Evaluating the efficacy of bivariate extreme modelling approaches for multi-hazard scenarios

摘要

Modelling multiple hazard interrelations remains a challenge forpractitioners. This article primarily focuses on the interrelations betweenpairs of hazards. The efficacy of six distinct bivariate extreme models isevaluated through their fitting capabilities to 60 synthetic datasets. Theproperties of the synthetic datasets (marginal distributions, taildependence structure) are chosen to match bivariate time series ofenvironmental variables. The six models are copulas (one non-parametric, onesemi-parametric, four parametric). We build 60 distinct synthetic datasetsbased on different parameters of log-normal margins and two differentcopulas. The systematic framework developed contrasts the model strengths(model flexibility) and weaknesses (poorer fits to the data). We find thatno one model fits our synthetic data for all parameters but rather a rangeof models depending on the characteristics of the data. To highlight thebenefits of the systematic modelling framework developed, we consider thefollowing environmental data: (i) daily precipitation and maximum wind gustsfor 1971 to 2018 in London, UK, and (ii) daily mean temperature and wildfirenumbers for 1980 to 2005 in Porto District, Portugal. In both cases there isgood agreement in the estimation of bivariate return periods between modelsselected from the systematic framework developed in this study. Within thisframework, we have explored a way to model multi-hazard events and identifythe most efficient models for a given set of synthetic data and hazard sets.