Earth and Planetary Science Letters 333–334 (2012) 238–249
Contents lists available at SciVerse ScienceDirect
Earth and Planetary Science Letters
journal homepage: www.elsevier.com/locate/epsl
Modelling the liquid-water vein system within polar ice sheets as a potential microbial habitat
K.G. Srikanta Dani a,1, Heidy M. Mader a,n, Eric W. Wolff b, Jemma L. Wadham c a School of Earth Sciences, University of Bristol, Wills Memorial Building, Queen’s Road, Bristol BS8 1RJ, UK b British Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET, UK c School of Geographical Sciences, University of Bristol, University Road, Bristol BS8 1SS, UK article info abstract
Article history: Based on the fundamental and distinctive physical properties of polycrystalline ice Ih, the chemical and Received 12 October 2011 temperature profiles within the polar ice sheets, and the observed selective partitioning of bacteria into Received in revised form liquid water filled veins in the ice, we consider the possibility that microbial life could survive and be 21 March 2012 sustained within glacial systems. Here, we present a set of modelled vertical profiles of vein diameter, Accepted 6 April 2012 vein chemical concentration, and vein water volume variability across a range of polar ice sheets using Editor: G. Henderson Available online 22 May 2012 their ice core chemical profiles. A sensitivity analysis of VeinsInIce1.0, the numerical model used in this study shows that the ice grain size and the local borehole temperature are the most significant factors Keywords: that influence the intergranular liquid vein size and the amount of freeze-concentrated impurities polar ice cores partitioned into the veins respectively. Model results estimate the concentration and characteristics of polycrystalline ice the chemical broth in the veins to be a potential extremophilic microbial medium. The vein sizes are psychrophilic bacterial metabolism vein system estimated to vary between 0.3 mmto8mm across the vertical length of many polar ice sheets and they temperature depression may contain up to 2 mL of liquid water per litre of solid ice. The results suggest that these veins in polar ice sheets could accommodate populations of psychrophilic and hyperacidophilic ultra-small bacteria and in some regions even support the habitation of unicellular eukaryotes. This highlights the importance of understanding the potential impact of englacial microbial metabolism on polar ice core chemical profiles and provides a model for similar extreme habitats elsewhere in the universe. & 2012 Elsevier B.V. All rights reserved.
1. Introduction Fig. 1(a)–(e) shows the geometry of the intergranular vein system found in natural polycrystalline ice Ih (hexagonal ice) that The presence of bacteria and archaea in glaciers and ice sheets makes up the bulk of temperate glaciers and polar ice sheets on has been reported by numerous researchers (Abyzov, 1993; Karl Earth (Nye and Frank, 1973). The liquid water phase exists in solid et al., 1999; Siegert et al., 2001; Foght et al., 2004; Gaidos et al., ice because the ice lattice tends to reject foreign ions (impurities) 2004; Abyzov et al., 2005; Kastovska et al., 2007; Hodson et al., as water is frozen. In other words, the solubility of compounds in 2008). For many years it was thought that the extremely low the individual ice grains is generally very low and the expelled temperatures within natural ice deposits on Earth and the lack of foreign ions from the growing grains remain in the liquid water liquid water, light and nutrients in them presented too harsh an and become more concentrated with decreasing temperature. environment to sustain any form of life. More recently however it Ultimately, an ice polycrystal is formed which contains an inter- has been proposed that the intergranular aqueous vein system connected network of highly concentrated water-filled veins found in ice could provide a habitat capable of sustaining around ice grains and films on grain boundaries. microbial life (Price, 2000; Mader et al., 2006; Rohde and Price, Experiments have demonstrated that microorganisms partition 2007) on Earth and that similar habitats might exist on other icy preferentially to the veins during grain growth (Fig. 1(f)–(i), Mader heavenly bodies such as Mars and some icy moons of Jupiter and et al., 2006; Amato et al., 2009). Even at extreme sub-zero tempera- Saturn (Price, 2002, 2007; Parkinson et al., 2008; Newman et al., tures in the veins, bacteria can find both liquid water and much 2009). higher concentrations of nutrients (impurities) than the average impurity concentration estimated in bulk ice. Moreover, some psychrophilic microbes show active growth at temperatures as low n Corresponding author. Tel.: þ44 0117 9545445. as 12 1C(Breezee et al., 2004). Some are shown to be metabolically E-mail address: [email protected] (H.M. Mader). 1 Present address: Department of Biological Sciences, Macquarie University, active down to 20 1C(Gilichinsky, 2002; Price and Sowers, 2004) North Ryde, Sydney, NSW 2109, Australia. and 39 1C(Bakermans, 2008). However, the mechanisms that
