The previous episode documented over a century of scientific warnings -- predictions made by physicists and chemists who looked at the machine described in Episode 2 and concluded, from the basic behaviour of carbon dioxide in the atmosphere, that measurable warming was not merely possible but physically inevitable. Those predictions were specific. They were quantitative. They were published.
This episode is about what has been measured since.
Not by one group. Not with one instrument. Not from one country. What follows is the record produced by five independent temperature reconstructions maintained by competing institutions on four continents, using different raw data and different mathematics -- and by ice cores, ocean-sensing robots, radar satellites, microwave imagers, gravity-measuring spacecraft, glacier surveys, and seawater chemistry. Every one of these systems uses a different physical principle. They share no common method that could produce a shared error. They produce the same answer.
I’m Allen Schyf, and this is Polite Disputes. You’re listening to part 4 of The Acceleration, a grounded view of what human-caused climate change actually looks like.
To understand why the agreement matters, it helps to understand what a global temperature record actually is and how you build one.
You start with thermometers. Thousands of them -- at weather stations on land, on ships, on buoys drifting in the ocean. Each one records the temperature at its location, at its time, in its local conditions. A thermometer in downtown Tokyo and a thermometer on a prairie in Saskatchewan are not measuring the same thing. They are measuring the temperature of the air at two specific points on the surface of a very large sphere.
To turn thousands of local readings into a single global number, you need to do two things. First, you need to organize those readings into a grid -- divide the Earth’s surface into boxes, average the readings within each box, and then average the boxes. This is called gridding, and every research group does it slightly differently. Some use large boxes. Some use small ones. Some weight the boxes by area. The choices matter, because the thermometers are not evenly distributed. Europe and North America have dense coverage. The oceans, the poles, and much of Africa and Central Asia have far fewer stations. What you do about the gaps -- the boxes with no thermometer in them -- is one of the biggest methodological decisions in the field.
Second, you need to account for the fact that the conditions around a thermometer change over time. A weather station that was on the edge of a small town in 1920 may now be surrounded by concrete and asphalt, which absorbs heat and raises the local temperature. That warming is real -- it is physically happening at that location -- but it is not climate change. It is the station’s environment changing. A ship that measured ocean temperature by hauling a bucket of water onto the deck in 1920 is measuring something different from a modern buoy floating at a fixed depth. The process of identifying and correcting for these non-climate changes in the record is called homogenization. Every group does it differently, using different statistical techniques, and this is the second major source of methodological divergence.
Here is where it gets interesting -- and here is where the argument begins.
Five institutions have independently built global temperature records. NASA’s Goddard Institute for Space Studies in New York. NOAA’s National Centers for Environmental Information. The UK Met Office Hadley Centre, working with the Climatic Research Unit at the University of East Anglia. The Japan Meteorological Agency. And Berkeley Earth, a privately funded group in California. Each of these groups made its own decisions about gridding, about gap-filling, about homogenization, about which raw data to include and which to exclude. They are not copying each other’s homework. They are doing the same assignment independently, with different methods, and in some cases with different data.
NASA fills in the gaps aggressively. Their method estimates the temperature of an empty grid box by looking at stations within 1,200 kilometres -- a smoothing radius large enough to cover most of the Arctic, where stations are sparse but where warming has been fastest. This gives NASA near-global coverage, but it means some of their data points are estimates based on distant neighbours, not direct measurements. NOAA historically left more gaps unfilled -- if there was no thermometer in a grid box, that box was simply empty. Their global average was based on the parts of the Earth they could actually measure, which meant the Arctic was underrepresented. HadCRUT, the UK record, took a similar conservative approach for years, though their most recent version introduced statistical gap-filling that brought their results closer to NASA’s. The Japan Meteorological Agency maintains its own independent reconstruction from Tokyo, using its own blend of observations and its own analytical method.
And then there is Berkeley Earth.
In 2010, a physicist at the University of California named Richard Muller decided that the existing temperature records could not be trusted. Muller was not a climate scientist. He was a particle physicist -- trained in the culture of experimental rigour that produces Nobel Prizes at places like CERN and Fermilab. He had publicly criticized the methods used by NASA, NOAA, and the UK Met Office, arguing that their homogenization procedures introduced biases, that their station selection was questionable, and that the urban heat island effect had not been adequately addressed. He believed the warming trend was likely overstated.
So he built his own record from scratch.
