Introduction

This dissertation attempts to explore the influence of near-snout overdeepening on ice flow by linking the various properties that give evidence of flow differences on the glacier surface of an outlet glacier of the ice-cap Eyjafjallajökull in Southern Iceland. Previous studies have suggested the idea of overdeepening beneath the glacier terminus of outlet glaciers flowing out of ice-caps on a steep slope gradient and ending in relatively soft sediments (Spedding, 1997).

The study of ice flow has a long tradition within glaciology and glacial geology which has lead to useful modelling of ice-flow (Nye 1952; Glen 1955; Paterson 1994). However two traditions have been developed, namely glacial geology and glaciology, both of which work perfectly well within their range of methods but fail to link processes and forms. Whilst glacial geology (Boulton, 1987, 1986) gives special importance to field data and the study of forms, i.e. concentrating on landforms and landscape, glaciology focuses on controlling processes (Clarke, 1987) and currently tends to attempt "abstract reductionist studies" (Spedding, 1997, p. 3) or sophisticated computer simulations. In trying to link form and process, glacial geomorphology plays a specific role in the field of geo-science (Spedding 1997). Like many authors -predominantly those from recent years, e.g. Sugden et al. (1997 cited in Spedding, 1997, p. 3) - I would confer that this link is crucial, not just for the understanding of changes in the landscape and for giving a meaningful explanation for the ways in which landforms have been produced, but also for their influence on processes. In this context, I intend to quantify changes in ice-flow throughout the glacier terminus and the influence of bed topography on ice flow as well as to qualify ‘visible’ properties of variability in the flow pattern. Using a specific set of field methods to quantify the influence of overdeepening on ice flow, I will follow the methodology of glaciology, whilst relating ice flow to the ‘visible’ properties (such as icefall, crevasse patterns, moulins, supra-glacial melt-water channels, closed-up conduits, debris bands and tephra outcrop, and eventually the lateral-terminal moraine) as in the glacial geomorphology tradition. The complexity of the process-form system does not permit a full explanation of the ‘key process relationships’. Generalisation as well as simplification is therefore inevitable.

Finally, the ice flow results will be used to back up and to evaluate the theory of overdeepening suggested by Spedding (1997). The latter will be considered in my final conclusion where the key process-landform relationship is set into the wider context of landscape and where the glacier’s sensitivity towards change is demonstrated. I will show that glaciers do not only change through space but also through time. I will discuss in a final step the implications for landforms (Brundsden & Thornes, 1979; and others) for instance, the historical implications.

This dissertation was inspired by the Iceland Fieldwork Course of 1997 when I first worked on Gigjökull. It was then that I decided to expand the project and do further research. It was of Nick Spedding’s thesis (1997) in particular, which aroused my interest in verifying, or at least testing, his hypothesis of overdeepening.

Chapter 1 Background

In this section, I will give a brief review of the literature relevant to this work. I will include some work that does not primarily correlate with the ‘ice flow’ but which will be important for the understanding of the further implications which can be inferred, and which will be discussed at the end of this dissertation. Therefore, the purpose of this chapter is to give the necessary background information and to introduce the reader to the study site.

1.1 Previous research

The area of interest is located at about 63° N in southern Iceland at a very low altitude (200-1400 m). Gígjökull is a temperate (i.e. warm-based) outlet valley glacier.

This allows the use of many substantive observations from past research which can be considered to be well founded. The substantive observations stem from past research, with a long tradition in the investigation of ice flow, glacier movement, etc. Principal explanations of glacier physics are provided by Paterson (1994) and Hook (1998). Good descriptions of stress and strain are provided by Paterson (1994), Twiss & Moores (1992) and Hutter (1983) wherein the latter gives more detailed analysis. The calculation of longitudinal stress will be of particular importance to this dissertation. Nye (1969) and Paterson (1994) provide the theoretical methods for calculating longitudinal stresses. In fact, as Bindschadler et al. (1977) have demonstrated by empirical studies, the influence of longitudinal stress can even be incorporated into the equation for stress (cf. Chapter 3) by taking into account the surface slope a . Flow laws that describe the response of ice to stress have mostly been developed in laboratory studies. The most often used is Glen’s Flow Law (1955; 1958), which was first adapted to ice by Nye (1957) (cf. Chapter 3 and 4). For the analysis of the data I will be using this relationship in particular. However, to believe that one single flow law could describe glacier movement precisely is unrealistic, since several processes contribute to creep and may change in their importance and influence through space and time. The driving forces of ice movement are tensile and compressive stresses induced by gradients at the glacier surface, bed and in the basal motion patterns (Kamb & Echelmeyer, 1986).