0012-821X/$ - see front matter & 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.epsl.2012.04.009 K.G.S. Dani et al. / Earth and Planetary Science Letters 333–334 (2012) 238–249 239
Fig. 1. Ice vein geometry and partitioning of microbes into liquid veins in ice (for a detailed figure caption with explained notations, please see supplementary material 1. (a) SEM of a typical ice vein cross section at a triple junction (crystal edge). (b) Diagram of a vein cross-section. (c) Semi-regular truncated octahedron (represents an ice grain) and a sketch of vein network surrounding it (after Price, 2000). (d and e) Transmitted white lightphotographs of the vein system in laboratory-grown ice (from Mader, 1992a); veins are 100 mm across. It is a curious point to note that tetrahedral geometry of nodes complements the inherent structural tendency towards stability of water molecules that cluster to form hydrogen bonds in a tetrahedral geometry. (f) Light and (g) fluorescence micrographs showing 1.9 mm fluorescent beads (size equivalent of bacteria) lined up along an ice vein running into a node (from Mader et al., 2006). (h) Bright field and (i) 510–560 nm epifluorescent micrographs showing yeast cells (44 mm) within an ice vein triple junction of size 410 mm (reproduced with permission from Amato et al., 2009). The scale on (f) applies to all the 4 photographs. enable maintenance of microbial intracellular fluidity at subzero A sensitivity analysis of the numerical model (VeinsInIce1.0) temperatures remain uncertain (Russel, 2006). There are very limited allows us to draw inferences about the relative impact of the experimental and/or modelled data on the interaction of ice bio- parameters tested in the model on the features of the habitat. For chemistry and the microbial activities at extremely low temperatures ice veins to develop in polycrystalline ice, the deposited snow has (below 20 1C) within glacier ice (Rivkina et al., 2000; Junge et al., to be preserved and compressed over several hundred years, 2004; Bakermans, 2008). which is not observed in shallow perennial snowpacks (up to Inorganic impurities from polar ice cores have been studied in 50 m deep) and some temperate glaciers, which are mostly detail in the scientific literature because they can act as proxies characterised as either superficial fresh snow or firn ice. The for palaeoclimate reconstructions (e.g., Wolff et al., 2006, 2010). polar ice sheets are volumetrically the most significant on earth. It is known that temperature and chemical concentrations in For these reasons we consider only deep (100–3000 m) polar ice natural polycrystalline ice regulate the intergranular vein size sheets for our analysis to establish the conditions in the veins. (Mader, 1992b; Paterson, 1994) and hence control their water volume and potential microbial population carrying capacities. Furthermore the chemical concentrations control the diffusion rates of chemical impurities through grain boundaries and veins 2. Definitions and ultimately govern the microbial metabolic reactions, which in turn could impact on vein chemistry. Anomalies observed in polar In the following sections, it is important to distinguish care- ice core gas records (e.g. N2O, CH4) have been attributed to fully between various symbols and terms that are used to refer to potential in situ microbial activity (Sowers, 2001; Tung et al., impurity concentration. 2005). In addition, some models suggest that the rate of down- The bulk impurity concentration C is given by the total mass of ward diffusion of impurities along the liquid veins within ice impurity (usually given in moles) divided by the total ice volume sheets could be significantly more than the rate of rheological ice (grainsþveinsþgrain boundary films). In this paper the term bulk is flow in ice sheets and this can lead to a major displacement of always used for concentrations averaged across the total ice volume climate signals trapped in ice cores (Hubbard et al., 2003; Rempel in this way. The vein impurity concentration is cv and is simply the et al., 2001; Rempel and Wettlaufer, 2003). total mass of impurity that partitions to the veins divided by the We utilise selected site-specific data sets on temperature and total vein volume. It is assumed that this is also the impurity chemical impurity profiles of polar ice cores and combine these concentration in the grain boundary films. We can define an datasets with a numerical model (VeinsInIce1.0) that encapsulates associated bulk veinþfilm impurity concentration Cb which is the our knowledge of the physical