Estimating background contributions and US anthropogenic enhancements to maximum ozone concentrations in the northern US

Parrish, David D.; Ennis, Christine A.

doi:https://doi.org/10.5194/acp-19-12587-2019

Articles | Volume 19, issue 19

https://doi.org/10.5194/acp-19-12587-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/acp-19-12587-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 19

Research article

|

09 Oct 2019

Research article |

| 09 Oct 2019

Estimating background contributions and US anthropogenic enhancements to maximum ozone concentrations in the northern US

David D. Parrish and Christine A. Ennis

Download

Final revised paper (published on 09 Oct 2019)
Supplement to the final revised paper
Preprint (discussion started on 13 Nov 2018)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'review of acp-2018-1174', Anonymous Referee #1, 09 Dec 2018
- AC1: 'Response to interactive comment by Anonymous Referee #1', David Parrish, 11 Dec 2018
- AC3: 'Response to interactive Comments by Anonymous Referees', David Parrish, 08 Jan 2019
RC2: 'review of acp-2018-1174', Anonymous Referee #2, 04 Jan 2019
- AC2: 'Response to interactive comment by Anonymous Referee #2 on “Estimating background contributions and U.S. anthropogenic enhancements to maximum ozone concentrations in the northern U.S.” by David D. Parrish', David Parrish, 07 Jan 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by David Parrish on behalf of the Authors (18 Jan 2019) Author's response Manuscript

ED: Reject (25 Feb 2019) by Qiang Zhang

ED: Referee Nomination & Report Request started (12 Apr 2019) by Ulrich Pöschl

RR by Anonymous Referee #3 (25 Apr 2019)

Suggestions for revision or reasons for rejection

Parrish uses a simple mathematical fit and an overly specific interpretation. The author has responded to previous comments by adding an acknowledgment of weaknesses of the technique and interpretation. The acknowledgment is not sufficient given the reliance of the manuscript on the specific interpretation.

In response to previous comments, the author asserts that there is likely no persistent emissions. The NEI shows a decay of emissions from 2002 to 2014 (a comparable period) with a lifetime of 18 years (see https://gispub.epa.gov/neireport/2014/) -- which is in reasonable agreement with the authors tau. However, the interpretation requires two more assumptions. The first assumption is that the trends are explainable by single exponential decay function. The second major assumption is that ozone trends will continue responding to future reductions as they have now.

First assumption: The interpretation by Parrish is based on exponential decay of local contributions, which are related to emissions. If the emissions reduction is not reasonably fit by a single exponential function, then it is over interpretation to conclude that y0 is background. The NEI trends data shows that some sectors are increasing while others are decreasing. Individual exponential fits would create tau values ranging from -50 to 22 years. For example, "Industrial and Other Process" category has gone up Nationally by 25% over the time period being analyzed (tau=-50years). If each sector is separately allowed to grow to 66 years (3 x tau in this manuscript), using the single exponential function would predict just 12% of the sum of individual functions. The "Industrial and Other Process" sector is likely linear not exponential; even holding it constant would mean that the single exponential function would predict just 34% of the future emissions. This highlights that the exponential decay is actually the sum of equations. Even the exponential decay of certain sectors could be due to an exponential decay in controlled processes and a constant or increasing component from uncontrolled processes.

Suggesting that an exponential decays that corresponds with specific sector controls can be used to characterize all sector contributions is over interpretation. There is evidence of this over-interpretation in the manuscript itself. Figure 7 is an excellent example of where the interpretation is obviously flawed. If this approach were robust, you could apply it to a time interval of observations and get the same result as if you applied it again after accumulating more data. If you had applied the model in 2003 (before the NOx SIP call took effect), the "Background ODV" appears as though it would have converged on 80-90 ppb. The exponential decay after 2003 highlights that the decay is in response to controls put in place over that time period. During that same time, motor vehicle emissions have also decreased. If controls targeted all sectors, the interpretation would be more reasonable. There is no evidence provided by the author in the manuscript.

Second assumption: If the sensitivity of ozone to NOx is not expected to remain constant, then this extrapolation is over-interpretation. As NOx-limitation becomes more sever, it is expected that the per unit NOx ozone production efficiency increases. This means that the remaining NOx will contribute more per unit than the NOx that has been controlled to date. Thus, any remaining NOx (the constant or increasing sectors) will contribute more than expected from present day inventory fraction.

