We will further specify this... it is meant for all scenarios as we realize not all models can be run emission driven.

Thank you for your thoughtful comments... we will consider them carefully. Regarding the more general point about speaking to 2 communities (and I would argue even the impact community as well): you are right that this might be a complicating factor - but that is also what makes ScenarioMIP unique. So we hope to keep that aspect (and think that the more detailed comments made by you can help us to make the text clearer).

This paper discusses a proposal for the upcoming CMIP7 ScenarioMIP.  It provides a presentation of the main aspects driving the design for the main experiments.  This paper is designed to be of use by both the IAMs and ESMs communities, which tends to make the paper somewhat confusing at times, as identified in some of my comments below.  Maybe the paper would be easier to read if the discussions for the two communities were separated? Furthermore, while it is vying to provide a complete picture, I have found many instances of unclear statements and unjustified decisions.  The paper will be significantly improved by considering the following comments, listed as they appear in the paper

Thank you for your thoughtful comments, we will consider them carefully. Regarding the more general point about speaking to two communities (in fact, we would argue three, including the impact research community): this is indeed a complicating factor, but that is also what makes ScenarioMIP unique. So we would like to keep that aspect, and think that the detailed comments made by you and other reviewers can help us to make the text clearer.

The title should indicate that this is a proposal, since that is made very clear in multiple places in the article

We would like to keep the current title for consistency with the CMIP6 version of the paper (O’Neill et al., 2016) that served the same purpose. By now the proposal is also final.

Lines 111-114: this seems like a much bigger role than what this paper describes, which is a set of scenarios. Similarly, line 199 should be “participants in ScenarioMIP”

As part of the scenarios’ forcings we plan to deliver those types of datasets, so this is within the purview of ScenarioMIP. Assuming the reference is to line 119, not 199, we are now referring to climate modeling groups participating in ScenarioMIP.

Line 137: instead of “IAM scenarios”, should it be “IAM-generated data based on the scenarios in this proposal”? Otherwise “IAM scenarios” should be defined.

We have expanded the text here to indicate that we mean the IAM emissions and land use scenarios based on the experimental design described in the paper.

Lines 163-165: this is an interesting point, although it seems that the end result is still with a high RCP8.5-like (or close) scenario. So, were the critiques unfounded and/or ignored?

Not ignored, but possibly one of the major critiques, having to do with climate impacts on the economy, remains unaddressed as methodological challenges still prevent us from representing them in the type of IAM simulations that will deliver the ScenarioMIP scenarios. That said, the results of “pushing the envelope of the upper end of emissions” results in a scenario that is far from reaching 8.5W/m², and is in fact even below 7W/m² according to FaIR preliminary estimates.

 Line 195: what is the rationale for this “agnostic” position? This seems like a step back towards the RCPs

This was always the principle behind the scenario framework, i.e., the fact that different SSPs could lead to the same RCP, and different RCPs could be compatible with the socio-economic assumptions of the same SSP. In CMIP6 the choice was made to explicitly mention the driving SSP of the marker scenario in the nomenclature of the forcing pathways and a posteriori we found this to be misleading. The assumption is that the socio-economic details of the SSPs are responsible only for spatial patterns of land use and SLCF emissions whose signal, in the multimodel context, is not robust. So, although one specific choice of SSP will underlie the marker chosen for each scenario, the expectation is that a different choice that still achieves the same global level of emissions of well mixed greenhouse gases would deliver a multi-model ensemble result that would be statistically indistinguishable, as far as both global and regional outcomes are concerned.

Line 231: this should be qualified as “CO2 concentrations”.

Thank you, done.

Lines 236-238: this is similar to when, for example, aerosol emissions and cloud-aerosol interactions were included. What is the approach suggested to help with the challenge in the interpretation of the results?  Who will be helping?

We are asking the E-driven participants to run at least one of the scenarios also in C-driven mode (we recommend the Medium scenario). We hope that this will help apportion the uncertainty to the different sources. We are aware of the possibility that the results would be dependent on the forcing levels, hence the “at least”, and we specify two more scenarios in concentration mode in Tier 2, but given the computational burden imposed on modelling centers we cannot require more.

Lines 252-254: shouldn’t there be a definition of which scenario? Is the implication that the concentration-emission comparison is scenario independent?  That seems hardly justifiable.

Since up to this point we have not described the scenarios, we are not mentioning which, but we will specify it later. We are also hoping the modeling centers will run more than one scenario in both modes, but computational constraints prevent us from requesting that. We agree that the comparison may be scenario dependent. Within the paper, ScenarioMIP is suggesting scientific directions of future research to the community, and this is one of those.

Lines 260-262: this would be a step backwards for most climate models since cloud-aerosol emissions are now a critical process that needs to be represented. There are multiple implementations of reduced chemistry relevant to aerosol only.

We have edited this paragraph to clarify our request to modelers, which is to run CO₂ as emissions-driven, other well-mixed GHGs as concentration-driven, and to provide aerosol and ozone precursor emissions so that models with atmospheric chemistry can run in emissions-driven mode. At the same time, we will provide ozone concentration and optical aerosol properties for those models that need to run these species in concentration-driven mode.

Line 264: what is proposed here in terms of AerChemMIP? To use the output of ESMs run with full chemistry?  Very confusing sentence.

We agree the sentence is confusing. Since the topic of future research directions is already covered in the concluding section, we have deleted the sentence here.

Lines 268-269: observations of what? CO2?

We’re refining this point to be less specific on what we demand of modeling centers with regard to their historical CO₂ concentration outcomes.

We replaced those lines with the following text:

“Running historical simulations in an emission-driven configuration presents a challenge, as a model's simulated CO₂ concentration can drift from observations. However, this approach is a deliberate and core feature of the CMIP7 experimental design.  Rather than take measures to artificially correct this bias, we encourage modeling groups to quantify CO₂ concentration evolution as a key diagnostic of model performance and carbon cycle sensitivity. To facilitate this analysis, the ScenarioMIP protocol includes a parallel concentration-driven simulations specifically to allow for a clean diagnosis of the impact of these emergent carbon cycle feedbacks on climate outcomes.”

Line 272: why still stop at 2100?  IAMs have been used over the last couple of decades and have always had 2100 as their end point.  One would therefore conclude that IAMs could be used for 100-year projections.  Why not here?

IAM simulations depend on driving datasets (e.g., future time series of population and GDP at country-level resolution) that have not been produced yet beyond 2100. In this phase of CMIP, given the tighter than usual timeline, it has not been possible to work on that aspect and explore IAM’s performance for a longer timeline.

Line 278: emissions are not observed, only estimated.

Reworded, thanks for pointing that out.

Line 285: why that request of 2150? Is there the expectation that something interesting will happen between 2100 and 2150?  This needs to be justified

The request to run simulations out to 2150 is to allow for a longer time horizon for adaptation planning and for the assessment of overshoots with all ScenarioMIP participants’ contribution.  Scenario experiments were first conducted in the 1990s, with a 2100 time horizon.  A similar projection length from present day requires advancing the time horizon of the simulation, getting closer to 2150.  While early plans for ScenarioMIP in CMIP7 were for IAMs to be run out to this date, technical barriers prevented this from happening as we mentioned above.  It was thus decided that, as it remains valuable to have a large number of ESMs extend the standard simulations out to 2150, the scenario extensions’ forcings could be used for this segment.  This will enable the assessment of climate commitments, and greater focus on long term impacts and reversibility of an overshoot, for example.

Line 292: why is there a need for a new high emission scenario? Can’t we just re-use SSP5-85 (or SSP3-70) with updated harmonization?  Looking at Fig 1a, the upward bend in CO2 emissions for this scenario seems hard to justify, unlike the M and ML scenarios

As far as we know, the 8.5 (and even 7.0) scenario have been criticized now for a long time. In addition, many conditions relevant to projections have changed, including most notably costs and rates of capacity addition of renewables, base-year data, anticipated trends in technology adoption and costs, GDP estimates, etc. Also, the CMIP6 SSP3-7.0 had higher loading of aerosols that were seen as “masking” warming and affecting the hydrological cycle, thus creating challenges in interpreting climate outcomes. The impact community asked for as high a scenario as plausible in terms of warming, not just CO₂ emissions, thus our new high scenario is not going to have high aerosol loading.

 Line 315: it is never “exact”, unfortunately

Agreed, we reworded as “actual.”

 Line 335: what will be the process to decide which emissions to use? Will ESMs be allowed to use whichever IAM output they want?

A single marker scenario will be chosen as in the past, and the choice and its rationale will be documented in papers that the IAM teams plan to publish in 2026, detailing the makeup of the various trajectories.

