Fire within the fire: The role of fire growth rate on patterns of fire refugia, severity and their drivers in forests of the Pacific Northwest

Cameron E. Naficy, Department of Forest Ecosystem & Society, Oregon State University, Corvallis, OR, Garrett W. Meigs, Washington Department of Natural Resources, Olympia, WA, Matthew J. Gregory, Department of Forest Ecosystems and Society, Oregon State University, Corvallis, OR, David M. Bell, Pacific Northwest Research Station, USDA Forest Service, Corvallis, OR and Meg A. Krawchuk, Forest Ecosystems and Society, Oregon State University, Corvallis, OR

Background/Question/Methods
Predictive models of fire severity are important tools for understanding the drivers of and future changes in fire severity. These models are often based on moderate- to coarse-scale biophysical predictors such as vegetation, topography, climate, and fire weather. However, even within a single fire event, dramatic shifts in the relative importance, scale, and interactions between endogenous and exogenous drivers may occur over the life of a fire as it grows and undergoes changes in critical behavior. This is especially apparent during large blowups, when major growth episodes may rapidly reorganize the drivers of fire severity. Here, we present results from an analysis of the drivers of fire severity in relation to fire growth rate from 2002-2017 in the Pacific Northwest to examine four central questions; (1) what are the geographic and temporal patterns of blowups, (2) are fire severity patterns contingent on fire growth rate and, if so, how do they respond to changes in fire growth rates, 3) how do the drivers of fire severity shift as a function of fire growth conditions, and (4) what biophysical factors contribute to blowup events. To evaluate these questions, we use Landsat burn severity estimates and MODIS fire detection and radiative power data to characterize temporal and geographic patterns of fire severity and growth. We pair these data with vegetation, topography, climate, daily weather, and fire behavior indices and use ecoregionally stratified boosted regression trees to evaluate shifts in the drivers of fire refugia, non stand-replacing and high-severity effects for forests of the Pacific Northwest.

Results/Conclusions
The frequency of blowups varied geographically and exhibited some temporal variation, but did not show a clear trend over the study period. Fire growth rate was a leading predictor in all fire severity models (e.g. fire refugia, non stand- fire, and high-severity fire), with higher probability of fire refugia and non stand-replacing severity and lower probability of high-severity fire where fire growth was slower. Differences in the drivers of fire severity were apparent for ecoregions with distinct structural and compositional characteristics. Fire growth models provided insights about the drivers of daily growth rates, but with reduced performance compared to fire severity models. This suggests that fire growth is likely driven by fine-scale endogenous fire behavior dynamics and macro-scale meteorological factors not captured in our models.