the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Aug 2025
Influence of secondary ice formation on tropical deep convective clouds simulated by the Unified Model
Abstract. Secondary ice production (SIP) plays an important role in tropical deep convection, yet its representation in models remains uncertain. This study incorporates multiple SIP mechanisms, including droplet fragmentation (Mode 1 and Mode 2) and ice–ice collisional breakup, into the CASIM microphysics scheme of the UK Met Office Unified Model, and evaluates their impacts through a real-case simulation of a Hector thunderstorm. SIP enhances ice number concentration in upper cloud layers, with values up to 3 orders of magnitude higher than the no-SIP case, particularly above −10 °C. Ice water content (IWC) increases by a factor of 3–5 in the anvil region, contributing to more extensive upper-level cloud coverage. These microphysical changes reduce outgoing longwave radiation (OLR) by ~3.2 W m−2 (1.3 %) and increase outgoing shortwave radiation (OSR) by ~4.5 W m−2 (1.8 %) over a 6-hour analysis period and a 110 km × 110 km domain. SIP modifies precipitation structure, enhancing local rainfall near the convective core while reducing domain-averaged precipitation by ~8 %. Peak rainfall rates remain only slightly affected, consistent with the minor changes (<1 m s−1) in maximum updraft velocity. Among the tested mechanisms, ice–ice collisional breakup shows negligible impact under warm, graupel-sparse tropical conditions. Ensemble experiments confirm that these effects are robust and exceed the influence of meteorological variability. These results highlight the importance of representing SIP processes in cloud-resolving models of tropical convection and accounting for their environmental dependence.
- Preprint
(2200 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 03 Oct 2025)
-
RC1: 'Comment on egusphere-2025-3158', Anonymous Referee #1, 08 Sep 2025
reply
Secondary ice production (SIP) is an uncertain process in different cloud types including the tropical deep convection. Incorporating multiple SIP mechanisms into the CASIM microphysics scheme, the results of a Hector thunderstorm show that SIP greatly increases ice number concentrations and ice water content, expands anvil cloud coverage, and modifies both radiation and rainfall patterns. These results highlight the need to represent SIP in cloud-resolving models to better capture tropical convection and its climatic impacts. As a strong positive, I appreciate the use of small modeling ensembles, which adds robustness to the results. The simulated cases are also well presented and compared with observations. However, the analysis remains somewhat limited and would benefit from greater depth. I recommend strengthening the analysis before the manuscript can be considered for publication.
Overall comments
In analyzing the mean values of different cloud properties, have you considered that the simulations may encompass different cloud volumes? Conditional sampling could introduce biases if only mean values are compared. In addition, the choice of modeling framework, along with the level of microphysical detail and spatial resolution, may strongly influence the results. These aspects should be discussed in greater depth.
Specific comments
Line 16: “…including droplet fragmentation (Mode 1 and Mode 2)…” Referring to the modes here is a technical detail that is not widely known. Please either provide a brief explanation of these modes or omit mentioning them from the Abstract.
Lines 74–75: Is the effect limited to increased ice loading, or is it possible that the anvil also extends over a larger area?
Section 2.2: A number of equations are presented, many of which appear to originate from earlier publications. If these are identical to previous formulations, please explain why they are repeated here. There is also an option to move these into supplementary material.
Lines 306–307: “…inclusion of secondary ice production leads to a reduction in mean reflectivity values in middle levels. This is mainly attributed to the smaller ice particles aloft.” What exactly does “mainly” mean here? Which hydrometeor categories are responsible for the change? Do all categories show decreased size and increased number, or is the effect limited to specific ones? Even if SIP generates new particles, those originating from primary freezing should still grow almost as fast as those without SIP if ice–ice collisional breakup is inefficient. Please clarify.
Lines 312–314: “Below 3 km, the model underestimates the frequency of reflectivity values exceeding 5 dBZ…” Does this mean that raindrops are evaporating too rapidly or are not large enough in the beginning, or could the discrepancy also stem from how reflectivity is calculated and conditionally sampled?
Lines 356–359: “The convective core is less pronounced…” Is it possible to get some statistics to support this. Visually this is not too evident.
Figure 8: How large is the uncertainty in the “observed” precipitation data?
