Technical note: The CAMS greenhouse gas reanalysis from 2003 to 2020

Agusti-Panareda, Anna; Barré, Jérôme; Massart, Sébastien; Inness, Antje; Aben, Ilse; Ades, Melanie; Baier, Bianca C.; Balsamo, Gianpaolo; Borsdorff, Tobias; Bousserez, Nicolas; Boussetta, Souhail; Buchwitz, Michael; Cantarello, Luca; Crevoisier, Cyril; Engelen, Richard; Eskes, Henk; Flemming, Johannes; Garrigues, Sébastien; Hasekamp, Otto; Huijnen, Vincent; Jones, Luke; Kipling, Zak; Langerock, Bavo; McNorton, Joe; Meilhac, Nicolas; Noel, Stefan; Parrington, Mark; Peuch, Vincent-Henri; Ramonet, Michel; Ratzinger, Miha; Reuter, Maximilian; Ribas, Roberto; Suttie, Martin; Sweeney, Colm; Tarniewicz, Jérôme; Wu, Lianghai

doi:https://doi.org/10.5194/egusphere-2022-283

Preprints

https://doi.org/10.5194/egusphere-2022-283

Preprints

18 Jul 2022

| 18 Jul 2022

Technical note: The CAMS greenhouse gas reanalysis from 2003 to 2020

Anna Agusti-Panareda, Jérôme Barré, Sébastien Massart, Antje Inness, Ilse Aben, Melanie Ades, Bianca C. Baier, Gianpaolo Balsamo, Tobias Borsdorff, Nicolas Bousserez, Souhail Boussetta, Michael Buchwitz, Luca Cantarello, Cyril Crevoisier, Richard Engelen, Henk Eskes, Johannes Flemming, Sébastien Garrigues, Otto Hasekamp, Vincent Huijnen, Luke Jones, Zak Kipling, Bavo Langerock, Joe McNorton, Nicolas Meilhac, Stefan Noel, Mark Parrington, Vincent-Henri Peuch, Michel Ramonet, Miha Ratzinger, Maximilian Reuter, Roberto Ribas, Martin Suttie, Colm Sweeney, Jérôme Tarniewicz, and Lianghai Wu

Abstract. The Copernicus Atmosphere Monitoring Service has recently produced a greenhouse gases reanalysis (version egg4) that covers almost two decades from 2003 to 2020 and will be extended in the future. This reanalysis dataset includes carbon dioxide (CO₂) and methane (CH₄). The reanalysis procedure combines model data with satellite data into a globally complete and consistent dataset using the European Centre for Medium-range Weather Forecasts’ Integrated Forecasting System (IFS). This dataset has been carefully evaluated against independent observations to ensure validity and point out deficiencies to the user. The greenhouse gas reanalysis can be used to examine the impact of atmospheric greenhouse gases concentrations on climate change, such as global and regional climate radiative forcing, assess intercontinental transport, and also serve as boundary conditions for regional simulations, among other applications and scientific studies. The caveats associated with changes in assimilated observations and fixed underlying emissions are highlighted, as well as their impact on the estimation of trends and annual growth rates of these long-lived greenhouse gases.

How to cite. Agusti-Panareda, A., Barré, J., Massart, S., Inness, A., Aben, I., Ades, M., Baier, B. C., Balsamo, G., Borsdorff, T., Bousserez, N., Boussetta, S., Buchwitz, M., Cantarello, L., Crevoisier, C., Engelen, R., Eskes, H., Flemming, J., Garrigues, S., Hasekamp, O., Huijnen, V., Jones, L., Kipling, Z., Langerock, B., McNorton, J., Meilhac, N., Noel, S., Parrington, M., Peuch, V.-H., Ramonet, M., Ratzinger, M., Reuter, M., Ribas, R., Suttie, M., Sweeney, C., Tarniewicz, J., and Wu, L.: Technical note: The CAMS greenhouse gas reanalysis from 2003 to 2020, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2022-283, 2022.

Received: 30 Apr 2022 – Discussion started: 18 Jul 2022

Anna Agusti-Panareda et al.

Status: closed

RC1:
'Comment on egusphere-2022-283', Anonymous Referee #2, 09 Aug 2022

Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.

This paper describes the greenhouse gases (CO2 and CH4) reanalysis as prepared for the Copernicus Atmosphere Monitoring Service (CAMS). Manuscript draft is well prepared for understading of the work that is carreid out. The method is well described but the results are not fully satisfacotry yet. Many traditional inversions produced better statistics for model simulated CO2 and CH4 concentrations by optimising the surface sources and sinks (e.g., Chandra et al., ACP 2022). However, the approach holds different promises for 4D concentration product of CO2 and CH4 in near realtime along with other meteorological parameters of ECMWF weather forecasting system. In general I recommend publication of the manuscript as "Technical note", after accounting for some of my comments below.

Line 48: are the SCIAMACHY and IASI data good enough for assimilation ? the growth rate may be fine but the spatial distribution are probably of poor quality.

I have found later that this is mentioned in the caveats. No action is needed but I am just making sure to mention the points of concern

Figure 1: I have several doubts on these plots.

1) the total column CO2 values on Dec 2020 are of similar magnitude of CO2 at MLO (414 pmm as per SIO flask data). I was thinking that global mean XCO2 will be a couple ppm lower than MLO value. Please confirm.

2) What is the advantage of showing 2003-2020 mean as opposed to just 2020 seasonal means ?

Figure 2: Should the two-way arrows between Forecast "Sphere" and Surface flxes "Maps" be one-way ? From Fluxes to Forecast

Table 2: You could use FF-CO2 emissions EDGARv6.0 etc. or the GridFED by UEA group. Similarly many of the flux components of CO2 and CH4 should be revised by using the most recent flux data sources. I not saying for this paper but in the near future.

Line 352: I think it would have been nice to add/subtract some CO2 flux to make the CO2 flux budget consistent with observed CO2 growth rate. At the very least the land biosphere + ocean exchange could be added/subtracted with an offset to balance all sources+sinks = 2.12 * Growth rate * 2.12. Then the statistics may be look that bad.

