Ice crystal images from optical array probes. Compatibility of morphology specific size distributions, retrieved with specific and global Convolutional Neural Networks for HVPS, PIP, CIP, and 2DS

Jaffeux, Louis; Breiner, Jan; Coutris, Pierre; Schwarzenböck, Alfons

doi:https://doi.org/10.5194/egusphere-2024-1910

Preprints

https://doi.org/10.5194/egusphere-2024-1910

Preprints

18 Sep 2024

| 18 Sep 2024

Ice crystal images from optical array probes. Compatibility of morphology specific size distributions, retrieved with specific and global Convolutional Neural Networks for HVPS, PIP, CIP, and 2DS

Louis Jaffeux, Jan Breiner, Pierre Coutris, and Alfons Schwarzenböck

Abstract. The convolutional network methodology is applied to train classification tools for hydrometeor images from optical array probes. Two models were developed in a previous article for the PIP and 2DS and are further tested. Three additional models are presented: for the CIP, HVPS, and a global model trained on a data set that includes all available data from all four instruments. A methodology to retrieve morphology-specific size distributions from the OAP data is provided. Size distributions for each morphological class, obtained with the specific or global classification models, are compared for the ICE GENESIS data set, where all four probes were used simultaneously. The reliability and coherence of these newly obtained machine learning classification tools are demonstrated clearly. The analysis shows significant advantages of using the global model over the specific ones, in terms of compatibility of the size distributions. The obtained morphology-specific size distributions effectively reduce OAP data to a level of detail pertinent to systematically identify microphysical processes. This study emphasizes the potential to improve insights in ice and mixed-phase microphysics based on hydrometeor morphological classification from machine learning algorithms.

Received: 21 Jun 2024 – Discussion started: 18 Sep 2024

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Preprint (PDF, 15125 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (15125 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

03 Jun 2025

Convolutional neural networks for specific and merged data sets of optical array probe images: compatibility of retrieved morphology-dependent size distributions

Louis Jaffeux, Jan Breiner, Pierre Coutris, and Alfons Schwarzenböck

Atmos. Meas. Tech., 18, 2311–2331, https://doi.org/10.5194/amt-18-2311-2025,https://doi.org/10.5194/amt-18-2311-2025, 2025

Short summary

Louis Jaffeux, Jan Breiner, Pierre Coutris, and Alfons Schwarzenböck

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-1910', Anonymous Referee #1, 08 Oct 2024

- suggestion: add classification results for the 3 general types (pristine crystals, intermediary particles and ultimate precipitating particles)
- section 2: a visualization of the CNN architecture would be helpful
- section 2.2.1 / 2.2.2 / 2.2.3: How many training samples are used for each class?
- Figure 2 / 3 / 4: As I understand, the results are for the training data. Please add the results for the test dataset.
- Line 203: It is not clear how the data assimilation was performed. Are the WD test samples added to the training data? Are the same WD samples used for the results presented in figure 5b?
- Line 236: Clarify LaMP

Citation: https://doi.org/10.5194/egusphere-2024-1910-RC1
- AC1: 'Reply on RC1', Louis Jaffeux, 17 Oct 2024
  
  - suggestion: add classification results for the 3 general types (pristine crystals, intermediary particles and ultimate precipitating particles)
  The classes that are gathered under the 3 general types may have widely different shapes, meaning they are subjected to various issues with respect to recognition. of their shapes. Therefore, adding these metrics is not pertinent to users of the trained models, neither to understand the obtained results nor to assess their performance.
  
  - section 2: a visualization of the CNN architecture would be helpful
  The article of Jaffeux et al. (2022) contains several of these representations. One of these figures will be added in a revised version of the manuscript.
  
  - section 2.2.1 / 2.2.2 / 2.2.3: How many training samples are used for each class?
  The exact number of images for each class and for each instrument is not thoroughly listed in the article, and these numbers only appear when necessary. A table can be added, with similar entries as Figure 1, but featuring these numbers instead of image examples. Readers have free access to the full, hand-labeled image folders in the dedicated Github repository.
  
  - Figure 2 / 3 / 4: As I understand, the results are for the training data. Please add the results for the test dataset.
  These figures provide the results of model predictions using an independent test data set, which was separated from the hand-labeled data set before training as described in Section 2. The overall accuracy obtained on the validation data sets is mentioned in each paragraph.
  
