Review of De La Torre Castro et al. 2025

This is a well-executed and scientifically valuable study on aerosol-cirrus interactions, focusing on results from the CIRRUS-HL campaign. It investigates why mid-latitude cirrus have higher ice crystal numbers than high-latitude cirrus despite similar aerosol abundances, attributing the differences to specific ice-nucleating particle types rather than total aerosol concentrations. The study highlights the dominant role of mineral dust at mid-latitudes and the potential influence of aviation black carbon at high latitudes, while also addressing model biases in convective ice representation. Overall, it demonstrates how combining observations with model simulations improves understanding and representation of cirrus clouds in climate models.

Recommendation: Major Revision

While the manuscript presents valuable results, several substantive points require clarification or additional detail before publication. These include conceptual clarifications, methodological considerations related to model calibration and evaluation against the same observational dataset, and further details on instrument characteristics, data processing, and figure interpretation. Some statements in the conclusions and discussion could also be more clearly supported by the data or relevant literature. Addressing these points will strengthen the manuscript and improve its clarity and impact.

Major comments:

• A key aspect that requires clarification is whether secondary ice production (SIP) mechanisms (such as rime-splintering, ice-ice collisions, or fragmentation) were considered as potential contributors to the observed cirrus ice number concentrations, in addition to the role of primary ice-nucleating particles. SIP has been shown to substantially increase ice crystal numbers not only in mixed-phase clouds but also in convective outflows that may feed cirrus anvils (Hallett and Mossop, 1974; Field et al., 2006, 2017; Korolev and Field, 2008).
• Even in the upper troposphere, SIP may influence ice crystal populations in cirrus formed from convective detrainment, which is particularly relevant for mid-latitude cirrus where high Ni values were observed during the CIRRUS-HL campaign. Neglecting SIP could lead to an incomplete attribution of the observed differences between mid- and high-latitude cirrus. I strongly recommend the authors explicitly discuss whether SIP processes were included in their analysis or simulations, and if not, acknowledge this limitation and its potential impact on the results.
• The methodology appears to adjust the model using observational data (e.g., for fit1 and fit2) and then evaluates the agreement with the same dataset. This approach risks overestimating model performance, as the apparent improvement may largely reflect the fact that the model was calibrated to these observations. The authors should clarify this point and, if possible, include an independent validation dataset to assess whether the fits genuinely improve predictive skill rather than just reproducing the observations used for adjustment (linked to Section 4.4, lines 479–481).

Minor comments:

