the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Dec 2023
Diagnosing Ozone-NOx-VOCs-Aerosols Sensitivity to Uncover Urban-nonurban Discrepancies in Shandong, China using Transformer-based High-resolution Air Pollution Estimations
Abstract. Narrowing surface ozone disparities between urban and nonurban areas escalate health risks in densely populated urban zones. A comprehensive understanding of the impact of ozone photochemistry processes on this transition remains constrained by our knowledge of aerosol effects and the spatial availability of surface monitoring. Here we developed a novel deep learning framework, which could perceive spatiotemporal dynamics from adjacent grids by multidimensional self-attention operation, integrating multi-sources data to estimate daily 500 m surface ozone, nitrogen dioxide (NO2) and fine particulate matter (PM2.5) concentrations. Subsequently, three distinct ozone formation regimes linked with its precursors, aerosols, and meteorology were delineated through an interpretable machine learning method. The evaluations of the framework exhibited average out-of-sample cross-validation coefficient of determination of 0.96, 0.92 and 0.95 for ozone, NO2 and PM2.5, respectively. In 2020, urban ozone levels in Shandong surpassed those in nonurban due to a more pronounced decrease in ozone in the latter where PM2.5 is the dominant anthropogenic driver. The ozone sensitivity to volatile organic compounds (VOCs), the dominant regime in urban areas, was observed to shift towards a NOx-limited when extended to rural areas. A third ‘aerosol-inhibited’ regime was identified in the Jiaodong Peninsula, where the uptake of hydroperoxyl radicals onto aerosols suppressed ozone production under low NOx levels during summertime. The reduction of PM2.5 would increase the sensitivity of ozone to VOCs, necessitating more stringent VOC emission abatement for urban ozone mitigation. Our case study demonstrates the critical need for advanced modeling approaches providing finer spatially resolved estimations.
-
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
(18459 KB)
-
Supplement
(16790 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(18459 KB) - Metadata XML
-
Supplement
(16790 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2640', Anonymous Referee #1, 26 Dec 2023
General comments
The manuscript titled “Contribution of Cooking Emissions to the Urban Volatile Organic Compounds in Las Vegas, NV” evaluates the contribution of commercial and residential cooking emissions to urban VOCs using mobile laboratory and ground site observations in Las Vegas, NV with supplemental observations made Los Angeles, CA and Boulder, CO. The authors found that long-chain aldehydes were significantly enhanced in restaurant plumes and long-chain fatty acids were also associated with cooking emissions, while the result of the contribution of the cooking source based on observation was not consistent with the result based on emission inventory. The results could contribute to the targeted reduction of the VOCs emissions to a certain extent. However, some statements should be more clearly and precisely, and some adjustments could facilitate the understanding of the paper. Therefore, some major issues need to be revised in the manuscript before it is considered for publication in this journal.
Detailed comments
- Line 54: The abbreviations tend to be capitalized and the ‘oxygenated VOCs’ are usually abbreviated to ‘OVOCs’.
- The description of the PTR-ToF-MS could be reduced because it is a common instrument and the technique was not improved in this study.
- The temporal changes of the mobile sources were irregular and the contribution during period 2 was lower than that during period 1. Could you explain the reason?
- The results of the inventory may be added a spatial distribution, which could contribute to the confirmation of the accuracy of the inventory.
- Conclusion should be general and succinct. The quotations may be moved to results or discussion chapters.
- Some punctuation-mark issues should be paid attention and modified. For example, ‘30 June– 27 July, 2021’ should be ‘30 June – 27 July, 2021’.
Citation: https://doi.org/10.5194/egusphere-2023-2640-RC1 -
AC1: 'Reply on RC2', Chenliang Tao, 20 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2640/egusphere-2023-2640-AC1-supplement.pdf
-
RC2: 'Comment on egusphere-2023-2640', Anonymous Referee #1, 05 Jan 2024
I am very sorry that I uploaded comments on other reserch earlier. Here is the refree for this article.
The authors developed a deep learning framework to estimate surface O3, NO2, and PM2.5 concentrations, and investigated urban-nonurban difference and ozone-NOx-VOCs-aerosols sensitivity for ozone pollution in Shandong. This manuscript needs to be revised before it can be published.
1.Theme: There are two logics based on the title, abstract, conclusion, and the last paragraph of the introduction. One theme is "High Resolution Air Pollution Estimation", while the spatial characterization of pollution and urban-rural differences are further investigated in order to illustrate the value of the application of this deep learning framework. The other theme is to study the spatial characteristics of pollution in Shandong, and a deep learning approach is used. Authors should consider the perspective of the writing.
2.The study is divided into two main parts: one on estimating ozone concentrations and studying urban-rural differences, and the other on ozone sensitivity. The logical relationship between the two parts is not very coherent. Ozone sensitivity does not adequately explain the variations and differences in ozone concentrations between urban and rural areas and in different years. In other words, if ozone concentrations are not estimated, it does not seem to affect the results of the ozone sensitivity study.
