The authors deeply appreciate the reviewer's careful attention to the previous version of the manuscript and have incorporated there suggested changes throughout.  Original reviewer comments in italics. Author response to reviewers are provided inline in Bold.

RC1: Anonymous:

CMIP has been a cornerstone of international Earth system modelling for the past 3 decades, delivering key science support to IPCC Assessments, while advancing the development of Earth system and climate models and their use to understand past and future evolution of the climate system. CMIP has been grappling with the dual demands of delivering science support (mainly future projections) to international climate change assessments, and the growing climate service sector, while also coordinating research-led experiments to advance Earth system models and scientific understanding. This dual set of demands has caused CMIP to grow significantly over its past two iterations (CMIP5 and CMIP6) in terms of MIPs, experiments to be run, and data to be archived, with consequences for contributing modelling groups. All of this has (and is) being done through short-term, uncoordinated (at an international level) funding, supporting the development of forcing data, realization of experiments, and maintenance of the underpinning infrastructure. Such a situation is difficult to maintain, something had to change in the organization of CMIP going forwards. CMIP7, as described in this paper, is a step towards such a change, with a first attempt to separate simulations intended to support international assessments (e.g. IPCC AR7 and the CMIP7 Fast Track) from other experiments intended to advance the science and modelling of the climate system (e.g. CMIP7 community MIPs). This paper is therefore timely and important. From this perspective the paper clearly needs to be published, though not in its present form. Below I outline a number of points that need addressing before the paper is suitable for publication. I hope this will increase its value for the research community and for CMIP more generally.

Major points.

The paper is very wordy, with lots of long sentences and lists justifying why things have been, or will be, done in a certain way. This makes the paper tedious to read. Addressing this could reduce the length of the paper (easily) by 25% and make it a more enjoyable read! As an example, lines 73 to 122 could be reduced to ~10 lines and still deliver the key messages.

Thanks for your comment. We have substantially edited the manuscript with this in mind.

Section 3.5 adds very little.

Agreed. We have deleted this section and refer to Appendix 3.

While the CMIP IPO is a very good development and is doing a great job supporting the development of CMIP7, I am not convinced much of section 4.1 is really needed in a paper.

We agree that the section was originally too long and have shortened it in the revision and added explicit mention to the task teams, we disagree on its outright removal.  The authors feel that the role of the IPO and associated task teams is critical to publicly acknowledge as part of the formal scientific record in response to WMO Resolution 67 and community surveys requesting increases in support for community engagement.

Section 4.2 is also very wordy and rambling. This is true for a lot of the introduction, which could be reduced in length without losing much.

Agreed. The text has been reduced and tightened throughout the paper.

There are quite a few examples of repetition. e.g. lines 84-86, lines 93-94, lines135 to 138, 144-145, 440-445. This needs to be reduced throughout the manuscript.

Agreed. We have reduced text in the suggested lines.

There are also numerous examples of sentences beginning with long justifications for what is to come based on what has already been said: e.g. Line 93: “In addition to the systematic characterization of climate mechanisms….” or line 110: “Beyond direct contribution to national and international climate assessments…and lots of similar examples. I don’t think these are needed and can be deleted in lots of places.

Agreed.  We have removed these sentences and reduced and revised the text substantially.

The paper has lots of examples explaining how CMIP has (and will) be supported by, aligned with, and deliver to, WCRP. While CMIP is a WCRP-sponsored activity and this is important, it is likely sufficient to say this once (most people know this already) and not have numerous motivations and links to WCRP listed. I suggest reducing these (examples include lines 60-63, 110-120, and others)

We have reduced these throughout the paper.

The 4 research questions are all interesting, and important, What the paper lacks is a clear link between these research questions and the experiments proposed (either as part of the Fast Track or within the community MIPs). Will there be new experiments designed to specifically address some of the research questions? How will the existing experiments advance understanding? In some cases this is clear (e.g. CO2-emission driven models will likely expose (and lead to improvement in) carbon-cycle biases and feedbacks more thoroughly than concentration-driven models) but in many instances it isn’t. The connection between the guiding research themes and the experiments planned in CMIP7 needs to be better explained.

We have added some context for the science questions, including explaining their provenance and connection to e.g. WCRP 2019-2028 Strategic Plan Science Objectives (https://www.wcrp-climate.org/wcrp-sp#sp5). We explain in particular that the questions are an assessment of timely opportunities rather than a constraint on the research agenda.  We also explicitly refer readers to section 3.4.5 where the connections between questions and experiments are made. 

In two places (line 180 and line 645) there is an assertion high ECS models in CMIP6 have been proven to be incorrect and by implication these models are worse than lower ECS models, or just wrong. I don’t agree with this assertion. A high ECS Earth (>5K) is very unlikely, but it has not been conclusively ruled out. If anything, recent increased warming and suggestions of a possible role for changing cloud-radiation processes in this increased warming, may increase the likelihood of a high ECS world. In addition, some of the CMIP6 models with high (increased relative to CMIP5) ECS have been shown to realize this because of improvements in specific cloud feedback processes that were previously (erroneously) balancing other incorrect feedback magnitudes leading to a lower ECS through compensating errors. With removal (improvement) in one aspect of this compensation, ECS has increased. While the higher ECS may not be correct, the underlying processes/feedbacks are likely simulated better. To me this is a model improvement. It would be a pity if CMIP7 discouraged groups from making such important model improvements, even if that risked increasing their model ECS value. I suggest modifying these two assertions.

