Great British Rainstorms – An events-based characterization of the properties of sub-hourly to daily annual maximum producing rainfall events

Abstract

This work presents an analysis of extreme rainfall in Great Britain (GB) using a new storm identification algorithm to obtain ~70,000 independent rainstorms associated with annual maxima (AM) for rainfall intensities from sub-hourly to daily durations. To do this a new rain gauge dataset with sub-hourly resolution was compiled and quality controlled. The high temporal resolution of rainfall data was found to improve the ability of quality control to identify faults in rain gauge records, as aggregated hourly or daily rainfall totals mask rapid changes to rainfall intensity that can indicate quality problems. The high temporal resolution of the dataset allowed for the first large scale characterisation of GB’s sub-hourly rainfall. The seasonality and geographical distribution of AM-generating rainstorms had good correspondence with known characteristics of GB rainfall at hourly and daily timescales. Extreme short duration (<4-hr) events have peak frequency during the summer, while the largest long duration (>12-hr) rainstorms occur in winter at rain gauges with high elevation. On average each rainstorm contributed ~3 AMs of different durations, this demonstrated important overlaps between AMs that pose a challenge for conventional rainfall design and analysis methods. Empirical annual exceedance probabilities were calculated at a regional level for rainstorms of different durations, this revealed similar quantiles for the most extreme short rainstorms across GB and major differences for long-duration rainstorms related to orographic effects and prevailing wind direction. Hyetographs of all AM-generating rainstorms were determined using sub-hourly resolution data, allowing an unprecedented exploration of the temporal characteristics of rainstorms in GB. These profiles showed significant variation with event duration, but events of similar duration but different seasons had similar profiles, contradicting key results from the 1975 Flood Studies Report (FSR). Hyetographs showed limited sensitivity to other variables such as rainstorm volume and geographical location. Only 22.6% of all rainstorms had centrally peaked hyetographs, therefore most events were unlike the FSR profiles currently used in UK rainfall design methods. This analysis has implications for all infrastructure and studies that use the FSR rainfall design profiles in the UK, including sustainable urban drainage systems and flood risk studies. A move away from centred design hyetographs towards the use of hyetograph ensembles is recommended, and current univariate extreme value frequency analysis methods are challenged.