Summary

This is a modelling study on the sensitivity of different cold air outbreak events to certain aspects of the microphysics package (CASIM) employed in the UK’s unified model (UM).  Specifically, this work examines the sensitivity to the droplet number concentration (serving as a proxy for aerosols and CCN), the ice nuclei concentration (through a scaling factor for the Cooper curve) and an efficiency setting for secondary ice production via the Hallett-Mossop process (rime splintering.)  The primary objective is to study the relative importance of aerosol-cloud-interactions (ACI) and ice production processes in two cold air outbreaks. 

Two extreme cold air outbreaks based on observations over the Gulf of Labrador have been chosen, a relatively warm outbreak (Oct 2022) and a cold outbreak (March 2022). They offer a very nice contrast arising from the difference in cloud temperature.  The analysis of the cold case is most novel, given that the cloud resides at temperatures colder than -15°C.  The author readily identifies other numerical studies of warmer cold air outbreaks, similar to the Oct 2022 case, where the Hallett-Mossop process is likely to be active. 

The paper is well organised and well presented.  I commend the author.  The discussion of the response/evolution of the simulated cloud field and the development of precipitation to the various microphysics variations is particularly done well.  The authors also do a very nice job in discussing the range selection for the sensitivity study. 

Recommendation – Major revisions

I have reached this decision as I do not think the evaluation of the control simulations are nearly adequate.  There is one evaluation figure of the cloud water path (CWP) (Fig 4), which is immediately dismissed because there is little confidence in the CWP from MODIS.  While that may be true, that means there is no meaningful evaluation.  It is simply unfair to ask me as the reviewer, and the wider scientific community to accept “a full model-observation comparison for this case is in preparation and will be shown in a subsequent paper.”  This is putting the cart before the horse. 

I am particularly concerned because I don’t think the ‘stratocumulus’ section of the March 15 simulation (the western portion of the simulation domain) shows sufficient skill to allow for the analysis presented – at least not without major caveats.  From the RGB imagery (Figure 3a), I see relatively shallow roll clouds that grow quickly, which are common at the start of cold air outbreaks, when dry Arctic air first passes over warmer water. The strong winds and latent heat flux produce these rolls which progress to open MCC.  The simulation, on the other hand, shows a thick, very cold stratocumulus, with the CWP two orders of magnitude more than MODIS CWP (Figure 4a and 4b).  This is far more than ‘within one order of magnitude’ stated in the manuscript.  The boundary layer in the simulated profile for this region (Figure 11, top row) extends to nearly 3 km.  This is a region renowned for multilayer clouds (e.g., Mace et al., 2009 https://doi.org/10.1029/2007JD009755), yet the CONTROL simulation has a single, well-mixed boundary layer with a 3 km thick cloud?  I need further evaluation work here. 

Are there any aircraft profiles through this region from the M-Phase field campaign?  Are there any upper air soundings from the coast of Canada? 

At the least, I require an evaluation of cloud top temperature and cloud top height against MODIS products with a discussion focussed on the stratocumulus region of 15 March.  In a perfect world, we would have a CALIPSO overpass to tell us the true cloud-top height and structure. 

I am requesting this, given that so much of the stratocumulus-to-cumulus transition (SCT) discussion is underpinned on there being genuine stratocumulus in this portion of the domain.  If the CONTROL simulation evolves from something rather inaccurate to something accurate, then aerosols and microphysics might not have anything to do with it, rather it’s simply the synoptic-scale dynamics in the operational analysis being downscaled from something poor to something accurate. 

While the Oct case study also needs a more rigorous evaluation, I am more comfortable with the quality of the CONTROL simulation and the ensuing discussion. 

Minor revision

Given the detailed discussion on the SCT, it is worthwhile to establish the environment beyond the un-evaluated cloud top temperature.  It would be worthwhile to consider other aspects of the boundary layer environment and the role they may play in this transition.  Please comment on the SST, downstream SST gradient, the boundary layer stability, the M parameter and the estimated inversion strength (EIS).  As this manuscript reads, one might think that the only thing that matter are the cloud temperature and the microphysics.

