Introduction

Glacier snowpacks can be regarded as solute stores. Winter precipitation scavenges sea salt and acid aerosols from the atmosphere, incorporating them into tile snowpack along with aeolian dust. This solute is subsequently released to catchment runoff as the snowpack melts. The purpose of this paper is to describe the chemical composition of the snowpack covering a High Arctic glacier, the provenance of the solute, and its impact on the composition of runoff from the glacier.

Site and Methods

Scott Turnerbreen is a 3.3 km² glacier, of altitude range 230-680 m a.s.l., located in central Spitsbergen (78° 06' N, 15° 57' E, Figure 1a), occupying a catchment of total area 12.8 km² (Fig. 1b). The catchment geology is dominated by gently dipping Tertiary sandstones, shales and siltstoncs (Reference Major and NagyMajor and Nagy, 1972). The glacier is non-temperate throughout, and drainage is essentially subaerial (Reference Hodgkins., Tranter and DowdeswellHodgkins and others, in press a).

Fig. 1. a. Location of Scott Turnerbreen within the Svalbard archipelago, b. The Scolt Turnerbreen catchment. Contours are in m a.s.l. Solid bold lines at the glacier snout indicate principal runoff routes.

Sampling Programme and Procedure

Ninety “dry” snow samples were obtained from six pits located along the glacier centre line, at altitudes from 240 to 560m a.s.l., between 27 April and 12 May, 1993, before seasonal melting began. The depth of the pits varied from 1.48 to 1.97 m, and a layer of depth hoar 0.19-0.28 m thick was present at the base of each. The mean density of the snowpack was 332 kg m^-3 Forty-one “wet” snow samples were obtained from ten pits situated near the transient snowline on the glacier surface, at altitudes from 235 to 310 m a.s.l., between 1 June and 9 July, following the onset of seasonal melting. The depth of the pits varied from 1.58 to 0.16 m. The snow consisted largely of granular crystals 1-2 mm in diameter. Depth hoar was absent and ice layers were frequently encountered, particularly after 18 June. The mean density of the melting snowpack during the period 1-18 June was 425 kg m ^-3. In addition, seventy-two meltwater samples were collected from the trunk stream draining the glacier, between 14 June and 10 July. The sample series commences with the onset of proglacial surface runoff. At least two samples per day were collected, at times as close as possible to minimum and maximum discharge (respectively 10.00 hours and 17.00 hours, on average).

Snow samples were sealed in polythene bags and subsequently melted by immersion in warm water, then immediately vacuum filtered through 0.45μm cellulose nitrate membranes. Two 60 ml aliquots of filtrate (one for cation anil Si, the other for anion and alkalinity analyses) were stored air-free in polypropylene bottles in cool, dark conditions until return to the U.K. Care was taken to prc-rinse the filtration apparatus and storage bottles with sample and filtrate as appropriate, to avoid contamination. Melt-water samples were collected in a pre-rinsed 500 ml polypropylene bottle, then immediately filtered and stored in the same way as the solid samples.

Stream stage was measured every 5 minutes with a Druck PDCR830 pressure transducer in a stable stretch of unbraided proglacial stream, c. 250 m downstream of the confluence of the ice-marginal streams which drain the glacier. Stream stage was logged as hourly means over the period 15 June-11 July. Discrete discharge measurements were undertaken at the same location by the relative dilution method and used to obtain a rating relationship for converting stage to discharge. The typical error of the relationship is ± 8%.

Laboratory Analyses

pH was determined on samples no later than 14 days after collection; air-free storage should have prevented significant change in sample pH. pH was determined with an Orion 290a portable pH meter with Ross combination electrode, calibrated with Orion low ionic strength buffers ol pH 4.0 and 7.0. The temperature compensation applied was that appropriate for the measured temperature of the buffer. The precision of the measurements is ± 0.2 pH units.

The major ion and Si concentrations of samples returned to the U.K. were determined as follows, file concentrations of major base cations (Na⁺, K⁺, Mg²⁺, Ca²⁺ and acid anions (Cl^-, NO₃ ^-, SO₄ ^2-) were determined by ion chromatography on a Dioncx 4000i ion Chromatograph with, respectively, CG12/CS12 and Fast Anion-1 columns, and MSA (methane sulphonic acid) and NaHCO₃/Na_sCO₃ elucnts. Sample loop volumes were either 50 or 200μ1, depending on concentration. The detection limits for these determinations are approximately 0.1μeql^-1. The precision of the analysis varies with the concentration of the sample, from ±3% at concentrations in excess of 50μeql^-1 to ± 100% at 1μeql^-1.

