the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Jul 2026
Enhanced warm-temperature marine ice-nucleating particles over the Arctic Ocean versus Asian-dust outflow at Jeju Island from real-time shipborne observations
Abstract. Ice-nucleating particles (INPs) influence mixed-phase cloud microphysics and the Earth’s radiative budget, yet observational records over the remote Arctic Ocean and contrasting continental-outflow environments remain sparse. This study presents INP concentrations measured by the Portable Ice Nucleation Experiment (PINE) over the Arctic Ocean during two RV(Research Vessel) Araon campaigns (ARA15A in 2024 and ARA16A in 2025) and at the Korea-Cloud Physics Experiment Chamber (K-CPEC) on Jeju Island, Korea (February–May 2024), totalling 41,524 expansion runs of which 35,370 passed rigorous quality control. INP temperature spectra differed markedly among the three datasets. At a representative warm temperature (−20 °C), median concentrations were 3.68 L−1 (ARA15A), 3.64 L−1 (ARA16A), and 0.89 L−1 at Jeju (Clean-day subset), with Arctic values exceeding Jeju; at a representative cold temperature (−28 °C), the pattern reversed — the Jeju Clean-day median rose to 8.22 L−1 while Arctic medians remained comparatively flat (ARA15A: 7.02 L−1; ARA16A: 6.39 L−1). Campaign-wide medians were comparable between the two Arctic years (6.83 and 6.57 L−1) but differed approximately two-fold in the 75–80° N latitude band (ARA15A: 1.56 and ARA16A: 3.19 L−1). These spectral contrasts are consistent with a marine biogenic-influenced regime at the Arctic warm-temperature end and an Asian-dust-influenced regime at the Jeju cold-temperature end, although direct compositional confirmation (heat treatment, chemical speciation) was not available in this dataset; all source attributions should therefore be interpreted as spectrally-consistent hypotheses rather than established source identifications. A dedicated ship-exhaust quality control scheme applied to both campaigns provides a replicable contamination-removal framework for future shipborne INP studies, and the combined dataset delivers new observational benchmarks for Arctic INP variability across the sampled 2024–2025 cruise periods.
- Preprint
(5976 KB) - Metadata XML
-
Supplement
(2943 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-3543', Anonymous Referee #1, 27 Jul 2026
-
AC2: 'Reply on RC1', Joo Wan Cha, 20 Oct 2026
We thank Referee 1. Our point-by-point answers are in the attached RC1QnA.pdf; the corrections that neither referee raised are recorded separately in AC1. Section, table and figure numbers in it are those of the revised manuscript, which will be submitted to the editor after this discussion and differs in places from the preprint.
The data processing was wrong, and correcting it changed the paper. Referee 1 inferred from Figs. 3 and 4 that Arctic values below about 3 L⁻¹ had been removed, and warned that a dataset biased high could be masking the very temperature dependence whose absence we were reading as evidence of biogenic influence. Referee 2 reached the same conclusion from the same figures. The inference was correct. Expansions that counted no ice crystal are now retained in the denominator throughout, and no measurement is removed for being small. At T = −20 °C the ARA15A daily-pooled median falls from 3.68 to 0.099 L⁻¹, and the temperature dependence is present in the corrected spectra.
The mechanism was not quite the one Referee 1 proposed — there is no hard cut at 3 L⁻¹ — and the true cause is worse than a removal threshold. RC1QnA.pdf, items A2 to A4, sets out what it was.
The correction factors are withdrawn. Referee 1 wrote that the methodology is flawed and the analysis does not belong in this paper. We accept that without qualification. The correction-factor appendix has been removed and no inter-method scaling is applied anywhere; published filter values now enter as lower bounds, unscaled. Following the referee's recommendation we also examined whether a fixed factor could be defended at all, and it cannot: the online–offline offset changes sign with aerosol regime, not only magnitude, so no single number can represent it.
