Intercomparison of global ground-level ozone datasets for health-relevant metrics

Wang, Hantao; Miyazaki, Kazuyuki; Sun, Haitong Zhe; Qu, Zhen; Liu, Xiang; Inness, Antje; Schultz, Martin; Schröder, Sabine; Serre, Marc; West, J. Jason

doi:10.5194/egusphere-2024-3723

Preprints

https://doi.org/10.5194/egusphere-2024-3723

Preprints

03 Jan 2025

| 03 Jan 2025

Intercomparison of global ground-level ozone datasets for health-relevant metrics

Hantao Wang, Kazuyuki Miyazaki, Haitong Zhe Sun, Zhen Qu, Xiang Liu, Antje Inness, Martin Schultz, Sabine Schröder, Marc Serre, and J. Jason West

Abstract. Ground-level ozone is a significant air pollutant that detrimentally affects human health and agriculture. Global ground-level ozone concentrations have been estimated using chemical reanalyses, geostatistical methods, and machine learning, but these datasets have not been compared systematically. We compare six global ground-level ozone datasets (three chemical reanalyses, two machine learning, one geostatistics) against one another and relative to observations, for the ozone season daily maximum 8-hour average mixing ratio, for 2006 to 2016. Results show significant differences among datasets in global average ozone, as large as 5–10 ppb, multi-year trends, and regional distributions. For example, in Europe, the three chemical reanalyses show an increasing trend while the other datasets show no increase. Among the six datasets, the population exposed to over 50 ppb varies from 60.8 % to 99 % in East Asia, 17 % to 88 % in North America, and 9 % to 77 % in Europe (2006–2016 average). These differences are large enough to impact health and other applications. Comparing with Tropospheric Ozone Assessment Report (TOAR) II ground-level observations, most datasets overestimate ozone, particularly at lower observed concentrations. In 2016, across all stations, R² ranges among the six datasets from 0.35 to 0.63, and RMSE from 5.28 to 13.49 ppb. Performance further declines when considering only stations with observations above 50 ppb. Although some datasets share some of the same input data, we found important differences among these datasets, likely from variations in approaches, resolution, and other input data, highlighting the importance of continued research on global ozone distributions.

Received: 27 Nov 2024 – Discussion started: 03 Jan 2025

Competing interests: Some authors are members of the editorial board for Atmospheric Chemistry & Physics. The authors declare that they have no other conflicts of interest.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2205 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (2205 KB)

Supplement (9376 KB)

Download & links

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

Journal article(s) based on this preprint

18 Nov 2025

Intercomparison of global ground-level ozone datasets for health-relevant metrics

Hantao Wang, Kazuyuki Miyazaki, Haitong Zhe Sun, Zhen Qu, Xiang Liu, Antje Inness, Martin Schultz, Sabine Schröder, Marc Serre, and J. Jason West

Atmos. Chem. Phys., 25, 15969–15990, https://doi.org/10.5194/acp-25-15969-2025,https://doi.org/10.5194/acp-25-15969-2025, 2025

Short summary

Hantao Wang, Kazuyuki Miyazaki, Haitong Zhe Sun, Zhen Qu, Xiang Liu, Antje Inness, Martin Schultz, Sabine Schröder, Marc Serre, and J. Jason West

Interactive discussion

Status: closed

RC1: 'Comment on egusphere-2024-3723', Anonymous Referee #2, 23 Jan 2025

Detailed comments can be found in the supplement.

Citation: https://doi.org/10.5194/egusphere-2024-3723-RC1
CC1: 'Comment on egusphere-2024-3723', Owen Cooper, 13 Feb 2025

This comment can be found in the attached pdf.

