Hyperspectral Facies Analysis as a Lithological Interpretation Tool for Carbonate Rocks
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsReview of “Hyperspectral facies analysis as a lithological interpretation tool for carbonate rocks”
This manuscript presents the results of a pilot study applying hyperspectral core-scanning techniques to carbonate lithostratigraphy, using drill-cores from the Irish Midlands. Given the challenges often encountered when objectively characterising and correlating lithofacies in such carbonate-dominated environments, this interesting study hopefully serves as a first step towards more widespread use of hyperspectral techniques in the region. The methods applied are robust and appropriate for the paper’s relatively modest conclusions. However, I found the structure of the manuscript a little confusing and difficult to follow, and the discussion rather lacking. Hence, I recommend accepting the manuscript for publication following the moderate revisions outlined below:
1. I suggest that the clarity of the manuscript would be greatly improved by adding a section “2. Geological setting and lithostratigraphy” in which (i) the regional and temporal setting of the discussed drillcores is established, followed by (ii) a relatively detailed lithological descriptions for each drill core (these are currently dispersed through the methods and results). This could then be followed by a dedicated section outlining the expected hyperspectral response of the logged lithofacies and the mechanisms behind this (e.g., Al-OH and CO3 absorptions, spectral quenching, etc.). Currently this important background information is dispersed through the methods and results, making it difficult to follow the authors’ key arguments by disrupting the flow of the narrative.
2. Given the availability of validation data (Table 2), a quantitative assessment of the proposed decision tree classifier’s accuracy should be presented (assuming that each thin section can be manually assigned to one of the spectral classes based on petrological observations). Presented as a confusion matrix, this would greatly facilitate the readers ability to assess the merits and robustness of the technique and derived lithostratigraphic logs. Additionally, it would allow Table 2 (which is overly large) to be shifted to the supplementary material, while still retaining a comparison between the hyperspectral and petrological lithofacies.
3. In addition to moving the contextual geological information into a dedicated introduction section, I suggest that the results section could be further clarified by first (i) establishing the relative accuracy / limitations of the proposed hyperspectral decision tree classifier (using e.g. a confusion matrix; cf. Point 2), and then (2) documenting the authors general lithostratigraphic/hyperspectral observations and hypothesised correlations.
4. The discussion is rather too brief, especially for a pilot study that should suggest avenues or limitations for future works. I suggest that this section could be extended to more broadly discuss the potential (and limitations) of hyperspectral imaging in this geological setting and, especially, define its potential with respect to currently outstanding challenges in Irish geology. Additional points that could be discussed include:
a. The strengths/limitations of the hyperspectral approach given its inherent bias towards mineralogy rather than other lithostratigraphic features (e.g., grainsize, texture, fossils, etc.).
b. The potential for sensors of different spectral ranges to better define lithostratigraphic units (e.g., integrating the VNIR for Fe2+ content, MWIR to mitigate issues with spectral quenching in dark carbonates and added sensitivity to quartz, etc.).
c. Implications of the proposed hyperspectral approach for (i) quantitatively measuring (potentially) systematic lithofacies changes in space and time, and (ii) exploration and minerals system analysis.
My other minor comments are as follows:
Line 16: Mississippian – please also provide absolute age ranges throughout (for us non-stratigraphers)
Line 87: Were the masks manually validated / updated? If not, do other materials in the core boxes (e.g., plastic or wooden meter markers, paper sample labels, plastic packing, etc.) not remain in the unmasked areas? How might this affect the results?
Line 103: Kaolin absorptions must show a double absorption though (~2160 and ~2200 nm), otherwise a feature at 2205 would still be interpreted as a clay or mica. Please clarify. Also, it is worth mentioning that coarse-grained carbonates often also show a higher-order absorption feature at ~2160 nm, which sometimes interferes with mapping and identification of accessory clays and muscovite / kaolinite group minerals.
Line 106: “Siliciclastic material”; consider rephrasing as “argillaceous material” or similar, given the SWIR range is not sensitive to the quartz or feldspar grains generally implied when using the term siliciclastic.
Figure 2: What is the “Sample” spectra in this figure? Please also (1) make the legend larger, and (2) use different line styles for a print and colourblind friendly figure.
Line 121: Clarify that the peak finding algorithm is being applied to identify local minima.
