LOGISTIC REGRESSION MODELING OF TSUNAMI-RELATED FIRE OCCURRENCE, BASED ON THE GREAT EAST JAPAN EARTHQUAKE

Abstract

 As shown in the Great East Japan Earthquake and other tsunami disasters, tsunami-related fire can occur even in tsunami inundation areas. Previous studies developed models to assess tsunami-related fire occurrence risk, using data from an assessment in the target area. A tsunami flow can, however, push flammable materials in the targrt area from other areas, that is an important feature of tsunamis. This study constructed logistic regression models in each of Yamada Town, Otuti Town and Kesennuma City to assess tsunami-related fire occurrence risk, using data from the Great East Japan Earthquake. The models included data that related with building damage in areas surrounding a target area, especially seaward areas, as well as that of the target area. Comparison of the constructed models showed that consideration of building damage in the seaward areas is important to predict the tsunami-related fire occurrence, while necessity of including topographic characteristics such as hills into stastical model is shown to futher improve the prediction accuracy.