• Lines 34-35: The statement that homogeneous nucleation occurs mainly in the uppermost troposphere and lower stratosphere is inaccurate. It primarily occurs in the cold upper troposphere under high ice supersaturation and is generally rare in the lower stratosphere. Please revise for accuracy.
• Lines 40-44: The paragraph beginning with “Understanding and accurately representing aerosol-cloud interactions…” is somewhat disconnected from the previous discussion of nucleation processes and INP uncertainties. Adding a transition sentence linking general nucleation uncertainties to Arctic observations or regional differences in cirrus properties would improve clarity.
• Lines 80-87: The paragraph describing the flights is somewhat ambiguous regarding convective systems. It states that two flights specifically targeted convective systems were excluded from the main analysis (except in Sect. 4.4), but the campaign also experienced generally unstable conditions with thunderstorms. Please clarify whether any other flights encountered convective systems and, if so, explain why they were included in the analysis while the two targeted convective flights were excluded.
• Lines 98-100: The previous paragraph notes that the CDP has a 1-second time resolution, but the time resolution or sampling frequency of the CIPgs and PIP instruments is not provided. Please include this information to allow proper comparison and assessment of the temporal representativeness of the measurements.
• Lines 98-106: The CIPgs and PIP instruments have overlapping particle size ranges. Please clarify how particle data in the overlap region were handled to ensure consistency between the two instruments.
• Lines 105-106: The manuscript mentions that one-pixel images and other artifacts were excluded. Please specify what types of artifacts were identified and removed from the dataset.
• Lines 107-110: The paragraph introduces the mass-dimension relationship and mentions that ice crystal number concentration (Nice) is derived from the combined dataset. Please clarify that D represents the ice particle diameter obtained from the probe measurements. Additionally, explain how Nice was derived from the combined instruments. Was it calculated directly from particle counts, or were corrections applied to account for overlapping size ranges and differences between probes?
• Lines 114-118: The size ranges for the CPC and OPC are provided, but for clarity, please indicate the overall maximum particle diameter measurable by AMETYST.
• Lines 150-153: The authors state that the simulations were performed in nudged mode. Could they discuss the potential impacts or limitations this approach may have on the results?
• Lines 174-175: Please clarify whether SSP3-7.0 was chosen instead of SSP2-4.5 or other scenarios. Was this selection specifically intended to explore a high-emission or strong-forcing pathway?
• Lines 22-211: Please clarify whether the global simulations used for the backward-trajectory analysis are the same as the EMAC + MADE3 model runs described earlier, or if a different model setup was used. Additionally, please provide more details on how clouds were classified as “liquid origin” versus “in situ origin” based on the presence of liquid water along the trajectory, such as how much liquid water, at which levels, and based on what threshold.
• Lines 224-225: I would be cautious with the statement that at lower altitudes, particularly at high latitudes, the median values are associated with larger uncertainties due to limited data availability. Figure 1 shows that this is not consistently the case across all aerosol size ranges. For example, for D > 14 nm the uncertainties appear of similar magnitude across altitudes, and for D > 12 nm the uncertainties at lower altitudes are even smaller than those at higher altitudes.
• Lines 238-240: This result is expected, since the tropopause is generally higher at mid-latitudes than at high latitudes. You may consider making this explicit in the text to clarify why cirrus are found at higher absolute altitudes in mid-latitudes, but at similar relative altitudes with respect to the tropopause.
• Lines 239-240: The authors note sparse high-latitude data below 225 hPa. Please clarify that this is largely due to the lower tropopause at high latitudes, so these levels correspond to the lower stratosphere where cirrus are rare, rather than just a limitation of the dataset.
• Lines 243-245: The statement that high-latitude cirrus contain fewer but larger ice particles appears to be a deduction based on the similar IWC despite lower Nice. Please clarify that this is an inferred conclusion from Fig. 2, or provide explicit evidence from the particle size distribution to support it.
• Lines 308-309: In the text, it is stated that “Climatological averages of these properties for June and July over the period 2014-2021 confirm this trend (see Fig. S2) and also serve as reference to contextualize the specific CIRRUS-HL episode.” However, Fig. S2 actually shows a model climatology, not observational data. It would be helpful to clarify in the main text that the climatological reference is based on model output, averaged over the spatial domain of the flights, rather than on observations. Additionally, it would be valuable to include a comparison with observations from satellite or other observational climatologies for variables such as Nice, IWC, and RHice. This would help place the CIRRUS-HL observations into a broader context.
• Lines 311-313: The manuscript states that “the observed ice properties at mid-latitudes in the cold regime (< 220 K) are mostly well captured in the model,” and that “at the lowest temperature of the measurements (≈ 212 K), the model predicts Nice and RHice comparable to the observations but overestimates IWC.” Since IWC is overestimated across all temperatures and remarkably in mid-latitudes, this seems somewhat contradictory. It would be helpful if the authors clarify that, while Nice and RHice are reasonably reproduced, IWC is systematically overestimated, to avoid potential misinterpretation.
• Lines 357: missing “at high latitudes” after “resulting cirrus” and before “(M-H)”.
• Lines 371-373: The final sentence beginning with “Ngo et al. (2025)” feels somewhat disconnected from the preceding discussion. I recommend adding a short transition or explanation to clarify how the findings of Ngo et al. relate to the results described here.
• Lines 376-378: The authors state that “ice crystal number concentrations and the frequency of events are very likely overestimated, as suggested by recent findings by Testa et al. (2024).” However, Testa et al. (2024) primarily investigate the poor ice-nucleating efficiency of aviation soot particles and do not directly demonstrate an overall overestimation of ice crystal numbers or event frequencies. I recommend revising this sentence to more accurately reflect the scope of the Testa et al. study, or clarifying how their findings support this interpretation.
• Lines 378-383: The paragraph discussing the potential impact of BCav as an INP is conceptually relevant but somewhat difficult to follow in its current form. The logical connection between BCav efficiency, cirrus formation at high latitudes, and the vertical distribution of black carbon concentrations could be clarified.

  It would also be helpful to briefly explain why BC and BCav behave differently despite their similar chemical composition. Clarifying this distinction would provide readers with a clearer understanding of the physical basis of the argument.

