Introduction
A wide range of processes can result in the entrainment of sediment into ice masses. Supraglacial material may fall onto the ice surface and be carried with the local flow vector, leading to burial in the accumulation area and exhumation in the ablation area. Large amounts of debris can also be entrained at the glacier margins and bed. This typically occurs through a debris-rich basal ice layer, which is formed by water freezing in the presence of debris near the basal interface (e.g. Reference Hubbard and Sharp.Hubbard and Sharp, 1989; Reference Alley, Cuffey, Evenson, Strasser, Lawson and Larson.Alley and others, 1997; Reference KnightKnight, 1997). Although the mechanism remains unproven, it has been argued that basal sediments may also be incorporated into the body of a glacier through folding and thrusting (Reference GoldthwaitGoldthwait, 1951; Reference Tison, Souchez and LorrainTison and others, 1989; Reference Hambrey, Bennett, Dowdeswell, Glasser and Huddart.Hambrey and others, 1999), and as basal crevasse-fills (Reference Mickelson and Berkson.Mickelson and Berkson, 1974; Reference SharpSharp, 1985; Reference Bennett, Hambrey, Huddart and Ghienne.Bennett and others, 1996; Reference Evans and Rea.Evans and Rea, 1999; Reference Ensminger, Alley, Evenson, Lawson and Larson.Ensminger and others, 2001; Reference Woodward, Murray and McCaigWoodward and others, 2002).
Extensive linear ridges of sediment-rich ice, commonly capped with a layer of melted-out debris, have recently been reported in the ablation area of numerous polythermal glaciers, particularly in Svalbard. These ridges are generally aligned either parallel to the ice-flow direction or perpendicular to it, and are respectively referred to hereafter as longitudinal and transverse supraglacial moraine ridges. These features are described in more detail below.
Longitudinal supraglacial moraine ridges
Longitudinal supraglacial moraine ridges, or medial moraines, located at the tongues of several composite Svalbard glaciers, have recently been interpreted by Reference Hambrey, Bennett, Dowdeswell, Glasser and Huddart.Hambrey and others (1999) and Reference Hambrey and Glasser.Hambrey and Glasser (2003) as products of folding along flow-parallel axes. The moraines are aligned parallel to longitudinal foliation and, with this structure, form an axial-planar relationship with the folding which develops in response to lateral compression as multiple flow units converge into a narrow tongue (Fig. 1). These authors reported that the debris-charged ridges associated with this folding emerge at the glacier surface near its terminus as fold hinges that plunge gently up-glacier.
Schematic illustration of the formation of longitudinal supraglacial moraine ridges at Svalbard glaciers (after Reference Hambrey, Bennett, Dowdeswell, Glasser and Huddart.Hambrey and others, 1999).
Reference Hambrey and Glasser.Hambrey and Glasser (2003) reported that the debris forming supraglacial moraine ridges at vestre Lovtjnbreen and austre Brøggerbreen can be either coarse and angular, indicating a supraglacial origin, or (more commonly) composed of diamicton containing subrounded, faceted and striated clasts, indicating a basal origin. Orientation measurements by these authors indicated that these basally derived clasts are strongly aligned parallel to the local longitudinal foliation and associated fold axis. In such cases, it was inferred by Reference Hambrey, Bennett, Dowdeswell, Glasser and Huddart.Hambrey and others (1999) that the source material (subglacial sediment or debris-rich basal ice) is folded along flow-parallel axes directly into the body of the glacier and reaches the surface as a consequence of this folding and subsequent ablation (Fig. 1).
Transverse supraglacial moraine ridges
Reference Glasser, Hambrey, Crawford, Bennett and Hud-dartGlasser and others (1998) associated transverse supraglacial moraine ridges on the glaciers of Kongsfjorden with structures interpreted as thrusts. As with longitudinal moraine ridges, the visual appearance of this debris is consistent with a basal origin, but few quantitative data support this inference.
Reference Woodward, Murray and McCaigWoodward and others (2002) recently presented an alternative hypothesis for the origin of the debris bands associated with transverse supraglacial moraine ridges at Kongsvegen. They argued that the features originated from debris-filled basal crevasses that have been exposed at the glacier surface by ablation of the overlying ice. However, in common with the thrust hypothesis outlined above, no new data relating to the physical properties of the materials concerned were reported.