 Lines 356-360: the recent IMO regulations of sulfur emissions have been shown to exert a very significant radiative forcing. See https://doi.org/10.5194/acp-24-13361-2024 for example

We are talking specifically about regionally detailed outcomes across ESMs, but we have added the point about global ERF and the citation, thank you.

 Line 367: who will be in charge of creating this set of emissions? SSP3-70 and SSP3-LowNTCF played that role in CMIP6

At the moment, there are no specific plans to run alternative scenarios in AerChemMIP. In the text, we now have left this open - if future requests would arrive. We have some variations in ScenarioMIP itself, as we have now more clearly indicated.

Lines 383-384: why only involve the IAM community?

That’s a fair question, but the reality is that these would be highly technical and IAM-centered evaluations of the consistency and assumptions behind the various pathways, characteristics that would be hard for a different community to assess.

Lines 424-442: this discussion does not seem to add anything useful to this paper. I suggest deleting

We respectfully disagree, as the role of emulators is becoming more and more important, and, through interactions with the CMIP chairs, this was identified as a topic to specifically address.

Table 2: does this table suggest that ScenarioMIP will be in charge of generating concentrations of all gases listed, aerosol optical depth (which by the way is not a very useful diagnostic to drive a climate model) and ozone?

We have edited the table to make this clearer. Where not otherwise noted, the emissions and concentrations listed will be provided by ScenarioMIP. Most gases’ concentrations will be obtained using a Simple Climate Model.  Aerosol optical depth and ozone concentrations will be provided by an atmospheric chemistry model that will be using IAM emissions as input.

 Lines 461-463: But the DECK only looks at the response to CO2 forcing and associated feedbacks.  Other MIPs should be in charge of other elements of comparison.

We stand by the decision not to be concerned with this aspect explicitly. In the past, experiments or even additional MIPs, like the reviewer suggests, were specifically designed to explore consistency of past and present phases, but the uptake of the question and the experiments by the community did not seem significant enough to warrant taking this as a constraint for the current design, and as a role for ScenarioMIP.

Lines 517-519: what kind of extreme events would lead to high emission future?

We agree this was opaque and we have deleted these lines.

Line 532: what is meant by “strong” here?

We reworded as  “strongly argued and well-founded storyline”

Lines 664-667: why is CDR measures as part of this discussion? If ESM produced temperature signals can’t tell the difference between the scenarios, why would those scenarios be run? This seems more like a topic of interest to the IAM community, but not ScenarioMIP. Maybe part of another MIP?

We have designed the scenario VL to be different enough from LN. Differences can come from CO2 emission pathways, non-CO2 gases, land use as well as from things related to CDR (both the biophysical consequences of land use and also the different representations of CDR in ESMs). However, as in the evaluation using small climate models we only look at the first two factors and the current formulation is confusion, we have now decided to simplify the text and only mention the two first factors..

Line 734: how will forcings be harmonized? Or is it meant to be emissions and LULCC that will be harmonized?

Yes, correct and corrected, thanks.

 Line 743: replace “small” with “simplified”

Thank you, this was an oversight of course.

Line 834: I don’t believe CDR is part of GeoMIP

Thank you for pointing this error out, we have indeed learned from the MIP site: “We are specifically not studying Carbon Dioxide Removal (CDR) methods, wherein carbon dioxide is removed from the atmosphere; potential methods include enhanced mineral weathering, ocean iron fertilization, or direct air capture. Model intercomparisons of CDR are being investigated by the Carbon Dioxide Removal Model Intercomparison Project (CDR-MIP)” and we corrected our text accordingly.

 Line 856: how would ESMs use BECCS resulting concentrations? ESMs don’t have a CO2 tracer just for BECCS.

We have clarified the sentences. The BECCS net result in terms of emissions  will be part of the information on  CO₂ emissions provided by the IAMs, so ESMs won’t process them separately but will respond to their effect as part of the overall outcome in total emissions. The mentioning of concentrations in this sentence referred to ESM models that cannot calculate concentration. These models will directly use the resulting CO2 concentrations calculated on the basis of IAM output (including emissions and CDR). However, directly coupling the word concentrations to CDR created a wrong impression.

Lines 867-869: doesn’t enhanced weathering depend on plant productivity (https://www.nature.com/articles/s41586-024-08429-2)?

Although enhanced rock weathering is performed on agricultural fields in this example, it does not significantly depend on plant productivity. It could influence plant productivity, though, via minerals and/or impacts on ozone. However, because for this section of the text the impact on plant productivity is not relevant, we have removed this part of the sentence.

Lines 874: do IAMs have a representation of wildfires? ESMs will definitely have that in the emission-driven runs.  This could lead to a significant mismatch

There are several reasons why there might be a mismatch, including the different growth rates in IAMs and ESMs, different saturation rates, the impact of climate change and also wildfires. In the experiments we hope to learn from these differences. We have now mentioned wildfires as a possible source of difference. In IAMs, wildfires might be part of the calibration of the forest growth rates (but indeed typically do not include a process representation).

 Lines 949: actually, this paper has proposed a limited set of scenarios (section 3), which, once modeled by IAMs, will lead to emissions, concentrations and LULLC.  Not the way it is stated

We have edited the text to correct this statement.

Lines 971-976: again, this ignores the fact that many ESMs have an explicit representation of aerosols. Furthermore, why would “a single atmospheric chemistry model” be the right approach?  Which one would ScenarioMIP pick? It seems that a tighter collaboration with AerChemMIP is required

Please see our response to the comment above on lines 260-262, which addresses the recommendation for emissions- vs concentration-driven runs. We edited the paragraph in section 2.2 to clarify our request to modelers. Since that request involves providing emissions of aerosol and ozone precursors so that ESMs can run these species in emissions-driven mode, we have deleted the text referred to here, which is no longer relevant.

Lines 977-982: are the authors arguing for the use of RCP8.5 as it was designed for CMIP5, or for SSP5-85 for CMIP6? Are those going to be harmonized to the same 2025 emissions/concentration data?  What about their extension to 2500? Is this part of CMIP7? Very confusing

We agree this point is confusing and have deleted it.

Thank you for the thoughtful comments. I am happy to add a more detailed response to several of them - but would like to consult the full author team (as some of your comments will also be looked at differently by different research communities and authors). Responding now a bit more open... thanks for pointing out to the probabilistic scenario literature and I agree we should add some references. At the same time, also these papers don't solve the problem that these are conditional probabilities at best (based on our current (political) situation. Regarding the role of CDR: I agree with the point made - but please note that it will be increasingly difficult to design scenarios without CDR use. 

Overall I found the article very interesting, and I think the overall experimental design of ScenarioMIP for CMIP7 is good. I think the proposal to focus on emissions-driven simulations is a good one. I applaud the co-chairs for the comprehensive consultation process and their diverse author team. My comments focus on particular aspects of experimental design and narrative.

The authors include a box (Box 1) on characterizing scenario likelihood. The box includes reference to likely scenarios, which have a relatively high probability of occurring. However, the authors state on lines 539-540 that the medium emissions scenario should not be considered a most likely scenario. The authors do not cite or discuss the substantial recent literature on probabilistic scenarios (e.g. Sarofim et al. (2024, https://www.nature.com/articles/s41467-024-52437-9), Moore et al. (2022, https://www.nature.com/articles/s41586-022-04423-8,), in which probabilities are assigned to future emissions, including papers which explicitly aim to characterize the likelihood of previous-generation ScenarioMIP scenarios e.g. Huard et al. (2022, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022EF002715). The latter study calculates likelihoods for each of the SSP-based scenarios used in the previous generation of ScenarioMIP, based on the results of other probabilistic scenario studies, and for example finds that intermediate scenarios, like the medium scenario, are more likely than some other scenarios. Note that such probabilistic scenarios offer substantial potential benefits for use in adaptation planning, since they can be used to make probabilistic projections which are not conditional on particular emissions scenarios. I am not suggesting that ScenarioMIP actually includes such ensembles of probabilistic scenarios, but I think it would be good to cite and discuss this literature to situate the MIP relative to this recent research. For example the authors could cite and discuss this literature and the potential benefits of such probabilistic scenarios for informing adaptation, and say that while they encourage individual modelling groups to run such scenarios, computational constraints preclude running ensembles of emissions across multiple models in ScenarioMIP – and the focus here is on sampling over model uncertainties for a limited range of plausible scenarios.

Authors: We agree that it is worthwhile to refer to the probabilistic projections literature and make clear how the ScenarioMIP approach to uncertainty relates to it. We have added a paragraph to the text in section 2.2, subsection on “a wide and plausible range,” that refers to recent literature ascribing probabilities to emissions projections (note that Sarofim et al. and Huard et al. do not develop probabilistic emissions projections, instead using projections developed in other studies, so we do not include these two references here). We note that we don’t take such an approach in the ScenarioMIP design, given the lack of wide agreement on methods and the necessity of subjective judgments. We believe this topic is better left as an open research question. However, we do now indicate that given the nature of the experimental design in which the highest and lowest emissions scenarios are intended to push the boundaries of plausibility, we anticipate that these bounding scenarios would be less likely than the others.

