Believe it or not? The challenge of validating large scale probabilistic risk models

The National Flood Risk Assessment (NaFRA) for England and Wales was initially undertaken in 2002 with frequent updates since. NaFRA has become a key source of information on flood risk, informing policy and investment decisions as well as communicating risk to the public and insurers. To make well informed decisions based on these data, users rightfully demand to know the confidence they can place in them. The probability of inundation and associated damage however cannot be validated in the traditional sense, due the rare and random nature of damaging floods and the lack of a long (and widespread) stationary observational record (reflecting not only changes in climate but also the significant changes in land use and flood defence infrastructure that are likely to have occurred). To explore the validity of NaFRA this paper therefore provides a bottom-up qualitative exploration of the potential errors within the supporting methods and data. The paper concludes by underlining the need for further research to understand how to robustly validate probabilistic risk models.