The Earth Test

Earth's surface averages 288 K, 15 °C. Is that set by the pressure and the lapse rate, or by the greenhouse effect? Same planet, same sunlight, same air, same lapse rate. Two runs, and an energy ledger that has to balance.

In 60 seconds

396 W/m²infrared emitted by the surface, measured
240 W/m²leave Earth: the same as the sunlight it absorbs
156 W/m²never reach space: the air traps them. That is the greenhouse effect
255 Kthe surface with the same air but no infrared absorbers (278 K even reflecting nothing)

Run

Sunlight reflected

With nothing reflected Earth absorbs all 340 W/m² that reach it, the most it could ever get.

Nothing in the air can radiate: air that doesn't absorb infrared doesn't emit it either, so the ground is the only thing radiating to space.

More absorbing gas, higher emission level. The measured profile reaches the emission temperature at 5.1 km. Above 11 km the troposphere ends and the profile stops cooling, so the model stops there.

Hover or touch the measured curve to read altitude, temperature and pressure; click or tap to pin the reading.

Energy ledger, per square metre, planet average

Earth as a whole
Sunlight absorbed
Infrared sent to space
Sent − absorbed
The surface
Infrared emitted by the surface, σT⁴
Sunlight absorbed by the surface
Must come back from the air
Measured back radiation (Trenberth et al. 2009; Wild et al. 2013: 342)
Greenhouse effect
Surface infrared that never reaches space (emitted − sent to space)
Measured (396 − 240)156 W/m²

Who does what: the published attribution

Share of the present-day greenhouse effect, GISS ModelE (Schmidt et al. 2010; Lacis et al. 2010).

Water vapour~50%
Clouds~25%
CO₂~20%
Other gases~5%

Water vapour and clouds do most of the absorbing, but they condense and rain out: their amount follows the temperature. CO₂, methane and ozone don't. In the GISS model, removing those non-condensing gases makes the water follow them out: within 50 years the global mean falls to −21 °C, 34.8 K colder, and sea ice grows from 4.6% to 46.7% of the globe.

"Other gases" is the 25% of the non-condensing gases minus CO₂'s 20%. Lacis et al. 2010, GISS research brief.

What leaves the surface, what leaves Earth

Planck's law, Mλ = 2πhc² / λ⁵ / (ehc/λkT − 1), per square metre; the area under each curve is σT⁴. Both curves follow the run and the emission height above. Shaded: the main CO₂ band, about 13–17 µm, and the window, 8–13 µm, where clear air absorbs least; the percentages are the share of the surface's emission that falls in each. The curves are smooth Planck curves: the real spectrum leaving Earth has a bite where the CO₂ band is ("Show what a satellite sees"), but the totals are set by the temperatures.

Isn't CO₂ saturated?

At the centre of its 15 µm band the air is already opaque: there, more CO₂ changes little near the ground. But Earth doesn't radiate the band to space from the ground. It radiates it from aloft, from the level where the air above is thin enough to let it out. More CO₂ lifts that level to colder air, so less leaves, until the column below warms up; the edges of the band, not yet opaque, widen too. The result is a forcing that grows with the logarithm of the concentration: every doubling adds the same, about 3.7 W/m². A saturated band would give a flat line.

ppm: pre-industrial, today, then one, two and three doublings.

Forcing since 1750
Emission level higher by
Surface warmer, before feedbacks
With feedbacks, at equilibrium

Today's 2.17 W/m² since 1750 matches the IPCC's assessed CO₂ forcing, 2.16 W/m². Each doubling then adds the same 3.7 W/m²: the line keeps climbing.

Forcing: 5.35 × ln(C / 280) W/m² (Myhre et al. 1998); the IPCC AR6 assesses 3.93 ± 0.47 W/m² per doubling, including fast adjustments. Before feedbacks: right after the CO₂ is added, Earth radiates from higher and colder air and sends out less; balance comes back when the column has warmed by forcing / (4σTe³) = forcing / 3.77, and the emission level has risen by that warming divided by the lapse rate, 6.5 K/km. In the run above: emission height 5.25 km instead of 5.10, and the surface goes from 288.2 to 289.1 K. With feedbacks, mostly water vapour: about 3 K per doubling, likely range 2.5–4 K (IPCC AR6), for CO₂ alone once the oceans have caught up; the warming measured so far also depends on the other gases, the aerosols and the oceans' delay.