Length. Referee 1 noted the paper is occasionally verbose, and the revision is still longer than the discussion paper: 44 pages against 41, typeset identically. The additions are detection limits attached to every reported value, statements of what could not be verified, and the record of what changed. A reduction pass has been run and removed what could go without losing an answer: the intercomparison evidence behind Sect. 4.4 now sits in Supplement Sect. S19 with its conclusion retained in the main text, and the account of what was repaired in the data deposit has moved to AC1, where it belongs, rather than occupying the data-availability statement. What remains long is Sect. 3.3, left alone deliberately because it carries the detection-limit argument that answers Referee 1's central point.
Both of the referee's primary concerns changed the paper's conclusions rather than its presentation, which is the outcome we would have wanted from review. We are grateful for the care in the report.
Joo Wan Cha, on behalf of all co-authors
-
AC2: 'Reply on RC1', Joo Wan Cha, 20 Oct 2026
-
RC2: 'Comment on egusphere-2026-3543', Anonymous Referee #2, 09 Sep 2026
Review of Joo Wan Cha et al., “Enhanced warm-temperature marine ice-nucleating particles over the Arctic Ocean versus Asian-dust outflow at Jeju Island from real-time shipborne observations”
This paper adds important INP measurements to the Arctic, especially those in the online variety, and over two seasons. Overall, it is well written, and many portions seem fundamentally sound and included a nice ship contamination identification scheme. The Jeju dataset seems good, though it distracts a little from the overall Arctic focus. My main concerns that I will present are potential Arctic INP attributions limited to only marine biogenic and mineral dust, mischaracterization of offline methods, and the unusually high lower limit of Arctic INPs that is not consistent with the individual runs in Figure 4b.
Lines 39-53: Probably a good idea to include the MOSAiC work here in the introduction, either from Barry et al., 2025 and/or Creamean et al., 2022.
Barry, K. R., Hill, T. C. J., Kreidenweis, S. M., DeMott, P. J., Tobo, Y., and Creamean, J. M.: Bioaerosols as indicators of central Arctic ice nucleating particle sources, Atmos. Chem. Phys., 25, 11919–11933, https://doi.org/10.5194/acp-25-11919-2025, 2025.
Lines 323-325 and others: Campaign-wide median INP concentrations are not recommended. Instead, it is more interpretable to give concentrations at specific temperatures, which are already done, so I would remove these.
Figure 4: Why does the individual run data for the Arctic observations in Figure 4b show values lower than 1 L-1, which makes sense as a limit of detection for the PINE, whereas the other plots show the lowest concentrations at around 3 L-1? This seems like a major point that affects interpretation of this dataset and really does not make sense.
Lines 396-397, 440-442, and others: The authors qualify that source attribution isn’t definitive, but there are strong statements in this paper suggesting that anything sampled that is not mineral is marine biogenic. I would like to point out recent Arctic studies, including those over the central Arctic such as during MOSAiC, have found indications of terrestrial INPs throughout the year. This is detailed in Barry et al. 2025. Besides replacing Barry 2024, which was a student thesis, with more updated work, this is consistent with other Arctic work like Tobo et al. 2024 that attributed increased INP concentrations in the summer with snow and ice free land, and Perring et al., 2023 that found evidence of terrestrial bioaerosols in the summer Arctic.
Barry, K. R., Hill, T. C. J., Kreidenweis, S. M., DeMott, P. J., Tobo, Y., and Creamean, J. M.: Bioaerosols as indicators of central Arctic ice nucleating particle sources, Atmos. Chem. Phys., 25, 11919–11933, https://doi.org/10.5194/acp-25-11919-2025, 2025.
Tobo, Y., Adachi, K., Kawai, K., Matsui, H., Ohata, S., Oshima, N., Kondo, Y., Hermansen, O., Uchida, M., Inoue, J., and Koike, M.: Surface warming in Svalbard may have led to increases in highly active ice-nucleating particles, Commun. Earth Environ., 5, 516, https://doi.org/10.1038/s43247-024-01677-0, 2024.
