the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Jul 2026
Constraining 2020 methane emissions in the Eastern Mediterranean and Middle East using TROPOMI satellite data
Abstract. Methane (CH4) is the second most abundant anthropogenic greenhouse gas after CO2. Emissions from the Eastern Mediterranean and Middle East (EMME) are important, as the region hosts a large share of global oil and gas production and thus offers significant mitigation potential. We use TROPOMI satellite observations in combination with the Lagrangian transport model FLEXPART and the inversion framework FLEXINVERT+ to estimate CH4 emissions in the EMME region for 2020. We perform sensitivity experiments to address methodological challenges and better quantify emission uncertainties.
Our inversions show robust downward corrections of emissions around the Persian Gulf relative to the EDGAR v8.0 inventory, consistent with the recently released EDGAR_2025_GHG dataset, yielding total emissions of 22.4 (±1.9) Tg yr−1 in the EMME region. At the national scale, we derive emissions of 4.6 (±0.4), 3.6 (±1.1), 3.5 (±0.8), 2.8 (±0.4), and 1.9 (±0.4) Tg yr−1 CH4 for Iran, Saudi Arabia, Turkey, Iraq, and Egypt, respectively. While many countries in the region appear to substantially underestimate their emissions, Greece, subject to mandatory reporting requirements, reports emissions comparable to our inversion estimates. Our results suggest a decrease in emissions in the EMME region during 2020, coinciding with reduced oil demand during the COVID-19 epidemic.
Our satellite-based inversions provide robust constraints in regions with strong emission signals and good observational coverage (e.g. Iran, Iraq), while other regions show greater sensitivity to uncertainty settings and prior estimates. Our sensitivity analyses highlight the importance of baseline optimization and its associated uncertainty.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Chemistry and Physics.
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.- Preprint
(7896 KB) - Metadata XML
-
Supplement
(12253 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3792', Anonymous Referee #1, 18 Aug 2026
-
RC2: 'Comment on egusphere-2026-3792', Anonymous Referee #2, 30 Sep 2026
Overall, this study represents an important contribution to the field of methane emission estimation using satellite observations within an atmospheric inversion framework. The extensive set of sensitivity experiments, exploring the impact of several inversion parameters and assumptions, is particularly valuable and provides a broad assessment of the robustness of the results. The manuscript is generally well structured and progressively introduces the main methodological elements needed to understand the inversion setup and the resulting emission estimates.
However, I believe that the clarity and traceability of the analysis could be further improved by being more explicit about some methodological choices and by more clearly linking specific findings to the corresponding figures and sensitivity experiments. I provide detailed comments on these points below. In addition, I consider it important that the final comparison with other studies places greater emphasis on estimates corresponding to 2020, whenever available. Given the exceptional conditions of that year, particularly the substantial changes in economic activity and fossil-fuel production associated with the COVID-19 pandemic, comparisons with estimates from 2019 or earlier years can provide useful context but are not directly comparable and should not be used to draw strong conclusions regarding agreement or disagreement with the present inversion results. The discussion and conclusions should therefore reflect this limitation more explicitly.
Lines 65–67: I suggest clarifying the statement that “the computational cost remains independent of the number of observations.” As currently written, this seems too general. For adjoint-based inversions, the main computational advantage is that the cost of computing the gradient is largely independent of the number of elements in the control/state vector, rather than of the number of observations. The treatment of observations (e.g. application of the observation operator, calculation of model–observation mismatches, averaging kernels, and data handling) may still scale with the number of observations, which can become relevant for large satellite datasets.
Line 110: The value of 30 km for the correlation length depends on the characteristics of satellite data retrieval, surface conditions, sources of instrumental error, etc. In this sense, could you briefly explain why this value, taken from a study on NO₂, is applicable to your study?
In Figures 1a and 1b, the addition of political boundaries (country borders) would help the reader to better locate the main features shown in the maps.
Line 158: The choice of spatial and temporal correlations needs to be justified and cited, especially given that each sector has a different spatiotemporal structure.
Line 213: The choice of 20 iterations for the sensitivity experiments and 30 iterations for the final inversions appears to be based mainly on visual inspection of the cost-function evolution. Could the authors provide a quantitative convergence criterion, for example based on the gradient norm.
Lines 250–260: The design of the sensitivity experiments is not entirely clear from the current description. Were these parameters varied one at a time relative to a reference configuration, while all other settings were kept fixed, or were combinations of the different parameter values explored? A full factorial combination would result in a very large number of experiments, so I assume a one-at-a-time sensitivity approach was used. Please clarify the experimental design. If parameters were varied individually, it would also be useful to briefly acknowledge that interactions between uncertain parameters are not explored.
Line 275: The sensitivity analysis explores a wide range of uncertainties assigned to the EGG4 baseline, but the inversion still relies on the same baseline product in all experiments. Given the apparent early-2020 bias in EGG4, it would be useful to discuss whether the posterior emissions may also depend on the choice of baseline product itself. If an alternative CH₄ background/initial field is available, a sensitivity experiment using an independent product would provide a complementary test of the robustness of the inferred emissions.
