Project

Back to overview

Risk and Complex Events: from Probability to Applications

Applicant Engelke Sebastian
Number 161297
Funding scheme Ambizione
Research institution Research Center for Statistics Université de Genève
Institution of higher education University of Geneva - GE
Main discipline Mathematics
Start/End 01.11.2015 - 31.01.2019
Approved amount 341'202.00
Show all

Keywords (6)

Risk quantification; Systemic risk on networks; Extreme value theory; Simulation; Multivariate dependence; Downscaling

Lay Summary (German)

Lead
Extreme Wetterereignisse sind weltweit für Überflutungen und Dürren verantwortlich, die nicht nur finanziellen Schaden verursachen, sondern auch regelmäßig Menschenleben fordern. Die letzte Finanzkrise hat außerdem gezeigt, dass das Unterschätzen der Wahrscheinlichkeiten von sehr seltenen Marktszenarien einen verheerenden Einfluss auf die Risikobewertung haben kann. Die Verkettung beziehungsweise Gleichzeitigkeit von Extremereignissen- starke Regenfälle in einem großen Gebiet oder simultane Insolvenzen von wichtigen Banken - stellt oft eine große Gefahr für ein System dar. Die möglichst genaue Quantifizierung dieser Risiken ist eine Herausforderung, für die neue statistische Modelle benötigt werden.
Lay summary

Inhalt und Ziel des Forschungsprojekts

Das übergeordnete Ziel dieses Projektes ist, effektive Methoden zur Analyse von komplexen Extremereignissen und deren Risikobewertung zu entwickeln. Dabei wird im Mittelpunkt stehen, theoretische Ergebnisse aus der sogenannten Extremwerttheorie in praxisrelevante Instrumente zu überführen. Im Detail werden (i) Simulationsmethoden von selten Ereignissen verbessert, die es erlauben, extreme Szenarien zu erzeugen und deren räumliche Abhängigkeitsstruktur zu bestimmen. Es soll außerdem untersucht werden, wie (ii) Ergebnisse aus Klimamodellen genutzt werden können, um die Gefahr von lokalen Wetterextremen zu quantifizieren. Ein weiteres Ziel ist, (iii) die Verkettung von einzelnen, extremen Ereignissen in komplexen Systemen zu analysieren, und das damit verbundene systemische Risiko zu bewerten.

Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojekts

Die Ergebnisse dieses Projektes werden eine verbesserte Bewertung der gesellschaftlichen und ökologischen Risiken durch Extremereignisse ermöglichen. Dies wird unter anderem Regulierungsbehörden valide Grundlagen zur Einschätzung der Gefahr durch Hochwasser oder andere Naturkatastrophen liefern und dadurch helfen, die genannten Risiken zu begrenzen.
 

 

Direct link to Lay Summary Last update: 15.10.2015

Responsible applicant and co-applicants

Employees

Publications

Publication
Extremal behaviour of aggregated data with an application to downscaling
Engelke Sebastian, De Fondeville Raphaël, Oesting Marco (2019), Extremal behaviour of aggregated data with an application to downscaling, in Biometrika, 106(1), 127-144.
Dependence properties of spatial rainfall extremes and areal reduction factors
Le Phuong Dong, Davison Anthony C., Engelke Sebastian, Leonard Michael, Westra Seth (2018), Dependence properties of spatial rainfall extremes and areal reduction factors, in Journal of Hydrology, 565, 711-719.
Optimal regionalization of extreme value distributions for flood estimation
Engelke Sebastian, Davison Anthony C., Asadi Peiman (2018), Optimal regionalization of extreme value distributions for flood estimation, in Journal of Hydrology, 556, 182-193.
Bayesian inference for multivariate extreme value distributions
Dombry Clément, Engelke Sebastian, Oesting Marco (2017), Bayesian inference for multivariate extreme value distributions, in Electronic Journal of Statistics, 11, 4813-4844.
Generalized Pickands constants and stationary max-stable processes
Debicki Krzysztof, Engelke Sebastian, Hashorva Enkelejd (2017), Generalized Pickands constants and stationary max-stable processes, in Extremes, 20, 493-517.
A characterization of the normal distribution using stationary max-stable processes
Engelke Sebastian, Kabluchko Zakhar (2016), A characterization of the normal distribution using stationary max-stable processes, in Extremes, 19, 1-6.
A Lévy-derived process seen from its supremum and max-stable processes
Engelke Sebastian, Ivanovs Jevgenijs (2016), A Lévy-derived process seen from its supremum and max-stable processes, in Electronic Journal of Probability, 21, 1-20.
Exact simulation of max-stable processes
Dombry Clément, Engelke Sebastian, Oesting Marco (2016), Exact simulation of max-stable processes, in Biometrika, 106, 303-317.
Robust bounds in multivariate extremes
Engelke Sebastian, Ivanovs Jevgenijs, Robust bounds in multivariate extremes, in Annals of Applied Probability.