The more precisely we attempt to calculate velocity and stress distribution of a glacier, the more complicated the relationship becomes. In regard to deviatoric stress gradients, Blatter (1995) suggests a simple and useful algorithm for the calculation in grounded glaciers. This will not be applied in this dissertation but could be applied for more detailed investigations.

Glaciers move as a result of permanent strain of the ice and the glacier bed in response to stress. Strain occurs either by deformation of the ice, deformation of the bed beneath the glacier or sliding at the ice-bed boundary (Benn & Evans 1998). The movement that can be recorded at the surface of the glacier is the cumulative effect of these processes. The latter may act on their own or in combination. The deformation of ice, represented by strain rate, results from creep or fracture (cf. Chapter 4.3.5). As Weertmann (1983) and Alley (1992a) have shown, ice creep can be thought to occur if the ice is close to its pressure melting point (cf. the behaviour of metals) and will be affected by a complex set of interacting processes.

Bed deformation can, in some cases, account for a substantial contribution to glacier movement and has to be considered here. Where this is the case, rocks or sediments underlying a glacier will permanently deform in response to stresses imposed by the ice (Benn & Evans 1998). Since bed deformation can result in unusually high rates of sub-glacial debris erosion, transport and deposition (Boulton & Hindmarsh, 1987), it represents an important geomorphological process which could explain the occurrence of the large lateral-terminal moraine of Gígjökull. However, as it will be shown later, the underlying geology determines this process in particular. The relationship between glacial erosion and debris supply caused by subglacial deformation is particularly discussed in Boulton (1996).

Sliding (Weertmann, 1957; 1964; 1979) of glaciers plays another important part in glacier movement. Three different kinds of sliding occur at the interface of the ice and the underlying bedrock:

• Adhesion describes the fact that owing to the adhesive strength of frozen glacier beds the drag of the latter will be higher than average. These can support larger shear stresses than the overburden ice so that ice movement is more likely to occur by ice creep. Throughout the glacier bed, patches of frozen and thawed ground occur irregularly, for example on the lee side of obstacles where the melting point is higher than average, owing to lower than normal pressures (Robin, 1976). Nevertheless, the theoretical work of Shreve (1984) and Fowler (1986) suggests that, even at sub-freezing temperatures, sliding should occur, although with very low velocities (Echelmeyer, 1987).
• Regelation has been recognised as a second key factor in glacier sliding. Its importance has been described on theoretical grounds (Deely & Parr, 1914; Clarke, 1987a) and has been verified to be well founded by numerous direct observations that have been made in sub-glacial cavities (Kamb & LaChapelle, 1964; Hallet et al. 1978, Hubbard & Sharp, 1993). The influence of the waterfilm in particular is the focus of Hallet’s considerations (1979).
• Enhanced creep occurs in the vicinity of obstacles, owing to the fact that locally high strain rates at the upstream side of obstacles cause the ice to accelerate around bumps. From theoretical work and observations, it can be assumed that there is a ‘critical obstacle size’ at which neither enhanced creep nor regelation occurs (Kamp & LaChapelle, 1964). Lliboutry (1993) suggests that in temperate ice, regelation can contribute to enhanced ice creep and he demonstrates the strong interrelationships involved in flow mechanics. A comparison between these relatively small-scale observations and the influence of large-scale obstacles on ice flow can provide a better understanding of the role such bedrock obstacles play in glacier movement (Knight et al., 1994).

All of these processes might contribute to the ice dynamics of Gígjökull, though regelation and enhanced creep will play the most important part since the glacier can be considered to be temperate, which means the ice is near the pressure melting point throughout most parts of the glacier. This will affect the parameter e (strain rate) which is highly sensitive to temperature changes in the glacier.

Research over the last few years has increased our insight into the processes of glacier motion, but there is still no explanation for the precise conditions that control strain rates, i.e. glacier movement. Some parameters affecting ice-flow interact in a complex way, such as temperature, debris content of the ice, bed roughness and water pressure. An important influence on ice flow characteristics is provided by rock particles. However, the effect of debris on ice deformation is very difficult to establish, mainly because a great number of factors are involved (e.g. grain size distribution and aggregate orientation). Different studies have been carried out and demonstrate either increasing strain rates with debris content (Nickling & Benett, 1984) or decreasing rates (Echelmeyer & Wang, 1987). Therefore, I will use the distribution of debris on/in ice only as evidence for a qualitative description. [For further reading see also Andrews (1972b)].

