This paper combines surface and aircraft measurements of atmospheric methane, together with modeling estimates to generate a reference dataset of methane column over a large region over the pacific ocean, south-east of Japan.  The resulting time series are analyzed to discuss the growth rate and seasonal cycle.

The paper is very well written.  The method is clearly described and the various uncertainties are discussed in detail.  The paper can be published with minimal changes.  I nevertheless offer some suggestions to the authors below :

Line 27 : I recommend to use ppb, consistently with the rest of the text, rather than %

Line 36 : You could say that methane is the second most important anthropogenic GHG after CO2 (rather than “one of the most”)

Line 208 : “Instantaneous lifetime as short as one year” for the summer condition is not clear.  Rather, you could provide the oxydatation fraction per month

Figures 3 and 4 are not clear.  I suggest to not show the shaded areas, but only the best estimates together with a single bars for the full period that would indicate the typical uncertainty range

The conclusion is more a summary than a conclusion.  It would be better to offer a real conclusion to the reader

In addition, the last paragraph is not a conclusion but rather a discussion.  Please correct

The authors develop approaches for generating XCH4 time series over ocean combining ship and aircraft measurements with model data. The observation-based XH4 data are compared with independent TCCON measurements and finally used for evaluating GOSAT measurements. The paper is well written and within scope of AMT.

I have some minor comments related to the uncertainty calculation that should be addressed in a revised manuscript:

• L82: What would be the required accuracy for a dataset assess the accuracy of trends and variations in XCH4 satellite observations over oceans? Moreover, what is the accuracy that is achieved with the dataset presented in this study?
• Figure 1 could already be mentioned in the beginning of Section 2.
• Section 3.3.1: The calculation of the tropospheric uncertainty of XCH4 is difficult to judge mainly because no profiles are shown in the manuscript. Please add a figure comparing the constructed CH4 profiles from measurements and the MIROC4-ACTM model with the HIPPO profiles.
• It would also be interesting to see how well your approaches can reconstruct a HIPPO profile when taking the three measurements (2 aircraft + 1 ship) from the HIPPO profile.
• L249: Can you explain why you only used only six profiles for assessing the MIROC4-ACTM simulations, while 20 profiles seem to be available in the study region?
• L255: Finally, it is unclear how the uncertainty in the profiles translate to XCH4 uncertainties. Do you do an error propagation or sensitivity study as for the tropopause uncertainty?
• L337: If you have three simulations and two agree with each other, it is not valid to conclude that the agreeing models are correct. The conclusion in this paragraph need therefore to be argued more carefully using previous results from literature (as done in Section 3.3.3) or conducting additional analyses (e.g., comparison with independent measurements).

Muller et al., developed three approaches to construct the CH4 vertical profiles, and then use them to calculate the XCH4. They then show how these observation-based XCH4s can be used to investigate the seasonal variation in CH4 and to evaluate the satellite observations of CH4. This study involves a large amount of data, including airborne and ship measurements, satellite observations, and model simulations. This manuscript is well-organized and is within the scope of AMT. I recommend its publication after the authors address the following comments:

1. Sampling bias is a big concern as the ship observations are 6 +- 4 days and the airborne observations are 2+-1 days each month. The uncertainty arising from limited measurements is not well covered in Sect 3.3.1.
2. Figure 2 includes a lot of information. I would add observation-based profiles from 3 approaches in different colors, instead of showing them in one symbol.
3. Figures 3 and 4 are very difficult to read. It is also unclear where is the gap in the dataset. In Figure 3, the linear regression fitted line is not in the legend and has the same color as blended obs/ XCH4.
4. Line 154: Why do you choose the data that only assimilates NOAA surface observation? How is it different from assimilation using both NOAA surface observations and GOSAT observations?
5. Line 367-370: Why the increasing trend in XCH4 is larger between 20-30N than 30-40N.