the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Apr 2023
Dynamics-based estimates of decline trend with fine temporal variations in China's PM2.5 emissions
Abstract. Timely, continuous, and dynamics-based estimates of PM2.5 emissions with a high temporal resolution can be objectively and optimally obtained by assimilating observed surface PM2.5 concentrations using flow-dependent error statistics. Annual PM2.5 emissions in China have consistently decreased of approximately 3 % to 5 % from 2017 to 2020. Significant PM2.5 emission reductions occurred frequently in regions with large PM2.5 emissions. COVID-19 could cause a significant reduction of PM2.5 emissions in the north China plain and northeast of China in 2020. The magnitudes of PM2.5 emissions were greater in the winter than in the summer. PM2.5 emissions show an obvious diurnal variation that varies significantly with the season and urban population. Improved representations of PM2.5 emissions across time scales can benefit emission inventory, regulation policy and emission trading schemes, particularly for especially for high temporal resolution air quality forecasting and policy response to severe haze pollutions or rare human events with significant socioeconomic impacts.
-
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
(7445 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(7445 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-755', Anonymous Referee #1, 30 May 2023
This study assimilates hourly air quality observations to provide detailed primary PM2.5 emissions inventory across China. Some insights about inter-annual, monthly and diurnal variation of the derived emissions, as well as impacts from COVID-19 lockdown are discussed. While the scope of this study is definitely within the scop of ACP, I have a serious reservation of publishing the manuscript at the present form. The main issue is about inadequate/insufficient discussion of the results and missing information of fundamentals which I will outline later. I am happy to re-review the revised manuscript, and will support the publication of this paper, if the following concerns can be adequately addressed.
Major points:
1) Assimilation approach regarding secondary PM2.5: it is hard for me to understand (Page 6-7) how "the impact of the secondary PM2.5 is ignored" during the assimilation. Secondary aerosol makes significant contributions (>50%) to observed PM2.5. This fact almost applies to all the hourly PM2.5 observations. Observations that are very remote and during less photochemically reactive periods may be less constituted by secondary aerosols, but I think these records make minor contributions to the air quality records this paper used which distribute mostly over eastern China. WRF-Chem simulation of PM2.5 includes important secondary species, which is fundamental to successfully capture PM2.5 spatiotemporal variations in China. So please further clarify:
What emission species are exactly optimized during the assimilation? Do you only constrain PM2.5 emissions and let PM2.5 precursors (SO2, NH3, NOx, VOC) to stay the same as the a priori? If so, how uncertain are the constrained PM2.5 emissions, if the a priori precursor emissions are incorrect and they make significant contributions to the observed PM2.5?
Lines 142 suggests that "hourly observed ... PM10, PM2.5, SO2, NO2, O3, and CO" are assimilated. So maybe these concentrations are used to also constrain the PM2.5 precursor emissions at the same time? If so, you should also briefly present the results of optimized precursor emissions, and how they affect the constraints on PM2.5 emissions.
Overall, I do not understand how successful constraints on PM2.5 emissions can be achieved with "ignoring secondary aerosol".
2) A missing piece of information is showing the improvement in model simulation/prediction after the assimilation? How is the agreement of simulated PM2.5 vs. observations improved after the assimilation? If you also constrained precursor emissions, comparison vs. observed air quality species other than PM2.5 should also be provided.
Another suggestion is to compare your results vs. the updated MEIC that has extended to more recent years (not just 2016). This discussion is especially necessary considering that MEIC contains detailed bottom-up information. Differences of the derived inter-annual and inter-month variations of emissions vs. MEIC will be indicative where and when MEIC might be unrepresentative and why.
3) Section 5: This section attributes the difference of 2020 emissions vs. the previous years to the COVID-19 lockdown. However, the 2020 emission vs. 2019 is not entirely stronger than the difference between other neighboring years (e.g., Figure 3 and Table 1). So how much of the 2020-2019 emission difference can also be contributed by continuous environmental policies (as discussed in Section 3)? Overall, Figure 11 does not provide continuous signal of lockdown either, as some provinces show temporary increases at certain phases. The authors discuss New Year firework. But how can they only occur in certain provinces (and not occurring during the first several days of New Year)? Overall, the attribution of 2020-2019 emission difference to COVID-19 lockdown and the relevant discussions about temporal changes of these differences are weak.
Specific comments:
1) Abstract: key quantitative results should be presented. The current form of abstract is too qualitative and less informative. Line 24-27 reads redundant and irrelevant, and is suggested to be replaced with a more concise sentence stating the significance of these results.
