The paper is a documentation of the latest model upgrade at the MetOffice. It is exceptionally well written and could almost be published as is. Since this is the documentation of a specific model, it is almost impossible to review it as an external, as a lot of the information is mainly relevant for the specific model. However, everything sounds sound and reasonable and I only have minor suggestions:

L43: “details the all”

L69: Maybe add a reference on ENDGame?

Section 2 reads a lot like a textbook, and it is not clear how much of this is actually needed. It does not serve as a full model documentation. Yet, it includes input that is not relevant for the model update. What is your aim of the section?

L723: “This its”

L882: “shown shown”

L896: “it primarily”

Figure 7: I do not understand the figure and I get the feeling that it would be better to just include the information in the text.

L910: “and a changes”

A message to the editor regarding L916: It is impossible to review the code with such a paper. It is therefore pointless to collect the code.

Review of “The Met Office Unified Model Global Atmosphere 8.0 and JULES Global Land 9.0 configurations” by Willett et al.

This is a model description paper of a new version of the UK Met Office Unified Model (GA8GL9). This is a world-leading model for NWP and climate applications, and therefore this is an important paper. The paper contains a comprehensive description of the updates relative to the previous version of the model. However, I think a number of changes are needed to the manuscript before finalisation.

A large part of Section 2, which contains a description of the model, is copied directly from Walters et al. (2019 https://doi.org/10.5194/gmd-12-1909-2019), including many sentences which are copied verbatim, and some others which are copied with only light edits. All of the subsection titles in Section 2 are the same in both papers. The authors should instead either omit the copied text and just cite Walters et al., or summarise it in their own words if needed, focussing on general description of changes relative to GA7GL7.

Section 3 is long and comprehensive and contains a description of all the changes made in the latest version of the model. I have the impression that the authors have collected a set of changes from a version control system, and described them all here. I got the impression that the section titles used in Section 3 were taken directly from a repository or issue tracker without any editing. I suggest editing the titles so that they are clearer to an outside audience.  While the description of some changes is quite clear, the description of others is rather difficult to follow, and I suspect might only fully make sense to readers who are familiar with the code. Moreover while all the sections describe the changes, some of the sections lack a full scientific justification. I encourage the authors to shorten discussion of technical details, make the text more accessible to outside readers, and add more justification and results of analysis to motivate the changes. Walters et al. (2019) include more such justification in the corresponding section of their paper. Parts of this section of the manuscript also contain many grammar errors which could be picked up by proofreading (I flag a few in my comments but there are many others).

While Section 3 of the manuscript is very long, it only contains a single paragraph on tuning – I suggest devoting more attention to this. Also the evaluation section of this manuscript is also very short – I suggest that the authors consider including more evaluation results in the manuscript.

Is it really the case that the UM source code cannot be publicly shared? Doing so would seem to align with the UKRI policy on making the results of publicly-funded research publicly accessible (https://www.ukri.org/what-we-do/supporting-healthy-research-and-innovation-culture/open-research/). The source code for several other climate models internationally is now publicly available, for example: https://github.com/ESCOMP/CESM, https://github.com/E3SM-Project/E3SM, https://github.com/EC-Earth, https://gitlab.com/cccma/canesm  among others. Parts of the manuscript feel as though they are written for readers who have access to the code – including references to particular parts of the codebase, and description of bug fixes – so it really would make more sense if readers could access the associated code.

The authors should check that all the acronyms they use are defined on first use.

It would be helpful to have more context on the planned uses of this model in the intro. Will this model be the basis of the UKMO CMIP7 climate model? Will it be used in NWP operations?

Specific comments:

Abstract: Were any of these versions used in UKESM (CMIP6 or CMIP7 versions?)

Ln 31: Can you name the CMIP6 models? UKESM1-0-LL and HadGEM3-GC3.1? 

Ln 35: If GL8 is a global land configuration why does it include changes in sea-ice drag?

Ln 36: Can you explain what ‘aggregate tile version’ means?

Ln 56: What does ‘held at the north and south poles’ mean?

Ln 116-123: It would be helpful to refer forward to Section 2.10 here. On reading this section I didn’t know how the aerosol size distribution was modelled, even though this is described later, in Section 2.10.

Ln 134-135: Consider also citing the original articles describing the McICA scheme, such as Pincus et. (2003), https://doi.org/10.1029/2002JD003322

Ln 232-234: The meaning of the sentence “There are additional non-local fluxes of heat and momentum in order to generate more vertically uniform potential temperature and wind profiles in convective boundary layers” isn’t clear to me. Does this mean that the model includes these non-local fluxes in order to achieve more vertically uniform potential temperature and wind profiles in convective boundary layers? If so, why?

