Dear editor and authors,

I reviewed your manuscript "Implementation of an interconnected fault system in PSHA, example on the Levant fault".

The article aims at improving PSHA for the Levant fault system including different scenarios about possible multiple activations of different faults by introducing connectivity between different seismogenic sources.

The manuscript is well written, clear and of wide and specific scientific interestand definitely deserves publication in NHESS after minor review. I reported my specific comments, suggestions and (minor) requests on the attached pdf for the sake of simplicity.

Thank you for considering my review

Review of the paper "Implementation of an interconnected fault system in PSHA, example on the Levant fault" by

Sarah El Kadri, Celine Beauval, Marlene Brax, and Yann Klinger

The manuscript presents an analysis of the Levant Fault System (LFS), a 1200 km-long left-lateral strike-slip fault, with the aim of improving regional Probabilistic Seismic Hazard Assessment (PSHA) models. By considering the interconnected nature of the LFS, the study challenges traditional approaches that treat faults as isolated segments. Building upon El Kadri et al. (2023), who developed a seismic hazard model for Lebanon using classical fault segmentation and a moment-balanced recurrence model, this research explores how fault connectivity affects source modeling and hazard estimates. It examines various rupture scenarios, both single and multi-segment, using the SHERIFS and OpenQuake Engine software.

The study provides valuable insights, but the structure of the manuscript could be improved for better clarity and coherence. One key concern is the inconsistency in the treatment of maximum magnitude (Mmax). While the study initially assumes Mmax = 7.5, different values are later tested without a clear rationale. A more structured approach would be to incorporate multiple Mmax values from the outset, considering both exponential and characteristic models within the interconnected fault system. This would enable a clearer comparison with observed seismicity rates and better illustrate model and parameter uncertainties, which are central to the study’s objectives.

Additionally, some assumptions are presented without sufficient justification, which could lead to confusion. Providing clearer explanations would enhance the readability and impact of the manuscript. A major revision is recommended to improve the coherence, transparency, and overall clarity of the study.

• The segmentation of the fault system into 43 sections provides valuable insight into its complexity. However, it would be helpful to clarify how this segmentation process influences the final results. Specifically, how do variations in segment length (min 4km max 40km) and number affect earthquake rate estimates and hazard assessment?
• The hypothesis that strike-slip fault segments are activated together with thrust fault sections (GF, MF, MLT) requires further clarification. Given the differences in kinematics and orientation between these structures, on what basis is this hypothesis supported? Are there geological, paleoseismological, or Coulomb stress analysis results that justify this assumption?
• The manuscript considers earthquakes up to Mw 8.1, despite paleoseismic records not exceeding magnitude 7.5. Could the authors justify the use of models that incorporate larger magnitudes?
• The study states that the interconnected fault model leads to increased hazard along secondary faults and lower hazard along the main strand compared to the classical implementation. However, variations in hazard levels are influenced not only by segmentation and connectivity assumptions but also by the annual probability of exceedance and return periods of individual ruptures, which vary with the maximum magnitude considered. It would be useful to clarify how these factors influence the observed differences in hazard estimates between the classical and interconnected models.
• For a maximum magnitude of approximately 7.9, the study states that the LFS fault system must be fully released. At a 475-year return period, PGA values reach around 0.3g within 20 km of the faults, with peak values of approximately 0.5g along specific sections. However, these relatively low PGA values appear inconsistent with an M7.9 earthquake, as recent smaller earthquakes have recorded PGA values close to 1.0g at similar distances. Could the authors clarify this discrepancy? Additionally, would it be beneficial to present hazard maps with a 2% probability of exceedance in 50 years to better capture the strong shaking potential of rare M7.9 events?
• Lines 63-67 discuss how recent earthquakes have demonstrated that ruptures can propagate across geometrical discontinuities, leading to larger-than-expected magnitudes. It may be helpful to include the 2023 Mw 7.8 and Mw 7.5 Kahramanmaraş, Turkey earthquakes, which ruptured a complex fault system just north of the study area. These events further support the need for an interconnected fault model in PSHA.
• The discussion on the Grand Inversion in UCERF and the SHERIFS algorithm is useful, but a brief comparison of their key differences would improve clarity. The explanation in lines 84-88, describing how SHERIFS constructs rupture scenarios, assigns occurrence rates, and distributes seismic moment, could also be better structured and clarified.
• Lines 161-163 introduce the term "seismic gaps" without a clear definition. Are the authors referring to a lack of recent seismicity, barriers to rupture propagation, or zones with unknown fault connectivity? Clarifying this would improve the interpretation of fault segmentation and its impact on hazard assessment. Similarly, in lines 200-202, the manuscript states that if section lengths are too heterogeneous, the algorithm subdivides longer sections to homogenize segment lengths. What criteria define heterogeneity? Is there a maximum allowable section length or a statistical or geological basis for these subdivisions? Furthermore, does this segmentation process impact rupture connectivity assumptions and earthquake rate estimates?
• Lines 205-208 suggest that large magnitudes are picked more frequently than smaller magnitudes, which contradicts the Gutenberg-Richter model, where lower-magnitude earthquakes should be more frequent. Could the authors clarify how magnitudes are sampled? Is there a weighting applied to moment rates that modifies the expected earthquake rate distribution?
• The slip rate increment is described, but its definition and application remain unclear. Is it sampled from a probability density function or applied in fixed steps? How does it relate to long-term slip rate constraints? Lines 237-240 also appear contradictory, as they state that the Gutenberg-Richter model is used while indicating that large magnitudes are picked more frequently. Since the maximum magnitude assumption (Mmax = 7.5) strongly influences hazard estimates, how does this assumption impact the b-value and the overall shape of the magnitude-frequency distribution?
• Figure 5 appears before Figure 4, and in lines 257-266, Figure 4 is referenced as illustrating the process at three different steps, but this does not seem to be the case. Did the authors mean Figure 3 instead?
• The discussion on aseismic slip (9%) suggests that it remains constant, but does it vary with Mmax choices? Additionally, in lines 440-442, the choice of 5% aseismic deformation is mentioned, but the rationale behind this assumption is unclear. Could the authors clarify the motivation for this choice? If a tapered Gutenberg-Richter model was used instead of a truncated Gutenberg-Richter model, how would it impact the slip distribution and moment balance?
• Figures 14 and 15 appear to be important as they provide a comprehensive comparison. The figure 14 include two different models, characteristic and exponential, and it seems that the characteristic interconnected model with 14% fits well with observed events for M > 6.0. Could the authors comment on this observation? Additionally, the figure captions and curve definitions are not clearly described, making them difficult to interpret. Improving these descriptions would enhance clarity and ensure a more accurate understanding of the presented results.

