Ruffling feathers: An appraisal of tail flood losses using grey swan scenarios in London, UK

There are several tools for assessing potential future insurance flood losses in the UK, including catastrophe models which seek to generate an annualised view of flood risk losses. These catastrophe models include plausible high impact and low frequency flood events in their stochastic event sets. The addition of events which are generally considered implausible, or grey swan scenarios, is useful to increase understanding of how re/insurers will perform should our understanding of what is plausible be incorrect.

The Thames Barrier has high levels of redundancy by design and it is generally considered implausible that the barrier would completely fail to operate. We propose three increasingly extreme scenarios for flooding in London as a consequence of the Thames Barrier and other defences across London failing. In all scenarios we assume a 1 in 250-year water level from coastal flooding, well within the standard of protection offered by defences through the city.

The following defence failure scenarios are then modelled using a coupled 1D-2D model: 1) the Thames Barrier is open but the river defences remain intact with only overtopping occurring; 2) the Thames Barrier is open and defences are breached upstream of the barrier; and 3) a worst case scenario composite of several flood event scenarios, where for upstream reaches of the barrier, breach and overtopping occur with the barrier open and for downstream reaches, breach and overtopping occur with the barrier closed.

JBA’s catastrophe model for the UK probabilistically models loss from river, surface water and coastal flooding. The model comprises 2D hydraulic modelled hazard maps at 5 metre resolution, a stochastic event set of 106,424 events generated from extreme value statistical analysis, and detailed vulnerability data derived from the Multi-Coloured Manual. The catastrophe model includes an occurrence exceedance probability curve for insurable residential properties, providing the wider context for estimating the loss return period of the scenario events. We present the modelled losses and the estimated loss return periods for the grey swan scenarios and make available the model for re/insurers for stress testing. The loss return periods for the three scenarios are: 1/50, 1/358, and 1/8813.