We accept the suggestion. The submitted text cites the immediate source (the CMIP6 request) and the CMIP5 standard output. We accept the point that the intellectual author should also be cited, at least for variables added in CMIP6. In addition to Griffies et al., 2016, this includes CMIP6 GMD papers for DynVarMIP, HighResMIP, Lmon, LS3MIP, SIMIP and VIACSAB.

We will add the levels in an appendix. The levels may be revised in CMIP7: we will add a note informing readers where to find current information.

We will add the levels in an appendix. The levels may be revised in CMIP7: we will add a note informing readers where to find current information.

The variable Oday.sos was included because it is "considered to be of high importance for characterising the ocean state".

Oday.zos was not initially a candidate because it did not feature in the CMIP6 Data Request. However, following this comment and the detailed arguments supporting inclusion of Oday.zos provided in CC4 we will add it to the list.

1: We accept this suggestion. Thank you for the thorough justification provided.

2: Thank you for the clarification on technical detail. Both Omon.thetao and Omon.bigthetao are included. A note on the need to provide Omon.thetao in addition to Omon.bigthetao will be added in table A4.

The growing importance of emission driven simulations was discussed at length within the authors team, as well as the fact that the identification of all relevant variables to analyse and constrain carbon cycles is an ongoing process. Beside the data request proposed in “Constructing a Global Carbon Budget” opportunity, there are a number of requests dealing with each Earth system realm that will certainly enforce the need to produce at least the variables related to the carbon fluxes. However, we agreed that it is still too early to clearly identify a minimal, consolidated set of carbon cycle variables.  

The Baseline Climate Variables is not intended to be complete, but it provides a starting point based on prior evidence matured from past cycles of CMIP. In particular, the BCVs principally deal with physical metrics that enable for a large interoperability across different models and were widely used/requested in previous climate studies.

We did not attempt to pre-judge the broader and emerging requirements of the community and the fact that AR7 Fast Track is requesting additional variables is in line with the intention here. Certainly, the next revision of the baseline will be able to evaluate a range of variables needed to diagnose emission driven runs which are becoming more established in CMIP7.

The growing importance of emission driven simulations was largely discussed at length within the authors team, as well as the fact that the identification of all relevant variables to analyse and constrain carbon cycles is an ongoing process. Beside the data request proposed in “Constructing a Global Carbon Budget” opportunity, there are a number of requests dealing with each Earth system realm that will certainly enforce the need to produce at least the variables related to the carbon fluxes. However, we agreed that it is still too early to clearly identify a minimal, consolidated set of carbon cycle variables.  

The Baseline Climate Variables is not intended to be complete, but it provides a starting point based on prior evidence matured from past cycles of CMIP. In particular, the BCVs principally deal with physical metrics that enable for a large interoperability across different models and were widely used/requested in previous climate studies.

We did not attempt to pre-judge the broader and emerging requirements of the community and the fact that AR7 Fast Track is requesting additional variables is in line with the intention here. Certainly, the next revision of the baseline will be able to evaluate a range of variables needed to diagnose emission driven runs which are becoming more established in CMIP7.

1. The download statistics are from the server log files which record successful responses to requests received over HTTP, including requests from scripts and from browsers, therefore the wget commands are counted as well. 

More specifically, these log files are sent in near real-time from each data node to the statistics service which is in charge of processing and aggregating the information.

However, currently, there is no mechanism for filtering repeated requests that may come from cronjobs or other tweaking processes: indeed, from the data statistics perspective, all logs with http status 200 are considered for the metrics calculation, with no ways to control the users’ behaviour.

In addition, multiple access to files held on shared file servers are not included, neither are downloads made using Globus. It is an index of usage which is repeatable and well defined from a provenance perspective, but it is not an accurate measure of usage.

We can will add text to clarify this.

2. On density, we are, essentially, following Griffies et al. who specified the approach to be taken for CMIP6. While that is unlikely to be the last word on this topic scientifically, it is the most recent review completed for CMIP.

3. While this is plausible, it is not clear that there is widespread demand. 

4. We can add to the variable long name, though not to the CF Standard Name.

5.  (a) Weighted: the time average is weighted by the time varying spatial coverage.

(b) Radiation: The radiation category is added to distinguish between the properties of the atmosphere itself and the electromagnetic radiation passing through the atmosphere. It is not the same as the cmip modelling realm, so perhaps the heading "realm" should be changed.

Thank you for the detailed and constructive review.

