Origin of stratified basal ice in outlet glaciers of Vatnajokull andOræfajokull, Iceland

GRAHAME J. LARSON, DANIEL E. LAWSON, EDWARD B. EVENSON, OSKAR KNUDSEN, RICHARD B. ALLEY ANDMANTHA S. PHANIKUMAR

BOREAS Larson, G.J., Lawson, D.E., Evenson, E.B., Knudsen, O., Alley, R.B. & Phanikumar, M.S. 2010 (July): Origin ofstratified basal ice in outlet glaciers of Vatnajokull and Oræfajokull, Iceland. Boreas, Vol. 39, pp. 457–470.10.1111/j.1502-3885.2009.00134.x. ISSN 0300-9483.

During the period 2000–2005, we collected samples of englacial ice, vent water, frazil/anchor ice and stratified basalice from warm-based outlet glaciers of Vatnajokull and Oræfajokull, Iceland, and analysed them for 3H, 18O andD. Results of 3H analyses show that the stratified basal ice contains 3H from atmospheric thermonuclear testingand is younger than the englacial ice. Results of the 18O and D analyses show that frazil/anchor and stratified basalice are both enriched by an average of 2.4% in 18O and 11% in D relative to vent water. These values are consistentwith fractionation during partial freezing of supercooled subglacial water in an open system, one in which theremaining water is continuously removed and replenished by water of similar composition. The isotopic data andfield observations do not support either a regelation or a thermal ad-freeze-on origin for the stratified basal ice.

Grahame J. Larson (e-mail: larsong@msu.edu), Department of Geological Sciences,Michigan State University, EastLansing MI 48824, USA; Daniel E. Lawson (e-mail: Daniel.E.Lawson@usace.army.mil), Terrain and CryosphericSciences Branch, Cold Regions Research and Engineering Laboratory, Hanover, NH 03755, USA; Edward B. Even-son (e-mail: ede0@Lehigh.edu), Department of Earth and Environmental Sciences, Lehigh University, Bethlehem,PA 18015, USA; Oskar Knudsen (e-mail: oskar@versco.is), The Commercial College of Iceland, Ofanleiti 1, IS 103,Iceland; Richard B. Alley (e-mail: ralley@geosc.psu.edu), Department of Geosciences, and Earth and EnvironmentalSystems Institute, The Pennsylvania State University, University Park, PA 16802, USA; Mantha S. Phanikuma(e-mail: phani@msu.edu), Department of Civil and Environmental Engineering, Michigan State University, EastLansing, MI 48824, USA; received 22nd July 2009, accepted 5th November 2009.

In 1998, Alley et al. presented a theory that glaciohy-draulic supercooling of water flowing up the adverseslope of an overdeepening produces ice that accretes tothe sole of the glacier, adding to the thickness and ex-tent of the basal zone. Furthermore, Alley et al. (1998)and Lawson et al. (1998) argued that the accumulatedice will trap debris and thereby generate discontinuouslayers of sediment-laden ice, forming the stratified fa-cies of the glacier’s basal zone. Evidence in support ofthis theory was originally restricted to the MatanuskaGlacier, Alaska (Strasser et al. 1996, Lawson et al.1998, Evenson et al. 1999), although Larson et al.(2006) argued on theoretical grounds that supercoolingshould have operated under both the Laurentide andScandinavian Ice Sheets. Roberts et al. (2002) extendedsupercooling to Skaftafellsjokull, Iceland, and Cooket al. (2010) have supported its operation as a means offorming stratified facies basal ice at both Skafta-fellsjokull and Svinafellsjokull, Iceland.

Herein, we present new isotopic data and field ob-servations from eight outlet glaciers draining Vat-najokull and Oræfajokull, Iceland, that support aglaciohydraulic supercooling origin for the stratifiedfacies of the glacier’s basal zone (often shortened in textto ‘stratified basal ice’). We also consider regelation andbasal ad-freeze-on as possible origins for the stratifiedbasal ice and conclude they are not supported by iso-topic data and field observations.

Glaciohydraulic supercooling and stratifiedbasal ice

The thermodynamics of glaciohydraulic supercoolingare well established (Rothlisberger 1968; Shreve 1972;Rothlisberger & Lang 1987) and have been applied towater flow at the bed of warm-based glaciers (Alleyet al. 1998) and to basal erosion and deposition alongthe ascending slope in the glacier’s bed (Alley et al.2003). Briefly stated, glaciohydraulic supercooling oc-curs when the pressure melting point increases morerapidly than the rise of temperature derived from vis-cous dissipation of water flow, geothermal flux, heatfrom sliding and latent heat if some water freezes. Ac-cording to Rothlisberger & Lang (1987), Hooke (1991),Hooke & Pohjola (1994) and Alley et al. (1998), thephysical requirements for glaciohydraulic supercoolingare sufficient water discharge flowing up an adverseslope of magnitude41.2 to 2.0 times that of the ice-surface slope, a condition commonly met where thethinning toe of a glacier is flowing out of a subglacialbasin (the lower magnitude is for air-saturated waterand the higher one is for pure water).

Glaciohydraulic supercooling produces frazil icein flowing water and the accretion of anchor ice onconduit walls and channel beds, potentially resultingin enough ice to constrict and close conduits(Rothlisberger & Lang 1987; Hooke 1991; Hooke &

DOI 10.1111/j.1502-3885.2009.00134.x Journal compilation r 2010 The Boreas Collegium.

No claim to original US government works

Pohjola 1994; Alley et al. 1998; Evenson et al. 1999). Inaddition, Alley et al. (1998) and Lawson et al. (1998)argued that the crystalline structure of frazil accumula-tions and anchor ice as it accretes to the glacier sole islikely to trap debris in water transport and form thin(millimetres thick), discontinuous layers of ice of vari-able debris content. Furthermore, they argued that,over time, multiple periods of accretion may lead to a‘stratified’ debris-laden accumulation of ice and thickenthe basal zone. According to Lawson et al. (1998), sub-sequent deformation of the accreted layers may alsoenhance stratification while destroying properties in-herited during freeze-on. For short transport distances,the original properties of the ice and sediment may bepreserved (Lawson et al. 1998).

The most compelling field evidence in support of aglaciohydraulic supercooling origin for stratified basalice remains the Matanuska Glacier in south-centralAlaska. Here, stratified basal ice 3 to 6m thick is pre-sent for several kilometres along the glacier’s westernterminus (Lawson 1979, 1995), where ice flows out ofseveral overdeepened basins just upglacier of the ter-minus (Arcone et al. 1995; Lawson et al. 1998). Alsopresent are anchor-ice terraces, some of which extendup to 20–30m beyond the active ice margin (Lawsonet al. 1998; Evenson et al. 1999). They form during themelt season where supercooled water emerges at ventmouths and builds up anchor ice up to several metresthick in stream beds.

Strasser et al. (1996) and Lawson et al. (1998) foundthat the stratified facies of the basal zone at the Mata-nuska contains thermonuclear-derived 3H, whereasenglacial ice above it does not. This observation ledthem to conclude that the stratified basal ice was rela-tively young and derived from amixture of englacial-icemelt and modern precipitation. Furthermore, theyfound that the stratified facies ice and anchor ice wereboth enriched in 18O by �2.7% and in D by �16% re-lative to water discharging from vents, amounts con-sistent with fractionation during partial freezing ofsupercooled subglacial water in an open system, one inwhich the remaining unfrozen water is continuouslyremoved and replenished by water of similar composi-tion (O’Neil 1968; Arnason 1969; Souchez & Jouzel1984; Lehmann & Siegenthaler 1991).

Outlet glaciers of Vatnajokull and Oræfajokull

Vatnajokull (Fig. 1) is Iceland’s largest ice cap andcovers several hydrothermal systems as well as Ice-land’s largest active volcanoes. It is warm-based andincludes a number of southward- and eastward-drain-ing outlet glaciers that range in size from piedmontglaciers 410 km wide to valley glaciers o1 km wide.Along the southern perimeter of Vatnajokull is Or-æfajokull, a small warm-based ice cap that drapes

Hvannadalshnukur, the largest active volcano andhighest peak (2110m a.s.l.) in Iceland. Draining Or-æfajokull are also several outlet glaciers. We in-vestigated basal ice at 11 of the outlet glaciers ofVatnajokull and Oræfajokull (Fig. 1).

