Simulating the expected value of wildfire potential outlooks: A decision problem

Annually, wildfires destroy millions of acres of land costing federal agencies billions for wildfire suppression. Over the past few decades wildland fires have become larger, more destructive and increasingly expensive. As a result of this trend, federal fire agency suppression costs have steadily increased. Well-planned, efficient initial suppression tactics reduce the ultimate size of a wildfire and the associated suppression costs. However, the ability to accomplish a timely initial attack of a wildfire depends on prior information received about wildfire risk. Wildfire risk is monitored by the Geographic Area Coordination Centers Predictive Services units, federal interagency organizations that provide informational products on wildfire risk to fire managers including daily area-specific updates to the Predictive Services 7-Day Fire Potential product. Probability of wildfire occurrence changes geographically and over time and as fire potential changes over different areas, fire managers reallocate wildfire suppression resources to better protect areas of concern. This thesis explores the expected value of 7 day forecasts of wildfire potential by simulating the differences in wildfire suppression costs between alternative allocations of resources using fire behavior software. Given different combinations of risk, expected suppression costs are simulated and a comparison is made between (1) a default allocation of resources (where risk information is not known) and (2) a least-cost allocation of resources (where risk information is provided). The difference between the two allocations is a proxy for the value of the information provided by Predictive Services.