Large Fire Sample Points, Santa Barbara County, 1980-2019

These sample points within large fires from 1980-2019 were used to train and test a large fire occurrence probability model for Santa Barbara county. We used fire perimeter data for 201 large (> 40-ha) fires from the State of California Fire and Resource Assessment Program (FRAP) fire history database, which covered our entire time period of interest. We generated a random sample of points within fire perimeters using the method developed by Davis et al. (2017): the number of random points generated within each fire perimeter was equal to the square root of the area within the perimeter divided by 40. We forced a minimum distance of 500 m between the random points (based on testing of best distance in the northern Sierra Nevada - Southern Cascades; Syphard et al. 2018) to increase spatial independence and reduce spatial autocorrelation and model performance inflation (Veloz 2009, Boria et al. 2014). This process resulted in establishing 840 total sample points within large fire perimeters. We reserved 20% of those points for model evaluation, leaving a total of 669 points for model training.