A machine learning approach to quantify meteorological drivers of recent ozone pollution in China
- 1School of Environmental Sciences, University of East Anglia, Norwich, NR47TJ, UK
- 2National Centre for Atmospheric Sciences, University of East Anglia, Norwich, NR47TJ, UK
- 3Grantham Institute – Climate Change and the Environment, Department of Physics, and the Data Science Institute, Imperial College London, London SW7 2AZ, UK
- 1School of Environmental Sciences, University of East Anglia, Norwich, NR47TJ, UK
- 2National Centre for Atmospheric Sciences, University of East Anglia, Norwich, NR47TJ, UK
- 3Grantham Institute – Climate Change and the Environment, Department of Physics, and the Data Science Institute, Imperial College London, London SW7 2AZ, UK
Abstract. Surface ozone concentrations have been increasing in many regions of China for the past few years, in contrast to policy-driven declines in other key air pollutants such as particulate matter. While the central role of meteorology in modulating ozone pollution is widely acknowledged, its quantitative contribution remains highly uncertain. Here, we use a data-driven machine learning approach to assess the impacts of meteorology on surface ozone variations in China for the years 2015 to 2019, considering the months of highest ozone pollution from April to October. To quantify the importance of various meteorological driver variables, we apply non-linear random forest regression (RFR) and linear ridge regression (RR) to learn relationships between meteorological variability and surface ozone in China, and contrast the results to those obtained with the widely used multiple linear regression (MLR) and stepwise MLR. We show that RFR outperforms the three linear methods when predicting ozone using only local meteorological predictor variables. This implies the importance of non-linear relationships between local meteorological factors and ozone, which are not captured by linear regression algorithms. In addition, we find that including non-local meteorological predictors can further improve the modelling skill of RR, particularly for Southern China, highlighting the importance of large-scale meteorological phenomena for ozone pollution in that region. Overall, RFR and RR are in close agreement concerning the leading meteorological drivers behind regional ozone pollution. For example, we find that temperature variations are the dominant meteorological driver for ozone pollution in Northern China (e.g., Beijing Tianjin Hebei region), whereas variations in relative humidity are the most important factor in Southern China (e.g., Pearl River Delta). Variability in surface solar radiation modulates photochemistry but was not considered as such in previous controlling factor analyses, and is found to be the most important predictor in the Yangtze River Delta and Sichuan Basin regions. In general, our analysis underlines that hot and dry weather conditions with high sunlight intensity are strongly related to high ozone pollution across China. This further validates our novel approach to quantify the central role of meteorology. By contrasting our meteorological ozone predictions with ozone measurements between 2015 and 2019, we estimate that almost half of the observed ozone trends across China might have been caused by meteorological variabilities on average. We highlight that these insights are of particular importance given possible increases in the frequency and intensity of weather extremes such as heatwaves under climate change.
- Preprint
(2323 KB) -
Supplement
(2866 KB) - BibTeX
- EndNote
Xiang Weng et al.
Status: closed
-
RC1: 'Comment on acp-2021-1075', Anonymous Referee #1, 13 Mar 2022
The authors present an exploration of observational and model data to explore the contribution of meteorology to ozone pollution events. They expand the period of the year considered in this analysis (April-October), compared to previous work in the same area. The way in which the authors explore multiple linear and non-linear techniques before justifying their chosen approach is exemplary. The manuscript feels more focused on the techniques taken rather than the application, but this is fine considering the novel techniques being used as well as the broad interest for their study. The approaches and general interest make this manuscript appropriate for ACP.
No notable changes are requested to the manuscript.
- AC1: 'Reply on RC1', Xiang Weng, 14 Apr 2022
-
RC2: 'Comment on acp-2021-1075', Anonymous Referee #2, 14 Mar 2022
Review of “A machine learning approach to quantify meteorological drivers of recent ozone pollution in China” by Weng et al.
The authors applied advanced statistical approaches to identify major meteorological drivers of ozone pollution over China. They also compared their results with the multiple linear regression methods. Moreover, they also found that by including the large-scale meteorology, their model skill will be improved relative to the model constructed by only local meteorological variables. Based on these regression models, the authors demonstrate the models’ capability and advantage in the understanding of major meteorological drivers of ozone pollution and in the isolation of meteorological effects from observed ozone trends.
