The school will provide advanced training in risk and uncertainty in natural hazards from the Cabot Institute's leading academics plus some special guest lecturers. Afternoons will include hands-on exercises.
The school is open to postgraduates, early career researchers and scientists from industry and government agencies.
Introduction to uncertainty - Professor Keith Beven
An introduction to the issues underlying uncertainty estimation. An R practical on generation of random variables and joint variables using copulas. Finishing with an example GLUE application.
Sensitivity analysis - Professor Thorsten Wagener
Sensitivity analysis investigates how the uncertainty in the model output can be apportioned to the uncertainty in the model input (including its parameters). We will introduce the most widely used approaches to sensitivity analysis and provide hands-on applications of these methods to natural hazard models of varying complexity.
Expert elicitation - Professor Willy Aspinall
This session covers background to the use of scientific experts' opinions in decision support; concept of Cooke's Classical Model for determining performance-based weights and differential pooling of opinions from a group of experts, with strong emphasis on the expression of uncertainty estimates; principles for obtaining a "rational consensus". Also described will be a complementary approach for eliciting qualitative rankings and preferences by a paired comparison approach coupled with probabilistic inversion to produce ranking metrics and consistency checks.
Stochastic modelling - Professor David Stephenson
Stochastic processes such as the Poisson point process can be incredibly useful for developing insight and quantifying the behaviour of hazards. In this session we will explore stochastic process modelling relevant to natural hazards, and show how stochastic processes can be used to quantify the clustering of hazards and rate dependence on time-varying factors. The concepts will be illustrated by application to European windstorms – which are well known to cluster (e.g. winter 2014!).
Calibrate your model - Dr Jonty Rougier
In this session we explore the principles of model calibration, and simple tools for the same: space-filling designs (eg latin hypercubes), visualisation with parallel coordinate plots, dealing with multivariate outputs (principal variables analysis), ruling out regions of parameter space (history matching), proceeding sequentially.
Postgraduate students: £300
The cost includes all course materials & lunch and your place on the course for the week.
Accommodation is not provided but we have a list of suitable venues for all budgets (PDF, 198kB). Please note the University does not endorse these venues.
You can register for this summer school by filling in the online application form. You will be notified by email if you have been successful, at which point we will send you the details to pay for your place.
This is always a popular summer school so we advise booking early to avoid disappointment as spaces are limited.
Deadline for applications is Monday 23 June 2014.
This summer school is being held in conjunction with the Cabot Institute, the University of Bristol’s environmental research institute and CREDIBLE, a NERC funded project which aims to better assess uncertainty and risk of natural hazards. Please contact firstname.lastname@example.org if you have any queries about the school.