ClassiFiRxe: A data-driven tool to support prescribed fire planning and implementation

Despite a growing need to expand the pace and scale of prescribed burning to address the national fire deficit, there is a lack of data-driven methods to incorporate fire danger information into prescribed fire planning. To address this gap, we combine nearly two decades of prescribed fire and wildfire records with spatial fire danger data across Montana and Idaho to characterize when and where each fire type has occurred. We implement a nonparametric cutpoint analysis to identify objective thresholds from a normalized fire danger index, the Energy Release Component (ERC), by treating pile burns, broadcast burns, and wildfires as ordinal classes along an ERC continuum. A novel bootstrapping approach generates distributions of optimal thresholds, enabling creation of “overlap zones” that quantify uncertainty between fire types. These methods are used to build a simple planning guide, ClassiFiRₓe. The model shows strong performance at distinguishing fire types, with a cross-validated balanced strict recall of 60.2% and balanced full recall of 73.7%. Finally, we demonstrate how these tools can reveal fire class seasonality, estimate prescribed burn windows at local scales, and map categories across large landscapes. Such ClassiFiRₓe can be developed wherever sufficient fire weather and occurrence data exist. This data-driven framework provides a repeatable, interpretable and robust tool to support prescribed fire planning, helping restore fire to landscapes.