Tuesday 15 Oct 2013: Understanding and Improving Model Predictive Performance in Water Systems
Dr Chris Hutton - University of Exeter
H 170 14:00-15:00
Models are widely applied to improve understanding of both natural and manmade water systems. However, the necessity of deriving a simpler representation of reality in mathematical form will lead to inevitable differences between what the model predicts and what really occurs. Such differences pose problems for how we use models to draw inferences about the world, and in turn how we react to or seek to influence system behaviour in light of model predictions. This presentation explores how predictive performance might be quantified and improved in model development, model calibration and real-time application, each explored in a different water system researched in CWS. First, in model development the issue of sub-grid scale parameterisation in overland flow and erosion modelling is considered, and a method presented to improve the process representation and predictive performance of coarser resolution models that for computational reasons often need to be applied at catchment scales. Second, in calibration formal and informal Bayesian approaches are compared as a means to quantify uncertainty and improve Water Distribution System Model predictions. Finally, the application of data assimilation methods for improving the predictive performance of urban rainfall runoff models in real-time is presented.
Chris is a Research Fellow in the Centre for Water Systems here at Exeter, and is currently working on two EC FP7 projects: PREPARED and iWIDGET. Chrisí primary research interests focus on methods for improving the robustness of model predictions, through model development, and quantification and reduction of uncertainty, both in calibration and in real-time application. In 2006 Chris graduated with a first class degree in Geography from the University of Sheffield and moved to the University of Exeter to take up an Exeter Graduate Fellowship (joint teaching and research) within the Geography department. In 2010 Chris completed his PhD, which concerned the application of numerical models to simulate catchment processes in semi-arid environments.