Statistics Seminars: Uncertainty quantification of geophysical models
9 December 2013 14:00 in CM221
In this talk, we show various strategies for the calibration and emulation of simulators having uncertain inputs and internal parameters, with applications to tsunami wave models and climate models. For a simple landslide-generated tsunami model, a fast surrogate is provided by the outer product emulator. It can enable either real-time warnings according to uncertain speed, position and shape of the landslide, or full uncertainty quantification for hazard assessment. We then show some new realistic simulations for earthquake-generated tsunamis in Cascadia (Western Canada and USA), using VOLNA. VOLNA is a solver of nonlinear shallow water equations on unstructured meshes that is now accelerated on the GPU system Emerald. Strategies for hazard assessment are discussed. The propagation of uncertainties in the bathymetry to maximum wave heights are also illustrated using VOLNA. Enhancements in terms of sequential design of computer experiments, where points in the design are chosen according to a new adaptive strategy, are shown and compared to current approaches in terms of statistical and computational efficiencies. Finally, we present a first attempt to calibrate the gravity-wave parameterizations in the US Community Earth System Model, using spherical representations for dimension reduction.
Contact email@example.com for more information