the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Oct 2023
The CO anthropogenic emissions in Europe from 2011 to 2021: insights from the MOPITT satellite data
Abstract. We have used the variational inversion drivers of the recent Community Inversion Framework (CIF), coupled to a European configuration of the CHIMERE regional chemistry transport model and its adjoint to derive carbon monixide (CO) emissions from the MOPITT TIR-NIR observations, for a period of over 10 years from 2011 to 2021. The analysis of the inversion results reveals the challenges associated with the inversion of CO emissions at the regional scale over Europe. Annual budgets of the national emissions are decreased by about 1–11 % over the decade and across Europe. These decreases are mainly due to negative corrections during autumn and winter. The posterior CO emissions follow a decreasing trend over the European Union + United Kingdom area with a trend of about -2.2 %/year, slightly lower than in the prior emissions. The assimilation of the MOPITT observation in the inversions indeed attenuates the decreasing trend of the CO emissions in the TNO inventory over areas benefiting from the highest number of MOPITT super-observations (particularly over Italy and over the Balkans), and particularly in autumn and winter. The small corrections of the CO emissions at national scales by the inversion can be attributed, first, to the general consistency between the TNO-GHGco-v3 inventory and the satellite data. Analysis of specific patterns such as the impact of the covid-19 crisis reveal that it can also be seen as a lack of observation constraint to adjust the prior estimate of the emissions. The large errors in the observations, and the lack of data over large parts of Europe are sources of limitation on the observational constraint. Emission hot spots generate a relatively strong local signal, which is much better caught and exploited by the inversions than the larger scale signals, despite the moderate spatial resolution of the MOPITT data. This is why the corrections of these hot spot emissions are stronger and more convincing than the corrections of the national and continental scale emissions. Accurate monitoring of the CO national anthropogenic emissions may thus require modeling and inversion systems at spatial resolution finer than those used here, as well satellite images at high spatial resolution. The CO data of the TROPOMI instrument onboard the Sentinel-5P mission should be well suited for such a perspective.
-
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
(3914 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3914 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1981', Anonymous Referee #1, 01 Nov 2023
Review of The CO anthropogenic emissions in Europe from 2011 to 2021: insights from the MOPITT satellite data by Audrey Fortems-Cheiney et al.
This study could provide interesting insights on an emission inversion framework for CO at the European scale. Posterior emissions are used to verify trends.
There is a lack of bibliographic references, which leads to misleading statements in the introduction and probably for some of the design of the study and the choice of parameters used in the inversions. There has been CO emission inversions at the regional scales, for instance Jiang et al. (2015) and Qu et al., (2022) performed regional MOPITT inversions at the grid cell level at 0.5° × 0.667°.
There are two main concerns:
- MOPITT errors reported on Fig. 3c (30 to 40 %) seems to be larger than usual. This is concerning because the main result of the paper indicates a lack of convergence of the emission optimization “The posterior simulation still presents positive biases compared to the observations, which can be partly explained by large errors in the MOPITT observations”. There are techniques to adjust the errors in order to reach convergence and at least an evaluation of the model data mismatch could be presented. The MOPITT data reports both the measurement error covariance and the smoothing error covariance in addition to the often use total retrieval error covariance. In this study, the averaging kernels (smoothing) are applied in the observation operator (as it should be) to estimate the columns, only the measurement error covariance should then be included, as done in Gaubert et al., (2023). The authors should at least consider showing some statistics on the convergence (e.g., chi-square) of the assimilation for the entire run and adjust the errors accordingly.
- Using a spatial correlations e-folding length of 50 km is effectively forcing the system to constrain hotspots only. Qu et al., (2022) used correlation lengths that varied by sectors with a 100 km to 200 km range. Only point sources from the energy sectors were considered to be at scales smaller than 100 km. Ma et al. (2019) and Gaubert et al. (2020) considers larger correlation lengths of 600 and 500 km on the basis that emission inventories are constructed at the province level (China) or at the scale of entire countries. While they are not inverting the CO emissions, Inness et al., 2023 show a global mean horizontal correlation length of 125 km at the surface level (Figure 1). This is important when the objective of the paper is to assessed “the ability of regional inverse systems to quantify CO budgets at the national scale from the MOPITT TIR-NIR satellite observations” and that the corrections of the hot spots are more “convincing”.
