Meteorological Detrending of Ozone at Three Sites in the Dallas-Fort Worth Area: Application of KZ Filter Method

Round 1

Reviewer 1 Report

The authors have addressed all of my comments and I recommend the publication of the paper in the current form. 

Author Response

Response: We thank the reviewer for recommending the publication of our paper.

Reviewer 2 Report

The authors have added significant amounts of new text but have not addressed any comments on level of English (sentences such as 'the linearity analysis of MDA8 ozone long-term component after removal of solar radiation influence exhibited the increasing linear trend over the study period for all sites.' or 'Besides, the gap between the adjusted and unadjusted trends was significant when solar radiation was significantly higher or lower than the average value.' remain) or novelty - the manuscript is still basically a repeat of other published studies, with little explanation of why a near repeat study is merited.  The methods remain unclear, the text is poorly written, the novelty difficult to assess, and the additional text has added little - 'whereas the seasonal component explained the seasonal phenomena better.' does not add anything.  It adds little unfortunately to the published work in JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION is higher quality 

Author Response

Response: Further changes were made to the sentences pointed out. In addition, we feel that the novelty of this study is not primarily from the methodology followed. Rather, it is because of the study region identified, and the important conclusions drawn from the results for the Dallas Fort-Worth region. As this region is one of the fastest-growing population centers in the USA, with a corresponding increase in mobile sources of air pollution, we believe that the results obtained in this study will assist in the redirection of regulatory efforts in the appropriate manner. Our study identifies that the significant reduction in Point source NOx/VOC emissions have only been able to stabilize the ozone levels at the two stations (C17, C61), with an increase at the third station. Any new emission controls enforced on point sources could only be marginally beneficial, and efforts need to be focused on mobile sources in order to observe any discernable trends in long-term meteorologically adjusted ozone.   For English, we obtained the help from a native English speaker who has a master’s degree in Technology Project Management.  She reviewed our manuscript and fixed some minor grammatical errors.

Reviewer 3 Report

Overview

The manuscript deals with the meteorological detrending of ozone at three sites in the Dallas-Fort Worth Area, USA, through the application of KZ Filter method.

I think that the paper presents an interesting statistical analysis of ozone and meteorological data. In my opinion, the document deserves publication, after the recommendations listed below are taken into account, which are essentially related to the interpretations of the results and are based on recent scientific knowledge regarding the ozone phenomenology.

General comments

I think that a weak point of the paper is that the levels of measured surface ozone are mainly related (or attributed) exclusively to the photochemical ozone production over the examined area. On the other hand, the atmospheric levels and the variability of background ozone within the boundary layer and the free troposphere, which might be important during spring and summer months at mid-latitudes of the northern hemisphere,  are not discussed. So, I would suggest that the issue of tropospheric ozone and its subsequent influence to the boundary layer and surface ozone concentrations should be also taken into account in the interpretation of the results.

For this purpose, I think that it would be quite helpful for the authors to consider a relatively recent extended review paper on tropospheric ozone on global scale, including USA, where in its eastern and southern part high tropospheric ozone levels are observed (Gaudel et al, 2018, Elem Sci Anth, 6: 39. DOI: https://doi.org/10.1525/elementa.291 and also references therein). From my perspective and based on my expertise of analyzing ozone episodes in the Mediterranean region, I would just point out that the possibility of vertical ozone transport in the troposphere influencing the boundary layer and surface ozone values (a major factor in the Mediterranean, especially in its eastern part during summer but also in its western part during spring) is not mentioned in the manuscript and so all measured ozone is considered to be produced by local photochemistry from precursor pollutant emissions emitted in the examined Dallas-Fort Worth area only. This might not be always the case, especially during the May-September period when the tropospheric influence to the boundary layer gets its maximum height while at the same time the tropospheric ozone maxima are observed during that period of the year, with subsequent influence to the boundary layer and surface ozone values depending on the prevailing synoptic meteorological conditions.

In relation to the above, it has also to be kept in mind that the good correlation of ozone with temperature and solar radiation has also to do with similar variability characteristics of the mixing height of the boundary layer, which is maximized at mid-day hours when the tropospheric ozone reservoir influences the most the boundary layer and surface ozone concentrations.

I would also suggest for a more effective interpretation to take into account the hourly in-situ NO2 measurements from the same stations providing the hourly ozone measurements, if they are available, in order to be able to check if the ozone increase is directly related to the corresponding NO2 decrease of surface concentrations, as it has been observed in many areas around the world.  

