14.2 — How We Know the Climate Is Changing

In 1856 Eunice Newton Foote, an American scientist, filled glass cylinders with different gases, put them in sunlight, and measured how hot each became.

The cylinder of carbon dioxide heated most and cooled slowest. She wrote that an atmosphere of that gas would give our earth a high temperature, and that if at one period of its history the air had mixed with it a larger proportion than at present, an increased temperature must have necessarily resulted.

Her paper was read at a scientific meeting in 1856 — by a man, because women did not present.

John Tyndall independently established the mechanism with far more precise instruments in 1859. Svante Arrhenius calculated the sensitivity in 1896 (Chapter 14.1).

The physics is 170 years old. This chapter is about how the observation caught up with it, and about how to tell a genuine uncertainty from a manufactured one.

The physical mechanism, once more

Chapter 2.1 gave it and it is worth stating precisely because it is where the argument should start.

The Sun radiates at short wavelengths; the Earth at long infrared wavelengths. A molecule absorbs infrared only if its vibrations change the distribution of electric charge within it. Symmetric two-atom molecules — nitrogen and oxygen, 99 percent of the atmosphere — cannot. Carbon dioxide, water vapour, methane and nitrous oxide can.

This is measurable in a laboratory and it is measured every day. The absorption spectra of these gases are used in industrial gas sensors, in medical capnography and in military infrared systems. If carbon dioxide did not absorb infrared at the wavelengths in question, a great deal of ordinary technology would not work.

The temperature record

Instrumental measurements from around 1850, from land stations, ships and buoys.

And the record requires correction, which is where the accusation of manipulation comes from, so the corrections should be stated openly.

Stations move, instruments change, and observation times shift. Ships changed from measuring intake water temperature to buckets and back, and the two methods differ systematically. Cities grow around stations, producing an urban heat island effect (Chapter 2.7).

Each of these is corrected, the methods are published, and the corrections are checked.

Three points settle the manipulation question.

Independent groups produce independent series. NASA, NOAA, the UK Met Office and Berkeley Earth all construct global temperature records with different methods and different station selections, and they agree closely.

Berkeley Earth is the specific case worth knowing. It was set up by the physicist Richard Muller, who was publicly sceptical of the existing records and of the urban heat island corrections, and was funded partly by donors sympathetic to that scepticism. His team reconstructed the record from scratch with new methods. The result matched the existing series, and Muller published a public statement that he had changed his mind. That is what the process is supposed to do and it did it.

And the urban heat island effect specifically has been tested by comparing rural-only stations, by comparing night and day trends, and by using satellite data — and the warming trend is present in all of them.

Satellite measurement of atmospheric temperature since 1979 is an independent system with its own correction history — early satellite datasets showed cooling until orbital decay and calibration errors were identified and corrected, and they now agree with the surface record.

And the current position. Global average surface temperature has risen by approximately 1.2 to 1.3 °C above the 1850–1900 baseline. The last ten years are the ten warmest in the instrumental record.

The other observations

A temperature series alone would be weak evidence. The strength is that everything else moves consistently.

Ocean heat content, measured by the Argo float network (Chapter 1.11), rising steadily — and this is the most important single indicator, because over 90 percent of the extra energy goes into the ocean and the ocean is far less noisy than the atmosphere.

Sea level, rising at an accelerating rate, from thermal expansion and from ice melt, measured by tide gauges and by satellite altimetry independently.

Ice. Arctic sea ice extent declining, particularly in summer. Greenland and Antarctic ice sheets losing mass, measured by satellite gravimetry. Glaciers retreating on every continent, documented photographically over more than a century.

Phenology. Spring events — leaf emergence, flowering, insect emergence, bird migration and arrival — occurring earlier, recorded in datasets some of which go back centuries. Japanese cherry blossom records from Kyoto extend over a thousand years and show the earliest flowering dates on record in the last two decades.

Species ranges shifting poleward and to higher elevations.

And humidity, sea surface temperature, growing season length and permafrost temperature all moving in the same direction.

The fingerprints

This is the part that distinguishes greenhouse warming from every alternative explanation, and it is the strongest evidence available.

If the Sun were the cause, the whole atmosphere would warm. What is observed is that the troposphere warms and the stratosphere cools (Chapter 2.1's layers). That is exactly what enhanced greenhouse trapping predicts: more heat retained below, less escaping to the layer above. A brighter sun cannot produce it.

If the Sun were the cause, days would warm more than nights. What is observed is that nights are warming faster than days, which is what reduced outgoing infrared predicts and what more incoming sunlight does not.

Winters are warming faster than summers, and high latitudes faster than the tropics — both predicted.