Minor comments:
Check that the figures appear in the same order in the text.
Figure 5’s colorbar indicates “ppb” while the maps show the number of observations.
References:
Gaubert, B., Emmons, L. K., Raeder, K., Tilmes, S., Miyazaki, K., Arellano Jr., A. F., Elguindi, N., Granier, C., Tang, W., Barré, J., Worden, H. M., Buchholz, R. R., Edwards, D. P., Franke, P., Anderson, J. L., Saunois, M., Schroeder, J., Woo, J.-H., Simpson, I. J., Blake, D. R., Meinardi, S., Wennberg, P. O., Crounse, J., Teng, A., Kim, M., Dickerson, R. R., He, H., Ren, X., Pusede, S. E., and Diskin, G. S.: Correcting model biases of CO in East Asia: impact on oxidant distributions during KORUS-AQ, Atmos. Chem. Phys., 20, 14617–14647, https://doi.org/10.5194/acp-20-14617-2020, 2020.
Gaubert, B.; Edwards, D.P.; Anderson, J.L.; Arellano, A.F.; Barré, J.; Buchholz, R.R.; Darras, S.; Emmons, L.K.; Fillmore, D.; Granier, C.; et al. Global Scale Inversions from MOPITT CO and MODIS AOD. Remote Sens. 2023, 15, 4813. https://doi.org/10.3390/rs15194813
Ma, C., Wang, T., Mizzi, A. P., Anderson, J. L., Zhuang, B., Xie, M., & Wu, R. (2019). Multiconstituent data assimilation with WRF-Chem/DART: Potential for adjusting anthropogenic emissions and improving air quality forecasts over eastern China. Journal of Geophysical Research: Atmospheres, 124, 7393–7412. https://doi.org/10.1029/2019JD030421
Inness, A., Aben, I., Ades, M., Borsdorff, T., Flemming, J., Jones, L., Landgraf, J., Langerock, B., Nedelec, P., Parrington, M., and Ribas, R.: Assimilation of S5P/TROPOMI carbon monoxide data with the global CAMS near-real-time system, Atmos. Chem. Phys., 22, 14355–14376, https://doi.org/10.5194/acp-22-14355-2022, 2022.
Jiang, Z., Jones, D. B. A., Worden, J., Worden, H. M., Henze, D. K., and Wang, Y. X.: Regional data assimilation of multi-spectral MOPITT observations of CO over North America, Atmos. Chem. Phys., 15, 6801–6814, https://doi.org/10.5194/acp-15-6801-2015, 2015.
Qu, Z., Henze, D. K., Worden, H. M., Jiang, Z., Gaubert, B., Theys, N., & Wang, W. (2022). Sector-based top-down estimates of NOx, SO2, and CO emissions in East Asia. Geophysical Research Letters, 49, e2021GL096009. https://doi. org/10.1029/2021GL096009
Citation: https://doi.org/10.5194/egusphere-2023-1981-RC1 -
AC1: 'Reply on RC1', Audrey Fortems-Cheiney, 19 Jan 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1981/egusphere-2023-1981-AC1-supplement.pdf
-
RC2: 'Comment on egusphere-2023-1981', Anonymous Referee #2, 01 Nov 2023
In their paper Fortems-Cheiney et al. present the assimilation of MOPITT CO data over Europe and discuss emission trends for a 10-year period. The paper is well written and well structured. I am in favour of publishing these results, but also have several requests for clarifications and minor adjustments, as listed below.
l 24: ..here, as well as satellite ..
l 98: "CO emissions from fires, .. not taken into account" It is difficult for me to understand the distribution shown in Figure 4. In particular the high values in Eastern Europe. Is this linked to fires or something else? is this inflow of CO through the Eastern domain boundary?
l 100: "CO biogenic emissions are assumed to be negligible and are not taken into account." But Table 1 mentions that MEGAN is used, and this seems to contradict this statement. Please explain. Also, Fortems-Cheiney (2021) contains a figure 3 which shows the importance of biogenic + anthro emissions. I could not connect this to Fig. 4 which indicates just a very small impact of the emissions.
l 113: Fig 2a is mentioned, but figure 1 is only referenced in line 227. Please change the order of the figures.
