Enhanced fine-scale satellite-based wildfire exposure assessment in wildland–urban interface communities using deep learning

Background Wildfires occurring within urban areas pose significant risks to structures and infrastructure. The wildland–urban interface (WUI) regions in Canada face heightened wildfire challenges due to climate change, emphasising the need for accurate fire exposure assessments. Aims This study aims to enhance the precision and scalability of WUI fire exposure assessment by automating fine-scale vegetation classification and integrating slope analysis using publicly available satellite imagery, aligned with Canadian national guidelines. Methods A computer vision framework is introduced, leveraging satellite imagery and convolutional neural networks to characterise vegetative fuels. Topographical factors, such as slope, are integrated to generate topo-fuel maps, enabling the extraction of exposure levels within each zone, and finally the assessment of fire exposure levels for individual buildings. Key results The framework is applied to the City of Edmonton as a case study, demonstrating its capability to assess WUI fire exposure levels at an urban scale. Conclusions The proposed method provides a scalable and automated solution for assessing fire exposure levels in WUI regions. Implications This framework supports fire management authorities in enhancing preparedness and mitigation efforts for WUI fire risks, contributing to safer urban planning and infrastructure resilience in wildfire-prone regions.