COMPARISON OF 2 MULTIVARIATE MODELING APPROACHES FOR THE SOURCE APPORTIONMENT OF INHALABLE PARTICULATE MATTER IN NEWARK, NJ

Two different receptor models were applied for source apportionment of inhalable particulate matter (IPM; D50 = 15-mu-m) in Newark, NJ, a complex urban airshed. The two models, a modified version of factor analysis/multiple regression (FA/MR) and regression on absolute principal components, were developed using 137 24-h samples collected during two summer and two winter seasons. Comparison of the respective model results indicated that the number and type of the sources resolved by the two approaches were similar. Source contribution estimates were within the respective 95% confidence ranges derived from the tracer or absolute principal component score regression coefficients. Larger differences in results were observed for source types contributing 1-mu-g m-3 or less to IPM. Source emission profiles determined by regression on principal components were also comparable, where applicable, to the partial profiles estimated from FA/MR results. In both cases, the estimated source emission profiles were in good agreement with reported source emissions composition. This study demonstrates the usefulness of using more than one modeling method as a tool for validating receptor modeling results. The approach should be further demonstrated for ambient aerosol components other than mass.