Mineralogical domaining of low grade and no grade zones using automated drill core logging

Globally, 400,000 km\\(^2\\) of land is impacted by the development of mineral deposits producing one of the largest industrial waste streams with an estimated 4 to 15 Gt of broken waste rock and tailings produced annually (Lottermoser, 2010; Hooke and Martin-Duque, 2012; Haas et al., 2015; L‚àö¬Æbre et al., 2017). The increased potential for acid and metalliferous drainage (AMD) from mining activities and mine waste poses a serious threat to the environment (e.g., land degradation and water contamination; Gurung et al., 2017; Anderson and Butler, 2017). If initial geoenvironmental sampling and ensuing AMD characterisation work is insufficient to inadequately predict risk, the costs of rehabilitation work (e.g., earthworks, capping, sealing, water management, topsoil replacement, fertilising) will far exceed the costs of thoroughly performing early life-of-mine (LOM) assessments (Pepper et al., 2014). For example, the minimum amount to start the earthmoving processes is estimated at AUD $20 per tonne (equivalent to 1 m3) and further, essential neutralising materials are costly to import (e.g., AUS $151‚ÄövÑvÆ350 per tonne acid neutralised; Taylor, 2005; Barker, 2015). The industry requires an updated approach for managing mine waste that considers and integrates all geoenvironmental properties of the gangue materials. Two of the most significant factors influencing contaminant release are the mineralogy and textural arrangement of these minerals in mine waste, so it is therefore critical to evaluate these for adequate geoenvironmental prediction. Although industry standard geoenvironmental practices (e.g., AMIRA, 2002; MEND 2009 guidelines; ASTM methods (i.e., D5744-13e1, D6234-13)) require the discrimination of acid-forming and acid-neutralising wastes as part of mine planning assessments, they do not mandate the routine assessment of mineralogy and texture. Instead, they rely heavily on chemical static tests that use sample powders (i.e., < 63 ˜í¬¿m or 125 ˜í¬¿m) whereby no textural features are preserved. Therefore, only limited understanding of how the materials may behave in a waste repository is ascertained. Further, these tests can produce erroneous results. For example, material with paste pH values of < 5 may contain acidic SO\\(_4\\)\\(^2\\)- salts that release acidity during ANC testing thereby obscuring part of the ANC and biasing NAPP calculations. Variations in NAG test pH values can also result from carbonate disequilibrium with atmospheric CO\\(_2\\) during the cooling phase resulting in excessively basic NAG pH test results (Weber et al., 2004; Charles et al., 2015). Such testing needs to be supplemented by mineralogical and textural data to adequately determine the geoenvironmental properties of future waste materials. By using new technologies (e.g., Corescan\\(^¬¨vÜ\\)) and optimising existing tools (i.e., laser ablation inductively coupled plasma mass spectroscopy (LA-ICPMS), micro X-ray fluorescence (˜í¬¿XRF)), detailed geoenvironmental investigations can be undertaken at the exploration stages of a mining project, allowing for deposit-wide geoenvironmental domaining. For example, long-term advantages for strategic mine planning can be realised if neutralising and capping materials are readily identified and can be correctly stored and used at the time of mine closure. The importance of geoenvironmental characterisation continues to be increasingly recognised by mine operators across the LOM. Therefore, evaluating the application of unconventional analytical techniques and how they can be integrated into deposit-scale mineralogical domaining is vital to promote innovation in this sector. This research aimed to explore how new mineralogical tools can be optimised to ensure early detection of AMD and to accurately identify the location, abundance, and chemistry of neutralising materials to support waste domaining early in the LOM (e.g., at exploration stages). Many new technologies now exist that enable rapid analyses of mineralogy and mineral chemistry of drill core, (e.g., laser Raman, LA-ICPMS, laser induced breakdown spectroscopy (LIBS), prompt gamma neutron activation analysis (PGNAA) and micro computed tomography (˜í¬¿CT)). Several technologies were selected for use in this research and applied to a suite of drill core materials (n = 100) from the Cadia East alkalic gold-copper porphyry and the Little Cadia skarn from the Cadia Valley Operations (CVO), New South Wales, Australia. The geoenvironmental characterisation practices of this site are typical of many mine sites in Australia in that waste properties are determined using total sulfur (%) and pH testing values. By relying only on these two tests, minerals and textural relationships are not characterised, which resulted in classification inconsistencies and expected behaviour of mine waste documented in several industry reports have been challenged (EGi 2002, 2010 and 2011). To