2014] showed that Sudden Stratospheric Warming (SSW) events are significantly more likely during La Niña winters. Cohen et al. [2021] further documented the link between polar vortex disruptions and mid-latitude cold extremes.

La Niña winters have a significantly higher probability of producing an SSW event — a complete collapse of the Polar Vortex. When that happens, the jet stream goes wavy, and Arctic air pours into mid-latitudes [Butler et al., 2014; Cohen et al., 2021].

6.2 Polar Vortex Disruption

That's exactly what happened. In January 2026, a weakened and distorted Polar Vortex caused extreme waviness in the polar jet stream, steering frigid Arctic air deep into Europe, North America, and Asia. At the mid-troposphere level, temperature anomalies exceeded −10°C over Central Europe and reached −15°C over parts of Germany, Poland, and eastern France [Copernicus, 2026].

This is standard meteorology. It's been observed repeatedly. It's well-modelled. And it fully explains the severe cold without invoking any AMOC hypothesis.

The irony is that the AMOC meltwater story sounds more dramatic, so it gets the headlines. The actual explanation — La Niña and polar vortex dynamics — is mundane by comparison. But being boring doesn't make it wrong.

7. Antarctic Volcanism — Far From Irrelevant

A common dismissal goes something like this: "Those Antarctic volcanoes are on the other side of the world. What's the relevance to Europe?"

The relevance is not geographic. It's systemic. Antarctic volcanism matters because it fundamentally changes the ice melt narrative. If geothermal heat from below is melting ice that gets blamed on warming from above, then the entire framework for "unprecedented" melt rates is wrong.

7.1 Active Magmatism Under Marie Byrd Land

Lucas et al. [2025] detected 285 seismic events from 2019–2024 across eastern Marie Byrd Land, including 14 events in a single day (19 November 2024) near the Kohler Range. Long-period earthquakes at crustal depths beneath Mount Takahe match signatures of active magma movement [Lucas et al., 2025].

Here's why this matters: the eastern Marie Byrd Land Volcanic Province spatially overlaps with five of the six major glaciers in the Amundsen Sea Embayment — the region that has undergone the most rapid ice mass loss in recent decades [Lucas et al., 2025]. The ice that's melting fastest sits on top of active volcanoes.

7.2 Squid Ridge — Holocene Volcanism in the Ross Sea

A 2025 expedition aboard the RV Nathaniel B. Palmer discovered "Squid Ridge," a recently formed fissure-driven eruptive seamount in the Terror Rift. The dredge recovered mantle peridotite-cored volcanic bombs. Explosive basaltic volcanism occurred during the Holocene, synchronous with ice sheet changes [Nature Communications Earth & Environment, 2025].

Active volcanism. Under the ice. During the current interglacial. Discovered in 2025. Anyone who says this is irrelevant to the ice melt discussion is not paying attention.

7.3 The Bransfield Strait Swarm

Between 2020 and 2021, approximately 85,000 volcano-tectonic earthquakes occurred near the Orca submarine volcano in the Bransfield Strait — previously considered inactive. This was the strongest seismic outburst ever recorded in Antarctica. The swarm was driven by the intrusion of 0.26–0.56 km³ of magma, caused by hot mantle material flowing through a tear in the subducting Phoenix plate [Poli et al., 2022].

7.4 Eighty-Five New Subglacial Lakes

In September 2025, ESA's CryoSat-2 data revealed 85 new active subglacial lakes, raising the total from 146 to 231. Geothermal heat is the primary driver of these lakes. Five new interconnected drainage networks were mapped [Nature Communications, 2025].

Subglacial meltwater forms due to geothermal heat from Earth's bedrock and frictional heat as ice slides over it. This water reduces friction between ice and bedrock, allowing ice to slide more quickly toward the ocean. This is a geothermal mechanism, not an atmospheric one.

7.5 Unexpected Lakes in East Antarctica

Seven active subglacial lakes were found under slowly moving coastal ice in Dronning Maud Land, East Antarctica — a region where models predicted a frozen bed. Something is producing more heat than models account for [Copernicus/The Cryosphere, 2025].

When models predict a frozen base and reality shows liquid water, the models are wrong. Not the water.

7.6 The Ice Unloading Feedback Loop

Coonin et al. [2024] modelled how ice loss reduces pressure on magma chambers, causing compressed magma to expand and increasing eruption frequency. The West Antarctic Rift System contains over 100 eruptive centres. Even long after ice unloading stops, eruption potential remains permanently elevated [Coonin et al., 2024].

This creates a positive feedback loop: melting ice triggers eruptions, which melt more ice, which triggers more eruptions. And none of this has anything to do wiht atmospheric CO2.

7.7 So What's the Relevance?

The relevance is this: when someone points to Antarctic ice loss as proof of atmospheric warming, they're ignoring the volcanoes, the geothermal heat, the 231 subglacial lakes, and the 85,000 earthquakes under the ice. You cannot attribute ice loss to a single cause when multiple heating mechanisms are operating simultaneously from below.

To be clear about scale: the average geothermal heat flux in West Antarctica has been measured at roughly 60–70 mW/m², with localised hotspots much higher [Maule et al., 2005]. That baseline flux alone melts only a few millimetres of ice per year. But that's the average. The new seismic data shows the flux is not evenly distributed — it's concentrated exactly where the fastest ice loss is occurring. The subglacial lakes, the magma movement, and the volcanic swarms are not background noise. They are localised heating events that the models didn't account for because the volcanoes were assumed to be inactive.