Recent research attempts to give answers to open questions concerning certain types of anomalously moving glaciers (Clarke 1987b; Boulton & Jones 1979). Surges represent the extreme of such anomalies (cf. Kamb & Engelhardt, 1987; Kamb et al., 1985; Sharp, 1988). Glacier movement is not a linear process and varies spatially throughout the glacier. This means that these differences do not only occur in time (cf. ‘seasonal variations in surface velocity’ Hook et al., 1989) but also within the glacier itself. The variations within glaciers have been investigated in detail by Paterson (1994) who developed relationships for the calculation of flow velocity at the surface of glaciers. According to these relationships the drag at the valley sides will retard the motion of the glacier margins and the result will be that velocities will be greatest near the centre-line of the glacier, even of each cross-profile along the length of the glacier (Menzies, 1995). This will be analysed in Chapter 4.1.

Special importance must also be given to the effect of water pressure on glacier movement (Bindschadler, 1983). Recent studies have demonstrated that times of increased glacier motion correlate with elevated basal water pressure. This insight has been gained from observations of the water levels in boreholes which connect with the sub-glacial drainage system (Müller & Iken 1973; Iken et al. 1989; Meier et al. 1995; Jansson, 1995). More recently, sliding theories have been developed by Lliboutry (1968, 1987), Fowler (1987a,b), Kamb (1987), Schweizer & Iken (1992) and Willis (1995). The distribution and pressure of water at the glacier bed is a very important factor in the control of temporal variations in ice flow (Willis, 1995) and there could be a link to the theory of melt-water controlling ice-marginal sedimentation (Spedding 1997). Especially in the context of icefalls water pressure plays an important role in glacier dynamics. Therefore, velocity variations and basal drainage can be presumed to be closely linked. This concept has been confirmed by several observations, such as the observation that an elevation in water pressure promotes an increase of the strain rate in sub-glacial deforming layers. This is because frictional strength decreases at the same time.

Another important effect of the elevated water pressure is that the spatial extent of the cavities filled with water enhances basal sliding rates (Iken & Bindschadler, 1986; Boulton & Hindmarsh; 1987, Schweizer & Iken, 1992; Willis, 1995). This concept could explain areas of high velocities that do not show very steep bedrock slope.

In general, changes in basal water pressure, and thus in glacier movement, primarily depend on water supply from the glacier surface and on switches (i.e. a reorganisation) of the basal drainage system (Benn & Evans, 1998). The fluctuations in water supply from the surface can either correlate with ablation cycles (daily or seasonal) or be due to anomalous weather conditions, such as extremely long periods of rainfall or clear sky (as is often the case on Iceland during summer). If more water is temporarily stored in distributed drainage systems beneath the glacier such as a braided channel system, Darcian flow or linked-cavity systems, it critically influences flow velocities (Iken et al., 1983; Iken & Bindschadler, 1986; Kamb, 1987; Schweizer & Iken, 1992). According to our present knowledge, glaciers can be considered as systems displaying a positive correlation between discharge and water pressure (i.e. water pressure elevates as the supply to the system increases). Hook (1989) in particular evaluates the role of englacial and subglacial hydrology.

Willis (1995) points out that if a ‘switch’ from a distributed to a channelised network occurs during the course of a melt season, drops in water pressure beneath the glacier and reduced sliding velocities can be expected. Such switches are considered by Spedding (1997) and could explain glacier movement of Gígjökull to some extent but certainly not the high debris transport through the glacier.. Echelmeyer & Harrison (1990) suggest that surface melt-water produced during the ablation season does not flow straight through the ice to the bed, but flows in most cases in englacial conduits and therefore at higher levels in the ice. If this is the case, fluctuations in melt-water supply from the surface would not affect basal sliding but could account for high surface velocities. Water flow in englacial conduits could also account for high sediment transport.

Creep rates that are in general only influenced by shear stress and ice temperature may also be influenced by water pressure, although only indirectly because variations in sliding rates controlled by water pressure on one part of the glacier will cause variations in longitudinal stresses elsewhere (Hutter, 1983).

Finally, high ice flow velocities are often associated with glacier steepening, as is the case in icefalls and maybe at the beginning of over deepenings. Here, the glacier tends to accelerate, thereby thinning the ice (Benn & Evans, 1998). This effect can be observed at Gígjökull.

