Uncertainty Estimation in Fluvial Flood Forecasting Applications

Abstract

A risk-based decision framework is outlined for the selection of uncertainty estimation techniques for use in real-time flood forecasting applications. The framework aims to achieve a balance between the operational requirements for probabilistic information and their use, and typical hydrological and operational constraints, such as catchment response times and computer processing power. A typology of techniques is introduced that employs the three overall categories of forward uncertainty propagation, probabilistic data assimilation and probabilistic forecast calibration. The relative strengths and limitations of each general category are contrasted, and illustrated using examples for a range of techniques and case study catchments across the UK. In conclusion, an indication is given of future research requirements in probabilistic flood forecasting.