Dear authors,

I thank you for your revisions and in particular your more in-depth analysis of the statistical soundness of your method. I have read all reviews, the interactive discussion and your new manuscript.

Overall, I see merit in comparing a geology-based approach of mapped fault lengths with geophysical observations from earthquake catalogues and trying to interpret the difference on a national-scale. However, such a comparison should be accompanied by careful enough wording and appropriate acknowledgement of the simplifications involved to be able to benefit science and society. In general I think your limited objectives and methods are described clearly in the paper. However, I still think that your formulations and some limitation descriptions - in particular with respect to anything relating to seismic hazard - should be improved and clarified at important locations in the text to not make suggestion beyond what your simplified method can do.

I articulate the important points in my comments below. I recommend to accurately and concisely include or address all comments.

Kind regards,

Ylona van Dinther


Comments:

You respond that your goal is not seismic hazard (response 1, reviewer 1), yet your abstract concludes with “Our work can provide a perspective time-independent seismic potential of faults” (l.16-18, p.1), which is a key ingredient of seismic hazard assessment. You claim - and your simplified method attest - that major implications for seismic hazard for Italy are not justified. I agree with that. So why are your last claims in abstract and conclusions about seismic potential relevant for seismic hazard assessment rather than about testing the usage of easily accessible fault databases on a national scale. I think such key pieces of formulation should be updated to reflect your goal better. I would suggest to not try to suggest implications beyond what the quality of your tests allows and remember the societal and political sensitivity such statements carry. Such worries are clearly stated by Valensise et al and reviewer 1, so please ensure that you remain concise and accurate about what your very simplified method can do and what it can not, particularly in the conclusions and implications you draw.

You write that you “benchmark” scaling relationships against catalogued magnitudes (l. 3, p.1), but I do not think you reach such a high level related to a scientific interpretation of that word. I rather think wording like “compare” is more appropriate.

p. 6, l. 11: You write you consider “all known faults”, while it is very clear from the long list of references from reviewer 1 and SC1-Valensise (p. C4-C5) that this is not the case. I think all descriptions of your fault database should be updated to reflect the actual situation accurately, while acknowledging other approaches and literature. For clarity, I do not think your database needs to be updated as you need to make assumptions to generate a database on some grounds. However, I think your text should be distinctly updated to acknowledge that you miss or miss-represent a distinct number of otherwise known faults. The worries of these experts should not be dismissed, but rather acknowledged appropriately.
From that same perspective I think the worries expressed by Valensise as summarized in the centre paragraph on C5 (SC1) should be better articulated in the paper. For example, for assumptions 3 and 4 in section 4.2.

Reviewer 1, Comment 6
I think it is fair to answer to the critique of missing faults with careful examples given by the reviewer. I understand your simplified method resolves the double counting aspect, but still there are two examples well total length may be underestimated. Addressing this limitation in the paper also gives the reader a better perception of the completeness of the databases you use. Here just saying they are as complete as possible is not really meaningful. Rather write exactly what you do (as you do), but than acknowledge potential though understandable flaws properly.

Your analysis does not account for rupture jumps that could happen from one fault to the next, as suggested by the “segmentation” examples of e.g. 2016 M7.8 Kaikoura earthquake. That would mean that earthquake with larger magnitudes than you estimated could occur. This is quite a likely scenario with major implications if that is too happen. However, you describe this limitation in only line (p. 11, l. 4) without any scientific references or appreciation of the process or consequences. I think that should be included.

I also very much appreciate you openly sharing the updated Italian fault database. However, I am also not able to find and download the fault database from the link given in the manuscript and rebuttal letter. Could you please ensure a long lasting and secured access?
http://pmd.gfzpotsdam.
de/panmetaworks/review/924b171fd21c78f295d58a7e9e321e8ad07667ab6201634b23d3
cb5a3f170d10/

I agree with Reviewer 2 (comment 10) and Valensise et al (comment 13) that using log-scale quantities such as magnitude introduces a bias and potential for misinterpretation of the size difference (see values in comment 13).
You agree with this comment, yet I do not see the suggested changes in the updated manuscript (l. 11, p. 13). I do not see exact numbers.
However, I find it more important that major differences often of 2 magnitude values (Fig. 10) are described as “having a good agreement …” (p. l. ), such that “this gives credibility to the used scaling relations” (p.1, l.16) or that these “can be benchmarked” (p. 1, l. 5). Based on the information in your abstract a 2 standard deviation range includes a magnitude range of 2*1.47 is almost 3 magnitude values, which require very large fault length differences. I think this rather shows large discrepancies, which could also be used to argue that something opposed. I have the impression you take care of this uncertainty by in the discussion saying that they show “either good agreement or some differences are observed”. I do not think that is very accurate. Could you please improve your formulation in such conclusions, abstracts and in the results?
I also like the suggestions to you seismic moment instead, which can indeed not be calculate directly for all earthquakes, but scaling relations similar to those used also exist to portray seismic moment. Doing that would better show the variation in values and would lead to less bias.

