Historical Analysis of U.S. Tornado Fatalities (1808–2017): Population, Science, and Technology

Abstract

The record of tornado fatalities in the United States for over two centuries (1808–2017) and decadal census records have been examined to search for historical trends. Particular attention has been given to the response to population growth and expansion into the tornado-prone regions of the country. The region selected includes the Tornado Alley of the central Great Plains, the Dixie Alley in the southeastern states, and the adjoining states in the Midwest that collectively encompass a 21-state rectangular region. The data record has been divided into two subintervals, Era A (1808–1915) and Era B (1916–2017), each of which consists of three equal-length periods. Era A is characterized by a growing and westward expanding population along with a basic absence of scientific knowledge, technology, and communications (for prediction, detection, and warning). This is followed by a renaissance of discovery and advancement in Era B that contributes to saving lives. The aforementioned periods are defined by a set of notable events that help to define the respective periods. A death per population index (DPI) is used to evaluate the 21 states in each era; there is a rise of mean DPI values to a maximum of 1.50 at the end of Era A and a subsequent fall to 0.21 at the end of Era B. It is also shown for all three periods in Era B that the deadliest tornado states, in ranked order, are Arkansas, Mississippi, Alabama, and Oklahoma. Suggestions are presented for ways to continue the decreasing trend in DPI, which would imply that the death rate increase is not as fast as the rate of population increase (or would even imply a decreasing death rate).

Predicting property damage from tornadoes with zero-inflated neural networks

“…Additionally, mean and maximum conditional property damage predictions were lower for Alabama than Illinois, $187,698 and $1,112,483. Although other tornado impact work found Oklahoma to be among the deadliest tornado states [49], the fact that Oklahoma's mean ($159,561) conditional property damage prediction breaks the impact trend found in [57] is likely because of Oklahoma's proximity to the Dallas metropolitan area, which was found to have the greatest potential for a tornado disaster among 4 other metropolitan areas [58] supported by our relatively large $1,411,481 prediction in that area. Additionally, our predictive maps display an increased amount of high probabilities of damage for the American south -an area found by climatological studies to be prone to EF2 (strong) and EF4 (violent) tornadoes [30,67] as well deadly nocturnal tornadoes [68].…”

“…Along with developing a neural network approach for predicting outcomes in a long-tailed, non-linear, and zero-inflated case, we compare the usefulness of several variable sets and produce 2019 predictions of damage occurrence and property losses conditioned on a tornado occurring. These predictions are presented in the form of maps, with additional emphasis on well-studied states, such as Kansas [48], Alabama [48,49,57], Illinois [8,50,57,58],…”

“…Oklahoma [57,58,49], and Florida [51,55]. In addition to in-print visualizations, we present prototype dashboards that communicate these predictions accessibly (online, open source) and in a way that could be used by planners without expertise in machine learning or spatial analysis [20].…”

Tornado Fatalities in Context: 1995–2018

“…For example, 2018 was the least active killer tornado season over the past 24 years, yet, ten people still lost their lives. Some previous studies have focused on the number and location of tornado fatalities (Grazulis 1990;Hammer and Schmidlin 2000;Ashley 2007;Fricker et al 2017b;Paulikas and Schmidlin 2017;Agee and Taylor 2019). For example, Grazulis (1990) provides a historical narrative of significant and killer tornado reports since the late nineteenth century, while Hammer and Schmidlin (2000) and Paulikas and Schmidlin (2017) detail vehicle-related tornado fatalities over the periods 1900-1998, and 1991-2015, respectively.…”

“…For example, Grazulis (1990) provides a historical narrative of significant and killer tornado reports since the late nineteenth century, while Hammer and Schmidlin (2000) and Paulikas and Schmidlin (2017) detail vehicle-related tornado fatalities over the periods 1900-1998, and 1991-2015, respectively. Ashley (2007 examines spatial and temporal aspects of all tornado fatalities between 1880-2005, while Fricker et al (2017b) use a dasymetric method to map all tornado fatalities across the most tornado-prone region of the United States over the period 1955-2016. More recently, Agee and Taylor (2019) identify which states have been the most impacted by tornado fatalities dating back to 1808 using a death per population index. Consistency in the results of these works can be seen in the spatial extent of tornado fatalities, which have occurred in the largest numbers in the Mid-South and Southeast United States, most notably in the states of Alabama, Arkansas, and Mississippi.…”