Further, glacial hydrology does not only explain variations in glacier flow (I will focus in this dissertation on the latter) but could explain the origin of the large lateral-terminal moraine of Gígjökull and hence help to link process and form. The glacio-fluvial transport of debris plays a major role in this context (Röthlisberger & Lang, 1987). The sediment transfer in turn highly depends on the structure of the subglacial drainage system and thus the surrounding topography (Sharp et al., 1989: 1993). Therefore, insight into the subglacial drainage system and a better understanding of the latter are not only important for explaining glacier movement but also to decide the implications for ice-marginal sedimentation (Sugden, 1984; Spedding, 1997). The high importance, and the albeit high complexity of the processes which control sub- and englacial drainage, have lead many scientist to focus particularly on the subglacial drainage system (Harbor et al., 1997; Hook, 1989; Meier et al., 1995; Sharp et al., 1989 and 1993; Spedding, 1997; Sugden et al., 1991; Röthlisberger & Lang, 1987).

In order to relate strain and velocity distribution on a glacier to the underlying bed topography and to calculate theoretical values for strain rates etc., some knowledge of ice thickness is required. Presently, the most useful method for establishing ice depth on glaciers and ice sheets is provided by radio-echo sounding (RES; cf. Chapter 3.3). Watts and England (1976) were some of the first to carry out radio-echo soundings on temperate glaciers. Comprehensive reviews of the principles of radar signal propagation and reflection can be found in Robin et al. (1969), Smith & Evans (1972) and Bogorodsky et al. (1985). Björnsson (1986) has applied this method to map the surface and bedrock topography of icelandic ice caps and has expanded the work on the thermal regime of sub-polar glaciers in general (Björnsson et al., 1996). The information which can be obtained by radio-echo sounding goes beyond the mere ice thickness but can give further information, such as internal layers of ice sheets, water lenses, debris nests, etc. (Yoshida et al., 1987).

1.2 Site description

1.2.1 Location

Gígjökull is one of the two biggest outlet glaciers of the small volcanic ice cap Eyjafjallajökull in Southern Iceland (63°40’ N; 19°37’ W) (Figure 1.1, Fig. 1.2 and Appendix 1). It is located approximately 150 km ESE of Reykjavík, at about the same longitude as the Vestmannaeyjer Islands and flows to the north into the Markarfljót-Valley. Gígjökull is fed from the ice cap that covers the central crater of the dormant volcano, Eyjafjöll. Over the course of its 6,5 km length the glacier descends ~ 1.450 m and covers an area of roughly 8,4 km². Most of the middle section of the glacier, i.e. between the crater lip and the relatively low glacier terminus, is dominated by a large icefall. Its properties are given by Spedding (1997, p. 154) as follows:

"~ 3.000 m in length, height drop ~ 1000 m, mean angle 16°, perhaps 25°+ at its steepest." The glacier terminus consists of a relatively gentle lobe ending in a lake called Lake Lónið (cf. Björnsson, 1976). This small lake is trapped between the terminus of Gígjökull and the unusually high lateral-terminal moraine (~ 74 m high). As Spedding (1997) suggests, this lake can be regarded as an indicator for the existence of a terminal overdeepening that has partially been exposed since Gígjökull’s retreat from its Neo-glacial maximum extent (19^th century). Since 1900, Gígjökull has undergone several distinct phases (Sigurðsson, 1992 cited in Spedding, 1997, 154):

• prior to c. 1960 it retreated
• between 1960 and c. 1990 it advanced again
• since 1990 it has been relatively stable but since 1997 a clear retreat has been observable (own observation).

It seems as if Gígjökull has been undergoing glacier fluctuations with intervals of approximately 15 years (Bob McCulloch, personal communication 1998). Gígjökull retreated several 100 m by 1947, as can be seen on the aerial photograph of 1947: the bedrock of the area is presently covered by ice and is most likely affected by the processes of overdeepening.

Figure 1.1
Iceland with its ice-caps, and the study site in Southern Iceland (Eyjafjallajökull and Mýrdalsjökull; cf. Geomorphological map Appendix 1.)

Figure 1.2
Gígjökull: extract from sheet 1812: III Eyjafjallajökull. Scale 1:50,000. Copyright: Defence Mapping Hydrographic/ Topographic Center, Washington DC, USA/ Iceland Geodetic Survey.                                                                                          

Figure 1.3
Geology: Extract from the Geological map Ísland 1:500,000. Legend overleaf.

2.2 Geology

The mountain ice-cap Eyjafjallajökull lies on a Pleistocene volcanic massif (Eyjafjöll) (Björnsson, 1979). The volcano underneath the ice cap last erupted in 1821-23, the second of only two historical eruptions, Dugmore (1989). Evidence for this last eruption can be found abundantly beyond the lateral-terminal moraine of Gígjökull, owing to the flood that resulted from the eruption. Geologically, the high cliffs which rise to both sides of the icefall consist of sequences of lava and pyroclastic flows (Spedding, 1997). Figure 1.3 shows the geology of the study site and its surroundings.