• Lines 418-419: The manuscript states that “fit1 seems to match the observations more closely, it may lead to an overcorrection of Nice.” At present, it is not clear how this potential overcorrection is determined. Is it inferred from deviations at other temperatures, from known model biases, or from the comparison with alternative fits (fit2) and climatological data? I recommend that the authors clarify the basis for this statement, for example by providing quantitative evidence or further explanation, so that readers can better understand why fit1 may overestimate ice crystal number concentrations.
• Figures 1, 4, and S1: I recommend that the authors include the upper bound of the measurement range in Figures 1, 4, and S1. Indicating the complete range will help readers interpret the variability shown and better understand the limits of the data for each instrument and particle-size range.
• Lines 447-459 and 484-486: The conclusion correctly emphasizes that differences in ice crystal number concentrations are not explained by total aerosol concentrations, but rather by the influence of specific INP types and their interaction with ambient supersaturation. I suggest that the authors highlight the need for further studies focusing on the roles of different INPs across regions and cloud types, rather than only total INP concentrations, to better understand regional differences in cirrus microphysics.
• Lines 450-452: The conclusion refers to “differences in total INP concentrations,” but it should be noted that the study only reports total aerosol concentrations (and particles >500 nm as a proxy for INPs). The actual concentrations of all INP types are not directly measured. I suggest clarifying this point, as the statement about total INP differences may overstate the available observational evidence.
• Lines 457-458: consider adding references that support the following satements: “denser aviation activity at mid-latitudes” and “which increases ice number concentration and reduces the effective particle diameter”.

References:

Field, P. R., Heymsfield, A. J., Bansemer, A., and Brown, P. R. A. (2006). Ice particle size distributions in cirrus and stratiform clouds. Atmos. Chem. Phys., 6, 2547–2563.

Field, P. R., Lawson, R. P., and Heymsfield, A. J. (2017). Observations and modeling of secondary ice production in clouds. Atmos. Chem. Phys., 17, 3363–3380.

Hallett, J., and Mossop, S. C. (1974). Production of secondary ice particles during the riming process. Nature, 249, 26–28.

Korolev, A., and Field, P. R. (2008). The effect of ice multiplication in ice clouds. J. Atmos. Sci., 65, 2797–2806.

Review by Minghui Diao, for “A combined observational and modeling approach to evaluate aerosol-cirrus interactions at high and mid-latitudes” by De La Torre Castro et al.

This study investigated the microphysical properties of cirrus clouds of mid- and high-latitudes based on the CIRRUS-HL campaign. In-situ observations and model simulations from a global model were compared with each other, and the simulations were further used to examine the influences of different types of aerosols on cirrus properties. Some specific types of aerosols were investigated as ice nucleating particles (INPs), including dust, aviation-emitted black carbon, etc. The model simulation was found to overestimate Nice in the midlatitudes, and underestimate large aerosol in the upper troposphere in the high latitudes. This bias can be corrected when an observation-based parameterization for convective ice crystals was applied to the model.

Overall, the manuscript is well written and nicely structured. The descriptions of the results are easy to follow, and the summary section nicely highlights the key findings. The combined usage of observations and simulations enhanced the interpretation of the observed results, as well as providing a pathway to improve the model. The reviewer has some minor comments below and recommends that these questions be addressed before the manuscript can be considered for publication.

1. Size restrictions to ice particles when comparing observations and simulations

• In Section 2.1, the authors mentioned that 3 cloud probes were used to measure ice particles, including CDP (2-50 micron), CIPgs (15-960 micron), and PIP (100-6400 micron). Then the authors mentioned that “to obtain a complete particle size distribution, data from the three instruments were combined”. The reviewer wonders how exactly these 3 probes were combined, such as X, Y, Z range provided by CDP, CIPgs, and PIP, respectively, since they have overlapping ranges.
• Later in the result section, line 280, section 4.2, an important size restriction was mentioned, that is, “only ice particles with diameters smaller than 200 micron” are used to calculate cirrus properties. The authors mentioned that > 200 micron particles will be categorized as snow crystals in the model and assumed to be removed within one model time step. This size cutoff at 200 micron will remove a significant amount of ice particles that would be the main contributor to ice water content (IWC). Evaluating how the model represents these >200 micron particles would be quite important for the overall model performance evaluation. Even though the model removes these particles in one time step, the authors can still show from a statistical point of view, before removal, how the model snow category’s IWC, Ni, and Di compare with those derived for > 200 micron particles in observations. In fact, the reviewer suspects that the model biases seen in these larger size range would suggest that the one-step removal is too fast, and the total IWC of cirrus may be underestimated by the model. This finding could be as significant as the model biases of Nice overestimation as discussed in the later section.
• If the combined observations provided 15 micron – 6400 micron of ice particle size range, the authors may compare 15 – 200 micron with the modeled ice category, and 200 – 6400 micron with the modeled snow category. Please be cautious of the larger size cutoff (i.e., 6400 micron), that is, the model size range should be trimmed at both ends to match the observations more closely. That is, instead of using model at 200 – infinity, please trim it to 200 – 6400 micron. Overall, the reviewer recommends that this size restriction method be added to Section 3.2 instead of only mentioned briefly in the result section.

