Fighting wildfires is a strategic operation. In fire-prone areas of the world, such as California and the Mediterranean Basin, it is important that managers allocate and position fire trucks, water bombers and human fire-fighters in locations that minimize the response time to reach new fires. Not only is this important to reduce risk to human lives and livelihoods, the financial demands of fighting a prolonged campaign against multiple fires demands that resources be used as economically as possible.
Characterizing the wildfire regime of an area (the frequency, timing and magnitude of all fires) can be very useful for this sort of planning. If an area burns more frequently, or with greater intensity, on average, fire-fighting resources might be better placed in or near these areas. The relationship between the frequency of fires and the area they burn is one the characteristics that is particularly interesting from this perspective.
As I’ve written about previously with colleagues, although it is well accepted that the frequency-area distribution of wildfires is ‘heavy-tailed’ (i.e. there are many, many more small fires than large fires), the exact nature of this distribution is still debated. One of the distributions that is frequently used is the power-law distribution. Along with my former advisors Bruce Malamud and George Perry, I examined how this characteristic of the wildfire regime, the power-law frequency-area distribution, varied for different regions across the continental USA (see Malamud et al. 2005). Starting with previously defined ‘ecoregions’ (area with characterized by similar vegetation, climate and topography) we mapped how the frequency-area relationship varied in space, finding a systematic change from east to west across the country.
More recently, Paolo Fiorucci and colleagues (Fiorucci et al. 2008) have taken a slightly different approach. Rather than starting with pre-defined spatial regions and calculating the frequency-area distribution of all the fires in each region, they have devised a method that splits a large area into smaller regions based on the wildfire data for the entire area. Thus, they use the data to define the spatial differentiation of regions with similar wildfire regime characteristics a posteriori rather than imposing the spatial differentiation a priori.
Fiorucci and his colleagues apply their method to a dataset of 6,201 fires (each with an area greater than 0.01 sq km) that burned between 1987 and 2004 in the Liguria region of Italy (5400 sq km). They show that estimates of a measure of the wildfire frequency-area relationship (in this case the power-law distribution) of a given area varies significantly depending on how regions within that area are partitioned spatially. Furthermore, they found differences in spatial patterns of the frequency-area relationship between climatic seasons.
Using both a priori (the approach of Malamud et al. 2005) and a posteriori (the approach of Fiorucci et al. 2008) spatial delineation of wildfire regime areas, whilst simultaneously considering patterns in the processes believed to be driving wildfire regimes (such as climate, vegetation and topography), will lead to better understanding of wildfire regimes. That is, future research in this area will be well advised to look at the problem of wildfire regime characterization from both perspectives – data-driven and process-driven. The approach developed by Fiorucci et al. also provide much promise for a more rigorous, data-driven, approach to make decisions about the allocation and positioning of wildfire fire-fighting resources.
Citation and Abstract
Fiorucci, P., F. Gaetani, and R. Minciardi (2008) Regional partitioning for wildfire regime characterization, Journal of Geophysical Research, 113, F02013
Wildfire regime characterization is an important issue for wildfire managers especially in densely populated areas where fires threaten communities and property. The ability to partition a region by articulating differences in timing, frequency, and intensity of the phenomena among different zones allows wildfire managers to allocate and position resources in order to minimize wildfire risk. Here we investigate “wildfire regimes” in areas where the ecoregions are difficult to identify because of their variability and human impact. Several studies have asserted that wildfire frequency-area relationships follow a power law distribution. However, this power law distribution, or any heavy-tailed distribution, may represent a set of wildfires over a certain region only because of the data aggregation process. We present an aggregation procedure for the selection of homogeneous zones for wildfire characterization and test the procedure using a case study in Liguria on the northwest coast of Italy. The results show that the estimation of the power law parameters provides significantly different results depending on the way the area is partitioned into its various components. These finds also show that it is possible to discriminate between different wildfire regimes characterizing different zones. The proposed procedure has significant implications for the identification of ecoregion variability, putting it in a more mathematical basis.