As a result of glacier–volcano interactions, subglaciallakes commonly develop beneath both Vatnajokull andOræfajokull. These lakes periodically drain catastro-phically through the outlet glaciers to producejokulhlaups that flood the proglacial landscape(Bjornsson 2002). Some outlet glaciers surge, but thereappears to be no correlation between their locationsand geothermal heat sources (Bjornsson et al. 2003).

Within their terminal areas, most outlet glaciers flowout of basins eroded into basaltic bedrock, or flow upsteep adverse grades associated with proximal slopes ofsandar. For example, radio-echo sounding data fromSkaftafellsjokull, Svınafellsjokull and Flaajokull revealoverdeepened basins at their terminal margins (Palssonet al. 1998; Palsson & Bjornsson 2000; Bjornsson et al.2003). Radio-echo sounding data from the lower partof Skeijararjokull show that it flows up a steep adversegrade at the terminus (Bjornsson et al. 1999, 2003), anddata from the lower part of Breijamerkurjokull showtwo small overdeepenings behind the glacier terminus(Nick et al. 2007: fig. 3). In addition, Kvıarjokull,Fjallsjokull, Skalafellsjokull, Heinabergsjokull andHoffellsjokull are currently retreating into basins(F. Palsson, pers. comm. 2003).

According to Flowers et al. (2003), subglacial melt-water at Vatnajokull is derived mainly from the meltingof surface snow and ice, and drains largely throughseasonally evolving conduit networks that discharge at

Fig. 1. Map showing the locations of the Vatnajokull and Or-æfajokull outlet glaciers examined in this study.

458 Grahame J. Larson et al. BOREAS

points (vents) along the glacier margins. These authorsalso argued that some of the meltwater is produced atthe glacier bed, and estimated that �5% of the meanannual discharge from Vatnajokull is geothermally de-rived. In addition, they estimated that up to �30% ofall meltwater in some parts of Vatnajokull may drainvia a groundwater system. Rainfall is also a significantcomponent of subglacial streams issuing from the gla-cier margins, as it makes up much of the annual pre-cipitation (1600 to 2000mm) falling on the glaciers.

Recently, Roberts et al. (2000, 2002) and Tweed et al.(2005) reported frazil-ice aggregates floating in streamsdraining Skeijararjokull, Skaftafellsjokull and Svına-fellsjokull, as well as growth of platy ice on suspendedtemperature probes and anchor-ice terraces arounddischarging vents. Temperature probes revealed waternear vents to be slightly supercooled, even when sum-mer air temperatures were well above the freezingpoint. Knudsen et al. (2001) also reported finding an-chor-ice terraces along the margin of Skaftafellsjokull,Kvıarjokull, Flaajokull and Hoffellsjokull and noted,as did Tweed et al. (2005), that bands of debris-ladenanchor ice in abandoned vents were texturally similarto stratified basal ice exposed along the glacier margins.Cook et al. (2007) found the physical properties of an-chor ice to be generally different from basal ice facies ofSvınafellsjokull; more recently, Cook et al. (2010) con-cluded that certain subfacies within what they recognizeas the stratified facies of Skaftafellsjokull and Svına-fellsjokull were formed subglacially as a result of gla-ciohydraulic supercooling.

Description of englacial ice, vent water, frazil/anchor ice and stratified basal ice

During the late winter of 2001–2005 and mid-summerof 2000 and 2004, we examined and sampled englacialice, vent water, frazil/anchor ice and stratified basal icefrom outlet glaciers draining Vatnajokull and Or-æfajokull. Clean vertical to near-vertical sections ofstratified basal ice were exposed at eight of the glaciersin late winter, when their terminal margins locallyramped up onto proglacial sediments and sublimationenhanced ice structures normally obscured by summermelting. We did not find stratified basal ice in ice-mar-ginal exposures of Falljokull, Breijamerkurjokull andFjallsjokull.

The englacial ice (the clean white ice above the basalzone) of the outlet glaciers is bubble-foliated at centi-metre to decimetre scales, with bubble-rich layers se-parating bubble-poor layers of clear ice. Ice crystals areseveral centimetres or more in diameter in terminalsections, depending on the glacier (range 2 to 10 cm).Englacial ice is generally free of debris, but with medialdebris septa and up-glacier-dipping transverse debrisbands sometimes present. Debris bands up to 2 cm wide

typically contain tephra derived from local volcaniceruptions, whereas thicker bands contain mostlycoarse-grained sands with occasional pebble-to-cobble-sized clasts. Bennett et al. (2000) and Swift et al. (2006)suggested that coarse-grained sandy debris was en-trained by tectonism, whereas Roberts et al. (2002)suggested that debris-laden, supercooled water welledup through ice hydro-fractures, producing frazil/an-chor ice that trapped the sandy debris within fractures.

The vent water issuing from the outlet glaciers con-tains considerable suspended sediment. Long-termstudies by Palsson & Vifgusson (1996) of suspended-sediment transport in many of the streams draining theoutlet glaciers recorded mean concentrations rangingfrom 78 to 6668mg�1, with concentrations exceeding11000mg/l during jokulhlaups.

During our investigation, we observed newly formedfrazil/anchor ice along the margins of Skeijararjokull,Skaftafellsjokull, Svınafellsjokull, Kviarjokull, Skala-fellsjokull, Flaajokull and Hoffellsjokull, both in thesummer and in late winter, regardless of air tempera-ture. The frazil ice appeared as ice-crystal aggregates(Osterkamp & Gilfillian 1975; Martin 1981; Daly 1984)adhering to vent and conduit walls, and as frazil ro-settes on rocks (Evenson et al. 1999). It also appeared asfrazil-ice aggregates (Evenson et al. 1999) floating inupwelling vent water. The anchor ice appeared as semi-circular terraces commonly 2 to 5m across around up-wellings (Fig. 2); the largest observed terrace was over10m across and 3m thick, at a vent feeding the SulaStream at the western end of Skeijararjokull. Debriscontent in the anchor ice was highly variable and dis-tribution varied from isolated clasts to a mixture ofclasts and fine-grained sediment in thin lenses and lay-ers that sometimes were crudely stratified.

Stratified basal ice was observed along the terminiof Flaajokull, Heinabergjokull, Hoffellsjokull, Kvıar-jokull, Skaftafellsjokull, Skeijarajokull, Skalafellsjo-kull and Svinafellsjokull and appeared similar incharacter from glacier to glacier. It varied from40.5 to

Fig. 2. Frazil rim around a �1-m-diameter upwelling at the marginof Skaftafellsjokull (March 2003).

BOREAS Stratified basal ice in outlet glaciers, Iceland 459

4m in thickness and was traceable for continuous dis-tances of up to 100m. Often, the stratified facies oc-curred near large conduits and vents and sometimesformed the walls of abandoned conduits exposed bywintertime uplift of the ice margin. In some locations,we observed ao10 to 30 cm thick clear-ice layer belowthe englacial ice (e.g. at Kvıarjokull, Skalafellsjokull,Hoffellsjokull). This horizon contained dispersed sandto granule-size debris and bubble clouds, with ice crys-tals 1 to 2 cm across and without internal layering. Asimilar clear-ice layer often existed where englacial icewas in direct contact with bedrock. We identify thishorizon as the basal dispersed facies of the basal zone(Lawson 1979).

The stratified basal ice includes generally bed-con-formable, discontinuous and broadly parallel layersand lenses of debris-rich iceo1 to 4 cm thick separatedby clear bubble-free iceo1 to 4 cm thick (Fig. 3). Theice is fine-grained, with grains ranging from 1 to 4mmacross, occasionally to 10mm across, whereas debris ismost commonly silty sand and sand-sized tephra. Deb-ris content ranges from �4 to 15.0% by volume, butoccasionally up to 44% by volume (Table 1). Individualdebris horizons occur in various modes (Lawson 1979)and include en echelon, thin (1 to 4mm), subparalleldiscontinuous planar lenses, subparallel o1 to 2 cmthick debris-rich layers that pinch out and merge lat-erally within less than a metre, and 1 to 4 cm thick pla-

nar clusters of silty sand aggregates 2 to 3 cm indiameter within clear, bubble-free ice. In addition, irre-gular to star-shaped 2 to 10mm diameter aggregates ofsilty sand may be uniformly distributed within clearbubble-free ice, and commonly located at the three- andfour-grain intersection of ice crystals. Clasts of pebble-to-boulder size are few, appear randomly distributed,and are dispersed throughout the stratified basal ice.However, clasts may be locally concentrated where thestratified basal ice is in contact with englacial ice (e.g.Skeijararjokull and Kvıarjokull; Fig. 4). These clastsare mostly subround to round, appearing to have beenshaped by fluvial action, and generally lack striations.

