b'AEGC 2021Short abstractsBHP Coal acquired their first 3D seismic survey in 1998 at theGeophysical signatures in and around iron oxide-copper-Goonyella mine site. The survey covered an area of 1.6 km andgold (IOCG) deposits are frequently associated with changes utilised a 360 channel cabled recording system and an explosivein redox. IOCG systems, in a geophysical sense, include source to record 15 fold data. Since then BHP has acquired ana spectrum of styles from reduced, pyrrhotite-dominant additional 25 3D surveys across its various coal assets. examples, sometimes referred as iron sulphide copper-gold In this paper we show how 3D seismic surveys have changed(ISCG) deposits, to intermediate, magnetite-pyrite dominant, over the years, both at BHP coal and within the larger contextand more oxidised, hematite dominant lithologies. Magnetite-of general developments in land seismic technology. Suchrich IOCG deposits, such as Osborne are relatively geophysically developments include the move from arrays to point receivers,simple, commonly displaying coincident reduced to pole (RTP) from explosive to Vibroseis sources, and from coarse geometriesmagnetic and gravity anomalies. In pyrrhotite-rich systems aimed at defining structure, to higher-density geometriessuch as Eloise, remanent magnetisation may cause offsetting giving far higher-resolution results as well as being suitable forof the RTP magnetic anomaly relative to the associated gravity more advanced quantitative analysis. As well as discussing theanomaly. In hematite-rich systems such as Starra the magnetic changes in acquisition parameters over the years we also showsignature may be subdued or even absent where bulk hematite data comparisons that show their impact. is of primary, hypogene origin. Whilst these basic geophysical principles are accepted, the recognition of redox gradients We finish by discussing the latest advances in seismicwithin mineral systems is more meaningful and can be related acquisition and how we see these improving Coal 3D seismicto geochemical processes involved during mineralisation. surveys in the future. Redox gradients have predictable geophysical signatures, but direct targeting of redox gradients using integrated 108: Automated fracture detection andpetrophysical and geophysical methods still is not common in IOCG exploration.characterisation from unwrapped drill-core images using Mask R-CNN Systematic, scale integrated petrophysical data (including density, magnetic susceptibility, remanent magnetisation, Miss Fatimah Al-zubaidi 1, Mr Patrick Makuluni1, Dr Stuartradiometrics and conductivity), mineralogical, and Clark2, Dr Jan Erik Lie3, A/Prof Peyman Mostaghimi1 and A/Profgeochemical data were collected from twenty-three deposits Ryan Armstrong1 and prospects as part of the Cloncurry METAL project. The data from numerous sites show clear relationships 1 School of Minerals and Energy Resources, University of Newbetween redox gradients, indicated by transitions between South Wales, Sydney, Australia pyrrhotite, magnetite and hematite bearing lithologies, and 2 UNSW steep gradients in magnetic susceptibility and/or remanent 3 Lundin Energy Norway AS, Lysaker, Norway magnetisation. Moreover, coincident Uranium-radiometric Drill cores provide the most reliable fracture information inspikes were shown to highlight zones of complex mineralogy subsurface formations as they present a clear and direct viewrelated to transitional redox. This diagnostic radiometric of fractures. Core observation and image log interpretation aresignature together with coincident magnetic gradients usually integrated for fracture analysis of underground layers.provides a more rigid geophysical proxy for differentiation of There has been a strong move towards developing automatedIOCG related signatures from false positives. Finally, the data fracture detection methods, however, the focus has been onprovides new insights into economic IOCG mineralisation and extracting fracture information from log images, such as acousticnew tools for targeting IOCG and related deposits in both or resistivity image logs. Such efforts using core images aresurface geophysical and downhole datasets.significantly less. This study presents a machine learning-based approach for automatic fracture recognition from unwrapped drill-core images. The proposed method applies a state-of-the- 111: Environmental site assessment at a mining art convolutional neural network for object identification andoperation in Western Australia using the Loupe TEM segmentation. The study also investigates the feasibility of usingprofiling systemsynthetic fracture images for training the model by creating two types of synthetic data using masks of real fractures and creatingDr Remkevan Dam 1, Mr Greg Maude1, Mr Brendan Ray1, Mr sinusoidal shaped fractures. The trained model is then used toGraham Jenke1 and Mr Russell Mortimer1detect fractures in real core images from two different boreholes and achieved a precision of 94.80%. The identified fractures1 Southern Geoscience Consultants (SGC)are further analyzed and compared to manually segmentedThe Loupe TEM is a transient electromagnetic profiling system fractures in terms of fracture dip angle and dip direction,for efficient mapping of near-surface ground conductivity. which achieved average absolute errors of 2.18 and 10.58,The system is operated by a two-person crew who carry the respectively. Overall, the study presents a novel application of antransmitter coil and 3-component receiver coils on backpacks. advanced machine learning algorithm for fracture detection andBenefits of the system include the high productivity and the analysis from unwrapped core images. ability to survey in various types of terrain and through dense vegetation.109: Geophysical proxies for redox gradients in IOCGIn mid-2020, we utilised the Loupe TEM system in the Pilbara systems: Cloncurry District, Qld, Australia region of Western Australia to investigate potential fluid Dr James Austin 1, Dr Tobias Schlegel1, Mr Andreas Bjork1 andpathways and contamination around an environmental Dr Clive Foss1 pond. A total of 23.5 km line kilometres of data was acquired on a dense grid with 20 m line spacing. We used a 75 Hz 1 CSIRO Mineral Resources base frequency and stacked 300 transients to improve the 79 PREVIEW AUGUST 2021'