the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 Jan 2024
Investigating Ground-Level Ozone Pollution in Semi-Arid and Arid Regions of Arizona Using WRF-Chem v4.4 Modeling
Abstract. Ground-level ozone (O3) pollution is a persistent environmental concern, even in regions that have made efforts to reduce emissions. This study focuses on the state of Arizona, which has experienced elevated O3 concentrations over past decades containing two non-attainment areas designated by the U.S. Environmental Protection Agency. Using the Weather Research and Forecasting with Chemistry (WRF-Chem) model, we examine O3 levels in the semi-arid and arid regions of Arizona. Our analysis focuses on the month of June between 2017 and 2021, a period characterized by high O3 levels before the onset of the North American Monsoon (NAM). Our evaluation of the WRF-Chem model against surface Air Quality System (AQS) observations reveals that the model adeptly captures the diurnal variation of hourly O3 levels and the episodes of O3 exceedance through the maximum daily 8-hour average (MDA8) O3 concentrations. However, the model tends to overestimate surface NO2 concentrations, particularly during nighttime hours. Among the three cities studied, Phoenix (PHX) and Tucson (TUS) exhibit a negative bias in both hourly and MDA8 O3 levels, while Yuma demonstrates a relatively larger positive bias. The simulated mean hourly and MDA8 O3 concentrations in Phoenix are 44.6 and 64.7 parts per billion (ppb), respectively, compared to observed values of 47.5 and 65.7 ppb, resulting in mean negative biases of -2.9 ppb and -1.0 ppb, respectively.
Furthermore, the analysis of the simulated ratio of formaldehyde (HCHO) to NO2 (HCHO/NO2; FNR), reveals interesting insights of the sensitivity of O3 to its precursors. In Phoenix, the FNR varies by a VOC (volatile organic compound)-limited regime in the most populated areas and a transition between VOC-limited and NOx-limited regimes throughout the metro area with an average FNR of 1.15. In conclusion, this study sheds light on the persistent challenge of ground-level O3 pollution in semi-arid and arid regions, using the state of Arizona as a case study.
- Preprint
(13489 KB) - Metadata XML
-
Supplement
(5747 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on gmd-2023-234', Anonymous Referee #1, 19 Feb 2024
This is a review for “Investigating Ground-Level Ozone Pollution in Semi-Arid and Arid Regions of Arizona Using WRF-Chem v4.4 Modeling.” The co-authors use WRF-Chem to understand and assess ozone concentrations and production chemistry in three cities in Arizona USA. The paper is well-written, the methods are clearly described, and the figures are easy to understand. The analysis includes extensive model evaluation using observations, sondes, and model reanalysis data, and an interesting assessment of FNRs in the region. I believe this paper to be suitable for publication following a few minor revisions as listed below following a brief list of my biggest lingering questions:
- Was windblown dust included in the emissions? Would this have an impact on ozone photolysis in AZ?
- What justification do you have for using the FNR values for theses AZ cities?
- What do your findings in the paragraph at lines 586-94 suggest for regulatory actions to reduce ozone concentrations? E.g. how should the cities determine emissions reductions strategies and under what ozone conditions?
Minor suggestions:
- Figure 1 caption lists two figure 1f, second should be figure 1h
- What was the CMAQ reanalysis data used for? “For evaluation” Please include a brief description of what CMAQ reanalysis data is used for in the methods.
- Were fire emissions data inputs year-specific?
- Line 445: I understand this is an average, but there can’t be partial exceedance days. I recommend rounding down.
- There is some duplicate information in the paragraphs starting at line 553 and line 565. For instance, lines 565-6 are a repeat of information in lines 558-60, and the description of the fire hotspots are repeated in the second paragraph. Please clean this up so as not to repeat your findings unnecessarily.
Citation: https://doi.org/10.5194/gmd-2023-234-RC1 -
RC2: 'Comment on gmd-2023-234', Anonymous Referee #2, 22 Feb 2024
The authors conducted a multi-year air quality simulation using WRF-Chem to describe and evaluate O3 and its influencing factors in three cities in Arizona, USA. They also analyzed ozone sources and chemistry. The paper is well-written and organized, with a clear introduction to the method, comprehensive use of observational and reanalysis data for evaluation, and thorough presentation of results. This study contributes to improving understanding of O3 formation, transport, and mitigation in arid and semi-arid regions. I am supportive of the theme of this manuscript and believe it can be published after minor revisions.
- The 4 km model-ready emissions from NEI2017 are used while the resolutions of the simulation are 3 km and 9 km. Please clarify how you made the emission regridding. Do you aggregate the elevated point source emissions to the surface layer or allocate them to the model 34 vertical layers?
