A multi-site methodology for understanding dependencies in flood risk exposure in the UK

Abstract

Recent large scale flood events in the UK and the continued threat of a major North Sea surge have motivated a re-appraisal of how flood risk is modelled. A new generation of flood risk models are starting to consider the spatial and temporal dependencies in flood events. This is important for a wide range of risk based decision making, with one of its most significant applications being the understanding of insurance exposure. The aim of this thesis is to increase understanding of flood risk exposure in the UK and identify areas where existing modelling capabilities and data limitations contribute to large uncertainties in the estimation of risk. Illustrating a successful collaboration between academia and the insurance industry, a case study of one company’s exposure from static caravans is used to develop a methodology for flood risk assessment at multiple sites nested within a national framework. This novel nested approach allows for greater detail to be included at sites of interest resulting in increased understanding of the risk driving processes while retaining the large scale dependence structure. This is demonstrated at high risk locations on the Lincolnshire and North Wales coastline and inland on the Rivers Severn and Thames. The proposed methodology takes a flexible component based approach and has potential adaptations to different receptors and end users. A systems based model is used which explicitly considers all key components of risk. Extreme fluvial and coastal events are modelled statistically using the conditional dependence model of Heffernan and Tawn (2004). Coastal flood defences are essential for the protection of static caravan sites however their inclusion in existing risk models contributes significant uncertainties. The quality of data available on flood defence heights is reviewed and a methodology to incorporate spatial variations is proposed. The failure of flood defences is modelled using fragility curves and inundation modelling is used to route water on the floodplain. Finally the damage to the static caravans is modelled using depth-damage curves with reference to the impact of limited observed data on flood damage for caravans. One of the biggest challenges of considering dependencies across multiple scales within a systems model is matching the data requirements across each component. To address this problem this thesis investigates the relationship between skew surge and wave height to estimate the total inshore water level, and develops a UK specific method to transform daily mean flow to peak flow. The modular structure of the proposed methodology means different component models can be used to suit the available data; here the integration of both 1D and 2D floodplain inundation models is demonstrated.