the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Sep 2025
The State of Diversity, Equity, and Inclusion in the Cloud Physics Community
Abstract. The Geosciences are amongst the least diverse research fields, where women and other underrepresented 19 groups face systemic biases. This paper presents the state of diversity, equity, and inclusion (DEI) in the cloud-physics community, by combining a metadata analysis of 6987 cloud-physics peer-reviewed journal articles published between 1970 and 2020 with responses from a survey of 198 participants from the cloud-physics community. Women first author contributions are evident only after 1997 and presently only ∼17 % of studies in the cloud physics field are led by women. Authors from the Global North dominate first and corresponding-author positions, with only ∼5 % of studies led by tropical affiliation authors. The latter’s participation was low even for study sites in the tropics, suggesting widespread practice of parachute science. Of the survey respondents, 23 % identified as a minority group and feel that being a minority has had a negative impact on their scientific career, in terms of collaborations, promotions, publishing, funding, salary, and citations. Although the survey data shows the general experiences of cloud physicists globally, the perspectives from this work can aid the cloud-physics community to develop strategies to improve DEI in institutions beyond a business case for a diverse science community. Rather we should consider an equity-centered approach by understanding our ethical responsibilities to benefit research of the climate system.
-
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
(1804 KB)
-
Supplement
(475 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1804 KB) - Metadata XML
-
Supplement
(475 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-4499', Anonymous Referee #1, 12 Oct 2025
This manuscript presents an important and timely assessment of diversity, equity, and inclusion (DEI) in the cloud physics community, combining a metadata analysis of 6,987 peer-reviewed papers (1970–2020) with a community survey of 198 respondents. The topic is highly relevant, and the effort to integrate quantitative and qualitative perspectives is commendable. However, several conceptual and structural issues limit the manuscript’s clarity and impact in its current form. In particular, the terminology used to describe researchers from the Global South, the framing of gender analysis, and the interpretation of small-sample survey results require substantial revision. Addressing these concerns would considerably strengthen the analytical rigor, inclusivity, and global sensitivity of the paper.
Major Comments
- Terminology – “Tropical Scientists”: In my opinion, the use of the term “tropical scientists” is highly problematic. I find it both unnecessary and ill-considered that the authors have chosen to introduce a new label that lacks conceptual grounding and adds no clarity to the discussion. As someone working within this geographic zone, I find this terminology technically inaccurate and, frankly, counterproductive—it conflates latitude with social and institutional realities, and risks exoticizing researchers from the Global South. In my view, this framing reflects a misunderstanding of the issue it seeks to address. Established terms such as Global North/South, high-, middle-, or low-income countries (as defined by the UN or World Bank), or developed/developing countries—though imperfect—are far more appropriate, as they rest on recognized metrics and are widely understood in academic and policy contexts. I strongly recommend removing the term entirely and clarifying the distinction between tropical study sites and researchers from the Global South, which are not interchangeable categories.
- Use of “Minorities” vs “Marginalized”: The manuscript frequently uses the term “minorities.” Given the global context of the study, where numerical minorities may differ by region, the term “marginalized” would be more appropriate and inclusive. It emphasizes systemic disadvantage rather than numerical representation. This change would also align better with the ethical framing of equity in the paper’s conclusion.
- Gender Analysis and Non-Binary Representation: The gender analysis appears to be strictly binary (men/women). While this may have been constrained by data availability (e.g., use of genderize.io), the discussion section should at least acknowledge and briefly discuss the exclusion of trans and non-binary individuals. For instance, the authors could note the limitations of algorithmic gender inference, and mention issues such as misgendering, the use of dead names in publication records, and the broader underrepresentation of trans scientists in the geosciences. This would strengthen the inclusivity and awareness reflected in the DEI framing.
- Survey Interpretation and Sample Size Limitations: The metadata analysis is the stronger and more robust component of the paper. However, the survey-based results (N=198) present significant limitations when further divided by gender, country, and ethnicity. When broken down into subcategories, the sample size per group, in my opinion, becomes too small for reliable interpretation or for drawing broad conclusions about regional or demographic patterns. I suggest restructuring the Results section so that the metadata analysis comes first—forming the backbone of the findings—followed by the survey results as complementary qualitative insight. The discussion could then focus on how these two strands inform each other, rather than treating both as equally representative.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC1 -
EC1: 'Thanks for RC1', David Crookall, 12 Oct 2025
Thank you for your excellent review.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC1 - AC1: 'Reply on RCI', Zamin A. Kanji, 15 Apr 2026
- AC1: 'Reply on RCI', Zamin A. Kanji, 15 Apr 2026
-
RC2: 'Comment on egusphere-2025-4499', Anonymous Referee #2, 28 Oct 2025
In this study, the authors present the current state and over time development of Diversity, Equity, and Inclusion (DEI) in the cloud physics community by considering two different data sets: metadata of 6,987 published papers between 1970 and 2020, and the survey responses from 198 participants. This research is significant for the advancement of the geosciences as it helps identify areas that need improvement. However, there are some issues with the framing of certain aspects, including the terminology used. Therefore, my general comments align with the Rev1.
General comments:- Underrepresented instead of the word minority or marginalised: Minority means that the smaller number or part. As Rev1 mentioned, minorities differ from region to regionIt is important to note that labelling the global south as a minority is a Western-centric viewpoint. The groups considered in this context are not smaller in number; they are underrepresented, overlooked, or lacking visibility. Additionally, "marginalised" implies insignificance or a peripheral status, which also doesn’t adequately convey the issue at hand.
- Terminology tropical scientists/countries: The term "tropical" carries connotations that could present an "exotic" portrayal, which is not appropriate here. I completely agree with Rev1. Additionally, framing a comparison of “tropical countries” with so so-called global north is a false dichotomy. One is defined by economic and historical factors under neo-colonial structures, while the other is geographical.
- The gender analysis presented is binary and fails to include non-binary and transgender individuals. This omission is not addressed in the paper. Although the potential for misgendering due to the algorithm is acknowledged, it still contains uncategorized (gender-blind) names. Furthermore, such gender-inferencing algorithms often struggle with the accuracy of non-Western names.
I will try to elaborate on what I mentioned in the general comments by giving specific examples from the first couple of pages, which are valid throughout the entirety of the paper.
line 46: ".... minority representation..." This gap exists for any underrepresented group, not just minorities. You might not be a minority, but underrepresented, you might be a minority if you are looking from a Western lens. We are great in numbers, but our work or existence is not represented or acknowledged.
line 65: same issue here with the use of minorities
line 70: Women are underrepresented but larger in number, exactly like the scientists from the so-called global south. They are underrepresented, and there is a lack of acknowledgement.
line 81: In the cited paper, this is mentioned as "USA, UK, ... However, these are not necessarily the most collaborative countries, if measured by their proportion of collaborative output... The analysis provided evidence that countries rated high in terms of scientific development were more likely to collaborate." This does not necessarily mean low income with low scientific development.
line 86&96: Egypt and South Africa are not situated in the tropics but the subtropics. When comparing these regions with the global north, framing it as a dichotomy could be misleading, aka a false dichotomy. Additionally, the term "tropical" carries connotations that could present an "exotic" portrayal, which is not appropriate here.
line 189: Latitude range includes Australia, which is considered to be a "global north" country. If the intention is to analyse underrepresentation, this could lead to misrepresentation in the analysis. Although the countries are highlighted later in the table, it would be beneficial to clarify this point earlier on.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC2 -
EC2: 'Thank you for RC2', David Crookall, 11 Nov 2025
Thank you for your excellent review.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC2 - AC2: 'Reply on RC2', Zamin A. Kanji, 15 Apr 2026
- AC2: 'Reply on RC2', Zamin A. Kanji, 15 Apr 2026
-
RC3: 'Comment on egusphere-2025-4499', Anonymous Referee #3, 11 Nov 2025
This paper presents an analysis into the current climate and feelings of belonging within the specific geoscience community of cloud physics. It utilizes two different approaches to address this question: a meta-analysis of peer-reviewed publications and an online survey. The question addressed in the paper is a unique contribution to the field, however there are several significant issues that need to be revised and addressed before publication. In addition to the previous two referee comments above, I will add the following:
- There has been a surplus of more recent work that highlights the their point in lines 45-47 (Although various programs, policies, and inclusive environments are being developed to address issues such as gender inequality in academia, many gaps in other minority representation still remain (e.g., Hispanic, black, and indigenous communities; Bernard and Cooperdock, 2018; NCSES, 2019; Gewin, 48 2020; Odekunle, 2020). Please include an updated literature review from the last 5 years in this.
- I agree with the previous two referees that “marginalized” is a more appropriate term than minority. Based on the descriptors provided, another alternative would be to say “historically excluded” groups. This is a better descriptor that is commonly used and recognizes the fact that these groups have been intentionally excluded by the system, as opposed to just individual actions.
- In line 55 – able-bodied individuals are also significantly more represented in all fields than in the general population. There is literature on this as well and should be included in the literature review.
