b'AEGC 2021Short abstracts100: A regional marker tool for the Walloon Coalthat could be related to the komatiite-hosted nickel deposits. Measures (Surat Basin) using detrital zircon U-PbBased on domain knowledge, we design a deep learning geochronology model that systemically combines geophysical and geological features. First, our model generates a feature map and initial Dr Pascal Asmussen 1,2, Dr Oliver Gaede2, Dr Thomas Flottmann3prospectivity map using geological data and geophysical and Mr Brett Pidgeon3 worms which could reveal the crustal structures. Next, the model produces a final prospectivity map that delineates 1 University of Queensland potential komatiite-hosted nickel deposits using whole data 2 Queensland University of Technology including geophysics. The model is trained with the locations of 3 Origin Energy the known nickel deposits.The coal seams within the Walloon Coal Measures (WCM,We divide the Yilgarn craton area into a train and test region to Surat Basin, Australia) are a significant target for Coal Seamvalidate our model. We adopt the AUC score and prospectivity Gas (CSG) production and the identification of the boundaryscore percentile of known deposits to evaluate our model in with the overlying Springbok Sandstone (SST) is critical to thevarious aspects. Our model achieved a high AUC score and stratigraphic framework of the basin, as well as the identificationpercentile score and it can be efficiently used for early-stage of groundwater flow units potentially associated with thisnickel exploration. The suggested workflow could be applied boundary. Identification of the contact on a local (tenement) andto the exploration of the other mineral types with a slight regional (basin) scale between the SST and WCM is challengedmodification reflecting the characteristics of the mineralizations.by (i) overall lithological similarity of the units (macro- and microscopically, whole rock geochemistry, wireline signature), (ii) lateral heterogeneity of lithologies on a regional scale due103: An automated system for preventing hydraulic to lateral facies and depositional environment variation in thevibrator tip-over during land seismic surveysuppermost WCM and lowermost SST and (iii) lack of consistent regional stratigraphic markers like erosion surfaces or correlatedMr Richard Barnwell 1, Mr Damien Barry1, Miss Megan tuff beds. Sediment provenance analysis via detrital zircon U-PbNightingale1, Dr Tim Dean2, Mr Hagay Haviv2 and Mr John Giles3geochronology allows to constrain spatial and temporal changes1 Terrex Seismicin sediment provenance by sampling across stratigraphic units2 BHPalong a lateral transect. Here we present the results of detrital3 Seismic Sourcezircon distributions from 27 samples that were sampled from four wells along a 300 km transect from the eastern EromangaHydraulic vibrators are the preferred source for land seismic Basin to easternmost extent of the Surat Basin. Our data show asurveys. The energy from the vibrators signal (commonly referred brief pulse of basement derived material at the base of the SSTto as a sweep) is transmitted into the ground via a baseplate that which is interrupting the volcaniclastic dominated sedimentis coupled to the surface by loading it with the weight of the provenance in the SST and WCM. This brief change in zirconvehicle. The baseplate is lowered into position by a stilt structure provenance is interpreted as a precursor to fluvial dominatedwhich continues to apply hydraulic pressure to ensure that the environments in the SST and can be utilised as a regional markerbaseplate does not become decoupled from the surface during across the Surat and Eromanga basin, suggesting a potentialthe sweep (the ground beneath the baseplate often becomes change in source provenance and basin dynamics associatedslightly compacted). If the vibrator is operated on a slope, or if with the SSTWCM boundary. the ground gives way in a significant and uneven manner, then the vibrator can continue to extend the stilt structure which in 102: Knowledge-guided machine learning forextreme cases can cause the vibrator to tip-over.komatiite-hosted nickel prospectivity mapping In this paper we describe a system developed by Terrex Mr Minsu Kwon 1 and Seismic Source, consisting of two inclinometers and an articulation sensor that can be fitted to a vibrator to 1 ENERZAi automatically retract the baseplate when the vibrator becomes overly-tilted. The system was recently deployed on a large Thanks to the brilliant progress in machine learning, many3D survey consisting of over 100,000 source points. Of these research works have conducted data-driven mineralsource points, 171 were skipped due to the tilt of the vibrator prospectivity mapping. However, it is challenging toexceeding the safe limit. Analysis of a LiDAR dataset showed integrate highly multidisciplinary geoscientific data withthat of these 171 points, less than 3.5% of crossline and 3.1% of machine learning algorithms. Especially, geological data areinline dips exceeded this safe limit. Indicating that if the LiDAR heterogeneous and non-numerical even though they are crucialdata had been utilised during the planning phase of the survey for mineral exploration. the majority of these positions would not have been excluded.In this work, we introduce how to preprocess the geoscientificThe success of the system on this initial survey has led to it data and design a machine learning model based on knowledgebeing adopted throughout the Terrex fleetin order to make the best use of both geoscientific information and the advantages of machine learning. We focus on the region-scale prospectivity mapping for the komatiite-hosted107: A brief history of 3D seismic acquisition in BHP nickel in Yilgarn craton, Western Australia. We extract secondCoal and some predictions for the futureand third-order features from geophysical data to enable machine learning models to capture various patterns of mineralDr Tim Dean 1, Mrs Margarita Pavlova1 and Mr Matt Grantdeposits. In terms of geology, faults, interpreted geology, and isotopic mapping data are converted into numerical features1 BHPAUGUST 2021 PREVIEW 78'