The present paper uses WRF-CHEM numerical model to consider the role of cloud condensation nuclei, originated mainly from sea salt aerosols (SSA), in three Medicanes. The aim is to assess the effects of an interactive calculation of SSA on the duration and intensity of Medicanes. To this end, simulations have been conducted both considering prescribed aerosols (PA) and interactive aerosol concentrations (IA). The results indicate that IA produces longer-lasting and more intense Medicanes. The paper deals with an issue really worth of investigation, at the forefront of the research in the field. Unfortunately, the paper suffers from many limitations; considering that a considerable amount of work is needed, my recommendation is for rejection, although I encourage the authors to complete and investigate more deeply the simulations.

We will complete the work as suggested. Thank you for your consideration. 

MAJOR POINTS:

1. The paper is mainly a model validation exercise, with very limited physics insight; thus, I do not think the paper in the present form fits the topics of ACP;

 The paper is not presented as a model validation exercise, and that was not definitely our intention with this research work. As you have pointed out, as well as we state in the manuscript (paragraph covering lines 269 to 283), no observations are used. Although this limits the scope of the work, the overall ability of the model to reproduce MTLC structures is not evaluated in any part of the presented work. This implies that a sensitivity study is performed in order to determine the role of an SSA interactive production in the MTLC intensification. To our knowledge, this topic has not been previously addressed, and thus, we agree in that it is an issue worth of investigation. However, we do not agree with the lack of physics insight. Although the important role of the SSA in warm-rain production has been widely studied, the explicit feedback mechanism studied in this contribution is worth being carefully taken into account for the rapid intensification stage of a MTLC event. Along with this mechanism, both the importance of the latent heat release contribution to the core heating, and the microphysical processes are also studied. Please note that this paper was initially conceived as a letter. We admit that the adaptation to an article structure may have not been fully accomplished, but we still think that the results are conclusive and seem to be robust enough for being published, as long as we succeed in accompanying them with a correct interpretation.

1. The analysis of the results is superficial; all figures should be analyzed and commented on more deeply  (in particular, Figs. 2 and 3); additional figures should be included to investigate the simulations more comprehensively;

Thank you for the comment. 

1. The role of nudging appears only marginally related to the rest of the paper. The link with the main part (aerosol-cloud interaction) appears very weak. Nudging simulation results confirm what is already known and do not add anything new. 

We did not found any paper addressing the role of nudging in the dynamics of a MTLC. If you do know any references that could be useful for the discussion, it would be really appreciated if you could provide them. Moreover, what is presented in the manuscript is only a part of the complete work, and the results presented with respect to the role of nudging have, to our knowledge, no previous consideration. Specifically, the nudging effect on the differences introduced by the interactive aerosols are the main point here, and is what we have centered the discussion into when it comes to the nudging effect on the development and intensification of a MTLC. As suggested by Referee 2, we will try to further dive into this structure breaking by providing a new figure showing the vertical structure of both no nudging and nudging cases. 

1. Also, what emerges from Lines 109-111 is not true: spectral nudging  is appropriate to downscale climate simulations in order to obtain finer scale climate fields, but it is not appropriate for individual case studies (as your runs confirm), since the dynamics are not allowed to make the cyclone evolve freely (without constraints) within the domain, affecting negatively the results.

In lines 109-111, we are echoing the information given in the referenced paper (Miguez-Macho et al., 2004). The discussion involved in the use of nudging for the case study of a MTLC lies in the fact that the initial conditions are critical for its development. In this sense, the selection of the initial time may not be trivial. What we wanted to show is that, for certain initialization times, spectral nudging can be important to ensure a proper “seed” for the medicane appearance. Besides, even with a proper initial conditions, nudging may be beneficial for ensuring a “correct” development of the system. However, what we found conversely is that, although its use may be beneficial in certain cases, the constraining on the medicane development is too strong. Ultimately, the presented results in the response to Referee 2, which will be included in the new manuscript version, are associated to the use of spectral nudging and show the importance of the medicane’s vertical structure and the fact that an uneven forcing in the vertical direction, associated to a vertical shear, is critical to the maintenance of its structure.