Sampling and isotopic analyses

A total of 63 samples of englacial ice (�300 g) were col-lected from near the glacier termini and up to severalkilometres upglacier along transects roughly parallel toice flow. The samples were taken by first removing a 3to 5 cm thick surface ice layer with an ice axe, and thenallowing the samples to melt in a sealed plastic bag. Atotal of 60 samples of upwelling vent water (250 g) werecollected from subglacial streams draining the glaciersusing polypropylene bottles. Samples of frazil/anchorice (�300 g) similar to that described by Evenson et al.(1999) for the Matanuska Glacier were collected fromwithin active and abandoned vents and conduits up-lifted at the glacier margins in winter and from anchor-ice terraces associated with vent openings. In the case ofthe uplifted active vents and conduits, it was possible toenter the larger ones during March, when meltwaterdischarge was low, and examine their complex geo-metry and the platy frazil ice lining channel walls. Atotal of 17 frazil/anchor-ice samples were collected fol-lowing the procedure used to sample englacial ice.

A total of 54 channel samples of stratified basal ice(�500g) were collected from exposed sections of the basalzone using a concrete saw (Table 1). Where the strati-fied basal ice was thick (1–4m), separation betweensamples was �0.5 m; where thin (o1m), it was�0.2 to 0.3m. Before sampling, we examined the basalzone in detail, identifying the facies and their varioussubfacies following Lawson (1979), and then restrictingsampling to the stratified facies. Recently, Cook et al.(2007) recognized several subfacies or types of basal icepresent at Svınafellsjokull, and we suspect that their‘sub-facies A and B’ are equivalent to our stratified fa-cies. We also avoided sampling ice of debris bands, suchas the transverse debris bands described by Swift et al.(2006) at Kvıarjokull, that may occur near the basal-iceexposures but are set within englacial ice and are not acomponent of the basal zone.

Seventeen englacial-ice samples and 24 stratified ba-sal-ice samples consisting of 250-ml aliquots were ana-lysed for 3H at the Michigan State University low-level

Fig. 3. Section of stratified basal ice exposed along the margin ofKvıarjokull, Iceland (March 2003). Note the concentration of gravel-size clasts at the upper contact and the apparent upglacier dip ofstratification. Ice flow right to left.

460 Grahame J. Larson et al. BOREAS

3H lab using a standard scintillation counting proce-dure following electrolytic enrichment of samples(Kessler 1988). Two vent-water samples and one frazil/anchor-ice sample were also analysed. In this in-vestigation, reported 3H activity is expressed in 3H units(TU), where one TU is equivalent to 7.2 dpm/l (disin-tegration per minute per litre), or one 3H atom perH�1018 atoms. The analytical detection limit of the labis 1 TU, and the precision is �1TU.

Approximately 10-ml aliquots from englacial-ice,vent-water, frazil/anchor-ice and stratified basal-icesamples were sent to MountainMass Laboratory,Colorado, for 18O and D analyses. The 18O/16O andD/H ratios were determined using the standardcalculation method: d(%)= [R(sample)�R(standard)/R(standard)]� 1000, where R is equal to the 18O/16O orD/H ratio for the experimental sample or laboratorystandard (e.g. SMOW). Reported analytical accuracy

from the lab is 0.1% and 1.0% for d18O and dD, re-spectively.

Differences in 3H between englacial-ice samples andstratified basal-ice samples were assessed by a t-test,with a set at 5%; all significance tests were performed inSYSTAT (version 12). Differences in d18O and dD amongvent-water, frazil/anchor-ice and stratified basal-icesamples were assessed by one-way analysis of variancefollowed by post hoc pair-wise comparisons assumingunequal variances (Tamhane’s T2). A Tamhane’s T2test adjustment was used as it is more conservative thanother adjustments (e.g. Dunnett’s T3). Homogeneity ofvariance was evaluated with Levene’s test, and nor-mality was evaluated by a normality plot and compa-nion statistics (Lilliefors, Shapiro–Wilk).

Isotopic characteristics of englacial ice, ventwater, frazil/anchor ice and stratified basal ice

Fig. 5A is a record of 3H in precipitation at Reykjavik(IAEA-WHO), located �180 km west of Vatnajokulland Oræfajokull. It shows that immediately followingthe beginning of atmospheric thermonuclear testingnear the end of 1952, 3H in precipitation was �10 to�30TU, but that it had climbed to almost 4000TU by1962. It also shows that since the enactment in 1962 ofthe international moratorium on atmospheric thermo-nuclear testing, 3H in precipitation has fallen; from1998 to 2004 the amount ranged from �10.0 to�2.0TU and averaged�4.9TU. A reconstruction of 3Hremaining in precipitation in 2004 based on a 3H half-life of 13.43 years (Clark & Fritz 1997) is also shown inFig. 5A. It reveals that any precipitation prior to thestart of atmospheric thermonuclear testing and

Table 1. Stratified basal-ice sections along margins of Vatnajokull and Oræfajokull outlet glaciers.

Outlet glacier (code) Section location Latitudelongitude

Length(m)

Thickness(m)

Debris content % by volume

Mean (no. samples) Range

Breidamerkurjokull (BM) - - - - - -

Falljokull (FA) - - - - - -

Fjallsjokull (FJ) - - - - - -

Flaajokull (FL) Terminal margin west end N6411904400

W1513304500100 o1–2 - -

Heinabergjokull (HN) Terminal margin southern end N6411702200

W1514003010 2–3 - -

Hoffellsjokull (HF) Terminal margin western end N6412502000

W151250050040 o1–3 15.0 (7) 3.0 to 44.0

Kviarjokull (KV) Terminal margin central part N6315603800

W161270410020 o1–3 9.1 (5) 7.7 to 12.0

Skaftafellsjokull (SKA) Terminal margin western end N6410102100

161560020020 2–4 11.5 (5) 5.3 to 16.4

Skeijararjokull (SKF) Terminal margin eastern end N6410100000

171060430010 o1–2 35.4 (1) -

Skalafellsjokull (SKE) Terminal margin eastern end N6411603400

W151400160060 o1–3 4.1 (7) 0.6 to 8.7

Svinafellsjokull (SV) Terminal margin southern end N6315903000

W161520200035 2–3 9.8 (6) 7.0 to 15.5

Fig. 4. Close-up photo of debris-rich and debris-poor horizons with-in the stratified facies of the basal zone, Kviarjokull (March 2003).

BOREAS Stratified basal ice in outlet glaciers, Iceland 461

preserved as snow or ice will have a 3H concentration ofo2TU, whereas that preserved after the start of testingwill have a concentration of Z2TU (Libby 1955; Le-venthal & Libby 1970).

Results of 3H analyses of the englacial-ice and strati-fied basal-ice samples are shown in Fig. 5B. With theexception of one sample with a value of 3.1�1TU,englacial-ice samples (Fig. 5B, upper panel) range from

Fig. 5. Isotopic composition of englacial ice, vent water, frazil/anchor ice and stratified basal ice from Vatnajokull and Oræfajokull outletglaciers. A. 3H in Reykjavik precipitation since 1953 (TU=tritium unit). Values from 1960 to 1966 and from 1992 to 2004 are from published3H analyses for Reykjavik precipitation (IAEA-WHO). Values from 1953 to 1959 are reconstructed using linear regression analysis of 3H datafor Reykjavik and Ottawa, Canada, and values from 1967 to 1991 are reconstructed using linear regression analysis of 3H data for Reykavikand Valentia, Ireland. Reconstruction of 3H remaining in the precipitation as of 2004 owing to radioactive decay based on a 3H half-life of 13.43years (Clark & Fritz 1997). B. 3H in englacial-ice and stratified basal-ice samples. TU values based on a detection limit of 1 TU and analyticalerror of �1.0 TU. Note: some TU values are negative owing to analytical error and spurious results at extremely low concentrations. d18Oand dD in englacial-ice (C), vent-water (D), frazil/anchor-ice (E) and stratified basal-ice (F) samples. Meteoric water line (MWL) based on d18Oand dD measured in Reykjavik precipitation since 1961. Most of the precipitation in Reykjavik is rainfall. BM=Breijamerkurjokull;FA=Falljokull; FL=Flaajokull; HN=Heinabergsjokull; HF=Hoffellsjokull; KV=Kviajokull; SKF=Skaftafellsjokull; SKA=Skalafellsjokull; SKE=Skeijararjokull; SV=Svinafellsjokull. Number of samples in parentheses.