An origin for debris bands by basal crevasse filling, however, has been advanced on the basis of research at other glaciers, from where data have been reported. Reference Ensminger, Alley, Evenson, Lawson and Larson.Ensminger and others (2001), for example, describe a finely layered (mm thick) sequence of debris-rich and debris-poor bands interpreted to have originated as basal crevasse fills at Matanuska Glacier, Alaska, U.S.A. In this case, the layers were associated with the injection of sediment-rich water into basal crevasses, where it froze in part as a result of supercooling upon emergence from a basal overdeepening. These authors also found that the debris within these basal crevasse fills was well sorted, silt-rich and fined with distance above the bed — consistent with grain settling through a viscous suspension during formation. Not all such debris injections, however, are considered to have involved sediment-laden water. Reference SharpSharp (1985), for example, argued that crevasse-squeeze ridges can form as a result of viscous soft-sediment injection into basal crevasses that open during surging. In such cases, the resulting features are more likely to include massive and poorly sorted debris, containing coarse clasts, than those associated with the less viscous flow described by Reference Ensminger, Alley, Evenson, Lawson and Larson.Ensminger and others (2001). However, these two cases most probably represent near end-members of a range of forms whose degree of sorting and layering corresponds to the extent of fluidization experienced during formation. Reference SharpSharp (1985) and Reference Evans and Rea.Evans and Rea (1999) also point out that the character of crevasse fills may be altered after formation by, for example, ice-deformation depositional processes.
Processes of incorporation and transport may also be investigated through analysis of the fabric of the clasts present within debris bands. For example, a spatially variable, but locally strong, fabric might be anticipated within a fluidized crevasse fill, similar to a viscous debris flow (Reference LawsonLawson, 1979a), whereas a more extensive and consistent fabric orientation might be anticipated within a thrust or thin shear zone.
Research approach
In this study we investigate the physical properties, particularly the sedimentology and isotopic composition, of debris bands associated with longitudinal and transverse supragla-cial moraine ridges at three Kongsfjorden glaciers: Kongsvegen, midre Lovenbreen and austre Brøggerbreen. In addition to characterizing these ridges, our aim is to use their physical properties to evaluate and refine theories of their formation.
Field Site and Methods
Fieldwork was undertaken at three glaciers, austre Brog-gerbreen, midre Lovenbreen and Kongsvegen, located on the Brogger peninsula (Broggerhalvoya), Kongsfjorden, in northwest Spitsbergen, Svalbard (Fig. 2). All three glaciers have been the subject of a long-term mass-balance monitoring programme by the Norsk Polarinstitutt (e.g. Reference Hagen and Sætrang.Hagen and others, 1991b), revealing a general thinning and recession (Liestol, 1988). Kongsvegen is a surge-type glacier which last advanced in 1948. Since then it has ceased to be active and, in contrast to fast-flowing (~700 m a) Krone-breen with which it shares a common tidewater terminus, it has a maximum surface velocity of ~8 ma-1 (Reference Hagen and Sætrang.Hagen and others, 1991b).
Kongsfjorden with sample sites numbered: (1) Kongsvegen transverse supraglacial moraine ridge and nearby glacier margin; (2−5) midre Love¤nbreen east margin (2), proglacial area (3), longitudinal supraglacial moraine ridge (4) and west margin (5); and (6) austre Br_ggerbreen longitudinal supraglacial moraine ridge.
Austre Brøggerbreen and midre Lovenbreen are also slow-moving glaciers, with equilibrium-line surface velocities of only a few metres per year (Reference BjörnssonBjörnsson and others, 1996). Close to their neoglacial maxima around 1890 they had vertical terminal cliffs and were considered by Liestol (1988) to be surging, an inference that is disputed by Reference Hambrey and Glasser.Hambrey and Glasser (2003), who cite the equivocal nature of Liestol’s (1988) evidence. However, proglacial geo-morphic evidence indicates that these glaciers were more dynamic at that time (Reference Glasser and Hambrey.Glasser and Hambrey, 2001). Their continual recession since 1890 is a reflection of climate warming through the 20th century. All three glaciers have substantial parts of their bed at sub-freezing temperatures, particularly austre Brøggerbreen where no temperate en-glacial ice is evident on radar images (Reference Macheret and Zhiravlev.Macheret and Zhuravlev, 1982; Reference Hagen and Sætrang.Hagen and Sstrang, 1991), and borehole temperatures indicate that most of the base is cold (Reference Hagen and Sætrang.Hagen and Sstrang, 1991). Midre Lovenbreen has a warm-based interior and a frozen terminus region, while tidewater Kongsvegen is wet-based throughout. The sample sites referred to below are summarized inTable 1, marked in Figure 2 and illustrated in Figures 3-5.
Summary of the features observed and sampled at the three glaciers studied. 
indicates that the feature was observed; S indicates that the feature was sampledfor its sedi-mentological characteristics; I indicates that the feature was sampledfor its isotopic composition
Transverse moraine ridge exposed (a) on the surface of Kongsvegen (viewing towards the north) and (b) at the nearby glacier margin (viewing towards the south). Figures for scale are in roughly the same location in both photographs.
