TY - JOUR

T1 - Spatial risk assessment for extreme river flows

AU - Keef, Caroline

AU - Tawn, Jonathan

AU - Svensson, Cecilia

PY - 2009/12/31

Y1 - 2009/12/31

N2 - The UK has in recent years experienced a series of fluvial flooding events which have simultaneously affected communities over different parts of the country. For the co-ordination of flood mitigation activities and for the insurance and reinsurance industries, knowledge of the spatial characteristics of fluvial flooding is important. Past research into the spatiotemporal risk of fluvial flooding has largely been restricted to empirical estimates of risk measures. A weakness with such an approach is that there is no basis for extrapolation of these estimates to rarer events, which is required as empirical evidence suggests that larger events tend to be more localized in space. We adopt a model-based approach using the methods of Heffernan and Tawn. However, the large proportion of missing data over a network of sites makes direct application of this method highly inefficient. We therefore propose an extension of the Heffernan and Tawn method which accounts for missing values. Furthermore, as the risk measures are spatiotemporal an extension of the Heffernan and Tawn method is also required to handle temporal dependence. We illustrate the benefits of the procedure with a simulation study and by assessing spatial dependence over four fluvial sites in Scotland.

AB - The UK has in recent years experienced a series of fluvial flooding events which have simultaneously affected communities over different parts of the country. For the co-ordination of flood mitigation activities and for the insurance and reinsurance industries, knowledge of the spatial characteristics of fluvial flooding is important. Past research into the spatiotemporal risk of fluvial flooding has largely been restricted to empirical estimates of risk measures. A weakness with such an approach is that there is no basis for extrapolation of these estimates to rarer events, which is required as empirical evidence suggests that larger events tend to be more localized in space. We adopt a model-based approach using the methods of Heffernan and Tawn. However, the large proportion of missing data over a network of sites makes direct application of this method highly inefficient. We therefore propose an extension of the Heffernan and Tawn method which accounts for missing values. Furthermore, as the risk measures are spatiotemporal an extension of the Heffernan and Tawn method is also required to handle temporal dependence. We illustrate the benefits of the procedure with a simulation study and by assessing spatial dependence over four fluvial sites in Scotland.

KW - Extreme value theory

KW - Missing data

KW - Multivariate extreme values

KW - River flows

KW - Spatial risk assessment

KW - Spatiotemporal extremal dependence

U2 - 10.1111/j.1467-9876.2009.00672.x

DO - 10.1111/j.1467-9876.2009.00672.x

M3 - Journal article

AN - SCOPUS:70350141749

VL - 58

SP - 601

EP - 618

JO - Journal of the Royal Statistical Society. Series C: Applied Statistics

JF - Journal of the Royal Statistical Society. Series C: Applied Statistics

SN - 0035-9254

IS - 5

ER -