Typos

Figure 1 legend:  2022 instead of 2024

Figure 8 legend:  24 October

In this manuscript, the authors examine a very important topic regarding the sensitivity of cold-air outbreak (CAO) cloud evolution and radiative impact to several highly uncertain properties of these clouds.  The cases they examine are quite distinctly different in terms of the temperature of the air mass.  Simulations are conducted using the Met Office Unified Model (UM) and parameters associated with the production of cloud liquid and ice phase particles are adjusted in multiple simulations for each case.  Satellite retrievals are used to establish the realism of the simulations and comparisons to aircraft data are shown in an appendix.  I find the paper to be generally well done.  It is logically presented and thoroughly described.  The conclusions are interesting and important where they find substantial sensitivity to several of the perturbed quantities and significantly different responses to these parameters depending on regime.   It is my opinion that this paper can be an important contribution to our understanding.  However, I have several questions regarding the simulations that should be addressed in what I think will amount to a major revision.

One concern I have is with the consistency between the ice water paths and the precipitation rates in both simulations.  The IWP in both simulations exceeds 1/2 kg/m2 as the clouds mature yet the precipitation rates seem to be quite small on the order of 5 mm/day in the mean.  Because the satellite data cannot constrain these aspects of the simulations, it would be helpful to present other evidence that the relationship between IWP and precipitation rates are reasonable.  My concern is that the model is too slow in making precipitation from the small ice that is nucleated in the supercooled environments.  In Figures 11 and 12, the authors show vertical profiles of the ice properties such as number concentration but they do not separate cloud ice from precipitation.  Some examination of the process rates would be interesting and might establish confidence since they could compare them to the process rates presented in other papers such as Karalis et al. (2022) whom they cite.

My primary concern is the validity of the March simulation.  This case is very cold and there is very little liquid water in these clouds  (~10% of the water path).  Essentially, the model is producing cirrus clouds in the MBL.  Comparison to satellite LWP from both MODIS and AMSR2 show that the simulated LWP is biased low by an order of magnitude.  I question the overall validity of this simulation and whether it is suitable for this paper.  It would help immensely in establishing confidence if the authors could present additional evidence that such ice dominated MBL cloud fields actually exist in nature. 

Minor comments:

1. Line 106:  Not sure what this means.  Why would ice and liquid be overlapped at 0.5?  Why not some other number like 0.9 since ice falls from liquid clouds in these environments.

2. line 144: When referring to quantities that vary vertically in the atmosphere one should not use above and below to indicate magnitude changes of a quantity.  For example, when referring to temperature, use warmer and colder since "above" can also refer to higher in the column creating ambiguity in the reader's mind.

3. What are the units of equation 1?

4.  SST and stability are relevant to understanding the events but neither are shown.

5.  Figure 4:  Seems that the LWP comparisons is not that great really.  It is very difficult to gauge what the water paths are to within an order of magnitude with the color bar used. The model does produce cloud streets and cellular clouds, but there seem to be pretty large differences in cloud fraction and water path in the March case.

6.  Figures 6 and 7 are very different but, according to the caption, are only different by 15 minutes in time.  I think there is some mistake here.

7.  Is Figure 8 March or October?

8.  Line 385:  I disagree with the contention that the March case LWP comparison to data are reasonable.  The figures show they are different by an order of magnitude.

Review of egusphere-2024-4070

Title: Different responses of cold-air outbreak clouds to aerosol and ice production depending on cloud temperature

Authors:  Xinyi Huang, Paul R. Field, Benjamin J. Murray, Daniel P. Grosvenor, Floortje van den Heuvel, and Kenneth S. Carslaw

Overview:

In this manuscript Huang and co-authors examine simulations of two Cold Air Outbreaks.  The analysis is largely focused on examining the sensitivity of the model simulations to changes in prescribed liquid cloud droplet number concentration, number of ice nucleating particles (INP), and the rate of ice product in a parameterization for the Hallet-Mossip secondary-ice formation processes.  The authors find that for very-cold  CAO events (in which there is little liquid water), variations in the prescribed liquid cloud droplet number concentration and Hallet-Mossip ice production rate have little impact, while large increases in INP leads to higher cloud cover, in-cloud IWP, albedo and SW flux at the top-of-atmosphere (the opposite of what happens in the warmer case).

