Climate Change: Simple, Serious, Solvable › Models and Data
Many Models, Many Futures
Thirty-five of the world's climate models ran this century four times, once for each of four scenarios of what people might do. Their answers differ. Pick a place, see every model's answer, and find out how much of the difference comes from our choices, how much from the models, and how much is just weather.
Loading the data…
| Scenario | Models | Their average | The middle 90% | Lowest; highest |
|---|
Each scenario is a what-if. It says what the climate might do if people follow that path. It isn't a forecast, and nobody has given any of the four a probability. The numbers in a name like SSP5-8.5 aren't degrees of warming either.
The spread of the models isn't the IPCC's range. The IPCC weighed the models against measurements and other evidence, and its assessed range for the world's temperature is narrower and a little cooler than the models' own. Choose the whole world, temperature and one of the IPCC's three periods to see the two side by side.
- Scenarios: how far apart their averages are
- Models: how far apart they are in one scenario
- Weather: how far a single year strays
All three are measured the same way: the average distance from an average. For the scenarios, it's how far each ticked scenario's average sits from the middle of them. For the models, it's how far each model sits from the average of its scenario. For the weather, it's how far a single year sits from its own model's 21-year average.
About this page
Climate modeling centers around the world run their Earth system models through the same experiments, so the answers can be compared. For the IPCC's Sixth Assessment Report, they ran the past from 1850 to 2014 and then four scenarios of the future to 2100. This page shows the answers of 35 of those models, for the whole world and for each of the IPCC's 58 regions, with no more than a year's average taken from the source. Chapter 7 of the book, "Climate Scenarios", explains where the scenarios come from and how the models work, and Chapter 8, "Future Climate Change", what they project.
On the page as it opens, it's the whole world's temperature in 2081–2100, compared with 1850–1900. Under the very high emissions of SSP5-8.5, the 34 models that ran it average 4.78 °C of warming, and their middle 90% runs from 3.60 to 6.32 °C. The IPCC's assessment for the same scenario and years is 4.4 °C, very likely between 3.3 and 5.7.
How to use this page
Choose a place, whether to show temperature or rain and snow, which baseline to compare with, which scenarios to show, and the 20 years to read. Three buttons jump to the IPCC's periods: 2021–2040, 2041–2060 and 2081–2100.
The "Every model" tab draws one thin line for each model and scenario: gray for the past and in the scenario's color from 2015. The bold line in each color is the average of the models. Under "Draw", choose 21-year averages, which smooth out the weather, single years, or a shaded band holding the middle 90% of the models. "Pick out one model" draws one model in black: its single years thin, its 21-year average bold. The table gives, for the 20 years being read, how many models ran each scenario, their average change, the middle 90%, and the lowest and highest. For the whole world's temperature in one of the IPCC's periods, it adds the IPCC's assessment.
The "Why they differ" tab measures three sources of spread in the same units: the scenarios, the models, and the weather, which Chapter 7 calls internal variability.
"Back to where the page opened" puts everything back. The address in your browser changes as you work, 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 Chapters 7 and 8.
The arithmetic
The page answers one question: how much do the models say a place will change, and why don't they agree? It takes each model's yearly averages and does four things.
1. An average over some years
For each model and scenario, the page averages the 20 years being read, T for temperature or P for rain and snow, and the years of the baseline, B. If more than two years are missing, it leaves that model out.
2. A change from the baseline
Each model's change is measured from its own baseline. Temperature changes are in degrees:
Rain and snow change as a percentage of what fell in the baseline:
Using each model's own baseline matters. Models disagree about how warm a place is today by a degree or more, and they agree much better about how much it warms.
3. The average of the models
The bold lines and the table's average are the plain average of the models that ran a scenario. Every model counts the same, whether it's good or not.
4. The middle 90%
Sort the n models' changes from lowest to highest, and number them from 1. The edges of the middle 90% sit at place
where s is 0.05 for the lower edge and 0.95 for the upper. When k falls between two models, L below and U above, the edge is that part of the way, r, from one to the other:
So 5% of the models sit below the band and 5% above it.
How far apart
The "Why they differ" tab measures spread as an average distance from an average. For the scenarios, it's how far each scenario's average sits from the middle of them. For the models, it's how far each model sits from its scenario's average, averaged over the scenarios. For the weather, it's how far each model's single years stray from its own 21-year average.
