A hybrid physical-statistical model for the estimation of insured loss from catastrophic severe storm activity over the United States and Canada

On average, severe thunderstorms and related hazards in the United States cause over 10 billion dollars of insured loss, annually. The (re)insurance industry typically bears the brunt of these losses to residential and commercial properties and automobiles, as typical homeowner insurance policies cover loss due to hail, wind, and tornado damage. In order to stay solvent in the event of a large-scale severe weather outbreak, insurance companies need a way to measure their exposure to these catastrophic events, especially events that are worse than can be anticipated using traditional methods of studying historical claims data alone. One method typically employed for this is the use of “catastrophe models”. These models consider not only the risk posed by the severe thunderstorm hazard, but also include complex modules that address construction practices, building response to hazard intensity, various insurance contract terms, and the inherent uncertainty that accompanies the modeling of such a complex system. This work will focus on the components of a recently developed model for estimating severe thunderstorm losses over the U.S. and Canada, with emphasis on the development of the hazard component of the model which incorporates both observational data, and meteorological index values through the use of a hybrid physical-statistical approach.