Suspended-sediment concentration (SSC) and discharge time series have been obtained for a proglacial stream draining from Scott Turnerbreen, a non-temperate glacier at 78° N in the Svalbard archipelago. These time series exhibit seasonal trends in diurnal hysteresis between SSC and discharge, which changes progressively from clockwise (SSC lead) to anticlockwise (SSC lag), and in the availability of sediment for fluvial transport, which increases during the season. These results contrast with those for streams draining temperate glaciers, in which a seasonal trend in diurnal hysteresis is typically not observed, and sediment supply is usually exhausted during the melt season. This contrast indicates differences in the processes of sediment acquisition and in temporal patterns of sediment transport between temperate and non-temperate glaciers, which reflect different drainage-system structures. The principal mechanism of suspended-sediment acquisition by meltwater is subaerial sediment supply to ice-marginal channels at non-temperate glaciers, whereas at temperate glaciers it is subglacial supply to basal channels. The hysteresis trend probably arises from a delaying effect, because of sediment circulation in pools and slack water adjoining the ice-marginal channels, which varies with discharge. Continuing subaerial sediment supply from heavily debris-covered ice-cored moraines during the melt season precludes sediment-supply exhaustion at Arctic glaciers but means that measurements of sediment flux give a misleading indication of the contemporary rate of glacial erosion.

Present understanding of the subglacial water system and its role in transporting solute and sediment is largely based on subglacial observations of water pressure, turbidity and electrical conductivity and on portal measurements. Such data reveal a wealth of intriguing phenomena, but convincing interpretations can be elusive. Although a proper mathematical description of the subglacial water system would unquestionably lead to a coupled system of non-linear partial differential equations, it is not fruitful to introduce this level of complexity until the important physical processes have been identified and quantified. Lumped-element models offer an efficient approach to examining the complex but dimly perceived physics of the subglacial water system. Water volume, hydraulic head, discharge and flow resistance have the respective electrical analogues of charge, voltage, current and ohmic resistance. Thus, subglacial hydraulic circuits can be approximated by electrical circuits. Mathematically, this circuit description commonly leads to a coupled system of algebraic and differential equations which can be solved numerically. It is straightforward to enrich this representation by adding sources and sinks of solute and sediment. To demonstrate the method, model results are compared to records of subglacial pressure, electrical conductivity and turbidity measured beneath Trapridge Glacier, Yukon Territory, Canada.

Field studies at Mueller and Tasman Glaciers in New Zealand and in Iceland have revealed extensive ice-marginal moraine sequences fed by a series of debris bands containing characteristic rounded debris. We interpret these debris bands as meltwater deposits from relict conduits. The process of sediment accumulation described requires that debris undergoing basal or high-level ice transport is entrained by running water, becoming comminuted and rounded before being returned to high-level ice transport and eventual deposition in marginal moraines. Tins sequence of events suggests a link between glacier hydrology and ice-marginal sedimentation, possible mechanisms for which are explored. Pronounced moraines with rounded debris can be expected if (a) large quantity of debris is carried within the englacial drainage network; (b) there is a tendency for this debris to be abandoned within the ice; and (c) ice-flow trajectories and ablation ensure that this debris congregates at the ice surface. Streams which flow at high level within the ice can intercept and entrain englacial debris derived from rock fall. However, debris sources at Gigjökull and Steinholtsjökull in Iceland are restricted to the glacier bed, suggesting that, in certain cases, the presence of a basal overdeepening acts as a key factor controlling the accumulation of these moraines. Water pressure rises as channels encounter an overdeepening, forcing debris-laden streams to leave the bed and take an englacial route. That debris may pass from ice to water and back into ice before deposition has gone largely unrecognised in accounts of glacial process systems, yet it provides an explanation of how temperate alpine glaciers can include water-worked debris in their marginal depositional facies.

