|Review of La Palma paper by Stéphane Abadie and co-workers|
This paper is an elaborate study of the effect of three scenarios due to slope failure from La Palma Island on tsunami impact in French territories. It reports large amount of analysis and work, and there has been surely large efforts behind producing these outputs, in particular given that some of the tools employed such as the CFD THETHIS code are demanding to operate for such purposes. It is an important study because of the practical implications. On the other hand, the elaboration is also a drawback of the paper. There are many different models used, to illuminate different types of physics, merged with an attempt to make an impact assessment (the authors uses the term hazard). Moreover, the paper seems to have undergone several previous reviews with large changes, and would benefit from a better organisation. Some related general comments are summarised briefly below, followed by a long list of line-by-line comments. These comments must be taken into account in a possible revision of the manuscript.
It is not clear why mane different models are used for various purposes. I would have liked a simpler strategy where the authors choose a simple set of models. The physics is well known: the tsunamis are dispersive, and we need nonlinear shallow water models for the inundation. Right now, there is a patchwork of models, even including analytical solutions (which I suggest to remove), and it is hard to understand why a given model is used where. While I would suggest that this is much simplified, I would probably expect that the authors would like to keep as much as possible of these results. Hence, as a minimum, a much tightened up introduction is needed to much better explain the scope and how the different models are used, and why. I would also suggest to better distinguish impact studies and studies of physical effects (e.g. dispersion, model comparisons).
Another major issue, in particular when reading the introduction, is that you sense that the hazard study is attempting to make a best estimate of a landslide motion and wave generation based on laboratory glass bead experiments. However, nature will not behave this way, and there is a considerable uncertainty related to the process and the sliding material. Granted, one cannot perhaps expect that the computations can cover all these uncertainties, but as a minimum, the authors must make it crystal clear that there can be a much larger variability related to the tsunami generation and tsunamigenic strength. This is a limitation of the study.
Finally, the title tsunami hazard is misleading, because the authors do not address return periods, in addition to lacking a proper treatment of the variability or sensitivity to landslide parameters as noted above. The title should hence be revised to take this into account.
Title: Probably the term "tsunamigenic strength from potential events" is better than hazard. After all, hazard refer to a temporal component, and should not really be used if return periods are not considered.
Page 1 line 5: "for 5 minutes" --> "after 5 minutes".
Page 2 line 8: "allow studying impact on France and Guadeloupe". Here you maybe emphasise more strongly that this is the scope? After all, the impact locally would be a more natural focus.
Page 1 - line 8: "Although the wave source seems to be reduced due to the rheology..." --> "Although the rheology applied in this study seemingly leads to smaller waves..."
Page 1 - line 9: add "mu(I)" ahead of rheology
Page 2 - line 7: It the term hazard is used properly, it would be useful to introduce a definition, and refer to at least one key paper. Use e.g. Grezio et al. (2017): Grezio, A., Babeyko, A., Baptista, M. A., Behrens, J., Costa, A., Davies, G., ... & Harbitz, C. B. (2017). Probabilistic tsunami hazard analysis: Multiple sources and global applications. Reviews of Geophysics, 55(4), 1158-1198.
Page 2 - line 13: On the complexity of these processes, please refer key review papers, Løvholt et al. (2015), Yavari-Ramshe and Ataie-Ashtiani (2016): Løvholt, F., Pedersen, G., Harbitz, C. B., Glimsdal, S., & Kim, J. (2015). On the characteristics of landslide tsunamis. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 373(2053), 20140376; Yavari-Ramshe, S., & Ataie-Ashtiani, B. (2016). Numerical modeling of subaerial and submarine landslide-generated tsunami waves—recent advances and future challenges. Landslides, 13(6), 1325-1368.
Page 3 line 9: Please update this sentence to say that you use mu(I). I think this is clearer than saying calibrated slide viscosity.
Page 3 - line 29: "first instance of the motion" --> "initial motion".
Page 3 - line 30: "solver code category" --> "type of solver".