properties of the vein system to total mass of impurity that partitions to the veins and films divided calculate vertical profiles of vein diameter, vein impurity con- by the total ice volume. Cb therefore represents the contribution to centration and total vein water content for these ice cores in order the bulk impurity concentration made by the impurities resident in to assess the vein system as a possible habitat for microbial life. the veins and films. Cb and cv are simply related via the total water 240 K.G.S. Dani et al. / Earth and Planetary Science Letters 333–334 (2012) 238–249 volume fraction (veinsþfilms) in the polycrystal denoted by f. measurements respectively. For GISP2 D and Dome C the number of data points per ice core is reduced by taking averages within each Cb f ¼ ð1Þ 10 m depth interval (with varying number of observations per depth cv interval). The ice core data from GRIP is discontinuous in some It is expected that the process of impurity partitioning requires regions of the ice core with occasional gaps of 20 m. The resolution of that impurity molecules form liquid or quasi-liquid layers at the ice the source data is 10 m and so no averaging across the depth interval grain boundaries as they are excluded from the solidifying ice is done in this case and data gaps are preserved. In some depth grain. We distinguish between the bulk film impurity concentration intervals, individual ion values that lie far from the main trend (more Cf (the total mass of impurities resident in the grain boundary films than about 0.5 standard deviations) are excluded from the analysis. divided by the total ice volume) and the bulk vein impurity We find that the 10 m depth interval successfully preserves the concentration Cv (the total mass of impurities resident in the veins observed trends in all the data sets with no noticeable loss of divided by the total ice volume). Clearly, there is then the following information. The data from Vostok is extracted from published simple relationship between the bulk impurity concentrations graphs; a value is extracted at every 25 m and that data is used to represent its respective 25 m depth interval. The original data set Cv ¼ Cb Cf ð2Þ fromSipleDomecoreisdifferentfromtheothersinthatithas and we can now also work out the vein water volume fraction from chemical profiles for the first 10 m of every 100 m depth interval (100–110 m, 200–210 m and so on). The average of those 10 points is Cv fv ¼ ð3Þ used to represent each 100 m depth interval. The Siple Dome dataset cv is used only to test for significant differences (if any) between the two calibration options available in the model. 3. Ice core data Grain sizes reported from these ice cores show complex variations with depth. However, in general it is seen that grains The ice core data used in this study, including depth range, become larger and coarser with increasing depth. The grain size number of measurements, their resolution (average distance between also shows consistent oscillating vertical trends for some ice cores measurements), bulk chemistry profiles, grain size profiles and their (Fig. 2(b)). Following such trends in grain size variation in original reference sources, are summarised in Table 1. Ice core available data sets, we assume a reasonable grain size (through chemical impurity profiles are available from the NOAA archive interpolation) wherever the information is unavailable. repository for Antarctic (Vostok, Dome C, Siple Dome) and Greenland We seek to determine the chemistry of the solution in the veins (GRIP, GISP2 D) ice cores. The original data sets available in the public and films and the vein size as a function of depth along the core. domain comprise chemical profiles that list concentration of indivi- No direct measurements of vein chemistry or size exist. Indeed, dual inorganic ions with depth (available online: the web link is given whilst the size of the veins within a natural ice sample at its in Table 1). Most of the ice core inorganic ion data sets, both at the original in situ temperature could in principle be measured in the levels of individual ions and the bulk concentrations examined and laboratory, no technique currently exists that would allow the used in the study are distributed normally about their means with or direct determination of the vein chemistry. Therefore, it is neces- without extreme values making positive skewness in sulphate a sary to rely on our knowledge of the bulk chemistry of the ice and noteworthy exception (Fig. 2(a)). The ice core data sets from GISP2 D make reasonable assumptions on how the different solutes parti- and EPICA Dome C have the maximum original resolutions (0.2– tion between ice grains, grain boundaries, and veins and couple 0.5 m and 0.55–1 m respectively) and produce 12,000 and 3000 this with our knowledge of the vein system.
