A note on scope: this piece explains the physics, which is mainstream science. What societies should do about it is a question of values and trade-offs, and it isn't covered here.
Energy in, energy out
Start with bookkeeping. Sunlight pours onto the Earth. About 30% of it bounces straight back to space off clouds, ice and bright surfaces, and the rest, about 240 watts per square metre averaged over the whole globe, is absorbed.
A planet can't keep absorbing energy without getting rid of it, or it would heat up forever. It gets rid of it the only way it can in the vacuum of space: by glowing, in infrared light you can't see. Every warm object does this, and the hotter it is, the more it radiates. The Stefan–Boltzmann law says the energy radiated goes up with the fourth power of temperature.
So ask: how warm must a planet be to radiate away 240 W/m²? The answer is 255 kelvin, about −18 °C. That is the Earth's "effective temperature", the temperature it has as seen from space.
Seen from space, the Earth glows like an object at −18 °C. Down on the surface it's 33 degrees warmer. The difference is the atmosphere.
At its real average of about 15 °C, the surface radiates roughly 390–400 W/m². Yet only about 239 W/m² escapes to space. Around 40% of what the surface sends upward is caught on the way out.
Why only some gases do it
Air is about 78% nitrogen, 21% oxygen and 0.9% argon. All three are almost completely transparent to infrared. The work is done by a small minority of gases, mainly water vapour and carbon dioxide, with methane and a few others. Their molecules can bend and stretch in ways that let them absorb infrared and re-emit it in all directions, including back down. Carbon dioxide is especially good at absorbing infrared near a wavelength of 15 micrometres, right where the Earth's own glow is strong. Satellites looking down see a clear "bite" taken out of the planet's infrared light at exactly that wavelength.
The "it's already saturated" objection, and why it fails
Here's a reasonable-sounding argument. Close to the ground, the air already absorbs about 90% of the infrared coming up from the surface. If CO₂ is already absorbing nearly everything, how can adding more make any difference?
The answer is that what matters is not where the infrared is absorbed, but where it finally escapes. Heat climbs through the atmosphere, absorbed and re-emitted again and again, until it reaches air thin enough that the radiation can get out to space. That escape happens high up, in the middle of the troposphere, where the air is much colder than at the surface.
| Simple picture | What actually happens | |
|---|---|---|
| Adding CO₂ | "Thickens a blanket that's already thick" | Raises the height from which infrared escapes |
| That height | — | Is colder, because air cools with altitude in the lower atmosphere |
| Result | "No effect" | A colder layer radiates less, so less energy leaves; the planet warms until it radiates 240 W/m² again |
Add CO₂ and the escape layer moves higher, to colder air, which glows more weakly. For a while the Earth sends out less energy than it absorbs, and it warms up until the books balance again. The effect never stops working; it just gets steadily weaker for each extra tonne, as explained below.
A warning about over-simple models. The classic textbook toy model, a single layer of air that absorbs all the surface's infrared, predicts a surface temperature of about 303 K (30 °C), too warm. The real atmosphere isn't a perfect absorber, and rising air (convection) carries a lot of heat upward directly. Serious climate models handle both.
Two centuries of the same physics
- 1824: Joseph Fourier argued that the atmosphere must keep the Earth warmer than it would otherwise be.
- 1856: Eunice Newton Foote found that sunlight heated moist air more than dry air, and air containing carbon dioxide most of all.
- 1859 onward: John Tyndall measured the infrared absorption of different gases and found that a small fraction of the atmosphere, mainly water vapour with CO₂ and others, does the absorbing.
- 1896: Svante Arrhenius made the first numerical estimate of the warming from doubling CO₂: 5–6 °C in total.
- 1901: Nils Ekholm first used the word "greenhouse" for it.
The name is a loose analogy. A real greenhouse stays warm mainly because its glass stops warm air rising and blowing away (it blocks convection). The atmosphere's effect works by slowing the escape of infrared radiation.
How much warming does extra CO₂ cause?
The effect of CO₂ is roughly logarithmic: each doubling adds about the same push, about 3.7 W/m² of extra trapped energy. Carbon dioxide was about 280 parts per million before industrialisation. NOAA's monthly average at Mauna Loa for September 2026 was 426 ppm, about 52% higher. By the standard formula (5.35 × ln(426 ÷ 280)) that is an extra push of about 2.2 W/m² from CO₂ alone.
How much warming that produces depends on feedbacks. On its own, a doubling of CO₂ would warm the Earth by roughly 1 °C. But a warmer atmosphere holds more water vapour, which is itself a greenhouse gas, and that amplifies the warming. Melting ice, which exposes darker surfaces, adds more. Clouds remain one of the biggest uncertainties. Pulling all the evidence together, the IPCC's 2021 assessment put the long-run warming from doubled CO₂ at about 3 °C, likely between 2.5 and 4 °C.