Out of balance: the warming in progress

The runs above show a planet in balance. The real Earth isn't, quite: it absorbs more than it sends back to space, and the difference is measured from orbit and in the oceans.

0.76 W/m²absorbed minus emitted, Earth's average for 2006–2020
×2the imbalance roughly doubled from 2005 to 2019, seen by satellites and by ocean floats alike
12 ZJa year, stored at that rate: 0.76 W/m² over the whole planet
89%of the heat gained since 1971 is in the oceans; 6% land, 4% melting ice, 1% air

The First Law, in time

Why does an imbalance last? Double CO₂ at once. Earth now radiates from 151 m higher, colder air, and sends out 3.7 W/m² less than it absorbs. That energy stays: it warms the ocean, and the surface with it, until the surface is 0.98 K warmer and the balance is back. Move the years.

Imbalance, absorbed − emitted
Surface warmer by
Energy stored, per m² of Earth

No feedbacks, one heat store: the top 100 m of ocean over 71% of the planet, C = 1025 × 3990 × 100 × 0.71 = 2.90 × 10⁸ J/(m²·K). The imbalance decays as 3.71 × e−t/τ W/m², with τ = C / (4σTe³) = 2.4 years; the energy stored is C × ΔT, the area under the imbalance curve. The real ocean is deeper and slower, and feedbacks, mostly water vapour, bring the final warming to about 3 K: the full response takes decades, and meanwhile CO₂ keeps rising. That is why the measured imbalance persists. Measured values: von Schuckmann et al. 2023 (0.76 ± 0.2 W/m², 2006–2020; 381 ZJ gained 1971–2020), Loeb et al. 2021 (CERES satellites and Argo ocean floats).

CO₂ or the Sun? The stratosphere decides

A brighter Sun and more CO₂ both warm the surface. They part ways higher up. Pick one, then look at what was measured.

Measured, 1979–2018: satellites (MSU, AMSU, SSU), radiosondes and GPS radio occultation (Steiner et al. 2020). The lower stratosphere, 15–25 km, cooled too, but there the main driver until the mid-1990s was ozone loss (IPCC AR6); the clean CO₂ signal is higher up, 25–50 km, and natural causes can't explain it (Santer et al. 2023). The Sun heats the stratosphere through the ultraviolet that ozone absorbs, so a brighter Sun would warm it. Natural drivers, the Sun included: −0.1 to +0.1 °C of the warming since 1850–1900 (IPCC AR6, SPM A.1.3). Arrows show the direction, not the size. Profile: U.S. Standard Atmosphere 1976.

The argument, step by step

  1. Earth is in energy balance. It absorbs about 240 W/m² of sunlight, averaged over the planet, and sends about the same back to space as infrared. The small difference, under 1 W/m², is the warming now under way.
  2. Without infrared absorbers, the surface can't be warmer than 255 K. Air that doesn't absorb infrared doesn't emit it either: the surface is the only thing that radiates, and all of its radiation escapes. So σT⁴, averaged over the surface, equals the 240 W/m² absorbed: 255 K, −18 °C, whatever the pressure, the winds, the convection or the lapse rate. A surface with hot and cold patches averages lower, not higher. Even reflecting nothing, the limit would be 278 K.
  3. The real surface emits 396 W/m², and only 240 leave. The other 156 W/m² of the surface's infrared is absorbed by the air. That is the greenhouse effect: the infrared radiation energy trapped by atmospheric gases and clouds. Raval & Ramanathan 1989 This step doesn't depend on how energy moves inside the atmosphere, only on how much leaves the surface and how much leaves Earth.
  4. The surface lives on what the air sends back. The surface absorbs about 161 W/m² of sunlight and emits 396. The air sends back down 333 W/m² of infrared, measured; the rest leaves the surface as rising warm air and evaporation. Trenberth et al. 2009; Wild et al. 2013
  5. The lapse rate fills in the shape. Earth radiates to space from about 5.1 km, where the air is at 255 K; below it the troposphere warms by 6.5 K per km: 255.02 + 6.5 × 5.097 = 288.15 K. The absorbers set the height, the lapse rate sets the slope. This step explains the profile; the 5.1 km and the slope are read from the measured profile, so 288 K comes out by construction. The proof is in steps 1 to 4.