West lateralmargin of midre Love¤nbreen: (a) general viewand (b) closer viewof basal solid sub-facies overlain by debrispoor, foliated glacier ice.
Longitudinal moraine ridge emerging from the surface of midre Love¤nbreen. Ice flow is directly out from the page.
Sample treatment and analysis
Ice, meltwater, frozen debris and unfrozen debris were sampled. Unfrozen debris was sampled by hand trowel, stored and transported in sealed plastic bags. Samples of frozen debris and debris-rich ice were removed from the glacier and broken up by ice axe and melted in clean plastic bags. Samples of debris-poor ice were recovered by ice screw. All samples were transported in sealed plastic bags to a field laboratory where they were filtered, dried and weighed within 24 hours of sampling. When filtrate was retained for isotopic analysis at the Geophysical Isotope Laboratory, Copenhagen University, Denmark, it was stored and transported in sealed, dark-brown bottles. Meltwater sampled as liquid in the field was press-filtered prior to storage in sealed, dark-brown bottles.
Debris textures were determined in the laboratory at 1ϕ intervals, from -4ϕ to 3ϕ by dry sieving, and from 3ϕ to “finer than 10ϕ” by settling analysis (SediGraph 5100, Mi-cromeritics). Results are presented on plots of size (ϕ) against weight (%) and as double logarithmic plots of particle diameter (d) against number of particles (N d). The latter of these allows the degree of self-similarity (expressed as the correlation coefficient (R) of the variables) and fractal dimension (m) of the debris to be calculated. Here, a selfsimilar distribution will define a straight line (R = 1.0000) on the plot of log d against log N d according to:
(Reference Hooke and Iverson.Hooke and Iverson, 1995), where N 0 is the number of particles of reference diameter d 0, and m, the fractal dimension, is given by the negative slope of the log-log plot. Although insensitive to minor changes in grain-size distribution (Reference Benn and Gemmell.Benn and Gemmell, 2002), the value of m summarizes the ratio of smaller particles to larger particles over the size range analyzed. The analysis therefore provides a useful single-value expression for the character of an entire particle-size distribution, and thereby provides a straightforward means to compare samples. A self-similar distribution of tessellating cubes has a fractal dimension of 2.58 (Reference Sammis, King and BiegelSammis and others, 1987), and samples of basally derived debris generally have fractal dimensions in the range 27−3.0 (e.g. Reference Hooke and Iverson.Hooke and Iverson, 1995; Reference Hubbard, Sharp and Lawson.Hubbard and others, 1996; Reference Fischer and Hubbard.Fischer and Hubbard, 1999; Reference Khatwa, Hart and Payne.Khatwa and others, 1999).
Clast macro-fabrics were recorded in the field for samples of 50 prolate clasts, each with an a-axis/c-axis ratio of >2 (Reference AndrewsAndrews, 1970). The data are plotted on Schmidt equal-area lower-hemisphere projections, and summarized using standard eigenvector analysis.
Statistical testing of differences between sample data is based on two sample t tests (large samples) or U tests (small samples), and results are expressed as the probability (P) of differences in the data being due to chance. According to the notation used, “similar at P > 0.1” means that there is greater than a 10% chance of the samples being from the same parent population, and “different at P < 0.01 ” means that there is less than a 1% chance that the samples are from the same parent population.
Results
Field sections
Transverse supraglacial moraine ridges
The transverse moraine ridge studied at Kongsvegen extended from the glacier margin for several tens of metres across the glacier surface. The surface debris formed an ice- cored mound of diamicton ~1 m high. Washing off the unconsolidated surface debris revealed a ~0.5 m wide debris band formed of thin alternating layers of debris-rich and debris-poor ice, the former containing a wide range of particle sizes. The debris-poor ice layers were largely bubble- free. Thus, it was visually similar to the stratified discontinuous sub-facies of the basal zone, as identified by Lawson (1979b) at Matanuska Glacier.
At midre Lovenbreen, a similar transverse moraine ridge extended from the glacier surface down the lateral margin, where it merged indistinctly into a layer of frozen basal sediment composed of muddy gravel. Similar to the surface debris band sampled at Kongsvegen, the band at midre Lovenbreen was formed of multiple debris-rich layers separated by relatively clean, bubble-free ice.
Longitudinal supraglacial moraine ridges
At midre Lovenbreen, the debris band associated with the longitudinal moraine studied forms part of a fold hinge that dips up-glacier at a shallow angle, consistent with Hambrey and others’ (1999) structural interpretation of these features (Fig. 1). This band was layered by debris concentration, with debris-rich layers containing polymodal diamicton with clasts up to boulder size.
At austre Brøggerbreen, the debris band forming the longitudinal moraine sampled at the surface was ~50 cm thick, and close inspection revealed that it was composed of alternating debris-rich and debris-poor layers. This band was visually similar to stratified basal ice and therefore similar in structure to other debris-charged ridges sampled for the present study.
