the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Jul 2024
Urban ozone formation and sensitivities to volatile chemical products, cooking emissions, and NOx across the Los Angeles Basin
Abstract. Volatile chemical products (VCPs) and other non-traditional anthropogenic sources, such as cooking, contribute substantially to the volatile organic compound (VOC) budget in urban areas. The impact of these emissions on ozone formation and urban atmospheric chemistry is uncertain. This study employs detailed Lagrangian box modeling and sensitivity analyses to evaluate ozone response to sector-specific VOC and nitrogen oxide (NOx) emissions in the Los Angeles (LA) Basin during the summer of 2021. The model simulated the photochemical processing and transport of temporally and spatially gridded emissions from the FIVE-VCP-NEI17NRT inventory that combines emissions from fossil fuels, VCPs, and other point sources and included updates to cooking emissions based on recent field observations. The model accurately simulates the variability and magnitude of O3 (hourly normalized mean bias = -0.03; R2 = 0.83), NOx, and speciated VOCs measured at a ground site in Pasadena, CA. VOC sensitivity analyses show that anthropogenic VOCs (AVOC) enhance daily maximum 8-hour average ozone in Pasadena by 12 ppb, whereas biogenic VOCs (BVOCs) contribute 8 ppb. Of the ozone influenced by AVOCs, VCPs represent the largest fraction at 44 % while cooking and fossil fuel VOCs are comparable at 28 % each. This study is the first to quantify the contribution of cooking emissions to urban ozone. NOx sensitivity analyses along trajectory paths indicate the photochemical regime of ozone varies spatially and temporally. The modeled ozone response is primarily NOx–saturated across the dense urban core and during peak ozone production in Pasadena, but transitions back to NOx–limited chemistry briefly during late afternoon hours. Lowering the inventory emissions of NOx by 25 % moves Pasadena to NOx–limited chemistry throughout the majority of the day and shrinks the spatial extent of NOx‒saturation towards downtown LA. Further sensitivity analyses in Pasadena show that using VOCs represented by a separate state inventory requires steeper NOx reductions to transition to NOx‒sensitivity, further suggesting that accurately representing VOC reactivity in inventories is critical to determine the effectiveness of future NOx‒reduction policies.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(1473 KB)
-
Supplement
(1286 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1473 KB) - Metadata XML
-
Supplement
(1286 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1899', Anonymous Referee #1, 17 Jul 2024
Stockwell et al. 2024 take advantage of a Lagrangian model armed with state-of-the-art VOC and NOx emissions and chemistry over a region undergoing degraded air quality to study the sensitivity of surface ozone to various emissions. The authors found some exciting new contributors to augmenting ozone in LA, such as the contribution of cooking emissions and personal products. Attributing what emission types control PO3 and where the non-linearities of PO3 are located is essential to better implementing emission mitigation. This type of work is usually practiced with CTM models because it can provide much larger data points and a more suitable treatment of non-local and local vertical mixing. However, having a more straightforward but effective tool is appreciated. The paper has some important new messages for controlling ozone pollution in the region; nonetheless, there are many ambitious points about the modeling framework, and the paper should extend the number of receptor points to capture better the complete picture of ozone sensitivity in the region. If an algorithm proves to be faster, it is likely to be reflected in more runs, necessitating the addition of more receptors. I recommend the publication of this work after major revisions.
Major comments
Poor description of the modeling framework: The description of the modeling framework is unclear and contains ambiguous points. The model calculates backward trajectories based on WRF-FLEXPART arriving at two separate stations in LA's downwind. For each 15-minute interval trajectory, Eq.1 is used to dilute any primary emissions over a box of 8 km x 8 km x PBLH, resolve chemical source/sinks by F0AM, and consider dilution/entrainment based on PBLH and some prescribed background concentrations. This model is a simplified version of a CTM model in a Lagrangian framework, ignoring cloud chemistry, vertical diffusion, convective transport, and dry deposition. The description of the modeling part needs serious refinement. For instance, the photolysis rates are mentioned in different subsections with a questionable correction factor stating that "the WRF-Chem calculated photolysis frequencies were scaled using observed NO2 photolysis rates at the ground site (jNO2) ratioed to WRF-modeled jNO2". How can a single measurement at a given location constrain photolysis rates in all backward trajectories going over different surface albedo, aerosol/cloud cover, and solar irradiance at different times? The same applies to meteorology. The authors stated: "The model was constrained with meteorological measurements of pressure, temperature, and relative humidity conducted at the ground site." How were meteorological measurements treated for all trajectories? Where are the Doppler LiDAR measurements, and how were they used to correct PBLH over this many trajectories? Are all these trajectories passing through the PBL region (because that’s how the authors have diluted the emissions)? How does the F0AM model work in this algorithm? Does the model cycle with the sun so that it can approach a steady state? Are we assuming a steady state for chemistry? Dry deposition is a large sink for surface ozone; where is its exclusion being compensated for? I highly recommend adding a flowchart explaining how each component is adjusted or constrained and its connections to the other elements.Excessive dilution within the PBL: From my understanding, any VOC or NOx emissions within the PBL (it could be mobile sources at the surface layer or a large chimney at higher altitudes) are diluted uniformly within the PBL (the first term in Eq1). This means that the authors are overly simplifying non-local and local vertical diffusion components typically parameterized in Eulerian models such as WRF-Chem. This will lead to some underestimation/overestimation issues, especially for short-lived species, including NOx and isoprene. The vertical distribution of these species rapidly declines within the PBL. Even under an expanded PBL height and considerable turbulence, it is unlikely for some of these critical compounds to be fully well-mixed. A vertically-mixed situation is only applicable to long-lived species such as ozone and CO. So, the current framework must have representative issues concerning photochemical ozone conditions because, realistically, there is a vertical dependency of PO3 sensitivity to NOx and VOC within the PBL height that is overlooked here. This representativity issue needs to be studied and carefully conveyed as a limitation in the conclusion.