Sec 2.1: Please explain how emissions are distributed on the vertical model layers for the different sectors.
Fig. 4: Please add the period in the caption. Is it a summer month?
difference between pirior and CO emis = 0 is very small?! Just few %.l 126: "MOPITT instrument version 8". Please change: the version refers to the CO retrieval code, not the instrument.
l 133: Why "surface" product instead of surface product? Please include a note on the sensitivity profiles or averaging kernels of MOPITT. How many degrees of signal are there typically in the combined profile retrieval?
l 133: "We choose to assimilate the MOPITT “surface” product." Why? Would column or profile assimilation give different results? The combination of NIR and TIR holds the promise of some vertical profile information, which could be extracted when the profiles are assimilated.
l 139: Superobservations are constructed by using the median observation in each 0.5x0.5 grid cell. I assume the median retrieval and median averaging kernel are used in the observation operator. But the process of using the median may/will remove noise from the MOPITT observations, and could result in smaller observation errors. Is the median observation error used, or a reduced error? Please state this explicitly.
l 140: What is "AK" (not defined before this point)? Please discuss the averaginging kernels: are these used in the observation operator or not? I assume they are. I would find it useful if some typical AK profiles are shown, in order to better understand Fig.3. Adding the modelled surface concentration and column to Fig.3 would be helpful.
Sec 2.2. How is the uncertainty of the superobservation determined. Is it equal to the error of the median retrieval? A superobservation may have a smaller error than the individual obervations. Since the uncertainty is important for the final result, the authors should explain this more clearly.
l 154: Why are emissions specified for 8 levels? Is this needed?
Does the analysis significantly change the vertical distribution of emissions?Sec 2.3: The consistency between B, R and the observation-minus forecast departures can be tested using chi^2. From the paper I have the impression this was not done. Why not?
Sec 3: I was surprised to read that emissions of CO need to be reduced. In the past models have struggled with a low bias in CO, e.g. Stein, 10.5194/acp-14-9295-2014, suggesting emissions should be increased, especially in winter. In Fig. 3 Chimere is generally higher tham MOPITT. But in the 2021 paper, fig.5, Chimere simulates smaller concentrations. Please explain the increase in CO in the model. Is it linked to the emissions, boundaries or model change? Please add some discussion on modelling bias in CO, including a few relevant papers.
Sec 3: Fortems-Cheiney (2021) mentions "local increments on CO emissions can reach more than +50 %, with increases located mainly over central and eastern Europe, except in the south of Poland, and decreases located over Spain and Portugal." Please contrast these earlier results with the results presented here.
Fig 3. In data assimilation the prior for a given month is often based on the posterior from the previous month. In contrast, the prior could also refer to a free model run with prior emissions. Maybe I missed it, but it was not clear to me how "prior" is defined in this paper. Please clarify if the results (e.g. emission adjustment) is passed on from one month to the next in the assimilation.
l 233: "The posterior CO emissions display a very similar decreasing trend than the prior emissions over the EU-27+UK area" If the assimilation uses the trends in the prior TNO emission database, and if the observations do not provide a strong forcing, then this may not be very surprising. (Would be nice to have results using fixed emissions, e.g. using 2015.)
l 203: "posterior simulation still presents positive biases" In fact only a minor part of the bias (order 20% over the continent) is removed if I understand Fig. 3-d correctly. Please be more quantitative here and mention these percentages.
l 213: Figure 1c should be 2b, I assume?
l 216: "Table 3" should be Table 2 I assume.
l 239: "steadily increasing" Please provide trend figures per country for 2011-2019 (extra table or combined in Table 2).
l 262: "decrease by about -1.3%" Could this be related to the remark on lime 203 that posterior emissions are still biased and capture only part of the model-MOPITT difference?
