Dear Editor and Authors,

Thanks for giving me the opportunity to review “Fault Interpretation Uncertainties using Seismic Data, and the Effects on Fault Seal Analysis: A Case Study from the Horda Platform, with Implications for CO2 storage" by Michie et al. The general topic of the paper should be of interest to the readership of Solid Earth, dealing as it does with the subsurface structural analysis of normal faults and rift basins. It should also be of interest to people concerned with conceptual and interpretational uncertainty, both of which are critical considerations when working with structural data of any type. I have written numerous comments directly on the manuscript, which I have scanned and attached to this review. The numbers below refer to specific numbers in the manuscript.

1. “…how detailed the surface is….” could be replaced with “…how rugose the surface is…”.
2. It might also be worth mentioning; (i) the seismic line trend relative to fault strike; and (ii) the seismic or more accurately bin spacing in the original survey (which dictates the line or seismic line spacing available to the interpreter).
3. Given that it is important and that it creeps into the paper Discussion, it might also be worth mentioning in the Abstract the detail vs. time trade-off.
4. Although a little pedantic, I would use “reflection” (i.e., the thing we observe and map in seismic reflection data) rather than “reflector” (i.e., the acoustic boundary or bedding surface that generates a reflection).
5. What is the difference between a “linked” and a “composite” fault segment? They sound like the same thing to me. Maybe use one rather than both terms?
6. There are some issues with the terminology you use here for the various fault growth models. I suggest you look at Childs et al. (2017) (https://sp.lyellcollection.org/content/439/1/1.abstract) and maybe use the terms therein, i.e., they refer to the “propagating” rather than “isolated” fault model. Granted, their Figs 1 and 2 are a little contradictory in terms of the terms they use, but I think the text (and subsequent papers by those and other authors) make this clearer.
7. Which other fault growth models are you referring to here? No references are provided to support this statement. I suggest you provide some or remove this statement.
8. As noted by Childs et al. (2017) and Rotevatn et al. (2019), it is important to remember that the constant-length model does not preclude relay-breaching, i.e., it simply envisages that relay formation and breaching occurs relatively rapidly after fault initiation, such that a single, large-scale slip surface is formed prior to significant displacement accumulation. This was discussed in Jackson et al. (2017), more specifically in the last paragraph of section 1 (https://www.sciencedirect.com/science/article/pii/S0191814117300652?via%3Dihub). To truly distinguish between these fault growth models requires the integration of growth strata and not simply geometric analysis (i.e., the identification of displacement minima; see Jackson et al., 2017), something that is not undertaken or at least presented in your paper. As such, I am very hesitant as to whether your paper can really say anything about which fault growth model is most appropriate to the Vette Fault. Overall, although I think the paper can say lots of useful tings about the present fault geometry, in particular the potential location of now-breached, high-permeability relays, I am not yet 100% sure what you can say about how that geometry came to be in a temporal sense.
9. As you discuss later, there is also the issue of failure to include so-called “continuous deformation” (folding) in the geometric analysis. This might be worth mention here, given it was again something we raised in Jackson et al. (2017) (the paper in response to the 2017 paper by Ze and Alves in JSG).
10. You could mention here that an outcome of relay breaching is the formation of abandoned hangingwall and footwall splays, which represent the now-deserted tips of the now-linked fault segments. These should be used in concert with simple fault bends to identify paleo-segmentation. In fact, do you see such things along the Vette Fault? It is noticeable that you do not show a single time-structure map in the paper, when such a map would be very useful in communicating aspects of the fault geometry (including, for example, the spatial distribution of folds, and the presence/absence of near-fault splays).
11. You have already stated this a couple of times already. This could be removed without loss to the content of the paper.
12. Precisely what about the hydrocarbon-water interface? The shape? The depth? Both? This is not clear, so I would encourage you to qualify this statement.
