Brief Communication : The reliability of gas extraction 1 techniques for analysing CH 4 and N 2 O compositions in gas 2 trapped in permafrost ice-wedges 3

Ji-Woong Yang, Jinho Ahn, Go Iwahana, Sangyoung Han, Kyungmin Kim and 4 Alexander Fedorov 5 School of Earth and Environmental Sciences, Seoul National University, Seoul, South Korea 6 International Arctic Research Center, University of Alaska, Fairbanks, USA 7 Melinkov Permafrost Institute, Russian Academy of Science, Yakutsk, Russia 8 North-Eastern Federal University, Yakutsk, Russia 9 *Now at: Laboratoire des Sciences du Climat et de l'Environnement, LSCE/IPSL, CEA10 CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France 11 **Now at: Division of Earth and Planetary Materials Science, Department of Earth Science, 12 Graduate School of Science, Tohoku University, Sendai, Japan 13 14 Correspondence: Jinho Ahn (jinhoahn@snu.ac.kr) 15


28
Permafrost preserves large amounts of soil carbon (C) and nitrogen (N) in a frozen state 127 For the control and HgCl2-treated wet extraction experiments, a melting-refreezing wet 128 extraction system at SNU was employed (Yang et al., 2017;Ryu et al., 2018). The gas extraction procedure is identical to the procedure described in Yang et al. (2017) and Ryu et al. 130 (2018), except for the sample gas trapping procedure (see below). Ice-wedge sub-samples of 131 ~40 g (composed of 10-12 ice cubes for each) were placed in a glass container welded to a 132 stainless-steel flange (sample flask), and the laboratory air inside the sample flasks was 133 evacuated for 40 min. The sample flasks were then submerged in a warm (~50°C) tap water 134 bath to melt the ice samples. After melting was complete, the meltwater was refrozen by 135 chilling the sample flasks with cold ethanol (below -70°C). The sample gas in the headspace 136 of each sample flask was then expanded to the volume-calibrated vacuum line to estimate the 137 volume of extracted gas, and trapped in a stainless-steel sample tube by the He-CCR device.

138
In this study, we attached the He-CCR device to our wet extraction line and the gas samples in 139 the flasks were cryogenically trapped. The reasons for using He-CCR instead of direct 140 expansion to a GC are twofold: 1) to better compare the dry and wet extraction methods by 141 applying the same trapping procedure, and 2) to maximize the amount of sample gas for GC 142 analysis, because the gas expansion from a large flask allows only a small fraction of gas to be 143 measured by the GC.

144
For biocide-treated tests, 1.84 mmol of mercuric chloride (HgCl2) was applied per unit 145 kilogram of soil, following established procedures for soil sterilization (Fletcher and Kaufman, 146 1980). We obtained the average dry soil mass (0.33 g) from the leftover meltwater samples of 147 the previous wet extractions, which were carried out for comparison between dry-and wet 148 extractions. Taking the average dry soil mass (0.33 g) into account, we added 24 μL of saturated 149 HgCl2 solution (at 20°C) to the sample flasks. The flasks with HgCl2 solution were then frozen 150 in a deep freezer at < -45°C to prevent the dissolution of ambient air into the solution during 151 ice sample loading. After the wet extraction procedure was complete, the extracted gas was 152 trapped in a sample tube and the CH4 and N2O mixing ratios were determined using the same 153 GC-ECD-FID system as the dry-extracted gas. The resulting CH4 and N2O mixing ratios have not been corrected for partial dissolution in ice melt in the flasks, because CH4 and N2O trapped 155 in refrozen ice are negligible compared to the ranges of the systematic blanks (see Appendix). The analytical methods described previously are for determining the mixing ratios of 159 CH4 and N2O in the extracted gas. To convert these mixing ratios into moles of CH4 and N2O 160 per unit mass of ice-wedge sample (CH4 and N2O content, respectively, hereafter) requires data 161 regarding the amount of gas extracted. The gas content is a measure of gas volume enclosed in 162 a unit mass of ice sample at STP (in mL kgice -1 ). Thus, the CH4 and N2O contents can be 163 calculated using the gas content, the total mass of the random cube ice, and the gas mixing 164 ratio. The gas content in the control and HgCl2-treated wet extraction experiments was 165 calculated from the temperature and pressure of the extracted gas and the internal volume of 166 the vacuum line. The details of the extraction system and correction methods used for 167 estimating gas content are described in Yang (2019). Similarly, the gas content of the dry 168 extraction samples was also inferred from the volume and pressure of gas inside the vacuum 169 line once the sample tube was attached to the line for GC analysis. The uncertainties of the 170 calculated CH4 and N2O contents were calculated by using error propagation of the blanks and  Dry soil content was measured using the leftover meltwater from the control-wet 176 extraction tests. After the control-wet extractions were complete, the sample flasks were shaken 177 thoroughly and the meltwater samples were each poured into a 50 mL conical tube. The 178 meltwater and soils were separated by a centrifugal separator at 3000 rpm for 10 min. The separated wet soils were wind-dried in evaporating dishes at approximately 100°C for 24 hours.

