Dear Dr Kirilova and coauthors,

I have now carefully read your revised paper and also considered James Gilgannon's review of that version of your manuscript and came to the conclusion that, in its current form, the manuscript is not yet ready to be published in SE.

I do think though that with a rework that addresses the concerns, your paper will make a robust and well-cited contribution, and I look forward to receiving your revised version.

Best regards,
Florian Fusseis

**********************************************

Comments

I see a certain disconnect between the data you present and your discussion, especially Section 5.3 - it isn't obvious to me that your data really support the conclusions you draw there. This doesn't mean that the conclusions are incorrect, but rather that the trends in the data you report are not overly obvious.

In particular, I have concerns about the validity of the shape analysis, which hinges entirely on the covariance matrix. I would argue that the matrix describes the shape of an ellipsoid, not a pore. Thus, for the conclusions to be robust, would one not need to demonstrate that the eigenvalues/vectors indeed are a good representation of pore shapes? Obviously, pores in deformed rocks can have very convoluted shapes which are difficult to describe with standard mathematical operations. There are, however, more sophisticated tools, such as Minkowski funtionals, which you could use.

My second major concern arises from the polynomial fitting - where really the R^2 values aren't too great. I suspect, on the basis of my own experience, that you could yield substantially different porosity values quite easily if you used slightly different fits.

Specific comments (line numbers)

98 I think it might be helpful if you very briefly, as obviously described in detail elsewhere, mentioned how these samples are related to the activity of the fault. A figure (BSE images and corresponding slices through uCT data) showing the microfabrics of the four samples should be added (and can replace some of Figs 1 and 2). Also, mineral compositions, grain sizes and proportions should be reported in a table. This is relevant for your section 5.2 below.

103 Please report on the sample-detector distance, as this is important to estimate a possible phase contrast in the data.

103 The sample size is only of minor relevance here, since the field of view will only be just over 2 mm high at 10x magnification (which I believe gives 1.3 um voxel size). What is the width of your FoV?

106 Why no phase retrieval in reconstruction?

111 What are the dimensions of the sub-samples, how did you choose them, and are they statistically relevant?

113 Why use a primitive thresholding tool, where much more sophisticated tools for local crack detection are available (e.g. Voorn et al., 2013, 2015, Ma et al., 2020)? This is especially relevant since the entire paper hinges on the analysis of your segmented porosity.

Voorn, M., Exner, U., & Rath, A. (2013). Multiscale Hessian fracture filtering for the enhancement and segmentation of narrow fractures in 3D image data. Computers & geosciences, 57, 44-53.

Voorn, M., Exner, U., Barnhoorn, A., Baud, P., & Reuschlé, T. (2015). Porosity, permeability and 3D fracture network characterisation of dolomite reservoir rock samples. Journal of Petroleum Science and Engineering, 127, 270-285.

Ma, X., Kittikunakorn, N., Sorman, B., Xi, H., Chen, A., Marsh, M., ... & Skomski, D.
(2020). Application of Deep Learning Convolutional Neural Networks for Internal Tablet Defect
Detection: High Accuracy, Throughput, and Adaptability. Journal of Pharmaceutical Sciences.

113 You also must have labelled the data, which degree of connectivity did you choose? This will significantly influence your pore statistics.

119-128 This approach could be discussed considering the findings of Keller et al., 2011, 2013a (and others) on the porosity distribution in clays (as these being related to gouges).

Keller, L. M., Holzer, L., Wepf, R., & Gasser, P. (2011). 3D geometry and topology of pore pathways in Opalinus clay: Implications for mass transport. Applied Clay Science, 52(1-2), 85-95.

Keller, L. M., Holzer, L., Schuetz, P., & Gasser, P. (2013). Pore space relevant for gas permeability in Opalinus clay: Statistical analysis of homogeneity, percolation, and representative volume element. Journal of Geophysical Research: Solid Earth, 118(6), 2799-2812.

121 "with limit of" is misleading, is this smaller or larger than 200 voxels?

130 10 voxels as a lower limit is too small, consider the limited number of shapes you could build out of 10 cubic Lego bricks - this inevitably introduces a bias in your analyses. Consider a lower limit of 5^3 voxels.

135-137 I disagree with the statement that the eigenvalues describe the shape of a pore - they describe the shape of an ellipsoid, which only under fortunate circumstances will be a good approximation of the actual pore shape. To make conclusive statements on the shapes of your pores, you need to go a step further and use, e.g., Minkowski functionals (Arns et al., 2002, Lehmann et al., 2006, Wildenschild & Sheppard, 2013).

Arns, C.H., M.A. Knackstedt, and K.R. Mecke. 2002. Characterizing the morphology of disordered media. p. 37–74. K. Mecke and D. Stoyan (ed.) In Morphology of condensed matter: Physics and geometry of spatially complex systems. Lecture Notes in Physics.
Springer, Germany.

Lehmann P et al. Tomographical imaging and mathematical description of porous media used for the prediction of fluid distribution. Vadose Zone J 2006;5(1):80.

Wildenschild, D., & Sheppard, A. P. (2013). X-ray imaging and analysis techniques for quantifying pore-scale structure and processes in subsurface porous medium systems. Advances in Water Resources, 51, 217-246.

155-157 The screenshot you have added in your rebuttal to J Gilgannon indicates that your segmentation captured only a very small proportion of the actual porosity. How have you chosen the threshold value for your segmentation?

164-167 Fig 7 - what is the reference frame for these stereoplots? I see you are speculating that they trace a foliation, but obviously the maxima are all roughly at the same orientation between the plots.

170 - "the gouges have composition..." - rephrase

177 - What are the criteria by which you distinguish cracks from other types of pores in your TEM data? It is not really evident from Fig. 8.

