Ecological correlates of fuel dynamics and potential fire behavior in former upland prairie and oak savanna

Fire behavior has changed drastically in North America following Euro-American settlement. We sought to identify potential fire behavior in former prairie and savanna following forest succession, and to predict potential fire behavior from plant community data. We collected data on fuel loads, soils, topography, and plant communities from 239 plots at seven sites in the southern Willamette Valley, Oregon, USA. We defined eight plant cover types; an ordination based on tree species basal area by diameter classes and the cover of ground layer functional types supported these community types. Extant prairie and savanna plots had the lowest fuel loads, but there were no consistent differences in fuels among wooded communities. We used a fire behavior model, BehavePlus, to simulate potential fire behavior in our study plots. Because plant community data, standard fuel models and the ordination axes were poor predictors of fuels and potential fire behavior, we explored a new method of incorporating ecological data into predictions of fire behavior. We used classification and regression trees (CARTs) to find groups of plots that differed in their potential fire behavior. Although the best CART explained only 32% of the variance in potential fire behavior, the CARTs suggest ways to more effectively manage fire behavior – for example, indicating the importance of the legacy effects of savanna trees. Our results suggest that the legacies of succession on historic prairie and savanna have led to novel and heterogeneous fuels complexes, making it difficult to predict potential fire behavior from standard fuel models or community types. CARTs could be useful in other regions, both to indicate locations that should be management priorities and specific management techniques to attenuate fire behavior.

► Fuel models do not summarize fuel loads for former grassland ecosystems.
► There were no consistent differences in fuel loads among wooded communities.
► Descriptions of community structure can help to predict simulated fire behavior.
► Prairie and savanna restoration may reduce the risks associated with wildland fire.