* Table 1: Yes, we will add reference to "research teams and individual researchers at all career stages" to avoid the unintended suggestion that it only applies to MIPs. We note that neither the usage statistics used as an initial guide nor the process of selecting authors relied on the MIP framework.

* Implications for exploiting BCVs for comparison with ECVs: This is a big topic. The drafting of this paper has run in parallel with a revision of the GCOS ECVs. At this stage the two processes are independent apart from some communication at an individual level between those taking part. The question of scientific comparison between models and observations is not picked up in this paper. The reviewer makes a valid point about the fact the climate models and observations are increasingly used together, but the engagement approach used to construct our list, which has been praised by the reviewer, necessarily looked at general metrics of utility and did not go into analysis of scientific use of each variable. The standardising of pressure levels has not bee discussed by this author team. There is such a discussion within the CMIP AR7 Fast Track data request author team for the atmospheric theme, and also in the impacts and adaptation theme. There is a similar, but independent, discussion about a set of standard ocean layers. This is perhaps a good example of how more domain specific issues can be better handled in more specialised settings. In this case the more specialised settings are ad hoc groups set up to create the AR7 Fast Track Data Request. 

* One the interests of indirect users: The indirect users have a significant interest in the objective of greater consistency within and among multi-model ensembles. Inconsistencies on primary outputs create problems in generation of products for indirect users. This is reflected on in the discussion of figure 4. Indirect users often use products which depend on multiple parameters, and lack of consistency in the selection of parameters provided can make it difficult to carry out such multi-parameter calculations consistently across many models. 

* On the link to observational datasets: This is a good question. One aim of this paper is to facilitate discussion of such issues by having the baseline list available ahead of finalisation of the more complete request and divorced from the complexity of specifications about differing experiments and usages. This has not been fully successful, as the Version 1.0 of the data request will be published this November, but the early discussion of this list has raised visibility of some issues.  The current structure of CMIP requires community groups to come forward with requests for variables to tackle science goals. These are not explicitly expressed in terms of specific observational variables. Generally, the description of science goals which is provided in the data request does not go into sufficient detail to identify specific observational datasets. 

* To maximise uptake: we need to continue advocacy, both with WCRP and in the broader community, of the list and the role it can play in enhancing consistency and interoperability. 

* Gaps in the interface between ECVs and ESM-BCVs: there are many, both at technical, scientific, and governance levels (or in terms of axiology, ontology, and epistemology). for example, at the technical/ontological level there is a need to agree a common, interoperable, syntax for recording our lists. A full analysis is out of scope here.

* The question of future iterations of the list is very open at the moment, and goes beyond the mandate of this author team.



Governance questions.

* For CMIP, the CMIP request will be a larger list which takes this as a starting point. The list provides a reference point. The DR communities need to decide on details of implementation, such as specific advice about high-volume variables. Action from the BCV community is not foreseen (there is some overlap of individuals).

* On revisions: as noted above, this is beyond the remit of this author team.

* Questions about exactly what will be needed for implementation in CMIP7, or any other activity, need to be picked up in CMIP. CMIP is itself an evolving process dependent on many community inputs. We cannot specify a precise set of requirements at this point.

* There are issues around consistency of approach to calculation of BCVs. We hope to encourage more consistent gathering of information both about approaches used and concerns of the community.

* Neighbour communities tend to be involved in the data request. For instance there is an active discussion about sea ice variables for comparison with GCOS sea ice variables.

• Omission of Eday.ta appears to have been an oversight. We will add it in the revised manuscript.
• 3hr sfcWInd was discussed and omitted in order to avoid duplication of high volume elements of the list.
• hurs has lower usage, perhaps because it is a variable which is difficult to compute reliably in models and has been plagued with problems. Until we have evidence that it can be computed reliably it is unlikely to be widely used. The 6hr version has been included despite being well down in the usage ranks, but promoting it further does not appear to be justified,
• The list provides for both the 8 and 19 level versions. The redundancy can be exploited or avoided by MIPs exploiting the list to create a request. Similarly for the hourly and 3-hourly redundancy. A short comment on the reasons for retaining redundant information will be added in section 3.1.
• msftyz and od550aer are important variables in their domain, but have not passed the test of broad interest. Similarly for hus and zg at 6hrPt: the specific use case should be dealt with by a specific request, not through inclusion in the baseline.

Pressure levels: yes, we will list these in an appendix.

Thank you for the detailed and stimulating comments.

