Was 2025 the Second or Third Warmest Year? Eight Datasets, Two Answers, No Argument.

On 14 January 2026, the World Meteorological Organization confirmed that 2025 was one of the three warmest years on record.

Not "the warmest." Not a single rank. That hedge is not diplomatic softness — it is the most accurate sentence available, and the reason why is worth understanding, because it applies to almost every climate number you will read this year.

The headline figures

• 1.44 °C ± 0.13 above the 1850–1900 average, from WMO's consolidated analysis of eight datasets.
• 2023–2025 are the three warmest years in all eight datasets. The three-year average is 1.48 °C ± 0.13.
• 2015–2025 are the eleven warmest years in all eight datasets.
• 2025 began and ended under La Niña — a pattern that normally cools the global average — and was still top-three.

That last point carries more information than the rank. La Niña is the thumb on the scale pressing down. It pressed, and 2025 came third anyway.

Why the rank is ambiguous

WMO doesn't run one thermometer. It consolidates eight independent datasets:

Six of these build from station, ship and buoy measurements, using statistical methods to fill gaps in sparsely-observed regions like the poles. Two — ERA5 and JRA-3Q — are reanalyses, which fuse observations with a physical model to produce a globally complete series.

Different methods, slightly different numbers, sometimes different ranks. Two of the eight are new to WMO's consolidation this year (DCENT and CMST), which is itself a reminder that the measurement is still being improved.

"Second or third" is not a fudge. It is the honest report of eight careful answers that do not perfectly agree — and they agree on everything that matters: the last eleven years are the eleven warmest, and 2023–2025 are the top three, unanimously.

The counterintuitive bit: we know the change better than the temperature

WMO puts the actual average global temperature in 2025 at about 15.08 °C — with an uncertainty of roughly ±0.5 °C.

Read that against the anomaly: 1.44 °C ± 0.13.

The uncertainty on how much it warmed is nearly four times smaller than the uncertainty on what the temperature is. That seems backwards until you see the mechanism: an anomaly is a difference, and a difference cancels most systematic error. If a station reads consistently half a degree high, that half degree is in both the baseline and today — and vanishes on subtraction.

This is why climate scientists talk in anomalies. Not to obscure anything. Because the anomaly is the number they actually know.

Where the uncertainty actually lives

Here is the detail almost nobody reports, and it inverts the usual assumption.

To express temperatures relative to pre-industrial, WMO computes each dataset's anomaly against 1981–2010, then adds an offset of 0.69 °C — the IPCC's estimate of the difference between 1981–2010 and 1850–1900.

The uncertainty in that offset is 0.12 °C. The combined uncertainty is 0.13 °C.

Almost all of the doubt in "1.44 °C above pre-industrial" comes from not knowing the nineteenth century precisely — not from any difficulty measuring 2025. Our thermometers are fine. Our time machine isn't.

So when someone says the warming figure is uncertain, they are right, and the uncertainty is in the baseline. That cuts both ways: 1.44 could be 1.31 or 1.57.

No, 1.5 °C has not been "breached" — and one widely-repeated claim is wrong

This is the most misreported sentence in climate coverage, so let's be exact.

The Paris Agreement's 1.5 °C limit refers to a long-term average, conventionally understood over decades — not any single year, and not a three-year stretch. A hot year is weather riding on top of climate. Exceeding 1.5 °C in one year does not mean the Paris limit is breached, any more than one cold winter would unbreach it.

We also want to flag a specific discrepancy, because we ran into it. Several outlets reported that 2025 completed the first three-year period averaging 1.5 °C above pre-industrial.

WMO's consolidated figure for 2023–2025 is 1.48 °C ± 0.13 — not 1.5.

The 1.5 framing may come from a single dataset rather than the consolidated eight, or from rounding. We are not going to print "1.5" as though it were the consolidated number when the primary source says 1.48. The difference is small. The difference between a number and a threshold everyone is watching is not.

And the honest caveat on our own caveat: 1.48 ± 0.13 includes 1.5. The uncertainty band straddles the line. What we can say is that WMO's central estimate is below it, and that a three-year average was never the test anyway.

The ocean took the heat, as it always does

About 90% of the excess heat from global warming ends up in the ocean, which makes ocean heat content a better long-run indicator than surface air temperature — it is far less jumpy.

A study in Advances in Atmospheric Sciences led by Lijing Cheng of the Chinese Academy of Sciences' Institute of Atmospheric Physics found:

• Global upper-2000 m ocean heat content rose by ~23 ± 8 Zettajoules from 2024 to 2025 — roughly 200 times the world's total electricity generation in 2024.
• About 33% of the global ocean ranked in its top three warmest conditions on record (1958–2025); about 57% in its top five.
• Sea surface temperature was 0.49 °C above the 1981–2010 baseline — 0.12 ± 0.03 °C cooler than 2024, consistent with La Niña, and still the third-warmest on record.

Note that last line does two things at once: it confirms La Niña did its job — the surface did cool relative to 2024 — and it shows what that bought. Third warmest.

Meanwhile the heat below the surface went up. The atmosphere cooled slightly and the system kept warming, which is exactly why the ocean number is the one to watch.

Why this is on AEDO1

Every reef in our coral bleaching report sits in that ocean heat content figure. The 23 Zettajoules is not an abstraction; it is the thing that bleached 84% of the world's coral.

But the reason this article exists is the eight datasets.

A field that publishes "second in two, third in six" instead of picking the better headline is a field being careful in public. The disagreement is small, documented, and explained — and it would be trivially easy to hide by quoting one dataset. The messiness is the credibility. WMO's full State of the Global Climate 2025 report, with greenhouse gases, sea level, glaciers and sea ice, is due in March 2026.