the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Feb 2026
Evaluating national, state, and urban Indian methane emissions using satellites
Abstract. Understanding the spatial distribution and magnitude of methane emissions is critical for developing effective mitigation strategies, particularly in rapidly growing economies like India, the world’s most populous country and a top global methane emitter with diverse emission sources. We quantify India’s 2021 methane emissions at up to 0.25°×0.3125° resolution using TROPOspheric Monitoring Instrument (TROPOMI) observations in a Bayesian inversion with the Integrated Methane Inversion framework (IMI). Prior emissions come from state-of-the-art global gridded bottom-up inventories and incorporate GHGSat-based estimates for eighteen landfills. The high-resolution inversion and incorporation of GHGSat data enable us to evaluate and interpret results at multiple policy-relevant scales. The national posterior emission estimate is 34.4 (32.0 – 40.4) Tg/yr, of which 31.5 (29.6 – 36.7) Tg yr-1 is anthropogenic, consistent with bottom-up prior emissions but 68 % higher than India’s UNFCCC inventory. National landfill and oil & gas emissions are 30 % higher, while coal emissions are 57 % lower than prior estimates. State-level analysis highlights seven states with higher emissions, with notably higher oil and gas emissions in Assam and Gujarat, and lower coal mining emissions in Rajasthan and Odisha. Urban-scale posterior estimates for fourteen cities reveal significant differences from the prior in ten cities. Posterior wastewater emissions are higher in nine cities, with the largest increases in Kolkata and Delhi, consistent 60–90 % of their populations lacking access to wastewater treatment facilities. GHGSat observations reveal landfills contribute 10–38 % of emissions in eleven cities, emphasizing the critical role of solid waste management. These results illustrate how satellite-based analyses can inform methane mitigation.
- Preprint
(1800 KB) - Metadata XML
-
Supplement
(1778 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-6528', Anonymous Referee #1, 05 Mar 2026
-
RC2: 'Comment on egusphere-2025-6528', Anonymous Referee #2, 22 Jun 2026
The authors apply IMI/GEOS-Chem (0.25°×0.3125°) with the blended TROPOMI+GOSAT product (Balasus et al., 2023) to estimate 2021 Indian methane emissions, supplemented with facility-level GHGSat landfill data, and report a national total of 34.4 (32.0 to 40.4) Tg yr–1 with a 30-member ensemble. The study is well written and makes real contributions, namely the high state-vector resolution relative to earlier national studies, the use of GHGSat facility data to improve landfill and wastewater separation, and the multi-scale policy framing. The national total is consistent with several independent estimates, which is reassuring for the aggregate. My concerns concentrate on the observation-error budget, the treatment of aerosol-related retrieval biases, the prior-dependence of the sectoral attribution, and the strength of validation. Also, several critical flaws exist in data preprocessing, observational error configuration, model validation and aerosol interference correction, which undermine the rationality and reliability of the research methodology. Because these concerns bear most directly on the sectoral conclusions, especially the -57% coal correction, rather than on the national total, and because they are addressable, I recommend major revision.
Major comments
1. The observation-error budget is likely optimistic and is not justified for Indian conditions
The super-observation error variance is reported as (11.23 ppb)2, built from = 15 ppb, = 0.55 and = 4.5 ppb, all transferred from a China inversion (Chen et al., 2023). There are two issues. First, these parameters were calibrated for a different domain, so the manuscript should demonstrate, rather than assume, their validity over India, where surface and aerosol conditions differ substantially. Second, an under-specified observation error inflates the effective weight of the observations in the cost function and therefore inflates the magnitude of the emission corrections, which are precisely the quantities (coal at -57% and oil and gas at +33%) that the paper reports as its headline findings. Given that residual surface-reflectance and aerosol artefacts in the TROPOMI XCH4 product are documented to introduce errors well beyond the random retrieval noise (Lorente et al., 2021, 2023), and that the reliability of TROPOMI-based emission estimates has been shown to depend strongly on how the observational error budget is constructed (Zheng et al., 2026), I recommend an ensemble member with a markedly larger and, ideally, spatially and seasonally varying observation error, with the sectoral posteriors re-reported.
