Explainer: The CMIP7 emissions scenarios – and how they explore future climate change

    </div>

                    </div>

In short, the socioeconomic world underlying the new scenarios is somewhat more crowded and less wealthy per person than the one the SSPs originally imagined.

Another notable change is the shift in the SSP that underlies the highest future emissions scenario. 

In the original SSPs, the “very high” SSP5-8.5 scenario was based on SSP5, while the new “high” scenario in CMIP7 is based on SSP3. 

The GMD study explains that this is because IAM teams that developed the scenarios found that SSP3 and SSP5 variants produced similar emissions. They judged that the “fragmented” SSP3 world – which is characterised by large challenges to adaptation – to be more relevant for exploring high-end risks.

No more ‘baseline’ scenarios and other changes

In another important change, the authors of the CMIP7 scenarios decided to eliminate “baseline” scenarios that assumed a world without any climate policy. These scenarios were previously used as a counterfactual against which to compare climate-changed worlds.

Instead, the range of future emissions scenarios starts with current policies and explores ways that they could be strengthened, weakened, or kept the same. The high scenario explores a plausible “rollback of current mitigation policies“. 

The medium scenario, by contrast, extends climate policies officially implemented as of 2025, without assuming countries achieve their Paris Agreement pledges – known as nationally determined contributions (NDCs) – or net-zero targets that are not yet backed by legislation. 

In their GMD paper, the authors of the CMIP7 scenarios emphasise that the medium scenario “should not be considered as a ‘most likely’ scenario”, but that it can provide a benchmark against which the effect of future policy strengthening or weakening can be measured. It is roughly analogous in its emissions levels to the old SSP2-4.5 scenario.

The new low scenario explores a world where climate policy is rapidly strengthened and warming by 2100 is limited to below 2C. This makes it analogous to the old SSP1-2.6 scenario. 

The very-low scenario limits global warming to around 1.5C by 2100, similar to the old SSP1-1.9 scenario. However, it involves a greater degree of overshoot mid-century, reflecting the fact that global emissions did not begin to rapidly decline in 2020 as envisioned by SSP1-1.9. As the authors of the GMD ScenarioMIP paper point out: “At this point of time, some overshoot of the 1.5C seems unavoidable.”

In addition, there are a number of scenarios that start on one path before undertaking rapid mitigation. These high-to-low, medium-to-low and low-to-negative scenarios are intended to explore futures where mitigation is further delayed, followed by a rapid turn-around later in the century.

The scenario developers noted that there is no specific likelihood or probabilities assigned to any scenario, but rather only a judgement that all are within the realm of plausibility given where the world is today. They also said that “there might be potential futures outside the ScenarioMIP scenario range”.

Timescales and other changes

In addition to the shift away from baseline scenarios, there are three other notable design changes made in CMIP7. 

First, CMIP7 models will be driven by emissions of CO2 and other greenhouse gases, rather than set atmospheric concentrations. 

In every previous generation of models, the ScenarioMIP experiments required that modelling groups simulate future climate using the same set of CO2 concentrations. For CMIP7, models with an interactive carbon cycle are asked to run in “emissions-driven” mode for CO2, calculating atmospheric concentrations themselves based on emissions. 

This is a significant improvement. It means that the substantial uncertainty in carbon-cycle feedbacks will now show up directly in the range of projected warming, rather than being overlooked. (The change applies to CO2 only; methane, nitrous oxide and halocarbons remain prescribed as concentrations.) 

Second, emissions match observations up to 2023. IAM modellers were asked to stay close to observed trends up to 2025 to avoid emissions diverging from reality before models were run. Scenario differences only open up after 2026, avoiding an earlier problem of scenarios diverging from reality years before the models were even run.

Finally, the period over which models are being run has been extended from 2100 to 2150. This is important as the world is already more than a quarter of the way through the 21st century. 

The extended model runs out to 2150 will provide a more thorough exploration of the warming that people born in the coming decades may experience within their lifetimes. 

In addition, all scenarios have extensions to 2500 where temperatures are eventually stabilised. These allow scientists to explore changes to long-term Earth-system processes, such as ice sheets and sea level, as well as whether warming is reversible. 

A narrower range of future CO2 emissions

Overall, the new scenarios provide a notably more narrow range of future CO2 emissions than the SSP scenarios used in CMIP6. 

The figure below shows net global CO2 emissions (combining fossil-fuel and land-use emissions) for the seven new scenarios, alongside the five SSP scenarios used for climate model runs in CMIP6 (e.g. SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5).

