The manuscript of Harrop et al., 2023 has two goals:

• to document changes in E3SM model needed to perform several cloud-radiation denial experiments and provide instructions on how to run them
• describe and interpret results of these experiments with the aim of giving guidance on what experiments needed.

While the utility of (1) is limited to E3SM model users, (2) provides several useful insights on cloud-radiation denial experiments with a broader relevance and is therefore appropriate for GMD. The manuscript is well structured, provides several interesting findings, and can be accepted after the comments are addressed.

General comments:

1. It would be great if the authors could provide a more systematic overview of what prescribing monthly mean radiative heating/CRE (Prescribed-Ht and Prescribed-CRE) methods can and what cannot do compared with Cloud-locking.
   
   In particular, it would be interesting to get more information about convection in the Prescribed-Ht and Prescribed-CRE experiments. What causes the changes in the simulated rainfall distribution over tropical land? Why does the frequency of high rain rates decrease? How does this compare to cloud locking?
   
   More discussion on such points would provide a lot of valuable information for researchers to decide which cloud decorrelation method to use in future studies.
2. Could the authors briefly comment on why the manuscript focuses on summer and winter averages, and not annual averages?

Specific comments:

Line 190 – 198, Page 9-10:

While I believe that the details of the SST prescription do not substantially change the surface temperature response, I am not sure that the plot really shows this. Qualitatively, the pattern is indeed similar. But without repeating the same experiment with E3SM, it would be hard to make a good statement. My suggestion is to either add such an E3SM experiment or to remove Figure 4 from the manuscript.

Page 10:

How does surface in “surface locking” compare to ocean-covered areas? In one case, the surface fluxes are prescribed, while in the other case, sea surface temperatures are prescribed. Why did you decide to prescribe the surface energy budget and not directly temperature?

Are substantial surface temperature anomalies a result of your relaxation of surface fluxes, that was in my understanding used to avoid numerical instability?



Also, how is the surface locking method of Lau et al. 2019 that is used in this work different from the prescribed land surface temperature method by Ackerley and Dommenget, 2016 (10.5194/gmd-9-2077-2016)?

Would “surface temperature locking” be a more appropriate method to use when clouds are set to be invisible for radiation?

Page 12:

Figure 5 makes a good argument for the use of Clouds-off ATM instead of Clouds-off LW, although is referenced only in 1 paragraph in the text. It may be interesting to add Surface-locking results in it. Are the anomalies further improved in Surface-locking? If not, that would give another reason for arguing that the surface-locking method is not worth the effort.

Page 14, Cloud-locking:

Could storing control simulation variables less frequently, e.g., every third radiation time step, along with some interpolation for model time steps between time steps with input data, be an alternative way to reduce the storage required by cloud locking?

Page 17:

Figure 8 is only mentioned in one paragraph of the text. It could probably be referenced a few more times.

Page 16, line 356-360:

I guess this is done in the same way that SST is considered in AMIP simulations? (interpolation of monthly means to intermediate time steps). Mentioning this may help in understanding the implementation of Prescribed-RadHt and Prescribed-CRE.

Page 18, line 381:

Why is radiative heating prescribed only within the troposphere? This seems to make the implementation a bit more complex. Would the results be substantially different if monthly averaged heating were prescribed at all model levels?

Page 24, lines 499-511

More reference to Figures 13 and 14 may help get the message across. The BSISO explanation is very detailed and could be left for a follow-up publication.

Page 28, lines 584-594:

Are prescribed-RadHt and prescribed-CRE really good enough to study the role of CRE on circulation?

Could you, based on your simulations, make more detailed statements/suggestions/guidelines about which method is most appropriate for a particular scientific question?

Conclusions:

Your work points at Clouds-off ATM as the clear winner of the COOKIE-style experiments. If so, this should be stated even more clearly in the conclusions.

Data and code:

In which E3SM branch on github can the code be found?

Could you add a branch-specific readme file with some information about the changes made (e.g. short summary, with a link to this manuscript), link to the manuscript(s) where the specific code was used?

Best regards,

Blaž Gasparini with comments from other members of the Climate Dynamics and Modeling team at the University of Vienna

This manuscript provides a detailed description of the experimental design of seven new cloud-radiation denial experiments using the E3SM model, which remove either the mean atmospheric cloud radiative effect (COOKIE/complete cloud-radiation denial experiments) or the covariance between the atmospheric cloud radiative effect and circulation (cloud locking/decorrelating experiments).  These experiments are designed to explore the impact of the cloud radiative effect on various aspects of Earth’s climate system.  The manuscript details the technical descriptions of how to set up and run these experiments, as well as providing some sample results.

Overall, this manuscript provides a nice summary of these cloud-radiation denial methodologies in the E3SM model.  A particular strength of the manuscript is a detailed methodological comparison of slightly different implementations of the COOKIE and cloud locking methodologies, which is something that has been absent from the literature and a very much needed contribution to better inform future studies in this area.  I have a number of minor comments below, which are suggested to improve the clarity of the manuscript.  In particular, I would encourage the authors to be clearer in describing variable names and namelist properties that are specific to E3SM, as this will broaden the reach of this study from E3SM users to other scientists who may wish to apply this methodology in other models.