- Lines 58-59 (The lack of diversity and inclusion across academic communities manifests itself in systematic bias of under-represented minorities). I disagree with how this comment is written – the system was intentionally designed to prevent lack of diversity and inclusion from the very beginning. The system is the problem, which further exacerbates the lack of diversity and inclusion we see today.
- I think it would also be beneficial to expand on the background of cloud physics as a research field. It seems that your assumption is that everyone in this field trained in geosciences, or is there a combination of physics, chemistry, geography, etc? Each of these fields has presented its own variations in why they have issues related to DEI and should be included as a background variable when considering the climate of cloud physics.
- Your response rates are not clearly defined and need to be included. How many individuals were included in the ICCP listserv and the other announcement sent? How many participants actually registered for the 2021 meeting? Comparing this number to other membership numbers is not sufficient – especially when the data is anonymous and you cannot guarantee that individuals did not take the survey more than one time. I don’t agree with the statement that it represents 30-47% of the community as a whole – you should only report on response rates from who the survey was disseminated to.
- Lines 137-139: you say the number of respondents who were Indian was higher – but how do you know that if the data is anonymous? Please clarify.
- Lines 206-208: What are the actual stats presented by Dutt, 2020? The ratio of white/non-white seems higher than what the ratio is in the US.
- Figure 1: The color purple is used to signify two different things – please use different colors for the “total” versus “other” gender category.
- The paper published by Marin-Spiotta et al (2023) in Earth’s Future provides survey results on the workplace climate in the geosciences. You should include this in your discussion/analysis as a comparison since their survey results include over 2000 survey responses.
- Did you ask if survey respondents had experienced bullying or harassment within a specified time frame or ever? This should be further clarified and justified on either approach.
- Line 324 – the leaky pipeline analogy is problematic. Berhe et al (2024) describes this in their Nature Geoscience paper and recommends a better analogy.
- Line 335-336. I disagree that you have the numbers to state that you are representing the cloud physicists globally, based on my previous comment about response rates. This should be removed.
- Given the broad demographics of who this survey was disseminated to, I think you should expand on potential limitations of your findings on feeling included. Your question was simply “Do you feel included in the cloud physics community?” This is really vague and could be interpreted in so many different ways – I am less inclined to say that this is a valuable finding with this approach. What do you mean when you say included? Just in the workplace, in professional societies, when in school? This is not well defined.
- Why did you only ask about working with women in your survey? This is not inclusive to the entire gender spectrum.
- What does “sensitive to gender issues” mean? Again as above – this is so vague and could be taken to mean so many things. I don’t agree that this should be presented in this dataset, it’s not a well designed survey question.
- Question 40 in your survey – were you only talking about race/ethnicity here? Or was this applicable to any and all identities? Not clearly defined, which is problematic.
- You inherently introduced a lot of bias in your survey by using the term “tropical” to describe an entire subset of people. This is problematic.
- You say that the majority of people said no to question 67. Why did you not include an additional clarifier on that to explain their perspective? There could be many reasons why participants selected no, I don’t think it’s appropriate to speculate on that in your paper.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC3 -
EC3: 'Thank you for RC3', David Crookall, 11 Nov 2025
Thank you for your excellent review. Much appreciated.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC3 - AC3: 'Reply on RC3', Zamin A. Kanji, 15 Apr 2026
- AC3: 'Reply on RC3', Zamin A. Kanji, 15 Apr 2026
-
RC4: 'Comment on egusphere-2025-4499', Anonymous Referee #4, 07 Jan 2026
This review is both significant and timely. However, it demonstrates the tendency within academia to use archaic language and framing when analyzing geographic, racial, gender, and sexuality DEI issues globally. To address these, significant revisions are needed before publication. Specifically, the manuscript will be more impactful if the authors address the implicit, inherited, and adopted racism and sexism entrenched in the sciences.
Gender Methodology Issues
Lines 182-186: "It is important to note that genderize.io assigns the gender based on the person's first name. Although this could be a good approximation, names cannot always be used to infer the gender (e.g., Sebo, 2021; Lockhart et al., 2023; Marty et al., 2023). Therefore, the results presented below must be taken with caution and considering the aforementioned bias."
Issue: This perpetuates binary gender assumptions and doesn't acknowledge how this methodology erases non-binary identities and reinforces cisgender normativity. The authors should explicitly state this limitation.
Problematic Categorizations
Lines 150-153: "Therefore, in further analysis, we classify the Caucasian, White or European categories as 'White', while the other ethnicities are classified as 'non-white' with the exception of participants who preferred not to identify their ethnicity."
Issue: This binary centers whiteness as the default and groups all other identities into a single "other" category, which obscures important differences in experiences among diverse racial/ethnic groups. This approach reinforces white supremacy frameworks.
Minority Definition
Lines 50-54: "The word 'minority' denotes negatively stigmatized, ostracized, oppressed, and outcast individuals or groups (Blanz et al., 1995). In the context of DEI, it refers to a numerically smaller population of a group within a larger population based on characteristics such as gender, race, religion, ethnicity, disability, sexual orientation, or other attributes. In the present study, 'minority' is used to describe an underrepresented or disadvantaged group, unless stated otherwise."
Issue: While acknowledging negative connotations, the definition still focuses primarily on numerical representation rather than power dynamics and systemic oppression. Missing discussion of how majority groups maintain power regardless of numbers.
Global South Terminology
Lines 66-69: "Systemic bias, under-representation of minorities, and a lack of recognition of researchers in and from developing countries (the so-called 'Global South') are reflected in the limited scientific collaborations between research groups from developed (the so-called 'Global North') and developing countries (Gewin, 2023)."
Issue: Uses "so-called" but doesn't interrogate the problematic colonial legacy embedded in these geographic/economic binaries or how they perpetuate deficit narratives.
Parachute Science
Lines 393-408: The discussion of parachute science identifies the practice but doesn't explicitly name it as a continuation of colonial extraction practices or discuss how implicit biases about expertise and knowledge ownership enable these dynamics.
Specific concern at lines 393-397: "Our results suggest that the so-called helicopter or parachute science... is unfortunately common in the cloud physics community; however, a deeper analysis is required to confirm this."
Issue: Needs explicit statement about colonial legacies and power imbalances, not just a neutral "deeper analysis required."
Intersectionality
Lines 284-286: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern."
Issue: Intersectionality is mentioned but not systematically applied throughout the analysis. The entire study design should have incorporated intersectional analysis from the beginning, not as an afterthought.
Harassment Analysis
Lines 276-278: "A higher fraction of women, 44% compared to 27% of men reported having experienced harassment, a difference that was statistically significant (χ²(1) = 5.70, p = 0.017, Cramér's V = 0.18), although the size of the gender difference is modest."
Issue: Doesn't explicitly discuss how sexism and misogyny create hostile environments or how power differentials enable harassment. Also minimizes the finding by calling a 17 percentage point difference "modest."
Recommendations for Addressing These Issues:
- Add a limitations section explicitly discussing how binary gender categorizations and racial/ethnic groupings reflect and potentially perpetuate systemic biases
- Incorporate intersectional analysis throughout, not just in one paragraph
- Explicitly name colonialism when discussing parachute science and Global North/South dynamics
- Acknowledge how the research team's own positionality may have shaped survey design and interpretation
- Discuss implicit bias training as insufficient without structural change
The dual-method approach (metadata analysis and survey) is valuable. However, in my review of procedures and statistical analyses, I've identified several significant methodological and statistical problems:
- Gender Inference Methodology (Critical Issue)
Lines 180-186: "Each co-author's gender was evaluated employing the genderize.io web server (Demografix ApS, 2022) and included in the analysis if the certainty estimate provided by the tool was larger than 75%."
Problems:
- The 75% threshold is arbitrary with no justification provided
- No discussion of error rates or validation
- Missing data from names below 75% certainty could introduce systematic bias
- Binary gender assumption excludes non-binary individuals entirely
- Cultural name variations may have different accuracy rates, potentially introducing geographic/ethnic bias
Lines 182-186 acknowledge limitations but don't explain how this affects the validity of conclusions.
- Survey Sampling and Representativeness Issues
Lines 132-143: "The survey was completed by 231 respondents, but the sample size used in the analysis here decreased to 198 after data quality assurance in which participants who did not identify with either the ice nucleation field and/or cloud physics field were removed."
Problems:
- 14% exclusion rate (33 participants) without detailed explanation of exclusion criteria
- No analysis of whether excluded participants differed systematically from included ones
- Potential selection bias not addressed
Lines 136-143: Claims representativeness based on conference attendance but:
- Conference attendees =/= field practitioners (excludes industry, non-conference-attending researchers)
- Three conferences from different years (2012, 2016, 2021) compared to 2021 survey—demographic shifts not accounted for
- "30-47% of the cloud physics membership" is a wide range suggesting uncertainty about denominator
- Acknowledged Overrepresentation Not Corrected
Lines 142-144: "Although the ICCP membership is representative of the international cloud physics community, there could be some unintended biases in the survey data coming from the overrepresentation of respondents born in India, Germany, and US (above 15% each) as well the overrepresentation of respondents younger than 40 years old (64%)."