1.2.3 Climatic regime

Gígjökull is located at about 63° N in Southern Iceland and is thus influenced by a maritime regime. According to Gläser and Schnütgen (1986),Vík and Mýrdal (20 m a.s.l.) have an average temperature of 5,7° C (1931-60) (an amplitude of 10,1° per year). The annual average precipitation is 2256 mm (245 days of rain per year). In general, the South West of Iceland has moderate wet-coolish conditions with highest annual average temperatures being 3,7-5,7° C and the annual amplitudes, due to the maritime location, not exceeding 8,9° (Vestmannaeyjer). Nevertheless, the precipitation rates on Mýrdalsjökull (850 m a.s.l.) are much higher than those at the coastline. On Mýrdalsjökull, the ice-cap to the east of Eyjafjallakökull, the highest precipitation rates in all of Iceland are recorded: 4000 mm/a (Gläser & Schnütgen, 1986). Appendix 3 shows the mean annual precipitation rates.

The period 1931-1960 is thought to have been the warmest during the past 2.500 years (Schutzbach, 1985). Since 1965, the weather conditions have deteriorated again, and the winter of 1965 was the first ‘ice winter’ for many years (drift-ice at the north coast!) followed by similar conditions in 1968 and 1969.

Winds come predominantly from the South and from the Southwest, bringing air masses with high moisture content that are then trapped by the mountains in the South and give rise to high precipitation rates on Eyjafjallajökull and Mýrdalsjökull. It can therefore be estimated according to Schutzbach (1985), that Gígjökull receives at least 3200-4000 mm rainfall per year. Comprehensive data is not available however. The exposure to onshore winds from the South-Southwest results in high accumulation and ablation rates and thus the total turnover can be presumed to be higher than average. Rist (cited in Spedding, 1997) gives some estimations as to the amount of snowfall on the neighbouring Mýrdalsjökull. Ablation averages of Spedding (1997) in August and September 1993 were ~ 6 cm/d. My own records of July 1997 and July 1998 revealed: ~ 7 cm (Table 4.5)

In conclusion, Gígjökull is a temperate glacier, meaning that ice below the pressure melting point is not a factor here (Shreve, 1972) - a fact that will be of more importance at a later stage. Finally, the equilibrium line of Eyjafjallajökull lies approximately at 1.100 m a.s.l. (Björnsson, 1979). However, it has to be mentioned that Björnsson’s AAR of 1,7:1 works out an ELA at ~ 1.150-1.200 m a.s.l. (Spedding, 1997)

Figure 1.4a
Looking south from beyond the glacier snout. On top: the ice-cap Eyjafjallakökull. Gígjökull as outlet-valley glacier flowing into the Márkarfljót-Valley calving into Lake Lònið (foreground). Clear view of the two small lateral-terminal moraines (ice-cored) and the present ice margin, as well as the rampart/ ‘gap’ at the glacier snout.

Figure 1.4b
View from Gígjökull’s western Neoglacial moraine (74 m) where the EDM total station tripold was positioned. Clear view on widespread NE-tephra-outcrop and radial compressional crevasses, as well as large spreads of relict conduit debris (left). Also to be seen: ‘Christmas-tree’-tephra loop (in foreground right).

Chapter 2. Conceptual rationale for the investigation of overdeepenings

2.1 The "frame"

Since overdeepening might have a critical effect on the routing of basal meltwater, and hence sediment supply towards the ice margin, the large moraine at the terminus of Gígjökull might be strongly connected with the process of overdeepening, or even exist as a result of it (Spedding, 1997). Besides influencing the sub-, en- and supra-glacial drainage system - mainly the sub-glacial one - overdeepening must also influence ice flow since glacier motion is strongly connected with topography and climate (Benn & Evans, 1998; Summerfield, 1991; Sugden, 1976). Therefore, properties visible on the glacier surface, such as crevasses and meltwater channels, can be expected to reflect the underlying bed topography to some extent.

The first aim of this work will be to quantify the impact of the bed topography on ice flow, i.e. stress and strain patterns, present within the ice as a result of a variety of other parameters which will be considered in more detail in chapter 3. As overriding aim, the analysis of calculable properties will lead to an evaluation of the hypothesis of overdeepening, and an attempted application of the insight gained into the processes of modern glaciers to historical implications. These two major overriding concerns will provide a framework for my work:

1. to link process and form, and
2. to show that change does not only occur through space but also through time.

