We use cookies to ensure that we give you the best experience on our website. You can change your cookie settings at any time. Otherwise, we'll assume you're OK to continue.

Durham University

Institute of Advanced Study

Past Events

Quantifying Output Uncertainty in Models used for Climatic Change Research Seminar - Environmental Extremes Estimated From Numerical Simulators

2nd February 2012, 13:00 to 14:00, Seminar Room, Institute of Advanced Study, Peter Challenor

This is part of the Quantifying Output Uncertainty in Models used for Climatic Change Research Seminars.

All interested in Climatic Change, whether past, current or potential future, and in the use of models in research in this field should find these seminars of interest. 

ABSTRACT: Environmental extremes are often calculated not from data but from the output of numerical simulators. For example we might be interested in extreme temperatures at 2050 under some future greenhouse gas concentration scenario. Clearly data are not available and we have to rely on numerical simulation to provide ‘data' which can then be analysed to produce estimates of extremes. Other examples include the use of hindcasting models to extend data back in time. Current practice is to pretend that the simulator output is data and to apply methods for the analysis of extremes. Although this methodology produces estimates of the extremes these estimates do not take into account any uncertainty arising from the simulator itself. In this talk we consider the effect of simulator uncertainty on the estimates of extremes. The methods we propose are extensions of those developed for deterministic computer simulators. These use ‘emulators: stochastic processes that are used to estimate the true simulator output. For extremes we need build emulators for stochastic rather than deterministic outputs and for very non-Gaussian distributions. This means that extremal processes may be more appropriate than Gaussian processes for our emulators.

Peter Challenor:  works in the Marine System Modelling group in the National Oceanography Centre, a NERC research laboratory. He is interested generally in environmental statistics and in particular in the statistics of complex numerical simulators, especially in simulators of the ocean and the climate system in general. Peter is a co-I on the Managing Uncertainty in Complex Models project and the PI on the RAPID-RAPIT project looking at the risk of the collapse of the thermohaline circulation.

Contact for more information about this event.