Finally: The author uses uncertainty in USB to suggest that USB estimates from this technique are reasonably within uncertainty. Much of the uncertainty in USB comes from natural process like stratospheric intrusions and wildfires. These processes are not expected to make substantial contributions in New York on exceedance days that are often typical local ozone production.

The author has received much of this feedback at various meetings. Before resubmitting again, consider that the feedback is consistent and from many sources.

Hide

RR by Anonymous Referee #4 (08 May 2019)

RR by Anonymous Referee #5 (26 May 2019)

Suggestions for revision or reasons for rejection

This paper attempts to determine values for the US background Ozone Design Value (ODV) as well as its anthropogenic enhancement, by fitting observed ozone timeseries with a simple exponential decay function, representing a reduction of anthropogenic ozone enhancement due to reducing emissions, with an asymptotic approach to a "true" background ODV.

The rationale for doing this is that numerical models of ozone are currently not able to provide satisfactory estimates of these values, which are important for policymaking in the US. Having an alternative, measurement-based approach for estimating these values would be of great interest to policymakers. Exactly how well this can be done is an open scientific question.

Two anonymous referees have both recommended that this paper be rejected after its initial review phase. Both argue that the approach used by the author is overly simplistic, and lacks a physical basis. While I can sympathise to some extent with their criticisms of this paper, on balance I think that the underlying issue (the ozone problem in the US) is important enough that alternative approaches should be explored.

The author has generally done a reasonable job of explaining the limitations of his approach, and provided an extensive discussion of how the results can and should be interpreted. He has generally anticipated the criticisms of the reviewers, but this has not stopped them from making them anyway. I think the author could perhaps have done more to highlight the uncertainties and potential biases in his revised version, but he is also justified in sticking to his guns, as he has done, since his original submission already covers these issues well enough.

Given the importance of the underlying issue, I have no doubt that future work will continue to develop different approaches for observation-based estimation of background ODVs and their anthropogenic enhancement. The current paper provides a good basis on which such future work can build.

Hide

RR by Anonymous Referee #6 (30 May 2019)

Suggestions for revision or reasons for rejection

Review of “Estimating background contributions and US anthropogenic enhancements to maximum ozone concentrations in the northern US” by D. Parrish

This manuscript attempts to provide observationally-based estimates of background ozone and anthropogenic ozone production in the northern US using time series of ozone design values (ODVs) for the rural Northwest and the northeastern US. The analysis is based on multiple unsubstantiated and likely incorrect assumptions. I agree with the editor’s decision to not publish precisely because of the author’s argument that this is likely to be a high-impact paper given that it estimates background ozone to be significantly higher than previous analyses in most areas. While the author argues that the potential importance of the paper means it should be published so that it can become part of the open debate regarding background ozone, the assumptions that underlie the analysis are, in my opinion, so problematic that it should not be published until they can be justified via modeling or some other formal analysis. The discussion of the potential weaknesses of the analysis does not provide sufficient caution regarding the interpretation of the results and the uncertainties associated with the estimates of background ozone and anthropogenic enhancements to ODVs are misleadingly small given the number of assumptions and the disregard for temporal and spatial covariations among the ODVs. I believe that the manuscript has a likelihood of skewing the conversation about background ozone because it seems to provide highly significant, observation-based estimates of the background but in fact is based on an overly-simplistic view of ozone chemistry and makes so many “leaps of faith” regarding ozone’s response to changing emissions that its results are meaningless without significant work to lay the grounds for those assumptions.

