SUMMARY

This manuscript describes the modifications to the DESCAM bin microphysics scheme whereby the predicted rime fraction for ice is added.  The authors then conducted high-resolution real-case mesoscale model simulations, using the original and modified scheme, for a heavy snowfall event that was well observed during the ICE-POP 2018 field campaign in South Korea.  The authors conclude that the modified scheme improves the simulated precipitation and other fields compared to observations.  This is a potentially valuable paper in a few ways.  First, although the introduction of prognostic rime mass is not novel in the microphysics modeling community, it is new for DESCAM and it represents an important development to that model.  Second, the use of microphysical data from field campaigns such as ICE-POP 2018 is very useful and interesting for examining detailed microphysics schemes.  In that and other regards, the manuscript is nice in lots of way, however, it has some major shortcomings (described below) that must be addressed in order for this to be considered for publication.

Given that I am recommending that a DESCAM be further developed (with explicit melting) in order to properly study the effects of predicted rime mass and that idealized tests/demonstrations be added, then to be followed by a modifications to the real-case simulation examination section, one possible path forward would be to re-cast this as a two-part paper:  1) description of new developments + idealized tests; 2) Real-case simulation of the ICE-POP case.  It may simply be too long as a single paper and I believe that the additions I am recommending are important.  I will leave that to the authors to decide, but the comments below must be addressed.  With that, I will recommend major revision.

SPECIFIC COMMENTS

1. The implementation of predicted rime mass in the DESCAM scheme is a major development. The authors go from describing the new method to attempting to illustrate the impacts through a full real case 3D simulation. This is a big leap.  Understanding and evaluating the changes to a microphysics scheme is complicated enough; the authors have gone directly to the most challenging approach.  Microphysical pathways in a 3D model are very complicated and evaluation based on comparison to observations is inherently challenging.  For a major development of the type presented in this study, the authors should really start by illustrating the behaviour of the modified scheme in a very simple context, such as a 0D or 1D model framework, in order to provide the reader a basic understanding of how the new scheme works and what it does, as well as to illustrate that the changes do indeed do what they are supposed to do.  I strongly recommend adding a section on idealized tests and demonstrations, even if it means reducing the amount that is presented for the 3D case.

2. The setup of sensitivity experiment is backwards. Normally in a control experiment one takes a baseline configuration, which is the control, and then sets out to conduct one or more sensitivity tests to examine the impact of one or more changes. In this case, the most logical setup would be to prescribe the control (or control simulation), which in this case would be the simulation with the unmodified DESCAM scheme, the control configuration, and define the experiment simulation to be the one with the modified code.  Following from point 1, the one could define the CTR (baseline DESCAM) and MOD (DESCAM with rime fraction) configurations and apply them to the idealized simulations/demos and then to the ICEPOP case simulations.

3. When one uses a real-case simulation and comparison to observations to illustrate the benefits (or any kind of impact) of a particular set of changes to a model, one first needs to demonstrate that the control (unmodified) simulation is sufficiently realistic that one may proceed to use the modeling framework to meaningly examine the impacts of sensitivity tests using the modified model. This is why it is necessary to start off with the baseline control. When jumping straight to comparisons between observations and simulations with the modified model, as the authors have done, one cannot tell if discrepancies between the observations and the model are due to limitations of the model set up (which could include a number of things, starting with the initial conditions) or negative impacts resulting from the changes.  Figures 5, 6, 7, 8, 10 compare observations and the simulation with the modified code, but no comparison to what should be called the control run (with no rime).  Every comparison here should include simulations from both the original and modified microphysics scheme.

4.  In the illustrations of model precipitation (Fig. 11), the observations are conspicuous in their absence. There was a dense network of surface precipitation observations in that region for ICEPOP. This needs to be added.

5.  The explanation of particle density needs to be expanded upon in section 2.2, not just summarized by a reference to Heymsfield et al. 2018 (line 107). What is the density of ice with a rime fraction of 1? Is there no distinction between graupel and hail?  I gather from line (“… it is currently not the case in DESCAM [variable density]”).

