Forest species composition and structure in northeastern Minnesota is tightly coupled with the size, frequency, and intensity of historic wildfires on this landscape as evidenced by the abundance fire-dependent forest species (e.g. Pinus banksiana Lamb.). Hence, fire is a salient disturbance agent on this landscape, in that it strongly influences nutrient cycling, carbon stores, and energy pathways between vegetation and the soil. While there are many positive forest ecosystem services associated with fire, there are also risks for an increasing population inhabiting the region. Because of this increase in the wildland-urban interface and shifts in fire frequency and severity induced by climate change, it is increasingly important to accurately model fire behavior and predict fire risk given the current state of fuels on the landscape. Such refinements in fire risk prediction will enable the development and implementation of efficient management strategies to maximize public safety.

Managers of the Superior National Forest (SNF) have faced lingering challenges in replicating historical wildfires without doctoring simulation inputs (e.g. increasing wind speeds). Crown fires were not propagating in simulations to the extent that they were in the field. As such, a pilot study was conducted over the course of a two-year period (2015-2017) to investigate fire behavior modeling issues faced by SNF forest managers.

The research presented here investigates the sensitivity of the fire area simulator FARSITE to these regionally-calibrated, spatially-explicit, landscape-scale fuel inputs. Initial tests focused on four new canopy bulk density (CBD) models derived from the 2015 pilot study. Additionally, we evaluated a canopy base height (CBH) model using the ground data from the 2015 pilot study as well as low-density LiDAR. Finally, we tested a crosswalked surface fuel model (FBFM) image based on a ruleset provided by managers of the Superior National Forest (SNF)

Two historical (2006) fires, Redeye and Famine, were used as proxies for simulation scenarios. A pairwise comparison of spatial correspondence metrics was used to analyze the recently calibrated forest fuels estimates to preexisting raster images provided by LANDFIRE. Results of this study provided evidence that the locally-calibrated images represented historical fire perimeters more accurately than LANDFIRE estimates for CBD and FBFM. However, the comparison between CBH estimates proved unsubstantial. These data products will allow a range of potential fire behavior options for the managers of the SNF to use in risk assessment and fuel management.