Near real-time CO2 fluxes from CarbonTracker Europe for high resolution atmospheric modeling
- 1Centre for Isotope Research, Energy and Sustainability Research Institute Groningen, University of Groningen, Nijenborgh 4, 9747 AG Groningen, the Netherlands
- 2Meteorology and Air Quality Group, Wageningen University, P.O. Box 47, 6700 AA Wageningen, the Netherlands
- 3ICOS ERIC, Carbon Portal, Geocentrum II, Sölvegatan 1222362 Lund, Sweden
- 4Max Planck Institute for Biogeochemistry, Hans-Knoell-Straße 10, 07745 Jena, Germany
- 5Department of Physical geography and Ecosystem Sciences, Lund University, Box 117, SE-221 00, Lund, Sweden
- 6Copernicus Institute of Sustainable Development, Utrecht University, Utrecht, the Netherlands
- 7Department of Climate, Air and Sustainability, TNO, P.O. Box 80015, 3508 TA Utrecht, the Netherlands
- 1Centre for Isotope Research, Energy and Sustainability Research Institute Groningen, University of Groningen, Nijenborgh 4, 9747 AG Groningen, the Netherlands
- 2Meteorology and Air Quality Group, Wageningen University, P.O. Box 47, 6700 AA Wageningen, the Netherlands
- 3ICOS ERIC, Carbon Portal, Geocentrum II, Sölvegatan 1222362 Lund, Sweden
- 4Max Planck Institute for Biogeochemistry, Hans-Knoell-Straße 10, 07745 Jena, Germany
- 5Department of Physical geography and Ecosystem Sciences, Lund University, Box 117, SE-221 00, Lund, Sweden
- 6Copernicus Institute of Sustainable Development, Utrecht University, Utrecht, the Netherlands
- 7Department of Climate, Air and Sustainability, TNO, P.O. Box 80015, 3508 TA Utrecht, the Netherlands
Abstract. We present the CarbonTracker Europe High-Resolution system that estimates carbon dioxide (CO2) exchange over Europe at high-resolution (0.1 x 0.2°) and in near real-time (about 2 months latency). It includes a dynamic fossil fuel emission model, which uses easily available statistics on economic activity, energy-use, and weather to generate fossil fuel emissions with dynamic time profiles at high spatial and temporal resolution (0.1 x 0.2°, hourly). Hourly net biosphere exchange (NEE) calculated by the Simple Biosphere model Version 4 (SiB4) is driven by meteorology from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis 5th Generation (ERA5) dataset. This NEE is downscaled to 0.1 x 0.2° using the high-resolution Coordination of Information on the Environment (CORINE) land-cover map, and combined with the Global Fire Assimilation System (GFAS) fire emissions to create terrestrial carbon fluxes. An ocean flux extrapolation and downscaling based on wind speed and temperature for Jena CarboScope ocean CO2 fluxes is included in our product. Jointly, these flux estimates enable modeling of atmospheric CO2 mole fractions over Europe.
We assess the ability of the CTE-HR CO2 fluxes (a) to reproduce observed anomalies in biospheric fluxes and atmospheric CO2 mole fractions during the 2018 drought, (b) to capture the reduction of fossil fuel emissions due to COVID-19 lockdowns, (c) to match mole fraction observations at Integrated Carbon Observation System (ICOS) sites across Europe after atmospheric transport with the Transport Model, version 5 (TM5) and the Stochastic Time-Inverted Lagrangian Transport (STILT), driven by ERA5, and (d) to capture the magnitude and variability of measured CO2 fluxes in the city centre of Amsterdam (The Netherlands).
We show that CTE-HR fluxes reproduce large-scale flux anomalies reported in previous studies for both biospheric fluxes (drought of 2018) and fossil fuel emissions (COVID-19 pandemic in 2020). After transport with TM5, the CTE-HR fluxes have lower root mean square errors (RMSEs) relative to mole fraction observations than fluxes from a non-informed flux estimate, in which biosphere fluxes are scaled to match the global growth rate of CO2 (poor-person inversion). RSMEs are close to those of the reanalysis with the data assimilation system CarbonTracker Europe (CTE). This is encouraging given that CTE-HR fluxes did not profit from the weekly assimilation of CO2 observations as in CTE.