While I agree with the point, raised in Box 2, that different assumptions about inequality between regions, reflected in different regional patterns of emissions, will not strongly affect the climate outcome, other aspects of physical climate change might be more sensitive to equity and justice assumptions. For example, integrated assessment modellers typically choose a cost optimization approach to reaching a certain level of radiative forcing in 2100, based on a discount rate which embodies intergenerational justice assumptions, and assumptions about the availability of CDR approaches in the future, many of which require extensive land (for example for afforestation, or growing bioenergy crops for use in BECCS), which may in turn impact food security or Indigenous land rights (e.g. Rubiano Rivadeneira and Carton, 2022 https://www.sciencedirect.com/science/article/pii/S2214629622002845). These scenarios typically offset relatively higher fossil fuel emissions in the near term with relatively larger CDR towards the end of the century. Different assumptions about intergenerational equity (i.e. discount rate), or availability of land for CDR, might change these scenarios in a way which would affect global climate evolution. Moreover the use of extensive CDR in the SSP-based scenarios consistent with Paris targets used in the previous iteration of ScenarioMIP is itself often used as a justification for the development of CDR – even though as just described other assumptions in the IAMs might have resulted in less CDR. Similarly, some papers argue that the strong use of BECCS in IAM scenarios is a product of the IAM structure and assumptions, and directly led to the stronger consideration of BECCS by the policy community (e.g. Creutzig et al., 2021 https://onlinelibrary.wiley.com/doi/10.1111/gcbb.12798). Overall, the current draft slightly gives the impression that the authors are trying to sidestep equity and justice considerations by arguing that the physical climate is not very sensitive to these assumptions. But even the selection of certain scenarios for inclusion in ScenarioMIP, with certain ranges of emissions, CDR, BECCS etc can send a message to the policy community, even if this isn’t the authors’ intention. So I suggest that the authors include a stronger caveat in this box saying that the scenarios selected for inclusion in ScenarioMIP reflect certain assumptions in the underlying IAMs, particularly with regard to equity and justice issues, and should not be seen as a full range of possible futures, particularly with regard to technologies such as CDR and BECCS.

Authors: We agree that there are more assumptions in IAMs (including discounting and the availability of CDR options) that have equity implications, so we have added text to Box 2 as suggested by the reviewer. There is also a clear indication to further research on equity-related aspects in future scenarios.

I was interested to see that the VLLO scenario may include ocean-based CDR strategies (ln 643-644), by which the authors presumably are thinking of geoengineering techniques such as ocean fertilisation or ocean alkalinisation (given that the potential carbon uptake of coastal measures is limited). Such measures stimulate uptake of CO2 by the ocean by altering ocean biogeochemistry on the large scale, and in ways which might interact with climate change induced changes. Also such measures could have implications for ocean ecosystems and ocean chemical/physical properties, including changes which could be simulated by ESMs. However, according to lines 247-250 “only afforestation and reforestation will be based on endogenous representation of land -based mitigation solutions in ESMs. For all other CDR options we will include their emission impact within the IAM emission output”. This seems like a potential miss from a science perspective to me. Also, omitting direct biogeochemical effects of ocean-based CDR in the simulations would mean that variables such as ocean pH or ocean carbon uptake, which are regularly examined and reported on including in IPCC reports, would not be correct in these scenario simulations, because they would be missing important biogeochemical processes (ocean alkalinization directly alters ocean pH, and both alkalinization and fertilization aim to directly increase ocean carbon uptake – whereas specifying the effects of these measures by simply reducing net global CO2 emissions would reduce ocean carbon uptake). I suggest either not including ocean-based CDR in this scenario, or if it is included, then either it should be modelled explicitly in ESMs, or if not then the authors need to include discussion of the limitations of not doing so. Separately, I think the inclusion of such measures in the VLLO scenario would need to be treated carefully to avoid the interpretation that only scenarios with ocean alkalinization/fertilization were able to keep global temperatures below 1.5°C by the end of the century.

Many thanks for this comment.  We agree that the VLLO and VLHO scenarios (now renamed as VL and LN, Very Low and Low-to-Negative, respectively) in particular would be strong candidates for a CDR component in the narrative, and it is a high scientific priority to simulate CDR from both land and ocean strategies in order to model both their process-resolved efficacy as carbon removal and their potential impacts on other elements of human and natural systems.

However, the implementation of these CDR strategies raises technical challenges in terms of scenario design (many IAMs do not currently fully represent ocean CDR as part of their mitigation portfolio), data pipeline (the data request to unambiguously and consistently define removal fluxes across different and diverse IAMs), downscaling (the mapping of coarse regional actions in the resolved IAM to spatially meaningful distributions for use in the ESM), ESM CDR process implementation, and sequestration and the reporting of sequestration fluxes within the ESM/CMIP data request.

Authors: None of these technical challenges is impossible, but the tight timescale of the ScenarioMIP process prevents us from being able to realize this dimension within the core ScenarioMIP set.  We hope, however, to work with CDRMIP to produce resolved CDR scenario experiments within the CMIP7 cycle, allowing for a small amount of time to implement the technical pipelines required to meaningfully represent both land and ocean CDR in the ESMs.

The text on lines 709-732 motivates extending simulations beyond 2100 based on understanding long-term dynamics of the earth system and exploring reversibility. Based on these considerations idealized extensions of the scenarios are proposed to 2500. These idealized scenarios include unrealistic features such as abrupt changes in the rate of changes of emissions, and extremely high levels of cumulative negative emissions (around 6000 GtCO2 in H-ext-OS) (Figure 2). But increasingly climate projections beyond 2100 are needed for climate adaptation (e.g. Lyon et al., 2024; https://pubmed.ncbi.nlm.nih.gov/34558764/ ; Park et al., 2025 https://www.science.org/doi/10.1126/sciadv.adn8819; Easterling et al., 2024 https://www.nature.com/articles/s41558-024-02085-0). And 2100 will only be about 70 years away by the time much literature based on the ScenarioMIP simulations is published. Literature containing plausible emissions scenarios beyond 2100 does exist  - for example Sarofim et al. (2024) https://www.nature.com/articles/s41467-024-52437-9 contains a probabilistic scenario to 2300 based on expert elicitation. Meinshausen et al. (2020) contained projections to 2500 for the SSP scenarios https://gmd.copernicus.org/articles/13/3571/2020/, based on assumptions of emissions evolution for each sector. Even the previous generation of ScenarioMIP used more plausible scenario extensions to 2300 than those proposed here (O’Neill et al., 2016; https://gmd.copernicus.org/articles/9/3461/2016/gmd-9-3461-2016.pdf). Even though uncertainties are no doubt larger for scenarios extending beyond 2100, I think it would be a disservice to the impacts and adaptation communities to choose scenario extensions beyond 2100 which are obviously implausible – for example imagine trying to defend coastal defense or mine reclamation plans post-2100 based on one of the scenarios in Figure 2. I encourage the authors to reconsider the extensions beyond 2100 and update with scenarios which are plausible and consistent with the assumptions underlying the scenarios up to 2100. And/or, if the authors retain some idealized scenarios which are not intended to be plausible, then I suggest that they clearly flag these so that they can be separated from those scenarios which are intended to be plausible post-2100. (I notice that Jean-Francois Lamarque raises a related issue in his comments).

Authors: Thanks indeed for your comments and for your well reasoned advocacy on behalf of the impacts and adaptation communities. We take your point that providing plausible, realistic scenarios beyond 2100 is a critical goal, and we agree that the initial design could be improved.

Firstly, though we agree that physical-societal evolution significantly beyond the 2100 timeframe should be a fundamental concern for society, the IAM community does not consider current integrated assessment models fit for purpose for multi-century scale applications.  Economic discounting assumptions which lie at the core of IAMs, combined with technological assumptions which are fundamentally grounded in present day systems, become less appropriate with longer lead times, and for this reason, the majority of IAMs do not attempt to simulate a multi-century timescale.  As such, meaningful models of climate-human-ecosystem evolution beyond 2100 would be a fascinating tool, but we would not consider that this can be achieved by simply extending the time horizon of existing IAMs for centuries.