The concentration of dissolved Si was determined by flow-injection analysis, using alècator FIAstar 5010 system with FIAstar 5023 spectrophotometer, V100 injector, Type III Chemifold and automatic sampler. 1 ml samples were injected into a stream of ammonium molybdate. oxalic acid and acidified stannous chloride to determine the concentration of molybdate- reactive Si, detected colorimetrically at a wavelength of 695 nm. The detection limit of this system is 1μmol l^-1, and the precision of the analysis ± 2μmol l^-1. Alkalinity (HCO₃ ^-1) was determined by titrating 25 ml of filtered sample to an end point pH of 4.5, using 0.01 M HCl (standardized with 0.01 MNa_sCO₃). detected with BDH mixed indicator. Titre volumes ranged from 0.10 to 6.35 ml. The precision of the analysis varies with titre volume, from ± 1 to 50%. HCO₃ ^-1 is overestimated at low concentrations.

Results

Snow Composition

The composition of snow samples is summarised in Table 1. Dry snow is generally dilute, although there is considerable variability between samples, with a range of c. 5000μeql^-1. Maximum concentrations correspond with a layer discoloured by aeolian dust, which was present in all snow pits. The dry snowpack is slightly acidic overall, with a mean pH of 5.4. and its composition is dominated by Na⁺ and Cl^-1. Si and NO₃ ^- were present in trace quantities, the formel probably being derived from aeolian dust and the latter from acid-nitrate aerosol. Dry-snowpack solute profiles are presented in Figure 2a-c; these are derived from the mean concentrations of all samples at equivalent proportional snow-pit depths (n= 6) for each of 15 depth intervals), standardized to the mean snow depth. Wet snow is more dilute than dry snow (mean total solute concentration 340 as opposed to 420 μeql^-1), and less acidic (mean pH of 5.9). This is consistent with the elution of ions from the snowpack by meltwater (Reference Johannessen and HenriksenJohannessen and Henriksen, 1978; Reference Davies, Vincent and BrimblecombeDavies and others, 1982: Reference Brimblecombe, Tranter, Abrahams, Blackwood, Davies and VincentBrimblecombe and others, 1985; Reference Tranter, Davies, Tranter and JonesTranter, 1991; Reference Goto-Azuma., Nakawo, Shimizu, Azuma, Nakayama and YokoyamaGoto-Azuma and others, 1993b), as previously observed on Svalbard (Reference Goto-Azuma, Enomoto, Takahashi, Kobayashi, Kamcda and WatanabeGoto-Azuma and others, 1993a; Reference HodsonHodson, 1994).

Table 1. Summary statistics for the compositions of dry (standard text) and wet (bold text) snow samples. Ratios of the concentrations of ions to Cl^-) in seawater are from Reference HollandHolland (1978)

Fig. 2. Dry snow profiles. a,pH; b, cations'sum; c, anions'sum. In each case, the plotted value is the mean of six measurements in different snow pits at equivalent fractional depths; each profile has been standardized to the mean depth of the dry snowpack.

The statistical significance of the differences in concentration between sample types cannot satisfactorily be tested (Reference Hodgkins., Tranier and DowdeswellHodgkins and others, in press b). A non-parametric test is required because of the highly skewed frequency distributions. A Mann-Whitney U test is not sufficiently sensitive to variations in concentration (differences of 1 and 1000 μeql^-1 being assigned equal importance through ranking), while a dilute sample in which all ion species are represented is not sensitively distinguished from a concentrated sample from which one or two arc absent. Furthermore, concentrations of ion species cannot be converted into mutually exclusive comparative frequency categories with sufficient cases to perform a valid X² test.

Runoff Composition

Significant temporal variations occur in the chemical composition of runoff. The overall trend is for the runoff to become more dilute as the melt season progresses but there are also significant variations in the relative proportions ol ion species (Fig. 3a). When runoff commences on 14 June, the composition of the bulk meltwater is dominated by Na⁺ and HCO₃, with minor proportions of SO₄ ^2-, Ca²⁺, Mg²⁺ and Cl^-1. By 23 June, the dominant ions are Na⁺ and Cl^-, with minor proportions of Mg²⁺, SO₄ ^2-, Ca²⁺ and HCO₃ ^-. From around 30 June to the end of the monitored period, HCO₃ ^- again becomes the dominant anion. K⁺ and especially NO₃ ^- are present in only trace quantities throughout.

Fig. 3. Runoff time series, a. Individual ion-species concentrations and meltwater discharge; b. Total solute concentration, snow-pack-derived solute concentration and discharge.

