In conjunction with Peter Young and the Environmental Systems group, we have long been working on the problems of rainfall estimation and runoff generation in flood prediction problems, including real-time flood forecasting, flood frequency analysis using continuous simulation methods and flood inundation modelling using hydraulic models for forecasting and flood risk planning purposes. In each type of problem, the methods aim to associate predictions with an estimate of uncertainty, including methods to constrain that uncertainty using data assimilation techniques in real-time applications. A flood forecasting system based on stochastic transfer functions based on these concepts was implemented in 1991 to provide warnings for Dumfries in Scotland. A refined methodology is currently being implemented in the UK National Flood Forecasting System in collaboration with WL|Delft Hydraulics.

Work on uncertainties in flood inundation forecasts has been funded by the European Flood Forecasting System and, more recently, by the EPSRC/Defra/EA Flood Risk Management Research Consortium.

Flood frequency estimation by continuous simulation is not limited by historical record lengths but depends on having good models of rainfall events, evapotranspiration and runoff generation processes. Models can be evaluated against the available historical data and, where shown to provide acceptable simulations, be used to make predictions of frequency (including under different climate change scenarios) together with an estimate of the uncertainties in the estimates. This approach has been applied at Lancaster to catchments in both the UK and abroad.

Some recent references