Evaluation of Emissions from R&D Facilities Using Stack Measurements

Abstract

Research and development (R&D) facilities may be required to estimate air chemical emissions to demonstrate compliance with federal and state regulations, or to manage emissions to avoid nuisance impacts from their operations. These emissions are difficult to estimate because R&D facilities typically use a large number of chemicals in small quantities and engage in numerous and diverse activities which can change over time. Although not required for compliance, the Pacific Northwest National Laboratory (PNNL) sampled air chemical emissions from facility stacks during 1998-2008. The purpose of the sampling was to provide data to compare estimated release fractions to those used for emissions estimates and to verify that methods used to determine compliance with air regulations and permits conservatively predict actual emissions. This unique data set was analyzed to compare emissions with regulatory criteria; determine relationships with chemical inventories, use quantities, and properties; and identify signatures of sources contributing to the emissions. For comparison with regulatory data, stack measurements were used as a basis to calculate 24-hr and annual average emissions and ambient air concentrations. The study included an extreme worst-case analysis maximizing emissions and alternate more realistic analyses using a Monte Carlo method that takes into account the full distribution of sampling results. The results from these analyses were then compared to emissions estimated from chemical inventories. Ambient air concentrations calculated from the measurement data were below acceptable source impact levels for almost all cases even under extreme worst-case assumptions. More realistic scenarios reduced the estimate significantly depending on the chemical and the mode of operation. Release fractions were calculated by dividing emission estimates obtained using a Monte Carlo technique on the measured data by a building chemical inventory quantity. Release fraction values had a wide range among chemicals and among data sets for different buildings and/or years for a given chemical. Regressions of release fractions and of mean emissions to chemical inventory and properties gave weak correlations. These results highlight the difficulties in estimating emissions from R&D facilities using chemical inventory data. Positive matrix factorization (PMF) was applied to stack measurements and, depending on the building, resulted in between 9 and 11 factors contributing to emissions. Some factors were similar between buildings, while others had similar profiles for two or more buildings but not for all four. At least one factor for each building was identified that contained a broad mix of many species, and constraints were successfully used in PMF to modify these factors to resemble more closely the off-shift concentration profiles.