In shearing sediment beneath glaciers, networks of grains may transiently support shear and normal stresses that are larger than spatial averages. Consistent with studies of fault-gouge genesis, we hypothesize that crushing of grains in such networks is responsible for surrounding larger grains with smaller grains. At sufficiently large strains, this should minimize stress heterogeneity, favor intergranular sliding and abrasion rather than crushing, and result in a self-similar grain-size distribution.

This hypothesis is tested with a ring-shear device that slowly shears a large annular sediment sample to high strains. Shearing and comminution of weak equigranular (2.0–3.3 mm) sediment resulted in a self-similar grain-size distribution with a fractal dimension that increased with shear strain toward a steady value of 2.85. This value is significantly larger than that of gouges produced purely by crushing, 2.6, but it is comparable to values for tilts thought to be deforming beneath modern glaciers, 2.8 to nearly 3.0. At low strains, under a steady mean normal stress of 84 kPa, variations in normal stress measured locally ranged in amplitude from 50 to 300 kPa with wavelengths that were 100 times larger than the initial grain diameter. Crushing of grains, observed through the transparent walls of the device, apparently caused the failure of grain networks. At shearing displacements ranging from 0.7 to 1.0 m, the amplitude of local stress fluctuations decreased abruptly. This change is attributed to fine sediment that distributed stresses more uniformly and caused grain networks to fail primarily by intergranular sliding rather than by crushing of grains. Sliding between grains apparently produced silt by abrasion and resulted in a fractal dimension that was higher than if there had been only crushing.

A size distribution with a fractal dimension greater than 2.6 is probably a necessary but not sufficient condition for determining whether a basal till has been highly deformed. Stress heterogeneity in subglacial sediment that is shearing through its full thickness should contribute to the erosion of underlying rock.

A regional hydrogeological study conducted by the Geological Survey of Canada acquired 35 line-km of 12-fold seismic reflection profiles on or adjacent to the Oak Ridges moraine, north of Toronto, Ontario, Canada. The three-dimensional geometry provided by these data aids in understanding the erosional and depositional processes that occurred beneath the Laurentide ice sheet during the late stages of glaciation. The seismic sections indicate large infilled channels in the subsurface which are interpreted as tunnel channels eroded by large, subglacial meltwater discharges. Two seismic profiles from different areas of the moraine show channel-cutting events of different ages and different types of infilling.

The volume of debris in the left-lateral, Little Ice Age (LIA: AD 1550–1850) moraine of the Feegletscher, Valais, Switzerland was compared with the actual volume being transported currently by the glacier. The latter is smaller by a factor of about two. In Tröllaskagi, north Iceland, a surface cover of debris on top of a very slow moving glacier ice mass (glacier noir, rock glacier) has been dated by lichenometry. The age of the oldest part is commensurate with LIA moraines in the area. Knowing the volume of debris of a given age allows an estimate of the debris supply to the glacier in a given time. Again, there appears to have been a significant reduction in debris to the glacier since the turn of the 19th century. Debris input in the early LIA seems to have been particularly copious and this may be important in the formation of some glacier depositional forms such as rock glaciers.

Seven basal ice facies have been defined on the basis of research at eleven glaciers in the western Alps. The concentration and texture of the debris incorporated in these facies are described. Grain-size distributions are characterised in terms of their: (i) mean size and dispersion, (ii) component Gaussian modes, and (iii) self-similarity.

Firnified glacier ice contains low concentrations (≈0.2 g 1−1) of well-sorted and predominantly fine-grained debris that is not self-similar over the range of particle diameters assessed. In contrast, basal ice contains relatively high concentrations (≈4–4000 g 1−1 by facies) of polymodal (by size fraction against weight) debris, the texture of which is consistent with incorporation at the glacier bed. Analysis by grain-size against number of particles suggests that these basal facies debris textures are also self-similar. This apparent contradiction may be explained by the insensitivity of the assessment of self-similarity to variations in mass distribution. Comparison of typical size–weight with size–number distributions indicates that neither visual nor statistical assessment of the latter may be sufficiently rigorous to identify self-similarity.