The discussion around high ECS has been removed. We also highlight that the CMIP7 effort to provide the Rapid Evaluation Framework allows for a better assessment of different aspects of model performance and simulation for different potential end users and applications.

The general aspiration for CMIP7 to separate out policy-relevant simulations (e.g. Fast Track for IPCC AR7) and longer-term MIPs aimed at specific research questions, is a good one. The paper could do a better job explaining and motivating this separation, including how modelling groups could best contribute to either or both parts of CMIP7.

We appreciate the support of the reviewer for this concept and have tried to make the experimental design and connection to the research questions more clear. 

Table 3 is very long and poorly explained. Could it be presented in a more engaging manner? If the main explanations for the different experiments are in the references listed in the table, please let the reader know that. Also, I think there may be some errors in the table. e.g. (i) are piClim-histaer and piClim-histall 30y AMIP or 172y AMIP runs? (ii) For piClim-X and SSPXSST-SLCF I don’t see how feedbacks can be assessed (as suggested in the table) if the models are run in prescribed-SST mode. At least the classical definition of a feedback modifying the SST-response to a given forcing and thus also modifying the forcing itself, cannot be realised in prescribed-SST mode. (iii) for piClimSLCF it is unclear what happens to the non-SLCF emissions. Are these held at PI values? A bit better explanation of this table would help the reader.

We have taken your comments on board and tried to streamline and clarify the information in the table and corrected for errors and inconsistencies.

Minor Points

On the “guiding research questions” I don’t understand why these are “ephemeral” (line 155).

We have made explicit why the questions are ephemeral: “These questions are more focused on the capabilities of current ESMs - and hence more ephemeral and timely - than those posed for CMIP6...”

Regarding the Fast Track experiments, it is not clear if groups are recommended to do everything in either emission-mode or concentration-mode. For example, are there plans for DAMIP to support both emission-driven and concentration-driven experiments? This is not made clear in the explanation of table 3.

The following clarification has been added to Table 3: The esm- prefix indicates experiments are forced by CO2 emissions rather than CO2 concentrations.

Line 494-495: How will DAMIP support analysis of individual forcings in the context of an interactive carbon cycle? Will DAMIP run a coordinated set of experiments for emission-driven ESMs?

Only the FastTrack DAMIP experiments (concentration driven) are discussed here. We now refer to the DAMIP v2.0 paper (Gillett et al 2025) where additional DAMIP experiments are discussed.

Line 128 talks about the lack of infrastructure for a sustained approach. This is also true with respect to funding of modelling groups to realize such regular simulations. This should also be highlighted.

Agreed, so highlighted.

Lines 221-223 on high resolutions models contradicts itself. Please make clearer what you mean here.

This section has been substantially revised.

In section 2.3 I am surprised that emission-driven ESM (scenarioMIP) projections are not discussed more. This seems an important development on CMIP6.

We have added a sentence highlighting the new focus on CO2-emissions forced scenarios.

Lines 266 to 267: while modelling groups suggest that increase in fire over this century (Allen et al. 2024) seems to be an incomplete sentence.

We have revised the text accordingly.

For section 2.4 more discussion on potential MIP contributions to addressing this seems appropriate (e.g. TIPMIP, CDRMIP, C4MIP). I am also surprised there isn’t more mention of global warming overshoot scenarios in this section.

Agreed.  Details of various MIP contributions to the research questions, including overshoot scenarios are included in 3.4.5.

Line 350: coupled carbon-climate ESMs importance in climate stabilization is mentioned. The importance in negative emission scenarios (warming overshoot) is likely even more important to mention.

Revised accordingly.

Lines 386-388: Will there be a coordinated effort to compare CMIP6 historical and scenario forcings to those in CMIP7? This would be a good thing to do (e.g. a forcing comparison MIP).

Comparison of CMIP6 historical forcings datasets with those of CMIP7 is underway as a Fresh Eyes on CMIP project. Comparison of model simulations driven by CMIP6 era versus CMIP7 era forcings is proposed in DAMIP v2.0 (Gillett et al, 2025).

Section 3.2. Will there be any stability/conservation requirements to meet for the piControl or esm_piControl runs?

We have revised this section to include the C4MIP criterion of 10 PgC/century per component and Irving et al., 2021 result comparing ocean heat content drift in CMIP6 piControls.

Lines 421 to 425: I don’t understand what is being proposed here. Please make it clearer.If model X is used in a given science MIP, is it still an entry-card that model-X also runs the CMIP7 DECK? This is not clear

We have revised the text accordingly to clarify that the DECK remains “mandatory” for inclusion in ESGF.

Line 628: The REF is mentioned and somewhere else this is defined as a Rapid Evaluation Framework. What the REF is, and what it is intended for, needs to be more clearly explained.

We have added the following paragraph to the manuscript that explains the idea and structure of the Rapid Evaluation Framework (REF). We also added the reference for the REF that will be available in the GMD CMIP special issue as well.

The authors deeply appreciate the reviewer's careful attention to the previous version of the manuscript and have incorporated there suggested changes throughout.  Original reviewer comments in italics. Author response to reviewers are provided inline in Bold.

