the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Nov 2022
Coupled mesoscale-LES modeling of air quality in a polluted city using WRF-LES-Chem
Abstract. To perform realistic high-resolution air quality modeling in a polluted urban area, the WRF (Weather Research and Forecasting) model is used with an embedded large-eddy simulation (LES) module and with online chemistry. As an illustration, a numerical experiment is conducted in the megacity Hong Kong, which is characterized by multi-type pollution sources as well as complex topography. The multi-resolution simulations from mesoscale to LES scales are evaluated by comparing to ozone sounding profiles and surface observations. The comparisons show that both mesoscale and LES simulations reproduce well the mean concentrations of the chemical species and their diurnal variations at the background stations. However, the mesoscale simulations largely underestimate the NOX concentrations and overestimate O3 at the roadside stations due to the coarse representation of the traffic emissions. The LES simulations improve the agreement with the measurements near the road traffic, and the LES with the highest spatial resolution (33.3 m) provides the best results. The LES simulations show more detailed structures in the spatial distributions of chemical species than the mesoscale simulations, highlighting the capability of LES to resolve high-resolution photochemical transformations in urban areas. Compared to the mesoscale model results, the LES simulations show similar evolutions in the profiles of the chemical species as a function of the boundary layer development over a diurnal cycle.
-
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
(14793 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(14793 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1208', Anonymous Referee #1, 12 Jan 2023
This article coupled mesoscale and large eddy scale models to simulate air quality in a densely populated city, shows the effect of spatial resolution on the model results and identifies the effect of turbulence on atmospheric chemistry. The content of the whole article is integrated and the model built in this study will effectively promote the air quality forecast to reach the large eddy scale. It is suggested to accept this manuscript with a minor revision. But I still have the following suggestions and questions about the article:
- The simulation period in this article is short. Although it is difficult for CFD and LES models to run a longer simulation, but as a state-of-art study, could the authors try to simulate or give some discussion about the simulation of the pollution processes, for instance, the pollution for several consecutive days and its elimination?
- How is the urban canopy scheme and its parameters set up in this study? It is suggested to clarify in the method section.
- There is no verification result of simulated and observed meteorological data in the article, and it is impossible to explain the phenomenon well in terms of meteorological factors. Will it be added or explained in the supplement?
- The article introduces the simulation results of roadside stations and ordinary stations, but it is not intuitive enough. Could the authors add time series diagrams for comparative analysis?
- Regarding the overestimation of NOx simulation and the underestimation of O3 simulation at some sites, could the authors further analyze it from the aspect of VOCs and explain it in combination with the actual industrial distribution?
- The vertical profiles in Fig 5 do not show a significant difference between mesoscale WRF and LES-WRF. In other words, the simulation accuracy of LES-WRF is not higher enough as we expected. Could the authors further show some comparison of potential temperature, water vapor mixing ratio, wind speed, wind direction, ozone mixing ratio, etc. inside the PBL or city surface layer?
- Could the authors discuss some potential bottlenecks for the use of WRF-LES-Chem in future air quality predictions?
Citation: https://doi.org/10.5194/egusphere-2022-1208-RC1 - AC1: 'Reply on RC1', Yuting Wang, 14 Apr 2023
-
RC2: 'Comment on egusphere-2022-1208', Anonymous Referee #2, 13 Mar 2023
EGUsphere-2022-1208
Coupled mesoscale-LES modelling of air quality in a polluted city using WRF-LES-Chem
Wang et alSummary
This paper presented high-resolution air quality simulation model using the WRF model with 7 nesting levels, where the innermost 3 domains are modelled through WRF-LES, using on-line chemistry with the RADM2 mechanism. The model was evaluated in Hong Kong, which has a complex topography and multi-type chemical sources. The LES model performed better than coarser models in terms of reproducing ozone concentration profiles and diurnal variations in mean concentrations at observation stations. However, the model had limitations in reproducing the NOX concentrations and overestimating O3 at roadside stations due to the coarse representation of traffic emissions. Despite this, the Authors asserted the potential of the multiscale approach (using WRF and WRF-LES) for accurate air quality forecasting, as it provided better details in pollutant distributions the explicit representation of energy-carrying turbulence structures.
