Review of Austin et al. manuscript essd-2023-464, “Integration by design: Driving mineral system knowledge using multi modal, collocated, scale-consistent characterization”

Randy Enkin, Geological Survey of Canada

With an extraordinary collection of 1590 well-chose rock samples from in and around 20 mineral deposits from the Conclurry District in NE Australia, the authors have compiled a consistent set of geological, geochemical, and petrophysical measurements.  Drawing on a decade of work, the resulting data set is a globally unique resource to test a whole range of exploration vectors in a geologically complicated region. 

    There are a few cases of this approach taken on individual deposits.  I am particularly thinking of the Canadian Mining Innovation Council’s “Footprints Project”, focused on 3 individual deposits.  While many exploration vectors were developed through that project, the unified approach was not as well-achieved as in Austin’s manuscript.

     The complete data set is easy to download and analyse. I can attest to the quality of the petrophysical data, and their relationship to the pXRF chemistry.  The Henkel plot (logarithm of magnetic susceptibility against density) is quite fantastic, with the IOCG deposits, like Ernest Henry and SWAN, plotting just above the Quartz-Feldspar-Calcite + Magnetite line of Enkin et al. (2020), in a similar manner to our Great Bear Magmatic Zone samples.  I explored the chemical relationships a bit, and see the expected anticorrelation of density with Silicon concentration, and the correlation of magnetic susceptibility with iron.  I was pleasantly surprized to see how well the electric conductivities plotted against density.  I had given up on using electromagnetic measurements of conductivity in my lab, but I have never had such an excellent collection of mineralized samples.

    I appreciated the descriptions of the lab methods, however I believe there are more details to include in the density section.  The Conclurry samples are not very porous, however the method of weighing the immersed samples can be quite sensitive, and should be described. 

    Note, I disagree with the statement (Line 365): “Users should be aware that the instrument can realistically only measure 10 cm³ samples up to susceptibilities up to ~2.25 SI. In some cases, particularly in magnetite-rich or mussketovite-rich ironstones, susceptibilities are likely much higher, probably in the range of 10-20 SI (Clark, 1988).”  We are not interested in the intrinsic magnetic susceptibility of rocks, but rather the external susceptibility which is limited by their demagnetization field.  The only way to get magnetic susceptibilities above 3 SI is for the magnetite grains to be elongated – which happens to a small degree in IOCG magnetite metasomatism.  But there is no apparent problem with the reported measurements.

     I believe the geological map (Figure 3) would be improved with some age information, and the symbology in Figure 4 should be described with a legend.  I feel the eight deposit-scale maps (Fig.s 5-12) are of less value, and could be moved to an electronic supplement.  Figure 14 would be improved if the Magnetic Susceptibility were plotted on a logarithmic scale (a Henkel plot).

     The important context for this paper is outlined in the multidisciplinary Venn diagram (Fig. 25), its description.  The authors are perfectly correct that combinations of different techniques should be done on a consistent set of samples to have confidence in the relationships.  The scaling involved in different geological and geophysical analyses have to take a consistent approach as argued in this manuscript.  And finally, the paper promotes the multidisciplinary approach, which is an essentially human ability, to be more successful with a new generation of mineral exploration.

    “The dataset that presented here, provides a unique opportunity to examine this complex mineral system through quantitative and scale-consistent means. We believe that this style of dataset is a pre-requisite to gain useful quantitative insights into the Cloncurry District, which will, hopefully, lead to some step changes in how we explore in this highly complex piece of the Earth’s crust.”

    Thank-you for giving me the opportunity to read this paper and examine its data set.  This will be a reference paper in many complementary fields.

General comments

Well done by the author team for putting together this extensive dataset – it must have taken a lot of effort, and as highlighted by some commentary in the manuscript, combining data from different sources and time periods may be challenging. The outcome seems coherent, and the data collection has been systematic enough (some minor variation in the equipment, but these are addressed in the manuscript).

I do like the “integrated by design” approach and agree with the authors that too often are geodata combined without considering the differences in scales, resolution, and data type (for example, direct measurement data vs interpreted data such as inversion). With the approach taken by the authors, we are likely as close to being able to integrate various datasets seamlessly as we’ll be able to get with current technologies. However, I would argue that also this approach is not fully scale-consistent: for example, bulk density of a sample may not correspond to pXRF data, which is a point measurement in comparison. I think this has probably a small overall effect in the results, as the sample selection has been done carefully and samples are representative (and hopefully quite homogeneous), but all in all, I think this uncertainty should still be addressed, as it could have perhaps even significant effect in the data quality in similar future campaigns, if not considered in sample selection and preparation.

All in all, it's an impressive dataset with a supportive manuscript. The final data excel needs some attention and minor fixes, see the notes below. This is not an exhaustive listing, and I suggest the data file could be inspected once more for possible inconsistencies.