RC2: Chris Jones:

Review of CMIP7 documentation paper, by Dunne et al. Firstly to say that the CMIP panel and authors here are to be congratulated on the way they have approached the task of developing CMIP7 plans in a complex landscape of requirements. CMIP has had a lot of success historically but requirements have grown and that growth is not sustainable so the new approach to consult with both users and providers and hence prioritise a more manageable, but still vital, set of simulations has been extremely welcome. The outreach, consultation and dissemination of information has been excellent throughout and this paper contributes to that process. CMIP is a huge undertaking and changes the deployment of resource (both personal and computing/technology) in many, many modelling and research centres around the world. Careful design of what is requested and why is essential. I perform this review mainly in the context that the main aspects of CMIP7 and the Fasttrack, are already determined and too late to make substantial changes. Therefore, I focus on the presentation and explanation aspects with a few suggestions of things which could still be tweaked or clarified. My major comment is to ask for more details on where the “Guiding Research Questions” came from? Are these the result of a consultation on the priority climate science questions? They resemble, but are not the same as, past WCRP grand challenges (e.g. on extremes or carbon cycle).

We have expanded the context for the research questions. As we explain the questions were developed by the CMIP panel as a way of making connections among experiments proposed during initial planning. They represent timely opportunities, based on new observations and evolving modeling capabilities, but do not constrain the research agenda. They are specific to ESMs and hence narrower than e.g. WCRP Grand Challenges.

The way the paper is presented implies you started with these as a guiding set of questions and designed CMIP7 to answer them. But in practice that wasn’t how I recall it happening – so have these questions been retro-fitted to the experiments? E.g. line 132 says that CMIP7 design came from consultation and surveys – this is certainly true of the experiments – but did this consultation also take place for the science questions?

The questions were developed in parallel to the the fast track experimental design.  We have comprehensively rewritten the paragraph to better contextualize them.  The fast track experiments were proposed by the strategic ensemble design task team, in iterative consultation with the MIPs, stakeholders, and CMIP Panel.  The questions, on the other hand, were developed within the CMIP Panel driven by our assessment of the opportunities that recent developments and additional years of observations provide for enhanced scientific understanding consistent with WCRP priorities as they apply specifically to modeling. 

When I look over the CMIP7 web page there are lots of details and further links to the experiments, the task teams, the data request, the REF etc. Your figure 3 is replicated on the website, which mentions the science questions linked to each FT experiment - but I cannot see the questions described or explained anywhere. It feels like these questions have been added after the experiment design. If these really are “guiding questions” that have guided, and are intended to keep guiding, CMIP I think they and their derivation need more prominence. It is not clear, for example, why you identify SST patterns over, say, cloud feedbacks, as a key driver of system sensitivity?

The text has been revised and context added, as described above.

Also, when you discuss a “carbon-water nexus” – is this just a catch-all for things not included in the other questions? The paragraphs of description of this question (sec 2.3) don’t appear to cover interactions between carbon and water cycles as implied by the “nexus” tag.

We have now clarified that the reason for calling it a “water-carbon nexus” is that the CMIP Panel sees water and land carbon as the scope of a set of fundamentally linked problems.

So overall it would be good to articulate maybe how these priorities were arrived at. I am not querying the importance of these questions – they are clearly crucial. But other aspects (for example on aerosol forcing and cloud processes) could also be seen as equally important, and CMIP7 will address many more than just these. Maybe it is better to present the experiments first and then give some example high priority questions as examples of things which CMIP7 may help address – but it feels to be overselling the tag of “guiding questions” to imply that these came first and led to the CMIP7 design.

In addition to the responses above, we have renamed the questions from ‘guiding research questions’ to ‘Fundamental Research Questions motivating Coupled Model intercomparison’ to avoid confusion with how they were developed.

Other suggestions I think are important: Model/simulation quality i. ii. Lines 374-375 – it feels reasonable to suggest a degree of stability of a control run: +5ppm is probably OK – but better as a rate than an absolute – is this +-5ppm per century for example? In CMIP6 C4MIP requested drifts of less than 10 PgC per century in the main pools.

Agreed, we have revised this as was also requested by RC1.

But it would be consistent to also request stability criteria for other metrics – e.g. global T must drift by no more than +- XX degrees, or AMOC within XX Sv. It would be good to treat all major climate components similarly.

We have added the C4MIP guidance on carbon system equilibration, ocean heat content drift, and surface temperature.  Beyond presenting these basic global metrics and requesting additional metrics be saved in the spin-ups (Appendix 1), the authors feel that adequate treatment of individual climate components is outside the scope of this paper.  We also have added context that the Rapid Evaluation Framework will allow for a better assessment of different aspects of model performance and simulation for different potential end users and applications, to support more comprehensive assessment of model performance by global mean temperature alone. 

More importantly – I think it is unwise, however, to suggest arbitrary quality criteria for historical runs. Many ESMs may not hit the historical CO2 within 5ppm. See e.g. Hajima et al (https://egusphere.copernicus.org/preprints/2024/egusphere-2024-188/) for thorough evaluation of CMIP6 models in this respect. What happens if a model does not hit you 5ppm bounds – is it excluded from analysis?

