High-resolution inversion of methane emissions over Europe using the Community Inversion Framework and FLEXPART

Mengistu, Anteneh Getachew; Tsuruta, Aki; Berchet, Antoine; Thompson, Rona; Tenkanen, Maria; Lindqvist, Hannakaisa; Markkanen, Tiina; Leppänen, Antti; Laitinen, Antti; Martinez, Adrien; Fortems-Cheiney, Audrey; Höglund-Isaksson, Lena; Aalto, Tuula

doi:10.5194/egusphere-2025-5877

Preprints

https://doi.org/10.5194/egusphere-2025-5877

Preprints

19 Jan 2026

| 19 Jan 2026

Status: this preprint is open for discussion and under review for Atmospheric Chemistry and Physics (ACP).

High-resolution inversion of methane emissions over Europe using the Community Inversion Framework and FLEXPART

Anteneh Getachew Mengistu, Aki Tsuruta, Antoine Berchet, Rona Thompson, Maria Tenkanen, Hannakaisa Lindqvist, Tiina Markkanen, Antti Leppänen, Antti Laitinen, Adrien Martinez, Audrey Fortems-Cheiney, Lena Höglund-Isaksson, and Tuula Aalto

Abstract. Constraining methane (CH₄) emissions at high spatial and temporal resolution is critical for accurate European greenhouse gas budgets and mitigation policy. We use the Community Inversion Framework to estimate monthly CH₄ fluxes across Europe (2017–2022) at 0.2° × 0.2°, coupling the FLEXPART and assimilating observations from 46 in situ stations, including ICOS and non-ICOS sites. Prior emissions combine GAINS and EDGARv8 anthropogenic inventories with GFED biomass burning, JSBACH-HIMMELI wetland fluxes, and climatological natural sources. The inversion markedly improves agreement with atmospheric observations (r² = 0.87, RMSE = 24.35 ppb, mean bias = –2.14 ppb), performing best at northern European stations. Posterior EU27+3 CH₄ totals 23.28 ± 0.36 Tg CH₄ yr⁻¹, 6.6 % above the prior. Anthropogenic emissions average 17.6 ± 0.3 Tg CH₄ yr⁻¹, exceeding GAINS by 11 %, EDGARv8 by 4 %, and UNFCCC NGHGI (2023) by 3 %, consistent with recent studies. Country-level differences are substantial: emissions are higher in BENELUX (+54 %), Germany (+37 %), and France (+10 %), and lower in the UK (–11 %), Romania (–25 %), Poland (–16 %), and Italy (–11 %) compared to UNFCCC NGHGI (2023). Sectoral changes primarily reflect agricultural increases in western and central Europe, with reductions in northern wetlands and southern geological sources. Sensitivity tests highlight the influence of horizontal correlation length and the value of dense observational networks for refining regional CH₄ budgets.

Received: 28 Nov 2025 – Discussion started: 19 Jan 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Status: open (until 02 Mar 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2025-5877', Anonymous Referee #1, 09 Feb 2026 reply
In “High-resolution inversion of methane emissions over Europe using the Community Inversion Framework and FLEXPART,” Mengistu and co-authors use the CIF system and Flexpart model to perform a 4D-Var data assimilation over Europe. The state vector includes CH4 concentrations and CH4 surface fluxes. Authors distinguish different sectors in the inversions and perform the inversions at 0.2x0.2 degrees. Results are compared to observations and compared to other inventories, including countries’ UNFCCC reporting. Overall I think this is a valuable study and has potential to be of strong interest to ACP readers. However, improvement is needed in the handling of uncertainty and description of the methods, particularly how prior error covariances are developed for the sectoral attribution.

General comments
Sectoral results heavily rely on sectoral correlations B, which contains correlation structure C. How is matrix C constructed? Are the same assumptions used for all sectors? There are many subjective choices for building this correlation matrix. It is not clear from the manuscript what these choices are.
How sensitive are the optimized fluxes to the observations? Can authors show the footprint map from the Flexpart simulations, are all areas of the map sensitive to the observations? I am particulary curious about Italy and the southern Europe region, where there are few observing stations but very large corrections.
Uncertainty on the results is not stated when results are given. The authors performed a sensitivity inversion, but only for 1 month due to computational cost. What are the implications for overall uncertainty on the year-to-year estimates? In addition to discussing the uncertainty in the 1-month estimates, can authors describing the implications for uncertainty levels on the overall results, and include that with the results? For example, on the bar chart in Figure 9 and time series in Figure 5. Or, can authors use posterior uncertainties described in lines 274-276 to present the results with their uncertainties?
In general, using 2 decimal places for CH4 concentrations in ppb implies that the model and measurements have precision down to the hundredth of a ppb. I doubt this is the case, and in my opinion using 0 or 1 decimal places would be much better. Similar for emissions, describing emissions down to 0.01 Tg precision implies a high degree of confidence in the results, but I am not convinced this level of precision is warranted. Including uncertainty ranges as mentioned above could help with this.

