This paper sets out to demonstrate the use of Tomofast-x for inversion of gravity data for bathymetry beneath ice shelves and poorly surveyed open water regions around Antarctica. Using both a synthetic model, and a real case study in Vincennes Bay the authors demonstrate the method and discus its limitations. In addition they implement their new bathymetry within a local ocean circulation model to show the impact of the revised bathymetry on ice shelf basal melt rate, and hence the relevance of this type of inversion for understanding potential ice sheet stability.

The paper is well written and clearly laid out. The methods are clearly explained, and results appear convincing.

I have a few minor comments relating to other work and clarity of the text laid out below.

It might be interesting to contrast your results with the recent work of Charrassin, et al 2025 https://www.nature.com/articles/s41598-024-81599-1 who used an inversion of the continent-wide ANTGG dataset to recover bathymetry all around the continent. I think it may reveal the benefit of using more local inversions for specific important use cases.

L125-129 It might be worth stating here that it is the removal of the interpolated misfit between forward and inverse gravity model which is the method used in this study for minimising the impact of both long wavelength effects such as isostatic variations in crustal thickness, and local variations in crustal density.

Figure 11 caption. (c) should be (b).

L424 to 426. When discussing uncertainty it may be worth mentioning that significantly higher errors may be found in areas with rugged topography, as the wavelength of the gravity data means that features <~6 km wavelength cannot be resolved.

L509 “Vincennes Bay meltwater pathways and implications”. It may be clearer to say something like “Vincennes Bay pathways for warm ocean water and implications for ice shelf melt”. Meltwater would be what is coming out from beneath the ice sheet, or the freshwater generated from the ice shelf, which I do no think is what is meant.

The submission entitled ‘Gravity-derived Antarctic bathymetry using the Tomofast-x open-source code: a case study of Vincennes Bay’ by Bird et al. presents a bathymetric model of the Vincennes Bay region, East Antarctica, including ice shelves and the open ocean. The resulting bathymetric model is combined with oceanic modeling to infer basal melt rates. The inversion is done by adapting an existing open-source software, Tomofast-x, towards the purpose of modelling (subglacial) bathymetry. In doing so, they provide a valuable alternative to existing licensed software, allowing improved accessibility and reproducibility.

Overall, the paper was a delight to review. It is well-structured and the results are very well presented. In the following are mostly minor general comments and some line-specific comments/suggestions.

General comments:

Bathy model, constraints:

- Figure 4c shows available minimum depth constraints from meop data in the model area. Here, seal dives are shown beneath the ice shelf. Is this due to positioning issues of the meop data? (In Fig. 4d of McMahon et al., 2023, these sub-ice shelf data are not shown). If it is due to bad positioning, these constraints are not trustworthy and should be removed.

- In IBCSO V2, there are a number of ‘isolated soundings’ in the area. Was it a conscious decision not to include these as constraints?

Bathy model, regional gravity field:

The interpolation of the regional gravity field in Fig. 9c does not appear to be ideal. Values below -20 mGal mostly appear in interpolated areas and rarely in constrained parts. This is quite noticeable at the trough continuation of the Vanderford Glacier, and offshore of BG and ANG. Did you compare different interpolation methods and respective results, other than MC?

Bathy model, gravity misfit:

The gravity misfit in Fig. 10d should be described and discussed, especially:

- the high misfits close to the grounding line; this is likely caused by keeping the ‘hard constraints’ fixed close to the grounding line. The misfits could potentially be minimized by allowing the areas close to the grounding line to move (possibly within the vertical resolution; ±50m).

- high positive gravity residuals in central part of VB ice shelf and at UG; does your model suggest grounding here? Either way, discuss these areas please.

Bathy model, setup of initial bathymetry:

This is regarding the difference of mapped bathymetry and your bathymetric model. With a mean diff of -34 m, it shows a clear trend. One way to mitigate that would be to use a first iteration of your bathymetric model instead of IBCSO V2 to fill in the gaps between constraints and then recalculate the forward model, and subsequently the regional gravity field. If IBCSO V2 is generally too shallow, then your modelled bathymetry might end up being too shallow, especially close to and at constraints.

Bathy model, overall:

I’m not necessarily suggesting re-modelling here, but the regional gravity interpolation and gravity residuals should be discussed more thoroughly. Same goes for the water column. Please add a grid of it to one of the figures. And discuss it, especially if there are aeas where your model suggests grounding, but the ice surface doesn’t.

Basal melt rates:

How do the basal melt rates you modelled compare with the melting rates derived from satellites (e.g. Davison et al., 2023). Is 13e closer to these than 13d?

Specific Comments:

- ll 7-9: This sentence should be slightly rephrased to clarify. You are not showing the deep offshore trough for the first time in ‘[y]our new bathymetry’, you are showing its continuation beneath the ice shelf, correct?

- l 31: IBCSO V2 also includes steering points beneath some ice shelves to mimic the continuation of onshore troughs (this is the case for the VB ice shelf).

- ll 52: ‘there is a desire to provide …’ is a bit vague, albeit true. It could be rephrased by giving some benefits of that approach instead (e.g., higher accessibility without software costs, improved reproducibility without the black-box approach).

- l 56: include reference for Tomofast-x here (Ogarko et al., 2024?).

- l 64: end with ref to sect. 6, to stay consistent.

- ll. 65 ff.: the last paragraph is not necessary here and could be dropped.

- l 357: ‘the the’

- l 385: do you mean Fig. S2 and Fig. S3?

- l 396: What’s the difference between ‘Vanderford Glacier ice shelf’ and ‘Vincennes Bay ice shelf’? If there is none, please stay consistent throughout the manuscript and only use one. If there is one, differentiate the two more clearly at the start (esp. in Fig. 3).

- l 420: how do you get to ‘±34–104m’ here? If you refer to the sentence before, it states a mean difference of ‘-34’ (not ‘±34’) and rmse of 95.  

- l 511: replace or scratch one of the two ‘accurate(ly)’.

- Figure 11: (c) should be (b). and (d) should be (c).

- Table 1: The model padding in the synthetic model is 10k in every direction. That reduces the model core by 20k Easting and Northing. Shouldn’t a model padding of 20k in the Vincennes Bay Model then reduce the model core by 40k? And not by 20k Easting and 10k Northing?