The authors of this manuscript present an observation-based model of ground-based ozone measurements across the U.S., including a closer analysis of the U.S. southwest region and several urban areas of interest within it. Using a regression model fit to observational data, they seek to separate U.S. anthropogenic from U.S. background ozone, and perform an analysis of trends in both ozone categories over time. Their results highlight the decline in U.S. anthropogenic ozone and the dominance of background ozone in the southwestern U.S. They also advocate strongly for this observation-based modeling approach over the use of chemical transport models (CTMs).

My general comments are as follows, with line-by-line comments below.

First, the introduction provides a critique of CTMs and speaks strongly in favor of the observation-based model that this paper will use. From the reader’s perspective, this seems out of order given that we haven’t yet come to understand how their model works and what it offers to justify favoring it over CTMs. I suggest relocating some of this content to the Discussion section. At the same time, it seems that there should be a place in the air quality research community for a combination of observations, observation-based models, and CTMs in the effort to understand ozone behavior and spatiotemporal trends. Rather than ranking one approach over another, they should be viewed as complementary. Moreover, ozone benefits from high spatial coverage of ground-based measurement networks that lends itself to observation-based modeling, whereas many other air pollutants do not have nearly this much measurement data upon which to judge regulatory efficacy. It seems beneficial to continue developing CTMs that could be validated against the large networks for ozone and PM that could then also be applied to more sparsely-measured pollutants. Discounting the validity and utility of CTMs seems counter-productive.

Second, the work presented here hinges heavily on earlier work published by the same group of authors in 2021 and 2022, as well as prior works by larger groups of authors in 2017 and 2019. Could the authors comment on (a) how the present manuscript builds upon but is also distinctly different from these earlier works, an (b) how their approach and outcomes are comparable with other groups pursuing the same questions in the literature? Lines 337-343 address question (a), but this comes at the start of the Results section and it may help the reader to understand these points earlier in the manuscript.

Third, throughout the manuscript the authors oscillate between the terms baseline ozone, background ozone, and ozone design values (ODVs). While their analysis focuses on ODVs, they also discuss both background and baseline in relation to their own work and the work of others. Over the course of the manuscript, when jumping between terms even in the same paragraph, for a reader it gets difficult to recall which is which. The authors could handle this more carefully, as these terms are easily confused but are also related.

Abstract

Line 16 refers to ODV as “an extreme value statistic of relatively rare maximum ozone concentrations upon which the NAAQS are based”. At first read, this came across as a critique of the meaning/utility of this metric for the NAAQS. Is it meant in this way? It may be of greater utility to define it first in the way that the NAAQS do (which appears to be the same as MDA8 based on the first paragraph of the introduction) and then if the authors want to emphasize that these values are extreme and rare, this could be added after the definition of the term.

Line 25 onward: The five points made here read more like findings, rather than implications (as they are named on line 25). I suggest the authors reorganize this to first clearly state the findings of their analysis, and then follow with the implications of those findings. These five points also do not need to be numbered and can just be provided as sentences incorporated into the text. Also, is point #3 only applicable to the southwestern U.S.?

Introduction and Background

Line 57: Can the authors clarify why they included Texas with the rest of the southwestern U.S.? Is it because the urban areas in Texas violate the NAAQS? Could they also define southwestern U.S. by listing the states in parentheses?

Lines 70-71: Could the authors offer a very brief explanation of how Parrish et al (2017) isolated the U.S. anthropogenic ODV? This seems like an important piece of earlier work upon which the current method in this manuscript hinges.

Lines 75-78: Can the authors offer a citation for this statement that baseline ozone constitutes the large majority of the U.S. background ODV? Is this also based on the Parrish et al. 2017, 2019, 2020, and/or 2022 manuscripts that are cited here?

Lines 84-86: Are these the same definitions of anthropogenic and background that are used in other similar studies? It seems very notable that domestic precursor emissions from agricultural soils, livestock, and VCPs fall under “background” by the definition here and it would be helpful to know if that is consistent with or different from the broader literature.

