Multiscale local definition of the wildland-urban interface to mitigate fire risk: an evidence-based approach

The Wildland-Urban Interface (WUI) is the spatial manifestation of the coupling of human communities and ecosystems, and wildfire is the most prominent issue. The WUI accounts for large percentages of fire prevention and suppression expenditures because it is where most human fatalities and structure losses occur. Therefore a fire-risk based definition of the spatial delimitation of the WUI may be critical to properly distributing prevention action and management investments to obtain the maximum social return. We present the first methodological approach that can be used to delineate the WUI based on a fire risk assessment. To accomplish this, we developed a geographical framework to model fire risk with the most prominent drivers and their interactions to define spatial explicit thresholds of the WUI. We built a Bagged Decision Tree (BDT) model to quantify fire risk based on Human Activity, Geographic and Topographic, and Land Cover variable interaction with fire ignition. For national and subnational threshold definition, we used Partial Dependence Plots (PDP) to analyze relationships between individual variables and predicted responses. A PDP can show the inflection point where a management action could potentially attain the best social return for decreasing fire risk. We find that the spatial threshold can vary more than double between subnational areas using the local fire risk-based approach. Subnational threshold definition accounts for 52% of fires in 3.4% of the national territory where lives 63% of the human population versus the conventional threshold or even nationally defined threshold that accounts for 36% and 54.4% of fires but in 3.3% and 4.3% of the land respectively. This multi-scale approach can be used to identify both general thresholds for large-scale applications as well as local thresholds for defining the WUI both operationally and empirically to determine optimal management areas.