the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Feb 2023
Remotely Sensed and Surface Measurement Derived Mass-Conserving Inversion of Daily High-Resolution NOx Emissions and Inferred Combustion Technologies in Energy Rich Northern China
Abstract. This work presents a new model free inversion estimation framework using daily TROPOMI NO2 columns and observed fluxes from the continuous emissions monitoring systems (CEMS) to quantify three years of daily-scale emissions of NOx at 0.05°×0.05° over Shanxi Province, a major world-wide energy producing and consuming region. The NOx emissions, day-to-day variability, and uncertainty on a climatological basis are computed to be 1.83, 1.01, and 1.06 Tg per year respectively. The highest emissions are concentrated in the lower Fen River valley, which accounts for 25 % of the area, 52 % of the NOx emissions, and 72 % of CEMS sources. Two major forcing factors (10th to 90th percentile) are horizontal transport distance per day (66–666 km) and lifetime of NOx (6.7–18.4 h). Both of these values are consistent with NOx emissions to both the surface layer and the free troposphere. The third forcing factor, the ratio of NOx / NO2, on a pixel-by-pixel basis is demonstrated to have a significant correlation with the combustion temperature and energy efficiency of large energy consuming sources. Specifically, thermal power plants, cement, and iron and steel companies have a relatively high NOx / NO2 ratio, while coking, industrial boilers, and aluminium oxide show relatively low ratio. Variance maximization is applied to daily TROPOMI NO2 columns identifies three significant modes, and successfully attributes them both spatially and temporally to (a) this work’s computed emissions, (b) remotely sensed TROPOMI UVAI, and (c) computed transport based on TROPOMI NO2.
-
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
(6513 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(6513 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2', Anonymous Referee #2, 15 Feb 2023
This paper uses a mass balance approach to interpret observations of TROPOMI NO2 columns and CEMS NOx emission flux, enabling the propagation of measured flux at in situ sites to build seamless monthly NOx emission estimates across Shanxi Province, China. It also further interprets the variability of derived model parameters in their framework that represent NOx/NO2 emission ratio (alpha1), NOx lifetime (alpha2) and horizontal advection rate (alpha3). To my knowledge, this is a pioneering study that evaluates and interprets an established space-borne emission estimation method, which has rarely been validated using densely distributed flux observations. The paper is overall well-written with sound methods and results. At the same time, some critical details are missing, and structural changes of the contents are needed. I support the publication of this manuscript, provided that the following comments can be addressed.
Major comments:
1) The MFIEF approach is largely originated from similar box-modeling ideas in previous studies (Beirle et al., 10.1126/sciadv.aax98, 2019; Kong et al., 10.5194/acp-19-12835-2019, 2019), while differs to some extent in details about assumptions on each source/sink process. Line 84-86 presents these studies in the overall "previous study" category, which seem to diminish this connection. I suggest the authors to introduce these approaches in the end of this paragraph, and acknowledge the similarity of idea used in this paper. Also, certain discussions about the uniqueness and capabilities of MFIEF (e.g., using measured emissions, fitting variable parameters that were fixed in previous approaches, etc.) relative to these previous methods should also be added in the Introduction and/or Discussion Section.
2) The UVAI is used as a proxy of OH and NOx lifetime to interpret EOF2. However, UVAI is a measure of aerosol absorption, while the actual radiation flux reach surface is also sensitive to aerosol scattering, cloud extinction, and solar angles. I did not find existing literature reporting strong correlation between UVAI and OH or NOx lifetime, so it is not convincing for me to justify the interpretation related with Figure 11. Please provide stronger evidence to justify the use of UVAI, or switch to use other parameters (e.g., alpha2?).
3) Section 3.3: besides introducing the total uncertainties, contributions from each factor should also be included. One particularly important source is the performance of the fitting and the consequent errors in each parameter. This is the most fundamental information to evaluate the fidelity of the MFIEF framework. How much variability of observed VNO2 and ENOx can be explained by Eq. (3)? As the fitting is performed including all observations, is it unbiased for all months and grids?
4) Section 3.4: The derived emissions are representative of ambient fluxes (instead of initial emissions from the furnace), so the rapid NO-NO2 conversion and consequently the NOx/NO2 ratio is dependent on not only the combustion environment factors discussed in the manuscript, but also ambient chemistry (e.g., photolysis rate and ozone concentration). I assume the latter factor might be more important in driving the seasonal variations in Figs. 12 and 14. Due to the lack of full consideration of all driving factors as well as the lack of outstanding hotspot of alpha1 from certain month or factory, the current discussion of alpha1 in this section is relatively more conjectural than the other part of the paper. My overall suggestion is to greatly reduce the amount of discussion and focus on 1-2 most convincing observations that can be concluded from existing data, with acknowledgement of various factors driving alpha1 variability and precluding a full explanation of all revealed variabilities.
Specific comments:
1) Line 22-23: As outlined before, the statement of "significant correlation with combustion temperature and energy efficiency" might be too strong here.
2) Line 39: delete "are more serious".
3) Line 48: delete "(2015, 2020)".
4) Citations in the Introduction Section:
Line 42: should also cite Zhang et al., 10.1073/pnas.1907956116, 2020; Wang et al., 10.1073/pnas.2007513117, 2020; Li et al., 10.1016/j.scitotenv.2021.150011, 2022; Wei et al., 10.5194/acp-23-1511-2023, 2023.
