Hybrid Targeted/Untargeted Screening Method for the Determination of Wildfire and Water-Soluble Organic Tracers in Ice Cores and Snow

Wildfires can influence the earth’s radiative forcing through the emission of biomass-burning aerosols. To better constrain the impacts of wildfires on climate and understand their evolution under future climate scenarios, reconstructing their chemical nature, assessing their past variability, and evaluating their influence on the atmospheric composition are essential. Ice cores are unique to perform such reconstructions representing archives not only of past biomass-burning events but also of concurrent climate and environmental changes. Here, we present a novel methodology for the quantification of five biomass-burning proxies (syringic acid, vanillic acid, vanillin, syringaldehyde, and p-hydroxybenzoic acid) and one biogenic emission proxy (pinic acid) using solid phase extraction (SPE) and ultrahigh-performance liquid chromatography coupled with high-resolution mass spectrometry. This method was also optimized for untargeted screening analysis to gain a broader knowledge about the chemical composition of organic aerosols in ice and snow samples. The method provides low detection limits (0.003–0.012 ng g–1), high recoveries (74 ± 10%), and excellent reproducibility, allowing the quantification of the six proxies and the identification of 313 different molecules, mainly constituted by carbon, hydrogen, and oxygen. The effectiveness of two different sample storage strategies, i.e., re-freezing of previously molten ice samples and freezing of previously loaded SPE cartridges, was also assessed, showing that the latter approach provides more reproducible results.


SI3. Solid-phase extraction optimization
The solid phase extraction procedure used in this method is similar to the one described by Vogel at al., 2019. However, due to the chromatographic column change (we used here an Organic Acid column, while Vogel et al., used a RP-MS column), we modified the SPE elution solutions. In Vogel et al., the elution was performed using a solution of 250 µL 0.5% HCl and 5% formic acid in methanol (hereafter, solution A), followed by a solution of 500 µL 5% formic acid in methanol (hereafter, solution B). Following the same approach, we observed a detrimental effect on the ionization performances, likely due to the presence of HCl residues after the evaporation and the non-suitability of the Organic Acid column with traces of a strong acid. Comparing the results of two extractions performed on three 0.1 ng g -1 UPW standard solutions, we observed that for the cartridges eluted using both solution A and B, the intensities of syringic acid, syringaldehyde, p-hydroxybenzoic acid and vanillic acid were 0.5, 0.8, 0.6 and 0.5 times the intensities obtained using 750 µL of solution B, only (not shown). For vanillin and pinic acid, the intensities were similar. In light of this, we eluted our cartridges using uniquely solution B.

SI4. UHPLC-HRMS optimization
Considering the acidic and polar properties of the compounds targeted in this study (i.e. methoxyphenols and secondary organic aerosol tracers), the chromatographic column used was the Acclaim TM Organic Acid column, a silica-based reverse-phase column designed for the retention and following detection of hydrophilic aromatic and aliphatic acids. The gradient program was optimized in order to achieve the better sensitivity for the targeted compounds and to maximize the number of molecules that can be detected following an untargeted approach. After testing different gradients (Table S1), we opted for the 15 min gradient (#1 in Table S1) because: a) the sum of the targeted compounds intensities was higher, b) the number of identified compounds following an untargeted approach was the largest, and, c) the analysis time was the shortest.
Table S1 -Different elution programs (#1-#6) resulted in different target compound intensities and in different numbers of identifications. Here we considered only the compounds that showed an area higher than 1E7. To perform this study we used a sample from the Colle Gnifetti. The values were normalized for the maximum area among the different testing conditions for every single compound (e.g. 0.6 means that the intensity was 0.6 times the highest intensity). The finally applied program #1 is marked in bold. n.d. = not detected.  (Table S2). We did not observe an overall improvement in sensitivity when the post-column addition of aqueous ammonia was included, probably due to the formation of salt adducts that had a detrimental effect on the ionization efficiency.
We found that the overall optimal eluent modifier concentration for the six target species was 0.001% formic acid. However, we opted for the 0.01% concentration that improved the sensitivity up to 7 times compared to previous methods (0.2% formic acid) (Vogel et al., 2019), and it was only 13% lower than the 0.001% option. This choice was to ensure a high sensitivity also for pinic acid and, consequently, for other similar aliphatic carboxylic acids that can be detected following a NTS approach.

SI5. Freezing tests
Besides applying the method to investigate the chemical composition of snow and ice core samples, we investigated the preservation of organic molecules when two different sample storage approaches are used: a) re-freezing of a previously molten sample and b) freezing of previously loaded SPE cartridges. With our study we provide evidences on the preservation of the chemical species following both a target approach ( §3.3.1 main text) and a NTS approach ( §3.3.2 main text).

Re-freezing samples in glass vials (target approach)
To evaluate whether the six target compounds were preserved after the refreezing of previously molten ice samples, we carried out two sets of experiments with snow collected from Jungfraujoch and b) as for a), but the spiked concentration was 0.1 ng g -1 ; c) 2x30 mL sample aliquots from the Colle Gnifetti ice core were spiked to reach a final added concentration of ≈ 0.03 ng g -1 . One aliquot was immediately extracted and analysed, while the other was kept frozen for seven days (-20°C), then extracted and analysed. This procedure was repeated for six different samples. Six blanks were also collected and analysed.