2. Aerosol contamination is addressed only by excluding “hazy days”; partial / sub-pixel aerosol coverage in the non-monsoon season is the larger, unaddressed problem
The hazy-day exclusion (63 days, heavily clustered in late October and November) does not address spatially heterogeneous, partial aerosol and cloud contamination on ordinary days across the long non-monsoon period. This matters for two reasons. First, aerosol-induced biases in the 2.3 µm window can be large and sign-dependent on aerosol type and surface albedo, and they are now shown to be specifically enhanced in the presence of aerosols (Somkuti et al., 2025; Huang et al., 2020; Lorente et al., 2023); the magnitude and sign of this effect depend on absorbing-aerosol optical properties that vary strongly across polluted Asian regions and are themselves uncertain (Wang et al., 2021; Guan et al., 2026). Second, the most-affected regions, namely the Indo-Gangetic Plain and the eastern coal belt (Odisha and Jharkhand), coincide with the most consequential corrections. The blended product (Balasus et al., 2023) corrects some biases via machine learning but does not eliminate aerosol-driven errors, and the residual structure is not characterized here. I recommend a sub-scene or per-pixel AOD screening test rather than day-level exclusion alone, an analysis of how residual aerosol bias propagates into sectoral attribution, and explicit acknowledgement that this is a leading uncertainty for the northern-India and coal results.
3. The −57% coal correction is presented more confidently than the evidence supports
This is the most striking sectoral result, and it currently rests on assumptions that should be reconciled. First, regarding prior dependence, the sectoral attribution (Sect. 2.8) assumes that grid-cell-level relative sectoral contributions in the prior are accurate. With a total DOFS of 296 over roughly 1785 elements, the split carries substantial prior information, and coal, although spatially concentrated, is co-located with the heaviest aerosol loading discussed in point 2. The coal prior itself derives from national UNFCCC-based reporting via GFEI (Scarpelli et al., 2020), so the inversion is correcting a specific bottom-up construction rather than an independent baseline. Second, regarding the direction-of-evidence tension, independent work has repeatedly found Chinese and global coal-mine methane to be under-reported or revised in complex directions (Sadavarte et al., 2021; Gao et al., 2021), so a large downward correction warrants engagement with that literature rather than a one-sided surface-mining argument. Third, the proposed mechanism, namely a shift toward less methane-intensive surface mining not yet reflected in activity data (Wright et al., 2024), is plausible but is not a demonstration, and the competing explanation, in which aerosol-contaminated retrievals over the coal belt are read as an emission reduction, is not ruled out. I recommend independent corroboration, whether facility-level, isotopic, or in-situ, or else substantial hedging, before the coal claim is stated as strongly as it is in the abstract.
4. Validation is weak and only partially independent
The GOSAT evaluation (Fig. 2, S7) is not independent, because GOSAT is the reference used to construct the blended product that is assimilated (Balasus et al., 2023), and this should be stated plainly where the GOSAT comparison is presented. Even so, it retains large structure, with monthly differences of roughly −30 to −40 ppb in summer that far exceed the approximately 11 ppb assumed observation error, which points to seasonally varying model and retrieval error not represented in the annual, spatially uniform error treatment. The single surface site at Nainital shows only marginal improvement, from an MAE of 73.3 to 70.5 ppb, with individual residuals exceeding 200 ppb. While representation error explains much of this, given that a 0.25° column-oriented inversion is compared against weekly flasks at a mountain site (Nomura et al., 2021), that is precisely why one such site cannot constitute validation. I recommend genuinely independent evaluation, such as additional in-situ records or a withheld-data cross-validation.
6. Annual-mean optimization and imposed seasonality limit diagnosis of the above
Optimizing annual 2021 emissions with EDGAR-imposed seasonality, combined with the substantial boundary-condition adjustments shown in Fig. S5, leaves little freedom to absorb seasonally structured mismatch as model and retrieval error rather than as emissions, which raises the risk of aliasing. The use of nested GEOS-Chem with annual-mean optimization follows established practice (Varon et al., 2022; Chen et al., 2022), but the interaction between the annual-mean framing and the seasonally varying aerosol bias of point 2 should be discussed. Please also clarify the temporal sampling of the super-observations, since the text emphasizes spatial aggregation while leaving unstated both the per-cell temporal sampling and whether super-observations are formed per overpass.
7. Homogenization of emission signals
The original daily and grid-level satellite observations were spatially aggregated for the assimilation and inversion analysis. Although this approach reduces random observational uncertainties, it leads to homogenization of emission signals: strong emission hotspots tend to be underestimated while weak emission areas are overestimated. Even if the total emission of each aggregated grid remains unchanged, the spatial characteristics of local emissions are distorted, which adversely affects the high-resolution emission analysis at state and city levels. The authors have not analyzed or discussed this issue.