Line 54: (Beirle et al., 2011) is not a paper studying NOx forming aerosol.
Line 56: Besides China, some other regional inventories (e.g., McDonald et al., 10.1021/es401034z, 2013; Xing et al., 10.5194/acp-13-7531-2013, 2013) might be worth citing.
Line 92: can cite (Zheng et al., 10.5194/acp-18-14095-2018, 2018) for MEIC.
5) Line 85-86: As outlined before, should clarify that your approach improves these assumptions to some extent.
6) Line 91-92: These two are bottom-up inventories, so should follow after Line 73? I do not see clear connection of this sentence with the previous text.
7) Line 98: What idea from bottom-up inventory is used in your approach?
8) Line 107: As outlined before, using UVAI as a proxy of UV radiation seems not appropriate.
9) Fig. 3c and 4c: set log-scale for x-axis might increase the readability of the figure. Also, 28% of days are absent in 2019 so would that affect the sampling and representativeness of data in Fig. 3?
10) Line 211: this is true for daytime and locations with strong NOx emissions only. See (Kenagy et al., 10.1029/2018JD028736, 2018) for nighttime sinks, and (Romer Present et al., 10.5194/acp-20-267-2020, 2020) for possible significant daytime sink via reactions with RO2. As Eq. 3 relates VNO2 at afternoon overpass to 24-h mean emissions, the lifetime should also reflect all hours during the day.
11) Equation 3: Since VNO2 is a snapshot of afternoon overpass while ENOx is 24-h average, so alpha1-alpha3 all contain the conversion from overpass time to 24-h mean. Should acknowledge this fact.
12) Fig. 5: As outlined before, alpha1 is not just determined by type of source.
13) Fig. 6: a scatter plot of Fig. 6a vs. Fig. 3a will provide an insight about how representative Eq. 3 is. Certain locations with strong emissions while unmeasured by Fig. 3a should also be discussed (e.g., are these exactly locations of missed stationary sources?).
14) Fig. 8: What spatial extent is used to calculate the city-mean emissions? If the range is too small, the difference between MEIC and CEMS could be dominated by dilution by the large grid cell.
15) Fig. 9: Are the spatial distribution of EOF1 correlating with that of alpha1? How about EOF2 vs. alpha2? EOF3 vs. divergence?
16) Sections 3.1 and 3.2: Alpha1-alpha3 all exhibit certain spatial and temporal variabilities. What are the implications on previous methods that have simpler (e.g., fixed) assumptions?
17) Line 462: should mention possible benefits from Geostationary instruments that can be promising to resolve the expectedly strong diurnal variability of alpha1-alpha3.
Citation: https://doi.org/10.5194/egusphere-2023-2-RC1 - AC1: 'Reply on RC1', Jason Cohen, 04 Apr 2023
-
RC2: 'Comment on egusphere-2023-2', Anonymous Referee #3, 13 Apr 2023
General comments:
This study presents a new model-free method to constrain NOx emissions using TROPOMI NO2 and ERA-5 wind data. The new method is based on mass balance theory and considers the NOx/NO2 ratio, NOx lifetime and NOx transport. Based on this approach, daily NOx emissions over Shanxi province are estimated during 2019-2021. Some comments should be addressed before its publication. Additionally, the authors apply EOF to TROPOMI NO2 and relate the first three PCs to NOx emissions, UVAI, and NOx transport. The method and conclusions are important, but some comments should be addressed before its publication.
Specific comments:
- When TROPOMI NO2 is not available due to cloud or other reasons, how do you deal with it? My understanding is when TROPOMI NO2 is not available, Eq. 3 does not work. And How the missing data affect the estimated emission inventory?
- In Sect 3.1, why a priori emission inventory (CEMS or MEIC) is needed to estimate new emission inventory? According to Eq. 3, it does not require a priori emission inventory.
- How EOF is applied to the daily TROPOMI NO2 columns when data are not available in some grids?
- I’m curious why the seasonal variation of NOx lifetime shown in Fig. 13b is so small. For example, Lamsal et al. (2010) estimated that the lifetime is NOx is 7.6 h in summer and 17.8 h in winter, while this study showed that lifetime is ~12 h regardless of season.
Lamsal, L. N., Martin, R. V., van Donkelaar, A., Celarier, E. A., Bucsela, E. J., Boersma, K. F., Dirksen, R., Luo, C., & Wang, Y. (2010). Indirect validation of tropospheric nitrogen dioxide retrieved from the OMI satellite instrument: Insight into the seasonal variation of nitrogen oxides at northern midlatitudes. Journal of Geophysical Research: Atmospheres, 115(D5). https://doi.org/10.1029/2009JD013351
- The authors concluded that “Thirdly, the general variability in geography, month of the year, and years before and after COVID-19 are all consistent with what is known.”, while readers cannot find any analysis that is related to COVID-19 in the manuscript.
