Simulations of winter ozone in the Upper Green River Basin, Wyoming, using WRF-Chem
- 1Department of Atmospheric Science, University of Wyoming
- 2Institute of Environment and Sustainability, University of California Los Angeles
- 3Desert Research Institute
- 1Department of Atmospheric Science, University of Wyoming
- 2Institute of Environment and Sustainability, University of California Los Angeles
- 3Desert Research Institute
Abstract. In both the Upper Green River Basin (UGRB) of Wyoming and the Uintah Basin of Utah, strong wintertime ozone (O3) formation episodes leading to O3 concentrations exceeding the 8-hour O3 NAAQS (70 ppb) have been observed over the last two decades. Wintertime O3 events in the UGRB were first observed in 2005 and since then have continued to be observed intermittently when meteorological conditions are favorable, despite significant efforts to reduce emissions. While O3 formation has been successfully simulated using observed volatile organic compound (VOC) and nitrogen oxide (NOX) concentrations, successful simulation of these wintertime episodes using emission inventories in a 3-D photochemical model has remained elusive. An accurate 3-D photochemical model driven by an emission inventory is critical to understand which emission sources have the most impact on O3 formation. In the winter of 2016–2017 (December 2016–March 2017) several high O3 events were recorded with concentrations exceeding 70 ppb. This study uses the Weather Research Forecasting model with chemistry (WRF-Chem) to simulate one of the high O3 events observed in the UGRB during March of 2017. The WRF-Chem simulations were carried out using the 2014 edition of the Environmental Protection Agency National Emissions Inventory (EPA-NEI 2014v2), which includes estimates of emissions from non-point oil and gas production sources. Simulations were carried out with two different chemical mechanisms: the Model for Ozone and Related Chemical Tracers (MOZART) and the Regional Atmospheric Chemistry Mechanism (RACM), and the results were compared with the observed data from 7 weather and air quality monitoring stations in the UGRB operated by Wyoming Department of Environmental Quality (WYDEQ). The simulated meteorology compared favorably to observations in terms of predicting temperature inversions and surface temperature and wind speeds. Notably, because of snow cover present in the basin, the photolysis surface albedo was modified in all simulations. Without this modification, none of the simulations formed O3 exceeding 70 ppb, though the models were relatively insensitive to the exact photolysis albedo if it was over 0.65. The MOZART simulation produced more O3 in the basin than the RACM simulation and compares better with the observations. However, while O3 precursors NOX and NMHC are predicted similarly in simulations with both chemistry mechanisms, simulated NMHC mixing ratios are a factor of six lower than the observations, while NOX mixing ratios are also underpredicted but are much closer to the observations within the region of oil and gas production. The results show that both the RACM and MOZART chemical mechanisms were able to produce O3 even though the NMHC mixing ratios in the model were a factor of six too low, an intriguing result for future studies.
Shreta Ghimire et al.
Status: final response (author comments only)
-
RC1: 'Comment on acp-2022-456', Anonymous Referee #2, 30 Aug 2022
This paper looks at high ozone events that occur in wintertime in the Upper Green River Basin. High wintertime ozone concentrations have been discussed in previous studies and are attributed to emissions from oil and gas extraction in combination with temperature inversions and enhanced photolysis fluxes due to snow covered ground. This paper discusses to what extent a regional chemical transport model (WRF-Chem) is able to represent these conditions by conducting sensitivity simulations with two different chemical mechanisms as well as a simulation where dry deposition has been turned off. The authors find that despite a significant underestimate in the modeled VOC concentrations, either chemical mechanism was able to represent the enhanced ozone concentrations.