Ozone pollution issue in China is of great concern. This study adds insights into the better understanding of recent ozone trend in China. I think the major novelty of this work is its new methods. However, several places should be improved in order to highlight this strength. Please find my comments below.
The Abstract should be revised. Firstly, I am surprised that there are almost no quantitative sentences to show the advantages of the machine learning approach. It prevents the readers from easily understanding the contribution of this work in current version. Then, a large fraction of the Abstract is the description of leading meteorological variables. However, these results are not new and have been reported by a lot of studies previously.
Section 3.4. As I mentioned above, this section shows a lot of previously-reported knowledge on the major meteorological drivers of ozone pollution. I suggest the authors to make this concise and to highlight your new findings.
Section 3.5. Another contribution of this work is the quantification of meteorological role in ozone trends. However, I failed to find the comparison between machine learning method and MLR method in this Section. At least, I suggest the authors to list the MLR-based estimates in Table 3.
The “recent” in the Title is not clear. It is better to be replaced by “2015-2019”.
L41: Brief information on VOC emission changes should be added.
- AC2: 'Reply on RC2', Xiang Weng, 14 Apr 2022
Status: closed
-
RC1: 'Comment on acp-2021-1075', Anonymous Referee #1, 13 Mar 2022
The authors present an exploration of observational and model data to explore the contribution of meteorology to ozone pollution events. They expand the period of the year considered in this analysis (April-October), compared to previous work in the same area. The way in which the authors explore multiple linear and non-linear techniques before justifying their chosen approach is exemplary. The manuscript feels more focused on the techniques taken rather than the application, but this is fine considering the novel techniques being used as well as the broad interest for their study. The approaches and general interest make this manuscript appropriate for ACP.
No notable changes are requested to the manuscript.
- AC1: 'Reply on RC1', Xiang Weng, 14 Apr 2022
-
RC2: 'Comment on acp-2021-1075', Anonymous Referee #2, 14 Mar 2022
Review of “A machine learning approach to quantify meteorological drivers of recent ozone pollution in China” by Weng et al.
The authors applied advanced statistical approaches to identify major meteorological drivers of ozone pollution over China. They also compared their results with the multiple linear regression methods. Moreover, they also found that by including the large-scale meteorology, their model skill will be improved relative to the model constructed by only local meteorological variables. Based on these regression models, the authors demonstrate the models’ capability and advantage in the understanding of major meteorological drivers of ozone pollution and in the isolation of meteorological effects from observed ozone trends.
Ozone pollution issue in China is of great concern. This study adds insights into the better understanding of recent ozone trend in China. I think the major novelty of this work is its new methods. However, several places should be improved in order to highlight this strength. Please find my comments below.
The Abstract should be revised. Firstly, I am surprised that there are almost no quantitative sentences to show the advantages of the machine learning approach. It prevents the readers from easily understanding the contribution of this work in current version. Then, a large fraction of the Abstract is the description of leading meteorological variables. However, these results are not new and have been reported by a lot of studies previously.
Section 3.4. As I mentioned above, this section shows a lot of previously-reported knowledge on the major meteorological drivers of ozone pollution. I suggest the authors to make this concise and to highlight your new findings.
Section 3.5. Another contribution of this work is the quantification of meteorological role in ozone trends. However, I failed to find the comparison between machine learning method and MLR method in this Section. At least, I suggest the authors to list the MLR-based estimates in Table 3.
The “recent” in the Title is not clear. It is better to be replaced by “2015-2019”.
L41: Brief information on VOC emission changes should be added.
- AC2: 'Reply on RC2', Xiang Weng, 14 Apr 2022
Xiang Weng et al.
Xiang Weng et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
509 | 199 | 20 | 728 | 38 | 7 | 11 |
- HTML: 509
- PDF: 199
- XML: 20
- Total: 728
- Supplement: 38
- BibTeX: 7
- EndNote: 11
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1