Climate Change: Simple, Serious, Solvable › Models and Data
Where the Carbon Goes: Air, Ocean, Land and Warming to the Year 3000
Every year we burn fossil fuel, and the carbon goes into the air. The ocean and the land take some of it back out. The rest warms the world. Set how the ocean, the land and the climate respond, see whether the model matches the past, then follow the carbon a thousand years ahead.
Loading the data…
The past: carbon dioxide and warming since 1850, beside the measurements
- This model
- Measured
Move over the graph, or select it and use the arrow keys, to read values.
| Billions of tonnes of carbon | This model | Carbon budget |
|---|---|---|
| Burned | ||
| Stayed in the air | ||
| Went into the ocean | ||
| Went into the land, after clearing |
Only the burning is given. The model is told how much fossil fuel was burned each year, and nothing about the carbon dioxide in the air. It works that out: what the ocean and the land don't take up stays in the air. Then it works out the warming. So the thin gray lines are a test.
Make it match. The model as it opens runs high in carbon dioxide from the 1950s on. Try the land's settings: more CO₂ fertilization, more room for wood, less clearing. Several different settings match the past about equally well. Which is right matters for the future, as the next views show.
The budget doesn't add up either. The Global Carbon Budget's own counts of clearing and of the land's growth miss what the air and the ocean say by about 1.6 billion tonnes a year in the 1950s. This model, which follows the same records of clearing, runs high in the same years.
The future: six scenarios, with the settings you chose
Move over the graph, or select it and use the arrow keys, to read values.
The scenarios are the IPCC's. Each sets the world's fossil emissions year by year to 2100, and how they end after that, by 2250 at the latest. Pick the years up to 3000 to see what happens after the burning stops.
Warming stops when emissions stop, but it doesn't go away. Carbon dioxide falls slowly once the burning ends, as the ocean takes more up. The warming hardly falls at all, because the ocean that was holding it back is warming too.
Where the carbon goes: the air, the ocean and the land
- Burned so far
- Still in the air
- In the ocean
- In the land
- In the air, without the removal
Move over the graph, or select it and use the arrow keys, to read values.
Much of it stays in the air while emissions are rising: more than half of the fossil carbon burned so far, because clearing forests has added carbon too. The ocean takes up most of the rest. When emissions fall, the share in the air falls, but slowly: the ocean takes centuries to mix the carbon down, and some of it stays in the air for thousands of years.
What does "comes back" mean? Three scenarios take carbon back out of the air late in the century: SSP1-1.9, SSP1-2.6 and SSP5-3.4-overshoot. For those, a thin dashed line shows the air as it would be without the removal, and the gap between the two air lines is what stays out. As CO₂ in the air falls, the ocean and the land take up less than they would have, or give some back. That takes centuries, as the deep ocean and the forests catch up, so the share that comes back keeps growing long after the removal stops.
The land can give carbon back. When emissions stop, the land's yearly uptake drops below zero in most scenarios. Look at the lower graph after 2100.
The land sink: what it's made of, and when each part stops
- The land, all together
- More CO₂
- Nitrogen from burning
- Warming
- Clearing and logging
Move over the graph, or select it and use the arrow keys, to read values.
Each part has a reason to stop. More CO₂ makes plants grow faster, but the extra stops once CO₂ stops rising and the forests catch up. Nitrogen from burning stops when the burning does. Warming helps the cold north at first and then hurts. Forests that grow back after clearing stop gaining once they're grown.
The parts needn't add up to the whole. Each is worked out by running the land again with that one cause held as it was in 1750. The causes work on each other: the extra growth from CO₂ needs nitrogen, for one.
Settings that match the past can disagree about the future. Go back to the first view and find two quite different settings that both match. Then come back here and compare them.
About this model
This is a model of the carbon cycle and the climate together, for the whole Earth, one year at a time. Fossil fuel puts carbon into the air. The ocean and the land take some of it back out: these are the carbon sinks. The carbon that stays in the air warms the world, and the warming changes how much the sinks take up. The model follows all of that from 1750 to the year 3000.
It has four parts. The air is one box of carbon. The ocean is four boxes of water: a warm surface, the cooler water under it, a cold surface near the poles and the deep ocean. Water carries carbon, heat and nutrients between them. The climate is an energy balance: the push from greenhouse gases and smog, against the heat the Earth sends to space and the heat the ocean takes in. The land is three regions, the cold north, the temperate middle and the tropics. Each has wood, dead leaves, soil and old stable soil, and each is cleared, farmed, logged and left to grow back as the records say.
It goes with Chapter 9 of the book. It began as Scott Denning's EarthCarbon, a Shiny app his students have used for years, and was rebuilt in November 2026.