13. First, I am not sure I follow why larger faults have lower critical threshold values. Can you please outline why this is the case? Second, and more generally, there is a *lot* (c. 2 pages) of material in this general review of fault sealing, which delays the reading in getting to the study results. My question is whether all this material is needed to provide a basis for the analysis, results, and discussion that follow? If not, can some of it be removed without loss?
14. Could you include a regional cross-section to illustrate the geological setting of the study area better? At present it is really hard, from text alone, to get a good feeling for this. My comment about a time-structure map of, for example, the base syn-rift surface (which would record the cumulative rift-related deformation), also relates to this.
15. What do you mean here by “good”? I suggest you estimate (using the extracted peak seismic frequency and the interval velocity at the depth of analysis) the actual seismic resolution. This will give the reader a better, more quantitative view of how good the resolution is, which I think is critical to convince people that variations in, for example, fault throw are really and not simply related to resolution-imposed picking limitations. Again, we discussed this in Jackson et al. (2017) when considering the very small throws (relative to seismic resolution) presented by Ze and Alves (2017).
16. Do you mean “normal” in the context of the SEG standard? See here: https://agilescientific.com/blog/2012/4/5/polarity-cartoons.html.
17. It is a bit unclear what you mean here, i.e., did you not undertake such a QC for the other line spacings? Did you only do it when picking every 25 metres?
18. How do you think a timeslice-based interpretation approach might have differed to your section-based approach? You say here timeslices were used to guide the interpretation, but if you do not see it as a valid option to replace the section-based approach, it might be worth stating precisely why not. For example, if variance timeslices are too noisy, compared to a standard cross-section, I think it would be valuable to say so. I say this because I think your paper has great value in guiding how people interpret seismic reflection data; in this case, also suggesting why *not* to use something is valuable too!
19. I am no expert in gridding approaches, so this section of text is a little hard for me to assess. However, it is not 100% clear to me from the text and Fig. 4 how the approaches vary in terms of what they’re doing to the physically picked fault sticks, or what their material impact on the fault geometry is. I guess this becomes clearer a lot later in the paper, especially through the use of Fig. 16.
20. A non-expert might not know what all these terms mean, especially fault stability and slip tendency, so I encourage you to include some supporting references…or provide a Supplementary Materials section containing information on how these various parameters are derived, what they mean, etc.
21. The term “polygon” (as used here and throughout the paper) is very confusing to me. I *think* you are using it in the sense that Badleys use it, whereas most interpreters will view it as a map-view feature that outlines the faut and essentially shows the fault heave at a specific structural level. Fig 5 does not currently really help me visualise what you mean by “polygon”, so perhaps this and the text could be modified to make this clearer?
22. Again, this material related to the present stress state is not really my area-of-expertise.
23. Why do you say this, i.e., why might dip-linkage not be associated with a change in fault strike? Is this because the upper and lower segments are inferred to have the same strike prior to linkage? In any case, I wonder why you bring up dip-linkage when, later in the paper, this is not really something you talk about.
24. What do you mean by “overlap” in this context? If the faults overlapped (prior to linking), then abandoned splays should be present. See my comments above.
25. Which structural level are you showing in Fig. 6? Top Sognefjord Formation? This is not clear in the text or figures.
26. It is not clear to me why you need to normalise the plots in Fig. 7, given that the analysed length of the fault is the same in all picking strategies (i.e., 25 m to 800 m). Please could you clarify why this was done?
27. I am confused by your use of the term “should be” in this context. Why “should” these minima be picked in the 800 m sampling case? Also, I can’t see the minima in the black circles in the 800 m spacing plot. Or is the point that the black circles are only seen on the 100 m case and *not* the 800 m case?