We furthermore compare CO2 observations at the Dutch Lutjewad coastal tower with high-resolution STILT transport to show that the high-resolution fluxes manifest variability due to different sectors in summer and winter. Interestingly, in periods where synoptic scale transport variability dominates CO2 variations, the CTE-HR fluxes perform similar to low-resolution fluxes (5–10x coarsened). The remaining 10 % of simulated CO2 mole fraction differ by > 2ppm between the low-resolution and high-resolution flux representation, and are clearly associated with coherent structures ("plumes") originating from emission hotspots, such as power plants. We therefore note that the added resolution of our product will matter most for very specific locations and times when used for atmospheric CO2 modeling. Finally, in a densely-populated region like the Amsterdam city centre, our fluxes underestimate the magnitude of measured eddy-covariance fluxes, but capture their substantial diurnal variations in summer- and wintertime well.
We conclude that our product is a promising tool to model the European carbon budget at a high-resolution in near real-time. The fluxes are freely available from the ICOS Carbon Portal (CC-BY-4.0) to be used for near real-time monitoring and modeling, for example as a-priori flux product in a CO2 data-assimilation system. The data is available at https://doi.org/10.18160/20Z1-AYJ2.
Auke Marijn van der Woude et al.
Status: closed
-
RC1: 'review comment on essd-2022-175', Anonymous Referee #1, 31 Jul 2022
General comments.
The manuscript presents a construction of surface CO2 flux dataset for European domain, designed for near-realtime updates and based on recent energy statistics and weather data. The flux dataset was evaluated via comparing the transport model-simulated timeseries to atmospheric observations and comparing fluxes to an urban flux tower data and demonstrated a reasonably good performance. The manuscript only requires minor revisions and can be accepted after implementing corrections and amendments based on suggestions provided below.
Detailed comments.
It is advisable to enhance the part on validation of the SiB4 CO2 terrestrial biospheric fluxes (in terms as midday fluxes or daytime mean, where fluxes are more curtain) against data-driven products available at similar resolution (~0.1): eg FLUXCOMM (Jung et al. 2020), Zeng et al (2020) or others, which can be reported summarily by region or dominant vegetation type. It is mentioned that Smith et al. 2020 did evaluation, but they did not have advantage of using a downscaled product which makes matching resolutions easier.
Technical corrections.
L1 ‘We present the CarbonTracker Europe High-Resolution system’. General reader familiar with CTE would suspect the CTE-HR is an inverse modelling system like CTE, but here the name CTE-HR is given to a set of [prior] fluxes, so it is better to explain the difference in the abstract, eg write something like: ‘We present the CarbonTracker Europe High-Resolution system fluxes’, and note that these fluxes are unoptimized (either in the abstract or in the text, eg lines 115-118)
L49 For Paris example (Breon et al 2015), one can cite a recent paper by Nalini et al 2022
L54 N2 -> N2O. In case of Swiss methane emissions, Henne et al 2016 is more widely cited.
L86-89 As for high resolution, operational NRT biospheric flux products one can mention SMAP L4C (Jones et al 2017)
L96 CAMS is associated with a wide variety of products related to the topic of this paper, better give more specific name like CAMS-REG. Also, the paper by Kuenen et al 2022 appears to document CAMS-REG-AP (air pollutant) inventory, while GHG portion called CAMS-REG-GHG was set aside.
L168-184 It is not clear, what data is used for spatial emission patterns.
L215 Final revised paper for GFAS is Di Guiseppe et al. 2018 (https://doi.org/10.5194/acp-18-5359-2018)
L231 Can cite here the (Chevallier et al 2019) method as ‘poor man’s inversion’ as done by Chevallier et al. (2010, 2019)
L249 “Emissions by” can be omitted.
L255 ‘exact monthly growth rate’ may draw doubts, as the rate is not that exact, can write ‘exactly follow monthly growth rate’ instead.
L280 Need to add spatial/vertical resolution at which IFS winds are used, and STILT domain geographical boundaries.
L471 Suggest clarifying the text: “Currently, CTE-HR only provides CO2 fluxes”. Better be more specific: eg biogenic/biospheric/net ….
L524 Here, fluxes ‘will be made available ', while on L518 same fluxes ‘are available’
References
Chevallier, F., et al.: CO2 surface fluxes at grid point scale estimated from a global 21 year reanalysis of atmospheric measurements, J. Geophys. Res., 115, D21307, doi:10.1029/2010JD013887, 2010
Chevallier, F., Remaud, M., O'Dell, C. W., Baker, D., Peylin, P., and Cozic, A.: Objective evaluation of surface- and satellite-driven carbon dioxide atmospheric inversions, Atmos. Chem. Phys., 19, 14233–14251, https://doi.org/10.5194/acp-19-14233-2019, 2019.