462 Grahame J. Larson et al. BOREAS

�1.5 to 1.5�1TU with a mean of �0.1 TU�1.18(standard deviation). The exception is believed to be asample of superimposed ice mistaken for englacial ice.In contrast to the case for englacial-ice samples, 3H inthe stratified basal-ice samples (Fig. 5B, lower panel)ranges from 1.9 to 6.1�1TU with a mean of3.8 TU�1.14. The difference in means is 3.2 TU, andresults of a t-test indicate that there is a significant dif-ference in 3H between the englacial-ice samples and thestratified basal-ice samples (Po0.001). The two vent-water samples yielded 2.8 and 4.8�1TU, and the frazil-ice sample yielded 4.7�1TU.

The results of the d18O and dD analyses of the eng-lacial-ice, vent-water, frazil/anchor-ice and stratifiedbasal-ice samples are shown Figs 5C–F, respectively,together with d18O and dD for monthly Reykjavik pre-cipitation (IAEA-WHO) and the meteoric water line(MWL) for Reykjavik. For the englacial-ice samples,d18O ranges from �11.94 to �8.07% with a meanof �10.50�0.70%, and dD ranges from �90.60 to�60.14% with a mean of �76.32�5.46%. When thed18O and dD for the englacial-ice samples are plottedagainst each other they form a cluster of points (Fig.5C) that straddles the Reykavik MWL and overlapsd18O and dD points for winter precipitation. The meand18O and dD of the englacial-ice samples result in apoint (open cross) that falls on the MWL. For the vent-water samples, d18O ranges from �13.21 to �8.41%with a mean of �10.99�1.09%, and dD ranges from�96.15 to �63.93% with a mean of �78.95�7.25%.When the d18O and dD for the vent-water samples areplotted against each other (Fig. 5D), they also form acluster of points that straddles the MWL and overlapsthe cluster of points associated with the englacial-icesamples. The mean d18O and dD of the vent-watersamples result in a point (open cross) that falls on theMWL and near to that of the englacial-ice samples.

For the frazil/anchor-ice samples, d18O ranges from�11.25 to �6.66% with a mean of �9.05�1.16%, anddD ranges from �82.56 to �49.86% with a mean of�67.61�9.10%. When plotted against each other (Fig.5E), the d18O and dD for the frazil/anchor-ice samplesform a cluster of points that straddles the MWL. Thecluster is generally less negative than the cluster asso-ciated with the vent-water samples, and the mean d18Oand dD of the frazil/anchor-ice samples result in a point(open cross) that falls just below the MWL. The differ-ence between the mean d18O for the frazil/anchor-icesamples and vent-water samples is 1.94%, and the dif-ference between the mean dD values is 11.34%.

For the stratified basal-ice samples, d18O ranges from�10.19 to �7.05% with a mean of �8.58�0.75%, anddD ranges from �82.80 to �53.43% with a mean of�66.98�6.92%. When plotted against each other thed18O and dD for the stratified basal-ice samples form acluster of points that lies mostly below the MWL. Thecluster is also generally less negative than the cluster

associated with the vent-water samples, and the meand18O and dD of the stratified basal-ice samples result ina point (open cross) that falls just below theMWL, verynear to that of the frazil/anchor-ice samples. The dif-ference between the mean d18O for the stratified basal-ice samples and the vent-water samples is 2.41%, andthe difference between the mean dD values is 11.96%.

Analyses of variances indicate significant differencesamong vent-water, frazil/anchor-ice and stratified ba-sal-ice samples for d18O and dD (d18O: F3,391=160.6,Po0.001; dD: F3,391=2.67, Po0.001). Results frompair-wise comparisons for each d18O and dD indicatesignificant differences between the frazil/anchor-icesamples and vent-water samples, and stratified basal-icesamples and vent-water samples (Po0.001 for d18Oand dD), but not between frazil/anchor-ice samples andstratified basal-ice samples (P=0.135 for d18O andP=0.999 for dD).

Discussion

The englacial-ice 18O and D data for the outlet glaciersin Fig. 5 are remarkably similar, despite the fact thatthe accumulation area of each glacier has a differentelevation range that would affect the isotopic composi-tion of snow. The similarity can be explained in part bythe fact that the accumulation areas of the glaciers lieabove �1050m a.s.l. but below �1650m a.s.l. (Flowerset al. 2003: figs 1, 4). Based on a global d18O lapse rateof �0.28%/100m (Poage & Chamberlain 2001), the al-titude effect over a maximum elevation range of 600mwould result in only a �1.68% change in the 18O com-position of snow. Converting the d18O change to d18Dusing the expression for the global MWLdD=8d18O110 (Craig 1961) would result in only a�23.44% change in the D composition of snow. For theoutlet glaciers draining Oræfajokull, however, the alti-tude effect on the 18O and D composition would begreater than that for the outlet glaciers of Vatnajokull,as the accumulation areas of these glaciers lie above�1250m but below �2100m a.s.l. (Flowers et al. 2003:figs 1, 4). The altitude effect over a maximum elevationrange of 850m would result in a�2.38% change in the18O composition of snow and a 29.04% change in the Dcomposition. These changes in isotopic compositionare still relatively small and within the range of valuesrecorded for englacial ice.

In our investigation of the stratified basal ice of theeight Icelandic outlet glaciers, we noted similarities tothe stratified basal ice of the Matanuska Glacier (Law-son 1979; Lawson et al. 1998; Evenson et al. 1999; Lar-son et al. 2006). At both the outlet glaciers and theMatanuska the stratified facies of the basal zone isdecimetres to metres thick and composed of bed-con-formable, discontinuous, broadly sub-parallel layersand lenses of debris-rich ice and clear bubble-free ice

BOREAS Stratified basal ice in outlet glaciers, Iceland 463

that dip upglacier, and includes pebble-to-boulder sizeclasts randomly and widely distributed throughout.Also at both locations, a sharp contact commonly se-parates the stratified basal ice from the overlying eng-lacial ice.

We also observed some differences between the stra-tified basal ice of the outlet glaciers and that of theMatanuska. At the outlet glaciers, the lateral extent ofstratified basal-ice exposures is generallyo100m,whereas at the Matanuska exposures extend for hun-dreds of metres. Another difference is in the debris tex-ture and concentration. At the eight outlet glaciers thedebris in the stratified basal ice is mostly fine-to-med-ium sand with lesser amounts of silt and occasionalgravel. The debris concentration averages 4.1 to 15.0%by volume. At Matanuska the debris is mostly silt withminor sand and some gravel, and averages �25% byvolume (Lawson 1979). We attribute the difference indebris texture mainly to fundamental differences in thebedrock underlying the glaciers as a debris source,which in Iceland is mainly basalt and hyaloclastic ba-salt (Johannesson & Sæmunjsson 1999), whereas at theMatanuska Glacier it is mainly low-grade metamorphicgraywacke with lesser amounts of sedimentary volca-nics (Winkler 1992). We cannot, however, rule out thepossibility that subglacial processes such as hydro-dynamic sorting of subglacial sediments (e.g. Vivian1970) and grain fracture in deforming subglacial sedi-ments (Hooke & Iverson 1995; Iverson et al. 1996) mayalso have affected debris textures.

We do not know why the debris concentration inthe stratified basal ice of the outlet glaciers is lower thanthat of the Matanuska. We suspect that it is relatedto the grain-size distribution of the suspended sedimentload and the geometry of the subglacial drainagenetwork, thus affecting the locations where debris-trapping frazil-ice aggregates and anchor ice are accu-mulating.

