Vertical structure and controlling factors of cloud condensation nuclei activation over eastern China: Insights from aircraft measurements and interpretable machine learning

Lu, Yichen; Wang, Honglei; Liu, Deyu; Chen, Yue; Peng, Chong; Song, Can

doi:10.5194/egusphere-2026-4682

Preprints

https://doi.org/10.5194/egusphere-2026-4682

Preprints

06 Aug 2026

| 06 Aug 2026

Vertical structure and controlling factors of cloud condensation nuclei activation over eastern China: Insights from aircraft measurements and interpretable machine learning

Yichen Lu, Honglei Wang, Deyu Liu, Yue Chen, Chong Peng, and Can Song

Abstract. Cloud condensation nuclei (CCN) play a critical role in cloud droplet formation and microphysical processes. Based on aircraft observations, this study investigated the factors controlling CCN number concentrations (NCCN) under different aerosol vertical structures (Decrease, Increase, and Stable) and supersaturation (SS) conditions using generalized additive models (GAM) combined with SHapley Additive exPlanations (SHAP).NCCN reached up to 10³ cm⁻³ near the surface and generally decreased with altitude, while aerosol vertical structures modulated its abundance, with the Increase structure showing higher NCCN than Stable and Decrease structures. CCN activation ratios increased with SS and exhibited a non-monotonic vertical variation, with no consistent ranking among aerosol structures, indicating that supersaturation dominates CCN activation. Activated CCN droplet spectra showed unimodal distributions, with peak diameters increasing from ~2 μm at SS = 0.2 to ~5 μm at SS = 1.0. Although spectral shapes were similar among different structures, higher small-size aerosol concentrations (SA) enhanced CCN peak concentrations. GAM results identified temperature (T), SA, relative humidity (RH), and horizontal wind speed (WS) as important explanatory variables for NCCN variations, with contributions of 20 % – 56 %, 9 % – 45 %, 9 % – 40 %, and 3 % – 19 %, respectively. SHAP analysis revealed that the contributions of T varied among different aerosol vertical structures, showing positive associations under Decrease and Increase structures but negative associations under Stable conditions, whereas SA consistently exhibited positive contributions. RH showed nonlinear relationships with NCCN, with an inflection point near 60 %.

Received: 03 Aug 2026 – Discussion started: 06 Aug 2026

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 paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 1876 KB)

Supplement (4555 KB)

Download & links

Yichen Lu, Honglei Wang, Deyu Liu, Yue Chen, Chong Peng, and Can Song

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-4682', Anonymous Referee #1, 09 Sep 2026
Comment to “Vertical structure and controlling factors of cloud condensation nuclei activation over eastern China: Insights from aircraft measurements and interpretable machine learning”

This manuscript presents a valuable observational study investigating the vertical distribution of cloud condensation nuclei (CCN) and their controlling factors over eastern China, using aircraft measurements combined with interpretable machine learning techniques (Generalized Additive Models, GAM, and SHapley Additive exPlanations, SHAP). The topic is highly relevant to aerosol–cloud–climate interactions, and the integration of aircraft observations with explainable AI methods represents a timely and scientifically sound approach. The writing is generally clear. I would like to recommend its acceptance for publication with necessary modifications.

General comments
As most aircraft observation based studies, the sample size is too limited. The analysis is based on only four aircraft flights conducted over a single week in March 2018 over eastern China. While the authors acknowledge this limitation, the restricted temporal and spatial coverage raises concerns about the generality of the findings. Seasonal variations, different synoptic regimes, and regional differences across eastern China are not captured.

The CCNC cycled through five supersaturation levels sequentially (5 minutes each) rather than measuring them simultaneously. As the authors note, this creates a partial coupling between altitude and SS because the aircraft changed altitude during the 25-minute measurement cycles. This complicates the interpretation of vertical profiles and may introduce artifacts in the composite statistics.

In aircraft based observations, temperature and altitude are physically correlated. Although the GAM/SHAP framework identifies temperature as the dominant predictor, the manuscript appropriately cautions that this reflects statistical association rather than direct causation. However, the discussion could benefit from a more explicit analysis of partial effects or sensitivity tests that decouple these variables.

Provide confidence intervals or bootstrap estimates for the SHAP contribution percentages to better convey the robustness of the factor rankings.