On another note, we have not really concluded that nudging is not appropriate for our case studies. The use of nudging is important if the main goal of the simulation is to reproduce the real track. However, provided that our aim with this work was to prove that a physical mechanism as it is the case of the wind-SSA production feedback is of paramount importance for a proper evolution of a medicane, this was conflictive with using the nudging to force the medicane tracks to resemble that of the real storms. In this regard, we decided to center the discussion for this paper in the role of the mentioned feedback in the development of the MTLCs, but not in the “correctness” of their track or intensities, meaning by “correctness” being close to those found in reality. Still, we considered interesting to study the role of nudging in what attains their intensification and maintenance. The role of nudging in the task of achieving more realistic MTLC simulations, along with that of the initial conditions (run-up time), has obviously been omitted provided the difficulty to validate MTLC simulations with observations. However, we will try to address this point with sufficient detail in the new manuscript version. Thank you for the comment. 

1. Line 67, 74: I understand (but I am not sure from what is written in the manuscript) that in the WRF-alone approach you are using the double moment microphysics with the single-moment approach (progn = 0, i.e. a constant concentration of an aerosol with a prescribed size). If this is the case, you need to make intermediate runs with: - double moment microphysics fully active but with chemistry switched off,  - single moment microphysics and chemistry switched on. In that way, you will take one step at a time, otherwise the comparison is not fair, i.e. you are changing both microphysics and chemistry.

We are using a double-moments microphysics in both cases. In PA, a prescribed total number concentration (distributed with a lognormal size distribution) of aerosols is used, with double-moments in the distribution of the five hydrometeor types considered. This would correspond to using the double-moments Morrison microphysics (progn=1, mp_physics=10) as described in (Morrison et al 2009). For the IA case, GOCART is used and the concentration of natural aerosols is interactively calculated. Thus, the comparison of both cases seems legitimate since the same microphysics is used for the two. The single-moments simulations are available upon request, but not presented or analyzed in this manuscript. Provided that this case is related to another microphysical solving, this is considered as part of another sensitivity study and not appropriate to be approached here. However, we will revisit our explanation on these details for the sake of clarity. Thank you for pointing that out.  

1. Line 84: As discussed in Liu et al. (2012), Veron et al. (2012), Rizza et al. (2021), when intense winds are generated, as in the presence of Medicanes, the range of sea salt size should reach 200 μm, not 20 μm!

From these papers, it seems like sea spray droplets up to 200 micrometers are produced under intense winds conditions. However, their permanence time in the atmosphere is short provided their large size and subsequently highly probable gravitational deposition. Besides, what we refer to is sea salt aerosols, part of sea spray droplets but not the same. Thus, only sea salt aerosols up to 20 micrometers are usually considered to serve as CCN (see e.g., Andreae and Rosenfeld 2008), and hence the SSA bins considered in the GOCART scheme. Thank you for your comment. 

1. The discussion should be less qualitative and more quantitative: e.g., the comments at Lines 186-188 are difficult to identify in Fig. 1; the inclusion of a Table would be helpful, showing for each cyclone the average track length and cyclone intensity for IA and PA runs, for both nudging and no-nudging runs.

Thank you for your recommendations. We will try to improve these points for the sake of clarity and preciseness. 

1. Figure 2 and related comment:

-Line 197: what do you mean with “limiting the intensification potential of the medicane”? 

The nudging, as introduced in our simulations, forces the wind (above the PBL) in scales larger than a thousand kilometers to resemble the reanalysis wind field. This is exactly what makes the nudging simulations produce, in general, weaker deep convection and, hence, less robust MTLCs structures, as can be appreciated in Figures S1 to S6.

Apparently, one of the most intense phases of Cornelia (when the cyclone moves toward Sicily; Reale and Atlas, 2001) is reproduced only in the nudging runs;

We will further investigate into this to clarify this fact in the next manuscript version. Thank you for pointing it out. 

