TA9 - Modeling and Optimization for Wildfire Operations
May 12 2026 10:30 – 12:10
Location: Luc-Poirier (green)
Chaired by Vittorio Nicoletta
4 Presentations
Predicting the Spread: A Multimodal Benchmark for Fire Propagation Modeling
Deploying resources against wildfires demands accurate daily propagation forecasts. To support this effort, we present a Canadian multimodal dataset, combining weather, fuel features, and satellite imagery, to train deep learning models on daily fire propagation. The work is currently underway to establish a robust benchmark on this dataset.
From Imagery to Insight: Automated Building Feature Extraction for Wildfire Risk Assessment in the Wildland–Urban Interface
The analysis of built-environment features plays a key role in supporting vulnerability assessments for natural hazards in wildland–urban interface areas. Conventional manual inspection methods are often time-intensive and challenging to scale. This study introduces a machine learning–driven tool that automatically derives building attributes from street-level and satellite imagery.
Understanding Evacuation Behavior in Wildfire Events: A Survey-Based Analysis
This study analyzes original survey data collected in three Quebec communities that experienced a first-time evacuation during the 2023 wildfire season. Using regression models, we examine preparation time and actions, mode of transport, and risk perception. Findings aim to support emergency planning and improve behavioral modeling of evacuation dynamics in remote, high social bound communities.
An Interactive Network Optimization Tool for Evacuation Planning and Visualization
We present an interactive decision-support tool for evacuation planning based on network optimization and traffic assignment. The system integrates shortest-path routing and BPR congestion functions to estimate clearance times and flow patterns, enabling scenario comparison and supporting methodological research in optimization under emergency conditions, especially related to wildfires.