On the more pragmatic question of whether the time horizon can be extended somewhat to account for the shortening timeframe to 2100 specifically, we fully agree that this would be desirable, and in the early stages of the ScenarioMIP planning process there was ambition to explore a 2150 end date for IAM simulations.  However, achieving this in practice requires assembling a large library of input data/drivers which define time-dependent boundary conditions; for example, defining scenario-dependent technological cost evolution, population and GDP trajectories with country-level resolution.  Unfortunately, the tight timescale of the ScenarioMIP process precluded this plan from being finalised in time for ScenarioMIP CMIP7, but it remains an ambition for future scenario development and climate model intercomparisons.

Given this, for CMIP7, the role of the post-2100 extensions is largely to provide a framing for emissions trajectory extrapolations which are continuous with the 21st century scenarios and span a range of futures which are both plausible and of scientific interest to diverse communities.  As you note, the impacts and adaptation communities are key examples, but the extensions are also of interest for fundamental understanding of carbon-climate dynamics on a range of timescales which can help to inform process understanding and guidance on the limitations of policy tools such as carbon budgeting based on the cumulative emissions accounting.

Based on your feedback, we have made two major changes which can allow us to increase the plausibility of our scenario extensions while staying within the necessarily idealised framework in which we’re constrained. First, we have removed the most extreme overshoot scenario (Hext-OS), from the protocol and replaced it with HLext, which represents a smaller and more plausible transition from a High scenario to one aligned with long-term Paris Agreement goals.

Second, we have added a new section to the paper (4.6) that comprehensively addresses the issue of "unrealistic features such as abrupt changes." Across all extensions, we have implemented methodologies to ensure smooth transitions from the 21st-century pathways and updated the figures with these new results. This includes ensuring continuous derivatives for both CO₂ and non-CO₂ emissions.   We are actively attempting to implement land use assumptions which will not result in unrealistic step changes (an iterative process with the LUH2 team).  We hope that these revisions will result in a set of extensions that better serve the communities you highlight.

Detailed comments:

Ln 268-269: How are modelling groups expected to “keep deviations from observations in the historical period to a minimum”? Does this refer to CO2 concentration alone, or CO2 concentration and temperature? As written it sounds like the authors might be suggesting nudging temperature and CO2 towards observations or similar, but I don’t think this is part of the historical experimental design. Probably the authors mean through model tuning – if so, state this explicitly. Also, the use of historical climate change to tune climate models is controversial, with many groups aiming to only tune the present day climate, and future warming being an emergent property of the model arising from the physics of the system, not something which is set by tuning (e.g. Hourdin et al., 2017 https://journals.ametsoc.org/view/journals/bams/98/3/bams-d-15-00135.1.pdf). Are the authors endorsing tuning models to reproduce historical warming here?

Thanks for this point.  We agree that this text presented an unrealistic request for model development teams. We have replaced the text with what we hope is a more nuanced description of the challenge introduced: “Running historical simulations in an emission-driven configuration presents a challenge, as a model's simulated CO₂ concentration can drift from observations. However, this approach is a deliberate and core feature of the CMIP7 experimental design.  Rather than take measures to artificially correct this bias, we encourage modeling groups to quantify CO2 concentration evolution as a key diagnostic of model performance and carbon cycle sensitivity. To facilitate this analysis, the ScenarioMIP protocol includes at least one parallel concentration-driven simulations specifically to allow for a clean diagnosis of the impact of these emergent carbon cycle feedbacks on climate outcomes.”

Ln 269-270: The authors argue that it is expected that analysis of future projections will focus on deviations compared to the start year of the simulations. This was not generally the case for analysis of future projections in the IPCC AR6 WGI report, because of the interest in climate change relative to preindustrial, as referenced in the Paris Agreement. For example, Figures SPM 4, 5, 7, 8 and 10  in the summary for policymakers of the AR6 WGI report all use quasi-preindustrial baselines for projections.

We have deleted that statement in the process of rewording the entire paragraph where it belonged, as just described. 

Ln 299: The text refers to the high emissions scenario as reflecting “slow development of mitigation technologies and diffusion”. My understanding is that high emissions scenarios such as SSP3-7.0 or SSP5-8.5 reflect no new mitigation actions at all, and as just stated on line 293, a reversal of trends towards mitigation. I recommend repeating the language concerning a reversal of current trends towards mitigation here.

We agree this text was confusing and we have removed it, adding the reference to reversal of mitigation policies as the reviewer suggests.

Ln 378-388: The timeline for provision of forcings from the IAMs seems a bit vague here, and the text makes it sound hypothetical and far in the future. But on lines 474-475 the authors say that the IAMs are running over the period September 2024 – June/August 2025, in order to meet the needs of IPCC AR7. So the IAM runs are presumably well underway. Can the authors add more here on the progress on the IAM runs so far? Are any outputs of early IAM runs available for inclusion in this paper? I find it a bit unsatisfactory not to have actual IAM outputs included in this paper – the equivalent description paper for CMIP6 (O’Neill et al. 2016) did include actual forcing timeseries from IAMs.

This issue has become even more important by now, as the IAM scenarios are now close to being final. However, as they are not final yet, and will be published in papers of the IAM community (with a much more detailed description), we have not included them in this paper. We have, however, revised significant parts of the paper based on the current status of the process. We have therefore produced new versions of the figures showing illustrative emissions pathways and their results concentrations, forcing, and temperature change, based on preliminary IAM and SCM results. We have also edited the paper throughout to change the perspective so that it is from the current scenario status. This has led to edits most prominently in sections 1 (introduction), 2.3 (scenarios), and 2.4 (design issues related to IAM runs).

Ln 512: A ‘competition scenario case’ is mentioned here, but this term hasn’t been introduced, and there is no description of what this means.

The text has been changed significantly, and the term is no longer used. .

Ln 517-518: Are the extreme events referred to here climate/weather extreme events? If so, this statement that extreme events could lead to high emissions seems to contradict what the authors say on lines 210-211, and again immediately following this on lines 521-524 – namely that all the scenarios used assume no climate impacts.

We agree the text was confusing and have deleted it.

Ln 589-591: This text references uncertainty in the carbon cycle response as a source of uncertainty in whether or not the global mean temperature can be returned to 1.5°C. But uncertainty in the physical climate response is also very important here, given for example the still relatively large uncertainty in equilibrium climate sensitivity. I would guess that physical climate uncertainty might be even more important than carbon cycle uncertainty for determining whether or not 1.5°C at the end of the century can be achieved.

Fair point, we have reworded as “(although it should be noted that the actual behavior of the ESMs is at the moment uncertain, due to both uncertainty in the physical climate response and the added uncertainty in the carbon cycle response, when models are run in emission driven mode). “

Ln 665-667: The text here refers to additional impacts from CDR measures. But according to lines 247-250 only afforestation and reforestation will be represented endogenously. Do the authors just mean additional impacts of afforestation and reforestation here? Or if not, how will these impacts be represented in ScenarioMIP? Clarify.

The scenarios include afforestation, but also other CDR measures. In the ESMs afforestation will be represented explicitly and thus indeed, the other implications of afforestation were referred to in this sentence. However, for other CDR measures the response in ESMs to net negative emissions may also lead to scenario differences. The wording of the sentences has been rephrased to make it clearer.

Ln 682: The text here says “the mechanisms and extent of CDR deployment will have ESM-specific efficacies”, but lines 247-250 say “only afforestation and reforestation will be based on endogenous representation of land -based mitigation solutions in ESMs. For all other CDR options we will include their emission impact within the IAM emission output”. If these statements are both true, then the first statement must surely refer only to afforestation and reforestation, since all other CDR options are apparently just represented by reducing anthropogenic CO2 emissions. Please clarify.

We have made the text more precise. Indeed, afforestation and reforestation are represented process-wise in ESMs.

Ln 748-751: Is H-ext really plausible as stated here? There is now extensive literature on the low likelihood of SSP5-8.5 (see Chen et al. (2021), https://www.ipcc.ch/report/ar6/wg1/chapter/chapter-1/, Hausfather (2025) https://journals.sagepub.com/doi/10.1177/29768659241304854, Sarofim et al. (2024) https://www.nature.com/articles/s41467-024-52437-9 among many other sources). This scenario has a similar level of radiative forcing in 2300 as SSP5-8.5 (Meinshausen et al., 2020).

Several impact research areas are interested in the H-ext extension to explore the long term Earth System response to emissions above current policy estimates.  Please also note that our H scenario is significantly lower than SSP5-8.5 in emissions, and therefore expected forcings and temperature outcomes.

Ln 748-749: I would encourage the authors to consider making the M-ext simulation a high priority simulation. Recent literature suggests that emissions of around the medium level are most likely (e.g. Sarofim et al. (2024) (https://www.nature.com/articles/s41467-024-52437-9), Moore et al. (2022) https://www.nature.com/articles/s41586-022-04423-8, Huard et al. (2022) (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022EF002715), Hausfather (2025) https://journals.sagepub.com/doi/10.1177/29768659241304854). And as discussed above, many adaptation planners will require scenarios post-2100. Relying only on H-ext and VLLO-ext (the two planned high priority extensions) means relying on two low likelihood/less plausible scenarios.

