This manuscript presents a new public data set of global peatland depths, compiled from published literature, national or regional datasets, and some unpublished data.  This is an important new data set for global carbon cycle assessments and earth system modeling activities.  ESSD is an appropriate journal outlet. The paper is very well written.   

I have no major questions/issues with the manuscript. All of my comments and questions below are minor ones.   

Line 60: ‘often contain errors’ seems a bit harsh for a ubiquitous problem.  maybe just ‘can contain errors’? 

Line 61: ‘outdated measurements’: I don’t know what this means. 

Line 68: Not sure if this is necessary, but it might be worth noting that most land-surface models need peat depth data for initiation as they do not simulate the multiple millennia needed to accumulate the peat, and the peat stocks are not necessarily in equilibrium. They cannot be ‘spun up’ to an equilibrium state. 

In Fig. 1, step 1 is to ‘confirm basal depths’. To me this implies confirm that the values are correct (how?) but maybe it means ‘confirm reported peat depths are basal’?  Also, step 3 includes ‘remove unnecessary information’, which makes sense.  In the best of all possible data worlds, it would be easy to align Peat-DBase data with other data sets (e.g., one of basal ages, or one of age-depth profiles).  The burden of creating this best-of-all-possible-worlds shouldn’t fall on you, but it seems to me that it must be something you have thought about. Would any of this ‘unnecessary data’ be useful in this regard?  Would that introduce uncertainties that you don’t want to (and shouldn’t be expected to) have to manage?  Could this be discussed in the ‘future work’ section, perhaps identified as ‘future work for the community to move the field forward', not specifically as future work for Peat-DBase and this manuscript’s authors. 

Line 125: are some of the peat depth data in the ocean from Treat et al. 2019, which included coastal shelf peatland data from the last glacial?  If so, these are not in error, just not relevant for Peat-Dbase. Also, there is not universal agreement on where the land ocean boundary is, and it certainly depends on spatial resolution. 

Table 2. ‘depth_cm’: is that reported to any particular significant figure?  nearest integer cm? 

Table 2 and elsewhere: for the sample duplication flag, one data sample is the ‘first instance’.  How is this selected/determined: earliest published, first that you acquired, first that you entered, ...?  It likely doesn’t matter in terms of the database, but it would be good to explain what you mean by ‘first instance’ as priority in time is an important currency in academic publication. 

Fig. 3 caption: Bin sizes are 3° in longitude; how many in latitude? 

Line 213: ‘enhanced growth, stagnation, or erosion’ -- I think that ‘loss’ (i.e., from ‘excess’ decomposition, not erosion) should be added to this list. 

Fig. 4 caption: it would be useful to add binning sizes (50 cm?, 1° latitude?) 

Fig. 5: The color scale is not too helpful here, at least for me, as the x-axis is number of measurements.  I think it would be helpful to label the bar at the top ‘zero depth’, since it doesn’t really work on the vertical log scale (I initially mis-interpreted this as the sum of all other data, but, of course, it didn’t add up).  In the caption: I don’t understand why a focus on depth distributions leads you to remove data points labeled as drained or modified. Particularly since you cannot make this assessment for all data. Nonetheless, if you maintain this, I suggest adding an ‘n = NN’ in parentheses for number excluded. 

More generally on the drainage/disturbance question: you say in the captions to Figs. 4 & 5 that drainage information is only available for the NatureScot dataset.  How is that noted in the Peat-Dbase (it is not mentioned in Table 2)?  Others may also want to make the exclusion that you did. 

-Steve Frolking