“A method for estimating localized CO₂ emissions from co-located satellite XCO₂ and NO₂ images” by B. Fuentes Andrade et al. is an interesting new study presenting a method for estimating CO₂ emissions from localized sources such as power plants. XCO₂ observations from OCO-3 and NO₂ observations from TROPOMI are used to estimate CO₂ emissions from the Bełchtów power station in Poland on 9 dates from 2019-2022. The method presented is based on the cross-sectional flux approach which is extended to estimate the emissions along with uncertainties. The results are compared with those of Nassar et al. (2022), which used a Gaussian Plume Model based approach for the same power plant on many of the same dates. This is a useful study, complementary to the Nassar et al. (2022) study, presenting some advantages and limitations to the cross-sectional flux method that has been developed. Continually improving our understanding of methods, biases, uncertainties and limitations for emission estimation in advance of CO2M is useful for the scientific community gearing up to deliver operational emission estimates from satellite data in the near future. The approach presented here is sound and rigorous. I see no major issues with this study and thus would recommend publication after some relatively minor revisions.

Specific points

Line 14: “possible thanks to” would better be rephrased as “made possible by”

Line 60 and 61: capitalization of ENVISAT and TANSO is the advised, although TANSO-FTS is the complete name of the instrument.

65: A Gaussian plume model does not account for eddies, however, it relies on the reasonable assumption that their effects are negligible for multi-kilometer spatial scales. It is recommended that the sentence is expanded to clarify this fact.

Line 116: “instantaneous hourly” would be more informative than just “hourly” to distinguish from an hourly average value.

Figure 1 caption “gross” should be “cross” or X.

Line 156: Is there any justification of the requirement of less than 5 hours?  Obviously a shorter offset in time is better, but are there any studies to quantify the effect that might justify this value? Both wind speed and direction could change significantly over a period of 5 hours, as discussed later around line 190.

Line 208: This approach to account for swath bias is interesting and likely contributes to an improvement in emission estimates, however, should the swath numbering be “j = 1,2,… n”, rather than only going up to n-1? Is it n-1 since the first swath has no offset, so j = 0,1,2 … n-1, where s₀ = 0?

Line 277: 1-3 hours for the characteristic time used to determine the bottom-up value is consistent with the findings of Nassar et al. (2021, https://doi.org/10.1016/j.rse.2021.112579, e.g. Figure 1 and sec 2.5), which considered the plume extent, time since emissions to derive a time-weighted or ‘dynamic’ bottom-up value. This similar analysis is worth mentioning very briefly and citing.

Section 2.3, uncertainty. Is there any uncertainty related to the observations? It was not entirely clear to me if this was indirectly included in the dispersion or sensitivity terms. The sensitivity term does account for uncertainty in the observations for background, but not necessarily the plume. Can the authors clarify?

Line 559: “lead” should be “led”

Line 595: It is not surprising that the difference between applying quality filters and ignoring them reduced when observations near the Bełchatów lignite pit were excluded. The digital elevation model for OCO-3 v10 data does not account for recent anthropogenic effects on topography such as this, so biased XCO₂ data will result through erroneous surface pressures. Although no DEM will be perfectly up to date with respect to anthropogenic effects on topography, the Copernicus DEM which will be used in OCO-3 v11 data will reduce the problem and thus the difference between quality-filtering and not, will be reduced.

Reviewer comments provided in the supplemental file.