In our interaction with the impact research community it was found that the two bracketing extensions remain the most desirable. Also, please note that the experimental design requires all scenarios, and therefore the M scenario too, to be run out to 2150 using the first 50 years of the extensions forcings. Therefore, we hope to balance service to the impact community and computational demands on modeling centers by exploring the extremes of the range, but also providing more than 100 years’ worth of projections according to the M scenario among all others.

Ln 977-982: The authors might also note that consideration of concentration-driven SSP5-8.5 simulations was in the past motivated by the possibility that stronger carbon-climate feedbacks than those included in MAGIC might give rise to SSP5-8.5 CO2 concentrations even if emissions were lower (e.g. Chen et al., 2021, IPCC AR6 WGI). This time around ScenarioMIP is recommending emissions-driven simulations, so even though the High scenario emissions are below those of SSP5-8.5, the CO2 concentrations could be at similar levels in models with a high airborne fraction/strong carbon-climate feedbacks.

Thank you, this was a good point, but no longer relevant as the text has been deleted.

Ln 1028-1033: Another area where emulators may be useful is for simulating future climate change under probabilistic ensembles of scenarios. The authors could mention this here.

Great point, now added to the text.

Thanks... I see the issue - and we should discuss within the author team how to solve this (which might very well follow your suggestion).

Many thanks for this suggestion.  We fully take your point that the current pipeline will introduce an inconsistency between the resolved carbon fluxes associated with afforestation and other land management activities as they are resolved within the Integrated Assessment Models and the Earth System Models.

However, we consider this to be a critical part of the emissions-driven experimental design for ScenarioMIP CMIP7.  There is - as you note - a disconnect between the process representation in the land surface models in IAMs and ESMs in both directions - the IAMs exclude climate-biosphere interactions, while the ESMs often exclude large refinement of energy crops and effects of technological investment in yields.

As such - we see a role of the ESMs in this system as being a source of objective uncertainty in the land use emissions outcomes from land use decisions outlined in the IAM.  Just as the Earth System Model will simulate a different global mean temperature outcome to the original IAM, we expect that consequences of an afforestation strategy - for example - will have a range of consequences for land carbon uptake (or loss) in the ScenarioMIP ESM ensemble.

ESMs will already be computing emissions associated with land-use change in the historical simulations (e.e.g expansion of cropland, wood harvesting, etc). It makes sense to continue this in the future scenarios, now with the addition of land-use change associated with land-based CDR such as afforestation. In fact it would be very difficult to separate emissions reductions from afforestation in an ESM from other land-use change emissions. Although the ESMs will calculate emissions associated with land-use change from expansion of bioenergy crops, the emissions reductions from the associated CCS cannot easily be computed by most ESMs at this point, so it makes most sense to have this component of BECCS emissions provided by IAMs.

Thanks for this suggestion. We agree that a High-Low scenario would be useful for exploring the consequences of late mitigation following a high emissions path, leading to a high temperature overshoot outcome. We have added such a scenario to the ScenarioMIP design. We have adopted the suggested design feature of emissions declining to net-zero CO₂ by 2100, but note that this would not lead to temperature overshoot this century, but only after 2100 (when the extension of this scenario includes net negative CO₂ emissions).

First, while the different scenarios are well well-motivated and well described qualitatively, there appears to be only weak geophysical constraints on them. For some scenarios there are relatively clear constraints (such as the magnitude and timing of maximum warming in an overshoot scenario and the target long-term temperature post-overshoot). But, for others there does not appear to be clear geophysical constraints (e.g. global warming bounds, global mean radiative forcing bounds) for the IAM teams to aim for. Figure 1 gives one example of some geophysical outcomes for the proposed example scenarios Should the results there be taken as constraints by the IAM teams? In CMIP6 (SSPs) and CMIP5 (RCPs) the main geophysical constraint was global mean top of atmosphere (TOA) radiative forcing anomaly at 2100. So, I wonder whether a clear set of geophysical guardrails (upper and lower targets at certain time points) need to be provided for each scenario? There are numerous places in the paper with comments such as “IAM teams should explore measures that minimize trade-offs and exploit synergies” or “IAM teams are encouraged to explore VLLO emission trends under different equity assumptions”. Isn’t there a risk that the resulting IAM scenarios span too broad a range of geophysical outcomes?

Authors: We have edited Table 1 to describe the “Primary design characteristics” for the scenarios, which specifies the emissions or temperature outcomes that are desired for each. The low scenarios are defined by their intended temperature outcomes, while the medium and high scenarios are defined either by their assumed policies (Medium scenario) or their emissions characteristics.

Linked to my first concern, it is not clear what the decision-making process is for deciding which of a potentially large number of different IAM scenarios produced for each proposed scenario, will ultimately be used in CMIP7. ESMs will want one emission and land use data set per ScenarioMIP scenario. Given the relatively weak geophysical constraints on the different scenarios and the need for one “winner” a clearer explanation of the decision-making process for selecting the “winner” for each scenario would be helpful.

Authors: We have described the marker scenario selection process in the introduction. There will be papers from the IAM community going into the details and technical considerations of this.

Linked to both concerns, I see at (Lines 137-139) “the IAM scenarios based on this proposal are developed in the period Sept 2024 to summer 20205, so climate model simulations can start after summer 2025.” I note the word “after” summer 2025. Is the aim that the ESM-ready scenarios will be available by “summer 2025” or just the mix of IAM scenario submissions? If the ESM-ready scenarios, this seems ambitious but if it is the aim then I think some more guidance on (i) geophysical constraints per scenario and (ii) the decision-making processes to go from; numerous IAM scenario per scenario pathway to the single data set to be used by ESMs, needs to be included.

Authors: The text has been edited to update the timeline to the current status of the process, indicating that IAM marker scenarios will be finalized by the end of 2025.

Minor points:

1. Why was 2100 chosen as the end data for the IAM-based scenario? Previously IAMs have generated scenarios for 85 (CMIP6) and ~100 years. This seems to be reduced now to ~79 years. I am very supportive of the scenario extension plans, and understand the IAM scenarios become increasingly less well-founded with time into the future, but based on past CMIP cycles, I wonder why 2125 was not chosen as the end date?

We fully agree that this would be desirable, and in the early stages of the ScenarioMIP planning process, there was ambition to explore a 2150 end date for IAM simulations.  However, achieving this in practice requires assembling a large library of input data which defines time-dependent boundary conditions - for example defining scenario-dependent technological cost evolution, population growth, GDP growth at country-level resolution.  Unfortunately, the tight timescale of the ScenarioMIP process precluded these updates/extensions from being finalised in time for ScenarioMIP CMIP7, but it remains an ambition for future scenario development and climate model intercomparisons.

1. Lines 191-193 talks about the CMIP6 emission scenarios being linked to different socio-economic futures (SSPs). At the time this was deemed to be beneficial for use of projections and scenarios linked through to societal impacts. So will the final CMIP7 scenarios (those chosen as the single representative emission scenario) also be linked to an assumed socoi-economic (SSP) future?

In hindsight, the naming choice of CMIP6 has been a failure in communication, because there should be no one-to-one mapping of SSPs to RCPs. So in this phase we try not to emphasize the linkage as before. Individual scenarios will be by necessity produced using specific SSP assumptions and we will describe that, but the aim is to decouple SSPs from the emission outcomes as much as possible, since multiple SSPs can be used to develop the same RCP trajectory, and a single SSP can be used to develop multiple RCP trajectories. Hence the choice of names, this time around.

1. Lines 267-269 the authors say: “Finally, it should be noted that models running emission-driven simulations can have different temperatures and concentration levels in the start year of the experiments. The ESM teams are strongly encouraged to keep deviations from observations in the historical period to a minimum.” I assume this is encouraging ESMs to try and have an accurate simulation of historical temperatures and CO2. This they generally try to do, without overtly tuning to every up and down in the historical records. As this paper is about the IAM scenarios I find this comment somewhat out of place here and I recommend dropping it.

Thanks for this point.  We agree that this text presented an unrealistic request for model development teams.

We have modified the text as follows: “Running historical simulations in an emission-driven configuration presents a challenge, as a model's simulated CO₂ concentration can drift from observations. However, this approach is a deliberate and core feature of the CMIP7 experimental design.  Rather than take measures to artificially correct this bias, we encourage modeling groups to quantify CO₂ concentration evolution as a key diagnostic of model performance and carbon cycle sensitivity. To facilitate this analysis, the ScenarioMIP protocol includes at least one parallel concentration-driven simulations specifically to allow for a clean diagnosis of the impact of these emergent carbon cycle feedbacks on climate outcomes.”

