Evacuating from wildfires
Forest fires, wildfire risk and smoke appear on 48 recent papers, split among economics, AI and operations research, and evacuation planning on 4. Fires now reach cities, and evacuations have failed with deadly results. Do evacuation models fit how people actually leave, and where should the roads, shelters and warnings be?
What prompted this
Since October 2025, 48 papers carry a wildfire, forest-fire, bushfire or evacuation keyword (19 in economics journals, 15 in AI, 10 in operations research): "forest fires" on 13, "wildfire smoke" on 7, "wildfire risk" on 6, "wildfire mitigation" on 4, "evacuation planning" on 4, "wildfire suppression" on 3. The economics papers measure damages, the AI papers detect fires, and the few evacuation models are rarely validated against real evacuations.
Questions
- In recent evacuations, when did people leave relative to the orders, which routes did they take, and where did congestion form, compared with what the models assume?
- Can mobile phone traces calibrate departure-time distributions by neighbourhood and warning type?
- Which investments (road capacity, staged warnings, shelters in place) reduce expected deaths most per dollar in the highest-risk towns?
Data that could help
Mobile phone mobility data during evacuations, evacuation order timings, road network data and fire spread records.
Next step
Reconstruct the departure times and routes of one recent evacuation from mobility data and compare them with the output of a standard evacuation model.