Reviewer 1

Thank you very much for spending the time and effort to thoroughly review our study, which helped to improve the manuscript. Based on the reviewer's comments, we rewrote most of the manuscript and shortened it significantly. We refined the research questions and give more insights into how our measurements are suited for model evaluation. Specifically, we worked on the description of the synoptic conditions during the MCAO event and of how the two flights probed different flavors of it. In the following, we reply (blue) to all reviewer‘s comments (black). Text passages from the manuscript are in italics. In our answers, we always refer to the line numbers of the revised manuscript.

One change that does not correspond to a reviewer's comment needs to be highlighted first. We found a problem in the roll circulation detection algorithm. For radar reflectivity (Ze) profiles with more than one cloud layer, the hydrometeor depth of the upper cloud layer was used by the algorithm previously. However, a few radar profiles observe two cloud layers with a very short cloud-free distance in between the layers, which likely occurs due to limited radar sensitivity. For these cases, the methodology to derive a roll circulation “object” ignored the lower cloud part. Thus, the updated version of the algorithm counts two cloud layers as one as long as the cloud-free distance between them is thinner than 50 m (see line 147). This update increases the hydrometeor depth for these cases. However, this modification of the algorithm only leads to small changes. Previously, we identified 364 and 109 cloud circulation objects in the ’cloud street’ regime on 01 and 04 April, respectively. This changes to 356 and 112 objects which is corrected in line 251 .

Major comments

First of all, the manuscript is much too long and contains too much detail not necessary for the reader to know. Much of this could just go all together, while some could be put in supplementary information.

We reduced the length of the manuscript by removing the synthesis section and moving the most important information into other sections for a better narrative. Moreover, we moved details about the methodology into the Appendix and reduced the amount of details in all sections, particularly in Sect. 4.4. Some extra figures can be found in a supplement.

Second, the discussion is sometimes very confusing, jumping from conditions in one of the cases to the other and back, and it is very easy to loose track on which case is actually discussed. It would be much better if at least initially, with all the background conditions, the two cases are discussed entirely separately.

We improved the flow of the manuscript by shortening it. Now, both cases are first discussed separately in Sect. 4.1 before focusing on similarities and differences between the two cases in the following sections.

Moreover, there is too much details in the figures, that are also much too small, and the figure captions are very disorganized. Figure captions should be very succinct and organized and – in contrast to the text – not have a narrative; just the bare bone facts about what is in the figure.

We reduced the details within Fig. 6 and 8 and better organized Fig. 2 and 3. We also divided Fig. 7 into two figures. We hope that this reduces the details in the figures and makes them easier to understand. Furthermore, we revised the figure captions so that they only describe necessary information.

Third, the author seems to work hard to make the case that the cases are really different, while what strikes me is how similar they are. Sure, magnitudes are very different but looking at – for example – trends along the fetch, I find the underlying structure surprisingly similar.

Similar and not similar highly depends on what type of clouds are compared. In our study, we focus on clouds in marine cold air outbreaks, which are of a specific nature, and general characteristics of clouds in MCAOs (trends along fetch) are well known. Thus, we state in line 408: „The evolution of cloud microphysics with fetch is similar on both days, however, thermodynamic conditions modify the intensity of the parameters.“ Note that we integrated more synoptic aspects into the paper following reviewer 2. This shows that both cases belong to one long-lasting MCAO event but have slightly different flavors. More details are provided in Sect. 2.1. In the conclusions, we also discuss that looking at two events with many commonalities but significant differences in cloud properties provides an interesting benchmark for testing models.

Finally, the conclusions and the answers to the research questions, given initially, are very trivial and reveal nothing new, that couldn’t have been guessed given just the basic information.

We agree, that some of the conclusions might have looked trivial as these agree with the general concept of the cloud evolution in cold air outbreaks. However, our study is able to confirm theory by detailed airborne observations that were not available before in terms of instrumentation but also in terms of resolution close to the sea ice edge, i.e., the initial stage of cloud formation in MCAOs. We sharpened the reformulated research questions:

I. What are the differences between the environmental conditions on both flight days and what are their implications on the cloud development?

II. Can we identify characteristic changes in cloud and precipitation properties perpendicular to cloud street orientation, i.e., within the roll circulation?

III. How do roll circulation, cloud, and precipitation properties evolve with fetch in the initial MCAO phase, e.g., up to travel times of four hours?

Furthermore, we reformulated the introduction to emphasize that we provide unique data and evaluation metrics to test MCAO simulations which is further discussed in the conclusions.

Minor comments:

Line 35-36: Unclear; I don’t understand why ice cannot grow at the expense of the liquid just because the humidity is above saturation w.r.t. ice.

We removed this statement.

Line 45: “flow divergence” how and where? Line 51: Why – and where - does the fluxes increase with a sharper MIZ?

Both statements were removed when we shortened the introduction.

Line 64-65: While the open water distance passed by an air parcel increases with open leans and open water over the MIZ, it is unclear what this means in terms of fluxes; this depends on the water temperature which to some extent will be controlled by the neaby ice.

