Uncertainties in prescribed fire emissions and their Impact on smoke dispersion predictions

Prescribed burning (PB) is practiced throughout the Southeastern U.S. for its important ecological and safety benefits such as preparing for tree seeding and planting, controlling disease and tree competition, managing understory debris, perpetuating fire-dependent plant species, protecting endangered animal species, improving wildlife habitat and forage for grazing, and reducing wildfire risk. However, smoke and air quality concerns constrain PB operations. Smoke-related incidents and accidents are the major local concerns while contributing to air quality problems in nearby urban areas is the primary regional concern. The U.S. EPA lists PB as the third largest source of fine particulate matter (PM2.5) in the U.S., emitting 12% of the total PM2.5 mass nationwide. With other sources being controlled and trending downwards, PB emissions are poised to be the single most important source of PM2.5 in Southeastern U.S., in the future.

Responsible practice of PB requires accurate predictions of the smoke and air quality impacts. Several modeling tools are available for use in PB management today. These models estimate how much smoke or pollutant emissions a burn will produce, how high the smoke plume will rise, how it will disperse under current meteorological condition, where it will be transported, and whether the pollutants it contains would burden air quality in urban areas. Some models are simple while others are complex. In general, complex models employ more accurate representations of the emission, dispersion, and transport processes, but they require greater computer resources and more computation time. One commonality to all models is that their predictions are subject to uncertainty.

There are several sources of uncertainty in model predictions and this paper will focus on emission related ones. Emissions are estimated from the fuel loads, the fraction of the fuels that would combust under actual field conditions, and the amounts of pollutants emitted by burning each fuel type involved. Understory fuel amounts are estimated typically from the type of the growth and the fire return interval. Fuel moisture is a significant factor in determining the completeness of combustion. Several field and laboratory studies provide emission factors as pollutants emitted per unit mass of fuel consumed. Each one of these parameters and factors has uncertainty associated with it. Other sources of uncertainty in smoke dispersion include wind speed, wind direction, and atmospheric stability.

The Prescribed Fire Combustion and Atmospheric Dynamics Research Experiment (Rx-CADRE) conducted at Eglin Air Force Base near Niceville, Florida between 4-12 February 2011 provided a unique opportunity to evaluate the uncertainties in emissions and their impacts on smoke dispersion predictions. As part of Rx-CADRE, fuels were sampled at two burn plots. For various Fuel Characteristic Classification System (FCCS) fuel bed components, the amounts of fuel both before and after the burn were measured. The fuel loads were also estimated using the photo series for Southeastern United States. Fuel consumptions were estimated both from the difference of pre- and post-fire amounts and by using version 3.0 of the CONSUME model. Emissions from the fires were measured above the canopy with the aid of an aerostat lofted and tether maneuvered instrument package, along the downwind edge of the burn plots. The height of the smoke plume was measured with a LIDAR ceilometer positioned 1-2 km downwind of the burns and ground-level smoke concentrations were measures at various downwind locations.

Direct comparison of the measured fuel loads to those estimated by using photo series provided examples of the uncertainty levels introduced by the photo series approach. Similarly, comparison of the consumption estimated by CONSUME to the amount of fuels consumed based on the difference of pre- and post-burn amounts provided examples of the uncertainty in the consumption estimates. Finally, by comparing emissions measured using the tethered aerostat to the emissions per unit mass of fuel consumed from laboratory and other field studies, one can also estimate the uncertainty in emission factors. The uncertainties in fuel load, fuel consumption, and emission factors constitute the major sources of uncertainty in fire emissions. By propagating the emission uncertainties in dispersion models, their impact on dispersion predictions can be assessed. Emission uncertainties combine with model and parametric uncertainties, as well as uncertainties in meteorological inputs, to yield the uncertain dispersion predictions. These other uncertainties will be the subject of future work.

The emissions calculated using the fuel load and fuel consumption measurements along with the measured emission factors will be used as the nominal value in this study. These emissions will be input to Daysmoke, which is a plume dispersion model that calculates three-dimensional downwind smoke concentrations as a function of time. Then, the nominal emissions will be perturbed by the amounts of uncertainties attributed to the use of photo series for fuel load, CONSUME 3.0 model for fuel consumption, and emission factors from different sources. The spatial and temporal smoke distributions obtained by perturbing emission inputs will be compared to the base distributions that the nominal emissions yielded. The differences will be evaluated relative to the measured smoke dispersion parameters such as plume height and downwind smoke concentrations.

Prescribed burning is a necessary activity in the Southeastern U.S. but the resulting emissions can contribute to exceedances of air quality standards. In a dynamic management environment, decisions have to be made to moderate the potential impacts on air quality from PB, and to obviate the air quality constraints that limit PB. Knowing the uncertainties in emissions and how these uncertainties will impact smoke predictions is critical for effective decision making. Presumptively, predictions with known uncertainties should lead to fewer exceedances of air quality standards and subsequent benefits to human health and welfare, and allow more acreage to be treated by fire and more frequently, thereby improving ecosystem health and wildlife habitat, and decrease the risk of wildfires.