Technical corrections:
Line 202: the first three PC account for less than 50%
Line 26: NO2 columns identifies -> NO2 columns, which facilitates to identify
Line 47: However -> Moreover
Line 49: also impacting -> that impact
Line 51: Nitrogen Monoxide -> nitric oxide
Line 60: statistics on -> statistics for
Line 67: differences -> the differences
Line 84-85: Kong et al. (Kong et al., 2019) and Beirle et al. 85 (Beirle et al., 2019) -> Kong et al. (2019) and Beirle et al. (2019)
Line 88: costly -> cost
Line 106-109: “The fact that …… variations observed.” This sentence is not easy to understand; it is better to rewrite.
Line 126: Henk Eskes, 2021 -> Henk et al., 2021
Line 126-128: “Furthermore, …… being discarded.” This sentence is not easy to understand; it is better to rewrite.
Line 130: 2021are - > 2021 are
Line 149: as discharged -> emitted
Line 151: NOx concentration measuring -> measuring NOx concentrations
Line 163: 24 is convert -> 24 is used to
Line 168: highest -> the highest
Line 179: uniformity -> uniformly
Line 201: (Björnsson and Venegas, 1997) and (Cohen, 2014) -> Björnsson and Venegas (1997) and Cohen (2014)
Line 208: transport to -> transport
Line 222: basis. -> basis
Line 225: and not -> rather than
Line 232: 𝛼1 𝛼2 and 𝛼3 -> 𝛼1, 𝛼2, and 𝛼3
Line 259: area. -> area,
Line 387: Fig. 12(a) -> Fig. 12a
Citation: https://doi.org/10.5194/egusphere-2023-2-RC2 -
AC2: 'Reply on RC2', Jason Cohen, 10 May 2023
Responses (italics) to Reviewer’s comments (normal text)
General comments:
This study presents a new model-free method to constrain NOx emissions using TROPOMI NO2 and ERA-5 wind data. The new method is based on mass balance theory and considers the NOx/NO2 ratio, NOx lifetime and NOx transport. Based on this approach, daily NOx emissions over Shanxi province are estimated during 2019-2021. Some comments should be addressed before its publication. Additionally, the authors apply EOF to TROPOMI NO2 and relate the first three PCs to NOx emissions, UVAI, and NOx transport. The method and conclusions are important, but some comments should be addressed before its publication.
Specific comments:
- When TROPOMI NO2 is not available due to cloud or other reasons, how do you deal with it? My understanding is when TROPOMI NO2 is not available, Eq. 3 does not work. And how the missing data affect the estimated emission inventory?
This paper introduces a new methodology and makes a first attempt on the combine use of NO2 column loadings and high spatial and temporal frequency observations of ground emissions, within the confines of a first order approximation to the overall mass balance framework. You are correct in that when and where there is missing data, that the emissions cannot be calculated at that exact place and time. We have examined the PDFs of the output emissions at each location and found that they are relatively smooth. For these reasons, unless the missing NO2 observation were statistically very high or very low compared with the other values that already exist, they would not make a large difference in the overall emissions. However, including more observations from other existing and new observation platforms, or using other remotely sensed species in tandem will also help to improve the emissions estimate. Thank you for this suggestion, as it provides a path for future work.
- In Sect 3.1, why a priori emission inventory (CEMS or MEIC) is needed to estimate new emission inventory? According to 3, it does not require a priori emission inventory.
In order to fit the first order terms approximating thermodynamics α1, chemistry α2, and transport α3 in Eq. 3, an initial guess of emissions is required to complete the multiple linear regression. This then allows the distributions of the parameters α1, α2, and α3 to be subsequently used in Eq. 3 to calculate the final emissions. It also allows for error analysis, since the fitted terms themselves have a range of possible solutions. In this work, two different emissions a priori were selected, with the goal being to demonstrate what differences this would have on the computed emissions.
This procedure is similar to how chemical transport models (including GEOS-Chem, WRF-Chem, etc.) have their initial uncertain variables fitted. The major differences being in this work the variables are sufficient simple so as to be flexible and presented in an open way. This allows for a wider range of possible emissions datasets to work within the model environment, which may not be possible with more heavily fitted or constrained modeling approaches. We believe the work herein demonstrates robustness as an entire system.
- How EOF is applied to the daily TROPOMI NO2 columns when data are not available in some grids?
When TROPOMI NO2 columns are not available in some grids, the climatological average value in that grid is assigned in order to compute the EOF. The grid is also tagged and after the EOF is computed, the grid in space and time is reset to NaN, following Cohen (2014).
- I’m curious why the seasonal variation of NOx lifetime shown in Fig. 13b is so small. For example, Lamsal et al. (2010) estimated that the lifetime is NOx is 7.6 h in summer and 17.8 h in winter, while this study showed that lifetime is ~12 h regardless of season.
Lamsal, L. N., Martin, R. V., van Donkelaar, A., Celarier, E. A., Bucsela, E. J., Boersma, K. F., Dirksen, R., Luo, C., & Wang, Y. (2010). Indirect validation of tropospheric nitrogen dioxide retrieved from the OMI satellite instrument: Insight into the seasonal variation of nitrogen oxides at northern midlatitudes. Journal of Geophysical Research: Atmospheres, 115(D5). https://doi.org/10.1029/2009JD013351
The median values of NOx lifetime do demonstrate a range from 9.0 hours to 14.7 hours in different months. The 10th and 90th percentile values match with your reference paper quite well, being 7.1 hours and 18.1 hours respectively. The results in this work are based on the total column values, which includes temperature, UV, climate, and aerosols which are observed in Shanxi. Based on the results herein, the largest values are found in June or July and the smallest values are found in September. This is due to the complex local conditions and forcing factors including the complex boundary layer height, the variable aerosol loading, cloudiness, and other factors.