supplement desk-based studies focused on understanding waste properties, the drill core assessment techniques presented in this thesis are intended to be the first pass tools to use when defining geoenvironmental properties. In this research, shortwave ultraviolet fluorescence (SWUV), long-wave ultraviolet fluorescence (LWUV), and carbonate staining techniques are used, along side the acid rock drainage index (ARDI) on drill core materials (n = 100) with these results screened against bulk mineralogy and ABA classifications. Carbonate staining significantly improves identification of calcite producing a distinguishable pink to red stain on calcite. SWUV fluoresces carbonates and feldspars a range of colours (blue, blue-green, red, pink) while in LWUV carbonates fluoresce pink to red only, making this the preferred analytical mode for carbonate identification. A new python code for quantifying this carbonate UV response was developed and validated against calcite measured by XRD and demonstrated strong positive correlation (R\\(^2\\) = 0.82). ARDI assessments defined the potassic alteration group as the most acid forming and the skarn alteration group to have the highest neutralising capacity potential. Overall, ARDI values are lower than expected (max. 34/50) due to abundant carbonate present throughout the deposit. Geochemical models plotted using multi-element assay data supports visual ARDI drill core logging, confirming that the potassic alteration group has the lowest neutralising potential. A more extensive element suite was collected using a field portable X-Ray fluorescence (pXRF) instrument and shows Al, Cu, Fe and Zn were consistently high (maximum 40 %) and As, Cd, Hg, Ni, and Pb were low (maximum 400 ppm). Portable XRF element suits shows the propylitic and potassic alteration groups hold the largest reserves for potential metal leaching upon weathering. In comparison to published geoenvironmental ore deposit models (e.g., Cox et al., 1986; Plumlee, 1999; Silitoe, 2010) pXRF elemental concentrations show that samples from Cadia East are below reported averages for other porphyry deposits. Collectively, these tools allow for effective forecasting of waste properties (i.e., SWUV, LWUV, ARDI, pXRF). Hyperspectral scanning data for geoenvironmental domaining using the Corescan\\(^¬¨vÜ\\) HCl-3 instrument provides classified mineralogy data using SWIR with data viewed in supporting proprietary software Coreshed. By scanning and viewing classified mineralogy data in this manner, there are opportunities to improve geoenvironmental sampling as several kilometres of drill core can be viewed easily allowing segregation of mineralogy into geoenvironmental domains based on mineralogy and mesotexture, thus supporting comprehensive sample selection. Using this information to inform geoenvironmental sampling improves confidence in the reproducibility and the representivity of samples collected for ABA testing. Using hyperspectral data, a new geoenvironmental domaining algorithm, termed the geoenvironmental domaining index (GDI) is developed in this thesis. The GDI inputs include the relative mineral abundances calculated from the raw data output from the HCl-3 instrument, absolute neutralising potential or acid forming potential values (Jambor et al., 2007; Parbhakar-Fox and Lottermoser, 2014), and published relative reactivity values at pH 5 (Sverdrup, 1990). Calculated Corescan\\(^¬¨vÜ\\) relative mineral abundances were validated against semi-quantitative XRD and calculated calcite from total carbon (measured in percent). A comparison of calcite abundances measured in by these techniques highlights that Corescan\\(^¬¨vÜ\\) data is sufficient for geoenvironmental domaining. Two versions of the GDI are developed and an appropriate selection is made based on the quality of the available Corescan data. For the Cadia East study, three risk categories (i.e., potential risk, low risk, and potential acid neutralising capacity (PNC)) are proposed based on GDI values and correlating bulk mineralogy and ABA results. Comparing GDI values to geochemical and mineralogical data shows neutralising characteristics of differing alteration types can be predicted accurately from hyperspectral data. Furthermore, GDI values in conjunction with NAG pH or total sulfur can help identify neutralising and acid forming zones. An additional algorithm to automatically perform ARDI calculations (termed the A-ARDI) using hyperspectral data and the associated RGB images has also been developed, and when used in conjunction with the GDI, can facilitate accurate classification of future waste materials. Although Corescan\\(^¬¨vÜ\\) data in isolation can provide an insight into neutralising or acid forming charactersistics, it does not provide information on the potential metal leaching issues that may be associated with waste. Therefore, the opportunity to integrate trace-element chemistry data produced by LA-ICP-MS on unpolished drill core with quantitative mineralogy from Corescan\\(...