Climate Change: Simple, Serious, Solvable › Models and Data
A Million Years of Ice Ages: Sunlight, Ice and Carbon Dioxide
Earth's orbit sets the rhythm of the ice ages, but on its own it can't make them as big as they were. The ice and the carbon dioxide do the heavy lifting. See how a small model of all three matches the last 800,000 years, then add fossil carbon and see what happens to the next ice age.
Move over the graph, or select it and use the arrow keys, to read values.
Move over the graph, or select it and use the arrow keys, to read values.
The orbit sets the pace. The summer Sunlight at 65 °N swings between about 430 and 550 W/m². That alone can't make the ice ages as big as they were. The ice and the carbon dioxide feed back on each other and on the temperature, and turn it into ice sheets that come and go about every 100,000 years. Turn off the carbon dioxide's part, or the faster melting of old ice, and see what's left.
This is a small model. It has one equation each for the ice, the carbon dioxide and the temperature, and its numbers were chosen to match the last 800,000 years. The About tab has every equation.
About this model
This is a small model of the ice ages, by Stefanie Talento and Andrey Ganopolski of the Potsdam Institute for Climate Impact Research. It follows three things through time, 1,000 years at a step: the ice on land, the carbon dioxide in the air and the world's average temperature. The only thing that pushes it from outside is the summer Sunlight in the far north, which changes with Earth's orbit. For the future, you can add a pulse of fossil carbon.
Chapter 5 of the book, "Past Climate Change", says that the orbit sets the timing of the ice ages, and that the ice-albedo and carbon-cycle feedbacks make them big. This model puts numbers on that story. Its summer Sunlight changes by only about a tenth. But in the model, the ice sheets grow for tens of thousands of years and then melt in about 10,000, roughly every 100,000 years, while the carbon dioxide swings between 150 and 288 ppm and the world cools by as much as 6.4 °C. Its ice rises and falls with the record of sea level, with a correlation of 0.86.
Then it looks ahead. With no fossil carbon, ice starts to grow again in about 55,000 years. With 1,000 GtC of fossil carbon, it waits about 172,000 years.
How to use this page
The page has two views of the same model. "The last 800,000 years" puts the model beside records of the past: sea level, which shows how much ice there was on land; the carbon dioxide in bubbles of air trapped in Antarctic ice; and the world's average temperature, worked out from the temperature of the sea surface. "The next million years" starts now. Each view has four graphs: the summer Sunlight at 65 °N that drives the model, the ice on land, the carbon dioxide and the temperature.
You can change four things:
- Whether the carbon dioxide follows the temperature and the ice. Turn it off, and the carbon dioxide stays at 278 ppm (plus any fossil carbon).
- Whether old ice sheets melt faster. Turn it off, and ice melts at the same pace however long it has been there.
- Earth's orbit. Keep both parts of the summer Sunlight, as they were, or hold one of them at its average.
- The fossil carbon put into the air now, from 0 to 3,000 GtC, on "The next million years". The page opens at 1,000 GtC, and the gray lines always have none.
Move over a graph to read the numbers for one time. Or select the graph and press the left or right arrow key. The first press starts 21,000 years ago, at the height of the last ice age (or 100,000 years from now), and each press after it moves 1,000 years, or 10,000 with Shift. Home and End go to the ends. Below the graphs, a few lines sum up the view. "Back to where the page opened" puts every control back where it started.
The address in your browser changes as you use the controls, so you can bookmark a view or send it to someone. Nothing is sent anywhere as you work.
This page is part of the website for Climate Change: Simple, Serious, Solvable, an open textbook by Scott Denning. It goes with Chapter 5, "Past Climate Change".
The equations
You don't need any of this to use the page. But the whole model is five short steps, and you can follow each of them with a calculator.
The model asks one question, over and over: how much will the ice, the carbon dioxide and the temperature change in the next 1,000 years? It starts 999,000 years ago with a little ice, and asks again every 1,000 years until a million years from now.