Specific comments

Introduction

• I would suggest tightening (shortening) the Introduction a bit, I think it could be done without losing any information.
• The complexity levels listed in the introduction are 1) variable precision, 2) variable scale, and 3) multi-dimensionality. Sparsity is mentioned as a factor added by the third dimension (line 66 onward). I suggest sparsity could be an item on its own and deserves more attention. After all, also in the Cloncurry dataset the data may be at times sparse either due to sampling or to data acquisition.
• The introduction mentions (line 100) statistical approaches as the common way to use complex datasets. However, in mineral exploration mineral prospectivity modelling (MPM) is probably more common, although data uncertainty estimations are still not always considered, but literature on that is available, too. I suggest you could use here an example from mineral exploration, not reservoir modelling.
• It would be useful to the reader to be able to make the mental shift from general geoscientific dataset description to mineral exploration clearer in the Introduction.
• In general, the Introduction has quite few literature references. Has anything similar been done anywhere? What are the preceding studies? How does this database compare globally?
• Fig. 2 would benefit from adding scales to especially B) and C), as scale is a critical topic here. Note that in the figure, the letters A, B, C are missing.
• From line 130 onward it would be helpful to mention the general location (Mount Isa region, Australia) of the deposits here, when they are first introduced

Rest of the manuscript

• I agree with the comment by the other reviewer: the deposit pictures could be transferred to additional material. 2D images of 3D models are always tricky, and there are several ways these figures could be improved (resolution, overlapping texts, inconsistent visualisation of sampling etc.), but I know it’s very time-consuming. The figure captions are also not quite consistent in style. As additional material I think they would be ok. Figure 10 is especially difficult for me to translate into mental 3D.
• Page 7 discusses the sampling and the limitations within. I think occasional inconsistent areal coverage or sample spacing and occasional challenges in representativity may be relevant for data sparsity (see above). (This is just an additional comment on sparsity, no need to fix anything).
• Chapter 3.1 Sample preparation: it was unclear to me (until further chapters) whether all sample types were finally prepared similarly into c. inch-by-inch samples. I understand they were, but this could be clarified, as line 330 states that the 25-mm samples were prepared, which I took to mean the sample type 1 (25-mm diamond cores). But the block samples were also drilled with the 25-mm dill, and line 330 then means all sample types?
• The manuscript (line 400 and Figure 25 y-axis) and the dataset (see below) at times seem to indicate that induced magnetization equals to magnetic susceptibility. This should be corrected.
• Line 491: the explanation “simple excel spreadsheet” is not clear to me in this context
• Some more information could be added on the measurement procedure chapters
  • what was the gamma-ray measurement time and how was the small sample size taken into account
  • what was the measurement time on pXRF? Were the samples measured once or averaged from several measurements?
  • The pXRF drift was monitored – was there any?
  • Are the resined samples marked somehow in the database?

Technical corrections

Manuscript

• Line 135: “all major techniques used in mineral exploration” could be “all major techniques used in rock material analysis in mineral exploration” or something similar (bit more specific)
• Line 140: I don’t understand the “mining space”
• Line 194 extra comma?
• Ch3.1 NQ and HQ could be explained
• Line 335: there is no Figure 13c. AMS already defined.
• 22 caption text should probably say 0.132 instead of 0.123

Dataset

• There are two figures at the end of the datafile on the complete database file (cf. cells N1460 and N1550), at a glance they look like some QA/QC plots and if not necessary, can be removed
• What is column T? The header says “0.00” (cell T1)
• I would prefer unique column names – there are two columns named “N” and two columns named “Mean length (%)”, and they will be hard to process and distinguish from each other with, for example, Python or similar, or any software.
• Personally I would prefer kg/m³ as the density unit (SI convention)
• The description for column “Q bulk” is misleading, as it incorrectly suggests induced magnetization equals to magsus: “Koenigsberger ratio: Ratio of Remanent Magnetisation (mean NRM) to Induced Magnetisation (mean magnetic susceptibility)”. The calculation seems to be okay, i.e., NRM to induced magnetization.
• For pXRF data, it would be useful to explain in the descriptions what <LOD means, for users not familiar with geochemical data
• The hyperspectral data columns are named with the wavelength number without any letters except for “350 nm” and “2500 nm”. Consistency would make automated processing more straightforward.
• The description sheet contains descriptions “Reflectance measured in the 350 nm wavelength band” and “Reflectance measured in the 351 nm wavelength band” for channel names “Wavelength 350 nm” and “351”, respectively. It could be clarified that these are examples of channel names (i.e., not all spectral channels are included in the descriptions). Moreover, there is no data column “Wavelength 350 nm” but “350 nm”.