Lines 88-89: If a rare, transboundary event occurs on top of typical background conditions, then it would seem that the two are in fact additive. Can the authors explain this further?

Line 90: Can the authors elaborate on what they mean by “when accurately determined”? As a reader this raises the question of examples where U.S. background ODVs have not been accurately determined.

Line 94: At this point it is still unclear what “the above referenced observation-based analysis” actually is. Without having to also read Parrish et al (2022) in detail, could the authors provide a brief but more specific description of that analysis here? It seems to be a critical foundation upon which the current work builds.

Lines 108-144: (Related to the general comment above) This section spends a fair bit of text justifying the use of their observation-based model in favor of CTMs, but we haven’t actually seen what their model is or how it performs yet. So, the detailed CTM critique and comparison feels preemptive. Much of this section actually reads like discussion or conclusions. I suggest that the authors focus the Introduction more on the underlying ozone problem they seek to understand, and available methodologies to do so, which will inform the observation-based modeling approach they are about to show us. A discussion of what CTMs can and cannot currently do is warranted, but I suggest the authors withhold judgement on the strength of their methodology over CTMs until the conclusions when they have provided evidence to support this argument.

Data sets

Lines 158-168: Earlier it seemed that ODVs and MDA8 values are the same, but this paragraph would suggest otherwise since it is stated that MDA8 was used for the California air basins and ODVs for the rest of the U.S. Can the authors please clarify this? It is also unclear from this text whether exceptional events were included or excluded (either by choice of the authors or by reporting to the EPA database). Though it appears Section S6 elaborates more on the impact on the exceptional events, can the authors briefly state here how they determined that the analysis in this paper was not significantly affected?

Methods

Line 184: Is it the case that ODVs are in fact subject to diurnal and seasonal variations, given that the highest MDA8 values are almost certain to occur in summertime at midday? It is unclear how the 3-year averaging would eliminate this dependence.

Line 215: For readers unfamiliar with Parrish et al (2022) can the authors briefly explain what this approach entails? This would aid in following the subsequent presentation of their mathematical model. Or is it the case that the method discussed on line 216 is the same as the Parrish et al (2022) approach? What elements are the same and what elements are different?

Lines 231-236: Can the authors offer a bit more explanation on why the a, b, and c terms represent the stated components of temporal ozone behavior? How confident are the authors in the “direct physical interpretations” of these values?

Lines 244-245: Can the authors specify where the observations were generated that resulted in these values in Parrish et al (2020)? The “northern midlatitudes” are mentioned but is this from sites in the U.S.? Europe? Asia? Multiple continents? And how many sites? This would help readers to understand how applicable these values are likely to be for the sites in question in the present manuscript (since they are in fact used here).

Line 252: Is this statement about urban ODVs decreasing rapidly and approaching non-zero background based on results from this work or from an earlier publication? If it is the latter, then could a citation please be provided? This is important for readers to understand how the A*exp(-t/T) term was derived. Later in the paragraph they say that more information on the choice of the exponential function is described in the supplement but a brief explanation would be helpful here.

Line 257: Earlier the authors said that the value for c would be derived from the present work (Line 237) so how should readers interpret the use of a previously derived value for c here?

Line 279: Does the Iglesias et al (2022) citation apply only to the third point in this sentence, or to all three points? The authors have already cited the works by Westerling in support of point #2, but can at least one citation be provided in support of point #1?

Line 285: Where is the factor of 4 represented in the model? And how do the authors know the location specific ODV enhancement due to wildfires (the WF parameter)?

Line 288: Similar to the previous question, how do the authors know which locations have significant enhancements of ODVs due to wildfire emissions? Are these factors predicted by the data or informed by some a priori knowledge?