1. Linked to my main point above, should IAM teams treat Figure 1 as a “target” for their scenarios? This is not really made clear. I also think global mean TOA radiation anomaly would be helpful to add to Figure 1.

We have edited the text to indicate that Figure 1 is based on preliminary versions of the IAM marker scenarios (the figure is not a target).

1. With respect to the high emission scenario (H): Given recent geopolitical developments (e.g. US leaving the Paris Agreement), which of course happened after this paper was written, is there now a need for an increased sampling of the medium to high emission scenario space?

The H scenario was already looking at possible reversal of trends for its justification. A HL scenario has now been added to the design, which increases coverage of the high end of the range.

1. Lines 512-513 discusses a so-called “competition scenario” that sounds like a particular SSP world, which again begs the question as to whether a range of SSPs will be represented in the resulting scenarios used in CMIP7 or whether this aspect of the scenario development is being abandoned?

The text has been edited to give an up-to-date description of the scenarios, their basis in SSPs (all IAM markers are based on SSPs), and the SSPs underlying particular scenarios.

1. Text at lines 686 to 694, discussing options for the VLLO scenario, again reads to me as though there is a large amount of flexibility for what the resulting scenario is made up of. Without clear geophysical constraints I struggle to see what the decision process will be for selecting the single IAM scenario to be used by the CMIP7 ESMs.

With the additional requirement of enough differences between the scenarios and plausibility constraints, the flexibility is somewhat bounded. Still, we asked several IAM teams to elaborate this scenario and finally chose the scenario that fitted the description best - also vis-a-vis the other ScenarioMIP scenarios. This process will be described in detail in the subsequent IAM papers.

1. Scenario extensions: These are important. It would be helpful to have a timeline for the availability of these as well.

The scenario extensions will be available at the same time as the scenarios themselves.

1. With reference to table 6. Is there a target resolution for the gridded BECCS and Afforestation data sets? And how is the DACCS data set delivered in terms of spatial distribution?

We do not plan to deliver DACCS information to ESMs as part of ScenarioMIP.  BioEnergy crops and afforestation will  be passed through the LUH2 pipeline, and will be outputted at the native ¼ degree resolution of LUH2.  Sequestration information for BECCS and DACCS will not be part of the ScenarioMIP dataset - and dedicated efforts like CDRMIP will work to extend the ScenarioMIP land use datasets to include gridded capture and sequestration products.

1. A lot of what comes under Discussion and Conclusions is really discussing things that would be nice to do in the future and could be removed. For example Lines 977 to 982 discusses high emission scenarios and high climate sensitivity models. This doesn’t really seem to be part of this paper on emission scenarios. This is also the case for e.g. lines 990 to 994 and lines 1021 to 1025. I recommend shortening the text discussing future work here and instead more clearly summarize the main recommendations from the paper to give IAM teams a concise summary of what is aimed for and by when.

The ScenarioMIP activity, and therefore the paper, targets a diverse audience, and we think it is important to address here many of the scientific future directions that could build upon this phase and facilitate the next, beyond the focus on this phase’s IAM work. Some of the research efforts we encourage, if successful, could also alleviate criticisms we have been target of, and probably we will continue to be, about some aspects of the design that we are constrained to by necessity, i.e., lack of resources or time, in this phase of ScenarioMIP.  That said, we have eliminated that bullet point about climate sensitivity and high scenarios, which was opaque and, we agree, did not add anything significant to the discussion.

I disagree with Box 1' line of reasoning. The first paragraph mentions that the question under discussion is the "likelihood of a scenario approximating reality for a number of key output variables". But then the second paragraph states that for a scenario to be plausible, it has to be feasible based on the five dimensions of feasibility, which arguably cast a much wider net than "key output variables". I imagine a scenario could have an unfeasible demography with unfeasible technologies, that combined together would yield feasible emissions. I would argue that, from the ScenarioMIP perspective, the only dimensions that matter are the variables used as forcings for climate models, and assessments of plausibility should only look at those.

Authors: We have edited the text to make this point more clearly. It is true that ultimately what is most important for ScenarioMIP is the plausibility of the forcings provided to ESMs (emissions, land use, in some cases concentrations). However, to judge the plausibility of those outcomes, one needs to judge the plausibility of the conditions that produce them. For that reason, the dimensions of feasibility are relevant to judging the plausibility of the scenario outcomes for ESM forcings.

Stylized Pathways (L726)

The paper justifies the decision to use idealized extensions beyond 2100 by the "uncertainties that increasingly affect the socio-economic drivers of these trajectories". This feels arbitrary, and I believe the 2100 switch to stylized pathways is deeply problematic for multiple reasons.

• It breaks the first design principle outlined in the introduction: "internally consistent socio-economic and technological scenarios";
• It will damage the credibility of all climate impact assessments post-2100. Just imagine yourself explaining to a corporate board or elected officials that your climate risk analysis is based on a "stylized emission pathway";
• It creates a break in the provenance of climate projections. All regional, national and international web portals, including the IPCC Interactive Atlas, will have to explain why in 2100 there are abrupt and unrealistic changes in emissions. Communicating climate change projections is hard enough already, please don't make it harder by creating artificial breaks in the series;
• If modeling centers are also skeptical of these stylized pathways, they're less likely to run simulations, which will reduce ensemble size post-2100, hamper our ability to assess hazards, and lower the value of the whole exercise.

An alternative to stylized pathways would be to generate large Monte-Carlo IAM ensembles of emission pathways, and pick those that match best the intended stylized pathways. This would preserve internal consistency and would not muddy communications with end users.

Authors: Thank you for your thoughtful comments. We agree that the credibility and usability of the scenario extensions are paramount, especially for post-2100 climate impacts and risk assessment. Your concerns about the "deeply problematic" nature of abrupt, stylized transitions at 2100 have prompted a significant revision of our methodology.

While the absence of integrated assessment model simulations beyond 2100 necessitates a degree of stylization, we have taken several concrete steps to improve the internal consistency and plausibility of the extensions. Firstly, we will now implement smooth transition periods between the IAM-driven scenarios and the subsequent extensions. The pathways will be constructed to ensure a continuous derivative evolution of emissions, explicitly removing the abrupt changes that you rightly identified as a concern.

Furthermore, this principle of smooth transition will be applied across all forcing components. For non-CO₂ gases, we will use a set of documented, functional choices to ensure a smooth evolution from their pre-2100 behaviour to their long-term levels. For land use, we are replacing the previous step-function change at 2100 with a 50-year linear decline in AFOLU emissions to zero, which can be interpreted as a gradual phaseout of 21st-century forest management practices. We believe these changes directly address your critique and will provide a more credible and defensible set of long-term scenarios for the user community.

 Minor suggestions

170: I would add "internally consistent" to the first design principle, as in line 180.

Agreed, we have edited accordingly

173: As discussed above, I object to this "design principle", which anyway feels more like an implementation detail than a principle.

Please see our reply to the comments on this issue just above.

Box 1: The first paragraph mixes two different ideas: the infinite number of potential futures, and the "dimensions" of each of those futures. It clarifies the latter by saying that we're mostly interested in key forcing variables, but does not address the former. One option would be to focus on the relative likelihoods of ScenarioMIP scenarios (a small, finite, set).

The number of potential futures is addressed in the current text by the reference to a scenario “lying in a given range”; i.e., not exactly matching a potential future outcome. While the likelihood of exactly matching the future is zero, the likelihood of approximating it is finite. Also note that we now discuss the relatively likelihood of scenarios in a new paragraph added just after Box 1.

451: Not sure I understand what the message from this table is, and whom it is intended for. I think a list of data that ScenarioMIP plans to make available, irrespective of whether it is intended for climate models or VIA, would suffice.

We beg to differ, as this paper is aimed at a much broader, diverse, multidisciplinary community, not just ESM modelers, so these distinctions are important to preserve.

498: Iea -> IEA

Corrected, thank you

947: I suggest to include among further research directions something about the relative likelihood of key forcing variables, e.g., something like "The IAM community is encouraged to develop probabilistic assessments of ScenarioMIP' emission pathways to inform climate risk assessments and adaptation decision-making."

We have included a new paragraph in section 2.2 that discusses the probabilistic scenario literature and its relation to the ScenarioMIP approach of judging plausibility alone. We have also added a new item to the concluding section referred to here that encourages research on judging the relative likelihood of scenarios.

To expand on scenario likelihood, what decision makers are looking for can be translated as the (Bayesian) probability of the following statement being true: "Climate model drivers from scenario S are an accurate description of future conditions", for all S in the set of available scenarios. This is beyond the scope of this paper, but I hope the scientific community can finally tackle that question head-on. It's been 25 years since Moss and Schneider (2000) wrote: "We believe it is more rational for scientists debating the specifics of a topic in which they are acknowledged experts to provide their best estimates of probability distributions and possible outliers based on their assessment of the literature than to have users less expert in such topics make their own determinations." 

