A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Remote Sensing and Geographic Information Systems

Thesis (M. Sc.) - Asian Institute of Technology, 2023

For Phuket, Thailand, stakeholders have strengthened preparedness and improved community resilience through conventional tsumami mitigation practices. While these practices are excellent as tools to manage risks and communicate evacuation plans, they are inadequate when considering the lack of numerical insights achievable by executing drills or experimenting tabletop exercises. An agent-based tsunami evacuation simulation is reported by many literature pieces to be a powerful tool to quantitatively study tsunami evacuation mechanisms and evaluate evacuation measures. This thesis aims to develop a tsunami evacuation simulation model for Patong and Kamala, determine the degree to which several human behavioral parameters influence the inundation risks, and demonstrate the effects of restricting certain parameters, such as an averaged milling time, on the evacuation outcomes. The simulation model is built with an agent-based simulation approach with the focus on three behavioral parameters which are an average milling time, a proportion of pedestrians, and a proportion of evacuees with route knowledge. The degrees of influence are determined with Sobol's sensitivity measures with the safety rate as a dependent parameter to the three behavioral parameters. Key messages of this study are: first, the two study areas' patterns of associations between the three parameters and the safety rate are similar, but patterns of influences of those three behavioral parameters to the safety rate are different; second, experimenting with scenarios of pedestrian ratio restriction in Kamala shows that as pedestrian ratio decreases the safety rate continuously increases. While the most impactful mechanism of evacuation in both areas is milling time, the secondary influential mechanism differs for the two areas. In Patong, it is the proportion of evacuees with route knowledge, whereas, in Kamala, the proportion of pedestrians is the second most influential. Lastly, decision makers can use the simulation model and the sensitivity findings to evaluate and improve their evacuation plans. Although traffic congestion has been a concern for many regions at risk of near-field tsunamis, this study's findings established a speculation that congestion might not be a concern by areas prone to far-field tsunamis given that evacuation protocols are being followed promptly. Furthermore, this study also laid a foundation that these insights regarding mechanisms of evacuation are site-specific.