The stratified-facies ice forming the longitudinal debris band at austre Brøggerbreen was also characterized by a systematic pattern of debris incorporation. Here, thin laminae, defined by fine-grained red debris (of Carboniferous mudstone and sandstone), enveloped a core of massive, grey diamicton (of Proterozoic metamorphic rocks) containing only interstitial ice (Fig. 6). Thus, the unit incorporating the red mudstone was identical to Lawson’s (1979b) stratified discontinuous sub-facies, and the unit containing the grey metamorphic material was identical to Lawson’s stratified solid sub-facies. The latter is frozen debris containing interstitial ice and ice lenses.
Longitudinal moraine ridge exposed at the surface of austre Breggerbreen. Ice flow is away from the viewer.
Basal ice
Basal ice and debris were sampled from the margins of Kongsvegen and midre Lovtjnbreen (Table 1). In both cases the basal zone is composed of stratified-facies ice which is principally solid sub-facies. No basal ice was sampled at austre Brøggerbreen. A separate, planar, lamination some millimetres thick and containing only fine debris was also sampled at the west margin of midre Lovtjnbreen. This layer was parallel to the debris-charged ridge at the site and was visually similar to planar-facies basal ice identified by Reference Hubbard and Sharp.Hubbard and Sharp (1995) at Alpine glaciers.
Sedimentology
Debris concentration
Debris concentration (expressed as grammes of debris per litre of meltwater; g L-1) was calculated for 35 samples of ice-borne debris (Table 2; Fig. 7). The mean concentration of 13 samples of basal ice sampled from ice-marginal locations is 7000 gL-1, whereas that of 11 transverse and 11 longitudinal supraglacial moraine samples is 1466 and 1259 g L 1 respectively. Statistical analyses of these data indicate that the supraglacial debris-band concentrations are significantly lower than the basal ice concentrations (P < 0.01). In contrast, the transverse and longitudinal debris-band concentrations are not significantly different from each other (P > 0.01).
Summary of sediment concentration results, classified by sample source and glacier. # indicates number of samples, 
indicates the mean concentration (g L
-1), and σ indicates the standard deviation in concentration (g L
-1)
Box plots of debris concentration data, classified by sample type from all three glaciers. Markers denote the 0th, 1st, 5th, 25th, 50th, 75th, 95th, 99th and 100th percentile values.The open square denotes the mean value.
Debris particle-size distribution
Forty-nine debris samples were analyzed for their particle- size distributions. These are classified by glacier and sample type, and plotted as size against weight in Figures 8−10. These data are also summarized in Table 3 in terms of % weight represented in the gravel (−4 to−1ϕ inclusive), sand (−0.5 to 4ϕ inclusive) and silt and clay (>4ϕ) size classes, and in terms of the correlation coefficient (R) and inverse slope (m) of bivariate plots of log number of particles against log particle diameter.
Bivariate plots of weight against size class for Kongsvegen debris samples: (a) basal solid sub-facies; (b) transverse supraglacial moraine ridge; (c) ice-cliff transverse moraine ridge; (d) transverse supraglacial moraine ridge melt-out debris.The finer than 10ϕ size class is not plotted.
Bivariate plots of weight against size class for midre Love¤nbreen debris samples: (a) basal solid sub-facies; (b) transverse supraglacial moraine ridge; (c) basal planar facies.The finer than 10ϕ size class is not plotted.
Bivariate plots of weight against size class for austre Br_ggerbreen debris samples: (a) longitudinal supraglacial moraine ridge (red Carboniferous debris); (b) longitudinal supraglacial moraine ridge (grey Proterozoic debris). The finer than 10ϕ size class is not plotted.
Summary of particle-size distribution results, classified by sample source and glacier. Weight in size class (%) relates to standard size-weight plots, and m and R are the negative slope and correlation coefficient respectively of plots of log N d against log d
Data from Kongsvegen indicate broad similarity in the textures of the basal solid sub-facies debris, the transverse supraglacial debris band (whether sampled at the ice surface or the ice margin) and the melt-out debris forming the transverse supraglacial moraine (Fig. 8). Close inspection of Figure 8, however, indicates that the last of these has a slightly greater proportion of gravel-sized clasts (or a lower proportion of silt- and clay-sized clasts) than the debris entrained within the debris band. Analysis of these data indicates significant depletion (P < 0.05) of silt-sized particles in the surface moraine (14.5% silt and clay) relative to the debris bands sampled at the glacier surface (24.8% silt and clay) and margin (22.1% silt and clay). Debris entrained within the latter two sample groups is statistically similar for all three size classes (P > 0.1).