Origin of the stratified basal ice

The presence of thermonuclear-derived 3H in the stra-tified basal ice of the outlet glaciers and its absence inthe englacial ice indicates that the stratified basal iceformed relatively recently (after 1952) and is youngerthan the overlying englacial ice. The presence of 3H alsoshows that the stratified basal ice, like the frazil/anchorice forming today, is derived from a mixture of waters,mainly englacial-ice melt and modern precipitation.Given that the mean 3H concentration in the stratifiedbasal-ice samples is only �3.8 TU, we suspect that thesource water for the ice is mainly melt from pre-1952englacial ice and firn, with a significant contributionfrom modern rain and snow melt. The extent to whichmodern water sources have contributed to the forma-tion of the stratified basal ice is, however, difficult to

estimate, as 3H in Icelandic precipitation has variedfrom 400TU in 1962 to 2–3TU in 2004 (corrected fordecay).

The fact that frazil/anchor ice is enriched on averageby �1.94% in 18O and by �11.34% in dD relative tovent water and has an average d18O and dD composi-tion significantly different from that of vent water(Po0.001 for d18O and dD) can be explained by frac-tionation associated with partial freezing in an opensystem (Souchez & Jouzel 1984). Basically, the freezingof supercooled water issuing from vents preferentiallypartitions 18O and D into the developing frazil/anchorice by an amount governed by the fractionation factorsbetween ice and water for 18O/16O and D/H. Despitethe partitioning, however, the 18O and D compositionof the vent water stays relatively constant because thesource water is rapidly and continuously being re-plenished by new supercooled water and little ice isbeing produced relative to the water flux. Any sub-sequent ice formed from supercooling will thereforehave the same 18O and D composition as the originalfrazil/anchor ice (Titus et al. 1999).

Based on experimental results, Lehmann & Sie-genthaler (1991) showed the equilibrium fractionationfactors between ice and water for 18O/16O and D/H tobe 1.00291 and 1.0212, respectively, and calculated thatin an open, well-mixed system where the freezing velo-city is small (o2mmh�1), ice formed will be enrichedrelative to water by �2.91% in 18O and by �21.2% inD. They also found that when freezing velocities arehigh (such as during frazil/anchor-ice growth), theamount of enrichment significantly decreases becausean isotopic gradient develops near the ice–water inter-face. This probably explains why the frazil/anchor ice isenriched on average by only �1.94% in 18O and by�11.34% in D, relative to the vent water. At Svına-fellsjokull and Skaftafellsjokull, Cook et al. (2010: table2) found frazil/anchor ice enriched relative to vent wa-ter by �1.7–2.8% in 18O and by �9.9–16.9% in D, andat the Matanuska Glacier, Lawson et al. (1998: fig. 12c)found frazil/anchor ice enriched relative to vent waterby �2.7% in 18O and by �16% in D.

Fractionation associated with partial freezing of su-percooled water in an open system (Souchez & Jouzel1984) can also account for the d18O and dD composi-tion of the stratified basal ice. The stratified basal ice isenriched on average by �2.41% in 18O and by�11.96% in D and has an average d18O and dD com-position significantly different from that of the ventwater (Po0.001 for d18O and dD). The d18O and dDcomposition of the stratified basal ice, however, is notsignificantly different from that of the frazil/anchor ice(P=0.135 for d18O and P=0.999 for dD) and pro-vides circumstantial evidence for the link between thetwo types of ice. The difference between the average 18Oand D enrichment of the stratified basal ice and that ofthe frazil/anchor ice is probably a result of different

464 Grahame J. Larson et al. BOREAS

freezing rates beneath the ice and at the ice margin.At Svınafellsjokull and Skaftafellsjokull, Cook et al.(2010: table 2) found their stratified subfacies enrichedrelative to vent water by �1.4–3.5% in 18O and by�8.4–24.3% in D, and at the Matanuska Glacier,Lawson et al. (1998: fig. 12b) found the stratified faciesof the basal-zone ice enriched relative to vent water by�2.7% in 18O and by �16% in D.

Other possible origins for the stratified basal ice

Regelation melting and refreezing around small (lengtho1m) bedrock hummocks can also produce a basal-icelayer in warm-based glaciers (Weertman 1957, 1964).Typically, a regelation layer appears as a sheet at theglacier sole less than a few centimetres thick (Nye 1970)and is composed of multiple closely spaced laminae ofregelation ice separated by thin debris-rich planes con-taining fine-grained debris abraded from the bedrockhummocks (Hubbard & Sharp 1993). Where such alayer has been observed forming, the debris concentra-tion is generally very low, a few percent too10% byvolume (Kamb & LaChapelle 1964).

As originally defined, regelation occurs in a closedsystem (Weertman 1957, 1964). As a result, regelationice will have a bulk 18O and D composition equivalentto that of the melting ice at the upstream side of thehummock. Moreover, the ice will be 3H-free if the icemelting is 3H-free. However, if regelation and refreez-ing is disrupted by the removal of some water into thesubglacial drainage system, the ice re-formed can beslightly enriched in 18O and D relative to the melting ice(Hubbard & Sharp 1993). Furthermore, some 3H couldenter the re-formed ice if melt associated with regela-tion is mixed with tritiated film water (Weertman 1972)at the glacier bed. Tritiated film water could also diffuseupwards into the regelation ice via intergranular lensesand veins (Nye & Frank 1973; Raymond & Harrison1975) when film-water pressure is high, and then freezewhen pressure falls.

We expect any mixing of regelation melt and tritiatedfilm water to be minimal because heat released fromregelation refreezing (Weertman 1957, 1964) would es-cape into the subglacial drainage system, effectivelyshutting down the regelation process. Furthermore,diffusions of tritiated film water into the glacier solewould probably result in a gradual upward transitionfrom tritiated ice to 3H-free ice. 3H profiles fromKviarjokull and Svinafellsjokull (Fig. 6), however,show a step change in 3H near the glacier bed, a patternconsistent with a glaciohydraulic supercooling originfor the stratified basal ice (Strasser et al. 1996; Lawsonet al. 1998).

Vertical regelation of englacial ice through sedimentat the glacier bed is another way in which a basal-icelayer can form in warm-based glaciers (Iverson 1993;

Iverson & Semmens 1995; Rempel 2008). Theoretically,this process can produce a sheet of basal ice a few cen-timetres to41m thick (Iverson 1993; Iverson & Sem-mens 1995). Because regelation through sedimentrequires sediment grains to be in mutual contact (Iver-son 1993; Iverson and Semmens 1995), the resulting icelayer will most likely have a very high debris content(probably460%).

In a closed system, ice formed at the lee of sedimentparticles during vertical regelation will have an 18O andD composition equivalent to that of the infiltrating iceand will be 3H-free provided that the infiltrating ice is3H-free (Jouzel & Souchez 1982). However, if regela-tion refreezing is incomplete or some drainage occurs,the ice formed will have a composition enriched in 18Oand D relative to the infiltrating ice (Iverson & Souchez1996). Moreover, if mixing occurs between regelationmelt and tritiated groundwater or film water, iceformed from regelation refreezing may contain some3H.

We doubt that the stratified basal ice at the outletglaciers formed from vertical regelation of englacial ice,as it is both enriched in 18O and D and contains bomb3H. These characteristics would require incompletefreezing or some melt drainage, as well as mixing of re-gelation melt with tritiated groundwater or film water.Furthermore, nowhere have we observed average deb-ris concentrations in the stratified basal ice 460% byvolume; concentrations typically are far below this, andthe ice lacks the evidence of tectonism that would berequired to explain entrainment of clast-supported ma-terials followed by dilution through mixing with glacierice (Knight 1989). It is possible, however, that shear,together with vertical extension of ice near the glaciermargin, could lead to reduced debris concentration ofthe stratified basal ice.

Fig. 6. 3H profiles across stratified basal ice/englacial ice transition atKviarjokull and Svinafellsjokull.

BOREAS Stratified basal ice in outlet glaciers, Iceland 465

Still another way to form basal ice in warm-basedglaciers is through net basal adfreezing of saturated se-diment or meltwater close to the glacier margin duringwinter (Weertman 1961; Hubbard & Sharp 1995). If netbasal adfreezing occurs within a closed system, the ba-sal ice formed will have a 18O and D compositionequivalent to that of the groundwater or meltwater, butif it occurs within an open system where not all wateravailable freezes, it will be enriched in 18O and D re-lative to the groundwater or meltwater (Souchez &Jouzel 1984). Furthermore, the basal ice will contain no3H if the groundwater or meltwater is 3H-free, but willcontain 3H if the groundwater or meltwater is tritiated.

