Author(s): Joseph Kim; Tomoyuki Takabatake; Ioan Nistor; Tomoya Shibayama
Keywords: Evacuation Model; Agent-Based Model; GIS Model; Tsunami Inundation; Coastal Communities
Abstract: Emergency managers have studied and invested in structural mitigation measures to combat the dangers of tsunamis and floods. However, it is unlikely that any interventions can fully guarantee a community’s safety. Thus, soft measures such as evacuation planning have been recommended to complement the planning and construction of resilient infrastructure to mitigate the loss-of-life during tsunamis. Numerical simulation to assess a community’s evacuation potential has been done in largely two ways. Agent-based modelling (ABM) defines sets of rules that each agent in the simulation follows during the simulated evacuation. By doing this, the dynamic and complex interactions between the agents and the environment can be accounted for, but requires substantial specialized knowledge to implement. Geographical information systems (GIS) are much more common and widely used by city planners and emergency coordinators. While GIS has been previously used to simulate evacuations, it lacks the ability to incorporate the dynamic behaviours that ABMs can account for. The two evacuation modelling methodologies were compared through a case study for Tofino, BC, Canada, by assessing the state of tsunami evacuation preparedness and by locating the optimal location for an additional Vertical Evacuation Structure (VES). A direct comparison on the performance differences between the two methodologies has never been done before. The fatality rates and facility demand ratios calculated from the ABM and GIS evacuation simulations presented significant differences in magnitude, but showed similar distributions in results. Both methodologies converged on the optimal location for an additional VES to mitigate loss of life during a tsunami. It was found that while GIS may be able to provide initial insights on a community’s evacuation potential, ABM’s are generally recommended due to their ability to incorporate dynamic flood casualty estimations and agent interactions.