Detailed comments
Line 33-34, Regarding the aerosol impacts, recent review studies could serve as important supporting references, such as Zhao et al. (2024, doi: 10.1016/j.scib.2024.03.014) and so on.
Line 37-39, Regarding this point, Negative Aerosol-Cloud re Relationship from Aircraft Observations over Hebei has been found, which is worthy to mention.
Line 39-40, In addition to the decrease of precipitation, aerosol invigoration effect on precipitation should also be briefly introduced for fair.
Line 42, Citation format is not right, which should be corrected (also other citations).
Line 45-50, Actually, different even opposite findings regarding aerosol-cloud interactions have been demonstrated by previous studies, which should be briefly introduced.
Line 56, based on Köhler theory, not only aerosol size, but also aerosol chemical composition affects the CCN activity.
Line 61-62, There are also laboratory experimental characterizations.
Line 62-64, For fair, the limitation of aircraft observations should be also discussed.
Line 74-75, Supporting evidence should be given for this claim.
Line 79-80, What do the authors mean “CCN research”?
Line 81, Ren et al. 2025 is not listed in the reference list.
More background information about the aircraft observations should be given, particularly over east China regions.
Line 108-134, Uncertainties in the instrument measurements should be briefly introduced.
Line 135-140, How do the authors solve the measurement issues in the first or second bins, along with the shattering of ice crystals?
Line 145, How reliable for this cloud identification method? Or how sensitive are the results to threshold values?
Line 205, UTC time or Beijing time?
Citation: https://doi.org/10.5194/egusphere-2026-4682-RC1
- AC1: 'Reply on RC1', Honglei Wang, 08 Oct 2026
  
  Dear Editor and Reviewer,
  We sincerely thank you for the careful evaluation of our manuscript and for the constructive and valuable comments and suggestions. We have carefully read and considered all of the comments and have revised the manuscript accordingly. We believe that these revisions have helped improve the clarity, rigor, and overall quality of the manuscript.
  A detailed, point-by-point response to each comment, along with the corresponding revisions made in the manuscript, is provided in the attached PDF file. We sincerely appreciate the time and effort you have devoted to reviewing our manuscript and hope that the revised manuscript is now suitable for publication.
  Thank you again for your valuable comments and suggestions.
  Sincerely,
  
  The Authors
  
  Citation: https://doi.org/10.5194/egusphere-2026-4682-AC1
RC2:
'Comment on egusphere-2026-4682', Anonymous Referee #2, 23 Sep 2026

This study examines the factors associated with variations in cloud condensation nuclei (CCN) number concentrations across different altitudes, aerosol vertical structures, and supersaturation conditions. A key strength of the work is the application of nonlinear multivariate analysis using Generalized Additive Models (GAMs), together with SHapley Additive exPlanations (SHAP), an approach used to quantify the contribution of individual variables to model predictions, to aircraft observations. These methods allow the authors to evaluate the relative importance of multiple aerosol and meteorological variables while accounting for nonlinear relationships. The analysis provides useful insight into the factors associated with vertical CCN variability and contributes to the broader understanding of aerosol-cloud interactions. However, I have several comments regarding the methodology and interpretation of the results that should be addressed before publication.

Abstract
Several terms in the abstract should be more clearly defined. Specifically, the “Decrease,” “Increase,” and “Stable” aerosol vertical structures are introduced without explaining what these classifications represent or how they are defined. In addition, “small-size aerosol concentrations (SA)” is too vague; please specify the particle diameter range included in this definition here.

Line 24-26. “GAM results identified temperature (T), SA, relative humidity (RH), and horizontal wind speed (WS) as important explanatory variables for NCCN variations, with contributions of 20%–56%, 9%–45%, 9%–40%, and 3%–19%, respectively.”
The use of “respectively” here is difficult to follow given the number of variables and contribution ranges. Consider restructuring the sentence to pair each variable directly with its corresponding range for clarity.

Line 42.
Please revise the citation formatting when references are incorporated into the sentence. For example, “(Prabhakaran et al., 2025) found that...” should be written as “Prabhakaran et al. (2025) found that...”. Please check for and correct this citation formatting throughout the manuscript.

Line 47. ”…(2003)observed…”
Change to ”…(2003) observed…”. Please check that there is a space between citations and the surrounding text throughout the manuscript (also seen in Line 68).

Line 65. “… T structure…”
Please define temperature (T) at its first occurrence in the main text. Although T is defined in the abstract, abbreviations should also be defined when first introduced in the main text.

Line 81.
Ren et al. (2025) is missing from the reference list.

Line 184. “…while SA, MA, and LA were…”
Please define all these abbreviations here instead of Line 306. SA is only defined in abstract.

Lines 210-220.
The description of the CCN activation ratio as exhibiting an “increase-decrease-increase” pattern is overly qualitative. Please specify the relevant altitude ranges and quantify the corresponding range and magnitude of changes in the activation ratio across these altitudes. This would provide a clearer assessment of the vertical variability than simply describing the direction of change.

