Journal of Geophysical Research - Atmospheres. Supporting Information for

Transcription

1 Journal of Geophysical Research - Atmospheres Supporting Information for Observing and modeling the influence of layering on bubble trapping in polar firn Logan E. Mitchell 1,2,*, Christo Buizert 1, Edward J. Brook 1, Daniel J. Breton 3, John Fegyveresi 4, Daniel Baggenstos 5, Anais Orsi 5, Jeffrey Severinghaus 5, Richard B. Alley 4, Mary Albert 6, Rachael H. Rhodes 1, Joseph R. McConnell 7, Michael Sigl 7, Olivia Maselli 7, Stephanie Gregory 6, Jinho Ahn 8 1 Oregon State University, College of Earth Ocean and Atmospheric Sciences, Corvallis, OR University of Utah, Department of Atmospheric Sciences, Salt Lake City, UT University of Maine, Department of Physics and Astronomy, Orono, ME and University of Maine, Climate Change Institute, Orono, ME Department of Geosciences, and Earth and Environmental Systems Institute, Pennsylvania State University, University Park, PA Scripps Institution of Oceanography, University of California, San Diego, CA 92093, USA 6 Dartmouth College, Thayer School of Engineering, Hanover, NH 03755, USA 7 Desert Research Institute, Nevada System of Higher Education, Reno, NV School of Earth and Environmental Sciences, Seoul National University, Seoul, South Korea. * Corresponding author. address: logan.e.mitchell@gmail.com (L. Mitchell) Contents of this file Text S1 to S4 Figures S1 to S6 1

2 Text S1. Post-coring entrapment of air The CH 4 and V measurements are inversely correlated with each other (r = -0.78, p < 0.01, after de-trending with a spline fit) and there are two hypotheses to explain this relationship. The first hypothesis is that they are real variations, i.e. depths that contain a greater amount of air began trapping air at a relatively shallow depth, containing older air with a lower methane concentration versus depths that contain less air began trapping air at a relatively deeper depth, containing younger air with a higher methane concentration. The second hypothesis is that the inverse relationship could be caused by post-coring bubble closure, i.e. firn from depths that contain a greater amount of air had less open porosity when the core was extracted from the ice sheet and therefore had low contamination resulting in a relatively low methane concentration versus firn from depths that contain a lower amount of air had greater open porosity when the core was extracted from the ice sheet, which is susceptible to post-coring bubble closure, and therefore these depths trapped more modern air resulting in a higher methane concentration. A combination of these processes could also have occurred. To investigate the possibility of post-coring bubble closure, we measured δ 15 N at the exact same depths as some of our CH 4 samples. δ 15 N is a sensitive tracer of post-coring bubble closure for two reasons. First, throughout the diffusive column, gravitational fractionation causes a linear increase in δ 15 N with depth. This ceases within the LIZ since there is no vertical diffusion of air. Since most of the bubble trapping occurs within the LIZ, and only a small amount occurs just above the LIZ, there should be little natural variability of δ 15 N in the closed pores in the LIZ. Second, there are no other sources of enriched 15 N that could contaminate the signal. We measured 33 δ 15 N samples and rejected 4 because the V did not agree with the CH 4 samples at the same depth. There was a high correlation (R 2 = 0.95) in the V between the remaining δ 15 N and CH 4 data indicating that both sets of samples should have recorded the same signals. The samples had a δ 15 N of ± 0.018, which is lower than the observations in the open porosity as expected, but it is also lower and more variable than the model-predicted value of ± (between 70-80m, Fig. S1). To analyze the possibility of post-coring contamination, we plot δ 15 N vs. CH 4 (Fig. S2). If the low δ 15 N values were authentic they would indicate that the air was isolated in bubbles at a shallow depth and would therefore have to contain old air that would also have a low CH 4 concentration, yielding a positive relationship between δ 15 N and CH 4. However, we observe a negative relationship (r = -0.76, p < 0.01) indicating that the low δ 15 N values must be caused by entrapment of some modern air by post-coring bubble closure (Fig. S2). A comparison between δ 15 N and V reveals a positive relationship (r = 0.85, p < 0.01) indicating that the amount of contamination is related to the V and, by extension, the amount of open porosity in the sample (Fig. S2). Therefore, samples with more open pores had a greater amount of total contamination. To correct for the post-coring contamination we assumed that the δ 15 N should be equal to that expected in the LIZ. We used the model-predicted value of δ 15 N between m, δ 15 N real = ± 0.005, which agrees well with observations in mature ice of ± We assumed that the contaminant was modern atmospheric air (δ 15 N cont = 0 ), then used a mass-balance calculation to derive the amount of contaminated V (V cont): (SI Eq. 1) 2