Table 1 Polar ice core data sets used in this studya.
Ice core Location Depth Grain size Borehole Bulk ion impurity Resolution range range temperature range concentration
Cbi range Original Depth interval (m) (mm) (1C) (lM) (m) (m)
Vostokb Antarctica 781 28’S 1061 48’E 50–2080 1.6–18c 58 to 35d 5–30e –25 EPICA Dome C 751 06’S 1231 24’E 100–3200 1.8–14f 54 to 3g 5.6–15h 0.55–1 10 Siple Dome 811 65’S 1481 81’W 10–977 1–4 25 to 6i 1.9–4.5j – 100
GRIP Greenland 721 35’N 371 38’W 100–3021 1.6–26k 31 to 8l 2.2–35.6m 10 10 GISP2 D 721 35’N 381 28’W 110–3040 0.6–10n 32 to 9o 2.2–33p 0.2–0.5 10
a Sourced from the online ice core gateway maintained by NOAA at http://www.ncdc.noaa.gov/paleo/icecore.html. b Although Vostok is a much deeper ice core (3400 m), the available information is limited to 2080 m and, unlike other data sets, the values are extracted from graphs published in the cited peer-reviewed literature (Obbard and Baker, 2007; Vostretsov et al., 1984; Legrand et al., 1988). c Obbard and Baker, 2007. d Vostretsov et al., 1984. e Legrand et al., 1988. f Durand and Weiss, 2004. g Ritz et al., unpublished. h Wolff et al., 2006. i MacGregor et al., 2007. j Kreutz et al., 1999. k Thorsteinsson et al., 1997. l Johnsen et al., 1995. m Legrand et al., 1993. n Alley and Woods, 1996. o Clow et al., 1996. p Mayewski et al., 1997. K.G.S. Dani et al. / Earth and Planetary Science Letters 333–334 (2012) 238–249 241
Fig. 2. Features of ice core data. (a) Frequency distribution of sulphate ion concentration from EPICA Dome C (data from Wolff et al., 2006). Note the positive skewness and the most frequent concentration ranging between 2 and 3 mM. The distribution of bulk ion concentration Cbi is given in the inset showing the most frequent bulk value to be between 8 and 10 mM. (b) The consistent increasing trend and oscillatory behaviour of grain size with depth observed in EPICA Dome C (data from Durand and Weiss, 2004).
þ Current available evidence suggests that NH4 and Cl are cations are assume to be available for partitioning into veins. In such þ favourably partitioned to the ice grains where they can be incorpo- cases, the Cl trapped in grains is balanced by NH4 and, if needed, rated substitutionally in the ice lattice whereas Naþ ,Ca2þ ,Mg2þ , then by Naþ . In certain regions within some ice cores (see example 2 2þ SO4 ,andNO3 are expected to partition favourably to grain from GISP2 D) the cation concentration (especially Ca ) is several boundaries and veins (Mulvaney et al., 1988; Cullen and Baker, times more than the available anions. In such situations, after 2000, 2001; Barnes and Wolff, 2004). The ion partitioning is not balancing it with Cl trapped in grains we assume all the excess perfect and trace amounts of ions are expected in the regions where cations to be available in veins (see footnote in Table 2). Finally, the the ions are not assumed to be favourably partitioned. A further bulk veinþfilm impurity concentration of ions Cbi in mMiscalculatedby complication is that chemical data for a specific field setting is adding the concentrations of all cations and anions that partition to invariably incomplete with only the concentrations of certain ion the veins and films (5.032 mMþ2.759 mM¼7.791 mM). Then the bulk species measured. We can infer some of the impurity load that has veinþfilm impurity concentration Cb is calculated by dividing Cbi by 3 not been measured by requiring that both the grains and veins are (Cb¼2.597). This is done since the calibration used in the model is for separately charge neutral. To account for overall ionic charge balance, molecular H2SO4 (a single entity), which is made up of three molar þ þ þ þ 2 a certain concentration of protons (H ) is assumed. H ,NH4 equivalents of two ions (2H and 1 SO4 ). Note that we have þ (substitutional), and Na (assumed interstitial) concentrations are calculated Cbi assuming that all compounds are fully dissociated. Thus utilised in that order to balance Cl in the grain lattice. The ion we have not taken into account that some salts may precipitate from concentrations reported in parts per billion (ppb) are converted the solution as the temperature is lowered. Establishing salt pre- into mM. cipitation from the mixtures would involve detailed thermodynami- From these considerations we first calculate the bulk cal calculations, which go beyond the scope of this research. However, veinþfilm concentration of ions expressed as Cbi, which is then in cases, where much of the material consists of compounds (such as used to compute the bulk veinþfilm impurity concentration Cb as H2SO4) which individually have a low eutectic point, it is likely that a function of depth from the raw data using the method explained precipitation will be minimal, and our calculations of vein size will below. See Table 2 for three examples of the calculations for polar therefore be close to valid. By contrast, in conditions where most salts ice core chemical profiles. The values given in parenthesis would precipitate, the veins could be much smaller than we estimate.
