The manuscript ‘Present-day mass loss rates are a precursor for West Antarctic Ice Sheet collapse’ by T. van den Akker and co-authors presents a new method to improve the initialization of ice flow models that better captures their current trend and should improve the reliability of projections of future ice sheet evolution. This is a very nice improvement and the method should be applicable to a range of models, therefore having a large impact on this community. The results, based on an ensemble of simulations and the use of two different ice flow models, show that the current mass loss observed in the Amundsen Sea sector is a precursor to large and rapid changes in the future.

I find the idea presented interesting and simple to implement, therefore making it a very nice improvement to other models. However, I find the manuscript confusing at times, some reorganization and changes in the figures/tables would help a lot. I also think the conclusions are not always supported but the results and the manuscript should be clarified to better integrate both models and ensure a direct comparison of the results.

Major points:

The current mass loss in West Antarctica is said to be a precursor to large future changes, but there is no direct comparison or quantitative analysis of what would happen otherwise (without the current rate of change) and how it pre-determines the future changes. This should be better justified and quantified, as demonstrated by model results, or this conclusion should be removed from the manuscript.

Another point is that this new method is used by two different ice flow models, which is great to show its easy implementation and possible impact on a range of models. However, the results from the two models are barely compared and used in a different way. It is therefore difficult to assess if similar results are achieved: for example, showing how spatial mass loss are very different without this method but very similar after including it in both models would show the similarity. In order to quantify the impact of these changes and the improvements made, it would be important to compare the models more directly and systematically, both for the initialization and for the sensitivity experiments.

It is also not clear which experiments have been performed by which model and why such choices were made, and why only the results from CISM are shown most of the time, with barely any result from UFEMISM. This should be clarified to help the readers follow easily the results and conclusions. Similarly, the manuscript is confusing at times and could be made more straightforward (see detailed comments below).

Technical comments:

- l.20: it is not clear how the current mass loss is a precursor to future large changes, the mechanism or reason causing this should be mentioned

- l.28: MISI is not really uncertain, it is a mechanism reproduced in a number of models and analyzed in depth by many theoretical papers. Its likelihood however is uncertain. This should be rephrased.

- l.29: ‘unbuttressed outlet glaciers’ I though this was applicable to both buttressed and unbuttressed glaciers, even if the rates of fluxes are impacting by the buttressing

- it remains unclear whether or not these glaciers are already engaged in an unstable retreat, recent studies suggest it might not have reached such a point yet (Hill, Urruty et al., TC, 2023; Reese, Garbe et al., TC, 2023)

- l.37: higher than what?

- l.38: ‘a model study’ -> ‘model studies’ (several studies cited)

- l.50-55: another study based on a model ensemble and investigating the marine ice sheet instability mechanism worth discussing here is Robel et al. (2019)

- l. 60-65: it would be good to also add the uncertainty study by Seroussi et al. (2023)

- l.67: maybe ‘to iterate toward a specific state’ -> ‘capture conditions at a given time

- l.72-73: the problem is rather matching a single time vs a period of change (Goldberg et al., 2015)

- l.81: mention that this initialization is done by running very long spin-ups

- l.84: How does that impact the initialization?

- l.88: Explain the DA biases, which ones and why?

- l.90-100: it is not clear when each model is used for the different experiments, when only one model and when both are used? Also why are experiments done with one or both models? I don’t know if this is the best place but it should be clarified.

- l.110: maybe this should go in the UFEMISM section

- l.133: explain what you mean by `the size of each term’

- l.134: ‘linear interpolation’

- l.135: explain why you use these values and how you chose them

- l.146: why do you use p = 0.5? How is that constrained or calibrated?

- l.155: is this the same \tau or a different o? What is the impact since it is a different physics and orders of magnitude?

- l.169: How is that scaled? Is it done linearly, or bilinearly or something else?

