The manuscript, “Aerosol characteristics and polarimetric signatures for a deep convective storm over north-western part of Europe – modeling and observations,” details the development and evaluation of an extended modeling system. This enables high-resolution modeling of aerosol-cloud interactions in the Terrestrial Systems Modeling Platform (TSMP), after extension with a chemical transport model and the polarimetric radar forward operator. The TSMP was evaluated at convection-permitting horizonal resolution against observations of a thunderstorm which took place in July 2015 over northwestern Germany. Overall, the extension of the model with the chemical transport model and polarimetric radar forward operator adds valuable capability to model predictions.

Beyond this, however, this reviewer found it difficult to pull out what the key conclusions of this study actually were, and what benefits the extended TSMP platform provided. The manuscript suffers from too much unnecessary detail, and sometimes repeated itself, and not enough necessary detail. Not only did the lack of clarity and explanation make it difficult to identify the key findings of the study, it also made it difficult to evaluate the claims and evidence that were more clearly presented.

For example, none of the polarimetry aspects of the study were clearly explained; prior knowledge of both polarimetric radar and how they are interfaced to/emulated by models seems to be necessary. Even more specifically as an example, polarimetric radar measurements such as horizonal reflectivity, differential reflectivity, specific differential phase, and the cross-correlation coefficient are introduced and the TSMP’s performance evaluated for these parameters, but the manuscript does not explain to which meteorological or chemical parameters they relate, what these parameters tell us about the atmosphere and the model’s performance. This reviewer is familiar with regional and high-resolution modeling and aerosol-chemistry-convection interactions but not polarimetric radar, and assumes that this will be the case for at least some readers, so that much of the manuscript dealing with the polarimetric model implementation (even how it’s interfaced with the model is unclear) and the related results are a bit shrouded in mystery. As a somewhat less critical example, the manuscript does not define what it considers to be northwestern Germany for the purpose of the study, and later on adds in mention of two additional German cities/radar sites, with insufficient location information for the deep convective event itself; a map may have been useful. Methods and results are not clearly explained – how are the column NO2 observations from satellite swaths and AERONET point observations compared to a model grid box, as another example.  Why was this study undertaken in the first place is also left unclear (though such a modeling platform does have great potential).

These examples are meant to be illustrative, as this problem is ubiquitous in the manuscript, and greatly impedes identification, interpretation, and evaluation of results.

Two Specific Comments:

Line 111: It is not really accurate to say that AERONET is considered to have a better accuracy than MODIS – one instrument is essentially a point-based observation and so is often likely more accurate for that specific location than a satellite, while the other observes over a larger swath and so loses some of the horizontal detail but can fill in the gaps between AERONET instruments. And both require a retrieval to turn the raw observations into useful measurements of aerosol properties, which introduces its own uncertainties. Which instrument is “better” for a study depends on what you’re trying to compare or investigate. It's more accurate to say that MODIS and AERONET are complementary, and each has its advantages and disadvantages.

Line 245: Comparison of satellite-based column observations to the NO2 columns computed from the model output is more than simply a first-order evaluation, but this again depends to an extent on what exactly you want to evaluate and be able to say with your study. Column quantities are quite useful for many applications, and again, the satellite-based column observations fill in gaps between ground-based or other types of in situ instruments and can, for example, provide information on chemical transport and transformation. The brief evaluation of the NO2 column comparisons already demonstrate a number of potential results and areas for model refinement that could be classified as more than first-order. Also, why were other trace gases not evaluated, such as O3 or HCHO which are available from satellite data products? And equally importantly, the satellite products come with specified uncertainties and data flags, so it is not enough to only acknowledge the uncertainty in the satellite estimates – the uncertainty in the comparison can therefore be quantified and would facilitate evaluation of the comparison.  

Review of Shrestha et al, 

In this study a convective situation is simulated with an ensemble of TSMP-ART runs over the Bonn region of Germany. The simulations are validated using the Bonn Polarimetric radar, aeronet and modis observations. In general, the ensemble runs are able to simulate the convective cell structure reasonably well, while the aerosol concentrations and precipitation are underestimated. When the model runs are run through a Forward Operator, difference between the simulated and observed cloud structure are presented. The paper is extremely well written but the story line and figures are very complex. I recommend the article for publication in ACP once the comments below are addressed.

General comments:

The paper is quite technical with lots of details (e.g. discussion about specific features in the plots), which often distract from the overall research questions and ending conclusions.