Apparent fractal dimensions may indicate the relative importance of fines in a debris distribution. Subglacially derived basal facies debris has a mean fractal dimension of 2.74. This value suggests an excess of fines relative to a self-similar distribution of cubes, which has a fractal dimension of 2.58. Subglacial sediments sampled from the forefield of Skalafellsjökull, Iceland, have fractal dimensions of 2.91 (A-horizon) and 2.81 (B-horizon). Debris from the A-horizon, which is interpreted as having been pervasively deformed, both most closely approaches self-similarity and has the highest fractal dimension of any of the sample groups analyzed.

The advance of Hubbard Glacier, near Yakutat, Alaska, U.S.A., in spring 1986 blocked the entrance to Russell Fiord with an ice-and-sediment dam, behind which a lake formed. The water level in Russell Lake rose to 25.5 m a.s.l. The dam catastrophically failed in October 1986, releasing 5.4 km3 of water into Disenchantment Bay. High-resolution seismic-reflection profiles show a 7.5 km long channel system cut into and buried by glacimarine sediment, represented by continuous, parallel reflections. The chaotic seismic facies filling the channel is interpreted to be debris flow deposits. A gravity core from channel-overbank deposits contained sandy diamicton with mud clasts. Above the channel a 1–2 m thick sediment drape extends across the bay. Laminated mud, fining-upward sand beds and diamicton were recovered from this unit. The sediment-drape deposits were produced by suspension settling from turbid plumes and non-channelized turbidity currents generated by the outburst flood.

Investigations of the stratigraphy and facies within a 2.69 m long gravity core (PS1423–2) from the southwestern Weddell Sea, Antarctica, indicate a significant change in the character of glaciomarine sedimentation since grounded ice withdrew from the continental shelf. Based on visual description, X-radiography, clast shape, particle-size analysis, physical properties and geochemical data, the core used in this analysis comprises five distinct units, from top to bottom: (i) massive diamicton, (ii) weakly to well-stratified diamicton, (iii) millimetre-scale laminated muds, with little or on coarse-clastic input, (iv) well- to weakly stratified diamicton, (v) massive diamicton. This succession is attributed to the variation in sedimentation associated with the recession of the grounding line of a previously advanced glacier. Grounded ice decoupled from the continental shelf to form an ice shelf, probably initiated by a rise in sea level in response to global climatic changes. Following disintegration of the ice shelf, sedimentation was influenced by marked variations in iceberg production. AMS-derived 14C ages from the upper 46 cm of the core indicate that the succession has been deposited since the end of the most recent glacial maximum (late Pleistocene), a conclusion supported by published data.

Vitus Lake, the ice-marginal basin at the southeastern edge of Bering Glacier, Alaska, U.S.A., is a site of modern, rapid, glacial-marine sedimentation. Rather than being a fresh-water lake, Vitus Lake is a tidally influenced, marine to brackish embayment connected to the Pacifie Ocean by an inlet, the Seal River. Vitus Lake consists of five deep bedrock basins, separated by interbasinal highs. Glacial erosion has cut these basins as much as 250 m below sea level. High-resolution seismic reflection surveys conducted in 1991 and 1993 of four of Vitus Lake’s basins reveal a complex, variable three-component acoustic stratigraphy. Although not fully sampled, the stratigraphy is inferred to be primarily glacial-marine units of (1) basal contorted and deformed glacial-marine and glacial sediments deposited by basal ice-contact processes and submarine mass-wasting; (2) acoustically well-stratified glacial-marine sediment, which unconformably overlies the basal unit and which grades upward into (3) acoustically transparent or nearly transparent glacial-marine sediment. Maximum thicknesses of conformable glacial-marine sediment exceed 100 m. All of the acoustically transparent and stratified deposits in Vitus Lake are modern in age, having accumulated between 1967 and 1993. The basins where these three-part sequences of “present-day” glacial-marine sediment are accumulating are themselves cut into older sequences of stratified glacial and glacial-marine deposits. These older units outcrop on the islands in Vitus Lake.