Figure 7: It would be nicer to use a discrete colour scale (like in Fig. 7). Otherwise, it is hard to distinguish between medium blue/red from dark blue/red

Also consider showing a Taylor diagram, like I see bellow for the TCCON sites

Do you need this Figure 14 ? If needed I suggest a more detailed site information is shown before Fig. 7. The Surface, MBL, TCCON, NDAAC, AirCore etc. sites can be shown in different symbols

Citation: https://doi.org/10.5194/egusphere-2022-283-RC1
- AC2: 'Reply on RC1', Anna Agusti-Panareda, 08 Dec 2022
  
  We are very grateful for the insightful comments from the reviewer to improve the documentation of the CAMS greenhouse gas reanalysis and in the manuscript and the suggestions for potential improvements to be considered in future reanalyses. We have addressed the different comments of the reviewer (in bold) to clarify the different aspects of the reanalysis including the selection/quality of observations assimilation, the constrain of the annual global growth rates and the quality of the prescribed surface fluxes.
  Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.
  This paper describes the greenhouse gases (CO2 and CH4) reanalysis as prepared for the Copernicus Atmosphere Monitoring Service (CAMS). Manuscript draft is well prepared for understanding of the work that is carried out. The method is well described but the results are not fully satisfactory yet. Many traditional inversions produced better statistics for model simulated CO2 and CH4 concentrations by optimising the surface sources and sinks (e.g., Chandra et al., ACP 2022). However, the approach holds different promises for 4D concentration product of CO2 and CH4 in near real time along with other meteorological parameters of ECMWF weather forecasting system. In general I recommend publication of the manuscript as "Technical note", after accounting for some of my comments below.
  Line 48: are the SCIAMACHY and IASI data good enough for assimilation ? the growth rate may be fine but the spatial distribution are probably of poor quality.
  
  I have found later that this is mentioned in the caveats. No action is needed but I am just making sure to mention the points of concern.
  The rationale for the selection of the observations assimilated in the CAMS reanalysis and the importance of assessing their quality will be emphasized in the revised version of the manuscript.
  Figure 1: I have several doubts on these plots.
  
  1) the total column CO2 values on Dec 2020 are of similar magnitude of CO2 at MLO (414 pm as per SIO flask data). I was thinking that global mean XCO2 will be a couple ppm lower than MLO value. Please confirm.
  
  Figure 1 gives an overview of the CAMS GHG reanalysis dataset including global trends and spatial/temporal distribution at seasonal sales. The global mean has an error ranging from -0.7 to 3.5 ppm. Figure 14 shows that global mean error in 2020 is +1.35 ppm. This is consistent with the assessment of the reviewer. The problem of constraining the global mean and annual growth rate is documented in section 3.3 and mentioned in section 4 (point 6). The range of error in the global mean will be added in the revised version of the manuscript.
  
  2) What is the advantage of showing 2003-2020 mean as opposed to just 2020 seasonal means?
  The purpose here is to show the typical seasonal cycle. For this reason, using 2003-2020 mean as opposed to just 2020 seasonal means is better because individual years can be affected by the large inter-annual variability of biogenic fluxes (e.g. during el Nino years). This will be mentioned in the revised manuscript.
  Figure 2: Should the two-way arrows between Forecast "Sphere" and Surface flxes "Maps" be one-way? From Fluxes to Forecast
  The two-way arrows indicate the two-way coupling between the atmospheric model and the model of biogenic surface fluxes. The surface fluxes are affected by the forecast of temperature, radiation, humidity and soil moisture and the atmospheric CO2 in the forecast is affected by the surface biogenic fluxes. This will be clarified in the revised version of the manuscript.
  Table 2: You could use FF-CO2 emissions EDGARv6.0 etc. or the GridFED by UEA group. Similarly many of the flux components of CO2 and CH4 should be revised by using the most recent flux data sources. I not saying for this paper but in the near future.
  The current reanalysis cannot be changed as the reanalysis system requires a consistent configuration of the model and prescribed input fluxes throughout the whole reanalysis, but we will certainly use a newer version of the anthropogenic emissions in the next CAMS reanalysis planned for 2024. This will be mentioned in the revised manuscript.
  Line 352: I think it would have been nice to add/subtract some CO2 flux to make the CO2 flux budget consistent with observed CO2 growth rate. At the very least the land biosphere + ocean exchange could be added/subtracted with an offset to balance all sources+sinks = 2.12 * Growth rate. Then the statistics may be look that bad.
  This adjustment of the fluxes with the CO2 growth rate will be considered in the next CAMS re-analysis. There is the caveat that the growth rate will have to be adjusted in near-real time which might not be feasible. A note of this option will be made in the discussion section of the revised manuscript.
  Figure 7: It would be nicer to use a discrete colour scale (like in Fig. 7). Otherwise, it is hard to distinguish between medium blue/red from dark blue/red
  Also consider showing a Taylor diagram, like I see below for the TCCON sites
  We will change Figure 7 to use a discrete colour scale and add a Taylor diagram in the revised manuscript.
  Do you need this Figure 14 ? If needed I suggest a more detailed site information is shown before Fig. 7. The Surface, MBL, TCCON, NDAAC, AirCore etc. sites can be shown in different symbols
  The map of the MBL sites is useful to be able to interpret the observed global mean produced by NOAA. A detailed list of TCCON, The NDACC, AirCore is provided in the Appendix. In the revised manuscript the map in Figure 14 will be updated to show the different types of observations with different symbols and placed before Fig. 7 as suggested by the reviewer.
  
  Citation: https://doi.org/10.5194/egusphere-2022-283-AC2
RC2:
'Comment on egusphere-2022-283', Anonymous Referee #3, 19 Aug 2022
This article presents a description of a new reanalysis dataset of greenhouse gases (GHGs). This is the first GHG reanalysis dataset ever produced using data assimilation techniques that adjust GHG concentrations rather than GHG fluxes. As such, it is a formidable achievement but at the same time, there is considerable room for improvement (as noted by the authors in sections 4 and 5). Reanalyses are typically improved after feedback from the user community and/or with improved data assimilation techniques or increased numbers of observations. Thus an important aspect of reanalysis production is the dissemination and documentation of the products to encourage and inform the potential user community. That is the main purpose of this article. As such, the comments below are primarily aimed at improving the description of the product to potential users.

Specific comments

Section 2.2 and Table 1: It would be useful to know how the satellite instruments were selected. For example, OCO-2 was not used. In section 5, it is mentioned that it will be used in future versions, but it is useful for the user to understand the rationale behind the selection of instruments for this product.

Line 99 and section 2.3: Figure 2 shows 2-way interaction between the forecast and surface fluxes. Please explain how the forecast influences the surface fluxes in section 2.3 or line 99.

Section 2.3: There is no mention of Figure 3 in the text. It is useful to understand why this Figure is presented, and the main message behind it.

Line 184: What is the rationale behind the choice of EDGAR versus higher spatial resolution datasets for anthropogenic emissions?

Line 217: Typo: Tl255 should be TL255 presumably.

Line 302: I do not see the 20 ppm error in Figure 5. If the vertical scale in Fig. 5c (bottom panel) is linear then the green and red dashed curves seem to overlap after 2019, meaning errors of less than 5 ppm.

Lines 368-370: Taylor diagrams of Figs. 8b, 11b. What is the normalization used on the radial axes? Presumably it is the observed standard deviation for a given site. Why was this normalization needed? Standard Taylor diagrams do not do a normalization. Presumably it allows for better comparison among sites with very different observed variability.

Lines 457-8 and Fig. 13: How was the site for Fig. 13 chosen? Is it a typical example, or a good example? It is nice to see that the overall structure of the profile (esp. for CO₂) is well captured in boreal summer in France. It would be interesting to see the comparison in the southern hemisphere. Are the general biases of Fig. 12 more evident in New Zealand?