2) Why are observations before 2016 not used in the assimilation?
3) Line 145: what is the spatial autocorrelation before and after the selection of stations?
4) Line 196: What are the "weather effects" referring to?
5) Figure 3b: the winter seasons show a sharp change from increases in 2017 to decreases in 2018. Is it related to the coal ban for residential heating since the 2017-2018 winter?
6) Line 215-216: As I understand, the centralized heating system in North China has a fixed date of turning-on and turning-off during each heating season. So a sudden drop of emissions from March to April looks reasonable to me. Do you suggest that the turning-off date is variable in different places to smooth-out the differences, or residential heating does not contribute that much to the total emissions variations between these two months?
7) Section 4: some recent bottom-up developments have more details about diurnal emission variations (e.g., Du et al., 10.5194/acp-20-2839-2020, 2020, Figure 1). Discussion about comparison of your results vs. these recent diurnal profiles can be insightful.
8) Line 310: missing words here.
Citation: https://doi.org/10.5194/egusphere-2023-755-RC1 -
RC2: 'Reply on RC1', Anonymous Referee #2, 31 May 2023
This manuscript proposes an ensemble Kalman smoother to constrain the PM 2.5 emissions by incorporating the information of PM 2.5 observations. Results based on 5-year cycling assimilation provide quantitively estimates for annual and monthly variations of the
PM 2.5 emission. By assimilating the observations with the ensemble Kalman smoother, the influences of COVID are clearly displayed. Moreover, diurnal variations of the PM 2.5 emission for each month are provided, which can be a valuable contribution to the PM 2.5 forecast. The manuscript proposed an advanced data assimilation method to update the PM 2.5 emissions by both present and future PM 2.5 observations. Overall it is well written and presented. It could be very beneficial to the community of chemistry data assimilation. I have several minor comments below.1. An EnKS is proposed to update the emission along with the concentration. Are both the emission and the concentration updated by future observations?
2. The lagged length for EnKS is an important factor because it determines how many future observations are applied to constrain the current state. The lagged length K is set to 6 in this study. How this parameter is determined?
3. It is interesting to see the quick influences of COVID on PM2.5 (Figure 11). Can such a DA system be practical for real-time operations?Citation: https://doi.org/10.5194/egusphere-2023-755-RC2 -
AC2: 'Reply on RC2', Meigen Zhang, 20 Aug 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-755/egusphere-2023-755-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Meigen Zhang, 20 Aug 2023
-
AC1: 'Reply on RC1', Meigen Zhang, 20 Aug 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-755/egusphere-2023-755-AC1-supplement.pdf
-
RC2: 'Reply on RC1', Anonymous Referee #2, 31 May 2023
-
RC3: 'Comment on egusphere-2023-755', Anonymous Referee #3, 02 Jun 2023
This study develops and presents top-down estimates of high temporal (up to hourly) PM2.5 emissions using an ENKS. The goal of this study is adequate for a publication in this journal, and I expect this research would inspire other researchers and would lead to further advances in top-down estimates of pollutant emissions. However, there are some parts that can be misleading or need to be clarified. Also, I agree with two other reviewers who raised important issues, which are as follows.
1. It is not clear how secondary PM2.5 is ignored. Did you just assume that the increments or differences resulting from PM2.5 assimilation are all attributed to PM2.5 emissions? Or, the formation of secondary PM2.5 is ignored in the WRF-Chem modeling? I agree with the first reviewer who emphasized the importance of secondary PM2.5 formation. The authors should demonstrate how the ignorance of secondary PM2.5 can be justified and what potential errors are.
2. Related to the first comment and also commented by the second reviewer. Is PM2.5 concentration also updated? Or just PM2.5 emission?
3. I seriously doubt the diurnal variations in PM2.5 emission (Fig. 6). Many studies assume that high emission rates during daytime (working hours) and low emission rates during nighttime as in Fig. 8. I think the highest emission rate in the morning in Fig. 6 is attributable to 1) high emission during rush hours and 2) shallow boundary layer. In other words, the diurnal variations in PM2.5 emissions estimated in this study do include the effects of time-varying boundary layer (and height). So, the effects of boundary layer are not separated from the emission estimates. We would expect high emission rates in the afternoon (working hours) and also during the late afternoon (evening rush hours). Because boundary layer height is generally highest in the late afternoon, the estimated emission rates in the late afternoon are too low (Fig. 6). I think monthly emission estimates or yearly estimates would be fine because the diurnally varying boundary layer is all averaged out at monthly and yearly time scales. To verify this, you can take a closer look at emission rates near industrial complex where diurnal variations in emissions are expected to be small (e.g., power plants, steel and cement companies …). I understand the horizontal grid of 45 km is too coarse to examine this, but I expect that there are some regions where many factories are concentrated.