3.Ozone formation regimes: From Fig.8, the NOx-limited regime dominates, especially Fig.8D shows that the proportion of NOx-limited is almost 1.0, which is not quite consistent with the authors' conclusions.
4.Specification of figures: Do Figures 7A and B share a colorbar to indicate O3 concentration? In fact, Figure 7B contains the information from Figure 7A. In Figure 7D, what does the arrow next to PM2.5 mean? Figure 8C is a legend for Figures 8A and B, making it difficult to understand. In Figure 9, it is not appropriate to represent one variable in dots and one in columns as they are of the same kind.
Citation: https://doi.org/10.5194/egusphere-2023-2640-RC2 -
AC1: 'Reply on RC2', Chenliang Tao, 20 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2640/egusphere-2023-2640-AC1-supplement.pdf
-
AC1: 'Reply on RC2', Chenliang Tao, 20 Feb 2024
-
RC3: 'Comment on egusphere-2023-2640', Anonymous Referee #2, 13 Jan 2024
This study developed a novel spatiotemporal deep learning model for concurrent prediction of three air pollutants (ozone, NO2, PM2.5). The authors used the generated fine-scale concentrations to assess urban-nonurban differences and ozone-NOx-VOCs-aerosols sensitivity in Shandong, China. To facilitate the analysis, interpretable machine learning was employed to handle nonlinearity and isolate impacts of drivers relating to ozone photochemistry. The methodology is solid, and the findings are important for the development of ozone control strategies, though a few issues remain.
1. Line 259: Please explain the possible reason why NO2 has significantly lower out-of-site CV-R2 (0.75) than ozone and PM2.5 (>0.9), note that the out-of-sample CV results are comparable across all pollutants?
2. Lines 265-266: In evaluating stability and robustness of the model, it would be interesting to see if the CNEMC-trained model can obtain local concentration variations and interpretation outcomes similar to that from the CNEMC+SDEM-trained model.
3. Lines 322-323: The time span of the training data should be given, as that information is important to understand whether the good agreements between measurements and estimations reflect fitting or prediction performance.
4. How many monitoring stations are there in urban areas? A map highlighting the urban and nonurban areas is recommended for intuitive understanding.
5. There is a lack of validation for the XGBoost model, given that reliability of interpretation outcomes should be based on the model with high accuracy.
6. Please provide more explanations for the SHAP interaction values. The statement “lower NO2 … could diminish the formation of ozone under high PM2.5 concentrations” (Line 456) is difficult to follow. In Figure 7e, lower NO2 and negative PM2.5-NO2 SHAP interaction values are observed at lower PM2.5 levels.
7. Figure 8d shows that the NOx-limited regime dominates in urban areas. Please confirm.
8. Minor typos and grammar errors need to be corrected. For example, Line 355: the upper right area of E, M, and U; Line 370: are shown in Figure 5, etc.
Citation: https://doi.org/10.5194/egusphere-2023-2640-RC3 -
AC2: 'Reply on RC3', Chenliang Tao, 20 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2640/egusphere-2023-2640-AC2-supplement.pdf
-
AC2: 'Reply on RC3', Chenliang Tao, 20 Feb 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2640', Anonymous Referee #1, 26 Dec 2023
General comments
The manuscript titled “Contribution of Cooking Emissions to the Urban Volatile Organic Compounds in Las Vegas, NV” evaluates the contribution of commercial and residential cooking emissions to urban VOCs using mobile laboratory and ground site observations in Las Vegas, NV with supplemental observations made Los Angeles, CA and Boulder, CO. The authors found that long-chain aldehydes were significantly enhanced in restaurant plumes and long-chain fatty acids were also associated with cooking emissions, while the result of the contribution of the cooking source based on observation was not consistent with the result based on emission inventory. The results could contribute to the targeted reduction of the VOCs emissions to a certain extent. However, some statements should be more clearly and precisely, and some adjustments could facilitate the understanding of the paper. Therefore, some major issues need to be revised in the manuscript before it is considered for publication in this journal.
Detailed comments
- Line 54: The abbreviations tend to be capitalized and the ‘oxygenated VOCs’ are usually abbreviated to ‘OVOCs’.
- The description of the PTR-ToF-MS could be reduced because it is a common instrument and the technique was not improved in this study.
- The temporal changes of the mobile sources were irregular and the contribution during period 2 was lower than that during period 1. Could you explain the reason?
- The results of the inventory may be added a spatial distribution, which could contribute to the confirmation of the accuracy of the inventory.
- Conclusion should be general and succinct. The quotations may be moved to results or discussion chapters.
- Some punctuation-mark issues should be paid attention and modified. For example, ‘30 June– 27 July, 2021’ should be ‘30 June – 27 July, 2021’.
Citation: https://doi.org/10.5194/egusphere-2023-2640-RC1 -
AC1: 'Reply on RC2', Chenliang Tao, 20 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2640/egusphere-2023-2640-AC1-supplement.pdf
-
RC2: 'Comment on egusphere-2023-2640', Anonymous Referee #1, 05 Jan 2024
I am very sorry that I uploaded comments on other reserch earlier. Here is the refree for this article.
