Portal de Periódicos da CAPES

Layering of surface snow and firn at Kohnen Station, Antarctica: Noise or seasonal signal?

2016; Wiley; Volume: 121; Issue: 10 Linguagem: Inglês

10.1002/2016jf003919

2169-9011

Thomas Laepple, Maria Hörhold, Thomas Münch, Johannes Freitag, Anna Wegner, Sepp Kipfstuhl,

Landslides and related hazards

Journal of Geophysical Research: Earth SurfaceVolume 121, Issue 10 p. 1849-1860 Research ArticleFree Access Layering of surface snow and firn at Kohnen Station, Antarctica: Noise or seasonal signal? Thomas Laepple, Corresponding Author Thomas Laepple tlaepple@awi.de Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany Correspondence to: T. Laepple, tlaepple@awi.deSearch for more papers by this authorMaria Hörhold, Maria Hörhold Institute of Environmental Physics, University of Bremen, Bremen, Germany Now at Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this authorThomas Münch, Thomas Münch Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany Institute of Physics and Astronomy, University of Potsdam, Potsdam, GermanySearch for more papers by this authorJohannes Freitag, Johannes Freitag Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this authorAnna Wegner, Anna Wegner Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this authorSepp Kipfstuhl, Sepp Kipfstuhl Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this author Thomas Laepple, Corresponding Author Thomas Laepple tlaepple@awi.de Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany Correspondence to: T. Laepple, tlaepple@awi.deSearch for more papers by this authorMaria Hörhold, Maria Hörhold Institute of Environmental Physics, University of Bremen, Bremen, Germany Now at Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this authorThomas Münch, Thomas Münch Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany Institute of Physics and Astronomy, University of Potsdam, Potsdam, GermanySearch for more papers by this authorJohannes Freitag, Johannes Freitag Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this authorAnna Wegner, Anna Wegner Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this authorSepp Kipfstuhl, Sepp Kipfstuhl Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, GermanySearch for more papers by this author First published: 28 September 2016 https://doi.org/10.1002/2016JF003919Citations: 23AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract The density of firn is an important property for monitoring and modeling the ice sheets as well as to model the pore close-off and thus to interpret ice core-based greenhouse gas records. One feature, which is still in debate, is the potential existence of an annual cycle of firn density in low-accumulation regions. Several studies describe or assume seasonally successive density layers, horizontally evenly distributed, as seen in radar data. On the other hand, high-resolution density measurements on firn cores in Antarctica and Greenland show no clear seasonal cycle in the top few meters. A major caveat of most existing snow-pit and firn-core-based studies is that they represent one vertical profile from a laterally heterogeneous density field. To overcome this, we created an extensive data set of horizontal and vertical density data at Kohnen Station, Dronning Maud Land, on the East Antarctic Plateau. We drilled and analyzed three 90 m long firn cores as well as 143 one-meter-long vertical profiles from two elongated snow trenches to obtain a two-dimensional view of the density variations. The analysis of the 45 m wide and 1 m deep density fields reveals a seasonal cycle in density. However, the seasonality is overprinted by strong stratigraphic noise, making it invisible when analyzing single firn cores. Our density data set extends the view from the local ice core perspective to a hundred meter scale and thus supports linking spatially integrating methods such as radar and seismic studies to ice and firn cores. Key Points Extensive data set of vertical and horizontal firn density variations at EDML, Antarctica Even in low-accumulation regions, the density in the upper firn exhibits a seasonal cycle Strong stratigraphic noise masks the seasonal cycle when analyzing single firn cores 1 Introduction Density, as a physical property of polar firn, is important for many topics of polar research. The determination of the ice sheet mass balance by either modeling or monitoring of the ice sheets with ground-penetrating radar or with satellite laser altimetry is strongly related to firn density [Rott et al., 1993; Li and Zwally, 2002; Rotschky et al., 2006]. The evolution of the firn density further determines the air enclosure in ice at the firn-ice transition [Martinerie et al., 1992; Schwander et al., 1997]. Polar firn is characterized by layering, each layer generated by a single deposition event and associated with different physical properties. The resulting vertical density variability caused by this layering is responsible for the depth and time when air trapped in the firn and ice is sealed off from the exchange with the atmosphere, thereby determining the age difference between air bubbles in ice cores and their