4. Related to the effects of meteorology (or boundary layer), I would suggest some extra experiments (also related to the first reviewer’s 4th minor comment asking “weather effects”). Let’s fix anthropogenic emissions all the time, and only consider time-varying meteorology. Assume the observations that will be assimilated here are the model outputs with the same emissions but time-varying meteorology (not real observations). Then, assimilate these fake observations (actually model outputs) and estimate emissions. Would your estimated emissions be almost identical to the prior emissions that are fixed with time? I’m curious if your estimated emissions depend on / are influenced by meteorology. A month-long simulation would be enough for this type of simulation.
5. line 306-307. Did you mean that emissions in 2019 were higher than those in 2020? I think in 2020 there were few firework activities due to the lockdown. If this is true, the color for BTH and SCR in Fig 11e should be blue (lower emissions in 2020 than in 2019). If not, please clarify. In addition, some studies highlighted that the PM2.5 concentration during Feb. 2020 is due to unfavorable meteorological condition in the BTH region (Sulaymon et al. 2021). Le et al. (2020) also showed that for the severe haze in northern China during the lockdown is due to 1) anomalously high humidity that promoted aerosol heterogeneous chemistry, 2) stagnant airflow 3) uninterrupted emissions from power plants and petrochemical facilities, and 4) secondary aerosol formation associated with increased ozone.
6. Constant emissions can be misleading (line 149, line 183, line 233…). Did you mean time invariant emissions? That is, emission rates do not vary with time at all at a grid cell. If so, I would recommend saying time-invariant emissions or constant emissions with time because the constant emissions can be interpreted as spatially homogeneous emissions.
7. Figure 9. The x-axis label should be date, not time, right? And recommend representing mm/dd, format.
References
Sulaymon et al. 2021. Persistent high PM2.5 pollution driven by unfavorable meteorological conditions during the COVID-19 lockdown period in the Beijing-Tianjin-Hebei region, China. Environmental Research. https://doi.org/10.1016/j.envres.2021.111186
Le et al. 2020. Unexpected air pollution with marked emission reductions during the COVID-19 outbreak in China, Science. DOI: 10.1126/science.abb7431Citation: https://doi.org/10.5194/egusphere-2023-755-RC3 -
AC3: 'Reply on RC3', Meigen Zhang, 20 Aug 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-755/egusphere-2023-755-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Meigen Zhang, 20 Aug 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-755', Anonymous Referee #1, 30 May 2023
This study assimilates hourly air quality observations to provide detailed primary PM2.5 emissions inventory across China. Some insights about inter-annual, monthly and diurnal variation of the derived emissions, as well as impacts from COVID-19 lockdown are discussed. While the scope of this study is definitely within the scop of ACP, I have a serious reservation of publishing the manuscript at the present form. The main issue is about inadequate/insufficient discussion of the results and missing information of fundamentals which I will outline later. I am happy to re-review the revised manuscript, and will support the publication of this paper, if the following concerns can be adequately addressed.
Major points:
1) Assimilation approach regarding secondary PM2.5: it is hard for me to understand (Page 6-7) how "the impact of the secondary PM2.5 is ignored" during the assimilation. Secondary aerosol makes significant contributions (>50%) to observed PM2.5. This fact almost applies to all the hourly PM2.5 observations. Observations that are very remote and during less photochemically reactive periods may be less constituted by secondary aerosols, but I think these records make minor contributions to the air quality records this paper used which distribute mostly over eastern China. WRF-Chem simulation of PM2.5 includes important secondary species, which is fundamental to successfully capture PM2.5 spatiotemporal variations in China. So please further clarify:
What emission species are exactly optimized during the assimilation? Do you only constrain PM2.5 emissions and let PM2.5 precursors (SO2, NH3, NOx, VOC) to stay the same as the a priori? If so, how uncertain are the constrained PM2.5 emissions, if the a priori precursor emissions are incorrect and they make significant contributions to the observed PM2.5?
Lines 142 suggests that "hourly observed ... PM10, PM2.5, SO2, NO2, O3, and CO" are assimilated. So maybe these concentrations are used to also constrain the PM2.5 precursor emissions at the same time? If so, you should also briefly present the results of optimized precursor emissions, and how they affect the constraints on PM2.5 emissions.