Perring, A. E., Mediavilla, B., Wilbanks, G. D., Churnside, J. H., Marchbanks, R., Lamb, K. D., and Gao, R.: Airborne Bioaerosol Observations Imply a Strong Terrestrial Source in the Summertime Arctic, J. Geophys. Res.-Atmos., 128, e2023JD039165, https://doi.org/10.1029/2023JD039165, 2023.
Line 475 and others such as 660-662: Kawai et al., 2023 utilizes a parameterization from Tobo et al. 2019, which was from glacial outwash sediments. They attribute most of the IN activity to organic matter rather than mineral, which is inconsistent with your statements in the paper detailing minerals when discussing this reference.
Tobo, Y., Adachi, K., DeMott, P. J., Hill, T. C. J., Hamilton, D. S., Mahowald, N. M., Nagatsuka, N., Ohata, S., Uetake, J., Kondo, Y., and Koike, M.: Glacially sourced dust as a potentially significant source of ice nucleating particles, Nat. Geosci., 12, 253–258, https://doi.org/10.1038/s41561-019-0314-x, 2019.
Section 4.2: Best to give concentrations at associated temperatures.
Line 578: I would discuss a little more between the PINE and offline measurements made during PICNIC, such as the CSU Ice Spectrometer, to be consistent with the ending of the paragraph.
Lines 605-606; 640-642; 670-671; 704: Several places this comes up, but there isn’t the evidence or literature cited required to make these claims (or imply in some cases) that the INPs sampled were fragile marine that were degraded during storage, or undersampled by offline methods. These statements are way too strong and need to be scaled down. I agree that offline and online methods have been subjected to disagreement in many different environments, and especially in the Arctic, disagreements can be large and open questions certainly remain; however, this makes statements toward that the type is known and assumes things about how other datasets were handled (e.g. freeze-thaw denaturation during storage, background filter contamination) and implies that the others are incorrect. It would be good to look at the storage study done by Beall et al., 2020, of precipitation from coastal environments showing that -20 °C was the best storage practice. There are still questions about filter storage, for sure, but many Arctic studies show strong biological INP signatures, and a diversity of INPs at warm temperatures beyond fragile marine.
Beall, C. M., Lucero, D., Hill, T. C., DeMott, P. J., Stokes, M. D., and Prather, K. A.: Best practices for precipitation sample storage for offline studies of ice nucleation in marine and coastal environments, Atmos. Meas. Tech., 13, 6473–6486, https://doi.org/10.5194/amt-13-6473-2020, 2020.
Appendix: Make sure to update with Barry et al., 2025, and what is the difference between Filter IS and CSU Ice Spectrometer? They are identical but are shown separately here.
General: It might be useful to also compare with the Hartmann et al., 2021 online portion, detailing the SPIN.
Citation: https://doi.org/10.5194/egusphere-2026-3543-RC2 -
AC3: 'Reply on RC2', Joo Wan Cha, 20 Oct 2026
We thank Referee 2. Our point-by-point answers are in the attached RC2QnA.pdf; the corrections that neither referee raised are recorded separately in AC1. Section, table and figure numbers in it are those of the revised manuscript, which will be submitted to the editor after this discussion and differs in places from the preprint.
The data processing was wrong, and correcting it changed the paper. Referee 2 reached from Figs. 3 and 4 the same conclusion as Referee 1, that Arctic values below about 3 L⁻¹ had been removed. The inference was correct. Expansions that counted no ice crystal are now retained in the denominator throughout, and no measurement is removed for being small. At T = −20 °C the ARA15A daily-pooled median falls from 3.68 to 0.099 L⁻¹, and the temperature dependence is present in the corrected spectra. Our reply to RC1 sets out the mechanism.
The report found a second class of error, in how we read the literature. Three of its points were corrections to our citations rather than to our data, and all three are accepted:
- We described the glaciogenic-dust pathway of Kawai et al. (2023), which uses the parameterisation of Tobo et al. (2019), as a route to warm-temperature INP without a biological mechanism. Tobo et al. conclude the opposite — the ice nucleating ability of those sediments is governed by organic matter (<1 wt % organic carbon) rather than mineral components. Our text stated the reverse of its source.