We have changed this statement to “As background, guidance is that modelling centres seek to improve upon the the historical CO2 trend in their esm-hist relative to the CMIP6 ensemble which was found to be biased by -15 to +20 ppmv CO2 by 2014 (Gier et al., 2020) and has been the topic of much recent research  (e.g. Hajima et al., 2025)”.  We have also added the C4MIP guidance on “stable” PI Control carbon budget, but I think it was just a target, not a “requirement” to put data on ESGF, nor will any models be excluded from the ensemble if the criterion is not met. Users will decide whether to use the output in their analysis.

Again – as above, will you also specify acceptance criteria on other measures? – e.g. goodness of fit of the historical temperature record?

As now clarified per the above changes, the criteria are targets but not a “requirement” to put data on ESGF, nor will any models be excluded from the ensemble if the criterion is not met

  This would be a big change for CMIP – to specify acceptance criteria – I think it needs much more consultation before you introduce this.

We agree with this concern and hope the above changes make it clear that CMIP7 will not exclude models based on performance assessment, but the Rapid Evaluation Framework aims to make it easier for end users to assess individual model fitness for various applications and regions. This reflects consultations with developers and user groups in the SED task team which assessed that high level acceptance criteria or model subselection would not be appropriate for CMIP, due to the difficultly of defining all-purpose skill scores.

Ensembles – do you have any recommendations around generation of ensembles (from each model)? I realise you don’t want to rule out models by requiring large ensembles, but some experiments may benefit more than others from ensembles.

We agree with the reviewer and have added the following guidance “ While any size of ensemble is acceptable to meet the mandatory DECK compliance for submission to ESGF, submission of multiple ensemble members of historical and/or esm-hist simulations are highly encouraged as critical to a wide range of detection and attribution questions (see sections 2.1, 2.2, and 3.3). Large ensembles of the Atmospheric Model Intercomparison Project (AMIP) simulations forced by SST and Sea Ice Concentrations (SIC) are also encouraged.”

Line 510 says that the FT “promotes the generation of ensembles” – but it is not clear how? FT does not appear to mention ensembles at all – but it could be a good opportunity to do so. It might be useful to provide guidance on this without mandating.

While the previous language was referring to the CMIP ensemble, not the ensembling of a single model, the point is well-taken, and we have changed this to “The Assessment Fast Track experiments (Table 3) were chosen as a practical balance among the number of participating models, and the complexity, resolution, and  number of ensemble members for each model (Figure 1) to help distinguish the role of different processes and interactions and local versus remote drivers.”

Likewise you could guide on choice of initial conditions (e.g. branch points best taken >XX years apart from the control run).

We have added a section 3.4.6 on the aspiration and best practices for initial condition ensemble generation on revision.  We agree with the reviewer that a strategy which samples states of low frequency climate variability (such as 20 year intervals from esm-pictrl) is preferable to incremental perturbations to avoid aliasing internal variability in the pre-industrial ensemble mean.  We will also highlight the importance of using a sufficiently spun-up control state when branching by recommending a desired maximum drift tendency in section 3.2.   

As an example, quantifying TCRE from flat10 is a relatively large signal-to-noise activity. Ensembles may add little value to this. But quantifying ZEC from the flat10-zec simulation is a very small signal-to-noise and ensembles of this run could be really useful. See e.g. Borowiak et al (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2024GL108654) which shows that ZEC derived from CMIP6 ZECMIP are subject to a level of uncertainty which CMIP6 did not consider due to lack of ensembles.

We have added clarification that we are referring to multi-model ensembles to assess structural variability rather than single-model ensembles to assess internal variability should resolve this.  We also agree with the reviewer that ZEC in particular (here assessed with esm-flat10-zec), is a strong candidate for additional ensemble members for those centers who can afford it.  However, general practice for CMIP7 is that such decisions on ideal extended ensemble size are the responsibility of the corresponding MIP - in this case, C4MIP.

Spin-up. I’m not sure I understand the request to submit numerical results from the spin-up of the models. What is the goal of this – how will they be used? “for curation” sounds like an odd phrase – why do these need curating? And what does “curation” involve – is this the same as archiving on a public database like ESGF?

We agree that we need to provide a better justification than “curation” and have changed this to “public dissemination”– indeed, our hope was that a “Fresh Eyes” team would perform an analysis of this dataset and that it would be in general useful information to researchers doing analysis on the potential role of spinup as a form of “structural uncertainty” and “internal variability”

Model selection. I think you are very wise not to do any prior screening or selection of models. The “hot models” paper you cite in Appendix 3 by Hausfather et al is rather simplistic to provide a table of “Y” and “N” on model screening based on sensitivity. A more nuanced analysis by Swaminathan et al (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2024EF004901) shows clearly that many metrics of crucial interest are not related to ECS. Many high sensitivity models have very good evaluation scores on many metrics and vice versa – having a lower ECS is certainly not a measure of quality. Any screening or selection needs to be much better understood and carried out case-by-case for the application in question. It cannot (yet) be done at the scale of CMIP which has so many downstream uses of the outputs.

We have moved the entirety of the model sub-selection section to Appendix 3 and added reference to the Swaminathan analysis.