This paper provides a good overview of previous work and in general the approaches and results are well presented. What I see missing is, however, a more in-depth analysis of the model results and an attempt to shed light into why the model despite a significant low bias in VOCs simulates ozone concentrations relatively well. The paper could be strengthened significantly by including more information on the NOx and VOc sensitivities in the model (e.g. looking at HCHO/NO2 ratios, modeled chemical tendencies etc.) and how they vary between the model simulations and also vary temporally and spatially. The model could also be compared to HCHO/NO2 ratios at the Boulder site if speciated VOCs are available (from the data set description it is not clear what type of VOC measurements were collected). It further would be valuable to focus on individual VOC species and not just the total VOCs since the reactivity of different VOCs and their role in ozone production varies widely. The modeled VOC bias might be driven by only a few VOCs that have abundant emissions but play little role in ozone production.
Specific comments:
Line 143: I would disagree in that a valuable model should be able to represent conditions under any emission scenarios and VOC:NOx levels
Section Model Setup: Table 2 lists only a few of the settings and Figures A2 and A3 will only be meaningful to readers who are very familiar with WRF-Chem. I suggest extending Table 2 and explicitly stating some of the main information there. Additional information is also needed on the model configuration, e.g. what was used as chemical boundary conditions, was the meteorology in the model constrained and if so how, …
Some questions to A2 and A3: The RACM setup does not use biogenic and fire emissions and also have_bcs_chem is set to false? There are a number of other differences between the MOZART and RACM simulation, so this means that the seen differences are not just related to the chemical scheme. Please elaborate on this and also provide justification behind these settings.
In addition, this is a fairly small domain and I wonder how do chemical boundary conditions influence the ozone concentrations in the Basin? How were the chemical initial and boundary conditions treated (related also to comment above)?Section 2.3: More detail on the measurement techniques and the accuracy of the measurements is needed.
Line 198: MOZ17 has not yet been defined
Line 209: Was a spin-up period considered and if so how long?
Line 310: Is the model able to represent the diurnal variability and day-to-day variations? Is there a significant difference between daytime and nighttime performance?
Line 321: I suggest replacing “accurately” with adequately given that the model has clear shortcomings in representing observed conditions
Line 339: I suggest to also define an acronym for the RACM simulation without dry deposition, e.g. RACM17_nodep to be consistent with the naming of the other simulations
Line 374: I suggest a phrasing of “... do not show a strong sensitivity …”
Line 376: I would say that despite missing data there seems to be a clear overprediction in modeled NOx. Have the authors looked at whether the type of mapping the model 4km data to the site location could explain some of the differences between measured and modeled concentrations?
Line 377: I suggest changing “removed” to “further away”
Line 379: Is this a boundary layer issue or an issue in the diurnal cycle of the emissions or is there any other reason?
Line 381ff: Have you looked whether model grids surrounding the Boulder site have higher NMHC mixing ratios? More information is also needed on how the intercomparison was done and how it was ensured that the modeled total NMHC indeed reflect the same type of information as the measured NMHC (i.e. that it really is an apple to apple comparison)
Line 388: What is the statement about NMHC removal based on?
Figures 9-14: It is really hard to see any details in the NOx and VOC graphs, maybe a different color range could help? The paper also first discusses the ozone plots from Figure 12 and then looks into the NOx graphs. You might want to consider swapping the order of the Figures.
-
CC1: 'Comment on acp-2022-456', Seth Lyman, 15 Sep 2022
On lines 383-384, does the Boulder station measure speciated NMHC, or just methane and total NMHC? If it measures only total NMHC, I wouldn't trust the magnitude of the measurement. TNMHC GCs tend to strongly overestimate the amount of NMHC in the air compared to speciated measurements of individual compounds, so a comparison with that measurement is not likely to be useful.
-
RC2: 'Comments on acp-2022-456', Anonymous Referee #3, 05 Nov 2022
The high wintertime O3 pollution in the Upper Green River Basin (UGRB), Wyoming is simulated in the study. During some years in winter months high O3 pollution in oil and gas producing basins of Utah and Wyoming have been observed. Numerous field campaigns and modeling studies have been conducted to understand the emissions and processes causing these high O3 pollution events. It is important for the air quality models to accurately simulate the wintertime O3 in UGRB, which could also help to develop mitigation strategies in the future. Here the authors deploy the-state-of-the-art WRF-Chem model to simulate high O3 during March, 2017. There are several aspects of the study that could make an important contribution to the field. The authors also conduct rigorous evaluation of the meteorological simulations. However, there are some shortcomings of the study that need to be addressed.