28. Could these corrugations also be associated with ‘drift’ in the picking by the human interpreter? Is this worth mentioning here? Oh, and I also would like to flag this interesting paper, which also reveals and discusses the origins of corrugations on seismic imaged normal faults: https://onlinelibrary.wiley.com/doi/abs/10.1111/bre.12146.
29. I am not sure I agree with this statement, given that the dominance of red and blues colours on the ‘low-resolution’ fault surface indicate that the fault still has an overall N-S strike.
30. Although you are not the first or only people to claim this, I have to say I disagree with the interpretation that corrugations are related to strike segmentation of normal faults. For this to be plausible, each segment would need to be relatively short (i.e., the corrugation wavelength, which in your case is a few hundred metres according to Fig. 8)…but very, very tall (i.e, the full fault height, which in your case is several kilometres, given the corrugations extend from the top to the bottom). This would result in faults with implausibly low aspect ratios (see Nicol et al., 1996) - https://onlinelibrary.wiley.com/doi/abs/10.1111/bre.12146. So, ultimately, I do not think picking strategy, “…may limit the interpretation of fault growth…”.
31. Or, perhaps, where fault bends occur as a result of out-of-plane propagation of the fault tips or complex early nucleation patterns? See https://www.sciencedirect.com/science/article/pii/S0191814106000320?casa_token=RfeVEhPMYFUAAAAA:97im5Akf7MWLjEkeslN3xtmWdRsR9RlNN-dpyDz1MG6GWnh7NejPQ0BzekKq3Fh5vZTyYFWEubg and https://www.sciencedirect.com/science/article/pii/S019181410600191X?casa_token=hBmvUeFBEXIAAAAA:e9K7bQaMHsUShAMzVWx7nv1TU8jsglkp5g9ekW95TKDzgYU3bYH4Lctm29GO711ROJyMhZrUkNA
32. Related to comment 25, which levels on the strike projection in Fig. 9 does the T-D plot come from? I ask because the black dashed lines are shown as vertical whereas the corrugations plunge slightly southwards, meaning they do not line up. Or at least they do not line up along the entire dip extent of the fault. Maybe they line up with the level at which the T-D plot is constructed, which is the reason to ensure that this is stated in the text, figures caption, and if possible labelled on the figure.
33. Regarding the comparative T-D plots for different picking strategies, maybe I’m misunderstanding something here, but shouldn’t there be locations where the values are exactly the same? For example, every 32^nd 25 m-picking line would lie up with an 800 m-picking line? If so, why is this seemingly not clear on Fig 10E?
34. Where is 0.4 on Fig. 10D? Do you mean values less than 40%? If so, there are hardly any values less than 40%. I’m a little lost here, so some clarification might be worthwhile.
35. Can you perhaps state how big these “patches” are and, most crucially, give a sense as to how they are distributed at and above the critical Sognefjord reservoir level? Surely, in the case you present here, the variability in these locations, arising from the various picking strategies, is what’s really key?
36. Why do you propose 100 m? What do you not believe it real, geologically speaking, at spacings less than 100 m? And/or what do you not think is important, in terms of CO2 storage and potential leakage, at these very small scales?
37. This whole section is a little ‘wordy’ and I think the language could perhaps be streamlined and simplified a little There are many, many occurrences of the terms “dilation and slip tendency”; could these perhaps be used only when essential?
38. Rather than stating “lines” in terms of seismic line spacing, could you stick with the terminology used throughout the paper up to this point, i.e., horizontal spacing of 25 m, 100 m, 800 m, etc.
39. What do you mean by the inlines and crosslines “not tying precisely”?
40. Is it not rather obvious that a relatively thin reservoir, in a relatively muddy sequence, next to a relatively large fault would lead high SGRs and a high likelihood of fault sealing? In that sense, I do not find it surprising that the picking strategy was an important control on this.

In summary, this is a very interesting and important work, which I very much enjoyed reading. Overall, the paper is well-structured, well-written, and the work is thorough and, for the most part, convincing. However, as I hope is clear from my comments above, some revisions will, I hope, help further improve the manuscript. In general, the English and grammar are very good, although there are a few places where these could be improved. I am more-than-happy for the authors to contact me to discuss any of the issues raised in my review.

Yours sincerely

Christopher Jackson (Christopher.jackson@manchester.ac.uk