Using formvar to capture atmospheric ice crystals and retrieve roughness parameters

Celebi, Omer; Smedley, Andrew R. D.; Connolly, Paul; Webb, Ann R.

doi:10.5194/amt-19-2621-2026

Articles | Volume 19, issue 7

https://doi.org/10.5194/amt-19-2621-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/amt-19-2621-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 7

Research article

|

17 Apr 2026

Research article |

| 17 Apr 2026

Using formvar to capture atmospheric ice crystals and retrieve roughness parameters

Omer Celebi, Andrew R. D. Smedley, Paul Connolly, and Ann R. Webb

Download

Final revised paper (published on 17 Apr 2026)
Preprint (discussion started on 16 Jan 2025)

Interactive discussion

Status: closed

RC1:
'Comment on amt-2024-200', Anonymous Referee #1, 03 Feb 2025

Review of ‘Using Formvar to Capture Ice Crystals and Retrieve Roughness Parameters’
from Celebi et al. (2025)

Celebi et al. present a proof-of-concept for using the formvar replica technique to preserve the shape of ice crystals to analyze the surface roughness offline, e.g. using an optical profilometer. SEM imaging was also done. The comparison of roughness parameter derived from salt crystals and their replica showed agreement despite of the skewness. By testing the technique with real ice crystals generated in the Manchester ice cloud chamber (MICC), it turned out that the mean roughness increased by increasing crystal length and levels off at around 50 µm, which can be considered as a reasonable result. However, I don’t fully understand what the aim of this study is. Is it planned as an add-on measurement technique for MICC to determine surface roughness of ice crystals which were formed in the chamber? How many ice crystals have to be analyzed to gain a representative statistic of the crystals and their roughness in the chamber? Is it possible to achieve this with the here presented method? Or what is the objective? I think it has to be shown that this method is robust to study a large number of ice crystals. I guess this is needed for any application. In principle, I see the potential of the here presented method as proof-of-concept. However, as there are some shortcomings using the actual formvar method and the fact that the objective is not defined, the value of the study needs to be presented more clearly. After overcoming these problems, which are probably largely due to the formulation and description of the study, the manuscript can be published in AMT, as the formvar replica technique has potential and fits within the scope of the journal.
Specific comments:
Abstract: It would strengthen the paper if the abstract stated that this was a proof-of-concept study. In addition, giving information of the comparison between salt and salt replica as a test would eventually increase confidence in this method.
Introduction: The introduction is well written and gives a good overview about the necessity to study ice crystal surface roughness and the state-of-the-art of surface roughness determination with formvar. However, a short paragraph of the state-of-the-art of surface roughness using other methods is missing. I am sure there are more methods and publications on that topic, but the study of Magee et al. (2014) [1] applying ESEM, Ulanowski et al. (2014)[2] Voigtländer et al. (2018)[3] and Järvinen et al. (2018)[4] using SID-3 instrument could be mentioned.
Methodology part & Figure 1: The methodology part is very detailed. There are some repetitions. In my opinion Figure 1 is not important. I suggest to shorten the section and leave Fig. 1 out.
P9 l208-210: The authors gave the information that many features in the formvar are not a result of accreted ice crystals but artefacts from the formvar process. Do the artefacts affect the determination of surface roughness of ice crystals? Can always be distinguished between artefact and ice crystal? Do the artefacts influence the counting statistics of ice crystals when the fall on a spot with an artifact? Can the process been improved in order to reduce or avoid these artefacts?
P9 l210-215: I do not understand why it solves the problem that ice crystals do not sink when using a thicker formvar layer.
Fig. 5: What relevance do those ice crystals have that not fully sink into the formvar? Could it be that a certain type of ice crystals always not sink into formvar and would be missing in the analyzes leading to basis of this method?
P12 Fig.5: Please add a scale bar.
P13 l283-286: I agree, but it would be good to have more information of the formation and history of the ice crystals which were examined (later it is discussed, but I missed the information already at this point in the text). Maybe the roughness is different due to different environmental conditions the ice crystals experienced [3, 4] before settled on the formvar. Further, the number of investigated ice crystals is not very high (caption of Fig. 9 ‘n=7 for each size bin’). It seems that this offline method -using formvar for ice crystal surface roughness detection- is very complex and time-consuming. How will it be possible to analyse a larger sample size? What are the advantages of this introduced method compared to online measurements of single airborne ice crystals such as e.g. SID-3?
P14 l316-317: ‘However, if appropriate mitigation methods are implemented, the occurrence of these artefacts can be greatly reduced, resulting in the most accurate possible roughness measurements.’ For such a statement a comprehensive analysis of already published methods is needed, which should be given in the introduction, see comment introduction of this review. And maybe it should be reformulated to ‘one of the most accurate…’. I would expect SEM to have a higher spatial resolution.