6.  Line 349, “… in DESCAM [it is assumed that] ice particles melt instantaneously at the 0C isotherm”. This is a huge weakness in the DESCAM microphysics scheme in terms of modeling ice – it is not just a minor simplification. That might be fine for tiny crystal, but certainly not for ice that would be considered to be “snow” (large crystals or aggregates) or graupel.  Presumably it would make more sense to address this deficiency in the scheme before adding predicted rime fraction.  I definitely think the calculation of explicit melting should be added (it should be added anyway) in order to examine the impacts of rime fraction in the context of a case like ICE-POP.  The rime fraction will affect the ice fall speed, which will affect the horizontal distance ice is transported before completely melting, which will therefore affect the spatial distribution of precipitation, particularly in the mountainous regions.  So in order to understand the impact of predicted rime fraction, melting has to be treated more rigorously.

MINOR POINTS

1.  The title could be improved.“Heavy snowfall event during ICE-POP 2018” by itself does not mean much; it is just a noun.

2.  Line 310, “The overproduction of these small ice particles likely originating from a numerical artifact.” This sounds like a guess and is not very satisfactory, particularly given the negative impacts on the simulation.

3.  The discussion on the impact the fixed rime density could be expanded upon. If variable density were to be added in DESCAM, this would exploit the predictive aspect of aerosols since the liquid droplet size is important for the rime density.

Given the magnitude of the major comments, I will stop with the minor points and address them in detail if/when a revised manuscript is submitted.

This article describes the implementation of a prognostic rime fraction in the bin microphysical scheme DESCAM, and demonstrates its performance for a simulation at kilometric resolution of one event from the ICE-POP 2018 field campaign. Although already available in some other microphysical schemes, this new development for DESCAM is an important improvement, as it enables a better representation of ice hydrometeors characteristics. The real-case evaluation and comparison to observations is also very interesting, highlighting impacts of the prognostic rime fraction on cloud composition, but also on the cumulated precipitations at the ground and their type. However, I think this paper does not put enough emphasis on the microphysical processes on the one hand, and I have some concerns about the evaluation in comparison to the selected ICE-POP case on the other. It felt a bit awkward, somewhere between a classical demonstration of model abilities and a dedicated case study. Thus, I believe this article requires substantial modifications before it can be accepted for publication.

Major comments

1. Paper structure

As this paper presents the newly available prognostic rime fraction for icy hydrometeors in DESCAM, I expected a more thorough demonstration of how this impacts not only the general characteristics of the cloud, but also microphysical processes. I think the paper should have a full section dedicated to the differences between the two simulations (with and without the predicted rime fraction), rather than some elements of this comparison here and there in what seems to be more a case study. This section could go into more details about microphysical processes, even without involving observations. For example, how are the secondary ice production mechanisms affected? How different is the melting layer (there is a mention of the melting layer, although it does not appear on the displayed radar observations)? How different is the supercooled liquid water? What about the vertical structure of clouds? The cross sections in fig. 16 are very interesting but the discussion remains really short. For example, they suggest that the rime fraction is strictly limited to regions where there is liquid water and below, are there no updrafts in the deep cloud able to lift rimed particles a bit higher up?

2. Evaluation in comparison to ICE-POP observations

Overall, I would like to have more information about the studied case, I give some examples of missing things in the minor comments below. But I also have specific concerns about some parts of the comparison:

- The simulated W band reflectivities in fig. 5d show that the deep and shallow clouds in the shallow system are not separated by a layer of cloud-free air. This configuration typically leads to contamination in the low cloud from particles precipitating from the cloud above, and/or strong radiative influence. This is never discussed in the article, and may largely disturb the shallow cloud composition. Does that invalidate all or some comments about the shallow clouds? This should at least be discussed in the conclusion.

- In section 4.3, the differences between control and no-rime in fig. 13 are not clear enough to state that there is an improvement with control regarding the overestimation of large ice particles?

