REVIEW FOR: ‘INTERACTIONS BETWEEN TRADE-WIND CLOUDS AND LOCAL FORCINGS OVER THE GREAT BARRIER REEF: A CASE STUDY USING CONVECTION-PERMITTING SIMULATIONS’, ZHAO ET AL.

Firstly, I would like to declare up front that it’s Sonya Fiddes here doing this review, which, as you read my comments, will become clear why I have done this. This paper presents a nice exploration of how trade-wind clouds over the Great Barrier Reef are influenced by the local mountain range, sea surface temperatures and aerosol sources. It is also one of very few atmospheric modelling studies focused over the Great Barrier Reef. Its results show that the topographical setting is the most important contributor to shallow cloud and precipitation with local SSTs and aerosol having a lesser impact. As efforts continue to protect the Great Barrier Reef, a better knowledge of how clouds interact with their local environment is essential. For this review, I have a few main comments, and then further general comments on the results. I think if these can be addressed (which I’m certain they can be), then this paper is a valuable contribution to the literature.

SPECIFIC COMMENTS:

1. More information about the WRF model set-up and how these choices may be impacting your results would be beneficial, particularly with respect to the microphysical scheme and how the aerosols are allowed to interact and influence cloud formation/radiation. For example, is the microphysics scheme a double moment scheme (or single moment), and how are aerosol allowed to interact with cloud and radiation (I see that you have mentioned how the aerosol are activated which is part of the indirect effects, but can you be more explicit here? And what about direct effects?)? Why have you decided to use a scheme that only represents ‘water friendly’ and ‘ice friendly’ particles, and not a scheme that represents aerosol sources more comprehensively? It would be beneficial to the reader to gain a more comprehensive understanding of the model being used and its limitations.
2. As mentioned, there are very few modelling studies that have the Great Barrier Reefs as their focus, especially with respect to aerosol/cloud modelling. However, you have not mentioned two of the most recent papers on this very topic (see below). I acknowledge that I have some conflict in this given the second paper is my own, however, the similarities between the paper presented for review and this prior paper I think should speak for themselves (eg. use of the WRF model, testing impact of aerosol, etc.). I would request that the authors take some time to consider how the results of their work either supports or disagrees with these prior papers to show how they are advancing a relatively small body of knowledge.
   • Jackson RL, Woodhouse MT, Gabric AJ, Cropp RA, Swan HB, Deschaseaux ESM and Trounce H (2022) Modelling the influence of coral-reef-derived dimethylsulfide on the atmosphere of the Great Barrier Reef, Australia.  Mar. Sci.9:910423. doi: 10.3389/fmars.2022.910423
   • Fiddes, S. L., Woodhouse, M. T., Utembe, S., Schofield, R., Alexander, S. P., Alroe, J., Chambers, S. D., Chen, Z., Cravigan, L., Dunne, E., Humphries, R. S., Johnson, G., Keywood, M. D., Lane, T. P., Miljevic, B., Omori, Y., Protat, A., Ristovski, Z., Selleck, P., … Williams, A. G. (2022). The contribution of coral-reef-derived dimethyl sulfide to aerosol burden over the Great Barrier Reef: a modelling study. Atmospheric Chemistry and Physics, 22(4), 2419–2445. https://doi.org/10.5194/acp-22-2419-2022
3. I think this paper would benefit from some more quantitative analysis, e.g. some statistics. There are some number values presented but reporting values more consistently (including %) would be good to help the reader gain an idea of how much these experiments are actually changing the fields of interest. I know that presenting significance values for case studies is difficult, so it would also be good to include in some of your discussion how confident you are in these changes given the model’s capabilities/limitations. (e.g. can you be sure it is not just ‘noise’?).
4. I think your conclusions need to include a bit more of a rounded discussion rather than just a summary of the results. E.g. what limitations does your model have in how it can represent the processes you are exploring (eg. what resolution are your SSTs/orography – would higher res change your results? Does the very simplified aerosol representation impact your results?). Does this work support or disagree with prior work? What would these results suggest for future work? Does it have implications for the RRAP work that is funding this paper? Can you comment on other times of year, and other cloud regimes? Do you think the results would be similar in these instances?

TECHNICAL COMMENTS

Line 49: The Fiddes et al. (2021) paper actually shows that there is ‘no robust evidence that coral-reef-derived DMS influences global and regional climate’. I would suggest re-reading this paper also, as there may be some interesting differences or similarities with respect to cloud response to aerosol.

Line 60: Fiddes et al. (2022) did provide a higher temporal resolution study to Fiddes et al. (2021). Analysis of diurnal cycles can also be found here: https://minerva-access.unimelb.edu.au/items/f7b061c0-0bf8-5574-aaa2-e9f83f5fe854

Figure 3: Is there a way you can simplify this plot – it took me a moment to figure out what was going on. E.g. can you combine say a & b and c & d?