Technical corrections:
P2 l49: Insert a blank between ‘(1951)used’.
P5 l115: Insert a half blank ‘0.6wt%’.
References: The AMT citation style has to be used.

[1]        N. B. Magee, A. Miller, M. Amaral, and A. Cumiskey, "Mesoscopic surface roughness of ice crystals pervasive across a wide range of ice crystal conditions," Atmos. Chem. Phys., 2014, doi: 10.5194/acp-14-12357-2014.
[2]        Z. Ulanowski, P. H. Kaye, E. Hirst, R. S. Greenaway, R. J. Cotton, E. Hesse, and C. T. Collier, "Incidence of rough and irregular atmospheric ice particles from Small Ice Detector 3 measurements," Atmos. Chem. Phys., 2014, doi: 10.5194/acp-14-1649-2014.
[3]        J. Voigtländer, C. Chou, H. Bieligk, T. Clauss, S. Hartmann, P. Herenz, D. Niedermeier, G. Ritter, F. Stratmann, and Z. Ulanowski, "Surface roughness during depositional growth and sublimation of ice crystals," Atmospheric Chemistry and Physics, 2018, doi: doi:10.5194/acp-18-13687-2018.
[4]        E. Järvinen, H. Wernli, and M. Schnaiter, "Investigations of Mesoscopic Complexity of Small Ice Crystals in Midlatitude Cirrus," Geophysical Research Letters, 2018, doi: https://doi.org/10.1029/2018GL079079.

Citation: https://doi.org/10.5194/amt-2024-200-RC1
- AC1: 'Reply on RC1', Omer Celebi, 04 Apr 2025
  
  We would like to sincerely thank the reviewers for their valuable feedback and constructive comments, which have significantly contributed to improving our manuscript. We have provided detailed responses and necessary explanations in the accompanying file titled "Reviewer 1 Document.pdf".
  
  Citation: https://doi.org/10.5194/amt-2024-200-AC1
RC2:
'Comment on amt-2024-200', Alexey Kiselev, 25 Feb 2025

The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2024-200/amt-2024-200-RC2-supplement.pdf

Citation: https://doi.org/10.5194/amt-2024-200-RC2
- AC2: 'Reply on RC2', Omer Celebi, 04 Apr 2025
  
  We would like to sincerely thank the reviewers for their valuable feedback and constructive comments, which have significantly contributed to improving our manuscript. We have provided detailed responses and necessary explanations in the accompanying file titled "Reviewer 2 Document.pdf".
  
  Citation: https://doi.org/10.5194/amt-2024-200-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Omer Celebi on behalf of the Authors (04 Apr 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (10 Apr 2025) by Maximilian Maahn

RR by Anonymous Referee #1 (05 May 2025)

Suggestions for revision or reasons for rejection

Second review of ‘Using Formvar to Capture Ice Crystals and Retrieve Roughness Parameters’
from Celebi et al. (2025)