Solute Provenance

Solute within the snowpack has heen partitioned into sea-salt aerosol, acid aerosol and crustal components following Reference Sharp, Tranter, Brown and SkidmoreSharp and others (1995). The mean ionic concentrations of dry snow were the basis of the calculations, full details of which have been given by Reference Hodgkins., Tranier and DowdeswellHodgkins and others (in press b). Briefly, all NO₃ ^- was assumed to be aerosol-derived (precipitation-scavenged acid-nitrate aerosol) and all Cl^-1 to be sea-salt-derived (precipitation-scavenged sea-salt aerosol). Sea-salt contributions to Na⁺, K⁺, Mg²⁺ and Ca²⁺ were then determined from the Cl^- concentration using standard sea-water ratios (Reference HollandHolland. 1978). Residual concentrations of these ions were attributed to a crustal source (aeolian dust); Si is entirely crustally derived. A snow pack-derived SO₄ ^2- concentration was determined from the Cl^- concentration using the mean ratio (assumed constant) of these species in the pre-melt season snowpack (0.125); the sea-salt component of this concentration was determined from the standard sea-water ratio, and the remainder is apportioned to precipitation-scavenged acid-sulphate aerosol. The resulting partitioned concentrations in units of μeql^-1 were converted to units of kg m^-3 and then to masses in kg by multiplying by the water-equivalent volume of the glacier snowpack. Although snow-density data are not available for all samples, mean concentrations are here believed to be a reasonable approximation to volume-weighted means, as measured snowpack density profiles are relatively uniform (standard deviation 0.05 kg m ^-3). The glacier area is 3.3 X 10⁶ m², the mean glacier snow depth is 1.6 m (162 point measurements) and the snow density 330 kg m ^-3 (40 measurements in four separate snow pits). The glacier snow volume is therefore 5.4xl0⁶m³ and the water-equivalent glacier snow volume 1.8 X 10⁶m³. The results are given in Table 2. The snowpack composition is clearly dominated by sea salt, which accounts for about 98% of total solute by-mass. This result is supported by the ratios of the concentrations of ions to CI in the dry snowpack, which arc very close to those in sea water (Table 1). The total solute mass of the glacier snowpack corresponds to a yield of 23 000 kg km^-2 a^-1.

Table 2. The provenance and yield of solutes from the 1992-93 Scott Turnerbreen snowpack. Tace quantities of Si are present

Snowpack Solute in Runoff

Reference Hodgkins., Tranter and DowdeswellHodgkins and others (in press a) have described the hydro-chemistry of runoff from Scott Turnerbreen in detail and have indicated that high but rapidly declining solute concentrations in the earliest days of the melt season (e.g. Na⁺, HCO₃ ^- and SO₄ ^2- , Fig. 3a) can be attributed to solute acquisition from a proglacial icing, which represents concentrated, stored winter runoff. The behaviour of Cl^- in runoff is distinctive: its concentration increases from less than 200 μeql^-1 when runoff commences on 14 June, to over 800 μeql^-1 on 23 June, declining thereafter to about 50 μeql^-1 at the end of the monitored period (Fig. 3a). Mg²⁺ follows a similar pattern, and so, to a lesser extent, do Na⁺ and Ca²⁺ (Fig. 3a).

The only source of Cl^- to the bulk meltwater is the snowpack (precipitation-scavenged sea-salt aerosol). Having established that elution of solutes occurs within the snowpack, the liming of the Cl^- peak relative to the onset of sustained high discharge of meltwater strongly suggests that this represents the discharge of concentrated snowmelt. The three-five-fold concentration of Cl^- in the runoff, relative to the snowpack, is consistent with Held and laboratory studies of the composition of snowmelt and its concentration by elution within snow (Reference Johannessen and HenriksenJohannessen and Henrikscn, 1978; Reference ColbeckColbeck. 1981: Reference Davies, Vincent and BrimblecombeDavies and others, 1982; Reference Brimblecombe, Clegg, Davies, Shooter and TramerBrimblecombe and others. 1987; Reference Tranter, Davies, Tranter and JonesTranter, 1991).