Lines 302-309: This in part addresses an earlier comment about the exclusion of these precursor emissions from the definition of U.S. anthropogenic ozone. It would be helpful to circle back on this point for specific regions or urban environments in the discussion of results in this manuscript. It seems like this exclusion would have more notable impacts on some urban areas (e.g. Denver) more than others, and it would help to acknowledge this based on the body of literature that has explored some of those locations specifically.

Results

Lines 332-336: This paragraph would fit well at the end of the introduction or start of methods, so that the objective of the methods is easier for readers to understand. (The end of line 334 also appears to contain a typo).

Line 346: Please also state for the reader what sections 4.3 and 4.4 will analyze.

Line 370: Earlier it was stated that coefficient b was derived from earlier work and applied here, whereas coefficient c would be derived here (lines 244-245). But this sentence implies different b coefficients by site. Can the authors clarify this?

Line 415: change “caused” to “cause”

Figure S1 is referenced frequently throughout section 4.2, and is the subject of the first sentence of the discussion online 569. I suggest the authors consider moving this figure to the main body of the text if it is going to be a focal point of discussion and not simply supporting information. If the current Figure 3 is sufficient to make their points, then the Results and Discussion should focus on this figure, with a direction to the supplement for more detail. But in the current format it seems that the reader needs to see all of Figure S1 to understand the discussed outcomes.

Line 453: Could the number of rural sites considered be added in parentheses?

Line 528: There appears to be a typo at the end of this sentence: “in southern and eastern Texas in the Southwest Texas”

Section 4.5 reads as supplementary information, and could be reduced to a couple of sentences for the main body of the paper.

Discussion and Conclusions

(see comment above about moving Figure S1 to the main text)

Lines 573-574: How do the authors know that 0.6-1.0 km is “high enough to avoid continental influences as air is transported ashore and low enough to represent air mixed into the convective boundary layer over the continent”. Is this a novel suggestion from these authors or supported by the literature (for which citation(s) could be provided)?

Lined 596-606: These sentences provide very helpful introductory material to set up the motivation for trying an observation-based model and highlighting the challenges or missing elements of CTMs. I suggest to move these sentences to the Introduction, and in exchange move comments from the Introduction that compare the observation-based model to CTMs to some part of Section 5, now that we can see how the authors’ model performs.

Much of Section 5.2 also reads as introductory background material, and very minimally integrates discussion of the new results obtained in this work (including Figure 9, which is mentioned here but as far as I can tell is not derived from the model they developed for this manuscript). I think this section would actually set the stage nicely for why the authors are motivated to do this observation-based model analysis, and suggest it be integrated into the introduction as helpful framing for the reader.

Line 668: Have other CTM analyses been published more recently with higher resolution models, as model development has continued to improve? Or is 200x200km the best available right now?

Line 687: Text refers to the “Crestline” site but Figure 12 caption refers to “Crestone”. One of these appears to be a typo?

Line 716 (or elsewhere): what changes in ozone were observed in the urban areas examined here during the initial COVID-19 lockdowns? In the absence of local transportation emissions, did urban NO2 and ozone decrease in any of these cities? This information may not be readily available for all cities under consideration, but would be an interesting piece of evidence to consider with regards to the magnitude of the US anthropogenic contribution. It may, however, also be the case that the lockdowns were not long enough and not during peak ozone season for such information to be impactful. But it may be worth investigating.

Line 734: The comparison of ozone to ionizing radiation is an interesting one. But is it not a bit of an apples-oranges comparison when you consider how ozone is transported, photochemically produced, and thus a secondary transboundary air pollutant that can cross state lines? Where do you draw the boundary for a southwestern U.S. standard?

Line 746: this recommendation to offset wildfire ozone by lowering urban NOx seems like it would depend on whether the urban environment was NOx or VOC limited. Can the authors comment on this for Denver or any other urban area?