The depth of winter ‘cold wave’ penetration into themargins of the outlet glaciers can be estimated using aheat-conduction model (Paterson 1994: p. 206) and asinusoidal forcing function derived from winter tem-peratures recorded for the period November to March2001–2006 at a meteorological station (94m a.s.l.)located just west of the terminus of Skaftafellsjokull(Fig. 1).

We use two approaches to estimate the depth of thewinter cold wave. In the first approach, we use Fourieranalysis to identify the dominant frequencies in the ob-served temperature data in order to calculate the pene-tration depth. In the second approach, a directnumerical solution based on the heat-conductionequation is used.

The one-dimensional heat-conduction equation isgiven by

@T

@t¼ a

@2T

@z2; ð1Þ

where z and t denote the depth and time, T denotes thetemperature and a the thermal diffusivity. For the fol-lowing sinusoidal variation in surface temperature,

Tð0; tÞ ¼ Ta þX1n¼1

TSn sin notþ jnð Þ; ð2Þ

where TSn is the amplitude corresponding to the nthharmonic, Ta is the average temperature in a given per-iod and o is the angular frequency of the periodic tem-perature fluctuations, Equation (1) has the followingsolution:

Tðz; tÞ ¼ Ta þX1n¼1

TSn exp �zffiffiffinp

=D� �

sin notþ jn � zffiffiffinp

=D� �

:

ð3Þ

In Equation (3),

Dn ¼ffiffiffiffiffiffi2ano

r¼ Dffiffiffi

np ð4Þ

is the depth of penetration of the nth harmonic, andD ¼ 2a=oð Þ1=2is the depth of penetration correspond-

ing to the first harmonic. Because the amplitude of theoscillations decreases with depth and larger frequenciesdamp out faster than smaller frequencies (according toEqn. 3), we used spectral analysis to identify the mostdominant frequencies in the data in order to calculatethe depth D. Temperature data for the six-year periodwas first de-trended (by removing the annual trend) toproduce a stationary time series for Fourier analysis.To present the results of the analysis in a way that pre-serves the variance of the original time series, the powerspectral density (PSD) was computed using a Tukey–Hanning window (Press et al. 2007) in MATLAB (TheMathworks Inc., Natick, MA 2008). The PSD (Fig. 7)shows the main spectral peaks in the data, their relativemagnitudes, and the frequencies corresponding to thevarious peaks. The diurnal frequency (one per day) isthe most dominant signal in the data, followed by asemi-diurnal signal and other high-frequency compo-nents (e.g. with 8- and 6-h periods). Using a thermaldiffusivity of 1.1�10�6m2/s for ice (Hooke et al. 1983),the penetration depth corresponding to the diurnal sig-nal (i.e. o=2p/86400=7.272�10�5 s�1) is estimatedto be only 0.18m, which is insufficient to ad-freeze 2 to3m of sediment and water to beneath the margins of theoutlet glaciers. Semi-diurnal and other higher-fre-quency components produce even smaller depths (e.g.D2=0.12m).

To provide an independent check on the values of Destimated using Fourier analysis, we also solved theheat-conduction equation numerically using the ob-served temperatures as a boundary condition at thesurface (z=0), assuming that the temperature im-mediately above the glacier termini is reflected by tem-peratures recorded at the meteorological station. Weassumed that ice in the outlet glaciers was at 01C;therefore, if surface temperatures were greater thanzero, a value of 01C was used for the boundary condi-tion at z=0. A zero-gradient boundary condition (@T/@z=0) was used at the bottom boundary at z=10m.This boundary condition had little influence on thecomputed temperature profiles near the surface, as thebottom boundary was located sufficiently far from thesurface. Equation (1) was solved using an implicit fi-nite-difference scheme with a vertical resolution of 5 cmand a timestep of 1 h. The resulting system of tri-diag-onal matrix equations was solved using the Thomas al-gorithm (Press et al. 2007). The model was first testedby comparing the numerical solution with the analy-tical solution (Eqn. 3) and an excellent agreement wasobtained (not shown here). Figure 8 shows the simu-lated vertical temperature profiles for the period 1stNovember 2005 to 31st March 2006. Although manysimplifying assumptions are involved in the applicationof this model (e.g. heat generated by refreezing of meltwater was ignored, as were complexities associated withthe air–ice interface, including ablation or accumula-tion and variable thermo-physical properties), results

466 Grahame J. Larson et al. BOREAS

(shown here only for the top 3m for clarity) indicatethat major changes are confined to within 0.5m ofthe surface. Furthermore, we have not observedany sediment with very low ice content in blockswithin the stratified basal ice of the outlet glaciers thatwould be suggestive of freezing en mass of subglacialsediment.

Glaciological and geological implications

The limited lateral extent of the stratified basal iceof the outlet glaciers in comparison to that of theMatanuska Glacier may be explained in many ways.We consider two hypotheses to be more likely thanothers.

Fig. 7. Results of spectral analysis appliedto the time series of air temperature recordedat Skaftafell.

BOREAS Stratified basal ice in outlet glaciers, Iceland 467

First, as argued by Alley et al. (1998), if the angle ofthe adverse slope of an overdeepening has a magnitudemore than a critical value that falls between �1.2 to 1.7times the surface slope, glaciohydraulic supercoolingcan disrupt the efficiency of subglacial channelizeddrainage systems, increasing the hydraulic pressure andleading to a distributed subglacial drainage systemfrom which sediment is entrained to the base of theglacier. The rate of accretion increases with the magni-tude of the bed-slope to surface-slope ratio above thecritical value. However, geological processes are ex-pected to limit the beds of most overdeepenings to an-gles close to this critical value, with climatically inducedreduction of the ice-surface slope over the over-deepening required to cause the larger ratios that pro-duce extensive supercooling as at Matanuska Glacier(Lawson et al. 1998; Alley et al. 2003). At the Icelandicoutlet glaciers, however, we suspect that the magnitudeof the slope ratio is closer to the critical threshold thanat the Matanuska, producing less freeze-on and fewerof the associated phenomena than at Matanuska. Thismay explain why we found most of the stratified basal-ice exposures around subglacial vents in Iceland, incontrast to Matanuska Glacier, where the stratifiedbasal ice can occur without association to a subglacialpoint of discharge.

Another possibility to help explain the paucity of ex-posures of the stratified basal ice is that such ice is pre-sent in Iceland but is not well exposed. At MatanuskaGlacier, many of our observations were made on basalice exposed by strong wintertime advances that pro-duced extensive up-thrusting, including local overhangsthat allowed direct observation of the sole of the gla-cier. Such behaviour was not prominent on the Icelan-dic glaciers during the years of our study.

Whatever the reason for the paucity of exposures, thestratified basal ice at the outlet glaciers does not appearto have much influence on landform development,whereas it does at the Matanuska (Lawson 1979, 1981,

1982; Larson et al. 2006). Where the ice is exposed tosun and air, the debris released by ablation accumulateson the ice surface, but does not build to a thick, in-sulating cover except where occasionally silt is a majorcomponent of the debris (e.g. Hoffellsjokull). In addi-tion, because of its sandy texture, lack of cohesion andbetter drainage characteristics, it is not subject to re-distribution by sediment flows as at the Matanuska.Our limited observations during the melt season sug-gest that the accumulated debris disaggregates duringhigh rates of melt or rainfall and is washed into ice-marginal streams, leaving behind a lag of coarse mate-rial. The exposed basal zone recedes without muchburial of ice, as is the case at the Matanuska terminus.Furthermore, where basal ice is preserved, the lowerdebris content and coarse texture of the debris in thestratified basal ice are not favourable to melt-out tillformation, but rather to reworking to produce a depositlacking structure. Thick sedimentary sequences com-posed of multiple re-sedimented diamictons are also notlikely to develop from release of the basal debris alongthe steeper ice faces, and thus only a narrow band alongthe ice margin is characterized by relatively few areas ofactive reworking of the deposited basal debris.