surrounding ice [Landais et al., 2006; Mitchell et al., 2015]. It was previously demonstrated that the layering changes with depth, showing a seasonal cyclicity in depths below 20–30 m that is highly correlated with seasonally varying impurity concentration [Hörhold et al., 2012; Freitag et al., 2013a]. It has been speculated since then that impurities, featuring a seasonal cycle in their concentration, reshape the layering by changing the mechanical properties of the snow and the firn and thus influence the densification of single layers [Hörhold et al., 2012; Freitag et al., 2013a; Fujita et al., 2014; Gregory et al., 2014]. The horizontal and vertical density variabilities at the surface of ice sheets are unclear. On one hand, radar images suggest an image of layers that undulate at a kilometer scale but are nevertheless continuous [Arcone et al., 2005; Anschütz et al., 2007; Arthern et al., 2013]. Thus, the firn column on top of the ice sheets is often understood as a sequence of successive layers [Gow, 1965; Palais, 1984; Kreutz et al., 1999]. Accumulation rates extracted from ground-penetrating radar or satellite images are one example for this concept of successive horizontally distributed annual layers [Winebrenner et al., 2001; Anschütz et al., 2007; Eisen et al., 2008; Arthern et al., 2013]. The concept of annual layer deposition is further more supported by measurements of impurity concentrations or stable water isotopes, which exhibit a seasonal cycle in the atmosphere prior to deposition. These properties show a seasonal cycle within the snow and firn at many polar sites [Gow, 1965; Benson, 1971; Palais, 1984; Alley, 1988; Goektas et al., 2002; Svensson et al., 2015], suggesting homogenous formation of successive layers throughout each year. On the other hand, especially in low-accumulation regions, considerable horizontal variability is observed [Libois et al., 2015]. This is reflected by the different stratification found in nearby snow pits [Fisher et al., 1985; Karlöf et al., 2006]. Further, high-resolution density analysis of firn-core material [Hörhold et al., 2012; Freitag et al., 2013a] did not reveal a seasonal cycle in the surface firn density, suggesting that re-deposition and wind-scouring dominate over the successive deposition of seasonal layers. In this manuscript, we reconcile the previous findings of a lack in seasonality in the top meters of firn cores with the observed large-scale seasonal layering of the firn. Analyzing density on vertical and horizontal scales from the centimeter to the 100 m scale, we demonstrate that the question of layering and seasonality depends on the analyzed spatial scale. 2 Materials and Methods In austral summer 2012/2013, close to the European Project for Ice Coring in Antarctica (EPICA) Dronning Maud Land drill site (EDML), firn cores and trenches were sampled and analyzed for density (Figure 1). The site is characterized by a mean temperature of −44.5°C and a mean accumulation rate of 62–73 kg/m2/yr depending on the age interval used in the accumulation estimate [Oerter et al., 2004; Klein, 2014]. The isotope data of the trenches are described in Münch et al. [2016]. Figure 1Open in figure viewerPowerPoint Map of the firn-core and snow trench positions. Firn-core sites are shown as green filled circles, trenches as red lines. The drilling site of the EPICA Dronning Maud Land (EDML) ice core is shown as black star. 2.1 Density Profiles From Firn Cores Four firn cores of at least 90 m length were drilled in the close vicinity of the EDML deep-drilling site and analyzed for density variations. We include three of the cores in this study (Figure 1) as the core quality of the first core of the season (B40) was considerably lower than for the remaining cores. The firn cores were transported in 1 m pieces to the cold laboratory at the Alfred-Wegener-Institute (AWI) in Bremerhaven, Germany. Density of the firn cores was measured at a 0.5 mm resolution using high-resolution X-ray computer tomography (CT) [Freitag et al., 2013a]. 2.2 2-D Density Profiles From Shallow Firn Trenches To study the horizontal density structure, close to the firn-core positions, two 1.2 m deep, 1.2 m wide, and around 45 m long trenches named T1 and T2 were excavated by using a snow blower (Table 1). Each trench was directed perpendicularly to the local snow-dune direction. In contrast to other studies, the vertical zero position for each trench was not set to the snow surface, but an absolute height reference was defined as a horizontal line between the highest points of the surface level of each trench. This was established by using short bamboo poles every 60 cm, aligned by a spirit level and a laser device. Unfortunately, due to technical problems, no absolute height reference between the two trenches could be established. A coarse measurement based on stacked laser level measurements showed that the vertical difference between the trenches is less than 20 cm. Table 1. Summary of the Snow Trenches Analyzed in this Study Trench Name Position (NW End) Bearing Length Sampled No. of Profiles T1 75.00641SS, 0.074978E 140°N 45.6 m 62 T2 75.00849S, 00.08693E 130°N 49.1 m 81 