Problems:
- Authors identify bias but don't apply any weighting or stratification to correct for it
- No sensitivity analyses to test how overrepresentation affects conclusions
- Geographic bias particularly problematic for study examining Global North/South dynamics
- Statistical Power Issues for Subgroup Analyses
Lines 284-287: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern."
Problems:
- Intersectional analyses attempted despite insufficient power
- No pre-registration or power calculations presented
- Multiple comparisons conducted throughout without correction for multiple testing
- Small cell sizes (e.g., 10 respondents with disabilities, line 269) make percentages misleading
- Inconsistent Statistical Testing
Problem: The study applies statistical tests inconsistently:
- Sometimes reports chi-square tests with effect sizes (e.g., lines 244-245, 277-278)
- Sometimes reports percentages without tests (e.g., lines 230-237)
- Sometimes explicitly states differences are "not statistically significant" (e.g., line 257-258)
- Sometimes presents differences without testing (e.g., lines 238-242)
No clear decision rule for when statistical testing is applied vs. descriptive statistics alone.
- Metadata Analysis Issues
Lines 157-171: "Although 8199 papers were pre-selected (from the documents available in Scopus on December 2021), those papers published in Astronomical and Astrophysical journals and other papers not related to cloud physics or written in a language different than English were removed leaving us with a total of 6987 papers"
Problems:
- a) Language bias explicitly introduced (lines 166-167): "We note that the number of papers (i.e., 97) removed because of language could slightly impact the outcome of the present analysis; however, this was necessary because the authors of the present study were unable to translate them"
- This systematically excludes non-English publications, likely disproportionately affecting Global South researchers
- Authors acknowledge this but proceed anyway
- b) Database limitations acknowledged but not addressed (lines 168-171): "the conclusions raised from the metadata analysis could be biased as Scopus, and most databases, does not cover all areas of interest leaving out important information from e.g., non-English-speaking countries or Global South countries"
- This is exactly the population the study aims to understand
- Critical methodological flaw for a DEI study
- Tropical Country Classification
Lines 187-192: "For a country to be considered tropical in this study, at least 50% of its inhabitants need to live in the tropics, i.e., between 23.5° N and 23.5° S."
Problems:
- Arbitrary 50% threshold not justified
- Geographic location used as proxy for "Global South" but these aren't equivalent
- India (2.86% of publications, Table 1) coded as tropical, but Taiwan (0.64%) also tropical—very different development contexts
- Conflates geography with economic development and colonial history
- Impact Factor as Outcome Variable
Lines 379-388: "The present data shows that in addition to the low number of studies published by tropical coauthors, on average, their indexed studies get published in lower-impact journals (impact factor IF=3.91, 95% confidence interval (CI): 3.72, 4.10 (n=287)) compared to their non-tropical peers (IF=4.58, 95% CI: 4.28, 4.88 (n=6638))."
Problems:
- Uses Impact Factor as measure of quality without acknowledging its limitations and biases
- Impact Factor favors established, English-language, Global North journals
- No discussion of how IF itself reflects and perpetuates the inequities being studied
- No statistical test comparing these confidence intervals
- Different sample sizes (287 vs 6638) not accounted for
- Career Retention Analysis Issues
Lines 444-460 (Figure 6 and discussion): The analysis of "years until last registered paper" as proxy for leaving the field.
Problems:
- Assumes last publication = leaving field (but could be career pauses, database lag, non-Scopus publications)
- Right-censoring issue: Recent entrants (2015-2020) have less opportunity to leave
- No survival analysis or time-to-event methods applied despite temporal data structure
- Compares retention across time periods with different follow-up durations
Line 453-455: "However, in recent years, after 1995, we see no difference in years between the genders in how long they remain in the cloud physics field judging by publication activity."
- This conclusion unsupported—visual inspection of Figure 6 without statistical test
- Doesn't account for different entry cohort sizes
- Missing Data and Non-Response
Lines 196-198: "Sixty seven percent of the participants were < 40 years old and 57.1% of the participants identified themselves as men, 41.4% as women, with 1.5% preferring not to respond"
Problems:
- 1.5% non-response on gender question acknowledged but not analyzed
- 5.6% non-response on sexual orientation (line 200-201)
- 6 respondents (3%) opted not to disclose ethnicity (line 207)
- No analysis of whether non-response patterns relate to other variables
- No sensitivity analyses for missing data
- Temporal Confounding
Lines 75-78: "the inclusion of researchers from new geographical locations was not significant from 1990 to 2000 (Wagner, 2005) but it was significant between 1990 and 2013 (Wagner et al., 2017)"
Problem: The study combines 50 years of data (1970-2020) but:
- Academic publishing, databases, and geopolitics changed dramatically over this period
- No time-stratified analyses
- Cold War effects on collaboration (1970-1991) not distinguished from post-Cold War
- Internet and digitization effects not accounted for
- Chi-Square Test Assumptions
Throughout the document, chi-square tests are used (e.g., lines 244-245, 277-278, 291-293), but:
- No verification that expected cell counts >5 (required assumption)
- With small subgroups (LGBTQIA+ = 8%, line 200), this assumption likely violated
- Should use Fisher's exact test for small samples
- Correlation vs. Causation
Lines 349-358: Discussion of women first authors having more women co-authors.
Problem: States "This may not be the result of them being more equitable but could rather mirror a preference for working with people similar to oneself" but:
- No longitudinal data to establish temporal precedence
- No controls for confounders (field subspecialty, institution, career stage)
- Alternative explanations not tested (e.g., mentorship networks, discrimination by male-led teams)
Recommended Corrections:
- Re-do gender analysis with validated methods and report uncertainty ranges
- Apply survey weights to correct for known overrepresentation
- Conduct power analysis and clearly state which analyses are exploratory
- Use consistent statistical testing with multiple comparison corrections
- Perform sensitivity analyses excluding English-only restriction
- Apply survival analysis methods to career retention data
- Stratify by time period in metadata analysis
- Use appropriate tests for small sample sizes
- Acknowledge Impact Factor limitations or use alternative metrics
- Pre-register any confirmatory hypotheses for future work
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC4 -
EC4: 'Thanks for RC4', David Crookall, 07 Jan 2026
Many thanks indeed for a tremendous review.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC4 - AC4: 'Reply on RC5', Zamin A. Kanji, 15 Apr 2026
-
RC5: 'Reply on RC4', Anonymous Referee #4, 27 Feb 2026
The following is a revision to my first review that was flagged as "AI-generated." My process for that version was to use the internal commenting tool while reading the manuscript, and then use AI to consolidate my comments into a single, readable file. When I was alerted that my review had been flagged, I ran that text through an AI detector, and it flagged quotes from the manuscript, equations, and associated comments on statistical assumptions and limitations, as well as basic textbook knowledge.
The following text is based on the original comments that I left in the downloaded manuscript, which I modified somewhat after re-reading for this review. My biggest concerns are the lack of acknowledgment of the limitations of statistical methods and the language around representation.
______________________________________________
PART 1: This review is both significant and timely. However, it demonstrates the tendency within academia to use archaic language and framing when analyzing geographic, racial, gender, and sexuality DEI issues globally. To address these, significant revisions are needed before publication.
- Include a Limitations section that explicitly addresses how categorizing gender as binary and grouping racial/ethnic reflects, and potentially perpetuates, systemic bias (this is why intersectionality is a critical parallel thread to your analysis).
- Include Position Statements from the authors. This clarifies how author backgrounds inform survey design and interpretation. Position statements are standard operating procedures in education and sociological research.
Lines 50-54: "The word 'minority' denotes negatively stigmatized, ostracized, oppressed, and outcast individuals or groups (Blanz et al., 1995). In the context of DEI, it refers to a numerically smaller population of a group within a larger population based on characteristics such as gender, race, religion, ethnicity, disability, sexual orientation, or other attributes. In the present study, 'minority' is used to describe an underrepresented or disadvantaged group, unless stated otherwise." Issue: I appreciate that you have acknowledged the negative connotation of the term “minority”. However, the term is appropriately used when discussing percent representation. When a group of people is not in the majority, they are often marginalized/oppressed due to systemic power structures.
Lines 66-69: "Systemic bias, under-representation of minorities, and a lack of recognition of researchers in and from developing countries (the so-called 'Global South') are reflected in the limited scientific collaborations between research groups from developed (the so-called 'Global North') and developing countries (Gewin, 2023)." Issue: Drop the use of "so-called".
Lines 150-153: "Therefore, in further analysis, we classify the Caucasian, White or European categories as 'White', while the other ethnicities are classified as 'non-white' with the exception of participants who preferred not to identify their ethnicity." Issue: This binary centers whiteness as the default and groups all other identities into a single "other" category, which obscures important differences in experiences among diverse racial/ethnic groups. This approach reinforces white supremacy frameworks.
