| The aim of this study was to study the possibility of using a principal component analysis (PCA) model with nonnegative constraints to apportion sources of polycyclic aromatic hydrocarbons (PAHs) in central Lake Michigan (CLM), and of polychlorinated biphenyls (PCBs) in the Milwaukee Harbor Estuary sediments. Seven sediment cores (A, C, D, E, F, G, and H) from the CLM areas were collected and analyzed for time records of PAH inputs to sediments. VC 6, 9, and 12, from the Milwaukee Harbor Estuary, were analyzed for eighteen PCB congeners. Time of deposition of sediment layers was calculated based on {dollar}rmsp{lcub}210{rcub}Pb{dollar} and {dollar}rmsp{lcub}137{rcub}Cs.{dollar} In addition, the PCA model was applied to several simulated data sets created by a Monte Carlo technique (cases 1-6, U.S. energy consumption records, and PCB congener profiles), and it was able to regenerate the original profiles.; PAH time record profiles of the seven CLM cores were unmixed and focus-corrected by {dollar}rmsp{lcub}210{rcub}Pb{dollar} and/or {dollar}rmsp{lcub}137{rcub}Cs{dollar} focusing factors. For the Milwaukee Harbor Estuary, 8 PCB congeners, and 7 layers from VC 6, 9 layers from VC 9, and 8 layers from VC 12, were used as inputs for the PCA and CMB model. Unmixed profiles revealed a number of features that were not seen in the original data, especially for cores F and G. Without unmixing, a number of maxima and minima for the CLM cores would have not been recognized. From the PCA results, Wisconsin petroleum and coal consumption records were identified for the first and the second factors, respectively. The predicted and the Wisconsin coal profiles exhibited similar trends with peaks for the 1946-51 and 1968-73 periods. The patterns of predicted and actual profiles for petroleum were also similar, however, predicted values were lower than actual values after 1964. Results from the petroleum profile analysis may indicate a limitation of the PCA model when coefficients of variation are not constant over the entire period of study. Factor score results showed a separation of the cores into two groups. Cores A, C, D, and E, were affected significantly {dollar}({lcub}>{rcub}36%){dollar} by coal sources (Port Washington) whereas cores F, G, and H, were impacted mainly {dollar}({lcub}>{rcub}76%){dollar} by petroleum sources (Milwaukee). This is consistent with the prevailing southwestern wind direction and chemical mass balance (CMB) model results for individual PAH compounds. The CMB results also demonstrated the dominance of coal from 1925-76, and of highway from 1983-92 for most of the seven cores. For the Milwaukee Harbor Estuary sediments, Aroclor 1016 and 1242 were major contributors to PCBs in cores VC 9 and VC 12 and Aroclor 1242 alone was dominant in VC 6. Aroclor 1254 and 1260 were of minor significance, but were identified as sources in all three cores. Both factor loadings and factor scores were examined and were found to give similar outcomes with respect to source identification. Results from the PCA model were also supported by the analysis based on a CMB model.; The PCA model appears to be a viable tool for PAH and PCB source apportionment. Although the PCA model can be operated with a minimum of information about sources, inputs such as results from the CMB model and source profiles proved to be used to validate the model prediction. The source apportionment results are essential for the design and implementation of a pollution source control program. |