Representativeness of the whole LA Basin: This study focused on two receptor sites to generalize the sensitivity of PO3 to different VOC types and NOx emissions. While one of these sites is a supersite measuring a vast number of geophysical variables, they are influenced by a limited number of atmospheric conditions (both chemistry and meteorology) along with the trajectories that may only be representative of some physiochemical processes transpiring in the LA basin. In other words, a limited number of trajectories (which move by time and space and do not sample uniformly like how an Eulerian model does) limits the degree of generalizability of the conclusions made from this work. If, hypothetically, we had 100 supersites uniformly spreading over the domain, would the conclusion change? Some of the statistics related to the supersite are also not indicative of how the model performs over the LA. For instance, it is challenging to attribute the isoprene biases over the Pasadena site to the isoprene emission over the whole region. The last air parcel over the supersite doesn't fully remember all the physicochemical processes happening back in the trajectory. The authors included another site with limited observations to expand their analysis, but this has raised a legitimate question: if this modeling framework is much more efficient than WRF-Chem, what is preventing the authors from applying the same algorithm on many EPA’s surface sites to boost the confidence in the results and attribution of LA ozone to LA’s emissions.
Minor comments:
I highly recommend replacing the NOx-saturated regime with the VOC-sensitive term throughout the paper. This is primarily because “NOx-saturated” tends to overemphasize the positive effect of NOx reduction on PO3 instead of the negative effect of VOC reduction. As a matter of fact, even in NOx-saturated regimes, it is important to reduce NOx as the outflow of NOx mixing with suburban and rural areas can cause rapid PO3 growth; additionally, a continuous and consistent reduction in NOx can eventually bypass the non-linearities and make a large contribution to the reduction of PO3. Almost every city in the US used to be NOx-saturated in the late 1990s, but the steady reduction in NOx now is helping a lot at curbing PO3 and, thus, O3.
L43: Based on the EPA’s report, point source emissions have decreased too.
L48. Please add additional sentences about CTM’s works, observationally constrained box models, and the use of OHR to detect unmeasured VOCs with significant errors on PO3. Your introduction has brushed off a large domain of scientific endeavor.
L51. Abdi-Oskouei et al. (2022) isn’t a box model study.
L55. Replace O3 with PO3 (ozone production rates).
The paragraph containing 55: In both cases (NOx-sensitive and VOC-sensitive), you have radical termination either through HO2+HO2, RO2+RO2, RO2+HO2 (NOx-sensitive), or NO2+OH (VOC-sensitive). The authors need to mention the effect of RO2 and HO2 on OH formation in the presence of NO. Also, the authors need to mention that these categories (including the transitional regimes because we don’t live in a binary world) are limited to photochemically active environments. Under low light conditions (radical-limited), we essentially see the portioning between NO-NO2-O3 without any noticeable effects of VOC.
L69. Wasn’t Jiang et al. 2019 debunked by Silvern et al. 2019?
The paragraph containing L95: More references need to be mentioned here about using observationally-constrained box models and using the LROx/LNOx indicator.
L99. Poor reasoning. CTMs require proper input.
Section 2.1. Where is the LiDAR measurement located?
Figure 1. Please mention why you are showing D5-siloxane emissions in the caption.
L162. What does a limited area version mean?
L131. Dry/wet deposition?
L173. Please stick to the FLEXPART-WRF description. The F0AM part needs to be moved to another section.
L199. Including the lockdown effect?
L229-232. Seems irrelevant and disjointed.
L294. Unclear what you mean by “limited” here.
Figure 2. Adding HCHO to these plots and discussing them in the text would be interesting, as it is a good proxy for VOCR.
Section 3.3. Almost everything is included in the supplementary, which is a bit distracting. I highly suggest bringing HCHO to the discussion and the plot.
L350. R2 isn’t a correlation coefficient.
