Developing a Prescribed Fire Forecasting Algorithm for the Southeast United States

Prescribed fires are common in the southeast U.S. during winter where they contribute to nearly half of annual smoke emissions in the contiguous U.S. These burns are ignited by fire managers to maintain forests, grasslands, and croplands and lower the risk of future wildfires. Due to their impacts on air quality, it has become increasingly necessary to provide forecast guidance for these burns. Dynamical air quality forecast models are now a mature part of weather and public health systems; however, they assume persistence of emissions through the multi-day forecast. Currently, few products exist to forecast the locations of prescribed fires multiple days in advance. Fire managers typically rely on a set of weather criteria for deciding whether to ignite a fire, which include threshold magnitudes for relative humidity, transport windspeed, Lavdas atmospheric dispersion index, and boundary layer depth. Here, we have developed a statistical algorithm that uses daily historical Hazard Mapping System satellite fire detections from 2015-2019 during winter months and calculates the probability of a fire being present on a given day for the number of the fire weather criteria that were met in NAM 12km analysis output. Preliminary results show that the probabilities of prescribed fires throughout the southeast generally increase as more of the fire weather criteria are met, implying that forecast meteorology can be used to determine whether fires are likely to occur days in advance. We also test the forecast skill of this algorithm against forecast meteorology and satellite fire detections with varying lead times in model initialization hour to quantify how much of the forecast error is due to meteorological forecast error versus that of our algorithm. In the future, this method could be useful in improving air quality forecast models by more accurately representing emissions from prescribed burns in the southeast U.S.