This dataset brings together ERS-1/2, Envisat, ICESat, Cryosat-2 and ICESat-2 observations into a new temporal mosaic of the mass balance of Greenland land ice over the period 1992-2022. The method of downscaling these datasets with higher resolution, and independent, velocity and hypsometry data is very novel. The product is evaluated at NASA ATM observations and compares well. I enjoyed the manuscript and below I provide some general point where I felt the reader would appreciate more clarification.

Elevation vs. Mass Balance – The title and Figures 4 and 5 highlight elevation change through 2023, but the main abstract-level findings and Table 2 are mass balance through 2022. Perhaps during the review process, the FAC products will be updated by another year to allow the mass balance conversion to be updated through 2023 as well, but I wonder if it would be helpful for Figure 4 and 5 to show mass balance patterns and time series, rather than elevation patterns and time series? The ultimate goal, and discussion point, is mass balance, so it would be nice if this is visualized.

Altimetry Data – It might be helpful, either in the Introduction or Section 2, to give some characteristics of the different satellite missions to place them in context. For example, perhaps footprint size at a given elevation, along track shot spacing at a given elevation, temporal repeat, track spacing at a given latitude, etc. The lasers altimeters are really quite different from the radar altimeters in this regard, and perhaps it would be helpful for the reader to understand any differences between the radar altimeters?

Filtering – There a several filtering steps. Sections 3.5 and 3.7 both have sequential 10-sigma and +/- 10 m residual filters. I guess this probably removes only a small fraction of observations. But Sections 3.8 and 3.9 both refer to rejecting data beyond the inter-quantile range, meaning below 25^th and above 75^th. This seems to be happening sequentially. Does this mean removing 50% in 3.8 and then 50% again in 3.9, leaving only 25% of the original observations? In Table 1, you have a count of how many observations per mission are used in the cross over analysis. It would be super helpful to have a similar table showing how many observations you start with for each mission, and how many observations remain after filtering, or even the steps therein. It is hard to tell which filter is the main QC, but I guess it is 3.9.

Spatial Resolution – There is a stated “native output” of 1920 m, with several intermediary resolutions. I see the slope correction (Section 3.2) is being done at 2 km resolution, the topographic separation (Section 3.3) is being down at 500 m resolution, the scattering correction (Section 3.4) is done at 1 km. I am not sure what resolution is being used for the critical step of the regression merge (Equation 1; Section 3.8). This is presumably happening at the native output resolution, but this should be made explicit. How are the intermediary corrections and higher resolution regression datasets (velocity and DEM) are being interpolated? Is there an influence on the final product if intermediary corrections are done at better/differing spatial resolution?

Correlation Lengths – Somewhat related to spatial resolution above, there are different correlation lengths adopted in different sections. For example, in the multi-mission stacking (Section 3.5) the correlation length is 2 km. With a 1.92 km resolution this is effectively one grid cell? In the multi-mission merging (Section 3.8) the correlation length is 20 km. Then in the post-filtering of the product (Section 3.9) there is reference to a correlation “filter length of nine points”, perhaps meaning 9x1920 = 17.28 km? Then the error analysis (Equation 3) uses a correlation length of 40 km. Lastly, the FDM models are said to have a correlation length of 200 km, based on a covariance analysis. Perhaps the reader would appreciate just a couple sentences explaining why specific correlation lengths are being chosen? Or, put the other way, why is there not a uniform correlation length to represent the length-scale of ice-sheet change in this multi-step analysis?

Velocity and Hypsometry Regression – This approach of using higher resolution DEM and velocity to downscale altimetry observations seems super novel and promising! The reader would like to see the “clear relationship” between velocity and elevation change in the main methods here (L192). If this relation is linear, then why is it weighted as an inverse squared in the regression? The reader could also use a better description of the velocity and hypsometry data being used. Presumably they are each single products applied to the entire time period? Or is some attempt made to acknowledge that velocity and hypsometry have changed over the 30-year period? This also makes we wonder, if you are downscaling with velocity and DEMs of better than 500 m resolution, and topographic separation can be at 500 m resolution, why isn’t the final product one of these better resolutions, instead of the rather curious 1920 m resolution?

Novelty – The novelty of these result are clearly perhaps their long temporal span and high spatial resolution, but the opportunity to actually highlight this is perhaps missed. For example, the Results begin (L304): “Our estimates of elevation change … are in line with previous studies … for areas of large change, such as Jakobshavn, Helheim, Kangerlugssuaq and Storstrømmen glaciers. However, by incorporating the background model, guided by velocity and hypsometry, we have achieved higher spatial resolution in regions characterized by pronounced dynamic thinning, such as Jakobshavn, Helheim, and Kangerlussuaq glaciers.” This is confusing. Are the results the same or different for these big glaciers? Perhaps focus the reader on the new insight and higher spatial resolution in the pre-2003 period, specifically?

Simonsen2021 (https://doi.org/10.1029/2020GL091216) – This previous study also pulled together Ku band ERS / Envisat and Cryosat radar altimetry observations to create a 1992-2020 altimetry record of the ice sheet. It used also a non-traditional methodology, albeit very different methodology than presented here. It might be helpful reader to the Greenland altimetry reader to place the current multi-mission assessment in the context of this past multi-mission assessment. For example, at least a simple compare and contrast of ice-sheet wide timeseries of mass loss since 1992. Perhaps probing differences, if there are any, at sector scale?

Nilsson2022 (https://doi.org/10.5194/essd-14-3573-2022) – I appreciate that this is a companion paper to a previous Antarctic assessment, but at present there are perhaps >35 references to Nilsson2022. This limits the reader’s ability to quickly understand this new Greenland dataset with a stand-alone read. I would encourage the authors to scrutinize how many of these references could possibly be removed by adding a sentence or two of clarification into the current article.

These are all suggestions for the authors to consider.

An excellent and clearly written paper, with the a sound methodology of remove-restore. It is especially to see how useful the CryoSat SARin mode has performed relative to IceSat-2. This is a landmark paper on Greenland ice sheet changes.

A few comments:

It would be useful to mention the use of outlet glacier ice  velocity for elevation changes already in the abstract, since this is not a "clean" elevation change paper. Also, could be useful to include in one of the figures the outline of the peripheral glaciers - not all readers know the Bolsch et al reference, and the definition of "peripheral glaciers" there are controversial (a large part of the central East Greenland ice sheet south of Scoresbysund is e.g. classified as "glacier" even though it is not (just an area of ice sheet with Nunataks). Therefore the split quoted between "glaciers" and "ice sheets" is somewhat ambitious.