L403. Due to the short lifetime of isoprene, the adjustment based on MACR-MVK may not be appropriate to fully match up the isoprene concentration.
L450. Do you really need to get an agreement with other studies focusing on different times and locations? I would just drop this part. You can highlight the contrasts and explain why there are some differences, but you shouldn’t look for validation based on results from different environments.
L451. Do cooking emissions consider propane leakage?L470. But the inflation of BVOCs wasn’t enough to match up isoprene, so I’m unsure if we would call it an upper bound.
L476. This is an overgeneralization. You can’t statistically say this using two data points. There are indeed large spatial variances associated with photochemical conditions. If you want to include this, please expand the analysis over other sites (see the major comment).
L485. This is not true, see Tao et al. (2022) (https://doi.org/10.1021/acs.est.2c02972) and Souri et al. (2020) (https://doi.org/10.1016/j.atmosenv.2020.117341)
L 487. It is an overstatement to say that a Lagrangian model can provide dense spatial information, especially relative to Eulerian frameworks.
Figure 5. This question must have been raised sooner, but which altitude do trajectories represent? Are they all within the PBL?
Section 4.2. Is studying chemical conditions meaningful in the early morning and later afternoon when HOx-ROx is rather inactive?Editorial comments
L136. 3D model retrievals -> 3D model simulations
The FIVE-VCP-NEI17NRT acronym isn’t easy on the eye. Can you make it more compact?L370. What do you mean by “enhanced PBL heights”?
L407. Formed chemically through secondary pathways…
Figure 3: The color lines in the legend do not match the plots. But I don’t know what the best way to fix this is.
Figure 4. Please insert 12 ppbv somewhere in the pie plot (maybe in the middle as a text box).
Line 436. Anthropogenic-induced instead of anthropogenic
Line 478. The magnitude of photochemically produced O3 through AVOC and BVOC…
Figure 5. You should clearly mention in the caption that the map is based on backward trajectories to avoid mistaking this for a continuous Eulerian framework. Because there are so rapidly elongated patterns that may seem nonsensical if we wrongly see them as separate boxes simultaneously.
Section 5. It is choppy; please break this long paragraph into smaller chunks and provide more quantitative numbers. Please provide more details about where this type of analysis, such as including cooking and VCP in PO3 sensitivity, becomes a necessity (don’t limit your audience to people interested in air quality issues in LA).Citation: https://doi.org/10.5194/egusphere-2024-1899-RC1 -
RC2: 'Comment on egusphere-2024-1899', Anonymous Referee #2, 09 Sep 2024
A Box model with updated VOC chemistry is used to determine the sensitivity of O3 to NOx and VOC emissions from various sources, including VCPs, fossil fuels, cooking, and biogenic sources, in the LA basin. FLEXPART, with input from WRF-Chem, was used to determine the trajectory of the air parcels that arrived at the two sites. Cooking emissions were added to the FIVE-VCP-NEI17NRT emission inventory and used in the box model. Various emission sensitivity experiments with the box model were conducted to assess the sensitivity of O3 concentration and MDA8 to changes in anthropogenic VOC emissions.
Key point conclusions from section 1:- FIVE-VCP-NEI17NRT inventory shows a good representation of LA emissions in space and time when compared to O3, NOX, and speciated VOC measurements in Pasadena (I am not convinced; details below).
- They concluded that the anthropogenic VOC contributes up to 12 ppb to the total MDA8 in Pasadena (what % of the total MDA8 is this?).
Of this 12ppb:
- 44% (5.28ppb) is attributed to VCPs [5.2 +/- 0.6 ppb]
- 28% (3.36ppb) to fossil fuel VOCs. What is the uncertainty?
- 28% (3.36ppb) to cooking. What is the uncertainty?In the second part of the study, they conducted more experiments with varying NOx emissions (and VOC). They concluded that the urban core of LA basin, including Pasadena, is primarily NOx-saturated for most of the daytime and shifts to NOx-limited farther east of the LA basin.
Using an updated box model is valuable for better understanding the O3 formation and O3 regime but requires a more careful experiment setup. For example, looking at one month of simulation is typically done with a CTM, which can capture variability in the transport patterns, meteorological conditions, background O3 (and other species), etc. Box models are not typically designed to capture this type of variability.
I see two ways to improve this work and overcome some of the limitations:
- Clean up the box model input data and only include days and hours with typical transport patterns (reduce uncertainties). This may reduce the scope of the work but will significantly increase the confidence in the conclusion
- Add CTM analysis to complement this study. In my opinion, adding a CTM analysis, especially to evaluate the performance of the new emission inventory, is essential.This paper contains many missing details and contradictory conclusions. I tried my best to understand and guess some of the assumptions by reading the references and going back and forth in the paper many times. I highly recommend reorganizing the paper and adding more information and details about the methods, assumptions, and, most importantly, limitations of this study.
2.1. Campaign description
You mentioned that the trailer was away from Aug 2-6 and then from Aug 31 to Sep 3. Fig 2a, shows measurements until Sep 6th. Was Sep 3-6 was included in the study?
