Please see the attached PDF of my comments

The authors use a combination of datasets to estimate the grounding line discharge of Antarctica from 1996 to 2023. This is an extraordinarily complex problem because of the heterogeneities of the datasets, uncertainties in ice thickness, elevation changes, correction for thickness, location of flux gates, etc. and how error propagate to that chain, which means in practice that a blind approach to the problem is likely to fail to identify errors and/or their cause. Identifying local uncertainties is especially relevant to focus future observations.

Overall the study is well written, the details of the methodology are properly explained and relatively well thought through, including correction for polar stereo distance, but it is not clear that the mixing various models of unknown performance levels necessary yields a more reliable estimate of errors since model performance is not included in the weighting of the different inputs. The authors also offer no analysis of the changes in ice flux from their approach compared to other methods, so it is unclear where the differences are and what is their cause. As a general guideline for science, when updates of prior results are presented, it is a valuable exercise to compare with prior work, identify the differences, their impacts, and their sources. In this regard, the paper is lacking details. While I am not asking the authors to fill that void (that’s their decision in the end), this gap will make it challenging to incorporate it in broader assessments which generally try hard to understand those details in order to generate further advances.

I have a few major comments and even fewer minor comments:

Major comment: 

The Frankbed is essentially BedMachine v2 (BedMachine v3.7 is available, would be good to update the study with it). Huss and Farinotti (2014) is a (relatively old) model-based thickness, with unknown skills, because there are very few quality thickness data in the Peninsula other than at the GL based on flotation and precise surface elevation, so the error of Frankbed compared to BMv2 is unknown in the study.

BMv2 already takes FC into account on the ice shelves, why remove FC again? What errors are introduced this way and is there a way the authors could evaluate their correction (or validate)?

SMB models have varying levels of performance in Antarctica, mixing their estimates could provide an indication of the uncertainty in SMB but does not really replace evaluating the skill of any of these models. RACMO2.3 is notoriously superior to other methods like MAR; and the HIRHAM is a notch below in Antarctica; but even various versions of RACMO2.3 disagree, especially in East Antarctica; and these differences are not discussed. There is also a specific version of RACMO2.3 for the Peninsula, but not sure the authors are aware.  

FrankenBedabj includes corrections for match elevation changes over 1996-2014, which is a large source of uncertainty since altimetry over that time period mixes radar and laser instruments onboard various platforms, with varying level of performance, and surely low skills in steep areas, e.g. Peninsula. A comparison of this topography with actual data would help establish the actual performance level of this Frankenstein version of BedMachine. I could not find any element in the study that would convince me why this bed topography would be better than BMv2. It might very well be, but this has to be discussed more quantitatively.

Later on, the authors correct REMA with altimetry 1992-2023, which introduces the same type of uncertainty at the surface, but they also correct for anomalies in SMB, but thickness is already corrected (twice if not once), so the entire picture is confusing. Would be nice to clarify.

The balance discharge from 3 x SMB models is problematic, especially in East Antarctica where RACMO2.3p2 has a lower performance than RACMO2.3p1, and in the Peninsula where only RACMO3.2 at 5.5 km has skills at reconstructing SMB. I do not have a sense of how these errors propagate in various estimates.

Minor comments:

Some comments seem arbitrary: “who manually selected different approaches to determine grounding line discharge and who modified ice thickness in a non-algorithmic manner in order to produce the required ice flux.” Some specifics should be defined here rather than these blanket statements; not sure what “non-algorithmic manner” means, nor what is meant by “manual” approaches. This is a case where one would question: how do you know that your estimate is more precise than a prior one, and give us as many examples as possible for the most important basin. For instance, discussing uncertainties for Conger ice shelf is almost irrelevant. But discussing uncertainties for the Amundsen Sea Embayment sector of West Antarctica would seem to be a key priority.

Line 125: It is unclear how the authors mix these velocity data from multiple sources into a cohesive dataset. How do the errors from these different products propagate? Monthly velocity maps at 100 m from velocity tracking seem almost impossible to achieve with S1. Why are they using ITS annual velocity maps but not the MEASURES annual velocity maps? Why is the multiyear map from 2020-2022 not used? Etc. The product quality varies significantly, so it is not clear what the aggregation of all of these estimates does to the analysis. An assessment should be made.

“Given that Antarctic Ice Sheet mass changes estimated from gravimetry and altimetry agree much more closely with each other than either do with the single available input-output mass change estimate (Otosaka et al., 2023)” ? This is not correct. Most of the disagreement is in East Antarctica. I would like to understand how the authors substantiate this comment.

Conclusions:

Overall, the paper has value but glosses over the performance level of the various components used in the analysis to offer a holistic approach with little insights about differences, uncertainties, and causes. The comparison with prior estimates is too brief. Thickness ought to be a major one but the mixing of various velocity estimates could be another one, esp. given the large uncertainty of ITS-Live coarse feature tracking. For major basins in West Antarctica and East Antarctica, the authors ought to compare their results with prior studies, identify the differences and explain them, but the results are very important to the ice sheet mass balance estimates. It would help justify why the current estimates are qualified by the authors to be “the best”, which is amusing and unjustified.