Parker and coauthors investigate the performance of the JULES land surface model for simulating wetland methane emissions with particular attention paid to poorly simulated African wetland regions. They run several different model setups along with different model forcings. The JULES estimated methane emissions are evaluated against atmospheric CH4 retrievals such as GOSAT/TOMCAT. I found this paper to be an enjoyable read and congratulate the authors on that. It was well-written, the arguments were sensible and well-laid out and I was able to easily follow the storyline. I appreciated the work with CaMa-Flood as that is a nice attempt to address a difficulty that land surface models have with wetlands (how to get the water to a location without it needing to fall from the sky in that grid cell). I think this paper is good for publication with only some minor revisions based upon my comments below.

Major comments:

From how I read the paper, there is a dependence upon accurate anthropogenic/other natural/fire CH4 emissions for the attribution to wetlands from the GOSAT/TOMCAT retrievals. It appeared to me that those non-wetland sources were assumed to be perfect (along with the atmospheric inversions). I would have liked to see some attempt to understand how reasonable these other CH4 source estimates were as all error terms were then pushed into the wetland methane emissions. I think it could be worthwhile to check on how much this included error affects the evaluation, perhaps byby some sensitivity tests. For example, changing the source strenghts of the other sources and checking if the wetland source distribution remains stable and doesn't change appreciably in location or strength. At the very least, please include some discussion on the impact of these assumptions, what errors this could be ignoring, and how it may change the evaluation.

Minor comments:

- Line 78: 'A deep layer of restrictive water flow' - does that just mean that you provide a no flow condition at 3 m?

- L109: Why is the time series scaled to 180 in particular? Why is this step necessary or desired?

- Fig 2: Do all of those in the grid actually give 180 Tg/yr in 2000? Ones like the bottom left seem to hardly be able to (although I realize the time shown is Aug 2011)

- L 137: When a single C pool is used does that mean both the litter and soil (humified) C are tracked in only one pool?

- L 150: Why use the SWAMPS dataset by itself, with its known inability to detect saturated, but not inundated, wetlands, and not make use of something like WAD2M? I see you use WAD2M later so are definitely aware of it.

-  line 179 - fix ref.

-  Fig 4 - what are the units?

-  L 320 - WAD2M uses more than microwave remote sensing. Perhaps give a bit more detail here otherwise it sounds like it is just SWAMPS (which does form the seasonality but there are other important differences)

-  Fig 12 - missing reference at end? (Fig: boxplot)?

-  Line 492 - chimney venting? Is this aerenchymal transport that is meant?

-  Code availability - user account required limits reviewers ability to check over code (should they wish to remain anonymous). 

-  L 532- doesn't quite make sense. Needs rewording.

This manuscript presents an interesting analysis of GOSAT observed seasonal cycles of methane and its use for evaluating various configurations of a simplified model for representing methane emissions from natural wetlands. The focus is on emissions from Tropical wetlands, for which the representation of meteorology turns out to be most critical. However, since a range of meteorological variables are used, it remains unclear which of the variable is most critical. The use of a hydrological model that is capable of river routing is found important also, which makes sense for the African wetlands that are studied in detail.

The approach of using GOSAT data for evaluating wetland emissions is an interesting option for the Tropics, where surface measurements are scarce. But the proposed method is tricky also, as I will explain below. Some disclaimers are missing to make sure that the interested reader is aware of its potentially important limitations. With these issues sufficiently well addressed, which will require substantial revisions, the manuscript should be ready for publication.

• The TOMCAT model is used to bridge between methane emissions and GOSAT observed column averaged mixing ratios. A model run without wetland emissions serves as a reference that is subtracted from the GOSAT data to derive a ‘observational’ dataset that is used to evaluate different configurations of the wetland model. It should be made clearer that this evaluation depends critically on the validity of the TOMCAT simulation without wetlands. The uncertainty of that simulation should receive more attention. The implicit assumption is that this uncertainty is small compared to uncertainties due to wetland emissions. However, no evidence is presented in its support. It would have been easy to include a figure comparing TOMCAT to background measurements and assess whether the model – data mismatch is consistent with wetlands as the most uncertain component. It is true that some important other sources do not show a strong seasonality, but due to seasonal variations in atmospheric transport their impact on total column methane will nevertheless vary seasonally.
• Even if the model performs well against background measurements, this is not a guarantee that GOSAT – model differences are due to wetland emissions. This needs to be acknowledged somewhere.
• An ensemble of wetland configurations is used to represent the uncertainty of wetlands, including an alternative representation of meteorology. However, it is unclear why the alternative meteorology is only used to drive the emission computation and not its transport in the atmosphere.
• It would have been useful to include another representation of the global methane sink and methane sources other than wetlands in the ensemble.  
• It is unclear how the TOMCAT model tracers are initialized. If the model starts at 2009, when also the comparison with GOSAT starts, the initialization needs to be very good to do without a spin-up to bring the global methane source and sink for each tracer in balance. An explanation is needed of how this was done.
• Figure 5: It is unclear why the correlation color legend starts at zero. How would negative correlations show up?
• How are regional averages in Figure 5 taken? I suppose that model has been sampled to the coordinates of the GOSAT soundings? But then the global and other region averages are weighted by the uneven coverage of the GOSAT data. In addition, the impact of regional emissions on the total column is not limited to the region where the emission takes place. It could be that emissions from another region contribute more to the reported variability of methane over a region that the sources that are located there. How is this issue dealt with?
• Related to the previous point: what could be the influence of the seasonally varying coverage of the GOSAT measurements on the derived seasonality for a particular region? How do you avoid that spatial differences between GOSAT and TOMCAT "alias" into apparent seasonal differences?   
• Figure 6: Why are changes in correlation coefficient and standard deviation only in positive direction? What happens if the correlation coefficient or standard deviation of the subset is less? If these plots represent the absolute value of changes than the explanation in the caption about improvement or worsening makes no sense. An extended explanation is needed here.
• Given its importance, it would be useful – without much work – to differentiate the impact of meteorology further. Is precipitation the dominant factor?
• Figure 8: The two ensemble member need more distinct colors to be able to see which is which.
• Figure 10, 12 and 13: the references to subfigures in the captions is wrong. What is the color legend of the MODIS imagery, is this RGB?