the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Sep 2026
Monitoring Indian CO2 Emissions: An Integrated Satellite-Based Dual-Species Approach Using TROPOMI and CO2M Retrievals.
Abstract. Accurate quantification of CO2 emissions from large point sources is essential for tracking progress in emission-reduction strategies under the Paris Agreement. However, quantifying these emissions remains challenging across the Indian subcontinent due to dense clustering of co-located sources, spatially heterogeneous emissions, and a sparse observational network coupled with insufficient modeling capabilities. This study presents an NO2-guided CO2 plume detection and Cross-Sectional Flux (CSF) emission quantification methodology using high-resolution TROPOMI NO2 and synthetic CO2M XCO2 retrievals. The plume detection and quantification system is evaluated through an Observing System Simulation Experiment conducted across 14 coal-fired thermal power plants (TPPs) in India. CO2M XCO2 retrievals are generated using the Weather Research and Forecasting model integrated with the Greenhouse Gas module by sampling through simulated CO2M satellite tracks. Several scenarios are examined to assess the potential of the quantification system and the impact of background XCO2 estimation error on emission estimates. For the highest emission TPP in this study, Vindhyachal, our CSF approach estimates emissions with remarkable accuracy, with an uncertainty of 9 % in a background-error-free scenario. Since erroneous background XCO2 estimations can have a considerable impact on the above emission estimates, two different methods are implemented to better decouple the anthropogenic contribution from the total measured quantity amid CO2M-level pixel noise, and their effectiveness is assessed with and without the noise-filter algorithm. The assessment of multiple power plants shows that sources exceeding 20 Mt of annual CO2 are reliably quantified across multiple seasons, establishing the operational lower quantification bound of our methodology at ~ 20 Mt CO2 yr-1. By combining TROPOMI observations with future CO2M retrievals, this study evaluates the potential of these data to detect and quantify CO2 emissions from Indian thermal power plants under real observational conditions. Overall, the study demonstrates the operational readiness of the quantification methodology for upcoming satellite missions, as well as for integrated Measurement, Reporting, and Verification (MRV) systems within the Copernicus CO2 Monitoring and Verification Support (CO2MVS) initiative.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.
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
(6034 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Review on egusphere-2026-3926', Anonymous Referee #1, 02 Oct 2026
- AC1: 'Reply on RC1', Jithin Sukumaran Kumar, 04 Nov 2026
-
RC2: 'Comment on egusphere-2026-3926', Anonymous Referee #2, 06 Oct 2026
This study presents a satellite-based framework that combines TROPOMI NO2 observations with simulated CO2M XCO2 retrievals to identify power-plant plumes and quantify CO2 emissions over India. The topic is timely and relevant, particularly for future satellite-based monitoring of large point sources. The manuscript also includes several useful methodological developments for plume identification and emission estimation.
However, several aspects of the methodology and its evaluation need clarification and more thorough testing before the main conclusions can be fully supported. In particular, the added value of using NO2 to guide CO2 plume detection should be demonstrated more explicitly, and the sensitivity of the plume-boundary definition and quality-screening procedure deserves further analysis. I therefore recommend major revision and encourage the authors to address the following points.
Major comments:
- Since one of the main ideas of this study is to use NO2 as a guide for identifying the XCO2 plume, I think it would be useful to show more clearly how much this actually improves the result. A simple comparison between retrievals with and without NO2 guidance would already be very informative. For example, does NO2 mainly help with plume detection, plume-boundary definition, or the final emission estimate? It would also be useful to show the consistency between the NO2 and CO2 plumes on a case-by-case basis. Since the two species have very different atmospheric lifetimes, their plume shapes and downwind extents may not always match well.
- The definition of the final plume boundary in the BEC procedure is still not fully clear to me. The manuscript says that the outermost enhancement contour that remains spatially connected to the SCI core is taken as the final boundary. However, this boundary may depend quite strongly on the selected enhancement levels, contour spacing, and the way spatial connectivity is defined. Please explain more clearly how these choices are made, and what happens if the expanding contour merges with another nearby plume or background enhancement. I would also suggest adding a sensitivity test to show how much the final plume area and retrieved emission rate change when these settings are varied.