In Celebi et al., a proof-of-concept study is presented to investigate the spatially high-resolved surface roughness of salt and ice crystals—generated in the Manchester Ice Cloud Chamber—using the formvar replica technique. As this is the second round of the review, I will avoid repetition and instead provide concise feedback. A significant portion of the earlier criticisms has been addressed in the revised version; for example, the authors have more clearly articulated the advantages of their study. However, there are still a few issues that, in my opinion, need to be addressed before the manuscript can be considered for publication in AMT.
I remain unconvinced that it is necessary to include the formvar brush technique in the main text, as it does not replicate the full crystal (as shown in Figures 4 and 5). I suggest either omitting this section or moving it to the supplementary information. Alternatively, if it is irrelevant because only one side of the crystal is used for roughness analysis, this should be clearly stated and justified.
Regarding Table 1: Why is there no direct comparison between the salt crystals and their corresponding replicas? Since only four crystals were analyzed, such a comparison seems feasible. If there is a reason this was not done, it should be explained.

Specific comments:

Abstract (l7-8): I would delete ‘weather and’ as the direct connection is not clear regarding how ice crystal surface roughness will affect weather forecast.
l29: This sentence is very general and therefore not meaningful. Please rephrase or leave it out.
L37-38: What is the specific outcome in these studies? ‘That these studies compared in many aspects’ is very unspecific and hence not very meaningful.
L43: ‘are not affected too much’ is very unspecific and a more precise description is needed.
L46-47: The introducing sentence sounds very generic and not very meaningful. Be a bit more precise and focus on the main points.
L237: Insert ‘salt’ before ‘crystal’.
L261-262: ‘initiate condensation under a supersaturated environment’ do you mean droplet formation prior to freezing here?
L336: Insert ‘salt’ before crystal as it has only been shown for salt crystals.
L366-368: Can this 1D information been extrapolated to 2D structures? This should be clarified or discussed.
L365-373: The whole text section sounds generic and lacks precision. For example: What is meant by ‘cloud behavior’? I do not see any link between the ice surface roughness analyzed in this study and a general ‘cloud behavior’. This needs to be improved. The conclusion and outlook section would improve when is more related to the specific topic.

Technical corrections:

L17: ‘Hong, Y’ Citation style is not correct.
L74: Delete ‘Meanwhile’.
References: ‘dio’ references are sometimes missing.

Hide

RR by Alexey Kiselev (09 May 2025)

Suggestions for revision or reasons for rejection

I appreciate the efforts the authors made to answer the reviewers' comments and to revise the manuscript accordingly. I therefore recommend its publication after addressing the following minor comments:

1. The manuscript states (lines 146-147) that "Samples were initially examined with a lateral resolution less than 10 nm and vertical resolution less than 5nm, to ensure that the crystal replicas remained intact...". I don't quite understand why such a high resolution is required for initial examination, especially if the final target resolution for roughness measurements is stated to be 100 nm. Maybe it is "µm" and not "nm"? Please check.

2. Second, from my own experience with formvar replicas of ice I know that the replicas of complex shapes (such as bullet rosettes or aggregates) are often deformed beyond recognition after solidification of formvar. The deformation arises when ice crystal's melting and evaporation proceeds along the hardening of formvar. For SEM, this is an insurmountable problem since the electron beam cannot penetrate the formvar layer; for laser scanning microscope that should not present such a difficulty. I would therefore strongly encourage authors to add a figure showing examples of optical images of the replicas to the manuscript. This would also strongly corroborate their statement that the replicas remain intact under the experimental conditions of their method. The SEM images suggest that was not always the case.

3. It is unclear from the description of the method if the ice crystals were allowed to sublimate or melt after immersion into formvar. Was the bell jar kept at the MICC temperature conditions or was it installed outside of the cold room? In any of this cases, if the ice crystals are completely immersed into formvar, where does the water go? Please add these details to the manuscript.

4. Line 127: from the blue line in figure 1, the vapor pressure of chloroform at -20°C is reduced by factor 8, not 5.5 as compared to room temperature: ~200 hPa at +20°C, ~25 hPa at -20°C. Not that it matters a lot...

5. The References do not comply with the AMT template. The DOIs are missing everywhere. Some references are incomplete: "Cheric .: Vapor pressure of chloroform, 1995.", "Finnegan W. and Pitter R.: Atmospheric ice crystal processes, 1988.", "Hong Y L.: The impact of two coupled cirrus microphysics–radiation parameterizations on the temperature and specific humidity biases in the tropical tropopause layer in a climate model, Journal of Climate, 2015.", "Centre of Atmospheric Science, 2024."