Lines 182-186: "It is important to note that genderize.io assigns the gender based on the person's first name. Although this could be a good approximation, names cannot always be used to infer the gender (e.g., Sebo, 2021; Lockhart et al., 2023; Marty et al., 2023). Therefore, the results presented below must be taken with caution and considering the aforementioned bias.". Issue: This perpetuates binary gender assumptions and doesn't acknowledge how this methodology erases non-binary identities and reinforces cisgender normativity. The authors should explicitly state this limitation.
Lines 276-278: "A higher fraction of women, 44% compared to 27% of men, reported having experienced harassment, a difference that was statistically significant (χ²(1) = 5.70, p = 0.017, Cramér's V = 0.18), although the size of the gender difference is modest." Issue: In this reviewer’s opinion, a 17% difference is not “modest”. Sexism and misogyny, which is well-understood to be underreported in STEM fields, create hostile environments and power differentials and leads to work-place environments prone to harassment.
Lines 284-286: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern." Issue: Intersectionality is mentioned but not systematically applied throughout the analysis. The entire study design should have incorporated intersectional analysis from the beginning, not as an afterthought.
Lines 393-408: As a practice, parachute science is a continuation of colonialism. Be sure to include how biases (implicit and explicit) around the “ownership of knowledge and expertise” contribute to parachute science. Consider FAIR data practices and cite Ramírez-Castañeda, et al. (2022). PNAS 119 (34). At the very least, identify colonialism as the founding principle of parachute science.
PART 2. The dual-method approach (metadata analysis and survey) is valuable. However, in my review of procedures and statistical analyses, I've identified several significant methodological and statistical problems.
- Throughout the document, chi-square tests are used (e.g., lines 244-245, 277-278, 291-293), but there is a lack of verification regarding required assumptions (cell counts are greater than 5). Consider using Fisher’s exact test (Kim, H. Y. (2017). Statistical notes for clinical researchers: Chi-squared test and Fisher's exact test. Restorative dentistry & endodontics, 42(2), 152.) for small subgroups where cell counts are less than 5. For example, the subgroup for LGBTQIA+ = 8%, (line 200).
- Be sure to apply statistical tests consistently. Report all chi-square tests with effect sizes, percentages with (n=xx) values and describe differences as significant or not significant (support with appropriate test). Be clear about when you are using descriptive vs. applied statistics.
Lines 75-78: "the inclusion of researchers from new geographical locations was not significant from 1990 to 2000 (Wagner, 2005) but it was significant between 1990 and 2013 (Wagner et al., 2017).” Issues/Suggestions: This combines 50 years of data (1970-2020) without addressing dramatic changes in academic publishing, databases, and geopolitics over this period. Consider time-stratified analyses, the effects of geopolitics on collaboration, and the rise of the Internet and digitization effects.
Lines 136-143: Claims representativeness based on conference attendance but conference attendance is not a good measure for STEM practitioners, as it excludes industry, researchers who cannot attend due to cost, distance, timing, etc. Additionally, the 3 conferences included were from different years (2012, 2016, 2021) and then compared to 2021 survey and any demographic shifts (especially the aging out of white males). Shishkova, E., Kwiecien, N. W., Hebert, A. S., Westphall, M. S., Prenni, J. E., & Coon, J. J. (2017). Gender diversity in a STEM subfield–analysis of a large scientific society and its annual conferences. Journal of the American Society for Mass Spectrometry, 28(12), 2523-2531.
Lines 142-144: "Although the ICCP membership is representative of the international cloud physics community, there could be some unintended biases in the survey data coming from the overrepresentation of respondents born in India, Germany, and US (above 15% each) as well the overrepresentation of respondents younger than 40 years old (64%)." Issues/Suggestions: Consider applying some weighting or stratified sampling corrections, conducting a sensitivity analysis to test how this potential overrepresentation affects conclusions. The geographic bias is particularly problematic for a study examining Global North/South dynamics
Lines 157-171: "Although 8199 papers were pre-selected (from the documents available in Scopus on December 2021), those papers published in Astronomical and Astrophysical journals and other papers not related to cloud physics or written in a language different than English were removed leaving us with a total of 6987 papers.” Issue/Suggestion: (1) Why didn’t the authors utilize translation software or library services to translate papers? They state, “We note that the number of papers (i.e., 97) removed because of language could slightly impact the outcome of the present analysis; however, this was necessary because the authors of the present study were unable to translate them.” This systematically excludes non-English publications, likely disproportionately affecting Global South researchers. I suggest that the authors include some text about how they would recommend addressing this challenge in the future. (2) In (lines 168-171): "the conclusions raised from the metadata analysis could be biased as Scopus, and most databases, does not cover all areas of interest leaving out important information from e.g., non-English-speaking countries or Global South countries.” My concern here is that this is exactly the population the study aims to understand. This is a methodological flaw for a DEI study.
Lines 180-186: "Each co-author's gender was evaluated employing the genderize.io web server (Demografix ApS, 2022) and included in the analysis if the certainty estimate provided by the tool was larger than 75%." Issue/Suggestion: This reviewer suggests that you provide some justification for the 75% threshold and discuss the potential error rates for (1) using AI to determine gender from names (given that you are using the male/female binary system to sort names); (2) cultural variations in naming; and (3) explain how limitations may affect the validity of conclusions.
Lines 196-198: "Sixty seven percent of the participants were < 40 years old and 57.1% of the participants identified themselves as men, 41.4% as women, with 1.5% preferring not to respond." Issues/Suggestions: Provide some text on what this non-response could mean (see Porter, J. R., & Ecklund, E. H. (2012). Missing data in sociological research: An overview of recent trends and an illustration for controversial questions, active nonrespondents and targeted samples. The American Sociologist, 43(4), 448-468.). Is there any pattern of non-response to other variables?
Lines 284-287: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern." Issues/Suggestions: I agree that the “non-White women” phrase indicates an intersectionality issue. To support the statement about the sample size, include a parenthetical phrase (n=xx). I am concerned that no power calculations are presented, nor are corrections for multiple rounds of testing. Also, in line 269, where the authors refer to respondents who reported disabilities, there should be some statement about how disabilities are underreported (Schneiderwind, J., & Johnson, J. M., 2021). Broadening the equity lens for STEM teacher education: The invisibility of disability. AAAS bulletin.)
Lines 349-358: Discussion of women first authors having more women co-authors. Issues/Suggestions: The authors state, “This may not be the result of them being more equitable but could rather mirror a preference for working with people similar to oneself,” but there is no longitudinal data to establish temporal precedence, no controls for confounders (field subspecialty, institution, career stage), or alternative explanations not tested (e.g., mentorship networks, discrimination by male-led teams).
Lines 379-388: "The present data shows that in addition to the low number of studies published by tropical coauthors, on average, their indexed studies get published in lower-impact journals (impact factor IF=3.91, 95% confidence interval (CI): 3.72, 4.10 (n=287)) compared to their non-tropical peers (IF=4.58, 95% CI: 4.28, 4.88 (n=6638))." Issues/Suggestions: As stated above (Lines 157-171), there is potential for significant biases introduced to these data by the exclusion of papers from the Global South and non-English language journals. Furthermore, there are limitations and biases in the concept of Impact Factor making it problematic for use in research (Seglen, P. O. (1998). Citation rates and journal impact factors are not suitable for evaluation of research. Acta Orhopaedica Scandinavica, 69(3), 224-229.)
Lines 444-460 (Figure 6 and discussion): The analysis of "years until last registered paper" as proxy for leaving the field. Issues/Suggestions: This assumes the last publication is equivalent to leaving field, but it could also be career pauses/shifts (Researcher to Admin), database lag, non-Scopus publications. There is a bias toward younger researchers where those who have entered the field recently (2015-2020) have been publishing dissertations and for tenure review, meaning they are a maximum productivity and have had less opportunity to leave. Include a survival analysis or time-to-event test given the temporal data structure and compare retention across time periods with different follow-up durations.
Line 453-455: "However, in recent years, after 1995, we see no difference in years between the genders in how long they remain in the cloud physics field judging by publication activity." Issues/Suggestions: This conclusion is unsupported. A visual inspection of Figure 6 indicates no statistical test. Also, this does not account for different entry cohort sizes.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC5 -
EC5: 'Reply on RC5, note on AI', David Crookall, 02 Mar 2026
Thank you for your updated RC on this MS. I think that I should mention the following:
Editors should not upload manuscripts and/or peer review reports to AI tools because this violates the distribution licence agreement as the AI tool does not give the required attribution to the authors. Editors' decisions should be entirely based on their scientific expertise and not rely on AI-generated feedback. If editors suspect peer review reports being generated by AI, they should notify the Copernicus Editorial Support Office. The assistive use of AI (punctuation, grammar, and spelling checks) to improve the structure and readability of editor reports is allowed.
to which I have been alerted. It is difficult to know how far your particular use of AI was assistive. My first impression is that it was (in fact or in spirit), but only you can say for certain.
