Venue: United College Quadrangle/School VI - Map
Title:
Copula-based Data Fusion
Invited Speaker:
Dr Stephanie Vogl, Institute of Meteorology and Climate Research – Atmospheric Environmental Research, Karlsruhe Institute of Technology, Germany
Bio:
Stephanie Vogl is a researcher at the Institute of Meteorology and Climate Research – Atmospheric Environmental Research at the Karlsruhe Institute of Technology and the University of Augsburg, Germany. She has a PhD from the Ludwig-Maximilians University in Munich on the topic of “Hurricane boundary layer models” and a previous degree in Mathematics from Regensburg University. Her broad research interests include Tropical Cyclones, Boundary Layer Meteorology and Dynamic Meteorology. She is currently working on the following topics:
- Local refinement and bias-correction of regional climate scenarios
- Copula based data assimilation and bias-correction
- Assimilation of radar, gauge and microwave link attenuation to derive precipitation fields
Abstract:
Many applications in geoscience are ill posed inverse problems where the lack of qualified information in the space or time domain leeds to tremendous uncertainties in the reconstructed quantities. Satellite observations e.g. for land use or soil moisture have too coarse spatio-temporal resolution to allow for a high resolution in the reconstructed fields and observations of meteorological quantities such as air temperature, humidity or precipitation are traditionally only available at specific station locations. Whenever different measurement devices observe the same quantity the question rises how these measurements can be merged so that their specific strengths are preserved. To bridge the gap between observations and the desired spatio-temporal resolution and to merge different observation types, the dependence structures between multiple variables or observations at multiple locations can be used as an additional constraint. As these dependencies can be strongly non-Gaussian and complex, standard correlationbased methods are not able to fully describe them. Therefore Copulas, i.e. specific multivariate distribution functions, are a promising tool as they are suitable to model non-linear behaviour. They further allow for revealing hidden interlinks and can be used to additionally include this valuable stochastic information. The added value of Copulas is shown in two examples: first for the local refinement of precipitation fields modelled by a regional climate model (RCM) and second for an improved estimation of modelled soil moisture fields by assimilating observation data.