Conclusion

The results of this investigation extend observations onthe connection between glaciohydraulic supercoolingand the stratified facies of the basal zone to temperateterrestrial glaciers in a maritime environment. We ob-served anchor-ice terraces developing from supercooledwater issuing from subglacial vents along terminalmargins of seven outlet glaciers draining Vatnajokulland Oræfajokull throughout the melt season. Withinthe glacier’s basal zones, we found that stratified basal icealso occurs in eight outlet glaciers, particularly near vents.In each instance, the glaciers flowed up adverse slopes.

Both the anchor-ice terraces and the stratified basalice of the Icelandic glaciers have physical characteristicssimilar to those of anchor-ice terraces and stratifiedbasal ice described at the Matanuska Glacier, Alaska.Isotopic analyses of samples of the stratified basal ice ofthe outlet glaciers show that it contains thermonuclear-derived 3H and formed relatively recently. The analysesalso show that it is enriched in 18O by �2.4% and in Dby �12% relative to subglacial water emerging fromvents, characteristics consistent with fractionation dur-ing partial freezing of supercooled water continuouslyflowing through an isotopically open subglacial drai-nage system.

Acknowledgements. – We thank the National Science FoundationOffice for funding of our research on glaciohydraulic supercooling.We also wish to acknowledge the US Army Cold Regions Researchand Engineering Laboratory (CRREL) for support, including

Fig. 8. Simulated vertical profiles of temperature for the period 1stNovember 2005 to 31st March 2006.

468 Grahame J. Larson et al. BOREAS

funding for the isotopic analyses. Special thanks go to Teryn Ebert,Sarah Kopczynski and Michiel Kramer for their assistance in thefield, and to Helgi Bjornsson and Finnur Palsson for informative dis-cussions about Icelandic glacial systems and their radar profiling dataof the outlet glaciers. We are particularly grateful to Ragnar FrankKristjansson of the Skaftafell National Park for housing and logis-tical support, and for lively discussions of Icelandic historical ob-servations of glacial activity. We thank Peggy Ostrom and the Centerfor Statistical Training and Consulting at Michigan State Universityfor their help with statistical analyses. Finally, we thank Neal Iversonand an anonymous reviewer for their constructive reviews of this ar-ticle, Simon J. Cook for his helpful comments on an earlier version ofthe article, and Jan A. Piotrowski for editorial comments.

References

Alley, R. B., Lawson, D. E., Evenson, E. B., Strasser, J. C. & Larson,G. J. 1998: Glaciohydraulic supercooling: A freeze–on mechanismto create stratified, debris–rich basal ice: II. Theory. Journal ofGlaciology 44, 563–569.

Alley, R. B., Lawson, D. E., Larson, G. J., Evenson, E. B. & Baker,G. S. 2003: Stabilizing feedbacks in glacier-bed erosion.Nature 424,258–260.

Arcone, S. A., Lawson, D. E. & Delaney, A. J. 1995: Short-pulse ra-dar wavelet recovery and resolution of dielectric contrasts withinenglacial and basal ice of the Matanuska Glacier, Alaska. U.S.A.Journal of Glaciology 41, 68–86.

Arnason, B. 1969: Equilibrium constant for the fractionation of deu-terium between ice and water. Journal of Physical Chemistry 7,3491–3494.

Bennett, M. R., Huddart, D. & Waller, R. I. 2000: Glaciofluvial cre-vasse and conduit fills as indicators of supraglacial dewateringduring a surge, Skeijararjokull, Iceland. Journal of Glaciology 46,25–34.

Bjornsson, H. 2002: Subglacial lakes and jokulhlaups in Iceland.Global and Planetary Change 35, 255–271.

Bjornsson, H., Palsson, F. & Magnusson, E. 1999: Skeidararjokull:Landslag og rennslisleidir vatns undir spordi. RaunvısindastofnunHaskolans RH-11-99, 20 pp.

Bjornsson, H., Palsson, F., Singurjsson, O. & Flowers, E. 2003: Sur-ges of glaciers in Iceland. Annals of Glaciology 36, 82–90.

Clark, I. & Fritz, P. 1997: Environmental Isotopes in Hydrogeology.328 pp. Lewis, New York.

Cook, S. J., Knight, P. G., Waller, R. I., Robinson, Z. P. & Adam,W.G. 2007: The geography of basal ice and its relationship to glacio-hydraulic supercooling: Svınafellsjokull, southeast Iceland. Qua-ternary Science Reviews 26, 2309–2315.

Cook, S. J., Robinson, Z. P., Fairchild, I. J., Knight, P. G., Waller, R.I. & Boomer, I. 2010: Role of glaciohydraulic supercooling in theformation of stratified facies basal ice: Svınafellsjokull and Skafta-fellsjokull, southeast Iceland. Boreas, 39, 24–38.

Craig, H. 1961: Isotopic variations in meteoric waters. Science 133,1702–1703.

Daly, S. F. 1984: Frazil ice dynamics. US.Army Corps of EngineersCold Regions Research and Engineering Laboratory, CRRELMonograph 84-1, 46 pp.

Evenson, E. B., Lawson, D. E., Strasser, J. C., Larson, G. J., Alley, R.B., Ensminger, S. L. & Stevenson, W. E. 1999: Field evidence forthe recognition of glaciohydraulic supercooling. In Mickelson, D.M. & Attig, J. W. (eds.): Glacial Processes Past and Present:Boulder Colorado, 23–35. Geological Society of America, SpecialPaper 337.

Flowers, G. E., Bjornsson, H. & Palsson, F. 2003: New insights intothe subglacial and periglacial hydrology of Vatnajokull, Iceland,from a distributed physical model. Journal of Glaciology 49,257–270.

Hooke, R. LeB. 1991: Positive feedbacks associated with erosion ofglacial cirques and overdeepenings. Geological Society of America,Bulletin 103, 1104–1108.

Hooke, R. LeB. & Iverson, N. R. 1995: Grain-size distribution in de-forming subglacial tills, role of grain fracture. Geology 23, 57–60.

Hooke, R. LeB. & Pohjola, V. A. 1994: Hydrology of a segment of aglacier situated in an overdeepening, Storglaciaren, Sweden. Jour-nal of Glaciology 40, 140–148.

Hooke, R. LeB., Gould, J. E. & Brzozowski, J. 1983: Near-surfacetemperatures near and below the equilibrium line on polar andsubpolar glaciers.Zeitschrift fur Gletscherkunde und Glazialgeologie19, 1–25.

Hubbard, B. & Sharp, M. 1993: Weertman regelation, multiple re-freezing events and isotopic evolution of the basal ice layer. Journalof Glaciology 39, 275–291.

Hubbard, B. & Sharp, M. 1995: Basal ice facies and their formationand deformation. Arctic and Alpine Research 27, 301–310.

Iverson, N. R. 1993: Regelation of ice through debris at glacier beds:Implications for sediment transport. Geology 21, 559–562.

Iverson, N. R. & Semmens, D. J. 1995: Intrusion of ice into porousmedia by regelation: A mechanism of sediment entrainment byglaciers. Journal of Geophysical Research 100, 10219–10230.

Iverson, N. R. & Souchez, R. 1996: Isotopic signature of debris-richice formed by regelation into a subglacial sediment bed. Geophysi-cal Research Letters 23, 1151–1154.

Johannesson, H. & Sæmunjsson, K. 1999: Island Jarjfræjikort (Ice-land Geological Map), Natturufræjistofnun Islands (Icelandic In-stitute of Natural History) scale 1:1,000.000. IcelandMinistry of theEnvironment, Reykjavik, Iceland.

Jouzel, J. & Souchez, R. A. 1982: Melting–refreezing at the glaciersole and the isotopic composition of the ice. Journal of Glaciology28, 35–42.

Kamb, B. & LaChapelle, E. 1964: Direct observation of the mechan-ism of glacier sliding over bedrock. Journal of Glaciology 5,159–172.

Kessler, M. J. 1988: Effective use of low level liquid scintillation ana-lysis. In Proceedings of the 2nd International Seminar for LiquidScintillation Analysis, 256–301. Packard Japan KK, Tokyo.

Knight, P. G. 1989: Stacking of basal debris layers without bulkfreezing on: isotopic evidence from west Greenland. Journal ofGlaciology 35, 214–216.