• In addition to IWC and Ni, can the author also quantify Dice (such as number-weighted mean or mass-weighted mean diameter) in Figures 2 and 5? Previously, observation-model comparisons such as Patnaude et al. 2021, Maciel et al., 2023, and Ngo et al., 2025 included the statistical distributions of IWC, Ni, and Di in midlatitudes and high latitudes as well. A quick comparison of CIRRUS-HL results with these previous observations obtained from other geographical regions would be helpful to understand the regional variabilities. In those previous studies, CESM2/CAM5 climate model was found to have too many but too small ice crystals (i.e., modeled Nice is too high and Dice is too low). It would be interesting to see if EMAC/ECHAM5 has a similar issue.

2. Size range of aerosol measurements: In the comparison between observations and simulations, the authors used three ranges, i.e., D > 12 nm, > 14 nm, and > 250 nm (such as Figures 1 and 3). The usage of D > 250 nm seems to be due to the aerosol probe size range, although typically the D > 500 nm has been used as a proxy for INPs, like the references that authors cited. The reviewer wonders if there is a specific reason that the authors chose to focus on the > 250 nm instead of > 500 nm (even though > 500 nm was briefly discussed in Fig. S1). For instance, is there a concern of the data quality of > 500 nm, which is why the main analysis focuses on > 250 nm?

3. The reviewer assumes that the aerosol measurements used in this study included both in-cloud and clear-sky conditions. Thus, the reviewer recommends that a sensitivity test being conducted to verify if using only the aerosol measured at clear-sky conditions (excluding in-cloud aerosols) would produce similar main conclusions. Even though at cirrus temperatures, the interference of cloud hydrometeors on aerosol measurements is likely smaller than compared with that at higher temperatures, a data quality check on the aerosol measurements would still help to verify whether or not the aerosol number concentrations are affected by shattering of ice particles.

4. Thermodynamic/dynamic influences: The authors examined RHice in addition to cloud properties, and the reviewer recommends that the vertical velocity distributions (such as sigma_w, like the equation used to derive sigma_w for every 50 seconds or 500 seconds of observations in Patnaude et al., 2021; Ngo et al., 2025) also be examined besides RHice. In addition, it is interesting to see that the authors applied the back trajectory tool to separate the two origins of cirrus – liquid origin and in-situ, but these two types of cirrus were not investigated separately, such as in Figure 2 (just observations) and Figure 5 (comparison with simulations). It would be helpful to see if the two origins of cirrus also show significantly different microphysical properties, and if the model tends to show better or worse results for one type of cirrus. For example, if the model convective cirrus parameterization has been improved in Section 4.4, one would expect to see better result for the liquid-origin cirrus comparison between observations and simulations.

5. The reviewer wonders if the model output variables used for comparisons are grid mean or in-cloud variables? Also, is the spatial averaging applied to both in-cloud and clear-sky segments of observations? It would be helpful to provide the model variable names and their meaning in section 3.1 or 3.2. In some previous studies of model evaluation such as Patnaude et al. (2021) and Maciel et al. (2023), the model grid-mean variables were used to compare with spatially averaged observations that included both in-cloud and clear-sky conditions. If one uses model variables that represent the in-cloud IWC and Ni, then those should be compared with in-cloud averaged IWC and Ni derived from observations.

6. Spatially averaging the observations to 15 minutes seems a reasonable choice of scale. It is worth noting that in addition to the temporal frequency of the output, using similar spatial scales (i.e., model grid spacing scale versus the scale of the spatially averaged observations) is also important. Assuming that HALO flies at 250 m/s at the cirrus temperatures, 15 minutes * 250 m/s is about 225 km, which is comparable to the 2.8 deg * 2.8 deg lat * lon resolution of the model configuration as well. This similarity in spatial scale is worth noting too.

7. In Figure 5, should the black dots in panels c and f, actually be orange color? Please clarify. It is a little confusing why the simulation data (orange) is broken up. Is it because of fewer vertical bins?

8. A minor comment on Figure 7’s line format: The dotted and dashed lines have small gaps, which are readable on computer screen, but when printed out, the lines look identical within each panel. The lines also used colors that are too similar within each panel. The reviewer recommends using slightly different colors for various lines to enhance the readability of this figure.

9. Line 38, “INPs, (Kanji et al.,” has an extra comma after INPs.