Spatial and temporal scaling of extreme rainfall in the United Kingdom

Extreme rainfall estimates for ungauged areas contribute to improved resilience to flooding. This study applies a sub-daily rainfall scaling method to annual daily maximum rainfall series from 102 UK weather stations. We analyse resultant parameters by season, homogenous rainfall region, urban area and geographic factors to investigate how scaling varies temporally and spatially. Dummy regression models are built using these variables to predict the sub-daily scaling parameter for any location in the United Kingdom. Estimated rainfall intensities are validated with observations and yield Mean Absolute Errors of 3.0, 1.9 and 0.9 mm/h for 1-, 2- and 6-h events, respectively. We also demonstrate intensity-duration-frequency curves at a site in Oxfordshire for scaled and observed data and find that 1- to 6-h, 20-year rainfall intensities are estimated to be within 9.4%. With such unified scaling relationships, it is possible to derive extreme rainfall for specified durations and return periods at ungauged locations. According to our cross-validation of estimated and observed intensities, more than 88% of sites fall within 10% error bounds. This method offers a means of generating design rainfall series as input to flood simulation models to evaluate pluvial flood risks in urban areas. Our research examined UK extreme rainfall scaling properties and its variation with space and time linked to season, homogenous rainfall region, urban area, and geographic factors. We then developed a dummy regression model using these variables to estimate a local scaling parameter and extreme rainfall intensities at more than 100 sites. After our model was cross-validated, we demonstrated how the scaling can be used to create an intensity-duration-frequency for unseen sites.image