Line 496: Since this is a reanalysis, why were the insitu data not assimilated? Would they be enough to better constrain the global growth rate? Will this be done in the future?
Citation: https://doi.org/10.5194/egusphere-2022-283-RC2
- AC3:
  'Reply on RC2', Anna Agusti-Panareda, 08 Dec 2022
  We are very grateful for the insightful comments from the reviewer to improve the documentation of the CAMS greenhouse gas reanalysis for potential users. We have addressed the comments of the reviewer (in bold) to clarify the different aspects of the reanalysis including the selection/quality of the assimilated observations, the different model components and the characteristics of the prescribed surface fluxes.
  Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.
  This article presents a description of a new reanalysis dataset of greenhouse gases (GHGs). This is the first GHG reanalysis dataset ever produced using data assimilation techniques that adjust GHG concentrations rather than GHG fluxes. As such, it is a formidable achievement but at the same time, there is considerable room for improvement (as noted by the authors in sections 4 and 5). Reanalyses are typically improved after feedback from the user community and/or with improved data assimilation techniques or increased numbers of observations. Thus an important aspect of reanalysis production is the dissemination and documentation of the products to encourage and inform the potential user community. That is the main purpose of this article. As such, the comments below are primarily aimed at improving the description of the product to potential users.
  Specific comments
  Section 2.2 and Table 1: It would be useful to know how the satellite instruments were selected. For example, OCO-2 was not used. In section 5, it is mentioned that it will be used in future versions, but it is useful for the user to understand the rationale behind the selection of instruments for this product.
  
  The main rationale for the selection of the data was the operational availability of NRT data, as the re-analysis is foreseen to catch up and run in near-real time eventually. Currently there is no near-real time product of the OCO-2 data and that is why OCO-2 has not been used in the current CAMS near-real time products (including re-analysis). We are working with LSCE to have a near-real time OCO-2 product based on a neural network retrieval (Breon et al., 2022; https://amt.copernicus.org/articles/15/5219/2022/amt-15-5219-2022-discussion.html). This will allow us to use OCO-2 in the next CAMS re-analysis. The rationale for the observation data selection will be clarified in the revised manuscript.
  Line 99 and section 2.3: Figure 2 shows 2-way interaction between the forecast and surface fluxes. Please explain how the forecast influences the surface fluxes in section 2.3 or line 99.
  
  The two-way arrows indicate the two-way coupling between the atmospheric forecast and the biogenic surface fluxes. The surface fluxes are affected by the forecast of temperature, humidity, radiation and soil moisture and the atmospheric CO2 in the forecast is affected by the surface biogenic fluxes. This will be clarified in the revised version of the manuscript.
  Section 2.3: There is no mention of Figure 3 in the text. It is useful to understand why this Figure is presented, and the main message behind it.
  
  Figure 3 shows the seasonal, inter-annual variability and trend of the surface fluxes. This is important because one of the caveats of the re-analysis is that CH4 emissions are kept fixed from 2010 onwards, while for CO2 anthropogenic fluxes we apply an extrapolation, and the biogenic fluxes are modelled and therefore have an inter-annual variability. In the revised manuscript, Figure 3 is mentioned and referred to when explaining the differences between different fluxes in terms of inter-annual variability and seasonal cycles.
  Line 184: What is the rationale behind the choice of EDGAR versus higher spatial resolution datasets for anthropogenic emissions?
  
  EDGAR produces global emissions for both CO2 and CH4 at a relatively high resolution of 0.1 degrees (compared to 80km resolution of the CAMS re-analysis). The problem with EDGAR is the extension to NRT and the caveats associated with large inter-annual variability during the COVID period. This will be clarified in the revised manuscript.
  Line 217: Typo: Tl255 should be TL255 presumably.
  
  This will be corrected in the revised manuscript.
  Line 302: I do not see the 20 ppm error in Figure 5. If the vertical scale in Fig. 5c (bottom panel) is linear then the green and red dashed curves seem to overlap after 2019, meaning errors of less than 5 ppm.
  
  Yes, indeed. There is a typo here, 20ppm should be 2ppm. The sentence “accompanied by a large standard deviation error (~20ppm, cf. dashed lines in same panel and figure)” has been removed as the degradation associated with IASI is mainly detected from the large step change in the averages of the departures.
  Lines 368-370: Taylor diagrams of Figs. 8b, 11b. What is the normalization used on the radial axes? Presumably it is the observed standard deviation for a given site. Why was this normalization needed? Standard Taylor diagrams do not do a normalization. Presumably it allows for better comparison among sites with very different observed variability.
  
  The reviewer is correct. Assuming we have 20 sites, then there are 20 time series for the model and the observed NDACC ground based data (gb), each with its own variability. In order to plot them together, the standard deviation of the model time series is used to normalize the timeseries (i.e for a site, the time series model or gb is divided with the std of the model time series ): that is why the model time series is always at 1 for each site. The correlations are independent of such a normalization and from the location of a site in the Taylor plot you can deduce if the model has higher variability (the site is plotted with a distance <1 to the origin) or lower (the site is plotted at a distance >1 from the origin) compared to the variability in the gb data. This will be clarified in the revised manuscript.
  Lines 457-8 and Fig. 13: How was the site for Fig. 13 chosen? Is it a typical example, or a good example? It is nice to see that the overall structure of the profile (esp. for CO₂) is well captured in boreal summer in France. It would be interesting to see the comparison in the southern hemisphere. Are the general biases of Fig. 12 more evident in New Zealand?
  
  The site of Traînou (France) has been used as a good example of the synoptic variability during the summer, more than the systematic errors. As the reviewer suggested, we have plotted the only profile available from Lauder (New Zealand) (see Figure S1 in supplement) which shows much larger errors in September, particularly near the surface and in the stratosphere. In the revised manuscript we will add an extra figure as a Supplement to illustrate the errors at different sites across different latitude bands.
  
  Figure S1. Vertical mole fraction profiles of CO₂ and CH₄ from the CAMS GHG reanalysis (dash line) and AirCore observations (solid line) at Lauder (New Zealand, see Table A3) over the period in June 2019.
  
  Line 496: Since this is a reanalysis, why were the in situ data not assimilated? Would they be enough to better constrain the global growth rate? Will this be done in the future?
  
