1 BHS Third International Symposium, Managing Consequences of a Changing Global Environment, Newcastle 2010 © British Hydrological Society Adding value to catastrophe models: A multi-site approach to risk assessment for the insurance industry L. Speight 1 , J. Hall 1 , C. Kilsby 1 , P. Kershaw 2 1 School of Civil Engineering and Geosciences, Newcastle University, UK 2 Catlin, London, UK Email: linda.speight@ncl.ac.uk Introduction The focus of this paper is the development of a multi-site based approach to lood risk assessment which considers all possible random factors inluencing lood risk. Over the past decade the importance of taking a risk-based approach to lood management has been realised and new methodologies have been developed covering different national and international contexts (Hall et al., 2003; Apel et al., 2006). However, as yet there is no integrated risk model that incorporates all contributing factors and the spatial and temporal dependencies between them. Risk is deined as the probability of an event multiplied by the consequence, which in this study are the insured losses. Thus far development has mainly focused on luvial looding. The importance of multiple sources of lood risk is acknowledged and consideration is given to including dependencies between luvial and coastal extreme events in the future. Catastrophe modelling Recent large scale lood events in the UK and the continued threat of a major North Sea storm surge have motivated a reappraisal of how well lood risk is estimated by insurance companies. The continued provision of insurance depends on insurance companies being able to accurately predict lood risk. The current means of doing this is through Catastrophe (Cat) models. Cat models are a particular type of process-based model which provide a means of “event-speciic stochastic modelling of highly correlated multi-location loss” (Muir- Wood, 1999, p1). In the UK, Cat models exist for luvial and pluvial looding, wind storm, and North Sea storm surge. There are three main Cat modelling companies; Risk Management Solutions (RMS), AIR Worldwide and EQECAT, who each produce different models for the above. Each of these models produces different estimates of risk due to differences in the model structure and the datasets employed. While these differences are known by the Cat modelling companies, details are not widely available to end users. When considering extremes it is dificult to know which representation of risk is most accurate. Rather, the most suitable representation depends on the particular asset of interest, its geographical location and vulnerability, and the insurance company’s appetite for risk. Therefore, it is useful for an insurance company to increase its understanding of risk outside of the Cat modelling framework. This additional knowledge can then be compared with data produced by the Cat models and used to inform risk-based decision-making regarding the management of insurance policies. Case study The interest in this project is in risk to an insurance portfolio covering a signiicant proportion of the static caravans in the UK, whose total values are in excess of £2bn. The caravans are located across the country and therefore require analysis over large spatial scales. Within the national portfolio there are clusters of known high risk, for example along the Lincolnshire coastline, an area dependent on extensive lood defences. Hence considerable insight can be gained by developing multi-site models for these risk clusters nested within the national framework rather than using a national scale model which restricts the amount of detail included at any one area. Five risk clusters have been selected which typify particular issues within the portfolio and provide broad spatial coverage across the UK; these are shown in Figure 1. Three signiicant issues have been identiied as particularly Abstract The UK is unique in that lood insurance is widely available to property owners. In order to continue to provide insurance, companies need to be able to accurately model and understand risk. The standard means of assessing risk is through Catastrophe models. These are complex process based models which use a mixture of numerical and statistical methods, making it dificult for the end user to fully understand the underlying processes. Using a case study of one company’s exposure from static caravans, we propose a methodology for lood risk assessment at multiple sites nested within a national framework. Following a source-pathway-receptor approach, a mixture of statistical and physically based methods is used in a systems based model which incorporates the most important random processes associated with lood damage. Within the system, meteorological inputs are modelled statistically using a conditional dependence model; the water level, loodplain inundation and damage calculations are deterministic; and the impact of lood defence failure is considered probabilistically. The methodology explicitly couples spatial dependencies between variables affecting lood risk at national and local scales. The output is greater understanding of risk, and the associated uncertainties, which can be used to inform decision-making.