Nonstationary Markov Chain Model for Stopping-by Behavior on the way to Tsunami Evacuation Site

Abstract

When tsunami occurs, residents and visitors in coastal area should evacuate as soon as possible to an evacuation site or a safe building, in order to ensure safeness from tsunami inundation. However, for various reasons, it has been frequently observed in the past disasters that people did not evacuate immediately to the evacuation site, and lost time due to late departure or stopping-by behavior on the way to the evacuation site. In this study, Markov chain model for stopping-by behavior in tsunami evacuation with nonstationary transition probability was developed. The model includes three state, that are a state of being at home, a state of being out from home for one's business, and a state in an evacuation site. As a result of simulation, it was found that approximately 21 % of evacuees are exposed to high risk of human safety from tsunami inundation, in condition that tsunami arrival time is at 900 s after an earthquake. And it was also found that proportion of evacuees who are exposed to high risk from tsunami inundation would decrease at approximately 4 %, when residents who were not at home go evacuation sites directly, instead to go back home.