180
The weight of each individual evaporating dish was pre-measured before use. The dry soil 181 content was calculated by subtracting the weight of the evaporating dish from the total weight 182 of the dried soil sample plus the evaporating dish.

186
The results from the wet and dry extractions were compared using 23 ice-wedge 187 samples (21 for N2O) from Alaska and Siberia. In both the CH4 and N2O mixing ratio analyses, 188 we found that the wet and dry extraction results did not differ significantly (p > 0.1), regardless 189 of sampling site or soil content (Figure 1, a to d). We note that the heterogeneous distribution  Shown are scatter plots between wet-and dry (hit5) extraction results of CH4 (a and b) and N2O (c and d), and between control-and biocide-treated wet extraction results for CH4 (e) and N2O (f). The 'hit5' denotes the dry extraction with five times hitting (see Section 3.3). Left panels (a, c, and e) and (f) present in mixing ratios of gas in bubbles, while right (b) and (d) panels in moles of gas in a unit mass of ice (gas content). The sampling locations are indicated by different symbols. The color of each data point indicates the dry soil weight in the subsamples used in control wet extraction. The 1-sigma uncertainties of the mixing ratios (a, c, e, and f) are magnified by 5x, 20x, 100x, and 500x as denoted as blue error bars (see Appendix). The error bars are not visible where the error bars are smaller than markers. The grey dashed lines are 1:1 reference line. Note that the units of the axes of the insets in (e) and (f) are identical to the original plots. The p-value of two-sided Students' t-test of each comparison is denoted at the bottom right corner of each plot.

201
To test the microbial production of CH4 and N2O during wet extraction more accurately, 202 we conducted wet extraction experiments on samples treated with HgCl2, a commonly used 203 effective biocide (e.g., Torres et al., 2005), and compared the results with those of untreated 204 (control) wet extractions. We prepared 12 additional ice-wedge samples using the random cube 205 method for these tests (see Materials and Methods section). We found no significant differences 206 between the control and HgCl2-treated wet extraction results for both CH4 and N2O mixing 207 ratios (Figures 1e and 1f), indicating that the bias due to microbial activity during 208 approximately an hour of the melting-refreezing procedure is not significant. This is further 209 supported by tests on an additional 12 ice-wedge samples (using the random cube protocol)  The two-sided t-test for the CH4 data indicates an insignificant difference between the two 214 results (p > 0.9). Data from individual sampling sites also do not show significant differences 215 (p > 0.9 for the Alaskan samples and p > 0.5 for the central Yakutian samples).

216
According to microbial sequencing studies that have shown the presence of viable 217 microbes in permafrost and ground ice (e.g., Katayama et al., 2007), it is likely that culturable 218 microbes exist in the ice-wedge samples used in this study. However, considering that at least 219 14 days and up to 3 months of culturing was required to identify microbe colonies extracted

224
One limitation of our needle crushing dry-extraction technique is the inability to completely extract gas from ice samples, because small ice particles and/or flakes placed in the 226 space between the needles are not fully crushed. The gas extraction efficiency of the SNU 227 needle crusher system has been reported as ~80-90% for polar ice core samples (Shin, 2014).

228
However, the gas extraction efficiency has not been tested for ice-wedge samples. Depending

233
To estimate the biases arising from incomplete gas extraction, we designed a series of 234 tests to identify the differences of the CH4 and N2O mixing ratios and contents between the 235 crushed and uncrushed sample portions. Each ice-wedge sample that was randomly collected 236 was first crushed by the regular dry extraction procedure (by hitting it five times with the needle 237 system, 'hit5'), and the gas liberated from the sample was trapped in a sample tube. Then we 238 performed an additional 100 hits on the leftover ice ('hit100'), monitored the amount of 239 additional gas liberated, and trapped the additional gas in a separate sample tube. Comparisons 240 between the hit5 and hit100 results are summarized in Table 1.

