Modeling seasonal variation in the total probability of wildfires

Abstract:

An equation is developed to estimate the probability of either a lightning-caused (PLF) or human-caused wildfire (PHF) occurring in the Black Hills region of South Dakota and Wyoming. Previous research has focused on obtaining probabilities for either type of wildfire by developing logistic regression equations separately. In this research, a total probability (PT = 1 – (1-PHF)(1-PLF)) is given for the likelihood of either a lightning-caused and/or a human-caused wildfire occurring during different months of the typical wildfire season and at a specific remote weather station. The probability of a lightning strike is derived from a logistic regression equation for each month of the convective season and is then used as a predictor for PLF. Similarly, predictors are developed for the human-caused wildfire equations through periodic logistic regression analysis to account for the day of the week and time of year. Other predictors used to calculate both PLF and PHF and thus PT, include daily fuels and weather observed in the Black Hills region.