The 5 km coincidence

A popular rule puts Earth's −18 °C at the "average" height of the atmosphere, the middle of its mass, and adds the lapse rate: no greenhouse effect needed. On Earth it almost works, because two different heights happen to be close.

EarthVenus
Emission levelwhere the measured profile reaches Tem5.10 km71.38 km
Middle of the masshalf of the air above5.48 km10.40 km
Middle-of-the-mass ruleTe + Γ zhalf255.12 + 6.5 × 5.477 = 290.72 Koff by −2.6 K226.63 + 7.145 × 10.40 = 300.94 Koff by +434.4 K
Emission-level ruleTem + Γ ze288.15 Kmeasured 288.15 K735.3 Kmeasured 735.3 K

On Earth the two heights are 0.4 km apart; on Venus, 61 km. The rule that works on both planets is the one where the absorbers set the height. Venus, in detail: The Venus Test.

If compression paid the bill

Suppose there were no greenhouse effect and the surface were kept at 288 K by the pressure. The surface loses more than the sunlight it absorbs, every second. Compression supplies energy only while the air is being compressed: gravity does work p₀ × v per square metre as the column sinks.

The air would have to sink at
That is, every year
Gravitational energy of the whole atmosphere, ∫ p dz
Its centre of mass sits at
All of it would pay the surface's bill for

A static atmosphere does no work: compression heats only while it lasts. Energy from the U.S. Standard Atmosphere 1976, 0–50 km, log-linear between levels, the column above 50 km added with a 7 km scale height; the bill is σT⁴ at 288.15 K minus the 161 W/m² of sunlight absorbed by the surface.

Radiation alone would be hotter

If the lapse rate set the temperature, taking convection away should cool the surface. It does the opposite. Manabe and Strickler computed it in 1964, with the water vapour, CO₂ and ozone of Earth's air:

Radiation only, clear sky332.3 K
Radiation + convection at 6.5 K/km, clear sky300.3 K
Radiation + convection, average clouds286.9 K
Measured288 K
No infrared absorbers, any lapse rate255.1 K

Convection doesn't heat the surface: it cools it, carrying heat up to where the air can radiate it away. The lapse rate is a cap, not a stove. And without infrared absorbers there is nothing to cap: 255 K. Venus shows the same at scale: 1603 K with radiation alone, 735 K with convection, in The Venus Test.

S. Manabe and R. F. Strickler (1964), Thermal equilibrium of the atmosphere with a convective adjustment, J. Atmos. Sci. 21, 361–385: clear sky, Fig. 4 and summary; average cloudiness, Fig. 8a. The last row is the "Pressure only" run above. Bars start at 200 K.

Dividing by 4 is not a flat Earth

The Sun lights Earth's disc, πR², and Earth radiates from its whole sphere, 4πR²: the ratio is 4. The 4 is the sphere.

Build the sphere from rings

Cut the globe into rings around the Sun–Earth line, each one a thin slice of a cone, and add them up. The light they catch together is always the light on their shadow. Their surface grows as the rings get thinner. Watch the ratio.

Sunlight caught by all the rings
Surface of all the rings
Surface ÷ shadow
Sunlight per m² of surface, on average

Ring k runs between the points of the circle at θk = kπ/N and θk+1 from the point under the Sun; its surface is π(rk + rk+1) × s, with r = R sin θ and s = 2R sin(π/2N). The light it catches is the light on its shadow, π(rk+1² − rk²) × 960.9 W/m² (for a ring across the day–night line, only its lit part), so the lit rings together always catch 960.9 W/m² × πR². All the rings together: 4πR² × cos(π/2N), which goes to 4πR². In the unrolled strip each ring's width is its surface and its height the sunlight per m² on it: the shaded area is the light caught, the total length is the surface. Archimedes, On the Sphere and Cylinder, third century BC: the surface of a sphere equals that of the cylinder wrapped around it, 2πR × 2R = 4πR².