  - Line 203: It is not clear how the data assimilation was performed. Are the WD test samples added to the training data? Are the same WD samples used for the results presented in figure 5b?
  WD test samples were obtained as a result of specific testing, which is a qualitative check beyond the presented methodology performed when using the models on newly acquired data sets. They were added to the whole hand-labeled data set before using the same methodology. In particular, a training set was separated from the data set, used to train the model, and finally evaluated with the independent test data.
  
  - Line 236: Clarify LaMP.
  Will be clarified as Laboratoire de Météorologie Physique (LaMP).
  
  Citation: https://doi.org/10.5194/egusphere-2024-1910-AC1
RC2:
'Comment on egusphere-2024-1910', Anonymous Referee #2, 17 Oct 2024

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1910/egusphere-2024-1910-RC2-supplement.pdf

Citation: https://doi.org/10.5194/egusphere-2024-1910-RC2
- AC2: 'Reply on RC2', Louis Jaffeux, 20 Nov 2024
  
  Major Comment #1:
  We have answered this comment in a previous response as follows:
  “Details concerning the CNN methodology have been added in two additional paragraphs in Section 2. The results are easily reproducible thanks to open access to data sets and trained CNN models in the linked GitHub repository. Our recent paper (Jaffeux et al. 2022) is primarily presenting the methodology of the implemented CNN.”
  We have added a condensed description of used CNN on half a page (2 paragraphs) in the revised manuscript, to our surprise this major comment seems to be still there. The image data selection, training, and testing methodologies are important for the present study, however, they follow point by point the methodology detailed in a dedicated and already published article (Jaffeux et al. 2022). This is the reason, why we don’t consider necessary to copy-paste details of Jaffeux et al. (2022) into this study. Additional CNN information has nonetheless been added in the form of a schematic as was stated in reply to RC1 (https://doi.org/10.5194/egusphere-2024-1910-AC1). This part of the article is easily reproducible thanks to the open access Github repository that contains all used images used for the CNN models.
  The study presented here is going far beyond training of CNNs (which has been the objective in Jaffeux et al. 2022) and presents unprecedented and unique scientific results when utilizing the CNN models to OAP datasets of the measurement campaign ICE GENESIS in order to produce morphology dependent quantitative number size distributions (see the entire section 3 of this study). These size distributions are produced for 4 OAP probes with possibility to build composite morphology dependent distributions from overlapping OAP probes . This was never done before. Likewise, the latter sentences above can be added to respond to the second major comment of the reviewer concerning the novelty of the study. The presented work cannot be done from CPI image data, since the CCD camera based instrument has such a small sample volume that you can’t retrieve size distributions in a heterogeneous cloud encountered on aircraft. Of course we are aware of the fact that CPI image data have better resolution with in addition 256 grey scales, which makes CNN different (and certainly extremely valuable) for high resolution CPI imager as compared to poor resolved OAP probe black and white images. CNN models for CPI probes produce different results than do CNN models for OAP probes, however, we have to keep in mind that quantitative results presented in section 3 of this study cannot be produced with CPI image data!
  
  Major Comment #2
  Likewise, we have responded to this major comment as follows:
  “The introduction has been improved in order to highlight the differences with previous methods that used OAP data. Note that obtaining morphology specific particle size distributions from OAPs (or any other instrument for that matter) was never shown in any other research publications. Przybylo et al. (2022) used CPI images, which cannot be used for quantitative concentration measurements (neither concentrations nor size distributions if not averaged over long time periods which is not compatible with cloud spatial heterogeneity encountered with aircraft); in particular, they do not produce particle size spectra (compare the 3 million usable images they obtained from eleven field campaigns to the 2.3 million used here and obtained in 7 flight hours; see next point).”
  The introduction mentions the few studies that have successfully applied CNNs to OAP images (Jaffeux et al. (2022) and Zhang et al. (2023)). Since OAP images they have very limited pixel numbers and also resolution, the CNN method for OAP images might be more challenging since the information content is not as rich as it is for CPI images (Pryzybylo et al. (2022)), and of course hardly comparable with diffraction patterns (Schmidt et al. (2024a,b)). Nevertheless, we will mention in the introduction these studies, having used a very similar methodology for the shape retrievals. These points will be made in the introduction as the methodology used in both papers is indeed similar. With the respect to novelty, the methodology was used for the first time to successfully train a single model based on image data sets from 4 different models of probes with different image resolution and pixel numbers. In addition, Jaffeux et al. (2022) and Zhang et al. (2023) limited their methodology to single probe data sets. Section 3 is beyond existing work on OAP based morphological analysis, as it takes advantage of the morphological information extracted from a coherent dataset using four different probes to reconstruct quantitative morphology-specific size distributions. These are presented for the first time and provide new insights in shape recognition with respect to observable size range and pixel resolution. Moreover, they are of great interest to the understanding of ice microphysics. The introduction will be corrected to reflect the importance of these latter points and in order to add some more selected relevant references.
  Minor Comment #1
  Acronyms are now defined from the beginning (which includes the abstract). In particular, probe and institute names will be fully written, the first time they are used.
  Minor Comment #2
  Grammar has been checked and corrected when needed
  
