the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Sep 2024
Arctic Multilayer Clouds Require Accurate Thermodynamic Profiles and Efficient Primary and Secondary Ice Processes for a Realistic Structure and Composition
Abstract. Multilayered clouds are frequent in the Arctic but their detailed analysis is underrepresented. Here, we simulate two cases observed during the 2019/2020 MOSAiC expedition using the ICosahedral Non-hydrostatic (ICON) model to explore the most accurate representation of these multilayer clouds. With a limited area setup, we investigate how cloud layers respond to perturbations in cloud droplet activation, primary ice, and secondary ice production (SIP). Using the measured aerosol concentration, we constrain our model through a new immersion freezing parameterisation. We find that multilayered clouds are challenging to simulate in remote areas without locally assimilated thermodynamics and that large-scale biases in the global forcing carry over to high-resolution simulations. Regarding cloud microphysics, warm-temperature ice nucleating particles (INP) are crucial to model mixed-phase clouds. However, constraining the model to the observed INPs is insufficient; a factor of 106 is required to reach observed ice mass concentrations, which is also achieved by including SIP. Breakup upon ice-ice collisions is explosive and can increase the integrated cloud ice number concentration by a factor of 105. Furthermore, the seeder-feeder mechanism significantly boosts snowfall by a factor of 103. An accurate representation of these microphysical processes is crucial to simulate multilayer clouds.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Chemistry and Physics and the authors also have no other competing interests to declare.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(17125 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Review comment on egusphere-2024-2988', Anonymous Referee #1, 18 Nov 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-2988/egusphere-2024-2988-RC1-supplement.pdf
- AC1: 'Reply on RC1', Gabriella Wallentin, 31 Jan 2025
-
RC2: 'Comment on egusphere-2024-2988', Anonymous Referee #2, 02 Dec 2024
This manuscript addresses the topic of Arctic multilayer clouds, and specifically ice related processes, using a nested modeling approach and case studies from the MOSAiC expedition. Broadly the topic of Arctic clouds, and mixed-phase clouds specifically, is an important one because of ongoing modeling challenges in representing Arctic cloud phase partitioning. Moreover, most mixed-phase cloud studies have focused on the arguably simpler single-layer stratiform cloud structure, while relatively little focus has been given to multi-layer cloud systems. One of the major challenges with Arctic clouds, single or multi-layer, is the formation and properties of ice crystals. A great deal of research is now pointing to the potentially important role of secondary ice production (SIP) in shaping the phase composition of Arctic clouds, yet there are few modeling tools to study SIP and its role in cloud structure. Through a series of simulations, this manuscript examines different factors that impact the ice within multilayer clouds. Thus, thematically, the manuscript is timely and focused on a topic that is important for improving our understanding of Arctic clouds and their representation in models. The topic is appropriate for ACP.
The manuscript itself has a number of issues (“major revisions”) that will need to be addressed before it is ready for publication. Many of the issues, listed below, are a matter of interpretation or description and should be straightforward to address. There are, however, two more significant issues that will require more work to address.
The first is interpretation of results. There are multiple examples outlined in the General Comments below where the authors speculate on why a given situation occurred. However, the speculation is not clearly labelled as such. Moreover, there is no need to speculate here because the results are from a model and the model should provide all information needed to clearly state why the given situation occurred within the model (which may or may not reflect nature). A good example is around lines 327-330 where there is an explanation given for the appearance of a second cloud layer. In my opinion there are physical inconsistencies in the explanation concerning evaporative cooling and mixing. It is possible that related mechanisms are in operation, but instead of speculating it is better to simply look at the tendency terms from the model to definitively state why the model formed the second cloud layer and then why it later went away. Temperature tendency terms (radiative, latent, mixing, etc.), water tendency terms (evaporation, condensation, etc.), and/or a buoyancy analysis would be very informative in this regard. Please have a look at the many areas where “interpretation” is provided and then include supporting evidence from the model instead of just speculation.
