Non-targeted analysis and suspect screening of organic contaminants in temperate snowfall using liquid chromatography high-resolution mass spectrometry.

Abstract

Contaminants released into the atmosphere that undergo regional and long-range transport can deposit back to Earth through snowfall. When snow melts, these contaminants re-enter the environment, sometimes far from their original emission sources. Here we present the first comprehensive characterization of organic contaminants in snow from North America. Fresh snowfall samples were collected in the central United States over a three-year period and measured by liquid chromatography high-resolution mass spectrometry for suspect screening and non-targeted analysis. The resulting data set was screened against experimental MS/MS libraries and underwent supplemental in silico MS/MS analysis. In total, 91 possible compounds were tentatively identified in snow, and 17 were successfully confirmed and semi-quantified with reference standards. These contaminants were mostly anthropogenic in origin and included six herbicides, three insect repellants, one insecticide metabolite, and one fungicide. The most prominent compounds present in all samples were N-cyclohexylformamide (known contaminant in tire leachate), DEET (insect repellent), and dimethyl phthalate (plasticizer), with median deposition fluxes of 4032, 284, and 262 ng m^-2, respectively. Three additional compounds were detected in 100% of samples: coumarin (phytochemical and fragrance additive), 5-methylbenzotriazole (antifreeze component), and quinoline (heterocyclic aromatic). The Peto-Peto test revealed statistically significant differences in deposition fluxes for these six contaminants (p < 0.05), with weak but statistically significant positive associations between coumarin and DEET and between coumarin and quinoline according to a Kendall's tau correlation analysis. These findings demonstrate the utility of in silico analysis to complement MS/MS matching with experimental databases. Even so, thousands of unidentified features remained in the data set, highlighting the limitations of current strategies in non-targeted analysis of environmental samples.