- In section 4.3 again, the comparison of observed riming index and simulated rime mass fraction (fig. 15) is used to demonstrate the improvement with with a predicted rime fraction. However, looking at this figure and fig. 14, I would argue that some data of fig. 15 should be removed when the number of ice particles is too low (specifically, observations around 6 UTC, and simulations after around 12 UTC). Then, I wonder if the argumentation about increasing rime index with diminishing number concentrations) remains valid?

- In section 4.4, I do not understand why these times were chosen (00 UTC, 14 UTC). According to fig. 5, 00 UTC corresponds to an “anomaly” in the deep cloud, when reflectivities are much lower than at other times. And 14 UTC is the very end of the shallow cloud event at MHS (figs. 5 and 14). The discussion about the change in large scale winds probably remains valid, but what about the cloud composition?

3. Discussion

The discussion could be extended. For example, is the fill-in hypothesis valid for all drop/droplet sizes, or is there a dependance on drop size? Since a parametrization of particle liquid fraction and progressive melting is available in DESCAM and mentioned in the article, why was it not used in this study? Possible biases in the aggregation are mentioned, how is aggregation influenced by the predicted rime mass?

Minor comments:

- l.59: DESCAM could be introduced before, citing some recent works, instead of appearing here in this sentence giving the article’s purpose.

- l.81+: give slightly more details about ice nucleation parametrization. Since aggregation is discussed in the article, briefly present its implementation, and should it be modified based on ice particles density? Also, is there a water shedding mechanism for particles experiencing wet growth?

- l.92: what type of crystals / clouds were sampled by Fontaine et al 2014?

- l.198-206: Since the paper is built around a case study, more information should be provided here. Maybe a weather map? I also missed a cross section, as that from fig. 16, to show the relative altitude of all sites more clearly than on fig. 4.

- l.226-236: how many vertical levels do you use? Why do you assume that aerosols are sulfate, so close to the sea? Maybe mention the chosen parametrizations for other processes, such as radiation and turbulence? These can also largely influence the production of liquid droplets and the availability of supercooled liquid water.

- l.237-240: Usually, the “control” simulation uses the original model configuration, and an “improved” simulations shows the new model version. I understand the choice here because the article focuses much more on the simulation using the predicted rime fraction, but it was slightly disturbing at first, and should be changed back if a section of the paper is dedicated to the differences between the two simulations.

- l.250: Do you need to change the particle properties in the radar simulator depending on the rime fraction? Does it account for attenuation?

- fig. 5 (and in other places): I feel that some information is missing, mostly regarding the temperature. It is essential to have that information, as it will largely influence the availability of supercooled liquid water, ice production (heterogeneous nucleation or secondary production), etc.

- l.274: can this be due to attenuation?

-fig. 7: I think you could enhance the figure by restricting the time period considered a bit more, to avoid the signal above 2km in the observations (only use observations after 8 UTC for example). And in the simulations, maybe only consider the reflectivities between 10 and 13 UTC, when the “gap” between the shallow and high clouds is the most important?

-l.304: Since you discuss particle sizes (and it is important indeed), it would be nice to have a cross section of some measure of particle sizes, compared between control and no-rime?

-l.324-325: delete this sentence (duplicate from the one before)

-fig. 9: why are the simulation curves blue and red? Use the same colors for all figures

-l.340: DESCAM is consistent with the observations

-l.348: the melting layer is not visible on any of the figures. It would be a nice addition.

-fig. 11: Is it really necessary to have 3 panels, as the liquid and ice precipitations are mostly exclusive of each other? Also, maybe merge with fig. 12?

-l.377: the bimodal size distribution is also apparent in the shallow system. Is it also a feature of the no-rime simulation? Could the numerous small crystals (often largely overestimated when we compare to the observations) result from (secondary) ice production, or is it really due to aggregation as suggested?

-fig. 14: Can you explain, or try to explain, why the control and no-rime simulations are almost identical until around 2UTC, and then so different? What happens then?

-fig. 16: the IWC and LWC contours are hard to read. At first, I even thought that there was ice between the two dotted lines in fig.16c, and almost nothing at MHS or above.

-l.424: the rime fraction barely exceeds 0.5 at MHS in fig.16c, and does not exceed 0.5 at YPO, if I read the colors correctly.