the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Sep 2026
Recent changes in the O3 and CO trends in the troposphere as observed by IAGOS
Abstract. IAGOS has been measuring Ozone (O3) and Carbon Monoxide (CO) since 1994 and 2002 respectively. This dataset allows us to calculate O3 and CO trends in the upper troposphere across nine regions of the Northern Hemisphere. Trends are computed over multiple time periods to highlight recent changes.
O3 growth in the upper troposphere (UT) of the Northern Hemisphere was more pronounced between 1994 and 2013 (e.g 0.34 ± 0.09 ppb/y in East US and 0.32 ± 0.07 ppb/y in Europe). These trends are significantly higher than those calculated from 1994 to 2023 (0.17 ± 0.07 ppb/y in East US and 0.16 ± 0.05 ppb/y in Europe). This difference in trend of O3 appears to be associated with high positive anomalies of O3 in the early 2010s, as well as significantly lower O3 levels after 2020, at least partially linked to the COVID-19 crisis.
For CO, we observe a gradual reduction in its rate of decline over the years, until a point where trends are no longer detected or are associated with low positive values since the 2010s at mid-latitudes in the Northern Hemisphere.
Further studies will be necessary to constrain the causes of recent O3 and CO variations. Continued IAGOS monitoring over the coming years will be essential, to further update this analysis and to determine whether the recent change in the observed trends are temporary or represent a sustained change in the upper tropospheric air composition.
- Preprint
(2181 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 22 Oct 2026)
- RC1: 'Comment on egusphere-2026-4770', Anonymous Referee #1, 15 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-4770', Anonymous Referee #2, 19 Oct 2026
reply
This paper provides a partial update to the analysis of Cohen et al. (2018), but it lacks the thoroughness of Cohen et al. (2018), and has many deficiencies that require a major revision before the paper is acceptable for publication in ACP. My detailed comments are listed below.
As a first step, the authors need to identify their target audience and ensure that the paper meets the needs of the audience. As the analysis focuses on ozone and CO in the upper troposphere, the findings are not relevant for studies of surface pollution. Instead, the findings would be of interest to modelers who are trying to understand global atmospheric chemistry. The modelling community is very aware of the limitations of most in situ records of ozone and CO, due to low sampling rates and limited spatial representativeness (Logan, 1999; Saunois et al., 2012; Lin et al., 2015; Barnes et al., 2016; Fiore et al., 2022). Therefore, the first question modelers will have when reading this paper is whether or not the sampling frequency is sufficient for trend detection. While IAGOS is well-known for having high sampling frequencies in many regions of the world, the sampling rate seems to have decreased in recent years, bringing the robustness of the data set into question. I elaborate on this point below.
Second, the authors need to ensure that they are up-to-date in terms of best statistical practices as they relate to trend detection. While the authors have followed some of the advice from the second phase of the Tropospheric Ozone Assessment Report (TOAR-II), the authors ignored other pieces of good advice, which makes the analytical methods out-of-date in comparison to recent TOAR-II studies that have also used IAGOS data. Positive aspects of this trend analysis include the application of quantile regression, the use of signal-to-noise ratio, and avoidance of the expression “statistically significant”, which has been shown to be very problematic (Wasserstein et al., 2019). Deficiencies with the trend analysis are as follows:
1) Following the advice from TOAR-II and Wasserstein et al. (2019), all trends need to be reported with the 95% confidence interval and the p-value and/or signal-to-noise ratio. While Figures 2, 3, 6, 7 and 8 roughly indicate the magnitude of the trend and the 95% confidence interval, the exact values are difficult to determine, and the signal-to-noise ratios are not shown (it is not enough to just indicate one signal-to-noise range with hatching). To provide the reader with this necessary information, please report the trends in tables, following the recent TOAR-II papers by Gaudel et al., 2024 (see their table 1), and Van Malderen et al., 2025 (see their table 6). Please note that these TOAR-II studies use IAGOS data and have IAGOS co-authors who are familiar with these methods.