The second significant concern is related to the model itself. Is it possible to evaluate the model’s sensitivity to the specified ice properties and SIP mechanisms when the model representation of the liquid water is so far from reality? In general, there appears to be little sensitivity to the ice processes, and this lack of sensitivity might not be realistic. Based on many past model studies of Arctic clouds, there should typically be sensitivity to the specification of ice processes. For example, numerous papers and model intercomparisons have shown how increasing the ice nucleating particle concentration in the model leads to more ice formation (number and mass) and eventually full glaciation of the cloud. While different models have different thresholds for glaciation based on their own specific set of parameterizations, this basic behavior seems to be consistent across most models. Why is there so little sensitivity in this model? It could be that 1) the simulated cases are just so warm that there is not much that can be done to promote ice formation, or 2) the general model set up (the parameterizations and how they are implemented) is not fit for the purpose of simulating these mixed-phase clouds. There are other, contemporary model studies (not yet published) of very warm mixed-phase clouds that are having difficulty simulating ice formation, so #1 could possibly be true. But it is also important to ensure that this model can represent the basic processes that are known to occur in these Arctic clouds. The model should be run on a colder mixed-phase cloud case, like one of the classics from MPACE or ISDAC or a MOSAiC case from earlier in the year. At these colder temperatures, is the partitioning of phase better? Is there sensitivity to INP concentration? If the model cannot represent this arguably easier situation at colder temperatures, then there is clearly a problem with the model that would inhibit it from successfully assessing the sensitivity to various ice processes like SIP. If the model is able to represent reasonable sensitivities to INPs or SIP at the colder temperatures, then the challenges experienced for these warm cases might simply be due to the fact that the model’s specific parameterizations are themselves not suited for the warmest temperatures.
My final summary point is related to the last comment above. If the first order goal is to examine model sensitivity to ice nucleation and SIP processes, it is probably best to do so in a temperature range where these processes are known to be active and effective. At temperatures near 0 C, all ice processes are greatly diminished and even the WBF is not very effective. Thus, the selected cases are not actually great conditions for understanding INP/SIP sensitivities. It would be much preferable to examine these processes at -6 to -20 C where ice is clearly more significant and where a variety of SIP processes are expected to operate.
General comments
Title: This is a detailed title, but I’m not sure it clearly represents the paper. First it is not clear that any of the simulations arrived at a “realistic structure and composition”, so it is hard to say what is required to produce those. Additionally, as noted in some of the comments below, “efficient primary” ice processes do not appear to be in operation in these simulations, in large part because apparently the only primary ice nucleation mechanism is rain freezing, which is inefficient in the model and should be inefficient in these clouds. Lastly “…and secondary ice processes” is also not reflected in this paper, as really the paper only dealt with one SIP process and did not examine the (presumably tunable) efficiency parameters embedded in that SIP parameterization. Thus, I think it is best to come up with a more representative title.
Line 47: There are “at least six” SIP mechanisms. Recent work has suggested more.
Line 56: “shown” is a bit of a stretch here. At best, some studies have inferred that these processes might play a role, but little about SIP has been definitively shown in natural clouds based on observations.
Line 66: It is better to say “near the north slope of Alaska” as the flights themselves were often over the adjacent ocean.
Line 66-67: It seems that a definition for cloud needs to be given somewhere. Individual layers within a cloud sounds like it could be two clouds or it could be one, depending on the definition. Since this whole paper is about multi-layer clouds, it is important to give a clear definition for what is meant.
Line 91: “moored to an ice floe”
Line 95: LWP is “retrieved” not “recorded”
Line 96-99: There are additional uncertainties for this type of cloud product specifically related to the cloud type classification, and unfortunately these are unquantified. For example, while a given IWC retrieval might have a quoted uncertainty of 40%, that is when the retrieval is applied to the appropriate cloud. But if it is applied to the wrong type of cloud the uncertainty can be much higher. Cloud type classification is the challenge here and Cloudnet has some challenges in that regard. If nothing more, it is worth mentioning that there are other uncertainties associated with the full way in which the cloud retrievals are applied.
Figure 1: The caption discusses degrees C while the axis label is in K. It would be best to have a consistent temperature unit used throughout the paper.
Line 122-125: It would be very useful to know the vertical resolution in the boundary layer and/or at cloud level. I understand that the resolution changes in the vertical, but some information is needed on how well this model set up is able to resolve the appropriate cloud structures.
Line 135-136: It is not clear what this statement means. Typically, the “spin up” is to spin up the turbulence while the thermodynamic state is largely advected into the domain based on the model forcing. Certainly, there is also interaction between the turbulence and thermodynamic state. Please clarify.
Line 141: We have gotten to the end of the description but so far there is no documentation of the spatial scale of the different model domains. This spatial scale information (similar to the vertical resolution information) is needed to understand how well the cloud systems are resolved within the domain. Without knowing this information, it is difficult for me to comment on the appropriateness of the applied domains and resolutions.