Lines 485–486: “Ice–ice collisional breakup remains largely insignificant…” Why is the discussion restricted primarily to graupel? The process involves all ice hydrometeors. Given that the all-SIP simulation produces the highest ice concentrations, how can you rule out a contribution from collisional breakup after other processes increase concentrations?
Figure 10: Is the amount of snow reduced because of SIP? Not evident from the ice particle concentrations in Figure 9. Can this affect the outcome, for example related to changes in reflectivity?
Figure 10 (Maximum updraft velocity): Could you compare the simulated maximum updraft velocities with those from higher-resolution studies, e.g., large-eddy simulations by Dauhut et al. (2015)? The values here appear smaller than might be expected, which could have implications for SIP efficiency and overall precipitation formation.
Lines 569–570: “SIP diverts condensate away from warm-rain processes into less efficient ice-phase pathways.” Please clarify this statement. Ice particles typically grow faster than liquid droplets, so in what sense is the pathway “less efficient”?
Line 578: “…the updraft velocity (Wmax) increases by ~10%… and does not substantially intensify peak convection.” A 10% change in cloud dynamics due to microphysical processes could be considered significant. What threshold would you regard as “substantial”?
References
Dauhut, T., Chaboureau, J.-P., Escobar, J. and Mascart, P. (2015), Large-eddy simulations of Hector the convector making the stratosphere wetter. Atmos. Sci. Lett., 16: 135-140. https://doi.org/10.1002/asl2.534
Citation: https://doi.org/10.5194/egusphere-2025-3158-RC1 -
RC2: 'Comment on egusphere-2025-3158', Anonymous Referee #2, 10 Sep 2025
reply
The paper numerically investigates the possible role of secondary ice production (SIP) in forming overall ice number concentration, and its influence on cloud properties in a convective case observed during ACTIVE campaign in early December 2005. For this, the authors incorporated three SIP processes (Hallett-Mossop rime-splintering, ice-ice collision, breakup of freezing raindrop) in UK Met Office Unified Model’s 2-moment CASIM microphysics scheme. The study finds that, through increased ice number and mass in the upper region, SIP can modify the anvil structure in the simulated thunderstorm, and changes the precipitation formation, especially associated with the convective core. The ensemble simulations are also performed that illustrates the robustness of the presented results. The introduction is comprehensive, and sufficiently discusses the recent advancements and challenges in SIP research, and the methodology is sound. The simulated properties, such as the radar reflectivity, OLR, and precipitation are compared and validated against the observations. While the overall presentation is good, the paper could benefit from a more detailed discussion and comparison of how the simulated microphysical properties agrees well with the observations. Nevertheless, the study addresses a timely and important topic and may be considered for publication after satisfactorily addressing the following concerns.
General comments
More details of radiation and microphysics scheme used are needed. The presented validation is reasonable but could have benefited from additional comparison, such as observed liquid/ice properties, if such observations (from satellite or other platforms) are available. Also, many findings presented in the manuscript can be supported by some previous studies, and can also be acknowledged in the introduction section to strengthen the proposed research questions.
Specific comments
Abstract
Line 14: Since the study mainly quantifies the impacts of SIP on the simulated clouds without attempting to modify/improve the existing SIP parameterizations (beyond the use of revised Ф from James et al. 2021), I would suggest rephrasing this sentence, as the focus is not on quantifying and reducing uncertainties in the modelled SIP processes.
Line 16: Mode 1 and 2 of drop shattering are scheme-specific terms and are not widely recognized. These are proposed by Phillips et al. (2018) to represent drop shattering in collision between rain/drizzle drop with ice particle. Other schemes of SIP in drop shattering (e.g., Sullivan et al. 2018) only consider shattering of raindrop during freezing, initiated due to immersed INP, without separating mode 1 and 2. Better to omit using mode 1 and mode 2 and say only ‘drop fragmentation’.
Line 25: change <1 to < 1.
Line 26: Not sure about the context of this sentence. On which process/property ice-ice collisional breakup has negligible impact? On ice concentration or dynamics?
Introduction
Lines 34-35: Citing more recent studies of observational evidences of SIP would be beneficial (e.g., Korolev et al. 2022).
Lines 35-43: Where the term SIP is introduced, please mention the region (mixed-phase) where it mainly occurs in clouds.