The reason is that it is already known that for most urban stations the sum NO2+O3=Ox, called also “potential ozone” is constant (e.g. Kley et al., 1994; Kalabokas et al., 2000), so that any NO2 decrease is directly related with an exactly equivalent increase in ozone (in ppb) through the reaction of NO with ozone (NO titration), which is very rapid and actually “masks” the photochemical ozone production in urban/suburban air pollution monitoring stations.

Actually, I think that a similar statistical analysis of the Ox quantity would produce interesting results.

Overall and dealing with the interpretation of the trend of the long term-component of MDA8 ozone (Fig. 15) it should be also mentioned that it is in agreement with the generally upward trend of tropospheric ozone (influencing the surface ozone background) and the generally downward trend of NOx emissions in the eastern USA, leading to ozone destruction in the first place and subsequently to an increase of ozone concentration levels.    

Specific comments

Figs 5, 6, 7: The downward trend of the long term-component for temperature at station C402 (Fig. 7), in contrast to the same shape upward trend for stations C17 and C61 (Figs 5, 6) might imply measurement problems, as temperature variability is expected to be more uniform.

Technical comments

Page 3, line 113: “Theis”, please correct.

Page 23, line 505: “It suggested”, please correct to “It is suggested”.

Author Response

Please check our response in the attached document. Thanks.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

It is largely derivative of a previous study, and as such has no new material.  It does not merit publication unless greatly expanded.

The last MS I saw had none of the changes that I required included.

Author Response

Please check our response in the attached document.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors response and the associated text modification regarding the issues of the photochemical ozone potential in the examined area as well as the influence of the tropospheric ozone background and the NOx concentrations on the surface ozone levels (at the rural and urban stations respectively) is satisfactory, in principle.

So, I would suggest publication of the submitted revised manuscript.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Review of “Meteorological Detrending of Ozone at Three Sites in 3 the Dallas-Fort Worth Area: Application of KZ Filter 4 Method” by Upadhaya et al.

General summary

The authors used Kolmogorov-Zurbenko (KZ) filter for air quality observations at Keller, Arlington, and Red Bird monitoring stations in the DFW area from 2003 to 2017 to perform temporal separation, regression analysis, and meteorological detrending for maximum daily 8-hour average ozone, maximum temperature, daily averaged solar radiation and wind speed. The authors find that meteorologically detrended long-term ozone had an increasing trend with the largest increase at Red Bird followed by Keller and Arlington. The slightly increasing trend was associated with increasing mobile emissions in DFW area. The study is well designed and I recommend publication after minor revisions. Specifically, I could not find sufficient details to understand the method. For example, several symbols used in the equations are not defined at all and no explanation is given about how different equations connect to ozone and meteorological variables. My specific comments are listed below.

Specific comments:

Line 14: Revise to: DFW that experiences high temperature and intense solar radiation fall in….

Line 44: USEPA (2008) is a very old reference. I suggest citing a more recent EPA report of non-attainment areas because emissions controls are decreasing surface ozone continuously in the US.

Line 48: I think boundary layer height and precipitation should also be included in this list.

Lines 57-58: Can you provide some references for the use of KZ for trend analysis?

Line 109: What are i and j in equation (1)?

Line 121: Please clarify the time-scale for the short-term variations: Diurnal, few days, a week?

Line 131: Does (15,5) for (m,k) mean 15 days and 5 iterations.

Section 2.2: Could you provide more details about the four models? It is not clear how the influence of temperature, solar radiation, and wind speed was removed from the ozone observations.

Line 153: Why did you use natural log of MDA8 ozone?

Equations 11-15: Some of the variables used in the equations 11-15 are not explained. Specifically, I did not understand what ???(15,5) (? + ?), ?0(?), ???(?) mean, and how they relate to ozone and weather variables.

Line 167: Change Parson to Pearson.

Line 169: Up to this point, you have used (365,3) for (m,k) and suddenly you say you have used 15 iterations. Please clarify.

Line 192: Remove “in natural log scale”.

Line 222: Could you say something about why the long-term temperature shows a decreasing trend at C402.

Figure 7-9: Why solar radiation shows a long-term decreasing trend at C17 and C402 but not at C61.

Section 3.3 and Table 2: Can you say something about very poor r2 values for wind speed.

Line 415: Change decreased to decreasing.

Author Response

Please check our response in the attached document and refer the line numbers to the word document with tracking changes.

Author Response File: Author Response.pdf

Reviewer 2 Report

This manuscript aims to study the detrending of ozone with multiple meteorological variables. They have a solid dataset studied, including annual emissions, daily meteorological variables, and MDA8 for more than a decade in the DFW region. Despite following several studies on the same topic, this manuscript fails to state their method clearly and has little insights and relevance in air quality study and control. Some of the discussions are not supported by their analysis presented. Also, the level of their English writing is not acceptable for publication.