Henne, S., Brunner, D., Oney, B., Leuenberger, et al. Validation of the Swiss methane emission inventory by atmospheric observations and inverse modelling, Atmos. Chem. Phys., 16, 3683–3710, https://doi.org/10.5194/acp-16-3683-2016, 2016.
Jones, L., et al.: The SMAP Level 4 Carbon Product for Monitoring Ecosystem Land-Atmosphere COâ Exchange. IEEE Transactions on Geoscience and Remote Sensing. 55. 6517-6532. 10.1109/TGRS.2017.2729343. 2017.
Nalini, K., Lauvaux, T., Abdallah, C., Lian, J., Ciais, P., Utard, H., et al. (2022). High-resolution Lagrangian inverse modeling of CO2 emissions over the Paris region during the first 2020 lockdown period. Journal of Geophysical Research: Atmospheres, 127, e2021JD036032. https://doi.org/10.1029/2021JD036032
Zeng, J., et al. Global terrestrial carbon fluxes of 1999–2019 estimated by upscaling eddy covariance data with a random forest. Sci Data 7, 313 (2020). https://doi.org/10.1038/s41597-020-00653-5
-
RC2: 'Comment on essd-2022-175', Anonymous Referee #2, 30 Sep 2022
The study by van der Woude et al. (2022) described the CarbonTraker Europe High-Resolution (CT-HR) system that estimates carbon dioxide exchange over Europe at high-resolution (0.1 x0.2) in near real-time (about 2-months latency). The system includes a dynamic fossil fuel emission model, hourly net biosphere exchange (NEE), and an ocean flux. All these flux components are hourly. The high spatiotemporal resolution is realized by incorporating easily available statistics on economic activity, energy use, weather, and land cover map. The study evaluated the CT-HR in several aspects, including: 1) comparing the simulated CO2 to atmosphere CO2 observations at ICOS sites; 2) analyzing the 2018 Europe drought signal; 3) analyzing fossil fuel emission anomalies due to the COVID lockdown; and 4) comparing the hourly fluxes against observations from a flux tower at a city center. Overall, the paper is well written. The low-latency high-resolution data will facilitate the high-resolution atmospheric flux inversions, and the near-real time aspect can provide rapid information on carbon cycle changes over Europe. My main concern is the uncertainty quantification and the evaluation of the high spatiotemporal resolution aspect of the product. Please see my detailed comments below.
- For each of the flux components, it would be critical to have uncertainty estimation in order to use them in the atmospheric flux inversions and provide useful information on carbon cycle changes, which are the two applications envisioned by the authors. To some degree, the uncertainty is as important as the absolute flux estimation itself for atmospheric flux inversion application. So I would suggest adding uncertainty estimation in the product that propagate the uncertainty of the input data to the uncertainty of the end products.
- The evaluation of the high spatiotemporal feature of the product is not sufficient. The paper underwent extensive effort to evaluate the product, but the evaluation is mainly on large spatial and monthly time scale (except the tower comparison and partially the ICOS comparison), which are actually mainly from the input dataset. For example, the 2018 drought was evaluated on an aggregated region, the COVID signal was also evaluated on the continental scale by comparing to the published results. As the high spatiotemporal feature is the main advancement, I would recommend adding more evaluation at finer spatiotemporal scales. For example, the observations from flux towers can be used to evaluate the hourly NEE from SIB4.
- It is not clear how the diurnal cycle of SIB4 was calculated, and what is the temporal frequency of SIB4 output. Is the original SIB4 output monthly or daily?
- Figure 6 shows that the HR almost has the same mean RMSE as the PRI, but the text claims that the CT-HR performs better. It would be informative to include the exact RMSE from PRI and CT-HR when comparing to ICOS network in the text.
- Figure 7 shows that all three fluxes give very similar results during well-mixed hours, but it is hard to tell which one has better performance at the bottom two panels since the scale is so large. I would suggest adding panels to show the difference between the simulated run and the measurements. I would also suggest discussing RMSE of CO2 forced by these three fluxes during well-mixed hours and during night and early morning.