Recommendation:  Accept pending minor revisions.

General Comments:

• How common are the Warm/Cold (March/October) like cases?

The manuscript criticizes previous studies for only focusing on one case, and yet only two cases are explored here.  These cases have (without sufficient support in my view) been presented as “archetypal”.   In particular, I think the manuscript should address how common both types might be, or at a minimum, present some evidence that very cold events like the March case are not unusual.

• Comparison of the control simulations to observations

The comparison of the control simulation to the observations is quite limited:  

1. A) On line 197 you write “A full model-observation comparison for this case is in preparation and will be shown in a subsequent paper.” In this event, why not wait for the “full” comparison to be completed in order to build appropriate confidence that these simulations are reliable and perhaps better establish a nominal control case?
2. B) Line 379. Given that events lasted days, why is the analysis restricted to essentially one A-train overpass, and even then, the only microphysical quantity used is LWP from AMSR? What about geostationary datasets?  Yes, one has to be careful are higher latitudes and larger solar zenith angles, but these data are far from useless.  I note the CERES SW, LW, Albedo data that are used, depend on cloud cover, optical depth, etc. from satellite datasets that are not used here (because it is suggested they are too uncertain).
3. C) The comparison to in situ data is largely confined to a small Appendix (B) without much detail. In Fig. B1 panel C. What are the symbols?  How do the satellite data compare to the aircraft data?  In general, TWC data appears to span orders of magnitude.  I’m not sure how one can justify the model as having done “well” based on this comparison. Please quantify (with appropriate statistics) and explain why this is “well”.  Is the position of the aircraft (relative to cloud base) accounted for in any way? In Fig B1 panel B, it appears that perhaps the aircraft was using ramps from which profiles of LWC/IWC and total liquid/ice water path might be obtained.  If yes, it looks like there might have been two-cloud-layers on two-of the ramps?  Are the later aircraft data from constant-altitude-legs (that might allow comparison for of obs/models at a fixed altitude)?   Was there no other microphysical information coming from the aircraft beside data from the Nevzorov probes?  Even if yes, why not compare/discuss the relative abundance of liquid vs. ice?

Minor Comments:

Abstract Line 2. Not entirely sure who is “our”.  The authors of this study?   The larger scientific community?   Perhaps simply change to read “Recent case studies of CAO events suggest that increases …”.

Abstract Line 3. Yes, on a reduction in total water content, but not sure this is true for reflectivity.  What radar wavelength?  Depending on the wavelength, reflectivity is largely controlled by precipitation particle size or amount (rather than differences in dielectric constant between liquid and ice) such that more precipitation will often lead to a larger reflectivity.

Line 22. Perhaps “a key” rather than “the key”?

Line 26. Similarly, perhaps “major” rather than “the main”. There are a lot of uncertainties.

Line 41. Here and next sentence, perhaps credit authors upfront rather than parenthetically. For example, “Field et al. (2014) found an improvement …”

Line 45. In my view, “Stratocumulus-to-cumulus transition (SCT)” is not a processes, and is a description (it doesn’t “affect the amount of stratocumulus and cumulus clouds in the cloud field). Perhaps simply, “Stratocumulus-to-cumulus transitions (SCT) in CAOs have an important radiative effect.”

**Line 58 (and others). Is “Murray and the MPhase Team, 2024” really the best available reference for the campaign? This is just a conference abstract with no links depicting flights, conditions observed, instruments, science plan OR information on how to obtain data, etc.  Frankly, a campaign web site might be more useful.

