review of manuscript acp-2021-786

Title: Changes of Anthropogenic Precursor Emissions Drive Shifts of Ozone Seasonal Cycle throughout Northern Midlatitude Troposphere

Authors: Henry Bowman et al.

General comments:

The manuscript presents a comprehensive analysis of the shifts in timing and strength of the seasonal ozone cycles at remote locations in the Northern Hemisphere midlatitudes. The paper is within the scope of ‘Atmospheric Chemistry and Physics’ but – to my mind – requires some revisions prior to its acceptance for ACP. See my specific comments below.

Abstract, line 19: use different word from "amplitude" as it here represents the double of the amplitude as (mathematically) defined in equation 3 and shown in Fig. 5.

Line 103: Reider et al. needs to read Rieder et al., here and in the references.

Lines 146 – 154: as seen later in the results, the model output varies considerably among the different models. However, this paragraph provides hardly any information about the different models. There is reference made to the table in the supplementary material that provides some details but some short summary of the differences of the models might help to interpret differences in the model performance below. How about the spatial resolution of the models, number of vertical layers, different parametrization, different meteorological input etc.? Reference to the Earth System Grid Federation should be added in the acknowledgements. Are there no model simulations after 2014? Since the most dynamic part is the last 30-40 years, an extension to the most recent past would be appreciated.

Table 1: I do not believe that O3 sondes are launched at Jungfraujoch. Please correct.

Caption, figure 2: add "simulated" that it reads "… time series of simulated monthly mean ozone concentrations …"

Caption, figure 3: write "Harmonic analysis of simulated preindustrial seasonal cycles …"

Line 366 ff.: the authors report "… quantitative differences in simulations among models …" but the authors do not provide any further ideas about the causes of the differences. An elaboration on this topic is highly requested.   

Figure 4 (legend) is somehow misleading as it makes the reader believing that observations above Jungfraujoch (FT 5-6 km) are available. Only the last sentence of the caption explains that the black curve is based on the mean of the European sonde measurements (which are made in Uccle, Hohenpeissenberg and Payerne (see lines 274-275)). Why do you use the model output above the Alps (which might be more difficult to interpret due to the underlying topography and potentially associated issues in the models) when the three observational sites are largely located North of the Alps. Why not choose another grid box of the model that better represents the centre among the three sonde sites (which will be above a less complex terrain)? Again, what's the spatial resolution of the global models? How does the model topography (of the Alps) influence the model output?

Line 454: "We average over all three European locations …"; write "… all three European sonde locations …" if this is meant here (i.e. Uccle, Hohenpeissenberg, Payerne).

Figure 8: does it need panels c and d? To my understanding, the widths of the Gaussian fit are hard to interpret. The authors may consider removing c and d or adding some interpretation of the findings.

Figure 8: there seems to be some kind of systematic pattern in the simulations results for the six models seen in panels a, c, and e (increase from left to right). Maybe this is by chance as the models are sorted alphabetically, but again, is it possible to explain some of the behavior of the different models? The width of the Gaussian fit seems to correlate with the maximum amplitude increase.

Table 3, caption: write "... second two rows give parameters for fits to observed and model simulated seasonal ozone cycles" to make clear that the lower two lines refer to ozone while the upper two lines refer to ozone precursors. Why are the numbers for the simulations averaged over all six sites (which ones? I am confused) while the observations are averaged over Hohenpeissenberg and the European Alpine sites only?

Discussions and conclusions are overall rather descriptive and do not provide many explanations. E.g. paragraph on lines 524 – 547, which starts "The model simulations exhibit large variability, both among models and locations …", doesn't hardly provide any conclusions for the reasons for the disagreements and the differences. Only the very last sentence of the paragraph briefly attempts to do so.

Lines 553-557: is this statement really needed?

I am wondering if some of the speculation (like on lines 571 ff. "… we hypothesize that the changing ozone precursor emissions is the cause of the shifts in the seasonal ozone cycle throughout northern midlatitudes." could be confirmed by looking into other world regions where the O3 precursors behave differently over time. The model output should be available globally, and there are some long-term ground-based O3 observations in pristine environments also available in tropical regions and the Southern Hemisphere. Also the SHADOZ program could provide a valuable dataset of O3 sonde observations for comparison with the model output.