Also in terms of Denver, the definition of U.S. anthropogenic in this manuscript does not include agriculture or oil and gas – this is missing a key source of local ozone in the Denver Basin that is well-documented in the literature and does not seem to fit the category of “background” given that local efforts could in theory reduce the contribution of these sources to Denver’s ozone production. Can the authors please comment?

Lines 765-771: Can the authors also comment on how (if at all) their interpretation would change if the ozone NAAQS were further lowered from 70 ppb to, say, 65 ppb? Will areas other than the southwestern U.S. still be required to reduce their local anthropogenic precursor emissions to remain in attainment? It may be difficult to speculate on something that hasn’t happened at this point, but a standard lower than 70 ppb was considered in the past and we just saw a tightening of the PM2.5 primary standard which has analogous regional effects on the challenges of emission controls. Is there a related point to be made for ozone, should something similar happen in the future?

Summary:

Background ozone is an important consideration for meeting national air quality standards and future policy. A key finding of this work is that background ozone is a major contributor to observed ozone levels in the southwestern US and that anthropogenic enhancements are small relative to this background ozone contribution. This paper brings value and insight into how we quantify background ozone in the southwestern US, how we think about anthropogenic enhancements on top of that background in this region, and what the implications are for continued emissions reductions in this region in the broader context of national air quality standards.

Comments

L81: The text box of key terms is a helpful visual aid in the intro. However, the current definitions in this table are confusing. For example, US background ozone is defined as “USB” in the text-box, but that definition is never used again. Further, “ODV” is used to define two different things in the text box. It would help to re-define or denote the terms in a way that better assists the reader in recognizing the differences right away. For example, background ODV could be denoted as ODV subscript BKG, reported ODVs could be denoted as ODV subscript reported, and anthropogenic enhancements could be denoted as ODV subscript enh.  

Even with this table of terms, it was occasionally hard to follow when the authors are referring to ODVs reported, background ODV, or the anthropogenic enhancement in ODVs above background. So, it could also help to have more consistency in the usage and definitions of these terms.

It was somewhat confusing to find the fit parameters in this initial table in the intro. Breaking up this information and moving the descriptions of the fit parameters to a separate text box in the methods section where the equations are discussed might be a better way of presenting this content. Adding the equations to this text box with brief descriptions of the utility each could also help serve as a quick reference guide to refer back to when reading this paper.

L94-144: This section seemed like it would be better served in the discussion in section 5. Many of the same concepts are addressed in more detail in Section 5, and it seems like a better flow and usage of space to simply incorporate this content into that section. This also allows for the novel part of this work to be established sooner than L145.

In general, it was my impression that the overall flow of the paper and ease of readership could be improved with some modest rearrangement. For example, the opening sentences of Sections 4.3 (L446-451) and 4.4 (L502-507) seem better placed in the introduction for context about ozone issues in the Southwestern US. It also seems like some sections of the main paper could have gone in the SI, while some sections of the SI should have gone in the main paper (specifically key points about uncertainties and the consideration of exceptional events).

L159: It would help to establish early on in this section that the study period is limited to summertime (i.e., May through September). This could be easily added to the first sentence of this paragraph.

L167: If I have this correct from this section (and Section S6), the authors choose not to select the “Exclude exceptional events data” option that excludes all flagged exceptional events regardless of concurrence. Thus, all reported exceptional events in the EPA AQS datasets regardless of whether EE are reported by the state agency that monitors ozone in a given area are included in all ODV calculations in this work. It seems appropriate to include exceptional events (even if each agency’s treatments of EE at individual sites may vary), since excluding them could lead to a significant difference in the annual ODV’s determined for each location.

However, the authors make a subsequent statement in Section S6 that says “the analysis of this paper is not significantly affected by consideration of exceptional events”, which leaves me still questioning: 1) whether the authors did or did not actually include exceptional events, and 2) how they know that it does or does not affect the results and conclusions. Can you comment on how different your results would be with and without selecting the “Exclude exceptional events data” option? An observation-based sensitivity test of this, for say a few selected urban and rural locations, could be an interesting addition to this work. This seems like an important consideration since exceptional events have been found to greatly impact the southwestern US (See David et al., 2021, https://iopscience.iop.org/article/10.1088/1748-9326/abe1f3) and ozone concentrations can be substantially elevated on smoke impacted days (e.g., Buysse et al., 2019, https://doi.org/10.1021/acs.est.9b05241).