Every 60 cm, 1 m long firn cores were taken by carefully pushing a glass fiber liner with 98 mm diameter vertically into the firn at ~5–10 cm distance to the trench wall. The full liners were carefully recovered by digging them out. During the process of pushing the liners into the snow, care was taken that the snow inside the liner did not get compressed. Liners with more than 1 cm compression, visible as a reduction of the snow level at the top relative to its surroundings, were remeasured by moving the position 15 cm perpendicularly to the trench wall. In some cases, the replicate measurements again showed compression, and thus, no data could be obtained for this position. The vertical distance from the top of the liner to the absolute reference was recorded with an accuracy of 1 cm. The snow-filled liners were carefully transported to the EDML processing trench. 2.3 Dielectric Profiling and Density Estimates The dielectric stratigraphy was measured in the processing trench applying the dielectric profiling (DEP) technique using the device described in Wilhelms et al. [1998]. Measurements were performed at 250 kHz frequency in 5 mm increments. The effective resolution is limited by the finite width of the 1 cm wide electrode and was found to be around 2 cm based on analyzing the air to ice transition at the top of the core. The top and bottom 4 cm of the DEP measurements are affected by edge effects due to the air to ice transition. Sometimes, a loss of snow at the top or bottom of the liner in the packing or transporting phase could not be completely avoided. The maximum-recorded loss was 4 cm at the top and 1 cm at the bottom. We therefore only analyze the part from 8 to 95 cm depth. For each measurement, the cable stray capacitance was subtracted and the capacitance record was divided by the free air capacitance, to derive the relative permittivity. To convert the DEP measurements into density, we use the real-valued Looyenga mixing model [Looyenga, 1965]. where is the density of pure ice and εice = 3.17 is the relative permittivity of ice at frequencies above 100 kHz [Glen and Paren, 1975]. This mixing model is a good approximation in the top meters of the firn, and a comparison to gamma-ray-based density measurements showed deviations of less than 5% in the top 40 m of a firn core [Wilhelms, 1996]. 2.4 Comparison of DEP-Derived Density and CT Density It is known that effects other than density also influence the dielectric stratigraphy including strong impurity changes or changes in the snow structure [Denoth, 1989]. Further, CT-derived densities have a true resolution of 0.5 mm, whereas DEP-derived densities are always integrated over a width of several millimeters and are more sensitive to core breaks. As we use both density measurement techniques in our study, we tested comparability of the methods by comparing the CT- and DEP-derived density for the firn core B41 (Figure 2). We analyzed the top 10 m for they cover a similar density range (250–550 kg/m3) as found in the trench. CT-derived densities were smoothed over 25 mm by using a 50-point running mean filter to result in a comparable resolution. In the remaining manuscript, we will always use the 25 mm running mean smoothed CT data except for the spectral analysis, which is performed on the raw data. Figure 2Open in figure viewerPowerPoint Comparison of CT and DEP-derived density of the top 10 m of B41. (a) Comparison of the depth series. The grey shows the areas in which a low core quality was noted during the DEP measurements. Both axes have the same scaling. (b) Scatterplot of DEP versus CT density. The 1:1 line is shown in red. Points with low core quality (= grey in Figure 2a) are not included in the analysis. Both methods show a strong agreement in variations on the centimeter to meter scale (R2 = 0.91). A small offset (12 kg/m3 difference in the 0–10 m mean density) between the CT and DEP density is observed, with the DEP showing a smaller density than the CT, but this does not affect our results as we only rely on relative density variations in this study. The main deviations occur at core breaks and at the boundary of the 1 m pieces. By integrating over several centimeter, the DEP device is more sensitive to small geometrical deviations from core catchers and breaks than the CT measurements. As the trench DEP measurements were performed by using firn inside liners, the firn/snow is better preserved, and we expect a higher quality of the DEP-derived density of the trenches compared to the firn-core results shown here. 2.5 Ion Measurements Within trench T2, discrete sampling at core positions 0.3, 10, 29.80, and 40 m (distance along the trench) was conducted. The snow surface was cleaned by removing the outermost layers. The snow was sampled in 3 cm increments and collected in Whirlpak plastic bags by using a precleaned Teflon spatula. The samples were closed carefully and sealed again for transportation back to Germany. During sampling protective Tyvek cloths were worn to avoid contamination. Samples were shipped in cool containers to Bremerhaven, Germany, and kept frozen until analysis. The analyses of the samples were carried out in the ion chromatography laboratory facilities at AWI by using DIONEX IC-S 2100. Blanks were checked regularly and typically showed below 0.6 ppb for Na+ and below the detection limit for methyl sulphonate (MSA). The ions of Na+ and MSA are presented here. Ca2+ measurements suffer from contamination and are thus not included in the analysis. Because impurity concentrations are always positive but show an asymmetry toward large values [e.g., Bigler et al., 2011], we use the logarithm of the ion concentration in all cases. 