Modelling fire perimeter formation in the Canadian Rocky Mountains

Abstract

Wildfires produce a mosaic of burned and unburned patches across varying temporal and spatial scales and provide a range of essential ecosystem services. Fire perimeters mark the separation between the burned and unburned matrix of a fire. Analysis of fire perimeters in the United States, Australia, and Alberta, have identified several key factors that influence the formation of a fire boundary, including fire environment variables, such as fuel, weather, and topographic conditions, as well as anthropogenic factors. We used matched case-control conditional logistic regression to assess the fire environment's influence on the formation of fire boundaries on the 2017 Verdant Creek Fire in the western Canadian Rocky Mountain region. Results indicated that fire boundary formation was strongly influenced by non-fuels, alpine and subalpine vegetation, wetland areas, and low and sparse shrub assemblages generally associated with avalanche paths. Fire cessation was more likely near waterways; however, the fire weather conditions that characterized most burning periods likely overrode other topographic influences. Fire cessation was most likely to occur one day following precipitation events when Vapour Pressure Deficit and Fire Weather Index values decreased. To explore the potential influence of spatial data resolution and fire mapping limitations on our results, we varied the distance between matched sample points selected on each side of the mapped fire perimeter during statistical modelling. Results demonstrated matched pairs of burned and unburned sample points were effective at representing discrete states when separated by at least 100 m and that insufficient or excessive separation between matched pairs confounds results, highlighting the importance of sensitivity analysis for determining the appropriate separation distance to represent burned and unburned states. We demonstrate a practical application of the model to predict spread potential, enabling rapid visual assessment of landscape locations most likely to limit fire spread.