(italicised) below relate to the example data set from EPICA Dome Finally, we have also treated H2SO4 as fully dissociated, even though C given in Table 2, unless stated otherwise. it is likely that the second proton is not dissociated in the strong acid ThechargeinmeqL 1 is calculated, taking account of the valency solutions that are experienced in the veins. This is implicitly taken 2þ of the relevant ions (e.g. [Ca ] 0.092 mMresultsinachargeof into account by our use of the actual H2SO4 freezing point curve. 0.184 meqL 1). The overall charge difference is calculated by subtract- ing the total cationic charge ([Na þ]þ[Ca2þ]þ[Mg2þ ]¼1.718 meqL 1) 2 fromthetotalanioniccharge([Cl ]þ[SO4 ]þ[NO3]þ[other 4. The numerical model VeinsInIce1.0 anions]¼5.852 meqL 1)([anions] [cations]¼4.134 meqL 1). If anions are in excess (which is true in general), the positive charge deficit is We used a numerical model called VeinsInIce1.0,whichwas assumedtobesuppliedbyHþ .ApartofthetotalHþ needed for developed to estimate some of the basic physical and chemical pro- 2 overall charge balance is transferred to SO4 to form HSO4 perties of the vein environment. The equations utilised by the pro- (2.518 meqL 1 of [Hþ ] will form bisulphate) and the remaining Hþ gramme have been reported and discussed in earlier studies (Mader, (1.616 meqL 1)isbalancedagainstCl (0.576 meqL 1)trappedin 1992a; Barnes et al., 2003). The programme takes the inputs as: grains. Any excess Hþ (1.040 meqL 1) is assumed to be in the veins. This way, the overall charge balancing is observed in both the grains depth h of the ice from the surface in [m], as well as the veins. If cations are in excess (which is rare), the excess the local temperature depression y below 0 1Cin[1C], 242 K.G.S. Dani et al. / Earth and Planetary Science Letters 333–334 (2012) 238–249
Table 2
Example calculations for bulk veinþfilm impurity concentrations Cb and their chemical breakdown. Individual ion impurity concentrations are given in mM and their charge is given in meqL 1 in parenthesis. Shaded ions (italicised) are assumed to partition favourably into ice grains.
Note: The conditions within veins are expected to be highly acidic (pHo3, Price, 2002) and the excess Hþ concentration in most cases is consistent with that requirement. Although in general, charge equivalence is observed after assuming H þ to compensate for positive charge deficit, it is important to note that in certain regions along the depth range of ice cores (e.g., GRIP and GISP2 D, 41500 m), certain cations (Ca2þ ,Mg2þ ) are in great excess making the charge deficit negative. It is possible that these ions might form micro inclusion bodies stuck within ice grains (Ohno et al., 2005). However, we assume that the excess cations are available for partitioning into veins. þ Bearing in mind that cations are better mapped within ice cores than anions, the actual charge difference is most likely to be less than the computed one.NH4 concentration is not available for EPICA Dome C, which otherwise offers the most comprehensive data set. Wherever available, it is included in the calculations. Trace ions such as formate and acetate are assumed to be in their native forms as weak organic acids and their concentrations are added to Cb wherever available (GRIP).