Hide

ED: Reconsider after major revisions (09 May 2025) by Maximilian Maahn

AR by Omer Celebi on behalf of the Authors (31 Jul 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (01 Aug 2025) by Maximilian Maahn

RR by Anonymous Referee #1 (13 Aug 2025)

RR by Alexey Kiselev (14 Aug 2025)

Suggestions for revision or reasons for rejection

I appreciate the improvements of the manuscript. After rereading the manuscript I do, however, have a few comments.

I noticed that neither the title nor the abstract mentions what kind of ice crystals are captured and characterized for roughness. It is also not necessarily clear from the context, so maybe a mention of "atmospheric cloud ice crystals" or something to this effect would be appropriate.

I still have trouble understanding how the roughness of ice crystals obtained with the optical profilometer compares to the literature values and how it can be used for deriving optical properties of ice crystals. First, the authors do not cite any other values of roughness measured in previous studies. Instead, they more or less put aside all previous measurements by stating “Direct measurement of the surface roughness of ice crystals has been elusive, …” (line 79 of the revised manuscript). Elusive or not, roughness of atmospherically relevant particles has been measured for mineral dust grains, as cited by the authors (Collier et al., 2016) and for ice crystals grown in SEM (Magee et al., 2014, Butterfield et al., 2017). Second, and most importantly, the dimensional roughness parameters defined in the manuscript by equations 2 and 3 cannot be compared to any previous measurements or to the values used in the optical simulations, because this is not how roughness of ice crystals is represented in light scattering models.

To simulate scattering of light by ice crystals with rough surfaces, roughness has been introduced either as a variation of tilt angles of the patches (facets) on the ice crystal surface, a method originally introduced by (Cox and Munk, 1954) and afterwards extensively used by modelling community (e.g. Yang et al., 2013), or by defining "the correlation length which describes the dominant spatial frequencies, and a standard deviation, which describes the variation in height" (Collier et al., 2016; Ulanowski et al., 2014). I am puzzled why authors have chosen a different, rather technical definition of roughness which is hardly applicable for optical calculations, but the measurement data is undoubtedly still available! With the 3D surface profiles of the formvar replicas and the full analytical power of modern computer software, I am sure this will be a simple exercise to derive one or both roughness metrics from the profilometer data.

I have partly addressed these issues in my review of the original manuscript, but the authors rebutted with the argument that the study is a “proof of concept”. From my point of view, however, the “concept” includes showing that the method allows measurements of the parameters needed for optical simulations, as mentioned by authors several times in the manuscript, and I strongly encourage the authors to perform such calculations. Roughness metrics derived in this way would be directly comparable to the data from previous studies and with the values used in optical simulations, thus immediately increasing the potential impact of this manuscript.

For the same reasons the sentence (lines 75-77) “The roughness values obtained from the sand grains were found to be similar to those retrieved by SID-3 for ice crystals, making sand a suitable medium for capturing roughness then applied to scattering models” is misleading because it implies that the roughness (as defined by equations 2 and 3) has been retrieved for both particle types from the SID-3 measurements. This is not so. In fact, the conclusion of roughness similarity between the mineral dust grains (“sand”) and ice crystals has been drawn by comparing diffraction patterns of both particle types measured with SID-3 instrument (Ulanowski et al., 2012; Ulanowski et al., 2014) and then, based on this observations, the roughness parameters obtained from the AFM measurements of “sand” particles (Collier et al., 2016) have been used to model the optical properties of ice crystals. Even then, the surface roughness parameters used in these studies were defined differently: as a combination of a correlation length and standard deviation of heights, both values not directly comparable to definitions given in (Gadelmawla et al., 2002). It is also important to keep in mind (and this is explicitly mentioned in (Collier et al., 2016)) that the similarity of the “scattering pattern derived roughness” holds only for similarly sized grains.