Specific comments
Line 16 – there is a more recent citation for the Global Methane Budget (Saunois et al. 2025 https://doi.org/10.5194/essd-17-1873-2025)

Line 112-113 – the sentence is confusing. Is the purpose of the sentence to state that analytical methods give a closed form expression for the posterior error covariance, while variational methods do not provide a posterior error covariance? If so, this should be stated. Does CIF-Flexpart give a posterior error covariance? Please clarify.

Line 143 – The methods are a bit unclear. Are two separate inversions being performed, one for total flux and another for specific sectors? Or are all sectors included in the state vector and all optimized with concentrations in a single inversion?

Line 161 – some more detail on the 3D OH fields are needed as these vary quite a bit (a citation to the study or model version is probably sufficient).

Line 183 – please state the significance of there being ICOS and non-ICOS observations – what is the difference between these networks? Are the measurements of different quality?

Lines 191-194 – Are the measurements at the sites in the same grid boxes comparable? Or, can authors provide some other justification for choosing arbitrarily? Are results sensitive to this choice?

Line 208 – again the ICOS definitions are unclear. Please define “ICOS levelling,” and the significance that data are ICOS labeled after a certain time.

Line 274-275 – How posterior error covariances are specific is unclear. Please give a brief description or equation.

Lines 276-277 – what are the expected impact of only doing the sensitivity tests in the summer? Methane’s seasonality in both emissions and background concentrations in the northern midlatitudes is significant, affecting background and wetland sources, please comment on the significance of this. Even better would be to do winter senstivity tests but this is likely computationally prohibitive.

Could authors please comment on the very large GEO decrease over Italy? The signal is a similar magnitude as wetland and anthropogenic emissions which is surprising.

Lines 404-410 – authors may be interested in and should consider citing two recent papers regarding seasonal cycles of energy sector emissions Varon et al. 2025 ES&T (https://doi.org/10.1021/acs.est.5c08745) and Hu et al. 2025 ES&T (https://doi.org/10.1021/acs.est.4c14090)

Lines 476 – Authors may also be interested in and should consider citing East et al. 2025 Nat Comm (https://doi.org/10.1038/s41467-025-67122-8) comparing inversion results to UNFCCC reports, and whether they see similar changes

What is the temporal resolution of the optimizations?

Lines 501-502 – could authors clarify what is meant by “fine-scale emissions patterns that are not accessible through conventional analytical inversion techniques”? I’m not sure this is a true statement.

Technical corrections
Abstract contains many undefined acronyms

Line 83 – parentheses around ICOS are missing

Figure 2 – I appreciate authors showing all of the data, but it is difficult to parse apart Fig 2 and make sense of it. Markers are quite small and cover one another, and the colors are difficult to distinguish and keep track of. Please consider if there is any way to make the plot more clear to aid readability.

Figure 9 – Y-axis is not labeled, what do the different bars with the same color represent?

Fig 7 – This figure is unclear. Where are the ranges coming from for the boxplots, from the posterior uncertainty? Or temporal variability? What is the yellow shaded area that is referred to in the description, could there be a typo?

Reply
Citation: https://doi.org/10.5194/egusphere-2025-5877-RC1

Viewed

Total article views: 352 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
188	147	17	352	13	20

HTML: 188
PDF: 147
XML: 17
Total: 352
BibTeX: 13
EndNote: 20

Views and downloads (calculated since 19 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	163	113	14	290
Feb 2026	25	34	3	62

Cumulative views and downloads (calculated since 19 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	163	113	14	290
Feb 2026	25	34	3	62

Viewed (geographical distribution)

Total article views: 318 (including HTML, PDF, and XML) Thereof 318 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 12 Feb 2026

Short summary

Our manuscript presents a six-year, high-resolution inversion of European methane emissions using the Community Inversion Framework and FLEXPART. Leveraging an expanded in situ network, we reconcile inventories with observations, reveal regional biases, refine European methane budgets, and highlight the dominance of agricultural emissions. Results provide policy-relevant insights for mitigation and inventory verification.


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