Citation: https://doi.org/10.5194/egusphere-2024-3723-CC1
RC2: 'Comment on egusphere-2024-3723', Anonymous Referee #1, 14 Feb 2025

L23: Given the large bias errors in these data sets, comparing the population exposed to a threshold value, like 50 ppb is meaningless. Do these differences impact "health" as stated or one's analysis of health effects. This is a bit sloppy writing.
L25: very good point, but you should compare the gridded data you have with Schnell's gridding of the EU and NAm. Comparing points to grid-cell averages that you have from the global data sets is a serious science problem -? not the way you treat it here.
How can you get an R2 for surface sites vs grid-cell means?? This is not sensible.
L27: You just said your data is worst at overestimating O3 at low abundances, but here you say it is worse for >50 ppb??
L29: "highlighting the importance of continued research on global ozone distributions"
L38: Oh really. The number of regions could be much much greater if you picked smaller regions. The key issue is the area fraction NOT the number of regions.
L42: 30 ppb is basically the minimum background level – this is not a useful statement and it implies that pollution is the cause here.
L44: The quality of writing (logic, not English) is poor: You just quoted all these results that rely on estimates of surface ozone and then you say you lack knowledge of surface ozone.
L70: Great. This is the most important statement. Could be up front
L71: "Potential" – there are most assuredly inconsistencies.
L75: the biases and errors certainly come from the process. I hope you are not using 'data' to describe the assimilation modeling here.

L85-95: This exposes the fundamental flaws in the focus of this paper. The use of OSDMA8 totally washes out the key fundamental information about ozone that can be tested with the real surface direct observations. The 24-hour diel cycle is a must that needs to be simulated in any modeled ozone product (all of your six sets are modeled products). Likewise the variability of ozone (including MDA8) is critical in evaluating health/agric. impacts and there needs to be a test of your six 'sets' as to their ability of match extremes.
L195ff: ibid. This is a mistake to smooth out the fundamental ozone cycles (diel and synoptic).
L220ff: "We adopted a point-to-grid evaluation approach, where the data from each TOAR-II observation site was matched with a corresponding grid cell in each dataset. For grid cells with a TOAR-II observation but no valid estimate in a dataset (NA value), we used the nearest valid estimate instead." This seems to ignore the previous TOAR-related work by Schnell where for the high-density of surface sites in EU and N.Am., a 1ºx1º grid-cell averaged, hourly surface ozone product was created.

This data set was used to assess extremes and to test the CMIP model's accuracy in seasonal and diel cycle of ozone. The cell average is the only way to do a fair comparison with the surface sites because of their irregular – sometimes oversampling and sometimes under sampling – many regions. Comparing surface sites with model cells is dangerous, especially since in this paper their appears to be a lack of understanding of the problems with this approach. The Schnell data are the obvious choice to validate your six model-data sets, even if it is only for EU and NAm:
doi:10.5194/acp-14-7721-2014
doi:10.5194/acp-15-10581-2015
doi:10.1002/2016GL068060
doi:10.1002/2017GL073044
doi: 10.1073/pnas.1614453114.
Then you can go after the rest of the world (which is very important).
This paper is based on comparing 6 different modeled surface ozone dataset with one another and with the TOAR set of surface sites (Table 3). The comparison of individual sites with grid-cell averages fails to recognize the difficulty of the task and ignores the extensive efforts to develop unbiased grid-cell means from high-density observations. The authors further corrupt the data set by averaging and smoothing to destroy the fundamental information on ozone variability that is critical for testing the modeled ozone products. The use of these 6 sets, varying in resolution from 0.1 to 2.5 degrees, to map population exposure is premature.
I can not recommend publication of this work as is.

Citation: https://doi.org/10.5194/egusphere-2024-3723-RC2
AC1: 'Comment on egusphere-2024-3723 - Response to Referee 1', Jason West, 15 Apr 2025

please see attached file

Citation: https://doi.org/10.5194/egusphere-2024-3723-AC1
AC2: 'Comment on egusphere-2024-3723 - Response to Referee 2', Jason West, 15 Apr 2025

please see attachment

Citation: https://doi.org/10.5194/egusphere-2024-3723-AC2
AC3: 'Comment on egusphere-2024-3723 - Response to Community Comment', Jason West, 15 Apr 2025

please see attachment

Citation: https://doi.org/10.5194/egusphere-2024-3723-AC3

Interactive discussion

Status: closed

RC1: 'Comment on egusphere-2024-3723', Anonymous Referee #2, 23 Jan 2025

Detailed comments can be found in the supplement.