At midre Lovétjnbreen (Fig. 9) the texture of the debris within the supraglacial transverse debris band is generally similar to that of the ice-marginal basal solid sub-facies. However, in detail, the former was significantly (P < 0.01) depleted in silt- and clay-sized material (4.1% silt and clay) relative to both the latter (23.0% silt and clay). Both were significantly (P < 0.01) depleted in silt- and clay-sized material relative to the planar facies sampled at the margin of the glacier, which was very well sorted and fine-grained (82.7% silt and clay). Corresponding inverse statistical differences exist between these sample groups in the gravel- size fraction.
At austre Brøggerbreen (Fig. 10) our data indicate a marked difference between the texture of the red and the grey debris within the longitudinal supraglacial debris band. Thus, the red sediment is significantly (P < 0.01) depleted in gravel-sized material (39.3% gravel) and enriched in silt- and clay-sized material (23.9% silt and clay) relative to the grey sediment (69.2% gravel; 7.3% silt and clay).
Summary data of the bivariate plots of log number of particles against log particle diameter (Table 3) indicate slopes or fractal dimensions (m) that are in the range 2.6− 2.9 with a few notable exceptions. The grey debris-charged ridge material at austre Brøggerbreen has a fractal dimension of 2.53, consistent with the general depletion in fines noted above. Similarly, debris sampled from the surface debris-charged ridge at midre Lovtjnbreen has a fractal dimension of 2.45. Conversely, the fine debris sampled from the planar facies at midre Lovtjnbreen has an apparent fractal dimension of 3.97, although this is questionable since the log−log bivariate plot is clearly not linear (R = −0.973) (Table 3).
Clast macro-fabrics
Eight sets of clast macro-fabric data were recorded from within the debris bands sampled at the three glaciers studied (Fig. 11; Table 4). At Kongsvegen the two samples recovered from the transverse supraglacial debris band (one from the marginal ice cliff and the other from the glacier surface; Table 1) are similar to each other, characterized by strong unimodal fabrics (first eigenvalues = 0.82 and 0.70) with an azimuth of ~20° and a dip of ~18°. These directions are parallel to the plan-form orientation of the supra- glacial moraine and its associated debris band, i.e. transverse to the direction of ice flow. In contrast, the local ice-marginal basal diamicton is characterized by a weaker fabric (first eigenvalue = 0.62). Fabrics measured in the basal solid sub-facies located around the margins of midre Lovtjnbreen were also spatially variable, characterized by spherical variances of 0.67−0.88 (Table 4). At austre Brog- gerbreen, the longitudinal debris-charged ridge sample from the glacier surface was characterized by a strong unimodal fabric with a first eigenvalue of 0.83.
Summary of clast macro-fabrics as plotted on equal-area lower-hemisphere projections (Figure 11), classified by sample source and glacier
Schmidt equal-area lower-hemispheric projections of clast fabric samples presented by sample type and glacier: (a)Kongsvegen basal solid sub-facies (unfrozen); (b) Kongsvegen moraine ridge from ice cliff; (c) Kongsvegen supraglacial moraine ridge; (d) midre Lovénbreen basal solid sub-facies (unfrozen) from east margin; (e) midre Love¤nbreen basal solid sub-facies from west margin; (f) midre Lovénbreen proglacial diamicton (unfrozen) from east margin; (g) midre Lovénbreen basal solid sub-facies from west margin; and (h) austre Br_ggerbreen supraglacial moraine ridge. Points are contoured at 5%intervals per 1% of area, and arrows indicate the local ice-flow direction.
Oxygen isotope composition
Oxygen isotope data are calculated as δ18O in ‰, which expresses the ratio of the abundance of the isotope 18O to 16O in the sample relative to that of Standard Mean Ocean Water (SMOW):
Analysis of 110 ice and water samples yielded a mean value of −11.75% and a standard deviation of 0.76% (Table 5). There is little variation in the sample means between the three glaciers studied; the mean isotopic composition of ice samples was −11.61% (n = 39) from Kongsvegen, −11.81% (n = 50) from midre Lovénbreen and −11.87% (n = 21) from austre Brøggerbreen.
In order to investigate these data further, samples are subdivided by glacier and by sample type, summarized in Table 5 and Figure 12. These data reveal significant and systematic patterns in sample group isotopic composition.
Summary of oxygen isotope results, classified by sample source and glacier. # indicates number of samples, indicates the mean δ18 O value (standard deviation of δ18 O values (‰)
Box plots of δ18Ocomposition of ice facies by sample type and glacier: (a)Kongsvegen, (b) midre Loveénbreen and (c) austre Brøggerbreen. Markers denote the 0th, 1st, 5th, 25th, 50th, 75th, 95th, 99th and 100th percentile values.The open square denotes the mean value. SDBin axis labels stands for supraglacial debris band.