So 1361 × (1 − 0.294) / 4 = 240 W/m² is the sunlight absorbed, averaged over the sphere. It is bookkeeping on a round Earth, not a claim that Earth is flat or evenly lit. Here is a planet with no infrared absorbers, lit on one side. Each spot radiates what it gets: its own sunlight, plus a share spread evenly by winds and oceans.

Average temperature
Warmest spot
Coldest spot
Infrared sent to space, average

No heat stored: each spot radiates what it receives, (1 − s) × 960.9 × cos θ on the day side, θ being the Sun's angle from overhead, plus s × 240.2 W/m² everywhere, s being the share spread evenly; 960.9 W/m² is 1361 × (1 − 0.294). The horizontal axis is area, so the average temperature is the average height of the curve, and the average of σT⁴ is always 240.2 W/m². On an uneven planet the average of T is below the fourth root of the average of T⁴, so below 255 K. With nothing spread, the night side is at 0 K: a body that stores heat, like the Moon, is less extreme, and still averages below its black-body temperature. Even reflecting nothing, the even planet would be at 278 K.

Try to beat 255 K

A planet with no infrared absorbers, cut into six zones of equal area. Set the temperatures of five; the sixth adjusts so that on average the planet still sends 240 W/m² to space, the sunlight it absorbs. Any pattern counts: day and night, rotation, oceans, winds, heat stored or not. Make the average temperature beat 255 K.

Zone 6, set by the budget
Infrared sent to space, average
Average temperature
The ceiling

Each zone sends ε σT⁴ to space; the six together must average 240.2 W/m², what the planet absorbs. The ceiling is the temperature of the even planet, (240.2 / εσ)1/4: 255.1 K, or 258.4 K with emissivity 0.95. The average of T can never exceed the fourth root of the average of T⁴ (the power-mean inequality), and equals it only when every zone has the same temperature. So no pattern of temperatures, however it comes about, lifts a planet with no infrared absorbers above the ceiling. The real Earth averages 288 K.

This game was suggested, without meaning to, by @KaiserSzecg7i on X. Thanks!

The Moon test: same distance from the Sun, no air

The Moon gets the same sunlight as Earth, reflects less of it and has no atmosphere.

MoonEarth
Sunlight arriving1,361 W/m²1,361 W/m²
Bond albedo (reflected)0.110.294
Sunlight absorbed, planet averageS (1 − A) / 4303 W/m² 26% more than Earth240 W/m²
Black-body temperature270.4 K its average can't be higher254.0 K NASA, older albedo; 255.1 K with 0.294
Average surface temperaturebelow 270 K288 K
Day–night temperature range95–390 K at the equator283–293 K

The Moon absorbs more sunlight than Earth and is colder on average, swinging by almost 300 K between day and night. Earth is warmer and swings by 10 K. Same Sun, same distance: the difference is an atmosphere that absorbs infrared.

NASA NSSDC Planetary Fact Sheets, Moon and Earth. "Its average can't be higher": for a body radiating what it absorbs, the mean of T is at most the fourth root of the mean of T⁴.

The Mars test: 95% CO₂, and cold

Mars's air is 95% CO₂, and Mars is cold. Proof that CO₂ doesn't warm? Look at how much CO₂ there is, at what pressure, and how much sunlight.

MarsEarthVenus
Sunlight absorbed, planet averageS (1 − A) / 4110 W/m²240 W/m²150 W/m²
Black-body temperature210 K255 K227 K
Greenhouse effectabout 5 K33 Kabout 510 K
Surface pressure0.006 bar1 bar92 bar
CO₂ share of the air95%0.04%96.5%
CO₂ above each m²p / g × CO₂ share by massabout 160 kg25 times Earth's6.6 kgabout 1,000 t
Water vapour210 ppmabout 1%, variable20 ppm

Mars has 25 times more CO₂ above each square metre than Earth, and a greenhouse effect about a sixth of Earth's. Three reasons:

The share of CO₂ in the air says little. The amount, the pressure it sits at and the water around it decide. Same CO₂ physics on three planets: about 5 K, 33 K, about 510 K.