What has been measured so far: the IPCC found that the decade 2011–2020 was about 1.1 °C warmer than 1850–1900, and stated that human activities, mainly greenhouse-gas emissions, have "unequivocally" caused global warming.
Venus: the same physics, turned up
Venus shows what the same physics does with a lot more CO₂. Its atmosphere is about 96–97% carbon dioxide, with a surface pressure about 92 times Earth's. It reflects about 77% of the sunlight that reaches it, so despite being closer to the Sun it absorbs less sunlight than Earth does. Without its greenhouse effect its surface would sit around −41 °C. With it, the surface is about 462 °C.
Why this is a beautiful piece of physics
It is worth pausing on how much follows from so little: a law about how warm things glow, the fact that a few kinds of molecule can absorb infrared, and the fact that air cools with height. From those, you can work out the temperature of a planet, see why a gas measured in parts per million matters, and understand why Venus is an oven. Fourier, Foote and Tyndall had worked out the essentials before anyone had flown in an aeroplane.
The takeaway
The Earth absorbs about 240 W/m² of sunlight and must radiate the same back as infrared. A body doing that would average −18 °C, but the surface averages about 15 °C, because water vapour, CO₂ and a few other gases absorb outgoing infrared. Adding CO₂ doesn't "saturate". It raises the altitude from which heat escapes to colder air, which radiates less, so the planet warms until it balances again. Each doubling of CO₂ adds about 3.7 W/m², which feedbacks turn into a best estimate of about 3 °C of long-run warming. CO₂ has risen from about 280 to 426 ppm, and the world has warmed about 1.1 °C (2011–2020 compared with 1850–1900).
Try this: look up NOAA's Mauna Loa carbon dioxide page and find this month's number. Then use the formula above, 5.35 × ln(today's ppm ÷ 280), to work out the extra W/m² yourself. You'll be repeating, with a calculator, the arithmetic Arrhenius did by hand in 1896.
Sources & Further Reading
Sources
- Energy budget figures (about 30% reflected, ~240 W/m² absorbed, ~398 W/m² emitted at the surface, ~239 W/m² escaping, the 159 W/m² or 40% difference), the −18 °C (255 K) and 15 °C (288 K) temperatures, the 33 °C greenhouse effect, the ~90% absorption of surface infrared, the emission-height explanation, the 15 µm CO₂ band, the greenhouse/convection distinction, the history (Fourier 1824, Foote 1856, Tyndall 1859, Arrhenius 1896 with 5–6 °C, Ekholm 1901) and the Venus figures (about 462 °C, 96–97% CO₂, ~92 atm, 77% reflected, about −41 °C without its greenhouse effect): Wikipedia, "Greenhouse effect", and the sources it cites.
- Dry-air composition (78.08% nitrogen, 20.95% oxygen, 0.93% argon): Wikipedia, "Atmosphere of Earth".
- CO₂ forcing: the logarithmic dependence and Myhre et al. (1998) formula ΔF = 5.35 ln(C/C₀) W/m² (Wikipedia, "Radiative forcing"); about 3.7 W/m² per doubling, roughly 1 °C of no-feedback warming, the water-vapour feedback and IPCC AR6's 3 °C best estimate (likely 2.5–4 °C) (Wikipedia, "Climate sensitivity").
- CO₂ at Mauna Loa: NOAA Global Monitoring Laboratory, monthly mean for September 2026, 426.07 ppm (updated 9 October 2026). Pre-industrial ~280 ppm.
- Observed warming of 1.1 °C (2011–2020 vs 1850–1900) and the "unequivocally" statement: IPCC Sixth Assessment Report, as summarised in Wikipedia, "IPCC Sixth Assessment Report".
- Computed in Python: (240 ÷ σ)^¼ = 255.1 K with σ = 5.670 × 10⁻⁸ W m⁻² K⁻⁴; σ × 288⁴ = 390 W/m²; single-layer model 2^¼ × 255 K = 303 K; 5.35 × ln 2 = 3.71 W/m²; 5.35 × ln(426.07 ÷ 280) = 2.25 W/m²; 426.07 ÷ 280 = +52%.
Go deeper
- 🌐 NOAA: Trends in atmospheric carbon dioxide (Mauna Loa) — the live record, updated monthly
- 🌐 Greenhouse effect — the energy budget and history in detail
- 📘 The Discovery of Global Warming — Spencer Weart (how scientists pieced this together over 150 years)
- 📘 Global Warming: Understanding the Forecast — David Archer (the physics, done properly, for non-specialists)