The letters, and what they stand for:
| What | Its value |
|---|---|
| v, the ice on land. 0 is as much as today, and 1 is the most there was in the last ice age, enough to lower the seas about 120 m (Chapter 5) | worked out |
| C, the carbon dioxide in the air | worked out, in ppm |
| T, the world's average temperature compared with preindustrial | worked out, in °C |
| p and o, the two parts of the summer Sunlight at 65 °N | from Earth's orbit, in W/m² |
| E, the fossil carbon put into the air now | you choose, in GtC |
| At the start, 999,000 years ago | v = 0.05, C = 300 ppm |
Every number in the equations is from Talento and Ganopolski's best version of the model, rounded to four figures. They chose their numbers so that the model would match the last 800,000 years. Each step uses the ice, the carbon dioxide and the Sunlight at the end of the step before.
1. How strong is the summer Sunlight?
F is how far the summer Sunlight at 65 °N is above its average, in W/m². Summer Sunlight there decides whether last winter's snow survives (Chapter 5). F comes in two parts, each worked out from Earth's orbit with the equations of the Milankovitch model. p is the strongest Sunlight of the summer, in a day's average, with the tilt held at its average of 23.32°: it changes only with precession and the stretch of the orbit. o is the Sunlight on the June solstice if the orbit were a perfect circle: it changes only with the tilt. 493.10 and 491.65 W/m² are their averages over the two million years the model runs.
The tilt part counts 4% extra (the 1.04): Talento and Ganopolski weighted the two parts that way because it best matched how fast the real ice grew and melted. Choose "Tilt only", and the model leaves out the first part; choose "Precession only", and it leaves out the second.
2. How warm is the world?
T is in °C, compared with preindustrial. The first term is the cooling from the ice, the ice-albedo feedback of Chapter 4: bright ice sheets reflect Sunlight back to space. Ice as big as in the last ice age (v = 1) cools the world by 3 °C. The second term is the greenhouse effect of carbon dioxide, which grows with the logarithm of C (Chapter 4). At the height of the last ice age, with v = 1 and C = 194 ppm, T = −3 − 2.00 = −5.0 °C: the 5 °C of Chapter 5.
3. How much carbon dioxide is in the air?
C is in ppm, and it's never allowed below 150 ppm. m is how much v shrank in the 1,000 years before (0 if it grew), and Cf is the fossil carbon dioxide still in the air (step 5). Each term is a way the ocean, the land and the ice change the carbon dioxide:
- Warmer, less carbon in the ocean. Warm water holds less carbon dioxide than cold water, the way a soft drink goes flat on the kitchen counter (Chapter 5), and the ocean's mixing changes too. Each °C of warming adds 17.28 ppm.
- Bigger ice, less carbon in the air. Ice sheets change the carbon in several ways. A saltier ocean and fewer forests let carbon dioxide out, while iron-rich dust feeds the ocean's plankton, which take it up. In the model, more ice means less carbon dioxide overall.
- A puff as the ice melts. When ice melts fast, its fresh water floods the North Atlantic and can shut down the ocean's overturning there. In many models, that raises the carbon dioxide by 10 to 20 ppm.
- 278 ppm is the carbon dioxide with no extra ice and no warming: the world before industry.
4. How fast does the ice grow or melt?
Δv is how much v changes in the next 1,000 years, and the new v is v + Δv. v is never allowed below 0, today's ice. G adds up four things that grow or melt the ice:
- Snow on the ice. Snow falls on an ice sheet's whole area, so a bigger ice sheet gathers more of it. The 0.2196 v is that snow, less the melting that grows in step with the ice.
- Melting at the edges. A bigger ice sheet reaches farther south, into warmer summers, so its melting grows faster than the ice does: the v3/2. This keeps the ice from growing forever.
- Summer Sunlight melts snow and ice: the 0.0008 F.
- Carbon dioxide warms the summers too: the ln C, about 5.63 at 278 ppm.
Then comes the old ice. While the ice is growing, M is 0. While it's shrinking, M is about the average of v over the last 30,000 years. Dividing by 1 − 0.1841 M makes old, big ice sheets melt faster once they start: at the end of the last ice age, with M about 0.87, it made the budget G 19% bigger. Talento and Ganopolski added this term so that the model's ice would melt as fast as the real ice did at the end of each ice age.
