The manuscript “Sensitivity analysis of an aerosol aware microphysics scheme in WRF during case studies of fog in Namibia” investigates how droplet activation of cloud condensation nuclei (CCN) affects simulated evolution of fogs. The analysis is done for two observed fog cases which are simulated using different assumptions for initial CCN concentrations affect the evolution of fog properties. The paper presents a thorough investigation of the topic and is within the scope of Atmospheric Chemistry and Physics. The main issue with the paper is that it is not clear to me, what is the scientific value of the study and in order to be published, this shoud be clarified.

In addition, the following points should be addressed:

Major comments:

- In many cases, the model setups and their effects on results are explained too ambiguously to be understandable for the reader. I was not able to understand properly the description of the setup for initial CCN concentrations. Section 3.3.2 discusses the vertical profiles of CCN and refers to the WRF user guide for an explanation. However, it is still unclear to me how this initial vertical distribution results in such a large difference in CCN over land and over ocean for Case CCN_300. Is the initial column intergral of CCN concentrations (CCN burden) significantly different depending on the terrain? Is the CCN concentration initialized in the beginning of the spin up or at the beginning of the actual simulation?

- Page 15, Line 310 says that “The initial CCN concentration for scenario CCN_300_landsea shows a clear contrast, with lower concentration over the ocean than the land (Fig. 10c). The lower concentration over the ocean counteracts the accumulation of CCN over time, as seen in CCN_300, resulting in a more balanced mean CCN concentration between land and ocean (Fig. 10d).” Is the accumulation of CCN in CCN_300 only because there are more CCN than in CCN_landsea? In what way there is a more balanced mean CCN concentration between land and ocean? The land-sea contrast at the south boundary seems quite high also in CCN_300_landsea.

The droplet activation parameterization is shown in Figure 2 to be sensitive to CCN concentration and updraft velocity. However, fog formation is also affected by non-adiabatic cooling. Poku et al., (2021) have suggested that instead of using simulated updrafts, it would be better to calculate the change in saturation due to non-adiabatic processes. In the current paper, only the effect of changing the minimum updraft speed was tested. Wouldn’t it have been fairly straight forward to for example convert the cooling rates to corresponding updraft speeds to have a more physical representation of fog droplet activation? On Page 21, Line 392 it is said “Their proposed work around is to include cooling tendency as proxy for updraft speed and then assigning a speed that will activate the appropriate number of droplets. This may come with a new set of problems in terms of early activation but this remains to be seen.” To me this seems a very good approach and if there is a new set of problems, would that point to problems in other physical processes of the model and in itself is not a good justification for not using this approach?

Minor comments:

Page 6, Figure 2: Is the activation sensitivity the activated fraction of CCN?

Page 11, Line 266: “Therefore, assigning a minimum updraft speed of 0.1 m s-1 can be a reasonable assumption, as it falls within the median of activation at the site 0.56” Did you compare the distributions of activated fractions for different minimum updrafts?

Page 15, Line 307: “Furthermore, the boundary conditions for scenario CCN_300 had relatively lower concentrations of CCN.” Lower concentrations compared to what? Why are they lower?

Page 21, Line 391: “In addition, the threshold updraft speed is often higher than the 0.01 ms-1 used in the T14 scheme, which effectively results in a higher super saturation and excess droplet activation than would be expected for a fog event.” Please add references to such studies / approaches.

The motivation for showing Figures 13-16 is not clear to me.

Techical comments:

Fonts in figures are extremely small.

This paper presents an analysis of two well-observed fog cases over Namibia with multiple versions of an aerosol-aware microphysics parametrization in WRF. The result show some interesting issues with the microphysical parametrization and its representation of fog, which are certainly worth reporting, although the manuscript could be clearer in explaining what these issues are and identifying possible further work to address them. I suggest the paper could be suitable for publication with some revision to improve this aspect.

Major points:

The authors appear to view the minimum updraft speed used for CCN activation as a tuning parameter. It is not. Whilst it is fine to adjust this parameter as part of a sensitivity analysis, the authors need to be clearer on the reasons for doing this, i.e. it is highlighting deficiencies elsewhere in the model. If insufficient activation is achieved for the physically-based default setup (obtained from parcel model analysis), then one of 2 things must be happening:

• The model updrafts themselves are underestimated. This point is not mentioned at all in the paper, and should be. Whilst I suspect (as usually happens in fog), the model updrafts are correctly small, it would be worth discussing - especially if you have observations of the near surface vertical velocity variance available.
• The updraft is not the process causing the aerosol activation. Whilst this point is mentioned briefly in the paper, it needs to be made clearer, and could be further discussed, e.g. what cooling rate does the change in minimum updraft velocity they try imply, and is this realistic?

Some specific (but not exhaustive) examples of text that needs addressing in this regard:

• L260-267 - this could be clearer in explaining the motivation, i.e. you're picking the minimum updraft to achieve the observed activation, but this doesn't imply that the updraft is the reason for the activation, i.e. it achieves the right answer but for the wrong reason
• L450-457 - again, need to be clearer here that the right answer is being achieved for the wrong reasons, and add some discussion on what the right way to achieve the desired result could be

Minor points:

• L14 - I'd say "is used" rather than "is parametrized".
• L73 - spelling of "Boutle"
• L112 - is 34m really low enough for the lowest model level in fog? Some more discussion on this might be useful - did you do any sensitivity studies to this? It implies that the fog must be at least 34m deep before it is present in the model, which could have significant effects on its early development. I'd suggest the authors look at https://acp.copernicus.org/articles/22/319/2022/ to see what effect the lowest level height can have on model development, e.g. the FV3 results with a lowest level at 21m are quite poor.
• Fig 3, 6 etc - it's usually helpful to plot visibility on a logarithmic axis, due to its highly nonlinear nature, to better show the low visibility events.
• L300-315 - I don't really understand here how the CCN is evolved in time in the different experiments, so it would be useful to explain this further. My assumption was that it was a prognostic variable, advected by the flow and processed by the physical parametrizations? But the text seems to suggest that it is somehow diagnosed from the boundary layer depth over land - why? And why is this not applied in the simulations where the CCN is initialised from a climatology - What happens to the CCN when the BL depth adjustment is not used? If it's important enough to discuss, why not process the CCN in the same way in all experiments (with or without this BL depth adjustment) for consistency? I think this is just complicating the results for no good reason, and would be better to be consistent.
• Fig 10 - would be helpful to use the same scale for panels a-d, rather than varying the left and right columns
• Fig 10 - it would be worth discussing somewhere why there is such a large discrepancy between the simple initialisations and the analysis - are the simple setups just really bad for this area of the world, so the analysis is the correct thing to use, or how much do we trust the analysis for fog initialisation?