Lines 39-41: Cite previous literature (e.g., Lohmann et al.; Kudzotsa et al. 2016; Han et al. 2024; Waman et al. 2025) supporting this.
Lines 41-43: I do not see that the manuscript attempt to improve the representation of SIP processes in numerical model. Rather, the effect of SIP processes is quantified in deep convective clouds using existing parameterizations. Please rephrase.
Line 50: Can the authors comment and acknowledge findings of recent study by Seidel et al. 2024, which see no experimental evidence of rime-splintering, especially in convective conditions. Considering the findings of Seidel et al. 2024, what is the relevance rime-splintering process and its existing parameterization in representing SIP at such warmer subzero levels?
Line 57: Please cite relevant previous studies that used Unified framework to study SIP.
Line 64: Waman et al. 2022 do not explicitly quantify the impact of SIP on the mentioned ice growth processes. Please correct/clarify more.
Methodology
Line 113: ‘are’ instead of ‘is’?
How are cloud droplets activated in CASIM? Does the scheme explicitly account for the activity of soluble aerosols as CCN, or is the CCN spectrum prescribed from observations? What is the nature of aerosols (continental/marine)?
Line 115: I believe with Cooper, only immersion mode of heterogeneous ice nucleation is represented. Can the authors clarify how other heterogeneous ice nucleation modes (e.g., deposition), that can be crucial at colder temperatures, are treated in CASIM? Additionally, is homogeneous freezing of aqueous aerosols represented separately from homogeneous droplet freezing, and if not, what are the implications for ice formation at cirrus temperatures?
Line 116: ‘rime-splintering’ instead of ‘riming splintering’?
Line 118: ‘other newly implemented SIP processes’ instead?
Line 135: What is the value of rime fraction used to represent ice-ice collision in Eq. 3? Also, can the authors comment on how the re-fitted values of the parameters in Eq. 3, given by Grzegorczyk et al. 2023 (Table 3) would influence the predictions from ice-ice collisional breakup?
Eq. 6: should be ?
General comment: Overall, the considered SIP processes are described adequately. However, the study does not appear to consider SIP during sublimation of ice particles in subsaturated cloudy environments (e.g., Deshmukh et al., 2022; also see Korolev and Leisner 2020 for limitations). While I understand this mechanism is still under active investigation, it could be relevant in tropical anvil outflow/downdraft regions (Waman et al. 2022). Could the authors briefly comment on the reason for excluding this process and its possible implications for the presented results? Also, adding a table of symbols used would be helpful.
Line 217-218: I do not understand what really makes Hector as an ideal case for studying SIP.
Line 221: Would be nice to mention cloud-base (LCL) and cloud top from Fig. 2.
Results
Line 291: space between ‘(CFAD)’ and ‘of’.
Line 306: ‘SIP’ instead of ‘secondary ice production’?
Figure 4: For better comparison, is it possible to show isotherms also in (a)? Also show in the form of text ‘0oC’, ‘-20oC’, ‘-40oC’, ‘-60oC’ (The same can be followed for Figs. 9 and 10).
Figure 5: Also mention date (1 December 2005) in the caption.
Lines 351-356: Can the authors comment on how well the model captures the observed surface precipitation? Overall, I see that the model significantly underpredicts the surface precipitation, both in all-SIP and no-SIP experiments.
Line 356: ‘diffused’ instead of ‘diffuse’?
Line 357: ‘The convective core is less pronounced…’ This is not clear. How is it interpreted? Additional analysis would be helpful to support this.
Line 358-359: This is also not quite clear. How all-SIP shows a more localized and organized precipitation? Also, I do not agree that the all-SIP experiment resembles the overall observed convective core as the simulations substantially fails to capture the observed precipitation features.
Figure 6d: is this all-SIP minus no-SIP? Mention clearly in the caption.
Line 366: Cite Figure 6d.
Line 367: I do not see these features; can the authors describe this more? How exactly no-SIP shows more evenly distributed precipitation than all-SIP?
Line 369: ‘realistic reproduction’: Both all-SIP and no-SIP rather captures more localized precipitation events in the simulated domain and not the overall precipitation. Please rewrite as precipitation differ significantly between the simulations and observation.
Line 371: What does ‘focus’ mean here? How convective rain(fall) is identified?
Line 377: ‘increase’ instead of ‘increases’?
