PhD: robust identification of excursion sets with application to flooding risk

The risk of coastal flooding is aggravated by the failure of coastal defenses (either natural
like dunes or artificial like dykes). In numerical simulations, such processes are
typically accounted for by defining a set of scenarii describing the possible
spatial location, time duration, erosion height / width of the failures. In the current PhD, we
propose to develop a systematic mathematical procedure to characterize the possible
combinations of conditions (named excursion set) that lead to flooding. It involves the
inversion of the numerical models that simulate the floods. In order to alleviate the
computational cost of this task, we build upon the combination of metamodeling techniques
(kriging/Gaussian processes) and active learning specifically dedicated to the estimation of
excursion set. The PhD aims at improving the existing methods in two ways:
(1) methodologically, by making the inversion robust to extreme-but-rare events and
accounting for uncertainties in the numerical models; (2) operationally, by assessing how this
approach can help in the communication and the management of the risk through better high
dimensional visualization of the excursion set and the decomposition of the uncertainties.
The application cases will focus on marine (BRGM) and river (IRSN) flooding. The PhD
candidate will participate to the continuation of the OQUAIDO collaborative project
( and benefit from numerous interactions
with other researchers in the same scientific domain.

Job location: 
Ecole des Mines de Saint-Etienne
42100 Saint-Etienne
Contact and application information
Tuesday, June 30, 2020
Contact name: 
Rodolphe Le Riche