References

Collier, C. T., Hesse, E., Taylor, L., Ulanowski, Z., Penttilä, A., and Nousiainen, T.: Effects of surface roughness with two scales on light scattering by hexagonal ice crystals large compared to the wavelength: DDA results, Journal of Quantitative Spectroscopy and Radiative Transfer, 182, 225-239, https://doi.org/10.1016/j.jqsrt.2016.06.007, 2016.

Butterfield, N., Rowe, P. M., Stewart, E., Roesel, D., and Neshyba, S.: Quantitative three-dimensional ice roughness from scanning electron microscopy, Journal of Geophysical Research: Atmospheres, 122, 3023-3041, https://doi.org/10.1002/2016JD026094, 2017.

Cox, S. C. and Munk, W. H.: Measurement of the roughness of the sea surface from photographs of the sun’s glitter, J. Opt. Soc. Amer., 44, 838-850, 1954.

Ulanowski, Z., Kaye, P. H., Hirst, E., Greenaway, R. S., Cotton, R. J., Hesse, E., and Collier, C. T.: Incidence of rough and irregular atmospheric ice particles from Small Ice Detector 3 measurements, Atmos. Chem. Phys., 14, 1649-1662, https://doi.org/10.5194/acp-14-1649-2014, 2014.

Ulanowski, Z., Hirst, E., Kaye, P. H., and Greenaway, R.: Retrieving the size of particles with rough and complex surfaces from two-dimensional scattering patterns, Journal of Quantitative Spectroscopy and Radiative Transfer, 113, 2457-2464, https://doi.org/10.1016/j.jqsrt.2012.06.019, 2012.

Gadelmawla, E. S., Koura, M. M., Maksoud, T. M. A., Elewa, I. M., and Soliman, H. H.: Roughness parameters, Journal of Materials Processing Technology, 123, 133-145, https://doi.org/10.1016/S0924-0136(02)00060-2, 2002.

Magee, N. B., Miller, A., Amaral, M., and Cumiskey, A.: Mesoscopic surface roughness of ice crystals pervasive across a wide range of ice crystal conditions, ACP, 14, 12357-12371, https://doi.org/10.5194/acp-14-12357-2014, 2014.

Hide

ED: Reconsider after major revisions (15 Aug 2025) by Maximilian Maahn

AR by Omer Celebi on behalf of the Authors (11 Oct 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (13 Oct 2025) by Maximilian Maahn

RR by Alexey Kiselev (04 Nov 2025)

Suggestions for revision or reasons for rejection

I acknowledge the changes made by the authors in response to my comments. Specifically, I agree that the addition of a paragraph citing the previous works by (Collier et al., 2016, Ulanowski et al., 2014, and Butterfield et al., 2017) sets the presented method of measuring surface roughness into the framework of optical properties of cirrus clouds. However, the connection between the roughness measurements and calculation of optical properties has not been addressed in a satisfactory way.

1. The authors added a sentence where they “…explicitly state that the profilometer approach can yield roughness metrics directly comparable to those used in optical scattering studies, namely correlation length and height standard deviation (σ)…” (I am citing from their response to my previous comments). This “explicit statement” reads as follows (lines 96 -97 of the revised manuscript): “Building on all these approaches, the formvar replication method combined with optical profilometry enables such surface features – including correlation length and RMS height – that are directly comparable to values used in scattering simulations.”

This statement, again, is a just a statement, which is neither supported by any further arguments, nor by comparing the actual data obtained in this manuscript with the data of other studies. This statement is also inconsistent with what authors write in response to my comments: the manuscript contains the values of RMS (root mean square) of height coordinate (equation 3); the standard deviation of height was not shown anywhere in the manuscript.

2. Further, in response to my comments, the authors write: “Finally, in lines 315–326, we present our dimensionless results which was produced by following Collier et al. (2016) and normalized to λ = 532 nm in a table and explicitly compare these values with those reported in prior studies, noting that they fall within the ranges used in light scattering simulations (e.g. Collier et al., 2016; Butterfield et al., 2017).”

