The paper presents a new seismic hazard model for Central Asia, released by an international cooperation between Italian experts and local experts, funded by European Union.

The study follows the state of the art in seismic hazard assessment with many data collected and updated during the elaboration of the model.

GENERAL COMMENTS

The number of citations is extremely small with respect to the argument. Seismic hazard assessment is a complex procedure that consider the previous experiences and the many data used for building the model; many affirmations made by the authors without an adequate reference must be referred to the authors themselves, but often this is not the case. In the following some examples.

The previous model for this area has been released more then 10 years ago (in 2012 the EMCA project released its model in the frame of the GEM activities). As the authors refer, after 2012 many studies have been realized in the single countries: in the period 2018-2021 all the country involved in this project were studied by several authors, in some cases with applications of the model to the building code. So, it’s not clear why a new study is necessary and what are the main criticalities of the previous works, if any. Please, add a comment on this issue.

NOTES ON THE MANUSCRIPT

On row 107, the authors comment that the hybrid approach allows a more realistic representation of the seismicity. In my opinion, in general this is true, but it depends on the practice of design large seismic source area due to the poor knowledge about the seismogenic processes.

Section 3 contains very few information about the definition of the source models. As far as I understood, the definition of seismic source zones (figure 1) is based only on seismic information. There is not any seismotectonic consideration. Is it correct?

Row 132: what does it means that this is the accepted version of the source model? From whom? Is it this information useful or necessary?

Also, the definition of the tectonic groups it’s not fully explained. When the authors write that the groups “are assumed to have comparable behavior…” on what basis their judgment is based? Only the earthquake catalog or other data? It is also missing any comment or comparison with previous source model.

About Section 3.2 (Deep seismicity zones), in figure 2 the position of the letters L and H are over the same zone. I understand that the two zones are overlapping, but from the caption I assume that the letter H refers to the wider area with the pale color.

At row 147, I suggest the use of the term “deeper” instead of “less”. Even if English is not my mother tongue, as written I understand the deep earthquakes occur at 20 or 30 km.

Section 4.2 (Occurrence rate model). The definition of seismicity rates is crucial in any seismic hazard model and object of many assumptions and operational choices by the modelers. In this field, it is normal to refer to analogue experiences. On the contrary, in this section there is only one reference about the Mmax estimation. I would like ask to the authors what is the approach adopted for the declustering; most used approaches (Gardner & Knopoff, 1974 or Reasemberg & Jones, 1985, among many others) lead to numbers of removed events very different.

The determination of b-value in two steps was adopted in many studies. Anyone to mention?

At row 184 it is reported a sentence that I have to dispute: “It should be additionally noted that the width of the non cumulative magnitude bins is not required to be uniform”. In my opinion, based on more than 30 years of expertise, It’s the first time that I read something like this. The bin width is a delicate point of the analysis, since it determines the b-value (Marzocchi et al., 2020; doi:10.1093/gji/ggz541). Even more so, the variable width is not acceptable. Let’s suppose that in the bin for magnitude 7 +- 0.5, all the events reported in the catalog have magnitude greater than 7: if you use 2 bins (with width 0.5) instead of 1 (with width 1), you will obtain 2 points with the same value in the cumulative curve, and this change the resulting fit. For me the assumption made by the authors it’s not acceptable.

Row 187: It’s true that most of rates models start at magnitude 4.5, mainly for completeness reasons. In some cases, we know damaging earthquakes for magnitude 4 or less (as an example in volcanic areas with very shallow hypocenters) Could you quote any papers that affirms what you are saying?

I understand that Mmax is based only on the information reported in the catalog, i.e., the maximum observed magnitude. Why was the maximum geological magnitude not considered? One example is contained in Woessner et al., 2015 (doi:10.1007/s10518-015-9795-1). Or do you think that magnitudes larger than the observed events are not possible?

Row 270: the smearing effect due to the adoption of seismic areas depends on the approach adopted to design the areas: smaller are the areas and more the hazard is concentrated on the epicentral areas. The design of areas should contain a sort of “prediction” for those zones with poor knowledge about the historical seismicity.

Figure 10: I wonder why the tectonic regions are different with respect to the groups of figure 1. As an example, source zone 5 in figure 1 has a different classification in figure 10 if I consider the other zones of group A. There is an explanation?