Fig.7. Why is the sea/ocean yellow in 2020 (positive increment) and not in 2019?
Citation: https://doi.org/10.5194/egusphere-2023-1981-RC2 -
AC2: 'Reply on RC2', Audrey Fortems-Cheiney, 19 Jan 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1981/egusphere-2023-1981-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Audrey Fortems-Cheiney, 19 Jan 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1981', Anonymous Referee #1, 01 Nov 2023
Review of The CO anthropogenic emissions in Europe from 2011 to 2021: insights from the MOPITT satellite data by Audrey Fortems-Cheiney et al.
This study could provide interesting insights on an emission inversion framework for CO at the European scale. Posterior emissions are used to verify trends.
There is a lack of bibliographic references, which leads to misleading statements in the introduction and probably for some of the design of the study and the choice of parameters used in the inversions. There has been CO emission inversions at the regional scales, for instance Jiang et al. (2015) and Qu et al., (2022) performed regional MOPITT inversions at the grid cell level at 0.5° × 0.667°.
There are two main concerns:
- MOPITT errors reported on Fig. 3c (30 to 40 %) seems to be larger than usual. This is concerning because the main result of the paper indicates a lack of convergence of the emission optimization “The posterior simulation still presents positive biases compared to the observations, which can be partly explained by large errors in the MOPITT observations”. There are techniques to adjust the errors in order to reach convergence and at least an evaluation of the model data mismatch could be presented. The MOPITT data reports both the measurement error covariance and the smoothing error covariance in addition to the often use total retrieval error covariance. In this study, the averaging kernels (smoothing) are applied in the observation operator (as it should be) to estimate the columns, only the measurement error covariance should then be included, as done in Gaubert et al., (2023). The authors should at least consider showing some statistics on the convergence (e.g., chi-square) of the assimilation for the entire run and adjust the errors accordingly.
- Using a spatial correlations e-folding length of 50 km is effectively forcing the system to constrain hotspots only. Qu et al., (2022) used correlation lengths that varied by sectors with a 100 km to 200 km range. Only point sources from the energy sectors were considered to be at scales smaller than 100 km. Ma et al. (2019) and Gaubert et al. (2020) considers larger correlation lengths of 600 and 500 km on the basis that emission inventories are constructed at the province level (China) or at the scale of entire countries. While they are not inverting the CO emissions, Inness et al., 2023 show a global mean horizontal correlation length of 125 km at the surface level (Figure 1). This is important when the objective of the paper is to assessed “the ability of regional inverse systems to quantify CO budgets at the national scale from the MOPITT TIR-NIR satellite observations” and that the corrections of the hot spots are more “convincing”.
Minor comments:
Check that the figures appear in the same order in the text.
Figure 5’s colorbar indicates “ppb” while the maps show the number of observations.
References:
Gaubert, B., Emmons, L. K., Raeder, K., Tilmes, S., Miyazaki, K., Arellano Jr., A. F., Elguindi, N., Granier, C., Tang, W., Barré, J., Worden, H. M., Buchholz, R. R., Edwards, D. P., Franke, P., Anderson, J. L., Saunois, M., Schroeder, J., Woo, J.-H., Simpson, I. J., Blake, D. R., Meinardi, S., Wennberg, P. O., Crounse, J., Teng, A., Kim, M., Dickerson, R. R., He, H., Ren, X., Pusede, S. E., and Diskin, G. S.: Correcting model biases of CO in East Asia: impact on oxidant distributions during KORUS-AQ, Atmos. Chem. Phys., 20, 14617–14647, https://doi.org/10.5194/acp-20-14617-2020, 2020.
