the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Feb 2025
Urban Ozone Trends in Europe and the USA (2000–2021)
Abstract. Trends in urban O3 and NO2 across Europe and the United States of America were explored between 2000–2021. Using surface monitoring site data from the TOAR-II and European Environment Agency databases, piecewise quantile regression (PQR) analysis was performed on 228 O3 time series (144 European, 84 USA) and 322 NO2 times series (245 European, 77 USA). The PQR analysis permitted 2 break points over the 23 year period to balance the intent to describe changes over a large time period, while still capturing the abrupt changes that can occur in urban atmospheres. Regressions were performed over quantiles ranging from 0.05 to 0.95 and indications of a slowing in the increase of high European O3 levels was observed. In Europe, more trends were found to having an increasing O3 trend between 2015–2021 compared with 2000–2004. The reverse was true in the USA, with a reduction in the number of sites with increasing O3 trends when the same periods were compared. An analysis of the change points revealed a large proportion of sites in Europe, were the second change point in NO2 switched from a positive to negative trend, occurred in 2020 (41/43 second change points in this year). This was attributed a reduction in NO2 due to the COVID-19 pandemic, however, in some cases these increasing trends have sustained beyond the recovery from restrictions.
- Preprint
(14339 KB) - Metadata XML
-
Supplement
(48 KB) - BibTeX
- EndNote
Status: open (until 25 Apr 2025)
-
RC1: 'Comment on egusphere-2024-3743', Anonymous Referee #1, 02 Mar 2025
reply
This manuscript describes a trend analysis of urban ozone in Europe and USA over 2000-2021. I believe the authors have great ambition and spent a great deal of time putting the analysis together. Unfortunately, their approach and discussion appear to be premature and unskilled for ACP. In general, I will expect a vast and extensive rewritten if this manuscript is not rejected. Some major issues are pointed out as follows:
- Methodology: Although using nonlinear methods such as Loess can visually identify the change points, are the authors really inspecting all the ozone and no2 time series (>500) and recording the change points for each individual location?
- Methodology: AIC is most useful to avoid overfitting, so it can be used to determine if a model with change points (more parameters) is actually better than a model without change points (fewer parameters). But AIC does not tell us the optimal changepoint location, the authors should discuss how they select the change point locations. More importantly, how to select change points objectively (eg, Muggeo 2003; Chen et al., 2011), given that some change points may be hindered by data variability and not visually detectable. The authors should also properly define what they mean regarding change points. To me, the authors merely compare the trends between different periods, and see which locations have large trend differences. This does not really change point analysis in statistics.
Chen, C. W., Chan, J. S., Gerlach, R., & Hsieh, W. Y. (2011). A comparison of estimators for regression models with change points. Statistics and Computing, 21, 395-414.
Muggeo, V. M. (2003). Estimating regression models with unknown break‐points. Statistics in medicine, 22(19), 3055-3071.
- Majority of US ozone studies show the ozone reductions since 2000 in response to emissions controls, but this study shows contradicted results. The fundamental problem of this study is that they use daily median ozone to conduct trend analysis, which is neither relevant to human health nor generally interesting. Medians can occur either daytime in one day or nighttime in another, mixing two makes the trends really ambiguous and difficult to interpret. I don't understand why the authors do not use MDA8 or daytime/nighttime observations, especially since this study is focused on urban ozone.
- Scientific interpretations are essential for ACP. Most discussions have only scratched the surface and not provided sufficient interpretations. For example, NOx is not the only proxy for ozone production, disproportionate seasonal trends (winter increases and summer decreases), wildfires, and weather (2003 heat wave in Europe and 2012 heatwave in the eastern US) also play important roles (Cooper et al. 2012; Simon et al., 2015; Seltzer et al., 2020; Wells et al., 2021; Chang et al., 2023), but none of these factors are discussed.
Chang, K. L., Cooper, O. R., Rodriguez, G., Iraci, L. T., Yates, E. L., Johnson, M. S., ... & Tarasick, D. W. (2023). Diverging ozone trends above western North America: Boundary layer decreases versus free tropospheric increases. Journal of Geophysical Research: Atmospheres, 128(8), e2022JD038090.
Cooper, O. R., Gao, R. S., Tarasick, D., Leblanc, T., & Sweeney, C. (2012). Long‐term ozone trends at rural ozone monitoring sites across the United States, 1990–2010. Journal of Geophysical Research: Atmospheres, 117(D22).
Seltzer, K. M., Shindell, D. T., Kasibhatla, P., & Malley, C. S. (2020). Magnitude, trends, and impacts of ambient long-term ozone exposure in the United States from 2000 to 2015. Atmospheric Chemistry and Physics, 20(3), 1757-1775.
Simon, H., Reff, A., Wells, B., Xing, J., & Frank, N. (2015). Ozone trends across the United States over a period of decreasing NOx and VOC emissions. Environmental science & technology, 49(1), 186-195.
Wells, B, Dolwick, P, Eder, B, Evangelista, M, Foley, K, Mannshardt, E, Misenis, C, Weishampel, A. 2021. Improved estimation of trends in US ozone concentrations adjusted for interannual variability in meteorological conditions. Atmospheric Environment 248: 118234. http://dx.doi.org/10.1016/j.atmosenv.2021.118234.
- If the authors aim to study extreme ozone events (Section 3.2), they should know the difference between the 95th percentile of daily medians and the 95th percentile of daily MDA8 or hourly observations, these are completely different concepts. Using daily medians to study extremes is completely misleading and unreasonable.