How to use this page
"1. The past" runs the model from 1750 to 2024 and draws its carbon dioxide and warming beside the measurements. The table under it says where the carbon burned since 1850 went. The settings are in groups. The climate: how much the world warms for doubled carbon dioxide, and how much smog cools it. The ocean: how fast the warm surface mixes with the water below. The land: how much faster plants grow with more carbon dioxide, how much wood a forest can hold, how short of nitrogen plants are, how fast dead matter rots when it's warmer, how warming changes growth, and how much clearing and logging there was. The record: whether the world burned less than it reported.
"2. The future" carries on to 2100, 2300 or 3000 in six scenarios. "3. Where the carbon goes" follows the chosen scenario's carbon into the air, the ocean and the land. In the three scenarios that take carbon back out of the air (SSP1-1.9, SSP1-2.6 and SSP5-3.4-overshoot), a thin dashed line shows the air as it would be without the removal, and the lines under the graph say how much of what was taken out stays out. "4. The land sink" splits the land's uptake into its causes. The group called The future picks the scenario and the last year shown.
Move over a graph, or select it and use the arrow keys, to read the numbers for one year. The line of text under it gives them. The address in your browser keeps your view and settings, so you can share them or come back to them.
The model goes with Chapter 9, "The One-Way Thermostat", of Climate Change: Simple, Serious, Solvable.
The equations
The model answers one question each year: how much carbon is in the air, and how warm is it? You choose the settings in the groups and the scenario. Everything else comes from the records and the scenarios, and from fixed numbers that are listed in the code with where each came from.
Step 1. Where does this year's carbon go?
C is the carbon in the air, in billions of tonnes (GtC). E is the fossil carbon burned this year, O what the ocean takes up and L what the land takes up after clearing, all in GtC a year. Whatever the sinks don't take stays in the air. The carbon dioxide in the air in parts per million, c, is the carbon divided by 2.18 GtC per ppm:
Step 2. How hard does the air push on the climate?
F is the climate forcing, in W/m². Doubling carbon dioxide from its 1750 level, c0, gives F2× = 3.93 W/m², and each doubling after that the same again: that's why the logarithm. a is the smog setting, Fs smog's forcing and Fo everything else: methane, nitrous oxide, ozone, the Sun and volcanoes.
Step 3. How warm does it get?
S is the climate sensitivity you set, and λ (lambda) the heat the Earth sends to space for each degree it warms, in W/m² per °C. T is the warming. The Earth gains F − λT, and 91% of that goes into the ocean. H is the heat the ocean takes up this year, in W/m² of the whole Earth, and it depends on T, because the ocean's surface warms with the world. So the model finds the T that makes the two sides equal.
The land warms faster than the sea, because the sea spends much of its energy evaporating water. The model's land warms 1.6 times as much as the warm sea. The cold sea near the poles is under ice at first and warms only as the ice goes.
Step 4. How much carbon does the ocean take?
For each surface box, the carbon going in each year is a rate, k, times the difference between the carbon dioxide in the air, c, and the carbon dioxide the water holds up against it, p. Both are in ppm. k grows with the box's area, how fast gas crosses the surface and how well CO₂ dissolves in cold water.
What sets p is chemistry. Most of the carbon in seawater isn't dissolved gas: it's bicarbonate and carbonate. As carbon is added, more of it stays as gas, so p rises faster than the carbon does. The model works this out with the constants every oceanographer uses. Water moves between the four boxes, sinking at the poles and rising again, and it carries the carbon, heat and nitrate with it. Plankton take carbon down as they sink. Warming makes the layers harder to mix, so less carbon goes down.
Step 5. How much does the land grow?
f is how much faster plants grow at carbon dioxide c than at c0. Below cc = 40 ppm plants can't grow at all. The curve flattens as carbon dioxide rises; h sets how soon, and the model works it out from the fertilization you set. Then growth is held back by nitrogen when plants run short, and changed by warming along a bell curve that peaks at 25 °C. The north grows faster as it warms; the tropics are near the top of the curve already.
Step 6. How fast does it rot?
Each pool of wood, dead leaves, soil and stable soil gains I a year from the pool before it and loses a share r of itself. τ (tau) is the pool's turnover time before any warming: about 40 years for wood, a few years for dead leaves, decades for soil and thousands of years for stable soil. Warming speeds rot by Q10 for each 10 °C. P0 is the pool last year and P this year. The loss is worked out from this year's pool, which keeps the model steady however short τ is.
Step 7. What is the land sink made of?
The model runs the land again four times, each with one cause held as it was in 1750: more carbon dioxide, warming, nitrogen from burning, and clearing and logging. The difference from the full run is that cause's share.