The last number, 0.5260, nearly balances the carbon dioxide's term. With no ice and 278 ppm, G = −0.09525 × 5.628 − 0.0008 F = −0.5360 − 0.0008 F, so Δv = −0.0100 − 0.0008 F. Ice can start to grow only when that's more than 0: when the summer Sunlight falls more than 12.5 W/m² below its average. That's the threshold for a new ice age. More carbon dioxide moves it: at 300 ppm, the Sunlight has to fall 21.6 W/m² below its average, and at 350 ppm, 40.0.
5. How much fossil carbon dioxide is still in the air?
Cf is in ppm. Each GtC of carbon in the air adds 0.469 ppm of carbon dioxide (Chapter 9 works out about 0.47). A is the share of the fossil carbon still in the air t years after it goes in, from an Earth system model by Natalie Lord and colleagues. It adds up five parts, each fading at its own pace, from a few years to a quarter of a million. Together they follow the long tail of Chapter 9, as the ocean, the dissolving of old seashells and, slowest of all, the weathering of rocks take the carbon back out of the air. Each part's share a and its time τ (tau), in years, change a little with the size of the pulse. For 1,000 GtC they are:
| Part | Share, a | Time, τ |
|---|---|---|
| 1 | 0.2101 | 5.3 years |
| 2 | 0.2119 | 55.8 years |
| 3 | 0.3563 | 225.9 years |
| 4 | 0.1181 | 3,602 years |
| 5 | 0.0742 | 246,989 years |
So 10,000 years after 1,000 GtC goes in, A is 0.0786, and Cf = 0.469 × 1,000 × 0.0786 = 36.9 ppm. The last part lasts hundreds of thousands of years.
Where the equations stop
- It can't melt today's ice sheets. v never goes below 0, so the model leaves Greenland and Antarctica as they are, and it says nothing about the sea-level rise ahead (Chapter 8).
- Steps of 1,000 years. The model puts all the fossil carbon into the air in one step, before the ocean has taken up any, and its next steps overshoot. So the graphs leave out the carbon dioxide and the temperature for the first 3,000 years after it goes in. Chapter 9 follows those first centuries.
- A short tail. In the Earth system model Lord and colleagues used, about 8% of a 1,000 GtC pulse is still in the air after 10,000 years. Chapter 9 says other models leave 10% to 40%. With more left in the air, the next ice age would wait longer.
- Small pulses. Lord and colleagues built their response from pulses of 1,000 to 20,000 GtC. Smaller pulses, like the 750 GtC we've emitted so far, use it a little below the range it was made for.
- A climate sensitivity of 3.85 °C. Doubling C warms the model's world 5.556 × ln 2 = 3.85 °C, the average of the climate models of the time. The IPCC's best estimate is 3 °C (Chapter 4), so the model's temperature changes about 28% more for each doubling. Talento and Ganopolski fitted the model's other numbers with the larger one, so it can't simply be swapped here.
- Too little carbon dioxide in the deepest ice ages. The model's carbon dioxide falls to its floor of 150 ppm, while the ice cores never go below about 174 ppm. Whenever the model sits on its floor, the ice cores show 30 ppm or more at the same time, so on the graph its blue line drops below the gray one in the deepest ice ages.
- Warm spells no warmer than today's. The model's temperature comes only from its ice and its carbon dioxide (step 2), so over the last 800,000 years it's never more than 0.2 °C warmer than preindustrial. The temperature record is warmer than that in two of the warm spells between ice ages: 1.0 °C warmer 127,000 years ago, in the last one before ours, and 1.0 °C warmer 404,000 years ago. In the last ice age, the model is at its coldest 6.4 °C below preindustrial, and the record 6.0. Over the 800,000 years, the two go together with a correlation of 0.68.
- Before about 410,000 years ago, more ice probably stayed on land between the ice ages than there is today. The model can't work that out from the orbit, so it keeps v at 0.05 or more until then.
- Thresholds. The ice ages hang on thresholds, so a small change can move the end of an ice age from one peak of summer Sunlight to the next. Rounded to three figures instead of four, the model's numbers end some of its ice ages, past and future, at different times.