Citation: https://doi.org/10.5194/egusphere-2024-3723-RC1
CC1: 'Comment on egusphere-2024-3723', Owen Cooper, 13 Feb 2025

This comment can be found in the attached pdf.

Citation: https://doi.org/10.5194/egusphere-2024-3723-CC1
RC2: 'Comment on egusphere-2024-3723', Anonymous Referee #1, 14 Feb 2025

L23: Given the large bias errors in these data sets, comparing the population exposed to a threshold value, like 50 ppb is meaningless. Do these differences impact "health" as stated or one's analysis of health effects. This is a bit sloppy writing.
L25: very good point, but you should compare the gridded data you have with Schnell's gridding of the EU and NAm. Comparing points to grid-cell averages that you have from the global data sets is a serious science problem -? not the way you treat it here.
How can you get an R2 for surface sites vs grid-cell means?? This is not sensible.
L27: You just said your data is worst at overestimating O3 at low abundances, but here you say it is worse for >50 ppb??
L29: "highlighting the importance of continued research on global ozone distributions"
L38: Oh really. The number of regions could be much much greater if you picked smaller regions. The key issue is the area fraction NOT the number of regions.
L42: 30 ppb is basically the minimum background level – this is not a useful statement and it implies that pollution is the cause here.
L44: The quality of writing (logic, not English) is poor: You just quoted all these results that rely on estimates of surface ozone and then you say you lack knowledge of surface ozone.
L70: Great. This is the most important statement. Could be up front
L71: "Potential" – there are most assuredly inconsistencies.
L75: the biases and errors certainly come from the process. I hope you are not using 'data' to describe the assimilation modeling here.

L85-95: This exposes the fundamental flaws in the focus of this paper. The use of OSDMA8 totally washes out the key fundamental information about ozone that can be tested with the real surface direct observations. The 24-hour diel cycle is a must that needs to be simulated in any modeled ozone product (all of your six sets are modeled products). Likewise the variability of ozone (including MDA8) is critical in evaluating health/agric. impacts and there needs to be a test of your six 'sets' as to their ability of match extremes.
L195ff: ibid. This is a mistake to smooth out the fundamental ozone cycles (diel and synoptic).
L220ff: "We adopted a point-to-grid evaluation approach, where the data from each TOAR-II observation site was matched with a corresponding grid cell in each dataset. For grid cells with a TOAR-II observation but no valid estimate in a dataset (NA value), we used the nearest valid estimate instead." This seems to ignore the previous TOAR-related work by Schnell where for the high-density of surface sites in EU and N.Am., a 1ºx1º grid-cell averaged, hourly surface ozone product was created.

This data set was used to assess extremes and to test the CMIP model's accuracy in seasonal and diel cycle of ozone. The cell average is the only way to do a fair comparison with the surface sites because of their irregular – sometimes oversampling and sometimes under sampling – many regions. Comparing surface sites with model cells is dangerous, especially since in this paper their appears to be a lack of understanding of the problems with this approach. The Schnell data are the obvious choice to validate your six model-data sets, even if it is only for EU and NAm:
doi:10.5194/acp-14-7721-2014
doi:10.5194/acp-15-10581-2015
doi:10.1002/2016GL068060
doi:10.1002/2017GL073044
doi: 10.1073/pnas.1614453114.
Then you can go after the rest of the world (which is very important).
This paper is based on comparing 6 different modeled surface ozone dataset with one another and with the TOAR set of surface sites (Table 3). The comparison of individual sites with grid-cell averages fails to recognize the difficulty of the task and ignores the extensive efforts to develop unbiased grid-cell means from high-density observations. The authors further corrupt the data set by averaging and smoothing to destroy the fundamental information on ozone variability that is critical for testing the modeled ozone products. The use of these 6 sets, varying in resolution from 0.1 to 2.5 degrees, to map population exposure is premature.
I can not recommend publication of this work as is.