The text has been shortened here. We now also show SST and sensible heat flux from ERA5 and dropsondes in Fig. 3. Note, that in the marginal sea ice zone the sea surface temperature (fixed to roughly -2°C) is always higher than the ice temperature and thus open water always enhances turbulent heat fluxes compared to closed ice (e.g., Liu et al., 2006).

Line 66: What cloud-street charateristics are variable and how?

Removed when shortening the text.

Line 74: Unprecise; the spatial resolution of CloudSat along the track is quite good; however, the swats are far apart and nearby swats occur sparsely and swats are generally not aligned with CAOs.

The statement was removed when we shortened the text. Note, that coarse is always relative: in respect to our study the roughly 10 times finer resolution of the airborne measurements is able to resolve details within the cloud streets which is not possible by MODIS or Cloudsat.

Line 128-129: Unclear; I have a hard time seeing the convergence line discussed here.

We added a synoptic map to Fig. 1 and explicitly mark the convergence zone, which also corresponds to the cloud band of enhanced MODIS LWP.

Line 130: Maybe “different” instead of “varying”; the latter would be changes within each case while the former changes between the cases.

We use different.

Figure 2: An interesting feature in this figure is that the reflectivity seems not to be symmetric; the maximum seems to be shifted to the right (in the figure) and is not centered in between the downdrafts. This is not even discussed, and I have no idea if this was just spurious or is a systematic feature.

Yes, we see it as an advantage of our algorithm that no symmetry needs to be assumed as updraft and downdraft regions are diagnosed independently. We now discuss this fact in line 248 (“According to our definition, the maximum updraft (maximum Ze_0.7) does not necessarily need to be centered between the two detected edges of our roll circulation object.”) and in line 345 (“As explained before, objects are not necessarily symmetric. However, most clouds form centered around the updraft of the circulation: around 50% of the time, maxima of Ze_0.7 occur within the central tercile of the cloud and only rarely within the tercile closest to the lateral cloud boundary (7%).”).

Line 160: Unclear what you mean by “below 30 g m-2”. Do you mean the accuracy is “better than 30 g m-2” or that there is something special with the accuracy “at values below 30 g m-2”? Moreover, 0.5 30 g m-2 is also “below 30 g m-2”, so if the first is true, it doesn’t say anything.

The maximum uncertainty of LWP induced by the retrival process is below 30 g m⁻². See Ruiz-Donoso et al. (2020) for detailed information about how this uncertainty is understood. A detailed paper about LWP observations during the HALO-(AC)³ campaign is in preperation. In the manuscript, we eliminated the part on the detection limit and reformulated the statement about the maximum uncertainty: “Depending on atmospheric conditions, the maximum uncertainty is below 30 g m⁻² (Ruiz-Donoso et al., 2020).“ (line 156).

Line 164: I thing you mean “… non-constant manner.” or maybe “… non-constant way.”

Corrected.

Line 187: What does it mean that you use 850 hPa (rather than some lower value)? Both cases are rather shallow at the distances you study, so why not use e.g. 925 hPa.

Our definition of the MCAO index follows the definition of previous studies like Walbröl, et al. (2024), Dahlke et al. (2022), Papritz et al. (2015), and Kolstad (2017). We take the potential temperature at 850 hPa to be able to compare our MCAO indices with indices from other events analyzed in previous studies. Moreover, we want to be sure that the layer of advected warm air above ABL over open water on 04 April, which does not represent the temperature of the free troposphere, does not influence the MCAO index.

Section 3.2: For something as basic as a trajectory calculation, this section is exceeeeedingly long. It can be shortened to a third of its current length.

We shortened this section.

Line 204-209: This paragraph is very confusing and it is unclear what the argument made really is. Moreover, the assumption that the cumulative flux is independent of wind speed, since length and time are interchangeable, is based on an assumption that the temperature difference between air and sea surface is constant, which is almost certainly not the case.

We deleted this paragraph and just make the point that fetch and travel time can be analyzed interchangably for our data. For Fig. 8, we show both fetch and travel time to enable the reader to better compare with other studies. See line 206: „For flows unaffected by land masses, travel time over open water and fetch can be linearly converted and are both valid to study.“

Line 213: I don’t know what and integrated time means.

We replaced integrated time by travel time over open water.

Line 215-217: I beg to differ; the IFS has significant biases in almost all boundary-layer parameters and especially in the turbulent fluxes. This is borne out by the quoted 200 m error in BL growth (line 217) which to me in these conditions is not a small error.

We removed the statement about the BLH as it does not fit the story line. Note, that most HALO dropsondes have been assimilated in ERA5 leading to an improved performance for our study case. This is now mentioned in Sec. 2.3, line 195. Thus, ERA5 sensible heat flux compare well with estimates from the dropsondes (Fig. 3).

Line 218: I don’t understand the argument of this resampling, which provides trajectories at an almost ridiculous resolution, far exceeding the resolution in the input data from ERA5.

