What this is. A JavaScript port of
METEOR 's generation
kernel, driven by a compact bundle exported from a model trained offline on
CMIP6 output. The whole payload is about 300 KB and there is no server: the
emulator runs on your machine.
What the range shows. Each climate model gives a
somewhat different answer, and the chart shows how much they agree:
the line is the middle model, the darker band the middle half of the
models, and the lighter band the middle 90%. These are the models'
long-term responses to each scenario, without the ups and downs of
individual years, measured from each model's own 1850–1900 level.
Precipitation change is a percentage of that level. The
Expert view shows year-to-year variability, one model at a
time, with maps, other baselines and more.
Places. The 58 AR6 reference regions, and 252
cities: every national capital of more than half a million people,
and every other city of more than 2.5 million. A city is the
nearest gridbox of each model that is more than half land, one to
three degrees across, so it describes the city's surroundings rather
than its streets.
Two generations of scenario. The CMIP6 SSPs and the
CMIP7 ScenarioMIP markers were built years apart and harmonized to
different vintages of history, so they are not interchangeable. Both
are driven here through the same simple climate model, CICERO-SCM,
which is what makes comparing them meaningful at all: taking CMIP7's
forcing from its own MAGICC assessment instead would have put a
0.74 W/m² model difference inside what looks like a scenario
difference.
Baselines. Temperature, and the maps, are shown as
change from a reference period: 1850–1900, the pre-industrial
convention warming levels are defined against, or 2005–2024, recent
history. Either is the model's own forced response averaged over the
period under CMIP7 Medium, and every scenario shares it, so the
difference between two scenarios does not depend on the choice.
Measuring from recent history removes the part of the spread between
models that each inherited from its own historical warming — the
models here warmed by between 0.7 and 1.6 °C from 1850–1900 to
2005–2024 — so switching baselines is a quick way to see how much of
their disagreement about the future is disagreement about the past.
The precipitation timeseries stays in absolute mm/day.
Degree days. Heating and cooling degree days, base
18 °C, count how far daily temperatures fall below or rise above the
point where buildings need heating or cooling. They start from the
observed climate of 1995–2014 (W5E5, daily) and add each model's
warming for each month, so they do not inherit a model's biases in
absolute temperature or in how much days vary. For a region they
average its land gridboxes' degree days; for a city they use the
nearest land gridbox of the observations, half a degree across. They
are not offered for the sea or the global mean.
Crop yields. Wheat, maize, rice and soybean, through
the GGCMI Phase 2 emulators (Franke et al. 2020): for each of up to
nine crop models, a polynomial giving its long-term rainfed yield at
every half-degree cell from CO₂, the change in temperature and in
precipitation since 1980–2010, and nitrogen. Each climate model's
change since 1980–2010 drives them, not its absolute climate, and
CO₂ comes from the same simple climate model as the warming. A
region's yield is weighted by where the crop is grown (MIRCA2000
rainfed harvested area), and only the globe and regions that grow
enough of it get one. The yields include CO₂ fertilization, no
adaptation of crop varieties, and the same nitrogen everywhere; the
emulators were trained on warming of up to 6 °C and CO₂ of up to
810 ppm, and inputs beyond that are held at the limit. They give a
climate's yield, not a harvest's, so there are no realizations.
Crop models. Pick one crop model, or all of them,
whose spread then takes the place of an ensemble's. Nitrogen can be
set to GGCMI's three levels (10, 60 and 200 kg/ha) or the default
100; CARAIB, JULES and PROMET ran one level, so it does not change
them. The multi-model mean of crop yields is precomputed, at the
default nitrogen, and its map is the simple view's.
What it is not. The band around any one model is that
model's internal variability — not model uncertainty, and not the
full uncertainty in a projection. Compare models to see how much they
disagree; with the 2005–2024 baseline, that is disagreement about the
future rather than about the past. The difference between two models'
maps is taken on the first model's grid, the second interpolated onto
it. Maps show the forced response only;
variability at a region you draw yourself would need a further 11 MB of
EOF maps that this page does not load. Custom emissions need a
simple climate model, which is not here either, so the scenarios on offer
are the ones the bundle was built with. Precipitation covers 2015–2100
only, the window its distribution transform was fitted for.
Sources. CMIP7 scenario emissions: ScenarioMIP-CMIP7
IAM quantification, harmonized and infilled marker scenarios, v0.2,
10.5281/zenodo.19825038
(Kikstra et al., forthcoming, “Harmonized and gridded emissions
dataset for ScenarioMIP-CMIP7”; Riahi et al., forthcoming, “Overview
of the new SSPs for ScenarioMIP”), used under the
ScenarioMIP
Explorer licence . This page shows figures and forcing derived from
them; the emissions themselves are available from the
ScenarioMIP-CMIP7
Explorer . Emulators trained on CMIP6 model output from the WCRP
Coupled Model Intercomparison Project. AR6 reference regions from
Iturbide et al. (2020),
ESSD 12, 2959–2970 ;
coastlines and cities from Natural Earth. Degree days are corrected with W5E5
v2.0 (Lange et al. 2021,
10.48364/ISIMIP.342217 ;
Cucchi et al. 2020,
ESSD 12, 2097–2120 ),
CC BY 4.0.
DIAMOND and FUTURA are funded by the European Union under Horizon Europe
grant agreements 101081179 and 101295156. Views and opinions expressed
are however those of the author(s) only and do not necessarily reflect
those of the European Union or the granting authority. Neither the
European Union nor the granting authority can be held responsible for
them.