Recent ozone trends and changepoints in the Greater Region: The role of meteorology and NOₓ emission reductions

Trebs, Ivonne; Pinheiro, Philippe; Krein, Andreas; Junk, Jürgen

doi:10.5194/egusphere-2026-5341

Preprints

https://doi.org/10.5194/egusphere-2026-5341

Preprints

29 Sep 2026

| 29 Sep 2026

Recent ozone trends and changepoints in the Greater Region: The role of meteorology and NOₓ emission reductions

Ivonne Trebs, Philippe Pinheiro, Andreas Krein, and Jürgen Junk

Abstract. We analysed long-term records of NO₂ and O₃in Luxembourg and the Greater Region with the Bayesian Estimator of Abrupt change, Seasonality, and Trend (BEAST) to identify probabilistic changepoints in time series. Seasonal components of NO₂ and long-term components of O₃revealed a significant changepoint in 2017/2018 that is more prominent than the effect of the short-term emission reductions in 2020/2021 and the subsequent post-Covid rebound in urban areas. NO₂ concentrations decrease by at least 50 % for all station categories from 2009 to 2024. Linear trends for O₃ during this period based on quantile regression and Theil-Sen trend estimator were positive, with highest values at urban stations (~0.75 µg m^-3 yr^-1), while increases at peri-urban, rural, and forest sites were smaller (~0.4–0.6 µg m^-3 yr^-1). Spring and autumn are the dominant seasons contributing to the O₃ increase. The decline of the positive weekend effect, together with the reduced MDA8 O₃ sensitivity to maximal air temperatures above ~20 °C after 2018, indicates an ongoing transition from predominantly VOC-limited towards NOₓ-limited chemistry, particularly at peri-urban, rural, and forest stations during summer. However, our results suggest that the impact of reduced O₃ titration by NO, particularly at low to moderate temperatures has sustained increasing O₃ in the Greater Region. Simultaneously, the reduction in peak O₃ due the influence of NO_x-limited chemistry was partly offset by the increase in daytime maximum boundary layer height observed after the 2017/2018 changepoint.

Received: 03 Sep 2026 – Discussion started: 29 Sep 2026

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

Download & links

Ivonne Trebs, Philippe Pinheiro, Andreas Krein, and Jürgen Junk

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-5341', Anonymous Referee #1, 17 Oct 2026
This manuscript presents a comprehensive analysis of long-term NO2 and O3 observations spanning up to 25 years (2000–2024, with a detailed focus on 2009–2024) across 15 monitoring stations in the Greater Region of Western Europe (Luxembourg, Belgium, France, and Germany). The study investigates seasonal and long-term trend changepoints using primarily the Bayesian Estimator of Abrupt change, Seasonality, and Trend (BEAST), Quantile Regression with Moving Block Bootstrap (QR-MBB), and non-parametric Theil-Sen slope estimates combined with SSA and STL seasonal decompositions. The authors report a consistent downward trend in NO2 concentrations (>50% reduction from 2009 to 2024 across all station categories) alongside widespread positive trends in surface O3. A distinct probabilistic changepoint was identified around 2017/2018, driven by shifts in winter/autumn NO2 seasonality, fleet composition changes, and meteorological changes such as deeper daytime BLH. Overall, the manuscript is well-structured, thoroughly executed, and addresses crucial policy and chemical-regime questions regarding surface ozone formation in Western Europe.

Definition of changepoint; This study does not adequately distinguish between abrupt changes and changes of trend magnitudes. Abrupt changes typically refer to step changes, not necessarily implying the long-term trends have changed. Changes of trend magnitudes can indicate turnaround, acceleration, and deceleration. Currently those two concepts are mixed badly in the manuscript. See Weatherhead et al. (2017), Lund et al. (2023) and Chang et al. (2025) for further discussion. Since the authors hardly discuss trends before changepoints, This is not obvious to me if authors want to discuss abrupt changes or trend changes. The authors are required to report the scenario for each site type, e.g., which sites truly have trends of turnaround, merely an anomalous year (see my next comment), or other cases (changes of data variability). A plot similar to Figure 2 of Chang et al. (2025) would be helpful.

