Prior to European settlement and 20th century management, low- to moderate-severity fires frequently occurred across the landscape, maintaining a significant area in old park-like and multi-story forest patches. Old multi-story forest patches, resulting from protracted fire-free intervals, were structurally complex, and represented later successional stages of development, provided integral habitat for a variety of species, and added functional diversity to the landscape. Many of these patches were subsequently eliminated by 20th-century timber harvesting or wildfires.
Historically, these “refugial” old forest patches persisted due to unique topographic and edaphic features that restricted fire access. Furthermore, characteristics of the surrounding vegetation and fuels matrix may have reduced the likelihood of high-severity fire, thereby contributing to their persistence. The objectives of this project were to identify pre-management era fire refugium patches within mixed conifer forests of the eastern Cascade Mountain region and to use a niche modeling framework to relate biophysical variables and characteristics of the surrounding vegetation matrix to their occurrence.
Unique vegetation patches were delineated from early 20th century aerial photography. Fire refugium patches were identified using criteria related to the amount of large tree cover, the number of canopy layers, and the size and canopy cover of understory tree layers, among others. Potential predictors of fire refugia included (1) biophysical variables defining topographic position and complexity, soils, climate, and potential productivity of each refugial patch, and (2) neighboring vegetation, fuel and topographic features immediately surrounding each patch. Traditional and machine learning algorithms were calibrated using these cofactors, and separate models were built (1) using various definitions of refugium (the binary response variable), and (2) for individual ecoregions within the study domain. Models were used to identify main cofactors associated with the presence of fire refugium patches, and to predict a continuous probability surface of fire refugium presence. Results suggest that strong topo-edaphic controls influenced fire refugium occurrence and that the presence of neighboring low flammability patches increased the probability of refugium occurrence. Results also indicate potential ecoregional differences in fire refugium models, suggesting top-down climatic and physiographic controls on spatial fire patterns. Model results are compared to current landscapes to highlight the most important functional transitions in the modern landscape. The implications of the landscape positions of refugial patches, characteristics of their surrounding neighborhood, and their relevance in the modern era are discussed.