The authors have captured the essence of my comments and those of other reviewers regarding the need to frame this study mechanistically before diving in to demonstrate relations to aerosols. Thus, the paper and its general focus are much better now. I understand the utility of the added case studies in addressing other reviewer concerns, even though I did not personally find them to add greatly to the paper. The introduction regarding nucleation processes still needs a little work in my opinion, and hence I make a few suggestions for consideration in laying out the basis for thinking about how the high ice concentrations could ensue. There is no discussion at all of homogeneous freezing, how it could come about, and whether composition really matters, preferring still to focus on heterogeneous nucleation. Instead, as in the response to reviews, the authors prefer to call them all ice-forming particles. That is not accurate. Some particles trigger ice formation, while ice formation ensues in others due to an ice formation process that cares very little about the particle except that there is dilute water in it. One reason that poor INPs such as biomass burning particles and urban pollution can impact ice formation in these clouds so strongly is most easily argued to be a process that is less deferential in selecting them for freezing (i.e., not due to the special properties of the particles). Excepting the regions influenced by mineral dusts, where one could plausibly say that thousands per liter might activate before the onset of homogeneous freezing as droplets are lofted, one apparently only needs high aerosol numbers and strong updrafts in order to promote more ice formation in the presence of more particles. This could be a working hypothesis for anyone pursuing this topic further, after reading this paper. For example, the argument now made is that sea salt can be ignored on a number basis alone. Well, there a second point made about its preferential liquid-phase scavenging in lower cloud regions, an argument that appears to ignore the role of cloud dynamics on impacting CCN activation and scavenging by any aerosol entering deep convection. In general, the paper lacked an expressed appreciation for how dynamics can overcome restrictions on CCN activation. Under strong updrafts, one could easily posit that chemistry as a player in CCN hygroscopicity is likely irrelevant, and that is why BB and urban aerosols become important for anvil microphysics over the tropics. If this had been considered, Fan et al. (Science 359, 411–418, 2018) might have been referenced as a case for even small pollution particles as likely CCN that can freeze in upper cloud regions, due to cloud dynamics and the role of coalescence in driving supersaturations in elevated cloud regions. But I am not trying to rewrite the paper, only state what is apparent to a reader.
Recommend publication with very minor revisions.
A few more specific comments
Line 21: Acronym SOFT-IO not defined in abstract.
Lines 30-31: Heterogeneous and homogeneous freezing. Saying both would make it clearer that both are likely involved.
Lines 81 paragraph: It is a little odd to start this paragraph this way, since you have already mentioned heterogeneous and homogeneous freezing above this point. It would be more appropriate to speak to both mechanisms then at this point. The paragraph only mentions heterogeneous INP concentrations and ice crystal concentrations, ending presently with a very nice statement about that. But why not say that homogeneous freezing would be driven by strong updrafts that send condensed water not already frozen by limited heterogeneous freezing or limited consumption by ice growth into the regime where remaining drops can freeze? Noting the role of updraft on activating particles pre- and post-coalescence could add some context as to why this mechanism might be particularly powerful in affecting upper anvil ice concentrations.
Lines 105-107: Not a nuanced discussion. Would one not expect some major fraction of all particles to be available as CCN from combustion sources, but some more limited fraction available as INPs prior to onset of homogeneous freezing conditions? I say this only because it otherwise sounds like very little is known, but that is not the case.
Line 125-127: Contrast the above discussion with the apparent need to state that rBC particles are coated. This is probably a nearly irrelevant factor since supersaturations could be high in deep convective updrafts (especially following initial coalescence). Also, is it not the case that rBC is but a small fraction of all biomass burning particles, all of which are available to act as CCN? You later rule out sea salt for similar number-based arguments.
Line 230: “Concentrations” or “Mixing ratios”?
Lines 285-292: The statements made on not considering sea salt remains speculative. It will “likely” be less of a factor would solve this for now.
Line 317: 0.01 cm-3 as a lower bound on ice concentrations. I thought the lower limit of detection used in this study was 50 per liter?
Lines 414-415: Dust acting as good INPs by “deposition” of water vapor to their surfaces? If all of the cirrus discussed were of liquid-origin (also stated as a conclusion on lines 584-585), then why mention a mechanism that is unrelated to liquid droplet activation?.
Lines 440-441: This repeats the same point about dust, but is better-spoken here. Could say this just once, in one spot or the other.
Line 454: Second mention of sulfate “mixed with dust”, but it is unclear why this is at all important, or if it is important (which I doubt, if CCN activation is the concern – i.e., big dust > smaller CCN for activation under strong forcing, all else being equal and no matter the sulfate).
Line 577: Discussion of ice crystal residuals as “ice nuclei”. You could stop that phrase at their “composition”, since “ice nuclei” infers a component that freezes heterogeneously. It could say ice crystal nuclei, but ice residual nuclei is already stated, and a better way to discuss it.
Lines 587-588: “…cloud chamber and field studies have shown that some fraction of the BB and UP aerosol are hygroscopic and can serve as CCN.” Just how hygroscopic do they really need to be in these circumstances? Would kappa of 0.1 not be sufficient? See Twohy et al. (2021; https://doi.org/10.1029/2021GL094224) on how easily smokes are activated even in modest cumuli.