Our aim is to attribute a fetch to each airborne measurement having a resolution of 1 second. To do so, we do not resample the minutely resolved fetches but rather assign a fetch to all observations within the respective minute. We have rewritten the statement in the manuscript: „Specifically, we calculate back trajectories for the previous 12 hours for every flight minute and assign them to the observations within each minute.“ (line 211). The aspect of resolution (and its limitation) is discussed in line 219: “Note that due to the resolution of ERA neighboring trajectories are rather similar (Fig. 1g, h). Differences in fetches between two neighboring trajectories mainly come from differences in SIC along the trajectories. The median of the relative change between two adjacent fetches is 9.6%.”.

Line 231: How do you handle the SST in the MIZ?

In the paragraph the reviewer refers to, we only calculate the distance traveled over open water where the geometric effect of sea ice is taken into account not the SST. In Fig. 3, we also include a map of SST (from ERA5) to illustrate the SST in the region. In the MIZ, it is basically fixed to freezing point of sea water. Quantitatively, we only need SST for the calculation of the MCAO index from the dropsondes. For this purpose, AVHRR observations of SST are used (see Sect. 2.3). As we only calculate MCAO indices for the cloud street regime the MIZ is not affected.

Section 3.2: This is one of the most interesting introductory sections, but it is also much too long and detailed. Describe the principles and leave out the details, or move them to an appendix or supplementary information.

We reduced the details within this section and moved the details about the algorithm and the selection of the height threshold (0.7 of the hydrometeor depth) into the Appendix. To provide the reader with an overview we illustrate the principle of the algorithm by a simple flow diagram added to Fig. 2.

Line 244: Why 0.7? Are the results sensitive to this choice? Why not a constant distance into the cloud? Entrainment does not necessarily penetrate deeper into deeper clouds.

Our aim is to identify the height with the strongest particle growth induced by the strongest updraft. Herein, we have to avoid the influence of the entrainment at cloud top and that of precipitating ice towards cloud bottom. The MCAO clouds develop highly dynamically, and therefore, their depth spans a wide range (10th percentile = 30 m, 90th percentile = 625 m). Therefore, a fixed distance from the cloud top might not lie within thin clouds anymore. Thus, we consider a dynamic adaptation of the height.

We improved the text and moved the details into the Appendix. This includes a discussion about the height selection based on a sensitivity study (see Table A1).

Line 241-242: Does D constitute a cloud depth? Considering boundary-layer scaling, why not use the surface as a lower limit instead? Are the results sensitive to the -5 dBZ threshold to distinguish cloud from precipitation? What about when signals reach into the surface clutter.

D is the hydrometeor depth that includes cloud and precipitation hydrometeors. Taking the surface as a lower limit could result in the situation that the radar reflectivity is evaluated at a height where no cloud is present. The results are not sensitive to the -5 dBZ threshold as shown in Fig. A1. For all profiles, only signals above the blind zone (150 m) are evaluated. The same holds for the hydrometor depth that starts at 150 m regardless of whether signals reach below 150 m height or not.

Line 255: This is not a retrieval; “Extract“ is better. Done.

Line 259 & 261: Number of samples are usually integers; there is no such thing as 2.9 samples.

The fitting procedure provides the width in terms of the number of time steps (samples), which does not necessarily have to be an integer. For clarity, we mention the corresponding horizontal distance of 230 m in line 510.

Line 281: “particle concentration” is better; in reality I guess it has more to do with the cross-section area…We use particle concentration.

Line 285-287: All these precipitation or size against reflectivity relationships are a bit arbitrary; an expert on radar meteorology once told me that estimating ice concentration from radar reflectivity is uncertain to about an order of magnitude. So maybe phrase all this in a less distinct way.

The reviewer is completely righ. The dependency of the precipiation amount at Ny-Alesund on the Z-S relation was shown by Maahn et al, 2014. We emphasize the uncertainty stemming from the Z-S relation in line 151: „Note that these S estimates are inaccurate since Z-S relations highly depend on ice habits, which are very variable within cloud streets (Maherndl et al., 2023a; Moser et al., 2023).“.

Line 296-306: From where is the information in this paragraph about ascending and descending air coming? Much of this appears speculative to me; besides drop “mass” - “air mass is a different thing.

The information about subsidence is from ERA5 reanalysis and now shown in Fig. 3g, h. We dropped the word “mass”.

Line 309-311: This is a lot of words to say its colder on one day than the other; the temperature at 2 km has little impact on the boundary layer.

We highlight the exact temperature ranges because temperatures lie within and outside of the dendritic-growth zone on 01 and 04 April, respectively. The different temperature ranges might thus affect microphysics. Moreover, we mention the temperature above 2 km height to highlight that the temperatures are generally lower on 01 April and not only within the ABL due to processes within the ABL. The sentences in the updated manuscript are as follows: „On 01 April, temperatures are lower than -20°C throughout all altitudes over sea ice (fetch <15 km) and for parts over open water (Fig. 4a)“ (line 279) and „all temperatures below 2 km height lie within −20 to −10°C (Fig. 4f) and θ of the free troposphere is on average higher by about 5 K compared to 01 April (Fig. 4b, g).“ (line 292).

Line 321: What do you mean by “weakens less”? Compared to what? No other weakening inversion is mentioned in the text.