At the bottom of the range, the new scenarios closely track their predecessors: the very-low scenario reaches net-zero CO2 around mid-century much like SSP1-1.9, while the low scenario lands close to SSP1-2.6. 

The chart below shows total emissions for the same scenarios for the period 2024-2100.

The lowest emissions scenarios now have somewhat higher total emissions, reflecting the failure of the world to rapidly reduce emissions after 2020 that occurred in the lower SSP emissions scenarios, such as SSP1-1.9 and SSP1-2.6. The very-low scenario results in 310bn tonnes of CO2 (GtCO2) cumulative emissions between 2024 and 2100, compared to around 110GtCO2 in SSP1-1.9.

At the top end, the change is particularly dramatic. The high scenario in CMIP7 reaches 55GtCO2 per year in 2100. The previous high scenario, SSP5-8.5, by contrast, reached around 126GtCO2 per year in 2100. 

In cumulative terms – which is what matters most for global warming – high reaches around 3,820GtCO2 over 2024-2100, half the roughly 7,600GtCO2 of SSP5-8.5 and about three-quarters of the 5,140GtCO2 of SSP3-7.0.

To put it another way: the top of the new scenario range sits between SSP2-4.5 and SSP3-7.0 in cumulative emissions terms, which is territory that CMIP6 treated as its middle ground.

To make the scale of this shift clear, Carbon Brief analysed the CO2 emissions trajectories in each of the prior generations of high-end emissions scenarios, using the same simple climate model – FaIR – to calculate future warming.

Below, four different generations of emissions scenarios are examined. The SRES scenarios were originally published in 2000 and used in the IPCC’s third (2001) and fourth (2007) assessment reports (and the corresponding CMIP3 model runs). The RCPs were developed in the early 2010s and used in the IPCC fifth assessment report (AR5; 2013) and CMIP5, while the SSPs were developed in the late 2010s and used in the IPCC AR6 report and CMIP6.

Over the past two decades, the highest emissions scenarios all resulted in comparable amounts of warming in 2100: SRES A1FI (the highest SRES scenario) reached 4.6C in 2100 (5-95% range; 3.5-6.1C), RCP8.5 reached 4.9C (3.7-6.5C) and SSP5-8.5 reached 4.6C (3.5-6.2C). 

(RCP8.5 edges out its successor despite lower CO2 emissions because it assumed considerably more methane and nitrous oxide.) 

CMIP7’s high scenario comes in remarkably lower, at 3.3C (2.6-4.4C).

The downward revision of future emissions in CMIP7 reflects two key changes since RCP8.5 was published back in 2011. First, the plausible baseline of a repeal of current policy has fallen. Cheap solar, wind and batteries, a global plateau in coal use and more than $2tr per year in clean-energy investment mean that a rollback in climate policy would not result in coal deployment levels assumed in the RCP8.5 scenario.

The GMD study states that CMIP6’s high-emission levels “have become implausible, based on trends in the costs of renewables, the emergence of climate policy and recent emission trends”. 

(For more, see Carbon Brief’s recent factcheck of false claims around the retirement of the SSP5-8.5 emissions scenario. Also see Carbon Brief’s recent interview with Prof Detlef van Vuuren, a key architect of both the old SSPs and new scenarios.)

Second, part of the apparent decline reflects a correction of how scenarios are communicated – rather than real-world progress. The old high-end scenarios always represented an estimate of worst-case scenarios at the time, rather than likely outcomes. 

Genuine progress in reducing emissions probably accounts for around 0.7C of the roughly 1.7C gap between SSP5-8.5 and today’s current-policy trajectory, with the remainder reflecting that the baseline was never particularly likely.

What the new scenarios mean for future warming

To compare warming outcomes across scenario generations on a like-for-like basis, Carbon Brief ran both the seven CMIP7 scenarios and the CMIP6 SSP emission scenarios through the same simple climate model. 

(This is FaIR v2.2, using the 841-member ensemble calibrated and constrained to match the assessment of climate sensitivity in IPCC AR6, historical warming and ocean heat uptake). 

These values may differ from the ultimate results that are found by CMIP7 climate models, but give a sneak peak of what those results may look like when they become available.

The seven scenarios produce warming in 2100, relative to pre-industrial (1850-1900), that ranges from 1.6C (with a 5-95% range of 1.1-2.5C) in the very-low scenario to 3.3C (2.6-4.4C) in high, with the current-policy medium scenario reaching 2.9C (2.2-3.9C). Warming also continues after 2100 in both the medium and high scenarios.

