IN040-06
pyrolite: Tools for Data Driven Geochemistry
Tuesday, 15 December 2020: 19:15
Virtual
Morgan J Williams1, Louise Schoneveld1, Laura Miller2, Yajing Mao3, Kaarel Mänd4, Justin Gosses5, Hayden Dalton6, Adam Bath1 and Steve John Barnes1, (1)CSIRO Mineral Resources, Perth, Australia, (2)Monash University, Melbourne, Australia, (3)Institute of Geology and Geophysics, Chinese Academy of Geosciences, Beijing, China, (4)University of Alberta, Edmonton, Canada, (5)Science Application International Corporation Houston, Houston, TX, United States, (6)University of Melbourne, Parkville, Australia
Abstract:
pyrolite is an open-source Python package developed for multivariate geochemical data analysis. The core features of the package are functions for geochemical data processing, transformation, and visualisation. In addition to common geochemical transformations (e.g. converting elements to oxides), pyrolite includes functions for compositional data analysis, enabling more robust statistical analysis of geochemical datasets. The visualisation tools include common diagrams (e.g. ternary, spider and data-density diagrams), plot templates (e.g. TAS diagram) and utilities for dimensionally-reduced visualisations of multivariate geochemical data. The package also includes a number of reference datasets, including for ionic radii, the compositions of some common rock forming minerals, and large-scale geochemical reservoirs. Extensions to the core package are being developed for specific use cases, including for linking geochemical datasets to alphaMELTS models (pyrolite-meltsutil).
pyrolite is built upon core data analysis packages within the scientific Python ecosystem, and exposes relevant interfaces such that users can easily integrate pyrolite into existing data analysis and visualisation workflows. This integration provides a solid foundation for prototyping, developing and interlinking tools from data processing through to modelling and machine learning. The documentation includes a gallery of examples and tutorials which demonstrate the use of key features. pyrolite is under active development, and we welcome and encourage questions and all forms of user contributions (e.g. code, documentation, bug reports and feature requests). The package has recently been openly peer-reviewed and published through pyOpenSci and the Journal of Open Source Software (doi: 10.21105/joss.02314); in this presentation we briefly discuss the advantages of this publication pathway for relevant research software projects.