Line 148: As stated here, CCN activation is based on vertical velocity. Yet, one of the conclusions of this paper is that the horizontal resolution doesn’t matter much. It is not clear how this can be true unless the CCN is an insensitive parameter in this model set up. At 1.6 km resolution the individual eddies (i.e., updrafts) in these clouds are not well resolved, such that the grid-scale vertical motions are likely much smaller than the actual vertical motions that occur at the (smaller) cloud scales. Thus, the CCN activation is likely less and more homogeneous across the cloud compared to the spatially inhomogeneous way that CCN are activated in natural Arctic stratiform clouds. It seems that a discussion of this point is quite relevant somewhere in the paper, especially its implication on the apparent insensitivity of the simulated clouds to the CCN perturbations. Additionally, it would be very informative to show vertical velocity results to provide insight into how well the model resolves cloud-scale processes as a function of resolution. There is literature (including some of the papers in the references section) that can provide insight into the expected magnitude of vertical air motions in these clouds.
Line 153-154: This sentence is repetitive with the following sentences and can be removed.
Line 161-162: Is there some justification for why “rain freeze” is the only primary nucleation mechanism for T > -12 C? Most (all?) of the pertinent clouds in these simulations are within this temperature regime, such that immersion freezing, and other nucleation mechanisms are not important at all. This would then require rain to form before ice could start forming. “Rain” is not common in these cloud as there is simply not the moisture and dynamics to form rain drops. There can be supercooled drizzle at these temperatures, so is that included in the “rain freeze” mechanism? If so, then this point should be discussed more clearly. If not, then it is not surprising that the lowest level cloud is typically almost entirely comprised of liquid water. Ice in that layer would only start to form due to what should be rare formation of rain or seeding from above (which could all then be multiplied by SIP). If the model says that there is a lot of rain forming in these clouds, then it is probably not properly representing natural clouds and the issue should be better understood. Finally, there is discussion (i.e, Line 193-196) about scaling all of the ice nucleation modes by a factor of 0.05. But when are deposition and immersion freezing active? The clouds simulated in the case studies are all warmer than the cut off thresholds for these two nucleation mechanisms. Then in lines 197-198 there is a statement about adding INPs at high temperatures (up to -7C), but it is not clear if this is an adjustment to immersion freezing so that it can occur up to -7 instead of only -12C. That point should be clarified. Even if immersion is adjusted to be active up to -7C, it still is not active in a lot of the clouds that are simulated.
Line 164-166 (and Line 440-441): “is known” is too strong here. Perhaps “is hypothesized.” The community simply does not understand SIP well enough right now to know what mechanisms are in action under what conditions, and what their net impact is on the ice properties. There is a “gap” between measured INP concentrations and measured ice crystal number concentrations, and people speculate that SIP might fill this gap. The “gap” between observations and models is another thing altogether. Surely it is possible to build and tune SIP parameterizations to fill any gaps that are present, but this does not confirm that SIP processes are actually the reason for the gap.
Line 171-172: Same as my above comment. There are a number of statements throughout this paper that tend to push the conclusions about SIP beyond what can really be concluded. In this sentence “has been shown to have a considerable impact” is a challenge because these are model studies. In a model a given SIP parameterization can be tuned such that it has an impact on the modelled clouds, but that does not mean that those same processes are important in natural clouds. Some of the language here should be tempered to more closely reflect the state of understanding based on observational and laboratory studies, which is not definitive at this point.
Figure 5: It is hard to read the contour labels, please replot with larger font.
Line 246: I’m not sure I’ve heard “hydrometeor content” before. How about just “water content” as in the labels?
Line 263-264: What does this mean? Is there data assimilation at other times?
Line 264-265: I’m not sure that the strength of the inversion is what sustains the cloud. It is likely more so the other way around. The inversion is not present because the cloud is not there to radiatively cool and drive vertical mixing. I also do not see the justification for the next statement about excessive vertical mixing being the culprit. Please provide further clarification / justification.
Line 268-269: From Appendix A it is not possible to determine if the simulations are improved. On the 1st, the figure set up and chosen contours do not allow one to see the potential impact on resolving cloud-scale motions and variability, which would be expected with higher resolution. On the 3rd, it does look like the higher resolution is starting to resolve some pulses of ice formation. I believe that starting to resolve some of these structures is actually a step in the right direction.
Line 278-279: Based on some of the comments above, it is not surprising that there is no large impact on the lowest cloud. Since only rain freezing is possible at the given temperature of this cloud, and rain formation should be very rare in these clouds, it doesn’t really matter how many INPs are present.