The figure below shows the range of 2100 warming (5th to 95th percentile) relative to the preindustrial period expected in each of the old SSP scenarios and the new CMIP7 ones, along with a central estimate (white dots).

The largest changes are, unsurprisingly, at the top. CMIP7’s high scenario (3.3C in 2100) produces less warming than SSP3-7.0 (3.7C in the same ensemble) and far less than SSP5-8.5 (4.7C). 

The entire CMIP6 “high” tier (e.g. SSP5-8.5 and SSP3-7.0) now sits above anything in the new scenario set, at least up to 2100. Extended beyond 2100, however, high keeps climbing towards levels the previous extreme scenarios reached earlier.

At the low end, the picture is more similar. The very-low scenario (1.6C in 2100) lands close to SSP1-1.9 (1.5C) and low (1.8C) is essentially indistinguishable from SSP1-2.6 (1.8C) in 2100. 

However, the new low scenario involves more rapid late-century emissions reductions and greater amounts of carbon removal than its SSP analogue, while the very-low scenario involves greater overshoot of 1.5C mid-century.

Crossing warming thresholds

In addition to calculating 2100 and 2150 warming, Carbon Brief has calculated the likelihood of passing different global warming levels (2C, 2.5C, 3C, 4C and 5C) over time in the new CMIP7 scenarios. 

The chart below uses the IPCC approach of calculating the crossing year based on a 20-year average, rather than when a single year exceeds the warming level.

Under the medium scenario, which reflects a world where current policies are maintained, passing 1.5C is essentially locked in. 

Most models cross the threshold by the late 2020s or early 2030s. The 2C limit is crossed around 2050 on average and 3C by around 2110. The chance of exceeding 4C is around one-in-four by 2150, but, ultimately, rises to roughly 50% if emissions continue after that point.

Under the high scenario, 2C arrives in the 2040s, 3C in the 2080s and the chance of exceeding 4C by 2150 is around 60% (and around 95% by 2300). Even 5C is reached by 2150 in roughly 20% of climate model simulations.

The lower scenarios tell a different story. In the very-low scenario, the chance that peak warming (which the IPCC determines using a 20-year average of warming) ever exceeds 1.5C is around 90%. This reflects the fact that passing 1.5C is almost unavoidable at this point. 

However, the chance of surpassing 2C sits at around 30% and the scenario has warming falling after mid-century as more CO2 is removed from the atmosphere than is added. 

Carbon dioxide removal

Every scenario that has global warming peaking and declining requires pulling CO2 back out of the atmosphere. Otherwise, warming from CO2 emissions will persist for millennia. 

CO2 removal (CDR) remains one of the few levers available to reduce future temperatures – particularly given additional warming caused by cuts to aerosol pollution. 

The chart below shows the total CDR deployment in each of the different scenarios by year, reflecting the sum of both land-based and engineered approaches (top), as well as the total CDR deployment between 2024 and 2150 (bottom).

Every scenario that deeply cuts global emissions in CMIP7 also involves a large amount of CDR. 

The low-to-negative scenario pulls a cumulative 2,360GtCO2 out of the atmosphere by 2150, roughly 60 years of today’s emissions run in reverse. 

The high-to-low scenario has around 1,480GtCO2 cumulative CDR, medium-low has 1,450GtCO2 and low has 1,360GtCO2. 

Even the very-low scenario, which seeks to minimise CDR use, requires 655GtCO2 of removals between 2024 and 2150.

The degree to which scenarios rely on “engineered” removals – such as the use of biochar or direct air capture – or land-based removals – including afforestation and reforestation – ranges  across models. 

In the low scenario, roughly one-third of the removals is from the land “sink”, while low-to-negative relies almost entirely on engineered methods, with direct air capture alone reaching around 16GtCO2 per year by 2100.

The chart below shows the deployment of engineered removals by year (top), as well as the total engineered CDR used between 2024 and 2150 (bottom). The lower plot also includes a breakdown between the portion of CDR that requires geologic storage (e.g. DAC and BECCS) and the portion that does not (e.g. enhanced weathering and biochar) and compares the total to a recent “prudent” total CO2 storage limit published in the scientific literature.

(For more on limits to carbon storage capacity, see Carbon Brief’s 2025 guest post.)

The amount of CDR going toward geological storage is most highest in the low-to-negative scenario, which injects around 1,750GtCO2 of BECCS and direct-air-capture CO2 underground by 2150. 