Line 378: Why the mode 2 results in more pronounced convective core?
Lines 366-405: Can the authors explain briefly in the manuscript what possible factors SIP alters that result in the predicted change in precipitation? Although paragraph (lines 405-413) explains the overall influence of a combination of various SIP processes, the exact cause in each case (in Fig. 7) is not discussed.
Lines 457-460: Although mode 2 is less efficient and more confined than mode 1, for what possible reasons does RS+M2 produces more precipitation (Fig. 7e)?
Line 470: ‘increase’ instead of ‘increases’?
Line 473-474: A suggestion: Time-height maps of total ice concentrations in all-SIP and no-SIP, and a similar difference plot (all-SIP minus no-SIP) would be helpful to visualize increased extensiveness over longer period in all-SIP case.
Line 532: Previous work by Qu et al. (2022) and Grzegorczyk et al. (2025) can be cited here.
Line 569-570: This needs more clarification, as both warm and cold rain processes can happen simultaneously at subzero levels.
References
Korolev, A., DeMott, P.J., Heckman, I., Wolde, M., Williams, E., Smalley, D.J. and Donovan, M.F., 2022. Observation of secondary ice production in clouds at low temperatures. Atmospheric Chemistry and Physics, 22(19), pp.13103-13113.
Lohmann, U., 2006. Aerosol effects on clouds and climate. Space Science Reviews, 125(1), pp.129-137.
Kudzotsa, I., Phillips, V.T. and Dobbie, S., 2016. Aerosol indirect effects on glaciated clouds. Part 2: Sensitivity tests using solute aerosols. Quarterly Journal of the Royal Meteorological Society, 142(698), pp.1970-1981.
Waman, D., Jadav, A., Patade, S., Gautam, M., Deshmukh, A. and Phillips, V., 2025. Mechanisms for Indirect Effects from Ice Nucleating Particles on Continental Clouds and Radiation. Journal of the Atmospheric Sciences.
Seidel, J.S., Kiselev, A.A., Keinert, A., Stratmann, F., Leisner, T. and Hartmann, S., 2024. Secondary ice production–no evidence of efficient rime-splintering mechanism. Atmospheric Chemistry and Physics, 24(9), pp.5247-5263.
Korolev, A. and Leisner, T., 2020. Review of experimental studies of secondary ice production. Atmospheric Chemistry and Physics, 20(20), pp.11767-11797.
Grzegorczyk, P., Yadav, S., Zanger, F., Theis, A., Mitra, S.K., Borrmann, S. and Szakáll, M., 2023. Fragmentation of ice particles: laboratory experiments on graupel–graupel and graupel–snowflake collisions. Atmospheric Chemistry and Physics, 23(20), pp.13505-13521.
Grzegorczyk, P., Wobrock, W., Canzi, A., Niquet, L., Tridon, F., and Planche, C. (2025). Investigating secondary ice production in a deep convective cloud with a 3d bin microphysics model: Part II - effects on the cloud formation and development. Proceedings of the National Academy of Sciences, 314.
Qu, Z., Korolev, A., Milbrandt, J., Heckman, I., Huang, Y., McFarquhar, G., Morrison, H., Wold, M., and Nguyen, C. (2022). The impacts of secondary ice production on microphysics and dynamics in tropical convection. Atmos. Chem. Phys., 22(18):12287–12310.
-
RC3: 'Comment on egusphere-2025-3158', Pierre Grzegorczyk, 12 Sep 2025
reply
This study investigates the influence of secondary ice production (SIP) on a tropical deep convective cloud (Hector) using the 2-moment microphysics scheme (CASIM) of the UK Met Office Unified model. The paper highlights significant effect of SIP on important cloud properties such as the ice particles number concentration, ice water content, precipitation amount and repartition as well as radiative properties of the cloud system. Overall, the paper is well written, and I enjoyed reading it. The results bring interesting and important results about the importance of SIP mechanisms. I especially liked the ensemble simulation performed in the study which gives even more robustness to the numerical sensitivity tests. I have a few questions, remarks and suggestions about the results and their interpretation. I think that my comments are more suggestive and can considered to be relatively minor. The current version is, in my opinion, almost ready for a final publication even if some specific points can be strengthened and clarified.