Overall, I do not understand how successful constraints on PM2.5 emissions can be achieved with "ignoring secondary aerosol".
2) A missing piece of information is showing the improvement in model simulation/prediction after the assimilation? How is the agreement of simulated PM2.5 vs. observations improved after the assimilation? If you also constrained precursor emissions, comparison vs. observed air quality species other than PM2.5 should also be provided.
Another suggestion is to compare your results vs. the updated MEIC that has extended to more recent years (not just 2016). This discussion is especially necessary considering that MEIC contains detailed bottom-up information. Differences of the derived inter-annual and inter-month variations of emissions vs. MEIC will be indicative where and when MEIC might be unrepresentative and why.
3) Section 5: This section attributes the difference of 2020 emissions vs. the previous years to the COVID-19 lockdown. However, the 2020 emission vs. 2019 is not entirely stronger than the difference between other neighboring years (e.g., Figure 3 and Table 1). So how much of the 2020-2019 emission difference can also be contributed by continuous environmental policies (as discussed in Section 3)? Overall, Figure 11 does not provide continuous signal of lockdown either, as some provinces show temporary increases at certain phases. The authors discuss New Year firework. But how can they only occur in certain provinces (and not occurring during the first several days of New Year)? Overall, the attribution of 2020-2019 emission difference to COVID-19 lockdown and the relevant discussions about temporal changes of these differences are weak.
Specific comments:
1) Abstract: key quantitative results should be presented. The current form of abstract is too qualitative and less informative. Line 24-27 reads redundant and irrelevant, and is suggested to be replaced with a more concise sentence stating the significance of these results.
2) Why are observations before 2016 not used in the assimilation?
3) Line 145: what is the spatial autocorrelation before and after the selection of stations?
4) Line 196: What are the "weather effects" referring to?
5) Figure 3b: the winter seasons show a sharp change from increases in 2017 to decreases in 2018. Is it related to the coal ban for residential heating since the 2017-2018 winter?
6) Line 215-216: As I understand, the centralized heating system in North China has a fixed date of turning-on and turning-off during each heating season. So a sudden drop of emissions from March to April looks reasonable to me. Do you suggest that the turning-off date is variable in different places to smooth-out the differences, or residential heating does not contribute that much to the total emissions variations between these two months?
7) Section 4: some recent bottom-up developments have more details about diurnal emission variations (e.g., Du et al., 10.5194/acp-20-2839-2020, 2020, Figure 1). Discussion about comparison of your results vs. these recent diurnal profiles can be insightful.
8) Line 310: missing words here.
Citation: https://doi.org/10.5194/egusphere-2023-755-RC1 -
RC2: 'Reply on RC1', Anonymous Referee #2, 31 May 2023
This manuscript proposes an ensemble Kalman smoother to constrain the PM 2.5 emissions by incorporating the information of PM 2.5 observations. Results based on 5-year cycling assimilation provide quantitively estimates for annual and monthly variations of the
PM 2.5 emission. By assimilating the observations with the ensemble Kalman smoother, the influences of COVID are clearly displayed. Moreover, diurnal variations of the PM 2.5 emission for each month are provided, which can be a valuable contribution to the PM 2.5 forecast. The manuscript proposed an advanced data assimilation method to update the PM 2.5 emissions by both present and future PM 2.5 observations. Overall it is well written and presented. It could be very beneficial to the community of chemistry data assimilation. I have several minor comments below.1. An EnKS is proposed to update the emission along with the concentration. Are both the emission and the concentration updated by future observations?
2. The lagged length for EnKS is an important factor because it determines how many future observations are applied to constrain the current state. The lagged length K is set to 6 in this study. How this parameter is determined?
3. It is interesting to see the quick influences of COVID on PM2.5 (Figure 11). Can such a DA system be practical for real-time operations?Citation: https://doi.org/10.5194/egusphere-2023-755-RC2 -
AC2: 'Reply on RC2', Meigen Zhang, 20 Aug 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-755/egusphere-2023-755-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Meigen Zhang, 20 Aug 2023
-
AC1: 'Reply on RC1', Meigen Zhang, 20 Aug 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-755/egusphere-2023-755-AC1-supplement.pdf
-
RC2: 'Reply on RC1', Anonymous Referee #2, 31 May 2023
-
RC3: 'Comment on egusphere-2023-755', Anonymous Referee #3, 02 Jun 2023
This study develops and presents top-down estimates of high temporal (up to hourly) PM2.5 emissions using an ENKS. The goal of this study is adequate for a publication in this journal, and I expect this research would inspire other researchers and would lead to further advances in top-down estimates of pollutant emissions. However, there are some parts that can be misleading or need to be clarified. Also, I agree with two other reviewers who raised important issues, which are as follows.