- We attributed to Porter et al. (2022) and Creamean et al. (2022) the identification of heat-labile marine-biogenic INP as the dominant warm-temperature type. Creamean et al. conclude that local marine and episodic long-range transported terrestrial emissions together control the central Arctic INP population.
- We bounded the online–offline offset with a statistic that compares two online instruments. Referee 2's phrase “to be consistent with the ending of the paragraph” identifies exactly that gap.
On source attribution, Referee 2 is right that we were treating “not mineral” as “marine biogenic”. We have narrowed the qualifier wherever the claim was ours rather than the literature's, and added terrestrial bioaerosol as a fourth candidate mechanism, with the three references the referee supplied. Reading them sharpened the point: Perring et al. (2023) measured over the Bering and Chukchi Seas — the transit sector of both our campaigns — and found bioaerosol loadings consistently low in the marine boundary layer, with the elevated loadings aloft traced to passage over land.
What we have not done is weaken the biogenic part of the claim, because Barry et al. (2025) strengthens it: biological INP dominated the entire temperature spectrum in summer, close to 100 % heat-labile at −15 and −20 °C. The correction is to the qualifier, not to the mechanism.
The referee's concerns changed the paper's conclusions rather than its presentation, which is the outcome we would have wanted from review. We are grateful for the care in the report.
Joo Wan Cha, on behalf of all co-authors
-
AC3: 'Reply on RC2', Joo Wan Cha, 20 Oct 2026
-
AC1: 'Comment on egusphere-2026-3543 -AC', Joo Wan Cha, 20 Oct 2026
Final response, part 1 of 3: corrections of record
Our final response is posted in three parts. This comment records the corrections we found ourselves and the items that remain open. The point-by-point replies to the two referees are posted under each report, as replies to RC1 and RC2, with their answers attached. Following the journal's instruction, the revised manuscript is not attached to the discussion; it will be submitted to the editor in the usual way.
Title and central claim. The title changes from “Enhanced warm-temperature marine ice-nucleating particles over the Arctic Ocean versus Asian-dust outflow at Jeju Island from real-time shipborne observations” to “Low warm-temperature marine ice-nucleating particle concentrations over the Arctic Ocean versus Asian-dust outflow at Jeju Island from real-time shipborne observations”. The enhancement the preprint reported between −19 and −22 °C was not a property of the air: it was the single-crystal counting quantum of the temperature-bin estimator, which returns the reciprocal bin volume 1/V, about 4.42 L⁻¹, for any bin in which one crystal is counted, while more than half of the Arctic expansions counted none and were dropped (RC1 A4). On the daily-pooled estimator of the revised Sect. 2.6, which keeps those zeros, the Arctic concentrations at −22 °C are 0.100 and 0.126 L⁻¹ against 1.949, 9.722 and 7.439 L⁻¹ for the Jeju Clean, Polluted and Asian Dust categories, and there is no crossover at any temperature. The title now states what the data show; the abstract is rewritten on the same basis, and the campaign-wide medians it quoted are withdrawn (RC2 R2-2).
Corrections we found ourselves. AC1_Add.pdf records 62 items that neither referee raised, including a figure caption that still described the previous version of its figure, twelve cross-references to a supplement section that did not exist, and an uncertainty attributed to a paper that does not contain it. We list them because a reader of the preprint has nothing else to read against it.
Scale of the revision. Roughly half the manuscript text is new. All sixteen figures have been regenerated; a figure-by-figure list of what changed in each, and the item that explains it, is at the end of AC1_Add.pdf. The automated checks that recompute manuscript numbers from the archived products grew from 43 to 207, and each addition was made because it caught something. A marked-up version of the manuscript and of the Supplement against the preprint, and a consolidated author response keyed to the revised page and line numbers, will accompany the revised submission to the editor; following the journal's instruction, the revised manuscript is not attached here.