  The CAMS reanalysis and generally the reanalysis at ECMWF aim to eventually run in near-real time. Since the availability of quality controlled in situ data is not generally available in near real time and we use a global model with a short window data assimilation (i.e. 12 hours) that adjusts only the GHG concentrations (i.e. not the fluxes), the impact of the sparse in situ data would be small as the increments are very localised in time (i.e at the beginning of the data assimilation window) and space (around the station near the surface). Preliminary tests showed that the increments around the in situ stations are dispersed by the atmospheric transport in less than 12 hours. In the future, we will explore the possibility of using in situ data with the inversion capability currently being developed in the CoCO2 project. The flux adjustment will result in a longer-lived impact of the impact of the in situ observations, as we expect the resulting atmospheric CO2 and CH4 corrections will not be as localised in time and space, particularly when using a longer data assimilation window with the inversion capability. All these aspects related to the future potential assimilation of in situ data will be clarified in the revised version of the manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2022-283-AC3
- AC4:
  'Reply on RC2', Anna Agusti-Panareda, 08 Dec 2022
  We are very grateful for the insightful comments from the reviewer to improve the documentation of the CAMS greenhouse gas reanalysis for potential users. We have addressed the comments of the reviewer (in bold) to clarify the different aspects of the reanalysis including the selection/quality of the assimilated observations, the different model components and the characteristics of the prescribed surface fluxes.
  Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.
  This article presents a description of a new reanalysis dataset of greenhouse gases (GHGs). This is the first GHG reanalysis dataset ever produced using data assimilation techniques that adjust GHG concentrations rather than GHG fluxes. As such, it is a formidable achievement but at the same time, there is considerable room for improvement (as noted by the authors in sections 4 and 5). Reanalyses are typically improved after feedback from the user community and/or with improved data assimilation techniques or increased numbers of observations. Thus an important aspect of reanalysis production is the dissemination and documentation of the products to encourage and inform the potential user community. That is the main purpose of this article. As such, the comments below are primarily aimed at improving the description of the product to potential users.
  Specific comments
  Section 2.2 and Table 1: It would be useful to know how the satellite instruments were selected. For example, OCO-2 was not used. In section 5, it is mentioned that it will be used in future versions, but it is useful for the user to understand the rationale behind the selection of instruments for this product.
  
  The main rationale for the selection of the data was the operational availability of NRT data, as the re-analysis is foreseen to catch up and run in near-real time eventually. Currently there is no near-real time product of the OCO-2 data and that is why OCO-2 has not been used in the current CAMS near-real time products (including re-analysis). We are working with LSCE to have a near-real time OCO-2 product based on a neural network retrieval (Breon et al., 2022; https://amt.copernicus.org/articles/15/5219/2022/amt-15-5219-2022-discussion.html). This will allow us to use OCO-2 in the next CAMS re-analysis. The rationale for the observation data selection will be clarified in the revised manuscript.
  Line 99 and section 2.3: Figure 2 shows 2-way interaction between the forecast and surface fluxes. Please explain how the forecast influences the surface fluxes in section 2.3 or line 99.
  
  The two-way arrows indicate the two-way coupling between the atmospheric forecast and the biogenic surface fluxes. The surface fluxes are affected by the forecast of temperature, humidity, radiation and soil moisture and the atmospheric CO2 in the forecast is affected by the surface biogenic fluxes. This will be clarified in the revised version of the manuscript.
  Section 2.3: There is no mention of Figure 3 in the text. It is useful to understand why this Figure is presented, and the main message behind it.
  
  Figure 3 shows the seasonal, inter-annual variability and trend of the surface fluxes. This is important because one of the caveats of the re-analysis is that CH4 emissions are kept fixed from 2010 onwards, while for CO2 anthropogenic fluxes we apply an extrapolation, and the biogenic fluxes are modelled and therefore have an inter-annual variability. In the revised manuscript, Figure 3 is mentioned and referred to when explaining the differences between different fluxes in terms of inter-annual variability and seasonal cycles.
  Line 184: What is the rationale behind the choice of EDGAR versus higher spatial resolution datasets for anthropogenic emissions?
  
  EDGAR produces global emissions for both CO2 and CH4 at a relatively high resolution of 0.1 degrees (compared to 80km resolution of the CAMS re-analysis). The problem with EDGAR is the extension to NRT and the caveats associated with large inter-annual variability during the COVID period. This will be clarified in the revised manuscript.
  Line 217: Typo: Tl255 should be TL255 presumably.
  
  This will be corrected in the revised manuscript.
  Line 302: I do not see the 20 ppm error in Figure 5. If the vertical scale in Fig. 5c (bottom panel) is linear then the green and red dashed curves seem to overlap after 2019, meaning errors of less than 5 ppm.
  
  Yes, indeed. There is a typo here, 20ppm should be 2ppm. The sentence “accompanied by a large standard deviation error (~20ppm, cf. dashed lines in same panel and figure)” has been removed as the degradation associated with IASI is mainly detected from the large step change in the averages of the departures.
  Lines 368-370: Taylor diagrams of Figs. 8b, 11b. What is the normalization used on the radial axes? Presumably it is the observed standard deviation for a given site. Why was this normalization needed? Standard Taylor diagrams do not do a normalization. Presumably it allows for better comparison among sites with very different observed variability.
  
  The reviewer is correct. Assuming we have 20 sites, then there are 20 time series for the model and the observed NDACC ground based data (gb), each with its own variability. In order to plot them together, the standard deviation of the model time series is used to normalize the timeseries (i.e for a site, the time series model or gb is divided with the std of the model time series ): that is why the model time series is always at 1 for each site. The correlations are independent of such a normalization and from the location of a site in the Taylor plot you can deduce if the model has higher variability (the site is plotted with a distance <1 to the origin) or lower (the site is plotted at a distance >1 from the origin) compared to the variability in the gb data. This will be clarified in the revised manuscript.
  Lines 457-8 and Fig. 13: How was the site for Fig. 13 chosen? Is it a typical example, or a good example? It is nice to see that the overall structure of the profile (esp. for CO₂) is well captured in boreal summer in France. It would be interesting to see the comparison in the southern hemisphere. Are the general biases of Fig. 12 more evident in New Zealand?
  
  The site of Traînou (France) has been used as a good example of the synoptic variability during the summer, more than the systematic errors. As the reviewer suggested, we have plotted the only profile available from Lauder (New Zealand) (see Figure S1 in supplement) which shows much larger errors in September, particularly near the surface and in the stratosphere. In the revised manuscript we will add an extra figure as a Supplement to illustrate the errors at different sites across different latitude bands.
  
  Figure S1. Vertical mole fraction profiles of CO₂ and CH₄ from the CAMS GHG reanalysis (dash line) and AirCore observations (solid line) at Lauder (New Zealand, see Table A3) over the period in June 2019.
  
  Line 496: Since this is a reanalysis, why were the in situ data not assimilated? Would they be enough to better constrain the global growth rate? Will this be done in the future?
  