This concentration peak is the product of processes that concentrate solutes on the surface of individual ice grains, i.e. melt-freeze cycles (Reference Bales, Davis and StanleyBales and others, 1989) and vapour transfer (Reference Cragin, Hewitt and ColbeckCragin and others, 1993), and the initial distribution (Reference Bales, Davis and StanleyBales and others, 1989) and subsequent redistribution by percolating meltwater (Reference ColbeckColbeck, 1981) of solutes within the snow pack. Snowmelt therefore becomes relatively dilute for the remainder of the melt season. The snowpack-derived concentration of a given ion in the runoff, C _{X-snow/runoff} can be estimated from the product of the concentration of Cl^- in runoff, C_{cl-snow/runoff}, and the ratio of the ion to Cl^-1 in the dry snowpack, K, assumed to be constant, i.e. C _{X-snow/runoff} = K[C_{cl-snow/runoff]} . The assumption that K is constant implies that significant preferential elution of ion species does not occur: reference to the ratios of ion concentrations to Cl_- in Table 1 suggests that this is reasonable. Total and snowpack-derived solute concentration time series are shown are Figure 3b, which illustrates the rapid early melt-season “flush” of concentrated snowpack solute in runoff, ami its subsequent dilution.

Discussion

The results presented in this paper confirm the dominance of sea-salt-derived solute in the snow cover of Svalbard. Reference Arkhipov, Moskalcvsky, Glaznvsky, Maycwski, Whitlow, Pulina and ErasoArkhipov and others (1992) sampled the snow cover on Kongs-vegen and Amundsenisen, and concluded that sea-salt aerosols dominated the composition of the snowpack. Reference Semb, Baekkan and JorangerSemb and others (1984) concluded that the observed concentrations of ions in snow were consistent with precipitation scavenging from a near-surface atmospheric layer, and that sea salt was the dominant source of solute. By partitioning snow pack solute according to provenance, this paper has quantified the relative contributions of sea-salt aerosol, acid aerosol and aeolian dust to the snow pack: 97.5, 2.1 and 0.4%. respectively Table 2).

The composition of snow sampled at Scott Turnerbreen prior to and following the onset of seasonal melting is consistent with the elution of ions from the snowpack by meltwater. The results presented in this paper enable t he impact of this process to be traced quantitatively from the snowpack to the runoff. The proportion of snowpack-derived solute in total meltwater solute is 12% when surface runoff commences on 14 June; it increases linearly to a maximum of 75% on 24 June, then decreases quasi-exponentially to 17% at the end of the monitored period on 10 July (Fig. 3b). The contribution of snowpack solute to runoff can therefore be demonstrated to be a brief but intense concentration episode early in the mell season.

Reference Goto-Azuma., Nakawo, Shimizu, Azuma, Nakayama and YokoyamaGoto-Azuma and others (1993b) found that 10-60% of snowpack SO₄ ^2- at Austrc Broggerbreen in Svalbard was from non-sea-salt sources. Other measured non-sea-salt SO₄ ^2- concentrations in snow from Svalbard are in the range 4.5-11 μeql^-1 (Reference GjcssingGjcssing, 1977; Reference Semb, Baekkan and JorangerSemb and others, 1984). The results presented in this paper quantify the amounts of solute derived from sources other than sea salt, principally acid aerosols. The snowpack acid aerosol solute yield amounts to 140 kg km ^-2 a^-1, or 2.1% of the total snowpack solute yield (Table 2). The High Arctic islands arc sinks for long-range, trans-boundary atmospheric pollution (Reference Hansen, Hanson and NorrisHansen and others, 1996). The supply of atmospheric pollutants to Svalbard exhibits strong seasonality, and is at a maximum during February April (Reference Rahn and McCaffreyRahn and McCaffrey, 1980; Reference Joranger and OltarJoranger and Ottar, 1984), corresponding to the winter precipitation maximum in this part of the archipelago (Reference Hanssen-Bauer, Solas and SteffensenHanssen-Bauer and others, 1990). Winter prevailing winds in the Svalbard archipelago are easterly and southeasterly (20% and 40% frequencies, respectively; Reference Hanssen-Bauer, Solas and SteffensenHanssen-Bauer and others. 1990), i.e. from the direction of industrial Eurasian regions. Deposition of precipitation-scavenged acid sulphate and nitrate aerosols therefore occurs in the Svalbard snowpack and is a potential contributor to meltwater acidification. A threc-fivc-fold increase in H⁺ in runoff, consistent with the concentration of solutes due to snowpack flushing described in this paper, could decrease the mean pH of runoff from 7.3 to 6.6-6.8, other factors remaining equal (since-log₁₀ (H⁺) = pH, where H⁺ is in units of mol l^-1). Though aquatic ecosystems are not typically well developed in glacierized catchments, the buffering capacity of such catchments is also characteristically limited, so meltwater acidity is unlikely to be effectively neutralized.