Rewritten: „ the capping inversion over the sea ice close to its edge is stronger and weakens less with fetch on 04 than on 01 April due to a layer of warm air above BLH.“ (line 295).

Figure 5. In fact, my interpretation of the temperature and moisture profiles is that the last profile for 4 April shows a deeper boundary layer, at almost 1 km albeit less distinct, than on 1 April that has a slightly more distinct top at maybe 600 m.

We do not share this opinion. The median of the cloud top height, which is often capped by BLH, is 300 m and not 1 km. Due to wind shear, the air above this BLH came from Svalbard while the air below originated from the north. Therefore, we strongly believe that air with different characteristics was advected above ABL.

Line 327: “directional shear”, “northerly” & “westerly”. Done.

Line 328: I disagree; I can’t see that the shear is systematically stronger for any of these heights

We have rewritten the sentence: „ A directional shear from northerly wind at the surface to westerly wind occurs at all heights, which is strongest at BLH (Fig. 4i).“ (line 295). The green and blue lines in Fig. 4i show a change in wind direction at 150 and 400 m, respectively.

Line 329-330: Again, very many words to say there is a low-level jet at 200 m.

Rewritten: „which is capped by a low level jet“ (line 282) and „Even though a low level jet exists at 200 m over sea ice as before (Fig. 4j), flow conditions differ compared to 01 April“ (line 294).

Line 332: I can’t see any Ekman spiral here. Dropped.

Line 334-335: Explain the discussion on the different angles and the convective instability.

We dropped the discussion about the alignment of the cloud streets as it does not fit the characterization of the ABL conditions with fetch.

Line 335-336: How does this statement agree with the previous angles discussed?

We dropped the discussion about the alignment of the cloud streets.

Line 341: It is not the profiles that precipitate!

Rewritten: “On 04 April, Ze rarely exceeds -5 dBZ even below 500 m reducing the frequency of precipitation compared to 01 April” (line 308).

Line 351: “weaker advection” is better Done.

Figure 7: How is the normalized width calculated; mirroring in the center? See my question earlier about the center (or the maximum updraft) appearing to be closer to one edge than the other. Also, why no LWP for 4 April; there are LWP values in the fetch plot.

See answer regarding Fig. 2 on page 3 of this document.

We now show LWP in the Fig. 7 and 8.

Line 392-393: I suggest ”… a factor of two or more …” Done.

Line 393: I have a hard time seeing any stagnation in this plot

Fig. 8 was revised and we hope that the new style simplifies seeing the reduced growth rate at fetches between 75 and 120 km.

Line 427-433: Suggest to drop this as you show no observations beyond 160 km.

Sorry, we meant 140 km which is included in Fig. 8. We corrected the statement in line 484.

Line 434-454: This can be dropped; if you decide to keep it shorten it and move to discussion or conclusions; it has no place here. We dropped it.

Section 5. I don’t really see the value of this section here. Reading this after the enormously detailed previous sections, this feels incredibly thin.

Riming preconditions cloud and precipitation evoluion. Thus, we think it is an important aspect to study. We moved the analysis of riming to Sect. 4.2 as riming preconditions the conditions similar to the ABL conditions in Sect. 4.1.

Section 6.: The whole Synthesis can be shortened significantly

We deleted this section and shifted the most relevant information to other sections.

Line 534-550: Is this what you learned from this study; it seems exceedingly trivial and I could have told you all of this, saving you a lot of time. There has to be more than this!

We significantly reworked the introduction, including motivation and reformulation of the research questions. In this way also our conclusions were adapted to highlight the significance of our work.

Liu, A.Q., Moore, G.W.K., Tsuboki, K. et al. The Effect of the Sea-ice Zone on the Development of Boundary-layer Roll Clouds During Cold Air Outbreaks.Boundary-Layer Meteorol 118, 557–581 (2006). https://doi.org/10.1007/s10546-005-6434-4.

Reviewer 2

Thank you very much for spending the time and effort to thoroughly review our study, which helped to improve the manuscript. Based on the reviewer's comments, we rewrote most of the manuscript and shortened it significantly. We refined the research questions and give more insights into how our measurements are suited for model evaluation. Specifically, we worked on the description of the synoptic conditions during the MCAO event and of how the two flights probed different flavors of it. In the following, we reply (blue) to all reviewer‘s comments (black). Text passages from the manuscript are in italics. In our answers, we always refer to the line numbers of the revised manuscript.

One change that does not correspond to a reviewer's comment needs to be highlighted first. We found a problem in the roll circulation detection algorithm. For radar reflectivity (Ze) profiles with more than one cloud layer, the hydrometeor depth of the upper cloud layer was used by the algorithm previously. However, a few radar profiles observe two cloud layers with a very short cloud-free distance in between the layers, which likely occurs due to limited radar sensitivity. For these cases, the methodology to derive a roll circulation “object” ignored the lower cloud part. Thus, the updated version of the algorithm counts two cloud layers as one as long as the cloud-free distance between them is thinner than 50 m (see line 147). This update increases the hydrometeor depth for these cases. However, this modification of the algorithm only leads to small changes. Previously, we identified 364 and 109 cloud circulation objects in the ’cloud street’ regime on 01 and 04 April, respectively. This changes to 356 and 112 objects which is corrected in line 251 .