Major comments:

Even though this study focuses on “detrending,” it is unclear how they did the meteorological detrending exactly. What is the process of “meteorological adjustment”? Are these two concepts the same? Please explain with mathematical equations. Also, the natural log of the MDA8 is not reflected in the equations. Please revise it accordingly.  

There are a lot of problematic discussions on the trend in this study. The authors should deliberate on how they interpret the linear trends of the long-term component in log scale. If they are physically meaningful, are the trends statistically significant? Also, it’s not necessary to say a linear trend is steady, e.g., Line 197. See nice discussions of trends as in the following studies. The long-term components should not be called “trend” in the Fig 10-13 as well as in their discussion since they have clear cyclic signals embedded. “±” is mistaken to be “+” in a lot of cases. Please explain how the uncertainty in the latter part is derived in Table 3. The small R2 indicates the linear trend is possibly not significant. Please show which of them are statistically significant.

Wu, Z., Huang, N.E., Long, S.R. and Peng, C.K., 2007. On the trend, detrending, and variability of nonlinear and nonstationary time series. Proceedings of the National Academy of Sciences, 104(38), pp.14889-14894.

Luo, H., Astitha, M., Hogrefe, C., Mathur, R., and Rao, S. T.: Evaluating Trends and Seasonality in Modeled PM2.5 Concentrations Using Empirical Mode Decomposition, Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2019-1079, in review, 2020.

The paper contributed a lot of paragraphs to the discussion of the maximum and minimum of the ozone distribution or its components, which is not that important in an air quality study because of the stochastic nature of the atmosphere. The maximum and minimum of the log-transformed concentrations are even harder to understand. The ranges are also not all correct based on their figures, e.g., the ranges of the log-transformed MDA8.

A lot of details are missing from statements such as in Line 22 “The contribution of the long-term component to total variation was low”. Please check similar problems throughout the paper and fix all the language issues. The authors might want to seek some professional editors’ help.

Minor comments:

1. It seems all three authors work in the same institute, yet they represented themselves using repeated institute names.
2. Please make clear what meteorological variables are used. For example, Line 18 has “daily” solar radiation and wind speed, but not for maximum temperature. Add the “(MDA8)” after its full name. Line 19 should be “ozone” instead of “air quality” since they’ve only investigated ozone. How did the authors choose to study the “maximum” t but the “average” solar radiation and wind speed? Is it justified by previous studies? Or is it because of data availability? The mixing height is an important factor in the transport of ozone (paragraph starting from Line 65). Have the authors sought out for the mixing height data? Please elaborate on what is the “resultant” of wind speed.
3. Line 23: “on” the maximum t and solar radiation?
4. Line 57: “dilution” is mainly for the liquid phase and should be deleted.
5. Add citations in Line 58.
6. Line 62: delete “main.”
7. Typos: KZ as KG in Line 68, extra “.” in Line 93, extra “,” in Line 120…
8. Please include the location settings for the studied area for readers who might not be familiar with the region. A land use/topography map with the three sites marked will do. What “spatial distribution” is considered when selecting these sites?
9. The representation of parameters in the equations and the name of the sites should be uniform throughout the paper for straightforward interpretation. For example, use t in Eq 1 as in Eq 2, KZ (15,5) for (15,5), e for LT, and Keller for C17.
10. Does “baseline” mean the summation of long-term and seasonal components?
11. Line 178-179 is a repetition of the previous paragraph, which consists of only one sentence. Please revise.
12. Where are the figures for the decomposition of wind speed for all stations, as mentioned in Line 217? Moreover, the correspondence of the figures is not correct in this sentence.
13. “Narrow” is not an important feature of the long-term component.
14. What are the readers supposed to take home for numbers of the peaks in different components?
15. Please elaborate on how the share of variance of the three components in each meteorological parameter is related to their impact on ozone based on Table 1. Is it based on the linear models or derived by assuming the parameters are linearly correlated to ozone?
16. Please clean all the references to a uniformly accepted format of the journal. Wrong citation format in Line 280 and so on.
17. It’s better if the authors have shown the regression results in some form before the discussion starting Line 317. Is the difference in the impact of wind speed stem from the urban/rural difference? What could be the cause if not?
18. Arrange the colors of Fig 10-13 so that readers can relate them quickly to the left- or right-side y-axis.
19. Please elaborate on the reasoning of the trends in the paragraph starting from Line 396. Are they based on some previous studies? They should be moved together to Section 3.5. Overall, the analysis is weak to relate the ozone problem to specific sources in Section 3.5.
20. Line 404: “precursors” instead of “factors.”
21. Please be specific with “summer” in Line 415. Is it Mar-May? Is it the same as the “ozone season” in the manuscript? I don’t think the ozone season used in most studies only refers to summertime.
22. Line 419: Just say mobile source as there are many other sources in addition to those two shown in Fig 14 and 15.

