General comments

This manuscript by Kelka et al reports an application of automatic lineament detection of a large-scale subsurface fault network in Australia. Authors use a compelling geophysical dataset and provide a precious comparison between  various automatic lineament detection and manual lineament detection. The manuscript is well-written, rather efficient in delivering its message, and figures are important and well designed, even if some are too small, given the richness of the details they bear. This manuscript is an important contribution as large-scale mapping of lineaments (here faults) is an area of growing impact in geology, wether it is for resource exploration, but also for structural integrity of an area and seismogenic risk. I think it fits perfectly the thematic of the special issue it aims at but there is some points that would benefit some more explanations. Hereinafter are reported some relatively minor questions, remarks and suggestion I wish the authors can answer before publication

Specific comments

• What is the depth reached by the subsurface dataset? Could not subsurface and surface data combined to produce a 3D mapping that also take into account the dip of the faults? I guess they here are vertical, but such approach will be usable in other tectonic context, hence it would be an added value to be able to get the dip of faults from this kind of datasets. Also would the use of a subsurface data help in areas with limited outcrops due to vegetation?
• My main problem is that this paper present a workflow, but there is no actual workflow diagram that would make it easier for the reader to understand what are the inputs, treatments and outputs at the different steps. Each steps are individually well illustrated and explained, but I missed a nice overall figure for the workflow.
• It seems that the authors restrict this workflow to regolith, and its interest to Australia. I understand that this is a special issue, but Solid Earth is a large-scope publication, and so I suggest authors extend the interest of this study beyond Australian mining in regoliths in the introduction and discussion
• There are been numerous studies that use drone/lidar images to map fractures from automatic lineament detection, should it be discussed what exactly does the subsurface geophysical dataset add to the table? There is two lines about that around l. 297 but it is very shallow, could you develop?
• I miss a brief discussion about the uncertainty and resolution of the method (e.g. on the orientation of lineaments, length, or % of lineament actually existing)
• The area of study has a long and complex geological history. So how do you know that the lineaments are indeed from the same long-term tectonic history which would be relevant to the fluid flow and related ore deposit?
• Is there anyway to interpolate between the large gaps by giving the machine a specific regional trend?
• Finally, some paragraphs are really hard to get because of the overuse of jargon, unexplained terms that the non-specialized reader will left puzzled with, without any reference. I am referring to the treatment part mainly: “10 Gaussian to the Kernel”; “Akaike information criteria”; “Sobel filtering”; “Green function”; “Canny edge detection”. What are those? Please explain in a few words and add references for more in-depth explanations.

Technical comments:

• The abstract should summarize the main results of the study please replace l. 10 onwards with the actual finding of your study.
• 1 why is it challenging only in Australia?
• 24 « « these and existing » suggest these don’t actual exist, please reword
• 30-33: complicated long sentence, could you add a break?
• 42:what exactly is a targeting map?
• Can the Olympic Dam located on fig 2?
• Figure one is small and the located area reported in a too subtle way.
• Is there a cross-section available for the area
• Line 101: not sure transported cover is the good wording, what do you mean?
• 114: what analytical products? That is rather vague
• 139: parenthesis missing after figure 4
• 150 it is under explored yet exploited for 125 years??
• 155 157 : please remove the teaser

This paper presents a workflow to firstly identify lineaments (faults and shear zones) by different techniques and later on plot their intersections in an effort to create density maps that can be used as exploration tool in areas with a ‘thick’ sedimentary cover.

In my opinion, the paper is very poorly written, especially the geological part. The text is very often redundant. Some parts are repeated constantly throughout the text (e.g.,  relationship between mineralization and faults). There are millions of commas missing. The expression ‘the study area’ is constantly repeated (I have counted up to 36 times).  Figures 13 and 14 are not discussed in the text and only referred to in table 1 or in figure 15. They should be at least minimally discussed.

From a geological perspective, the Geological Overview is not well organized and it is almost impossible to get an image of the geology of the area. Structures and zones cited in the text are often missing in figures so it is impossible to follow the description. The differentiation between shear zones and faults the authors make is meaningless. And the paper lacks a description about the age of faults/shear zones and that of mineralization. If the latter is older than most of the structures, what is the point of doing this work. Contrarily, if mineralization is younger, it could use any pathway. That needs to be described in the introduction. And if this is not known, you need to say it and use it as support for your work.

In relation with the technical part, the work has a sound mathematical basis used with little geological/rational support. The datasets have different resolution, implying that identified features have different length and are going to be plotted in different positions (something that is key to make intersection maps). Upward continuation uses values that are not explained (why 2000 m or 900 m and what does it imply). Why not using vertical derivatives to picture gradients and boundaries/lineaments? These are good at spotting gradients, which in potential field data are evidence of lithological boundaries (by fault, shear zones or purely compositional). And last but no least, the places that authors point as potentially interesting areas for mineral prospecting (Fig. 15), very seldom coincide with mineral occurrences, so we have a problem.

In summary, although mathematically the paper is correct, it lacks a good geological background. Processes are applied over datasets that might not be useful for the purpose of the paper. And in case it was going to be published, it needs to be rewritten, mostly the geological part. Also, datasets need to be further discussed and errors and limitations considered.  So in my opinion, the paper has to be rejected in its present form although resubmission  of a new MS could be an option given the goal and background of the paper

Ahead there are some comments that I made before I realized it was better to make annotations in the MS. So use both, this text and the annotated MS for review.

Some comments. More in the tex.

Line 81: The key geological features should be enough….

Line 87: What does Lake Harries Greeenstone Belt (figure 1 and 2b) have to do with the lithologies? You haven’t cited this belt before. Cite figure 1 so we know what you refer to.

Line 91: Rocks of the Hiltaba Suite (figure 2).

Lines 91 to 95: No need to mention four times ‘study area’

Line 96: Granitic gneisses are orthogneisses?

Line 97: What is a low magnetic signature? Low amplitude?

Line 98: Comma missing after Suite.

Lines 117-120: How can you distinguish the magnetic signature of shear zones and faults? That paragraph responds to your own work or there references missing. Fluid flow also demagnetizes or remagnetizes shear zones. Also, shear zones also juxtaposes blocks of rocks with different magnetic character.

Line 123: I don’t see, as such, a NW trending Mulgathing Trough in figure 2a. In general, there is little relationship between geology as described in the text and shown in the maps. This needs to me highly improved.

Fig. 1: Add a rectangle in the Australia map in figure 1b delimiting the area shown in the bigger figure 1a. Maybe figure 1a should be figure 1b and viceversa.