Indicator parameters for PCDD/PCDF from electric arc furnaces

The unintentional formation and release of persistent organic pollutants (POP) from industrial sources is of environmental concern and efforts are now made to reduce these emissions [The Stockholm Convention on persistent organic pollutants; United Nations Environment Programme: Geneva, 2001]. The emissions of chlorinated trace organics from electric arc furnaces (EAF) have been monitored on a regular basis in Sweden since the 1980s. Most analyses have encompassed not only polychlorinated dibenzo-p-dioxins (PCDD) and dibenzofurans (PCDF), but also chlorinated benzenes and phenols. Emissions of 2,3,7,8-substituted PCDD/PCDF from municipal solid waste incinerators (MSWI) can be modelled and predicted from analyses of chlorinated benzenes and phenols, which are suspected to be precursors in the formation process. The purpose of this investigation was to extend and update previously reported models with new samples from EAF, to describe the main sources of variation and to compare multivariate calibration with univariate regression. The measurement data consisted of 27 samples collected between 1987 and 2002 and analysed by two different laboratories. A general multivariate calibration model was able to describe 96% of the variation in the toxic equivalent quantity (TEQ) value over five orders of magnitude. Univariate regression models cannot account for changes in the congener pattern and thus gave a poorer performance. In plant-specific applications, the univariate approach did, however, perform equally well. It was therefore concluded that both multivariate and univariate regression models can be used in process optimisation studies, but that multivariate models are better suited for emission monitoring and evaluation of removal efficiencies in the off-gas cleaning systems.