The project received funding from multiple sources, including the Charles Koch Foundation -- a foundation controlled by one of the two brothers whose industrial and political network has been among the most prominent funders of organizations questioning mainstream climate science. The funding was not hidden. It was disclosed. The implication was clear to everyone involved: This was a project designed to find the errors in the existing temperature records, backed by money from people who had reason to hope errors would be found.
Muller assembled a team that included his daughter Elizabeth Muller as project manager and Saul Perlmutter, a Nobel laureate in physics, as a collaborator. They gathered the largest collection of raw temperature station data ever assembled -- over 36,000 stations, roughly five times the number used by the other groups. They developed a new statistical method for homogenization that they called the “scalpel” approach: Instead of manually adjusting station records when a known change occurred -- a station moved, an instrument was replaced, a building went up nearby -- they let the algorithm detect discontinuities in the data automatically and break the record at those points. No human judgment about which adjustments to make. The statistics handled it.
The results were published in 2012. Muller wrote an opinion piece in the New York Times titled “The Conversion of a Climate-Change Skeptic.” His opening line: “Call me a converted skeptic.” The Berkeley Earth analysis confirmed that global land temperatures had risen approximately 1.5 degrees Celsius over the past 250 years. The warming curve matched what NASA, NOAA, and the Met Office had been reporting for decades. The most thorough, most transparently funded, most methodologically independent audit of the temperature record in the history of the field had reproduced the result it was designed to challenge.
In 2024, all five groups -- along with a sixth, the European Centre for Medium-Range Weather Forecasts -- independently reported the same finding: 2024 was the warmest year in the instrumental record. The World Meteorological Organization, consolidating all six datasets, put the figure at 1.55 degrees Celsius above the 1850 to 1900 pre-industrial average. The previous ten years -- 2015 through 2024 -- were the ten warmest years on record. The five groups did not coordinate their analysis. They used different methods. In some cases, they used different raw data. They arrived at the same conclusion.
Berkeley Earth’s own annual summary notes that all monitoring groups “produce a similar understanding of recent climate change” and that “different groups use different data and methods, but arrive at broadly similar conclusions.” The remaining disagreements are small and concentrated in the nineteenth century, where data are sparse. In the modern era, the records are effectively indistinguishable.
The thermometer record goes back to the mid-1800s. To see further, you need a different kind of instrument.
In Antarctica, snow falls and does not melt. Each year’s snowfall compresses under the weight of the next, and the next, and the next, until it becomes ice. As the snow compresses, it traps tiny bubbles of air -- actual samples of the atmosphere at the time the snow fell. Drill a core of ice from the Antarctic ice sheet and you hold, in your hands, a physical archive of ancient air. The European Project for Ice Coring in Antarctica -- EPICA -- drilled to 3,260 metres near bedrock at Dome C and recovered an atmospheric record stretching back 800,000 years.
Over those 800,000 years, atmospheric CO2 oscillated between approximately 180 parts per million during ice ages and approximately 280 to 300 parts per million during warm periods. Eight full glacial cycles. The CO2 level never exceeded 300. The current level -- above 425 parts per million -- is not at the high end of the range. It is entirely outside the range. It is roughly 50 per cent above the ceiling that held for the entire span of time captured in the ice.
Thermometers measure the temperature of the air. Ice cores measure the composition of ancient atmospheres. But the most important number for understanding how much extra energy the climate system is actually holding is not in the air at all. It is in the ocean.
When greenhouse gases trap additional heat, more than 90 per cent of that energy goes into the water. The ocean is the planet’s heat ledger -- the running total of how much extra energy has accumulated in the system. Surface air temperature fluctuates from year to year with weather patterns, volcanic eruptions, and cycles like El Nino. Ocean heat content does not. It is the signal without the noise.
Measuring it requires an instrument that can reach into the deep ocean and report back. Since 2000, the Argo network has provided this. Approximately 4,000 autonomous robotic floats are distributed across the world’s oceans. Each float drifts at a parking depth -- typically around 1,000 metres -- and every ten days adjusts its buoyancy to dive to 2,000 metres, then slowly rises to the surface, recording temperature, salinity, and pressure as it ascends. At the surface, it transmits the profile via satellite, then sinks again.