Citation: https://doi.org/10.5194/egusphere-2024-3723-RC2
AC1: 'Comment on egusphere-2024-3723 - Response to Referee 1', Jason West, 15 Apr 2025

please see attached file

Citation: https://doi.org/10.5194/egusphere-2024-3723-AC1
AC2: 'Comment on egusphere-2024-3723 - Response to Referee 2', Jason West, 15 Apr 2025

please see attachment

Citation: https://doi.org/10.5194/egusphere-2024-3723-AC2
AC3: 'Comment on egusphere-2024-3723 - Response to Community Comment', Jason West, 15 Apr 2025

please see attachment

Citation: https://doi.org/10.5194/egusphere-2024-3723-AC3

Peer review completion

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

AR by Jason West on behalf of the Authors (15 Apr 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (27 May 2025) by Chul Han Song

RR by Anonymous Referee #1 (27 May 2025)

Suggestions for revision or reasons for rejection

Re-review of EGUSPHERE-2024-3723, Wang, West, and others.
Re-reviewing this manuscript with the authors' responses to the first review does not provide any more encouragement that the paper adds significant new knowledge to the atmospheric chemistry domain. In terms of regulatory issues, it may be fine, but the extreme limitation of their diagnostics makes the paper useless for diagnosing the problems with modeling surface ozone. Further, calculating human impacts from 6-month OSDMA8 ozone when you show the incredible biases of the modeled ozone seems like going too far, e.g., the numbers in Table 3 have no uncertainties related to the obvious model bias, so how can they be published?
"The ozone seasonal daily maximum 8-hour average mixing ratio " = six-month running monthly mean!). This diagnostic totally obscures all issues of extremes and even hides the seasonal cycle, removing most of the inter-month variability. The use of a six-month DMA8 because some of the models only did that is still a poor excuse. Not looking at the diel cycle and extreme days means you cannot understand why the models fail to match DMA8. The CAMS and TCR-2 results seem sensible (2-hr and 3-hr ozone) and at a more reasonably resolution to make a useful comparison.
Looking at these two responses shows that there is little understanding of the problem:
"However, the goal of our work is to assess the accuracy of gridding products in estimating measured ground concentration point values.
"That is why we have adopted this grid-to-point evaluation approach. Any gridded product, such as that of Schnell et al, uses an interpolation of a point value that introduces its own uncertainties and biases. To avoid these additional uncertainties, we directly compared observed point-level ozone values to the nearest available grid estimates.
You did not "avoid" the uncertainties with the problem that Schnell dealt with, you simply ignored them. There is nothing here that attempts to deal with the problems of creating a high-resolution map based on the site measurements and then integrating it to get the cell average. This problem is what Schnell's first paper spent most of its effort on. I may have missed it but I find no serious effort to optimize the point-area comparison.
“Previous research has adopted a 1º×1º grid-cell-averaged hourly ozone data from TOAR observations to evaluate global chemistry model performance over North America and Europe, which is suitable for analyzing extremes and validating seasonal and diel ozone cycles (Schnell and Prather, 2017; Schnell et al., 2015).”
“We adopted a grid-to-point evaluation approach, where the data from each TOAR-II observation site was matched with a corresponding grid cell in each dataset. For grid cells with a TOAR-II observation but no valid estimate in a dataset (NA value), we used the nearest valid estimate instead.”
Minor point – Schnell gathered all the AQ station data in N.Am. and EU, not specifically the TOAR data. Really, a "grid-to-point evaluation" is simply saying that every point in a cell should have the same value as the mean of that cell. This is quite apparently false when you have several nearby sites. The algorithm here (last sentence) does not really address how "far" a single station can reach? or why? "nearest" valid estimate is not provide a scientific comfort level.
"our focus is explicitly on assessing whether global ozone mapping products can reasonably estimate point-level concentrations at locations lacking monitoring stations.
I cannot see how you can begin to do that without models that resolve station-to-station differences (1 km) and so you really cannot say this.
2
The authors have done nothing to address the primary problems with this analysis: the use of 6-month averages of MDA8 to compare with models; and the fundamental methodology of how to compare a mean grid-cell value with point measurements, especially when there are several in a cell. The latter is clearly major unresolved problem (here at least). What happens if you have three different sites within a cell, each with three different values – how does one compare and derive R2? I realize that the authors may not be able to do this given the material they have, and thus this paper belongs in an air quality management journal that deals with meeting regulations, not in a science journal like ACP.

