Modelling of space–time rainfall for three UK regions

In the context of flood management in urban areas in the UK, the Gaussian displacement spatial–temporal model (GDSTM) is used to generate a continuous rainfall field. The model makes the assumption that storms arrive according to a Poisson process in space and time, triggering another Poisson process of cell arrivals over the storm duration. These cells are displaced from the storm centre according to a normal distribution. Historic rainfall events are identified from a 3·5-year record of Met Office weather radar data from three different radars and the characteristics of the interior of each event are represented by 11 parameters. The process of event arrival is described by six additional parameters. For each month within the radar data record, a library of model parameters is obtained and used to develop simulations of 100 years of continuous, spatially varying rainfall at 5 min intervals and 1 km spatial resolution over three regions: London, Bradford and Glasgow. The objective is to use the synthetic data as input to existing urban drainage models to examine system performance. It is possible to reproduce standard statistics over the London region but the extreme values are underestimated when compared with statistics from the Flood Estimation Handbook (FEH). The mean rainfall is overestimated at the other sites but more agreement in simulating the extremes is observed when compared with the FEH.