Gaubert, B.; Edwards, D.P.; Anderson, J.L.; Arellano, A.F.; Barré, J.; Buchholz, R.R.; Darras, S.; Emmons, L.K.; Fillmore, D.; Granier, C.; et al. Global Scale Inversions from MOPITT CO and MODIS AOD. Remote Sens. 2023, 15, 4813. https://doi.org/10.3390/rs15194813
Ma, C., Wang, T., Mizzi, A. P., Anderson, J. L., Zhuang, B., Xie, M., & Wu, R. (2019). Multiconstituent data assimilation with WRF-Chem/DART: Potential for adjusting anthropogenic emissions and improving air quality forecasts over eastern China. Journal of Geophysical Research: Atmospheres, 124, 7393–7412. https://doi.org/10.1029/2019JD030421
Inness, A., Aben, I., Ades, M., Borsdorff, T., Flemming, J., Jones, L., Landgraf, J., Langerock, B., Nedelec, P., Parrington, M., and Ribas, R.: Assimilation of S5P/TROPOMI carbon monoxide data with the global CAMS near-real-time system, Atmos. Chem. Phys., 22, 14355–14376, https://doi.org/10.5194/acp-22-14355-2022, 2022.
Jiang, Z., Jones, D. B. A., Worden, J., Worden, H. M., Henze, D. K., and Wang, Y. X.: Regional data assimilation of multi-spectral MOPITT observations of CO over North America, Atmos. Chem. Phys., 15, 6801–6814, https://doi.org/10.5194/acp-15-6801-2015, 2015.
Qu, Z., Henze, D. K., Worden, H. M., Jiang, Z., Gaubert, B., Theys, N., & Wang, W. (2022). Sector-based top-down estimates of NOx, SO2, and CO emissions in East Asia. Geophysical Research Letters, 49, e2021GL096009. https://doi. org/10.1029/2021GL096009
Citation: https://doi.org/10.5194/egusphere-2023-1981-RC1 -
AC1: 'Reply on RC1', Audrey Fortems-Cheiney, 19 Jan 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1981/egusphere-2023-1981-AC1-supplement.pdf
-
RC2: 'Comment on egusphere-2023-1981', Anonymous Referee #2, 01 Nov 2023
In their paper Fortems-Cheiney et al. present the assimilation of MOPITT CO data over Europe and discuss emission trends for a 10-year period. The paper is well written and well structured. I am in favour of publishing these results, but also have several requests for clarifications and minor adjustments, as listed below.
l 24: ..here, as well as satellite ..
l 98: "CO emissions from fires, .. not taken into account" It is difficult for me to understand the distribution shown in Figure 4. In particular the high values in Eastern Europe. Is this linked to fires or something else? is this inflow of CO through the Eastern domain boundary?
l 100: "CO biogenic emissions are assumed to be negligible and are not taken into account." But Table 1 mentions that MEGAN is used, and this seems to contradict this statement. Please explain. Also, Fortems-Cheiney (2021) contains a figure 3 which shows the importance of biogenic + anthro emissions. I could not connect this to Fig. 4 which indicates just a very small impact of the emissions.
l 113: Fig 2a is mentioned, but figure 1 is only referenced in line 227. Please change the order of the figures.
Sec 2.1: Please explain how emissions are distributed on the vertical model layers for the different sectors.
Fig. 4: Please add the period in the caption. Is it a summer month?
difference between pirior and CO emis = 0 is very small?! Just few %.l 126: "MOPITT instrument version 8". Please change: the version refers to the CO retrieval code, not the instrument.
l 133: Why "surface" product instead of surface product? Please include a note on the sensitivity profiles or averaging kernels of MOPITT. How many degrees of signal are there typically in the combined profile retrieval?
l 133: "We choose to assimilate the MOPITT “surface” product." Why? Would column or profile assimilation give different results? The combination of NIR and TIR holds the promise of some vertical profile information, which could be extracted when the profiles are assimilated.
l 139: Superobservations are constructed by using the median observation in each 0.5x0.5 grid cell. I assume the median retrieval and median averaging kernel are used in the observation operator. But the process of using the median may/will remove noise from the MOPITT observations, and could result in smaller observation errors. Is the median observation error used, or a reduced error? Please state this explicitly.
l 140: What is "AK" (not defined before this point)? Please discuss the averaginging kernels: are these used in the observation operator or not? I assume they are. I would find it useful if some typical AK profiles are shown, in order to better understand Fig.3. Adding the modelled surface concentration and column to Fig.3 would be helpful.
