Analysis of spatiotemporal patterns of significant tornado damages in the United States

• Main Author: Dong, Luojie
• Other Authors: Long Cheng
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Earth and Environmental Sciences; Tornado; Extreme weather; Data analysis
• Online Access: https://hdl.handle.net/10356/175219
• Institution: Nanyang Technological University

Tornadoes are destructive meteorological phenomena that cause severe damage to property and lives in the United States every year. This study analyses the spatial and temporal patterns of significant tornado damage, based on the absolute damage and the Fujita (F) / Enhanced Fujita (EF) scale. It is discovered that within the contiguous United States, the ratio of damage caused by the top 1% most costly tornadoes shows an upward trend from the year 1970 to 2023. A slightly declining trend is identified in the ratio of damage caused by EF4+ tornadoes. Within the three regions of interest of this study - the Great Plains, the Southeast United States and the Upper Midwest US, similar trends are observed; however, in the Southeast US and the Upper Midwest US, the ratio of damage caused by the 1% most costly tornadoes increased drastically over the years - doubling and tripling roughly, from 1970 to 2023. The decreasing share of the strong tornadoes objectively defined by the EF scale suggests the improvement in preventing losses from tornadoes, which could be due to improving forecasting systems and community resilience; however, the increasing share of damage of the most costly tornado suggests that extreme tornadoes are getting more extreme, or the tornadoes that are not so extreme are causing more damage over the years. Whichever scenario it might be, it is worth further investigating. The seasonal variability of the tornado occurrences and damages are studied as well. The number and damage of late-fall and winter tornadoes are found increasing. The project also identified the tornado outbreaks using fine-tuned parameters with the HDBScan clustering algorithm and discovered that the tornado outbreak damages are taking a growing share within all tornado damages. This might suggest a change in the characteristics of the tornadoes happening in the United States and warrants future studies on it.