- l.171: ‘including a few applications’ … ‘ e.g., Yu et al.. (2019)’

- l.172: I think it is the same problem for Greenland? Or if not you could explain why it is  different between Greenland and Antarctica.

- l.176: why was 1 m chosen? What would be the impact of choosing a higher threshold?

- l.181: it would be easier if the main experiment had a name or something easy to refer to

- l.184: the sliding is also different

- l.185: ‘targeted high resolution’: what does that mean?

- l.197: what else is different between CESM and UFEMISM?

- l.207: ‘with observations like the result of the equilibrium’ -> ‘with observations, similar to the result of the equilibrium’

- l.210-215: this explication is a bit confusing, since this is a key element of the method, you should add a schematic equation to explain it more clearly

- l.226: ‘ would make the ice sheet theoretically more stable’: why is it more stable? In your set-up or in the experiments? If you want to show that this method provides more mass loss, what is the best way to demonstrate that? Adding something like that would go a long way.

- l.235: is the ‘default evolution’ similar to a case with constant climate conditions?

- l.237: in which case is it negligible?

- l.246: which parameterization is that? Add a reference or equation to explain it

- Overall I am confused about the forward simulations. The abstract mentioned projections, but I only see the constant climate conditions with and without the mass correction or something like that. I would be important to clarify and describe these experiments with more details.

- Fig.1: the model used to produce these results should be mentioned at the beginning of the caption. Remind that e and f are observations and which ones were used to calibrate the basal melt for the model.

- l.287: one important method is the difference between the two initialization methods. It should be shown more and not just in the supplementary material. It should also be show for the two models to compare the impact in both cases and how both methods display similar trends or not with the new initialization

- l.287: how low is the bias? Add numbers

- l.290: same here, how well does it agree? Add numbers

- l.294: are you refereeing to the models from the ISMIP6 ensemble using data assimilation or spin-up?

- l.308: how do these other corrections compare?

- Fig.2: why is the grounding line in the Siple Coast and most of Filchner-Ronne not changing at all? How does that compare to previous results?

- l.340: ‘sea level change equivalent’

- l.333: what is so special about this ridge? There is another one at about 100 km that is just as high?

- l.353: ‘a large ice shelf has formed instead’: how stable is this ice shelf? How thick/thin is it? I don’t remember seeing a mention of what is done for the calving for both models and I am wondering how this result would be impacted by the choice of calving?

- Fig.3: Again, why just show results from CISM here? If the idea is to look at the similarities between the two models and the possible timeline of collapse, comparing results from the two models would go a long way. It would be interesting to see if the new initialization method allows to better reconcile results from the two models.

- l.370: are there figures showing these results?

- l.376: again here, how is that impacting by the choice of calving?

- l.384: Which former case is discussed here? The previous sentence discussed melting caused in part due to the variability and in part to anthropogenic forcing, so it is a little unclear what former refers to.

- l.395: Why is it incorrect? It is not clear if this is just a different evolution and how the “correctness” can be established. What are the conditions or criteria to decide whether or not this is correct.

- l.396: “set of model choices”: which choices are being considered? There should be a clear list of choices and parameters.

- l.410: “weaker” and “stonger” should be discussed and quantified. What parameters are changed?

- l.414: by how much is it delaying the collapse?

- l.415: where are these results shown?

- l.421: in what sense is “linear” used here? Is it the same rate of retreat? Or evolution of grounded/floating areas? And which parameters impact it?

- Fig.4: why are only 2 results from UFEMISM shown on this figure? Also, would there be a way to organize the runs by colors and symbols according to parameters changed or something more intuitive? It is a bit difficult to read in this format. The legend mentions some broad categories but is not very clear (e.g. what does forcing refer to since there is also a basal melt category, which is a big part of the forcing).

- l.486: what about the availability of the code to reproduce the results presented and reproduce the figures presented in this manuscript?

- Supplement movie: Clarify time in supplement movie (e.g., 10001.0, not clear what this time corresponds to)

Please see attached PDF.