Review of “Long range transport of course mineral dust: an evaluation of the Met Office Unified Model against aircraft observations” by Ratcliffe et al. for publication in Atmospheric Chemistry and Physics

The paper presents data on observed dust particle size distribution vertical profiles from three airborne campaigns: Fennec, AER-D, and SALTRACE. The data present evidence for the existence of super-coarse and giant sized dust particles from from their origins over the Sahara desert. The data from these different campaigns is synthesized to be evaluated in context of climate model simulations performed with the HadGEM3 model. The model is shown to have systematic biases with respect to how it apportions coarse and fine mode dust, emphasizing too much the fine mode dust at the expense of coarse particles. This is a result of model tuning (mentioned, not shown) that itself corrects a bias toward the model's excessive loss of coarse dust particles compared to observations. In short, the model is tuned to satellite AOD. A statistical analysis shows the positive correlation relationship between dust AOD and dust concentration is robust and that there is no strong evidence of a relationship between particle size and dust AOD (in other words, the dust particle size is similar regardless of loading).

The author’s are writing on a timely topic and using the best available airborne observations. The modeling methodology needs further discussion and I suggest a major revision only because I’m going to ask them to do some additional calculations, but I don’t think they should be too burdensome:

Further detail is needed in the paper to describe aspects of the modeling. First, what is the initial size distribution for dust emission assumed? Then, what is the effective radius of the dust size bins that is used in the settling calculation? These could be added straightforwardly to Table 5. But further I suggest the analysis will be improved if a fundamental model bias is corrected at the start. Specifically, as Figure 4a shows the model doesn’t even agree with the size distribution over the source region. I think this is consistent with Figure 3b in the Woodward et al. (2022) paper referenced in this paper. It does not seem as though the model would under any circumstances be capable of reproducing the observed size distributions because at the source the particles are biased toward smaller sizes. It is not clear to me from reading the Woodward paper how the emitted size distribution is calculated; the paper seems to suggest it is related to the horizontal dust flux at the surface, but I don’t think that’s right (or indeed what is shown).

My request then is that here the authors make a version of Figure 4 in which they rescale the modeled dust according to the initial dust particle size distribution observed in the Fennec measurements over the Sahara. At least near the surface. Applying those same scaling factors it should be possible to make a version of Figure 4 that looks in some way more like the observed dust, but still makes the point that the coarse mode particles fall out too quickly. It would demonstrate that the model itself has this deficiency and it is not simply a function of incorrect assumptions for the initial particle size distribution. The model would need to be retuned if such were to be used more routinely, but since in this paper everything is scaled and normalized to facilitate the comparisons it is just one more example of that. I recommend this would be clarifying for the literature on this topic.

Line 37: Haven’t yet identified why the simulation in models is “challenging.” Suggest you remove this phrase here.

Line 55: See Nowottnick et al. 2010 for an example of this that was found to require improvement of wet scavenging processes.

Nowottnick, E., P. Colarco, R. Ferrare, G. Chen, S. Ismail, B. Anderson, and E. Browell (2010), Online simulations of mineral dust aerosol distributions: Comparisons to NAMMA observations and sensitivity to dust emission parameterization, J. Geophys. Res., 115, D03202, doi:10.1029/2009JD012692.

Line 61 - 74: There is no mention here about how the processes are represented in the models, i.e., the numerics of the settling calculation, for example. See for example Ginoux (2003) who shows significant difference between simple upwind advection scheme for calculating particle vertical transport versus a higher order scheme (his Figure 4b). This has to matter to solving this problem. What kind of scheme is used in your model?

Ginoux, P., Effects of nonsphericity on mineral dust modeling, J. Geophys. Res., 108 (D2), 4052, doi:10.1029/2002JD002516, 2003.

Line 125: You mention sizes up to 300 um here (and on line 142) but I don’t see that in Table 3. What do you mean here? Do you have any quantitative information about such large sizes?

Table 5: Could you provide some additional details of the model here? For example, how total vertical flux is partitioned into each size class (i.e., could your model potentially *ever* match the size distribution of observed dust shown in Figure 4), and also what the radius used for the settling velocity calculation is (and/or what is the settling velocity for each bin at, say, 1-2 km altitude).

Line 266: I’m don’t understand what you are saying differently in this paragraph than in the following one (line 274). Here you explicitly refer to spatial distribution, but the result is presented in terms of mean AOD differences. You say those are comparable between the campaign and long-term averages except for the Canaries. You make these same points in the next paragraph, although there you also say SALTRACE observed dustier than typical conditions. How do you assess comparable spatial patterns? And what is it is really you are seeing? Dustier than average conditions, or typical conditions? This could be clarified.

Line 294: The stated null hypothesis reads differently than what is written in caption 6. The statement in the caption reads more sensibly than what is written here. Please adjust. You conclude statistically significant differences in all cases, although the Canaries case shows no significant difference when M and H cases are compared; why is that?

Line 305: You mention Figure 3 before Figure 2. Suggest you just reorder the figures.

Lin 325: The appears to be an inconsistency here between the stated mean mass concentrations and the mean plotted on Figure 3. For example, the stated mean concentration over the Sahara is 347 ug m-3, but Figure 3a shows values in excess of 500 ug m-3 from the surface to 5 km. Similar for other values. What is going on? Is the model calculation consistent?

The paper presents relevant data from various observation campaigns on dust particle size distribution vertical profiles, which are synthesized and analyzed to verify and contrast how can be crucial for the radiative aspects of climate models. The information and analysis from the campaigns are significant for furthering our understanding of Saharan dust transport, and the comparison with the model is particularly interesting for identifying uncertainties and limitations.

The manuscript is well-written, well-organized, and the presented analysis is accurate. I would like to propose a few questions to the authors that I believe are pertinent to the work conducted, as well as some minor considerations to enhance the reader's understanding of the paper.

Questions:

• What type of emission data is input into the model, and in what format, spatial resolution, temporal resolution, etc.? One key reason for the model's discrepancies regarding size distribution could be improper emission inputs.
• Given that the measurement campaigns are conducted on specific days and hours, have the results of these specific campaigns been compared with the model outputs for those concrete days? Perhaps the unaveraged model output is less representative but could provide clues to identify some of its limitations. Along these lines, can CTM models, which can be used with hourly resolution, be compared similarly to the climate model?
• What is the vertical resolution used? The number of layers is mentioned, but I think it's relevant to include this to understand how the model outputs have been averaged. Are these models sensitive to increasing vertical resolution to improve results?
• The discrepancy between the model and observations is a recurring issue in the manuscript; it might be beneficial to summarize this in one or two paragraphs towards the end, if the authors agree.
• The final paragraph of the conclusions feels somewhat weak and lacks impact. Given the importance of this work, it would be advisable to strengthen this last paragraph.

Recommendations to facilitate understanding of the paper:

• Do SABL and SAL mean the same thing?
• In line 388, "too low" is mentioned; I recommend providing height values as a reference for what is considered "too low."
• Figure 3: I understand that the scales are correct due to the significant difference in mass concentration between the obsservations and the model; nonetheless, it would be helpful to include a note or warning regarding the scales in the figure.
• Figure 5: It might be useful to mark the start of the distributions in Figure (a), since (b) and (c) start at 0° and (a) starts at 10°E. Alternatively, align (b) and (c) with the scale of (a) for better visual consistency.
• The values in Table 8 are mainly referenced as sums in the following paragraph; consider indicating this either in the paragraph or in the table itself.