2) The authors have not provided a convincing analysis that demonstrates that their minimum sample size of 4 sampling days per month per region is adequate to capture the true monthly mean. A recent TOAR-II paper (Chang et al. 2024) provides a very clear demonstration of the impact of low sampling frequency on trend detection. Low sampling rates of 4 profiles per month generate unreliable trends; their figure 6 shows that 16-20 profiles per month are needed for reliable trend detection (but even then, there is some uncertainty). As they show in their introduction and conclusions, this finding is not new, and it corroborates the findings of other studies that go back to the 1980s. It also corroborates a well-known study based on IAGOS data (published in ACP, with an IAGOS co-author) that showed that once-a-week sampling is unreliable for calculating monthly means (Saunois et al., 2012). Based on the information shown in Figure 4 of the submitted paper, the current minimum sample size seems to be inadequate. The highest sampling rate is found above Europe, and we see in Figure 4 that the monthly anomalies are mainly within +/- 10 ppb. But when we examine other regions with lower sampling rates, such as the western Mediterranean, the Middle East and Siberia, we see that the monthly anomalies have much higher variability and many more extreme values; this is consistent with the noise associated with a low sampling rate. The authors need to conduct an analysis to determine an accurate minimum sampling rate. For example, they could simply conduct random sampling of the data above Europe and gradually reduce the sample size to 4 days per month. Based on the findings of Chang et al. (2024) the authors will likely see that the monthly anomalies become more variable with smaller sample sizes, and once the sample size is reduced to 4 days per month the trends will likely become unreliable.
3) The authors are basing their confidence in the trends solely on the signal-to-noise ratio, which is not correct given the uncertainty generated by small sample sizes and months with missing data (the p-value and the signal-to-noise ratio do not take these limitations into account). As discussed in recent TOAR-II papers (Chang et al., 2024; Gaudel et al., 2024; Van Malderen et al., 2025), the reported confidence needs to be adjusted to account for small sample sizes and missing data; please see Table 6 in Van Malderen et al. (2025) and Table 1 in Gaudel et al. (2024). This method of reporting trend confidence based on a thoughtful consideration of sample size and data availability is now becoming standard practice and the modelling community needs this type of valuable information. For example, when I look at Figure 4 of the submitted manuscript and see the extreme variability in monthly means, and when I see many months with missing data, my confidence in the trends becomes quite low, and I doubt that the reported trends are useful for evaluating model data.
Additional comments:
Given that the European dataset has the most data, and therefore the most reliable trends, why not place more emphasis on the very clear ozone increase in the European boundary layer? This is a credible result and would be of great interest to the research community, especially since some studies have suggested that highest values of observed surface ozone are decreasing across much of Europe, as shown by a recent TOAR-II paper (Nelson and Drysdale, 2025).
Cohen et al. (2018) reported trends for western North America, but the current study does not. Why?
Here I highlight the time series with large or frequent data gaps, which diminishes confidence in the trends. In the last 5 years of the record (2019-2023), NE Asia has ozone observations in just 17 months out of 60; this is a temporal coverage of just 28%. In addition, the reader has no idea how sparse the sampling is in the 17 months with reported data. Likewise, Siberia has data coverage in 11 out of 48 months in 2020-2023, or 23%. South Asia has data coverage in 19 out of 60 months in 2019-2020, or 32%. North Atlantic has data coverage in 27 out of 48 months in 2020-2023, or 56%. W Med has data coverage in 20 out of 48 in 2020-2023, or 42%. Tropical Africa has data coverage in 27 out of 48 months in 2020-2023, or 56%. Figure 4 is difficult to understand because grid cells with missing data are white, which can be confused with grid cells that have no anomaly. To distinguish between these types of cells, please shade the grid cells with missing data as black or gray.
Chang et al. (2023) show that quantile regression can be applied to different percentiles. Given that Cohen et al. 2018 looked at trends for different percentiles, and noted a stronger ozone trend at lower percentiles (also noted by Ryoo et al., 2026), why does your paper not consider this important aspect of ozone and CO?
The Introduction is not well written and needs to be completely revised.
1) The first paragraph begins abruptly, describing the relevance of CO on atmospheric composition, but the reader is left wondering where all this is going. It’s as if the original version of the manuscript had an introductory paragraph describing the importance of long-term monitoring of trace gases, but this paragraph was inadvertently deleted.
2) The third paragraph describes stratospheric ozone depletion and the Vienna Convention, but this paper is not focused on the stratosphere. It’s as if this paragraph once belonged to an entirely different paper.
3) The fourth paragraph discusses global tropospheric ozone trends but misses some key references, some of which relied heavily on IAGOS data (Zhang et al., 2016, Gaudel et al., 2020; Chang et al., 2022; Wang et al., 2022; please also see the second TOAR-HEGIFTOM paper by Van Malderen et al., which reported regional ozone trends, with contributions from IAGOS data).
