Methods All of the ice cores represented in this study were drilled without the use of fluid, except for the Huascaran core, for which ethanol was used. After core sections were retrieved during drilling, they were immediately placed in plastic sleeves and then cardboard tubes for transportation to the Byrd Polar Research Center and subsequent storage in a -30 deg C freezer. All samples used in this investigation were collected from below the firn layer, and were thus out of contact with the atmosphere and organic aerosols at the time of coring. Furthermore, the ice was sub-sampled from the center of core sections and the outer surfaces of the samples were rinsed with purified (Milli-Q) water in order to minimize entrainment of contamination resulting from drilling or storage. All subsequent sample manipulations were performed in a HEPA-filtered laminar flow hood. For each post-industrial sample, 10, 20 and 30 ml aliquots of melted ice were decanted from a Teflon(R) bottle into separate clean 40 ml glass vials for SBSE and analysis. In addition, a 10 ml sample from Sajama and a 20 ml sample from Kilimanjaro were run in duplicate. Pre-industrial ice was transferred in entirety to sample vials and allowed to melt prior to extraction. To improve SBSE recoveries, methanol and HCl were added to each sample. Extracted solutions consisted of 25% methanol, as an analysis of fatty acid (C16-C24; even chain lengths) standards indicated that optimal SBSE recoveries occurred at about this concentration. Based on standard recoveries and the results of Pfannkoch et al. (2003), a dilute HCl/H2O(Milli-Q) solution was added to each sample to achieve a pH of 3-4. A pre-conditioned (300 deg C under an He flow for 2 hours) Twister stir bar (Gerstel, Inc.; 10 mm long, 0.5 mm phase thickness) was then added to each sample, and after capping the vials all samples were stirred at 750 rpm for 10+ hours at room temperature to ensure maximal extraction (Popp et al., 2001). Following removal from the samples, the stir bars were transferred into clean glass desorption tubes and loaded onto the TDS-A autosampler (Gerstel). Re-extraction tests of ice core samples and an assessment of stir bar carry over using standards revealed that nearly all of the extractable organic matter in a sample is removed after one SBSE treatment and entrainment of analytes between analyses is minimal. All analyses were performed using the TDS 2 ThermoDesorption System (Gerstel) mounted on a TRACE GC, which was coupled to a Tempus TOF-MS (Thermo Scientific). A Gerstel CIS3 GC inlet with cryo-cooling and an Agilent DB5-MS column (20 m length, 0.18 mm i.d., 0.18 micrometer film thickness) were used. GC oven temperature was initially held at 40 deg C for 1 minute, then increased by 30 deg C/minute to a final temperature of 320 deg C, which was held for 5 minutes. The CIS3 injector temperature was maintained at -80 deg C during thermal desorption and held there for 0.1 minutes at the beginning of the GC run, then increased at 12 deg C/second to 320 deg C, which was held for 4 minutes. It was operated in solvent venting mode with 1 minute of splitless time at the start of the GC run. Thermal desorption of the Twister stir bars was performed prior to the GC run under splitless conditions with a He flow of 50 ml/minute. During this process, transfer line temperature was held at a constant 325 deg C, and the TDS2 was operated with an initial temperature of 30 deg C for 1 minute, then heated by 60 deg C/minute to 300 deg C, where it was held for 4 minutes. These analytical parameters were optimized through iterative testing with standards. Compound detection was achieved by TOF-MS and identification was accomplished through comparison of the individual mass spectrum generated for each analyte (Fig. S2) with the NIST/EPA/NIH Mass Spectral Library. Characteristic ions (Table S2) were then used to generate an extracted ion chromatogram for each compound so that a clean chromatographic peak with no co- eluting species was produced. To perform quantification, the resultant peaks were manually integrated using the Xcalibur(TM) (Thermo Scientific) software package. Standards consisting of n-alkanes (C14-C19, C22, and C24), fatty acids (C16-C24 even chain lengths, C28, and C30), and PAHs (2-6 rings) were prepared in H2O(Milli-Q)/methanol/HCl solutions and processed in the same manner as the samples. Where they occurred in the samples, these compounds were quantified by comparison with the authentic external standards. Authentic standards were not measured for the other observed compounds, for which concentrations were estimated based on response factors of the available standards. Nevertheless, abundance comparisons between samples for the analytes without standards are robust because the methods of quantitation were applied consistently. Mean compound abundances were calculated from the results of all sample aliquots measured for a given ice core site (and depth within the core) and are reported throughout. Reproducibility was similarly determined (as standard deviation) and is reported in Table S2. Two blanks were created from purified (Milli-Q) water and were analyzed along with the ice core samples. One blank was frozen and treated in the same manner as the ice that was sampled from the cores, from initial sub-sampling through analysis, and the other was treated as a laboratory/analytical blank in order to assess the contamination entrainment potential of a melted sample. Both blanks were subjected to the same laboratory processing as the melted ice core samples, including methanol and HCl addition, and were also analyzed in the same manner. For each compound analyzed, the highest concentration observed in either of the blanks is reported in Table S2. Blank levels and their impacts on biomarker interpretations are discussed in Section 3., although the reported values (Fig. 1 and Table S2) have not been blank-corrected.