2.2. Box model configuration
L141- Does the location of the air parcel’s starting point change for each trajectory?
L142- You described the typical transport in the LA basin and how variability in the transport pattern can introduce uncertainties. Why not only include trajectories with typical transport patterns?L147- How are the initial mixing ratios determined? This is an important piece of information that is missing from the text. There is a mention of ozone concentrations being initialized using measurements from the Westchester SCAQMD site during the day (why not at night?) in section 2.2.4. Is this used for all the trajectories? This site is next to LAX (major pollutant source); how does this impact your conclusions? If the initialization is done using WRF-Chem values, how is the performance of this model? In general, how sensitive are your conclusions to the initial values you pick for your box model?
L148- “emissions encompassed by the area of the box…”. To clarify, the box moves based on the back-trajectory information. You then extract the emission rates from the inventory for that given lat/lon (8x8km box) and add that to equation 1?
So, if there are mismatches in the location of emissions, will they not be added correctly to the box model? If this is true, then this is another important limitation of this work. This points out the importance of evaluating the new emission inventory using CTM prior to running box models.L149- Is PBLH assumed to be the same throughout the trajectory as the measured values at the Pasadena site? Line 151 says observations and 3D-model estimates constrain the PBL. Which observation? Where model estimates are used? Again, very critical information is missing. Is this a reasonable assumption?
Fig1. I highly suggest adding maps of emissions by sector in the paper or in SI.
2.2.1. FLEXPART-WRF back-traj analysis
If you decide to run WRF-Chem for this study, you can use the trajectory monitoring feature in WRF-Chem and skip using FLEXPART. See here for details: https://www2.acom.ucar.edu/sites/default/files/documents/Trajectory.desc_.pdf
L165- Verryken et al. (2024) paper is in prep. Please add information regarding the WRF-Chem performance in capturing transport in this paper.L170- How many air parcels were released from the site, and at what altitude? Please add a map of back-trajectories, similar to Fig S8.
What do you mean by “parcel locations were derived as the center-of-mass from the main particle cluster”? What if the air particle cluster is spread?
With FLEXPART, you can easily create an ensemble run of air parcels and quantify transport uncertainties.L180- Another reason to only include certain hours (mid-day with filly developed boundary later) in the analysis.
2.2.2. Emissions
A lot of important information is missing from this section.
Please add details about the diurnal variability of emissions from each sector (can be a diel plot).
Please add details about the spatial variability of emissions from each sector (it can be a map), especially the cooking sector. How are population density and temporal profile for human activity used for the spatial distribution of cooking emissions?Have the cooking emissions been evaluated in any CTM model? If not, how does this limit your conclusions?
Where are oil and gas emissions? Part of fossil fuel?
2.2.4. Meteorology, dilution, and entrainment
L284- How exactly are the measurements from the Westchester site used? Please mark it on the map in Fig 1.
3. Model evaluation
I wish the method section had more details so I could review sections 3, 4, and 5 more thoroughly.3.1. Ozone.
L303- “the general agreement…” Drawing this conclusion from only NMB and R2 is not accurate. Can you compare O3 and NOX concentrations between the box model and WRF-Chem along the trajectory and at the site?Fig 2. You can exclude days after Sep 3rd if VOC obs is not available then?
Fig 3. X-axis label is not correct.
3.3. VOCs
Low ethane concentrations in the model compared to observation. Did you include oil and gas emissions in your study? Is it part of fossil fuel emissions?Formaldehyde matches observation well. Give the low bias in isoprene, probably for the wrong reasons.
4.1. Contribution of Anthro and bio VOCs to O3
In my opinion, to correctly capture the impact of all the anthropogenic sectors, you need to run a scenario with all anthropogenic emissions removed instead of using Equation 4. Same goes with 10% reduction.
Another point: In reality, you cannot turn off some pollutants from a sector (only VOCs, not NOX) and asses the contribution of this sector on O3 concentration/production or MDA8. As you mentioned, the system is non-linear; thus, including both VOC and NOx simultaneously matters. Changes in the O3 chemical regime are part of the reality and need to be studied.
L435:
Can you add details about the variability and standard deviation of the data? How many ppb is 28%?L437:
Coggon et al showed fossil fuels contribution to ozone production from AVOC at midday was 60%. This is different from the 28% contribution to MDA8 ozone in this work. Please be more precise about your conclusions in this paragraph.What is the total mass of VOC and NOX emissions in each city? Is fossil fuel emission lower in LA?
L453:
What is the contribution of each sector to OH reactivity? Adding more details on this can be valuable for the discussion.
4.2. Spatial and temporal ozone sensitivity to NOx.
Similar to scenarios in the previous section, preturbing only NOx does not show us the full picture of the impact of mitigation policies on O3 concentrations.