- The authors concluded that “Thirdly, the general variability in geography, month of the year, and years before and after COVID-19 are all consistent with what is known.”, while readers cannot find any analysis that is related to COVID-19 in the manuscript.
The results herein show clearly that the emissions before COVID-19 were higher than after COVID-19. In specific, the time series shows that while there is a variation as a function of the time of the year, there is also a disturbance in this variation due to the timing of onset of COVID-19. What is important is that the emissions results match well with what is known by the community in terms of month-to-month changes, and geographic diversity, variability, and consistency across different industrial sources and under different oxidative and transport conditions. We have reorganized here in the following way:
“Thirdly, the variability of emissions in terms of different geographic location, source types, special events which changed the emissions levels (such as the onset of COVID-19), and general oxidative, photochemical, and transport conditions of the atmosphere on a monthly-scale, are all consistent with what is known.”
Technical corrections:
Line 202: the first three PC account for less than 50%
The first point is that the community acknowledges there is an uncertainty in TROPOMI observations of NO2 which ranges as high as 30% to 50%. In this case, 30% to 50% of the PCs will be representative of this uncertainty, meaning their signal pattern while mathematically correct, is physically meaningless. Therefore, the results herein represent nearly all of the remaining variability. The three spatial modes [EOF1, EOF2, and EOF3] contribute 29.4%, 8.4%, and 4.4% respectively (accounting for 42.2%), while the fourth mode onward all contribute less than 4.0%. Next, the first three modes all have a high degree of correlation with known underlying driving phenomenon, while the fourth mode and onward show no such relationship. There are many other factors that affect the pollutant column loading in the atmosphere, and if we had more data to analyze, a way to bring in more variables, or a way to reduce the uncertainty, we would also like to search for and work more on attribution. Thank you for helping to carefully guide and clarify our thought process.
Line 26: NO2 columns identifies -> NO2 columns, which facilitates to identify
Thank you, it has been modified.
Line 47: However -> Moreover
Thank you, it has been modified.
Line 49: also impacting -> that impact
Thank you, it has been modified.
Line 51: Nitrogen Monoxide -> nitric oxide
Thank you, it has been modified.
Line 60: statistics on -> statistics for
Thank you, it has been modified to “statistics representing”.
Line 67: differences -> the differences
Thank you, it has been modified.
Line 84-85: Kong et al. (Kong et al., 2019) and Beirle et al. 85 (Beirle et al., 2019) -> Kong et al. (2019) and Beirle et al. (2019)
Thank you, it has been modified.
Line 88: costly -> cost
Thank you, “costly to run” has been modified to “computationally intensive”.
Line 106-109: “The fact that …… variations observed.” This sentence is not easy to understand; it is better to rewrite.
We have reorganized here in the following way.
“This method has been used in different situation such as over different months, over multi-year changes in the environment, under different actinic flux and atmospheric oxidation conditions, under complex meteorological domains, and over sources which are both thermodynamically stable as well as unstable. That permits this study to explore the full range of variations”.
Line 126: Henk Eskes, 2021 -> Henk et al., 2021
The reference format has been modified.
Line 126-128: “Furthermore, …… being discarded.” This sentence is not easy to understand; it is better to rewrite.
We have reorganized here in the following way.
“Furthermore, an additional filter is applied to set all individual gird of NO2 column which is less than 1.4×1015 molec cm-2 to be NaN. This is done to avoid issues where the observed signal may be smaller than the uncertainty of the signal itself (J.H.G.M Van Geffen, 2021; Qin et al, 2022)”.
Line 130: 2021are - > 2021 are
Thank you, it has been modified.
Line 149: as discharged -> emitted
Thank you, it has been modified.
Line 151: NOx concentration measuring -> measuring NOx concentrations
Thank you, it has been modified.
Line 163: 24 is convert -> 24 is used to
Thank you, it has been modified to “24 is used to convert units from hours to days”.
Line 168: highest -> the highest
Thank you, it has been modified.
Line 179: uniformity -> uniformly
Thank you, it has been modified.
Line 201: (Björnsson and Venegas, 1997) and (Cohen, 2014) -> Björnsson and Venegas (1997) and Cohen (2014)
Thank you, it has been modified.
Line 208: transport to -> transport
Thank you, it has been modified.
Line 222: basis. -> basis
Thank you, it has been modified.
Line 225: and not -> rather than
Thank you, it has been modified.
Line 232: 𝛼1 𝛼2 and 𝛼3 -> 𝛼1, 𝛼2, and 𝛼3
Thank you, it has been modified in the whole paper.
Line 259: area. -> area,
Thank you, it has been modified.
Line 387: Fig. 12(a) -> Fig. 12a
Thank you, it has been modified in the whole paper.
References
Cohen, J. B.: Quantifying the occurrence and magnitude of the Southeast Asian fire climatology, Environ. Res. Lett., 9, https://doi.org/10.1088/1748-9326/9/11/114018, 2014.
J.H.G.M. van Geffen, H. J. Eskes, K.F. Boersma and J.P. Veefkind: TROPOMI ATBD of the total and tropospheric NO2 data products, Open File Rep., 2021.
