the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
High-resolution large-eddy simulation to understand ozone formation and atmospheric oxidation capacity in Houston, Texas
Abstract. Highly reactive volatile organic compounds (HRVOCs) from mobile and petrochemical sources are important players in atmospheric photochemistry that contribute to the formation of ozone (O3). In a typical elevated O3 episode, we applied a high-resolution large eddy simulation (LES), coupled with the Weather Research and Forecasting model with chemistry (WRF-LES-Chem) to understand the mechanism of high O3 production over the Houston area. Our modeling was constrained and evaluated using field measurements from the NASA Tracking Aerosol Convection Interactions ExpeRiment – Air Quality (TRACER-AQ) project, Texas Commission on Environmental Quality (TCEQ), and vertical column density observations from Pandora spectrometers. The modeling results show enhanced performance in the LES domain, compared to the mesoscale models in simulating key chemicals. O3 sensitivity in the Houston urban area demonstrates a nearly homogenous early morning VOC-limited regime and transits to a noontime NOX-limited regime. As the day progresses into the afternoon, the atmospheric oxidative capacity (AOC) increases with major contribution from hydroxyl (OH) radical (90 %). High concentrations of alkenes also increased O3 (8–10 %) contribution to AOC in the late afternoon. The OH reactivity (KOH) is dominated by isoprene (35.76 %), carbon monoxide (CO; 12.98 %), formaldehyde (HCHO; 12.21 %), and alkanes with C > 3 (6.29 %), thus accelerating the production of hydroperoxyl (HO2) and peroxy (RO2) radicals. The concentrations of short-lived VOCs such as HCHO and acetaldehyde from the oxidation of HRVOCs, increased in the afternoon, which elevated O3 production rates under a NOX-limited regime. The oxidation of isoprene also accelerated the production of HCHO and contributed to the production of HO2 radicals, thus leading to a high O3 production rate. This study suggests the possible impacts of NOX-O3-VOC sensitivity on O3 production rates in polluted urban areas with high emission of HRVOCs, and also provides insights on radical chemistry that drives the photochemical processes of O3 formation. Ultimately, the study underlines the need to control anthropogenic emissions such as alkenes and HCHO and also highlights the role of naturally emitted isoprene species in elevated urban O3 levels.
- Preprint
(32848 KB) - Metadata XML
-
Supplement
(23674 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1190', Maarten Krol, 27 Jul 2024
This paper presents a detailed analysis of the photochemistry in Houston, Texas. The (too long?) paper presents detailed analyses of some relevant quantities linked to ozone formation, such as the ratios Ln/Q and FNR, OH reactivity, Atmospheric Oxidizing Capacity (AOC), which is interesting to read. I note, however, that all these metrics are derived from model data, and not linked to similar observational quantities. The authors do some effort to compare the model to observational data from meteorology and atmospheric composition and conclude that the “D04-LES-300m” domain performs better than the DO2-mesoscale domain. They also write " LES simulation 300 m spatial resolution adequately captures the surface concentration of O3, NOX , and HCHO”. They base this conclusion mainly on Figures 4, 5, and 6, and some figures in the Appendix. Based on this conclusion, they proceed with the detailed analysis of the model output from D04.
The comparison shows still large discrepancies with observations. For instance, the ozone profiles presented in figure 6 show large discrepancies. Moreover, figure 5 shows only averages (bit unclear). Apparently, there is a wealth of information on e.g. diurnal variations of isoprene, but we do not get to see these. I would therefore strongly advise to make a more convincing comparison to observational data.
From a modelling point of view, the authors take a big step to run on high resolution (300 m). Apart from concluding that on this resolution the model performs better, there is unfortunately little analysis of the reasons why this is the case. Is this because the chemical contrasts get larger, causing more extreme chemical regimes (e.g. larger disruptions of chemistry because high NOx emissions)? There are some hints that the boundary layer mixing is better resolved on 300 m resolution, but this aspect is not well worked out either.
In the end, the authors draw some conclusions that both isoprene is important (e.g. largest OH-reactivity) next to anthropogenic VOC emissions. VOC concentrations basically are not evaluated with observations. Moreover, I was shocked that existing NOx inventories had to be scaled by ~0.2 to get meaningful results. This hardly receives attention in the abstract and conclusions. How does the model behave if the emissions of anthropogenic VOCs are off by a similar amount?
In conclusion, the paper presents an interesting analysis of model data in terms of ozone production and chemical reactivity, but the analysis is based on poorly evaluated models results, and fails to address the reasons why a high-resolution simulation would be required to analyse valuable data from measurement campaigns.
Minor Issues
Referencing: it seems referencing is restricted to inner circle papers. There is a long history of research lines that address the effects of resolution on atmospheric chemistry that is totally missing.
Other comments are in the annotated manuscript.
Appendix:
Figure S1: Unclear how to interpret the bottom plots.
Figure S5: Panels d-f not in the caption.
Figure S6: Idem: lower panels are not explicitly mentioned in caption.