A worked example
On "The last 800,000 years", select the graph and press the right arrow key eight times, to 14,000 years ago, near the end of the last ice age. Here's where the page's numbers come from. Fifteen thousand years ago, at the end of the step before, v = 0.61727, and 1,000 years before that it was 0.70335. C = 181.611 ppm, p = 502.91 W/m² and o = 501.24 W/m².
- The summer Sunlight (step 1). F = (502.91 − 493.10) + 1.04 × (501.24 − 491.65) = 9.81 + 1.04 × 9.59 = 19.7836 W/m², well above average.
- The temperature (step 2). ln(181.611/278) = −0.42575. T = −3 × 0.61727 + 5.556 × −0.42575 = −1.8518 − 2.3655 = −4.2173 °C.
- The carbon dioxide (step 3). The ice shrank by m = 0.70335 − 0.61727 = 0.08608. C = 17.28 × −4.2173 − 31.95 × 0.61727 + 120 × 0.08608 + 278 = −72.9 − 19.7 + 10.3 + 278 = 195.7 ppm.
- The ice's budget (step 4). ln 181.611 = 5.20187, and 0.617273/2 = 0.48497. G = 0.2196 × 0.61727 − 0.2874 × 0.48497 − 0.0008 × 19.7836 − 0.09525 × 5.20187 = 0.13555 − 0.13938 − 0.01583 − 0.49548 = −0.5151.
- The old ice (step 4). The ice is shrinking, so M is the average of v over the 30,000 years before: 0.8684. 1 − 0.1841 × 0.8684 = 0.8401. Δv = −0.5151 ÷ 0.8401 + 0.5260 = −0.613 + 0.5260 = −0.087.
- The new ice. v = 0.61727 − 0.087 = 0.530.
The page says that 14,000 years ago, the ice in the model was 0.53, the carbon dioxide 196 ppm and the temperature 4.2 °C colder than preindustrial. The ice is melting fast, and as it shrinks, the carbon dioxide and the temperature climb.
A few details
- The order of a step. The temperature comes first, from the ice and the carbon dioxide at the end of the step before. Then the carbon dioxide, from that temperature and the ice. Then the ice, from the ice, the summer Sunlight and the carbon dioxide at the end of the step before. That's the order in Talento and Ganopolski's own code.
- The old ice, exactly. M adds up v over the 30 steps before, counting the first and the last of them half, and divides by 30. That's how a trapezoid adds up an area.
- The first 30,000 years. The model holds v, C and T at their starting values for its first 30 steps, then runs for 169,000 years before the graphs begin, long enough to forget where it started.
- The Sunlight. The parts p and o use the book's solar constant, 1,361 W/m². Added up as p + o − 491.65, they're never more than 1.8 W/m² from the strongest summer Sunlight itself.
- Ice and sea level. The sea-level record is turned into v by dividing it by its lowest point, −130 m, 24,000 years ago, as Talento and Ganopolski did. The page's sea level for the model's ice is v × 120 m, to the nearest 5 m, or the nearest meter below 10 m. The record's last few thousand years sit a few meters above today's sea level, which its authors put down to mud stirred by burrowing animals, so there its ice dips a little below 0.
See the equations at work
- Hold the carbon dioxide still. On "The last 800,000 years", turn off "Carbon dioxide follows the temperature and the ice". The ice never grows bigger than 0.41, and the 100,000-year ice ages are gone. What's left rises and falls with the 41,000-year tilt and the 23,000-year precession (steps 3 and 4).
- Keep the old ice. Turn on the carbon dioxide again, and turn off "Old ice sheets melt faster". The ice grows to 1.07 and never shrinks below 0.71 after 700,000 years ago: the ice age never ends (step 4).
- One part of the orbit. Turn the old ice back on, and choose "Tilt only", then "Precession only". The model still makes long ice ages, but they end at the wrong times. Its ice and the record go together with a correlation of only −0.07 with the tilt and 0.27 with precession. The ice and the carbon dioxide set the length of an ice age, and it takes both parts of the orbit to set the timing (step 1).
- The next ice age. On "The next million years", set the fossil carbon to 0. Ice starts to grow again 55,000 years from now: Chapter 5 says the orbit would keep this warm spell going for about another 50,000 years. With 1,000 GtC, it waits until 172,000 years from now, and with 3,000 GtC, until 508,000 (step 5).
