Extreme meteo-oceanic events

Abstract : This PhD on published works aims at unifying the works carried out on the topic of extreme metocean events since 2009, while working for SOGREAH then ARTELIA.As these works went along, a leading theme progressively appeared: the notion of event, such as a storm. This concept provides a sound and relevant framework in particular in the case of multivariate extremes (such as joint probabilities of waves and sea levels), as well as a better understanding of the notion of return period, much used for design in the field of engineering.The main results of the works carried out in the last decade are as follows:- updating of the methodology for determining extreme wave heights or wind speeds:- development and justification of a two-step framework for extreme univariate over-threshold modelling introducing the concept of event and the separation of the physical and statistical thresholds,- proposal of practical tools for choosing the statistical threshold,- introduction of the parametric bootstrap approach for computing confidence intervals,- identification of a problematic issue in the behaviour of the Maximum Likelihood Estimator and proposal of a solution: use of 3-parameter distributions along with the L-moments estimator,- application of the POT framework to the Joint Probability Method for determining extreme sea levels:- distinction between sequential values and event peaks through extremal indexes for surge and sea level,- construction of a mixture model for the surge distribution,- refinements for handling tide-surge dependence,- application of the POT-JPM framework for the joint analysis of wave height and sea level:- proposal of an alternative sampling procedure,- separate analysis of tide and surge in order to model the dependence between wave height and surge to be incorporated in the joint distribution of wave height and sea level thanks to a 2D1D convolution operation,- use of extreme-value copulas,- improved presentation of the chi-plot,- introduction of a new classification for multivariate analyses:- Type A: a single phenomenon described by different physical quantities that are not of the same kind,- Type B: a phenomenon made of different components, described by physical quantities of the same kind between one component and another,- Type C: several phenomena described by physical quantities that are not of the same kind,- interpretation of the meaning of multivariate events:- link with the sampling procedure,- link with the different definitions of the return period,- in the bivariate case: transformation of the joint distribution of event-describing variables into the joint distribution of sequential pairs,- generation of alternative output plots such as contours of density for sequential pairs;- a dedicated R package, artextreme, for implementing the methodologies presented above
Document type :
Theses
Complete list of metadatas

https://pastel.archives-ouvertes.fr/tel-01743728
Contributor : Abes Star <>
Submitted on : Monday, March 26, 2018 - 4:30:12 PM
Last modification on : Thursday, May 16, 2019 - 6:29:58 PM
Long-term archiving on : Thursday, September 13, 2018 - 10:11:02 AM

File

TH2017PESC1148.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01743728, version 1

Collections

Citation

Franck Mazas. Extreme meteo-oceanic events. Environmental Engineering. Université Paris-Est, 2017. English. ⟨NNT : 2017PESC1148⟩. ⟨tel-01743728⟩

Share

Metrics

Record views

341

Files downloads

361