Reading, Anya1,2,3, Morse, Peter2,3, Staal, Tobias2,3
1School of Natural Sciences (Physics), University of Tasmania, Hobart, Australia; 2School of Natural Sciences (Earth Sciences, CODES), University of Tasmania, Hobart, Australia; 3Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Australia
Parallel Coordinate Visualisations (PCVs) are a powerful means of exploring multivariate, high-dimensional, numerical datasets. Their great strength is the ability of such plots to reveal relational patterns in data that would be hard to identify using conventional geostatistical tools such as scatter plots. Although high-dimensional datasets are commonplace in the geosciences, PCVs have had limited previous usage for addressing questions related to the Earth and environment. Practical considerations such as the ordering of axes, and the need to change and experiment with such ordering, leads to difficulties in usage with PCVs as static plots. A productive way forward is to make use of interactivity, now enabled in many plotting environments. Thus, PCV plots can be handled in a dynamic way, enabling the exploration of data through high-dimensions as originally intended.
This presentation illustrates the process of knowledge generation using PCVs for geoscience examples at two scales. In the first example, a dataset produced by a rock sample map generated by Laser Ablation ICP-MS analysis is explored. We show that while some patterns shown in the PCV are evident in the maps for individual masses (elements) the PCV also reveals previously unseen patterns. In the second example, a large scale dataset, a multilayer geophysical compilation from Antarctica relational patterns are shown that can be taken forward from this reconnaissance analysis to more detailed study by conventional means. Interactive PCV plots may be manipulated on a conventional desktop display screen, however, extended reality (XR, immersive) platforms may also be used. We review the benefits of using this technology, and in particular the types of datasets for which the (relatively small) additional investment might be useful. We find that interactivity is vital to the successful use of PCVs. Up to 3-4 most evident data relations are readily revealed by interactive desktop displays and the XR platforms enable reconnaissance of up to twice that number of relational patterns, should they be present in the data. In summary, PCVs are a useful addition to the geoscientific tool box provided the analyst can make use of an interactive desktop display. The use of XR may be indicated if there is a need to identify more subtle relations in the presence of other dominant relational patterns.
Anya Reading leads the Compute Earth / Physical Sciences Group at the School of Natural Sciences (Physics), University of Tasmania as Professor of Geophysics.
Reading’s innovative approaches to data science build on a foundation of experimental field seismology in challenging regions such as Antarctica and outback Australia.