To draw a more comprehensive conclusion about the NOx vs VOC sensitivity in the region, you can use metrics such as FNR, LROx/LNOx, or methods as described in Vermeuel et al. 2019(https://doi.org/10.1029/2019JD030842) (they also used a box model). Box models can provide details about PO3, OH reactivity, etc. Please use this information in your discussion about the O3 regime.
L576
Why FIVE-VCP has 50% lower off-road VOCs?Figure 6.
A better comparison is the Red scenario + cooking emission.Thank you, and good luck!
Citation: https://doi.org/10.5194/egusphere-2024-1899-RC2 - AC1: 'Author Comment on egusphere-2024-1899', Chelsea Stockwell, 31 Oct 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1899', Anonymous Referee #1, 17 Jul 2024
Stockwell et al. 2024 take advantage of a Lagrangian model armed with state-of-the-art VOC and NOx emissions and chemistry over a region undergoing degraded air quality to study the sensitivity of surface ozone to various emissions. The authors found some exciting new contributors to augmenting ozone in LA, such as the contribution of cooking emissions and personal products. Attributing what emission types control PO3 and where the non-linearities of PO3 are located is essential to better implementing emission mitigation. This type of work is usually practiced with CTM models because it can provide much larger data points and a more suitable treatment of non-local and local vertical mixing. However, having a more straightforward but effective tool is appreciated. The paper has some important new messages for controlling ozone pollution in the region; nonetheless, there are many ambitious points about the modeling framework, and the paper should extend the number of receptor points to capture better the complete picture of ozone sensitivity in the region. If an algorithm proves to be faster, it is likely to be reflected in more runs, necessitating the addition of more receptors. I recommend the publication of this work after major revisions.
Major comments
Poor description of the modeling framework: The description of the modeling framework is unclear and contains ambiguous points. The model calculates backward trajectories based on WRF-FLEXPART arriving at two separate stations in LA's downwind. For each 15-minute interval trajectory, Eq.1 is used to dilute any primary emissions over a box of 8 km x 8 km x PBLH, resolve chemical source/sinks by F0AM, and consider dilution/entrainment based on PBLH and some prescribed background concentrations. This model is a simplified version of a CTM model in a Lagrangian framework, ignoring cloud chemistry, vertical diffusion, convective transport, and dry deposition. The description of the modeling part needs serious refinement. For instance, the photolysis rates are mentioned in different subsections with a questionable correction factor stating that "the WRF-Chem calculated photolysis frequencies were scaled using observed NO2 photolysis rates at the ground site (jNO2) ratioed to WRF-modeled jNO2". How can a single measurement at a given location constrain photolysis rates in all backward trajectories going over different surface albedo, aerosol/cloud cover, and solar irradiance at different times? The same applies to meteorology. The authors stated: "The model was constrained with meteorological measurements of pressure, temperature, and relative humidity conducted at the ground site." How were meteorological measurements treated for all trajectories? Where are the Doppler LiDAR measurements, and how were they used to correct PBLH over this many trajectories? Are all these trajectories passing through the PBL region (because that’s how the authors have diluted the emissions)? How does the F0AM model work in this algorithm? Does the model cycle with the sun so that it can approach a steady state? Are we assuming a steady state for chemistry? Dry deposition is a large sink for surface ozone; where is its exclusion being compensated for? I highly recommend adding a flowchart explaining how each component is adjusted or constrained and its connections to the other elements.Excessive dilution within the PBL: From my understanding, any VOC or NOx emissions within the PBL (it could be mobile sources at the surface layer or a large chimney at higher altitudes) are diluted uniformly within the PBL (the first term in Eq1). This means that the authors are overly simplifying non-local and local vertical diffusion components typically parameterized in Eulerian models such as WRF-Chem. This will lead to some underestimation/overestimation issues, especially for short-lived species, including NOx and isoprene. The vertical distribution of these species rapidly declines within the PBL. Even under an expanded PBL height and considerable turbulence, it is unlikely for some of these critical compounds to be fully well-mixed. A vertically-mixed situation is only applicable to long-lived species such as ozone and CO. So, the current framework must have representative issues concerning photochemical ozone conditions because, realistically, there is a vertical dependency of PO3 sensitivity to NOx and VOC within the PBL height that is overlooked here. This representativity issue needs to be studied and carefully conveyed as a limitation in the conclusion.
Representativeness of the whole LA Basin: This study focused on two receptor sites to generalize the sensitivity of PO3 to different VOC types and NOx emissions. While one of these sites is a supersite measuring a vast number of geophysical variables, they are influenced by a limited number of atmospheric conditions (both chemistry and meteorology) along with the trajectories that may only be representative of some physiochemical processes transpiring in the LA basin. In other words, a limited number of trajectories (which move by time and space and do not sample uniformly like how an Eulerian model does) limits the degree of generalizability of the conclusions made from this work. If, hypothetically, we had 100 supersites uniformly spreading over the domain, would the conclusion change? Some of the statistics related to the supersite are also not indicative of how the model performs over the LA. For instance, it is challenging to attribute the isoprene biases over the Pasadena site to the isoprene emission over the whole region. The last air parcel over the supersite doesn't fully remember all the physicochemical processes happening back in the trajectory. The authors included another site with limited observations to expand their analysis, but this has raised a legitimate question: if this modeling framework is much more efficient than WRF-Chem, what is preventing the authors from applying the same algorithm on many EPA’s surface sites to boost the confidence in the results and attribution of LA ozone to LA’s emissions.
