Seminar - Characterising high river flow clustering: UK analysis for insurance risk
Floods are the most common and costly natural hazard worldwide and recent, individual events in the UK have cost insurers over £500 million. Following several recent, large-scale flood events in the UK (including central and eastern England in Easter 1998; northern England in autumn 2000; northern and central England in 2007; and Scotland and northern England in 2009), research to explain their occurrence has shifted away from “searching for a trend.” Instead, new studies have demonstrated that in the UK, high river flows are temporally clustered into flood-rich and flood-poor periods over decadal time-scales (e.g. Lane, 2008; Wilby et al., 2008; Raven et al., in prep). Risk assessments rely on the reliable estimation of event probabilities under the assumption that flow data is statistical stationary or in a predictable trend. Hence, understanding and correctly charactering the spatial and temporal occurrence of high river flow clustering is important for flood frequency estimation for policy-making, risk management and the insurance industry.
Using analysis of river flow time-series from the longest available records in the UK, we explore these decadal length flood-rich periods and explore the best ways to characterise them for use in risk analysis? Our methodology includes: (1) counts of peaks within a moving window of length ranging from one to forty years; (2) obtaining a dispersion statistic to ascertain whether any inter-annual clustering of peaks are statistically greater than that expected due to natural variability; and (3) exploring the time duration between peaks including a Monte Carlo based analysis to compare the observed time differences from randomly sampled records. We consider additional factors including flow return period, geographical location and catchment size during the analysis.
Speaker: Dr Emma Raven, IHRR, Dept of Geography
Contact email@example.com for more information about this event.