For sedimentary macroscopic charcoal analysis, 1 cm3 of wet sediment was dissolved in water, sieved through a 150-µm mesh. Under a stereomicroscope, macroscopic charcoal of three size classes (150-300, 300-500, and ≥500 µm) was counted continuously throughout the core. To estimate a proxy error that combines sampling, preparation and macrocharcoal counting uncertainties, we continuously sampled short core JC11-K2 between 35-55 cm core depth (n = 20), i.e., interval 1840-1875 CE, that could be linked to core JC11-K5 by four marker layers as determined from varve counting. Samples were processed in the same way as for JC11-K5. The numbers of absolute particles cm-³ were compared with the JC11-K5 samples of the same time interval (n = 31) to determine an overall mean relative standard deviation of 0.8 % of each sample.
The topmost 75 samples (1780-2010 CE) were also analyzed for monosaccharide anhydrides (MAs) (n = 75, 1780-2011 CE): 125-250 mg dry sediment were extracted with a DIONEX Accelerated Solvent Extractor (ASE 200, 100 °C, 7.6×106 Pa) using a 9:1 solvent mixture of dichloromethane (DCM):methanol (MeOH). As an internal standard, 2.5-5 ng deuterated levoglucosan (dLVG) was added. The total lipid extracts were separated on an unactivated SiO2 gel column (Merck Si60, grade 7754) using sequential elution with DCM:MeOH (9:1) and DCM:MeOH (1:1). The 1:1 fractions were re-dissolved in 95:5 acetonitrile:H2O and filtered using a 0.45 µm polytetrafluoroethylene filter before analysis. The MAs were analyzed by ultra-high pressure liquid chromatography-high resolution mass spectrometry using a method adapted from earlier HPLC-ESI/MS2 methods (Hopmans et al., 2013). Authentic standards for LVG, GAL and MAN were obtained from Sigma Aldrich, and that for dLVG (C6H3D7O5) from Cambridge Isotope Laboratories, Inc. Integrations were performed on mass chromatograms within 3 ppm mass accuracy. Concentrations were corrected for relative response factors to dLVG of 0.997, 0.822, and 2.137 for LVG, MAN, and GAL, respectively. Instrumental (standard) errors for LVG, MAN, and GAL were 4 ± 3, 14 ± 15, and 28 ± 38% (1σ), respectively.