3 Results 3.1 2-D Density Data From the Trenches The large number (160) of density profiles analyzed from the two trenches allows creating a two-dimensional image of the horizontal and vertical density variations (Figure 3). The density distribution in both trenches shows a high vertical and horizontal variability with very similar magnitudes in both dimensions (sdvertical = 20.7 kg/m3 and sdhorizontal = 19.7 kg/m3). Figure 3Open in figure viewerPowerPoint 2-D (vertical and horizontal) structure of the DEP-derived density variations of both trenches. The depth is relative to an imaginary horizontal surface, which was arbitrarily set to 0 at the mean actual snow surface. The actual snow surface is shown as grey line. The removal of areas affected by edge effects in the DEP measurements lead to the missing area below the snow surface. Dots at the bottom of each panel depict where a profile was taken. The vertical dashed lines indicate the positions of the profiles shown in Figure 4. For displaying reasons, we limit the color scale to the 99% quantile range of the density variations. Roughly four layers can be identified as alternations of high- and low-density regions. Interestingly, as previously found while analyzing the isotope data from trench T1 [Münch et al., 2016], the density layers below 30 cm seem to be on average horizontal and do not follow the actual snow surface. Indicative for this behavior is the smaller horizontal variability below 30 cm when all profiles are aligned to an absolute height (sd = 17.8 kg/m3) compared to the horizontal variability relative to the snow surface (sd = 20.1 kg/m3). Therefore, in the remainder of this study, we use and display the profiles relative to an absolute height reference and not to the snow surface. The layers show undulations of several centimeters, similar in magnitude to the surface undulations that were present during the time of the sampling. Some density variations could be explained by vertically compressing or stretching the same density profile. However, in addition, other features are visible. Prominent are high-density anomalies (400–420 kg/m3) with a horizontal extension of 2–5 m, which are especially frequent in the 20–70 cm firn depth range. 3.2 Correlation Structure of Trench and Firn-Core Data To visualize the density variability and representativity of single profiles, we show four arbitrarily picked density profiles from the trenches T1 and T2 together with the density profiles of the top 5 m from the firn cores (Figure 4). This mimics the results we would have obtained in a classical study analyzing single snow pits or firn cores. The first meter does not show an increase in density. Furthermore, no clear seasonal cycle is visible in the uppermost meter. Interestingly, all shown density profiles are uncorrelated to each other (R < 0.2 for all possible correlation pairs), demonstrating that the density profile of a single core is not representative for a larger region, at least on the vertical centimeter to meter scale shown here. Figure 4Open in figure viewerPowerPoint Examples of vertical density profiles from the trench and firn-core data. Profiles from trench T1 at the horizontal position 5 and 45 m, from T2 at 5 and 45 m, and the top 5 m from firn cores B41, B42, and B50 are shown. There is no significant correlation between any of the profiles (R < 0.2 using detrended data). Similar results are obtained from analyzing all profiles by correlating all the possible pairs of individual profiles from trench T1 with individual profiles from T2. This results in a mean correlation of 0.0. Allowing for a shift in the depth of + −12 cm for every single profile, and thus for potential undulations of the present snow surface, increases the mean correlation to 0.31. However, this is roughly what is expected by chance (R = 0.28 using surrogate data with the same autocorrelation structure as the sample autocorrelation of the real data) since allowing for shifts always leads to spurious positive correlations. While these horizontally separated profiles are uncorrelated, a different picture is obtained for nearby profiles. Estimating the correlation for different separation distances (Figure 5) shows that nearby ( = 10 m Distance) Signal-to-Noise Ratio R/(1-R) Lag to Density in cm (~months) Density 0.13 0.15 δ18O 0.53 1.1 −3 cm (−2 months) MSA 0.30 0.43 −1 cm (−1 month) Na+ 0.31 0.45 9 cm/−10 cm (+5/−6 months) a Correlations for density (this study) and δ18O [Münch et al., 2016] are the mean values of both trenches, whereas MSA and Na+ (this study) were only measured in Trench 2. Lags were determined by cross correlation with negative values indicating a lead relative to density. For Na+, the maximum cross correlation for negative and positive lags is the same inside uncertainty; therefore, both values are given. 