Ln 318-323: It wasn’t clear to me why lakes are represented as a canopy. Doesn’t this imply that the lake interacts with the underlying surface in the model, like a tree canopy? Is the same approach used for large lakes like the Great Lakes or Caspian Sea? Does the approach change in conditions where lake ice would be present?

Ln 345-348: If freshwater outflow from inland basins, distributed evenly across all sea outflow points, is an important component of the thermohaline circulation, isn’t this a concern? Physically in the real world there is no such transfer of freshwater – doesn’t freshwater only leave inland basins by evapotranspiration?

Ln 344-353: How is runoff over ice sheets handled?

Ln 368-369: The preceding text gives a physical justification for the use of stochastic physics schemes, but this seems to be contradicted here by the statement that their formulation ‘lacks a sound physical basis’.

Ln 389-391: Are these GMED tickets publicly accessible somewhere? If not, perhaps it is not so helpful to list them here. If they are available, give a reference or URL.

Ln 515-534: This paragraph doesn’t say explicitly what the new value of puns is. I re-read the paragraph and gathered that it is 1.0, but this wasn’t clear on first reading. Also the phrasing ‘value for puns of 0.5 has been used in GA7’ makes it sound like GA7 is the topic of the current paper. I suggest ‘0.5 was used in GA7’.

Ln 539: ‘the time-damping is increments are defined as’ – grammar error – delete ‘is’.

Figure 1: x-axis should be labelled ‘Fraction of time’.

Ln 535-536: This text explains that the convective scheme is intermittent, but it doesn’t explain why this is unrealistic. Isn’t convection intermittent in the real atmosphere? Are the authors arguing that the convection was unrealistically intermittent compared to the real atmosphere?

Ln 558-559: Check grammar “and showed that not only diffusive but also that it was only computational stable”

Ln 562: occasion -> occasional

Ln 581: dimension -> dimensional

Caption to Table 6: ‘season mean diurnal cycles’ -> ‘seasonal mean diurnal cycles’

Ln 625: single level -> a single level

Ln 647: Not clear what ‘otherwise it will be ignored’ is referring to.

Ln 680: The results -> This results

Ln 685: ‘CMT’ is not defined.

Ln 690: What is ‘ProgEnt’?

Ln 694-702: Some comment on the importance of freezing convective rain would be helpful. Does this change reduce model biases? Was it done to better simulate freezing rain at the surface?

Ln 715-720: Can you provide any more justification for the change in refractive index of BC? Note that there are other estimates which have even higher absorption, such as Bescond et al. (2016): https://doi.org/10.1016/j.jaerosci.2016.08.001

Ln 744-74: Could the authors add a few more words to explain why a scaling on PMOA was applied by Gantt et al. (2012), but is not applied here?

Ln 748-758: Is volcanic aerosol diagnosed or prescribed? Section 2.10 indicates that stratospheric aerosol is prescribed, but this section suggests that it is simulated based on emissions. Or is this for stratospheric aerosols in the troposphere? Or are stratospheric volcanic aerosols simulated in the model but not used in the radiation scheme? Clarify.

Ln 757-759: This section should reference Section 2.2 where the loop structure is explained.

Ln 782: doubled counted -> double counted

Ln 810: convective parcel -> the convective parcel

Ln 829-839: The description of the tuning process for the model is very brief. It is helpful to have a list of the parameters that were tuned and some discussion of the target metrics , but it would be good to have a more detailed discussion, given the importance of the tuning process. What were the target metrics? Was any kind of systematic tuning process applied? There is now an extensive literature on model tuning approaches – for example: https://doi.org/10.1175/BAMS-D-15-00135.1, DOI: 10.1126/sciadv.adf2758, https://doi.org/10.5194/gmd-10-1789-2017 . Perhaps there were computational or people time limitations which prevented a more systematic approach to tuning, but it would be good to briefly discuss such approaches, and perhaps speculate on whether they might be applied to the UM in the future. Note that Walters et al. (2019) include a much more comprehensive description of the tuning of the previous version of this model. One other related question - are any of the model parameters scale-independent?

Ln 852: cooling relative -> relative cooling

Figure 3, caption: Why use ERA-Interim for model evaluation rather than ERA5? ERA5 generally has lower biases.

Ln 890: his counteracts -> this counteracts

Figure 7: What does the green rectangle represent?

Figure 8: are proportion to the change in RMSE -> are proportional to the change in RMSE

Ln 907: computation stability -> computational stability

Ln 978: lower Limit the CAPE timescale -> lower limit to the CAPE timescale