Causes of variability in pesticide and PCB concentrations in air near the Great Lakes

Data through 2001 from the Integrated Atmospheric Deposition Network (IADN) were used to investigate the causes of variability in gas-phase polychlorinated biphenyl (PCB) and pesticide concentrations measured near Lakes Michigan, Erie, and Superior. A multiple linear regression model that incorporates temperature and time was used explain the variability in the concentrations. Our approach used autocorrelation analyses of the residuals to help us determine the effectiveness of the regression. Autocorrelation plots for the in-use pesticide lindane indicated that an agricultural application cycle was also present in the regression residual data at all sites. The addition of parameters for this effect to the regression equation accounted for, on average, 16% more of the variability in the data. Similar analyses for the in-use pesticide endosulfan did not show an agricultural application effect. The banned compounds DDT and chlordane showed that temperature and time correctly accounted for the variability in the atmospheric concentrations of these compounds at all sites. In contrast to the other compounds, PCBs and hexachlorobenzene showed strong residual autocorrelation patterns near Lake Michigan of an unknown origin.