Minor comments

• Lines 102-107. This is a nice description of how CMIP has expanded and refined focus as both the expertise and need evolves. It feels that more knowledge of reversibility and symmetry is a big gap in our understanding of the climate system, and here could be a good place to articulate the need for more process exploration of how the system behaves under reversing of forcing.

We have added that the projections include “a range of increasing and recovery trajectories”

• Line 216 says that CMIP7 focus on emissions-driven runs allows for more exploration of extremes under stabilisation – can you explain how so?

We have clarified that “The increasing proportion of models driven by emissions rather than concentrations will allow for novel investigation of extremes under climate stabilization due to the demonstrated rigor of Transient Response to Cumulative CO2 Emissions (TCRE; Matthews et al., 2009) and climate stability under zero emissions (MacDougall et al., 2020). “

• Sec 2.4 on points of no return – is there a reason not to call this either “tipping points” or “irreversibility” which have become much more common phrases for these topics. Wood et al (2023 - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022EF003369) is a good reference here for the framing of high impact/low likelihood outcomes and the need for research spanning different dimensions of this topic.

We have changed “Points of No Return/Ratching” to “Tipping Points” and added reference to Wood et al., 2023.

• Line 297 onwards – describing the CMIP7 DECK intent. It is worth being explicit here that the goal is only to characterise the response to _increasing_ forcing. It was a deliberate decision not to add a DECK experiment to characterise the system response to reducing forcing. (This remains a gap in CMIP7 – noting that flat10-cdr can only be performed by ESMs)

Discussion of priorities for zero and negative emissions forcings experiments is included in the description of the “Assessment Fast Track”

• Table 1 is important. A couple of notes/suggestions - For esm-piControl the forcing is described as “emissions” - I wonder if this should be better described as “interactive CO2” or “simulated CO2” because of course there are no emissions. So even though we informally describe this as “emissions mode” it risks implying that there are some emissions being applied. Or at least specify that CO2 emissions are zero. -

We have clarified that we have “expanded protocol to facilitate participation with ESMs that close the carbon budget and are capable of running with interactive CO2 forced by emissions (including positive, zero, and negative scenarios) in addition to prescribed concentrations.” and added “zero emissions” to Table 1.

Typo – looks like the 1% and historical lines have transposed the solar/volcanic forcing entries

Fixed

• Line 355. Can you clarify the need for 100 years of control run before any experiments are branched off? I don’t recall this being requested in CMIP6

We have added “One change in CMIP7 is the explicit recommendation for modeling centers to provide at least 100 years of their piControl and/or esm-ctrl before the corresponding branching points for 1pct, 4xAbrupt and historical perturbations to allow users to better characterize drift.”

• Line 364 – can you explain why conc-driven control run is required if the esm-control is stable? That seems redundant 

We have changed this guidance to “ Note that a piControl simulation forced by the same CO2 concentration is also encouraged to account for any carbon-climate coupling differences between esm-poControl. “  The concern here is not only that the esm-piControl might not be stable, but that it may have a fundamentally different vegetation state than would be in the piControl depending on the treatment of canopy CO2 under the diurnal cycle and regional variability.

• Table 2 is useful – but it feels odd to name individuals. What happens as/when a person moves job etc? maybe a named group in an organisation is more useful.

These were placeholders in the previous version and have been replaced with citable references.

• Table 2, N deposition. Will this be speciated into dry/wet and oxidised/reduced reactive nitrogen?

This level of specificity cannot be answered at this time as it remains a placeholder until the dataset is provided.

• Line 405. The section on spin-up – it is not clear how the strap line “characterising model diversity” is relevant to this sub-section. Maybe just call the section “ocean and land spin-up” (where land here includes land ice/cryosphere?)

Removed

• Line 470 – is “SCP” a typo? “SSP”?

Yes, fixed

• Table 3 is super useful and important – it will be a very good easy-look-up of the whole set of FT simulations. But it is really big! It is important that it is produced and typeset to be easily readable given how big it is. I feel this comment may be more for the journal/typesetters than the authors – I hope you can find a way to make it well readable.

Agreed

• Table 3 – scenario time period. You quote that scenarios run to 2100 – is this decided? I thought it would be 2125, or at least this was still being discussed. (personal opinion – it drives me mad that IPCC figures and values can only ever quote a climate – i.e. 20-year average – for 2090. So an extension to a minimum of 2110 seems vital so that we can actually quote a 2100 value for projected results!)

We have clarified with ScenarioMIP that the formal IAM “Realistic Scenarios” are driven by population and Gross Domestic Product data that only extends to 2100.  However, all such “Scenarios” will continue past 2100 as more idealized “Extensions” to at least 2150 and in some cases beyond to 2500.

• Appendix 1 – requested spin-up metrics. As per my comment above I’m not yet convinced why you need to request these. But if you do, then to close the land carbon cycle you should also requested cProduct. Even if the control run has no land-use _change_ it will still have land use, and the product pools may well be non-zero. cLand is then the sum of cVeg+cLitter+cSoil+cProduct

Added to Appendix 1

The authors deeply appreciate these perspectives on the previous version of the manuscript and have incorporated most of the suggested changes  Original reviewer comments in italics. Author response to reviewers are provided inline in Bold.