- It is advisable to introduce the data, data processing, and quality control before the weather/air quality description of the study area. How did you calculate the monthly wind directory? Average them simply may result in biases.
Minor:
- T, RH: define them upon their first appearance.
- L317: reference for the BVOC contribution.
- L430: The PBLH in PHX shows the biggest negative bias (-509.7 m) while its O3 is underestimated. Are there any other reasons?
- Line 456: should be Figure 10
- L556~580: reorganize.
- L638: O3, subscript
Citation: https://doi.org/10.5194/gmd-2023-234-RC2 -
RC3: 'Comment on gmd-2023-234', Anonymous Referee #3, 24 Feb 2024
This work evaluates surface ozone simulated using WRF-Chem against several observational networks for three cities in Arizona. The analysis and paper itself are comprehensive. I recommend publication after several minor revisions specified below:
Page 7 line 184: Why use a 1 deg meteorological model for initial and boundary conditions when there are other higher resolution products for this too? Why not use a finer horizontal resolution meteorological model like NAM or HRRR? How might this impact your meteorological analysis later on? This would in particular impact wind direction and speed. How did the wind speed and wind direction from AQS compare with the model results? You show the observations in Figure 1, but do not show how this compares with the model results.
Page 8 line 213: Do you mean 30 days in June? If not, why not run for the entire June time period?
Page 9 line 217: For the selection of AQS sites can you describe this more? Did you only select sites with a certain amount of data available for the entire time window of 2017 – 2021? What latitude / longitude constraints did you use to define the region of a city? The AQS sites include metadata to characterize the site measurement scale, which gives you an idea of how representative that site is of a broader area. Some AQS sites are in locations that are less applicable to be simulated by a 3 km horizontal resolution model. Did you consider this at all in your choice of selected sites? And / or does this explain some of the biases you see in the results section?
Page 11 section 3.1: I’m assuming from Section 2.3, that this is an average of many sites in a given city. Can you provide a table in the supplement for which sites you included in this analysis? Can you do this for other measurements if needed too? Is there variability in hourly ozone and MDA8 ozone across the different sites in a given city and how well does the model represent this variability?
Table 1 and 2: Are these averaged for June 2017 – 2021 or just for a specific year.
Figure 5: Especially for isoprene and HCHO there do seem to be some differences between the model and observations. Can you add the median or mean bias to the plot to show this better?
Figure 5: For Figure 5 and in plots/analysis later on, for NO2 is this a direct comparison with NO2 in the model to NO2 in the observations or do you apply any correction for interferences that the AQS sites can have for NO2 (e.g., Dunlea et al., 2007, https://doi.org/10.5194/acp-7-2691-2007). If no correction was applied, do you think this could explain some of the biases you see?
Figure 6: Why not include all the different monitoring sites for these cities and zoom in more to show the regional variation across a city here rather than an average for all sites in a city? If this is an average, it would be good to be clearer here and in the text.
Figure 10: Including the WRF-Chem and reanalysis model simulated number of exceedance days corresponding with the AQS sites for direct comparison against the observations would be extremely useful for understanding the forecast skill of the WRF-chem model. It’s hard to discern this from Figure 10. Can this be added to the analysis in some way? Your conclusions and Section 3.2 state that WRF-Chem agrees quite well with the observational data for number of exceedance days, but statistics or a bias plot would support this conclusion better.
Figure S5: Can you double check the units for ozone along the trajectory. The color bar goes from 0, 2, 4, 6, 8, 1? And the units are in ppm? This seems too high?
Page 26 line 520 and also in your conclusion on page 31 line 618: Can you explain further why these results suggest that ozone exceedances on this day were caused by inter-state transport rather than local production? How far back in time do these HYSPLIT trajectories go? You also state that the PBLH was lower and temperature was higher which could cause higher local ozone formation too? Looking at the wind direction and speed from the model and observations seems important for evaluating this event. Have you looked at these metrics too? Adding this as a conclusion on page 31 and line 618 “that Arizona is substantially affected by inter-state transport of O3 from California” seems speculative. You need more analysis to state this, so I would strongly recommend rewording this sentence or doing more analysis.
Page 28 line 550: Can you provide the references for why you use these specific characteristics of FNR to describe the differences in the regimes? Additionally, there are several studies (e.g., Schroeder et al., 2017, https://doi.org/10.1002/2017JD026781) that demonstrate the uncertainties of using the FNR approach to approximate ozone production. Can you provide more context on the uncertainty of this approach? Are there other studies that have investigated the ozone production in your cities of interest in the recent past that you can also refer to? Do they agree with your conclusions using this FNR approach?