It is a pleasure from time to time to find a paper well done as the one by El Kadri et al. dealing with PSHA in the Levant fault system relaxing the assumption of isolated fault segments. The authors provide evidences that many ingredients are affected by large uncertainties, some choices are subjective, but the "exploration" character of the study is clearly stated in the text.  My comments are therefore to be intended as suggestions to the authors, not mandatory requests of changes.

1) chapter 2: the observations on trenches and earthquake catalog are mentioned in Fig. 2, but the description is given in chap. 6.1-6.2. Consider to anticipate part of these subchapters in chap. 2 (also fig. 13), leaving the comparison with models in chap 6.

2) chapter 3: the description of the SHERIFS steps is always an headache, quite intriguing the transformation of a classical G-R in a pdf of the relative contribution of mag bins in the moment rates within the system. I suggest to clearly state that the moment rate for a section is L*W*mu*sliprate  (according to the values listed in Tab.1), and the global budget of the system is the summation of the moment rates of all the sections. The intrinsic problem of modelling "independent" events from the statistical point of view, or full earthquake sequences remains open, and probably deserves to be mentioned, or commented when the "acceptable" aseismic budget is mentioned.

3) chapter 4: the distribution in space of the magnitude rates, given in fig. 5 is interesting, even if I am not sure that the "participation rate" of a section to a rupture can be considered a truly annual rate. Please specify what happens to those sub-sections in Tab. 1 (e.g 1-3, 43, 32-34) not allowing M>=6.0; is their modelling limited to a participation rate in bigger ruptures? Then, considering fig. 6a the difference in annual rates of moderate magnitude is not distinguishable as stated in lines 323-324. I wonder if the moment budget is fully preserved, as stated for the results showed in Fig. 8

4) chapter 5: I suggest to change some titles in this chapter, as it is really difficult to state which model is the most realistic, given the uncertainties in paleoseismic magnitudes, incompleteness of the catalog,  evaluation of the acceptable aseismic budget (considering also that the modelling is limited to M>=6, and some seismic moment budget could be spent by small ruptures too). Titles such as for example: "5. Sensitivity tests on Levant fault system; 5.1 Test with Mmax 7.9 and full connectivity; 5.2 Selection among different hypotheses" are more neutral and open to alternative interpretations (e.g. the char eq model, instead of G-R, that in my opinion could accomplish better the absence of moderate magnitude in the last century. 

5) chapter 6: I suggest to repeat in this chapter some basic data/assumptions used, e.g. at line 485, after "of occurrences" add "grounded on slip rates, and fault surfaces ad given in Tab. 1." and at line 487, after "given shape" the sentence "i.e. a G-R truncated model with b-value=1". Then you cannot sy that paleoseismic data were not used to derive the model, as I presume  that slip rates of fault segments are mainly controlled by these data. I also suggest to avoid "regularly" (at line 515), stress the uncertainties in assigning magnitude in trenches (the Central Italy sequence in 2016 is a lesson on how many events of 6<m<6.5 can generate a rupture of a single M=6.7).  The standard deviations assigned to M at line 535 are underestimated, in my opinion; I suggest to be less sharp also in the description of the model fit vs observations given in Fig. 14.

Here some technical, minor corrections:

a) line 27 and in the following text: I suggest to avoid the use of "secondary" faults, as it can make confusion with the terminology used in displacement hazard. I suggest to use the terms "splays" or "branches" that clearly address the departure from the main fault strand.

b) line 281, Figure 3 is figure 4, correctly mentioned in the text

c) line 321, figure 4a is is figure 6a

d) line 341, figure 4 is figure 6, I suggest to change the color of the dash gray line in frame (a) with black

e) add a scale bar in both Fig. 8 (line 382) and Fig. 12 (line 479)

f) at line 528, after "363" add "A.D.". At line 540 add "known" before "fault".

g) lines 601-603, Solid orange and dashed lines are yellow and orange solid lines, if I have correctly understood

Congrats to the authors.