CC1, Mark Zelinka:

Review of “An evolving Coupled Model Intercomparison Project phase 7 (CMIP7) and Fast Track in support of future climate assessment” by Dunne et al [egusphere-2024-3874]

Summary The authors motivate and describe the seventh iteration of CMIP, including the new Fast Track set of experiments which serves the IPCC. The paper is mostly effective in achieving these goals, but there are a few areas needing improvement. This review largely deals with issues relevant to the Cloud Feedback Model Intercomparison Project (CFMIP). Mark Zelinka Maria Rugenstein Alejandro Bodas-Salcedo Jennifer Kay Paulo Ceppi Mark Webb on behalf of the CFMIP Scientific Steering Committee

Major Comments

• Section 2.1 describes the first of four guiding questions in CMIP7, dealing with pattern effects. A large part of the reason the scientific community is interested in pattern effects is because of the science conducted by members of the CFMIP community (Andrews et al. 2015; Zhou et al. 2016; Andrews and Webb 2017; Ceppi and Gregory 2017; Andrews et al. 2018, 2022), facilitated by CFMIP experiments like amip-piForcing (Andrews 2014; Webb et al. 2017), and illuminated by CFMIP diagnostics (including satellite cloud simulator diagnostics that reveal the diverse cloud responses to warming patterns). The “Why expect progress now?” section completely excludes a role for CFMIP while instead mentioning the roles that can be played by DAMIP and AerChemMIP. The focus here seems to be more on what causes warming patterns (a worthy goal), but the understanding of the climate response (including but not limited to clouds) to diverse warming patterns is essential to this problem and should not be neglected. Moreover, the surface temperature response pattern is likely to be at least partly affected by how clouds and their radiative effects feed back on warming patterns (Myers et al. 2017; Erfani and Burls 2019; Rugenstein et al. 2023; Espinosa and Zelinka 2024; Breul et al. 2025) and are involved in teleconnections that propagate surface temperature anomalies from high to low latitudes (Kang et al. 2023; Hsiao et al. 2022). We suggest better acknowledging CFMIP contributions to the current understanding of the pattern effect and explicitly calling out the role that CFMIP can play in making progress. We also note that the first sentence of this paragraph is rather hard to parse and is formulated rather weakly (“xyz may all help” – it remains unclear with what and how). 

We have revised the discussion of all four questions to be more brief and pointed. In formulating question 1 we were anxious to communicate that the SST pattern problem was more expansive than cloud feedbacks but took that thinking too far. The revised text seeks to provide a more balanced and compact discussion of connections between ocean temperatures, clouds, and other processes.

The discussion of opportunities has likewise been sharpened to focus on general ideas rather than contributions from individual MIPs. We do want to emphasize new opportunities, precluding a discussion of previously-performed experiments and diagnostics no matter how valuable they’ve been.

• CFMIP requests that the abrupt CO2 experiments (4x, 2x, and 0.5x) be run out to a minimum of 300 years, and we strongly encourage modeling groups to run beyond that (which could be noted at L331). Note that CFMIP requested this minimum duration as part of the FastTrack consultation process, which was then adapted into the request for the abrupt CO2 experiments. (See the abrupt-4xCO2 request: https://airtable.com/embed/appVPW6XAZfbOZjYM/shrqq9I4NJThwOT9W/tblkc1lkKEtiY Kcho/viw9PLlrOnfUMcvHw/recl01t59HM8jz8ax.) Table 1 currently lists the abrupt-4xCO2 run as extending for “150+ (300)”, though it is not clear what this nomenclature means exactly.  We request that “150+” be replaced with “300+” to make it clear that 300 years is the desired minimum, and “(300)” be replaced with “(1000)”.

We have adopted this suggestion. 

The reasons for requesting that the abrupt CO2 runs be integrated for a minimum of 300 years with strong encouragement to extend beyond that are manifold:  ○ Better ECS quantification: Rugenstein and Armour (2021) quantified with 10 equilibrated CMIP5 and CMIP6 models that 400 years are necessary to estimate the true equilibrium climate sensitivity within 5% error. The model spread in equilibration is large and CMIP6/7 models probably need longer to equilibrate due to the "hot model problem" (Hausfather et al. 2022), which partly consists of temperature- and time-dependent feedbacks. Kay et al (2024) estimated an equilibrium timescale of 200+ years for 2xCO2 and 500+ years for 0.5xCO2, noting important implications for paleo cold climate constraints (e.g., LGM) that can only be understood if the simulations are long enough. ○ Understanding centennial coupled behavior: Simulations of at least 300 years are necessary for estimating the pattern effect, ocean heat uptake and convection (Gjermundsen et al. 2021), AMOC recovery (Bonan et al. 2022), and Equatorial Pacific response timescales (Heede et al. 2020). ○ Understanding and quantifying feedback temperature dependence: This is not well understood, could lead to tipping points and is, after the pattern effect and cloud feedbacks, the biggest unknown in estimating ECS, understanding hot models, and high-risk futures (Bloch-Johnson et al. 2021). It is very hard to quantify because it is obscured by the pattern effect, but is aided by longer simulations.  ○ Practical considerations: Running existing simulations for longer is typically easier than running new simulations. Thus, if computing time is available at modeling centers, it is strongly encouraged that pre-industrial control and abrupt CO2 runs be extended as long as possible. Anecdotally, many of the model centers contributing to LongRunMIP (Rugenstein et al. 2019) had independently run their simulations for longer than 150 years and had the data sitting around, suggesting that in many cases such long simulations are already being performed or are trivial to extend. Currently, ~52 groups are using the LongRunMIP simulations for studies on internal variability, global warming levels, feedback quantification, paleo climate, oceanography, and training for data-driven machine learning approaches.

