There’s something climate deniers consistently get right: they pick exactly the right year to start their graph. Not just any year. The perfect one for their purposes. 1998.
Why 1998? Because that year saw a massive El Niño — the weather phenomenon that temporarily warms the tropical Pacific waters and spikes the planet’s average temperature — which left a peak in the historical record that took nearly two decades to be permanently surpassed. If you take that peak as your starting point and measure through 2012, you can “prove” that warming has paused. Or even reversed. The statistical trick has a technical name: cherry-picking, selectively choosing data. It works exactly like claiming the stock market “hasn’t gone up” by picking the dot-com bubble peak of 2000 as your starting point.
The question that actually matters is different: how do we know exactly how much the Earth has really warmed? Across the entire planet, since reliable records began, with every bias identified and corrected.
The Planet’s Collective Thermometer
There’s no single global thermometer. What exists is thousands of weather stations spread across continents, buoys and ships recording ocean temperature, and satellites measuring atmospheric temperature from orbit. Three main organizations take all that information and integrate it into a coherent figure: NOAA (the US oceanic and atmospheric agency), NASA (specifically its GISS institute in New York), and HadCRUT (a collaboration between the UK Met Office and the University of East Anglia). All three use partially different data and different methodologies. And all three reach practically identical conclusions [1].
When institutions from different countries, with different interests, with teams of scientists who don’t coordinate with each other, arrive at the same result using independent methods, the probability of a shared systematic error — or of a conspiracy, which is what some people insinuate — approaches zero. This is the logic of independent replication, which is exactly the mechanism that makes science work better than any other way of knowing the world.
How is that record built? On land, surface weather stations are used: thermometers that have been running for decades, some for over a century, recording maximum and minimum temperatures. In the ocean — which covers 71% of the planet’s surface and can’t be ignored — data is combined from ships, anchored buoys, drifting buoys, and, since the nineties, the Argo network of automated floats that descend to 2,000 meters and surface to transmit their readings via satellite. There are over 3,000 of them currently drifting through the oceans right now. It’s not a thermometer. It’s a network of thermometers so dense that geographic coverage is, in the modern era, practically complete.
The bottom line, in numbers: in 2024, every group that measures global temperature independently — including Berkeley Earth and the European Copernicus service — confirmed it was the warmest year since instrumental records began around 1850 [1]. Global average temperature exceeded the pre-industrial baseline (1850-1900) by approximately 1.47°C according to NASA’s estimate [2], or 1.55°C according to the World Meteorological Organization [3]. The numbers don’t match to the decimal point. That’s not a problem: it reflects legitimate methodological differences in how ocean temperature is estimated for the 19th century, when data is scarcer. The uncertainty range is known and quantified.
What No One Tells You About the Adjustments: They Cool the Past, Not Warm the Present
One of denialism’s favorite arguments is that scientists “adjust” historical data upward, artificially inflating the warming. It’s an argument that sounds reasonable if you don’t know anything about the subject, and it falls apart the moment you actually look at the direction of the corrections.
The most significant corrections to the historical record run in the opposite direction from what the denialist argument assumes: they cool the past to make it comparable to the present, which in effect makes the observed warming look larger. Not because of any conspiracy. Because of a 19th-century naval engineering problem nobody anticipated, because nobody was thinking about climate when they designed it.
Until the 1940s, most ocean temperatures were measured using a method as quaint as it was problematic: sailors would throw wooden or canvas buckets into the water, haul them up on deck, and stick a thermometer in. The problem is that the water cools during the process, especially in uninsulated canvas buckets, depending on the ship’s speed, air temperature, and how long the measurement took. When, by the mid-20th century, measurement shifted mostly to engine cooling water intake — which comes directly from hull intakes and runs slightly warmer than surface water — the uncorrected record showed an artificial temperature jump that didn’t exist in reality [5].
Scientists at HadCRUT and other groups have spent decades modeling and correcting that bias. The method involves physically reconstructing how much water cools in each type of bucket (wood, canvas, rubber), comparing simultaneous measurements taken with different methods on the same ships, and cross-referencing that data with nighttime marine air temperature records, which serve as an independent reference point. It’s a piece of climate archaeology that’s anything but arbitrary [5].
The result is that the past ends up slightly cooler than the original thermometers recorded, and the observed warming turns out larger, not smaller. It’s not a matter of “manufacturing” the adjustments.
There are other equally well-documented types of corrections. Time-of-observation bias — whether a thermometer was read in the morning or afternoon affects the recorded maximum and minimum — instrument changes (from mercury thermometers to electronic sensors), station relocations. Every correction is logged, published, and can be audited by anyone. In fact, the great strength of the system is that the raw, uncorrected data is also public. Anyone who wants to run the numbers without corrections can do so and publish it. People have. The result is essentially the same [4].
The Urban Heat Island Effect: Real at the Local Scale, Irrelevant at the Global Scale
The urban heat island effect is a genuine, well-documented phenomenon: cities are warmer than the surrounding countryside because asphalt and concrete absorb more solar radiation than vegetation, buildings reduce ventilation, and humans and our activities generate heat directly. A weather station that sat on the outskirts of a mid-sized city in 1950 and is now surrounded by apartment blocks, shopping centers, and parking lots can register local warming that has nothing to do with global climate. The denialist argument says this contaminates the global record. That’s a reasonable argument that deserves a concrete answer, not a pat on the head.
That concrete answer arrived in 2012 from an unexpected source. Berkeley Earth was an independent review project of temperature records, founded by Richard Muller, a Berkeley physicist who described himself as a “properly skeptical” scientist with serious doubts about the solidity of existing records. His team included climatologist Judith Curry, at the time one of the most vocal critics of the climate consensus. They received partial funding from the Koch foundation, tied to the oil industry. Not exactly the team you’d expect to end up confirming global warming.
