b'AEGC 2023Short abstractsused to indicate the occurrence and to define the signature ofensembles of subsurface models, provide a tool for porphyry systems around this district. investigating the range of plausible models which can fit a given data set. Such petrophysical model ensembles can Geophysical Resources and Services (GRS) was commissionedbe difficult to interpret directly and to relate to geological by Codelco to acquire MIMDAS IP, and MT resistivity at severalentities. To properly utilise posterior model ensembles, of these prospects. The results presented herein are from theit is becoming increasingly common to extract and map Ministro Hales (MH) deposit. specific features of interest from them using post-processing The information that could be extracted from the geophysicalalgorithms. Application of machine learning methods such datasets available offered valuable knowledge about theas supervised and unsupervised classification to posterior geological setting, for instance the extent of the post- model ensembles in this role is an active area of research mineralisation cover and the location of the west fault,and promises to speed up workflows and reduce reliance nevertheless, IP data stands alone in presenting a directon subjective parameter specification. We argue that response to the deposit. Chargeability inversion results showingthese algorithms are currently typically not applied in an a strong relation with the mineralised ore. theoretically justified way because they use ensembles representing posterior uncertainty as if their members are In this work, we will focus on the importance of IP data to theactual data points. If not handled carefully, this can lead to MH mine with a brief review of the geological framework insignificant over representation of how much information is the district. We will also discuss the advantages of the otheractually available. To learn features correctly, we propose geophysical data sets available that the parameters defining them must be formulated as if they were hyper-parameters to the original inversion problem. We introduce a method for learning features from Revisiting legacy AEM data to generate newprovided posterior model ensembles which achieves this exploration opportunities in areas of outcrop or thinand is generally computationally inexpensive and easy to cover. implement. The method can be used as interpretive tool, for mapping selected geological features probabilistically, Andrea Viezzoli and for testing simple hypotheses regarding those features. Aarhus Geophysics To demonstrate, we present both real and synthetic tests mapping resistivity layers and boundaries using The last decade saw a general shift to greenfield explorationmagnetotelluric and airborne electromagnetic data.under thicker and thicker cover, for a series of solid reasons. We argue, however, that there is a still a huge potential inTrusted environmental and geological information areas of outcrop or limited cover, waiting to be unlocked. Suchstratigraphic frameworks for resource and opportunity stems largely from improper assessment of AEM datasets acquired in those areas over the last 20 years. The keyhydrogeological assessments.aspect we refer to, that often hid, or at least blurred that potential,Carmine Wainman, Darren Ferdinando, Stephen Hostetler, is the Induced Polarisation effect in AEM data (AIP). Both rigorousChris Gouramanis, Meredith Orr and Sarlae McAlpinesynthetic experiments and numerous case studies prove that not dealing with IP effects in the AEM data results in erroneousGeoscience Australiaconductivity models. What look like isolated conductors can be, in reality, stratigraphic units that happen to be partially hidden byThe Australian Governments Trusted Environmental and chargeable cover. Deep legit conductors, on the other hand, canGeological Information (TEGI) programme is a collaboration disappear completely or show deeper than they are. The need tobetween Geoscience Australia and CSIRO that aims to provide deal with AIP has now been accepted, at least by the Majors, asaccess to baseline geological and environmental data and a mean to obtain correct conductivities. However, we have notinformation for strategically important geological basins. The seen yet the major revision of legacy data that one could haveinitial geological focus is on the north Bowen, Galilee, Cooper, expected as a consequence of this uptake. Not to be forgotten, theAdavale, and their overlying basins. This paper presents seven chargeability retrieved through AIP modelling, although differentstratigraphic frameworks from these basin regions to enable from the one gathered with galvanic methods, provides a richgroundwater, environmental and resource assessments, identify model complementary to conductivity. Interestingly, IP effects inintervals of resource potential, and to manage associated risks AEM are stronger over outcrop, or in presence of moderately thickto groundwater resources and other environmental assets. The conductive cover; they become negligible for cover thickness construction of stratigraphic frameworks for this programme 100 m, where majority of explorers are focusing their efforts. Thebuilds upon existing lithostratigraphic schemes to the extent public legacy AEM datasets could very well be hiding several low- of the current state of knowledge. The frameworks incorporate hanging opportunities in many of Australians mineral districts. Theplay divisions for resource and hydrogeological assessments. only way to know is to revisit them and model the IP effects. A total of 33 play intervals are defined for the north Bowen, Galilee, Cooper, Adavale, and their overlying basins, using chronostratigraphic principles. Where possible, unconformities Automating feature extraction from model ensemblesand flooding surfaces are designated to define the lower produced by probabilistic geophysical inversion. and upper limits of plays. Data availability and geologic duration are considered in capturing significant changes in Gerhard Visser and Hol Seill gross depositional environments. The results from this work CSIRO enable the consistent assessment of shared play intervals between basins, and also highlight uncertainties in the age and Geophysical inversion is an inherently non-unique problem.correlation of lithostratigraphic units, notably in the Galilee and Bayesian probabilistic inversion methods, which producenorth Bowen BasinsFEBRUARY 2023 PREVIEW 148'