4) The fifth paragraph (with just two sentences) mentions COVID-19, but only says that the pandemic resulted “in a change in O3 values during that period.” This statement is so vague that the reader has no idea how the pandemic affected ozone. Did ozone increase or decrease? Was the impact the same in urban areas and in the free troposphere? Please read and summarize the recent review by Heald et al. (2026), as well as these key papers: Steinbrecht et al., 2021; Sokhi et al., 2022; Putero et al. 2024.
5) When summarizing the findings from the Tropospheric Ozone Assessment Report, please refer to the new assessment papers by Szopa et al. (2026) and Hubert et al. (2026).
6) The final paragraph discusses the importance of IAGOS, and mentions studies that have used IAGOS data only, but it ignores influential studies that have successfully merged IAGOS data with other data sets to produce robust, long-term trends based on large sample sizes, such as: Gaudel et al., 2020; Chang et al., 2022; Wang et al., 2022; Gaudel et al., 2024; Van Malderen et al., 2026; Ryoo et al., 2026.
References:
Barnes, E. A., Fiore, A. M., and Horowitz, L. W.: Detection of trends in surface ozone in the presence of climate variability, J. Geophys. Res.-Atmos., 121, 6112–6129, https://doi.org/10.1002/2015JD024397, 2016.
Chang, K.-L., et al. (2022), Impact of the COVID-19 economic downturn on tropospheric ozone trends: an uncertainty weighted data synthesis for quantifying regional anomalies above western North America and Europe, AGU Advances, 3, e2021AV000542. https://doi.org/10.1029/2021AV000542
Chang, K.-L., Schultz, M. G., Koren, G., and Selke, N.: Guidance note on best statistical practices for TOAR analyses, https://doi.org/10.48550/arXiv.2304.14236, arXiv:2304.14236 [stat], 2023.
Chang, K.-L., et al. (2024), Technical note: Challenges in detecting free tropospheric ozone trends in a sparsely sampled environment, Atmos. Chem. Phys., 24, 6197–6218, https://doi.org/10.5194/acp-24-6197-2024
Cohen, Y., et al. (2018), Climatology and long-term evolution of ozone and carbon monoxide in the upper troposphere–lower stratosphere (UTLS) at northern midlatitudes, as seen by IAGOS from 1995 to 2013, Atmospheric Chemistry and Physics, https://doi.org/10.5194/acp-18-5415-2018
Heald, C. L., Jesse H. Kroll, Jennifer G. Murphy, Delphine K. Farmer, Juliane L. Fry; Atmospheric Chemistry Insights from the Global COVID-19 Pandemic: A Review. Environ. Sci. Technol. 14 April 2026; 60 (14): 10393–10404. https://doi.org/10.1021/acs.est.5c16737
Fiore, AM., et al. (2022), Understanding recent tropospheric ozone trends in the context of large internal variability: A new perspective from chemistry-climate model ensembles, Environmental Research: Climate, https://doi.org/10.1088/2752-5295/ac9cc2
Gaudel, A. et al (2020), Aircraft observations since the 1990s reveal increases of tropospheric ozone at multiple locations across the Northern Hemisphere. Sci. Adv., DOI: 10.1126/sciadv.aba8272
Gaudel, A., et al. (2024), Tropical tropospheric ozone distribution and trends from in situ and satellite data, Atmos. Chem. Phys., 24, 9975–10000, https://doi.org/10.5194/acp-24-9975-2024
Hubert, D., et al. (2026), Tropospheric Ozone Assessment Report II: Past and present tropospheric ozone using satellite observations, Philosophical Transactions of the Royal Society A
Lin, M., et al. (2015), Revisiting the evidence of increasing springtime ozone mixing ratios in the free troposphere over western North America, Geophys. Res. Lett., 42, doi:10.1002/2015GL065311.
Logan, J. A.: An analysis of ozonesonde data for the troposphere: Recommendations for testing 3-D models and development of a gridded climatology for tropospheric ozone, J. Geophys. Res.-Atmos., 104, 16115–16149, 1999.
Nelson, B. S. and Drysdale, W. S.: Urban ozone trends in Europe and the USA (2000–2021), Atmos. Chem. Phys., 25, 16009–16026, https://doi.org/10.5194/acp-25-16009-2025, 2025.
