TsunaWAVE: coupling uncertain landslide-generated tsunami scenarios with agent-based evacuation modelling through an interactive web platform. Application to Mamoudzou (Mayotte, France)

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Pedestrian tsunami evacuation depends on wave arrival time, warning dissemination, the spatial configuration of the exposed area and population response. These factors, studied by distinct disciplines, are rarely integrated within a single framework, limiting the use of evacuation models for local planning. We present TsunaWAVE (Web interface for Agent-based modelling and Visualisation of Evacuation), an open web interface that links tsunami hazard scenarios with agent-based evacuation scenarios through a one-way sequential coupling, in which simulated arrival times define the time budget against which precomputed evacuation outputs are read, making both explorable by decision-makers. The agent-based model simulates warning dissemination, pre-evacuation delay and movement towards safe areas, while the published surrogate model (Landslide-Tsurrogate v1.0) provides first-wave arrival times for submarine landslide scenarios. The framework is applied to the Mamoudzou waterfront (Mayotte, France), where lead times are 10 to 16 minutes in two reference scenarios and 759 exposed individuals were counted using drone imagery, video line-crossing and manual counts. Ninety scenarios, each replicated 42 times, rank three operational levers (warning, safe-area accessibility and social contagion). Warning reliability is the first-order control. Under degraded warning without social contagion and a 10 min lead time, fewer than 8% of agents reach safety, whatever the safe-area configuration. Social contagion, implemented as neighbourhood-based imitation, ranks second; it cannot substitute for a reliable warning but partially compensates for a degraded one when time allows, raising the share reaching safety at 16 min from 19.4% to 43.8%. TsunaWAVE makes these findings openly available to local authorities at https://geodata.univ-montp3.fr/web-app/irima-mayotte/.

This publication has 37 references indexed in Scilit: