Date of Award

May 2019

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Civil Engineering

Committee Member

Weichiang Pang

Committee Member

Thomas E Cousins

Committee Member

Brandon Ross

Committee Member

Ravichandran Nadarajah

Abstract

During the period from 1950 to 2015, the United States experienced more than 60,000 tornadoes resulting in more than 900,000 injuries and about 6,000 fatalities (NOAA, 2016). Compared to hurricanes, the impact of a tornado is much localized and the probability of occurrence at a given location can be extremely low. Therefore, it is not feasible to use solely the raw historical data or tracks to quantify the risk of tornadoes for a given structure or a city that has not been affected by historical tornadoes. In order to properly quantify the risk of tornado, there is a need to develop a stochastic tornado simulation model to generate a large database of synthetic tornado tracks to quantify the tornado hazard. To carry out tornado risk assessment, both a methodology to perform stochastic simulation of tornado tracks and a tornado risk analysis framework are needed for the continental United States and the details of these frameworks will be presented in the following study.

In Chapter 2, a methodology to perform stochastic simulation of tornado tracks and parameters is presented. The stochastic simulation framework contains three sub-models, namely, genesis model, tracking model and wind field model. The genesis model utilizes the kernel density estimation method to simulate the annual number of tornadoes and touchdown locations. The tracking model is utilized to generate the tornado intensity, path width, path length, heading direction, intensity and time/date of spawn. The wind field model was used to compute the tornado wind speed along the tornado footprint. The tracking model was calibrated using the historical tornado information maintained by the NOAA Storm Prediction Center (SPC). A database of 1 million years of simulated tornado tracks was generated using the Clemson high performance computing facility. The final simulated tornado track parameters include the tornado occurrence rate, intensity (EF scale), touchdown location, touchdown time, and path direction. All these parameters are geographic dependent, in other words, the simulated parameters vary spatially and depending on its spawn locations.

Chapter 3 presents a framework of develop the tornado hazard maps in United States. Using the simulated tornado database (Chapter 2), Hazard maps in United States for EF0-EF5 wind speeds have been developed for several different target structure sizes, and the target include point target, 0.08 mi2, 0.03 mi2, and 0.5 mi2 circular target, respectively. Relationship between tornado striking probability and target size have been investigated, and tornado hazard for a specific structure in United States can be interpolated from given location and size using the hazard maps.

In order to predict the tornado damage and improve the community resilience performance, a new approach of tornado scenario selection and damage estimation is proposed in Chapter 4. The damage area and peak wind speed have been calculated, for each tornado tracks which impact the study domain, to estimate the corresponding mean recurrence interval (MRI). The building locations and dimensions are determined using an image segmentation algorithm, and the damage state is evaluated using the fragility curves. Damage estimation for three tornado scenarios, selected according to damage area, peak wind speed and both intensity measures were conducted with different hazard level (MRI).

Share

COinS