In January 2025, a team led by Lijing Cheng at the Chinese Academy of Sciences published the annual ocean heat assessment. In 2024, the upper 2,000 metres of the ocean absorbed an additional 16 zettajoules of energy compared to 2023 -- confirmed independently by two other datasets. A zettajoule is ten to the twenty-first power joules. To put that in a unit that might mean something: The additional heat the ocean absorbed in a single year was roughly 140 times the total electricity generated by every power plant on Earth in 2023. The ocean is absorbing, every year, an amount of surplus energy that dwarfs the entire global energy economy. And the trend is not levelling off.
Sea level is measured from orbit. Since 1992, a series of satellites -- TOPEX/Poseidon, then the Jason series, now Sentinel-6 -- have bounced radar pulses off the ocean surface and timed the return to calculate the distance between the satellite and the water below, referenced against precise knowledge of the satellite’s own orbit. The continuous record shows that global mean sea level has risen approximately 10 to 11 centimetres since 1993. The rate has accelerated from approximately 2.1 millimetres per year in the early 1990s to approximately 4.5 millimetres per year by 2023 -- more than doubling in three decades. Two mechanisms drive it: Water expands as it warms, and melting land ice adds water to the ocean. Both are operating simultaneously.
Arctic sea ice is measured by satellite microwave sensors, which work because ice and open water emit microwave radiation at different intensities -- a difference that can be detected regardless of cloud cover or darkness, making it possible to map ice extent continuously, year-round, from 1979 to the present. The long-term trend is a decline of approximately 12 per cent per decade in the September minimum -- the annual low point. The 19 lowest September minimums on record have all occurred in the last 19 years.
The twin GRACE satellites -- launched in 2002 and succeeded by GRACE-FO -- measure something different again. They orbit in formation, and by detecting tiny changes in the distance between them as they pass over different parts of the Earth, they map variations in the gravitational field -- variations that change when large masses of ice are gained or lost. The data show that Greenland is losing approximately 264 billion tonnes of ice per year. Antarctica is losing approximately 135 billion tonnes. These are not estimates derived from temperature. They are direct measurements of mass, made by detecting the change in the pull of gravity as the ice disappears.
Global glacier surveys, compiled by the World Glacier Monitoring Service, show that 2022 through 2024 was the largest three-year glacier loss on record. Ocean acidity -- measured directly by sampling seawater and testing its pH -- is increasing at 0.017 units per decade as the ocean absorbs CO2 from the atmosphere, a rate of chemical change faster than anything in the geological record of the past 300 million years.
Thermometers. Trapped air in ancient ice. Robotic ocean profilers. Radar altimeters in orbit. Microwave satellite sensors. Gravity-measuring spacecraft. Glacier surveys. Seawater chemistry. Each of these measurement systems uses a different physical principle. A thermometer has nothing in common with a gravity satellite. An ice core has nothing in common with a microwave sensor. There is no shared instrument, no shared algorithm, no shared institutional incentive that could produce agreement among all of them simultaneously. The only thing that can make them all point in the same direction is if the thing they are measuring is real.
They all point in the same direction.
The measurements describe what is happening to the physical system -- the atmosphere, the ocean, and the ice. What follows describes what happens when those changes arrive at the world people actually live in.
But first, a point that is easy to miss and essential to understand: The climate system did not become dangerous when we started adding energy to it. It was already dangerous.
The first episode of this series documented what the geologic record shows -- that the climate system is capable, on its own, of swings large enough to bury continents under ice or melt that ice entirely. The Younger Dryas, roughly 12,900 years ago, plunged temperatures back toward near-glacial conditions within decades. The Bolling-Allerod warming that preceded it produced several degrees of change within centuries. These are not ancient history in the sense that the Permian-Triassic is ancient history. They happened within the timeframe of anatomically modern humans. People were alive for them. The system that produced those swings is the same system operating today. It has always been capable of civilization-threatening variation. It has always been loaded.
What the measurements in this episode document is that we are adding energy to that system. Not a small amount. The ocean alone absorbed 16 zettajoules of additional heat in a single year. The atmosphere now holds 50 per cent more CO2 than at any point in 800,000 years. The ice is responding. The sea level is responding. The chemistry of the ocean is responding.
The effect of adding energy to a system that is already capable of extreme behaviour is not complicated. It is the same physics that governs any oscillating system: More energy means wider swings. Wider swings mean that the tail events -- the droughts, the floods, the heat waves, the storms that have always existed -- move from rare to regular. The system does not produce new kinds of disasters. It produces the old kinds more often, and at greater intensity.