The question is not whether the atmosphere is warming. The question is how much of the observed ice loss is atmospheric and how much is geothermal. Nobody has conclusively answered that question yet. Until they do, attributing all Antarctic ice loss to atmospheric warming alone is premature.

8. The EPA Endangerment Finding — "Opinion" or Evidence?

The paper "Reconsidering the 2009 EPA Endangerment Finding: A Scientific Review" published on this site has been dismissed by some as "an opinion piece only." This deserves a direct response.

8.1 What the Paper Actually Contains

The review draws on work by scientists including John Christy (University of Alabama in Huntsville, lead scientist for the UAH satellite temperature record), Judith Curry (former chair of Georgia Tech's School of Earth and Atmospheric Sciences), Steven Koonin (former Undersecretary for Science at the US Department of Energy), Ross McKitrick (University of Guelph, co-author of the CMIP6 warming bias paper), and Roy Spencer (principal research scientist at UAH, co-developer of the satellite temperature record).

These are not random bloggers. These are published scientists with decades of peer-reviewed work in climate physics, satellite measurement, and statistical analysis. Their individual publications are cited throughout this paper and are independently verifiable in the scientific literature.

8.2 What Constitutes "Opinion"?

Calling something an "opinion piece" is easy. Addressing the specific claims in it is harder. The paper makes testable claims:

• Climate models have overpredicted warming — confirmed by McKitrick & Christy [2020]
• US extreme weather has not gotten worse — supported by IPCC AR6 Chapter 12's own admission of "low confidence" in attributing most extreme weather trends
• CO2 provides real benefits to agriculture — supported by satellite-observed global greening [Zhu et al., 2016]
• Sea level rise has not accelerated beyond historical rates — debated, but tide gauge data supports modest acceleration at best

Dismissing a paper as "opinion" without addressing any of its specific claims is not a rebuttal. It's an avoidance strategy. If the claims are wrong, show the data that proves them wrong. If you can't, maybe the claims aren't wrong.

9. The Predictions That Never Came True

Climate science has a prediction problem. Not because making predictions is bad — it's essential. But because when predictions fail, they get quietly forgotten rather than honestly acknowledged.

9.1 Scientific and Institutional Predictions

9.2 Public and Media Predictions

Beyond the scientific literature, a number of high-profile public predictions have also failed. These weren't peer-reviewed, but they shaped public perception and policy urgency:

• Ice-free Arctic by 2013–2014 (Al Gore, Copenhagen 2009) — Still 4.6 million km² of ice in 2025. No new minimum since 2012.
• No permanent Arctic ice by 2022 (James Anderson, Harvard, 2018) — Ice still present.
• "Children won't know what snow is" (David Viner, CRU, The Independent, 2000) — Record snowfalls across the Northern Hemisphere, 2024–2026.

The distinction between scientific predictions and media predictions matters. But both shape public trust. When predictions consistently fail in one direction — too warm, too dramatic, too fast — that tells you something about the assumptions behind them. Systematic bias in one direction is not "settled science." It's a calibration problem.

10. On Moving Goalposts and Circular Arguments

There's a particular trick that comes up in climate discussions. It goes like this:

1. Person A makes an original point supported by data.
2. Person B introduces a tangential question or new topic.
3. Person A answers Person B's question thoroughly, with evidence.
4. Person B then accuses Person A of "losing relevance to their original point."

This is a textbook example of what's known as a red herring followed by burden-shifting. Person B created the tangent. Person A followed it in good faith. Then Person B blamed Person A for the deviation that Person B introduced.

It's effective because it looks like Person A has gone off-topic. But the transcript shows who introduced each topic. The evidence trail doesn't lie.

Good faith debate means:

• Addressing the specific claims made, not dismissing them with labels ("opinion piece")
• Providing data to support your position, not just asserting it as fact
• Acknowledging when a claim has been answered, even if you disagree with the answer
• Not introducing new questions and then complaining when they get answered

If you ask a question, you should expect an answer. And if the answer includes six peer-reviewed citations about Antarctic volcanism, maybe the problem isn't that the answer is irrelevant. Maybe the problem is that it was more thorough than you expected.

11. Conclusion

The claims examined in this paper don't survive contact with the data.

• IPCC predictions: All 38 CMIP6 models run too hot. The IPCC can't agree on a schedule for AR7. Their regional cold extreme projections are being challenged by four consecutive severe Northern Hemisphere winters.
• "Unprecedented" ice melt: The long-term Arctic decline is real, but the rate has flattened since the 2007 step-change. Antarctic ice recovered to near-average by end of 2025. "Unprecedented" in 47 years of satellite data is not unprecedented in geological time.
• AMOC meltwater causing cold: The AMOC has not declined since 1963 [Foukal et al., 2025]. The method that claimed decline (SST proxies) has been shown to be unreliable. European cold is driven by La Niña and polar vortex dynamics.
• Antarctic volcanism irrelevant: 285 seismic events under the fastest-melting glaciers. 85,000 earthquakes at a "dead" volcano. 231 subglacial lakes heated from below. A positive feedback loop between ice loss and eruptions. This is not irrelevant.
• EPA paper is "opinion": The paper draws on work by Christy, Curry, Koonin, McKitrick, and Spencer — published scientists with verifiable track records. Dismissing it without addressing its specific claims is not a rebuttal.

Science is not about consensus. It's not about authority. It's about data. When the data disagrees with the narrative, the narrative needs updating. Not the data.

And when someone answers your questions with six cited sources, the appropriate response is to engage with the evidence. Not to accuse them of losing their point.

References