185-187 This statement is incorrect, since this is the voxel size, which does not equate to the spatial resolution.

188-190 This could be contrasted with Keller et al (2013b)'s study on Opalinus Clay. Obviously, they investigated undeformed clay as opposed to highly sheared fault gouge, but the key finding suggests that, in clay-mineral rich portions, there may be a very substantial volume of porosity hidden. What are your arguments in support of this statement?

Keller, L. M., Schuetz, P., Erni, R., Rossell, M. D., Lucas, F., Gasser, P., & Holzer, L. (2013). Characterization of multi-scale microstructural features in Opalinus Clay. Microporous and mesoporous materials, 170, 83-94.

191 (also section 3.4) How were the locations for the FIB foils chosen? How did you ensure they were representative?

192-194 I have difficulties understanding how the first half of the sentence relates to the second, and how you arrive at that conclusion.

205-208 cf. Keller et al., 2013b

209 Please restructure this section for clarity and revise paragraph structure. The reader shouldn't have to read until the end to figure out what this is about.

213-214 How do you identify these clay minerals as authigenic?

214-216 Annotate these observations on the figure, and pls add references to support the pressure solution argument.

228 Would be good if you provided grain size distributions for the two materials, see comment for line 98.

233 Does your analysis really demonstrate significantly different porosity values between the two cataclasite units, given the R-squared values of your regression model?

240-251 This paragraph is a bit bold on the assumptions. First, low porosity is not necessarily consistent with low permeability, and I don't see how you verify permeability measurements? Aren't you over-interpreting a difference in porosity of 0.12%, given the uncertainties outlined in your own statistics plus the uncertainties introduced by the data processing steps and sub-sampling sites you have chosen?

254 If it is precipitation of hydrous minerals in pores (as opposed to veins), how much does the pore fluid pressure actually change as clay minerals form?

255 Neither Byerlee nor Sibson do actually talk about precipitation in pores (but of course about fluid pressurization), could you reformulate this slightly to avoid that impression?

262 Is there graphite present in these samples? - It isn't highlighted in Fig 8. If not, please clarify this statement.

Fig 2 isn't particularly helpful, since it doesn't show a workflow as such (sensu flow chart). I wonder whether 2.5D "thick slices" (see, e.g. Fig. 3 in Menegon et al., 2015, Geology) would be more better.

419 Fig caption "after the fractures" - "after removal of the fractures from the segmentated data".

Fig 2c, since this is a perspective figure, the scale bar does not apply to the entirety of the image, pls consider/correct.

438/439 I don't see a difference between pores in b) and d). What are your arguments for the pores in d) being fractures?

Dear Authors,

Thank you for your reply to my comments. Unfortunately, it does not convince me that your porosity estimates are robust and support your interpretations sufficiently. Your descriptions are neither precise nor detailed enough to allow for a reproduction of your analysis, and I do miss evidence that supports the choices you have made in your data processing. Just a few examples to justify these statements:

What data support your claim that binary thresholding is “more than sufficient for robust porosity segmentation”?
What threshold values did you use, and how were they chosen?
Where they the same between all four datasets?
What is the effect of the NLM filter on your porosity quantification?
Are your subvolumes statistically representative?
Where all of your pores labelled on the basis of face-connected pores?
What degree of voxel connectivity did you choose to label your pores?

All of this is critical, as your entire interpretation is based on your segmented porosity data. Your reference to Menegon et al., (2015) and Gilgannon et al. (2017) is to a degree justified, both do indeed use the same thresholding technique. However, we now have significantly more sophisticated algorithms freely available, and they should be used. The significance of the choice of segmentation technique has been demonstrated in a large number of publications. On the basis of these, I am actually certain that I could analyse your data with a different set of algorithms and arrive at significantly different numbers, which, I think, highlights the principal problem.

As is, I cannot recommend the paper for publication. I do you encourage you to address the concerns raised above, and I would further ask you to consider the review of your revised manuscript provided by James Gilgannon. I would further ask you to include the detailed microstructural sample description that must form the basis for your interpretation of the porosity in these rocks.

Kind regards,
Florian Fusseis

Dear Dr Kirilova,

Thanks for the submission of your revised manuscript and supplementary material, which I have read carefully.

I would like to remind you that the choice of the appropriate reviewer lies with the topical editor, and the choice of Dr Gilgannon was entirely justified, irrespective of the fact that he used to work in my group. I would also like to emphasise that Dr Gilgannon has contributed two extremely constructive reviews, both of which were available to you through the MS records website (the latter on August 6).

I understand that you feel that the review process focussed overly on technicalities, however, as described in SE's review criteria (https://www.solid-earth.net/peer_review/review_criteria.html), before a manuscript can be accepted for publication, the scientific methods presented there must be [...] clearly outlined (4), and the description of your work must be sufficiently complete to allow reproduction (5). I am satisfied that this is now the case, and will recommend your manuscript for publication pending a minor technical addition:

Given that you have chosen four substantially different threshold intervals for four different uCT datasets (DFDP-1B 69_2.54: 0 to 28; DFDP-1B 69_2.57: 0 to 44; DFDP-1B 58_1.9: 0 to 74; DFDP-1B 69_2.48: 0 to 46), I would ask you to include screenshots for all four samples with the segmented porosity highlighted (as you did for your sample DFDP-1B 69_2.54 in your rebuttal letter) in the supplementary material. I would ask you to further include the grey value histograms, showing the threshold value, with this figure.

Lastly, as I have done already in email exchanges with your coauthors Toy and Renard, I would like to apologize for the slightly irritating tone of my last communication.

With kind regards,
Florian Fusseis

Dear Authors,

based on the Topical Editor's report and overall assessment of the review process, I am pleased to accept your manuscript for publication in Solid Earth.

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

Sincerely,

Federico Rossetti