  The CAMS reanalysis and generally the reanalysis at ECMWF aim to eventually run in near-real time. Since the availability of quality controlled in situ data is not generally available in near real time and we use a global model with a short window data assimilation (i.e. 12 hours) that adjusts only the GHG concentrations (i.e. not the fluxes), the impact of the sparse in situ data would be small as the increments are very localised in time (i.e at the beginning of the data assimilation window) and space (around the station near the surface). Preliminary tests showed that the increments around the in situ stations are dispersed by the atmospheric transport in less than 12 hours. In the future, we will explore the possibility of using in situ data with the inversion capability currently being developed in the CoCO2 project. The flux adjustment will result in a longer-lived impact of the impact of the in situ observations, as we expect the resulting atmospheric CO2 and CH4 corrections will not be as localised in time and space, particularly when using a longer data assimilation window with the inversion capability. All these aspects related to the future potential assimilation of in situ data will be clarified in the revised version of the manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2022-283-AC4
AC1: 'Comment on egusphere-2022-283: Reply to reviewer 1', Anna Agusti-Panareda, 08 Dec 2022

We are very grateful for the insightful comments from the reviewers to improve the documentation of the CAMS greenhouse gas reanalysis for potential users. In the supplement files attached, we have addressed the comments of the reviewer 1 (in black) to clarify the different aspects of the reanalysis including the selection/quality of observations assimilation, the constraint of the annual global growth rates and the quality of the prescribed surface fluxes (see text in blue).

Citation: https://doi.org/10.5194/egusphere-2022-283-AC1

Status: closed

RC1:
'Comment on egusphere-2022-283', Anonymous Referee #2, 09 Aug 2022

Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.

This paper describes the greenhouse gases (CO2 and CH4) reanalysis as prepared for the Copernicus Atmosphere Monitoring Service (CAMS). Manuscript draft is well prepared for understading of the work that is carreid out. The method is well described but the results are not fully satisfacotry yet. Many traditional inversions produced better statistics for model simulated CO2 and CH4 concentrations by optimising the surface sources and sinks (e.g., Chandra et al., ACP 2022). However, the approach holds different promises for 4D concentration product of CO2 and CH4 in near realtime along with other meteorological parameters of ECMWF weather forecasting system. In general I recommend publication of the manuscript as "Technical note", after accounting for some of my comments below.

Line 48: are the SCIAMACHY and IASI data good enough for assimilation ? the growth rate may be fine but the spatial distribution are probably of poor quality.

I have found later that this is mentioned in the caveats. No action is needed but I am just making sure to mention the points of concern

Figure 1: I have several doubts on these plots.

1) the total column CO2 values on Dec 2020 are of similar magnitude of CO2 at MLO (414 pmm as per SIO flask data). I was thinking that global mean XCO2 will be a couple ppm lower than MLO value. Please confirm.

2) What is the advantage of showing 2003-2020 mean as opposed to just 2020 seasonal means ?

Figure 2: Should the two-way arrows between Forecast "Sphere" and Surface flxes "Maps" be one-way ? From Fluxes to Forecast

Table 2: You could use FF-CO2 emissions EDGARv6.0 etc. or the GridFED by UEA group. Similarly many of the flux components of CO2 and CH4 should be revised by using the most recent flux data sources. I not saying for this paper but in the near future.

Line 352: I think it would have been nice to add/subtract some CO2 flux to make the CO2 flux budget consistent with observed CO2 growth rate. At the very least the land biosphere + ocean exchange could be added/subtracted with an offset to balance all sources+sinks = 2.12 * Growth rate * 2.12. Then the statistics may be look that bad.

Figure 7: It would be nicer to use a discrete colour scale (like in Fig. 7). Otherwise, it is hard to distinguish between medium blue/red from dark blue/red

Also consider showing a Taylor diagram, like I see bellow for the TCCON sites

Do you need this Figure 14 ? If needed I suggest a more detailed site information is shown before Fig. 7. The Surface, MBL, TCCON, NDAAC, AirCore etc. sites can be shown in different symbols

Citation: https://doi.org/10.5194/egusphere-2022-283-RC1
- AC2: 'Reply on RC1', Anna Agusti-Panareda, 08 Dec 2022
  
  We are very grateful for the insightful comments from the reviewer to improve the documentation of the CAMS greenhouse gas reanalysis and in the manuscript and the suggestions for potential improvements to be considered in future reanalyses. We have addressed the different comments of the reviewer (in bold) to clarify the different aspects of the reanalysis including the selection/quality of observations assimilation, the constrain of the annual global growth rates and the quality of the prescribed surface fluxes.
  Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.
  This paper describes the greenhouse gases (CO2 and CH4) reanalysis as prepared for the Copernicus Atmosphere Monitoring Service (CAMS). Manuscript draft is well prepared for understanding of the work that is carried out. The method is well described but the results are not fully satisfactory yet. Many traditional inversions produced better statistics for model simulated CO2 and CH4 concentrations by optimising the surface sources and sinks (e.g., Chandra et al., ACP 2022). However, the approach holds different promises for 4D concentration product of CO2 and CH4 in near real time along with other meteorological parameters of ECMWF weather forecasting system. In general I recommend publication of the manuscript as "Technical note", after accounting for some of my comments below.
  Line 48: are the SCIAMACHY and IASI data good enough for assimilation ? the growth rate may be fine but the spatial distribution are probably of poor quality.
  
  I have found later that this is mentioned in the caveats. No action is needed but I am just making sure to mention the points of concern.
  The rationale for the selection of the observations assimilated in the CAMS reanalysis and the importance of assessing their quality will be emphasized in the revised version of the manuscript.
  Figure 1: I have several doubts on these plots.
  
  1) the total column CO2 values on Dec 2020 are of similar magnitude of CO2 at MLO (414 pm as per SIO flask data). I was thinking that global mean XCO2 will be a couple ppm lower than MLO value. Please confirm.
  
  Figure 1 gives an overview of the CAMS GHG reanalysis dataset including global trends and spatial/temporal distribution at seasonal sales. The global mean has an error ranging from -0.7 to 3.5 ppm. Figure 14 shows that global mean error in 2020 is +1.35 ppm. This is consistent with the assessment of the reviewer. The problem of constraining the global mean and annual growth rate is documented in section 3.3 and mentioned in section 4 (point 6). The range of error in the global mean will be added in the revised version of the manuscript.
  
