Canadian Forest Service Publications
Considerations for modeling burn probability across landscapes with steep environmental gradients: an example from the Columbia Mountains, Canada. 2012. Parisien, M.C.; Walker, G.R.; Little, J.M.; Simpson, B.N.; Wang, X.; Perrakis, D.D.B. Natural Hazards 66(2):439-462.
Issued by: Northern Forestry Centre
Catalog ID: 34230
Available from the Journal's Web site. †
† This site may require a fee
Fire and land management in fire-prone areas can be greatly enhanced by estimating the likelihood of fire at every point on the landscape. In recent years, powerful fire simulation models, combined with an in-depth understanding of an area’s fire regime and fire environment, have allowed forest managers to estimate spatial burn probabilities. This study describes a methodology for selecting input data and model parameters when creating burn probability maps in difficult-to-model areas and reports the results of a case study for a large area of the Columbia Mountains, British Columbia, Canada. In addition to having particularly mountainous topography, the study area is covered by vegetation types that are poorly represented in fire behavior systems, even though these vegetation types have experienced considerable (if highly irregular) fire activity in premodern times (before 1920). Parameterization of the fire environment for simulation modeling was accomplished by combining various types of fire information (e.g., fire history studies, reconstructed fire climatologies), new technologies (high-resolution remotely sensed data, wind flow modeling), and—a must in data-limited areas—ample expert advice. In this study, we made extensive use of personal accounts from experienced fire behavior officers for the creation of model inputs. Despite difficulties in validating outputs of burn probability models, the multisource model-building approach described here provides a conservative, yet informative, means of estimating the likelihood of fire. Due to the data-intensive nature of the modeling and paucity of input data, an argument is made that modelers must focus on the inputs that are the most influential for their study area.