Minor comments:
I highly recommend replacing the NOx-saturated regime with the VOC-sensitive term throughout the paper. This is primarily because “NOx-saturated” tends to overemphasize the positive effect of NOx reduction on PO3 instead of the negative effect of VOC reduction. As a matter of fact, even in NOx-saturated regimes, it is important to reduce NOx as the outflow of NOx mixing with suburban and rural areas can cause rapid PO3 growth; additionally, a continuous and consistent reduction in NOx can eventually bypass the non-linearities and make a large contribution to the reduction of PO3. Almost every city in the US used to be NOx-saturated in the late 1990s, but the steady reduction in NOx now is helping a lot at curbing PO3 and, thus, O3.
L43: Based on the EPA’s report, point source emissions have decreased too.
L48. Please add additional sentences about CTM’s works, observationally constrained box models, and the use of OHR to detect unmeasured VOCs with significant errors on PO3. Your introduction has brushed off a large domain of scientific endeavor.
L51. Abdi-Oskouei et al. (2022) isn’t a box model study.
L55. Replace O3 with PO3 (ozone production rates).
The paragraph containing 55: In both cases (NOx-sensitive and VOC-sensitive), you have radical termination either through HO2+HO2, RO2+RO2, RO2+HO2 (NOx-sensitive), or NO2+OH (VOC-sensitive). The authors need to mention the effect of RO2 and HO2 on OH formation in the presence of NO. Also, the authors need to mention that these categories (including the transitional regimes because we don’t live in a binary world) are limited to photochemically active environments. Under low light conditions (radical-limited), we essentially see the portioning between NO-NO2-O3 without any noticeable effects of VOC.
L69. Wasn’t Jiang et al. 2019 debunked by Silvern et al. 2019?
The paragraph containing L95: More references need to be mentioned here about using observationally-constrained box models and using the LROx/LNOx indicator.
L99. Poor reasoning. CTMs require proper input.
Section 2.1. Where is the LiDAR measurement located?
Figure 1. Please mention why you are showing D5-siloxane emissions in the caption.
L162. What does a limited area version mean?
L131. Dry/wet deposition?
L173. Please stick to the FLEXPART-WRF description. The F0AM part needs to be moved to another section.
L199. Including the lockdown effect?
L229-232. Seems irrelevant and disjointed.
L294. Unclear what you mean by “limited” here.
Figure 2. Adding HCHO to these plots and discussing them in the text would be interesting, as it is a good proxy for VOCR.
Section 3.3. Almost everything is included in the supplementary, which is a bit distracting. I highly suggest bringing HCHO to the discussion and the plot.
L350. R2 isn’t a correlation coefficient.
L403. Due to the short lifetime of isoprene, the adjustment based on MACR-MVK may not be appropriate to fully match up the isoprene concentration.
L450. Do you really need to get an agreement with other studies focusing on different times and locations? I would just drop this part. You can highlight the contrasts and explain why there are some differences, but you shouldn’t look for validation based on results from different environments.
L451. Do cooking emissions consider propane leakage?L470. But the inflation of BVOCs wasn’t enough to match up isoprene, so I’m unsure if we would call it an upper bound.
L476. This is an overgeneralization. You can’t statistically say this using two data points. There are indeed large spatial variances associated with photochemical conditions. If you want to include this, please expand the analysis over other sites (see the major comment).
L485. This is not true, see Tao et al. (2022) (https://doi.org/10.1021/acs.est.2c02972) and Souri et al. (2020) (https://doi.org/10.1016/j.atmosenv.2020.117341)
L 487. It is an overstatement to say that a Lagrangian model can provide dense spatial information, especially relative to Eulerian frameworks.
Figure 5. This question must have been raised sooner, but which altitude do trajectories represent? Are they all within the PBL?
Section 4.2. Is studying chemical conditions meaningful in the early morning and later afternoon when HOx-ROx is rather inactive?Editorial comments
L136. 3D model retrievals -> 3D model simulations
The FIVE-VCP-NEI17NRT acronym isn’t easy on the eye. Can you make it more compact?L370. What do you mean by “enhanced PBL heights”?
L407. Formed chemically through secondary pathways…
Figure 3: The color lines in the legend do not match the plots. But I don’t know what the best way to fix this is.
Figure 4. Please insert 12 ppbv somewhere in the pie plot (maybe in the middle as a text box).
