Colloquium Speaker:
Dr. Kevin Granville, Ph.D.
University of Windsor, Assistant Professor
Title: Modelling Multi-day Evacuation and Return Decisions for Isolated Communities Under Uncertain Wildfire Smoke Forecasts
Abstract: Wildfire and smoke frequently threaten people and periodically lead to evacuations. We model decisions to evacuate and return for scenarios representative of some situations for isolated northern Canadian communities. These decisions are important and difficult for reasons including but not limited to: (1) the impacts of both smoke exposure and evacuation are in different ways harmful and life-threatening, and (2) smoke conditions can change rapidly and there is significant uncertainty in smoke forecasts. Transition probabilities for smoke hazard levels are modelled using a multinomial logistic GAM, and the evacuation decision model is formulated as a stochastic dynamic programming problem for potential use for insight and training. Numerical results from the model illustrate characteristics of model-optimal policies given impact weighting and we contrast simulated distributions of impacts (smoke, evacuation, total) given uncertain or varying lengths of perfect forecasts.
Day & Time: Friday, March 27, 2026, at 3:00pm
Location: Lambton Tower, Room 9-118
Counts toward seminar attendance for MSc and PhD students in Math & Stats.