Qin, K., Shi, J., He, Q., Deng, W., Wang, S., Liu, J., and Cohen, J. B.: New Model-Free Daily Inversion of NOx Emissions using TROPOMI (MCMFE-NOx): Deducing a See-Saw of Halved Well Regulated Sources and Doubled New Sources, ESS Open Archive [preprint], https://doi.org/10.1002/essoar.10512010.1, July 26, 2022.
Citation: https://doi.org/10.5194/egusphere-2023-2-AC2
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2', Anonymous Referee #2, 15 Feb 2023
This paper uses a mass balance approach to interpret observations of TROPOMI NO2 columns and CEMS NOx emission flux, enabling the propagation of measured flux at in situ sites to build seamless monthly NOx emission estimates across Shanxi Province, China. It also further interprets the variability of derived model parameters in their framework that represent NOx/NO2 emission ratio (alpha1), NOx lifetime (alpha2) and horizontal advection rate (alpha3). To my knowledge, this is a pioneering study that evaluates and interprets an established space-borne emission estimation method, which has rarely been validated using densely distributed flux observations. The paper is overall well-written with sound methods and results. At the same time, some critical details are missing, and structural changes of the contents are needed. I support the publication of this manuscript, provided that the following comments can be addressed.
Major comments:
1) The MFIEF approach is largely originated from similar box-modeling ideas in previous studies (Beirle et al., 10.1126/sciadv.aax98, 2019; Kong et al., 10.5194/acp-19-12835-2019, 2019), while differs to some extent in details about assumptions on each source/sink process. Line 84-86 presents these studies in the overall "previous study" category, which seem to diminish this connection. I suggest the authors to introduce these approaches in the end of this paragraph, and acknowledge the similarity of idea used in this paper. Also, certain discussions about the uniqueness and capabilities of MFIEF (e.g., using measured emissions, fitting variable parameters that were fixed in previous approaches, etc.) relative to these previous methods should also be added in the Introduction and/or Discussion Section.
2) The UVAI is used as a proxy of OH and NOx lifetime to interpret EOF2. However, UVAI is a measure of aerosol absorption, while the actual radiation flux reach surface is also sensitive to aerosol scattering, cloud extinction, and solar angles. I did not find existing literature reporting strong correlation between UVAI and OH or NOx lifetime, so it is not convincing for me to justify the interpretation related with Figure 11. Please provide stronger evidence to justify the use of UVAI, or switch to use other parameters (e.g., alpha2?).
3) Section 3.3: besides introducing the total uncertainties, contributions from each factor should also be included. One particularly important source is the performance of the fitting and the consequent errors in each parameter. This is the most fundamental information to evaluate the fidelity of the MFIEF framework. How much variability of observed VNO2 and ENOx can be explained by Eq. (3)? As the fitting is performed including all observations, is it unbiased for all months and grids?
4) Section 3.4: The derived emissions are representative of ambient fluxes (instead of initial emissions from the furnace), so the rapid NO-NO2 conversion and consequently the NOx/NO2 ratio is dependent on not only the combustion environment factors discussed in the manuscript, but also ambient chemistry (e.g., photolysis rate and ozone concentration). I assume the latter factor might be more important in driving the seasonal variations in Figs. 12 and 14. Due to the lack of full consideration of all driving factors as well as the lack of outstanding hotspot of alpha1 from certain month or factory, the current discussion of alpha1 in this section is relatively more conjectural than the other part of the paper. My overall suggestion is to greatly reduce the amount of discussion and focus on 1-2 most convincing observations that can be concluded from existing data, with acknowledgement of various factors driving alpha1 variability and precluding a full explanation of all revealed variabilities.
Specific comments:
1) Line 22-23: As outlined before, the statement of "significant correlation with combustion temperature and energy efficiency" might be too strong here.
2) Line 39: delete "are more serious".
3) Line 48: delete "(2015, 2020)".
4) Citations in the Introduction Section:
Line 42: should also cite Zhang et al., 10.1073/pnas.1907956116, 2020; Wang et al., 10.1073/pnas.2007513117, 2020; Li et al., 10.1016/j.scitotenv.2021.150011, 2022; Wei et al., 10.5194/acp-23-1511-2023, 2023.
Line 54: (Beirle et al., 2011) is not a paper studying NOx forming aerosol.
Line 56: Besides China, some other regional inventories (e.g., McDonald et al., 10.1021/es401034z, 2013; Xing et al., 10.5194/acp-13-7531-2013, 2013) might be worth citing.
Line 92: can cite (Zheng et al., 10.5194/acp-18-14095-2018, 2018) for MEIC.
5) Line 85-86: As outlined before, should clarify that your approach improves these assumptions to some extent.
6) Line 91-92: These two are bottom-up inventories, so should follow after Line 73? I do not see clear connection of this sentence with the previous text.
7) Line 98: What idea from bottom-up inventory is used in your approach?
8) Line 107: As outlined before, using UVAI as a proxy of UV radiation seems not appropriate.
9) Fig. 3c and 4c: set log-scale for x-axis might increase the readability of the figure. Also, 28% of days are absent in 2019 so would that affect the sampling and representativeness of data in Fig. 3?