- All the figures with maps (eg Figs 3, 4, 9-12) have low quality.
l22 which WHO standards?
l49 please clarify which analysis will end in 2021 or 2023.
l63 medians are not calculated through averaging.
l64 please provide justifications. A visual inspection may be subjective.
l66 deseasonalized
l217 should the units of slope be ppb/decade or ppb/year?
Citation: https://doi.org/10.5194/egusphere-2024-3743-RC1 -
RC2: 'Comment on egusphere-2024-3743', Anonymous Referee #2, 15 Apr 2025
reply
This study presents an analysis of the multi-year trends in surface ozone (O3) and nitrogen dioxide (NO2) mixing ratios at a few hundred urban and suburban locations across Europe and the USA that have near-complete time series between 2000 and 2021.
The predominant statistical approach the authors use is piecewise quantile regression (PQR) on deseasonalised daily median values of O3 or NO2 at each site. The authors permit the PQR to have two change points in the trend over the time period, subject to some restrictions such as change points having to occur on a 1st of January, there is at least 5 years between change points, and there are no change points in the first and last 2 years of the time period. Change points are different between each pollutant and each site. The author summarise, mainly visually, the extent to which the collection of sites show increasing or decreasing (or non-significant) trends in O3 and NO2 across the whole time series, how the trends change in magnitude and/or sign at change points, and in what years the change points occur. The summaries are presented separately for the sites in Europe and in the USA.
General comments
Substantial effort has clearly been expended in undertaking all the PQR (and other) statistical analyses and in thinking of ways in which to visually summarise the resulting datasets of trend directions, magnitudes and changes. These visual summaries are inventive and helpful for appreciating the distributions within these summary datasets.
- However, the overarching issue with this work is “why”? Or, to reframe the “why” into two more specific questions: what is/are the scientific and/or policy questions motivating these particular analyses; and what do scientists and/or policy-makers learn from these analyses that isn’t known before? With respect to the first of these: as the authors note in the Introduction, a lot of analyses of O3 and NO2 trends have come before; but the Introduction doesn’t really identify what specific questions motivate this particular study, and why the authors’ approach is well-suited to answer such questions. With respect to the second: the paper is essentially a description of the summary statistics of the analyses – there is very little detailed discussion of what the reader learns scientifically or policy-wise from these analyses.
- A second issue requiring further justification is the use of daily median O3 and NO2 as the underlying measure of O3 and NO2 levels. The majority of metrics used to capture O3 levels are based on the daily maximum 8-hour mean. This latter way of defining a daily value for O3 level could as easily be calculated as the median daily value, so why was it not used? It also means that when this paper is referring to trends in high levels of O3 it is referring to trends in the highest daily median O3 levels which does not match the usual way of thinking about episodic high O3 levels in the literature or in air quality quantification.
- What is the rational for choosing a 5-y period (2000-2004) as a time range over which to summarise trends in daily medians at the start of the full time-range of the datasets, but a 7-y period (2015-2021) as the time period over which to summarise trends in daily medians at the end of the full time-range of the datasets? Doesn’t the length of time period used to quantify a trend potentially have some influence (bias) on the distribution of trend magnitudes and p-values? i.e., that this is not a like-for-like comparison?
- What is the scientific merit in comparing average trend values across those sites with significant positive trends (and average trend values across those sites with significant negative trends) between different time periods and between different geographical areas, given that there are different numbers (and different identities) of sites contributing to each of those average trend values?
At present there seems to be insufficient insight from the analyses to justify publication. A major revision would require substantial attention to the above-mentioned motivations and take-home messages, and attention to questions about use of median values.
Minor and editorial comments
L4: The time series are described here as being 23 years long, but the date range given in the title and in the first line of the abstract comprises 22 years.
L6: Need to specify is meant by “high” European O3 levels, i.e. what metric of O3 is used to define “high”.
L8: In this sentence, is the word “trend” still referring to high O3 levels, or is it now referring to trends in some form of average O3 level?
L8: Is a 5-year period of 2000-2004 long enough to be confident of the direction and magnitude of a trend? Ozone concentrations are notoriously temporally variable.
L9: typo in “where”
L28-30: I don’t understand the point being made in the last two sentences about citing locations with highest absolute values in 4MDA8 and NDGT70: all discussion to this point has been about trends not absolute values. The next paragraph returns to talking about trends again.
L31-L33: Again, I am struggling to understand the narrative here. In the first 3 sentences of this paragraph, it is noted how precursor emissions in in the USA and Europe have been declining since the late 80s/early 90s, but in a sentence at the end of the previous paragraph it talks about North America and Europe having highest precursor emissions. What is the point the authors want the reader to take away?
L49: There is an error in the time-series range quoted here (“2020-2023”). Neither the start or the finish year match those given in the paper’s title, nor are the constituent numbers the same but accidentally typed in the wrong order.
L137: typo in “where”
The captions of Tables 2 and 3 are not clear enough. It needs to be made clear that the data in the table are numbers of sites having the trends specified. The captions imply that the data in the table are the trends.
Section 3.3: Needs to be citation to Figures 9 and 10 in this section.
Section 3.3: The values for the trend switches are given in ppb, but if these values describe trends shouldn’t there by a temporal component to the unit?
Figure 13: Several sites don’t appear to have increasing NO2 levels from 2020 onwards, which is what the caption and associated manuscript text description state that this figure shores.
Citation: https://doi.org/10.5194/egusphere-2024-3743-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 208 | 31 | 11 | 250 | 18 | 4 | 5 |
- HTML: 208
- PDF: 31
- XML: 11
- Total: 250
- Supplement: 18
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 99 | 39 |
| China | 2 | 38 | 15 |
| United Kingdom | 3 | 24 | 9 |
| India | 4 | 14 | 5 |
| Germany | 5 | 11 | 4 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 99