The BEAST algorithm detects a major changepoint in autumn 2017/2018 for O3 trends and NO2 seasonality. The manuscript attributes this partially to vehicle fleet composition shifts (e.g., Euro 6 SCR systems, rising electric/hybrid registrations starting ~2017/2018) and partially to meteorological shifts (step-like increase in maximum daytime BLH). How confident are the authors that the BEAST seasonal/trend changepoint in late 2017 was not artificially accentuated by the extreme European heatwave of summer 2018? Indeed, based on Figure 3d, it looks like the gray stripes coincide with heatwave years 2003, 2006 and 2018 (much better correlations than 3a and 3c). Heatwaves appear to be the main driver of abrupt changes, instead of NO2.

Section 2.1 mentions that early NO2 records before 2008 were flagged for seasonal irregularities and removed from the final dataset for several stations. Could the authors specify which exact stations required pre-2008 NO2 removal, and how this affects the analysis of 2000–2024 time series in Figure 3 and relevant figures? I am very skeptical about such a data removal. Seasonal irregularities appear to be commonly observed in NO2 data. Recently available satellites and Pandora network often see seasonal irregularities in total column NO2. Why are those considered suspicious? Evidence and references are needed.

Data availability: I know the "data upon request" statement is now prohibitive for many journals, but I am not sure it is still allowed for ACP.

Minor comments: L14, specify the exact number of sites

Chang, K. L., McDonald, B. C., Harkins, C., & Cooper, O. R. (2025). Surface ozone trend variability across the United States and the impact of heat waves (1990–2023). Atmospheric Chemistry and Physics, 25(10), 5101-5132.
Lund, R. B., Beaulieu, C., Killick, R., Lu, Q., & Shi, X. (2023). Good practices and common pitfalls in climate time series changepoint techniques: A review. Journal of Climate, 36(23), 8041-8057.
Weatherhead, E. C., Harder, J., Araujo-Pradere, E. A., Bodeker, G., English, J. M., Flynn, L. E., ... & Woods, T. N. (2017). How long do satellites need to overlap? Evaluation of climate data stability from overlapping satellite records. Atmospheric Chemistry and Physics, 17(24), 15069-15093.
Citation: https://doi.org/10.5194/egusphere-2026-5341-RC1
RC2:
'Review Report on Trebs et al. 2026', Tabish Ansari, 02 Nov 2026
Trebs et al. 2026 present a comprehensive analysis of ground-based observations of the “Greater Region” with a focus on ozone trends and changepoints in the context of changing meteorology and reducing local NOx emissions. I thoroughly enjoyed reading this manuscript. It’s very well-written and presents the results from a purely observational point of view. I have several comments but they are mostly minor in nature, and once addressed, will make this manuscript fit to publish in ACP. My comments are listed below:

International readers of the journal may not be familiar with the generic term “the Greater Region” as mentioned in the title. Consider using something like “SaarLorLux” or “Greater Region in Western Europe”.

The introduction is very well written. It provides the state of the science and frames the motivation of the study well.

The meteorological data used in this study are not taken from ground stations but instead from gridded reanalyses dataset ERA5. Therefore it is not certain how representative the gridcells are for T2 and PBLH to individual ozone stations. Include a sentence with information on how unique these corresponding gridcells were. Also mention if meteorological data was pulled from the same gridcell for multiple ozone stations (in case they were close). Add a few sentences on uncertainty of using this approach rather than actual ground-measured meteorology.

L186: The authors mention that following an initial increase between 2000 and 2003, the NO2 concentrations remain “relatively stable” for approximately a decade. I do not see this in Fig 2a. I see a modest but consistent decrease, which intensifies in the second decade.

L189: Here the authors state that for urban stations NO2 concentrations decline to near 10 ug/m3 after 2020. It would be more accurate to say that they decline to 15ug/m3 and below.

L215: Here the authors note that the amplitude and phase of the O3 seasonal cycle remained comparatively stable over the 2000-2024 period. This is in contradiction with the findings of Ansari et al. 2025 who found a dip in the summertime peak with no change in the springtime peak (Ansari et al., 2025; Figure 15). It would be worth putting the results in context of the published literature and offer some reasons for the difference in results (e.g., the specific stations included in the analyses here vs the wider western European stations considered in Ansari et al., 2025).

Figure 3: The fact that the changepoints for NO2 and O3 occur at very different periods points to the fact that much of the O3 change is in foreign-transported ozone. Are the NO2 and O3 changepoints related in any way? This could be discussed.