Line 146: Perhaps you can also introduce the up/down wind domains here as well to help make sense of Figure 3.

Line 160: What are ‘history intervals’?

Line 166: Can you describe microphysics scheme in more detail (see main comment 1).

Line 168: Do you mean all model grid points with cloud in them?

Line 169: Can you describe the aerosol climatology a bit more? What resolution is it based on, what model derived them? What aerosol does it include (ie. sea salt, sulfates?), how does it include anthropogenic emissions (Fiddes et al. 2022 showed these to be dominant for the GBR region). Do you know if it is realistic for the GBR? Models are becoming increasingly aerosol aware these days, so answering these questions is becoming more and more important, especially if you are considering aerosol-cloud interaction!

Line 170: What scheme are you referring to? The microphysics?

Line 173: I’ve never really heard of aerosol being categorised as ‘water friendly’ and ‘ice friendly’. Some further justification/explanation of this might be good, i.e. what types of aerosol do these categories actually represent?

Line 194: Can you describe the SST data sets more? What are you using for the control? Is it the HadISST data set? What have you used to create the climatology? Why did you choose a 1˚C perturbation and not more or less? Also with your supplementary Figure (S2), I think reversing plot c would make more sense as you mention that the climatology is cooler than the control, but that plot would make you think it was warmer if you didn’t read the caption properly (also can you describe plot c a bit more clearly in caption – it’s not so much an anomaly distribution as just an anomaly plot).

Line 208: It’s great that you are evaluating the model against obs, but unfortunately, none of the obs you are using to evaluate the model are independent, i.e I am fairly certain that all of the obs you have used would have been ingested into the re-analysis. This is certainly worth mentioning.

Line 211: I’m not sure that comparing a model that is being driven by ERA5 to ERA5 is worthwhile – a good sanity check yes, but not really necessary to include as a result. If there are BoM pressure/wind obs, that would be better, but that faces the same problem as the other obs in that they are ingested into ERA5, so not really independent either.

Line 250: I think you mean Figure 6?

Line 259: I think you mean green lines?

Line 258: I think if you want to show the correlation of precip to altitude, a better plot might be a scatter plot with elevation on y-axis, and precip on x-axis? You could colour it by east/west aspect or something like that to take into account up/down wind orientation.

Figure 5: Can you expand the figure caption? Eg. what are the dotted/solid lines?

Line 284: Can you please explain what you mean here? ‘In this study, CF is defined as the proportion of total grid points in the target domain that are classified as cloudy grids, denoted by the binary number 1, for each model level ‘. Do you mean that the grid point has to have 100% CF to be included?

Figure 9: I know it’s often a personal preference, but I always get confused when precipitation is plotted so that less precip is blue and more is red, I intuitively think of it the other way around.

Line 380: I would say the spike isn’t that ‘notable’ – it would be good to mention that it is only in WFAx5, and it is preceded by a gradual ramp up? Why would a change in wind direction change the aerosol properties?

Line 388: Same results are found (though not shown) in Fiddes et al. 2022 (see last paragraph of results section). You can find plots for this here: https://minerva-access.unimelb.edu.au/items/f7b061c0-0bf8-5574-aaa2-e9f83f5fe854

Line 392: The increase in LWP is really only occurring on a few occasions and not uniformly – can you comment on this?

Line 395: Due to cloud lifetime effects? Does your model support these indirect effects?

Line 414: Can you provide some stats to support this ‘respond strongly’ statement? It would be useful to have a quantitative idea of this.

Review for the manuscript “Interactions between trade-wind clouds and local forcings over the Great Barrier Reef: A case study using convection-permitting simulations” by Zhao et al.

This study examines how trade-wind clouds and precipitation over the Great Barrier Reef (GBR) from a particular case study respond to three local forcings: orographic lifting, aerosol concentration, and sea surface temperature (SST). The authors employed the Weather Research and Forecasting (WRF) model, with three-level nested domains. The finest horizontal resolution in the smallest domain is 1 km. The simulations were driven with large-scale forcings from ERA5. During the period of this case study, April 2016, trade-wind cumulus was present northeast of Queensland, Australia, upwind of the mountains. The author performed sensitivity studies to three local forcings: the orography, aerosols, and sea surface temperatures. They compared how the cloud fraction and precipitation in the upwind and downwind regions of the mountain ranges respond to these three factors and discussed relevant explanations to the changes they found.

I really enjoyed reading about the orographic effects on trade-wind clouds over the GBR. The authors clearly demonstrated how the orography drives low-level flows which then affect the cloud-top boundary layer. However, for the aerosol and SST sensitivity tests, I find that some analysis and discussion are still missing - a reason for why I think this manuscript needs some major revisions. I believe that after these changes, especially those mentioned in major comments #2 and #3, this manuscript will be a valuable contribution to the literature on shallow cumulus clouds.

