The Impact of Rainfall Uncertainty on Flood Simulations

Abstract:

Pluvial flooding from highly localised extreme rainfall, commonly occurring over relatively short durations, causes significant damage across the UK annually. Our ability to simulate such events and provide robust flood predictions is hampered by a lack of reliable sub-daily rainfall products and effective techniques to quantify the non-stationary error within them. These issues could lead to type 1 and type 2 errors when trying to identify useful model simulation tools within a rejectionist framework. Uncertainties will then propagate non-linearly through rainfall-runoff simulations to inundation predictions. Therefore it is vital to understand where the uncertainties are greatest in such a model simulation cascade and how they impact flood predictions in order to direct future research.

We consider the impact of rainfall errors on flood predictions using a new gridded sub-daily (hourly) rainfall record for the UK from 1993 to 2011 from a network of tipping bucket data. We improve upon currently available daily gridded datasets, in order to provide a more suitable rainfall record with which to model flooding from intense storm events. We attempt to quantify the most influential sources of uncertainty within the gridded product in order to provide an ensemble of rain-field realisations with which to force a coupled hydrological – hydrodynamic model for the representation of flash flooding at regional scales.

Here we present the results from a test case in order to demonstrate the degree to which uncertainty in the sub-daily gridded rainfall product propagates through the modelling framework to influence the prediction of flooding over a variety of events. The results will provide an insight into the uncertainty in estimating catchment area rainfall, and its influence on our ability to predict flooding from intense storm events and avoid type 1 and type 2 errors.