We have made the change to request 300 years.  Discussion of LongrunMIP, however, is outside the scope of the present work.

Minor Comments

• L34: Should it be “...include experiments to diagnose historical…”?

We have rephrase from “include historical, effective radiative forcing, and focus on CO2-emissions-driven experiments” to “...evaluate historical changes and effective radiative forcing”.

• Introduction section: This section may be too long. The main audience of this paper is the science community that want to understand the rationale and details of the experimental design, not the history of CMIP iterations.

The introduction has been shortened and, we hope , sharpened.

• L90: should be Zelinka et al 2020

Corrected

• L125-127: Suggest being more specific and use “modeling community”, rather than “research community” as a whole. The research community benefits as a whole, but it doesn't share the burden.

Adopted

• L130” “... the present experimental design includes some components …” This point is hard to parse.

This section is modestly rewritten for clarity

The entire paragraph reads well though, but the role DECK plays in climate services might need more highlighting. The remainder of the paper is phrased mostly in terms of science questions and the role climate service plays in there remains somewhat unclear.

• L140: Would it be worth listing a few big questions which were answered mainly or only through past CMIP cycles?

We have actually done the opposite in removing much of the introduction motivation to reduce the length in response to the reviewers but cite Durack et al., 2025 for CMIP history.

• L265-266: something wrong with the phrasing here

Changed to “Tipping Points”

• Table 1: It's unclear why the request is for a small ensemble for historical and a large ensemble for amip

We have added a new section to give more explicit guidance on ensembles (3.4.6).

• Section 3.1.2: It would be helpful to see a plot of how the new forcing datasets differ from those used in CMIP6 during the 1850-2014 period.

Forcings will be the subject of their own set of publications.

• L310/Fig.2: This schematic might benefit from a vertical time axis. The current version leaves a lot of room for interpretation. What are the small orange arrows? What is the connection between DECK and AR7 Fast Track?

The figure has been revised

• L355: “year 100 or later of piControl” – is the rationale for this given anywhere in the manuscript?

Explained as similarly requested by RC2.

• L383: The historical and AMIP simulations end in 2021 according to Table 1.

Corrected

• L498: CFMIP deals with cloud and non-cloud feedbacks (all radiative feedbacks)

Corrected

• L501: Figure 3 excludes RFMIP from the “Characterization” box, yet it is highlighted in this Characterization section, which is confusing.

Corrected

• L510-511: Very hard to parse this statement

Clarified

• L516: “Forcing” should be “Feedback”

Corrected

• L517: I believe you mean “CFMIP” rather than (or in addition to) “CMIP” here

Corrected

• L541: Missing section number

Text has been moved to figure caption

• Table 3, amip-p4K: missing word here? “feedbacks observed” 

Column removed for space considerations.

• Table 3, amip-p4K: the number of years should be 44 (1979 - 2022)

Corrected to 43 (1979-2021)

• Table 3, amip-piForcing: the number of years should be 153 (1870 - 2022)

Corrected to 152 (1870-2021)

• L638: 4 should be 3

Corrected

• Appendix 1 table: Suggest specifying top of atmosphere albedo when referencing rsdt and rsut

Added

• L712-713: Might be some missing words here

Revised language

The authors deeply appreciate these perspectives on the previous version of the manuscript and have incorporated both reference to the many national assessments CMIP has supported as a webpage hosted by the IPO and changed the label of the "Assessment Fast Tract" to align with this more general utility Original reviewer comments in italics. Author response to reviewers are provided inline in Bold.

CC2, Cath Senior:

Comment on Dunne et al 2025: ‘An evolving Coupled Model Intercomparison Project phase 7 (CMIP7) and Fast Track in support of future climate assessment’  This is a very timely and important paper that lays out the evolution of the CMIP project and details the plans for its next phase, CMIP7. I have a couple of comments;

1. An important part of the design of CMIP7 that differs from earlier phases is the separation of policy relevant simulations (the Fast-Track) from the research orientated simulations designed to address the scientific questions and provide a rich characterisation of climate model capability to support future development. I feel the thinking behind this new development could be made more explicit. In particular, how this came about - at least in part- from the feedback from modelling groups about the burden of CMIP6 simulations. Engagement and support for modelling groups contributing to CMIP7 will be critical and documenting more clearly the influence they had on the design of CMIP7 will give reassurance to the community that they can achieve a balance between delivering to their national agendas as well as engagement in international community science.  

The motivation section has been comprehensively revised.

