Resampling (detrital) zircon age distributions for accurate multidimensional scaling solutions

Abstract

Multidimensional scaling (MDS) has become an important tool in detrital zircon geochronology for discriminating and comparing multiple samples, especially where large datasets are concerned. However, the non-parametric test statistics used in MDS (e.g., Kolmogorov–Smirnov [K-S] or Kuiper) to calculate differences between detrital zircon samples are sensitive to differences in grain age uncertainty, for example, when comparing compilations of U-Pb data produced via different analytical techniques. We illustrate the effect of different age uncertainties on comparison metrics and MDS analysis using synthetic populations. Additionally, we introduce two methods that reduce these limitations: an error resampling MDS (ERM) technique and bootstrap resampling with Procrustes rotation (BoPR) technique. These new resampling techniques reduce limitations associated with MDS. We demonstrate the utility of these new tools by comparing detrital zircon samples from the Mesoproterozoic Roper Group of Northern Australia with nearby potential source regions. Mitigating the limitations of MDS, and the underlying statistics by employing resampling techniques makes MDSa more potent means of addressing source to sink relationships in detrital zircon studies.