3 where δ 15 N meas and V meas are the measured values. This calculation reveals that 10.6 ± 6.1 % of the V meas was from post-coring contamination. Since we did not measure δ 15 N at all of the depths where we have CH 4 measurements, we fit a linear regression to the percent of contamination ( ) vs. V meas (r 2 = 0.75, Fig. S3) and used this relationship to calculate the V cont based on the V meas in the samples. We applied this linear fit to our δ 15 N measurements and note that the fit does not account for all of the variability, leading to some uncertainty in our correction, as seen in Fig. S1. We then assume that the modern air that is being trapped in the ice has a concentration of 2000 ppb. This is a reasonable assumption based on preliminary methane concentrations observed at the NOAA Boulder Atmospheric Observatory (designated BAO in the NOAA data base) air sampling station that average ~1900 ppb but have frequent spikes in excess of 2100 ppb during the time period when these cores were stored at the National Ice Core Laboratory (NICL) in Lakewood, CO. We can then calculate the corrected CH 4 (CH 4-corr) by: (SI Eq. 2) The original and corrected CH 4 measurements are shown in Fig. S4. Since there is a significant amount of uncertainty in the estimation of the amount of contaminated air, the absolute concentration of the CH 4 in the closed pores is uncertain. In Fig. S4 we show a shaded region of methane concentrations that assumes the contaminant CH 4 has a concentration of 1800 ppb to 2400 ppb, corresponding to end member estimates of the contaminant methane concentrations. Note that the range of the shaded correction region scales with the V. The correction also affects the relative changes between the samples, and causes the observed anti-correlation between CH 4 and V to decrease to r = -0.44, p < The CH 4 and V samples were measured in three batches in 2008, 2009 and Some sample depths from 2008 and 2009 were re-measured in 2012 and yielded consistent results suggesting that at least over this time there was not a significant amount of post-coring bubble closure. Therefore we hypothesize that most of the post-coring bubble closure occurred between the time of drilling during the 2005/2006 field season and Text S2. Comparison of discrete and continuous CH 4 measurements Continuous measurements of CH 4 were conducted on ice samples adjacent to the discrete measurements. A comparison of the discrete measurements and the continuous measurement is shown in Fig. S5. In this figure both the continuous and discrete measurements are corrected for solubility only (not for post coring contamination) to facilitate their comparison. With a few exceptions in the upper samples, there is excellent agreement between these sampling techniques. The close agreement demonstrates multiple points. First, this method confirms the observation of very large variations in CH 4 in the closed pores from the discrete measurements. Second, it demonstrates that the continuous measurement technique does not trap ambient air in the open porosity during the melting process, and can be used on samples with some open porosity with fidelity as previously hypothesized [Rhodes et al., 2013]. The non-standard shape of samples (smaller and shorter than is typically used) makes it likely that the disagreement between the continuous and discrete measurements between 72.5 and 73.2 m results from ambient air entering the melt stream at the melter head. We did not observe any relationship between the total V derived from discrete measurements and the sample pressure recorded prior to entry into the laser spectrometer 3

4 (not shown), suggesting that this parameter does not reflect the V of the ice core when measuring LIZ samples. Text S3. Cut-bubble correction The measured V in our samples is lower than expected from in-situ ice because the process of preparing samples results in the cutting of several of the bubbles that were sealed when the sample was in-situ. To estimate a cut-bubble correction we first prepared ~1.5 mm thin-sections of ice within the LIZ on samples adjacent to our other measurements. We then digitally imaged these sections and identified the bubbles using the Fovea Pro 4.0 plug-in for Photoshop 5 Extended Edition. This plug-in has built-in edge and feature-detection functionality that is capable of exporting data on the size and shape characteristics of the bubbles. This analysis was conducted on 44 cross sections throughout the LIZ with each cross section having a depth range of 3.5 cm. Individual bubbles were characterized by their form factor; a factor of 1 equating to round (spherical) and 0 equating to non-round. We assumed that the non-round features had a cylindrical shape since this is a good approximation of the complex shapes of bubbles near the top of the LIZ, as discussed in further detail by Martinerie et al. [1990]. Using a Fovea Pro form factor cutoff of 0.59 to represent the round/non-round delimiter, we grouped the digitized bubbles in each cross section into spherical and cylindrical groups. Bubble size characteristics of inscribed radius (spherical and cylindrical), and length (cylindrical only) were then computed for each cross section and input into equation 11 from Martinerie et al. [1990] to obtain the final values for the overall cut-bubble correction, which is shown in Fig. S6. We estimate that the uncertainty of the correction is approximately 10% of the magnitude of the correction. We used a smoothing spline to apply the cut-bubble correction to all of our V data. Text S4. Propagating uncertainty in fitting s cl/s vs. local ρ When fitting the s cl/s vs. local ρ curves in the left panels of Fig. 3, we must bear in mind that the ρ data themselves are subject to additional measurement and sampling uncertainties (σ meas), which also contribute to broadening of features in the data. We consider two contributions to σ meas that we assume to be independent. The first is the measurement precision, which can be estimated from replicate measurements. The second is the finite size of the samples, having potentially larger dimensions (5 cm) than the underlying density variability. We evaluated this by taking half of the average density difference between adjacent samples, which is the density variability in half a sample. We multiplied this value by, assuming density in both halves of the sample can vary independently from each other. In this way we found σ meas = 9.4 kg m -3 for the Summit data, and σ meas = 7.1 kg m -3 for the DE08-2 data. The red curves in the left panels of Fig. 3 were fitted with Eq. (5), where. Fitting of σ co was done by eye; the best fit to both curves was obtained with σ co = 7 kg m -3. We therefore recommend this value be used at other sites as well. No fitting was done on the data in the right panels of Fig. 3; when plotting the red curves in the right panels we use Eq. (6), where given above., with σ meas values as 4