• to make the analysis more complete, one should add the track of the cyclone as provided by satellite images;

They have been included in the KDEs plots, as shown in the response to Referee 2. Besides, we will add a short discussion related to them in the new manuscript version. Thank you. 

• I do not understand why you do not show all cyclones but you focus only on one case; for example, important differences emerge between no-nudging and nudging experiments for Celeno, but they are not commented on. I think Figure 2 must include all pictures in an 18-panel figure;

The nudging and run-up time effects are here considered secondary results that should not overshadow the main ones. The fact of having considered only a medicane for the discussion prevented from showing multiple figures for each results, which could end up producing an overextended manuscript. We do think that enough results are presented for the discussion to be valid for the three case studies, since the rest of the cases are presented in the Supplementary material. Still, we will consider your suggestion and try to improve the discussion. Thank you. 

• the difference between IA and PA runs is only superficially commented on.

We suppose Referee makes reference to the introduction and methodology. Thank you for the comment, we will try to improve this point. 

MINOR POINTS:

Line 27-28: Note that this is not true for all Medicanes. In some cases, the WISHE mechanism is not so important while the baroclinic development is the driving mechanism even in the mature stage. See Miglietta and Rotunno (2019) and Dafis et al. (2020). 

That’s right. Provided that the role of the WISHE mechanism is not totally proven to be the main driver in all the MTLC stages, we will try to avoid its use since no explicit discussion about the importance of the surface enthalpy fluxes is addressed in this work. We will also include a short reference to the existence of multiple principal driving mechanisms for MTLC. Thank you for pointing that out.

Line 44: SSA is not defined yet (apart from the abstract);

Thank you. 

Line 45: The role of SSA has been partially addressed already in Rizza et al. (2021);

This seems to be an interesting study that we will try to mention in the next manuscript version to account for their results and reinforce our main message: the sea spray production is an important factor for the medicane development and should be interactively calculated. Thank you for discovering this work to us.

Line 94: what do you mean with “with dominant categories recomputed”?

This means that both landuse and soil data are recomputed when running real. When manually modifying these input data, the variable surface_input_source should be set to 3 to ensure the changes persistence. This expression will be substituted in the new manuscript version. Thank you. 

Line 96: do you take SST from ERA-Interim as well?

Yes, SST is assimilated from ERA-interim every 6 hours. 

Line 99: Why not using ERA-5 as initial and boundary conditions?

To our knowledge, no information is available about the best reanalysis for this type of study, and we wanted to keep ERA5 ‘unseen’ by the model to be consistent with future studies where validation with this finer resolution reanalysis could be useful. 

Line 126: what is NRL?

United States Naval Research Laboratory (NRL), it will be included in the new manuscript version. 

Line 134: I am confused: you indicate here Rolf as a tropical cyclone, but earlier you mention it is a Medicane. So, what is the right classification?

Rolf was a medicane. We will try to clarify this. Thank you.

Line 141-145: This is not a disagreement, but a normal difference that may occur between observations and reanalysis for a small-scale system.

That’s correct. This sentence will be modified. ¡

Line 157-158: why not evaluating the 4th condition for Celeno???

Although formulated by Hart for tropical cyclones, much of the works that employ the Hart methodology for studying the phase space of MTLC do not consider the 4th condition (see e.g. Picornell et al, 2013).  In the tracking algorithm used in this work, the use of the 4th condition is optional. An interesting discussion around the necessity of imposing this condition associated to the warm core height may turn too extensive and thus is not considered for this work. However, as it was used for Rolf and Cornelia, and not for Celeno, new figures have been produced as a result of removing the 4th condition for the other two medicanes to ensure the consistency of the employed methodology across all the considered cases. These figures will be included in the new manuscript version. Thank you for your comment. 

Line 191: additional indications on KDE should be provided.

You are right. We will include a section in the Methodology to briefly define KDE. 

Figure 3 caption: “upper half”: do you mean farther from the center?

Yes, we will change the way of referring to it. 

Line 215: the word “essential” is not appropriate;

Right, thank you. 

