Long et al. present urban methane emission estimates for London, New York, and Los Angeles using an updated source pixel method and TROPOMI methane retrievals. The improved method uses a revised wind analysis to incorporate pressure-weighted PBL winds into the emission calculation, rather than the 10m wind data provided with the TROPOMI data. The manuscript is well written and thoroughly presented. There are, however, a few areas that I believe require attention prior to publication.

1. In the abstract, the work is motivated by the following statement: “Previous studies of urban methane emissions have employed a range of emission quantification methods, leading to inconsistent estimates and different associated uncertainties.” However, the conclusion does not address whether this work has helped improve these inconsistencies.
2. Section 2.2: Oversampling method. How are missing pixels handled with this approach? Is the oversample done on the column mixing ratios or on the calculated methane enhancements?
3. A major component of this work is the use of pressure-weighted PBL wind instead of 10m wind in the calculation of the emission rate. Unless I missed it, I do not see a comparison of how these two approaches compare in emission space. In the comparison with other studies for these cities, how do the emission rates calculated using 10 m winds compare? Put another way, how do you know the pressure-weighted PBL results result in a more accurate emission estimate?
4. Line 212: Is the methane enhancement calculation done for each overpass, or over the averaged column mixing ratio data? If it is the former, this should be clarified. If it is the latter, would this not skew the result towards overpasses with larger overall column retrievals?
5. Line 276: “While most previous studies using the source pixel method have analyzed an aggregated sets of observations through equally weighting each day in the average, we use the spatial coverage (in %) of valid observations in the source region to weight each overpass in an effort to reduce bias.” It is not clear to me which bias this will reduce. There may be more valid pixels in a part of the domain than another, but there will still be transport within the domain such that a column retrieval for a given pixel will include surface emissions from outside the pixel geometry.
6. Figures 4,5,6: What is driving the differences in the size of the error bars between different months? Some months have very tight error bars.
7. In the several months, there have been a few new studies examining urban methane using TROPOMI (https://doi.org/10.1073/pnas.2504211123, https://www.science.org/doi/full/10.1126/sciadv.adz9308). They are not included here, so I am left wondering whether or how this work fits within the context of these more recent works. Or perhaps, they are not strictly relevant to the work here?

This manuscript presents an improved source-pixel method to estimate urban methane emissions from single-pass TROPOMI observations over London, Los Angeles, and New York during 2021–2023. The use of PBL-weighted winds and an upwind background concentration is a useful development, and the manuscript is well organized and easy to follow. The manuscript would be strengthened by further clarification and justification of key method assumptions, particularly the wind fields and the calculation of background concentrations. I recommend publication after the authors address the following comments.

1. The methane enhancement observed by TROPOMI can reflect emissions accumulated and transported over several hours before the overpass. It would be helpful if the authors could discuss and test the sensitivity to a 3-hour pre-overpass mean wind compared with the 1-hour overpass-time wind.
2. As the authors noted, single-pass TROPOMI observations can be sparse and uncertain over urban areas, while temporally averaged fields are often used for emission estimates. Could the proposed approach be applied to time-averaged methane enhancements (e.g., monthly means) to increase the number of usable observations?
3. The approach implicitly assumes no significant methane sources in the background region, so that the methane enhancement can be attributed to the source region. However, Figure A1 suggests substantial emissions around some source regions, which could bias the background concentration and inferred emissions. It would be useful to discuss the sensitivity of the results to the size or location of the background region, or to screen the background regions using available inventories.
4. Negative ΔXCH4 values and negative emission estimates, especially for London, are not physically meaningful. A more careful treatment of these cases is needed, for example by testing sensitivity to alternative background regions or excluding them from averaged emission estimates.
5. The methane enhancement ΔXCH4 appears small in some cases, for example below 2 ppb in London. Do such weak enhancements meet the detection/signal-to-noise criteria discussed in Jacob et al. (2016), and how reliable are the resulting single-pass emission estimates?
6. For coastal cities such as Los Angeles and New York, the upwind background region may fall over the ocean, where TROPOMI retrievals are filtered out under Sect. 2.1.1. Does this occur, and if so, how are background concentrations calculated?
7. The apparent increase in Los Angeles emissions from 26 t h^-1 in 2021 to 62 t h^-1 in 2023 appears inconsistent with previous reported decreasing trends in this region (https://doi.org/10.1021/acs.est.5c09817, https://doi.org/10.1088/1748-9326/acb6a9, https://doi.org/10.1038/s41467-023-40964-w), and may reflect sampling or methodological biases. Similarly, the large error bars and limited data coverage limit confidence in the inferred seasonal cycles. Further discussion of the uncertainties in these temporal variations would be helpful.
8. The comparison with previous top-down studies in Sect. 3.4 needs more clarification. The statement that this study finds “generally lower, yet overall consistent” emission estimates could be misleading, because the urban domains used here are much smaller than those in previous studies. Although the authors rescale previous estimates using emission intensity, this assumes spatially uniform emissions, which not hold for heterogeneous urban sources such as landfills and gas infrastructure. A more direct, apple-to-apple comparison would be preferable; otherwise, the uncertainty introduced by the rescaling method should be acknowledged.

Minor Comments:

1.  In the Abstract, “combat climate changing emissions” would be better phrased as “mitigate climate change” or “reduce climate-forcing emissions.”

1. Some statements in the Abstract and Introduction overstate the limitations of previous studies and imply that this study provides a more accurate emission estimation method (lines 4-5, lines 60-62). I suggest softening these claims and focusing the motivation more on the simplicity, efficiency, and urban-scale applicability of the source-pixel approach.
2. Abstract and line 375: Factors of 0.1 and 0.3 indicate emissions lower than the inventory estimates, not higher. The two reported ranges are inconsistent and need clarification.
3. Line 22: Revise to “…the second most significant anthropogenic greenhouse gas…”
4. Line 30: Grammar issue in “natural and processes”.
5. Line 33: Tibrewal et al. (2024) refers to fossil fuel sectors rather than all sectors mentioned in the preceding sentence.
6. Line 53: “actual emissions” is too strong and could be removed.
7. Figure 1, line 156, and elsewhere: Please correct the resolution format, such as “0.25° × 0.25°”, “1 km × 1 km”, or “1 × 1 km²”.
8. Section 2.2: From my understanding, the authors mainly remap individual TROPOMI pixels onto a 0.1°×0.1° grid, which differs from oversampling that usually averages multiple overpasses to obtain a higher effective spatial resolution. I suggest using “regridding” or “remapping” instead of “oversampling”.
9. Lines 316 and 369: “Landfills and waste” is inaccurate, as landfills are part of the waste sector.
10. Line 377-379: Please clarify the reference used for the reported percent differences.
11. Figure 7: Please clarify why the adjusted Los Angeles estimate is larger than the original Nesser et al. (2024) estimate, given the smaller domain used here.
12. Previous intercomparisons suggest that de Foy et al. (2023) lies the upper bound of top-down urban methane estimates (https://www.science.org/doi/full/10.1126/sciadv.adz9308). This would be useful to acknowledge when discussing consistency with previous studies.