Row 371: what are the considerations that allow you to say that in stable continental crust zone an intermediate behavior between active shallow crust and stable crust is expected? I don’t say that it is not true, but I would like that you support this sentence with a reference or your comment.

Section 8. I don’t find any description in the manuscript about the 3 options for the assignation of b-value (b, b+0.5, b-0.5). With regard to Mmax, on contrary, in the manuscript I found only a sentence about the branch with Mmax+0.1. I think that the whole logic tree has to be described together with the strategies adopted for assigning weights.

Row 415: this is a clarification. The results of the calculation are only the hazard curves (not only in OpenQuake engine). Maps and UHS are possible representations!

Section 10. I don’t understand why the presentation of the model expressed in terms of macroseismic intensity is in this section and it’s not in ad hoc section.

Row 434: “Comparison with previous PSHA studies shows general agreement”. I don’t see the comparison! In the Supplement it is reported the map of GSHAP, released 25 years ago. Probably this is not the best test… In my opinion the comparison has to be performed with EMCA project (most recent study for the same area) or with recent national projects. Not only: I expect a quantitative comparison, not only a comparison of two figures.

Regarding the intensity maps, at row 464 you write: “All intensity maps are consistent with a shear wave reference velocity of 800m/s”. This is a strong statement and I ask you to cite a paper or discuss it. Most localities are built near rivers for access to water; this means soil conditions other than rocky ones.

In the conclusions, again, very few reference, but paper by Poggi et al.. When you talk of the strategies for assess seismic hazard at national scale, for example, you could quote Gerstenberger et al., 2020 (doi:10.1029/2019RG000653). For the international project, also, the references for CCRIF and ARC projects are missing.

To whom it may concern

The paper titled "Creation of a regional probabilistic seismic hazard model for Central Asia" has outlined the inputs and findings of probabilistic hazard analysis for the Central Asia region using the Cornell-McGuire classic method. The study focused extensively on the choice of seismic sources and historical seismic activity. The conducted analyses are deemed suitable, and this research is expected to hold significance in the region. The authors are appreciated for their work, and the following questions are posed to gain a clearer understanding of the subject matter.

Line 95: What does the word "size" mean? Refers to the dimensions, proportions, or magnitude.

Part 3: Is the division of the region into 7 groups and 61 regions based on the current work or previous works? If it has been done in this work, the method needs to be explained.

Line 146. How is the depth uncertainty included?

Line 151. H and L areas are not distinguishable in the map (Figure 2). In this case, as mentioned in the previous question, the discussion of depth uncertainty needs to be included in the analysis and text.

In Figure 3, the H and L zones are not consistent with the text. For example, is the depth of 150 the limit or the depth of 170?

Figure 4. The Gutenberg-Richter (in its logarithmic form) is a linear relationship; why is the fitted curve non-linear?

Line 170. Why the (one-side) truncated Gutenberg-Richter relation is used? Why the m_min is not included? In the rest of the text, contradictions can be seen in this field and the double truncated Gutenberg-Richter relationship is used.

Line 172. Is Gutenberg-Richter's relation applicable for values lower than completeness magnitude?

Line 177. Why is the list square method used? Is the data homogeneous? Considering the age of the countries in the region and the long history, are historical data included or not?

Line 187. The minimum magnitude of the Gutenberg-Richter of 4.5 does not match the one- side truncated Gutenberg-Richter relation. There seems to be a problem in the sentence " rate of earthquakes with magnitude greater than 0 "; because minimum magnitude of the Gutenberg Richter's has nothing to do with the potential of failure, but with the completeness magnitude. Therefore, the sentence seems to require revision.

In Table 2, the weights are assigned on what basis?

Table 3, what is the source (reference) of the table?

Line 304: Seismic coefficient = 0.1, why?

In the manuscript, the experts' opinion is repeatedly mentioned. What was the mechanism of use and criteria?

In Tables 5, 6 and figure 12, what is the method for assigning weights?

In Table 6, most of the weights are 0 and 1, and it doesn't seem that the two-step method (line 380) has much effect on the results.

How does Figure 11 help present the paper?

Thanks