8. The use of GHGSat estimates as annual prior emissions requires justification
The authors use instantaneous GHGSat landfill emission estimates from 2021–2022 to modify the annual prior emissions for the 2021 inversion, but it is unclear how a limited number of satellite observations can represent annual mean emissions. The authors should report the number of observations, detections and non-detections, individual emission estimates, and associated uncertainties for each landfill, and justify the use of 2022 observations to represent 2021 emissions. In addition, setting EDGAR landfill emissions within 25 surrounding grid cells to zero and concentrating the GHGSat estimate into a single grid cell may remove other unobserved small or diffuse waste sources and artificially alter the prior spatial distribution. Sensitivity tests using alternative replacement strategies and an inversion without GHGSat-based prior adjustments are needed. Because GHGSat information is already incorporated into the prior, agreement between the posterior and GHGSat estimates should not be presented as independent validation. The conclusions regarding the improved separation of landfill and wastewater emissions should therefore be stated more cautiously.
Specific comments
EQ 6 in Sect. 2.7. Report the spatial distribution of the super-observation error and the number of super-observations per grid cell, since the domain total of 888,987 implies tens per cell per year, which would dispel any impression of heavy temporal aggregation.
Fig. S4. Individual grid points worsening after inversion should not be over-interpreted, because with = 0.1 the inversion deliberately does not fully fit the observations (regularized cost function; Brasseur and Jacob, 2017), so local degradation alongside an improved aggregate correlation, which rises from 0.89 to 0.93, is expected. I raise this so that the authors can pre-empt it.
Sect. 2.9. Report the sectoral, and not only the total, sensitivity to hazy-day exclusion, since the text gives only the +0.7 Tg yr–1 total change.
Sect. 3.4 (Talcher coal field). The factor-6.5 local reduction rests on a few grid cells in a high-aerosol region, so please show the averaging-kernel sensitivity and aerosol conditions for these cells.
Figure S7 presents the monthly mean biases between inversion simulations and GOSAT observations, with the maximum monthly bias approaching 10 ppb. The actual biases of daily and individual grid observations are inevitably larger than this monthly average value, exceeding the preset observational error range in the manuscript.
Figure S8, large fluctuations exist between in-situ measurements and model simulations, with partial deviations exceeding 200 ppb. There needs in-depth analysis on the causes of such substantial discrepancies.
Technical
Table 1 (Supplement): the seeps sensitivity entry, given as 0.1 with a range of 0.3 to 0.6, appears inconsistent, so please check it.
Clarify in Sect. 3.1 the partition of the bias and RMSE improvement between boundary-condition optimization and interior emission adjustment.
References
Balasus, N., Jacob, D. J., Lorente, A., et al. (2023). A blended TROPOMI+GOSAT satellite data product for atmospheric methane using machine learning to correct retrieval biases. Atmos. Meas. Tech., 16, 3787–3807. https://doi.org/10.5194/amt-16-3787-2023
Brasseur, G. P., and Jacob, D. J. (2017). Modeling of Atmospheric Chemistry. Cambridge University Press. https://doi.org/10.1017/9781316544754
Chen, Z., Jacob, D. J., Nesser, H., et al. (2022). Methane emissions from China: a high-resolution inversion of TROPOMI satellite observations. Atmos. Chem. Phys., 22, 10809–10826. https://doi.org/10.5194/acp-22-10809-2022
Chen, Z., Jacob, D. J., Gautam, R., et al. (2023). Satellite quantification of methane emissions and oil–gas methane intensities from individual countries in the Middle East and North Africa: implications for climate action. Atmos. Chem. Phys., 23, 5945–5967. https://doi.org/10.5194/acp-23-5945-2023
Gao, J., Guan, C., Zhang, B., and Li, K. (2021). Decreasing methane emissions from China’s coal mining with rebounded coal production. Environ. Res. Lett., 16, 124037. https://doi.org/10.1088/1748-9326/ac38d8