I think that the authors should, therefore, not be bound by this review. However, if the authors find that some suggestions are likely to help improve the MS, then they should use the suggestion and not feel bound to reference the suggestion. If the authors feel that a suggestion goes beyond the scope of the MS or that it is not fully relevant, then they should be at liberty to ignore the suggestion and not be obligated to justify ignoring it.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC5 - AC4: 'Reply on RC5', Zamin A. Kanji, 15 Apr 2026
- AC4: 'Reply on RC5', Zamin A. Kanji, 15 Apr 2026
Peer review completion
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-4499', Anonymous Referee #1, 12 Oct 2025
This manuscript presents an important and timely assessment of diversity, equity, and inclusion (DEI) in the cloud physics community, combining a metadata analysis of 6,987 peer-reviewed papers (1970–2020) with a community survey of 198 respondents. The topic is highly relevant, and the effort to integrate quantitative and qualitative perspectives is commendable. However, several conceptual and structural issues limit the manuscript’s clarity and impact in its current form. In particular, the terminology used to describe researchers from the Global South, the framing of gender analysis, and the interpretation of small-sample survey results require substantial revision. Addressing these concerns would considerably strengthen the analytical rigor, inclusivity, and global sensitivity of the paper.
Major Comments
- Terminology – “Tropical Scientists”: In my opinion, the use of the term “tropical scientists” is highly problematic. I find it both unnecessary and ill-considered that the authors have chosen to introduce a new label that lacks conceptual grounding and adds no clarity to the discussion. As someone working within this geographic zone, I find this terminology technically inaccurate and, frankly, counterproductive—it conflates latitude with social and institutional realities, and risks exoticizing researchers from the Global South. In my view, this framing reflects a misunderstanding of the issue it seeks to address. Established terms such as Global North/South, high-, middle-, or low-income countries (as defined by the UN or World Bank), or developed/developing countries—though imperfect—are far more appropriate, as they rest on recognized metrics and are widely understood in academic and policy contexts. I strongly recommend removing the term entirely and clarifying the distinction between tropical study sites and researchers from the Global South, which are not interchangeable categories.
- Use of “Minorities” vs “Marginalized”: The manuscript frequently uses the term “minorities.” Given the global context of the study, where numerical minorities may differ by region, the term “marginalized” would be more appropriate and inclusive. It emphasizes systemic disadvantage rather than numerical representation. This change would also align better with the ethical framing of equity in the paper’s conclusion.
- Gender Analysis and Non-Binary Representation: The gender analysis appears to be strictly binary (men/women). While this may have been constrained by data availability (e.g., use of genderize.io), the discussion section should at least acknowledge and briefly discuss the exclusion of trans and non-binary individuals. For instance, the authors could note the limitations of algorithmic gender inference, and mention issues such as misgendering, the use of dead names in publication records, and the broader underrepresentation of trans scientists in the geosciences. This would strengthen the inclusivity and awareness reflected in the DEI framing.
- Survey Interpretation and Sample Size Limitations: The metadata analysis is the stronger and more robust component of the paper. However, the survey-based results (N=198) present significant limitations when further divided by gender, country, and ethnicity. When broken down into subcategories, the sample size per group, in my opinion, becomes too small for reliable interpretation or for drawing broad conclusions about regional or demographic patterns. I suggest restructuring the Results section so that the metadata analysis comes first—forming the backbone of the findings—followed by the survey results as complementary qualitative insight. The discussion could then focus on how these two strands inform each other, rather than treating both as equally representative.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC1 -
EC1: 'Thanks for RC1', David Crookall, 12 Oct 2025
Thank you for your excellent review.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC1 - AC1: 'Reply on RCI', Zamin A. Kanji, 15 Apr 2026
- AC1: 'Reply on RCI', Zamin A. Kanji, 15 Apr 2026
-
RC2: 'Comment on egusphere-2025-4499', Anonymous Referee #2, 28 Oct 2025
In this study, the authors present the current state and over time development of Diversity, Equity, and Inclusion (DEI) in the cloud physics community by considering two different data sets: metadata of 6,987 published papers between 1970 and 2020, and the survey responses from 198 participants. This research is significant for the advancement of the geosciences as it helps identify areas that need improvement. However, there are some issues with the framing of certain aspects, including the terminology used. Therefore, my general comments align with the Rev1.
General comments:- Underrepresented instead of the word minority or marginalised: Minority means that the smaller number or part. As Rev1 mentioned, minorities differ from region to regionIt is important to note that labelling the global south as a minority is a Western-centric viewpoint. The groups considered in this context are not smaller in number; they are underrepresented, overlooked, or lacking visibility. Additionally, "marginalised" implies insignificance or a peripheral status, which also doesn’t adequately convey the issue at hand.
- Terminology tropical scientists/countries: The term "tropical" carries connotations that could present an "exotic" portrayal, which is not appropriate here. I completely agree with Rev1. Additionally, framing a comparison of “tropical countries” with so so-called global north is a false dichotomy. One is defined by economic and historical factors under neo-colonial structures, while the other is geographical.
- The gender analysis presented is binary and fails to include non-binary and transgender individuals. This omission is not addressed in the paper. Although the potential for misgendering due to the algorithm is acknowledged, it still contains uncategorized (gender-blind) names. Furthermore, such gender-inferencing algorithms often struggle with the accuracy of non-Western names.
I will try to elaborate on what I mentioned in the general comments by giving specific examples from the first couple of pages, which are valid throughout the entirety of the paper.
line 46: ".... minority representation..." This gap exists for any underrepresented group, not just minorities. You might not be a minority, but underrepresented, you might be a minority if you are looking from a Western lens. We are great in numbers, but our work or existence is not represented or acknowledged.
line 65: same issue here with the use of minorities
line 70: Women are underrepresented but larger in number, exactly like the scientists from the so-called global south. They are underrepresented, and there is a lack of acknowledgement.
line 81: In the cited paper, this is mentioned as "USA, UK, ... However, these are not necessarily the most collaborative countries, if measured by their proportion of collaborative output... The analysis provided evidence that countries rated high in terms of scientific development were more likely to collaborate." This does not necessarily mean low income with low scientific development.
line 86&96: Egypt and South Africa are not situated in the tropics but the subtropics. When comparing these regions with the global north, framing it as a dichotomy could be misleading, aka a false dichotomy. Additionally, the term "tropical" carries connotations that could present an "exotic" portrayal, which is not appropriate here.
line 189: Latitude range includes Australia, which is considered to be a "global north" country. If the intention is to analyse underrepresentation, this could lead to misrepresentation in the analysis. Although the countries are highlighted later in the table, it would be beneficial to clarify this point earlier on.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC2 -
EC2: 'Thank you for RC2', David Crookall, 11 Nov 2025
Thank you for your excellent review.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC2 - AC2: 'Reply on RC2', Zamin A. Kanji, 15 Apr 2026
- AC2: 'Reply on RC2', Zamin A. Kanji, 15 Apr 2026
-
RC3: 'Comment on egusphere-2025-4499', Anonymous Referee #3, 11 Nov 2025
This paper presents an analysis into the current climate and feelings of belonging within the specific geoscience community of cloud physics. It utilizes two different approaches to address this question: a meta-analysis of peer-reviewed publications and an online survey. The question addressed in the paper is a unique contribution to the field, however there are several significant issues that need to be revised and addressed before publication. In addition to the previous two referee comments above, I will add the following:
- There has been a surplus of more recent work that highlights the their point in lines 45-47 (Although various programs, policies, and inclusive environments are being developed to address issues such as gender inequality in academia, many gaps in other minority representation still remain (e.g., Hispanic, black, and indigenous communities; Bernard and Cooperdock, 2018; NCSES, 2019; Gewin, 48 2020; Odekunle, 2020). Please include an updated literature review from the last 5 years in this.
- I agree with the previous two referees that “marginalized” is a more appropriate term than minority. Based on the descriptors provided, another alternative would be to say “historically excluded” groups. This is a better descriptor that is commonly used and recognizes the fact that these groups have been intentionally excluded by the system, as opposed to just individual actions.
- In line 55 – able-bodied individuals are also significantly more represented in all fields than in the general population. There is literature on this as well and should be included in the literature review.
- Lines 58-59 (The lack of diversity and inclusion across academic communities manifests itself in systematic bias of under-represented minorities). I disagree with how this comment is written – the system was intentionally designed to prevent lack of diversity and inclusion from the very beginning. The system is the problem, which further exacerbates the lack of diversity and inclusion we see today.
- I think it would also be beneficial to expand on the background of cloud physics as a research field. It seems that your assumption is that everyone in this field trained in geosciences, or is there a combination of physics, chemistry, geography, etc? Each of these fields has presented its own variations in why they have issues related to DEI and should be included as a background variable when considering the climate of cloud physics.
- Your response rates are not clearly defined and need to be included. How many individuals were included in the ICCP listserv and the other announcement sent? How many participants actually registered for the 2021 meeting? Comparing this number to other membership numbers is not sufficient – especially when the data is anonymous and you cannot guarantee that individuals did not take the survey more than one time. I don’t agree with the statement that it represents 30-47% of the community as a whole – you should only report on response rates from who the survey was disseminated to.
- Lines 137-139: you say the number of respondents who were Indian was higher – but how do you know that if the data is anonymous? Please clarify.