Line 270-271: As discussed above, at least the track of the cyclone as derived from satellite are available and must be used. Also, some data for Cornelia (Reale and Atlas, 2001), Rolf (Ricchi et al., 2017) and Celeno (Lagouvardos et al., 1997) are available.

We will review these references for the sake of completeness. Thank you. 

Line 277-281: the paper does not provide sufficient evidence to support the comments provided here.

Please note that we are not concluding that even if PA simulated medicanes are closer to the “real” ones, we have to use IA. We are concluding that provided the high sensibility that the medicane formation and evolution have on the use of interactive aerosols, it should, at least, be taken into consideration when simulating medicanes. We will still consider to rephrase these lines to provide conclusions in line with the presented results. 

This is an interesting modelling study about the influence of sea salt aerosols on the development of Mediterranean tropical-like cyclones (medicanes). The effects of different initialization times and spectral nudging are also investigated. The results were based on a large ensemble of 72 experiments, produced by the simulations of 3 medicanes (Celeno in 1995, Cornelia in 1996 and Rolf in 2011) x 2 aerosol approaches (prescribed and interactive aerosols) x 6 different initialization times x 2 nudging approaches (no nudging and spectral nudging). The use of English is very good and the abstract is concise. The figures are necessary and with good quality, although some minor corrections are suggested. However, I have some concerns about some points in the methodology and the depth of the analysis. It is suggested that this article may be acceptable for publication after a number of major corrections is performed.

Thank you very much for your kind words and for the invaluable time spent in a detailed and insightful reading of our manuscript. 

Major corrections:

1. lines 165-170: A) I agree with the notion of “compact set”. However, the authors have to justify the need of the compact set in this article, since it was not actually used here. The compact size was only shown in Figures 1 and S1-S6, but without being discussed or used. Moreover, it was not used in the summary diagram of figure 3, in which the individual times with a medicane were used. In the current form of the paper, I would suggest to remove the compact size calculation and description. The following 3 corrections (B, C, D) are relevant if the compact size is actually utilized and discussed in the paper. B) line 163: it is suggested to use “output steps (e.g. hourly)” instead of “time steps”, because they may be confused with the model time steps. C) In the formula of Q (line 167), why does m start from i+1, instead of i? (i) This is not properly defined when i=Nt. In this case, the sum starts from m=Nt+1 (i.e. outside the simulation), and (ii) when j takes its first value (i.e. =i) the sum is counted backwards (from m=i+1 to m=j=i). D) line 166: why does j ends at Nt-1, instead of Nt? Why is the final time of the compact set not allowed to be Nt, even if there is a continuous period with a simulated medicane? This appeared in Figure 1, in the IA simulation with run-up time equal to 108 hours.

There is a misspelling in the description of the indices prior to the formula definition. It should read “For each i in 1:(Nt-1), and each j in (i+1):Nt, find pair [i,j] such that:”. Thank you for noticing out. 

Besides, we will keep the compact set definition and use it in the discussion, provided that a discussion of the track length is what it was conceived for in the first place. 

2) Discussion of figure 6: Does an azimuthally-averaged, radius vs height, plot of these variables (that takes into account the whole storm) produce the same differences (at least qualitatively) between PA and IA runs?

Yes, these differences are preserved if we do an azimuthally-averaged radius-height plot, as shown in the figure below. We will include these figures in the new manuscript version. Thank you very much for the interesting suggestion.

Please see Figure 1 in the supplement provided with this comment.

3) lines 264-266: No figures were shown in the article in order to support the physical explanation which is provided here. It is suggested to provide and discuss these results about the effects of spectral nudging.