The main research questions are not really answered and it comes off more of as, “we have these observations, let’s see how our more complex model is doing” rather than geared to address a particular research question e.g. does using prognostic aerosols improve the reproducibility of the storm, does constraining the cloud droplet distribution improve storm signatures, does a Forward Operator allow us to understand model deficiencies more easily? Especially as research question 2) is never really answered, as it is not clear if the “capabilities” lead to an improvement to simulations without these additions. There is no doubt that getting these components of the model to run together is a complex and cumbersome process, but without showing that this is an improvement or how the Forward Operator polarimetric variables improve the assessment of aerosol-cloud-precipitation interactions, the paper is lacking a clear direction.

One of the main findings from the aerosol module is that the convective updrafts produce aerosol towers and that the properties of the aerosols change after cloud processing. However, these themes are not carried through the manuscript. There is also very little discussion on how the aerosol cloud processing is parameterized or compared to previous studies. Also, it is not immediately clear if these aerosols will remain interstitial when exposed to such high updrafts. If they are activated as cloud droplets/nucleated into ice crystals, will they be scavenged and removed?

Are the polarimetric variables actually helping point to why the model is not matching the observations? There is a lot of discussion on how these variables differ but not to what questions they can answer. At the same time, as the Forward Operator introduces many uncertainties in itself as mentioned in the discussion, is it worth including it as a diagnostic tool? Consider building on the advantages of including this in the modelling process/ post processing.

At the same time, one of the main hypothesis for the discrepancy between observed and simulated polarimetric variables is the cloud droplet size distribution. Even though this explicitly tested, the discussion on this topic is not well connected. Even when running this sensitivity study, the polarimetric variables do not improve significantly. Does this really mean it is necessary to run simulations with prognostic aerosol as stated in the conclusion? Rather is it better to further adjust the cloud droplet size distribution?

Many of the figures are poorly labeled and described in the captions as noted below. This makes the discussion difficult to follow. 

Minor comments:

Line 60: remove “s” from lightning.

Line 63: would be nice to have a domain of the modelling in the manuscript, rather than just a reference to a previous study.

Line 137-149: As this study looks at the aerosol transport in convective updrafts, the description on aerosol activation (both CCN and INP), subsequent removal via precipitation and aerosol scavenging should be described. There is no doubt that updrafts help to lift boundary layer aerosol into the free troposphere but the fraction of these aerosol, especially large ones that do not act as CCN/INP should be discussed.

Line 163: “the” -> “that”

Line 179: remove “this”

Figure 1: please add a and b in the caption

Figure 3: In panel (a) add units for number conc. Panel (c), is that the 0 degree isotherm? Why is the center latitude for the cross section not at the center of the box? Legend for altitude lines are very hard to see. Also, is crust meant to be dust? Make it clearer that the same colored lines are from ensemble runs, especially in the PM2.5 vertical distribution in panel (d).

Line 267: Is the storm/ air flow moving to the north east? So the wind would be southwesterly or in a northeasterly direction? If this is not the case, would be nice to have sort of meteorological overview e.g. wind barbs to understand the direction of the flow.

Line 268: Why However? Are the updrafts not also helping to distribute the aerosol. Additionally, it looks like the role of updrafts is far more important at heights above 2km. That’s probably not important, but it begs the questions as to why the 2 km height is chosen for this analysis (later in the text the 6 km level is discussed).

Figure 4: the legend should reference Figure 3 and see comments for Fig. 3. CRUST seems cutoff in panel (b).

Line 269: “to determined” add “be”

Line 287: Please define “ABL”, is this aerosol boundary layer?

Line 288-290: But on the previous day, the peak occurred earlier and lasted well into the early morning hours. It is not evident that this statement is supported from this two-day simulation period. At the same time, it is not really clear why this is important.

Line 295: should it be PM2.5 mass (concentrations) or is PM2.5 concentration always reported in mass?

Figure 10: panels a and b the colorfill units are not immediately clear.  Why are there so few aerosols in the vertical cross section at ranges less than 10 km in panel f at the 6 km height while in panel e, the concentration at 6 km is close to 2000 cm-3?

Figure 11: Please label the red gamma distribution as the CDSD. Also, it is hard to easily compare the CFAD ensembles with the CDSD and lognormal droplet size. Consider making a difference plot of the CFADS.

Line 441-442: it would be nice to have a discussion about how changing the cloud droplet size distribution assumptions had on this here.

Line 445: second Poll ref is missing a year.

Line 458-460: Again here, did things improve when a different cloud droplet size distribution was assigned? This is touched on later in the section but the discussion could be shortened to combine the influence of the CDSD runs.

Line 484-486: There is no doubt that the strong updrafts help to loft aerosols to higher levels. However, with such strong updrafts (e.g. > 10 m/s), is it realistic that the aerosol are still interstitial and not activated as cloud droplets?

Line 492: Consider rephrasing to: “…can also not be neglected”