The high-to-low and low scenarios each commit around 800GtCO2 to storage by 2150. This is within the range of available geologic storage, but would require that the storage industry handles more CO2 than the mass of oil currently moved by the fossil-fuel industry. 

That said, there are other potential CDR approaches – such as enhanced rock weathering, surficial mineralisation and ocean alkalinity enhancement – that do not require injection of CO2 into geologic formations. In-situ mineralisation approaches that inject CO2 into alkaline rock formations such as basalt or peridotite could also open up more potential CO2 storage.

It is worth noting that the amount of CDR deployed in these scenarios would require planetary-scale engineering at the cost of trillions of dollars, while many of the engineered CDR approaches are still relatively early-stage technologies. 

No single climate future

The goal of scenarios is to span a range of possible futures. While it may be tempting to treat current climate and energy policies – and the medium scenario – as a forecast, there is no reason to expect that they will not change in the future. 

It is likely that policies will continue to be strengthened, as has been the case over the past two decades. However, they may also be weakened if national priorities or politics change, as has happened in the US during the two terms of the Trump administration.

In the new CMIP7 scenarios there is no “business-as-usual” scenario, but rather a narrower range of futures than was available in CMIP6, reflecting greater clarity among scientists on where the world is heading in terms of future energy use and emissions. 

The fact that the worst-case scenarios of the past have become increasingly implausible is good news. However, this is tempered by the fact that the very-low emission scenarios have, in turn, become harder to achieve given that global emissions have yet to decline.

There is also real uncertainty in the climate-system response to emissions. This is due to uncertainty around how sensitive the climate is to a build-up of CO2 in the atmosphere, as well as how the carbon cycle will respond to emissions. 

The CMIP7 medium scenario – which has a central estimate of 2.9C of warming by 2100 – still has around a 3% chance of reaching 4C by that date. If emissions continue, those odds increase to 25% by 2150. This remains far outside anything resembling a safe outcome for the climate system.

The scenarios are now being run using the new CMIP7 models, whose emissions-driven runs will fold carbon-cycle uncertainty directly into projections. These projections will subsequently be analysed in the reports of AR7. 

Ultimately, it will be decisions made by governments, businesses and individuals that decide which of these seven futures become closest to reality.

Methodology

Emissions scenarios shown in this article are the seven CMIP7 ScenarioMIP scenarios set out in van Vuuren et al. (2026), harmonised to observed 2023 emissions, with rule-based extensions to 2500 generated using the FLEX methodology. Emissions through 2100 match the ScenarioMIP database; extension trajectories are indicative and may differ from the final published extensions.

Temperature projections use FaIR v2.2 with the fair-calibrate v1.4.5 constrained ensemble (841 members set out in Smith et al. (2024), which matches the AR6 assessed climate sensitivity (ensemble ECS median 3C, 5-95% 2.0-5.1C), historical warming and ocean heat content. 

Historical emissions (1750-2022) use the FaIR historical emissions dataset, with scenario emissions spliced in after 2023. 

Solar and volcanic forcing are updated through 2025 from the Climate Indicator forcing timeseries; future volcanic forcing ramps to the 1850-2021 climatological background by 2035 (following the CMIP7 protocol) and solar forcing follows a SOLARIS-HEPPA-derived cycle projection to 2300. 

All warming is expressed relative to 1850-1900.

SSP comparisons run the RCMIP-harmonised CMIP6 scenario emissions through the FaIR ensemble, which yields 2081-2100 warming 0.1-0.3C below the AR6-assessed values at the high end (e.g. SSP5-8.5: 4.2C vs 4.4C assessed), reflecting differences between the AR6 assessment and the FaIR configurations used here. Updating the volcanic dataset to use CMIP7 values (which revises the eruption-rich 1850-1900 baseline period) raises all reported anomalies by 0.03-0.05C.

For CDR, the scenario database reports the technology split (for example, BECCS, direct air capture, enhanced weathering, biochar, ocean-based, soil carbon management). Agriculture, forestry and other land-use  (AFOLU) removals are available only as a net flux, so are shown as the net sink where negative. Soil-carbon management is grouped with land-based rather than engineered removal, and the geological storage comparison uses BECCS plus direct air capture only. 

The figure showing high-end scenarios for the past four CMIP generations runs SRES A1FI through the same ensemble using the A1G MiniCAM model from the SRES database v1.1, spliced onto historical emissions at 2000, and covering CO2 (fossil and land use), methane, nitrous oxide and sulphur; SRES-era ozone-precursor projections (nitrous oxide, carbon monoxide and volatile organic compounds) lie outside the calibrated range of FaIR, so RCP8.5 values are used instead. RCP8.5 uses RCMIP v5.1 emissions, with 13 minor halogenated gases absent from the RCP database following SSP5-8.5.