Major comments:

1. Simulation setups: More information on the simulation setups is necessary. Are the simulations nudged to ERA5 large-scale forcings, and if so, what is the time scale? What is the vertical resolution in the smallest domain of your simulations? And with the finest resolution, 1 km, this is still quite coarse for marine shallow cumulus and stratocumulus especially if you are using the convection-permitting mode / turning off the cloud parameterization. I recommend you discuss other modeling studies of marine shallow clouds that compare their results when using coarse (~O(1 km) ) vs. finer (~O(100 m) ) resolutions. Based on their findings, that may help justify your choice of resolution and the robustness of your model output. You could look at these studies and the reference therein for example: Saffin et al. (2023, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022MS003295). It should also be noted that in their studies, Saffin et al. (2023) found that a longer spin-up time helps them reproduce a better agreement with the observations.
2. Aerosols: Your sensitivity studies showing how the shallow clouds over the GBR respond to the aerosols emitted off the coast. But for a complete story, I find that some discussion is still missing. For example, why do you consider “water friendly” and “ice friendly” aerosol categories? Are you assuming all aerosols are hygroscopic? Are your aerosols size-dependent? What is the PDF of the aerosol size distribution, and does it change uniformly when you increase the aerosol concentrations? The sizes of the aerosols will play an important role when you consider the cloud fraction and precipitation changes. For example, if you increase finer particles, you may have more clouds but not rain, but if you increase coarser particles, you may have more rain but not clouds. Therefore, more information on the aerosol type and size distribution is needed for you to justify your aerosol sensitivity tests. For references, see Hoffmann and Feingold (2021, https://doi.org/10.1175/JAS-D-21-0077.1) for example.
   
   
   
   Furthermore, aerosols do not only interact with clouds via microphysical processes but also change the turbulence through radiation. The latter has been shown to affect marine shallow cumulus clouds in a recent study, i.e., the second half of Narenpitak et al. (2023, https://doi.org/10.1029/2022MS003228).
3. Local SST: Typically, reduced SST leads to more shallow clouds or higher cloud fraction. There are some studies that have found otherwise, just like in your simulations. In this section, I recommend you add more analysis as to why the clouds in your simulations reduce with lower SST. 
   
   
   
   In general, warmer SST increases surface latent heat flux, sensible heat flux, and moisture flux (consistent with what you showed in Fig. 14). More is needed from there.
   
   
   
   (1) How does the inversion strength change? If the TWI is weakened, your clouds will deepen, and more dry-air entrainment will dry out your clouds with warmer SST. But if the TWI is stronger, this may prevent entrainment of dry air from the free troposphere, and this may still help retain the clouds. Precipitation also affects these processes further.
   
   
   
   (2) How does the temperature lapse rate change when the SST is changed in your simulations? More stable lapse rates may prevent formation of shallow clouds.
   
   
   
   (3) Does the humidity in the free troposphere in your simulations change? If the free troposphere is drier and the TWI is weaker, more dry air can be entrained into the clouds, depleting the cloud liquid water with warmer SST.
   
   
   
   For more processes linking changes in SST to cloud amount and precipitation, you can read Bretherton and Blossey (2014, http://dx.doi.org/10.1002/2013MS000250) and Vial et al. (2017, https://doi.org/10.1007/s10712-017-9418-2). A study that finds warmer SST increasing cloud fraction (similar to your simulations) is Narenpitak and Bretherton (2019, http://dx.doi.org/10.1029/2018MS001572); their reasoning might also be helpful in your analysis and discussion.
4. Inversion strength: It is great you are computing the trade wind inversion (TWI) and the inversion strength derived from the inversion base and top from the TWI. However, there are other indices to measure the inversion strength, which are commonly used in the literature of marine shallow clouds. Those are the “estimated inversion strength” (EIS, Wood and Bretherton, 2006, https://doi.org/10.1175/JCLI3988.1) and the “estimated cloud-top entrainment index” (ECTEI, Kawai et al., 2017, https://doi.org/10.1175/JCLI-D-16-0825.1). I suggest adding a calculation of either one of these indices, so the results from your study can be easily compared with other studies in the shallow cloud literature.
   
   
   
   The EIS is also a common index used in the literature when discussing about the SST’s effect on shallow cumulus and stratocumulus clouds. It might be helpful for other readers to compare your manuscript with the others if you decide to use it, but this change is not necessary.

Minor comments:  

Line 106: “Figure S2” – should be “Figure S1” since it is first mentioned.

Figure 2: I suggest using a different colormap that is not diverging, since you are showing something that ranges from 0 to 100. Having blue color representing zero precipitation is counterintuitive to me. It’s also hard to see the black topography contours on the blue (zero precipitation) color!