The authors do not explain even briefly how these “dimensionless results” have been obtained. There is also an intrinsic inconsistency in their statement (lines 316 – 317 of the revised manuscript): “…we derived the correlation length (Lc) and standard deviation of height (σ) for one representative ice crystal replica which was normalized to the incident wavelength (λ = 532 nm), following the approach of Collier et al. (2016).” If the Lc and σ have been derived “for one representative ice crystal”, then why are values for four different size parameters X are given? For a given wavelength, a single ice crystal can have only one size parameter, so where are the others are coming from?

3. The choice of size parameters is another mystery. From the definition of the size parameter and a wavelength of 0.532 µm, the largest ice crystal used for this derivation (X = 80) was not larger than 13.5 µm in diameter, and the smallest just above 3 µm (X = 20). However, all other images in the manuscript (figures 3, 4, 5, and 8) show ice crystals that are at least 50 µm in one dimension, corresponding to X = 300 for the wavelength of 0.532 µm. So how such a crystal could be declared “representative”?

4. It should also be mentioned, that Collier et al. (2016) used the experimental values of correlation length and standard deviation to simulate model rough ice particles and then to calculate their scattering properties using the DDA (Direct Dipole Approximation) method. The DDA method, being computationally very expensive, is not capable of dealing with size parameters above 100. To mitigate this issue, and to keep the number of dipoles in the DDA limited, the roughness parameters (correlation length and height variance) were scaled up (artificially !!!) with size parameter – see section 3 “DDA Results” of the Collier et al. (2016): “Four crystal size parameters were considered; 20, 40, 60 and 100, with roughness being scaled proportionately with size”. While this explains why the normalized correlation length and sigma both increase with simulated crystal size (see Table 1 of Collier et al. (2016), there is no physical reason to expect such proportional increase in your case. However, the Table 3 of the manuscript clearly demonstrate an increase of the normalized correlation length and standard deviation of height with the increasing size of ice crystals.

This argument renders the statement (lines 319 - 320 of the revised manuscript) “The values are consistent in magnitude with those reported for mineral dust analogues measured by an AFM (Collier et al., 2016) and with SEM-derived features of ice crystals (Butterfield et al., 2017)”, essentially, meaningless.

Finally, the comparison with SEM-derived features of (Butterfield et al., 2017) is not valid because (Butterfield et al., 2017) does not report values of correlation length or standard deviation of height. Reading (Butterfield et al., 2017) carefully, one immediately notices that their definition of Z (equation 1 in Butterfield et al., 2017) has nothing to do with height variability but gives local gradient of a tilted microsurface on a rough ice surface.

In my opinion, these inconsistencies can be resolved in two possible ways:

Option 1. The authors explain in great detail how the correlation length and sigma have been calculated from the actual measurement data; the original measurement data has to be made available in supplementary material or via data repository.

Option 2. The authors remove the newly added "dimensionless results" and reduce the claim that the method is capable of providing roughness characteristics required for calculations of optical properties of rough ice crystals. The scope of the manuscript is reduced to the description of a method to measure surface roughness of ice crystal replicas with a profilometer. The demonstration of similarity between the roughness parameters retrieved for replicas of salt crystals and of the original salt crystals implies the possibility to apply this method for characterization of atmospheric ice crystals, but the applicability for optical calculations is yet to be demonstrated.

Hide

ED: Reconsider after major revisions (04 Nov 2025) by Maximilian Maahn

AR by Omer Celebi on behalf of the Authors (18 Dec 2025) Author's response Author's tracked changes Manuscript

ED: Publish as is (08 Jan 2026) by Maximilian Maahn

AR by Omer Celebi on behalf of the Authors (23 Jan 2026) Manuscript

Short summary

The surface structure of ice crystals in clouds influences how sunlight travels through the atmosphere and contributes to uncertainty in climate predictions. Because these features are difficult to observe, we present a practical approach for mapping ice crystal surface structure using an established capture technique. Ice crystals were produced in a laboratory cloud chamber and their surfaces were measured in detail. This approach enables reliable mapping of ice crystal surfaces.