The Value of AI-Driven Predictive Intelligence in Wildfire Relief Logistics: Evidence from the 2026 Ontario Thunder Bay 36 Wildfire

<title>Abstract</title> <p>The 2026 Thunder Bay 36 wildfire, the largest ever recorded in Ontario, burned more than 315,000 hectares and threatenedthe transportation corridors linking remote communities to essential relief supplies. Although AI-driven predictive intelligenceis increasingly proposed to support disaster management, its operational value for humanitarian logistics remains poorlyunderstood. We develop a hybrid framework that integrates wildfire-spread prediction with relief-routing optimization toevaluate when predictive disruption information improves logistics decisions. The framework couples a physics-informedfire-spread model with a multi-period humanitarian routing model through dynamic road-closure probabilities, validated onthe 2026 northwestern Ontario wildfire event. The hazard model accurately identifies observed road disruptions (ROC-AUCapproximately 0.80), yet better prediction does not translate into better logistics performance: risk-aware routing producedplans identical to reactive planning, because available schedule slack allowed relief operations to finish before major corridorsclosed. Predictive intelligence creates operational value only when the timing of network disruption intersects active serviceperiods, and communities dependent on a single corridor remain vulnerable regardless of routing strategy. These findingsclarify the conditions under which AI-driven predictive intelligence enhances disaster logistics and guide emergency managersbalancing investment in forecasting, routing optimization, and transportation resilience. Simulation code is available on GitHub.</p>