The authors developed a deep learning framework to estimate surface O3, NO2, and PM2.5 concentrations, and investigated urban-nonurban difference and ozone-NOx-VOCs-aerosols sensitivity for ozone pollution in Shandong. This manuscript needs to be revised before it can be published.
1.Theme: There are two logics based on the title, abstract, conclusion, and the last paragraph of the introduction. One theme is "High Resolution Air Pollution Estimation", while the spatial characterization of pollution and urban-rural differences are further investigated in order to illustrate the value of the application of this deep learning framework. The other theme is to study the spatial characteristics of pollution in Shandong, and a deep learning approach is used. Authors should consider the perspective of the writing.
2.The study is divided into two main parts: one on estimating ozone concentrations and studying urban-rural differences, and the other on ozone sensitivity. The logical relationship between the two parts is not very coherent. Ozone sensitivity does not adequately explain the variations and differences in ozone concentrations between urban and rural areas and in different years. In other words, if ozone concentrations are not estimated, it does not seem to affect the results of the ozone sensitivity study.
3.Ozone formation regimes: From Fig.8, the NOx-limited regime dominates, especially Fig.8D shows that the proportion of NOx-limited is almost 1.0, which is not quite consistent with the authors' conclusions.
4.Specification of figures: Do Figures 7A and B share a colorbar to indicate O3 concentration? In fact, Figure 7B contains the information from Figure 7A. In Figure 7D, what does the arrow next to PM2.5 mean? Figure 8C is a legend for Figures 8A and B, making it difficult to understand. In Figure 9, it is not appropriate to represent one variable in dots and one in columns as they are of the same kind.
Citation: https://doi.org/10.5194/egusphere-2023-2640-RC2 -
AC1: 'Reply on RC2', Chenliang Tao, 20 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2640/egusphere-2023-2640-AC1-supplement.pdf
-
AC1: 'Reply on RC2', Chenliang Tao, 20 Feb 2024
-
RC3: 'Comment on egusphere-2023-2640', Anonymous Referee #2, 13 Jan 2024
This study developed a novel spatiotemporal deep learning model for concurrent prediction of three air pollutants (ozone, NO2, PM2.5). The authors used the generated fine-scale concentrations to assess urban-nonurban differences and ozone-NOx-VOCs-aerosols sensitivity in Shandong, China. To facilitate the analysis, interpretable machine learning was employed to handle nonlinearity and isolate impacts of drivers relating to ozone photochemistry. The methodology is solid, and the findings are important for the development of ozone control strategies, though a few issues remain.
1. Line 259: Please explain the possible reason why NO2 has significantly lower out-of-site CV-R2 (0.75) than ozone and PM2.5 (>0.9), note that the out-of-sample CV results are comparable across all pollutants?
2. Lines 265-266: In evaluating stability and robustness of the model, it would be interesting to see if the CNEMC-trained model can obtain local concentration variations and interpretation outcomes similar to that from the CNEMC+SDEM-trained model.
3. Lines 322-323: The time span of the training data should be given, as that information is important to understand whether the good agreements between measurements and estimations reflect fitting or prediction performance.
4. How many monitoring stations are there in urban areas? A map highlighting the urban and nonurban areas is recommended for intuitive understanding.
5. There is a lack of validation for the XGBoost model, given that reliability of interpretation outcomes should be based on the model with high accuracy.
6. Please provide more explanations for the SHAP interaction values. The statement “lower NO2 … could diminish the formation of ozone under high PM2.5 concentrations” (Line 456) is difficult to follow. In Figure 7e, lower NO2 and negative PM2.5-NO2 SHAP interaction values are observed at lower PM2.5 levels.
7. Figure 8d shows that the NOx-limited regime dominates in urban areas. Please confirm.
8. Minor typos and grammar errors need to be corrected. For example, Line 355: the upper right area of E, M, and U; Line 370: are shown in Figure 5, etc.
Citation: https://doi.org/10.5194/egusphere-2023-2640-RC3 -
AC2: 'Reply on RC3', Chenliang Tao, 20 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2640/egusphere-2023-2640-AC2-supplement.pdf
-
AC2: 'Reply on RC3', Chenliang Tao, 20 Feb 2024
Peer review completion
Journal article(s) based on this preprint
Data sets
Surface Ozone, NO2, and PM2.5 Concentrations Estimated by the Deep Learning model (Air Transformer) based on Satellite data. Chenliang Tao https://doi.org/10.5281/zenodo.10071408
Model code and software
Air-Transformer Chenliang Tao https://github.com/myles-tcl/Air-Transformer
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 313 | 107 | 24 | 444 | 31 | 9 | 15 |
- HTML: 313
- PDF: 107
- XML: 24
- Total: 444
- Supplement: 31
- BibTeX: 9
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Chenliang Tao
Yanbo Peng
Qingzhu Zhang
Yuqiang Zhang
Bing Gong
Qiao Wang
Wenxing Wang
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(18459 KB) - Metadata XML
-
Supplement
(16790 KB) - BibTeX
- EndNote
- Final revised paper