  2) What is the advantage of showing 2003-2020 mean as opposed to just 2020 seasonal means?
  The purpose here is to show the typical seasonal cycle. For this reason, using 2003-2020 mean as opposed to just 2020 seasonal means is better because individual years can be affected by the large inter-annual variability of biogenic fluxes (e.g. during el Nino years). This will be mentioned in the revised manuscript.
  Figure 2: Should the two-way arrows between Forecast "Sphere" and Surface flxes "Maps" be one-way? From Fluxes to Forecast
  The two-way arrows indicate the two-way coupling between the atmospheric model and the model of biogenic surface fluxes. The surface fluxes are affected by the forecast of temperature, radiation, humidity and soil moisture and the atmospheric CO2 in the forecast is affected by the surface biogenic fluxes. This will be clarified in the revised version of the manuscript.
  Table 2: You could use FF-CO2 emissions EDGARv6.0 etc. or the GridFED by UEA group. Similarly many of the flux components of CO2 and CH4 should be revised by using the most recent flux data sources. I not saying for this paper but in the near future.
  The current reanalysis cannot be changed as the reanalysis system requires a consistent configuration of the model and prescribed input fluxes throughout the whole reanalysis, but we will certainly use a newer version of the anthropogenic emissions in the next CAMS reanalysis planned for 2024. This will be mentioned in the revised manuscript.
  Line 352: I think it would have been nice to add/subtract some CO2 flux to make the CO2 flux budget consistent with observed CO2 growth rate. At the very least the land biosphere + ocean exchange could be added/subtracted with an offset to balance all sources+sinks = 2.12 * Growth rate. Then the statistics may be look that bad.
  This adjustment of the fluxes with the CO2 growth rate will be considered in the next CAMS re-analysis. There is the caveat that the growth rate will have to be adjusted in near-real time which might not be feasible. A note of this option will be made in the discussion section of the revised manuscript.
  Figure 7: It would be nicer to use a discrete colour scale (like in Fig. 7). Otherwise, it is hard to distinguish between medium blue/red from dark blue/red
  Also consider showing a Taylor diagram, like I see below for the TCCON sites
  We will change Figure 7 to use a discrete colour scale and add a Taylor diagram in the revised manuscript.
  Do you need this Figure 14 ? If needed I suggest a more detailed site information is shown before Fig. 7. The Surface, MBL, TCCON, NDAAC, AirCore etc. sites can be shown in different symbols
  The map of the MBL sites is useful to be able to interpret the observed global mean produced by NOAA. A detailed list of TCCON, The NDACC, AirCore is provided in the Appendix. In the revised manuscript the map in Figure 14 will be updated to show the different types of observations with different symbols and placed before Fig. 7 as suggested by the reviewer.
  
  Citation: https://doi.org/10.5194/egusphere-2022-283-AC2
RC2:
'Comment on egusphere-2022-283', Anonymous Referee #3, 19 Aug 2022
This article presents a description of a new reanalysis dataset of greenhouse gases (GHGs). This is the first GHG reanalysis dataset ever produced using data assimilation techniques that adjust GHG concentrations rather than GHG fluxes. As such, it is a formidable achievement but at the same time, there is considerable room for improvement (as noted by the authors in sections 4 and 5). Reanalyses are typically improved after feedback from the user community and/or with improved data assimilation techniques or increased numbers of observations. Thus an important aspect of reanalysis production is the dissemination and documentation of the products to encourage and inform the potential user community. That is the main purpose of this article. As such, the comments below are primarily aimed at improving the description of the product to potential users.

Specific comments

Section 2.2 and Table 1: It would be useful to know how the satellite instruments were selected. For example, OCO-2 was not used. In section 5, it is mentioned that it will be used in future versions, but it is useful for the user to understand the rationale behind the selection of instruments for this product.

Line 99 and section 2.3: Figure 2 shows 2-way interaction between the forecast and surface fluxes. Please explain how the forecast influences the surface fluxes in section 2.3 or line 99.

Section 2.3: There is no mention of Figure 3 in the text. It is useful to understand why this Figure is presented, and the main message behind it.

Line 184: What is the rationale behind the choice of EDGAR versus higher spatial resolution datasets for anthropogenic emissions?

Line 217: Typo: Tl255 should be TL255 presumably.

Line 302: I do not see the 20 ppm error in Figure 5. If the vertical scale in Fig. 5c (bottom panel) is linear then the green and red dashed curves seem to overlap after 2019, meaning errors of less than 5 ppm.

Lines 368-370: Taylor diagrams of Figs. 8b, 11b. What is the normalization used on the radial axes? Presumably it is the observed standard deviation for a given site. Why was this normalization needed? Standard Taylor diagrams do not do a normalization. Presumably it allows for better comparison among sites with very different observed variability.

Lines 457-8 and Fig. 13: How was the site for Fig. 13 chosen? Is it a typical example, or a good example? It is nice to see that the overall structure of the profile (esp. for CO₂) is well captured in boreal summer in France. It would be interesting to see the comparison in the southern hemisphere. Are the general biases of Fig. 12 more evident in New Zealand?

Line 496: Since this is a reanalysis, why were the insitu data not assimilated? Would they be enough to better constrain the global growth rate? Will this be done in the future?
Citation: https://doi.org/10.5194/egusphere-2022-283-RC2
- AC3:
  'Reply on RC2', Anna Agusti-Panareda, 08 Dec 2022
  We are very grateful for the insightful comments from the reviewer to improve the documentation of the CAMS greenhouse gas reanalysis for potential users. We have addressed the comments of the reviewer (in bold) to clarify the different aspects of the reanalysis including the selection/quality of the assimilated observations, the different model components and the characteristics of the prescribed surface fluxes.
  Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.
  This article presents a description of a new reanalysis dataset of greenhouse gases (GHGs). This is the first GHG reanalysis dataset ever produced using data assimilation techniques that adjust GHG concentrations rather than GHG fluxes. As such, it is a formidable achievement but at the same time, there is considerable room for improvement (as noted by the authors in sections 4 and 5). Reanalyses are typically improved after feedback from the user community and/or with improved data assimilation techniques or increased numbers of observations. Thus an important aspect of reanalysis production is the dissemination and documentation of the products to encourage and inform the potential user community. That is the main purpose of this article. As such, the comments below are primarily aimed at improving the description of the product to potential users.
  Specific comments
  Section 2.2 and Table 1: It would be useful to know how the satellite instruments were selected. For example, OCO-2 was not used. In section 5, it is mentioned that it will be used in future versions, but it is useful for the user to understand the rationale behind the selection of instruments for this product.
  
  The main rationale for the selection of the data was the operational availability of NRT data, as the re-analysis is foreseen to catch up and run in near-real time eventually. Currently there is no near-real time product of the OCO-2 data and that is why OCO-2 has not been used in the current CAMS near-real time products (including re-analysis). We are working with LSCE to have a near-real time OCO-2 product based on a neural network retrieval (Breon et al., 2022; https://amt.copernicus.org/articles/15/5219/2022/amt-15-5219-2022-discussion.html). This will allow us to use OCO-2 in the next CAMS re-analysis. The rationale for the observation data selection will be clarified in the revised manuscript.
  Line 99 and section 2.3: Figure 2 shows 2-way interaction between the forecast and surface fluxes. Please explain how the forecast influences the surface fluxes in section 2.3 or line 99.
  
  The two-way arrows indicate the two-way coupling between the atmospheric forecast and the biogenic surface fluxes. The surface fluxes are affected by the forecast of temperature, humidity, radiation and soil moisture and the atmospheric CO2 in the forecast is affected by the surface biogenic fluxes. This will be clarified in the revised version of the manuscript.
  Section 2.3: There is no mention of Figure 3 in the text. It is useful to understand why this Figure is presented, and the main message behind it.
  