Line 128: How did you select the number of bands used for the quadratic fitting? How sensitive are the results to this choice? Given the importance of this fitting to the resulting spectral lithofacies classification, it’s worth expanding this into a full paragraph.
Line 138: I suggest that a small figure showing the decision tree classifier that is described here could be helpful to the reader.
Line 174: Add a short paragraph introducing the results section before Table 2.
Figure 3: Please use more visually distinguishable colours for calcite vs dolomite – as it is, it’s difficult to clearly spot dolomitised horizons. A scale indicating if the core trays pictured are 1 or 3 m long would also be useful.
Figure 4: This caption doesn’t seem to match the figure?
Figure 5: Consider using a majority filter (e.g., scikit-image’s rank.majority filter) to remove “speckles” in these classifications.
Line 238: Dolosparite? Please define.
Line 256: Surely increasing mineral variability would exaggerate not reduce facies differentiation?
Figure 7: Please add a legend?
Line 357: It would be interesting to add a paragraph or two here discussing / interpreting these results in a sequence stratigraphic context. E.g., is there any relation between the hyperspectral facies (especially clay-rich layers and/or thin dolostones) and sequence boundaries / maximum flooding surfaces?
Line 360: Which fault block?
Line 423: Carbonaceous normally refers to something containing organic carbon or graphitic material?
Please also see additional specific suggestions in the attached annotated pdf.
Kind regards,
Comments for author File: 
Comments.pdf
Please see additional english language suggestions in the annotated pdf attached.
Author Response
Thank you for your time and effort in making such constructive comments to improve this paper. I have responded point by point in the attached document.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsPaper: Hyperspectral facies analysis as a lithological interpretation tool for carbonate rocks
General assessment
This is an interesting case study that is well written and should be published with minor revisions. Below I make a few suggestions for modifications to the text and a figure to improve clarity and I leave it to the author to determine if they wish to go ahead with the recommended changes. I also make suggestions for additional information products to derive from HS data likely as part of future work (estimation of toc and showcasing wavelength displacement of carbonate feature). This study will fit well into a growing body of studies focusing on carbonate rocks using HS data. You might want to do another literature search because I know of at least two more carbonate studies (at outcrop scale, one in Africa & one in Canada) that might be worth quoting.
Detailed comments
Results
Figure 3: table 1 details 8 classes but in figure 3 they are displayed as 4 classes in shades of green and 4 in shades of blue, I can understand the geological motivation (to differentiate clean vs muddy) but from a perspective of discerning these classes in the figure its not great. I see a lot of spectral products and here at best I can discern 2 greens and 2 blues which defeats the purpose. I would encourage you to rethink your choice of colors. It may be a lot of work but I would improve the impact of the figures. I realise it works in the enlargements e.g fig4 but it precludes a synoptic view in fig 3. Also have you though to display the wavelength of the carbonate feature along side, it would be informative to assess the range of composition of do and cc as its basically treated as a 2 class problem here and perhaps it is not. It’s a suggestion, not a requirement
Figure 4 caption is unrelated to the figure, a mistake, you need a caption. And you don’t quote figure 4 in the text of section 3.1.1 and that applies to subsequent sections.
Line 256 you need to help the reader here no familiar with this stratigraphy when you refer to the Tournaisian. May point to a depth interval or a set of formations or restate “below the blue arrow”. Again you are not making good us (her for fig 7) of the enlargement in the body of the text, its not quoted anywhere.
Line 354 “allows for quick and easy differentiation” not apparent from fig 3 where much of the formations in the Tournaisian look the same though I realize this can be a function of scale but it makes me think there could be more to pull from the hs data for discrimination
Line 431 might be worth stating it can also guide sampling for more detailing investigations if hs analysis is conducted upstream.
As a final comment you might want to design a toc estimator, I suspect it would be of help in extracting further info from the muddy units and your QC units, because as it stands you are not getting as much from HS data in the Tournaisian
Author Response
Thank you for taking the time to review this article and provide constructed comments towards its improvement. I have responded to your comments point by point in the attached document
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsI would like to thank the authors for their revisions, and am happy to recommend acceptance of the manuscript in its current form.
Comments on the Quality of English LanguageThe writing quality is sufficient to ensure clear communication. Minor copy-editing is still required.