Satellites measuring outgoing infrared from space find reduced emission at exactly the wavelengths carbon dioxide absorbs, and instruments on the ground find increased downward infrared at those same wavelengths. The energy is going where the mechanism says it should.

And solar output has been measured directly by satellite since the late 1970s. It has been flat or slightly declining while temperature has risen sharply. The Sun and the temperature have been moving in opposite directions for four decades.

The models

What they are. Numerical simulations of the atmosphere and ocean on a grid, built from the same physical equations as weather models (Chapter 2.9), run for decades to centuries.

What they are not. They are not curve-fitting exercises. They are built from conservation of momentum, mass and energy, and they use parameterisations for processes smaller than the grid — with clouds the largest source of spread, exactly as in weather forecasting.

How they have been tested, and this is the part that matters.

Hindcasting. Run from 1900 with the actual historical forcings, they reproduce the observed temperature record including the mid-century flattening caused by industrial aerosols and the cooling steps after major volcanic eruptions.

Prediction. James Hansen's 1988 projections to the American Congress, run on the computers of the time, have been checked against what happened; the scenario closest to actual emissions tracks the observed warming reasonably well. A 2019 review of seventeen model projections published between 1970 and 2007 found that most were consistent with subsequent observations once the actual emissions were accounted for.

And Pinatubo. Hansen's group predicted the magnitude and duration of the cooling from the 1991 eruption before it happened (Chapter 1.7). The prediction was correct. That is a genuine out-of-sample test of the physics.

Where models are weakest: regional projections, precipitation, and the behaviour of clouds and ice sheets. These are stated as uncertainties in the assessments rather than concealed.

How the science was settled

Not by consensus being declared, which is the accusation, but by the alternatives failing.

Every proposed alternative explanation has been tested and each fails on a specific observation.

Solar variation — contradicted by direct measurement and by the stratospheric cooling fingerprint.

Cosmic rays affecting cloud formation — a real proposed mechanism, tested experimentally at CERN's CLOUD facility, and found too weak to account for the observed warming, with no trend in cosmic ray flux matching the temperature trend.

Natural ocean cycles — these redistribute heat within the system and do not add energy to it, and the ocean heat content is rising, which internal variability cannot produce.

Volcanoes — emit around a hundredth of human emissions (Chapter 14.1).

And "it is coming out of an ice age" — the orbital forcing (Chapter 2.8) has been in the direction of gradual cooling for several thousand years.

The consensus figure — that over 97 percent of publishing climate scientists attribute recent warming principally to human activity — comes from several independent studies of the literature and of scientist surveys. It is a description of where the evidence led, not an argument in itself, and the correct response to anyone citing consensus as proof is that the fingerprints above are the actual argument.

The genuine uncertainties

A chapter that claimed certainty about everything would be as misleading as one that denied everything, so here is what is actually open.

Climate sensitivity, with a likely range of about 2.5 to 4 °C — a factor that matters enormously for how much time there is.

Ice sheet dynamics. How fast Greenland and West Antarctica can lose mass, particularly whether marine ice cliff instability is real, which determines whether sea level rise this century is closer to half a metre or to more than a metre.

Tipping points. The Atlantic overturning circulation (Chapter 2.6), Amazon dieback, permafrost release. Whether thresholds exist, where they are, and whether crossing them is reversible are all open.

Regional projections, particularly for monsoon behaviour — which for India is the question that matters most and is among the least well constrained (Chapter 2.5).

And the social and economic responses, which are not physics at all and which dominate the projections of impact.

Where this shows up in your life

The distinction between a real and a manufactured uncertainty is the useful skill here.

A real uncertainty has a range, a stated method, and researchers actively working to narrow it. A manufactured one asserts that because something is uncertain, nothing is known.

And the history of the manufacture is documented. Internal documents from oil companies, published by researchers and by journalists, show that their own scientists produced projections in the 1970s and 1980s that were broadly accurate, while the companies funded public messaging emphasising uncertainty. A 2023 study in Science assessed Exxon's internal projections and found them as accurate as contemporaneous academic ones. The strategy is documented in internal memoranda from the tobacco industry, from which it was borrowed: the product is doubt, because doubt is sufficient to delay.

Chapter 16.4 makes the general method. The question to ask of any claim of scientific uncertainty is: what specific observation would settle it, has anyone made it, and what did they find.

What the next page covers

Chapter 14.3 covers what the warming actually does — heat and human physiology, water and the monsoon, food production, sea level and coastal cities, storms, ecosystems, and the specific consequences for India, which is among the most exposed large countries on almost every measure.