  Figure 3 shows the seasonal, inter-annual variability and trend of the surface fluxes. This is important because one of the caveats of the re-analysis is that CH4 emissions are kept fixed from 2010 onwards, while for CO2 anthropogenic fluxes we apply an extrapolation, and the biogenic fluxes are modelled and therefore have an inter-annual variability. In the revised manuscript, Figure 3 is mentioned and referred to when explaining the differences between different fluxes in terms of inter-annual variability and seasonal cycles.
  Line 184: What is the rationale behind the choice of EDGAR versus higher spatial resolution datasets for anthropogenic emissions?
  
  EDGAR produces global emissions for both CO2 and CH4 at a relatively high resolution of 0.1 degrees (compared to 80km resolution of the CAMS re-analysis). The problem with EDGAR is the extension to NRT and the caveats associated with large inter-annual variability during the COVID period. This will be clarified in the revised manuscript.
  Line 217: Typo: Tl255 should be TL255 presumably.
  
  This will be corrected in the revised manuscript.
  Line 302: I do not see the 20 ppm error in Figure 5. If the vertical scale in Fig. 5c (bottom panel) is linear then the green and red dashed curves seem to overlap after 2019, meaning errors of less than 5 ppm.
  
  Yes, indeed. There is a typo here, 20ppm should be 2ppm. The sentence “accompanied by a large standard deviation error (~20ppm, cf. dashed lines in same panel and figure)” has been removed as the degradation associated with IASI is mainly detected from the large step change in the averages of the departures.
  Lines 368-370: Taylor diagrams of Figs. 8b, 11b. What is the normalization used on the radial axes? Presumably it is the observed standard deviation for a given site. Why was this normalization needed? Standard Taylor diagrams do not do a normalization. Presumably it allows for better comparison among sites with very different observed variability.
  
  The reviewer is correct. Assuming we have 20 sites, then there are 20 time series for the model and the observed NDACC ground based data (gb), each with its own variability. In order to plot them together, the standard deviation of the model time series is used to normalize the timeseries (i.e for a site, the time series model or gb is divided with the std of the model time series ): that is why the model time series is always at 1 for each site. The correlations are independent of such a normalization and from the location of a site in the Taylor plot you can deduce if the model has higher variability (the site is plotted with a distance <1 to the origin) or lower (the site is plotted at a distance >1 from the origin) compared to the variability in the gb data. This will be clarified in the revised manuscript.
  Lines 457-8 and Fig. 13: How was the site for Fig. 13 chosen? Is it a typical example, or a good example? It is nice to see that the overall structure of the profile (esp. for CO₂) is well captured in boreal summer in France. It would be interesting to see the comparison in the southern hemisphere. Are the general biases of Fig. 12 more evident in New Zealand?
  
  The site of Traînou (France) has been used as a good example of the synoptic variability during the summer, more than the systematic errors. As the reviewer suggested, we have plotted the only profile available from Lauder (New Zealand) (see Figure S1 in supplement) which shows much larger errors in September, particularly near the surface and in the stratosphere. In the revised manuscript we will add an extra figure as a Supplement to illustrate the errors at different sites across different latitude bands.
  
  Figure S1. Vertical mole fraction profiles of CO₂ and CH₄ from the CAMS GHG reanalysis (dash line) and AirCore observations (solid line) at Lauder (New Zealand, see Table A3) over the period in June 2019.
  
  Line 496: Since this is a reanalysis, why were the in situ data not assimilated? Would they be enough to better constrain the global growth rate? Will this be done in the future?
  
  The CAMS reanalysis and generally the reanalysis at ECMWF aim to eventually run in near-real time. Since the availability of quality controlled in situ data is not generally available in near real time and we use a global model with a short window data assimilation (i.e. 12 hours) that adjusts only the GHG concentrations (i.e. not the fluxes), the impact of the sparse in situ data would be small as the increments are very localised in time (i.e at the beginning of the data assimilation window) and space (around the station near the surface). Preliminary tests showed that the increments around the in situ stations are dispersed by the atmospheric transport in less than 12 hours. In the future, we will explore the possibility of using in situ data with the inversion capability currently being developed in the CoCO2 project. The flux adjustment will result in a longer-lived impact of the impact of the in situ observations, as we expect the resulting atmospheric CO2 and CH4 corrections will not be as localised in time and space, particularly when using a longer data assimilation window with the inversion capability. All these aspects related to the future potential assimilation of in situ data will be clarified in the revised version of the manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2022-283-AC3
- AC4:
  'Reply on RC2', Anna Agusti-Panareda, 08 Dec 2022
  We are very grateful for the insightful comments from the reviewer to improve the documentation of the CAMS greenhouse gas reanalysis for potential users. We have addressed the comments of the reviewer (in bold) to clarify the different aspects of the reanalysis including the selection/quality of the assimilated observations, the different model components and the characteristics of the prescribed surface fluxes.
  Review of "The CAMS greenhouse gas reanalysis from 2003 to 2020" by Agusti-Panareda et al.
  This article presents a description of a new reanalysis dataset of greenhouse gases (GHGs). This is the first GHG reanalysis dataset ever produced using data assimilation techniques that adjust GHG concentrations rather than GHG fluxes. As such, it is a formidable achievement but at the same time, there is considerable room for improvement (as noted by the authors in sections 4 and 5). Reanalyses are typically improved after feedback from the user community and/or with improved data assimilation techniques or increased numbers of observations. Thus an important aspect of reanalysis production is the dissemination and documentation of the products to encourage and inform the potential user community. That is the main purpose of this article. As such, the comments below are primarily aimed at improving the description of the product to potential users.
  Specific comments
  Section 2.2 and Table 1: It would be useful to know how the satellite instruments were selected. For example, OCO-2 was not used. In section 5, it is mentioned that it will be used in future versions, but it is useful for the user to understand the rationale behind the selection of instruments for this product.
  
  The main rationale for the selection of the data was the operational availability of NRT data, as the re-analysis is foreseen to catch up and run in near-real time eventually. Currently there is no near-real time product of the OCO-2 data and that is why OCO-2 has not been used in the current CAMS near-real time products (including re-analysis). We are working with LSCE to have a near-real time OCO-2 product based on a neural network retrieval (Breon et al., 2022; https://amt.copernicus.org/articles/15/5219/2022/amt-15-5219-2022-discussion.html). This will allow us to use OCO-2 in the next CAMS re-analysis. The rationale for the observation data selection will be clarified in the revised manuscript.
  Line 99 and section 2.3: Figure 2 shows 2-way interaction between the forecast and surface fluxes. Please explain how the forecast influences the surface fluxes in section 2.3 or line 99.
  
  The two-way arrows indicate the two-way coupling between the atmospheric forecast and the biogenic surface fluxes. The surface fluxes are affected by the forecast of temperature, humidity, radiation and soil moisture and the atmospheric CO2 in the forecast is affected by the surface biogenic fluxes. This will be clarified in the revised version of the manuscript.
  Section 2.3: There is no mention of Figure 3 in the text. It is useful to understand why this Figure is presented, and the main message behind it.
  
