New Methods in Tornado Risk and Vulnerability Assessments

This dissertation includes a series of studies that present innovative methodologies to improve tornado risk and vulnerability assessments. Limitations of the historical tornado dataset are well known and relate to inconsistencies in data collection procedures, rating assessments, updates in technology, and public awareness. The limitations make it difficult to accurately evaluate tornado risk and vulnerability. Thus, the research presented in this dissertation aims to 1) improve tornado risk assessments using the historical dataset by accounting for known non-meteorological factors and 2) enhance tornado vulnerability assessments by utilizing a new dataset containing more precise damage survey data. This work includes three individual studies, two focused on risk and one on vulnerability, using different geographic scales. Tornado occurrence rates computed from the available reports are biased low relative to the unknown true rates. A method to estimate the annual statewide probability of getting hit by a tornado improves this low bias by using the average report density as a function of distance from nearest city center. The method is demonstrated on Kansas and then applied to 15 other tornado-prone states from Nebraska to Tennessee over the period 1950--2011. The adjusted rates are significantly higher than the raw rates and thus, the return periods are less than previously thought (closer to 1000 years). The expected annual number of people exposed to tornadoes has also increased for every state. The evaluation of tornado occurrences is improved using a statistical model that produces a smoothed regional-scale climatology. The model is applied to data aggregated at the county level, including annual population, annual tornado counts, and an index of terrain roughness. The model has a term to capture the smoothed frequency relative to the state average and is used to examine additional hypotheses concerning relationships of tornado activity with terrain roughness and County Warning Area. Tornado reports are found to increase by 13\% for a two-fold increase in population across Kansas after accounting for improvements in rating procedures. The pattern of spatially correlated errors also shows Kansas tornado activity to be consistent with the dryline climatology. The model is significantly improved by adding terrain roughness, which has a negative relationship with tornado activity and its flexibility is demonstrated by fitting it to data from Illinois, Mississippi, South Dakota, and Ohio. Advancements in technology have improved the collection of tornado damage survey data which can be used to enhance vulnerability assessments. The National Weather Service (NWS) Damage Assessment Toolkit (DAT) contains the most extensive GIS-based damage survey data available to the public which provides more precise damage path areas. These data are used with socioeconomic data in two statistical models. The models are developed to determine which factors are significant predictors of the incidence and magnitude of casualties while accounting for maximum EF Scale rating, total path area, and population density at the storm level. Percent unemployment is a significant predictor and produces the best model for the incidence of at least one tornado casualty. Although percent elderly generates the best model for predicting the magnitude of casualties, it is only marginally significant and its relationship is negative. The Southeast has the highest averages of the sensitivity factors considering all of the tornado events. These results highlight the need for heightened tornado awareness and preparedness as our exposure to these events increases due to our population continuing to expand. As demonstrated in this work, these methods can be used to enhance regional/local tornado forecasts, insurance risk estimates, public policy, urban planning, and emergency management and mitigation with the detection of spatiotemporal patterns in tornado activity (due to variations in climate) and vulnerability (due to changes in population demographics and urban sprawl). They can be employed to examine other geographic locations on multiple scales. They can also be adapted to study the patterns and relationships of other spatial and temporal phenomena.