- It is strange that the “Flat” fluxes seem to have the same performance as the “full res” in December, which seem to indicate potential large errors in boundary layer height. I would suggest evaluating boundary layer height information used in STILT and isolate the impact of transport and fluxes on the flux errors shown in Figure 7.
- AC1: 'Comment on essd-2022-175', Auke Van Der Woude, 02 Dec 2022
Status: closed
-
RC1: 'review comment on essd-2022-175', Anonymous Referee #1, 31 Jul 2022
General comments.
The manuscript presents a construction of surface CO2 flux dataset for European domain, designed for near-realtime updates and based on recent energy statistics and weather data. The flux dataset was evaluated via comparing the transport model-simulated timeseries to atmospheric observations and comparing fluxes to an urban flux tower data and demonstrated a reasonably good performance. The manuscript only requires minor revisions and can be accepted after implementing corrections and amendments based on suggestions provided below.
Detailed comments.
It is advisable to enhance the part on validation of the SiB4 CO2 terrestrial biospheric fluxes (in terms as midday fluxes or daytime mean, where fluxes are more curtain) against data-driven products available at similar resolution (~0.1): eg FLUXCOMM (Jung et al. 2020), Zeng et al (2020) or others, which can be reported summarily by region or dominant vegetation type. It is mentioned that Smith et al. 2020 did evaluation, but they did not have advantage of using a downscaled product which makes matching resolutions easier.
Technical corrections.
L1 ‘We present the CarbonTracker Europe High-Resolution system’. General reader familiar with CTE would suspect the CTE-HR is an inverse modelling system like CTE, but here the name CTE-HR is given to a set of [prior] fluxes, so it is better to explain the difference in the abstract, eg write something like: ‘We present the CarbonTracker Europe High-Resolution system fluxes’, and note that these fluxes are unoptimized (either in the abstract or in the text, eg lines 115-118)
L49 For Paris example (Breon et al 2015), one can cite a recent paper by Nalini et al 2022
L54 N2 -> N2O. In case of Swiss methane emissions, Henne et al 2016 is more widely cited.
L86-89 As for high resolution, operational NRT biospheric flux products one can mention SMAP L4C (Jones et al 2017)
L96 CAMS is associated with a wide variety of products related to the topic of this paper, better give more specific name like CAMS-REG. Also, the paper by Kuenen et al 2022 appears to document CAMS-REG-AP (air pollutant) inventory, while GHG portion called CAMS-REG-GHG was set aside.
L168-184 It is not clear, what data is used for spatial emission patterns.
L215 Final revised paper for GFAS is Di Guiseppe et al. 2018 (https://doi.org/10.5194/acp-18-5359-2018)
L231 Can cite here the (Chevallier et al 2019) method as ‘poor man’s inversion’ as done by Chevallier et al. (2010, 2019)
L249 “Emissions by” can be omitted.
L255 ‘exact monthly growth rate’ may draw doubts, as the rate is not that exact, can write ‘exactly follow monthly growth rate’ instead.
L280 Need to add spatial/vertical resolution at which IFS winds are used, and STILT domain geographical boundaries.
L471 Suggest clarifying the text: “Currently, CTE-HR only provides CO2 fluxes”. Better be more specific: eg biogenic/biospheric/net ….
L524 Here, fluxes ‘will be made available ', while on L518 same fluxes ‘are available’
References
Chevallier, F., et al.: CO2 surface fluxes at grid point scale estimated from a global 21 year reanalysis of atmospheric measurements, J. Geophys. Res., 115, D21307, doi:10.1029/2010JD013887, 2010
Chevallier, F., Remaud, M., O'Dell, C. W., Baker, D., Peylin, P., and Cozic, A.: Objective evaluation of surface- and satellite-driven carbon dioxide atmospheric inversions, Atmos. Chem. Phys., 19, 14233–14251, https://doi.org/10.5194/acp-19-14233-2019, 2019.
Henne, S., Brunner, D., Oney, B., Leuenberger, et al. Validation of the Swiss methane emission inventory by atmospheric observations and inverse modelling, Atmos. Chem. Phys., 16, 3683–3710, https://doi.org/10.5194/acp-16-3683-2016, 2016.
Jones, L., et al.: The SMAP Level 4 Carbon Product for Monitoring Ecosystem Land-Atmosphere COâ Exchange. IEEE Transactions on Geoscience and Remote Sensing. 55. 6517-6532. 10.1109/TGRS.2017.2729343. 2017.