Collaboration

Group / person Country
Types of collaboration
Seth Westra at University of Adelaide Australia (Oceania)
- in-depth/constructive exchanges on approaches, methods or results
Philippe Naveau at LSCE Paris France (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Paul Embrechts at ETHZ Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Chen Zhou at Erasmus University Netherlands (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Jevgenijs Ivanovs at Aarhus University Denmark (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Jonas Peters at University of Copenhagen Denmark (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Marco Oesting at University of Siegen Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Jenny Wadsworth at Lancaster University Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Marius Hofert at University of Waterloo Canada (North America)
- in-depth/constructive exchanges on approaches, methods or results
Clément Dombry at Université de Franche-Comté, Laboratoire de Mathématiques de Besançon France (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Thomas Opitz at INRA, Avignon France (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Zakhar Kabluchko at Münster University Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Jakob Zscheischler at ETHZ Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Anthony C. Davison at EPFL Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Raphael Huser at KAUST Saudi Arabia (Asia)
- in-depth/constructive exchanges on approaches, methods or results
Enkelejd Hashorva at Université de Lausanne Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Michael Leonard Australia (Oceania)
- in-depth/constructive exchanges on approaches, methods or results
Nicolai Meinshausen at ETHZ Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Department of Decision Sciences, HEC Montreal Individual talk Graphical models and structural learning for extremes 05.12.2018 Montreal, Canada Engelke Sebastian;
Applied Probability and Risk Seminar, Columbia University Individual talk Graphical models and structural learning for extremes 29.11.2018 New York, United States of America Engelke Sebastian;
Department of Statistical Sciences, University of Toronto Individual talk Graphical models and structural learning for extremes 08.11.2018 Toronto, Canada Engelke Sebastian;
Data, Models and Statistical Inference, workshop in honour of Anthony C. Davison Talk given at a conference Structural learning for extremes 15.09.2018 Lausanne, Switzerland Engelke Sebastian;
BIRS-CMO Workshop: Self-Similarity, Long-Range Dependence and Extremes Talk given at a conference Extremal (in)dependence structures of copulas with multiplicative constructions 19.06.2018 Oaxaca, Mexico Engelke Sebastian;
Rare Events, Extremes and Machine Learning Workshop Talk given at a conference Extremal (in)dependence structures of copulas with multiplicative constructions 25.05.2018 Paris, France Engelke Sebastian;
Seminar at School of Mathematics, Statistics and Physics Individual talk Extremal (in)dependence structures of copulas with multiplicative constructions 26.03.2018 Newcastle, Great Britain and Northern Ireland Engelke Sebastian;
5th British-German Frontiers of Science Symposium Poster Flood risk assessment: an extreme value approach 21.03.2018 Dorking, Great Britain and Northern Ireland Engelke Sebastian;
10th ERCIM Conference Talk given at a conference Extremal (in)dependence structures of copulas with multi- plicative construction 18.12.2017 London, Great Britain and Northern Ireland Engelke Sebastian;
Thiele Seminar, Aarhus University Individual talk Extremal (in)dependence structures of copulas with multiplicative constructions 30.11.2017 Aarhus, Denmark Engelke Sebastian;
Workshop "Modern Statistics for Complex Data Structures", KAUST Talk given at a conference Extremal (in)dependence structures of copulas with multiplicative constructions 13.11.2017 Thuwal, Saudi Arabia Engelke Sebastian;
Seminar at Risk Lab, ETH Zürich Individual talk Extremal (in)dependence structures of copulas with multiplicative constructions 19.10.2017 Zürich, Switzerland Engelke Sebastian;
10th Conference on Extreme Value Analysis Talk given at a conference Extremal behavior of aggregated data with an application to downscaling 29.06.2017 Delft, Netherlands Engelke Sebastian;
IDEA Workshop, Rotterdam Talk given at a conference Functional Data Analysis for Extremes 24.06.2017 Rotterdam, Netherlands Engelke Sebastian;
Swiss Statistics Seminar, University of Bern Talk given at a conference An entropy-based test for multivariate threshold exceedances 05.05.2017 Bern, Switzerland Engelke Sebastian;
Talk at Seminar for Statistics, ETH Zürich Individual talk Models for extremes on graphs 06.04.2017 Zürich, Switzerland Engelke Sebastian;
Seminar at Pierre and Marie Curie University (Paris 6) Individual talk Robust bounds in multivariate extremes 14.03.2017 Paris, France Engelke Sebastian;
Seminar at Department of Statistical Sciences, University of Toronto Individual talk Robust bounds in multivariate extremes 07.02.2017 Toronto, Canada Engelke Sebastian;
Seminar at Department of Mathematical Science, Chalmers University Individual talk Robust bounds for multivariate extreme value distributions 26.01.2017 Gothenburg, Sweden Engelke Sebastian;
Seminar at Research Center for Statistics, University of Geneva Individual talk Robust bounds for multivariate extreme value distributions 19.12.2016 Geneva, Switzerland Engelke Sebastian;
9th ERCIM Conference Talk given at a conference Robust bounds for multivariate extreme value distributions 09.12.2016 Sevilla, Spain Engelke Sebastian;
Workshop on Conditional Independence Structures and Extremes Talk given at a conference Extremes on river networks 12.10.2016 München, Germany Engelke Sebastian;
German Statistical Week Talk given at a conference Statistical regionalization for estimation of extreme river discharges 13.09.2016 Augsburg, Germany Engelke Sebastian;
Workshop on Extremes and Risks in Higher Dimensions Talk given at a conference Robust bounds for multivariate extreme value distributions 12.09.2016 Leiden, Netherlands Engelke Sebastian;
4th IMS-APR Meeting Talk given at a conference Exact Simulation of max-stable Processes 27.06.2016 Hongkong, Hongkong Engelke Sebastian;
Uncertainty Modeling in the Analysis of Weather, Climate and Hydrological Extremes, Banff Research Station Talk given at a conference Statistical regionalization for estimation of extreme river discharges" 13.06.2016 Banff, Canada Engelke Sebastian;
Workshop on Dependence, Stability and Extremes, Fields Institute Talk given at a conference Bayesian inference for multivariate extreme value distributions 02.05.2016 Toronto, Canada Engelke Sebastian;
Seminar of Risk Lab, ETH Zürich Individual talk Bayesian inference for multivariate extreme value distributions 26.02.2016 Zürich, Switzerland Engelke Sebastian;
Seminar of LMU München Individual talk Bayesian inference for multivariate extreme value distributions 13.01.2016 München, Germany Engelke Sebastian;
8th ERCIM Conference Talk given at a conference Bayesian Inference for Multivariate Extremes 12.12.2015 London, Great Britain and Northern Ireland Engelke Sebastian;
Seminar at King Abdullah University of Science and Technology Individual talk Bayesian inference for multivariate extreme value distributions 07.12.2015 Thuwal, Saudi Arabia Engelke Sebastian;