Line 664: Indeed, any boundary layer aerosols are lofted by strong convergence, and hence, one expects pollution, smoke, dust, and even sea spray particles (primary and secondary formed ones) to influence convective cirrus that dominate in the tropics. This paper confirms that.
The basic data set and analysis in this paper are very welcome additions to the literature, and are worth publishing in some final form. I enjoyed reading it, and I expect that others will. Nevertheless, I sense that the paper began as an investigation of aerosol effects on extreme ice events, as evidenced by the title, and also that it began with the notion that biomass and fossil fuel combustion particles were influencing these, but that the data do not back this up in more than a qualitative sense at best. Conflicting statements then remain in the text regarding the assured role of different aerosol types in affecting non-EIE and EIE, and the ice nucleation mechanisms are not at all resolved. Further, it only becomes evident later in the paper (obvious to the reader, and then finally stated) that there is really no distinguishing factor between EIE and non-EIE events from the aerosol standpoint, and what is most driving EIE is deep convection. Lots of strong updrafts lofting aerosols to low temperatures, potentially as liquid through homogeneous freezing conditions. Do the base aerosols even matter? The paper is still useful for framing and asking the questions at the end, but I suggest that the authors need to work a little harder at demonstrating such a conclusion, and using the provocative title that implies causal links (and will be referenced for such). The relation with aerosols is interesting but never proven to be causal. The remarkable consistency between regions and seasons in Fig. 2 does not to my mind speak to an obvious aerosol effect, and especially not a link to major cities. Hence, one wonders about whether plots of CO or delta-CO versus ice concentrations are not shown. All else being equal, would a relationship not be expected? And what would be expected if the aerosol effect were related to either heterogeneous versus homogeneous freezing nucleation? Would dust influences be distinguished from the others? The anomaly plots are not all especially revealing about what is driving things.
Hence, I suggest the need to:
1) Describe in the introduction how aerosols might influence cirrus concentrations via heterogeneous AND homogeneous freezing.
2) Discuss the complicating role of cloud dynamics and convection, and how the nature of the cirrus targeted (if all expected to be liquid-formed) matters. This and the first suggestion add context to the study, instead of diving immediately into aerosols as the only influence.
3) Consider if some truly quantitative analyses of relations between aerosols/gases and ice concentrations are possible, instead of only associations of fires, smoke and pollution areas with areas of EIE.
Finally, perhaps not for this paper, but would not a consideration of extratropical areas add to the understanding by taking away the immense role of tropical convection in driving EIE? I have a number of selected smaller comments, some related to these major ones.
Line 25: Using the term ice-forming aerosols is not exact, in that the aerosols may or may not be directly linked to the freezing mechanism. If homogeneous freezing, the factor of importance is simply that the particles carry liquid with them. If heterogeneous freezing, the nature of the particles truly matters. Perhaps, lofting aerosols that directly or indirectly lead to freezing, or better serve as seeds for heterogeneous and homogeneous freezing nucleation?
Lines 28-29: Why only heterogeneously if the cold clouds are of liquid origin? There would be a competition between heterogeneous freezing and homogeneous freezing, and what wins at cloud top will be determined by both cloud dynamics (updraft) supplying supersaturation and the propensity of particles for freezing heterogeneously and growing prior to the point where homogeneous freezing will ensue. Where would 5000 per liter INPs come from prior to -38°C? And what concentrations would be necessary freezing prior to that temperature to defeat water persistence to -38C that could then lead to further massive freezing? Consider the observations of Rosenfeld and Woodley (2000) in this regard. Deep convective clouds readily overcome the relatively low numbers of INP from the boundary layer.
Lines 81-84 paragraph: In reference to the point above, homogeneous freezing needs mention as a potentially very important process.
Lines 90-92: “Some fraction of particles emitted from biomass and fossil fuel burning will act as CCN or INP especially as they age while lofted to the UT…”. I consider the especially while they age part as not yet strongly demonstrated for the ambient atmosphere. Atmospheric measurements in this regard are not well-represented in the reference list. Recently, both Schill et al. (2020) and Barry et al. (2021) discuss ambient measurements related to biomass burning INPs, and production is mentioned in the latter study. Those measurements directly in plumes should constrain expectations on INP concentrations feasible from biomass burning, at least at temperatures in the mixed-phase regime prior to the homogeneous freezing threshold.