Author Response

Please check our response in the attached document and refer the line numbers to the word document with tracking changes.

Author Response File: Author Response.pdf

Reviewer 3 Report

This study presents a statistical analysis of data from 2003-2017 collected at three Dallas-Fort Worth air quality monitoring sites. The analysis uses data on ozone and the co-located measurements of temperature, solar radiation and wind speed/direction.  It is shown to be possible to decompose the original timeseries of measurements into three components using a Kolmogoroz-Zurbenko filter applied to the data.  These components are associated with long term, seasonal and short-term variations. The correlation of these component signals with other measurements are then presented. The paper is very similar to an earlier study by one of the authors (RRK) "Meteorological detrending of long-term (2003-2017) ozone and precursor concentrations at three sites in the Houston Ship Channel Region" and also to "Long-term meteorologically independent trend analysis of ozone air quality at an urban site in the greater Houston area"  only one of which is referenced

The figures themselves are clear but the supporting text is not sufficiently clear for me to make a clear assessment of the validity of the methods or the success of the analysis.  Based on a reading of the Contributions section, it would appear that those involved with producing the final draft (HD and RRK) would need to perform more extensive revision of PU's original draft to bring it from first draft into publication form, and I suggest the MS is returned to the authors, and given this list of Contributions it should probably fall to the senior authors to improve the description of the methods and results.  

Terms need to be defined consistently and used consistently (e.g.

baseline - baseline what?

consistent use of long-term component, using a noun, rather than just 'long-term' which is an adjective and makes the sentences harder to read).

In fact, it might help if the revised MS could be closely proof-read by a native english speaker.

There are some very long paragraphs which span more than one subject, in contravention to conventional style guides, and which are very difficult to read (L65-95, L342-378).

The title is satisfactory but the abstract is not of publishable quality, being not particularly clear on the study's aims or very clearly written.  The first two sentences are particularly hard to understand.  This needs to be redrafted, again with senior authors paying more attention to make sure it's clear to a more general audience. E.g. 'MDA8 Ozone' is not defined. The definition of statistical significance used is not clear, with no p value given.

Section 1 summarizes the previous work acceptably, but needs to add a section on the motivation for this particular study in more depth than the three lines buried at the end of a paragraph that is describing previous work.  Minor comments:

L18 The names of chemicals are not capitalized unless they are the first word of a sentence, per IUPAC rules.  The authors do however revert to lower case ozone here and there.  Use lower case ozone throughout.

L43 In which country are these counties?

L45 not normal usage

L59 no reference

L68 what is a KG filter? not defined.

L74 not grammatically correct

L78 ditto

L81 can the authors try to improve the scientific level of description (relatively flat? relative to what?  what does flat mean - zero gradient?)

L89 'It showed that the long term' [behavior?] of what?

L91 VOCs not VOC's (possessional apostrophe not appropriate)

Section 2 reviews the methods used. This was unsatisfactory being rather vague, lacking a clear structure, with many inconsistencies and ambiguities.

e.g.

in  equation 1 j is undefined,

Equation 2, K is undefined

Equation 3, the description is redundant and this style is not used for the other equations so is incongruous. 

Equation 4 M is not defined. 

Equation 5 upper case W is used in the equation, script w in the text, m is not defined.

L131 sentence is is not grammatically correct.

L132 baseline is used extensively for the rest of the text, is used first here and is not defined. This hampers understanding of its subequent usage, and given its centrality this is should be remedied.

L139 is not a sentence, having no verb, i(t) and O(t) not consistent with text above. 

L147 'meteorological adjustment' not defined.

Section 2.4 should be folded into the rest of the section to explain better how the formal methods were used in the study.

Section 3 outlines the results. 

3.1 lists the outcome of an analysis to determine long-term seasonal and short-term components in the total signal.  Analyses are performed for ozone (fig 1, 2, 3), temperature (4, 5, 6) and solar radiation (7, 8,9).  Contrary to the text (L217) did not find data on wind speed. This is an interesting section, and LT or baseline trends are of interest.

L180 - no commercial instrument can measure ozone to 0.01 ppb precision, so the last decimal place is not meaningful.