Thanks again for your work on this, I can't overstate how important this is.

Thank you

We have added a new subsection to section 2.4 that describes natural forcings (solar and volcanic) in the ScenarioMIP scenarios. The CMIP7 forcing task team has developed a new historical reconstruction for stratospheric aerosol optical properties and volcanic sulfur emissions covering the period from January 1750 to December 2023 (Aubry et al., in prep). For the future scenarios beginning in 2025, a specific protocol will be followed to ensure a smooth transition and a consistent baseline for future projections.

The future volcanic forcing will involve a 9-year linear ramp-up from the historical values at the end of 2024 to a new background level. This background is defined as the mean stratospheric aerosol optical depth (SAOD) from the pre-industrial control (piControl) simulations, which is calculated over the 1850–2021 period. This target is higher than the mean SAOD of the more volcanically quiescent 2000-2020 period. The SAOD will ramp up until 2033 and then hold steady at this piControl climatological value for the remainder of the future simulations. Notably, the major eruption of Hunga Tonga-Hunga Ha'apai in 2022 is deliberately excluded from the future forcing scenarios to avoid its large and anomalous cooling signal from complicating the analysis of the anthropogenic warming signal in the scenario runs. Full details of the methodology and dataset are forthcoming (Aubry et al., in prep).

I have a few comments I hope the authors will consider.

Lines 323-332. Figure 1 gives an illustration of the possible CO2 trajectories, and the global temperature response based on a FaIR model. But does it give a realistic estimate what would be expected from ESMs? Figure 7c in Sanderson et al 2024, shows an estimate of ESMs warming from CMIP6 models driven with CO2 emissions with a present day central 50% range of about 0.4C. This is much bigger than is suggested in Figure 1b. I suggest that this is addressed, so that readers will have some awareness of how much larger temperature ranges will be in the CO2 emission driven experiments, compared to the CO2 concentration experiments.

Authors: We have better underlined the approximate nature of this figure, including the large uncertainty bands around the trajectories.

L268-269 How do ESM teams practically keep "deviations from observations in the historical period to a minimum", especially if extra radiative forcing uncertainty (from CO2 emission configurations) is present?

Thanks for this point.  We agree that this presents an unrealistic request for model development teams. We have replaced the text as follows:

“Running historical simulations in an emission-driven configuration presents a challenge, as a model's simulated CO₂ concentration can drift from observations. However, this approach is a deliberate and core feature of the CMIP7 experimental design.  Rather than take measures to artificially correct this bias, we encourage modeling groups to quantify CO₂ concentration evolution as a key diagnostic of model performance and carbon cycle sensitivity. To facilitate this analysis, the ScenarioMIP protocol includes at least one parallel concentration-driven simulations specifically to allow for a clean diagnosis of the impact of these emergent carbon cycle feedbacks on climate outcomes.”

In O'Neill et al. (2016), there was a recommendation about the solar and volcanic forcing in the 21st century. What is the recommendation for CMIP7?

We have added a new subsection to section 2.4 that describes natural forcings (solar and volcanic) in the ScenarioMIP scenarios.

Solar Forcing

The solar forcing recommendations for CMIP7 are provided by the SOLARIS-HEPPA group. The historical forcing reconstruction covers the period up to December 31, 2023, and is extended through December 31, 2024. For future simulations from January 1, 2025, through 2299, the data uses an intermediate scenario from a stochastic ensemble of future solar activity, which is based on a surrogate analysis of cosmogenic isotope datasets. The complete recommendations and dataset construction are described in Funke et al. (2024) .

Funke, B., Dudok de Wit, T., Ermolli, I., Haberreiter, M., Kinnison, D., Marsh, D., Nesse, H., Seppälä, A., Sinnhuber, M., and Usoskin, I.: Towards the definition of a solar forcing dataset for CMIP7, Geosci. Model Dev., 17, 1217–1227, https://doi.org/10.5194/gmd-17-1217-2024, 2024.

Volcanic Forcing

The CMIP7 forcing task team has developed a new historical reconstruction for stratospheric aerosol optical properties and volcanic sulfur emissions covering the period from January 1750 to December 2023 (Aubry et al., in prep). For the future scenarios beginning in 2025, a specific protocol will be followed to ensure a smooth transition and a consistent baseline for future projections. The future volcanic forcing will involve a 9-year linear ramp-up from the historical values at the end of 2024 to a new background level. This background is defined as the mean stratospheric aerosol optical depth (SAOD) from the pre-industrial control (piControl) simulations, which is calculated over the 1850–2021 period. This target is higher than the mean SAOD of the more volcanically quiescent 2000-2020 period. The SAOD will ramp up until 2033 and then hold steady at this piControl climatological value for the remainder of the future simulations. Notably, the major eruption of Hunga Tonga-Hunga Ha'apai in 2022 is deliberately excluded from the future forcing scenarios to avoid its large and anomalous cooling signal from complicating the analysis of the anthropogenic warming signal in the scenario runs. Full details of the methodology and dataset are forthcoming (Aubry et al., 2025).

I.    It is incorrect to claim that the climate models “are not very responsive to underlying assumptions on equity and justice”- especially with respect to regional patterns of land use change (Box 2). In the case of global fossil fuel CO2 emissions, the corresponding warming levels and climate model outputs are dependent only on the emissions trajectory/forcing and not on the regional distribution of emission sources, and it follows that socio-economic assumptions underlying emissions trajectories can be decoupled from climate model outputs. However, this clearly does not hold for land use cover and land use change. Land Use scenarios for the corresponding emissions scenarios proposed in this paper are intended to be generated using Integrated Assessment Models- and therefore involve explicit assumptions about regional incomes, population growth, food demand, food prices, production costs and trade. The range of land use scenarios used as inputs to Earth System Models- and therefore the range of socio-economic assumptions underlying them- will constrain the possible set of scenarios that can be generated subsequently for assessments in the upcoming IPCC AR7 cycle. Socio-economic assumptions already embedded in the land use change scenarios will foreclose various future possibilities of regional distribution of mitigation burdens, and therefore the question of equity is explicitly involved. 

Therefore, we suggest that the ScenarioMIP for CMIP7:

1. State clearly the socio-economic assumptions underlying each land use scenario, and highlight implications for food demand, prices and costs and consequently food security, land availability for urbanisation, biodiversity conservation, agriculture, and energy crops.
2. Generate a baseline land use scenario consistent with present day land cover, that can be used as a control to isolate the effects of land use change by comparison with climate model outputs for other land use change scenarios.
3. Generate a set of land use scenarios that can be used in permutation with the proposed emissions trajectories, as opposed to fixing one specific land use scenario corresponding to each of the emissions trajectories.
4. The range of land use scenarios should necessarily include a set of scenarios with underlying assumptions of ensuring food security in the near term, and also meeting the Sustainable Development Goals.

We underline here that suggestions 1-4 will allow for the development of a wider range of scenarios. This is directly linked to allowing the possibility of constructing scenarios that take equity, justice, and sustainable development into consideration, as indicated by the approved outlines for AR7. Aspects of equity and sustainable development are directly embedded in land-use assumptions. The current proposal assumes that underlying assumptions on equity will not impact outcomes of Earth System Models. We submit that while even this claim must be examined (as discussed in the next set of points), it is undeniable that fixing land-use scenarios from the IAMs will inevitably narrow options for constructing equity-based emissions pathways.

Author's response: Thank you for these considerations. Clearly, socio-economic assumptions in scenarios (including those related to equity) do impact fossil fuel emissions, land use outcomes as well as mitigation costs and the impacts of climate change. We think it is critical to realize that the main focus of ScenarioMIP is to provide a set of reference scenarios that explore climate change patterns using ESMs as a function of emission and land use scenarios. Here, so-far, multimodel studies have - at best - found only mild robus signals from underlying socio-economic assumptions (see Section 2.2). That does not mean that it will not be important to look into this more. We had already pointed out that it will be important to do further work on establiishing the exact relationships between land use and air pollution patterns and local climate change (also see recommendations). It might be also be the interest of more specific MIPs  to look into alternative land use and air pollution patterns (like LUMIP under CMIP6 did run alternative land use cases, very similar to what you describe above). Please also note that in the papers that will be published by the IAM community on the ScenarioMIP scenarios more details on the socio-economic assumptions will be provided. We disagree that the provision to ESMs of specific land use outcomes for each scenario will narrow options for considering alternative equity-based emissions pathways. Such alternative pathways can and should be developed in research of the WGIII research community and coupled with the climate outcomes produced by the ScenarioMIP scenarios.