Major comments

Abstract: Why these 2 CAOs in particular? Are they representative? Where do they fall within the more comprehensive assessment of Murray-Watson for example? This is addressed later in the document, but would be good to bring into the abstract. Is there a new process insight or hypothesis developed from their study that is new to their understanding? It is not clear to me why modelers should choose these examples above others, other than the opportunistic sampling. The authors also don’t come back to this argument towards the end of the document.

We take these two cases because of two reasons: first, P5 had a special sampling strategy on these days flying perpendicular to the cloud streets. Second, even though both flights probed the same MCAO event (Walbröl et al. (2024)), both cases have different MCAO strengths that allow us to study the preconditioning by the strength of the MCAO.

We added to the abstract:

“The evolution and structure were assessed by flight legs crossing the Fram Strait multiple times, sampling perpendicular to the cloud streets.” (line 10).

“The two events, just three days apart, belong to a particularly long-lasting MCAO and occurred under relatively similar thermodynamic conditions. However, for the first event, colder airmasses from the central Arctic led to an MCAO index twice as high as for the second event” (line 6).

The setting with a high over Greenland and a low over Siberia is characteristic for most MCAO over the Fram Strait as shown by Dahlke et al (2022). Nevertheless, both cases are stronger than the 75th percentile of the MCAO index between 1979 and 2022 (Walbröl et al. (2024)). We also added information on the synoptic situation in Sect. 2.1.

To the abtract, we added: “though both events were stronger than the climatological 75th percentile for that period.” (line 8).

We added more discussion on the study by Murray-Watson et al. (2023) who analysed MCAO development over 30 hours in the introduction. We also added a time axis to Fig. 8, which highlights that we focus on the first four hours of the events. Contrarily to their study, for our cases mixed-phase clouds dominate (Tab. 2).

The revised text includes more details on why modelers should choose our cases. With our data, modelers can test whether their model can represent dynamics and microphysics at the same time. Moreover, our data offer the unique opportunity to test the MCAO evoultion of models close to the sea ice edge.

We have added these information to the abstract and made the argument about the model comparison study at the end of the manuscript: „Within our analysis, we developed statistical descriptions of various parameters (i) within the roll circulation and (ii) as a function of distance over open water. In particular, these detailed cloud metrics are well suited for the evaluation of cloud-resolving models close to the sea ice edge to evaluate their representation of dynamics and microphysics.“ (line 21)

„To answer the two last research questions, we established composite approaches to characterize the roll circulation (Fig. 7) and fetch (Fig. 8) depends. Such metrics can also be generated from cloud-resolving model output and be used to evaluate their performance to represent microphysics and dynamics in the initial phase of an MCAO. By considering the two cases with similar large-scale synoptic settings but differences with respect to microphysics, e.g., LWP and riming, insights into the simulation of cloud microphysics could be gained. In particular, it will be interesting to analyze whether such models successfully reproduce the observed factor of two in scaling found for several parameters between the two cases.“ (line 463)

It makes sense to me that the 2nd CAO is the weaker of the 2, as it occurred 3 days later, and likely experienced similar synoptic conditions. The authors treat the 2 CAOs as if they are independent of each other. What initial conditions did they share? A figure indicating the dominant synoptic situation and how it evolved would helpfully complement the first figure. I hypothesize a coastal low interacted with the northerly CAO flow to develop that convergence line, for example.

Both days were part of a longer MCAO which lasted for more than two weeks (see Walbröl et al., 2024 and their Fig. 11). During this period, slight changes in MCAO intensity occurred, though both days were above the 75^th percentile. We now show the synoptic situation on both days in Fig 1a, b and discuss the environmental conditions in Section 2.1. The major difference is that on 04 April, the flow came more from the east and was thus affected by Svalbard.

More detailed comments:

Abstract, line 12: “the liquid layer is wider” is unclear here. Do you mean a larger horizontal dimension of the rolls? Except the previous sentence says the width of the roll circulation is similar between the 2 cases. “Occurrence” -> occurrence of what? “are” -> “have”

We have dropped the first statement and rewritten the other ones, i.e., precipitation occurrence (line 15).

Line 30: “them” what is ‘them’? Define

The statement was removed when we shortened the introduction.

Lines 33-36: unclear to me how the 2nd sentence relates to the first. Isn’t the first sentence saying vapor depositional growth is favored when the vapor pressure> ice and liquid saturation?

Sentence removed during revision.

Line 37: awkwardly written. The statement was removed when we shortened the introduction.

Line 73: “were” -> “was” Done.

Line 83: “extend” -> “extent”. Same for line 391. And line 520. Done.

Line 85: awkward. Overall the sentence is vague. What are the induced changes in dynamics and clouds?

The statement was removed when we shortened the introduction.

Line 160-161: how do you know most of the liquid is near the top? As written this sentence sounds presumptuous. And what is the motivation from deriving LWP from the slanted profiles? Superficially the nadir profiles would make more sense.

We rearranged the data section to describe the retrieval of the lidar cloud top height, which is sensitive to the liquid on top, before the radar cloud top height. A discussion on the typical structure of mixed-phase clouds can be found in Shupe et al (2008).