Figure 14: Can you add lines to represent the different regime changes as specified in line 550 that you are assuming in this work? This would be useful further validation of these values.
Page 31 line 624 - 630: Can you be clearer in this paragraph what your main conclusion is with regard to the ozone production sensitivity including the approach used to determine it and the uncertainties of this approach? From a policy perspective you state both your correlation approach strongly suggests a VOC-limited regime while also saying the FNR analysis suggests VOC-limited or transitional regime. Which regime does your analysis support and what is the uncertainty on it? It is important to be clear what your analysis is suggesting and the uncertainty on your analysis for understanding the policy ramifications of your work.
Citation: https://doi.org/10.5194/gmd-2023-234-RC3 - AC1: 'Authors' response to all comments on gmd-2023-234', Yafang Guo, 02 Apr 2024
Status: closed
-
RC1: 'Comment on gmd-2023-234', Anonymous Referee #1, 19 Feb 2024
This is a review for “Investigating Ground-Level Ozone Pollution in Semi-Arid and Arid Regions of Arizona Using WRF-Chem v4.4 Modeling.” The co-authors use WRF-Chem to understand and assess ozone concentrations and production chemistry in three cities in Arizona USA. The paper is well-written, the methods are clearly described, and the figures are easy to understand. The analysis includes extensive model evaluation using observations, sondes, and model reanalysis data, and an interesting assessment of FNRs in the region. I believe this paper to be suitable for publication following a few minor revisions as listed below following a brief list of my biggest lingering questions:
- Was windblown dust included in the emissions? Would this have an impact on ozone photolysis in AZ?
- What justification do you have for using the FNR values for theses AZ cities?
- What do your findings in the paragraph at lines 586-94 suggest for regulatory actions to reduce ozone concentrations? E.g. how should the cities determine emissions reductions strategies and under what ozone conditions?
Minor suggestions:
- Figure 1 caption lists two figure 1f, second should be figure 1h
- What was the CMAQ reanalysis data used for? “For evaluation” Please include a brief description of what CMAQ reanalysis data is used for in the methods.
- Were fire emissions data inputs year-specific?
- Line 445: I understand this is an average, but there can’t be partial exceedance days. I recommend rounding down.
- There is some duplicate information in the paragraphs starting at line 553 and line 565. For instance, lines 565-6 are a repeat of information in lines 558-60, and the description of the fire hotspots are repeated in the second paragraph. Please clean this up so as not to repeat your findings unnecessarily.
Citation: https://doi.org/10.5194/gmd-2023-234-RC1 -
RC2: 'Comment on gmd-2023-234', Anonymous Referee #2, 22 Feb 2024
The authors conducted a multi-year air quality simulation using WRF-Chem to describe and evaluate O3 and its influencing factors in three cities in Arizona, USA. They also analyzed ozone sources and chemistry. The paper is well-written and organized, with a clear introduction to the method, comprehensive use of observational and reanalysis data for evaluation, and thorough presentation of results. This study contributes to improving understanding of O3 formation, transport, and mitigation in arid and semi-arid regions. I am supportive of the theme of this manuscript and believe it can be published after minor revisions.
- The 4 km model-ready emissions from NEI2017 are used while the resolutions of the simulation are 3 km and 9 km. Please clarify how you made the emission regridding. Do you aggregate the elevated point source emissions to the surface layer or allocate them to the model 34 vertical layers?
- It is advisable to introduce the data, data processing, and quality control before the weather/air quality description of the study area. How did you calculate the monthly wind directory? Average them simply may result in biases.
Minor:
- T, RH: define them upon their first appearance.
- L317: reference for the BVOC contribution.
- L430: The PBLH in PHX shows the biggest negative bias (-509.7 m) while its O3 is underestimated. Are there any other reasons?
- Line 456: should be Figure 10
- L556~580: reorganize.
- L638: O3, subscript
Citation: https://doi.org/10.5194/gmd-2023-234-RC2 -
RC3: 'Comment on gmd-2023-234', Anonymous Referee #3, 24 Feb 2024
This work evaluates surface ozone simulated using WRF-Chem against several observational networks for three cities in Arizona. The analysis and paper itself are comprehensive. I recommend publication after several minor revisions specified below:
Page 7 line 184: Why use a 1 deg meteorological model for initial and boundary conditions when there are other higher resolution products for this too? Why not use a finer horizontal resolution meteorological model like NAM or HRRR? How might this impact your meteorological analysis later on? This would in particular impact wind direction and speed. How did the wind speed and wind direction from AQS compare with the model results? You show the observations in Figure 1, but do not show how this compares with the model results.