Page 3 - line 33: "close but not completely equivalent to models, also use to simulate landslide tsunami generation" --> "more sophisticated with respect to the slide motion than models such as"
Page 4 - line 9: For a complete review, discuss also the model of Si et al 2018:
This model is more sophisticated material wise, but probably not able to tackle operational environments yet: Si, P., Shi, H., & Yu, X. (2018). Development of a mathematical model for submarine granular flows. Physics of Fluids, 30(8), 083302.
Page 4 line 22: Clarify where Newtonian and mu(I) rheologies are used, maybe reformulate: "Both Newtonian and mu(I) rheologies are used in the simulations".
Page 4 - line 23: The experimental results cannot necessarily represent the real case realistically (glass beads are far from a realistic rock slope material). Hence, all the different viscosities may represent the reality, and should not be calibrated towards a single dataset. This is actually a misconception, the hazard analysis should ideally include this as an uncertainty. Hence please reformulate.
Page 5 - first paragraph. Please see above comment. I dont believe a single calibrated result represent the reality realistically. This does not mean that new simulations should be done, but the authors should make the reader aware of this uncertainty.
Page 5 - line 15: Please explain that this is just possible value for the material parameter, and there is likely a rather large uncertainty that is not covered in our analysis. Otherwise, the reader gets the false impression that the wave generation is deterministic, which it is'nt.
Page 7 - line 3: Please clarify "can be upgraded"? Do you mean that it also contain dispersive features. In this case reformulate. It is BTW not clear why two types of dispersive models are used. Does this code have wetting and drying facilities?
Page 7 - line 25: Again, why is this model used? It is not clear why so many seemingly similar models are used, please elaborate.
Page 7 - line 34: "In this work..." do you refer to Telemac? The meaning is not clear.
Page 9 - line 17: This is not a proper hazard assessment. Impact analysis or scenario analysis are better terms.
Page 9 - line 22: Again, I miss the reasoning for choosing this model, and why other models are employed elsewhere. This is generally quite messy. You need a structured introduction upfront in the paper explaining these choices.
Page 9 - line 31: Again, this is not hazard, probably something else but not hazard... Please revise sentence.
Page 11 - line 4: Delete double "smaller"
Page 11 - new paragraph marked red: Not clear what this paragraph add, it is confusing. We have repeatedly shown the effect of dispersion in previous studies. I dont see the need for doing this again, it disrupts the text.
Page 11 line 31: This was analysed in more detail first by Løvholt et al. (2008), please notify and provide reference.
Page 13 - line 27: Delete double punctuation.
Page 13 - line 28: Again, this is not hazard assessment, but only an assessment of possible inundation or impact. Please revise title.
Page 13 - first three paragraphs of section 3.5: I find all this analytical analysis strange for a phenomena so strongly controlled by local phenomena. Why not limit the impact analysis to the local inundation study. I would suggest to skip this part, and only keep the part using NSW inundation analysis. The paper is overloaded with results, and this is for me a distraction. Moreover, such a rough analytical analysis could be worthwhile for assessing the hazard region, but not for a local analysis.
Page 15 - line 5: As said above, the authors does not seem to take into account that the dynamics and material behavior is uncertain, and that a simple glass bead experiment cannot be conveyed to real situation. The paragraph should be rewritten to better reflect this. Granted, the simulations fit better the experiments, but the authors have no guarantee that the slope failure will behave this way. Probably it will not.
Page 15 - line 15: Again, please replace the term hazard assessment with something more appropriate, such as an impact assessment. The study is not broad enough and does not cover return periods, so cannot be coined a hazard study.
Page 15 - line 31: This discussion of model effects is too long. I would suggest to shorten it dramatically, as results are shown above and the physics is well-known. Besides, the effects of dispersion have been investigated in previous studies. It can also be analysed with a dispersion number (e.g. Glimsdal et al., 2013)
Page 16 - line 23: Wynn and Masson found upward fining, which indicate long separations in time. This means that this was no real retrogression, but more likely separate events. On the other hand, I agree with the authors statement in the last part of this paragraph.
Page 17 - line 13: See comment above several times on uncertainty, and reformulate accordingly.
Page 17 - line 20: This sentence is not well formulated, I dont fully understand what you mean.
Figure 8: Slide contours are very difficult to read. I suggest fewer and larger figures allowing the reader to see the details.