General comments
Congratulations for undertaking such a challenging problem. As indicated by the authors, even in the presence of highly resolved spatiotemporal dynamics (in this case LES), the characterization of emission sources remains problematic, have reverberated our own research. The manuscript is generally well prepared, albeit there are few technical and rhetorical concerns, listed below, in addition to the following general concerns, which should be adequately addressed prior to publication.
- As the study concerns a heavily urbanized region, the type of urban canopy model that was used (in the RANS domains) should be indicated. The expectation is that these models, should provide representative surface fluxes in the urban region and allow explicit resolution in the LES domains. It might be worthwhile to provide some form of commentary on this and the impacts of using and not using an urban canopy model.
- My understanding on the emission input data presented in Section 2.2 and Figure 2 is that they are stationary (i.e., time-invariant) boundary conditions throughout the model period. While this has been rightfully pointed out as a deficiency for the study (e.g., on page 13 line 322), it would be more sensible to at least discuss or suggest the possibility of disaggregating these values based on, for instance, sector-relevant diurnal profiles (e.g., GNFR or equivalent)?
- It took me a few readings to understand sections 3.1 and 3.2 correctly. The authors might want to make these two sections more concise and clear. For instance, the authors might want to clearly specify, directly on Table 1, what species are being monitored at each station. In addition, I have to assume that “general” stations measure background concentrations. The authors should also indicate this for completeness, especially when these “general” stations are further categorized (e.g., urban, suburban, or rural background) which should also be indicated.
- In section 4.1, the authors presented the PBLH at the time of the sounding profile (13:55 local time) is made, as a form of model evaluation. Showing the PBLH as a single point measurement (in time and space) is inadequate, in my opinion, because the evolution of the PBLH in the model, and thus the vertical mixing, is not known. The discussion on vertical concentrations in Section 4.5 (and Figures 13 / 14) attempts to bring some comparison PBLH between the RANS and LES domains and use that as the basis to estimate the over- / under-predictiveness in the LES domains, but the temporal relationship of this is all but gone, largely due to how Figures 13 and 14 are presented, which makes the understanding quite difficult. I would suggest showing the diurnal profile of the PBLH in the domains considered, in addition to the vertical profiles, to aid the explanation.
- The explanation (i.e., first appearance) of abbreviated terms should be consistent. Sometimes to the full term is first referred and the corresponding abbreviations provided, while other times it is the other way around. In some instances there are no explanation provided for the abbreviated term. A quick example can be found on the first paragraph on Page 5, and the last paragraph of Page 7. I will let the authors sort this out.
- The captions for Figures 8, 10, and 12, as well as Table 3 should be expanded. While the current approach it is more succinct, it saves the forgetful reader (like yours truly) from needing to constantly refer to the corresponding captions which, given their length and the size of figure / table, can be very cumbersome. This naturally applies to other similar figure and tables on this manuscript not mentioned in this comment.
- Referral of the observational stations should be accompanied by a definite article. For example, “at Causeway Bay station” should be “at the Causeway Bay station”.
Specific comments
- Page 3 Line 76 : It should be made aware that INIFOR has restricted availability as it derives lateral meteorological profiles from proprietary data (COSMO). Instead, the authors are strongly encouraged to refer to WRF4PALM (Lin et al, GMD 14 2503-2524, 2021) as a generally available method for obtaining mesoscale meteorological and chemical boundary conditions from WRF and WRF-Chem model data, which aligns much more closely with the context of this manuscript.
- Page 5 Line 143 : A brief technical description on the cell perturbation method of Muñoz-Esparza et al in D05 should be provided, in particular, how the turbulent length scales are modelled / parameterized and how they represent the turbulent spectrum commonly encountered in atmospheric flows, different from other methods (e.g., Xie and Castro, Flow Turbul Combust 81 449-470, 2008; as implemented in Zhong et al, GMD 14 323-336, 2021).
- Page 5 Line 144 : For the reader’s benefit, the authors might want to elaborate on what “too coarse” means.
- Page 5 Line 147 : Is one hour of spin-up time sufficient for the LES domains? If this is solely due to constraints on computational resource, and are relying on the spun-up properties from the RANS domains (since they were set to 42 hours), the authors should indicate this explicitly. Otherwise, some substantial form of justification would be necessary.
