A quantitative, empirical method for generating fuel-bed maps from plant cover estimates reveals thresholds and one-to-many fuel-to-plant type relationships

Abstract Background Computer-model simulations of wildfire behavior and associated wildfire risk are critically needed to inform the design and implementation of fuels treatments for proactive management of landscape wildfire risks. Available computer-models require maps of fire behavior fuel models (FBFMs) to simulate wildfire behavior, which are challenging to create for mixed woody/herbaceous vegetation. This problem is compounded in the fire-prone sagebrush steppe of the western USA, owing to the vast and remote areas of concern. One solution is to crosswalk readily measured vegetation variables (i.e., cover data) to FBFMs. To demonstrate this approach, we related FBFM assignments from 3700 recently burned and unburned plots to plant cover estimated using either line-point intercept (LPI, 25–50-m field transects), or grid-point intercept from aerial photos (GPI, digital, high-resolution, 6 m 2 areas), or coarsely over 13-m radius area in the field without guides (so-called ocular estimation), throughout the Great Basin. Results Random forest models of the cover-to-FBFM relationships were generated and their accuracy assessed using data withheld from parameterization, and then used to create decision trees for selecting FBFMs from vegetation cover. Examples of findings from random forest-informed decision trees included (1) 7–10% shrub cover was a threshold for a site to be classified as a shrub FBFM, and (2) individual grass FBFMs could be comprised of very different blends of large statured perennial grasses, exotic annual grasses, or litter cover—all of which differ markedly in fire characteristics. Agreement between modeled and observed FBFM classification was greatest for the ocular sampling, likely due to its larger plot size. Conclusions The models provide an initial quantitative framework for classifying FBFMs from historic and contemporary cover data in order to conduct fire simulation modeling on sagebrush steppe landscapes.