NASA NSSDC Fact Sheets. Mars: 586.2 W/m², Bond albedo 0.250, 6.36 mbar, g = 3.73 m/s², CO₂ 95.1% by volume, mean molecular weight 43.49, water 210 ppm. Venus: 2601.3 W/m², Bond albedo 0.77, 92 bar, g = 8.87 m/s², CO₂ 96.5%, 43.45, water 20 ppm, average 737 K. Mars's greenhouse effect, about 5 K: Haberle 2015; Catling & Kasting 2017. Mars's plain average surface temperature, about 202 K, is below its effective temperature, about 208 K, because its surface is so uneven; the fair comparison uses the fourth root of the average of T⁴ (Haberle 2013). See "Dividing by 4".

Two centuries of the same physics

  1. 1824Joseph Fourier: sunlight gets through the air more easily than the ground's "dark heat" gets back out, so the air keeps Earth warmer.
  2. 1856Eunice Foote: in sunlight, a glass cylinder of CO₂ heats more than one of ordinary air, and stays hot longer. An atmosphere of that gas would give to our earth a high temperature.
  3. 1859John Tyndall measures in the laboratory that water vapour and CO₂ absorb and emit infrared, while oxygen and nitrogen hardly do.
  4. 1879–1884Josef Stefan and Ludwig Boltzmann: a body radiates σT⁴.
  5. 1896Svante Arrhenius: the first calculation of how much warmer Earth gets with more CO₂.
  6. 1900Max Planck: the spectrum of thermal radiation, the curves on this page.
  7. 1964Syukuro Manabe and Robert Strickler: radiation and convection together. Convection cools the surface; see "Radiation alone would be hotter".
  8. 1967Manabe and Richard Wetherald: doubling CO₂ warms the surface by about 2.4 K. Manabe shared the 2021 Nobel Prize in Physics for modelling Earth's climate.
  9. 1989Raval and Ramanathan: the greenhouse effect measured from satellites, as the infrared trapped by the air.
  10. 2001Harries et al.: Earth's infrared spectrum seen from space in 1970 and in 1997. The difference shows the greenhouse effect growing, gas by gas.
  11. 2015Feldman et al.: CO₂'s infrared reaching the ground, measured growing by 0.2 W/m² per decade.
  12. 2021Loeb et al.: Earth's energy imbalance roughly doubled from 2005 to 2019.

Objections

"It's the pressure: 1 bar of air keeps the surface warm."

Same 1 bar, same lapse rate, air that doesn't absorb infrared: 255 K. Press "Pressure only". Pressure decides how the air is arranged above the surface; it can't make the surface send more to space than Earth absorbs. A pressure that doesn't change does no work: see "If compression paid the bill".

"PV = nRT: the pressure sets the temperature."

The ideal gas law links pressure, density and temperature at each point; it doesn't say which one causes the others. At sea-level pressure, 1013 hPa, air at 288 K has a density of 1.225 kg/m³; at 255 K, 1.384 kg/m³. Both obey the law. Which temperature the surface gets is set by the energy ledger: what it absorbs and what it can send to space. The law then gives the density.

"−18 °C sits at the average height of the air, about 5 km: the lapse rate does the rest."

On Earth the middle of the mass (5.5 km) and the emission level (5.1 km) happen to be close, so the rule almost works. On Venus they are 10.4 and 71 km, and the same rule misses by 434 K. See "The 5 km coincidence".

"CO₂ is only 0.04% of the air."

0.04% of a lot of air. Above every square metre of Earth there are about 10 tonnes of air and 6.6 kg of CO₂. What matters for absorbing infrared is how much absorber there is, and at what pressure, not its share of the air: see "The Mars test". In the published attribution CO₂ is about 20% of the greenhouse effect.

"Water vapour does it all, not CO₂."

Water vapour and clouds do most of the absorbing, about 75%. But water condenses: its amount follows the temperature. In the GISS model, removing CO₂ and the other non-condensing gases makes the water rain out too, and the global mean falls to −21 °C in 50 years. See "Who does what".

"Cold air can't warm a warmer surface."

