Application of an ensemble-trained source apportionment method to speciated pm2.5 data at the st. louis midwest supersite
Maier, Marissa Leigh
MetadataShow full item record
Four receptor models and a chemical transport model were used to quantify the sources of PM2.5 impacting the St. Louis Supersite (STL-SS) between June 2001 and May 2003. The receptor models utilized two independent datasets, one that included ions and trace elements and a second that incorporated 1-in-6 day organic molecular marker data. Since each source apportionment (SA) technique has its own limitations, this work compared the results of five different SA approaches to better understand the biases and limitations of each. The source impacts predicted by these five models were then integrated into an ensemble-trained SA methodology. The ensemble method offered several improvements over the five individual SA techniques. Primarily, the ensemble method calculated source impacts on days when individual models either did not converge to a solution or did not have adequate input data to develop source impact estimates. Additionally, the ensemble method resulted in fewer days on which major emissions sources (e.g., secondary organic carbon and diesel vehicles) were estimated to have either a zero or negative impact on PM2.5 concentrations at the STL-SS. When compared with a traditional chemical mass balance (CMB) approach using measurement-based source profiles (MBSPs), the ensemble method was associated with better fit statistics, including reduced chi-squared values and improved PM2.5 mass reconstruction. A comparison of the different modeling techniques also revealed some of the subjectivities associated with applying specific SA models to the STL-SS dataset. For instance, positive matrix factorization (PMF) results were very sensitive to both the fitting species and number of factors selected for the analysis, whereas source impacts predicted in CMB were sensitive to the selection of source profiles to represent local metals processing emissions. Additionally, the different SA approaches predicted different impacts for the same source on a given day, with correlation coefficients ranging from 0.03 to 0.66 for gasoline vehicle, -0.51 to 0.85 for diesel vehicles, -0.29 to 0.86 for dust, -0.34 to 0.76 for biomass burning, 0.22 to 0.72 for metals processing, and -0.70 to 0.68 for secondary organic carbon. These issues emphasized the value of using several different SA techniques at a given receptor site, either by comparing source impacts predicted by different models or by utilizing an ensemble-trained SA technique.