- Lines 206-208: What are the actual stats presented by Dutt, 2020? The ratio of white/non-white seems higher than what the ratio is in the US.
- Figure 1: The color purple is used to signify two different things – please use different colors for the “total” versus “other” gender category.
- The paper published by Marin-Spiotta et al (2023) in Earth’s Future provides survey results on the workplace climate in the geosciences. You should include this in your discussion/analysis as a comparison since their survey results include over 2000 survey responses.
- Did you ask if survey respondents had experienced bullying or harassment within a specified time frame or ever? This should be further clarified and justified on either approach.
- Line 324 – the leaky pipeline analogy is problematic. Berhe et al (2024) describes this in their Nature Geoscience paper and recommends a better analogy.
- Line 335-336. I disagree that you have the numbers to state that you are representing the cloud physicists globally, based on my previous comment about response rates. This should be removed.
- Given the broad demographics of who this survey was disseminated to, I think you should expand on potential limitations of your findings on feeling included. Your question was simply “Do you feel included in the cloud physics community?” This is really vague and could be interpreted in so many different ways – I am less inclined to say that this is a valuable finding with this approach. What do you mean when you say included? Just in the workplace, in professional societies, when in school? This is not well defined.
- Why did you only ask about working with women in your survey? This is not inclusive to the entire gender spectrum.
- What does “sensitive to gender issues” mean? Again as above – this is so vague and could be taken to mean so many things. I don’t agree that this should be presented in this dataset, it’s not a well designed survey question.
- Question 40 in your survey – were you only talking about race/ethnicity here? Or was this applicable to any and all identities? Not clearly defined, which is problematic.
- You inherently introduced a lot of bias in your survey by using the term “tropical” to describe an entire subset of people. This is problematic.
- You say that the majority of people said no to question 67. Why did you not include an additional clarifier on that to explain their perspective? There could be many reasons why participants selected no, I don’t think it’s appropriate to speculate on that in your paper.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC3 -
EC3: 'Thank you for RC3', David Crookall, 11 Nov 2025
Thank you for your excellent review. Much appreciated.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC3 - AC3: 'Reply on RC3', Zamin A. Kanji, 15 Apr 2026
- AC3: 'Reply on RC3', Zamin A. Kanji, 15 Apr 2026
-
RC4: 'Comment on egusphere-2025-4499', Anonymous Referee #4, 07 Jan 2026
This review is both significant and timely. However, it demonstrates the tendency within academia to use archaic language and framing when analyzing geographic, racial, gender, and sexuality DEI issues globally. To address these, significant revisions are needed before publication. Specifically, the manuscript will be more impactful if the authors address the implicit, inherited, and adopted racism and sexism entrenched in the sciences.
Gender Methodology Issues
Lines 182-186: "It is important to note that genderize.io assigns the gender based on the person's first name. Although this could be a good approximation, names cannot always be used to infer the gender (e.g., Sebo, 2021; Lockhart et al., 2023; Marty et al., 2023). Therefore, the results presented below must be taken with caution and considering the aforementioned bias."
Issue: This perpetuates binary gender assumptions and doesn't acknowledge how this methodology erases non-binary identities and reinforces cisgender normativity. The authors should explicitly state this limitation.
Problematic Categorizations
Lines 150-153: "Therefore, in further analysis, we classify the Caucasian, White or European categories as 'White', while the other ethnicities are classified as 'non-white' with the exception of participants who preferred not to identify their ethnicity."
Issue: This binary centers whiteness as the default and groups all other identities into a single "other" category, which obscures important differences in experiences among diverse racial/ethnic groups. This approach reinforces white supremacy frameworks.
Minority Definition
Lines 50-54: "The word 'minority' denotes negatively stigmatized, ostracized, oppressed, and outcast individuals or groups (Blanz et al., 1995). In the context of DEI, it refers to a numerically smaller population of a group within a larger population based on characteristics such as gender, race, religion, ethnicity, disability, sexual orientation, or other attributes. In the present study, 'minority' is used to describe an underrepresented or disadvantaged group, unless stated otherwise."
Issue: While acknowledging negative connotations, the definition still focuses primarily on numerical representation rather than power dynamics and systemic oppression. Missing discussion of how majority groups maintain power regardless of numbers.
Global South Terminology
Lines 66-69: "Systemic bias, under-representation of minorities, and a lack of recognition of researchers in and from developing countries (the so-called 'Global South') are reflected in the limited scientific collaborations between research groups from developed (the so-called 'Global North') and developing countries (Gewin, 2023)."
Issue: Uses "so-called" but doesn't interrogate the problematic colonial legacy embedded in these geographic/economic binaries or how they perpetuate deficit narratives.
Parachute Science
Lines 393-408: The discussion of parachute science identifies the practice but doesn't explicitly name it as a continuation of colonial extraction practices or discuss how implicit biases about expertise and knowledge ownership enable these dynamics.
Specific concern at lines 393-397: "Our results suggest that the so-called helicopter or parachute science... is unfortunately common in the cloud physics community; however, a deeper analysis is required to confirm this."
Issue: Needs explicit statement about colonial legacies and power imbalances, not just a neutral "deeper analysis required."
Intersectionality
Lines 284-286: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern."
Issue: Intersectionality is mentioned but not systematically applied throughout the analysis. The entire study design should have incorporated intersectional analysis from the beginning, not as an afterthought.
Harassment Analysis
Lines 276-278: "A higher fraction of women, 44% compared to 27% of men reported having experienced harassment, a difference that was statistically significant (χ²(1) = 5.70, p = 0.017, Cramér's V = 0.18), although the size of the gender difference is modest."
Issue: Doesn't explicitly discuss how sexism and misogyny create hostile environments or how power differentials enable harassment. Also minimizes the finding by calling a 17 percentage point difference "modest."
Recommendations for Addressing These Issues:
- Add a limitations section explicitly discussing how binary gender categorizations and racial/ethnic groupings reflect and potentially perpetuate systemic biases
- Incorporate intersectional analysis throughout, not just in one paragraph
- Explicitly name colonialism when discussing parachute science and Global North/South dynamics
- Acknowledge how the research team's own positionality may have shaped survey design and interpretation
- Discuss implicit bias training as insufficient without structural change
The dual-method approach (metadata analysis and survey) is valuable. However, in my review of procedures and statistical analyses, I've identified several significant methodological and statistical problems:
- Gender Inference Methodology (Critical Issue)
Lines 180-186: "Each co-author's gender was evaluated employing the genderize.io web server (Demografix ApS, 2022) and included in the analysis if the certainty estimate provided by the tool was larger than 75%."
Problems:
- The 75% threshold is arbitrary with no justification provided
- No discussion of error rates or validation
- Missing data from names below 75% certainty could introduce systematic bias
- Binary gender assumption excludes non-binary individuals entirely
- Cultural name variations may have different accuracy rates, potentially introducing geographic/ethnic bias
Lines 182-186 acknowledge limitations but don't explain how this affects the validity of conclusions.
- Survey Sampling and Representativeness Issues
Lines 132-143: "The survey was completed by 231 respondents, but the sample size used in the analysis here decreased to 198 after data quality assurance in which participants who did not identify with either the ice nucleation field and/or cloud physics field were removed."
Problems:
- 14% exclusion rate (33 participants) without detailed explanation of exclusion criteria
- No analysis of whether excluded participants differed systematically from included ones
- Potential selection bias not addressed
Lines 136-143: Claims representativeness based on conference attendance but:
- Conference attendees =/= field practitioners (excludes industry, non-conference-attending researchers)
- Three conferences from different years (2012, 2016, 2021) compared to 2021 survey—demographic shifts not accounted for
- "30-47% of the cloud physics membership" is a wide range suggesting uncertainty about denominator
- Acknowledged Overrepresentation Not Corrected
Lines 142-144: "Although the ICCP membership is representative of the international cloud physics community, there could be some unintended biases in the survey data coming from the overrepresentation of respondents born in India, Germany, and US (above 15% each) as well the overrepresentation of respondents younger than 40 years old (64%)."
Problems:
- Authors identify bias but don't apply any weighting or stratification to correct for it
- No sensitivity analyses to test how overrepresentation affects conclusions
- Geographic bias particularly problematic for study examining Global North/South dynamics
- Statistical Power Issues for Subgroup Analyses
Lines 284-287: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern."
Problems:
- Intersectional analyses attempted despite insufficient power
- No pre-registration or power calculations presented
- Multiple comparisons conducted throughout without correction for multiple testing
- Small cell sizes (e.g., 10 respondents with disabilities, line 269) make percentages misleading
- Inconsistent Statistical Testing
Problem: The study applies statistical tests inconsistently:
- Sometimes reports chi-square tests with effect sizes (e.g., lines 244-245, 277-278)
- Sometimes reports percentages without tests (e.g., lines 230-237)
- Sometimes explicitly states differences are "not statistically significant" (e.g., line 257-258)
- Sometimes presents differences without testing (e.g., lines 238-242)
No clear decision rule for when statistical testing is applied vs. descriptive statistics alone.
