Regional likelihood of very large wildfires over the 21st century across the western United States: Motivation to study individual events like the Rim Fire, a unique opportunity with unprecedented remote sensing data

Studies project that a warming climate will likely increase wildfire activity in many areas (Westerling and others 2002; Flannigan and others 2005, 2009; Littell and others 2009). These analyses are often of aggregate statistics like annual area burned, which are insufficient for analyzing changes in seasonality of fire events, the temporal resolution useful for fire management and understanding what drives individual events. Stavros and others (in press, a) show that very large wildfires (VLWFs >20,234 ha ~50,000 ac) may account for only the top two percent of all fires burned in the western contiguous United States, but they constitute a substantial fraction (approximately 33 percent) of aggregate area burned from 1984 to 2010, thus providing strong motivation for understanding what drives them. Using composite records of mean and 95% confidence interval of climate indices for individual fires within a region vs. these same indices the weeks leading up to, including, and post ignition, Stavros and others (in press, a) investigate the spatial and temporal variability of the VLWF climate space. The VLWF climate space was used to define explanatory variables for logistic regression models of the probability of VLWF occurrence with high accuracy of area under the curve (AUC) > 0.80. Assessments of this climate space show that relationships between climate and aggregate area burned may be driven by how VLWFs respond to climate, and that climate and weather both before and after ignition determine fire growth to VLWF size.