A worked example: from 2024 to 2025
With the page's opening settings, in the scenario SSP2-4.5:
- In 2024 the world burned 10.31 GtC of fossil carbon.
- The ocean took up 3.38 GtC and the land 1.31 GtC.
- So the air gained 10.31 − 3.38 − 1.31 = 5.62 GtC.
- Divided by 2.18 GtC per ppm, that's 2.58 ppm more carbon dioxide.
- Carbon dioxide went from 441.26 ppm at the start of 2024 to 441.26 + 2.58 = 443.84 ppm at the start of 2025. "2. The future" shows it for 2025, rounded to 443.8 ppm.
Some details. Carbon is in billions of tonnes of carbon, not of carbon dioxide: a tonne of carbon makes 3.67 tonnes of CO₂. Warming is counted from 1850 to 1900. The model gives back a tiny share of the extra carbon in the air each year to the rocks, so slowly that it barely shows by 3000. It leaves out drought, fire, permafrost and any future change in farming.
See the equations at work
- On "1. The past", set the land's CO₂ fertilization to 0. Carbon dioxide in 2024 rises by about 31 ppm. Step 5: without it, the land's growth doesn't keep up with clearing.
- On "2. The future", choose SSP5-8.5 and the years up to 3000. Carbon dioxide in 3000 is still about 1,337 ppm, five hundred years after the burning stopped. Step 4: the ocean is slow.
- On "4. The land sink", choose SSP1-2.6 and the years up to 2300. The land's uptake falls below zero by 2100. Step 7: each cause has stopped, and the nitrogen sink has turned around.
Where the data come from
Fossil emissions, and the ocean and land sinks the table compares with: the Global Carbon Budget 2025. Friedlingstein, P., and others (2026). Global Carbon Budget 2025. Earth System Science Data 18, 3211–3288. doi:10.5194/essd-18-3211-2026. CC BY 4.0.
Measured carbon dioxide and the climate forcing of the past: Forster, P. M., and others (2025). Indicators of Global Climate Change 2024. Earth System Science Data 17, 2641–2680. doi:10.5194/essd-17-2641-2025. CC BY 4.0.
Measured warming: HadCRUT.5.1.0.0. Morice, C. P., and others (2021). Journal of Geophysical Research: Atmospheres 126, e2019JD032361. doi:10.1029/2019JD032361. Contains public sector information licensed under the Open Government Licence v3.0.
Clearing, farming and logging: LUH1. Hurtt, G. C., and others (2011). Harmonization of land-use scenarios for the period 1500–2100. Climatic Change 109, 117–161. doi:10.1007/s10584-011-0153-2. Data: Chini, Hurtt and Frolking (2014), ORNL DAAC, doi:10.3334/ORNLDAAC/1248. After 2004, HYDE 3.3: Klein Goldewijk (2024), doi:10.24416/UU01-94FNH0, CC BY 4.0.
The scenarios: RCMIP. Nicholls, Z., and others (2020). Reduced Complexity Model Intercomparison Project Phase 1. Geoscientific Model Development 13, 5175–5190. doi:10.5194/gmd-13-5175-2020. Their extensions after 2100: Meinshausen, M., and others (2020), Geoscientific Model Development 13, 3571–3605, doi:10.5194/gmd-13-3571-2020. The page's file of scenarios is derived from RCMIP's data and shares its licence, CC BY-SA 4.0.
How it was checked
The page's code is a line-by-line copy of a model written first in Python. The two give the same carbon dioxide and warming, to one part in a million, in eight runs to the year 3000 at very different settings. The Python model's ocean takes up 198 GtC over 1850 to 2024, the Global Carbon Budget's 200. Its seawater chemistry agrees with PyCO2SYS, the standard program, to within 1.5%. Its land matches the measured split of the land's uptake between the north and the tropics, and the wood that satellites saw it gain since 2000. Its warming matches the measured record to within about 0.1 °C a year on average since 1960.
The code
The model is in model.js: the steps above, in order. app.js is the page: the controls and the graphs.
model.js, the science
app.js, the page
Download model.js · Download app.js · All of the model's code on GitHub
Credits
By Scott Denning, for Climate Change: Simple, Serious, Solvable. The measurements are the work of the Global Carbon Project, the authors of the Indicators of Global Climate Change, the Met Office Hadley Centre and the Climatic Research Unit at the University of East Anglia, and the land-use teams of LUH and HYDE, who made this possible by sharing them.
The model is Scott Denning's, and it began as a web app written in R. Claude, from Anthropic, rebuilt it with him, first in Python and then to run in your browser. The rebuild gave the model its four-box ocean, its energy balance in place of a fixed delay, and its land of three regions, in which each sink has its own reason to stop.