Lines 606 ff. ("measurement records […] are so short that the precision […] of the seasonal cycle shift […] is limited"): the observation records used here stop in 2016 and 2017, respectively. Newer data are available in the public repositories. Why no newer data is used?

Lines 614-616: remove this statement as "… the uncertainty […] prevents a firm conclusion."

Lines 622-624: remove statement on determining the impact of the changing climate? This wasn't discussed/mentioned at all in the manuscript.

Line 625: the data availability statement is insufficient. The statement refers to "public archives referenced in the paper" but neither table 1 nor chapter 2.4 provide any indication how the data were accessed and where interested reader may find the data.

Bowman et al. present a comprehensive analysis of changes in the ozone seasonal cycle across the northern hemisphere. The manuscript is generally well prepared and fits the scope of ACP. I suggest some revisions and adjustments to the presentation of the methods and results before consideration for publication in ACP by the Editor. Specific comments are given below:

General Comments:

Section 1: Several studies (e.g. Clifton et al., 2014; Schnell et al., 2016; Jaidan et al. 2018; Rieder et al., 2018) have reported on potential future changes in the ozone seasonal cycle due to changes in emissions/climate. Some reference to these findings should be made either in the introduction or in the conclusion section, especially given the concluding remarks in L620ff.

Section 2.1: This section is important to the manuscript as it details the statistical treatment of the data sets. I think however, it could be presented in more concise form, e.g. keep the fundamental equations and narrative in the main manuscript and move other parts to the supplement.

Section 2.4: Why do the authors limit the observational data to  2016/17? or if already shorter than available data is used do not  align it to the same time frame of the historical ESM simulations?

Section 3.2: provides a comparison of the preindustrial seasonal cycle as derived from different ESMs, the authors however do not provide any insights (or hypotheses) regarding the drivers of differences among ESMS and which estimates might be more reliable (e.g. overall amplitude differences of factor ~3, the differences in North American and European FT sites)

Section 3.3: I am not convinced that fitting Eq (5) to the comparatively short observational time series adds all that much besides maybe for the European Alpine and HP data sets.

Section 3.4: Figure 8 and the text surrounding need some clarifications/modifications. It is not entirely clear what we learn from the width of the Gaussian in panels (c) and (d).

Discussion and Conclusions:

i) I would like the authors to address the spread among models a bit more in this section. While overall the models agree on the timing of the shifts in the amplitude and phase of the seasonal cycle the authors highlight substantial variability across models and sampled locations which remains widely unaddressed.

ii) The authors hypothesize that the main driver of changes in the seasonal cycle are changes in anthropogenic precursor emissions. Testing this hypothesis for other regions, with different emission patterns, in the NH (e.g. Asia) or SH would make an interesting addition to the manuscript.

Minor Comments:

Table 1: Jungfraujoch should be specified as surface site, as far as I am aware ozone sondes are in Switzerland only launched in Payerne

Table 1: provide Trinidad Head and 200km West of Trinidad Head as last entries to separate observations from the additional model grid cell analyzed.

L104: typo Reider àRieder  

L 273: typo Zugspitse à Zugspitze

L 278: I assume this refers to Parrish et al. (2014, 2020), but it is not entirely clear which “studies of western European baseline ozone concentrations” are refereed to.

Table 2: does not add to much in the main body of the manuscript, add to supplemental material

Figures: throughout: when showing model data add “simulated”

Figures: not all homogeneously labeled, compare e.g. order Fig. 6 and Fig. 8

Fig. 2: legend should show “red” not “yellow” color.

Figs. 5 and 6: I suggest moving the legends indicating model color coding outside the plotting frame for readability, do so also in the supplement

Figure 8: align sub panels, panels (f) and (e) are shifted compared to (a)-(d).  

Figure 9: maybe color code model results based on colors chosen in Figs. 4,5,6

References:

Clifton et al., 2014 https://doi.org/10.1002/2014gl061378

Jaidan et al., 2018 https://doi.org/10.5194/acp-18-9351-2018

Rieder et al., 2018 https://doi.org/10.1016/j.atmosenv.2018.07.042

Schnell et al., 2016 https://doi.org/10.1002/2016GL068060