- AC1: 'Reply on RC1', Yang Li, 30 Aug 2024
-
RC2: 'Comment on egusphere-2024-1190', Anonymous Referee #2, 08 Aug 2024
The study analyzes the O3 production regimes, OH reactivity, and atmospheric oxidative capacity (AOC) for a high-ozone episode in Houston using a large-eddy WRF simulation coupled with chemistry, which is named WRF-LES-Chem. The usage of WRF-LES-Chem is a highlight of the study, but all the primary analyses related to ozone formation and AOC are based on WRF-LES-Chem outputs, which are verbose and have been done in previous studies. I didn’t find the necessity to use a large-eddy simulation based on the current results, especially when the results from the large-eddy simulation are similar to previous studies using mesoscale simulations. A possible way to improve the study is adding the O3 formation and AOC analysis using the mesoscale simulation and comparing them with the results from the large-eddy simulation against observations and/or previous studies. The authors compared the large-eddy simulation and the mesoscale simulation against observed meteorology and concentrations of some chemical species, but those comparisons can’t convince the community that we need a large-eddy simulation to investigate O3 formation and AOC.
My second concern is the scaling of NEI2017 NOx emissions by 0.2 – 0.3 based on limited observations. Even if I can accept the scaling, why are the O3 concentrations still much larger than the observations in Figure 6? Doesn’t it mean that your scaling has problems? Notably, NOx emission has apparent spatial heterogeneity. The NEI2017 you used only has a resolution of 12 km, much coarser than the WRF-LES-Chem simulation at a resolution of 300 m. Therefore, much finer information about the NOx emission spatial distribution is missed. Without a compatible high-resolution emission inventory, WRF-ELS-Chem can only provide better meteorological conditions, which is helpful but cannot justify the running of the chemical model at such a high resolution. A region with spatially homogeneous NOx emissions would be better to justify the usage of WRF-LES-Chem.
My third concern is the uncertainties of those observations used in the study. I suggest adding more details about those observations, e.g., how the data were measured, their accuracy and precision, and the assumptions used to derive the data.
Minor comments:
Line 89: What do you mean by this sensitivity regime? The 120 ppb O3 concentration? If you referred to the general NOx-VOC-O3 sensitivity regime, the papers you cited right before this sentence have shown some results. Why did you say that it is poorly understood in the Houston area? Please consider rewriting this sentence.
Lines 136-137: Do you mean that you ran two separate simulations, one with D01 and D02 and the other with D03 and D04?
Lines 169-171: The simulation period has been mentioned in Lines 151-152.
Line 182: Add “Meteorological” before “initial conditions.”
Line 183: Why did you use D01 but not D02?
Lines 193-194: The scaling is either 0.3 or 0.2. How did you get a value of 11.32%, even smaller than 0.2? Or do you mean the speciation of NO and NO2 emissions? In addition, did you use the TCEQ and Pandora observations when determining the scaling factors?
Line 292: In Line 183, you said you used output from D01 as IC & BC for the inner domains. Why did you show D02 here?
Figures 3 and S2. The first rows of the two figures have the same contours. Are you sure the two figures represent different times? In addition, subplot (j) in both figures is unclear. Please consider a more straightforward way to show wind directions.
Lines 306-307: I see what you mean, but I’m not entirely convinced. The LES just shows finer structures with more significant spatial heterogeneity.
Line 322: “diurnal”? You only showed 9:00 and 15:00 CDT.
Line 325: Delete “slightly.”
Lines 341-343: Figures 4 and S3 show larger HCHO in the morning than in the afternoon!
Line 348: Are isoprene, MEK, and xylene concentrations better in D04 than in D02 compared to observations? Figure 5 doesn’t show that!
Lines 394-396: Firstly, the relative bias is larger for the LES simulation than the mesoscale simulation. Secondly, I can’t believe that a time lag of less than 20 minutes can explain such a large model bias in ozone concentrations throughout the boundary layer.
Figure S5. Could you please show the model results and observations consistently to facilitate comparison? It is really hard to read the figure. Anyway, the figure indicates that LES better captures the observed evolution patterns compared to the mesoscale simulation. Do you know the uncertainties of the NASA Lidar O3 measurements?
Lines 409-411: Do you have any explanations for that? This is out of my expectation. I think it is because Houston is heavily polluted. Photochemical production of O3 in the boundary layer exceeds the default O3 positive vertical gradient. But I don’t understand why the mesoscale simulation can’t reproduce it, considering NEI2017 has a resolution of 12 km, coarser than the mesoscale simulation.
Figure 9. Please change the legends to make the plots more straightforward.
Lines 484-486: Please rewrite this sentence.
Line 497: Photodissociation of NOx? Do you mean NO2 -> NO + O? The primary loss of NOx in the daytime is NO2 + OH -> HNO3.
Line 514: Isn’t it due to the limited vertical mixing of surface-emitted NOx into the upper boundary layer in the morning?
Lines 574-580: I don’t understand the logic. You cited two papers with different conclusions about the primary sources of HCHO. Then why did you choose the second paper as the basis of your analysis?
Citation: https://doi.org/10.5194/egusphere-2024-1190-RC2 - AC2: 'Reply on RC2', Yang Li, 30 Aug 2024
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
381 | 113 | 42 | 536 | 35 | 11 | 13 |
- HTML: 381
- PDF: 113
- XML: 42
- Total: 536
- Supplement: 35
- BibTeX: 11
- EndNote: 13
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1