Évaluation des outils de télédétection pour l'évaluation des risques d'incendie de forêt ferroviaire

Abstract

This study evaluates the capabilities and limitations of emerging drone and satellite-based wildfire risk assessment tools, supported by limited field trials conducted in collaboration with railways in Saskatchewan, British Columbia, and Northern Ontario. The objective is to assess how remote sensing technologies can enhance wildfire risk detection and monitoring, supporting evidence-based decision making for wildfire mitigation and safety planning. The drone-based tool focuses on small scale assessments of hazardous fire fuel vegetation over areas of less than one hectare. It employs artificial intelligence driven analysis of vegetation type, condition, color, and density to identify high risk areas on the rail right of way arising from factors such as highly flammable vegetation, infestation such as Mountain Pine Beetle, and dead vegetation. Field trials showed good performance in detecting stressed and dead vegetation and provided valuable insight into localized wildfire risk, although further refinement is needed to improve accuracy and distinguish more effectively between vegetation types and conditions. Stakeholder feedback suggests the tool could help validate risk at high priority sites and refine intervention strategies but would be most effective if integrated into a broader inspection program that also includes track and infrastructure condition monitoring, hydrology incident detection, landslide and flood risk assessment, and other safety and asset management applications. The satellite-based assessments complemented drone operations through large scale analysis. Fire exposure assessments were used to identify hot zones on the national rail network by mapping hazardous fuel layers and supporting mitigation planning and resource prioritization, while directional vulnerability assessments were performed at selected sites to evaluate potential fire spread trajectories and identify critical infrastructure and communities at risk. Examples of future work include advancing artificial intelligence algorithms for greater accuracy and more robust feature detection, adding additional functionalities to the drone platform to expand its inspection and monitoring capabilities, and integrating drone and satellite tools as a complementary two-level system for identifying hot zones at the network scale and confirming site specific risks. These advancements could strengthen wildfire risk assessment and mitigation efforts, ultimately enhancing public safety and improving the resiliency of Canada’s rail network.