The Value of Global Sensitivity Analysis for Diagnostic Evaluation and Uncertainty Quantification in Environmental Modelling
Thursday 15 September 2016 at 1.15pm
Room 8 West 2.5, University of Bath (Location and maps)
Computer models are a cornerstone of scientific exploration and practical decision-making in engineering, environmental science, ecology and earth systems modelling. Our group develops tools, theory and guidance for the use of Global Sensitivity Analysis (GSA) to analyse such computer models. I will present examples of how GSA can be used for (1) the diagnostic evaluation of hydrological models – how can we improve our model structures? (2) For understanding future landslide risk – how can we provide useful guidance for decision makers in the presence of unknown uncertainties? And (3) for quantifying the space-time varying uncertainty in flood inundation modelling – how can we understand the complex interactions of different sources of uncertainty? These examples highlight the variety of uses for GSA in the context of building and using models across a wide range of application areas.
The examples shown have been implemented using the SAFE Toolbox that integrates key GSA methods in a single environment. Free Matlab and R versions of the SAFE Toolbox for Global Sensitivity Analysis are available.
Co-authors: Francesca Pianosi, Fanny Sarrazin, Susana Almeida, Liz Holcombe