A figure showing the time-averaged (over the times in which a medicane structure is present) vertical extent of the warm core of the pairs no nudging/spectral nudging simulations for both PA and IA are included below for Rolf (starting on 2011 Nov 05) and Cornelia (on 1996 Oct 05) simulations (Celeno is discarded provided that no nudging and spectral nudging simulations of Celeno reproduce a totally different system). From this figure, it is clear that the warm-core is not as well developed and continuous over the vertical direction in the nudging runs as in the no nudging ones. We hope this sufficiently proves that, although nudging simulations may produce “better” tracks, the vertical structure of the reproduced cyclone is not as well-developed as in the cases where no nudging is used. Indeed, for Cornelia it can be seen that spectral nudging cools the core in the 700 hPa level and breaks the deep convective structure. 

Please see Figure 2 in the supplement provided with this comment.

4) Although section 5 states that this paper did not attempt to assess the simulated medicanes against the actual ones, I need to mention that (i) the medicane positions of PA runs in figure 2 seem to be closer to reality than the ones of IA experiments, since section 2.2 showed that Cornelia temporarily lost its structure before it moves over the Tyrrhenian Sea and strengthen again, and (ii) the tracks of Celeno in figure S12 do not agree with the track of the actual system in the literature. Therefore, some basic comparison of the simulations with the literature is necessary, because this is not an idealized modelling study.

Both notes are interesting and worth a discussion. First, it is certainly true that Celeno tracks are far from real when simulated without nudging. However, a more robust and symmetric MTLC structure is formed in this simulations. It would probably mean an entire second paper to address this fact and provide a feasible explanation on why this happens, since no easy interpretation seems to exist. Besides, literature suggests the structure break of Cornelia medicane before reaching the Tyrrhenian Sea. But, in fact, only the eye-like feature seems to have been lost (see Cavicchia and Von Storch, 2012 and Mazza et al, 2017), contrarily to what we wrote in section 2.2, and we will correct. Indeed, it seems like Cornelia did not entirely lost its tropical-like nature. 

Moreover, while it is true that a discussion of the tracks divergence from the “real” ones may be interesting, we found at least three problems that made us not including it: i) NOAA Tropical Bulletin Archive is available from 2007 to present. Thus, Celeno and Cornelia are not contained on it. ii) We were not able to find any paper providing an objective reproducible manner of tracking medicanes from satellite images. iii) We found no paper providing the tracks of these three medicanes. iv) As we argued in the response to Referee 1, it doesn’t seem like a valid approach to use three tracks from three different works obtained with different methods. 

Hence, in order to introduce the discussion suggested by both Referee 1 and 2, we have included in the KDE figures the tracks of the three medicanes obtained as a result of running the tracking algorithm on the ERA5 reanalysis dataset. These tracks will serve as reference for the discussion we will introduce in the new manuscript version. 

Please see Figure 3 in the supplement provided with this comment.

Minor corrections:

1. line 62: The phrase “No physics suite (…) is used for the model run…” is confusing. It is suggested to state that no physics suite is widely accepted for the simulation of medicanes.

You are right. Thank you for your comment. 

2) line 84: “ Single-moment microphysics …” 

What we meant here is that when single-moment microphysics is used, the effect of an interactive aerosol calculation is ignored, and no coupling is allowed by definition of a single-moment approach. We will include the sentence “Double-moments microphysics (progn=1) has been used for all the simulations object of this work” in the next manuscript version. 

3) section 2.1.1: Please provide the number of model vertical levels and the model top.

The model top is fixed at 1000 Pa, and the model vertical levels are 40 (not fixed by us, but instead selected by the model). 

4) line 94: Please explain the acronym WPS geogrid, which is a technical term of WRF not known to the users of other models.

You are right. Thank you for the comment. 

Moreover, what data sources were used to define land use, soil category, topography and land-sea mask?

The data sources are the ones downloadable from the WRF Users Page (WPS V4 geographical static data downloads), accessed on 4th December 2019. 

5) lines 118, 120, 122, 143, 217, Figure 1, etc.: It is suggested to use UTC, instead of GMT, throughout the manuscript.

Right, thank you. 

6) line 130: Please define NOAA and any other acronym at the first time it is used in the manuscript.

Thank you for pointing it out. 

7) line 138: It is suggested to use hPa, instead of mbar.

Right, thank you. 

