AN IMPROVED PSEUDO-DETERMINISTIC RECEPTOR MODEL (iPDRM) TO APPORTION AMBIENT PM CONSTITUENTS TO SOURCES IN TAMPA, FL

Abstract

In 2005, Park et al., developed a new Pseudo-Deterministic Receptor Model (PDRM) to apportion SO2 and ambient particulate matter (PM) constituents to local sources near Tampa Bay. Ambient pollutant measurements were fit to products of emission rates and dispersion factors constrained with a Gaussian plume model for individual sources. In our study, the original samples were reanalyzed by ICPMS for 10 additional elements to improve the resolving power. Chemical mass balance (CMB) terms were added to PDRM to allow fitting of background aerosol sources. More accurate, curvilinear plume trajectories were computed to predict arrival times in both surface and aloft layers. This allowed application of the PDRM complicated meteorological conditions, e.g. wind shifts. Predicted emission rates for particle-bound elements were constrained using chemical compositional information obtained from published source profiles for generic source types. Constraints applied to source emissions of known tracer species allowed the "conditioning" of predicted dispersion factors for those sources to tracer species concentration profiles to better determine the dispersion factor temporal profiles. This enabled the model to apportion pollutants to individual sources with intermittent emissions the omission of which in Park et al. lead to significant residuals. Excellent fits were obtained for all modeled pollutants: 14 of 22 species have Normalized Mean Square Error (NMSE) values of < 2.5% and 21 of 22 have values < 8%. These were improved for SO2 and 8 of 10 elements (by 7-35% for Al, Cu, Ni, Pb, and Zn) modeled by Park et al. Our predicted emission rates are in much better agreement with chemical compositions for generic source types. Key results include: (1) predicted SO2 contributions to ambient levels from a small, lead battery recycling plant that were reduced from 50-59% at its peak influence to a more reasonable 2-4%, (2) Pb/Zn ratios from that plant increased from 1.0 to 734 and better agree with published ratios of 67-440, (3) predicted Ni emission rates for one of the oil-fired power plants (OFPP) was increased by 100-fold (larger than Park's), and now better agrees with its published National Emissions Inventory (NEI) emission rate and with X/Ni ratios for generic OFPP emissions derived from EPA's SPECIATE database, and (4) our predicted emission rates for hazardous air pollutants and toxics from power plants agree with ~75% of those reported annual emission rates from NEI and Toxic Release Inventories (TRI) to within a factor of 5. This suggests that these reported data provide a good qualitative estimate of emissions, but should not be treated as accurate in a predictive model to quantify source emissions. It was also observed that the TRI values for As emission rates from coal-fired power plants are more accurate that their NEI values.