The paper by Koch et al considers the effect of peatland rewetting on GHG emissions in Denmark. It uses a large dataset of water table measurements to test and validate a model for national upscaling. It then looks at prioritising the rewetting of wetter for drier fields to make recommendations. The manuscript is interesting and timely, considering current interests in peatland rewetting for climate mitigation. It is well-written and I enjoyed reading it. I have only minor suggestions. Note however that I have no experience of machine learning so cannot comment at all on that part of the manuscript.

Section 2.1. It might be helpful to readers to specific the climate zone somewhere here. A few words of extra text around L82 specifying agriculture (cows + specific kinds of crops?) might also be informative.

Section 2.2. Perhaps I misunderstand something, but why is the median used on L89 but mean on L93?

L160. A slightly pedantic comment, but I’m not sure I’d describe static chamber sampling with syringe sampling and GC analysis (as in Petersen et al 2012) as “state of the art” (there’s nothing wrong with this method, but neither is it anything incredibly advanced).

L171. It is good to see that waterborne C losses are accounted for, using the IPCC defaults.

L180. 10% seems quite high here. The IPCC Wetlands Supplement has 0.05 (5%) as Fracditch for boreal/temperate drained grasslands and croplands.

Figure 6. I think measured vs modelled plots are more “honest” when plotted as square panels and the same scale on both axes. That gives a true visual representation of the model.

Figure 7. I wonder how much panel a is skewed by the data point around 1.2 m WTD? Without this point would the relationship be linear, as in Evans et al (2021)? Does it make sense for CO2 emissions to stabilise at 0.4 m WTD? This is mentioned in the discussion but I’m not sure I’m convinced.