Line 436. Anthropogenic-induced instead of anthropogenic
Line 478. The magnitude of photochemically produced O3 through AVOC and BVOC…
Figure 5. You should clearly mention in the caption that the map is based on backward trajectories to avoid mistaking this for a continuous Eulerian framework. Because there are so rapidly elongated patterns that may seem nonsensical if we wrongly see them as separate boxes simultaneously.
Section 5. It is choppy; please break this long paragraph into smaller chunks and provide more quantitative numbers. Please provide more details about where this type of analysis, such as including cooking and VCP in PO3 sensitivity, becomes a necessity (don’t limit your audience to people interested in air quality issues in LA).Citation: https://doi.org/10.5194/egusphere-2024-1899-RC1 -
RC2: 'Comment on egusphere-2024-1899', Anonymous Referee #2, 09 Sep 2024
A Box model with updated VOC chemistry is used to determine the sensitivity of O3 to NOx and VOC emissions from various sources, including VCPs, fossil fuels, cooking, and biogenic sources, in the LA basin. FLEXPART, with input from WRF-Chem, was used to determine the trajectory of the air parcels that arrived at the two sites. Cooking emissions were added to the FIVE-VCP-NEI17NRT emission inventory and used in the box model. Various emission sensitivity experiments with the box model were conducted to assess the sensitivity of O3 concentration and MDA8 to changes in anthropogenic VOC emissions.
Key point conclusions from section 1:- FIVE-VCP-NEI17NRT inventory shows a good representation of LA emissions in space and time when compared to O3, NOX, and speciated VOC measurements in Pasadena (I am not convinced; details below).
- They concluded that the anthropogenic VOC contributes up to 12 ppb to the total MDA8 in Pasadena (what % of the total MDA8 is this?).
Of this 12ppb:
- 44% (5.28ppb) is attributed to VCPs [5.2 +/- 0.6 ppb]
- 28% (3.36ppb) to fossil fuel VOCs. What is the uncertainty?
- 28% (3.36ppb) to cooking. What is the uncertainty?In the second part of the study, they conducted more experiments with varying NOx emissions (and VOC). They concluded that the urban core of LA basin, including Pasadena, is primarily NOx-saturated for most of the daytime and shifts to NOx-limited farther east of the LA basin.
Using an updated box model is valuable for better understanding the O3 formation and O3 regime but requires a more careful experiment setup. For example, looking at one month of simulation is typically done with a CTM, which can capture variability in the transport patterns, meteorological conditions, background O3 (and other species), etc. Box models are not typically designed to capture this type of variability.
I see two ways to improve this work and overcome some of the limitations:
- Clean up the box model input data and only include days and hours with typical transport patterns (reduce uncertainties). This may reduce the scope of the work but will significantly increase the confidence in the conclusion
- Add CTM analysis to complement this study. In my opinion, adding a CTM analysis, especially to evaluate the performance of the new emission inventory, is essential.This paper contains many missing details and contradictory conclusions. I tried my best to understand and guess some of the assumptions by reading the references and going back and forth in the paper many times. I highly recommend reorganizing the paper and adding more information and details about the methods, assumptions, and, most importantly, limitations of this study.
2.1. Campaign description
You mentioned that the trailer was away from Aug 2-6 and then from Aug 31 to Sep 3. Fig 2a, shows measurements until Sep 6th. Was Sep 3-6 was included in the study?
2.2. Box model configuration
L141- Does the location of the air parcel’s starting point change for each trajectory?
L142- You described the typical transport in the LA basin and how variability in the transport pattern can introduce uncertainties. Why not only include trajectories with typical transport patterns?L147- How are the initial mixing ratios determined? This is an important piece of information that is missing from the text. There is a mention of ozone concentrations being initialized using measurements from the Westchester SCAQMD site during the day (why not at night?) in section 2.2.4. Is this used for all the trajectories? This site is next to LAX (major pollutant source); how does this impact your conclusions? If the initialization is done using WRF-Chem values, how is the performance of this model? In general, how sensitive are your conclusions to the initial values you pick for your box model?
L148- “emissions encompassed by the area of the box…”. To clarify, the box moves based on the back-trajectory information. You then extract the emission rates from the inventory for that given lat/lon (8x8km box) and add that to equation 1?
So, if there are mismatches in the location of emissions, will they not be added correctly to the box model? If this is true, then this is another important limitation of this work. This points out the importance of evaluating the new emission inventory using CTM prior to running box models.L149- Is PBLH assumed to be the same throughout the trajectory as the measured values at the Pasadena site? Line 151 says observations and 3D-model estimates constrain the PBL. Which observation? Where model estimates are used? Again, very critical information is missing. Is this a reasonable assumption?
Fig1. I highly suggest adding maps of emissions by sector in the paper or in SI.
