Use of pattern recognition in the evaluation of PCDD and PCDF residue data from GC/MS analyses

We used the chemometric approach with SIMCA (soft independent modeling by class analogy) multivariate statistical programs, based on principal components modeling to examine complex PCDD and PCDF residues in environmental samples. This approach was employed to determine the acceptability of measured tetrachloro isotope ratios when three ion intensities, m^+., (m+2)^+., and (m+4)^+., were measured by monitoring multiple ions. Isotope labeled internal standards and native compounds were examined. Principal components contain information about similarity of samples and variability in measurement. The data for the observed ion ratios were merged with the theoretical ion ratios and a calculated set of data that deviated from the theoretical values over a range of +/− 10 % for each ion pair. Thus the plot of principal components sample scores contains the error boundaries for accepting data that meet this quality assurance criterion and identifies the samples and calibration data that are within the acceptable range. In addition, GC/MS peaks having more than one constituent with different degrees of chlorination were readily detected. Residues of PCDDs in eggs of herring gulls ( ) collected from colonies in the Great Lakes region in 1981 – 1984 were measured by GC/MS. Principal components modeling of the normalized residue data demonstrated that the sample residue profiles differed according to the sample's origin and that sample profiles remained remarkably similar over the 4-year period in which the samples were collected.