Tornado Occurrence Modelling and Equivalent Tornado Design Wind Profile for Canada

Qian Huang, The University of Western Ontario

Abstract

The tornadoes can be devastating to structures and infrastructure systems. They can cause fatalities and economic losses. In this thesis, the tornado hazard for Canada is systematically assessed. The use of three spatial point processes (i.e., Poisson, zero-inflated Poisson, and negative binomial models) are first investigated in a hierarchical structure to predict the tornado occurrence rate by considering the annual cloud-to-ground lightning flash density and annual thunderstorm days as covariates. An existing historical tornado catalogue with reported tornado events up to 2009 is considered and the population bias effect in estimating the tornado occurrence rate is included. Both the Bayesian technique and the maximum likelihood method are used to estimate the model parameters. It was shown that the NB model outperforms the other two judged based on statistical criteria such as the Akaike information criterion or Bayesian information criterion.

As the existing tornado catalogue is not up to date, a new tornado catalogue with reported events up to 2019 is developed. By using the existing and new tornado catalogue, the modeling of tornado occurrence rate focused on southern Ontario – a region that is prone to tornado in Canada – is carried out. Also, an assessment of the rate of the tornado striking a site of interest is estimated by incorporating the statistics of the tornado path characteristics. The striking rate and an adopted probabilistic tornado wind field model are then used to estimate the tornado wind velocity hazard maps. It is shown that for a point-like structure, a simple tornado design wind profile could be developed based on the specified return period value of the tornado wind velocity at different heights. This concept is extended for a line-like structure based on the return period values of the bending moment and shear force along the height of the structure. In all cases, the tornado design wind profile is expressed as a function of the tornado wind speed at 10 m height.

The development of the tornado wind hazard map and tornado design wind profile is extended to Canada. For the development, the Poisson model, zero-inflated Poisson model, and negative binomial model, as well as the adaptive Gaussian kernel smooth technique, are used to model tornado occurrence. The modeling again takes into account the tornado reporting bias due to population density and uses the cloud-to-ground lightning flash density and the thunderstorm days as the explanatory variables. The tornado wind velocity hazard maps for Canada, in terms of wind velocity at 10 m height above the ground surface for different return period T, V_T(10), are presented. Most importantly, it is shown that a simple equivalent tornado design wind profile can still be developed for regions with significant different tornado activities. The developed tornado design wind profile only depends on V_T(10), which greatly facilitates its potential use for structural design subjected to tornado wind load and its possible implementation in structural design code and standards. It is shown that the tornado winds could dominate the wind hazard as the length of the footprint of an infrastructure system or the area of the footprint of a structure increases. This indicates that the consideration of tornado winds is necessary for a spatially extended building complex and infrastructure system that are critical for the safe operation of the society.