Comments:
Line 51-52: I would argue the opposite, supercooled droplets and graupel are common in convective updrafts which favor strong supersaturations and transport supercooled liquid droplets. Evidence of riming in convective regions can be found, for example, in https://doi.org/10.1175/JAS-D-25-0021.1
Intensification of strong precipitation (Line 65-67 and Section 3.2): Our studies (Grzegorczyk et al., 2025b and Grzegorczyk et 2025c, https://doi.org/10.5194/acp-25-10403-2025) show that for two types of convective clouds strong precipitation (>40 mm) is especially reduced by SIP. It may be interesting to mention this to balance your results and those of Sullivan et al. (2018) in the results and discussion section?
Line 131: Is the riming rate obtained in a similar way as for ice-ice collision breakup (Eq. 2) and drop shattering (Eq. 8)? Since HM and DS both depend on collisions between ice particles and drops, do you consider that HM and DS can occur simultaneously when a certain number/mass of drops are freezing by collisions with ice particles?
Paragraph 2.2.3: It is a good point to describe the implementation of SIP in CASIM in detail. I am just wondering what values were considered for the ice-ice and ice-drop collision efficiencies? Additionally, how are the terminal velocities of ice, snow, and graupel calculated? I think it is important since it directly determines the rate of SIP processes.
Line 259-260 and lines 583-596: Great idea to use four ensemble members to make the results more robust.
Line 299-304 for Fig 4: The cloud top altitude seems to be lower in the no SIP case which is quite interesting as the same results were obtained in Qu et al. (2022) (https://doi.org/10.5194/acp-22-12287-2022) while we found the opposite in our study (Grzegorczyk et al. 2025b).
Fig 5: I am not a specialist in radiative transfer, but do you think that the changes in OLR and OSR du to SIP can be important at larger scale for the climate?
Line 406-409: Our study (Grzegorczyk et al (2025c) (https://doi.org/10.5194/acp-25-10403-2025) as well as the one of Dedekind et al. (2021) (https://doi.org/10.5194/acp-21-15115-2021) investigated the reasons for the SIP influence on ground precipitation, which may support your statement.
Fig. 9b: The graupel concentration (1-10 L⁻¹) seems high, and close to ice and snow concentration below 8 km. Even if I am not familiar with the number of graupels in models, I am unsure if this is realistic. Regarding lines 486 and 618-619 I am not sure that the reason for the weak ice-ice breakup is due to the lack of graupel collisions since their concentartion look relatively high. Could the weak ice-ice breakup result from snow-snow or graupel-snow collisions?
Fig 9e: The total increase in ice particle number with SIP seems relatively small (less than one order of magnitude). Is this concentration realistic for a deep convective cloud? The production of ice particles by SIP in CASIM could be maybe validated against in situ aircraft observations in a future study?
Fig 10: I think that it is important to plot the liquid water content and to see the effect of SIP on it. It can further explain differences in OSR of OLR as well as explain why M1 or M2 are strong.
Lines 612-613 and 656-657: In the original Phillips et al. (2017) formulation, ice production from snow breakup is weaker than from graupel collisions. In my opinion snow breakup is underestimated, as snow particles are more fragile than rimed ones. In Grzegorczyk et al. (2025c), I used different values (see the appendix of the paper) based on earlier experiments (Grzegorczyk et al., 2023) for Phillips et al. (2017) formulation. Testing these new values may be beyond the scope of this paper but it would be interesting to see whether it affects your results.
Minor comments:
Line 37: The recent study of Seidel et al. (2024) (https://doi.org/10.5194/acp-24-5247-2024) can be cited.
Line 39: Additional recent papers about ice-ice breakup can also be included: Yadav et al. (2025) (https://doi.org/10.5194/acp-25-8671-2025) and Gautam et al. (2024) (https://doi.org/10.1175/JAS-D-23-0122.1)
Line 41-43: Some references could be used to support that.
Line 85: Some references can maybe be cited for the ‘Hector-type’ cloud.
Line 164-165: I think that the structure of the sentence could be improved to be clearer
Citation: https://doi.org/10.5194/egusphere-2025-3158-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,590 | 23 | 6 | 1,619 | 14 | 19 |
- HTML: 1,590
- PDF: 23
- XML: 6
- Total: 1,619
- BibTeX: 14
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1