An Agent-Based Modeling Approach to Integrate Tsunami Science, Human Behavior, and Unplanned Network Disruptions for Nearfield Tsunami Evacuation

For the Cascadia Subduction Zone (CSZ) and other extreme near-field tsunami hazards, coastal residents and tourist must evacuate within 15 to 30 minutes immediately following intense ground-shaking and will be confronted with an array of choices: Should I evacuate on foot or by car? Alone, or find friends and family first? Head for high ground far away, or seek shelter a nearby building? How will the roads and bridges be affected by the preceding earthquake? In this project, we integrate the disciplines of tsunami inundation science, sociology, and civil engineering to investigate how decision-making by individual evacuees with respect to milling time, mode choice, and destination affects their life safety. We use an Agent-Based Model (ABM) to create credible scenarios for near-field tsunami evacuation. The ABM integrates (1) the time-dependent tsunami inundation computed separately using NOAA's ComMIT/MOST model, (2) population layers to account for variations in population density of residents and tourist, (3) evacuation route network including roads, bridges and foot paths for multi-modal transportation, and (4) evacuation destinations for horizontal and vertical evacuation. For this project, we apply the ABM at two locations: the city Seaside, OR, and South Beach State Park in Newport, OR. In the Seaside scenario, we show how unplanned network disruption - e.g. the partial or total failure of bridges due to the preceding earthquake - will affect life safety and show how the ABM can be used to provide retrofit strategies. For South Beach, we show how alternative routing can have a substantial impact on life safety. The ABM shows results that are initially counterintuitive. For the Seaside example, resource allocation for bridge retrofit favors investments in nodes and links not necessarily in close proximity to population centers. For the South Beach example, the routes which provide for the lowest risk (maximum life safety) are not always those with the shortest distance between the evacuee and the shelter area. Detailed investigations with the Agent Based Model reveal that the evacuation planning for both of these cases requires a synthesis of the tsunami inundation dynamics, social conditions and decision making, and the potential for unplanned disruptions of the engineered infrastructure.