  Citation: https://doi.org/10.5194/egusphere-2024-1910-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-1910', Anonymous Referee #1, 08 Oct 2024

- suggestion: add classification results for the 3 general types (pristine crystals, intermediary particles and ultimate precipitating particles)
- section 2: a visualization of the CNN architecture would be helpful
- section 2.2.1 / 2.2.2 / 2.2.3: How many training samples are used for each class?
- Figure 2 / 3 / 4: As I understand, the results are for the training data. Please add the results for the test dataset.
- Line 203: It is not clear how the data assimilation was performed. Are the WD test samples added to the training data? Are the same WD samples used for the results presented in figure 5b?
- Line 236: Clarify LaMP

Citation: https://doi.org/10.5194/egusphere-2024-1910-RC1
- AC1: 'Reply on RC1', Louis Jaffeux, 17 Oct 2024
  
  - suggestion: add classification results for the 3 general types (pristine crystals, intermediary particles and ultimate precipitating particles)
  The classes that are gathered under the 3 general types may have widely different shapes, meaning they are subjected to various issues with respect to recognition. of their shapes. Therefore, adding these metrics is not pertinent to users of the trained models, neither to understand the obtained results nor to assess their performance.
  
  - section 2: a visualization of the CNN architecture would be helpful
  The article of Jaffeux et al. (2022) contains several of these representations. One of these figures will be added in a revised version of the manuscript.
  
  - section 2.2.1 / 2.2.2 / 2.2.3: How many training samples are used for each class?
  The exact number of images for each class and for each instrument is not thoroughly listed in the article, and these numbers only appear when necessary. A table can be added, with similar entries as Figure 1, but featuring these numbers instead of image examples. Readers have free access to the full, hand-labeled image folders in the dedicated Github repository.
  
  - Figure 2 / 3 / 4: As I understand, the results are for the training data. Please add the results for the test dataset.
  These figures provide the results of model predictions using an independent test data set, which was separated from the hand-labeled data set before training as described in Section 2. The overall accuracy obtained on the validation data sets is mentioned in each paragraph.
  
  - Line 203: It is not clear how the data assimilation was performed. Are the WD test samples added to the training data? Are the same WD samples used for the results presented in figure 5b?
  WD test samples were obtained as a result of specific testing, which is a qualitative check beyond the presented methodology performed when using the models on newly acquired data sets. They were added to the whole hand-labeled data set before using the same methodology. In particular, a training set was separated from the data set, used to train the model, and finally evaluated with the independent test data.
  
  - Line 236: Clarify LaMP.
  Will be clarified as Laboratoire de Météorologie Physique (LaMP).
  
  Citation: https://doi.org/10.5194/egusphere-2024-1910-AC1
RC2:
'Comment on egusphere-2024-1910', Anonymous Referee #2, 17 Oct 2024

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1910/egusphere-2024-1910-RC2-supplement.pdf

Citation: https://doi.org/10.5194/egusphere-2024-1910-RC2
- AC2: 'Reply on RC2', Louis Jaffeux, 20 Nov 2024
  
