Hydrological ensemble prediction systems: From evaluating daily streamflow forecasts to exploring the impact of selected flood events in a future climate

Abstract

Worldwide, floods are the natural disaster causing the highest economic losses and casualties. Flood warnings are issued to inform of potential threats, and allow flood managing authorities time to take appropriate actions to prevent and reduce the impact. Flood forecasts are often uncertain, and statistical techniques are used to improve the forecasts. Climate studies indicate that flood magnitude and frequency will change. More knowledge to improve flood forecasts and to assess the outcome of future floods is needed. The aim of this thesis is to provide improved operational flood forecasts and raise awareness of future floods. An experimental setup of the Norwegian flood forecasting systems was used and included 145 river catchments. First, techniques to improve flood forecasts were applied and revealed seasonal and regional patterns in flood forecast performance. By exposing areas with low forecast performance, targeted and tailored corrections can be applied. Secondly, extreme flood events caused by atmospheric rivers in a future climate were compared to the most extreme events of the present climate. The future extreme events showed an increase in both flood magnitude and more rivers flooded concurrently.