The first assumption is that ODVs everywhere in the northern U.S. have decreased following an exponential curve in response to emissions controls. While this exponential decrease appears to be fairly well-defined for Los Angeles, it is so poorly defined for the regions in this analysis that a second and more objectionable assumption must be made that the timescale (tau) is the same for the rural northwest, the highly variable northeast, and Los Angeles. The need for this assumption arises from the fact that the ODVs have only been decreasing since 2000 in many of the time series analyzed and the decrease has been small relative to that in Los Angeles, so that the exponential decay is shallow and not well-defined. In fact, it is clear that in many cases a linear decrease would fit the ODV data as well if not better than an exponential. Assuming the exponential form and the same tau everywhere implies that 1) anthropogenic precursor emissions have decreased exponentially in response to emissions controls in all regions analyzed and that 2) “control strategies have produced approximately equal relative reductions in anthropogenic ozone enhancements through the country.” This is such an overly-simplified view of the highly heterogeneous system of emission control and of ozone’s highly nonlinear response to emissions changes that it is difficult to overstate how breathtaking it is. There are two throwaway statements in the paper regarding the possible influence of “volatile compounds”, but the role of VOCs in ozone chemistry, the fact that ozone can be either VOC- or NOx-limited and that the response to emissions controls can be opposite for two nearby locations depending on the NOx / VOC mix are completely ignored. So too are the differences in emissions controls that have been applied in Los Angeles and those that have been applied in the Northeast (there is a brief discussion of this issue on page 18, lines 20-26, but the author appears to believe that the question of whether precursor emission reductions have been similar between Los Angeles and the Northeast is unknowable, when in fact a simple analysis of emissions inventories would provide a great deal of relevant information) and the transition of substantial regions of the Northeast from VOC-limited to NOx-limited during the period from 2000 to the present. Furthermore, by the author’s own description, the equation used to describe the decrease in ODVs “assumes that all sectors of anthropogenic US ozone precursor emissions have been reduced by emission controls at approximately the same rate. However, in some respects this is a poor approximation in that some emission sectors have received lesser efforts than others. Any emissions that have not been reduced would tend to lead to an overestimate of US background ODV.” In fact, we know with absolute certainty that all precursor emissions have not been reduced at the same rate. Given this and the fact that any increase in tau also leads to a lower estimate of background ozone, the confidence limits provided for the background ozone estimates are ridiculously small and the estimates are not meaningful.

Another objectionable assumption is that background ozone has remained constant over the entire observational record and that all of the processes that might change background ozone act merely to produce variability in ODVs rather than trends. In fact, we have good reason to believe that there should be trends in background ozone associated with wildfires and international emissions, and that these should be highly regionally variable.

The author argues that two additional analyses demonstrate the robustness of the results despite the assumptions inherent in the equation describing ODV changes. The first such analysis is that shown in Figure 10, where ODVs are reconstructed from Equation 1 using the tau value from Los Angeles and a single, regionally-averaged value of background ozone (yo). The only thing proven by this analysis, in which there is only one free parameter remaining (A), is that a very shallow exponential curve reasonably fits ~15 years of data for the Northeastern US. Given that there has been a relatively steady decline in ozone in the Northeastern US since 2000, this is a given once you fix a value for tau and yo. The fit improves if you consider only the maximum ODVs in each state each year, but no justification is provided as to why this should be the case. The second analysis that is meant to demonstrate the robustness of the results is, to my mind, a direct argument against their robustness. This analysis also relies on unsubstantiated assumptions that 1) all ODV time series follow the same functional form (it is not clear why this should be the case given the heterogeneity of ozone precursor emissions in the region and the nonlinearity of ozone chemistry) and 2) all of the time series will approach a common US background ODV (we have no a priori expectation of this given expected gradients in influence from international transport from Canada, variability in natural emissions, etc). Furthermore, one of four time series that differs substantially from all of the other time series in the Northeast (i.e. the New York City urban area, which has a higher background ozone value that most of the rest of the region in Table 2) is used as the reference. The regression analysis produces background ozone estimates ranging from 25 to 62 ppb, but a single, averaged value is calculated and compared to the result from the exponential fit analysis. Yet if one were to assume that the variability in the derived background ODV is real (and no justification is provided as to why it should not be) then the derivation of A values (or the taus – it is impossible to say which) in the exponential fits would be completely different. The uncertainties of 3.9 ppb and 1.7 ppb on the background ozone estimates do not consider regional and temporal coherence of the ODVs nor the uncertainties associated with the assumptions underlying each analysis and are likely grossly underestimated. While the author is correct that there is not a good way of accounting for these additional uncertainties, the low values provide a misleadingly precise estimate of background ozone.

In Section 4.3, the author treats his “observationally-derived estimates” of background ozone as the gold standard against which model-derived values should be evaluated. This discussion illustrates the danger of this paper. There is a natural inclination to treat observationally-based estimates as “truth”, when in fact the unsubstantiated assumptions underlying this analysis and the gross underestimate of uncertainties are not likely to be clear to most casual readers. Until additional analysis has been done to provide a foundation for the assumptions on which the analysis is based, this work should not be published.

Hide

ED: Reconsider after major revisions (03 Jun 2019) by Ulrich Pöschl

AR by David Parrish on behalf of the Authors (17 Jul 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (25 Jul 2019) by Ulrich Pöschl

RR by Anonymous Referee #3 (02 Aug 2019)

Suggestions for revision or reasons for rejection

First, I would like to thank the author for taking the time to consider many different pieces of feedback.

