How do vegetation, fire weather, and land use control contemporary fires? (UCI)

Increased anthropogenic ignitions due to population growth and urban expansion, coupled with severe fire weather, result in high fire frequencies at the wildland-urban interface (WUI) (Keeley and Fotheringham 2003; Syphard et al. 2007). Most fire risk models use only hydroclimate and biophysical variables, and much less is known about how human activity and winds affect fire (Syphard et al. 2007; Westerling et al 2009; Spracklen et al. 2009). We will analyze the roles vegetation, climate, and land use play in controlling fire occurrence and the fire size, especially for large fires. We will develop separate empirical models for fire occurrence and burned area based on fire history, climate, vegetation, and land use. We will use spatially explicit burned areas derived from FRAP and augmented with Landsat back to 1983 (FRAP excludes most small fires), as well as vegetation cover data (such as species composition) annually derived from Landsat. We will first calculate the probability that a fire will occur at a monthly time scale. We will use a logistic regression method to derive a statistical probability of fire initialization at the first step. We will test the significance of population density and road accessibility as predictors in addition to the biophysical and hydroclimate variables such as antecedent climate variables and fire weather (especially Santa Ana winds) from the reconstructed climate data (Table 1). We will use the annual housing density data from the 1990 and 2000 U.S. Census (U.S. Census Bureau, 2002) at the finest demographic scale and the road data from the US TIGER 2000 GIS layers of roads (U.S. Census Bureau, 2002). We will explore grouping data on various levels of ecological units based on USDA Ecological Units map (Bailey et al. 1994) and identify the most significant variables which explain the most variances in fire occurrence within each ecological unit. The two dominant ecological sections are Southern California Coast including 10 subsections, and Southern California Mountains and Valleys with 19 subsections. We will calculate the percentage of area burned in each subsection on a monthly basis for each category. We will analyze how much variation in burned area can be explained by each variable and then build a linear regression model for the percentage of area burned with the most significant variables.