Review of "MIROC6 Large Ensemble (MIROC6-LE): Experimental design and initial analyses" by Shiogama et al

This study introduces an impressive new dataset that is available for community use with the MIROC6 model.  This dataset consists of large ensembles of historical and future projections with multiple scenarios as well as single forcing simulations.  The primary purpose of the paper is to introduce these simulations but it also presents some cursory, but useful, analyses of changes in global temperature, precipitation and their extremes and some assessment of non-linearity in the single forcing simulations and analysis of the number of members required to detect changes or differences between scenarios.  Overall, I think this is a useful and well written study that introduces this important new dataset and I have only minor recommendations to consider before publication.

l83: It could be worth providing a bit more information about the piControl and the initialization dates.  Firstly, it could be worth stating whether the piControl is still drifting at this stage and some more specifics about the initialization dates e.g., were they spaced by a certain number of years?

l87: It can sometimes be a bit confusing where biomass burning aerosols are represented.  I'm assuming that they are included in the anthropogenic aerosol contribution?  Even though there is a natural component to that.  It might be worth being clear about this.

l120-126 and Fig 3: I'm not sure what the motivation is for doing this assessment of non-linearity by using only samplings of 1 member.  It may be that there is a true non-linearity but you can only see it in the ensemble means.  You could do the same analysis but sample N members with replacement from each ensemble, where N is your original ensemble size, to determine whether there are any non-linearities that can be detected with the ensemble means.

l137: At the introduction to Fig 5, it might help readers to remind them what time period is being considered.  I think it's 2000-2020 minus 1850-1900?

l147-151: I think this text is describing the behavior of the hist-nat+ssp245-nat run in Figure 6, but it's not entirely clear.  Maybe reference that part of the figure when referring to the solar and volcanic contributions.

l210: It seems like another possibility beyond the interannual external forcings is inacuracies in the use of a linear trend?  If so, that could be mentioned too.

l224: I got confused by the wording here.  You refer to a "single-member estimate" but then proceed to discuss the method, which doesn't sound like a single member estimate at all.  The description sounds like an "N member estimate".  Suggest clarification.

l246: Presumably some measure has been chosen to quantify whether it has been "degraded".  Suggest being clear by what measure you are using here.

l310: Again, it seems you need to have chosen some threshold to quantify whether the amplitude of the internal variability is underestimated.  Suggest being clear about how you have determined that.

l315: There is an accompanying single forcing large ensemble for the CESM2-LE which I think would increase the number of years of simulation for CESM2 (https://www.cesm.ucar.edu/working-groups/climate/simulations/cesm2-single-forcing-le)

Typo's/wording:

l58: suggest changing "and" between the reference to the biomass burning simulations and the greenhouse gas simulations to "or" since it is not

both that are time evolving.

l131: "TX" --> "Tx"

l157: Here, and throughout, there's some inconsistency as to whether you refer to "ssp" or "SSP" and "ssp245" or "SSP-2.45".  Suggest being consistent.

l252: "variabilities than the best" --> "variabilities compared to the best"

l319: "federation grid" --> "grid federation'

Review of “MIROC6 Large Ensemble (MIROC6-LE): experimental design and initial analyses by Shiogama et al.

This paper is intended as a documentation paper for a new Large Ensemble of climate model simulations with MIROC6. This is an extensive set of new simulations, including historical & scenario simulations, as well as single forcing experiments. It presents the experimental design and overview of simulations, as well as presents some high-level analyses of global climate characteristics of this ensemble, in particular testing the linear additivity of the single-forcing experiments to make up the all-forcing simulations, and projected changes in ENSO variability.

The paper introduces a valuable new resource for the climate community, is well written, and presents some interesting first analyses of climate characteristics in this ensemble. This is a valuable contribution, and I would recommend publication subject to minor revisions. I have included my comments below. Comments 6) and 12) are slightly more substantial than the others but should be easy to address.

Experimental design:

1) Do the 50 historical ensemble members of the LE include the CMIP6 historical ensemble members, or are they separate? Is the model version identical to CMIP6 (i.e., can both ensembles be merged)?

2) How were the initial conditions years selected / spaced out in the piControl run?

Other comments:

3) Timeseries figures (1, 2, 6 & 7): it would be very helpful to draw a line at 0

4) L105: It might be worth noting that the effect of hist-stratO3, hist-totalO3 and hist-lu are small but not zero especially in the 2^nd half of the 20^th century

5) L108: do you mean to say small trend or no trend? I would suggest either using “the P of the historical runs only has a small trend” or “has little trend” depending on the intended meaning

6) L117-126: The median of the blue bars (corresponding to the linear sum of single forcing experiments) is in some cases not near the middle of the orange bars, suggesting some non-linearities in the medians / ensemble means. There could still be important non-linearities here that would be evident in the ensemble means (or medians); overlap of the blue / orange ranges is not evidence of the absence of non-linearities in the ensemble means. This should be commented on here

7) L139 typo: faction -> fraction

8) Figure 4: I would suggest using a title for each row e.g. deltaTx(hist)-deltaTx(hist-nat) to help the reader

9) L153-155: I don’t think ‘nearly disappear’ for the aerosol signal by end of 21^st century in T is quite correct: substantially reduced yes, nearly disappear I’m not sure – it looks like -0.2 degrees C by the end of the century. That’s still about half of the strongest aerosol response in the late 20^th century (approx. -0.4C by eye)?

10) L 187: Internal variability (=variabilities) in plural sounds a bit unusual, I think you could use Internal variability singular here and in other places

11) L 210: single-ensemble – don’t you mean single-member?

12) The methods and terms used to describe the two methods are somewhat ambiguous (L190ff): I think it should be made clear that the single-member / multi-member name refers to the method for removing the forced response only, since the circle and crosses in Figures 10, 11 and 12 are both estimated from averages across multiple ensemble members. The authors could try using different names: multi-member-mean removed and single-member trend removed, for example

The authors also need to explicitly state how they calculate the variability once the forced response is removed in both methods: standard deviation across time for each member, and then averaged across all ensemble members? Are you averaging standard deviations or variances?

L 198 is especially ambiguous: does the “best estimate” ensemble averages apply to both methods?

13) L291: It might be worth stating explicitly that AGCMs cannot simulate changes in coupled modes of variability and SST patterns, and hence can only inform projections and attribution statements conditionally with respect to prescribed SST patterns

14) L297: perhaps worth inserting “future changes in ENSO variability" or "ENSO amplitude”

General comment:

It is somewhat obvious, but the authors state in a number of places that larger ensembles are needed to detect changes between ssp119 and ssp126. This is certainly true, but it might be worth saying somewhere something to the effect of “unsurprisingly, the smaller the difference in forcing, the larger the ensemble needs to be to detect differences in the forced response. This is just a suggestion, the authors can choose to take it or leave it!