Where the data come from
The model, its equations and its numbers are Talento and Ganopolski's. The page's code is new, and it was checked against their own code and runs, which they published with the paper:
Talento, S., and Ganopolski, A. (2021). Reduced-complexity model for the impact of anthropogenic CO₂ emissions on future glacial cycles. Earth System Dynamics, 12, 1275–1293. doi:10.5194/esd-12-1275-2021
The fossil carbon dioxide still in the air (step 5) follows:
Lord, N. S., Ridgwell, A., Thorne, M. C., and Lunt, D. J. (2016). An impulse response function for the “long tail” of excess atmospheric CO₂ in an Earth system model. Global Biogeochemical Cycles, 30, 2–17. doi:10.1002/2014GB005074
The summer Sunlight is worked out from Earth's orbit in the table that the Milankovitch model uses:
Laskar, J., Robutel, P., Joutel, F., Gastineau, M., Correia, A. C. M., and Levrard, B. (2004). A long-term numerical solution for the insolation quantities of the Earth. Astronomy & Astrophysics, 428, 261–285. doi:10.1051/0004-6361:20041335
The records drawn beside the model are sea level from up to seven reconstructions put together; the carbon dioxide in Antarctic ice cores, as in Chapter 5's Figure 5.11; and the world's average surface temperature from Clark and colleagues, who put together records of the sea surface's temperature kept in ocean sediments and scaled them to the whole surface with climate models. Its zero is the last few thousand years, which they say may be taken as preindustrial. The page shows each record every 1,000 years, and uses the temperature with the authors' permission.
Spratt, R. M., and Lisiecki, L. E. (2016). A Late Pleistocene sea level stack. Climate of the Past, 12, 1079–1092. doi:10.5194/cp-12-1079-2016
Bereiter, B., Eggleston, S., Schmitt, J., Nehrbass-Ahles, C., Stocker, T. F., Fischer, H., Kipfstuhl, S., and Chappellaz, J. (2015). Revision of the EPICA Dome C CO₂ record from 800 to 600 kyr before present. Geophysical Research Letters, 42, 542–549. doi:10.1002/2014GL061957
Clark, P. U., Shakun, J. D., Rosenthal, Y., Köhler, P., and Bartlein, P. J. (2024). Global and regional temperature change over the past 4.5 million years. Science, 383, 884–890. doi:10.1126/science.adi1908
How it was checked
A set of tests comes with the model, and they run every time the page changes. Here's what they hold it to.
- The authors' own runs. Given their Sunlight and their numbers in full, a second copy of the model, in another programming language, gives the same ice, carbon dioxide and temperature as all 29 versions of the model they kept, with and without fossil carbon, to within a hundred-millionth.
- The code, against that copy. With the Sunlight worked out here and the numbers to four figures, the page's code and the second copy agree to within a billionth, at every setting of the controls the tests try.
- The paper's results. With the Sunlight worked out here, the model's ice matches the record as closely as in the paper (a correlation of 0.86), and its next ice age starts at the same time.
- The fossil carbon's numbers. The 40 numbers behind step 5 are the ones in the table Lord and colleagues published with their paper. Their own code has the same numbers, to more figures.
- The Sunlight, against the Milankovitch model. The two parts of the summer Sunlight are worked out again with the Milankovitch model's own code.
- This tab, against the model. Every equation above is tried with the model's own numbers. The worked example is done again from the rounded numbers a calculator would use. And each number in the text is worked out again from the model.
The code
The code is small enough to read. Two files do all the work, and they use no outside software. The comments in model.js use the same step numbers and letters as the equations above.
model.js is the science: the five steps, and the loop that runs them.
app.js is the page: the controls, the graphs and the lines under them.
To keep a copy, save model.js, app.js or the data, data.json.
All of the model's code is on GitHub, at github.com/airscottdenning/climate-models. You're free to use it, change it and teach with it.
Credits
By Scott Denning, for Climate Change: Simple, Serious, Solvable.
The model is Stefanie Talento and Andrey Ganopolski's (2021), and the fossil carbon dioxide follows Natalie Lord and colleagues (2016). Claude, from Anthropic, wrote it to run in your browser and checked it against their published code and runs.