At Kongsvegen, the mean composition of glacier ice and supraglacial meltwater is −12.14% (n = 23), and the mean composition of the (debris-rich) ice within the supraglacial debris band is −10.82% (n = 6). The respective values at midre Lovtjnbreen are −12.38% (n = 21) and −11.40% (n = 20). At both glaciers, ice within the supraglacial debris bands is isotopically enriched (P < 0.01) in 18O relative to glacier ice and surface meltwater samples. The ice sampled from the supraglacial debris bands is isotopically similar (P > 0.1) to that sampled from the debris-rich basal layer (or frozen subglacial sediment) located at the margin of these glaciers: −10.85% (n = 10) at Kongsvegen and −11.05% (n = 4) at midre Lovénbreen.
At austre Brøggerbreen, the isotopic composition of the ice forming the longitudinal supraglacial debris band (δ18O = −11.83%; n = 11) is similar to (P > 0.1) that of glacier ice (δ18O = −11.90%; n = 10). However, if the samples recovered from the debris band are reclassified by sub-facies, the solid sub-facies (grey debris) is depleted in 18O relative to the discontinuous sub-facies (red debris) (P < 0.05). Neither sub-facies has a significantly different isotopic composition from glacier ice. However, if an anomalous glacier ice sample of-14.86% in •18O is discounted from the analysis, the solid sub-facies becomes significantly lighter than the remaining nine glacier ice samples (P < 0.01).
Discussion
Certain consistent relationships between the supraglacial debris bands and other sample types emerge from the evidence presented above.
Transverse supraglacial debris bands and moraine ridges
Transverse debris bands at Kongsvegen and midre Lovénbreen contain debris that is generally of similar particle-size distribution to that within basal ice at these and other glaciers (e.g. Reference LawsonLawson, 1979b; Reference Hubbard and Sharp.Hubbard and Sharp, 1995). At both Kongsvegen and midre Lovenbreen, many of the clasts entrained within the transverse debris bands are striated and faceted. They are also characterized by a strong unimodal fabric in which the clasts are aligned parallel to the plane of the supraglacial moraine ridge. At Kongsvegen, this preferred orientation is remarkably consistent at two sites, one exposed on an ice cliff and the other ~30 m distant on the glacier surface (Fig. 11b and c). At Kongsvegen and midre Lovenbreen, ice contained within the basal solid sub-facies and the supraglacial debris bands (whether at the glacier margin or glacier surface) is enriched in •18O by ~1-2 % relative to local glacier ice and supraglacial meltwater. Since glacier ice (or basal meltwater derived from it) is the most likely source for the basal ice and debris-band ice, it is probable that these latter groups have been isotopically altered during their formation and/or transport. Such enrichment is consistent with open-system, or incomplete, freezing of meltwater in the presence of debris at the glacier bed (Reference Jouzel and Souchez.Jouzel and Souchez, 1982; Reference Souchez and JouzelSouchez and Jouzel, 1984). This is supported by the absence of any significant difference between the isotopic composition of the debris-band ice and that within the subglacial basal solid sub-facies at Kongsvegen and midre Lovénbreen.
In summary, these sedimentological data provide strong evidence that the debris incorporated within the transverse debris bands and supraglacial ridges at Kongsvegen and midre Lovenbreen was derived from the beds of these glaciers. Further, the isotopic data are consistent with the ice matrix of these debris bands also originating by refreezing at the glacier bed.
These data may also be used to shed some light on the processes responsible for forming the transverse debris bands concerned; in particular on the competing hypotheses of formation as thrusts or as basal crevasses. The main obstacle to such an interpretation is that both processes could produce features with physical and compositional similarities. Both, for example, involve the same subglacial debris and water source, and both can result in the development of strong clast fabrics within the bands they form. However, we believe the data from this study are more consistent with an origin as thrusts than with an origin as basal crevasses for the following reasons.
Fluidized flow, however viscous, of soft sediments into basal crevasses would be characterized by some degree of local debris sorting. In this study, we neither observed nor measured any such sorting. At Kongsvegen, for example, almost identical polymodal diamicton was recovered from samples of the transverse supraglacial debris band located at the glacier surface and in an ice-cliff section tens of metres distant. Although these bands were layered by variations in debris concentration, the debris was not sorted in terms of its grain-size distribution.
Fluidized flow of soft sediments into basal crevasses (at debris-water concentrations of 4 1000 g L-1; Table 2) would be unlikely to result in spatially extensive, planar layering such as was observed in the present study. The transverse debris bands investigated at the surface of Kongsvegen and midre Lovénbreen were formed of extensive debris-rich layers separated by clean and bubble-free ice, identical to stratified-facies basal ice. These properties therefore indicate that the transverse supraglacial debris bands sampled at these glaciers are formed of pre-existing stratified-facies basal ice that has been elevated from the glacier bed to the surface without noticeable alteration. While such a mechanism is incompatible with the formation of these debris bands by basal crevasse filling, it is compatible with their initial formation as basal ice and their subsequent englacial transport by thrusting.