Line 284-286: This interpretation is not convincing for a couple of reasons. First, just based on Clausius-Clapeyron, sublimating ice will not provide enough moisture to then lead to liquid water saturation without a significant cooling of the air parcel. That cooling would have to come from vertical lifting of the parcel. But how would the parcel be lifted, ie. what provides the buoyancy? The text suggests that this is due to latent heating (i.e., condensational heating). But the source of moisture was from sublimation (cooling), which would cause the parcel to sink not rise. I don’t think it is possible to have both sublimation and buoyancy generated from condensational heating at the same time. Perhaps I’m missing something that needs to be clarified? In general, the model should provide the information that is needed to understand the thermodynamic balances at play as a function of height and to clearly distinguish why the model produced liquid water. Generally, there will need to be some convergence of moisture at that height, likely supported by advection (as suggested by the soundings), and the ice deposition rate must be small enough that some liquid water can form. Once that liquid water forms, the typical mixed-phase processes will kick in (radiation-turbulence-microphysics feedback) to allow the layer to persist for some time in the face of the low ice crystal concentration. To help understand this situation, it might be useful to do a simulation where ice is turned off altogether. I suspect that liquid water will form at that height supported by moisture advection. Then, as ice is turned on, and turned up, eventually the liquid cloud cannot sustain itself (as is shown by some of the simulations).
Line 287-288: Where is the evidence of this seeding? It looks like there is a full gap between layers without any falling ice in between.
Line 289-290: The fact that the cloud layers “seem quite impervious to perturbations” is concerning. There is a lot of literature on modeling studies that show clear sensitivities to ice. It should be possible to turn up the INP concentration high enough to achieve glaciation in the lowest cloud. So why does this model not show much sensitivity? It could be that 1) the cases are just so warm that there is not much that can be done to promote ice formation, or 2) the general model set up (the parameterizations and how they are implemented) is not fit for the purpose of simulating these clouds. What happens if this model runs a colder mixed-phase cloud case (like one of the classics from MPACE or ISDAC or a MOSAiC case from earlier in the year)? Is there sensitivity to INP concentration?
Line 308-310: WBF can only glaciate the cloud if there are enough ice particles. To better understand this point, what are the actual ice number concentrations? Is it the ice number concentration that increased by two orders of magnitude or only the INP concentration? In these simulations it seems apparent that primary nucleation is a limiting factor such that there are not enough ice particles to make WBF much of a factor; if the available INPs were to nucleate into ice crystals there should be plenty of ice for the WBF (although WBF is also limited at these warm temperatures). Generally, these results seem to suggest that the number of INPs is not the problem with this model set up, but it is rather the parameterizations for ice crystal nucleation.
Line 312: Here the CCN activation rate perturbation is introduced. It would be useful here or earlier when the simulations are introduced to more clearly outline why the different simulations are performed. In this case, for example, why is a perturbation of the CCN implemented? Is there some hypothesis that the properties of the liquid drop size number/distribution are important in the ice processes? Please explain so that the reader understands the logic behind the different simulations.
Line 319: It looks to me that the LWP is 3-4 times too high and not 5 times too high.
Line 327-330: The explanation for the appearance of this second layer seems to be speculative and it is not clear why evaporative cooling would lead to mixing. Additionally, there is speculation in the following sentences about the upper layer then radiatively shielding the lower layer so that it “dissipates.” It is possible that these mechanisms are in operation, but instead of speculation it would be best to look at the tendency terms from the model to definitively state why the model creates another layer and then why that layer goes away. For example, temperature tendency terms (radiative, latent) or water tendency terms (evaporation, condensation) will be informative in this regard.
Line 333: This statement about the upper layer causing the lower to dissipate is likely true and has been described by a few papers both observationally and within models. However, the description of this process has revealed the point that the authors tend to describe what is playing out in this Eulerian perspective as if it were a Lagrangian perspective. While the additional cloud layer only lasts for a couple of hours within the stationary vertical column, this perspective does not represent how long the cloud itself actually lasts. This could simply be a second cloud deck that advects into and then out of the vertical column of interest. When it leaves the vertical column, this does not necessarily mean that it dissipates as it is also advecting at the time. The authors should consider the difference between Eulerian and Lagragian perspectives when describing the interactions and transitions, both here and throughout the manuscript.
Line 336: Which increase in condensation?
Line 355: “ice concentration”: Does this mean ice number or mass concentration? Please be clear.