Hide

RR by Anonymous Referee #2 (02 Jun 2025)

Suggestions for revision or reasons for rejection

Review on the manuscript of egusphere-2024-3723 titled “Inter-comparison of global ground-level ozone datasets for health relevant metrics” written by Hantao Wang et al.

The authors have addressed some of the reviewers’ suggestions, and there are some improvements in the revised manuscript. However, several important issues remain and require further revision. Failure to address the comments from the first round seriously compromises the value of this study. Unless these critical points are properly dealt with, the contribution of the work remains questionable.
1. Impact of data uncertainty on related analysis and reorganizing structure (i.e. third comment in the 1st round): Regarding this issue, I do not agree with your response. While readers can draw their own conclusions, it is the authors’ fundamental responsibility to provide clear and specific interpretations based on the findings. Leaving key aspects of the analysis entirely up to the reader undermines the completeness and clarity of the work. This approach also deviates from the stated aims and title of the study, which emphasize evaluating uncertainty and accuracy through comparisons between observation and datasets. One of the core values of this research should be to account for these uncertainties and to draw meaningful conclusion – particularly about implications for public health and agriculture. Moreover, Sections 4 and 5 lack a logical flow, which further weakens the clarity and impact of the manuscript. Without a through and well-reasoned interpretation of the results (i.e., the uncertainty and its impact on trend analysis and ozone exposed population assessment), the manuscript does not fulfill its stated objectives and, in its current form, is not suitable for acceptance in ACP.
2. Again, regarding the fourth comment in the first round, I don’t think the authors thoroughly evaluate the performance of the datasets. The authors missed their scientific discussion and the implications.
3. The numbering of the supplementary table (Table S) is not presented in sequential order, which can cause confusion. Ensuring correct and consistent numbering of all tables and figures is a basic requirement that should be addressed prior to submission. Please revise the manuscript carefully to ensure that all Tables (S) and Figures (S) are accurate and properly ordered.
4. Units are missing in Table 2. For clarity and scientific accuracy, please add the relevant units to all applicable columns or data entries.
5. Even in the revised version, some statements still lack objective explanations based on consistent criteria. For example, the manuscript highlights a downward trend only for NJML (in Line 264); however, BME also shows a concurrent downward trend in both area- and population-weighted metrics. If the authors interpret the populated-weighted trend of BME (i.e., -0.04) as insufficient to indicate a clear trend, then by the same standard, none of the remaining five datasets show an increasing trend either. In this context, the statement in line 415 needs to be revised. More importantly, the manuscript lacks a clear definition of what constitutes a “clear” or “unclear” trend. The criteria used to make such classifications should be explicitly stated—whether based on slope magnitude, statistical significance, or another method. Without a consistent and objective basis for trend interpretation, the analysis risks appearing arbitrary and subjective. Please review the manuscript carefully to ensure that all claims are supported by clear, objective, and consistently applied analytical reasoning. In addition, the authors are advised to carefully review all statements throughout the manuscript.
6. (Sect 3.3, Line 224-225) Regarding 6th comment in the 1st round, the concern is not about missing data. Of course, reanalysis datasets like GEOS-CHEM, CAMS, and TCR-2 have complete spatial coverage. The issue is the lack of spatial representativeness when comparing coarse-resolution grid cells with single (or multiple) observation sites. A single or (multiple) monitoring station(s) may not adequately represent the entire of a coarse grid cell. The authors need to justify how the address this potential mismatch in spatial representativeness.
7. The amount of supplementary material is excessive and may overwhelm the reader. The authors are encouraged to condense and streamline the supplementary information by summarizing key findings and removing redundancies. Presenting only the most relevant and necessary content will improve the clarity and accessibility of the supporting materials. In particular, Figure S4 does not appear to add meaningful value.