241
Here we regard the ratio of gas content of hit100 to that of hit5 (hit100/hit5 ratio 242 hereafter) as a measure of the gas extraction efficiency of the needle crusher system. The results 243 demonstrate an average hit100/hit5 ratio of gas content of 0.40 ± 0.07 for the Zyryanka samples, 244 0.24 ± 0.07 for the Bluff samples, and 0.14 ± 0.11 for the Cyuie samples (Table 1). Despite the 245 fact that the number of samples was limited, the ice-wedge samples from the different sites 246 show distinct hit100/hit5 ratios of the amount of extracted gas. However, we observed that the 247 leftover ice from the Bluff and Zyryanka samples were not well-crushed, even after 100 hits 248 with the needle crusher. This was especially true if the ice sub-samples contained soil 249 aggregates: the frozen soil aggregates were barely crushed. In contrast, the Cyuie samples were relatively well-crushed, and the leftover samples were apparently finer-sized ice flakes. We 251 also observed that the hit100/hit5 ratios of gas content are highly variable within samples from 252 a particular site, implying that the extraction efficiency of the needle crusher not only depends 253 on site characteristics, but also on the individual ice sample hardness. When compared with the 254 dry soil content measured from the sub-samples used for wet extraction, no relationship was 255 observed between the dry soil content and the extraction efficiency (Figures 1 and A3). In 256 addition, in the case of samples uncrushed by the hit100 test, it is difficult to estimate the 257 extraction efficiency using the hit100/hit5 ratio of gas content, as the hit100 tests liberated only 258 a marginal portion of gas from these samples. This is because the large-sized uncrushed soil 259 aggregates or particles may have prohibited the needle crusher from crushing the small-sized 260 ice flakes or grains. The needles move up and down together, as they are fixed to a pneumatic 261 linear motion feedthrough device, thus if there is a sizable soil clod that cannot be crushed, it 262 blocks the needle crusher from moving further down. Therefore, we do not recommend using 263 a needle crusher system to measure gas contents in ice-wedge samples.

264
The hardness of the ice samples may also affect the gas mixing ratio analysis in the hit5 265 and hit100 procedures. The hit100/hit5 ratio of CH4 mixing ratio of Bluff and Zyryanka 266 samples are less than 1 in four out of six samples, yielding an average of 0.9 ± 0.5. However, 267 all five samples from the Cyuie ice-wedges have ratios greater than 1, with an average of 4.7 268 ± 2.6 ( Table 1). The higher hit100/hit5 ratio of CH4 mixing ratios of Cyuie samples indicates 269 that the gases extracted via the hit100 procedure have higher CH4 mixing ratios than the gases 270 extracted via the hit5 procedure. Considering these results with those discussed previously, we 271 speculate that there are three ways gas can be trapped in ice-wedge ice: enclosed in bubbles,  It is worth noting that friction between stainless steels could produce CH4 with carbon 304 from the damaged stainless-steel surface and hydrogen gas (Higaki et al., 2006). If needle 305 crushing causes contamination in this way, the dry extraction results should be affected by the 306 number of hits. To check the impact of the needle crushing procedure on ice-wedge CH4 and 307 N2O measurements, we carried out blank tests by changing the numbers of hits from 5 to 100.

308
The results of these tests show no systematic offset among the experiments with different 309 numbers of hits ( Figure A2), which implies that the crushing procedure does not affect the dry 310 extraction results for CH4 and N2O. Even though a small of contamination does exist, its effects 311 have already been subtracted via blank correction and taken into account in the overall error 312 estimation (see Appendix). Therefore, we consider that our findings are not artefacts of metal 313 friction during crushing.

314
To summarize, from the hit5 and hit100 comparison tests, we found that 1) the needle 315 crusher method is not able to fully crush the ice-wedge ice samples and thus is unsuitable for 316 measuring gas contents in a unit mass of ice, and that 2) weak crushing (e.g., a small number 317 of hits by the needle crusher system) may better reflect gas mixing ratios of the soft parts of 318 the samples (such as air bubbles) than strong crushing (e.g., a greater number of hits). Table 1. Results of dry extraction tests with 5-and additional 100 times hitting ice-wedge samples, denoted as 'hit5' and 'hit100', respectively. 'hit100/hit5' 320 is the ratio in extracted gas content or gas mixing ratio of 'hit100' to 'hit5' cases. Also shown are gas content results from both experiments, where the hit100 321 values are given both in the unit of ml kg -1 at STP conditions and μmol/kg (in parenthesis). It should be noted that the 'hit100' gas content results indicate 322 the additional amount of gas extracted after 'hit5' crushing and evacuation.

