• Research

Rouba ISKANDAR thesis defense - An integrated seismic risk modeling approach including human behavior

on the July 13, 2022

IMAG-UGA Auditorium

Thesis directed by :
. Cécile CORNOU, Research Director, ISTerre-IRD
. Elise BECK, Associate professor, PACTE-Université Grenoble Alpes
. Julie DUGDALE, Professor, LIG, Université Grenoble Alpes

We are proud to anounce the thesis defense of our CDP Risk PhD Students. Rouba will defend her thesis entitled An integrated seismic risk modeling approach including human behavior, on Wednesday, July 13th, 2022 at 1.30pm, in IMAG-UGA. Like the 11 thesis of the Cross Disciplinary Programme Risk, her work has been co-directed: ISTerre, PACTE & LIG (UGA).


Numerous observations from previous earthquakes highlighted the influence of human behavior on seismic risk (L’Aquila 2009; Great East Japan 2011; Lorca 2011).
Individuals’ actions, or even inaction, may have detrimental effects on the social consequences of an earthquake. Yet, seismic risk assessment methodologies rarely take into account human behavior. This thesis explores how modeling human behaviors and mobility in a post-earthquake environment can contribute to the definition of dynamic seismic risk indices that integrate human behavior. We adopt an interdisciplinary approach, merging earth, social and computer sciences, to model seismic risk while accounting for both its physical and social aspects. Specifically, we develop an agent-based model for the simulation of pedestrian earthquake evacuation (PEERS).
PEERS integrates realistic physical components related to building damages and resulting debris. The social component is taken into account by integrating behavioral responses, represented by the evacuation and mobility decisions and the interactions between individuals that result in the formation of groups.
We choose Beirut, Lebanon as a study area and recreate a virtual seismic crisis environment. We define two seismic scenarios: the first one corresponding to the regulatory Peak Ground Acceleration (PGA) of 0.3 g and the second one corresponding to a PGA of 0.5 g. We construct a building database for Beirut, and then estimate the building damages for the defined seismic scenarios using Artificial Neural Networks. To estimate the damage-induced debris, we develop an approach to predict the distribution of debris around a building according to its damage level. We define open spaces, i.e. areas away from buildings where individuals are safe, and identify the constraints in the urban environment that might affect pedestrians’ mobility. Moreover, we recreate a synthetic population of individuals and households that replicates the available data on Beirut’s population. Finally, we calibrate the behavioral responses in PEERS based on survey data from the August 4, 2020 explosion at the port of Beirut.
We run several earthquake evacuation simulations in this virtual environment, and look at people’s safety in the aftermath of an earthquake, with safety defined as being in an open space. We investigate the capacity of the open spaces in Beirut for providing shelter for the population in the immediate aftermath of an earthquake. We also investigate the effects of more complex physical and social environments on the population’s arrivals to safe areas. We find that in Beirut, the distribution of open spaces in terms of size and location cannot ensure the safety of all of the population, even in ideal conditions with minimal social and physical constraints. Furthermore, we find that debris and human behaviors both significantly delay the arrival times to open spaces. Yet, the human behavior delays the arrivals by two times compared to the delay caused only by the presence of debris.
A similar approach can be adopted to identify and rank physical and social components of a dynamic seismic risk index.

Published on June 28, 2022

Practical informations


IMAG-UGA, Auditorium
700 Avenue Centrale
38400 Saint-Martin-d'Hères
Open to all on site
Online participation: contact sylvie.perrier@univ-grenoble-alpes.fr

Rouba Iskandar