Re-freezing samples in SPE cartridges (target approach)
To explore alternative strategies to sample re-freezing in glass vials, we also investigated the compound's stability once loaded on a SPE cartridge. We performed different experiments: a) We evaluated the sample recovery from frozen cartridges in UPW at three different concentrations (i.e. 0.03 ng g -1 [n=3], 0.1 ng g -1 [n=4] and 1 ng g -1 [n=4]) and compared the results with those obtained from unfrozen cartridges.
b) 2x30 mL sample aliquots from Colle Gnifetti ice core were spiked to reach a final added concentration of ≈ 0.03 ng g -1 . Both aliquots were loaded on two different SPE cartridges.
One cartridge (and one blank) was eluted and analysed the same day. The other cartridge (and an additional blank) was dried under vacuum for 5 minutes, then wrapped into two aluminium foils and stored for 7 days at -20°C. Before elution, the cartridge was thawed at room temperature for ≈ 30 minutes under Class-1000 laminar flow hood, then eluted and analysed. This procedure was repeated for seven different samples. Six blanks were also collected and analysed.

Re-freezing samples in glass vials and SPE cartridges (untargeted screening approach)
Finally, to extend our analysis to a wider set of compounds, we performed a NTS on three 30 mL aliquots from a Belukha ice-core sample. One aliquot was immediately extracted and analyzed. The second aliquot was frozen into a 50 mL glass vial for seven days at -20°C, then extracted with SPE, eluted and analyzed. The third aliquot was loaded on a SPE cartridge, frozen for seven days at -20°C, then eluted and analyzed. An UPW blank was prepared in parallel for each aliquot and treated the same as the sample.     Table S4 -Comparison between the methodological limit of detection of this method with the concentration range of the six target compounds in ice and snow samples from different locations. MDL = methodological limit of detection (this study). ND = not detected, below methodological LoD of the cited study. NA = not available.  Table S5 -Evaluation of the reproducibility from a selected Colle Gnifetti ice core sample (named 169) divided in two 30 mL aliquots (169A and 169B). The aliquots were spiked with the six targeted compounds to achieve a final added concentration of ≈ 0.03 ng g -1 .

Sample_169B
/ng g -1  Figure S1 -Pinic acid was synthesized according to previously published procedure (Steimer et al., 2018). Its NMR spectrum is shown below. The signal at 3.78 ppm is associated to dioxane. Afterwards the product was further dried. The pinic acid measured exact mass of [M-H]was 185.0820 (Δ<1 ppm). Figure S2 -Chromatographic separation of the targeted species at a concentration of 10 ng g -1 .

Figure S3
-Calibration curves for the six targeted species over the range 0.5-15 ng g -1 . Data are presented as area ratio (i.e. the ratio between the area of the targeted species and the area of the internal standard) vs concentration ratio (i.e. the ratio between the concentration of the targeted species and the concentration of the internal standard). The internal standard is vanillin-(phenyl-13 C 6 ). More details are reported in the main text. Figure S4 -Details on the recovery experiments performed on fresh UPW samples prepared at 0.03 (blue bar, n = 3), 0.1 (orange bar, n = 4) and 1 ng g -1 (yellow bar, n = 4).  Figure S7 -Boxplot that shows the differences between unfrozen (n=4) and frozen (n=4) aliquots for the six targeted compounds for the 0.03 ng g -1 experiment (samples refrozen in glass vials). The dotted green line represents the procedural limit of detection.

Figure S8
-Boxplot that shows the differences between Day 1 (n=4) and Day 2 (n=4) for the six targeted compounds for the 0.1 ng g -1 experiment (samples refrozen in glass vials). The dotted green line represents the procedural limit of detection.

Figure S9
-Comparison between unfrozen (black solid lines) and frozen samples (red dashed lines) from the Colle Gnifetti ice core. Samples were frozen in the glass vials before extraction and they were all spiked to reach a final spiked concentration of ≈0.03 ng g -1 . Procedural limit of detection is given (green line).

Figure S10
-Comparison between unfrozen (black solid lines) and frozen samples (red dashed lines) from the Colle Gnifetti ice core. Samples were frozen in the SPE cartridges before the elution and analysis, and they were all spiked to reach a final spiked concentration of ≈0.03 ng g -1 . Procedural limit of detection is given (green line). Sample 5 for syringic acid and vanillic acid is not reported due to an observed contamination. Figure S11 -The recovery of the six targeted species between frozen and non frozen cartridges is compared at three different concentrations (0.03 ng g -1 , 0.1 ng g -1 and 1 ng g -1 ). Solid blue, yellow and red bars refer to recoveries from not frozen cartridges at 0.03 ng g -1 , 0.1 ng g -1 and 1 ng g -1 , respectively. Dashed blue, yellow and red bars refer to recoveries from frozen cartridges at 0.03 ng g -1 , 0.1 ng g -1 and 1 ng g -1 , respectively.