The performance of rainfall nowcasts from the Nationale Regenradar for extreme-event rainfall and runoff prediction : A quantitative analysis using FEWS and 3Di

Abstract

Extreme rainfall can potentially lead to inundation and even damage. Decision-makers can take measures if they have both timely and accurate rainfall forecasts. A shortterm rainfall forecast based on the extrapolation of radar images is called nowcast. This thesis studies the performance of nowcasts from the Nationale Regenradar. The historic rainfall is compared with the nowcasted rainfall at several lead times for 3 events: a cold front (13-10-13), a cloudburst (28-07-14) and a thunderstorm (30-05-16); all three events lead to inundation in urban areas. Performance statistics are calculated in FEWS (Flood Early Warning System, a central database containing various modelling tools) and runoff and flooding are modelled in 3Di. The first observation is that the performance of the nowcasts decreases for longer lead times. The maximum lead time for nowcasts found reliable enough is 20(±10) minutes before the extreme rainfall event, depending on the event, region and performance statistic used. The second observation is that the performance also decreases for higher rainfall intensities. The third observation is that the nowcasts underestimate the historic rainfall, varying from 1.3 to 3.3 times, due to errors in the real time radar images. In order to determine the level of nuisance, the nowcasted maximum rainfall is very important. This thesis suggests an improvement of the nowcasts by improving the real time radar images. This can be done by including a correction factor amongst others. The conclusion of this thesis is that there is a potential for using nowcasts as a warning system in urban areas, provided that proper bias-corrections are applied.