Page 8 line 213: Do you mean 30 days in June? If not, why not run for the entire June time period?
Page 9 line 217: For the selection of AQS sites can you describe this more? Did you only select sites with a certain amount of data available for the entire time window of 2017 – 2021? What latitude / longitude constraints did you use to define the region of a city? The AQS sites include metadata to characterize the site measurement scale, which gives you an idea of how representative that site is of a broader area. Some AQS sites are in locations that are less applicable to be simulated by a 3 km horizontal resolution model. Did you consider this at all in your choice of selected sites? And / or does this explain some of the biases you see in the results section?
Page 11 section 3.1: I’m assuming from Section 2.3, that this is an average of many sites in a given city. Can you provide a table in the supplement for which sites you included in this analysis? Can you do this for other measurements if needed too? Is there variability in hourly ozone and MDA8 ozone across the different sites in a given city and how well does the model represent this variability?
Table 1 and 2: Are these averaged for June 2017 – 2021 or just for a specific year.
Figure 5: Especially for isoprene and HCHO there do seem to be some differences between the model and observations. Can you add the median or mean bias to the plot to show this better?
Figure 5: For Figure 5 and in plots/analysis later on, for NO2 is this a direct comparison with NO2 in the model to NO2 in the observations or do you apply any correction for interferences that the AQS sites can have for NO2 (e.g., Dunlea et al., 2007, https://doi.org/10.5194/acp-7-2691-2007). If no correction was applied, do you think this could explain some of the biases you see?
Figure 6: Why not include all the different monitoring sites for these cities and zoom in more to show the regional variation across a city here rather than an average for all sites in a city? If this is an average, it would be good to be clearer here and in the text.
Figure 10: Including the WRF-Chem and reanalysis model simulated number of exceedance days corresponding with the AQS sites for direct comparison against the observations would be extremely useful for understanding the forecast skill of the WRF-chem model. It’s hard to discern this from Figure 10. Can this be added to the analysis in some way? Your conclusions and Section 3.2 state that WRF-Chem agrees quite well with the observational data for number of exceedance days, but statistics or a bias plot would support this conclusion better.
Figure S5: Can you double check the units for ozone along the trajectory. The color bar goes from 0, 2, 4, 6, 8, 1? And the units are in ppm? This seems too high?
Page 26 line 520 and also in your conclusion on page 31 line 618: Can you explain further why these results suggest that ozone exceedances on this day were caused by inter-state transport rather than local production? How far back in time do these HYSPLIT trajectories go? You also state that the PBLH was lower and temperature was higher which could cause higher local ozone formation too? Looking at the wind direction and speed from the model and observations seems important for evaluating this event. Have you looked at these metrics too? Adding this as a conclusion on page 31 and line 618 “that Arizona is substantially affected by inter-state transport of O3 from California” seems speculative. You need more analysis to state this, so I would strongly recommend rewording this sentence or doing more analysis.
Page 28 line 550: Can you provide the references for why you use these specific characteristics of FNR to describe the differences in the regimes? Additionally, there are several studies (e.g., Schroeder et al., 2017, https://doi.org/10.1002/2017JD026781) that demonstrate the uncertainties of using the FNR approach to approximate ozone production. Can you provide more context on the uncertainty of this approach? Are there other studies that have investigated the ozone production in your cities of interest in the recent past that you can also refer to? Do they agree with your conclusions using this FNR approach?
Figure 14: Can you add lines to represent the different regime changes as specified in line 550 that you are assuming in this work? This would be useful further validation of these values.
Page 31 line 624 - 630: Can you be clearer in this paragraph what your main conclusion is with regard to the ozone production sensitivity including the approach used to determine it and the uncertainties of this approach? From a policy perspective you state both your correlation approach strongly suggests a VOC-limited regime while also saying the FNR analysis suggests VOC-limited or transitional regime. Which regime does your analysis support and what is the uncertainty on it? It is important to be clear what your analysis is suggesting and the uncertainty on your analysis for understanding the policy ramifications of your work.
Citation: https://doi.org/10.5194/gmd-2023-234-RC3 - AC1: 'Authors' response to all comments on gmd-2023-234', Yafang Guo, 02 Apr 2024
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 311 | 84 | 23 | 418 | 26 | 16 | 11 |
- HTML: 311
- PDF: 84
- XML: 23
- Total: 418
- Supplement: 26
- BibTeX: 16
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