1. There are numerous references to the critical role that CMIP has played in underpinning the IPCC assessments. This is absolutely right and an important point to be made. I also think the authors have tried to carefully lay out that CMIP has - and will continue to support the national and international science communities. However what is perhaps missing is a third important role that the policy relevant simulations have played in supporting the national assessments of many countries. A quick question to ChatGPT(!) gives the following 12 countries/communities that have used CMIP scenarios to deliver their national assessments. It would be good to document this important role emphasising the support CMIP plays for national agendas.  a. United States • National Climate Assessment (NCA) • Led by the U.S. Global Change Research Program (USGCRP) • Uses CMIP5 and CMIP6 projections for national and regional climate impact assessments. • Latest report: Fifth National Climate Assessment (NCA5, 2023) b. United Kingdom • UK Climate Projections (UKCP) • Developed by the Met Office Hadley Centre • Uses CMIP5 (UKCP18) and CMIP6 (UKCPNext) to provide probabilistic and highresolution UK-specific projections. c. European Union • European Climate Risk Assessment (EUCRA) • Managed by Copernicus Climate Change Service (C3S) and European Environment Agency (EEA) • Uses CMIP6 projections within EURO-CORDEX for downscaled regional assessments. d. Canada • Canada’s Climate Change Report (CCCR) • Produced by Environment and Climate Change Canada (ECCC) • Uses CMIP5 and CMIP6 for projections at the national level. e. Australia • State of the Climate Report (by CSIRO & Bureau of Meteorology) • Climate Change in Australia Projections • Uses CMIP5 and CMIP6, downscaled for Australian conditions. f. Germany • GERICS Climate Fact Sheets (by the Climate Service Center Germany) • German Climate Change Assessment Report • Uses CMIP6 projections, often combined with EURO-CORDEX downscaling. g. France • Drias Future Climate Scenarios (by Météo-France) • GREC (Regional Climate Group) Reports • Uses CMIP5 and CMIP6, combined with CNRM-CM models and EURO-CORDEX. h. China • China’s Third National Climate Change Assessment Report • Uses CMIP5 and CMIP6 within China’s regional modeling framework (BNU-ESM, FGOALS). i. Japan • Climate Change in Japan Report (by Japan Meteorological Agency, JMA) • Uses CMIP6 and the JRA-55 reanalysis dataset. j. New Zealand • NIWA Climate Change Projections • Uses CMIP5 and CMIP6, often with regional downscaling via VCSN (Virtual Climate Station Network). k. South Africa • South African Risk and Vulnerability Atlas (SARVA) • Uses CMIP5 and CORDEX-Africa for regional climate projections.

In acknowledgment of the broad utility of this effort for assessment beyond just the IPCC, we have changed “AR7 Fast Track” to “Assessment Fast Track” and the IPO can set up a web page on the use of CMIP in National Assessments at https://wcrp-cmip.org/cmip-use-in-policy/

The authors deeply appreciate these perspectives on the previous version of the manuscript and have incorporated most of the suggested changes  Original reviewer comments in italics. Author response to reviewers are provided inline in Bold.

CC3, Annalisa Cherchi:

Broad and comprehensive article to describe forthcoming CMIP7 effort. Some comments below:

- among the challenging questions, section 2.3 about the water-carbon-climate nexus does not fully exploit water and the importance of the hydrology processes. We know there are still weaknesses and limitation in this (i.e. Douville et al 2021 in last IPCC AR6 and beyond) but there are now more efforts in modelling centres in this direction;

Similar to comment by RC2, we have tried to make this link more explicit.

- Fig 1: the term multiverse seems not fully appropriate as what is shown need and depends on coupling and feedbacks between components and processes. Even here the hydrology part is not fully exploited/described. For example, monsoons are missing among the phenomena. The land interaction is expressed mostly in terms of vegetation and carbon cycle but land is also interaction with the atmosphere via moisture and heat exchanges. In the caption of the figure, red and blue are mentioned as colors for atmosphere and ocean, what about land and cryosphere for example?

We have revised the Figure.

Also related to this, in lines 50-53  model development need to consider and properly represent the coupling between the new components, cryosphere but also improved land-hydrology

We have added mention of cryosphere and land-hydrology interactions as key efforts in improving model comprehensiveness.

- In term of outline of the paper, the key points highlighted in the abstract (lines 33-38) are not fully exploited within the text, either in terms of sectioning and mostly in the summary.

The abstract has been revised.

In addition the summary (section 5) is not a real summary but mostly contains points of discussion and also new features of this CMIP cycle not described in the sections before, ie. Fresh Eyes on CMIP.

The comment on Fresh Eyes has been moved to the earlier sections.

Also the concept of emulators would deserve a bit more of clarification/explanation. Eventually these new aspects could be more extensively described in this manuscript, leaving some details of the experiments to forthcoming papers.

We have added a statement on emulators to the introduction and reframed this section.

For example, there are references to details of ScenarioMIP that is not published yet. There is probably no need for those details at this stage as they will described and explained in details once the reference papers will be ready.

Now that the ScenarioMIP manuscript has been released, we have gone back and made sure this description is in alignment and not redundant or conflicting.

A description (outline) of the content of the manuscript could be useful at the end of the Introduction.

We end the introduction with a sentence describing the following sections

- Overall there are some repetitions (mostly of concept) that could be avoided to simplify the reading (for example, lines 60-76 contains repetitions in the two paragraph and the text could be rewritten and lightened), there are some typos in section 5 (section numbering).

We have deleted these repetitions.

[L60 delete: As an international research activity within WCRP,

Corrected

Line73 delete: As a publicly available ensemble including state-of-the-art coupled model contributions from centers around the globe,

Corrected

L557-563 could be deleted].

Deleted