In response to Valensise (SC1, p. C7-C9 & comment 16) I think you should also stress more the low probability of occurrence for such large earthquakes. Common people that do not in detail read the whole paper should readily understand this low probability to prevent confusion and potential follow-up mis communication at all levels. The emphasis on such information should thus be done in the main text at multiple key locations and by extending the disclaimer with more information than this being for scientific purposes only (e.g., the missing component of probability of occurrence and risk that is missing in all such figures). Considering the sensitivity of this topic for society I think such additions are useful and required.

Your new section structure makes it hard for the reader to understand your flow and appreciate what you did. Section 5 describes “Results and discussion” all together in one section. Section 6 then describes a “Statistical test of FLEM values”, which is again an analysis of the results. Could you update your structure and section headings a more standard scientific paper setup to facilitate the reader in finding the information needed?

As a follow-up on comment 17 by Valensise:
I agree lengths in Fig. 7 and 8 seem to represent the required fault length and there are no isolated cells with too large a magnitude
However, in Fig. 9 for the observations I do see quite many of such isolated cells, as I guess these values have ended up only in the cell of the hypocenter. This means that when assessing the difference in Fig. 10,11, and 12 you are also including this missing information and are thus underestimating Mcatalogue and overestimating deltaM.

- Your sentences can be quite long (e.g. p. 14 l.2-7). I recommend another look at the sentence length and language of the manuscript.

- Occasionally (p. 16, l. 15; p.1, l. 14) you still use scaling law instead of scaling relation.

Dear Andrea Billi,

Thank you for in most cases appropriately addressing the last comments. I found three occasions where your response was missing or not accurate, which you can best find in the attached PDF. I hence suggest technical corrections that need to address three points. Upon including these I do not need to check your manuscript again and it can proceed to the publication process.

Best regards,

Ylona

From that same perspective I think the worries expressed by Valensise as summarized in the centre paragraph on C5 (SC1) should be better articulated in the paper. For example, for assumptions 3 and 4 in section 4.2.
>> Not addressed (see start p. 2)

As a follow-up on comment 17 by Valensise:
I agree lengths in Fig. 7 and 8 seem to represent the required fault length and there are no isolated cells with too large a magnitude
However, in Fig. 9 for the observations I do see quite many of such isolated cells, as I guess these values have ended up only in the cell of the hypocenter. This means that when assessing the difference in Fig. 10,11, and 12 you are also including this missing information and are thus underestimating Mcatalogue and overestimating deltaM.
R: Dear Editor, we have re-checked Figs. 7 and 8 and we do not see isolated cells characterized by high FLEM values (and therefore by long faults). In particular, we refer to red cells that are never isolated in Figs. 7 and 8. Please, note that cells with lower FLEM values (orange-yellow to blue cells) are characterized by shorter faults and hence can occur also as isolated. In other words, isolated orange-yellow to blue cells are compatible with the used method whereas isolated red cells would be incompatible (they are not isolated indeed) with our method as pointed out by Valensise et alii. We are obviously available for further improvements and clarifications.
>> You have not addressed the key point, which is the second paragraph that refers to Fig. 9, not 7 or 8. I already wrote that I agree with your assessment of Fig. 7 and 8.

- Occasionally (p. 16, l. 15; p.1, l. 14) you still use scaling law instead of scaling relation.
R: There are no “scaling laws” in our manuscript. We have found three “scaling equations” that have been changed into “scaling relationships” (Lines 16 P. 10; 22 P. 10; 28 P. 10).
>> A simple search does show them at key locations in the manuscript. For example, see p.1, l.1 qand p. 16, l. 25.

Dear Authors,
based on the Topical Editor's report and overall assessment of the review process, your revised manuscript can be now accepted for publication in Solid Earth, pending the amendments outlined by the Topical Editor.

Many thanks for choosing Solid Earth as a platform for publishing your research

Sincerely,
federico rossetti