- Metadata Analysis Issues
Lines 157-171: "Although 8199 papers were pre-selected (from the documents available in Scopus on December 2021), those papers published in Astronomical and Astrophysical journals and other papers not related to cloud physics or written in a language different than English were removed leaving us with a total of 6987 papers"
Problems:
- a) Language bias explicitly introduced (lines 166-167): "We note that the number of papers (i.e., 97) removed because of language could slightly impact the outcome of the present analysis; however, this was necessary because the authors of the present study were unable to translate them"
- This systematically excludes non-English publications, likely disproportionately affecting Global South researchers
- Authors acknowledge this but proceed anyway
- b) Database limitations acknowledged but not addressed (lines 168-171): "the conclusions raised from the metadata analysis could be biased as Scopus, and most databases, does not cover all areas of interest leaving out important information from e.g., non-English-speaking countries or Global South countries"
- This is exactly the population the study aims to understand
- Critical methodological flaw for a DEI study
- Tropical Country Classification
Lines 187-192: "For a country to be considered tropical in this study, at least 50% of its inhabitants need to live in the tropics, i.e., between 23.5° N and 23.5° S."
Problems:
- Arbitrary 50% threshold not justified
- Geographic location used as proxy for "Global South" but these aren't equivalent
- India (2.86% of publications, Table 1) coded as tropical, but Taiwan (0.64%) also tropical—very different development contexts
- Conflates geography with economic development and colonial history
- Impact Factor as Outcome Variable
Lines 379-388: "The present data shows that in addition to the low number of studies published by tropical coauthors, on average, their indexed studies get published in lower-impact journals (impact factor IF=3.91, 95% confidence interval (CI): 3.72, 4.10 (n=287)) compared to their non-tropical peers (IF=4.58, 95% CI: 4.28, 4.88 (n=6638))."
Problems:
- Uses Impact Factor as measure of quality without acknowledging its limitations and biases
- Impact Factor favors established, English-language, Global North journals
- No discussion of how IF itself reflects and perpetuates the inequities being studied
- No statistical test comparing these confidence intervals
- Different sample sizes (287 vs 6638) not accounted for
- Career Retention Analysis Issues
Lines 444-460 (Figure 6 and discussion): The analysis of "years until last registered paper" as proxy for leaving the field.
Problems:
- Assumes last publication = leaving field (but could be career pauses, database lag, non-Scopus publications)
- Right-censoring issue: Recent entrants (2015-2020) have less opportunity to leave
- No survival analysis or time-to-event methods applied despite temporal data structure
- Compares retention across time periods with different follow-up durations
Line 453-455: "However, in recent years, after 1995, we see no difference in years between the genders in how long they remain in the cloud physics field judging by publication activity."
- This conclusion unsupported—visual inspection of Figure 6 without statistical test
- Doesn't account for different entry cohort sizes
- Missing Data and Non-Response
Lines 196-198: "Sixty seven percent of the participants were < 40 years old and 57.1% of the participants identified themselves as men, 41.4% as women, with 1.5% preferring not to respond"
Problems:
- 1.5% non-response on gender question acknowledged but not analyzed
- 5.6% non-response on sexual orientation (line 200-201)
- 6 respondents (3%) opted not to disclose ethnicity (line 207)
- No analysis of whether non-response patterns relate to other variables
- No sensitivity analyses for missing data
- Temporal Confounding
Lines 75-78: "the inclusion of researchers from new geographical locations was not significant from 1990 to 2000 (Wagner, 2005) but it was significant between 1990 and 2013 (Wagner et al., 2017)"
Problem: The study combines 50 years of data (1970-2020) but:
- Academic publishing, databases, and geopolitics changed dramatically over this period
- No time-stratified analyses
- Cold War effects on collaboration (1970-1991) not distinguished from post-Cold War
- Internet and digitization effects not accounted for
- Chi-Square Test Assumptions
Throughout the document, chi-square tests are used (e.g., lines 244-245, 277-278, 291-293), but:
- No verification that expected cell counts >5 (required assumption)
- With small subgroups (LGBTQIA+ = 8%, line 200), this assumption likely violated
- Should use Fisher's exact test for small samples
- Correlation vs. Causation
Lines 349-358: Discussion of women first authors having more women co-authors.
Problem: States "This may not be the result of them being more equitable but could rather mirror a preference for working with people similar to oneself" but:
- No longitudinal data to establish temporal precedence
- No controls for confounders (field subspecialty, institution, career stage)
- Alternative explanations not tested (e.g., mentorship networks, discrimination by male-led teams)
Recommended Corrections:
- Re-do gender analysis with validated methods and report uncertainty ranges
- Apply survey weights to correct for known overrepresentation
- Conduct power analysis and clearly state which analyses are exploratory
- Use consistent statistical testing with multiple comparison corrections
- Perform sensitivity analyses excluding English-only restriction
- Apply survival analysis methods to career retention data
- Stratify by time period in metadata analysis
- Use appropriate tests for small sample sizes
- Acknowledge Impact Factor limitations or use alternative metrics
- Pre-register any confirmatory hypotheses for future work
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC4 -
EC4: 'Thanks for RC4', David Crookall, 07 Jan 2026
Many thanks indeed for a tremendous review.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC4 - AC4: 'Reply on RC5', Zamin A. Kanji, 15 Apr 2026
-
RC5: 'Reply on RC4', Anonymous Referee #4, 27 Feb 2026
The following is a revision to my first review that was flagged as "AI-generated." My process for that version was to use the internal commenting tool while reading the manuscript, and then use AI to consolidate my comments into a single, readable file. When I was alerted that my review had been flagged, I ran that text through an AI detector, and it flagged quotes from the manuscript, equations, and associated comments on statistical assumptions and limitations, as well as basic textbook knowledge.
The following text is based on the original comments that I left in the downloaded manuscript, which I modified somewhat after re-reading for this review. My biggest concerns are the lack of acknowledgment of the limitations of statistical methods and the language around representation.
______________________________________________
PART 1: This review is both significant and timely. However, it demonstrates the tendency within academia to use archaic language and framing when analyzing geographic, racial, gender, and sexuality DEI issues globally. To address these, significant revisions are needed before publication.
- Include a Limitations section that explicitly addresses how categorizing gender as binary and grouping racial/ethnic reflects, and potentially perpetuates, systemic bias (this is why intersectionality is a critical parallel thread to your analysis).
- Include Position Statements from the authors. This clarifies how author backgrounds inform survey design and interpretation. Position statements are standard operating procedures in education and sociological research.
Lines 50-54: "The word 'minority' denotes negatively stigmatized, ostracized, oppressed, and outcast individuals or groups (Blanz et al., 1995). In the context of DEI, it refers to a numerically smaller population of a group within a larger population based on characteristics such as gender, race, religion, ethnicity, disability, sexual orientation, or other attributes. In the present study, 'minority' is used to describe an underrepresented or disadvantaged group, unless stated otherwise." Issue: I appreciate that you have acknowledged the negative connotation of the term “minority”. However, the term is appropriately used when discussing percent representation. When a group of people is not in the majority, they are often marginalized/oppressed due to systemic power structures.
Lines 66-69: "Systemic bias, under-representation of minorities, and a lack of recognition of researchers in and from developing countries (the so-called 'Global South') are reflected in the limited scientific collaborations between research groups from developed (the so-called 'Global North') and developing countries (Gewin, 2023)." Issue: Drop the use of "so-called".
Lines 150-153: "Therefore, in further analysis, we classify the Caucasian, White or European categories as 'White', while the other ethnicities are classified as 'non-white' with the exception of participants who preferred not to identify their ethnicity." Issue: This binary centers whiteness as the default and groups all other identities into a single "other" category, which obscures important differences in experiences among diverse racial/ethnic groups. This approach reinforces white supremacy frameworks.
Lines 182-186: "It is important to note that genderize.io assigns the gender based on the person's first name. Although this could be a good approximation, names cannot always be used to infer the gender (e.g., Sebo, 2021; Lockhart et al., 2023; Marty et al., 2023). Therefore, the results presented below must be taken with caution and considering the aforementioned bias.". Issue: This perpetuates binary gender assumptions and doesn't acknowledge how this methodology erases non-binary identities and reinforces cisgender normativity. The authors should explicitly state this limitation.
Lines 276-278: "A higher fraction of women, 44% compared to 27% of men, reported having experienced harassment, a difference that was statistically significant (χ²(1) = 5.70, p = 0.017, Cramér's V = 0.18), although the size of the gender difference is modest." Issue: In this reviewer’s opinion, a 17% difference is not “modest”. Sexism and misogyny, which is well-understood to be underreported in STEM fields, create hostile environments and power differentials and leads to work-place environments prone to harassment.
Lines 284-286: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern." Issue: Intersectionality is mentioned but not systematically applied throughout the analysis. The entire study design should have incorporated intersectional analysis from the beginning, not as an afterthought.
Lines 393-408: As a practice, parachute science is a continuation of colonialism. Be sure to include how biases (implicit and explicit) around the “ownership of knowledge and expertise” contribute to parachute science. Consider FAIR data practices and cite Ramírez-Castañeda, et al. (2022). PNAS 119 (34). At the very least, identify colonialism as the founding principle of parachute science.
