Investigation of smoke exposure during the fire season in Australia: Importance of quantifying plume injection heights

Wildfires can have a significant impact on air quality in Australia during severe burning seasons, but incomplete knowledge of the injection heights of smoke plumes poses a challenge for quantifying smoke exposure. In this study, we use two methods to quantify the percentage of fire emissions injected above the planetary boundary layer (PBL), and we further investigate the impacts of plume injection heights on daily mean surface concentrations of fine particulate matter (PM2.5) from smoke in 11 key cities in southeastern and northern Australia from 2009 to 2020. For the first method, we rely on assimilated PBL heights from the NASA Modern-Era Retrospective analysis for Research and Application version 2, together with the climatological monthly mean vertical profiles of smoke emissions from the Integrated Monitoring and Modelling System for wildland fires. For the second method, we develop a novel approach based on satellite observations and a random forest, machine-learning model that allows us to directly predict the percentage of smoke injected above the PBL in each grid cell. We apply the resulting injection percentages to the smoke PM2.5 concentrations simulated by the Stochastic Time-Inverted Lagrangian Transport model for each target city. Results show that smoke PM2.5 contributed 2% to 51.5% to total PM2.5 at the surface in key Australian cities on average during wildfire seasons over the last decade, driving the seasonal and interannual variability of surface PM2.5. Characterization of the plume injection heights greatly affects estimates of surface daily smoke PM2.5, especially during severe wildfire seasons, when intense heat from fires can loft smoke high in the troposphere. Both plume injection methods diminish the model overestimate of daily surface smoke PM2.5 compared to observations. However, using climatological vertical profiles does not capture well the spatial and interannual variability of injection heights. The use of machine learning to predict the percentages of plumes injected above the PBL leads to the best agreement of daily smoke PM2.5 with surface measurements, especially in northern and southeastern Australia, where many fires occur.