Episode 1 introduced the characters who are vulnerable to this: The body, calibrated to consistency at the molecular level. The society, anchored to geography it cannot move and planning horizons that cannot see what is coming. The biosphere, already under rate-of-change pressure that matches the opening signature of previous mass extinction events. These are rate-limited systems. Their vulnerability is not to any particular temperature or any particular event. It is to the frequency and intensity of events arriving faster than the systems can absorb and recover from them.
The claim being made here requires precision, because imprecision is where credibility is lost. No individual weather event can be attributed solely to the long-term increase in atmospheric CO2. Weather has always been volatile. Droughts have always occurred. Floods have always occurred. The claim is not that any specific disaster was “caused” by climate change. The claim is that a system holding more energy oscillates more widely, and a system oscillating more widely produces extreme events more often. This is physics, not politics. It is the same principle that makes a pot of water on a higher flame splash more than a pot on a lower one.
The pattern that emerges from the documented cases is consistent. Environmental pressure exceeds adaptive capacity. Food systems fail or infrastructure breaks. People move. The receiving communities strain. Institutions buckle. Jobs disappear. Competition for diminishing resources produces the responses that competition has always produced.
In 2022, monsoon rainfall submerged roughly one-third of Pakistan. Thirty-three million people were affected. More than 1,700 were killed. Eight million were displaced. 1.2 million head of livestock drowned. The World Bank assessed damages and losses at over 30 billion US dollars. A rapid attribution study found that the heaviest rainfall over the affected provinces was approximately 75 per cent more intense than it would have been had the climate not warmed by 1.2 degrees. The flooding did not occur because of climate change. Monsoon floods have occurred in the Indus basin throughout recorded history. The flooding was intensified because the atmosphere now holds more moisture -- approximately seven per cent more for every degree of warming, following the Clausius-Clapeyron relation, one of the oldest established results in thermodynamics. Eight to nine million more people were pushed toward the poverty line. Recovery, years later, remains incomplete.
In 2023, Canada experienced its worst wildfire season on record. Approximately 15 million hectares burned -- more than twice the previous national record. More than 200 communities were evacuated. Over 230,000 people were displaced. Smoke crossed the Atlantic and reached Europe. In northeastern British Columbia, a single two-year period burned as much forest as the preceding 72 years combined. Two years earlier, the village of Lytton, BC had recorded 49.6 degrees Celsius -- the highest temperature ever measured in Canada -- and burned to the ground the following day. Over 600 people died in British Columbia during the broader heat event. A subsequent attribution study found the heat dome that produced those temperatures would have been, in the study’s language, “virtually impossible” without the long-term warming trend.
In 2022, drought dropped the Rhine -- the river that carries roughly 80 per cent of Germany’s inland waterway freight -- below navigable thresholds. Barges that normally carry full loads were forced to run at 25 to 30 per cent capacity, because drawing too much water would run them aground. The Kiel Institute for the World Economy estimated that every day the water stayed below the critical level at the Kaub chokepoint cost German industrial production roughly one per cent. It would take approximately 40 trucks to replace one barge. The comparable low-water event in 2018 cost German industry nearly three billion euros. This is not a story about a river. It is a story about supply chains, manufacturing schedules, heating fuel deliveries, and the jobs that depend on all of them.
In the Mekong Delta -- home to 18 million people and over half of Vietnam’s rice production -- saltwater is intruding five to seven kilometres further inland than normal, and arriving earlier in the season than it used to. The mechanism is straightforward: Sea level rises, upstream dams reduce freshwater flow, groundwater extraction causes the land itself to sink, and the saltwater advances. Farmers whose families have grown rice on the same land for generations are watching their fields become too saline to plant. The agricultural zone is moving, and the infrastructure -- the irrigation systems, the processing facilities, the communities -- cannot move with it.
The scholarship on climate and conflict is deliberately cautious, and the caution is important to preserve. Syria’s 2006 to 2010 drought -- the worst in the modern instrumental record for that region -- destroyed agriculture in the northeastern breadbasket and displaced approximately one million rural people into cities already under pressure. The civil war erupted in 2011. A widely cited 2015 study linked the drought to a warming-driven drying trend and to the social vulnerabilities it exposed. One of the study’s co-authors, the climate scientist Richard Seager, was explicit about the limits of the claim: “We’re not saying the drought caused the war.” Other scholars argue that political mismanagement, economic policy, and sectarian tensions were more fundamental. The honest framing -- and the only one the evidence supports -- is that the drought was a contributing amplifier in a system that had multiple pre-existing fractures. The pattern is the same: Environmental pressure compounding existing institutional fragility until something breaks.
