Reviewed by Minghui Diao

Review of Huang et al. Microphysical processes producing high ice water contents (HIWCs) in tropical convective clouds during the HAIC-HIWC field campaign: dominant role of secondary ice production.

In this work, the authors use observations from the High Altitude Ice Crystals (HAIC)-HIWC experiment to evaluate WRF model simulations based on the P3 cloud microphysics scheme. Specifically, model performance for representing high ice crystal number concentrations (Nice) observed during convective activity is examined. Several model setups are investigated, including the default P3 microphysics scheme, P3 scheme with various horizontal grid spacings, and P3 scheme that includes several secondary ice production (SIP) processes, such as (1) the rime-splintering or Hallett–Mossop (HM) process, (2) ice–ice collision fragmentation (IICB), (3) raindrop freezing breakup (RFZB). The results show that the P3 scheme that includes all three SIP mechanisms (HM, RFZB, and IICB) provides the most comparable results to the observed Nice / ice water content (IWC) ratio and radar reflectivity. The dominant SIP mechanism at various temperature ranges is also evaluated.

Overall, the manuscript is well written. The model experimental design is straightforward and easy to follow. The reviewer recommends the manuscript being considered for publication after making the following revisions.

Major comments:

(1) The sensitivity tests conducted for WRF mainly focus on the inclusion of various SIP processes. There are other factors that potentially affect the distribution of IWC and Nice significantly, such as the threshold of relative humidity with respect to ice (RHice) used to initiate ice nucleation. As far as the reviewer knows, P3 scheme still uses the ice nucleation formulation of Cooper (1986), similar to Morrison double moment scheme. If so, the reviewer thinks that the P3 scheme may be subject to similar problems seen in previous WRF simulations of convective systems.

Previously, D'Alessandro et al. (2017) evaluated several double moment schemes (i.e., Morrison 2-moment, Thompson 2-moment, and Thompson-Eidhammer aerosol aware scheme) in WRF model, and Diao et al. (2017) evaluated the Morrison 2-moment scheme in the NCAR CM1 cloud-resolving model (similar to WRF model). Both studies showed that the Cooper parameterization used to initiate ice nucleation has a RHice threshold that is too low (D'Alessandro et al., 2017, doi:10.1002/2016JD025994; Diao et al., 2017, doi:10.1175/JAS-D-16-0356.1). When the Cooper parameterization is used in Morrison 2-moment, it does not allow clear-sky ice supersaturation to exceed 108%. When comparing WRF simulations against aircraft observations of anvil clouds during the NSF DC3 field campaign, such activation of ice nucleation at 108% leads to an underestimation of the occurrence frequency of ice supersaturation in both clear-sky and in-cloud conditions. Consequently, it allows ice nucleation to happen too early when RHice is still relatively low and leads to higher ice water content and ice crystal number concentrations than those seen in the observations.

The reviewer suggests testing the activation of Cooper 1986 parameterization at a higher ice supersaturation threshold (such as RHice = 125% or 130%). This would allow more clear-sky ice supersaturation to exist without turning into ice crystals too early. It would be interesting to see the impacts of such revision compared with the inclusion of SIP processes.

(2) Following the first comment, evaluations of thermodynamic conditions (such as RHice) would be helpful besides the evaluation of Nice and IWC in relation to vertical velocity. It would be valuable to show the distributions of ice microphysical properties in relation to RHice with or without SIP processes included as well as how these simulated distributions compare with observations.

(3) The current study includes three SIP mechanisms, while previous literature pointed out other existing SIP mechanisms as well. If the current three SIP mechanisms already provide a similar amount of Nice as the observed value, wouldn’t adding other SIP mechanisms in the future lead to Nice that is too high compared with observations? The reviewer suggests adding some discussions on this potential problem when more SIP processes are included in the WRF model.

(4) Since Nice is a key observed variable used to evaluate WRF simulations, what is the range of uncertainty associated with Nice, as it is affected by various potential problems such as shattering and poorly defined depth of field? Can the authors give a range of observed Nice at various temperatures by using a more rigorous versus a less rigorous quality control procedure?

(5) Can the authors elaborate on how liquid droplets or raindrops are separated from ice crystals in 2DS and PIP measurements? The paper of Huang et al. (2021) briefly mentioned that “The two optical array probes, 2D-S and PIP, recorded 2D images of ice crystals nominally in the size range of 10–1280 and 100–6400 µm, respectively.” But this statement is not entirely true because 2D-S and PIP can capture liquid phase as well.

(6) The analysis uses the size range (Dmax) of 0.1 to 12.845 mm, which is slightly higher than the range of 0.05 – 12.845 mm used in the companion paper (Huang et al., 2021). Is there a reason that a higher minimum threshold of Dmax (0.1 mm) is used in this work?

In addition, is the model output of ice microphysical properties (e.g., IWC, Nice, mass-mean diameter, etc.) re-calculated based on this partial size range in order to match the size range of the observations? A partial size selection procedure has been applied to model output as shown in previous studies, e.g., Fridlind et al. (2007), Eidhammer et al. (2014), Patnaude et al. (2021) https://doi.org/10.5194/acp-21-1835-2021, Yang et al. (2020) DOI: 10.1002/essoar.10504450.1.

Since the smaller ice particles dominate the Nice value, if the authors only evaluate the range of 0-0.05 mm (given that the observations at this range have large uncertainties), do the simulations show too many or too few ice particles at this size range? This size range (0-0.05 mm) is quite important because if the simulations already provide several orders of magnitude of higher Nice than observed Nice at 0-0.05 mm, getting a more similar Nice for > 0.05 mm by adding SIP processes can potentially lead to worse results when considering the entire size range of ice crystals. Alternatively, if the default P3 simulations provide smaller Nice at 0-0.05 mm compared with observations, it can be interpreted as either the observed Nice is overestimated due to shattering, or the default P3 simulations really underestimate Nice even at small ice crystal sizes.