In 1967, as the result of a major surge, glacier ice completely filled all five basins. Subsequent terminus retreat, which continued through August 1993, exposed these basins, providing new locations for glacial-marine sediment accumulation. A correlation of sediment thicknesses measured from seismic profiles at specific locations within the basins, with the year that each location became ice-free, shows that the sediment accumulation at some locations exceeds 10 m year−1.

Investigations of grounding-line sedimentation in front of tidewater termini of temperate valley glaciers demonstrate that sediment yields and dynamics provide a second-order control on glacier stability by influencing water depth at the grounding line. Sediment is delivered to the grounding line by two routes: (1) debris transported in, on and beneath the glacier, and (2) sediment transported in glacial outwash streams. Glacial streams in Glacier Bay, Alaska, U.S.A., deliver 106 to 107 m3 year−1 of sediment to the grounding lines. The glacial debris flux transports 105 to 106 m3 year−1 of debris to the ice cliffs, where approximately 10% is released at the grounding line, the remainder being transported downfjord by iceberg-rafting. An additional 105 m3 year−1 of sediment may be transported to the grounding line by shearing and advection of a deformable bed.

A robotic submarine was used for the first observations of a grounding-line area of a floating glacier. The site was Mackay Glacier which terminates as a floating glacier tongue in the Ross Sea, at latitude 77°S. Half of the 20 m thick basal debris layers in Mackay Glacier are deposited as subglacial till in the last 1.8 km that the glacier remains grounded. Subglacial till observed at and beyond the grounding line varies rapidly in texture and rheology spatially, occurring as a flat sheet, as flow-parallel flutes, or as bank forms into which it has been pushed at the grounding line. Very little free- flowing subglacial water was present during the observations, and no major subglacial water discharges appear to have occurred in the past. The other half of the basal debris is melted out up to 1.5 km in front of the grounding line, producing a sheet of glacimarine sediment as shelfstone diamicton and mud draped on subglacial till. Both till and glacimarine sediment may be turbated by icebergs. This simple model of till overlain by shelfstone diamicton and mud is a direct contrast to sedimentary depositional systems at tide-water termini of temperate glaciers.

Detailed measurements of sediment flux in meltwaters draining from the portal of Gornergletscher, Valais, Switzerland, each ablation season between 1987 and 1990 indicate the nature of interactions between the developing basal drainage system and sediment stored at the glacier subsole. Sediment-flux maxima, which were generally smaller and less frequent later in a season, occurred at times of rising discharge. Overall, sediment flux first increased then decreased during a season. A simple conceptually based model is proposed for the subglacial interaction between flowing meltwater and sediment derived from glacial abrasion. The model was developed in order to allow investigation of relationships between suspended-sediment flux and discharge in meltwaters draining from temperate glaciers, In the model, water spreads out over a grid-square bed such that the area of subsole integrated with flow is a function of discharge. Sediment is abraded from the bed at a uniform rate in cells not covered by water. In winter, sediment accumulates over almost all the bed. Periods of rising flow at the start of the ablation season lead to drainage net expansion, much entrainment of stored sediment and large flux events in portal meltwaters. The model allows redistribution and deformation of stored sediment in two directions. Sediment squeezed to the margins of the wetted area throughout summer provides the background level of sediment flux between events. Reduction in total channel-margin length with progressive changes in drainage-network topology explains declining background flux later in summer. The model replicates well the timing, but not the absolute and relative magnitudes, of sediment-flux maxima in portal meltwaters draining from Gornergletscher.