We state: “Lidar backscatter is highly sensitive to hydrometeors, especially to liquid, which, in our case, is always super-cooled.“(line 123)

“Cloud top height is also derived from the radar profiles at the height of the uppermost radar reflectivity signal above the noise level. Comparing this height with CTH from lidar allows us to assess the supercooled liquid layer thickness (LLT). Here, we exploit the fact that the lidar is more sensitive to particle amount (liquid), whereas the radar is more sensitive to particle size, i.e., ice particles (Ruiz-Donoso et al., 2020).” (line 141)

“As shown by the lidar backscatter and its strong attenuation close to cloud top and in accordance with Shupe et al. (2008) we assume that most liquid resides in the uppermost few hundred meters of the cloud. “ (line 158)

You are absolutely right, nadir profiles would be the best. However, the LWP is only available on slanted paths and contrarily to the radar profiles not corrected to nadir. Thus, we correct the profiles to nadir via geometrical considerations.

In the manuscript we state:

“Both measurements are taken with 25° backward inclination of the instruments. While the vertically resolved radar measurements are reconstructed to nadir measurements, the passive measurements represent a slant path.” ( line 130)

“While radar reflectivities are corrected to nadir profiles, the TB and thus LWP measurements are along the slant path (Mech et al., 2022a).“ (line 157)

Line 240: wouldn’t the strongest updrafts generate liquid drops that are less obvious to the radar than the radiometer? Why expect ice production and liquid production to be collocated?

We expect ice production within updrafts because there saturation with respect to ice frequently occurs (Korolev and Field, 2008).

See line 234: “Here, frequent saturation with respect to ice and thus the formation of cloud droplets and growth of both liquid and ice particles occurs (Korolev and Field, 2008).”

Line 256: 2.9 samples = what horizontal distance?

This corresponds to a horizontal distance of 230 m (line 510).

Fig. 4: this is the first introduction of the measurement regimes, so you may want to identify them in the caption, it’s hard to understand this figure otherwise. Are these the same as provided in Table 1? the legend and symbol shapes are hard to see. Besides increasing them in size, would also suggest increasing the size of the overall figure.

Overall, the new layout of the figure does not include the regimes anymore.

Line 281: what is the uncertainty in the LLT metric?

The uncertainty in LLT is difficult to quantify. Geometrically, the vertical resolution of the instruments (e.g., 7.5 m for the lidar) limits the lower end. See manuscript line 144: „Due to limited vertical resolutions of the instruments and resulting uncertainties in CTH, the CTH of the lidar has to exceed the CTH of the radar by at least 10 m to be defined as liquid topped“.

Line 290: “precipitating” -> precipitation Done.

Section 4.1: there really should be a synoptic analysis and figure provided, this would be a good spot for it. Done, see Sect. 2.1 and Fig. 1a, b.

Line 296: was VIIRS imagery, of higher spatial resolution, not available?

We show the corrected reflectance in true color of Modis that uses bands 1, 3, and 4, which have a resolution of at least 500 m. The bands (l1, m4, m3) that are used for the true color corrected reflectance obtained by VIIRS have a resolution of 375 and 750 m. Thus, this product is not finer resolved than the respective Modis image having a resolution of at least 500 m. The corrected reflectance of VIIRS in false color (that is shown in Fig. 1g, h) takes the bands m3, l3, and m11 into account. Here, the m bands have a resolution of 750m, and the l bands of 375m. The corrected reflectance from Modis is thus finer resolved.

Table 2 is nice. Does the BLH top correspond to the cloud top, or the mixed-layer height? Indicating the range of cloud top heights would be useful also.

We now provide the acronyms in the Table caption and added as further information the median and the interquartile range (IQR) of the cloud top height. “BLH and CTH stand for atmospheric boundary layer height, i.e., the inversion height of potential temperature, and cloud top height, respectively.”. Boundary layer height and CTH are thus derived independently from dropsondes and radar, respectively.

Lines 301-304: fig. 4 is difficult to follow to be honest. The dropsonde and ERA5 values don’t agree well, the disagreement is more than I would have expected based on Seethala et al. 2021.

We updated the old Fig. 4 by the new Fig. 3 which helps to compare dropsonde and ERA5 data for MCAO indices and sensible heat fluxes. Generally, we can see a good spatial consistency of both.

Line 308: I don’t see mean thermodynamic profiles provided as part of fig. 1e,f,…..these figures only indicate the locations of the dropsondes.

Sorry, the reference did not fit here. We dropped the reference.

Line 303: One also can’t see from Fig. 4 that the MCAO follows the SST (SST isn’t shown) and basically we are just seeing a few values for the ERA5 dataset, because of the coarse resolution. I would suggest increasing the size of the dropsonde-derived values in fig. 4, and including SST, and reducing the size of the ERA5 values. I am confused why there are so many ERA5 values. The interesting aspect of this figure is the sharp increase in the MCAO index for 1 April (isn’t it)? How much of the fetch is over the MIZ? It would be useful to include the beginning point in the figure 4.