10) Line 211: this is true for daytime and locations with strong NOx emissions only. See (Kenagy et al., 10.1029/2018JD028736, 2018) for nighttime sinks, and (Romer Present et al., 10.5194/acp-20-267-2020, 2020) for possible significant daytime sink via reactions with RO2. As Eq. 3 relates VNO2 at afternoon overpass to 24-h mean emissions, the lifetime should also reflect all hours during the day.
11) Equation 3: Since VNO2 is a snapshot of afternoon overpass while ENOx is 24-h average, so alpha1-alpha3 all contain the conversion from overpass time to 24-h mean. Should acknowledge this fact.
12) Fig. 5: As outlined before, alpha1 is not just determined by type of source.
13) Fig. 6: a scatter plot of Fig. 6a vs. Fig. 3a will provide an insight about how representative Eq. 3 is. Certain locations with strong emissions while unmeasured by Fig. 3a should also be discussed (e.g., are these exactly locations of missed stationary sources?).
14) Fig. 8: What spatial extent is used to calculate the city-mean emissions? If the range is too small, the difference between MEIC and CEMS could be dominated by dilution by the large grid cell.
15) Fig. 9: Are the spatial distribution of EOF1 correlating with that of alpha1? How about EOF2 vs. alpha2? EOF3 vs. divergence?
16) Sections 3.1 and 3.2: Alpha1-alpha3 all exhibit certain spatial and temporal variabilities. What are the implications on previous methods that have simpler (e.g., fixed) assumptions?
17) Line 462: should mention possible benefits from Geostationary instruments that can be promising to resolve the expectedly strong diurnal variability of alpha1-alpha3.
Citation: https://doi.org/10.5194/egusphere-2023-2-RC1 - AC1: 'Reply on RC1', Jason Cohen, 04 Apr 2023
-
RC2: 'Comment on egusphere-2023-2', Anonymous Referee #3, 13 Apr 2023
General comments:
This study presents a new model-free method to constrain NOx emissions using TROPOMI NO2 and ERA-5 wind data. The new method is based on mass balance theory and considers the NOx/NO2 ratio, NOx lifetime and NOx transport. Based on this approach, daily NOx emissions over Shanxi province are estimated during 2019-2021. Some comments should be addressed before its publication. Additionally, the authors apply EOF to TROPOMI NO2 and relate the first three PCs to NOx emissions, UVAI, and NOx transport. The method and conclusions are important, but some comments should be addressed before its publication.
Specific comments:
- When TROPOMI NO2 is not available due to cloud or other reasons, how do you deal with it? My understanding is when TROPOMI NO2 is not available, Eq. 3 does not work. And How the missing data affect the estimated emission inventory?
- In Sect 3.1, why a priori emission inventory (CEMS or MEIC) is needed to estimate new emission inventory? According to Eq. 3, it does not require a priori emission inventory.
- How EOF is applied to the daily TROPOMI NO2 columns when data are not available in some grids?
- I’m curious why the seasonal variation of NOx lifetime shown in Fig. 13b is so small. For example, Lamsal et al. (2010) estimated that the lifetime is NOx is 7.6 h in summer and 17.8 h in winter, while this study showed that lifetime is ~12 h regardless of season.
Lamsal, L. N., Martin, R. V., van Donkelaar, A., Celarier, E. A., Bucsela, E. J., Boersma, K. F., Dirksen, R., Luo, C., & Wang, Y. (2010). Indirect validation of tropospheric nitrogen dioxide retrieved from the OMI satellite instrument: Insight into the seasonal variation of nitrogen oxides at northern midlatitudes. Journal of Geophysical Research: Atmospheres, 115(D5). https://doi.org/10.1029/2009JD013351
- The authors concluded that “Thirdly, the general variability in geography, month of the year, and years before and after COVID-19 are all consistent with what is known.”, while readers cannot find any analysis that is related to COVID-19 in the manuscript.
Technical corrections:
Line 202: the first three PC account for less than 50%
Line 26: NO2 columns identifies -> NO2 columns, which facilitates to identify
Line 47: However -> Moreover
Line 49: also impacting -> that impact
Line 51: Nitrogen Monoxide -> nitric oxide
Line 60: statistics on -> statistics for
Line 67: differences -> the differences
Line 84-85: Kong et al. (Kong et al., 2019) and Beirle et al. 85 (Beirle et al., 2019) -> Kong et al. (2019) and Beirle et al. (2019)
Line 88: costly -> cost
Line 106-109: “The fact that …… variations observed.” This sentence is not easy to understand; it is better to rewrite.
Line 126: Henk Eskes, 2021 -> Henk et al., 2021
Line 126-128: “Furthermore, …… being discarded.” This sentence is not easy to understand; it is better to rewrite.
