Copula geostatistics – because ‘normal’ isn’t always the best choice

Speaker: Dr Sebastian Hoerning, Research Fellow, The University of Queensland's Centre for Natural Gas

Abstract: Traditional geostatistical techniques such as Kriging or Sequential Gaussian Simulatioe complex environmental systems. A key problem of these approaches is that they are all based on the assumption of multivariate normal (or Gaussian) distributions which, by definition leads to spatial symmetry. Spatial symmetry however is rather rare and thus its assumption is often unrepresentative of the true spatial dependence structure. Spatial copulas are a novel geostatistical tool which go beyond the assumption of multivariate normal distributions. They enable the analysis and subsequent modelling of asymmetric spatial dependence structures which are omnipresent in environmental systems. Improved modelling of these asymmetries leads to more realistic characterisations of the system under study, which in turn can lead to improved modelling capabilities of dependent variables (such as groundwater flow and transport behaviour). This seminar aims to introduce the concept of spatial asymmetry and how copulas can serve as a tool to detect and to model this asymmetry. Examples will be used to visualise spatial asymmetry and to demonstrate that normal isn’t always the best choice in geostatistics.

Bio: Dr Sebastian Hoerning holds a BSc, a MSc and a PhD in Environmental Engineering from the University of Stuttgart, Germany. After finishing his PhD on ‘Process-oriented modelling of spatial random field using copulas’ in 2016, he started as a Research Fellow at the Centre for Natural Gas at The University of Queensland. Sebastian’s research focusses on developing geostatistical techniques and inversion approaches which allow more realistic representations of environmental systems and their dependent variables.