Fatality and Injury Prediction Model for Tornadoes

Abstract

Over the last 1 years, tornadoes have caused the highest number of fatalities among natural hazards in the United States. Supercell-spawned tornadoes can be in excess of 1 km wide and often have long tracks that can pass through an entire community, resulting in numerous casualties and potentially several billions of dollars in direct and indirect damages for a single event. For example, the EF5 tornado in Joplin, Missouri on May 22, 211 is the deadliest and costliest tornado in the US since 195. In order to study the resilience of a community, life safety constraints should be satisfied along with other resilience metrics at the community level. In this study, a multivariate regression model is presented to assess the expected number of injuries and fatalities caused by a tornado as a function of tornado intensity, the number of people located in the tornado path, and the tornado path length. Moreover, the time of the day and month of the year when a tornado happens and the property damage caused by a tornado were used in predicting injuries and fatalities. In this regard, the United States tornado database and the US census database at the block level were used to provide a dataset for the regression model. Two prediction models for injuries and fatalities were developed for moderate tornadoes (i.e., EF and EF1), strong tornadoes (i.e., EF2 and EF3), and violent tornadoes (i.e., EF4 and EF5). The proposed models outperform existing predictive models in that they estimate casualties better with fewer explanatory variables. This is primarily due to the use of block level census data instead of county level, which is shown to be critical at tornado model scale. These models can be used to design a resilient community or upgrade an existing community by applying a probabilistic life safety constraint at the community level beside the resilience metrics.