Testing probabilistic adaptive real-time flood forecasting models

Operational flood forecasting has become a complex and multifaceted task, increasingly being treated in probabilistic ways to allow for the inherent uncertainties in the forecasting process. This paper reviews recent applications of data-based mechanistic (DBM) models within the operational UK National Flood Forecasting System. The position of DBM models in the forecasting chain is considered along with their offline calibration and validation. The online adaptive implementation with assimilation of water level information as used for forecasting is outlined. Two example applications based upon UK locations where severe flooding has occurred, the River Eden at Carlisle and River Severn at Shrewsbury, are presented.