Defensible Space, Housing Density, and Diablo-North Wind Events: Impacts on Loss Rates for Homes in Northern California Wildfires

If a house is exposed to a wildfire, what is the probability that it will be destroyed? How is the risk of loss affected by vegetation cover near the home (i.e., defensible space), the proximity to other homes, and wind levels? This study addresses these questions with an analysis of 36,777 single-family homes involved in ten recent Northern California wildfires. Two logistic regression models are constructed, one for Diablo-North Wind (DNW) fires and another for fires with more moderate winds. Vegetation cover within 50 meters and housing density within 100 meters of each house are identified as statistically significant variables. But the models including those two variables alone are relatively weak predictors of structure loss. The addition of an autocovariate derived from the outcomes for nearby houses substantially improves prediction accuracy. The autocovariate partially accounts for events during fires, such as wind changes or structure-to-structure fire spread, which influence the fate of multiple homes in close proximity. The effect on classification accuracy is illustrated for the Coffee Park neighborhood in the 2017 Tubbs Fire. Increases in housing density appear to have little effect on loss rates in moderate wind fires, but can raise loss rates by 35% in DNW fires. A 10% reduction in vegetation cover near homes is estimated to reduce loss probability by 4-6% in most situations, but by only 1-2% when high winds are combined with high housing density. Loss rates are 20-60% higher in DNW fires compared to moderate wind fires for the same levels of vegetation cover and housing density. Previous studies and Red Flag Warning data indicate that the San Francisco Bay Area is most at risk for Diablo-North winds, followed by the Northern Sierras. The higher elevations found in the Sierras south of Lake Tahoe tend to reduce the chances for DNW-type events.