The layout of Fig. 4 (now Fig. 3) has completely changed. The updated panels include also SST and the locations with 80 % sea ice concentration. Still, the differences between MCAO indices retrieved from dropsonde observations and ER5 data are visible close to sea ice edge that arise due to the less sharp increase in ERA5.

Line 314: where are you showing the surface flux values you refer to?

We integrated ERA5 fluxes into Fig. 3 and also calculated them from dropsondes. The reference has been adapted.

Line 315: here you define the BLH. This would be useful to include in the caption of table 2 as well.

Done.

Lines 328-330: I rather think a spatial plot of the ERA5 winds and Geopotential height would be more informative here.

We now provide a map of the geopotential potential height at 500 hPa, mean sea level pressure and 850 hPa equivalent potential temperature in Sect. 2.1 (Fig 1a, b). This plot does not include wind fields as they can be derived from the shown variables.

Line 339: that the clouds only reach 1km and 500m respectively is interesting too. The CAOs documented at COMBLE tend to reach higher.

Yes, we now discuss this in more detail in our introduction. During COMBLE, observations were taken at around 500 and 1000 km fetch while we are focusing on the first 150 km. Thus, they capture a later stage of evolution than our data. This might explain the higher cloud top heights during COMBLE. For one MCAO during COMBLE, Geerts et al. (2022) retrieved from satellite observations that the BLH close to sea ice edge reaches also only a few hundred meters.

See line 65: “Therefore, the Cold-Air Outbreaks in the Marine Boundary Layer Experiment (COMBLE) in 2021/2022 (Geerts et al., 2022) established two ground stations at Andenes and Bear Island, Norway, which provided important insights into cloud properties (Mages et al., 2022; Lackner et al., 2023) and supported model evaluation (Geerts et al., 2022). However, these stations were located about 1000 km away from the sea ice edge. Thus, only open and closed cellular convection but no cloud streets have been observed.”

Line 342: awkward writing.

Rewritten: „The shorter the fetch on 01 April, the stronger is the decrease in the mean Ze profile close to the surface (not shown).“ (line 304).

Line 343: you could look at the near-surface RH to examine if more evaporation is happening.

We investigated the profile of the relative humidity with respect to ice (see Fig. R1 at the end of this document). The figure nicely shows that for small fetches on 01 April (a, light blue), the relative humidity decreases to values below 100 % close to the surface. This confirms the hypothesis that ice particles might sublimate under these conditions.

In line 305, we added a statement to the manuscript: „Thus, near-surface ice particles might experience stronger sublimation on 01 April when the mixing ratio is comparably small and relative humidity with respect to ice below 100 % (not shown).“ .

Line 347: not convinced this is the best place to mention Morrison et al 2012. That is a modeling study based on more weakly-forced clouds. If you are going to mention it, better left to a discussion later on. We dropped the sentence.

Fig. 7: this is a nice figure. The caption should state clearly what the 3 groupings represent. Is it the same for all panels? The lowest one is labeled differently. How wide/coarsely resolved are these distances? I just wonder if ice production is truly colocated w liquid production, or perhaps they get lumped together through a coarseness in the grouping.

We agree that it was irritating that the x-axis was not the same for all panels. The axis of the first row differed from the other rows. Thus, we separated the first row from the rest of the figure and put it into the supplement.

In the caption, we state: „The median (horizontal line) and lower and upper quartile (box edges) are displayed at the boundary of the clouds, the updraft position, and in between.“

Moreover, we added to line 349: “More precisely, we group cloud properties according to their distance from the maximum updraft region (Ze_0.7) into three regions: the central updraft region, the region close to cloud boundary, and the region in between.“

The resolution of the bins depends on the width of the cloud streets and is not universal. It might be true, that ice and liquid production is lumped together. However, since we take cloud streets that are at least 5 radar profiles wide, a finer grouping is impossible.

Line 367: not following this sentence. What do you mean by extreme?

Rewritten: „Strong riming events might explain the frequent high extremes of S (snowfall).“ (line 360).

Line 368: 87 -> 87% Done.

Line 370: spell out what ‘0.6 of D’ means (it’s easy to lose track of the acronym meanings).

Done.

Line 375: is the measurement limit known to the reader somewhere earlier? I don’t recall seeing it.

The measurement limit is already known to the reader, it was first introduced in Sec. 2.2, line 156: “Depending on atmospheric conditions, the maximum uncertainty is below 30 gm² (Ruiz-Donoso et al., 2020). ”.

Line 380: I think the clouds looked at within Shupe 2008 are different enough from CAOs that this analogy isn’t quite right. It could just be a coincidence. You could say in the discussion section that examining if CAO clouds are a truly unique cloud regime that behave differently from non-CAO MPC is a topic worthy of future study (I suspect the differences in surface fluxes do genuinely separate CAO and non-CAO clouds).

We dropped this statement as an analogy cannot be drawn.

Lines 381-383: if you want to keep this, it would be better suited for a discussion section towards the end.

We dropped this paragraph.

Line 398-399: vague. Might help to indicate the distances of ‘large’ and ‘small’ fetches.