- For overpasses classified as “excellent”, “good”, or “poor”, is this ranking actually used when calculating the final emission estimate? For example, are better-quality overpasses given larger weights, or are all accepted cases treated equally and then summarized using the mean and one standard deviation? If all accepted cases are treated in the same way, then it is not entirely clear what the practical role of this classification is. It would be helpful to explain this more clearly and, if possible, show whether the different quality classes have different retrieval errors or uncertainties.
- In Section 3.4, the manuscript mainly shows results for TPPs with quality_flag = 0. I think it would also be useful to briefly look at the cases that do not pass this quality check. What are the main reasons for failure? Are they mostly related to weak emissions, clouds, unfavorable wind conditions, large background variability, overlapping plumes, or insufficient satellite coverage? This could help readers better understand the practical limits of the current method and under what conditions it is likely to work well.
Specific comments:
- Line 30: “uncertainties in transport modelling over India.” Is there any specific reason why transport modelling is particularly challenging over India? For example, does this refer to complex topography, monsoon circulation, boundary-layer dynamics, or strong spatial variability in meteorological conditions?
- Line 35: “anthropogenic enhancements” is a little vague. I suggest using something more specific, such as “anthropogenic emission-induced concentration enhancements” or simply “XCO2 enhancements caused by anthropogenic emissions”. It would also be helpful to give a rough quantitative value here, for example, the typical plume enhancement in ppm compared with the background XCO2 level.
- Line 45: Please give the full name of TROPOMI when it first appears.
- Line 45: Please define XCO2 when it first appears, for example, as the column-averaged dry-air mole fraction of CO2.
- Line 85: What does “IST” stand for? Please define it when first used.
- Line 90: Please give an approximate lifetime of NO2. Since NO2 is used here as a guide for CO2 plume identification, it would also be useful to briefly mention that its lifetime varies with photochemical and meteorological conditions.
- Figure 3(a): There seems to be an extra “1” in panel (a).
Citation: https://doi.org/10.5194/egusphere-2026-3926-RC2 - AC2: 'Reply on RC2', Jithin Sukumaran Kumar, 04 Nov 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 339 | 179 | 53 | 571 | 55 | 55 |
- HTML: 339
- PDF: 179
- XML: 53
- Total: 571
- BibTeX: 55
- EndNote: 55
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
“Monitoring Indian CO2 Emissions: An Integrated Satellite-Based Dual-Species Approach Using TROPOMI and CO2M Retrievals” by Sukumaran et al. is a potentially useful manuscript that explores the possibility of monitoring CO2 emissions from space based on real TROPOMI NO2 observations and simulated XCO2 observations corresponding to the future CO2M satellite constellation.
The manuscript makes a valuable qualitative conclusion about approaches to background definition, but the quantitative conclusions of the paper are questionable related to 2 key factors that the authors do acknowledge as caveats: namely the neglect of the contribution of transport errors and the coarse resolution of the model in comparison to the satellite data, mentioned again later in this review.
Specific Points
Abstract: The reported “lower limit of quantification” of 20 MtCO2/year is quite high and seems like a pessimistic/conservative lower limit for annual emission quantification from CO2M. Bovensmann et al. (2010) looked at the capability of quantifying emissions from a mid-sized power plant considered to be ~13 MtCO2/year from CarbonSat, which eventually evolved into the 3-satellite CO2M constellation (with added aerosol and NO2 observations further enhancing capabilities). Nassar et al. (2025) have demonstrated quantification of sources on the order of ~5 MtCO2/year from a single overpass in the U.S. without even bringing in supporting NO2 information, but they only conducted and instantaneous emission estimate rather than constrain annual emissions. Figure 12 of Bovensmann et al. (2010) shows the power plant emission statistics for countries including India (reasonably current at the time of that publication). With a cut-off of 20 MtCO2/yr, this would only cover about 20% of India’s power plant emissions. With a cut-off of 5 MtCO2/yr, about 75% of India’s power plant CO2 emissions would be included. Thus, the lower limit of quantification is quite an important point, thus the authors should refrain from reporting a number if there are major caveats related to it.
Line 35: Introduction should mention Nassar et al. (2017) following Bovensmann et al. (2010), since Nassar et al. (2017) was the first publication to quantify power plant CO2 emissions and in fact did so for Indian power plants, unlike Bovensmann et al. (2010).