- Page 7 Figure 3 : Station names on both figures are very hard to see (and I already zoomed it in to 250%). Consider the following measures: 1) apply a background color to each name label, preferably white, and introduce a transparency in the background. 2) use lines and arrows to provide additional spacing between closely clustered stations. 3) Remove lat / long indicators on both figures to maximize figure real estate (as the coordinates are already indicated in Table 1 it is not necessary to state them again graphically).
- Page 12 Line 298 : If the NOx emission dataset were, indeed, time-invariant (see corresponding general comment), then this statement of overestimated road emissions during rush hours cannot be true, as the diurnal disaggregation will very likely result in a higher emission than presented during this period. This, in turn, leaves the only interpretation, that the NOx emissions are in fact overestimated all day long, which is not the case here. Please revise this statement accordingly. Also see general comment on PBLH for further information.
- Page 12 Line 302 : Now the authors refers to a suburban station. Are they officially classified as such? Please refer to general comment on Section 3.1/3.2 for further details to ensure consistency.
- Page 27 Line 567 : As a rhetorical question, what exactly are “large” turbulent eddies?
Rhetorical remarks
- Page 5 Line 125 : Did the authors mean to write “IGBP-MODIS” or “IGBP MODIS”? If the former is intended the abbreviation should be explained together so that it does not look like a typographical error. Also see general comment on abbreviations.
- Page 7 Line 189 : “heighs” > “heights”
- Page 7 Line 177 : “Megan” > “MEGAN”
- Page 9 Line 221 : Would it not be easier to explain what theta_vs is directly, instead of explaining first theta_v, and then the subscript s, since the subscript s only exists in conjunction with theta_v?
- Page 11 Line 264 : “carbonyls” > “RCO” (to be consistent with RH and ROx in the other reactions.)
- Page 12 Lines 277-279 : The sentence with the with the multiple slashed adjective choices becomes too confusing to read. Consider rewriting to something like this: “In addition, the air parcels with low O3 values are clearly transported to higher attitudes by the updrafts produced by […]. On the other hand, the downdraft produced by […] transport the air parcels with high O3 values to the lower attitudes.”
- Page 13 Lines 320-321 : Based on the discussion, I think the authors mean to write “.. high-resolution model does not necessary provide better predictions …” as opposed to “… much better predictions”.
Citation: https://doi.org/10.5194/egusphere-2022-1208-RC2 - AC2: 'Reply on RC2', Yuting Wang, 14 Apr 2023
-
AC3: 'Comment on egusphere-2022-1208 (track-change&supplement)', Yuting Wang, 14 Apr 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-1208/egusphere-2022-1208-AC3-supplement.pdf
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1208', Anonymous Referee #1, 12 Jan 2023
This article coupled mesoscale and large eddy scale models to simulate air quality in a densely populated city, shows the effect of spatial resolution on the model results and identifies the effect of turbulence on atmospheric chemistry. The content of the whole article is integrated and the model built in this study will effectively promote the air quality forecast to reach the large eddy scale. It is suggested to accept this manuscript with a minor revision. But I still have the following suggestions and questions about the article:
- The simulation period in this article is short. Although it is difficult for CFD and LES models to run a longer simulation, but as a state-of-art study, could the authors try to simulate or give some discussion about the simulation of the pollution processes, for instance, the pollution for several consecutive days and its elimination?
- How is the urban canopy scheme and its parameters set up in this study? It is suggested to clarify in the method section.
- There is no verification result of simulated and observed meteorological data in the article, and it is impossible to explain the phenomenon well in terms of meteorological factors. Will it be added or explained in the supplement?
- The article introduces the simulation results of roadside stations and ordinary stations, but it is not intuitive enough. Could the authors add time series diagrams for comparative analysis?
- Regarding the overestimation of NOx simulation and the underestimation of O3 simulation at some sites, could the authors further analyze it from the aspect of VOCs and explain it in combination with the actual industrial distribution?
- The vertical profiles in Fig 5 do not show a significant difference between mesoscale WRF and LES-WRF. In other words, the simulation accuracy of LES-WRF is not higher enough as we expected. Could the authors further show some comparison of potential temperature, water vapor mixing ratio, wind speed, wind direction, ozone mixing ratio, etc. inside the PBL or city surface layer?
- Could the authors discuss some potential bottlenecks for the use of WRF-LES-Chem in future air quality predictions?