  Figure 3 shows the seasonal, inter-annual variability and trend of the surface fluxes. This is important because one of the caveats of the re-analysis is that CH4 emissions are kept fixed from 2010 onwards, while for CO2 anthropogenic fluxes we apply an extrapolation, and the biogenic fluxes are modelled and therefore have an inter-annual variability. In the revised manuscript, Figure 3 is mentioned and referred to when explaining the differences between different fluxes in terms of inter-annual variability and seasonal cycles.
  Line 184: What is the rationale behind the choice of EDGAR versus higher spatial resolution datasets for anthropogenic emissions?
  
  EDGAR produces global emissions for both CO2 and CH4 at a relatively high resolution of 0.1 degrees (compared to 80km resolution of the CAMS re-analysis). The problem with EDGAR is the extension to NRT and the caveats associated with large inter-annual variability during the COVID period. This will be clarified in the revised manuscript.
  Line 217: Typo: Tl255 should be TL255 presumably.
  
  This will be corrected in the revised manuscript.
  Line 302: I do not see the 20 ppm error in Figure 5. If the vertical scale in Fig. 5c (bottom panel) is linear then the green and red dashed curves seem to overlap after 2019, meaning errors of less than 5 ppm.
  
  Yes, indeed. There is a typo here, 20ppm should be 2ppm. The sentence “accompanied by a large standard deviation error (~20ppm, cf. dashed lines in same panel and figure)” has been removed as the degradation associated with IASI is mainly detected from the large step change in the averages of the departures.
  Lines 368-370: Taylor diagrams of Figs. 8b, 11b. What is the normalization used on the radial axes? Presumably it is the observed standard deviation for a given site. Why was this normalization needed? Standard Taylor diagrams do not do a normalization. Presumably it allows for better comparison among sites with very different observed variability.
  
  The reviewer is correct. Assuming we have 20 sites, then there are 20 time series for the model and the observed NDACC ground based data (gb), each with its own variability. In order to plot them together, the standard deviation of the model time series is used to normalize the timeseries (i.e for a site, the time series model or gb is divided with the std of the model time series ): that is why the model time series is always at 1 for each site. The correlations are independent of such a normalization and from the location of a site in the Taylor plot you can deduce if the model has higher variability (the site is plotted with a distance <1 to the origin) or lower (the site is plotted at a distance >1 from the origin) compared to the variability in the gb data. This will be clarified in the revised manuscript.
  Lines 457-8 and Fig. 13: How was the site for Fig. 13 chosen? Is it a typical example, or a good example? It is nice to see that the overall structure of the profile (esp. for CO₂) is well captured in boreal summer in France. It would be interesting to see the comparison in the southern hemisphere. Are the general biases of Fig. 12 more evident in New Zealand?
  
  The site of Traînou (France) has been used as a good example of the synoptic variability during the summer, more than the systematic errors. As the reviewer suggested, we have plotted the only profile available from Lauder (New Zealand) (see Figure S1 in supplement) which shows much larger errors in September, particularly near the surface and in the stratosphere. In the revised manuscript we will add an extra figure as a Supplement to illustrate the errors at different sites across different latitude bands.
  
  Figure S1. Vertical mole fraction profiles of CO₂ and CH₄ from the CAMS GHG reanalysis (dash line) and AirCore observations (solid line) at Lauder (New Zealand, see Table A3) over the period in June 2019.
  
  Line 496: Since this is a reanalysis, why were the in situ data not assimilated? Would they be enough to better constrain the global growth rate? Will this be done in the future?
  
  The CAMS reanalysis and generally the reanalysis at ECMWF aim to eventually run in near-real time. Since the availability of quality controlled in situ data is not generally available in near real time and we use a global model with a short window data assimilation (i.e. 12 hours) that adjusts only the GHG concentrations (i.e. not the fluxes), the impact of the sparse in situ data would be small as the increments are very localised in time (i.e at the beginning of the data assimilation window) and space (around the station near the surface). Preliminary tests showed that the increments around the in situ stations are dispersed by the atmospheric transport in less than 12 hours. In the future, we will explore the possibility of using in situ data with the inversion capability currently being developed in the CoCO2 project. The flux adjustment will result in a longer-lived impact of the impact of the in situ observations, as we expect the resulting atmospheric CO2 and CH4 corrections will not be as localised in time and space, particularly when using a longer data assimilation window with the inversion capability. All these aspects related to the future potential assimilation of in situ data will be clarified in the revised version of the manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2022-283-AC4
AC1: 'Comment on egusphere-2022-283: Reply to reviewer 1', Anna Agusti-Panareda, 08 Dec 2022

We are very grateful for the insightful comments from the reviewers to improve the documentation of the CAMS greenhouse gas reanalysis for potential users. In the supplement files attached, we have addressed the comments of the reviewer 1 (in black) to clarify the different aspects of the reanalysis including the selection/quality of observations assimilation, the constraint of the annual global growth rates and the quality of the prescribed surface fluxes (see text in blue).

Citation: https://doi.org/10.5194/egusphere-2022-283-AC1

Anna Agusti-Panareda et al.

Data sets

CAMS global greenhouse gas reanalysis (EGG4) Copernicus Atmosphere Monitoring Service https://doi.org/10.24380/8fck-9w87

Anna Agusti-Panareda et al.

Viewed

Total article views: 1,019 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
647	349	23	1,019	15	17

HTML: 647
PDF: 349
XML: 23
Total: 1,019
BibTeX: 15
EndNote: 17

Views and downloads (calculated since 18 Jul 2022)

Month	HTML	PDF	XML	Total
Jul 2022	215	90	6	311
Aug 2022	133	48	7	188
Sep 2022	49	47	0	96
Oct 2022	40	18	1	59
Nov 2022	33	32	1	66
Dec 2022	55	33	2	90
Jan 2023	38	23	0	61
Feb 2023	51	34	2	87
Mar 2023	33	24	4	61

Cumulative views and downloads (calculated since 18 Jul 2022)

Month	HTML	PDF	XML	Total
Jul 2022	215	90	6	311
Aug 2022	133	48	7	188
Sep 2022	49	47	0	96
Oct 2022	40	18	1	59
Nov 2022	33	32	1	66
Dec 2022	55	33	2	90
Jan 2023	38	23	0	61
Feb 2023	51	34	2	87
Mar 2023	33	24	4	61

Viewed (geographical distribution)

Total article views: 1,027 (including HTML, PDF, and XML) Thereof 1,027 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 26 Mar 2023

Short summary

We present a global dataset of atmospheric CO₂ and CH₄, the two most important human-made greenhouse gases, which covers almost two decades (2003–2020). It is produced by combining satellite data of CO₂ and CH₄ with a weather and air composition prediction model, and it has been carefully evaluated against independent observations to ensure validity and point out deficiencies to the user. This dataset can be used for scientific studies in the field of climate change and the global carbon cycle.


Total:	0
HTML:	0
PDF:	0
XML:	0