This paper needs major revisions.  It reports unusual responses to black carbon injections, but does not diagnose the climate response to explain why.  It seems that your model does not loft the BC into the stratosphere nor rain it out, and so it persists somehow in the upper troposphere, heating it and probably evaporating clouds.  Is this physically realistic?  Please show the vertical distribution of the BC and clouds over the regions that experience surface warming in the first JJA.  Without looking at the distribution of clouds, precipitation, and BC, how can we figure out why there is warming.  Is it downward emission of longwave from the BC cloud?  How is this possible if there are water clouds?  The paper calls for future work on that, but this paper needs that analysis.  Otherwise, the strange result is not explained.

Why does the paper choose the BC injections that they did?  They call for model intercomparison studies, but do not use the same injection scenarios as other studies.  It seems they are not aware of Toon et al. (2019) which examined scenarios with the same range as this paper, 5 Tg, 15 Tg, 27 Tg, and 37 Tg.  This paper should have used the same ones so they could be compared.

Toon, Owen B., Charles G. Bardeen, Alan Robock, Lili Xia, Hans Kristensen, Matthew McKinzie, R. J. Peterson, Cheryl Harrison, Nicole S. Lovenduski, and Richard P. Turco, 2019: Rapid expansion of nuclear arsenals by Pakistan and India portends regional and global catastrophe. Science Advances, 5, eaay5478, doi:10.1126/sciadv.aay5478.

They also should be aware of Oman et al. (2006).  They found summer warming over Africa and Asia after a volcanic eruption, and the explanation was a weaker summer monsoon, which produced fewer clouds.  This paper needs to show the distribution and anomalies of clouds and precipitation in the first JJA, and diagnose whether the response they found was due to a dynamic response, or simply an evaporation of the clouds due to upper troposphere heating.

Oman, Luke, Alan Robock, Georgiy L. Stenchikov, and Thorvaldur Thordarson, 2006:  High-latitude eruptions cast shadow over the African monsoon and the flow of the Nile. Geophys. Res. Lett., 33, L18711, doi:10.1029/2006GL027665.

The BC lifetime in this simulation is much shorter than in previous work.  This has to be explained, with figures showing the BC distribution vertically.  What was not washed out immediately should have been transported to the upper stratosphere and lasted for years.  Previous work had e-fording lifetimes of 5-7 years.  Why is it so short here?

The terminology of “low-moderate” injections, “standard high-end cases,” and “relatively weak” injection needs to be changed.  There is no standard case, and all incidents of nuclear war, including 5 Tg injections would be horrific, and would not be low or moderate.

The caption for Fig. 4 says sea ice is shown by orange lines, but I don’t see any.

There are 23 additional comments on the attached annotated manuscript which also should be addressed.

Review by Alan Robock

The authors explore large scale climatic impacts of low-to-moderate black carbon injections associated with a nuclear conflict. While the manuscript is adequately written,  I’m not convinced about the robustness of some of the key results discussed in this paper, and I believe some more analysis and discussion is needed before the paper can be published.

Major comments: 

(i) The authors claim they find a non-linear response across the different injection rates. However, 3 our 4 simulations (including the 12 and 24 Tg injection cases used to produce the results in Table 1) consist of only 3 ensemble members and the analysis is done for single individual time periods  (e.g. JF mean in year 1, JJA mean in year 1, etc), leading to 3 data points in total contributing to each response. While this should be sufficient to estimate the changes in BC burden and related variables (e.g. AOD) with relative confidence, I’m not convinced this is sufficient to estimate the response in global mean SW and LW fluxes or surface temperature with confidence sufficient to confidently claim whether linearity or non-linearity is found. As is also evident in Fig. 2 (a,c,d), the interannual variability on most of these variables is not negligible. While Table 1 reports 95% confidence intervals, it is not clear to me how are these calculated. Also, how big is the ensemble spread of these values? Is it adequately represented by the way the authors calculate the 95% confidence interval? I would expect even the value of standard deviation calculated based on just 3 data points will be subject to large uncertainty..

On a related note – Section 2.2 discusses the nuclear winter experiments performed, but nowhere does it say what data are used for calculating the responses. Is it the preindustrial control simulation? If so, how many years and/or ensemble members are being averaged?

(ii) I agree with the other reviewer that better understanding of the drivers of the reported boreal summer warming in year 1 is required before publication. In fact, I’m not convinced by the authors’ claim that this is driven predominantly by the change in downward LW forcing, as the change in downward SW is at least factor of ~2 larger and opposite sign. And so considering just changes in direct downward radiative forcing, the overall change in RF should be still negative. And so the changes have to be driven by indirect processes, either changes in the upward  LW/SW and/or latent and sensible heat fluxes. In fact, to me the pattern of surface temperature changes looks very much like a dynamically induced response (I was thinking something to do with NAM maybe), or a BC-induced modulation of the model’s internal modes of climate variability.   In fact in figure 2d, it looks like an injection drives am oscillatory behavior in the model, that is largest for high BC injections. How realistic is this response, and could that somehow explain what is happening at the surface? Also, I see that the authors do principal component analysis of the DJF Tas response (to comment on links with AO modulation). I think it might be helpful to do the same for the JJA in year 1, my guess would be that the pattern obtained would mimic the response to BC simulated in JJA that year (this is just my guess, but maybe worth checking).

On a related note – Fig. 6 shows the map of changes in DLRF and net RF response. If one shows DLRF, I think it is important to also show DSRF (since as discussed above DSRF is much larger so it’s important to see whether that is also true locally over the region in question), as well as decompose the net RF budget calculation. The authors try and discuss parts of the latter in the text, but the reader can’t verify (or visualize) the accuracy of the statements without seeing the plots somewhere (even if it’s just in the supplementary material)

Minor changes:

-    L. 17-18 “reduced net long wave and turbulent fluxes”. Please state direction. Throughout most of the manuscript it is argued that the downward LW increases, but in the abstract you say the net LW actually decreases - I guess that’s because net is defined as upward, but it is confusing to the reader here.

-    Table 1 - the number of decimal places that are reported is too high, given all the uncertainties. 

-    Table 1 – “the confidence intervals for the response in years 6-15 are calculated taking into account autocorrelation”. Does it mean you use data for years 6,7,8,9,10,11,12,13,14 and 15 separately, instead of mean over years 6-15? If so, this is not clear, and given there is a strong trend present within that period, I’d expect the 95% confidence interval to be much larger in that case (as I would expect it to reflect that trend)  

-    Fig. 2 caption – please make sure you include description of all inserts, and with correct labels (some are missing, and some are mixed up I think)

-    Fig. 2 – “shading indicates standard error of the mean” – but the plot shows the difference in two means? (i.e. with and without BC injection)

-    Fig. 5 – what are orange lines?

-    Section 2 – while the aerosol microphysics scheme used is a bulk one, we are still missing some information of the assumed BC aerosol properties, including size.