A Spatial Approach to Modeling United States Wildfires and Property Loss Implications

With climate undergoing unprecedented changes within the past several decades, wildfires have only become more frequent across the globe, with the United States being no exception. Sweeping wildfires have been the cause of total loss of entire neighborhoods, amassing billions of dollars lost in property every year. Despite recent advances in statistical methods and the field of actuarial science (the discipline of applying mathematical and statistical methods to assess risk in insurance), there is still a gap in accurately pricing wildfire policies in the insurance industry. This gap, in part, is due to the difficulty in predicting the frequency, severity, and location of catastrophic events. This thesis attempts to bridge the gap by applying the geostatistical technique, kriging, to wildfire data from the United States over the span of 25 years. With the wildfire size estimates produced from the kriging model, the potential construction of a predictive model of property loss that has a penalty based on the size estimates is explored. Building on this, a closer look at California is used as an example as to what data would be needed and how the data would be matched to the kriged values, providing a guideline for future applications.