Guan, L. Y., Cohen, J. B., Wang, S., et al. (2026). Improving aerosol absorption estimates via size-resolved constraints based on AERONET and in situ measurements. Geophys. Res. Lett., 53, e2025GL117418. https://doi.org/10.1029/2025GL117418
Huang, Y., Natraj, V., Zeng, Z.-C., Kopparla, P., and Yung, Y. L. (2020). Quantifying the impact of aerosol scattering on the retrieval of methane from airborne remote sensing measurements. Atmos. Meas. Tech., 13, 6755–6769. https://doi.org/10.5194/amt-13-6755-2020
Lorente, A., Borsdorff, T., Butz, A., et al. (2021). Methane retrieved from TROPOMI: improvement of the data product and validation of the first 2 years of measurements. Atmos. Meas. Tech., 14, 665–684. https://doi.org/10.5194/amt-14-665-2021
Lorente, A., Borsdorff, T., Martinez-Velarte, M. C., and Landgraf, J. (2023). Accounting for surface reflectance spectral features in TROPOMI methane retrievals. Atmos. Meas. Tech., 16, 1597–1608. https://doi.org/10.5194/amt-16-1597-2023
Nomura, S., Naja, M., Ahmed, M. K., Mukai, H., Terao, Y., Machida, T., Sasakawa, M., and Patra, P. K. (2021). Measurement report: Regional characteristics of seasonal and long-term variations in greenhouse gases at Nainital, India, and Comilla, Bangladesh. Atmos. Chem. Phys., 21, 16427–16452. https://doi.org/10.5194/acp-21-16427-2021
Sadavarte, P., Pandey, S., Maasakkers, J. D., Lorente, A., Borsdorff, T., Denier van der Gon, H., Houweling, S., and Aben, I. (2021). Methane Emissions from Superemitting Coal Mines in Australia Quantified Using TROPOMI Satellite Observations. Environ. Sci. Technol., 55, 16573–16580. https://doi.org/10.1021/acs.est.1c03976
Scarpelli, T. R., Jacob, D. J., Maasakkers, J. D., Sulprizio, M. P., Sheng, J.-X., Rose, K., Romeo, L., Worden, J. R., and Janssens-Maenhout, G. (2020). A global gridded (0.1°×0.1°) inventory of methane emissions from oil, gas, and coal exploitation based on national reports to the United Nations Framework Convention on Climate Change. Earth Syst. Sci. Data, 12, 563–575. https://doi.org/10.5194/essd-12-563-2020
Somkuti, P., McGarragh, G. M., O’Dell, C., Di Noia, A., Vogel, L., Crowell, S., Ott, L. E., and Bösch, H. (2025). Surface reflectance biases in XCH4 retrievals from the 2.3 µm band are enhanced in the presence of aerosols. Atmos. Meas. Tech. Discuss. [preprint], https://doi.org/10.5194/amt-2024-145
Varon, D. J., Jacob, D. J., Sulprizio, M., Estrada, L. A., Downs, W. B., Shen, L., Hancock, S. E., Nesser, H., Qu, Z., Penn, E., Chen, Z., Lu, X., Lorente, A., Tewari, A., and Randles, C. A. (2022). Integrated Methane Inversion (IMI 1.0): a user-friendly, cloud-based facility for inferring high-resolution methane emissions from TROPOMI satellite observations. Geosci. Model Dev., 15, 5787–5805. https://doi.org/10.5194/gmd-15-5787-2022
Wang, S., Wang, X., Cohen, J. B., and Qin, K. (2021). Inferring polluted Asian absorbing aerosol properties using decadal scale AERONET measurements and a MIE model. Geophys. Res. Lett., 48, e2021GL094300. https://doi.org/10.1029/2021GL094300
Wright, C., Patel, B., Modadugu, R., Setawan, D., and Yeman, C. (2024). Coal mine methane’s critical moment in India. Ember [report]. https://ember-energy.org/latest-insights/coal-mine-methanes-critical-moment-in-india
Zheng, B., Cohen, J. B., Lu, L., et al. (2026). How can we trust TROPOMI based methane emissions estimation: calculating emissions over unidentified source regions. Atmos. Chem. Phys., 26, 1931–1946. https://doi.org/10.5194/acp-26-1931-2026
Citation: https://doi.org/10.5194/egusphere-2025-6528-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,468 | 1,070 | 94 | 2,632 | 221 | 74 | 150 |
- HTML: 1,468
- PDF: 1,070
- XML: 94
- Total: 2,632
- Supplement: 221
- BibTeX: 74
- EndNote: 150
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This paper provides a detailed study using an inverse model to update the methane emission inventory for India based on satellite data from GOSAT, TROPOMI and GHGSat. The study seems very thorough, the paper is well written and the results are clearly presented. Furthermore, they are relevant to policy making and provide an important advance for developing methane reduction strategies.
My one quibble is that GHGSat data are mentioned, but we never really see a map of the retrievals. It would be nice to see some of these above the cities with increased landfill emissions.
I am happy to recommend publication.