Line 130: 2021are - > 2021 are
Line 149: as discharged -> emitted
Line 151: NOx concentration measuring -> measuring NOx concentrations
Line 163: 24 is convert -> 24 is used to
Line 168: highest -> the highest
Line 179: uniformity -> uniformly
Line 201: (Björnsson and Venegas, 1997) and (Cohen, 2014) -> Björnsson and Venegas (1997) and Cohen (2014)
Line 208: transport to -> transport
Line 222: basis. -> basis
Line 225: and not -> rather than
Line 232: 𝛼1 𝛼2 and 𝛼3 -> 𝛼1, 𝛼2, and 𝛼3
Line 259: area. -> area,
Line 387: Fig. 12(a) -> Fig. 12a
Citation: https://doi.org/10.5194/egusphere-2023-2-RC2 -
AC2: 'Reply on RC2', Jason Cohen, 10 May 2023
Responses (italics) to Reviewer’s comments (normal text)
General comments:
This study presents a new model-free method to constrain NOx emissions using TROPOMI NO2 and ERA-5 wind data. The new method is based on mass balance theory and considers the NOx/NO2 ratio, NOx lifetime and NOx transport. Based on this approach, daily NOx emissions over Shanxi province are estimated during 2019-2021. Some comments should be addressed before its publication. Additionally, the authors apply EOF to TROPOMI NO2 and relate the first three PCs to NOx emissions, UVAI, and NOx transport. The method and conclusions are important, but some comments should be addressed before its publication.
Specific comments:
- When TROPOMI NO2 is not available due to cloud or other reasons, how do you deal with it? My understanding is when TROPOMI NO2 is not available, Eq. 3 does not work. And how the missing data affect the estimated emission inventory?
This paper introduces a new methodology and makes a first attempt on the combine use of NO2 column loadings and high spatial and temporal frequency observations of ground emissions, within the confines of a first order approximation to the overall mass balance framework. You are correct in that when and where there is missing data, that the emissions cannot be calculated at that exact place and time. We have examined the PDFs of the output emissions at each location and found that they are relatively smooth. For these reasons, unless the missing NO2 observation were statistically very high or very low compared with the other values that already exist, they would not make a large difference in the overall emissions. However, including more observations from other existing and new observation platforms, or using other remotely sensed species in tandem will also help to improve the emissions estimate. Thank you for this suggestion, as it provides a path for future work.
- In Sect 3.1, why a priori emission inventory (CEMS or MEIC) is needed to estimate new emission inventory? According to 3, it does not require a priori emission inventory.
In order to fit the first order terms approximating thermodynamics α1, chemistry α2, and transport α3 in Eq. 3, an initial guess of emissions is required to complete the multiple linear regression. This then allows the distributions of the parameters α1, α2, and α3 to be subsequently used in Eq. 3 to calculate the final emissions. It also allows for error analysis, since the fitted terms themselves have a range of possible solutions. In this work, two different emissions a priori were selected, with the goal being to demonstrate what differences this would have on the computed emissions.
This procedure is similar to how chemical transport models (including GEOS-Chem, WRF-Chem, etc.) have their initial uncertain variables fitted. The major differences being in this work the variables are sufficient simple so as to be flexible and presented in an open way. This allows for a wider range of possible emissions datasets to work within the model environment, which may not be possible with more heavily fitted or constrained modeling approaches. We believe the work herein demonstrates robustness as an entire system.
- How EOF is applied to the daily TROPOMI NO2 columns when data are not available in some grids?
When TROPOMI NO2 columns are not available in some grids, the climatological average value in that grid is assigned in order to compute the EOF. The grid is also tagged and after the EOF is computed, the grid in space and time is reset to NaN, following Cohen (2014).
- I’m curious why the seasonal variation of NOx lifetime shown in Fig. 13b is so small. For example, Lamsal et al. (2010) estimated that the lifetime is NOx is 7.6 h in summer and 17.8 h in winter, while this study showed that lifetime is ~12 h regardless of season.
Lamsal, L. N., Martin, R. V., van Donkelaar, A., Celarier, E. A., Bucsela, E. J., Boersma, K. F., Dirksen, R., Luo, C., & Wang, Y. (2010). Indirect validation of tropospheric nitrogen dioxide retrieved from the OMI satellite instrument: Insight into the seasonal variation of nitrogen oxides at northern midlatitudes. Journal of Geophysical Research: Atmospheres, 115(D5). https://doi.org/10.1029/2009JD013351
The median values of NOx lifetime do demonstrate a range from 9.0 hours to 14.7 hours in different months. The 10th and 90th percentile values match with your reference paper quite well, being 7.1 hours and 18.1 hours respectively. The results in this work are based on the total column values, which includes temperature, UV, climate, and aerosols which are observed in Shanxi. Based on the results herein, the largest values are found in June or July and the smallest values are found in September. This is due to the complex local conditions and forcing factors including the complex boundary layer height, the variable aerosol loading, cloudiness, and other factors.
- The authors concluded that “Thirdly, the general variability in geography, month of the year, and years before and after COVID-19 are all consistent with what is known.”, while readers cannot find any analysis that is related to COVID-19 in the manuscript.
The results herein show clearly that the emissions before COVID-19 were higher than after COVID-19. In specific, the time series shows that while there is a variation as a function of the time of the year, there is also a disturbance in this variation due to the timing of onset of COVID-19. What is important is that the emissions results match well with what is known by the community in terms of month-to-month changes, and geographic diversity, variability, and consistency across different industrial sources and under different oxidative and transport conditions. We have reorganized here in the following way:
“Thirdly, the variability of emissions in terms of different geographic location, source types, special events which changed the emissions levels (such as the onset of COVID-19), and general oxidative, photochemical, and transport conditions of the atmosphere on a monthly-scale, are all consistent with what is known.”
