This manuscript is a challenging approach for the detection and emission amount estimation for CH4 by using a satellite-based dataset.

If this method is well performed, we can extend to a similar method to other materials. 

However, within this manuscript, some manuscript presentation is not well organized. 

Maybe, on this round, I suggest that the methodology description has to be improved, especially. 

1) Abstract in L4-L8: The detailed satellite specification is not essential to include on Abstract session. Please simplify.

2) Introduction L34-37: This sentence is not perfect. Please re-sentence it.

3) Introduction in L75-84: In this paragraph, the author identified the details of study. However, it must include and emphasize the purpose and main results through the developed automated algorithm.

4) Data in L102-109: For using the XCH4 by TROPOMI, why the period from 2018- to 2021 was used? In addition, for the spatial grid resolution, is 0.1 degree * 0.1 degree statistically significant and valid? Because the TROPOMI resolution is 7*7 km2, the 0.1-degree resolution is not sufficient for pixel sampling. 

5) Section 2.2: How the author improve the spatial resolution for reanalysis products? It would be driven the error during the resolution improvement.

6) Section 2.4: It is too simple to explain the each datasets and its purpose for use in this study. Please improve and add the details of data characteristics and purpose of using.

7) Figure 1: The author includes the overall of flowchart for automated algorithm in Figure 1. However, each part of the process is difficult to understand. Although the Section 3.3 explained the details of respective processes, I suggest that the Figures will add the flowchart of respective process to supplement of Section 3.3 explanation.

8) Section 3.3: This session has to be reorganized. Some part of parameter is explained before the parameter definition, and some parts are hard to understand because of the lack of importance of used variables. 

9) Caption of Figure 4 and 5: Figures' captions are too long. Because the author explained the details of figures in figures' captions, the explanation in the body of manuscript is insufficient. 

10) Session 3.3.2 : For the threshold selection, the author did not explain the reason. So it is hard to agree that the threshold is valid.

11) Session 3.4 in L313: During the gridding process from original pixel to gridding, is it valid to ignore the data distortion of XCH4?

12) Session 3.6: Please add the list of source types by Table or something.

13) Table 1: For PPSR analysis, the author includes the 1sigma uncertainty. Is this a methodological error? If so, can you tell how much smaller this error is compared to other methods?

12) Session 3.4 in L338: For the scale height H is assumed as 8.5 km. Do you have some references or reasons? 

Vanselow et al. developed a data-driven algorithm for the identification of regions with persistent methane enhancements based on TROPOMI XCH4 data. The method looks intriguing, and the results look consistent with prior knowledge about methane source regions. However, Section 3 about methodology is currently hard to follow. I also have some other concerns about the manuscript in current version.

Major concerns:

1. Methane concentrations are subject to long-term and short-term dynamics. For example, the research period of 2018 to 2021 contains two periods of methane trends before and after 2020. There was also a sudden change in atmospheric oxidation capacity caused by the COVID-19 pandemic. Wetland emissions show seasonality, which could also depend on latitudes. The authors identify regions with persistent enhancements based on monthly XCH4 data from all 48 months. I would expect some discussion on the impact of these dynamics on the choice of parameters (e.g., the fraction of months with enhanced anomalies) in the algorithm.

2. The emission inventories could have large uncertainties. For example, it is kind of surprising that there are no persistent oil and gas-related source regions in Russia. It would be necessary to include some consideration about addressing the uncertainties from emission inventories in the source type assignment.

Specific points:

Is there any treatment on the stripes in TROPOMI XCH4 data? How would the stripes affect the estimate of persistent enhancements?

Fig. 1: This is a very busy figure and is not very helpful for following the later explanation of the algorithm. Each step of the algorithm is complicated and requires an individual flow chart or diagram (something like Fig. 4). So I would recommend simplify this to just show the general steps.

L38-42: A recent paper could fit well in this discussion about point sources: “He et al. Increased methane emissions from oil and gas following theSoviet Union’s collapse. PNAS. 2024”.

L43-45: A recent paper is suitable to cite here: “Chen et al. African rice cultivation linked to rising methane. Nat. Clim. Chang. 2024”.

L168 and Fig. 2: Is the threshold of Nobs>3 too low for the calculation of monthly average XCH4? It would be good to see XCH4* filtered with other thresholds, e.g., Nobs>7, Nobs>15, etc.

L102-106: Is this data set related to the reprocessed XCH4 data?

L229-230: Would 3x3 grids large enough to account for meteorological effect on anomalies? This should also depend on the size of the filter, right?

Fig. 3: Please comment on the high anomalies over the Southern Ocean and in Antarctica.

Section 3.2: A schematic diagram explaining the different steps in the calculation of anomalies could be helpful for an improved clarity here. It could be something similar to Figure 4.

Fig. 9: please add a subplot zooming in over China as there are 3 regions with top 10 emissions.  Also, as the authors mentioned in the paper, would it help to add some pie charts showing the percentage of different source types as source regions could blend different types?

L440-446: The discussion should be slightly expanded here. The disproportionate distribution of methane sources agrees well with prior knowledge. Adding some references could help here, e.g., the Frankenberg et al. 2016 PNAS paper.

Fig. 11: The colormap makes (e), (f) and (g) hard to read.

L504-512: The He et al. 2024 PNAS paper mentioned above could be added here.