Knudsen, O., Roberts, M. J., Tweed, F. S., Russell, A. J., Lawson, D.E., Larson, G. J., Evenson, E. B. & Bjornsson, H. 2001: Five ‘‘su-percool’’ Icelandic glaciers. EOS, Transactions of the AmericanGeophysical Union, Fall Meeting Supplement 82, p. 557.

Larson, G. J., Daniel, E., Lawson, D. E., Evenson, E. B., Alley, R. B.,Knudsen, O., Lachniet, M. S. & Goetz, S. L. 2006: Glaciohydraulicsupercooling in former ice sheets? Geomorphology 75, 20–32.

Lawson, D. E. 1979: Sedimentological analysis of the western termi-nus region of the Matanuska Glacier, Alaska. US Army Cold Re-gions Research and Engineering Laboratory, CRREL Report 79-9,112 pp.

Lawson, D. E. 1981: Distinguishing characteristics of diamictons atthe margin of the Matanuska Glacier, Alaska. Annals of Glaciology2, 78–84.

Lawson, D. E. 1982: Mobilization, movement and deposition of ac-tive subaerial sediment flows, Matanuska Glacier, Alaska. Journalof Geology 90, 279–300.

Lawson, D. E. 1995: Sedimentary and hydrologic processes withinterrestrial valley glaciers. In Menzies, J. (ed.): Modern Glacial En-vironments: Processes, Dynamics and Sediments, 337–363. Butter-worth-Heinemann, Boston.

Lawson, D. E., Strasser, J. C., Evenson, E. B., Alley, R. B., Larson,G. J. & Arcone, S. A. 1998: Glaciohydraulic supercooling: a free-ze–on mechanism to create stratified, debris–rich basal ice. I. Fieldevidence. Journal of Glaciology 44, 547–561.

Lehmann, M. & Siegenthaler, U. 1991: Equilibrium oxygen- and hy-drogen-isotope fractionation between ice and water. Journal ofGlaciology 37, 23–26.

Leventhal, J. S. & Libby, W. F. 1970: Tritium fallout in the PacificUnited States. Journal of Geophysical Research 75, 7628–7633.

Libby, W. F. 1955: Tritium in nature. Journal of the WashingtonAcademy of Sciences 45, 301–314.

Martin, S. 1981: Frazil ice in rivers and oceans. Annual Review ofFluid Mechanics 13, 379–397.

Nick, F. M., van der Kwast, J. & Oerlemans, J. 2007: Simulation ofthe evolution of Breidamerkurjokull in the late Holocene. Journalof Geophysical Research 112, B01103, doi:10.1029/2006JB004358.

BOREAS Stratified basal ice in outlet glaciers, Iceland 469

Nye, J. F. 1970: Glacier sliding without cavitation in a linear viscousapproximation. Proceedings of the Royal Society of London, SeriesA 315, 381–403.

Nye, J. F. & Frank, F. C. 1973: Hydrology of the intergranular veinsin a temperate glacier. In Stelczer, K. (ed.): Symposium on the Hy-drology of Glaciers, Cambridge, 7–13 September 1969, Interna-tional Association Hydrological Sciences Publication 95, 157–161.

O’Neil, J. R. 1968: Hydrogen and oxygen isotope fractionationbetween ice and water. Journal of Physical Chemistry 72,3683–3684.

Osterkamp, T. E. & Gilfillian, R. E. 1975: Nucleation characteristicsof stream water and frazil ice nucleation.Water Resources Research11, 926–928.

Palsson, F. & Bjornsson, H. 2000: Vatnsrennsli undan eystri hlutaFlaajokull. Raunvısindastofnun Haskolans RH-14-00, 22 pp.

Palsson, S. & Vifgusson, G. H. 1996: Gagnasafn aurburjarmælinga1963–1995. Orkustofnun Report OS-96032/VOD-05 B, 270 pp.

Palsson, F., Bjornsson, H.,Magnusson, E. & Eydal, G. 1998: Konnunrennslisleida vatns ur Skramuloni, undir sporj Svınafellsjokull.Raunvısindastofnun Haskolans RH-08-98, 4 pp.

Paterson, W. S. B. 1994: The Physics of Glaciers. 480 pp. Pergamon,Oxford.

Poage, M. A. & Chamberlain, C. P. 2001: Empirical relationshipsbetween elevation and the stable isotope composition of precipita-tion and surface waters; considerations for studies of paleoeleva-tion change. American Journal of Science 301, 1–15.

Press, W. H., Teukolsky, S. A., Vetterling, W. T. & Flannery, B. P.2007: Numerical Recipes in C11: The Art of Scientific Computing.1265 pp. Cambridge University Press, New York.

Raymond, C. F. & Harrison, W. D. 1975: Some observations on thebehavior of the liquid and gas phases in temperate glacier ice.Journal of Glaciology 14, 213–233.

Rempel, A. W. 2008: A theory for ice–till interactions and sedimententrainment beneath glaciers. Journal of Geophysical Research -Earth Surface 113, F01013, doi:10.1029/2007JF000870.

Roberts, M. J., Russell, A. J., Tweed, F. S. & Knudsen, O. 2000: Icefracturing during jokulhaups: implications for englacial floodwaterrouting and outlet development. Earth Surface Processes andLandforms 25, 1429–1446.

Roberts, M. J., Tweed, F. S., Russell, A. J., Knudsen, O., Lawson, D.E., Larson, G. J., Evenson, E. B. & Bjornsson, H. 2002: Glaciohy-draulic supercooling in Iceland. Geology 30, 439–442.

Rothlisberger, H. 1968: Erosive processes which are likely to accent-uate or reduce the bottom relief of valley glaciers. International As-sociation of Hydrological Sciences Publication 7, 87–97.

Rothlisberger, H. & Lang, H. 1987: Glacial hydrology. InGurnell, A.M. & Clark, M. J. (eds.): Glacio-fluvial Sediment Transfer: An Al-pine Perspective, 37–44. John Wiley, Chichester.

Shreve, R. L. 1972: Movement of water in glaciers. Journal of Gla-ciology 11, 205–214.

Souchez, R. A. & Jouzel, J. 1984: On the isotopic composition in dDand d18O of water and ice during freezing. Journal of Glaciology 30,369–372.

Strasser, J. C., Lawson, D. E., Larson, G. J., Evenson, E. D. & Alley,R. B. 1996: Preliminary results of tritium analysis of basal ice, Ma-tanuska Glacier, Alaska, U.S.A.: evidence for subglacial accretion.Annals of Glaciology 22, 126–133.

Swift, D. A., Evans, D. J. A. & Fallick, A. E. 2006: Transverse eng-lacial debris-rich ice bands at Kviarjokull, southeast Iceland. Qua-ternary Science Reviews 25, 1708–1718.

Titus, D. D., Larson, G. J., Strasser, J. C., Lawson, D. E., Evenson,E. B. & Alley, R. B. 1999: Isotopic composition of vent dischargefrom the Matanuska Glacier, Alaska: Implications for the origin ofbasal ice. In Mickelson, D. M. & Attig, J. W. (eds.): Glacial Pro-cesses Past and Present: Boulder Colorado, 37–44. Geological So-ciety of America, Special Paper 337.

Tweed, F. S., Roberts, M. J. & Russell, A. J. 2005: Hydrologic mon-itoring of supercooled meltwater from Icelandic glaciers. Qua-ternary Science Reviews 24, 2308–2318.

Vivian, R. 1970: Hydrologie et erosion sous-glaciaires. Revue de Geo-graphie Alpine 58, 241–264.

Weertman, J. 1957: On the sliding of glaciers. Journal of Glaciology 3,33–38.

Weertman, J. 1961: Mechanism for the formation of inner morainesfound near the edge of cold ice caps and ice sheets. Journal of Gla-ciology 3, 965–978.

Weertman, J. 1964: The theory of glacier sliding. Journal of Glaciol-ogy 5, 287–303.

Weertman, J. 1972: General theory of water flow at the base of a gla-cier or ice sheet. Reviews of Geophysics and Space Physics 10,287–333.

Winkler, G. R. 1992: Geologic map and summary geochronology ofthe Anchorage 11�31 Quadrangle, southern Alaska. U.S. Geologi-cal Survey, Miscellaneous Series Map I-2283, scale 1:250,000.

470 Grahame J. Larson et al. BOREAS