It doesn't heat it: heat, the net flow, always goes from the warm surface up. What the air does is send back 333 W/m² of infrared, so the surface loses less. The surface balances what it emits with what it absorbs, from the Sun and from the air. Take the air's share away and the net loss jumps; the surface cools to 255 K.

"Back radiation doesn't exist: it has never been measured."

It is measured every minute. Pyrgeometers at surface stations around the world record it day and night, and the global average is estimated at 333 to 342 W/m². Spectrometers looking up see it with the shape of its sources, the CO₂ band at 15 µm and the water vapour bands. From 2000 to 2010, two of them, in Oklahoma and in Alaska, measured the CO₂ part of it growing by 0.2 W/m² per decade, while CO₂ rose by 22 ppm (Feldman et al. 2015).

"The greenhouse effect is just a theory: nobody has ever seen it at work."

Everyone has. A cloudy night stays warmer than a clear one: the clouds absorb the ground's infrared and send part of it back. Measured, they add on average 28 W/m² to the infrared coming down, from 314 W/m² under clear skies to 342 (Wild et al. 2019). Dry deserts show the other side: little water vapour overhead, little infrared back, and the ground cools fast after sunset. Same Sun by day, different nights: what changes is the absorber overhead.

"Wood's 1909 experiment proved there is no greenhouse effect."

Wood warmed two boxes in the Sun, one under glass and one under rock salt, which lets infrared through, and found them almost equally warm: a greenhouse stays warm mainly because its lid stops the warm air from rising away. That is right, and it's why "greenhouse effect" is a poor name. But Earth has no lid. As a whole it can lose heat only by radiating to space, and it does so from about 5 km up, where the air is colder: that is the effect measured at the top, the 156 W/m² that never reach space. Wood himself wrote that he did "not pretend to have gone very deeply into the matter".

"It's the Sun."

The Sun supplies the 240 W/m² that Earth absorbs, and Earth sends the same back. The Moon, at the same distance, absorbs 26% more sunlight and is colder. See "The Moon test". And the stratosphere tells the two apart: a brighter Sun would warm it, CO₂ cools it, and it has cooled. See "CO₂ or the Sun?".

"You tuned the second run to give 288 K."

Yes: its height and slope are read from the measured profile, so it matches Earth by construction. It is there to show the shape. The proof doesn't use it: it uses the surface emission, the energy that leaves Earth and the sunlight the surface absorbs, all measured.

"The CO₂ band is saturated: adding more changes nothing."

The centre of the band is: there, more CO₂ changes little near the ground. But Earth radiates the band to space from aloft, and more CO₂ lifts that level to colder air: less leaves, the column warms. The forcing grows with the logarithm of the concentration and doesn't stop: about 3.7 W/m² for every doubling. See "Isn't CO₂ saturated?".

"CO₂ cools the planet: it radiates heat to space."

It does radiate to space, and that is where the warming comes from. Earth's CO₂ band leaves from high, cold air, so it sends out less than the warm surface would. Add CO₂ and the band leaves from higher still, colder: less goes out, and the air below warms (see "Isn't CO₂ saturated?"). In the middle and upper stratosphere, about 25 to 50 km, more CO₂ does cool: there it radiates more than it absorbs. Satellites measure that cooling, a human fingerprint that natural causes can't explain (Santer et al. 2023). A brighter Sun would warm the stratosphere, not cool it. See "CO₂ or the Sun?".

"Dividing the sunlight by 4 means a flat Earth with no night."

4 is the ratio between the disc that catches the sunlight, πR², and the sphere that radiates, 4πR². It averages energy; it doesn't say the temperature is even. And the even planet is the warmest case: day and night make the average colder, never warmer. Move the slider in "Dividing by 4".

"Mars is 95% CO₂ and freezing."

Mars has about 25 times more CO₂ per square metre than Earth, at 160 times lower pressure, with almost no water vapour, and it absorbs less than half the sunlight. Its greenhouse effect is about 5 K; Earth's is 33 K, Venus's about 510 K. The share of CO₂ in the air says little. See "The Mars test".

Found an error?

Every number on this page comes from the equations and data listed below, and the code is public. If one is wrong, show which one, the value you get instead and how you get it. The report opens on GitHub, already filled in with a link to the view you are looking at.

See the reports