Understanding broadscale wildfire risks in a human-dominated landscape

Abstract

Broadscale statistical evaluations of wildfire incidence can answer policy-relevant questions about the effectiveness of microlevel vegetation management and can identify subjects needing further study. A dynamic time series cross-sectional model was used to evaluate the statistical links between forest wildfire and vegetation management, human land use, and climatic factors in Florida counties. Four forest wildfire risk functions were estimated: one for fires regardless of ignition source, and three others for fires of specific ignition sources: arson, lightning, and accident (unintentional anthropogenic). Results suggest that current wildfire risk is negatively related to several years of past wildfire and very recent site prep burning, and risk is positively related to pulpwood removals. The effect of traditional prescribed burning on wildfire risk varies by ignition source. El Niño-Southern Oscillation (ENSO) sea surface temperature (SST) anomalies were also significantly linked to forest wildfire risk, but a measure of the wild land-urban interface was not significant. Although these county-level results hold promise for aggregate risk assessment, modeling at finer spatial and temporal scales might further enhance our understanding of how land managers can best reduce the longer term risk of catastrophic wildfire damages.