Publication Date:
2018-04-23
Description:
Natural hazard loss portfolios with exposure over a region are sensitive to the dependency between extreme values of the key hazard variable at different spatial locations. It is therefore important to correctly identify and quantify dependency to avoid poor quantification of risk. This study demonstrates how bivariate extreme value tail dependency methods can be used together in a novel way to explore and quantify extremal dependency in spatial hazard fields. A relationship between dependency and loss is obtained by deriving how the probability distribution of conceptual loss depends on the tail dependency coefficient. The approaches are illustrated by applying them to 6103 historical European windstorm footprints (spatial maps of 3-day maximum gust speeds). We find there is little evidence of asymptotic extremal dependency in windstorm footprints. Furthermore, empirical extremal properties and conceptual loss distributions between pairs of locations are shown to be well reproduced using Gaussian copulas but not by extremally-dependent models such as Gumbel copulas. It is conjectured that the lack of asymptotic dependence is a generic property of turbulent flows, which may extend to other spatially continuous hazards such as heat waves and air pollution. These results motivate the potential of using Gaussian process (geostatistical) models for efficient simulation of hazard fields.
Electronic ISSN:
2195-9269
Topics:
Geography
,
Geosciences
