|This manuscript describes a downscaling exercise of three major Alpine flooding events in three different centuries. The main findings of the study, in my opinion, are (i) the major role of the relationship between cyclone location and moisture fluxes towards the Alpine slope in determining the regional precipitation patterns and (ii) the insight that a pre-selection of ensemble members according to these large-scale forcing factors (cyclone position, moisture flux) may be useful prior to dynamical downscaling. While point (i) has already been made in previous studies, the new aspect here is that its relevance is demonstrated consistently through a downscaling chain.|
In this revised version, the authors have put a lot of effort in incorporated comments and suggestions by three reviewers. Note that I have not been part of the first review round and have thus looked at the manuscript for the first time after these revisions. Possibly also related to this later involvement, my opinion on the manuscript is somewhat different from the previous assessments. While I think that the findings summarized above are clearly important and worth publishing, and the paper is technically and methodologically sound, the way the manuscript is put together, in my view, is not very appropriate. My main criticism is related to the fact that the paper is full of technical details and it is very difficult to extract the main messages and conclusions. As it is often the case, the number of details has even been increased by incorporating the comments from the first review round. I'd be really afraid that many readers will give up during the first technical part before being able to appreciate the interesting results described towards the end of the paper. I’d thus invite the authors to take a step back and think again which of their results provide useful insights for other researchers in the fields and which steps, on the other hand, may have been technically important during the preparation of this study, but do not necessarily have to be described in full detail in the paper. Here a few more concrete comments that might be helpful in this respect:
- The entire section 3 reads more like a work report than a scientific paper (the structure, in short, is “we started with a specific approach, than tried something else that turned out to be better, and finally tested a few more sensitivities”). For a paper, such a “chronological” order is often not the best choice, and I’m really not sure what a reader should learn from this lengthy description (as you only analyzed a few cases, your specific results, e.g., on the choice of parameterizations can hardly be generalized anyway). I’d suggest to start with the description of the main setup that was used as a basis for the process analysis in section 4 (lead time of 1 day) and afterwards briefly (!) summarize the sensitivities w.r.t. lead time and parameterizations (removing most of the details, e.g., provided in tables 1 and 2, or moving them to the appendix).
- Much related to the previous point: I’m really not surprised by the fact that shorter lead times lead to better results. This first-order relevance of the initial conditions is very well known from weather prediction. Compared to this, soil moisture conditions are clearly of secondary importance, and in addition it may take much longer than 10 days for the soil moisture to equilibrate if it is not properly initialized. In this sense, the author’s first approach has been somewhat naive, and I do not think that it is necessary to dedicate a lot of space to describing the fact that shorter lead time yield better results.
- The description of sensitivities to the parameterization schemes is spread over the manuscript (e.g., end of section 2, section 3.2), which should be avoided.
- The potential pre-selection of ensemble members based on their representation of the relevant large-scale conditions (my point (ii) in the first paragraph) is only mentioned at the very end of the paper, where it might be overseen. I think that this is an important general conclusion from the paper that should be highlighted more specifically. I would add a note on this at the end of the abstract and also formulate a corresponding research question/hypothesis in the introduction section to prepare the reader for this finding.
- I do not find it very helpful to add the specific case study at the end of section 4. This again shifts the focus to the details (“On 15 June 1910, 8 UTC…”) and away from the more general conclusions. If you really want to keep this, I would move it to the beginning of the results section to set the scene for the following, more quantitative analysis.
Other minor/technical comments (page and line numbers refer to the version with marked changes in the response letter):
P 1, l 22-23: The sentence “The analyses show…” is difficult to understand at this point because it is not generally clear what “cyclone fields” are. I would also make clear that “deviations” in the second part refer to deviations between ensemble members (and not with respect to some ground truth).
P 2, l 6: “For this reason” (add reason); bracket missing
P 2 , l 14, “transported” or “advected” instead of “led”
P 2, l 36 – P 3, l 5: This paragraph is rather technical for an introduction section, and it left me with more questions than answers (which approach did you use? which is more suitable?)
P 3, l 12: “losing control over” is too unspecific; the information from the initial conditions is lost
P 3, l 14-17: I’d avoid the term “cloud-resolving” (although it has been previously used in the literature), as a parameterization of cloud microphysics is still required. In this context, also “explicit production of precipitation” is imprecise.
P 4, l 20: I don’t understand the term “1st-step deviations”
P 5, l 16: “parameterizing” instead of “resolving”
Section 4.1: You put some effort in identifying cyclones on different vertical levels, but it is not entirely clear to me what you learn from this in the end. Can you formulate a conclusion that could not have been obtained from detecting cyclones on one level only?
P 16, l 9-12: I cannot follow this. Doesn’t the interpolation from the coarse input data apply a height correction? How do warm temperatures “produce” water vapor?
P 24, l 21: “parameterizations are turned off”: no, only the convection scheme
P 25, l 18-20: It is impossible to say if the “patterns and dynamics” are reproduced by some ensemble members also for the 1876 case, as there is no ground truth for comparison. From the ensemble spread, you can only learn that the uncertainties are larger because of missing observational constraints.
Figures: Most figures contain a lot of different information, which reduces their accessibility. Please check all figures once again and remove the content that is not absolutely necessary to demonstrate the main findings (for instance, showing results from an ERA-Interim driven experiment in Fig. 5 is again very much on the technical side; I’m also not sure why it should be helpful to mention this experiment in the main text on page 16).