Comprehensive Fuel and Emissions Measurements Highlight Uncertainties in Smoke Production Using Predictive Modeling Tools

Abstract

Predictive modeling tools, such as the First Order Fire Effects Model (FOFEM), are used to generate estimates of the effects from wildland fires, including fuel consumption and smoke emissions. Given the use of such models in planning and forecasting for wildland fires, coupled with the adverse health and climate impacts of smoke, there is a need to understand the sensitivity to model inputs and processes, evaluate smoke production, and identify critical uncertainties. In this work, FOFEM was applied to a series of prescribed burns at the Blodgett Forest Research Station (BFRS), a western mixed coniferous forest in northern California, adapted to a frequent low-severity fire regime. We evaluated the sensitivity of predicted smoke emissions to parametric uncertainty in model inputs, including fuel characteristics (composition, loading, and moisture) and emission factors (EFs), and structural uncertainty in the consumption model. The results of the modeling simulations and comparison with a unique and comprehensive suite of fuel and emissions measurements suggest that in this application of FOFEM, fuel loadings based on land cover maps had the highest uncertainty and resulted in the largest sensitivity in predicted smoke emissions. The use of land-cover-based fuel loading values significantly underpredicted gas and particle emissions from the prescribed burns by up to ∼80% for carbon monoxide (CO) and carbon dioxide (CO₂) and by up to ∼85% for fine particulate matter (PM_2.5). Improvement in the predicted smoke emissions could specifically be achieved by more accurate fuel loading data, particularly for duff and coarse wood, the consumption of which generated the majority of gas (∼50–70%) and particle (∼65%) emissions. For individual gaseous nonmethane organic compounds (NMOCs), predicted emissions were additionally sensitive to uncertainty in EFs, demonstrating that the accurate prediction of these NMOCs requires accurate representation of fuel consumption as well as representative EFs.