Line 43: “Signal-to-noise ratio” may not be the best term to use here. The high contrast with the ambient NO2 background is what the authors are talking about, which is analogous to SNR, but when using the term SNR in this context, the authors have not specified the specific parameters that they consider signal or noise.
Line 49: It is worth expanding to “cross sectional flux and gaussian plume models” since Nassar et al. (2021) and (2022) along with Lin et al. (2023) all used the Gaussian plume model approach rather than the cross-sectional flux approach.
Line 96: Is there actual “chemistry” such as chemical reactions, or just “tracer transport”? I believe it is actually the latter.
Table 1: Surprisingly absent from the table is the Sasan Ultra Mega Power Plant (https://www.gem.wiki/Sasan_Ultra_Mega_Power_Project) near Vindhyachal and Rihand. Furthermore, is there any objective system behind the power plant listed order? Perhaps re-order from largest installed capacity to smallest or north to south latitude.
Figure 1: Nassar et al. (2021) Figure 7 shows the region of the Rihand Reservoir (Govind Vallabh Pant Sagar Reservoir) with multiple power plants: Sasan, Vindhyachal, Singrauli and Rihand. The additional power plants would be relevant secondary sources in emission estimation. One would expect a paper focused on emission estimation from Indian coal power plant to include these other known sources. Singrauli apparently has a planned expansion of 2 additional 800 MW units: https://www.gem.wiki/Singrauli_Super_Thermal_Power_Station
Line 111: The fossil fuel scaling factors of Nassar et al. (2013) are commonly referred to as “TIMES (Temporal Improvements for Modelling Emissions by Scaling)”. It would be clearest for readers if the scale factors were referred to with this name.
Figure 2: Panel e) EDGAR seems to show two distinct sources. The source at 24.5°North and 83.0°East is clearly smaller, but the WRF-GHG simulation has no real plume there, although there is a feature that is at about the level of the noise. Would the authors not expect a secondary plume in CO2 and NO2 based on the EDGAR emissions?
Line 159: There is inadequate discussion of the temporal offset between TROPOMI NO2 and CO2M CO2 observations. The authors need to elaborate more on their assumptions and the implications of a temporal offset when using observations from a different satellite compared with fully coincident observations in both space and time.
Line 218-219: Are both the speed and direction of the representative wind used? How?
Figure 3b – surprising that T tied for lowest number of “Excellent” and had second highest number of “Poor”. For Vindhyachal, secondary sources contaminating the plume would give systematically high emissions, while sources contaminating the background would give systematically lower emissions, which the only possibility that the authors seem to mention.
Line 313: What were the criteria for an emission estimate to be considered valid/invalid?
Line 322: It should be noted that by applying the Nassar et al. (2013) TIMES scaling factors to the model simulation, the emissions observed during daylight/midday will be higher by a small percentage compared with the reported annual emissions disaggregated uniformly.
Line 370: Explanation says that CPR and KDI sources are embedded in dense industrial regions where neighbouring sources contribute to the detected plume. Do they also contribute to EDGAR (True emissions)? CPR ET emissions are extremely overestimated.
Line 377: Bovensmann et al. (2010) do not use the CSF method. In fact, it is not even mentioned in their paper.
Line 385: The explanation for underestimation seems to be an important finding. Is there a set of figure panels resembling Figure 2 that could be used to illustrate this better for the reader rather than just a text description?
Line 408: In fact, the OCO-2 swath is usually less than 10 km, depending on the position of the sun an orientation of the satellite. Perhaps replace “10 km” with “≤ 10 km”.
Line 422: Missing beginning of sentence or word: “he contrast …”
Line 429: VDL is not really the most isolated emitter, with Rihand nearby according to figure 1, which does not even account for all nearby sources.
Line 449: It might not be worth reporting a detection limit of 20 MtCO2/year based on the inconclusive finding of this study in contrast to the Kuhlmann et al. (2021) value of 3/5 MtCO2/year.
Line 474-476: Neglecting transport error could have a major impact on results.
Line 476-478: Surely, the fact that the model resolution at 10-km is coarser than the satellite observations will introduce an error that could prevent a fully quantitative interpretation of the results.
Line 485: I believe this sentence should refer to “utilizes synthetic CO2M observations”, not just synthetic CO2 observations in general.