We changed to: „The aspect ratio of the circulation decreases with fetch for fetches smaller 50 km and stays constant for larger fetches.“ (line 459).

Line 407: to me it makes more sense to look at how cloud microphysics depends on the surface fluxes, as opposed to the exposure to open water. You can actually estimate the surface fluxes from the dropsondes, this was done within Seethala et al 2021. It would be interesting to know those values.

Following Seethala et al. (2021) we now calculated the fluxes from dropsondes and compared them to ERA5 data in Fig. 3. A discussion can be found in line 259 ff. The figure shows the strong changes in flux strength close to the sea ice edge and a good consistency between dropsondes and ERA5. As the dropsonde locations were rather different on both days, ERA5 is better suited to get the overall picture.

Line 408-409: this sentence doesn’t make sense.

Rewritten: „On this day, 90 % of the profiles containing liquid-topped cloud streets have LLT smaller than 100 m, which is more than on 01 April (70 %).“ (line 409).

Line 413: here is where I think the reader learns the LWP uncertainty. Better to say the LWP values are beneath the maximum uncertainty of 30 g/m2.

Here, we removed the statement. However, the measurement limit is already known to the reader, it was first introduced in Sec. 2.2, line 156: “Depending on atmospheric conditions, the maximum uncertainty is below 30 gm² (Ruiz-Donoso et al., 2020). ”.

Lines 427: how do you deduce the subsidence? From satellite imagery?

We deduce subsidence from ERA5 and show it in Fig. 3. In the manuscript we write: „Even though ERA5 reanalysis with its coarse resolution shows a rising air mass at these fetches (Fig. 8), we suggest that the air mass subsides and suppresses cloud development.“ (line 385)

Lines 434- : I appreciate the effort to place the 2 CAOs into context and not claim the noted features are universal (though they well might be). It would be nice to see this sentiment expressed in the abstract as well.

See our answer to your first major point

Line 445: decreaseing -> decreasing Done.

Line 452: Khanal et al 2020 examine MPCs using MODIS, cited below.

We dropped the statement. We do not cite Khanal et al. (2020), because their analysis did not focus on Arctic MPCs.

Line 475: since updraft speeds are correlated to surface fluxes, you could examine the dependence of riming on dropsonde-derived surface fluxes to put this statement on a stronger footing. Unless you have in-situ vertical velocities from one of the planes (it doesn’t seem like it though).

The variability in riming is on a spatial range of 700-1100 m while the dropsondes probe at distant locations compared to the aircraft with sondes falling for roughly 5 min through the atmosphere. Thus, matching radar measurements and sondes is rather complex.

Line 485: the -> these, before ‘conditions’ Done.

Lien 486: do you show CTH<BLH somewhere? I don’t recall seeing that.

In line 377 we state: ”The comparison of boundary layer height BLH derived from dropsondes and closely located airborne measurements showed that CT H is generally only by 8.5 m lower than BLH”

Line 508: isn’t dendritic growth through vapor deposition, followed by aggregation, a different growth process than riming? I’m not sure I understand how aggregation encourages more riming.

Sorry, some poor English phrasing led to a misunderstanding. We reframed the sentence as follows: „On this day, cloud top temperatures are colder than or at the low end of temperatures within the dendritic-growth zone (DGZ; −20 to −10°C) and hence too cold for aggregation to be dominant (Chellini et al., 2022). On 04 April, contarily, riming is not significant because cloud top temperatures lie within the DGZ that favors aggregation.“ (line 329).

Line 508: and how do you know SST is rising? You haven’t shown SST anywhere.

We included SST maps from ERA5 in Fig. 3a, b.

Line 510: this is difficult to follow, not sure what this sentence is saying.

We have simplified the sentence: „In contrast, a decrease of LLT (32 %) by 20 m can be seen.“ (line 355).

Line 511: I think liquid droplets are typically more readily raised, because they are smaller than ice particles. Or maybe you just need to be more precise about the particle size - but I’m not sure you have the microphysical measurements to back up this statement. All in all, this paragraph is too speculative. Please dial it back.

We now emphasize that the statement is just our hypothesis: „We hypothesize that updrafts transport ice particles into higher parts of the clouds. The mixed-phase region thus increases at the expense of the liquid layer and riming is enhanced (Fig. 4.2). Riming increases ice particle size, Ze, and S in updrafts. The observed LW P increase in updrafts might indicate that condensation is more favored than depletion of liquid.“ (line 356).

Line 516: you are contradicting what you said on line 508.

This is again due to the misunderstanding. As already mentioned, we have rewritten the sentence.

“On 04 April, contrarily, riming is not significant because cloud top temperatures lie within the DGZ that favors aggregation.“ (line 331).

Line 549: I don’t believe temperatures have to be colder than -20C for riming to be

active, this is certainly not universally true, nor do you explicitly show this.

You are right, we do not explixitly show it. We deleted the statement.

Fig. R1: Averaged dropsonde profiles from HALO and P5 of relative humidity with respect to ice (RH_ice) binned by fetch on 01 April (a) and 04 April 2022 (b). The shaded areas represent the standard deviation of each category. The color coding follows the categorization of the manuscript.