Technical corrections:
Line 202: the first three PC account for less than 50%
The first point is that the community acknowledges there is an uncertainty in TROPOMI observations of NO2 which ranges as high as 30% to 50%. In this case, 30% to 50% of the PCs will be representative of this uncertainty, meaning their signal pattern while mathematically correct, is physically meaningless. Therefore, the results herein represent nearly all of the remaining variability. The three spatial modes [EOF1, EOF2, and EOF3] contribute 29.4%, 8.4%, and 4.4% respectively (accounting for 42.2%), while the fourth mode onward all contribute less than 4.0%. Next, the first three modes all have a high degree of correlation with known underlying driving phenomenon, while the fourth mode and onward show no such relationship. There are many other factors that affect the pollutant column loading in the atmosphere, and if we had more data to analyze, a way to bring in more variables, or a way to reduce the uncertainty, we would also like to search for and work more on attribution. Thank you for helping to carefully guide and clarify our thought process.
Line 26: NO2 columns identifies -> NO2 columns, which facilitates to identify
Thank you, it has been modified.
Line 47: However -> Moreover
Thank you, it has been modified.
Line 49: also impacting -> that impact
Thank you, it has been modified.
Line 51: Nitrogen Monoxide -> nitric oxide
Thank you, it has been modified.
Line 60: statistics on -> statistics for
Thank you, it has been modified to “statistics representing”.
Line 67: differences -> the differences
Thank you, it has been modified.
Line 84-85: Kong et al. (Kong et al., 2019) and Beirle et al. 85 (Beirle et al., 2019) -> Kong et al. (2019) and Beirle et al. (2019)
Thank you, it has been modified.
Line 88: costly -> cost
Thank you, “costly to run” has been modified to “computationally intensive”.
Line 106-109: “The fact that …… variations observed.” This sentence is not easy to understand; it is better to rewrite.
We have reorganized here in the following way.
“This method has been used in different situation such as over different months, over multi-year changes in the environment, under different actinic flux and atmospheric oxidation conditions, under complex meteorological domains, and over sources which are both thermodynamically stable as well as unstable. That permits this study to explore the full range of variations”.
Line 126: Henk Eskes, 2021 -> Henk et al., 2021
The reference format has been modified.
Line 126-128: “Furthermore, …… being discarded.” This sentence is not easy to understand; it is better to rewrite.
We have reorganized here in the following way.
“Furthermore, an additional filter is applied to set all individual gird of NO2 column which is less than 1.4×1015 molec cm-2 to be NaN. This is done to avoid issues where the observed signal may be smaller than the uncertainty of the signal itself (J.H.G.M Van Geffen, 2021; Qin et al, 2022)”.
Line 130: 2021are - > 2021 are
Thank you, it has been modified.
Line 149: as discharged -> emitted
Thank you, it has been modified.
Line 151: NOx concentration measuring -> measuring NOx concentrations
Thank you, it has been modified.
Line 163: 24 is convert -> 24 is used to
Thank you, it has been modified to “24 is used to convert units from hours to days”.
Line 168: highest -> the highest
Thank you, it has been modified.
Line 179: uniformity -> uniformly
Thank you, it has been modified.
Line 201: (Björnsson and Venegas, 1997) and (Cohen, 2014) -> Björnsson and Venegas (1997) and Cohen (2014)
Thank you, it has been modified.
Line 208: transport to -> transport
Thank you, it has been modified.
Line 222: basis. -> basis
Thank you, it has been modified.
Line 225: and not -> rather than
Thank you, it has been modified.
Line 232: 𝛼1 𝛼2 and 𝛼3 -> 𝛼1, 𝛼2, and 𝛼3
Thank you, it has been modified in the whole paper.
Line 259: area. -> area,
Thank you, it has been modified.
Line 387: Fig. 12(a) -> Fig. 12a
Thank you, it has been modified in the whole paper.
References
Cohen, J. B.: Quantifying the occurrence and magnitude of the Southeast Asian fire climatology, Environ. Res. Lett., 9, https://doi.org/10.1088/1748-9326/9/11/114018, 2014.
J.H.G.M. van Geffen, H. J. Eskes, K.F. Boersma and J.P. Veefkind: TROPOMI ATBD of the total and tropospheric NO2 data products, Open File Rep., 2021.
Qin, K., Shi, J., He, Q., Deng, W., Wang, S., Liu, J., and Cohen, J. B.: New Model-Free Daily Inversion of NOx Emissions using TROPOMI (MCMFE-NOx): Deducing a See-Saw of Halved Well Regulated Sources and Doubled New Sources, ESS Open Archive [preprint], https://doi.org/10.1002/essoar.10512010.1, July 26, 2022.
Citation: https://doi.org/10.5194/egusphere-2023-2-AC2
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 420 | 103 | 18 | 541 | 3 | 15 |
- HTML: 420
- PDF: 103
- XML: 18
- Total: 541
- BibTeX: 3
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| China | 1 | 187 | 34 |
| United States of America | 2 | 174 | 32 |
| undefined | 3 | 23 | 4 |
| Germany | 4 | 22 | 4 |
| France | 5 | 16 | 2 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 187
Xiaolu Li
Hong Geng
Liling Wu
Xiaohui Wu
Chengli Yang
Rui Zhang
Liqin Zhang
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(6513 KB) - Metadata XML