Impact of Measurement Uncertainties on Receptor Modeling of Speciated Atmospheric Mercury.

Receptor models have been used in the source apportionment of atmospheric mercury (Hg) and other air pollutants. These models make use of ambient monitoring data as well as back trajectory data in some models for the purpose of identifying potential types of emission sources, atmospheric Hg processes, and locations of major source regions. A summary of the major receptor methodologies and findings from speciated atmospheric Hg studies have been published in a review paper1. One of the recommendations from the review paper is assessing the extent that uncertainties in gaseous oxidized Hg (GOM) and particle-bound Hg (PBM) measurements affect receptor modeling results. GOM and PBM are operationally-defined Hg(II) compounds, such as mercuric halides, mercuric sulfate, mercuric nitrite, and mercuric hydroxide2. Measurement of GOM and PBM in ambient air is challenging because they are typically present in trace concentrations3 (pg m−3 range). The exact chemical forms have not been determined; therefore, calibration methods are not available to determine the accuracy and uncertainties of GOM and PBM measurements. Currently the uncertainties are estimated from intercomparison of measurements from different Hg measurement methods. Other sources of measurement uncertainties during sampling include ozone and water vapor interferences4,5,6, chemical reactions2,7,8, and variable collection efficiencies depending on the form of GOM9,10. These uncertainties are believed to be contributing to the underestimation of GOM. PBM measurements may be biased low or high and the extent of the uncertainties are not well-established3. Furthermore, GOM may not be easily separated from PBM during sampling and analysis due to the temperature dependency of Hg(II) gas-particle partitioning. Therefore, it may be considered more accurate to combine GOM and PBM into a single reactive Hg parameter3,11. In this study, the effect of GOM measurement uncertainties on Principal Components Analysis (PCA), Absolute Principal Component Scores (APCS), and Concentration-Weighted Trajectory (CWT) receptor modeling results were investigated. The study utilized the same receptor methodologies as in previous studies12,13; however, the atmospheric Hg data have been modified in two ways to improve the data quality. The first approach combines GOM and PBM into a single reactive Hg compound, instead of separating GOM and PBM as in previous studies. The second approach excluded low GOM measurements, which is a method of data quality control to handle below detection limit (DL) data. Comparison of the receptor model results from this study to previous studies may determine which approach would be most effective in reducing the receptor model uncertainties and could be an alternative method for analyzing atmospheric Hg data until calibration methods become available. Results from this study can also provide an uncertainty assessment of the previously-obtained atmospheric Hg receptor modeling results.