L190 The use of log scale makes things difficult to understand and not justified in the text. 

L220 A reduced range in the data appears to be the desired outcome - why is this desirable?

The discussion of section 3.1 is to describe the data only, with no attempt at speculation/attribution of the reasons for these results, which should be addressed.

Section 3.2 performs variance analysis. This allows for the attribution of which parameters control variance and which control concentration. After a promising start, in which the rationale is explained (this was good to see), the section is overall not well written e.g
L269 the short component. Not clear.
L271 The short-term component was more dominant in wind speed, with 68 % - what does this mean? Do you mean that the short-term component explained more of the variance? I'd avoid more/less dominant.
L274 starts to talk about trends. Trend in what? Trend in contribution to the variance per year? Again, not clear.
L281 'Their study indicated that the seasonal component was the primary contributor to its total variance of MDA8 Ozone, TMAX, and solar radiation and that the short-term was the primary contributor for wind speed' To what does 'its' refer to here? 
L287-290 was very confusing as the authors do not clearly discuss how their results differ from the previous studies - 'our DFW analysis showed that the short-term component of wind speed was the primary contributor to the MDA8 Ozone [concentrations, I presume, and in this section they should be clearer when discussing variance and when discussing concentration]' indicates how WS controls O3, whereas L281 is about which is the dominant component overall to the variance.

3.3 Regression analysis.  This has a similarly incomplete feel.

L292 - The single and multiple linear regression analysis was performed among the original, baseline, and short-term components of Ozone and meteorological components for each site. What was regressed onto what?

L295 - 'were correlated' do you mean that you found that they were correlated or that you determined the correlation between these?
Overall this section, appears to be an assessment of regression of the various ozone against meteorological parameters but it's not clear. No plots are given.  The terminology is again quite variable - consistent terms should be used. The discussion turns on sentences like

L302 'The R2 value of the correlation between [raw?which?] solar radiation and raw Ozone was weaker than that with baseline data.'  It is completely impossible to understand what is going on (what do you mean by 'baseline data'?) without a working baseline definition, so this needs expanding, and which baseline data? Either use baseline as a noun, or if adjective say e.g. baseline ozone data.

3.4 I simply do not understand the goal of this section which is a shame as it's the most interesting, potentially. I'd be interested to know what the trend in the LT component is, and whether there is any trend in the ST component. Previous work such as Primarily this is because what 'meteorological adjustment' is is never defined (L107 mentions it first in reference to R code) and 'meteorologically independent' is never defined. In this section, unlike S3.2, the authors do not provide any introduction, just start by discussing another study.

L347 'A similar correlation was found with solar radiation as well. The linear trend of the adjusted Ozone was evaluated using linear regression and found that linearity was not applicable for model 1, i.e., Ozone without the influence of temperature at the Keller and Arlington sites due to very low R2' is not grammatical.  Again, this needs a re-write to explain the reasons for the analysis, the goals of the analysis, and a clear summary of what was found.

L358 the linearity of the sites... again, you don't mean this, you mean the linearity of the data

I couldn't get to the end of section 3.4 without getting more confused each time I read it.

3.5 presents an analysis of the emissions variation over the course of the study with various graphs to support this. There is no reference back to the sites, or discussion of how the ozone trend would be expected to reflect these.

In summary, I want to close what I am sure has been quite a bruising review by addressing some comments for the individual authors. I have pleasure in noting that this MS features some nice and interesting analysis with well-presented figures. My compliments to the author PU on the preparation of these and in bringing a difficult study together into a first draft, and, for their sake, I state that my comments in no way reflect on this author's contribution.

But, the submitted MS has three main problems: firstly, novelty - The strong similarity between this MS and "Meteorological detrending of long-term (2003-2017) ozone and precursor concentrations at three sites in the Houston Ship Channel Region" by Venkata S. V. Botlaguduru and RRK and also "Long-term meteorologically independent trend analysis of ozone air quality at an urban site in the greater Houston area" Venkata S.V. Botlaguduru, Raghava R. Kommalapati   & Ziaul Huque is also concerning as the present MS represents only a minor update to those other articles, choosing a second set of three sites for analysis.  Are they really so different? 

Secondly, significance of content: as written, the MS does not allow the reader the ability to judge the aims of the study, the appropriateness of the analysis and the success of the analysis in determining the underpinning reasons for the behavior. 

Lastly: quality of presentation.  For these reasons, major revision is necessary.  

Author Response

Please check our response in the attached document and refer the line numbers to the word document with tracking changes.

Author Response File: Author Response.pdf