final_program_abstracts[1]

11 IMSC Session Program
An application based approach to modelling the spatial
dependencies of extreme flood risk
Friday - Parallel Session 7
L. Speight 1 , J. Hall 1 , C. Kilsby 1 and P. Kershaw 2
1
School of Civil Engineering and Geosciences, Cassie Building, Newcastle
University, Newcastle upon Tyne, UK
2
Catlin, London, UK
The floods of summer 2007 caused widespread disruption across the UK. Temporary
flood defences could not be deployed and national resources were stretched to their
limits. For those involved in risk assessment and emergency planning it highlighted
the need for greater understanding of extreme events to answer questions such as, “if
town A is flooded, what is the probability of town B also flooding?”, or “what is the
probability of an extreme event affecting more than 50% of the country?”. We
develop an application based approach to modelling the spatial dependencies of
extreme flood risk in the UK to address these issues.
Traditional methods of extreme event analysis recommend the use of peaks over
threshold data. However when looking at large spatial scales it is unlikely that all
gauges in the network will be extreme at the same time. An alternative is to extract
continuous flow data for all gauges when one or more gauges are extreme. Taking an
example gauge with 20 years of 15 minute flow data this equates to over 700,000 data
points to analyse. Scaling this up to a moderate network of gauges soon results in an
unmanageable amount of data. The third option is to use daily mean flow data which
provides daily data for all gauges regardless of the extremeness of the event. This is
the dataset which is used in the following methodology.
Spatial dependencies in extreme flow events are characterised using the Heffernan
and Tawn 1 model for conditional dependence as used in a recent study of flood risk
by Keef et al 2,3 . The methodology also enables simulation of large scale events that
are more extreme than anything that has been observed in the gauged data.
However for risk assessment these results are of limited use since daily mean flow
data is not able to capture the full details of extreme events. In particular the flood
peak, which is likely to be most significant in terms of flood defence overtopping, is
not reproduced. To overcome this problem we use the ratio between daily mean flow
and flood peaks to estimate peak flows.
The strength of this methodology is that it can be used for flood risk assessment at site
specific locations nested within the national framework, for example an electricity
company may wish to know the risk across their network of substations or an
insurance company may wish to know the risk to a particular portfolio. This final
stage of the methodology involves transferred the flow estimates from gauges to sites
of interest while maintaining the spatial distribution of the extreme event.
1 Heffernan, J., and Tawn, J. (2004) A conditional approach for multivariate extreme
values, Journal of the Royal Statistical Society, Series B, 66 (3), p497-546
2 Keef, C., Svensson, C., Tawn, J. (2009) Spatial coherences in extreme river flows
and precipitation for Great Britain, Journal of Hydrology, 378, p240-252
Abstracts 339