The industry whose entire business model depends on accurately pricing risk has already reached its own conclusion. Swiss Re, one of the world’s two largest reinsurers, reported that global insured losses from natural catastrophes reached 137 billion US dollars in 2024. The following year, insured losses again exceeded 100 billion -- the sixth consecutive year above that threshold. Munich Re, the other major global reinsurer, titled its 2024 annual report on natural catastrophe losses “Climate change is showing its claws.” Total economic losses in 2024 were estimated at 320 billion US dollars, of which approximately 140 billion were insured. These are not projections. They are booked losses -- money that has already been paid out. The institutions that price risk for a living have repriced.
The repricing does not stay in boardrooms. It arrives at the scale of daily life.
In regions where wildfire or flooding risk has increased, insurers are not simply raising premiums. They are withdrawing coverage entirely -- declining to write new policies or refusing to renew existing ones, because the actuarial math no longer supports insuring structures in those locations at any price the market will bear. A homeowner who cannot obtain insurance cannot obtain a mortgage. A community where homes cannot be mortgaged cannot sustain property values. The mechanism is financial, but the effect is geographic: Certain places are becoming, in economic terms, uninhabitable -- not because a disaster has struck, but because the probability of one has crossed the threshold at which the institutions that absorb risk will no longer do so.
Agricultural zones are shifting. This is not a projection about what might happen under various emissions scenarios. It is a description of what crop suitability maps already show. In late 2024, arabica coffee futures reached a 47-year high -- rising over 70 per cent in a single year -- driven by drought in Brazil and harvest failures in Vietnam, the world’s two largest coffee producers. Multiple studies project that roughly half the land currently suitable for arabica cultivation could become unsuitable by 2050 as the viable growing zone moves upslope. Coffee is an example, not an exception. The same logic applies to any crop whose viable range is defined by temperature and rainfall patterns that are now shifting faster than agricultural infrastructure can follow.
In Australia, the Great Barrier Reef experienced its most spatially extensive bleaching event on record in 2024 -- the fifth mass bleaching since 2016, part of the fourth global bleaching event. The Australian Institute of Marine Science recorded the largest annual coral cover declines in 39 years of monitoring. A reef system that took thousands of years to build is degrading in annual increments. The fisheries, the tourism, the coastal protection the reef provides -- all of these are economic activities that employ people, sustain communities, and depend on an ecosystem that is failing faster than it can recover between events.
None of this is abstract. A garden that can no longer be planted to the same schedule. A grocery bill that reflects harvest failures on another continent. An insurance renewal that arrives with a number that no longer makes sense. A job in an industry that depends on a supply chain that depends on a river that depends on a snowpack that is no longer reliable. The volatility documented in the first half of this episode does not arrive as a headline. It arrives as a change in the cost, the availability, and the reliability of things that were never supposed to change.
The measurements documented in this episode were produced by independent instruments, operated by competing institutions, in different countries, using different physical principles. Thermometers, ice cores, ocean robots, radar altimeters, microwave sensors, gravity satellites, glacier surveys, seawater chemistry. Every one of them points in the same direction: The system is accumulating energy, and the accumulation is producing the effects that were predicted -- by Arrhenius in 1896, by Callendar in 1938, by Revelle and Suess in 1957, by the President’s Science Advisory Committee in 1965, by Sawyer in 1972, by the Charney panel in 1979, and by Hansen before the United States Senate in 1988.
The energy does not arrive at human societies as a number on a graph. It arrives as volatility -- as seasons that no longer follow the patterns that infrastructure, agriculture, and daily life were built to expect. The disasters are not new. The frequency is. The disruptions do not stop at borders, and they do not distinguish between countries that produced the emissions and countries that did not. They follow the logic that disruption has always followed: Pressure finds the weakest point, and the weakest point is wherever the gap between what the system expects and what it receives is widest.
The first episode of this series established that rate of change -- not change itself -- is what the geological record shows is dangerous. The second described the machine producing that rate. The third documented the warnings. This episode has documented the confirmation. Every tool the series has built now points at the same question, and it is not the question most people expect.
The question is not whether the planet will survive what is happening. The planet has survived worse -- five times. The question is what “surviving” looks like at the scale that matters to us, and whether the word means what we think it does when applied to a rock that has been orbiting a star for four and a half billion years.
That is the subject of the final episode.
This has been an episode of Polite Disputes. Thanks for listening.