This reviewer cannot overlook the assumptions necessary to support the implications and conclusions. In many ways, the additional mathematical and uncertainty estimates obfuscates the real problem. In my estimation, the value of the author's work is the predictive power and not the physical interpretation. Yet, the paper's implications and conclusions focus on the interpretation. The authors application and interpretation is not unlike Hubbert's peak oil curve, which likewise sparked much controversy around the pop cultural interpretation.

The similarity of your fit to Hubbert's peak oil curve has both strengths and weaknesses. Even when fitting observations, both curves equally support two competing interpretations. The first interpretation is, as the author focuses, y0 occurs when US contribution is zero. By analogy, Hubbert's peak oil curve can be interpreted as approaching zero when there is no oil is left in the US (only other countries). The second interpretation is that y0 is when US regulations have been applied to the sectors where and the extent to which it will be applied -- leaving some residual (USRES). By analogy, Hubbert's peak oil curve can be interpreted as no *practically recoverable* oil is left in the US. In the second interpretation, y0 is really the sum of y_{USB} + y_{USRES}. The strength of such the simple mathematical fits is their predictive power, and not the conjecture about the composition of the intercept.

There is value in the predictive power of the exponential model proposed. The exponential model does fit recent data and one may conjecture that it will fit near-term future reductions. As the author points out, the curve may represent diminishing reductions of US contribution. The predictive power is independent of the physical interpretation of the composition of y0. Parrish and Ennis project an attainment date of 2021 and that prediction is not predicated upon the physical interpretation y0, which does seem valuable -- as was their prediction for SoCal.

Regarding your response to my critique of fitting to data for Massachusetts (old Fig 7, new Fig 6). If in 2003 you been applying your exponential model, you would have only had the historical record to that date. My visual interpretation of the historical record between 1984-2003 is that your curve would have had some limited explanatory power. If I am not much mistaken, your y0 would have been substantially higher (80 ppb?) because the record as a local plateau between 1995 and 2003. This illustrates clearly that the physical interpretation of y0 predicted from that segment (1984-2003) would have been better interpreted as y_{USB} + y_{USRES}. For the 2003 inflection point, it is easy to guess that is caused by the NOx SIP call. How can we justify assuming there are not other controllable sectors that would produce a similar inflection in the future?

The additional sections with all the analysis suggest that this fundamental problem of interpretation can be solved by more math. The largest uncertainty in y_{USB} is how well it is approximated by y_0. If y_{USRES} is large, then the uncertainty is swamped by the interpretation. If y_{USRES} is small, then the interpretation is reasonable. y_{USRES} is some combination of the technologically challenging reductions and, as illustrated by the previous paragraph, the sectors that may not yet have been the focus of controls. No effort is made to consider or estimate the magnitude of US emission sectors that are not decreasing. In response to a previous revision, this point was made by highlighting specific sectors that are not decreasing. No amount of regression uncertainty analysis will address the uncertainty associated with a physical interpretation.

In my view, this manuscript has valuable analysis mixed with what seems like an arbitrary interpretation. The author has made token statements that allow for an alternative interpretation, but the conclusions and implications are so intertwined with the interpretation and those have not changed. This manuscript would be far easy to accept if the conclusions and implications it promotes/asserts were not dependent on an arbitrary physical interpretation of a parameter with an unknowable composition. Given that there are two equally likely physical interpretations (y0 = y_{USB} or y0 = y_{USB} + y_{USRES}), this reviewer cannot support a manuscript whose conclusions and implications are based on a seemingly arbitrary choice between the two.

Hide

RR by Anonymous Referee #5 (06 Aug 2019)

RR by Anonymous Referee #4 (06 Aug 2019)

ED: Publish subject to minor revisions (review by editor) (20 Aug 2019) by Ulrich Pöschl

AR by David Parrish on behalf of the Authors (29 Aug 2019) Author's response Manuscript

ED: Publish as is (30 Aug 2019) by Ulrich Pöschl

AR by David Parrish on behalf of the Authors (05 Sep 2019)

Download

Article (9753 KB)
Full-text XML

Short summary

Background ozone transported into cities contributes greatly to urban concentrations. Based on projections of past trends, the largest ozone concentrations on which the 70 ppb National Ambient Air Quality Standard is based will reach that standard by ∼ 2021 in the New York City area but much later (∼ 2050) in the Los Angeles region. The much smaller background contribution in New York City (45.8 ± 1.7 ppb) than in Los Angeles (62.0 ± 2.0 ppb) is the primary reason for this large difference.