It is likely that fluidized flow of soft sediments into basal crevasses would be characterized by some degree of fining with distance from source, as identified by Reference Ensminger, Alley, Evenson, Lawson and Larson.Ensminger and others (2001). This effect was not observed in the present study.
Basal crevasses would be expected to cut sharply across other basal ice layers at a high angle (consistent with crevasse orientation being broadly orthogonal to the glacier bed and basal ice layers being broadly parallel to it). This effect was not observed in the present study. Conversely, we did observe continuity in the structure of individual transverse debris bands between the surface and margins of midre Lovtjnbreen. In this case, the bands merged indistinctly into the debris-rich basal ice layer present at the base of the lateral margin of the glacier (Fig. 4). This pattern is consistent with local ductile deformation contributing to, and occurring between, initially low-angle thrusts initiating near or at the ice−bed interface.
The heavy-isotope enrichment of the debris bands by <3 % in δ18O relative to glacier ice and supraglacial meltwaters is consistent with basal ice formation by open-system refreezing at the glacier bed. Indeed, such enrichment has commonly been reported in basal ice studies (Reference Lawson and Kulla.Lawson and Kulla, 1978; Reference Hubbard and Sharp.Hubbard and Sharp, 1989). In contrast, once injected into a basal crevasse, a meltwater suspension is more likely to freeze without renewed water turnover, essentially closing the system. Sampling ice frozen in a closed system should result in a wide range of isotopic values, from slightly heavier (≤3 in δ18O) to substantially lighter (>6 % in 618O as freezing nears completion) than the composition of the water in the slurry from which they formed (Reference Jouzel and Souchez.Jouzel and Souchez, 1982). This effect was not measured in the present study.
Although none of the individual lines of evidence presented above can be interpreted as unequivocal proof of transverse supraglacial debris-band formation as thrusting of basal ice from the glacier bed, the weight of evidence favours such a mechanism over that involving formation as sediment-filled basal crevasses. Indeed, Reference Hubbard and Sharp.Hubbard and Sharp (1995) interpreted planar facies basal ice sampled in the Alps as healed crevasses, probably containing aeolian debris sourced from the glacier surface. The planar facies sampled from midre Lovenbreen is similar to these features and we interpret it similarly. However, it is possible in both cases that the facies forms as a basal fracture into which fine subglacial debris may be introduced by flushing in suspension (Reference Knight and Knight.Knight and Knight, 1994).
One further observation at Kongsvegen was that the unconsolidated material sampled from the surface of the supraglacial moraine ridge lacked fines relative to that sampled from the underlying, and ice-marginal, debris band. We interpret this effect in terms of the preferential eluviation of fine particles from surface moraine ridges by rainfall and meltwater. Similar effects were reported by Reference Boulton and Dent.Boulton and Dent (1974) and Reference Fischer and Hubbard.Fischer and Hubbard (1999).
Longitudinal supraglacial debris bands and moraine ridges
The longitudinal supraglacial debris band sampled at austre Brøggerbreen contains debris that is polymodal, has a typically basal particle-size distribution and contains clasts that are faceted and striated. As with transverse debris bands at Kongsvegen and midre Lovenbreen, therefore, we interpret this material as being basally derived.
The longitudinal supraglacial debris band at austre Brøggerbreen is formed of two sub-facies: a central, solid sub-facies enveloped by a discontinuous sub-facies (Fig. 13). Associating this pattern with Hambrey and others’ (1999) structural interpretation of longitudinal debris bands (Fig. 1) indicates the presence of a basal ice layer composed of two sub-facies at the bed of this glacier. Moreover, the position of the sub-facies at the surface of austre Brøggerbreen indicates that at the glacier bed the discontinuous sub-facies overlies the solid sub-facies (Fig. 13). This implies that the former was incorporated up-glacier of the latter and/or before the latter. This interpretation is consistent with the strong lithological contrast between the debris incorporated within the different sub-facies.
Schematic illustration of the distribution of basal ice sub-facies associated with the longitudinal supraglacial moraine ridge sampled at austre Bmggerbreen (depicted in Fig. 6.).