279: Replace (b) the relative error reduction by (b) the relative uncertainty reduction
Figure 5: The spatial distribution of posterior emissions is somewhat difficult to interpret because most grid cells appear close to the lower end of the color scale, while only a few emission hotspots are clearly visible. The authors may consider adjusting the color-scale normalization or using units such as mg m⁻² d⁻¹ to improve readability. It would also be helpful to include a figure showing the relative percentage change in the increase.
Line 302-308: Given the large number of sensitivity tests performed, I understand that organizing the corresponding results and discussion is challenging. However, Section 3.3 would be easier to follow if the text more explicitly indicated which figure or supplementary figure supports each of the main findings discussed. In particular, between lines 302 and 308, several conclusions are presented without reference to a specific figure in the Appendix or Supplement. Adding these references would help the reader follow the analysis and identify the evidence supporting each statement.
Line 308 and Figure B1: The sensitivity tests using different a priori emission uncertainties are very informative. As shown in Fig. B1, the inversion increments are clearly sensitive to the assumed prior uncertainty, both in magnitude and spatial distribution. Could the authors clarify over which range of a priori uncertainties the posterior emissions can be considered robust? In other words, is there a range of values for which the main spatial patterns and regional emission estimates remain relatively stable? This would help justify the uncertainty chosen for the reference inversion.
Figure 6: suggests that the spread among posterior estimates is substantially reduced compared with the spread among the different a priori inventories for several regions. However, the term “remarkably robust” appears to be used qualitatively. Could the authors quantify the robustness of the posterior emissions, for example by reporting the spread, standard deviation, or coefficient of variation across the inversions using different prior inventories? This would also help distinguish regions where the posterior is largely independent of the prior from cases such as Saudi Arabia, where substantial prior dependence remains.
In addition, Saudi Arabia appears to be a notable exception to the overall consistency across prior inventories. Could the authors discuss why the posterior remains more dependent on the prior in this region.Line 313-314: Please indicate which figure supports or illustrates this statement
Line 315-316: Same, please indicate which figure supports or illustrates this statement
Line 596: The sensitivity of the inversion to the assumed spatial error-correlation length is clearly demonstrated. However, it is less clear how the tested range (50–500 km), and particularly the value adopted for the reference inversion, were selected. Could the authors provide a physical or statistical justification for these correlation scales? For example, were they informed by the spatial structure of the prior inventory, comparisons among inventories, variogram analyses, or previous inversion studies? Just as an example, I'm sharing a paper that proposes different values for correlation length depending on the emission sector and also mentions other studies that use different reference values.
Montenegro, N., Berchet, A., Martinez, A., Thanwerdas, J., Bousquet, P., Pison, I., and Saunois, M.: Capability of current observing systems to monitor CH4 emissions from the regional to the global scales, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-5923, 2025.
Line 326: Could the authors please clarify the total number of inversions underlying the final emission estimate? If each of the eight inversion setups (4 prior inventories × 2 error distributions) includes a Monte Carlo ensemble of 100 members, does this correspond to 800 Monte Carlo inversions in total.
Figure 7: Given that the inversion specifically targets 2020, I strongly encourage the authors to include comparisons with independent emission estimates for the same year wherever available. This is particularly important because 2020 was an anomalous year, with substantial changes in economic activity and oil and gas production associated with the COVID-19 pandemic. Comparisons with 2019 or earlier estimates are useful for context, but they cannot provide a direct like-for-like evaluation of the 2020 posterior emissions.
Lines 414–416: The attribution of the late-summer CH₄ maximum over Iran to enhanced wetland and agricultural emissions appears somewhat stronger than the supporting evidence. Mousavi et al. (2024) identify a pronounced seasonal cycle in XCH₄ over Iran, but the interpretation in terms of wetland and agricultural emissions relies largely on broader/global literature (Saunois, 2024). Given that wetland CH₄ seasonality can depend strongly on local hydrological and biogeochemical processes, including potentially substantial lags between precipitation, inundation and emissions, I suggest either providing more region-specific evidence for this attribution or softening the statement to indicate that wetlands and agriculture are possible contributors rather than established drivers.
Recent work has also shown that wetland CH₄ emissions can exhibit substantial lags relative to precipitation and inundation, highlighting that the timing of wetland emissions may be highly region dependent. Reference: Ruoqi Liu et al. ,Unveiling cascading lag effects of wetland methane emissions: Evidence from Lake Chad in Africa.Sci. Adv.12,eadx9866(2026).DOI:10.1126/sciadv.adx9866
Line 458: In line with the suggested revisions regarding the comparison with studies from other years, please make the corresponding changes to the conclusions so that they are consistent with the revised discussion.Citation: https://doi.org/10.5194/egusphere-2026-3792-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 213 | 118 | 42 | 373 | 75 | 36 | 38 |
- HTML: 213
- PDF: 118
- XML: 42
- Total: 373
- Supplement: 75
- BibTeX: 36
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors conducted the methane emission inversion with Lagrangian particle dispersion forward model and TROPOMI methane observations over the Eastern Mediterranean and Middle East. They performed comprehensive sensitivity tests for the parameters used in the inversion framework. The scope is important and the method is sound. However, the work can be strengthened upon further clarifications/investigations on the
General comments
Specific comments
Technical comments