Sec 2.2. How is the uncertainty of the superobservation determined. Is it equal to the error of the median retrieval? A superobservation may have a smaller error than the individual obervations. Since the uncertainty is important for the final result, the authors should explain this more clearly.
l 154: Why are emissions specified for 8 levels? Is this needed?
Does the analysis significantly change the vertical distribution of emissions?Sec 2.3: The consistency between B, R and the observation-minus forecast departures can be tested using chi^2. From the paper I have the impression this was not done. Why not?
Sec 3: I was surprised to read that emissions of CO need to be reduced. In the past models have struggled with a low bias in CO, e.g. Stein, 10.5194/acp-14-9295-2014, suggesting emissions should be increased, especially in winter. In Fig. 3 Chimere is generally higher tham MOPITT. But in the 2021 paper, fig.5, Chimere simulates smaller concentrations. Please explain the increase in CO in the model. Is it linked to the emissions, boundaries or model change? Please add some discussion on modelling bias in CO, including a few relevant papers.
Sec 3: Fortems-Cheiney (2021) mentions "local increments on CO emissions can reach more than +50 %, with increases located mainly over central and eastern Europe, except in the south of Poland, and decreases located over Spain and Portugal." Please contrast these earlier results with the results presented here.
Fig 3. In data assimilation the prior for a given month is often based on the posterior from the previous month. In contrast, the prior could also refer to a free model run with prior emissions. Maybe I missed it, but it was not clear to me how "prior" is defined in this paper. Please clarify if the results (e.g. emission adjustment) is passed on from one month to the next in the assimilation.
l 233: "The posterior CO emissions display a very similar decreasing trend than the prior emissions over the EU-27+UK area" If the assimilation uses the trends in the prior TNO emission database, and if the observations do not provide a strong forcing, then this may not be very surprising. (Would be nice to have results using fixed emissions, e.g. using 2015.)
l 203: "posterior simulation still presents positive biases" In fact only a minor part of the bias (order 20% over the continent) is removed if I understand Fig. 3-d correctly. Please be more quantitative here and mention these percentages.
l 213: Figure 1c should be 2b, I assume?
l 216: "Table 3" should be Table 2 I assume.
l 239: "steadily increasing" Please provide trend figures per country for 2011-2019 (extra table or combined in Table 2).
l 262: "decrease by about -1.3%" Could this be related to the remark on lime 203 that posterior emissions are still biased and capture only part of the model-MOPITT difference?
Fig.7. Why is the sea/ocean yellow in 2020 (positive increment) and not in 2019?
Citation: https://doi.org/10.5194/egusphere-2023-1981-RC2 -
AC2: 'Reply on RC2', Audrey Fortems-Cheiney, 19 Jan 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1981/egusphere-2023-1981-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Audrey Fortems-Cheiney, 19 Jan 2024
Peer review completion
Post-review adjustments
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 294 | 109 | 27 | 430 | 15 | 11 |
- HTML: 294
- PDF: 109
- XML: 27
- Total: 430
- BibTeX: 15
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 121 | 28 |
| France | 2 | 52 | 12 |
| China | 3 | 48 | 11 |
| Germany | 4 | 36 | 8 |
| Netherlands | 5 | 21 | 4 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 121
Audrey Fortems-Cheiney
Gregoire Broquet
Elise Potier
Robin Plauchu
Antoine Berchet
Isabelle Pison
Hugo A. C. Denier van der Gon
Stijn N. C. Dellaert
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3914 KB) - Metadata XML