Minor Comments:

Lines 44-45, 82-83, 316–323, and hereafter: I think better distinction needs to be made throughout this manuscript between cloud locking experiments for which the control simulation is a different climate (such as is often done in investigating the role of CRE in the climate change response; e.g., Ceppi and Hartmann 2016; Albern et al. 2019), versus cloud locking experiments for which the control simulation is the same climate (in this case, the CRE climatology remains identical, but the covariance between CRE and circulation is decoupled; e.g., Rädel et al. 2016; Grise et al. 2019).  In the former (climate change) case, the mean CRE does not necessarily remain the same.  For example, a control run (T0) could be run with clouds either prescribed to the present-day (C0) or warmer climates (C1).  In this case, the mean CRE is dependent on the climate to which the clouds are locked (C0 or C1).  In the latter case, the CRE is always locked to C0, but the individual years are scrambled so that the clouds do not co-vary with the circulation features.

Lines 72-73: It would be good to specify the length of all model runs used in this manuscript (i.e., how many years is each run in Table 1?).

Lines 207–214: More explanation of these results is needed.  Why is there an increase in snowfall for both runs, even though the changes in winter surface temperature in these regions are very different between the runs (Fig. 2)?

Lines 251-255: In this paragraph, you also need to discuss what the presc_sfc_flux_cycle_yr and pertlim namelist settings mean.

Line 347 (Equation 2): It’s more conventional to calculate the mass streamfunction integrating downward from the top of the atmosphere, rather than upward from the surface.  See Chapter 6 of Hartmann’s Global Physical Climatology textbook.  This prevents issues with model representation of surface pressure from impacting the streamfunction calculation over the depth of the troposphere.

Line 385: It’s good to note which vertical level indices correspond to 25-80 hPa, as level indices are listed in the p_radht_coefs variable given on lines 392-398.

Lines 386-389: Why is cpair multiplied by qrs_input and qrl_input, but not by qrs and qrl?  I see the note on Line 425, but you may need to make a similar note here.

Lines 421-425: Why do the fsnt, flnt, fsns, and flns cloud variables have dimensions of (i,1), but the clear-sky variables have dimensions of (i)?  (Note: I now see why after reading Appendix A, but it may be good to refer readers here.)

Lines 443-444: It would be good to define q here as well.

Lines 482–484:  See my earlier comment about the length of the runs.  If the dynamical responses to ACRE are small relative to the magnitude of internal variability in these regions, you would need a relatively lengthy model run to be able to discern the signal from the noise.  So, if the runs are not long enough, some of the discrepancy here could be internal variability, rather than being physically meaningful.

Lines 499-511, Section 4.3:  See first comment above.  Here, if I understand correctly, it appears the authors are locking the clouds to the clouds of the same climate, but with the clouds and circulation de-coupled (similar to Rädel et al. 2016).  Again, better clarity is needed when describing what is done in these locking-type simulations, as the discussion in Section 3.2 is focused almost entirely on describing the cloud locking methodology for the climate change response (as shown in Fig. 16), and not the type of experiments shown in Figs. 15 and 17.

Line 511: See also Benedict et al. (2020), who examined the role of cloud-circulation interactions on modes of tropical intraseasonal variability.

Section 5.2: The text needs to be clearer that this entire section is discussing the circulation response to warming.

Lines 528-530:  I find it difficult to follow what the authors are doing here, even after referring back to the Voigt and Albern (2019) paper.  I would suggest showing the equation for the SST response in the manuscript.

Lines 545-548:  I don’t understand what the authors are arguing in this sentence.  The radiative heating and CRE locking runs also show a poleward shift in the SH Hadley cell edge, consistent with the poleward shift in P-E = 0.

Lines 585-586: I would disagree with this assessment.  For example, there are fairly large differences in the precipitation distributions in Fig. 17 and in the circulation responses to climate change shown in Fig. 16.  While these new methodologies are intriguing, I think the verdict is still out as to whether they can be used to replace the cloud locking methodology.

Figure 5: Statistical significance should be noted on this figure.  It seems important to know whether the differences between clouds-off-LW and clouds-off-ATM are robust, or relatively small compared to internal variability.

Figure 8: Again, it’s important to know where the responses from the control are statistically significant.

Figures 13-14: Are these calculations for annually averaged P-E or summertime P-E?  Also, statistical significance should be noted in Fig. 13, as in earlier figures.

Typos:

Lines 7-8: necessary to implement

Line 113: Missing )

Line 156: Do you mean atmospheric heating rates or atmospheric layer heating rates?

Line 228: I think lwup should be FLUS to be consistent with Line 241.

Lines 382: based --- this seems like an incomplete sentence, based on what?

References:

Albern, N., Voigt, A., & Pinto, J. G. (2019). Cloud-radiative impact on the regional responses of the midlatitude jet streams and storm tracks to global warming. Journal of Advances in Modeling Earth Systems, 11, 1940–1958. https://doi.org/10.1029/2018MS001592

Benedict, J. J., Medeiros, B., Clement, A. C., & Olson, J. (2020). Investigating the role of cloud-radiation interactions in subseasonal tropical disturbances. Geophysical Research Letters, 47, e2019GL086817. https://doi.org/10.1029/2019GL086817

Ceppi, P., and D. L. Hartmann, 2016: Clouds and the Atmospheric Circulation Response to Warming. J. Climate, 29, 783–799, https://doi.org/10.1175/JCLI-D-15-0394.1.

Grise, K. M., Medeiros, B., Benedict, J. J., & Olson, J. G. (2019). Investigating the influence of cloud radiative effects on the extratropical storm tracks. Geophysical Research Letters, 46, 7700–7707. https://doi.org/10.1029/2019GL083542

Rädel, G., Mauritsen, T., Stevens, B., Dommengat, D., Matei, D., Bellomo, K., & Clement, A. (2016). Amplification of El Niño by cloud longwave coupling to atmospheric circulation. Nature Geoscience, 9(2), 106–110. https://doi.org/10.1038/ngeo2630