Effect of distance, meteorology, and burn attributes on ground-level particulate matter emissions from prescribed fires

Multiple air monitoring campaigns were conducted from 2003 to 2007 during the months of December through April in which time-integrated monitoring of PM2.5 was performed during 55 prescribed burn (PB) events administered under select meteorological conditions. The data were analyzed using a generalized additive mixed-model (GAMM) where the logarithm of PM2.5 concentrations were modeled as the sum of linear and non-linear functions of our covariates of interest after accounting for the confounding associated with the grouping structure of our data and spatial dependency. Our model explained 61% of the variance of log transformed PM2.5. The GAM component of the model (i.e. our covariates of interest) explained 40% of that variance and the mixed-model component (i.e. random-effects and spatial correlation) explained 21%. Within burn events, concentrations across our air monitoring grid were significantly influenced by the distance from burn perimeter to air monitor and the proportion of time the air monitor was in line with the wind field over the burn. Among burns, variation was significantly influenced by burn duration, burn size, mode of ignition, and regional background PM2.5; indicating the importance of these for fire management. Measures of burn intensity and meteorology were found to be insignificant in our model; however, it is possible that some of these effects were absorbed by the random-effects of our model. These findings demonstrate that downwind exposure is largely a function of wind direction combined with distance from the burn and that emissions from one burn to the next are dependent on burn size, duration, and mode of ignition.