3.4 Spectral Analysis of the Vertical Firn-Core Density Evolution The four layers in the trench density data set (Figure 3), the reproducible density variations in the mean profiles, and the correlation with the δ18O profile (Figure 6) that is usually interpreted as a temperature signal suggest that the density variations contain a seasonal signal. This is further supported by the observation that the layer thickness of around 20 cm (Figure 3) agrees with the annual accumulation rate at the site. In a first view, this seems to contradict earlier findings [Hörhold et al., 2012; Freitag et al., 2013a] where it was shown that high-resolution density data in the top meters of the firn do not show any clear seasonality as diagnosed by analyzing the depth variability in the spectral domain. As these previous studies have been obtained at different sites and with different data sets, we repeat the wavelet analysis by using the same methods as described in Hörhold et al. [2012] for the CT density data of all three firn cores. To be able to investigate variations in respect to time as the seasonal cycle, the depth scale is converted into water equivalent depth (w.eq.) and averaged to 5 mm w.eq. resolution. Low-frequency variations in the density records were removed by using a finite impulse response high-pass filter [Bloomfield, 1976] (cutoff frequency of 0.5 m w.eq. −1). The wavelet sample spectrum was estimated by using the Morley wavelet (sowas package) [Maraun and Kurths, 2004] to analyze the depth-dependent behavior of the density in the frequency domain. Local significance was again tested against a red-noise null hypothesis. The mean wavelet for all three cores is shown (Figure 7), but similar results are obtained with any of the three cores. We note that this mean wavelet spectrum is a different quantity to that obtained by stacking the three cores first and than calculating the wavelet. Figure 7Open in figure viewerPowerPoint Spectral analysis of the depth dependency of density variability. (a) Mean wavelet spectrum of the firn-core density variations. The black contour lines enclose the regions with significant concentrations of spectral power. (b) Mean power spectra of density variations at three depth intervals. Significance levels are shown as dashed lines. The approximate range of annual layer thickness is shown as horizontal grey lines in Figure 7a and vertical grey lines in Figure 7b. Significance (p = 0.05) in Figures 7a and 7b tested assuming a first-order autoregressive (AR(1)) model. The result of the wavelet analysis (Figure 7a) is very similar to the ones previously obtained from other Antarctic and Greenlandic cores [Hörhold et al., 2012], showing a broad continuum of variability at the surface, a decrease in variability with a minimum around 10–20 m w.eq. and statistically significant patches of energy (black contours in Figure 7a) appearing at depth close to the frequency of the modern accumulation rate range. To complement the results of the wavelet analysis, we estimate the power spectra of the density variability of the firn cores at different depths (Figure 7b) using the multitaper technique. Local significance was tested against a red-noise null hypothesis. The spectrum of the upper 10 m w.eq. of the core shows a variability continuum from the millimeter scale up to 0.3 m w.eq. length scale with increasing energy toward longer variations and no clear, statistically significant signal at the indicated present-day accumulation rate. In contrast, the two other depth intervals show a distinct signal at the frequency of the accumulation rate. This in good agreement with former findings, where wavelet analysis revealed no distinct frequencies at the surface but a development of a significant peak at the frequency of accumulation rate with depth. We note that the statistically nonsignificant local maximum at the low-frequency end of the spectrum (2–3 m w.eq.−1) might be an artifact of the core processing as 1 m pieces correspond to ~0.34 m w.eq. at the surface and more deeper in the core. Therefore, we can summarize that using spectral analysis, no clear seasonal cycle can be detected in the near surface firn (upper 10 m w.eq.) when analyzing single cores. 4 Discussion We present an extensive high-resolution density data set characterizing the near-surface density variability in an Antarctic low-accumulation region. In the following, we will discuss potential mechanisms for the build up of the layered snow column, its link to the layered structure of compacted firn, and the representativeness of single firn-core measurements. 4.1 Spatial Variability of Density in the Upper Snow Column The re-deposition due to interaction with wind is one of the reasons for the high lateral density-variability of surface snow [Fisher et al., 1985; Birnbaum et al., 2010; Libois et al., 2015]. Our site is located within a region of light catabatic winds leading to a moderate mean snow density of 340 kg/m