1. It is not clear how secondary PM2.5 is ignored. Did you just assume that the increments or differences resulting from PM2.5 assimilation are all attributed to PM2.5 emissions? Or, the formation of secondary PM2.5 is ignored in the WRF-Chem modeling? I agree with the first reviewer who emphasized the importance of secondary PM2.5 formation. The authors should demonstrate how the ignorance of secondary PM2.5 can be justified and what potential errors are.
2. Related to the first comment and also commented by the second reviewer. Is PM2.5 concentration also updated? Or just PM2.5 emission?
3. I seriously doubt the diurnal variations in PM2.5 emission (Fig. 6). Many studies assume that high emission rates during daytime (working hours) and low emission rates during nighttime as in Fig. 8. I think the highest emission rate in the morning in Fig. 6 is attributable to 1) high emission during rush hours and 2) shallow boundary layer. In other words, the diurnal variations in PM2.5 emissions estimated in this study do include the effects of time-varying boundary layer (and height). So, the effects of boundary layer are not separated from the emission estimates. We would expect high emission rates in the afternoon (working hours) and also during the late afternoon (evening rush hours). Because boundary layer height is generally highest in the late afternoon, the estimated emission rates in the late afternoon are too low (Fig. 6). I think monthly emission estimates or yearly estimates would be fine because the diurnally varying boundary layer is all averaged out at monthly and yearly time scales. To verify this, you can take a closer look at emission rates near industrial complex where diurnal variations in emissions are expected to be small (e.g., power plants, steel and cement companies …). I understand the horizontal grid of 45 km is too coarse to examine this, but I expect that there are some regions where many factories are concentrated.
4. Related to the effects of meteorology (or boundary layer), I would suggest some extra experiments (also related to the first reviewer’s 4th minor comment asking “weather effects”). Let’s fix anthropogenic emissions all the time, and only consider time-varying meteorology. Assume the observations that will be assimilated here are the model outputs with the same emissions but time-varying meteorology (not real observations). Then, assimilate these fake observations (actually model outputs) and estimate emissions. Would your estimated emissions be almost identical to the prior emissions that are fixed with time? I’m curious if your estimated emissions depend on / are influenced by meteorology. A month-long simulation would be enough for this type of simulation.
5. line 306-307. Did you mean that emissions in 2019 were higher than those in 2020? I think in 2020 there were few firework activities due to the lockdown. If this is true, the color for BTH and SCR in Fig 11e should be blue (lower emissions in 2020 than in 2019). If not, please clarify. In addition, some studies highlighted that the PM2.5 concentration during Feb. 2020 is due to unfavorable meteorological condition in the BTH region (Sulaymon et al. 2021). Le et al. (2020) also showed that for the severe haze in northern China during the lockdown is due to 1) anomalously high humidity that promoted aerosol heterogeneous chemistry, 2) stagnant airflow 3) uninterrupted emissions from power plants and petrochemical facilities, and 4) secondary aerosol formation associated with increased ozone.
6. Constant emissions can be misleading (line 149, line 183, line 233…). Did you mean time invariant emissions? That is, emission rates do not vary with time at all at a grid cell. If so, I would recommend saying time-invariant emissions or constant emissions with time because the constant emissions can be interpreted as spatially homogeneous emissions.
7. Figure 9. The x-axis label should be date, not time, right? And recommend representing mm/dd, format.
References
Sulaymon et al. 2021. Persistent high PM2.5 pollution driven by unfavorable meteorological conditions during the COVID-19 lockdown period in the Beijing-Tianjin-Hebei region, China. Environmental Research. https://doi.org/10.1016/j.envres.2021.111186
Le et al. 2020. Unexpected air pollution with marked emission reductions during the COVID-19 outbreak in China, Science. DOI: 10.1126/science.abb7431Citation: https://doi.org/10.5194/egusphere-2023-755-RC3 -
AC3: 'Reply on RC3', Meigen Zhang, 20 Aug 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-755/egusphere-2023-755-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Meigen Zhang, 20 Aug 2023
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 331 | 114 | 22 | 467 | 9 | 11 |
- HTML: 331
- PDF: 114
- XML: 22
- Total: 467
- BibTeX: 9
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Zhen Peng
Lili Lei
Zhe-Min Tan
Aijun Ding
Xingxia Kou
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(7445 KB) - Metadata XML