5 δ15n δ15n Figure S1. Overview of δ 15 N (top) and detailed view of the data (bottom) between 70 and 74 m. Orange squares and dashed line are the measured [Battle et al., 2011] and modeled δ 15 N from the open porosity. Grey circles and thick line are the originally measured values. Blue circles are the corrected δ 15 N. Thin grey and thick black lines are the modeled δ 15 N based on local and bulk density profiles shown in Fig. 1. 5

6 1700 CH 4 (ppb) R 2 = R 2 = δ15n ( ) Air Content ( m L kg - 1 ) Figure S2. δ 15 N vs. methane (CH 4, top) and air content (V, bottom). Note that these CH 4 data have not been corrected for post-coring contamination. 6

7 30 Y = * X R 2 = * V con t /V meas (%) V meas ( m L kg -1 ) Figure S3. Relative amount of contaminated air vs. measured air content (V), and the linear fit used to correct the CH 4 and V data. 7

8 V ( m L kg - 1 ) CH 4 (ppb) V ( m L kg - 1 ) CH 4 (ppb) Figure S4. Overview of original and corrected values (top) as well as an expanded section (bottom). Orange squares and dashed line are the open porosity measurements [Battle et al., 2011] and model. Blue line is the model CH 4 in the closed porosity using the new stochastic parameterization. Grey circles are the original measurements. Blue circles are the measurements corrected for post-coring contamination (V and CH 4) and the cut bubble effect (V only). The light blue shaded region is the range of possible CH 4 values determined by assuming the contaminated CH 4 values range from 1800 ppb to 2400 ppb. 8

9 CH 4 ( ppb) Figure S5. Comparison of discrete (blue symbols and line) and continuous (green line) CH 4 measurements. CH 4 measurements from the open porosity (orange symbols, [Battle et al., 2011]) as well as modeled CH 4 in the open porosity (orange dashed line) and the closed porosity (black line) are also shown (as in Fig. 2). Both the discrete and continuous data are corrected for solubility, but neither of them is corrected for post-coring bubble closure. 9

10 Cut Bubble Correction ( %) Cut Bubble Correction ( %) Figure S6. Cut-bubble correction and spline fit that was used to correct the V data. Inset shows a detailed view of the correction within the LIZ. Corrections below 100m are derived from WDC06A data [Fegyveresi et al., 2011]. References Battle, M., J. P. Severinghaus, E. D. Sofen, D. Plotkin, A. J. Orsi, M. Aydin, S. A. Montzka, T. Sowers, and P. P. Tans (2011), Controls on the movement and composition of firn air at the West Antarctic Ice Sheet Divide, Atmos. Chem. Phys., 11(21), Fegyveresi, J. M., R. B. Alley, M. K. Spencer, J. J. Fitzpatrick, E. J. Steig, J. W. C. White, J. R. McConnell, and K. C. Taylor (2011), Late-Holocene climate evolution at the WAIS Divide site, West Antarctica: bubble number-density estimates, J. of Glaciol., 57(204), Martinerie, P., V. Y. Lipenkov, and D. Raynaud (1990), Correction of air content measurements in polar ice for the effect of cut bubbles at the surface of the sample, J. of Glaciol., 36(124), Rhodes, R. H., X. Faïn, C. Stowasser, T. Blunier, J. Chappellaz, J. R. McConnell, D. Romanini, L. E. Mitchell, and E. J. Brook (2013), Continuous methane measurements from a late Holocene Greenland ice core: Atmospheric and in-situ signals, Earth and Planetary Science Letters, 368(0),