<section class="cards wp-block animate cards--columns-4 alignfull">
    <div class="cards__inner">
                        <div class="cards__header">
                <h2 class="cards__heading">Related</h2>
                                </div>

        <div class="cards__items">
                                <article class="g-card g-card--size-medium g-card--size-mobile-small g-card--row has-image has-link">
<div class="g-card__inner">

                <div class="g-card__image">
            <img width="768" height="512" src="https://www.carbonbrief.org/wp-content/uploads/2026/05/3DHFT0T-1-768x512.jpg" class="attachment-medium_large size-medium_large" alt="" role="presentation" loading="lazy">            </div>

    <div class="g-card__content">

                        <h2 class="g-card__heading">Guest post: How CMIP7 will shape the next wave of climate science</h2>            
        <div class="g-card__meta">

                                <div class="g-card__date">
                    22.05.2026                    </div>

                                <div class="g-card__labels">
                        <a href="https://www.carbonbrief.org/topics/climate-modelling" class="g-card__label">Climate modelling</a>                    </div>

        </div>

    </div>

</div>

        <a href="https://www.carbonbrief.org/guest-post-how-cmip7-will-shape-the-next-wave-of-climate-science" class="g-card__link" aria-label="Read more about Guest post: How CMIP7 will shape the next wave of climate science"></a>

                <div class="g-card__image">
            <img width="768" height="512" src="https://www.carbonbrief.org/wp-content/uploads/2026/04/3A65T9F-768x512.jpg" class="attachment-medium_large size-medium_large" alt="" role="presentation" loading="lazy">            </div>

    <div class="g-card__content">

                        <h2 class="g-card__heading">Traditional models still ‘outperform AI’ for extreme weather forecasts</h2>            
        <div class="g-card__meta">

                                <div class="g-card__date">
                    29.04.2026                    </div>

                                <div class="g-card__labels">
                        <a href="https://www.carbonbrief.org/topics/climate-modelling" class="g-card__label">Climate modelling</a>                    </div>

        </div>

    </div>

</div>

        <a href="https://www.carbonbrief.org/traditional-models-still-outperform-ai-for-extreme-weather-forecasts" class="g-card__link" aria-label="Read more about Traditional models still ‘outperform AI’ for extreme weather forecasts"></a>

                <div class="g-card__image">
            <img width="768" height="512" src="https://www.carbonbrief.org/wp-content/uploads/2026/03/2Y49DH3-768x512.jpg" class="attachment-medium_large size-medium_large" alt="" role="presentation" loading="lazy">            </div>

    <div class="g-card__content">

                        <h2 class="g-card__heading">Limiting global warming to 2C would not ‘rule out’ extreme impacts</h2>            
        <div class="g-card__meta">

                                <div class="g-card__date">
                    25.03.2026                    </div>

                                <div class="g-card__labels">
                        <a href="https://www.carbonbrief.org/topics/climate-modelling" class="g-card__label">Climate modelling</a>                    </div>

        </div>

    </div>

</div>

        <a href="https://www.carbonbrief.org/limiting-global-warming-to-2c-would-not-rule-out-extreme-impacts" class="g-card__link" aria-label="Read more about Limiting global warming to 2C would not ‘rule out’ extreme impacts"></a>

                <div class="g-card__image">
            <img width="768" height="512" src="https://www.carbonbrief.org/wp-content/uploads/2026/01/3AFHX9N-768x512.jpg" class="attachment-medium_large size-medium_large" alt="" role="presentation" loading="lazy">            </div>

    <div class="g-card__content">

                        <h2 class="g-card__heading">State of the climate: 2025 in top-three hottest years on record as ocean heat surges</h2>            
        <div class="g-card__meta">

                                <div class="g-card__date">
                    14.01.2026                    </div>

                                <div class="g-card__labels">
                        <a href="https://www.carbonbrief.org/topics/climate-modelling" class="g-card__label">Climate modelling</a>                    </div>

        </div>

    </div>

</div>

        <a href="https://www.carbonbrief.org/state-of-the-climate-2025-in-top-three-hottest-years-on-record-as-ocean-heat-surges-9" class="g-card__link" aria-label="Read more about State of the climate: 2025 in top-three hottest years on record as ocean heat surges"></a>