Abstract

Extreme rainfall events pose considerable threats to society and critical infrastructure yet, by definition, these events are rare. Reliable estimates of the likelihood of such events are required to assist with impact quantification and risk management. Similarly, the detection of any changes in... Show moreExtreme rainfall events pose considerable threats to society and critical infrastructure yet, by definition, these events are rare. Reliable estimates of the likelihood of such events are required to assist with impact quantification and risk management. Similarly, the detection of any changes in the observed frequency or nature of these events is essential to facilitate appropriate adaptation actions to be taken by decision makers. Long established hydrological practice may no longer be appropriate if these changes are significant over the engineering design life. While extreme daily precipitation events are known to be seasonally over-dispersed, the dependent relationship between events is often ignored. Many statistical representations of extreme rainfall also explicitly ignore the complex relationships which exist between seasonality, atmospheric variables and extreme event frequency; thus a robust statistical tool is required to test the significance of any changes. Using newly defined extreme rainfall regions for the UK, a Vector Generalized Additive Model (VGAM) is presented which characterizes the inter-annual variability of extreme daily precipitation event frequency, and their associated magnitude. The modeling technique is one which could be applied in many regions of the world, but is specifically focused on an application to UK extreme daily precipitation. The seasonal behavior of daily extreme precipitation and its dependence on sea surface temperatures (SST), air temperature range and the North Atlantic Oscillation (NAO) are represented in flexible Generalized Pareto and Poisson distribution parameter estimates using VGAMs, to test the significance of changes in the temporal pattern of frequency and intensity. There is a strong negative correlation with monthly maximum diurnal air temperature range, reflecting heightened event intensity and probability when the diurnal temperature range is at its lowest. Event frequency is positively correlated with SST for all UK regions; while event magnitude is dependent on either SST in the south of the UK or the NAO in the northwest of the UK. While the timing of extreme events has not changed substantially, event probability has increased - resulting in greater seasonal over-dispersion. This result could have considerable implications for the future when projected increases in SST and decreases temperature range are accounted for, leading to increases in event intensity and frequency. Show less