General comments on introduction and Figure 1: EIEs appear to occur in all regions, and quite high values occur even over oceans. Realizing that your focus is on connecting certain sources and EIEs, I wondered if the IAGOS network coverage adds any particular bias. Does the absence of occurrences between Japan and the U.S. indicate a true absence or a limitation of the network? In this regard, I felt it would be helpful to see a supplemental figure of all of the flight paths. Then it would be easier to understand where flying occurred versus where high values were seen. This point about potential bias or absence of coverage is only otherwise brought up late in the paer on lines 412-413, in regard to absence of flights over a region in Africa. Yet, large ocean regions of the Pacific in both hemispheres are missing from assessment, in a region more remote from fire and urban influences. As a second comment, some of the data are from over the maritime continent and other open ocean regions. Is there a reason not to consider sea salt as an aerosol that could affect deep convective clouds? It has been noted as a freezing nucleus at low temperatures in laboratory studies (e.g., Wagner et al., 2018), and was identified in ice residuals in deep convective anvils (Cziczo et al., 2013). When one is dealing with aerosols and ice nucleation, abundance and activation potential are both factors to consider, so I do not see a reason to exclude something in favor of something more abundant like biomass burning particles. This also arises later when the aerosols of “relevance” are mentioned on lines 373-374. It is simply what you chose to focus on.
Line 138: This is an instance of another reference to the heterogeneous ice nucleating properties of the aerosols being the only matter of interest to relate to EIE. Not so.
Section 3.3 and Figure 6: I feel that a better explanation of the meaning of this figure is needed. What is event frequency? Is it any concentrations of ice crystals coinciding with a CO anomaly? I see nothing much distinguishing low-ice and extreme-ice, and the values of the median CO anomalies are extraordinarily low compared to say CO anomalies inside and outside of biomass burning plumes. How does this indicate impact, if at all? Or especially, how does it show that “…frequency distributions do suggest that emissions from UP sources are potentially a larger source of nucleating particles in the ice clouds, in general.“? To me, I interpret this figure to mean that clouds and aerosols will be associated, but there is no smoking gun for any particular aerosol type or its direct involvement in creating EIEs.
Lines 320-321: This inclusion of CCN here may be a nod to homogeneous freezing as a source of ice clouds, but only the INP connection is tendered earlier as a hypothesis. If the different mechanisms are made explicit in the introduction, this will all be resolved.
Line 345: AOD is an integrated measure. You do not know where in the vertical it resides, right? Most often it is in the boundary layer, although I understand that plumes can be elevated. And it seems that the full range of AOD underlies the EIE points. Like fire power and other relations, the correlation is only a spatial one as viewed from above.
Line 375: Fig. SMx? There is no AOD plot of this type in the supplement.
Line 393, paragraph: Not much is said about AOD over parts of Indonesia to Australia, which are striking for the apparent lack of any apparent influence. This is the regions that begs explanation, if it is to be contended that only certain types of particles are associated with EIEs.
Section 4.2: Not intending to beat on a point I raised already in summary, but convection so clearly shows the strongest correlation with EIE, regardless of aerosols. One has to ask for more than association with CO and other tracers in order to claim that any specific aerosol type is making a difference. Regardless, I felt that the convection link came far too late in this paper, and has to raise a question about the appropriateness of the title.
Section 4.4: A similar comment about a summary point. The discussion under this section finally acknowledges the ways that aerosols, ice formation and deep convection interplay, including mention of liquid origin ice processes and homogeneous freezing. Missing still is the fact that there must also be a relation between ice concentration and vertical velocity. And the mechanism will depend on that and on the freezing efficiency of heterogeneous INPs, as mentioned earlier. This should have been discussed up front, rather than alluding to the fact that the mechanism might be via INPs only.
Line 386: Intended reference is missing.
Line 485: “temperature of EIE..”
Barry, K. R., T. C. J. Hill, E. J. T. Levin, C. H. Twohy, K. A. Moore, Z. D. Weller, D. W. Toohey, M. Reeves, T. Campos, R. Geiss, E. V. Fischer, S. M. Kreidenweis, and P. J. DeMott, 2021: Observations of ice nucleating particles in the free troposphere from western US wildfires. Journal of Geophysical Research: Atmospheres, 126, e2020JD033752. https://doi.org/10.1029/2020JD033752.
Cziczo, D. J.; Froyd, K. D.; Hoose, C.; Jensen, E. J.; Diao, M.; Zondlo, M. A.; Smith, J. B.; Twohy, C. H.; Murphy, D. M. Clarifying the Dominant Sources and Mechanisms of Cirrus Cloud Formation. Science (80-. ). 2013, 340 (6138), 1320–1324. https://doi.org/10.1126/science.1234145.
Rosenfeld, D., Woodley, W. Deep convective clouds with sustained supercooled liquid water down to -37.5 °C. Nature 405, 440–442 (2000). https://doi.org/10.1038/35013030
Schill, G. P., S. H. Jathar, J. K. Krodos, E. J. T. Levin, A. M. Galang, B. Friedman, D . K. Farmer, J. R. Pierce, S. M. Kreidenweis, and P. J. DeMott, 2016: Ice nucleating particle emissions from photo-chemically-aged diesel and biodiesel exhaust, Geophys. Res. Lett., 43 (10), 5524-5531, doi:10.1002/2016GL069529.
Wagner, R.; Kaufmann, J.; Möhler, O.; Saathoff, H.; Schnaiter, M.; Ullrich, R.; Leisner, T. Heterogeneous Ice Nucleation Ability of NaCl and Sea Salt Aerosol Particles at Cirrus Temperatures. J. Geophys. Res. Atmos. 2018, 123 (5), 2841–2860. https://doi.org/10.1002/2017JD027864.