A worked example
Open the page. It's the whole world's temperature, compared with 1850–1900, in 2081–2100.
- One model. Pick out CanESM5, from the Canadian Centre for Climate Modelling and Analysis. Its whole world averaged 13.56 °C in 1850–1900, and 20.53 °C in 2081–2100 under SSP5-8.5. Its change is 20.53 − 13.56 = 6.97 °C, the highest of the 34 models.
- The lower edge of the band. 34 models ran SSP5-8.5, so the lower edge is at place 1 + 0.05 × 33 = 2.65, between the second and third models from the bottom. Their changes are 3.585 and 3.607 °C, so the edge is 3.585 + 0.65 × (3.607 − 3.585) = 3.60 °C.
The table says the middle 90% runs from +3.60 to +6.32 °C, and the highest is CanESM5 at +6.97.
Where the numbers stop
- The raw models, not the IPCC's assessment. The IPCC narrowed and cooled the models' range for the world's temperature, using measured warming and other evidence: some of the models warm more than the evidence allows. The page shows the IPCC's numbers beside the models' only for the world's temperature in its three periods.
- One run for each model. Each model ran each scenario once here. So the weather in it is one sample, and some of a model's difference from the others is just its weather.
- Not every model ran everything. The number of models changes from scenario to scenario, and fewer have rain and snow.
- Averages over big regions. A region's average hides what happens inside it, on a mountain or a coast.
- No 1901–1949. The source has those years for only some of the models, so the page leaves them out.
Things to try
- Choices or models? Open "Why they differ". In 2021–2040, the scenarios are only 0.04 °C apart, but the models are 0.27 °C apart. By 2081–2100 it's the other way around: 0.95 °C for the scenarios and 0.64 for the models. The scenarios first spread further apart than the models for the 20 years from 2059.
- Land and sea. Choose SSP3-7.0 alone and compare "The world's land" with "The world's oceans" in 2081–2100: 5.15 °C of warming over land, 3.36 over the oceans.
- The Arctic. Choose the Arctic Ocean: 11.55 °C of warming under SSP3-7.0 in 2081–2100.
- Wetter and drier. Show rain and snow under SSP3-7.0 in 2081–2100. The Mediterranean gets 17.4% drier on average, and every model in the middle 90% dries it. South Asia gets 17.1% wetter, but the models range from almost no change to much wetter.
- Weather. Draw single years, and pick out one model. Its own warm and cool years wander around its 21-year average, by 0.09 °C on average for the whole world.
Where the data come from
The models' regional averages were made and published by the IPCC for the Atlas of its Sixth Assessment Report, from the CMIP6 models' monthly output: the air temperature near the surface and the rain and snow, averaged over each reference region, for land boxes, sea boxes or both. The page uses one run of each model, for the past and SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5, and averages each year's twelve months. The data are CC BY 4.0.
Iturbide, M., et al. (2022). Implementation of FAIR principles in the IPCC: the WGI AR6 Atlas repository. Scientific Data, 9, 629. doi:10.1038/s41597-022-01739-y
The IPCC's assessed warming of the whole world is from its Table 4.5 and Table SPM.1:
IPCC (2021). Summary for Policymakers. In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press. doi:10.1017/9781009157896.001
How it was checked
- The arithmetic, against a second copy. The tests have the page's arithmetic written out a second time, in another programming language, and compare them for many places, scenarios and periods.
- The data, against the source. The reduced numbers were made from the IPCC Atlas's files, each checked by its fingerprint, and the page's files are made from them by a script whose output is checked the same way.
- The IPCC's numbers, against the report. Every assessed number was read from Table 4.5 and Table SPM.1.
- This tab, against the data. Every equation above is tried with the page's own numbers, and each number in the text is worked out again from the data.
The code
Two files do all the work, and they use no outside software.
numbers.js is the arithmetic: averages, changes, the middle 90% and the spreads.
app.js is the page: the controls, the graphs and the table.
To keep a copy, save numbers.js or app.js.
Credits
By Scott Denning, for Climate Change: Simple, Serious, Solvable. Claude, from Anthropic, built the page from the IPCC's data and checked it against the book. The modeling centers that ran the models are named in the "Pick out one model" list.