2.2 Scientific rationale and significance of this work

The scientific rationale for the detailed investigation into glacier dynamics can be found in the consideration of the implications of glacier fluctuations for the landscape as well as for the glacier itself (e.g. positive feedback). First, the understanding of the behaviour of modern glaciers, which can be gained, for example, by investigating the topography-hydrology-ice-flow relationship, allows one to explain how specific landforms such as the terminal moraine of Gígjökull are produced, and what implications ‘active’ glaciers have for these land forms in the short- as well as the long-term.

Secondly, the understanding of such changes has historical implications. In particular, three different aspects can be considered:

1. The circumstances in which old developments have taken place,
2. the wider relevance for similar topographic situations; that is for plateau and valley terminating glaciers, (mainly those flowing into the same kind of soft sediment),
3. the sensitivity to change, which is by far the most difficult, yet most challenging, point to derive from all this.

At first sight, Gígjökull seems to display an unusual phenomenon: a huge terminal moraine. This is surprising because the glacier does not belong to the type of glacier one would expect to produce such a big moraine. As has been suggested by Spedding (1997), this could be the result of an overdeepening beneath the terminal lobe of the glacier. A switch in the routing of basal and englacial meltwater, resulting from the process of overdeepening, could increase the amount of sediment that is pushed towards the ice margin where it would then accumulates.

However, the effect of this overdeepening on glacier flow at the terminus has not yet been investigated. There is also no explanation for the differences in ice morphology on the glacier surface beside the changing bed topography. However, the changed routing of meltwater, and hence enforced sediment supply might also influence the way in which the ice above moves downwards, for example by bed deformation of sub-glacial sediment sliding on water saturated sediment or even ‘hydraulic jack’ (Summerfield, 1991; Spedding, 1997; Röthlisberger & Iken, 1981). A significant part of ice flow might thus be caused by the deformation of sub-glacial sediment, or by sliding on water-filled cavities.

If sub-glacial and englacial meltwater and fine debris facilitate ice flow, there must be a strong correlation between areas of high surface velocity (and thus strain) and the sub-glacial and englacial drainage system and debris flows. It is important to consider sub-glacial meltwater and its debris content, in addition to regional parameters such as topography and climate (Sugden & John, 1976). Here, ice radar can provide necessary information, albeit indirectly (cf. Chapter 3.3 and 4.2).

Additionally, the near-snout overdeepening must affect the longitudinal ice flow velocity at the glacier terminus, since increasing bedrock gradients enhance ice flow velocities and thus compressive flow at the glacier terminus. This in turn facilitates the process of overdeepening. Longitudinal stresses are likely to vary critically since ice slowing down causes longitudinal stresses to be compressive and accelerating ice to be tensile.

A positive feedback could well exist in which more debris would be transported towards the ice margin, where it would be deposited and would build up a moraine, such as the one at Gígjökull. This would in turn constrain the ice. This would mean that, given an ice advance due to climatic change, the first reaction of the glacier would not be a longitudinal extension in the first instance but a thickening of ice. This in turn would enhance the englacial and sub-glacial water pressure and thus increase the debris supply to the ice margin which would then support further overdeepening (due to sediment accumulation) and ice thickening in the pro-glacial area (i.e. thickening of the sediment). It can therefore be presumed that, due to a steep bedrock slope, as well as the barrier effect of the moraine, the terminus represents an area of high compressive flow.

In the case of such a thickening, it can be presumed that the whole system would, at a certain stage, reach a threshold at which the system collapses; to reach a new steady-state the glacier would therefore have to break through the moraine. However, evidence of historical advances of Gígjökull (e.g. 1740) show that the small channel which was cut in the terminal moraine allowed sufficient drainage, and that the moraine itself could withstand the stress of the overlying ice. The current knowledge of ice-flow is already very precise (cf. Chapter 1), but the implications of the given situation, that is an "ice-fall/overdeepening regime" as suggested by Spedding (1997, p. iv), have not yet been investigated in detail. This study therefore explores the influence of near-snout overdeepening on glacier flow. It is ice flow that best reflects changes in bed topography. If the analysis of the different data sets gained during the fieldwork on Gígjökull in 1997 and 1998 gives further evidence or even proof for the argument of over-deepening beneath the glacier terminus, Nick Spedding’s thesis could be the starting point (‘the spring-board’) for further research.

2.3 Specific questions: So what?

I will briefly outline the specific questions that are raised by the problem based on the background theory of glacier dynamics (Chapter 2.1) and the overall aims described above:

1. Can the influence of near-snout overdeepening on ice flow be quantified?
2. Can differences in flow velocity between near-snout areas and mid-glacier areas be recorded?