It is generally accepted that solid sub-facies basal ice forms by the net adfreezing of unconsolidated subglacial sediments (Reference Hubbard and Sharp.Hubbard and Sharp, 1989). At polythermal glaciers, this is associated with temporal variations in the position of the freezing isotherm at the boundary between subfreezing basal conditions at the ice margins and temperate basal conditions beneath thicker ice up-glacier (Reference WeertmanWeertman, 1961). In contrast, thinly layered discontinuous sub-facies basal ice forms from repeated freezing events more likely to be associated with generally temperate basal conditions. Such freezing may involve a number of processes including: (i) the initial formation of finely laminated ice by closed- system regelation (Reference Kamb and LaChapelle.Kamb and LaChappelle, 1963; Reference Hubbard and Sharp.Hubbard and Sharp, 1993, 1995), (ii) more extensive freeze-on associated with ephemeral patches of cold basal ice (Reference Robin and deRobin, 1976), or (iii) the freezing of supercooled waters emerging from basal overdeepenings (Reference Alley, Lawson, Evenson, Strasser and Larson.Alley and others, 1998, 1999; Reference Lawson, Strasser, Evenson, Alley, Larson and Arcone.Lawson and others, 1998). We therefore infer from the patterns we record at austre Brøggerbreen that temperate basal conditions existed upflow of marginal freezing conditions at the time of the formation of the ice now exposed in the longitudinal debris band at the glacier’s surface. Since austre Brøggerbreen is currently largely cold-based (Reference Hagen and Sætrang.Hagen and Sstrang, 1991; Reference Hagen and Sætrang.Hagen and others, 1991a), it is likely that these basal ice sub-facies formed > 100 years ago, when the glaciers of the area were generally thicker and more dynamic than at present (Reference Glasser and Hambrey.Glasser and Hambrey, 2001).
The discontinuous sub-facies debris band is isotopically similar to glacier ice at austre Brøggerbreen, and both are isotopically heavier than the solid sub-facies debris band sampled at the glacier. The isotopic similarity of the discontinuous sub-facies to the glacier ice must be explained in the light of the size of the sample collected relative to the scale of individual freezing events (the latter being a unit of ice formed from a closed and isotopically uniform water body). Since the discontinuous sub-facies at austre Brøggerbreen contains millimeter-scale laminae, and the ice screw used to sample it was ~10 mm in diameter, no isotopic enrichment would be expected if the sub-facies formed by closed- system refreezing of water that was isotopically similar to current glacier ice (Reference Jouzel and Souchez.Jouzel and Souchez, 1982; Reference Hubbard and Sharp.Hubbard and Sharp, 1993). This and the physical structure of the discontinuous sub-facies are consistent with initial formation by Weertman regelation (Reference WeertmanWeertman, 1964), implying that the ice formed in an area of the glacier bed that was temperate and probably bedrock-based (Reference Kamb and LaChapelle.Kamb and LaChapelle, 1963; Reference Hubbard and Sharp.Hubbard and Sharp, 1993).
Two interpretations may be advanced for the relative isotopic lightness (by ~1‰ in δ18O) of the solid sub-facies relative to glacier ice at austre Brøggerbreen. First, the sub-facies may have formed by the open-system freezing of source water that was, at the time of formation, >1 % lighter in δ 8O than current glacier ice. Second, the sub-facies may have formed by the closed-system freezing of source water that was, at the time of formation, ~1‰ lighter in S18O than current glacier ice. In the latter case, for isotope samples to be of the restricted range in S 8O measured, the scale of each freezing event would have to be smaller than our sample size (~10 mm vertically). This is unlikely given the massive and undifferentiated nature of the solid sub-facies. We therefore favour formation of the solid sub-facies ice at austre Brog- gerbreen by the open-system freezing of water that was at least 1% lighter in δ8O than current glacier ice. However, these competing hypotheses can really only be evaluated with confidence in the light of more ice and water samples from the glacier, particularly from its base.
Conclusions
Physical properties of debris bands, from which supraglacial moraine ridges are formed, suggest all transverse bands and some longitudinal bands are sourced from the glacier bed. The sedimentology and isotopic composition of transverse bands indicate formation from pre-existing basal ice that has been elevated with little bulk modification into an englacial position. Our evidence suggests the process responsible for this elevation is more likely to be related to thrusting than to the filling of basal crevasses.
Longitudinal debris bands can also be sourced from the glacier bed, and one such band was observed at austre Brøg gerbreen to be formed of two distinct sub-facies. Isotopic analysis of these sub-facies indicates that the glacier was polythermal, with a temperate interior and a frozen margin, at the time of basal ice formation.
Acknowledgements
We thank T. Knudsen (University of Aarhus, Denmark) and C. Hammer (Geophysical Isotope Laboratory, Copenhagen University) for arranging the isotope sample analyses. We also thank D. Evans and D. Lawson for commenting on the manuscript, as a result of which it has been greatly improved. This work was partly funded by a U.K. Natural Environment Research Council (NERC) grant (GST022192). J.E. acknowledges funding by NERC studentship NER/S/A/2000/03690.
