The paper discusses the key issues of the Earthquake Risk Model of Switzerland (ERM-CH23). It is the result of huge work from the hazard, vulnerability and exposure sides. Authors summarize the main input and output of the study, including the comparison with other models. Such large research effort typically goes in research reports because a journal paper is often too short and also these projects implement state of the art research without actual scientific innovation. This is also the case of this manuscript in my opinion. Reproducibility is also impossible based on this work. Therefore I regret that I cannot recommend this manuscript for publication until it qualifies as a research paper. This notwithstanding, I appreciate the work and understand many of the authors are directly involved in the special issue this manuscript is submitted to. I also offer some comments as listed.

1. No equations are reported in the paper. I understand that the paper discusses the results of a huge work involving several expertise and alternative approaches (combined by the logic tree) but an analytical framework should be described.
2. The introduction section should briefly present the structure of the paper.
3. Figure 1 is never cited in the body of the text.
4. It is discussed that ERM-CH23 relies on 2015 Seismic Hazard Model (SUIhaz2015; Wiemer et al., 2016), and that some adjustments were made to it for use within the context of ERM-CH23 (lines 65-69). In turn, SUIhaz2015 relies on ESHM13, among others. This reviewer wonders whether one additional adjustment to SUIhaz2015 may be the replacing of ESHM 13 with ESHM20 (Danciu et al., 2021). Can authors comment on it?
5. Authors also discuss that ERM-CH23 is based on two main “sub-models”. One uses spectral accelerations, the other adopts macroseisimic intensity, and it seems that they are arranged in a logic tree framework. However, it is not clear, at least to this reviewer, how hazard results (hazard curves), in terms of different ground motion intensity measures, are finally combined.
6. What is the considered minimum magnitude? Lines 81-83, events of magnitude around 4 have been known to cause damage and casualties elsewhere, the justification for minimum magnitude 4.5 needs to be contextualized for Switzerland or justified quantitatively.
7. Line 84. But they are the most frequent earthquakes. Since the cited paper refers to PSHA, can the authors demonstrate that the neglected magnitudes do not contribute in terms of losses?
8. In the discussion of section 3.1.1, an important piece of information seems to be missing (or if it is implied, then it is unclear) and that is the weights assigned to the five branches. In fact, the first and last quantiles seem extreme enough to warrant justification if equal weights were adopted.
9. Section GG. Was spatial correlation motion of GMM residuals considered in the analyses?
10. Line 79. when later is used as an object to the sentence, the “t” is doubled.
11. Line 90. the reference Zhao et al. (2006) must be replaced by Zhao (2006).
12. Line 93. Wiemer is cited twice.
13. Table 1 lists, among others, two ECOS09 models, but they are not even mentioned in the paper (the not expert readers do not know that ECO09 is the Earthquake Catalogue of Switzerland). The other two models are calibrated based on data from metropolitan France (Baumont et al., 2018) and central Asia (Bindi et al., 2011). Can authors motivate such a choice?
14. The definitions of PGV and SA do not mention how the two horizontal components of motion are combined.
15. In figure 3, vulnerability curves expressed in terms of two different intensity measures (spectral acceleration at two different vibration periods) are presented in a single panel. This should be avoided, as it is typical for the reader to compare vulnerability by the relative shift of the curves along the abscissa (as the authors themselves do in section 3.2.4) and thus, grouping curves this way can be misleading. Also: in the caption spectral acceleration is denoted both as SA and Sa.
16. Between fragility functions and consequence models (3.2.2 and 3.2.3) some mention of damage states seems to be missing?
17. Line 240. The fact that earthquake recurrence is modelled via a Poisson stochastic process has nothing to do with whether or not “aleatory uncertainties” are considered in its modelling.
18. Line 244. This makes results not replicable. In other similar situations, an "average" branch able to approximate the results of the whole logic three was identified/constructed. Did the authors consider such opportunity? Moreover, it may be interesting to deepen the discussion of results describing the logic three branches that produce the lowest and the largest results.
19. Lines 250-251. Monte-Carlo simulation cannot be said to be “required” in this context. It is simply a matter of choice, and the authors would be well-advised to have the courtesy of explaining the motivation behind said choice, while being more careful in their phrasing.
20. In the opinion of this reviewer, section 3.4 should be moved after section 3.1.3, as it deals with the logic tree at the basis of the hazard assessment.
21. With reference to Figure 6 (showing the logic tree): what is the building mapping scheme? And what do RB and RF mean?
22. Figure 6 the SUIhaz15 logic tree branching is illegible and beyond the pdf file resolution.
23. Line 256. “probable maximum loss” lacks a definition (while it emphatically needs one).
24. Line 262. The term “epistemic distribution” is a neologism that should be avoided.
25. According to reviewer some additional information should be provided about bilinear capacity curve used in fragility assessment leaving to related papers the description of the numerical models.
26. Method used to derive fragility functions should be cited at least. Leaving it to the related paper makes the manuscript difficult to read.
27. How modeling uncertainties in fragility assessment have been accounted for in risk assessment? Paragraph 3.4 does not seem to treat this issue.
28. One would expect some more in depth discussion about the differences seen in Figure 7 between the SAM and MIM model risk estimates.
29. Caption to figure 9 needs some revising, should that be “epistemic variability”? The meaning of the sentence “if only…” is not clear. More in general, the figure is not clearly discussed in the text of the paper.
30. Line 359. some additional details about the tools should be provided.