L280: It’s not appropriate to use the term “stations” here since the data isn’t coming from stations but rather from corresponding gridcells of the ERA5 dataset. You may use something like.. “Meteorological data corresponding to urban stations…”

Figure 5: Is there any physical explanation (in terms of wider climatic phenomenon) for the elevated BLH over this region during the 2017-2024 period? Perhaps in the published literature? If so, it would be good to include it here.

L305: “BLH changepoint probabilities are mostly higher…..extreme events”. This sentence is not clear. Consider rephrasing.

Figure 7: These are very interesting results with implications for distinguishing foreign vs local O3 from a purely observational data perspective (given that foreign-transported O3 dominates the lower-end of O3 distribution while local-photochemically-produced O3 dominates the upper tail).

Figure 9: This is a very thoughtful analysis which, in a way, provides insight into the reasons for increasing ozone trends: since the increases are more pronounced in autumn/winter, it points to weakened local titration rather than increased hemispheric transport.

Figure 11 and its related text: The authors should find opportunities to relate the different trends across different o3 percentiles to fundamental physical and chemical processes in the atmosphere.

L503: Here the authors mention that they classified the o3 data based on 2.5oC bins and then the corresponding mean MDA8 o3 was calculated for each bin. This is not clear to me. Once you bin the o3 data based on temperatures, you have already time-scrambled the datapoints and then “MDA8” cannot really be calculated. MDA8 calculation requires consecutive hourly o3 datapoints. Please clarify.

L515: Here the authors discuss the concept of “temperature sensitivity of MDA8 O3” relating to lines in Figure 14. However, they refer to temperature sensitivity being higher when the colored line is above the dotted line. This is not how temperature sensitivity is defined. Instead it is (∆MDA8 O3 / ∆T) or the slope of the line. So, the temperature sensitivity is “higher” when the slope of the line is steeper, and it is lower when the slope of the line is flatter. Therefore, the statement that the air temperature sensitivity decreased at peri-urban, forest, and rural stations above 20oC is not quite correct, because it decreased for all temperatures (since the slope decreased despite the magnitude of MDA8 o3 being higher). This should be corrected. Basically, the dependence of MDA8 o3 has decreased throughout the o3 range and across all stations. It’s worth pondering (and discussing) why this is the case. Is it due to more foreign-transported ozone?

L508: This statement is not true (refer to comment 15).

Figure 15 and its description: Again, make it clear that the magnitude of MDA8 o3 under lower max-BLH has increased but its “dependence” on BLH hasn’t increased (because the slope is lower).

Figure 16: The solid lines representing the AOT40 for the recent period (2019-2023) remain lower than the AOT40 for the early period (2007-2011) but eventually merge with (for forest stations) or exceed them (for rural stations) towards the very end. This means that the AOT40 metric in late-summer was quite high. This is in contradiction with the decline in late-summer peaks of MDA8 O3 in Western Europe (Ansari et al. 2025 and others), which I can also see in your Figure 4b if I look closely. So, what’s going on in the AOT40 calculation that’s leading to this increase in late summer? Please explain and include a couple sentences in the manuscript to help the reader better interpret this metric in relation to MDA8 o3 changes.

L565: Can you provide a directional statement on the expected change if these parameters are considered for the AOT40 calculation?
Citation: https://doi.org/10.5194/egusphere-2026-5341-RC2

Ivonne Trebs, Philippe Pinheiro, Andreas Krein, and Jürgen Junk

Viewed

Total article views: 585 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
375	165	45	585	45	41

HTML: 375
PDF: 165
XML: 45
Total: 585
BibTeX: 45
EndNote: 41

Views and downloads (calculated since 29 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	140	37	8	185
Oct 2026	195	90	26	311
Nov 2026	40	38	11	89

Cumulative views and downloads (calculated since 29 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	140	37	8	185
Oct 2026	195	90	26	311
Nov 2026	40	38	11	89

Viewed (geographical distribution)

Total article views: 585 (including HTML, PDF, and XML) Thereof 585 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 13 Nov 2026

Short summary

Ground-level ozone is a harmful air pollutant that affects human health, crops and ecosystems. We analysed 25 years of air quality measurements from cities, rural areas and forests in the Greater Region. Despite a major decline in nitrogen dioxide, ozone concentrations continued to rise. Our results show that future air quality policies must address both pollutant emissions and climate change to reduce ozone effectively.


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