Self-organised

Title Date Place
Winter school on “Recent Advances in Extremes” 13.02.2017 Lausanne, Switzerland

Awards

Title Year
Honourable mention in Bernoulli Society New Researcher Award 2018
Early Career Investigator Award for Best Presentation, 9th ERCIM Conference in Seville, Spain. 2016

Abstract

Extreme value theory is an important and widely-used approach to multivariate risk assessment which provides mathematically justified estimates of small tail probabilities. It is a rapidly evolving area of research at the interface of probability theory, statistical methods and applications. Both theory and statistical methods for univariate tail analysis are well-developed. While the probabilistic foundations for multivariate theory and complex events are quickly advancing, there is still much potential to translate them into effective tools and statistical models. This is crucial for an improved application of extreme value theory for risk assessment in applied science and industry. The main objective of my research project is thus to provide tools and methods that facilitate risk quantification in complex applications.The first part of this project will tackle the as yet unsolved problem of generating random samples from multivariate extreme value distributions. Even though this is regularly needed for evaluation of risk measures via Monte Carlo methods, simulation studies of statistical methods or rare event generation,only inexact and time-intensive algorithms are currently available. In the second part, the question whether the risk of extreme weather events will change in future will be addressed. Long-term climate models predict meteorological variables only on very large scales. A method that allows for downscaling of the output of these models to information on local extreme events will be developed. The third part of the project aims at opening up new fields of application for multivariate extreme value theory. Complex systems like, for instance, river networks or systems of interacting banks, can often be represented as graphical structures. New models that take the connections in these networks into account will allow for an accurate assessment of systemic risk. A workshop on risk quantification with extreme value methods and their applications in related fields will be organized withinthe scope of this independent research project.
-