PART 2. The dual-method approach (metadata analysis and survey) is valuable. However, in my review of procedures and statistical analyses, I've identified several significant methodological and statistical problems.
- Throughout the document, chi-square tests are used (e.g., lines 244-245, 277-278, 291-293), but there is a lack of verification regarding required assumptions (cell counts are greater than 5). Consider using Fisher’s exact test (Kim, H. Y. (2017). Statistical notes for clinical researchers: Chi-squared test and Fisher's exact test. Restorative dentistry & endodontics, 42(2), 152.) for small subgroups where cell counts are less than 5. For example, the subgroup for LGBTQIA+ = 8%, (line 200).
- Be sure to apply statistical tests consistently. Report all chi-square tests with effect sizes, percentages with (n=xx) values and describe differences as significant or not significant (support with appropriate test). Be clear about when you are using descriptive vs. applied statistics.
Lines 75-78: "the inclusion of researchers from new geographical locations was not significant from 1990 to 2000 (Wagner, 2005) but it was significant between 1990 and 2013 (Wagner et al., 2017).” Issues/Suggestions: This combines 50 years of data (1970-2020) without addressing dramatic changes in academic publishing, databases, and geopolitics over this period. Consider time-stratified analyses, the effects of geopolitics on collaboration, and the rise of the Internet and digitization effects.
Lines 136-143: Claims representativeness based on conference attendance but conference attendance is not a good measure for STEM practitioners, as it excludes industry, researchers who cannot attend due to cost, distance, timing, etc. Additionally, the 3 conferences included were from different years (2012, 2016, 2021) and then compared to 2021 survey and any demographic shifts (especially the aging out of white males). Shishkova, E., Kwiecien, N. W., Hebert, A. S., Westphall, M. S., Prenni, J. E., & Coon, J. J. (2017). Gender diversity in a STEM subfield–analysis of a large scientific society and its annual conferences. Journal of the American Society for Mass Spectrometry, 28(12), 2523-2531.
Lines 142-144: "Although the ICCP membership is representative of the international cloud physics community, there could be some unintended biases in the survey data coming from the overrepresentation of respondents born in India, Germany, and US (above 15% each) as well the overrepresentation of respondents younger than 40 years old (64%)." Issues/Suggestions: Consider applying some weighting or stratified sampling corrections, conducting a sensitivity analysis to test how this potential overrepresentation affects conclusions. The geographic bias is particularly problematic for a study examining Global North/South dynamics
Lines 157-171: "Although 8199 papers were pre-selected (from the documents available in Scopus on December 2021), those papers published in Astronomical and Astrophysical journals and other papers not related to cloud physics or written in a language different than English were removed leaving us with a total of 6987 papers.” Issue/Suggestion: (1) Why didn’t the authors utilize translation software or library services to translate papers? They state, “We note that the number of papers (i.e., 97) removed because of language could slightly impact the outcome of the present analysis; however, this was necessary because the authors of the present study were unable to translate them.” This systematically excludes non-English publications, likely disproportionately affecting Global South researchers. I suggest that the authors include some text about how they would recommend addressing this challenge in the future. (2) In (lines 168-171): "the conclusions raised from the metadata analysis could be biased as Scopus, and most databases, does not cover all areas of interest leaving out important information from e.g., non-English-speaking countries or Global South countries.” My concern here is that this is exactly the population the study aims to understand. This is a methodological flaw for a DEI study.
Lines 180-186: "Each co-author's gender was evaluated employing the genderize.io web server (Demografix ApS, 2022) and included in the analysis if the certainty estimate provided by the tool was larger than 75%." Issue/Suggestion: This reviewer suggests that you provide some justification for the 75% threshold and discuss the potential error rates for (1) using AI to determine gender from names (given that you are using the male/female binary system to sort names); (2) cultural variations in naming; and (3) explain how limitations may affect the validity of conclusions.
Lines 196-198: "Sixty seven percent of the participants were < 40 years old and 57.1% of the participants identified themselves as men, 41.4% as women, with 1.5% preferring not to respond." Issues/Suggestions: Provide some text on what this non-response could mean (see Porter, J. R., & Ecklund, E. H. (2012). Missing data in sociological research: An overview of recent trends and an illustration for controversial questions, active nonrespondents and targeted samples. The American Sociologist, 43(4), 448-468.). Is there any pattern of non-response to other variables?
Lines 284-287: "A higher fraction of non-White women reported experiencing bullying and harassment, which aligns with the concept of intersectionality, where overlapping social identities and positions can increase the likelihood of discrimination (Crenshaw, 1991). However, the sample size prevents statistical verification of this pattern." Issues/Suggestions: I agree that the “non-White women” phrase indicates an intersectionality issue. To support the statement about the sample size, include a parenthetical phrase (n=xx). I am concerned that no power calculations are presented, nor are corrections for multiple rounds of testing. Also, in line 269, where the authors refer to respondents who reported disabilities, there should be some statement about how disabilities are underreported (Schneiderwind, J., & Johnson, J. M., 2021). Broadening the equity lens for STEM teacher education: The invisibility of disability. AAAS bulletin.)
Lines 349-358: Discussion of women first authors having more women co-authors. Issues/Suggestions: The authors state, “This may not be the result of them being more equitable but could rather mirror a preference for working with people similar to oneself,” but there is no longitudinal data to establish temporal precedence, no controls for confounders (field subspecialty, institution, career stage), or alternative explanations not tested (e.g., mentorship networks, discrimination by male-led teams).
Lines 379-388: "The present data shows that in addition to the low number of studies published by tropical coauthors, on average, their indexed studies get published in lower-impact journals (impact factor IF=3.91, 95% confidence interval (CI): 3.72, 4.10 (n=287)) compared to their non-tropical peers (IF=4.58, 95% CI: 4.28, 4.88 (n=6638))." Issues/Suggestions: As stated above (Lines 157-171), there is potential for significant biases introduced to these data by the exclusion of papers from the Global South and non-English language journals. Furthermore, there are limitations and biases in the concept of Impact Factor making it problematic for use in research (Seglen, P. O. (1998). Citation rates and journal impact factors are not suitable for evaluation of research. Acta Orhopaedica Scandinavica, 69(3), 224-229.)
Lines 444-460 (Figure 6 and discussion): The analysis of "years until last registered paper" as proxy for leaving the field. Issues/Suggestions: This assumes the last publication is equivalent to leaving field, but it could also be career pauses/shifts (Researcher to Admin), database lag, non-Scopus publications. There is a bias toward younger researchers where those who have entered the field recently (2015-2020) have been publishing dissertations and for tenure review, meaning they are a maximum productivity and have had less opportunity to leave. Include a survival analysis or time-to-event test given the temporal data structure and compare retention across time periods with different follow-up durations.
Line 453-455: "However, in recent years, after 1995, we see no difference in years between the genders in how long they remain in the cloud physics field judging by publication activity." Issues/Suggestions: This conclusion is unsupported. A visual inspection of Figure 6 indicates no statistical test. Also, this does not account for different entry cohort sizes.
Citation: https://doi.org/10.5194/egusphere-2025-4499-RC5 -
EC5: 'Reply on RC5, note on AI', David Crookall, 02 Mar 2026
Thank you for your updated RC on this MS. I think that I should mention the following:
Editors should not upload manuscripts and/or peer review reports to AI tools because this violates the distribution licence agreement as the AI tool does not give the required attribution to the authors. Editors' decisions should be entirely based on their scientific expertise and not rely on AI-generated feedback. If editors suspect peer review reports being generated by AI, they should notify the Copernicus Editorial Support Office. The assistive use of AI (punctuation, grammar, and spelling checks) to improve the structure and readability of editor reports is allowed.
to which I have been alerted. It is difficult to know how far your particular use of AI was assistive. My first impression is that it was (in fact or in spirit), but only you can say for certain.
I think that the authors should, therefore, not be bound by this review. However, if the authors find that some suggestions are likely to help improve the MS, then they should use the suggestion and not feel bound to reference the suggestion. If the authors feel that a suggestion goes beyond the scope of the MS or that it is not fully relevant, then they should be at liberty to ignore the suggestion and not be obligated to justify ignoring it.
Citation: https://doi.org/10.5194/egusphere-2025-4499-EC5 - AC4: 'Reply on RC5', Zamin A. Kanji, 15 Apr 2026
- AC4: 'Reply on RC5', Zamin A. Kanji, 15 Apr 2026
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 4,134 | 1,661 | 311 | 6,106 | 599 | 193 | 180 |
- HTML: 4,134
- PDF: 1,661
- XML: 311
- Total: 6,106
- Supplement: 599
- BibTeX: 193
- EndNote: 180
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Luis A. Ladino
Karin Ardon-Dryer
Diana L. Pereira
Ulrike Proske
Zyanya Ramirez-Diaz
Antonia Velicu
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1804 KB) - Metadata XML
-
Supplement
(475 KB) - BibTeX
- EndNote
- Final revised paper