Figure 1. Please show the domain simulated in Figure 1. More generally, why show most of Europe and parts of North Africa. In my view it would be better to show a narrower region that focuses on region of study so as to provide more detailed picture of the relevant meteorology.

Line 94. Fixed Nd may make interpretation easier but it also removes potential feedbacks created by aerosols/CCN being removed via coalescence of cloud droplets (precipitation formation). Surely this is worth a line or two of text and perhaps some considerations as regards future activities discussed in the final section.

Line 106. I do not understand what “… a mixed-phase overlap fraction is calculated, with a default value of 0.5 along with liquid and ice cloud fractions from the cloud scheme” is intended to convey.  Please expand or rephrase the description, and in particular, please address the implications this has for the results presented in the manuscript.

Line 119. Do you mean equivalent mass?  equivalent volume?

Line 138. Units?  Perhaps give concentrations in #/Liter?

Line 143. The slope matches, but figure 2a would suggest to me the control case should perhaps be Sinp = 0.01.  So, using Sinp values of 0.0001, 0.01, and 1 for sensitivity tests. 

Line 384. I am not sure this is the best way to interpret the comparison. 5 g/m2 out of what?  Presumably less than 10 g/m2 on average?  What is the minimum LWP that AMSR can reasonably identify?   How good do you expect AMSR LWP to be?  My take would be that AMSR-2 shows LWP is small and this is consistent with the model for this case.

Line 387. Your write “Small underestimation of SW flux …”.   Do you mean in the default run?  Why do you suspect cloud cover is the issue?

**Figure 5 & C. This is all model data, yes?  Perhaps show uncertainty in the mean and/or IQR or some measure of spatial variability.  Why is observational satellite data not shown … at least for cloud-cover, albedo, SW, LW (which you seem to trust).   Also please consider putting March and October lines in same panels (unless you are going to add observations and this makes it hard to see).

Figure 5e & C1d. How is 5e consistent with C1d for October (**I think perhaps October data in panels C1e and C1d been swapped)?

Lines 200 to 217. I think many individual references here to figures 4a to 4e are supposed to be 5a to 5e

** Line 244. You have already said there is little LWP several times.  I presume the simulations of Tornow and Able were for much warmer clouds and contained more Liquid.  If yes, I don’t know why you are expecting results to be similar.

** Line 278. You write, “The responses become complex and some even non-monotonic near the eastern end of the sub-domain”.   Yes, so what is going?   The discussion on the role of SIP just seems to end with “it is complicated”, which is not very satisfying or useful.

** Line 280-286. Somewhat similar to the above comment, the take away on the effect of EHM changes just seems to be “it is complicated”.  As best I can see there are no conclusions in this manuscript as regards the importance of the HM process. You might look at total number of ice particles and the production rates of ice changes due to SINP and EHM.  Are EHM driven changes relatively small or only happening in only part of the domain?

Figure 7. The caption claims these are results for 15 March, but I believe they are for the October Case.

Figure 8.  Appears to have the opposite problem.   Claims to be for the October case but is for March?   Please check your captions carefully.

Line 383.  When SZA & Hsigma are large this is legitimate concern.   Note from their Figure 2, however, this means Hsigma > 10.    Is this really the case here?   Surely not for the whole domain?   

Line 391 you write “The simulation with a high SINP in the October case agrees well with all satellite retrievals in general (Figure 14). There is a small overestimation of all-sky LWP …”.  I don’t think I agree with this assessment. For LWP the default and small Sinp simulations look good on the western boundary but all simulations look poor in the east and only high SNIP compares well near 52W (in the middle).  A similar comment as regards SW might be made (all look bad from 48W to 44 W with error of 25 to 50 watts in the SW).   As per earlier general comment, instead of using qualitative “agrees well”, perhaps quantify apparent error (with appropriate statitiscs) and explain why you consider this good agreement.

Line 476 (and others).  Shouldn’t a change in cloud phase driven by aerosols be called a “cloud-phase adjustment” rather than a feedbacks?