Hide

ED: Reject (05 Jun 2025) by Chul Han Song

ED: Reconsider after major revisions (14 Jul 2025) by Bryan N. Duncan

AR by Jason West on behalf of the Authors (09 Sep 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (09 Sep 2025) by Bryan N. Duncan

RR by Anonymous Referee #1 (04 Oct 2025)

ED: Publish as is (04 Oct 2025) by Bryan N. Duncan

AR by Jason West on behalf of the Authors (13 Oct 2025) Manuscript

Journal article(s) based on this preprint

18 Nov 2025

Intercomparison of global ground-level ozone datasets for health-relevant metrics

Hantao Wang, Kazuyuki Miyazaki, Haitong Zhe Sun, Zhen Qu, Xiang Liu, Antje Inness, Martin Schultz, Sabine Schröder, Marc Serre, and J. Jason West

Atmos. Chem. Phys., 25, 15969–15990, https://doi.org/10.5194/acp-25-15969-2025,https://doi.org/10.5194/acp-25-15969-2025, 2025

Short summary

Hantao Wang, Kazuyuki Miyazaki, Haitong Zhe Sun, Zhen Qu, Xiang Liu, Antje Inness, Martin Schultz, Sabine Schröder, Marc Serre, and J. Jason West

Supplement

https://doi.org/10.5194/egusphere-2024-3723-supplement

Hantao Wang, Kazuyuki Miyazaki, Haitong Zhe Sun, Zhen Qu, Xiang Liu, Antje Inness, Martin Schultz, Sabine Schröder, Marc Serre, and J. Jason West

Viewed

Total article views: 1,034 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
795	206	33	1,034	74	29	36

HTML: 795
PDF: 206
XML: 33
Total: 1,034
Supplement: 74
BibTeX: 29
EndNote: 36

Views and downloads (calculated since 03 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	191	38	4	233
Feb 2025	70	24	4	98
Mar 2025	30	7	0	37
Apr 2025	52	27	4	83
May 2025	24	10	4	38
Jun 2025	30	13	3	46
Jul 2025	33	11	3	47
Aug 2025	115	18	3	136
Sep 2025	189	15	4	208
Oct 2025	40	26	3	69
Nov 2025	21	17	1	39

Cumulative views and downloads (calculated since 03 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	191	38	4	233
Feb 2025	70	24	4	98
Mar 2025	30	7	0	37
Apr 2025	52	27	4	83
May 2025	24	10	4	38
Jun 2025	30	13	3	46
Jul 2025	33	11	3	47
Aug 2025	115	18	3	136
Sep 2025	189	15	4	208
Oct 2025	40	26	3	69
Nov 2025	21	17	1	39

Viewed (geographical distribution)

Total article views: 995 (including HTML, PDF, and XML) Thereof 995 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 18 Nov 2025

Download

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

Preprint (2205 KB)
Metadata XML

Short summary

We compare six datasets of global ground-level ozone, developed using geostatistical, machine learning, or reanalysis methods. The datasets show important differences from one another in ozone magnitude, greater than 5 ppb, and trends, globally and regionally. Compared with measurements, performance varies among datasets, and most overestimate ozone, particularly at lower concentrations. These differences among datasets highlight uncertainties for applications to health and other impacts.


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