Line 144: Does this mean the vertical resolution is roughly 100 m? It would be nice if the vertical resolution can be shown along with the vertical profiles in Fig. 4 or in the supplementary information.

Section 3.1: In general, do you nudge the simulations to ERA5? Or do you simply initialize the simulations and let them run? Please specify.

Lines 168-169: What is the “auxiliary aerosol climatology” and what is the aerosol concentration of the “multiyear (2001-2007) global model simulations”? What are the aerosol size distributions of the aerosols prescribed in your simulations? The latter is important for interpreting the aerosol sensitivity study.

Lines 173-175: What exactly are the “water friendly” and “ice friendly” aerosol scheme? As in, what do the “hygroscopic aerosols” and the “nonhygroscopic ice nucleating aerosols” do in the cloud microphysics / aerosol scheme? More information about the process, rather than the “water friendly” and “ice friendly” description is needed. Also, how will your results change if you use a different aerosol scheme that has different assumptions about the aerosol types?

Line 188: Figure S1 should be Figure S2 (swapped with the other figure in line 106, based on the order they are first introduced.

Line 193: Maybe show the result of a 500 m threshold in the supplementary information? It would be interesting to see.

Lines 195-196: I was confused when I first read the descriptions of the simulation names. The CTRL simulation has the warmest SST, followed by the SST-climatology simulation, and then the SST-cool is the coolest. Is that right? If so, that is worth mentioning here too since you show the SSTs only in the supplementary information. Additionally, it’s good to note that between CTRL and SST-climatology, the temperature different is not spatially uniform and it is cooler in CTRL between 18-19S, 149-151E. Is this where the clouds are observed or upstream of the air trajectory flowing to the shallow cumulus clouds?

Lines 199-202: Why do you only consider the “water-friendly” / hygroscopic aerosols? Aerosols that do not get activated as cloud condensation nuclei can also affect shallow cumulus cloud processes by changing the radiative heating profiles, which then alters the stability of the cloud layer and cloud fraction. This is different from the Twomey effect and the Albrecht effect, which you discussed earlier. See major comment #2 for details.

Line 243 and Fig. 5b: Based on the sounding profiles, the simulated temperature profile does not really show a temperature inversion layer. In a case like this, how do you compute the TWI?

Line 250: Do you mean Figure 6?

Line 258: Do you mean green contours, not black? Honestly the green contours are hard to see with the black background. I suggest trying a different colormap for the brightness temperature (maybe white-blue, rather than white-black) and use black contours instead of green.

Lines 262-264: There is a bias / discrepancy in observed and simulated precipitation, south of 17.5S, 146E, where the mountains are only 250 m tall. Maybe it’s worth discussing about this discrepancy, especially how this may or may not affect your sensitivity test.

Figure 4: There is also a disagreement in the wind speed east of Townsville on 30 April 2016 (bottom row). Might this be a cause for the precipitation discrepancy?

Lines 280-283: It would be good to see whether the subdomains (red boxes) in Fig. 3c and 3d overlap with each other. Please consider combining Fig. 3c and 3d, and show both of your subdomains in the same plot.

Line 286: Since you already compute mid and low cloud fraction, it will be helpful to show a spatial map of mid- and low-cloud fraction, similar to Fig. 9, but for both the upwind and downwind subdomains. It will be helpful to see where the clouds are before seeing the precipitation. This will also help your discussion (Lines 300-304) and Fig. 7-8. Right now it’s rather hard to follow that discussion.

Figures 7-8: Firstly, it’s counterintuitive to have red colors represent “more clouds” and blue “less clouds”. Please consider reversing your colormap. (The same goes for Fig. 9.) Secondly, it might be helpful if you can show the average height of the TWI base along with the PDF of the cloud fraction. This will help the audience see if the cloud changes are, at all, connected with the TWI, which is an important constraint for shallow clouds.

Figure 9: Is a colorbar for CTRL missing?

Line 330: I suggest introducing CAPE, w_diff, and 10-m wind convergence in the order you show in Figure 10.

Lines 341-344: This is a nice discussion. I’m curious to see the surface temperature differences between the CTRL and TOPO300 runs. A comparison figure could be added to the supplementary information. This will build up a nice discussion for your SST sensitivity part too.

Line 356: There’s an extra “that” in your sentence.

Lines 451-452: I suggest including an analysis of the TWI or EIS, and other relevant factors that might affect the cloud fraction when you perturb the SST. See the main comment for details.

Line 522 / Conclusions: Other forcings that might be relevant: There are other local forcings that affect cloud and precipitation in trade-wind cumuli, such as wind shear and surface wind speed. The mesoscale organization of these clouds may also matter. Will you be looking at other relevant forcings? Or why not?