b'Data trends Data trendsfruitful, although some oddities remain.spatial clustering, and it was not possible These oddities mostly disappeared whento discriminate where Wallaroo Group and samples were spatially joined to magneticGRV domains pop up inside neighbouring geological domains of similar age to thedomains (top right Figure1). The stratigraphy (seeFigure 1). unnamed samples (green) are clearly have Spatially the Wandearah Formationhigh density and high ferromagnetism separated easily. The spatial groupingscompared to those in the Gawler Range corresponded with two recognised sub- Volcanics Domain (magenta), and domains, which plotted with overlappingthe famous Hiltiba Suite (yellow) sits but surprisingly distinctive magneticsomewhere in between (bottom panel susceptibility profiles. Samples fromFigure 1). The Wallaroo Group (black) the Moonta Subdomain (cyan) wereand the Donington Suite (cyan) straddle almost exclusively highly ferromagnetic,the whole range and could be treated whereas samples from the Gawler Cratonas one unit for potential field modelling (magenta) were dominated by lowerpurposes, showing individual ranges are paramagnetic values, although therenot always necessary.Tim Keepingwere a similar number of ferromagneticKeep an eye out for GSSAs report on this Associate Editor for geophysicalrocks (see also Figure 1). analysis, as hopefully it will be of value datamanagement and analysis The Gawler Range Volcanics (GRV)in better constraining tenement scale technical-standards@aseg.org.au stratigraphy is a bit too complicated forgeophysical and geological models.Spatial classification of petrophysical dataAt the Geological Survey of South Australia we have been working on a report on drillhole petrophysical data from the Gawler Craton, which includes measurements from Olympic Dam kindly provided by BHP. For the curious, the theoretical average rock in South Australia has density of 2.79 g/cm3 and geometric mean susceptibility of 0.001 SI, teetering between unaltered and Fe-enriched paramagnetic minerals. But, as we are all aware, these values do not reflect random sampling, but the custom of targeting bullseyes in magnetic and gravity data.The density histograms sorted by stratigraphy generally have normal distributions with low variation. Any other distribution would indicate alteration, mineralisation or a need to further split the stratigraphic logs. Magnetic susceptibility histogram distributions are typically wider with large variance and long skew tails or multiple peaks. Less definitive signs than density, but still worth a second look.Lithology and mineralisation logs would obviously achieve more precise categorisation of petrographic samples, but this information is often not digitised in a format that is as easy to sort as one stratigraphic name. Machine learning style clustering of the sample data failsFigure 1.Drillhole locations of petrophysics samples from Wandearah Formation (top left) and Gawler as stratigraphic properties crossover, andRange Volcanics (top right). Colours in the scatter plots (middle and bottom) represent geological domains are not easily separated. Spatial clusteringcoincident with samples. The colours used in the scatter plots do not correspond with the colours used in the seemed like a silly idea but provedlocation diagrams.37 PREVIEW APRIL 2022'