Minor comments:

Line 19 – 22, some of the temperature ranges do not have the complete range. See suggestions in brackets: “… shattering of freezing droplets dominates ice particle production in HIWC regions at temperatures > −15C [temperatures between -15C and 0C??] during the early stage of convection, and fragmentation during ice–ice collisions dominates at temperatures > −15C [temperatures between ‑15C and 0C??]  during the later stage of convection and at temperatures < −20C [temperatures between ‑40C and -20C??] over the whole convection period.”

Line 218, “... interpolated to temperatures of -10, -30, …”, can the author explain what kind of interpolation was done?

Figures 6 and 7 have two colors that are very similar in the legend, that is, HM in red and Others in fuchsia. Even though the colors are readable from a computer screen, when printed out on paper these two colors look almost the same. The reviewer suggests changing “Others” to some other colors, such as green, yellow or light gray.

Review of “Microphysical processes producing high ice water contents (HIWCs) in tropical convective clouds during the HAIC-HIWC field campaign: dominant role of secondary ice production” by Huang et al.

Overview

In this study, simulations of tropical deep convective clouds performed with the Weather Research and Forecasting (WRF) model are evaluated against observations from the High Altitude Ice Crystals (HAIC)-HIWC experiment. As a companion paper to Huang et al., 2021, a closer look on the role of secondary ice production (SIP) to generate observed HIWC regions is taken. Three SIP mechanisms are investigated: the Hallett–Mossop (H-M) process, ice–ice collision fragmentation (IICB) and raindrop freezing breakup (RFZB). It is found that simulations including all three SIP processes successfully produces HIWC regions in all three temperature levels that were investigated. The results highlight the importance of SIP processes in controlling the ice water content in the studied tropical convective clouds. The paper is well written and the model experiments are easy to follow. There are only minor shortcomings that should be addressed before publication. First, the ice crystal observations would have deserved more discussion and ideally, the authors could have included their best estimation of the magnitude of the uncertainties related to these observations. Furthermore, the uncertainties related to the existing SIP parameterisations should be highlighted more. After these minor revisions the manuscript is recommended for publication.

Recommendation: It is recommended that this manuscript is accepted after minor revisions. 

General comments

1. For evaluation of the role of SIP it is crucial to have reliable observations of ice crystal number concentrations (Ni). However, the authors do not discuss the Ni observations during the HAIC-HIWC field campaign or, more importantly, their uncertainties. Although Huang et al., 2021 contains information of cloud microphysical observations, the relevant measurement methods should be summarised also in this manuscript. For example, Huang et al. (2021) states that the Ni measurements were derived from the Two Dimensional Stereo Imaging Probe (2D-S) and the Precipitation Imaging Probe (PIP) fro the size range for Dmax>50 µm but in this manuscript use a higher lower size limit of 100 µm. This choice should be discussed.
2. How is the model sampled in order to get Ni and IWC values in the size range of 0.1-12.845 mm? How is the ice particle size defined in the model? What is the possible error in Ni if the model and observations have a different definition for size?
3. The sensitivity of the model to horizontal resolution and aerosol profile is discussed in Sec. 3.2. The results are discussed in terms of Ni/IWC values but it is not well justified why the 250m-resolution model was chosen for the sensitivity studies including SIP processes. 
4. Fragmentation of ice in ice-ice collisions is shown to dominate the ice particle production rates outside the updraft core region even at temperatures warmer than -15°C. Do the authors consider this as a realistic outcome or a result of the way ice-ice collisions were implemented in the model?
   
   Laboratory studies suggest that the number of ice ejected in collisions has a strong dependence of temperature with a maximum around -18°C (Takahashi, Nagao & Kushiyama, 1995). The break-up of ice crystals in ice-ice collisions is explained by the different ice crystal surface properties (brittle surface with plate-like growth or dendrites colliding with compact graupel). Plate-like and dendritic growth contributing to fragmentation is taking place in the temperature region between -15°C and -20°C, which explains the observed maximum in ice production rate in the laboratory studies. Ice crystals around -5 and -10°C have more compact columnar shapes. Is this difference in shapes taken into account in the parameterisation?
5. It is important to state that there are severe uncertainties in the parameterisations of SIP mechanisms. Although there is a statement about this in Summary and Conclusions, this could be also stated earlier in the manuscript (e.g. in Sec. 2.1). Ideally, some additional sensitivity tests by tuning the different SIP parameterisations could have been performed.

Minor comments

Line 203: Can the authors explain the choice to discuss the results in the form of Ni/IWC? What is the additional value in this representation?

Figure 4: Can the observations be added to the corresponding subplots or even combine Figures 3 and 4?

Figure 5: Same as for Fig. 4. It would be helpful to have the observations included. In addition, please explain the acronyms HM, RFZB, IICB and SIPs in the figure caption to improve readability.

Line 256: What kind of significance test was performed? Please add this information.

Line 345: Some discussion of how generating more ice in the early or lower parts of the cloud though SIP will lead to HIWC regions with more small ice crystal at colder and higher part of the cloud would be helpful to visualise the dynamics of these systems. 

References

Takahashi, T., Nagao, Y., & Kushiyama, Y. (1995). Possible high ice particle production during graupel-graupel collisions. Journal of the Atmospheric Sciences. https://doi.org/10.1175/1520-0469(1995)052<4523:PHIPPD>2.0.CO;2