Self-organized patterns of wildfire sizes in California: Implications for landscape resilience

Background/Question/Methods Ecological resilience has been described as the amount of disturbance that an ecosystem could withstand without changing self-organized processes and structures. Wildfires are an important disturbance process in most western US ecosystems and are integral in shaping spatial and temporal vegetation patterns. It is theorized that self-organized systems are capable of withstanding a large range of disturbance sizes and intensities without significantly changing the resultant distribution of vegetation and disturbance patch sizes over space and time. In the past, power-law statistics have been used to describe self-organized behavior in wildfire event sizes, and we extend this research using several different methods to evaluate evidence for landscape resilience to wildfires during the 20th-century, throughout the state of California. We used an atlas of California wildfires (>1ha; 1950-2007) grouped at three levels within the Bailey ecoregion hierarchy to (1) identify self-organized patterns in wildfire size distributions across the state, (2) identify the process domain of the self-organized behavior, and (3) to quantitatively evaluate evidence for exogenous and endogenous controls on fire size distribution.

Results/Conclusions Within each ecoregion we found evidence for self-organized behavior in wildfire size distributions. Evidence included good fits of: (1) 2-3 parameter statistical distributions within the Pareto and Generalized Beta II (P/GB2) family of distributions over the entire range of fire event sizes; these distributions all have in common a power-law tail, (2) the Pareto I (power-law) distribution to the right-tail of the fire-size distributions, and (3) broken-stick regression models to the inverse cumulative distribution functions for fire sizes. For most ecoregions, self-organized properties were generally limited to fires within 100 to 10000 ha, indicating that meso-scale processes controlling fire sizes likely are acting at this scale. Scaling parameters averaged 1.7 for most ecoregions, which is well within prior estimates for wildfires in other regions. Potential bottom-up drivers, including topographic features such as aspect and slope patches, also fit well to P/GB2 distributions, and scaling parameter estimates matched closely with fire-size distributions. Significant differences in fire size distributions among ecoregions likely indicate top-down controls from broad-scale geological and climatic factors. These results suggest that ecosystems within California are likely resilient to wildfire disturbances <10000 ha and are best modeled using P/GB2 distributions. Likewise, wildfires across this range likely respond to topographic and broad-scale climate influences.