Putero et al. [2023] Fingerprints of the COVID-19 economic downturn and recovery on ozone anomalies at high-elevation sites in North America and western Europe, ACP https://doi.org/10.5194/acp-23-15693-2023
Ryoo, J.-M., et al. (2026), Origin of low ozone above western North America: an investigation of sources and trends, Atmos. Chem. Phys., 26, 13595–13615, https://doi.org/10.5194/acp-26-13595-2026
Saunois, M., Emmons, L., Lamarque, J.-F., Tilmes, S., Wespes, C., Thouret, V., and Schultz, M.: Impact of sampling frequency in the analysis of tropospheric ozone observations, Atmos. Chem. Phys., 12, 6757–6773, https://doi.org/10.5194/acp-12-6757-2012, 2012
Sokhi, R.S., Singh, V., Querol, X., Finardi, S., Targino, A.C., de Fatima Andrade, M., Pavlovic, R., Garland, R.M., Massagué, J., Kong, S. and Baklanov, A., 2021. A global observational analysis to understand changes in air quality during exceptionally low anthropogenic emission conditions. Environment international, 157, p.106818.
Steinbrecht et al. [2021] COVID‐19 Crisis Reduces Free Tropospheric Ozone Across the Northern Hemisphere, Geophys. Res. Lett. https://doi.org/10.1029/2020GL091987
Szopa S., et al. (2026), Tropospheric Ozone Assessment Report II: Ozone interactions with climate change: drivers and long-term trends, Philosophical Transactions of the Royal Society A
Van Malderen, R., et al. (2025), Ground-based tropospheric ozone measurements: regional tropospheric ozone column trends from the TOAR-II/HEGIFTOM homogenized datasets, ACP, 9905–9935, https://doi.org/10.5194/acp-25-9905-2025
Citation: https://doi.org/10.5194/egusphere-2026-4770-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 232 | 105 | 51 | 388 | 54 | 50 |
- HTML: 232
- PDF: 105
- XML: 51
- Total: 388
- BibTeX: 54
- EndNote: 50
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper by Lebourgeois et al. discusses long term-trends in CO and Ozone observed from the IAGOS data set which is available starting in 1994. Overall the paper and the analysis are sound and the results are of interest and within the scope of ACP. I have one major and some minor comments which I suggest should be addressed before publication. I can recommend the paper for publication after addressing theses comments.
I also have a few suggestions with respect to the presentation. In particular I would like graphics and figure captions to be more self-explanatory. E.g. some figures do not have axis (e.g. Fig 4/6) but these are only explained in the Figure captions. Also I suggest that Figure captions should be more informative so as to allow figures to stand more by themselves.
Major Comments
There are ground-based networks (e.g. ICOS in Europe, NOAA in North America) which monitor CO in the troposphere. The results of this study should be discussed more in conjunction with results from these networks and partially available emission estimates from such networks.
Specific comments.
Abstract: I was wondering why the trends as presented in the abstract where focussed on comparing the entire period (1994-2023) with the previous analysis which covered the time period up to 2013. I would like to see a trend focussed on the more recent period (e.g. 2013-2023) in the abstract. Note that such trends are presented in the paper.
l. 5. More pronounced than what? Maybe state first whether ozone has increased during which period.
l. 7. As noted above. Why not compare 1994 – 2013 with the trend over the last decade to emphasize changes in long term trends?
l. 10. Start by saying that a decrease in CO is observed.
l.26. avoid relative time statements. Like forty years ago. This will be outdated in 10 years from now.
l. 35: be careful; an increasing trend can be understood as a trend which is getting stronger (i.e. the 2nd derivative). Rather use positive and negative trend.
l. 38: I do not understand the causality in the phrase starting with although...
l. 45 emissions of what?
l. 50: again, I do not understand why this sentence starts with Nevertheless.
l. 74: shown, not showed
l. 75: explain the source of the PV values used for the tropopause criterion. There is a recent paper by Bauchinger et al. (https://acp.copernicus.org/articles/25/14167/2025/acp-25-14167-2025.pdf ) which discusses the use of different tropopause definitions. This actually shows that using 2 PVU is a good criterion in order to ensure that only tropospheric values are used.
l. 118: Some explanations are required to sustain this.
Section 2.2.: I suggest to start with a short explanation on quantile regression and why it is preferred over a linear regression.
Section 3. I suggest to show at least one plot of the time series in order to give the reader a better feeling for the underlying data. This could for instance be monthly means and standard deviations for ozone and CO over the entire period.
l. 172: change to in Siberia is observed
l. 180: this is a speculation and should be marked as such
l. 183: Fig 4 appears to .. sounds strange; change to Fig 4 highlights…
l 187: refrain from using relative references to Figures; give Fig Nr.
l. 192: unambiguous seems to contrast with qualitative; leave out.
Figure 4 and Figure 6: these Figure Captions need more detail
l. 223: what is meant by absolute maginitude here?
l. 240. The discussion now changes more towards attribution of the trends; maybe have a new sub-heading.