1. The meta-level view is important, but should include (c) make information more accessible to users from the climate impacts and adaptation communities. This aim can overlap with (a) model intercomparison and (b) comparing with observations, but also brings distinct new considerations. While data on model levels may not be the ideal way of comparing between models, it remains the best approach we have until a standard set of levels is defined and accepted. Such standardisation discussions take place within the atmospheric and oceanographic communities and are beyond the scope of this paper, but we can add a short clarification on this point. There may be progress in CMIP7 which could be incorporated into an update of this list.
2. The monthly mean cloud cover is a very highly used variables. There are a number of variables for which there are concerns both about how uniform the implementation is in models and about how strong the equivalence to nominally equivalent observational variables. We can add a caveat in the conclusions, together with comments on the relation to observations requested by other reviews. 
3. The issue of cloud ice and cloud water issue is being picked up in the author teams of the AR7 Fast Track data request, together with the related point on cloud forward models. See https://github.com/cf-convention/vocabularies/issues/52  . The variables included here, clivi and clwvi, are clearly not appropriate for comparison with observations. The fact that both are provided by a high number of CMIP6 models implies that they can be readily generated and as such can provide a good basis for model intercomparison. As for item 2, we need to warn against direct comparisons with observations. CMIP7 will include a more complete and clearly documented set of satellite based diagnostics (compatible with ISCCP and MODIS), but these variables have a more limited user community at this point.
4. Defining new model diagnostics for direct evaluation against satellite radiance measurements may be relatively simple, but it would involve the specification of a new forward model which is beyond the scope of this work. 
5. New ocean heat content diagnostics are being discussed by the more specialist team of authors convened for the CMIP AR7 Fast Track request. We can reference this.
6. On indices, yes, this would be very nice to have. However, calculation, or even enumeration, of such indices is beyond the scope of this paper, but it may be possible to indicate the potential. The challenge is the need for standardisation and the proliferation of different definitions. There is related discussion here:  https://wcrp-cmip.org/event/ref-project-launch/ . We can add a comment.

Thank you for the detailed and stimulating comments.

1. The meta-level view is important, but should include (c) make information more accessible to users from the climate impacts and adaptation communities. This aim can overlap with (a) model intercomparison and (b) comparing with observations, but also brings distinct new considerations. While data on model levels may not be the ideal way of comparing between models, it remains the best approach we have until a standard set of levels is defined and accepted. Such standardisation discussions take place within the atmospheric and oceanographic communities and are beyond the scope of this paper, but we can add a short clarification on this point. There may be progress in CMIP7 which could be incorporated into an update of this list.
2. The monthly mean cloud cover is a very highly used variables. There are a number of variables for which there are concerns both about how uniform the implementation is in models and about how strong the equivalence to nominally equivalent observational variables. We can add a caveat in the conclusions, together with comments on the relation to observations requested by other reviews. 
3. The issue of cloud ice and cloud water issue is being picked up in the author teams of the AR7 Fast Track data request, together with the related point on cloud forward models. See https://github.com/cf-convention/vocabularies/issues/52  . The variables included here, clivi and clwvi, are clearly not appropriate for comparison with observations. The fact that both are provided by a high number of CMIP6 models implies that they can be readily generated and as such can provide a good basis for model intercomparison. As for item 2, we need to warn against direct comparisons with observations. CMIP7 will include a more complete and clearly documented set of satellite based diagnostics (compatible with ISCCP and MODIS), but these variables have a more limited user community at this point.
4. Defining new model diagnostics for direct evaluation against satellite radiance measurements may be relatively simple, but it would involve the specification of a new forward model which is beyond the scope of this work. 
5. New ocean heat content diagnostics are being discussed by the more specialist team of authors convened for the CMIP AR7 Fast Track request. We can reference this.
6. On indices, yes, this would be very nice to have. However, calculation, or even enumeration, of such indices is beyond the scope of this paper, but it may be possible to indicate the potential. The challenge is the need for standardisation and the proliferation of different definitions. There is related discussion here:  https://wcrp-cmip.org/event/ref-project-launch/ . We can add a comment.

Thank you for the constructive comments.

1. Plans for expansion are, in the short term, covered by the CMIP AR7 Fast Track Data Request (https://wcrp-cmip.org/cmip7/cmip7-data-request/public-consultation/.) We will add a reference. 
2. We accept that this list does not support comprehensive analysis in many areas: supporting comprehensive analysis will need different lists tailored for different topics. That is the role of the fast track request referred to in item 1.
3. Mixed layer thickness is clearly an interesting parameter, but it is not at this point clear that it is sufficiently well defined to be a priority for model output. At this point the evidence does not point to strong demand for this parameter as a model diagnostic.
4. We will clarify the selection process.