Understanding and modelling extreme multi-hazard events

Student thesis: Doctoral ThesisDoctor of Philosophy

Abstract

The study of natural hazard interrelations exposes the complexity of extreme climatological, geophysical and hydrological processes and poses new science challenges. This PhD thesis is located at the confluence of multivariate statistics, climatology and natural hazard modelling and aims to provide new approaches to model and quantify natural hazard interrelations. Chapter 2 consists of a critical literature review of 146 sources. From these, the historical context for quantitative single-hazard and multi-hazard assessment is discussed, and 19 different modelling methods to model multi-hazard interrelations are identified and organized into three broad approaches (empirical, stochastic, mechanistic). Chapter 3 examines the multi-hazard landscape of the European Atlantic Region (EAR) but has global relevance in its application. A total of 16 relevant natural hazards for the EAR region are identified on three main criteria: (i) frequency of occurrence, (ii) spatial relevance, (iii) potential to impact energy infrastructures. Based on the knowledge of hazard interrelations and physical drivers, natural hazards are grouped into five multi-hazard networks. Through a review of 32 single hazard catalogues, 50 historic major multi-hazard events in the EAR are pinpointed for each network. Within each network, the prevalence of each hazard interrelation is discussed. After identifying the main modelling approaches and dominant hazard interrelations in the EAR, the abilities of a group of modelling method for multi-hazard modelling is assessed. Chapter 4 evaluates the efficacy of bivariate extreme modelling approaches for multi-hazard scenarios. Six bivariate extreme models are evaluated and compared by using each model’s fitting capabilities to 60 synthetic datasets. The properties of the synthetic datasets are matching bivariate time series of environmental variables. The systematic framework contrasts model strengths (model flexibility) and weaknesses (poorer fits to the data). The benefits of this framework are highlighted with two applications to natural hazard interrelation modelling. Using the findings of Chapter 3, two pairs of natural hazard are selected: extreme hot temperature–wildfire; extreme wind–extreme rainfall. Chapter 5 analyses the spatiotemporal features of hazard interrelations using climate reanalysis data for two hazards (extreme wind and extreme rainfall) for 1979–2019 within a region including Great Britain and the British channel. A clustering algorithm is used to create hazard clusters with extreme values (above the 99% quantile) of hourly precipitation and wind gust. A total of 4555 compound wind-rainfall clusters are detected for 1979–2019 by assessing the spatiotemporal overlap of the two hazards. The characteristics (e.g., size, duration, season, intensity) of created clusters are confronted with observations and analysed. One of the bivariate modelling methods assessed in Chapter 4 is used to estimate return periods of compound hazard events. The relationship between the return period of compound hazard events and the spatial and temporal attributes of compound hazards events is then analysed. Throughout the thesis, the following main aspects of a quantitative multi-hazard approach are addressed: interrelation characterisation, multivariate modelling, physical drivers, spatiotemporal overlap, data. Robust solutions to identify, discriminate and model hazard interrelations at different spatial and temporal scales are offered.
Date of Award1 May 2021
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorBruce Malamud (Supervisor), Hugo Winter (Supervisor) & Amélie Joly-Laugel (Supervisor)

Cite this

'