Their analysis used more than 36,000 weather stations, five times more than previous NASA and NOAA analyses, and explicitly included rural stations far from any urban area for comparison. Conclusion: the urban heat island effect is real at the local scale, but its contribution to the global average land temperature is indistinguishable from zero [4]. Cities cover less than 1% of the Earth’s surface. The global average is dominated by the ocean — which nobody has paved over — and by thousands of rural stations that don’t have this problem. Warming in urban and rural areas is statistically the same.
Muller, who had started the project as a skeptic, published the results in July 2012 in a New York Times column with a blunt title: “The Conversion of a Climate-Change Skeptic.” What’s notable isn’t that he changed his mind — the evidence was there for anyone who looked — but that he did it publicly and in that tone. That has a name: intellectual integrity. It’s in short supply.
The 2023-2024 Jump: Too Fast
Here’s something concerning: the warming of the last two years has been faster than models expected, and scientists don’t yet have a fully satisfactory explanation.
Part of the jump is well explained: a strong El Niño from late 2023 through mid-2024 pushed temperatures up. But when Gavin Schmidt, director of NASA’s GISS and one of the most respected climatologists in the world, looks at the numbers, he says the size of the jump exceeds what El Niño alone would justify. Berkeley Earth, in its 2024 temperature report, says something similar: the peak was “larger than expected” and known factors “do not fully explain the magnitude” [6]. That’s not great news.
At least two additional hypotheses are under investigation. The first has to do with sulfate aerosols. Cargo ships burned high-sulfur fuel for decades, which oxidized in the atmosphere to form particles that reflected part of the sun’s radiation, acting as an unintentional shade over the planet. In 2020, the International Maritime Organization mandated a drastic cut in the sulfur content of marine fuel. The result is that this artificial shading has decreased, and more radiation now reaches the surface. This isn’t a speculative hypothesis: the physical mechanism is well known. What’s unclear is how much of the recent warming can be attributed to it.
The second hypothesis involves the underwater volcanic eruption of Tonga in January 2022, one of the most energetic eruptions recorded in decades, which — unlike land eruptions that inject sulfur dioxide particles that temporarily cool the climate — injected unusual amounts of water vapor directly into the stratosphere. Water vapor is itself a greenhouse gas. If that stratospheric vapor has lingered longer than expected, it could have contributed to the anomalous temperatures of 2023 and 2024. This is a conjecture — anchored in real physics, but still unconfirmed as the main cause.
Pinpointing the cause matters, because if part of the recent warming is due to the reduction in marine aerosols, and that reduction has already happened and is permanent, then the baseline for global temperature has shifted up a notch that isn’t coming back down. And if climate models underestimated that contribution, they might also be underestimating future warming. The real issue is that if neither hypothesis turns out to be responsible, there’s a third, unidentified factor warming the planet — which is the concerning part.
Why the Consensus Matters More Than Any Single Data Point
It’s tempting, when discussing global temperature, to get stuck arguing over a single data point, a single year, a single station, a single correction. Deniers are very good at provoking exactly that kind of debate — possibly funded by companies affected by climate policy? — because there’s always something to question in the details. Science isn’t perfect, and historical records carry real uncertainties.
But there’s something harder to attack than a single data point: the pattern of convergence across independent sources. NASA, NOAA, and HadCRUT don’t share real-time data or coordinate their methodologies. Berkeley Earth was created by skeptics with oil-industry funding. The European Copernicus service is run by the European space agency. The Japanese have their own independent analysis. They all point in the same direction.
And the instrumental record isn’t alone. Climate proxies — indirect temperature data extracted from tree rings, corals, sediments, air bubbles trapped in glacier ice — paint the same picture: current warming has no precedent in the last two thousand years of documented climate history. Thermometers started recording in 1850. Antarctic ice cores go back 800,000 years. And none of those records show anything comparable to what has happened in the last century.
Whether temperature goes up or down in any given year isn’t the data point. A 1998 peak caused by El Niño isn’t the data point. The arrow is the data point. And that arrow has been pointing in the same direction for over a century, with a notable acceleration in recent decades, measured by instruments from five different countries, corroborated by natural climate archives spanning millions of years, reviewed by thousands of scientists who don’t know each other and who would be thrilled to find the error that brings the whole thing down.
REFERENCES
[1] NASA Goddard Institute for Space Studies. 2024 Was the Warmest Year on Record. NASA Science, January 2025. Reliable
[2] NOAA National Centers for Environmental Information. 2024 Was the World’s Warmest Year on Record. NOAA, January 2025. Reliable
[3] World Meteorological Organization. WMO Confirms 2024 as Warmest Year on Record at About 1.55°C Above Pre-industrial Level. WMO, January 2025. Reliable
[4] Berkeley Earth. Global Temperature Report for 2024. Berkeley Earth, January 2025. / Berkeley Earth FAQs: Urban Heat Island Effect. Available at berkeleyearth.org. Reliable
[5] Carella, G. et al. Measurements and models of the temperature change of water samples in sea-surface temperature buckets. Quarterly Journal of the Royal Meteorological Society, 2017. DOI: 10.1002/qj.3078. Reliable
[6] Berkeley Earth. Global Temperature Report for 2024 (section on additional factors in the 2023-2024 warming). January 2025. Caveats apply — the section on additional factors behind the 2023-2024 temperature jump (such as reduced shipping aerosols or the solar cycle) involves causal attribution still under active discussion, not the measurement of the record itself, which is well established.