Nalini, K., Lauvaux, T., Abdallah, C., Lian, J., Ciais, P., Utard, H., et al. (2022). High-resolution Lagrangian inverse modeling of CO2 emissions over the Paris region during the first 2020 lockdown period. Journal of Geophysical Research: Atmospheres, 127, e2021JD036032. https://doi.org/10.1029/2021JD036032
Zeng, J., et al. Global terrestrial carbon fluxes of 1999–2019 estimated by upscaling eddy covariance data with a random forest. Sci Data 7, 313 (2020). https://doi.org/10.1038/s41597-020-00653-5
-
RC2: 'Comment on essd-2022-175', Anonymous Referee #2, 30 Sep 2022
The study by van der Woude et al. (2022) described the CarbonTraker Europe High-Resolution (CT-HR) system that estimates carbon dioxide exchange over Europe at high-resolution (0.1 x0.2) in near real-time (about 2-months latency). The system includes a dynamic fossil fuel emission model, hourly net biosphere exchange (NEE), and an ocean flux. All these flux components are hourly. The high spatiotemporal resolution is realized by incorporating easily available statistics on economic activity, energy use, weather, and land cover map. The study evaluated the CT-HR in several aspects, including: 1) comparing the simulated CO2 to atmosphere CO2 observations at ICOS sites; 2) analyzing the 2018 Europe drought signal; 3) analyzing fossil fuel emission anomalies due to the COVID lockdown; and 4) comparing the hourly fluxes against observations from a flux tower at a city center. Overall, the paper is well written. The low-latency high-resolution data will facilitate the high-resolution atmospheric flux inversions, and the near-real time aspect can provide rapid information on carbon cycle changes over Europe. My main concern is the uncertainty quantification and the evaluation of the high spatiotemporal resolution aspect of the product. Please see my detailed comments below.
- For each of the flux components, it would be critical to have uncertainty estimation in order to use them in the atmospheric flux inversions and provide useful information on carbon cycle changes, which are the two applications envisioned by the authors. To some degree, the uncertainty is as important as the absolute flux estimation itself for atmospheric flux inversion application. So I would suggest adding uncertainty estimation in the product that propagate the uncertainty of the input data to the uncertainty of the end products.
- The evaluation of the high spatiotemporal feature of the product is not sufficient. The paper underwent extensive effort to evaluate the product, but the evaluation is mainly on large spatial and monthly time scale (except the tower comparison and partially the ICOS comparison), which are actually mainly from the input dataset. For example, the 2018 drought was evaluated on an aggregated region, the COVID signal was also evaluated on the continental scale by comparing to the published results. As the high spatiotemporal feature is the main advancement, I would recommend adding more evaluation at finer spatiotemporal scales. For example, the observations from flux towers can be used to evaluate the hourly NEE from SIB4.
- It is not clear how the diurnal cycle of SIB4 was calculated, and what is the temporal frequency of SIB4 output. Is the original SIB4 output monthly or daily?
- Figure 6 shows that the HR almost has the same mean RMSE as the PRI, but the text claims that the CT-HR performs better. It would be informative to include the exact RMSE from PRI and CT-HR when comparing to ICOS network in the text.
- Figure 7 shows that all three fluxes give very similar results during well-mixed hours, but it is hard to tell which one has better performance at the bottom two panels since the scale is so large. I would suggest adding panels to show the difference between the simulated run and the measurements. I would also suggest discussing RMSE of CO2 forced by these three fluxes during well-mixed hours and during night and early morning.
- It is strange that the “Flat” fluxes seem to have the same performance as the “full res” in December, which seem to indicate potential large errors in boundary layer height. I would suggest evaluating boundary layer height information used in STILT and isolate the impact of transport and fluxes on the flux errors shown in Figure 7.
- AC1: 'Comment on essd-2022-175', Auke Van Der Woude, 02 Dec 2022
Auke Marijn van der Woude et al.
Data sets
Near real-time, high-resolution CO2 fluxes over Europe Auke van der Woude, Remco de Kok, Ute Karstens, Wouter Peters https://doi.org/10.18160/20Z1-AYJ2
Auke Marijn van der Woude et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
615 | 288 | 18 | 921 | 12 | 9 |
- HTML: 615
- PDF: 288
- XML: 18
- Total: 921
- BibTeX: 12
- EndNote: 9
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1