The authors provide a report on an open dataset of terrain and seafloor topography created at global level by integration of various existing products. The results are to be praised as a valid contribution towards integrating data, which can be used for future investigations, as these terrain/seafloor models are key for many applications. The authors correctly report on the methodology including the validation but not much is reported about the existing data merged together, its own accuracy metrics. It should be noted that the error distribution, propagation and thus the relative reliability of values in the final product are key information.

The work reports on integrating and validating a 15-arc-second topography and bathymetry dataset that consists in the collection of multiple other products. It provides “bare-earth” dataset, as reported in the abstract, thus a terrain digital model (DTM), as opposed to a digital surface model. We therefore expect the altimetric data to refer to the ground without vegetation and buildings. This should be noted as a limiting factor to some technologies is the canopy cover. NASA’s ICESAT data are used to validate the altimetric values.

One thing to note is to assess reliability of the map, according to where the ICESAT values were used as validators, as the spatial distribution of high-quality elevation data might be biased to specific scenarios (e.g. low/high vegetation, flat terrain vs steep slopes etc…).  The challenge is two-fold: rigorous assessment of data also in complex scenarios, e.g. thick vegetation and/or mountainous terrains, and map reliability metrics to address areas were little information is available.

I commend the authors for producing this data set, and the comparison with ICESat-2.  They convincingly demonstrate that the new data sources (and probably improved techniques) greatly improve on the older version of ETOPO.  This data set directly matches SRTM15+, which should be acknowledged.  There should be a direct comparison, and short discussion of the differences, between ETOPO and SRTM15+.  I would suggest a map like Figure 9 for the differences, and an RMSE map like Figure 7.

I commend the data set (or its creators) for doing the following, which is unfortunately not universally done with similar data sets.

• Including the version number of the DEM in the file names
• Including the vertical datum with Geotiff key 4096 inside the file’s metadata.

Lines 165-170: Is this data set publicly available?  I checked the PE&RS paper, and can’t find any reference on where to actually get the data.  Most of the data sets I am familiar with, but it would be nice to have explicit download links for all of these which are public.

Line 180: how was the reprojection done?  Going from the US projections to EGM is non-trivial.  This is covered later in section 4.2.  I think VDATUM only works in the US, so that limitation might be mentioned.

Line 185: please provide the link.

Line 332, Figure 4: is the difference between the two cells due to latitude (higher coverage at high latitude), clouds, or something else?  The coverage would of course be greatly improved by using additional years data.

Line 359: delete “and” (the 4^th word)

Line 380, figure 7: the colors on the figure are very hard to read.  Can the color palette be improved?  Or use two figures with different ranges, one for the low RMSE, and another to show the large ones?

Line 439: put in date

Comments for the authors to think about for future work, which do not necessarily need to be addressed for this paper:

There are now many more years of ICESat-2 data, so Figures 4 and 5 could be greatly improved, but that is not a reason to revise the manuscript.

The following comments are based on my work with DEMs, most importantly the 1” global data sets.  I include them here so the ETOPO team can think about them, as the 1” DEMs are generalized and the ETOPO moves to smaller and smaller grid spacings, I see issues arising:

• I would prefer that the tile names start at the SW corner, which I think is more common with global DEMs, or include the full range as is done with the AW3D30 (N000W060_N005W055, admittedly for zipped directories, but could be done for tiles). One can of course keep track of the peculiarities of each DEM, but it would be nice for all of DEMs to be consistent, and for a quick glance at a file to be unambiguous).
• Most of the 1” DEMs have the pixel centroids aligned on parallels and meridians; a few have a ½ pixel offset. This is more than just the pixel-is-area or pixel-is-point, but is also affected by the starting grid corner.  This means that point elevations cannot be directly compared (for instance between CopDEM and AW3D30) without reinterpolation of one.  It would be nice to agree on the pixel representation moving forward, and have a consistent standard.