L172: Can you add a reference for this sentence about sensitivity tests?

L182: The phrase “during the entire periods of ozone monitoring” is a bit confusing. This seems to imply year-round measurements, yet this analysis focuses on summertime.

L190: The authors mention that the single largest MDA8 O3 concentration is considered. Do you know if the day that this highest concentration was measured was impacted by an exceptional event? Would it change your results and conclusions if it is?

L209: The authors imply that some years of data are excluded for select sites because they “do not appear to accurately represent the overall urban area”. Is this fair? What criteria is used to justify excluding this data?

L241: This seems like a good place to follow up with a note about not being able to forecast ODVs.

L294: There are ways to empirically estimate smoke impacted days at individual monitoring sites (e.g., Brey & Fischer, 2016, https://pubs.acs.org/doi/10.1021/acs.est.5b05218; Buysse et al., 2019, https://doi.org/10.1021/acs.est.9b05241; McClure et al., 2018, https://doi.org/10.1016/j.atmosenv.2018.09.021).

L310: Could oil and gas be contributing to this term? Prior studies show substantial increases in oil and gas development activities in parts of Texas and that these development activities influence much of the southwestern US (e.g., Koss et al., 2017, https://amt.copernicus.org/articles/10/2941/2017/; Dix et al., 2019, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019GL085866; Pan et al., 2023, https://doi.org/10.1080/10962247.2023.2266696).

L320 (and 485): The range of RMSD seems rather large compared to the anthropogenic enhancements (6 ppb on average). How does this impact the findings? Can additional bounds be put on background and anthropogenic contributions by propagating the errors associated with the variability in fits?  

L337 – 346 and parts of Section 4.1 seem like they could be better placed in the methods section and relabeled as “Section 3.5: Justifying assumptions for determining long-term ODV at rural sites”.

L373: The reason for normalizing the data in Figure 1 was not clear. Also, does the selection of what the data are normalized to (i.e., 71.4 ppb) matter? A sensitivity to this selection seemed to be indicated in L388, but again it was not clear if and how this matters. Does this selection bias the background ODVs derived in other fits using a similar normalized a parameter, as mentioned on L394?

L 415: Change “caused” to “cause”

L558: Does anomalously high ODVs in the Upper Green River Basin during winter impact the results of this summertime study? Is it fair to exclude summertime data from this site in this analysis? It also seems worth mentioning that this area is also notoriously impacted by oil and gas production activities (e.g., Ghimire et al., 2023, https://doi.org/10.5194/acp-23-9413-2023).

L643: Could recirculation of air associated with the “Denver Cyclone” (Reddy and Pfister, 2016, https://doi.org/10.1002/2015JD023840) lead to an overestimate of background and/or anthropogenic ODVs in parts of Colorado?

L 761: Change “his” to “this”

L 765: It is worth noting that reducing O3 precursors in general has more benefits related to air quality and climate than what has been put forth here. For example, reducing emissions from sources that generate VOC and NOx have the added benefits of reducing other criteria air pollutants and air toxics, co-emitted greenhouse gases, and precursors to fine particle formation and SOA which contribute to visibility and haze. While this is implied in Zhang et al. 2019, it could be more explicit stated in the concluding remarks.

L783: For completeness, support for the third author’s effort seems like it should also be included in the acknowledgements and legal notice section.