Citation: https://doi.org/10.5194/egusphere-2022-1208-RC1 - AC1: 'Reply on RC1', Yuting Wang, 14 Apr 2023
-
RC2: 'Comment on egusphere-2022-1208', Anonymous Referee #2, 13 Mar 2023
EGUsphere-2022-1208
Coupled mesoscale-LES modelling of air quality in a polluted city using WRF-LES-Chem
Wang et alSummary
This paper presented high-resolution air quality simulation model using the WRF model with 7 nesting levels, where the innermost 3 domains are modelled through WRF-LES, using on-line chemistry with the RADM2 mechanism. The model was evaluated in Hong Kong, which has a complex topography and multi-type chemical sources. The LES model performed better than coarser models in terms of reproducing ozone concentration profiles and diurnal variations in mean concentrations at observation stations. However, the model had limitations in reproducing the NOX concentrations and overestimating O3 at roadside stations due to the coarse representation of traffic emissions. Despite this, the Authors asserted the potential of the multiscale approach (using WRF and WRF-LES) for accurate air quality forecasting, as it provided better details in pollutant distributions the explicit representation of energy-carrying turbulence structures.
General comments
Congratulations for undertaking such a challenging problem. As indicated by the authors, even in the presence of highly resolved spatiotemporal dynamics (in this case LES), the characterization of emission sources remains problematic, have reverberated our own research. The manuscript is generally well prepared, albeit there are few technical and rhetorical concerns, listed below, in addition to the following general concerns, which should be adequately addressed prior to publication.
- As the study concerns a heavily urbanized region, the type of urban canopy model that was used (in the RANS domains) should be indicated. The expectation is that these models, should provide representative surface fluxes in the urban region and allow explicit resolution in the LES domains. It might be worthwhile to provide some form of commentary on this and the impacts of using and not using an urban canopy model.
- My understanding on the emission input data presented in Section 2.2 and Figure 2 is that they are stationary (i.e., time-invariant) boundary conditions throughout the model period. While this has been rightfully pointed out as a deficiency for the study (e.g., on page 13 line 322), it would be more sensible to at least discuss or suggest the possibility of disaggregating these values based on, for instance, sector-relevant diurnal profiles (e.g., GNFR or equivalent)?
- It took me a few readings to understand sections 3.1 and 3.2 correctly. The authors might want to make these two sections more concise and clear. For instance, the authors might want to clearly specify, directly on Table 1, what species are being monitored at each station. In addition, I have to assume that “general” stations measure background concentrations. The authors should also indicate this for completeness, especially when these “general” stations are further categorized (e.g., urban, suburban, or rural background) which should also be indicated.
- In section 4.1, the authors presented the PBLH at the time of the sounding profile (13:55 local time) is made, as a form of model evaluation. Showing the PBLH as a single point measurement (in time and space) is inadequate, in my opinion, because the evolution of the PBLH in the model, and thus the vertical mixing, is not known. The discussion on vertical concentrations in Section 4.5 (and Figures 13 / 14) attempts to bring some comparison PBLH between the RANS and LES domains and use that as the basis to estimate the over- / under-predictiveness in the LES domains, but the temporal relationship of this is all but gone, largely due to how Figures 13 and 14 are presented, which makes the understanding quite difficult. I would suggest showing the diurnal profile of the PBLH in the domains considered, in addition to the vertical profiles, to aid the explanation.
- The explanation (i.e., first appearance) of abbreviated terms should be consistent. Sometimes to the full term is first referred and the corresponding abbreviations provided, while other times it is the other way around. In some instances there are no explanation provided for the abbreviated term. A quick example can be found on the first paragraph on Page 5, and the last paragraph of Page 7. I will let the authors sort this out.
- The captions for Figures 8, 10, and 12, as well as Table 3 should be expanded. While the current approach it is more succinct, it saves the forgetful reader (like yours truly) from needing to constantly refer to the corresponding captions which, given their length and the size of figure / table, can be very cumbersome. This naturally applies to other similar figure and tables on this manuscript not mentioned in this comment.
- Referral of the observational stations should be accompanied by a definite article. For example, “at Causeway Bay station” should be “at the Causeway Bay station”.
Specific comments
- Page 3 Line 76 : It should be made aware that INIFOR has restricted availability as it derives lateral meteorological profiles from proprietary data (COSMO). Instead, the authors are strongly encouraged to refer to WRF4PALM (Lin et al, GMD 14 2503-2524, 2021) as a generally available method for obtaining mesoscale meteorological and chemical boundary conditions from WRF and WRF-Chem model data, which aligns much more closely with the context of this manuscript.