Lines 224-226.
The criteria used to classify adjacent bins as Stable (<100 cm^-3 absolute difference and <20% relative change) appear somewhat arbitrary. Please provide justification for these threshold values, including whether they are based on instrument uncertainty, prior literature, or the distribution of the observations.

Lines 256-268. “Fig. 4 shows that the size distributions of CCN-activated droplets measured by the CCNC are generally unimodal. The modal diameter is approximately 2 μm under low SS conditions (0.2 and 0.4), whereas it shifts to approximately 5 μm under higher SS conditions (0.8 and 1.0).”
The droplet size distributions do not appear strictly unimodal in all cases. Particularly at lower SS, there appears to be a smaller mode or shoulder at <2 μm in addition to a broader population at 2-3 μm. Could this mode becoming less apparent at higher SS also be related to the enhanced condensational growth described later in this section?

Line 305-306. “…medium-size aerosol numberconcentration (MA), and large-size aerosol number concentration (LA)...”
Please define SA when it is first introduced in the main text and provide the particle size ranges corresponding to SA, MA, and LA. Currently, SA is only written out in the abstract, and the size ranges for these categories are only provided in Fig. 5.

Line 346-349. “Specifically, in the Decrease and Stable scenarios, T exhibits the highest contribution, exceeding 25% in both cases. In the Increase scenario, T’s contribution ranking decreases under certain SS conditions but still remains above 15%, suggesting that T maintains high importance across different scenarios and is closely associated with variations in NCCN”
Figure S1 shows that CCN number concentration generally decreases with altitude, as also discussed previously in the manuscript. Because temperature is also expected to vary systematically with altitude, please address the potential covariability between altitude and temperature and how this may affect the inferred importance of temperature for CCN number concentration. In particular, the high contribution attributed to temperature may partly reflect the vertical dependence of CCN number concentration rather than a direct temperature effect. This is only discussed in the conclusion. If sufficient data are available, consider including altitude as an additional model variable to help separate these effects.

Figure S9.
It is difficult to evaluate the shape of the particle number size distributions in Fig. S9 because particle diameter is plotted on a linear rather than logarithmic scale. I recommend plotting diameter on a logarithmic scale to better assess the reported unimodal distribution.

Figure 6
I recommend using conventional SHAP summary (essentially a beeswarm) plots, which would show both the magnitude and direction of each predictor’s contribution to the modeled CCN concentration, rather than only the aggregated contribution percentages shown in Figure 6. Percentages can just be added as text on the plot.

Citation: https://doi.org/10.5194/egusphere-2026-4682-RC2
- AC2: 'Reply on RC2', Honglei Wang, 08 Oct 2026
  
  Dear Editor and Reviewer,
  We sincerely thank you for the careful evaluation of our manuscript and for the constructive and valuable comments and suggestions. We have carefully read and considered all of the comments and have revised the manuscript accordingly. We believe that these revisions have helped improve the clarity, rigor, and overall quality of the manuscript.
  A detailed, point-by-point response to each comment, along with the corresponding revisions made in the manuscript, is provided in the attached PDF file. We sincerely appreciate the time and effort you have devoted to reviewing our manuscript and hope that the revised manuscript is now suitable for publication.
  Thank you again for your valuable comments and suggestions.
  Sincerely,
  
  The Authors
  
  Citation: https://doi.org/10.5194/egusphere-2026-4682-AC2

Yichen Lu, Honglei Wang, Deyu Liu, Yue Chen, Chong Peng, and Can Song

Supplement

https://doi.org/10.5194/egusphere-2026-4682-supplement

Yichen Lu, Honglei Wang, Deyu Liu, Yue Chen, Chong Peng, and Can Song

Viewed

Total article views: 239 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
130	82	27	239	59	22	21

HTML: 130
PDF: 82
XML: 27
Total: 239
Supplement: 59
BibTeX: 22
EndNote: 21

Views and downloads (calculated since 06 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	43	13	3	59
Sep 2026	60	47	14	121
Oct 2026	27	22	10	59

Cumulative views and downloads (calculated since 06 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	43	13	3	59
Sep 2026	60	47	14	121
Oct 2026	27	22	10	59

Viewed (geographical distribution)

Total article views: 226 (including HTML, PDF, and XML) Thereof 226 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 10 Oct 2026

Short summary

Clouds are affected by tiny airborne particles that influence cloud formation and climate. Using aircraft observations over eastern China, this study reveals how the vertical distribution of atmospheric particles controls their cloud-forming ability. The findings improve our understanding of aerosol–cloud interactions and provide new insights for weather and climate modeling.


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