If one assumes that stress, strain rate and flow velocity are strongly linked to the underlying bed topography, i.e. if a steeper slope (at bedrock level), as is the case for an icefall/overdeepening, correlates with an increasing stress and flow velocity and if the existence of an overdeepened area can be confirmed at the same time, it must be possible to quantify the differences in ice flow. Flow patterns could thus be inferred.

In a further step, three more intriguing questions will be tackled:

3. What is the influence of meltwater?
4. Is there a correlation between areas of high flow velocities and those expected to be affected by certain drainage patterns?
   And, closely linked to this question:
5. Do areas of high strain, i.e. flow velocities, correlate with those areas from which I would presume englacial meltwater to be squeezed out to the ice margin?

While focusing on the first two questions, the following set of three will inevitably have to be borne in mind:

• What factors affect ice flow in general?
• How and why does ice flow vary?
• What are the links with other glacier processes?

In order to give answers to these questions, the overall flow pattern has to be examined, stress and strain rates have to be calculated and analysed and ice velocity has to be correlated with the underlying bed topography. Surface properties such as crevasses also have to be considered. In the following chapter, the field methods applied are discussed separately; however, in order to give a brief overview they are listed below:

1. Flow will be measured by using E.D.M. measurements of stakes placed into holes on the glacier;
2. deformation will be recorded by analysing strain-net data;
3. ice radar survey will provide ice depth in order to calculate stress and strain rates and to back-up the idea of a near-snout overdeepening, which then can be related to strain-net recordings;
4. surface slope angles have to be measured in order to calculate strain rates;
5. ablation will be measured to adjust the flow records and to infer the local climatic conditions;
6. last, but not least, aerial photographs will be analysed and a geomorphological map sketched.

Chapter 3. Field methods

3.1 Flow

The data required for the analysis of flow was primarily that of the rate of flow over most of the ablation area of Gígjökull. To measure flow, different shaped strain nets were placed on the ice at various sites during the fieldwork in July 1997 (Figure 3.1 and 3.2). Each strain net consisted of either 3 stakes arranged in a triangle whereby two triangles often had one stake in common, or of 4 stakes arranged in a diamond. In 1997 four sites were chosen for analysis:

The stakes were placed in holes created with an ice drill and were then numbered from
1-20. The sites were chosen by analysing aerial photographs from 1947, 1989 and 1994. The tephra exposure on these photographs indicated a non-uniform flow across the glacier; in particular the western side seemed to exhibit diverse and complicated flow patterns. Besides this, it appeared from aerial photographs that ice flow would diverge at this point. To make sure that the stakes were placed approximately equidistant a measuring tape was used; the distance between each stake was ~ 30 m.

The most accurate technology for determining a specific surface location is optical techniques such as Electronic Distance Measurer (E.D.M.), although logistics on small valley-glaciers are sometimes difficult. The initial position of the stakes was measured by using a Geodometer (model 400) total station as an E.D.M. The error incurred when using the geodemeter is two parts per million (2 mm km^-1). The E.D.M. was positioned on the

Figure 3.1 Study site on Gígjökull with the location of the strain nets (Site 1-4) and the two longitudinal ice radar profiles A and B. The outcrop of the Hekla-1947-tephra outcrop (H1947) and the routing ways of the supraglacial meltwater channels ( ) as well as the most important moulins (¤ ) are sketched. Note: The ‘Christmas-tree’ like shape of the tephra layer around site 3 and the englacial meltwater channel observed beneath site 3. Two supraglacial meltwater channels, running parallel to the centre-line, seem to drain most of the glacier. Some of the water disappears in big moulins (¤ ) on each side. On the ‘rampart’ at the snout, only two shallow channels drain supraglacially into the lake. Downglacier from approximately the height of the uphill end of the two lateral moraines conduits are concentrated (ð ï ). The two ‘active’ moulins are located at the same height. E and W represent the two meltwater channels found on both sides at approx. the height of half way between the weather station and the lake and could be a continuation of the supraglacial meltwater streams. Additionally, the ‘moulin of death’ (Ä) and its canyon-like trench ( ) are shown at the eastern flank. In front of the glacier snout the debris ‘barrier’ (db) can be seen. ^ indicates the position where water was forced out of the glacier by hydrostatic pressure in fountains of 1-2 m height.

adjacent westerly lateral-terminal moraine (~ 74 m height) in order to measure the exact position of the poles in three dimensions: East, North, and Altitude. Owing to the way in which the E.D.M. sites its co-ordinates, the following have to be distinguished:

• Eastings (Y-axis) represent the longitudinal co-ordinates of an International Grid system giving the exact position on a grid referenced map of Iceland. This grid coincides with the Icelandic equivalent of Ordnance Survey by setting the E.D.M. to fix a position which coincides with this grid reference and to measure all subsequent points from this axis. (= distances across the width of the glacier).
• Northings (X-axis) represent the latitudinal co-ordinates in the same grid systems as above and provide, together with the Eastings, the exact location of the marker poles (= distances along the longitudinal axis of the glacier with increasing values the further down the glacier one travels).
• Altitude is given as the height above sea level at which the marker pole is positioned.

