An Approach for Assessing Misclassification of Tornado Characteristics Using Damage

Abstract

Tornado characteristics (e.g., frequency, intensity) are challenging to capture. Assessment of tornado characteristics typically requires damage as a proxy. The lack of validation in the Enhanced Fujita (EF) scale and the likelihood of rural tornadoes suggests that tornado characteristics are not accurately captured. This manuscript presents an approach to quantify the potential misclassification of tornado characteristics using Monte Carlo simulation for residential structures in rural areas. An analytical tornado wind field model coupled with fragility curves generates degrees of damage (i.e., DOD) from the EF scale in a wind speed to damage approach. The simulated DODs are then used to derive damage to wind speed relationships built from the National Weather Service Damage Assessment Toolkit (NWS DAT). Comparisons are then made between the simulated tornado characteristics and those derived from damage. Results from the simulations show a substantial proportion of tornadoes were ‘missed’ and path width and path length on average are underestimated. An EF4 rating based on damage is favored for EF3 to EF5 simulated tornadoes. A linear regression was utilized and determined damagebased wind speeds of different percentiles, damage length, damage width and the number of structures rated at a particular DOD were important for prediction. The distribution of DODs was also used to predict wind speed and the associated intensity rating. These methods were tested on actual tornado cases. Tornadoes that have the same damage-based peak wind speed can be objectively assessed to determine differences in overall intensity. The results also raise questions about the level of confidence when assessing wind speed based on damage.

Post Disaster Damage Assessment Using Ultra-High-Resolution Aerial Imagery with Semi-Supervised Transformers

“…In addition to visual data, researchers have utilized other dynamic data sources like wind speed, ground motion data like PGA (Peak Ground Acceleration), and response spectra [43]. A paper by Lombardo et al presents an approach to the use of Monte Carlo simulation to quantify the misclassification of tornado characteristics by establishing a relationship between the degree of damage and wind speed [44]. Yuan et al introduced a 1D CNN-based approach for damage assessment [45,46].…”