The water and sediment output from Vitus Lake, in Front of Bering Glacier, was monitored starting in July 1994, Instrumentation was placed in the lake outlet to record stage, turbidity, conductivity and temperature and velocity of flow of river water. Two outburst floods punctuated the termination of the 1993–94 Bering Glacier surge in August 1994. The better-documented flood lasted about 10 d, during which the flood discharge averaged 1100 m3 s−1, in excess of normal discharge (1550 m3 s−1) during this part of the ablation season, and about 9.5 × 108 m3 of water drained from Bering Glacier. The excess water volume discharged during this flood corresponds to a 0.4 m thick layer of water extending over the 2500 km2 of Bering Glacier that was surging in early summer. The suspended-sediment flux from Vitus Lake during the summer of 1994 was two orders of magnitude less than rates of sediment production by other fast-moving glaciers in southern Alaska. This implies that most of the sediment produced is being stored in Vitus Lake, under the glacier, or in both locations.

Bakaninbreen, a 17 km long glacier terminating in fjord waters in central Spitsbergen (77°45′ N, 17°20′E), began to surge between the springs of 1985 and 1986. By summer 1994 the surge front had reached a position 3 km from the terminus and had almost ceased propagation. Structural investigations were undertaken to characterise the tectonic evolution of this thermally complex surge-type glacier, and the role played by thrusting and its effect on debris entrainment. Much of the glacier surface, particularly within and below the surge front, displayed transverse high-angle thrusts, defined by discrete fractures bounded by coarse clear ice. Some fractures were associated with a film of mud, whereas in others a discrete laver of diamicton, with interstitial ice several decimetres thick, was evident. Within the surge front, and genetically related to the thrusts, was a number of shear zones several metres wide. These were defined by fine-grained ice that was the product of the grinding up of crystals during shear (mylonitization). Three main sedimentary facies are associated with the thrusts: mud, gravelly mud and clast-rich muddy diamicton. The diamicton has the character of basal glacial debris: grain-size distribution ranging from clay to cobble size, clasts with a predominance in the sub-angular and sub-rounded classes, and striated and faceted clasts. Hot-water drilling through the glacier revealed several englacial layers above the surge front, and debris brought up on the drill stem suggests a basal origin. At least some of these englacial layers are probably the sub-surface continuations of the thrusts. The observed facies indicate that the glacier is moving over a soft, deformable bed and that thrusting is an important process in transferring debris to the surface, especially when the surge front is propagating down-glacier.

The greatest event in the 20th-century glaciological history of Iceland has been the glacier burst of Katla on Saturday 12 October 1918. Several cubic kilometres of water and ice were transported by the burst, and over 0.5 km3 of magma surfaced from the Katla caldera. The volcanic material was transported by air and water, and part of it probably formed pillow lava at the eruption site. The volcanic material transported by water was deposited mostly on the Mýrdalssandur plain and at the coast, but part of it was probably carried out to greater depths by gravity currents as the water entered the sea. The peak flow rate during the jökulhlaup was probably over 3 × 105 m3 s−1 of water, with a further 25 Kt s−1 of ice and a similar amount of sediment.

The classical theory of jökulhlaups used Röthlisberger’s earlier theory of ice-channel drainage to describe the development of the flood hydrograph. This theory has some drawbacks: the mechanism of initiation (breaking the seal) is opaque, the Manning roughness coefficient is too large and the hydrographs can reveal a sudden switching from channel opening to channel closure which is not simulated by the model. In this paper, we examine these features by exploring a more detailed model, which takes into account the physics of sediment erosion and its effect on channel morphology. We propose a theory in which channels need not be semicircular, but have shapes determined by a local balance between closure and melting, and in which erosion of the tunnel margins is taken into account; in particular, we derive theoretical predictions for sediment discharge, and we also propose a mechanism whereby the pressure seal over the caldera rim at Grímsvötn in Vatnajökull, Iceland, can be broken when the lake-level water pressure is still some 6 bar below the maximum overburden ice pressure.