  Major Comment #1:
  We have answered this comment in a previous response as follows:
  “Details concerning the CNN methodology have been added in two additional paragraphs in Section 2. The results are easily reproducible thanks to open access to data sets and trained CNN models in the linked GitHub repository. Our recent paper (Jaffeux et al. 2022) is primarily presenting the methodology of the implemented CNN.”
  We have added a condensed description of used CNN on half a page (2 paragraphs) in the revised manuscript, to our surprise this major comment seems to be still there. The image data selection, training, and testing methodologies are important for the present study, however, they follow point by point the methodology detailed in a dedicated and already published article (Jaffeux et al. 2022). This is the reason, why we don’t consider necessary to copy-paste details of Jaffeux et al. (2022) into this study. Additional CNN information has nonetheless been added in the form of a schematic as was stated in reply to RC1 (https://doi.org/10.5194/egusphere-2024-1910-AC1). This part of the article is easily reproducible thanks to the open access Github repository that contains all used images used for the CNN models.
  The study presented here is going far beyond training of CNNs (which has been the objective in Jaffeux et al. 2022) and presents unprecedented and unique scientific results when utilizing the CNN models to OAP datasets of the measurement campaign ICE GENESIS in order to produce morphology dependent quantitative number size distributions (see the entire section 3 of this study). These size distributions are produced for 4 OAP probes with possibility to build composite morphology dependent distributions from overlapping OAP probes . This was never done before. Likewise, the latter sentences above can be added to respond to the second major comment of the reviewer concerning the novelty of the study. The presented work cannot be done from CPI image data, since the CCD camera based instrument has such a small sample volume that you can’t retrieve size distributions in a heterogeneous cloud encountered on aircraft. Of course we are aware of the fact that CPI image data have better resolution with in addition 256 grey scales, which makes CNN different (and certainly extremely valuable) for high resolution CPI imager as compared to poor resolved OAP probe black and white images. CNN models for CPI probes produce different results than do CNN models for OAP probes, however, we have to keep in mind that quantitative results presented in section 3 of this study cannot be produced with CPI image data!
  
  Major Comment #2
  Likewise, we have responded to this major comment as follows:
  “The introduction has been improved in order to highlight the differences with previous methods that used OAP data. Note that obtaining morphology specific particle size distributions from OAPs (or any other instrument for that matter) was never shown in any other research publications. Przybylo et al. (2022) used CPI images, which cannot be used for quantitative concentration measurements (neither concentrations nor size distributions if not averaged over long time periods which is not compatible with cloud spatial heterogeneity encountered with aircraft); in particular, they do not produce particle size spectra (compare the 3 million usable images they obtained from eleven field campaigns to the 2.3 million used here and obtained in 7 flight hours; see next point).”
  The introduction mentions the few studies that have successfully applied CNNs to OAP images (Jaffeux et al. (2022) and Zhang et al. (2023)). Since OAP images they have very limited pixel numbers and also resolution, the CNN method for OAP images might be more challenging since the information content is not as rich as it is for CPI images (Pryzybylo et al. (2022)), and of course hardly comparable with diffraction patterns (Schmidt et al. (2024a,b)). Nevertheless, we will mention in the introduction these studies, having used a very similar methodology for the shape retrievals. These points will be made in the introduction as the methodology used in both papers is indeed similar. With the respect to novelty, the methodology was used for the first time to successfully train a single model based on image data sets from 4 different models of probes with different image resolution and pixel numbers. In addition, Jaffeux et al. (2022) and Zhang et al. (2023) limited their methodology to single probe data sets. Section 3 is beyond existing work on OAP based morphological analysis, as it takes advantage of the morphological information extracted from a coherent dataset using four different probes to reconstruct quantitative morphology-specific size distributions. These are presented for the first time and provide new insights in shape recognition with respect to observable size range and pixel resolution. Moreover, they are of great interest to the understanding of ice microphysics. The introduction will be corrected to reflect the importance of these latter points and in order to add some more selected relevant references.
  Minor Comment #1
  Acronyms are now defined from the beginning (which includes the abstract). In particular, probe and institute names will be fully written, the first time they are used.
  Minor Comment #2
  Grammar has been checked and corrected when needed
  
  Citation: https://doi.org/10.5194/egusphere-2024-1910-AC2

Peer review completion

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

AR by Louis Jaffeux on behalf of the Authors (20 Nov 2024) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (26 Nov 2024) by Wiebke Frey

RR by Anonymous Referee #2 (02 Dec 2024)

RR by Anonymous Referee #1 (07 Jan 2025)

ED: Publish subject to minor revisions (review by editor) (07 Jan 2025) by Wiebke Frey

AR by Louis Jaffeux on behalf of the Authors (31 Jan 2025) Author's response Author's tracked changes Manuscript

ED: Publish subject to minor revisions (review by editor) (14 Feb 2025) by Wiebke Frey

Dear Louis Jaffeux et al.,

thanks for providing the revised mansucript.
I have the feeling, that you did not fully use the potential of revising your introduction according to the referees suggestions. You might want to consider to try to add more context. Other than that, I have spotted some smaller issues, or needs for clarification, see below.