One open item bears on a reported number, and we state it rather than let it pass; four further bookkeeping items are listed at the end of AC1_Add.pdf. The MOSAiC comparison in Sect. 4.4 mixes estimators. Recomputing the archived CSU Ice Spectrometer record (DOE ARM, doi:10.5439/1804484) confirms the published figure exactly as a summer mean, 0.442 L⁻¹ at −20 °C over 18 filters. But that record is strongly skewed — its median over the same window is 0.080 L⁻¹ — and our own 0.099 L⁻¹ is a median. We have been setting a median against a mean. The comparison will be restated on one estimator. Those files carry a prior-permission stipulation from the PI, which we have requested; the revised manuscript will either quote the recomputed statistics with that permission or make the comparison qualitatively without them.
The preprint is not amended, on the handling editor's instruction; this comment, the two replies to the referees and their attachments are the correction of record for egusphere-2026-3543. The concentrations in the preprint's Table 2, Abstract, Figs. 3, 4 and 7, and the correction-factor treatment of Sect. 4.4 and Appendix A should not be used. Anyone citing the preprint should cite this comment with it.
Joo Wan Cha, on behalf of all co-authors
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 194 | 142 | 59 | 395 | 100 | 46 | 46 |
- HTML: 194
- PDF: 142
- XML: 59
- Total: 395
- Supplement: 100
- BibTeX: 46
- EndNote: 46
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “Enhanced warm-temperature marine ice-nucleating particles over the Arctic Ocean versus Asian-dust outflow at Jeju Island from real-time shipborne observations” by Joo Wan Cha and co-authors.
This paper investigates the distribution of INPs across contrasting remote-marine and continental-outflow environments. Using a Portable Ice Nucleation Experiment (PINE) online expansion chamber, researchers measured ambient INP concentrations over the Arctic Ocean during two shipborne research campaigns and compared them with observations on Jeju Island in South Korea. The authors contrast the measured INP concentrations at the two locations and provide analysis and discussion on the origin of the INPs (in the absence of compositional information). Finally, the authors include historical INP measurements from other studies to investigate a hypothesised relationship between sea ice retreat, marine activity, and INP concentrations in the Arctic. The paper is very well written — though perhaps a little verbose at times.
I have several major concerns with this study – to a degree that I will not provide a full review until my primary concerns have been addressed. I believe a portion of the PINE data measured in the Arctic may have been erroneously removed below an INP threshold. If this is the case, then the resulting dataset is severely biased, influencing all subsequent analysis, methodological choices, and conclusions. I also have major concerns about the justification for ‘correcting’ historical datasets to better fit with the PINE data. For these reasons, I do not recommend publication in its current form, but I look forward to the author’s responses in the hope that this can be published in the near future.
Data removal
For the Jeju expansions the procedure was 30s at a rate of 3Lmin-1, equating to around 1.5 L of air sampled. Assuming 1 ice crystal per volume sampled is the minimum that can be reported should give a minimum INP concentration of 0.66 L-1. This minimum is consistent with the typical reported minimum INP concentration reported by previous PINE users (see Herbert et al. 2026 ESSD – specifically Figure 2 with the ExINP-ZEP level 1 dataset). For the Araon expansions the procedure was 40s at a rate of 3Lmin-1, equating to around 2 L of air sampled per expansion and a minimum INP concentration of around 0.5 L-1. The two minimum measurable concentrations are similar, which is expected given the similar procedure.
Figure 4 (RHS – or 4b) confirms this – the lowest temperature-independent ‘stripe’ of INP data falls around 0.5 L-1 throughout both of the campaigns – which corresponds to 1 ice particle being measure during the expansion (The Jeju data minimum is lower than expected and will be discussed further down). So far, the Araon data presented looks sensible. However, comparing the Level 1 Araon data from Figure 4b with that of Figures 3b and 3d suggests that all Aaraon data below N_INP = 3 L-1 was removed – but only for the data measured within the Arctic (Figure 4 LHS – or Fig 4a). Section 2.6 alludes to the source of this – a Poisson-based LOD was applied to the data to ensure that more than 1 ice crystal was being measured during the 24 resampling period, which I presume ends up being around 3 L-1. If this is the case, then I have a few questions.