This manuscript perfectly describes the aggregation of all current knowledge and methods for characterizing seismic hazard, vulnerability and risk, applied to Switzerland. This manuscript compiles several decades of work, from different groups (seismology, engineering seismology, earthquake engineering and even communications), for a single product as target: the seismic risk map of Switzerland. The authors are all experts in their field, and have contributed in one way or another to the definition of one of the strategies in their area of expertise followed in this study: definition of source zones, selection of GMMs and associated logic tree, characterization of site effects, definition of vulnerability curves and loss models. Each step references a series of peer-reviewed publications, giving scientific validity to the study. 

So why this article? 

On reading this manuscript, I fail to see the scientific contribution or the added value for publication in NHESS. Throughout the manuscript, reference is made at every step to Wiemer et al. 2023, and the question arises as to how this article compares with Wiemer's paper and what is the original contribution with respect to this 2023 paper?.

A few attempts at "testing" with recently published models show that, in the end, the estimate is neither right nor wrong, good nor false,… but rather overestimates one and underestimates the other, without it being clear why. This testing section could have been extended, which would have improved the novelty of this manuscript: by identifying the source of the differences, by discussing the need for a regional model versus a local model, by discussing some options taken (notably in ESRM20) or by extending the comparison between models to specific target areas for which the site conditions and the exposure model are (perfectly!) known (e.g. Valais or maybe the Basel urban area).

The issue on the contribution of each step in the final uncertainties could also have been developed. We've known for some time that the uncertainty linked to the hazard assessment is certainly the factor with the greatest impact: this message now seems self-evident, but has yet to be confirmed. In particular, this may be due to a lack of understanding of vulnerability and exposure: if we were to benefit from the same  amount of experimental data as those used for ground motion modelling (the number of which has increased enormously over the last 20 years), wouldn't we end up with a shared contribution of hazard and exposure to the final uncertainty? Or is the variability of ground motion intrinsically (non-epistemic) greater than that of structural response? This question, debated and simulated thanks to the logic trees defined herein, could have been tested, bringing a novelty to this study.

Finally, by compiling and aggregating previously published studies, much information is missing in order to follow the flow of the manuscript: GMM equations are missing, parameters characterizing site conditions are missing, descriptions of vulnerability models are missing etc... and in the end, the paper looks like a compilation of studies, not a scientific paper.

Some general comments:

Introduction lines 22-27: very classic, and in a special issue like this one, we're convincing the convinced. In my opinion, a little added value, such as the fact that natural hazard management programs save $3-4 on average for every $1 invested, goes a long way towards justifying the production of seismic risk maps like the one presented here. 

Line 52-55: " ... is considered moderate with three to four earthquakes a day.... " This sentence (and many others) is too vague, certainly because of the systematic reference to already published papers. Relying on previous studies should not compromise understanding of the article as a whole.

Line 114: " site condition indicators... " which ones? 

Line 161: "...collected from other sources..." which ones?

Figure 1 is not cited in the body of the manuscript

Line 162: please specific the macroeconomic model. Do you think that structural/nonstructural damage-to-loss ratio from US can be transferred to Switzerland? In particular, this ratio could have been explored in greater depth with a few tests, so as to add something new to the study.

All in all, I think the paper could be eventually accepted if the issues set out above were strengthened and each step specified more precisely.