2.2.1. FLEXPART-WRF back-traj analysis
If you decide to run WRF-Chem for this study, you can use the trajectory monitoring feature in WRF-Chem and skip using FLEXPART. See here for details: https://www2.acom.ucar.edu/sites/default/files/documents/Trajectory.desc_.pdf
L165- Verryken et al. (2024) paper is in prep. Please add information regarding the WRF-Chem performance in capturing transport in this paper.L170- How many air parcels were released from the site, and at what altitude? Please add a map of back-trajectories, similar to Fig S8.
What do you mean by “parcel locations were derived as the center-of-mass from the main particle cluster”? What if the air particle cluster is spread?
With FLEXPART, you can easily create an ensemble run of air parcels and quantify transport uncertainties.L180- Another reason to only include certain hours (mid-day with filly developed boundary later) in the analysis.
2.2.2. Emissions
A lot of important information is missing from this section.
Please add details about the diurnal variability of emissions from each sector (can be a diel plot).
Please add details about the spatial variability of emissions from each sector (it can be a map), especially the cooking sector. How are population density and temporal profile for human activity used for the spatial distribution of cooking emissions?Have the cooking emissions been evaluated in any CTM model? If not, how does this limit your conclusions?
Where are oil and gas emissions? Part of fossil fuel?
2.2.4. Meteorology, dilution, and entrainment
L284- How exactly are the measurements from the Westchester site used? Please mark it on the map in Fig 1.
3. Model evaluation
I wish the method section had more details so I could review sections 3, 4, and 5 more thoroughly.3.1. Ozone.
L303- “the general agreement…” Drawing this conclusion from only NMB and R2 is not accurate. Can you compare O3 and NOX concentrations between the box model and WRF-Chem along the trajectory and at the site?Fig 2. You can exclude days after Sep 3rd if VOC obs is not available then?
Fig 3. X-axis label is not correct.
3.3. VOCs
Low ethane concentrations in the model compared to observation. Did you include oil and gas emissions in your study? Is it part of fossil fuel emissions?Formaldehyde matches observation well. Give the low bias in isoprene, probably for the wrong reasons.
4.1. Contribution of Anthro and bio VOCs to O3
In my opinion, to correctly capture the impact of all the anthropogenic sectors, you need to run a scenario with all anthropogenic emissions removed instead of using Equation 4. Same goes with 10% reduction.
Another point: In reality, you cannot turn off some pollutants from a sector (only VOCs, not NOX) and asses the contribution of this sector on O3 concentration/production or MDA8. As you mentioned, the system is non-linear; thus, including both VOC and NOx simultaneously matters. Changes in the O3 chemical regime are part of the reality and need to be studied.
L435:
Can you add details about the variability and standard deviation of the data? How many ppb is 28%?L437:
Coggon et al showed fossil fuels contribution to ozone production from AVOC at midday was 60%. This is different from the 28% contribution to MDA8 ozone in this work. Please be more precise about your conclusions in this paragraph.What is the total mass of VOC and NOX emissions in each city? Is fossil fuel emission lower in LA?
L453:
What is the contribution of each sector to OH reactivity? Adding more details on this can be valuable for the discussion.
4.2. Spatial and temporal ozone sensitivity to NOx.
Similar to scenarios in the previous section, preturbing only NOx does not show us the full picture of the impact of mitigation policies on O3 concentrations.
To draw a more comprehensive conclusion about the NOx vs VOC sensitivity in the region, you can use metrics such as FNR, LROx/LNOx, or methods as described in Vermeuel et al. 2019(https://doi.org/10.1029/2019JD030842) (they also used a box model). Box models can provide details about PO3, OH reactivity, etc. Please use this information in your discussion about the O3 regime.
L576
Why FIVE-VCP has 50% lower off-road VOCs?Figure 6.
A better comparison is the Red scenario + cooking emission.Thank you, and good luck!
Citation: https://doi.org/10.5194/egusphere-2024-1899-RC2 - AC1: 'Author Comment on egusphere-2024-1899', Chelsea Stockwell, 31 Oct 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 572 | 236 | 29 | 837 | 45 | 16 | 19 |
- HTML: 572
- PDF: 236
- XML: 29
- Total: 837
- Supplement: 45
- BibTeX: 16
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 434 | 52 |
| China | 2 | 85 | 10 |
| Germany | 3 | 43 | 5 |
| Belgium | 4 | 29 | 3 |
| France | 5 | 29 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 434
Chelsea E. Stockwell
Matthew M. Coggon
Rebecca H. Schwantes
Colin Harkins
Bert Verreyken
Congmeng Lyu
Qindan Zhu
Jessica B. Gilman
Aaron Lamplugh
Jeff Peischl
Michael A. Robinson
Patrick R. Veres
Andrew W. Rollins
Kristen Zuraski
Sunil Baidar
Shang Liu
Toshihiro Kuwayama
Steven S. Brown
Brian C. McDonald
Carsten Warneke
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1473 KB) - Metadata XML
-
Supplement
(1286 KB) - BibTeX
- EndNote
- Final revised paper