The removal of this data results in a very different distribution – as can be seen in Figure 3b and 3d – and influences the entire narrative of the paper. Removing the data below 3 L-1 strongly biases the dataset towards higher values – and may result in the largely temperature-independent INP spectrum used in all subsequent analysis in the study. The absence of strong temperature-dependence is routinely seen as evidence of biogenic influence, yet if the data is biased high then this may be masking the true temperature dependence. The authors need to look at this again and, if correct, provide more rationale as to why so much of this perfectly good looking data has been removed.
Conversely, the Jeju Level 1 dataset should have a similar INP distribution as that shown in Figure 4b for the Araon data – the lowest ‘stripe’ should fall around 0.75 L-1. However, inspecting Figures 3e to 3i and the box plots in Figure 4a suggest a lower limit of around 0.15 L-1. If this is from the Level 1 data then this implies a much larger volume was sampled (around 6 L). Can the authors check and confirm the data is correct? And can the level 1 Jeju data be included on Figure 4 as a third panel?
Herbert et al., 2026: https://essd.copernicus.org/preprints/essd-2026-41/
Including zero counts
The PINE expansions routinely result in an INP concentration of 0. This simply means that there were no ice crystals measured during the expansion. This is still useful data and should be included in the analysis – other PINE users have resampled the timeseries and taken means of the INP concentrations, including the zero INP expansions. Figure 2 from Herbert et al. 2026 demonstrates this. I suggest the authors consider performing this extra step before presenting their data as it will help remove some of the temperature independence that is seen in the dataset.
Correction factors
Although I agree that it would be very useful to ground the PINE observations with the historical record of filer-based methods, I don’t believe there is sufficient justification for combining INP datasets by including correction factors to bring them in line – to then look at something as fundamental as linking retreating sea ice with INP concentrations in the Arctic. The methodology is flawed and the current analysis does not belong in this paper.
From Section 4.4: "This integrated view frames the central methodological question of this section—how to place online expansion-chamber (PINE) and offline filter-based INP measurements on a common, quantitatively defensible scale — and both motivates and brackets the fixed harmonisation factor (CF = 16) derived in the remainder of this subsection." The simple answer is that they should not be manipulated and massaged to be put onto a common scale. If they do not match, then there are issues with one or more of the instruments, techniques, or analysis. I recommend the authors read Robinson et al. (2026) – they provide evidence for why PINE and other methods may give different results – especially at warmer temperatures where (for instance) the Wilbourn ENA comparison has the greatest discrepancy. Their results suggest there is a strong element of temperature-dependence to the differences, as well as from the INP species / material. If this is the case, then a systematic CF, as applied in this study, will not be able to account for these differences.
With regards to the justification in Section 4.4. “First, the PINE chamber samples a broader size and freezing-mode population than immersion-only offline filter assays”. What size ranges does PINE measure? How does this compare with filter-based sampling techniques? PINE has significant losses of particles around 5 micron (see Mohler et al., 2021 and Knopf et al., 2021 SI) – is this better than the filter-based methods? Is pore-condensation and deposition mode heterogenous nucleation truly likely to cause more freezing events in PINE than in other methods? This kind of statement needs to be supported by evidence.
Having said this – if the data below 3L-1 has been erroneously removed from the Araon datasets (see previous comment) then it might be that the PINE data falls closer to the filter-based data. I recommend the authors look into their data to ensure it is correct, then revisit this analysis to see whether CFs are even needed. I do not believe this paper will be publishable with its current use of correction factors.
Robinson et al., 2026: https://ar.copernicus.org/preprints/ar-2026-17/
Mohler et al, 2021: https://amt.copernicus.org/articles/14/1143/2021/
Knopf et al., 2021: https://doi.org/10.1175/BAMS-D-20-0151.1