1. Irrespective of whether ESMs are concentration-driven or emissions-driven, so long as land-use scenarios are generated from IAMs, the underlying range of socio-economic assumptions will, in principle, influence the range of ESM outputs for key climatic variables and carbon budgets. Whether the explicit representation of land processes in emission-driven runs will make outputs from climate models more sensitive to land use forcings and therefore to the underlying socioeconomic assumptions remains an open question.

The paper cites Tebaldi et al 2023 to state that “regional patterns of emissions of short lived forcers and land use… do not necessarily translate to a unique climate change pathway, since they have been shown to lack a robust regional climate signal in a multi-model context”. Tebaldi et al 2023 studies different land use change scenarios to show that the resulting ESM multi-model outputs do not give a clear regional signal in terms of changes in the occurrence of extreme climate events. This, in itself, is insufficient to claim that different regional patterns of land use give rise to equivalent climate forcings.

Modeling studies need to be done to clearly demonstrate the influence (or the lack thereof) of socio-economic assumptions on ESM outputs, rather than simply claiming apriori that ESM outputs are insulated from underlying socioeconomic assumptions.

Author's response: Again, thank you for your thoughtful considerations. In the introduction to the paper work is cited that has failed to find significant and coherent changes for average Temperature & Precipitation. Also, the paper does not consider grid-point-level ‘far-in-the-tail’ extremes. The extreme indices considered are designed to be impact-relevant but also do not ask much of ESMs in their extreme behavior, and the significance of the difference between experiments is computed using big regional averages, and choosing deliberately areas where aforestation/deforestation are most consistent and strong across ESMs. All these choices are trying to facilitate the emergence of the LUC signal, which remains elusive at best.

III. In considering the set of scenarios, we suggest the inclusion of a scenario that stabilizes at 1.7-18 deg. C. We understand that this depends on the sensitivity of ESM runs and if the ESMs can resolve the effects of this temperature difference significantly over internal climate variability. Since a separation of 0.25-0.3 deg. C was suggested for CMIP6, we suggest ScenarioMIP develop a scenario that stabilizes at 1.7-1.8 degrees by 2100.

This would be very policy relevant as overshoot (from 1.5 deg.C) scenarios imply a certain set of interventions required during the period of overshoot. On the other hand, the mitigation, adaptation, and other policy options available for a 1.7-1.8 deg. C stabilisation pathway would be significantly different as compared to a 1.5 deg. C overshoot pathway. The near-term implications of the responses required under both sets of scenarios (1.5 deg. C overshoot vs. 1.7-1.8 deg. C stabilisation) would also be considerably different, e.g. implications for food security if higher CDR is required in overshoot scenarios as compared to stabilisation scenarios. Instead of the current focus on overshoot pathways in the proposal, which in our view is disproportionate, the inclusion of a 1.7-1.8 deg. C stabilisation pathway will provide a more balanced and comprehensive view of future possibilities.

Author's response: Thank you. This is actually meant to be taken care off in the design, among others for the reasons you indicate. While the VL and LNeg scenarios look into different levels of overshoot, the current “Low (L)” scenario addresses the desired levels and is expected to end just below 2C. (Figure 1)

Since this comment, the trajectories with regard to air quality policies/air pollutants/SLCFs were discussed further with most of the researchers involved in this comment. Based on this, we have maintained the assumptions of improving air quality, and therefore overall consistent treatment, across the ScenarioMIP scenario set, of SLCFs. The reason was that it was found that the scenarios do cover a sufficiently large range in air pollution outcomes. We still believe that the best way to find out the exact impact of air pollution assumption is to test alternative patterns as variants of one another only with respect to SLCFs (as in the past happened with variants run by AerChemMIP of SSP1-2.6 and SSP3-7.0). If desired, the IAM groups plan to collaborate with AerChemMIP and produce these variants.

The CMIP7 ScenarioMIP proposal presented here misses the opportunity to study and understand equitable pathways for climate action.

1) Firstly, the proposal refrains from exploring equity and justice concerns in its scenario framework because its main goal is to provide emissions data for ESMs to study climate outcomes. This exclusion is based on multi-model studies (Westervelt et al. (2020) and Tebaldi et al. (2023)) that find no robust differences in climate outcomes when considering different regional distributions of emissions from short-lived forcers and land use. As the ESM assessments are primarily sensitive to parameters like global emission levels and land use, and not much responsive to how these emissions are distributed across regions or sectors, authors argue that equity-focused distributional concerns are outside the scope of ScenarioMIP activities (see section 2.2; Box 2).

Given this, it is surprising to find that later in the paper (see Box 3 and lines 1018-1020 on pg. 28), the authors note that, in fact, this assumption is supported by limited research and “further study of the degree to which variations in regional forcing due to short-lived climate forcers or land use could produce significantly different climate and impact outcomes would be useful” (lines 1018-1020). If the evidence on the impact of regional and sectoral distribution of emissions of short-lived forces and land use on climate outcomes remains limited, then the reasoning of excluding equity and justice considerations in ScenarioMIP becomes questionable.

Author's response. Thank you. There are many aspects of scenario-related research that will benefit from further work; we list several in the concluding section. We don’t believe it is inconsistent to establish the experimental design for ScenarioMIP based on the best current understanding of climate sensitivity to regional forcing, while also encouraging further research on this topic.

2) Secondly, the proposal misses the opportunity to bring in equity and justice discussions, which could be brought in even when discussing scenarios that result in the same forcing overall. The qualitative scenarios discussed in section 3 usually suggest the same policy and technology trends globally, without making any distinctions between developed and developing countries. However, there could be scenarios where emission trajectories between the two groups could diverge; for example, an increase in emissions in developing countries in the short- to medium-term, whereas a steep decline in developed country emissions.

• For example, authors state that the medium emission scenario (M), which assumes the continuation of current policy, can be used to answer questions like, “What are the future physical, socio-economic, and ecological risks implied by current levels of climate change policy?” (lines 540-543). However, there could be a medium emissions scenario where the emissions increase in developing countries to meet their needs for better living standards, which is partly compensated by strengthened mitigation efforts beyond the current policy pledges in developed countries. Similar framing could be suggested in other scenarios as well.
• Authors note that it would be interesting to explore medium (M) and medium-low (ML) scenarios “under a wider set of socio-economic assumptions (lines 564-565). This again presents an opportunity to explore differentiated assumptions on the basis of equity. This could be an important breakaway from previous scenarios that have ended up perpetuating existing inequities in energy access and carbon space between developed and developing countries into the future.
• For designing low emission scenarios, the authors emphasize accounting for feasibility and plausibility concerns regarding technology adoption. It would be critical to note here that such assessments aren’t only informed by technological considerations, such as availability of technology and its cost. Even for the same technology, the actual cost of adoption could be much higher in developing countries compared to developed countries due to the persistent inequities in climate finance. This creates gaps between the feasible levels of technology adoption between developed and developing countries, even when both have committed to stringent climate action.
• The authors mention that they remain ‘agnostic’ to the socio-economic pathways that are used in producing the emissions and land-use pathways (lines 195-195 on pg. 5). However, the qualitative scenario design is largely discussed within the SSP or SPP-like framework, especially for the mitigation levels discussed in low emissions scenarios. If the scenarios are largely produced within the SSP framework, it limits the opportunities to design and understand the climate outcomes of alternative socio-economic realities. However, given the tight timeline to produce scenarios based on this proposal (i.e., by summer 2025), how feasible would it be to incorporate alternative views in the scenario design?

We appreciate the several important research questions the reviewer identifies related to equity issues. We believe all of these are questions that should be pursued in future work, as we encourage in the concluding section of the paper containing a list of future research directions. We maintain that these questions are outside the scope of ScenarioMIP, which is focused on producing climate projections that can be used as input to such research and is forced, by computational constraints weighing on modeling centers, to choose a single trajectory per scenario to be used in ESM experiments..

3) Thirdly, it is curious to note that the authors are suggesting assessing SDGs along with the very low emission scenarios and not the other scenarios. The historical evidence, so far, suggests that many of the SDGs (such as poverty eradication, improved metrics of education and health) are correlated with increased energy use (and consequently, increased emissions). Achieving SDGs remains a key challenge for developing countries, and hence, it would be worthwhile to include them in consideration in other scenarios as well. At least, it would be useful to justify why these are included for low-emission scenarios only.

Our mention of the SDGs in relation to the VLLO (now VL) scenario is not meant to imply that SDGs are unimportant in other scenarios. Rather, it is meant to provide a further constraint on this scenario given the potential for rapid emissions reductions to impact development concerns. We have edited the text describing the VL scenario to make this clear, and note that we also point to future research needed on sustainable development pathways in the concluding section.