Figure 5, 6, and 8. It isn’t always clear which fit is which in the figures. It could help to add a legend and/or refer to the fits by name (or their description) rather than by equation number. Also, on several occasions throughout Section 4.2 and Section S1, the authors mention “excellent fits” of the data to equation 3. However, there are cases in these figures where the fits don’t appear to represent the data that well (e.g., the green dashed line and the red curve on the left side of Figure 5 and in some of the panels in Figure 6). Can you clarify what the quantitative benchmark is for an “excellent fit”?

Further, the red fits for many of the geographical areas shown in Figures 5 and 6 and in the top panels of Figure 8 exhibit a consistent downward trend even though there is a clear uptick in ODVs in recent years. To clarify, is this because data from recent years is not included in the fit, or is this because the fit is poor for these data points. Since one novel aspect of this work is to update the trends in the southwestern US to recent years, it seems like these data points should be included in the fits. Maybe I missed something, but it seems unfair to exclude them from the fit just because they do not fit the functional form. Given the physical reasoning for this uptick (wildfires, maybe increased impacts from oil and gas), should additional terms in the polynomial be considered?

Figure 12. Are some of the bars out of order (e.g., forward hash for 2020 comes before backward hash for 2000 for Phoenix and Las Vegas)?

Section S3: Using your illustrative example in this section, can you estimate how much ozone precursors would have to decrease and over what time frame to eliminate the less than 6 ppbv on average anthropogenic enhancement above background? This could add context to the statement in the main paper at L719.

Section S6: It is hard to believe that there was only one ozone exceptional event concurrence when there are so many reports in the literature demonstrating exceptional events associated with wildfires in the western US over the time frame of this study. For context, I’m thinking of Figures 1 and 2 in David et al., 2021 (https://doi.org/10.1088/1748-9326/abe1f3), where they show wildland fires have a significant impact on the number of exceedance days for ozone in the western US between 2000 and 2017 (specifically EPA regions 8,9, and 10). Other reports also show a substantial number of high ozone days (as many as half) at selected sites in Colorado and California are associated with wildfire smoke impacted days in more recent years between 2016 and 2022 (some examples include, but are not limited to: https://doi.org/10.1029/2022JD037700; https://doi.org/10.1029/2021JD035221; https://doi.org/10.5194/acp-22-9681-2022).

In general, it would be worth re-checking the manuscript for definitions of abbreviations and acronyms. In some cases, definitions are overly abundant and in other cases they are lacking. For example, starting on Line 354, definitions for National Park (NP), National Monument (NM), and United States (US) could provide helpful added context for an international reader, yet Table S6 seems unnecessary.  

The authors have conducted a very similar analysis to their previous work and I do not see major differences between this work and the previous work.   I would also say this approach has not been well accepted by the ozone community.  While background ozone is indeed a large contribution to the health thresholds, I do not believe this paper adds significant new information to the scientific literature.  The authors have presented a simplistic model that has a number of problematic assumptions.   Mainly:

1. Their model assumes that high ozone days from stratospheric intrusions or other background sources are the same days as high ozone days due to local photochemical production. This is clearly not the case.
2. The model assumes that anthropogenic emissions (of NOx) are approaching zero. This assumes that our inventories are accurate and that sources such as agricultural emissions are going down at the same rate as other emission sources.  I do not believe this is  a good assumption.
3. The model treats agricultural emissions and wildfire contributions in a extremely simplistic way (e.g. equation 4).
4. The authors statements about the policy implications are not correct. States can work with the EPA to exclude high ozone data that is not under their control thru the exceptional events policy.    So for example, on line 30:   "Together, the US background plus wildfire contribution approach or exceed the US NAAQS for ozone of 70 ppb (implemented in 2015) and 75 ppb (implemented in 2008); consequently, in the southwestern US NAAQS achievement has been precluded."  This is really not true.   The EPA has a mechanism to exclude background O3.  Its the exceptional events rule.   Its not easy to get these excluded, but it is possible.  Many states have used this to exclude days that have either a strong strat influence at the surface or wildfire influence.

For these reasons I recommend the manuscript be rejected.