- Page 5 Line 143 : A brief technical description on the cell perturbation method of Muñoz-Esparza et al in D05 should be provided, in particular, how the turbulent length scales are modelled / parameterized and how they represent the turbulent spectrum commonly encountered in atmospheric flows, different from other methods (e.g., Xie and Castro, Flow Turbul Combust 81 449-470, 2008; as implemented in Zhong et al, GMD 14 323-336, 2021).
- Page 5 Line 144 : For the reader’s benefit, the authors might want to elaborate on what “too coarse” means.
- Page 5 Line 147 : Is one hour of spin-up time sufficient for the LES domains? If this is solely due to constraints on computational resource, and are relying on the spun-up properties from the RANS domains (since they were set to 42 hours), the authors should indicate this explicitly. Otherwise, some substantial form of justification would be necessary.
- Page 7 Figure 3 : Station names on both figures are very hard to see (and I already zoomed it in to 250%). Consider the following measures: 1) apply a background color to each name label, preferably white, and introduce a transparency in the background. 2) use lines and arrows to provide additional spacing between closely clustered stations. 3) Remove lat / long indicators on both figures to maximize figure real estate (as the coordinates are already indicated in Table 1 it is not necessary to state them again graphically).
- Page 12 Line 298 : If the NOx emission dataset were, indeed, time-invariant (see corresponding general comment), then this statement of overestimated road emissions during rush hours cannot be true, as the diurnal disaggregation will very likely result in a higher emission than presented during this period. This, in turn, leaves the only interpretation, that the NOx emissions are in fact overestimated all day long, which is not the case here. Please revise this statement accordingly. Also see general comment on PBLH for further information.
- Page 12 Line 302 : Now the authors refers to a suburban station. Are they officially classified as such? Please refer to general comment on Section 3.1/3.2 for further details to ensure consistency.
- Page 27 Line 567 : As a rhetorical question, what exactly are “large” turbulent eddies?
Rhetorical remarks
- Page 5 Line 125 : Did the authors mean to write “IGBP-MODIS” or “IGBP MODIS”? If the former is intended the abbreviation should be explained together so that it does not look like a typographical error. Also see general comment on abbreviations.
- Page 7 Line 189 : “heighs” > “heights”
- Page 7 Line 177 : “Megan” > “MEGAN”
- Page 9 Line 221 : Would it not be easier to explain what theta_vs is directly, instead of explaining first theta_v, and then the subscript s, since the subscript s only exists in conjunction with theta_v?
- Page 11 Line 264 : “carbonyls” > “RCO” (to be consistent with RH and ROx in the other reactions.)
- Page 12 Lines 277-279 : The sentence with the with the multiple slashed adjective choices becomes too confusing to read. Consider rewriting to something like this: “In addition, the air parcels with low O3 values are clearly transported to higher attitudes by the updrafts produced by […]. On the other hand, the downdraft produced by […] transport the air parcels with high O3 values to the lower attitudes.”
- Page 13 Lines 320-321 : Based on the discussion, I think the authors mean to write “.. high-resolution model does not necessary provide better predictions …” as opposed to “… much better predictions”.
Citation: https://doi.org/10.5194/egusphere-2022-1208-RC2 - AC2: 'Reply on RC2', Yuting Wang, 14 Apr 2023
-
AC3: 'Comment on egusphere-2022-1208 (track-change&supplement)', Yuting Wang, 14 Apr 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-1208/egusphere-2022-1208-AC3-supplement.pdf
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 386 | 213 | 18 | 617 | 7 | 6 |
- HTML: 386
- PDF: 213
- XML: 18
- Total: 617
- BibTeX: 7
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 193 | 30 |
| China | 2 | 97 | 15 |
| Germany | 3 | 75 | 11 |
| undefined | 4 | 44 | 6 |
| India | 5 | 18 | 2 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 193
Cited
2 citations as recorded by crossref.
Domingo Muñoz-Esparza
Jianing Dai
Cathy W. Y. Li
Pablo Lichtig
Roy C. W. Tsang
Chun-Ho Liu
Guy P. Brasseur
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(14793 KB) - Metadata XML