Major comments:
- The authors emphasize the importance of using the existing anthropogenic emission inventories to model the high winter O3 in UGRB, and claim that this is the main strength of this study. While it’s important to use the bottom-up emission inventories, the scientific community should not limit itself using the bottom-up inventories only. As Ahmadov et al. 2015 showed the EPA NEI inventory can grossly over/under-estimate the NOx/VOC emissions from an oil and gas producing region (Uintah Basin). Therefore, in my opinion it’s an underestimation of the importance of the study by focusing on the use of the emission inventory.
- Introduction: The statement about the shortfalls of other studies is somewhat misleading. Do the authors refer to the box modeling studies conducted in the past? The box models are designed to use measured concentrations of the chemical species, not emission inventories. As for the 3D air quality models Ahmadov et. al. (2015) demonstrated that the emission inventories can have huge uncertainties. Moreover, as I discuss below this study doesn’t prove that the NEI-2014 inventory accurately represents the emissions for the UGRB during March, 2017.
- Here two different gas chemistry schemes are used, MOZART and RACM. As the WRF-Chem namelists provided in SI show the MOZART simulation included aerosols and their feedback on radiation. However, in the RACM simulations the authors turned off aerosols. In the paper differences in the meteorological simulations between these two model cases are presented and attributed to the aerosol feedback, though simulated aerosol fields aren’t shown. I assume the aerosol concentrations in UGRB were relatively low.
- The two gas chemistry mechanisms also use different photolysis schemes (phot_opt). Such difference makes it hard to compare the results of these two model cases.
- Here the model simulations are presented for 5 days only. This is quite short. I suggest extending the model simulations to evaluate the model’s capability in simulating ozone and other chemical species other days in March, 2017. Even if O3 levels were low those days it's imporant to check the model's ability to simulate O3 and other species in different meteorological conditions by using the same model configuration and emission dataset.
- 350: Ahmadov et al. (2015) found that the reduced dry deposition of ozone over snow covered ground is one of key processes leading to high wintertime ozone buildup. It seems that the model has this snow impact on dry deposition in the MOZART scheme, but not in the RACM scheme in the version of the model used here. This discussion of the dry deposition needs to be revised.
- Although the model is able to simulate the high O3, the simulated VOC mixing ratios are a factor of six lower than the observed ones. The NOx simulations show underestimation too. This begs the question, does the model simulate high O3 for the right reasons? It’d be helpful to conduct sensitivity simulations by adjusting the NOx and VOC emissions to account for uncertainties in the NEI.
- 360: This is missing in the community version of WRF-Chem.
- For the mitigation strategies it’s helpful to understand the NOx/VOC sensitivity of the O3 formation. I suggest conducting sensitivity simulations by adjusting the emissions to show how the simulated O3 will respond to the NOx and/or VOC emission adjustments in UGRB.
- The advantage of using a tightly coupled meteorology-chemistry model such as WRF-Chem isn’t discussed here. As Ahmadov et al. showed this is essential to simulate the stagnation episodes and multi-day buildup of the pollutants in a basin.
Minor comments:
The CMAQ modeling paper by Matichuk et al. (https://doi.org/10.1002/2017JD027057) isn’t cited here.
The evaluation of the meteorological simulations can be moved to SI.
Shreta Ghimire et al.
Shreta Ghimire et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
290 | 99 | 9 | 398 | 2 | 2 |
- HTML: 290
- PDF: 99
- XML: 9
- Total: 398
- BibTeX: 2
- EndNote: 2
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1