3.4.Residual gas mixing ratios and contents after wet extraction 324
To examine how well the gas is extracted by wet extraction, we applied the dry 325 extraction method to refrozen ice-wedge samples after wet extraction. We first prepared

331
The wet-degassed ice was then inserted into the needle crusher and the crusher chamber was 332 evacuated. A specific amount of standard air was injected. Then, the wet-degassed ice samples 333 were hit 20 or 60 times by the needle crusher. The amount of gas and gas mixing ratio of the 334 additionally extracted gas from the wet-degassed ice are shown in Figure 2 and Table A1.

335
The tests using the wet-degassed ice show an additional gas extraction of ~12 to 20 ml 336 kgice -1 , which is ~43 to 88% of the amount of gas extracted during the initial wet extraction.

337
The additionally extracted gas from the dry extraction is referred to as residual gas hereafter.

338
This is remarkably in contrast to the less than 1% residual fraction of the SNU wet extraction low values compared to those from the initial wet extraction (Figure 2 and Table A1). These 355 results imply that most of the N2O in ice wedges is extracted by three melting-refreezing cycles,

356
such that only a small amount of N2O is left adsorbed or entrapped in ice-wedge soils. The 357 authors posit that this might be attributed to the high solubility of N2O to water compared to 358 CH4. However, it needs further investigation to better understand this.

359
In this section, we found that a certain amount of gas remained in ice wedges, even after 360 three cycles of wet extraction, which is extractable instead by needle crushing. This implies 361 that, unlike polar ice cores, wet extraction of ice-wedges does not guarantee near-complete gas 362 extraction, and therefore, precise measurements of the gas content of ice wedges are difficult 363 to obtain. The difficulty in measuring gas content imposes a large uncertainty in estimating 364 CH4 and N2O contents. Furthermore, we found that the residual gas has a similar order CH4 365 mixing ratio as the gas extracted by initial melting-refreezing, indicating that a comparable 366 amount of CH4 still remains unextracted in ice-wedges. Hence, a novel extraction method is 367 required to produce reliable gas content and gas mixing ratios in ice wedges. In contrast, our 368 results show that the N2O content of the residual gas is at trace levels, which may suggest that 369 most of the N2O in ice-wedges is extractable during initial melting-refreezing. Therefore, wet

384
In this study we carried out comparisons between wet and dry extractions, between 385 untreated and biocide-treated wet extractions, and gas extraction from the easily to extract and 386 difficult to extract parts of ice-wedge ice to better understand the characteristics of each 387 extraction method, in order to adequately analyse CH4 and N2O mixing ratios and gas contents 388 from permafrost ice wedges. Based on these comparisons, our major findings are summarized 389 as follows: water trap and cryogenically pumped into the sample tubes, using the He-CCR. The number of 431 hits did not significantly affect the systematic blank ( Figure A2) and the regression curve for 432 blank correction was fitted to the entire set of data points (red dashed curve in Figure A1).

433
For the wet extraction, a total of ~45 g of BFI cubes was placed into each sample flask.

434
The flasks were connected to the wet extraction line and sealed with a copper gasket, then The results of the blank experiments are shown in Figure A1. The systematic blanks 444 appear to be inversely correlated with the gas pressure in the sample tube. The systematic blank 445 test results were fitted using exponential regression curves (dashed lines in Figure A1), and 446 these regression curves were then used for systematic blank correction in our ice-wedge sample 447 analyses.

448
To calculate uncertainties of the blank corrections, the blank test data were fitted with 449 exponential regression curves ( Figure A1). The root-mean-square-deviations (RMSD) of the 450 data from the regression curves are taken as the uncertainties of blank corrections (Figure 1).

451
Since the ice-wedge data used in this study showed the pressure in GC sample loop of about 8 452 ~ 50 torr, the RMSD were estimated from the blank test data within this pressure range. The 453 uncertainty of the gas content measurement is calculated by error propagation from those of Figure A1. Systematic blank of the needle crushing (dry extraction) and melting-refreezing (wet extraction) methods for (a) CH4 and (b) N2O measurements in control and biocide (HgCl2) treated experiments. Also plotted are the CH4 blanks of BES-treated wet extractions. The dashed lines represent exponential regression curve fittings. Note that all data are plotted against the amount of gas trapped in the sample tube, presented here as the pressure in the GC sample loop when the sample gas is expanded. The grey shaded areas indicate the range of ice-wedge samples used in this study (see main text). The big-delta (Δ) notion in the y-axes indicate the offset from the values of the standard used. Figure A2. Influence of different number of hitting on the systematic blank of the needle crushing (dry extraction) system for (a) CH4 and (b) N2O measurements. Note that all data are plotted against the amount of gas trapped in the sample tube, presented here as the pressure in the GC sample loop when the sample gas is expanded (see main text). The big-delta (Δ) notion in the y-axes indicate the offset from the values of the standard used.