8) lines 142-143: Please provide the location and the time of Meteor’s measurement and the time of the SLP measurement at northern Malta. Was this SLP measurement at Malta associated with the medicane or its parent low? The center of actual Celeno did not pass close to Malta.

You are right, we will review this specific data and correct them as needed. 

9) lines 148-149: a) The sentence should become “The lowest model estimated atmospheric pressure …”, b) I think that no such information about the lowest pressure of Cornelia is found in Pytharoulis et al. (2000) and Cioni (2014), c) The paper of Cavicchia and Von Storch was published in 2012. This correction must also be made in line 310.

We will take these notes into consideration for the next manuscript version. Thank you for your recommendations and comments. 

10) section 2.3.1: Please specify the vertical layers and the radius used in the Hart diagrams. Was the radius constant (provide the value for each medicane) or variable? Although this information may be available in Pravia-Sarabia et al. (2020), some basic information must also be provided in the current article.

The radius was variable, calculated as the mean distance from the center to the zero-vorticity line in eight directions, with a minimum of six non-infinite directions. The vertical layers were 900-600 and 600-300, being the vertical levels in which the thermal wind parameters is calculated:  900, 850, 800, 775, 750, 725, 700, 675, 650, 625, 600, 550, 500, 450, 400, 350, 300.

11) section 2.3.2: The intensity must be removed from the title of this subsection because it is not discussed in 2.3.2.

That is absolutely true. Thank you for noticing. 

12) line 160, last word: Do you mean the number of time points, i.e. the number of output times? Please make this correction throughout the article, to avoid confusion with grid points.

Yes, we will make the correction, thank you. 

13) line 161: The phrase “… support will be the total length calculated …” is not clear to me. The term “support” was not used in the article.

We will try to use a more accurate term. Thank you for pointing out. 

14) line 162: “... serves as an objective …”.

Yes, thank you. 

15) line 182: “… for IA simulations without spectral nudging.”.

Yes, thank you. 

16) Figure 1, 3, 4, 5, 6, S1-S12: The labels must be enlarged because they cannot be read in the printed version.

We are aware of this. Thank you for pointing out. 

17) Figures 2 and S7-S12: It is suggested to overplot the track of each actual medicane.

As mentioned before, we have overplotted the medicane tracks as a result of running the tracking algorithm with the ERA5 reanalysis dataset. 

18) lines 207 and 208: It is suggested to use “medicane duration” instead of “medicane tracks”. 

Thats right.

Figure 3 shows the duration of medicane conditions and not the track’s length. A longer duration does not imply a longer track (because the translation speed may change).

When we use the term “track length”, we refer to its temporal length, which seems to be an equally valid and frequently used measure of its duration. Thank you. 

19) Figure 3, caption: it is suggested to replace “upper half” and “lower half”, with “outer half” and “inner half”, respectively, because the orientation of all ring portions is not the same.

Your recommendation will be implemented in the new manuscript version. Thank you. 

20) lines 215-218: Please justify the choice of this case (Rolf) and this initial time (00:00 UTC, 5 November 2011) for analysis of the SSA-wind feedback in section 4.

This case accomplished the best equilibrium between robustness and long-lasting MTLC structure among all cases. Besides, it has been the most widely studied medicane in the literature. The initial time was chosen provided that it produces the longest-lasting and most intense MTLC structure. 

21) line 224, figure 5: I think that the strongest effect in mid-low levels (800-500 hPa) equivalent potential temperature appears in the center and seems to be related to ‘eye’ dynamics.

Right, thank you for the note. We will study this in more detail. 

22) Figure 4: The label “UTC+01:00” is not clear to me. Does this figure use local time? Why?

It is an error; it should read UTC+00:00 and will be corrected in the next manuscript version. 

23) line 243: “... an ensemble of simulations has been …”

Yes, thank you. 

24) References: The link “https://doi.org” appears twice and must be removed from the references of Dafis et al. (2018), Gong (2003), Miglietta et al. (2013), Miguez-Macho et al. (2004), Pytharoulis et al. (2000).

Right, thank you.