Quantification of monosaccharide anhydrides by liquid chromatography combined with mass spectrometry: Application to aerosol samples from an urban and a suburban site influenced by small-scale wood burning

Levels of the monosaccharide anhydride (MA) levoglucosan and its isomeric compounds galactosan and mannosan were quantified in the PM10 fraction (particulate matter <= 10 mu m in aerodynamic diameter) of ambient aerosols from an. urban (Oslo) and a suburban (Elverum) site in Norway, both influenced by small-scale wood burning. MAs are degradation products of cellulose and hemicellulose, and levoglucosan is especially emitted in high concentrations during pyrolysis and combustion of wood, making it a potential tracer of primary particles emitted from biomass burning. MAs were quantified using a novel high-performance liquid chromatography/ high-resolution mass spectrometry-time of flight method. This approach distinguishes between the isomeric compounds of MAs and benefits from the limited sample preparation required before analysis, and no extensive derivatization step is needed. The highest concentrations of levogucosan, galactosan, and mannosan (Sigma MA) were recorded in winter because of wood burning for residential heating (Sigma MA(MAX) = 1,240 ng m(-3)). This finding was substantiated by a relatively high correlation (R-2 = 0.64) between the levoglucosan concentration and decreasing ambient temperature. At the suburban site, Sigma MA accounted for 3.1% of PM2.5 whereas the corresponding level at the urban site was 0.6%. The mass size distribution of MAs associated with atmospheric aerosols was measured using a Berner cascade impactor. The size distribution was characterized with a single mode at 561 nm. Ninety-five percent of the mass concentration of the MAs was found to be associated with particles < 2 mu m. A preliminary attempt to estimate the contribution of wood burning to the mass concentration of PM10 in Oslo using levoglucosan as a tracer indicates that 24% comes from wood burning. This is approximately a factor of 2 lower than estimated by the AirQUIS dispersion model.