Menu









SC8

The Role of Emulation in Managing Uncertainty in Complex Geophysical Models
Convener: Dan Cornford 
Wed, 06 Apr, 15:30–17:00 / Room 25
Add this session to your Personal programme

This short course will explore the role of emulation in managing uncertainty in geophysical models. It will provide an introduction to the concept of emulation, which is the creation of a surrogate statistical model to act in place of the geophysical simulator, and describe the state of the art in creating emulators, for example how to create emulators for multi-output simulators, as well as the limitations of the existing technology. We will present the typical stages necessary for the construction of an emulator of a geophysical simulator including: problem definition; prior elicitation; screening and experimental design; emulator learning and emulator validation. We will address the issue of using the emulator for uncertainty analysis and sensitivity analysis in a Bayesian framework. Using examples from the MUCM project (http://www.mucm.ac.uk) we will describe the practical implementation of emulation methods, linking to the MUCM Toolkit. By the end of the session we anticipate that attendees would know the potential and limitations of emulation, and have sufficient knowledge to start applying emulation to their simulators, using the freely available MUCM Toolkit for guidance.