U.S. Joint Fire Science Program


Date of this Version


Document Type



Final Report: JFSP Project Number 08‐1‐6‐04


US government work.


Emissions from wildland (wild and prescribed) fires add to the burden of air pollution and can have adverse impacts on air quality and public health. Numerical models for dispersion and chemical transport, also known as air quality models, can be used to investigate the fire plume evolution and the smoke impacts. However, it is important that the predictive skills of smoke models be evaluated under a wide range of applicable conditions through systematic simulations of past events with existing data. Three models were evaluated in this research: CALPUFF, DAYSMOKE and CMAQ. Different prescribed burn and wildfire episodes occurring throughout the Southeastern US were simulated with the models to evaluate their performance. The abilities of the models to predict the observed PM2.5 levels were evaluated in detail by pairing model predictions at monitoring locations with observations. Models were also assessed at a diagnostic level by analyzing whether they succeeded in predicting observed PM2.5 levels for the right reason or, if they did not succeed, why they failed. The sensitivities of PM2.5 predictions to fire emissions were estimated using brute‐force and decoupled direct methods. From the results of this study CALPUFF could not be determined to be a suitable model for simulating the air quality impacts of fires. DAYSMOKE matched the field measurements of plume tops and ground‐level PM2.5 concentrations quite well and can be used for injecting fire emissions into regional‐scale air quality models. Model evaluation indicated that DAYSMOKE can be turned into a reliable a short‐range smoke‐impact prediction tool for land managers. On a regional scale, PM2.5 impacts of prescribed burns and wildfires are best predicted by air quality models such as CMAQ. The standard version of CMAQ underestimated prescribed fire impacts and severely under‐predicted wildfire impacts. CMAQ‐APT, in its publicly available form, is not suitable for tracking fire plumes. AG‐CMAQ requires that adaptation parameters be adjusted for adequate modeling of wildfires. Additionally, CMAQ predictions were found to be sensitive to the vertical distribution of emissions and the contributions calculated from first‐order sensitivities fully explained the modeled PM2.5 impacts. An uncertainty between 30‐50% was determined for fire emissions estimated with existing tools and study results indicated that modeled air quality impacts are a linear function of emissions in this range of uncertainty