A Route-Memory Agent-Based Framework for Tourist Evacuation in Tsunamis

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Tourist evacuation in tsunami-prone coastal communities is strongly shaped by local road knowledge acquired before a warning is issued. This dependence becomes especially important where external guidance is limited, damaged, or unavailable during a near-field tsunami emergency. Existing evacuation models often represent tourists through static familiarity classes, complete ignorance, or simple heuristic rules. These assumptions limit their ability to capture how pre-disaster sightseeing, memory degradation, and route recall affect evacuation behavior. This study develops an agent-based evacuation framework that represents tourist road cognition as a dynamic route-memory state. Tourists first explore a road network, encode experienced road links, undergo empirically calibrated cue perturbation, and retrieve usable route knowledge through Hopfield-network associative recall. The framework is tested on an idealized (10×10) grid and then applied to Tofino, British Columbia, using GIS-derived road geometry, zoning-based population placement, tourist sightseeing areas, and actual evacuation destinations. The simulations show that incomplete route memory can substantially reduce tourist shelter arrival within a guidance-free near-field tsunami response window, while full local road knowledge improves but does not eliminate the tourist evacuation shortfall. The results link tourist evacuation outcomes to remembered route availability, warning-time location, milling delay, and congestion on limited corridors. The framework provides a way to examine where route memory constrains evacuation and where information-support strategies may improve planning in tourism-oriented coastal communities.