best wishes,
Wiebke Frey

(line numbers refer to the track change version)
line 1: "...and allows a more refined processing..."
This study allows processing? I think, 'allows' is not the right word here, please rephrase.

line 13: "from whatever OAP dataset" - remove 'whatever'

line 16/17: please remove "and thus extends..." I think that goes a bit too far.

line 47-49: Again, why do you change this sentence and its meaning? In the new version, it is not clear how knowledge of morphology is translating to improved weather forecast (do weather forecast model treat ice crystal habits at all?). Please change back to old version.

line 64ff: "CNNs ... automatically identify patterns and features ..." How do CNNs differ then from feature based algorithms (line 60)? Please clarify this.

line 71/71: "... sample volumes, however..." Please make it two sentences (...sample volume. However,...)

line 74: remove "also"

line 79: "This current study " -> The current study

lines 82-85: I think this sentence is not entirely clear, could you please rephrase or expand the explanation to make it better understandable.

line 90/91: remove sentence, "grey" literature; same for line 312 - grey literature should be avoided. To the least, the conference name and place should be given in the references.

line 94: why the change from "most of" the size range to "the full"?

Hide

AR by Louis Jaffeux on behalf of the Authors (24 Feb 2025) Author's response Author's tracked changes Manuscript

ED: Publish as is (04 Mar 2025) by Wiebke Frey

AR by Louis Jaffeux on behalf of the Authors (10 Mar 2025) Manuscript

Journal article(s) based on this preprint

03 Jun 2025

Convolutional neural networks for specific and merged data sets of optical array probe images: compatibility of retrieved morphology-dependent size distributions

Louis Jaffeux, Jan Breiner, Pierre Coutris, and Alfons Schwarzenböck

Atmos. Meas. Tech., 18, 2311–2331, https://doi.org/10.5194/amt-18-2311-2025,https://doi.org/10.5194/amt-18-2311-2025, 2025

Short summary

Louis Jaffeux, Jan Breiner, Pierre Coutris, and Alfons Schwarzenböck

Data sets

Public GitHub repository with data sets, codes, and trained CNN models Louis Jaffeux https://github.com/LJaffeux/JAFFEUX_et_al_AMT_2024

Louis Jaffeux, Jan Breiner, Pierre Coutris, and Alfons Schwarzenböck

Viewed

Total article views: 671 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
248	90	333	671	10	17

HTML: 248
PDF: 90
XML: 333
Total: 671
BibTeX: 10
EndNote: 17

Views and downloads (calculated since 18 Sep 2024)

Month	HTML	PDF	XML	Total
Sep 2024	56	7	23	86
Oct 2024	62	16	70	148
Nov 2024	34	13	54	101
Dec 2024	10	6	91	107
Jan 2025	22	7	83	112
Feb 2025	23	5	6	34
Mar 2025	11	16	0	27
Apr 2025	15	14	1	30
May 2025	12	6	5	23
Jun 2025	3	0	3

Cumulative views and downloads (calculated since 18 Sep 2024)

Month	HTML	PDF	XML	Total
Sep 2024	56	7	23	86
Oct 2024	62	16	70	148
Nov 2024	34	13	54	101
Dec 2024	10	6	91	107
Jan 2025	22	7	83	112
Feb 2025	23	5	6	34
Mar 2025	11	16	0	27
Apr 2025	15	14	1	30
May 2025	12	6	5	23
Jun 2025	3	0	3

Viewed (geographical distribution)

Total article views: 660 (including HTML, PDF, and XML) Thereof 660 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 03 Jun 2025

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (15125 KB)
Metadata XML

Short summary

Airborne cloud observation relies on high frequency black and white image information. The study presents automatic shape recognition tools developed with machine learning techniques and adapted for this type of images. Applied on a recent field campaign, these tools are compared across four instruments that cover different size ranges. The analysis show that the tools are performing well and are consistent across the different instruments.


Total:	0
HTML:	0
PDF:	0
XML:	0

Ice crystal images from optical array probes. Compatibility of morphology specific size distributions, retrieved with specific and global Convolutional Neural Networks for HVPS, PIP, CIP, and 2DS

Journal article(s) based on this preprint

Interactive discussion

Interactive discussion

Peer review completion

Suggestions for revision or reasons for rejection

Journal article(s) based on this preprint

Data sets

Viewed

Viewed (geographical distribution)