Statistics Seminars: Bayesian uncertainty analysis for natural hazards modelled by computer simulators
24 February 2014 14:10 in CM221
Natural hazards are frequently modelled by computer simulators. There are many sources of uncertainty in using such a simulator: examples include the choice of parameters at which to run the simulator, the uncertainties in the forcing functions, and the inherent differences between the simulator and reality. This talk will outline a Bayesian framework for quantifying these uncertainties, involving emulation of the simulator, experiments on the simulator to estimate "internal" model discrepancy, and careful assessment of "external" discrepancies. Examples will be taken from the PURE (Probability, Uncertainty, and Risk in the Environment) Programme, with particular focus on flood modelling.
Contact email@example.com for more information