Line 355-356: This statement could possibly be true under certain circumstances, but it should be shown with model results. This process would totally depend on the ice crystal number concentration, which is not given. In general, even for enhanced ice cases, the number concentration of ice crystals is still likely very low. When one considers the size of a crystal, and the number of ice crystals and liquid droplets per volume, there is typically a large physical spacing between ice crystals with many liquid droplets (i.e., lots of surface area for evaporation) in the vicinity of each ice crystal, providing ample vapor supply to grow the ice.
Line 381-383: this collisional breakup is a positive feedback, where more particles makes more collisions, which makes more particles. Thus, it is indeed an attractive mechanism to multiply ice concentration. This also means that the decisions made regarding the parameterized efficiencies (collision efficiency, breakup efficiency, number of resulting particles per breakup event, etc) are very important and some might be highly sensitive. This point should probably be discussed.
Line 384-391: Given that seeding plays only a small role and does not appear to be the culprit for glaciation towards the end of the case, why does this glaciation occur? The SIP processes are presumably occurring for many hours prior to the point of full glaciation. What is special about that transition? Line 398-399 further suggests that there is a time dependence to the SIP impact. First, it is difficult to interpret time-dependence in this Eulerian perspective. Second, if there is indeed time dependence, what is the mechanism for this?
Line 393-395: I think this sentence states that there are no sinks of ice and that the ice that occurs at the analysis point was initialized when the model was initialized and advected all the way to the analysis point (i.e., it is the same ice). However, I do not believe this is the case unless there is something strange with how the model represents fall speed. Ice crystals typically fall at about 1 m/s, so they would fall more than 3 km in 1 hour. Thus, there is a sink of ice crystals from any given layer of the atmosphere due to fall speed. Moreover, the residence time of ice in the atmosphere relative to advection from the domain boundary is something that can be determined directly from the model.
Line 433-435: It is not clear how the extent of these clouds should matter. Please explain further. There are many past model studies that have been set up in a similar way and have shown sensitivity to CCN, so why not in this model?
Line 437: “…does not survive for more than two hours.” Given the Eulerian perspective you can only say that the cloud does not exist in the analysis column for more than two hours. However, within the model you could track the cloud elements to see how long the cloud actually lasts along its Lagrangian trajectory, assuming the domain is sufficiently large.
Line 455-456: Indeed, the theta profile from the case suggest that there is not strong coupling with the near surface, such that the surface aerosol measurements are likely not representative of cloud level. In fact, the INP concentrations are likely larger aloft (based on some other work from MOSAiC and elsewhere).
Line 467: It is more useful to know the actual vertical resolution at cloud level than the number of layers in the simulation.
Figure A2: This comparison suggests that there is way too much ice/snow formed in the upper cloud. What does this say about other problems with ice nucleation and growth? Also, it is not possible from these plots to evaluate the impact of resolution on the lowest clouds, which should be the ones that are most impacted by resolution. It looks like no ice is forming in them, leaving way too much liquid.
Figure B1: What is the gray shading? That should be added to the caption.
Figure B5: What is the “heterogeneous freezing” and why is it at the bottom of the upper clouds? I assume this is the rain freeze mechanism and thus there is freezing at the bottom where there is the most rain? How are the particles at the top of the cloud formed? All of the particles will, on average, fall so there needs to be a particle source at the top, not just the base.
Line 567-568: This citation is incomplete. Should include: Atmospheric Research, 51, 45-75.
Line 576-577: This citation is incomplete. Should include: Atmos. Phys. Chem., 12, 9817-9854.
Line 639-640: This citation is incomplete. Should include: J. Atmos. Sci., 62, 1665-1677.
Line 646-647: This citation is incomplete. Should include: Nat. Geosci., 5, 11-17.
Line 671-672: This citation is incomplete. Should include: J. Atmos. Sci.
Line 691: This citation is incomplete. Should include: J. Atmos. Sci., 63, 697-711.
Line 723-724: This citation is incomplete. Should include J. Climate
Line 736-737: This citation is incomplete. Should include J. Atmos. Sci.
Citation: https://doi.org/10.5194/egusphere-2024-2988-RC2 - AC2: 'Reply on RC2', Gabriella Wallentin, 31 Jan 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 283 | 76 | 130 | 489 | 19 | 19 |
- HTML: 283
- PDF: 76
- XML: 130
- Total: 489
- BibTeX: 19
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 155 | 32 |
| Germany | 2 | 67 | 13 |
| Sweden | 3 | 32 | 6 |
| Netherlands | 4 | 28 | 5 |
| China | 5 | 19 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 155