Northings, Eastings and elevation readings are all related to a set of arbitrary co-ordinates, in this case set at (2000,2000,2000). This system is designed in order to avoid negative values whereby the E.D.M. itself is positioned at the co-ordinates (2000,2000,2000), and its exact position in the landscape may be found using differential G.P.S. positioning. It is important to note here that E.D.M. measurements show surface slope only. Assumptions have to be made using climatic data concerning the nature of basal flow in such an environment as that in which Gígjökull is situated.

In July 1997, the position of each pole was monitored over a sequence of 7 days: 28^th June, 2^nd July and 4^th July 97 (Appendix 4). Only measurements from the beginning and the end of the period of record are needed, in order to measure the total movement during a specific period. Readings from the middle day provide a useful check and were meant as a safe-check in case the weather conditions were too bad to take readings on the final day.

By comparing the position of the marker poles from one measurement to the following, a down-glacier movement can be seen. Graphing the relative movements of the stakes in each strain net revealed the flow patterns. To ensure uniformity between the readings, the prism staff was always held tightly against the stake on the down-glacier side.

Figure 3.2 Position of strain nets, Long Profile A and B, and Cross-profile e in relation to each other.

Additionally, four transverse sets of poles were set by Barnes et al. (1997) to study ice flow velocities across the width of the glacier (Figure 3.3 and Appendix 6a).

During fieldwork in July 1998, similar procedures were carried out (Wise & Stewart, 1998). This time transects were spread at approximately equal distances down the glacier in order to compare higher and lower areas of the ablation zone, as well as to investigate the conditions in the ‘over-deepened’ area in the lower section. In addition, a longitudinal section was set up near the centre line to measure flow down the middle of the glacier. Finally, another set of strain nets was concentrated at the western flank, as this area was presumed to have undergone a high degree of deformation. Results of these measurements were only used to compare with my own records.

Figure 3.3 The initial position of all transverse transects (Barnes et al. 1997).

3.2 Deformation

In order to monitor deformation rates within the (surface layer) of the glacier, strain nets were set up in 1997 to measure flow velocities (cf. 3.1 Flow) and the movement of each stake in relation to the other stakes of a single strain net (Appendix 7a, b, c). In 1998, a further set of 18 poles (18-35) was set up as strain nets in a triangular form, whereby different scales of nets were expected to allow a focus from larger areas to smaller ones so that deformation of both high and low extent could be revealed. As in 1997, the smaller scale triangles had one common point due to the limited number of marker poles available. Again, measurements were taken using the E.D.M. The locations of the strain nets set up in 1997 are shown in Figure 3.2, those of Wise & Stewart (1998) are not graphed separately since the data will only be used for comparison.

Having established the ice thickness, stress can be calculated according to the equation:

where t is stress, r the density of ice (c. 900 kgm^-3), g gravitational acceleration (9,81 ms-²), h the ice thickness (m) and a the surface slope of the ice for small bed slopes and comparable. For h = 100 m, the normal stress would be c. 882,9 kPa, and for h = 100 m and a = 6° the basal shear stress would be c. 92 kPa.

The equation above demonstrates that stress increases with ice thickness (h). To understand glacier dynamics it is important to know that from the surface towards the bed, normal and shear stress increase linearly. As the equation for shear stress shows, measurements of slope angles are required too (cf. 3.4 Surface slope).

In general, flow is best described by Glen’s Flow Law (1955) adopted by Nye (1957). The strain rate¹ may be calculated according to the equation shown below:

where e ´ represents the strain rate, A and n are constants, and t is the shear stress. A is a constant related to the temperature of the ice and the effective shear stress and n the exponent of a mean value of 3.

The factor A is highly sensitive to changes in the ice temperature. Since Gígjökull is expected to be a temperate glacier (i.e. ice temperature is near the melting point) a value of 6,8^-15 s^-1 kPa^-3 suggested by Paterson (1994) for 0°C will be used for the calculation of strain rates. Glen’s Flow Law is a robust model that can be widely used (albeit only for uniaxial shear). It shows that the strain rate is highly sensitive to the shear stress. Differences of field records from laboratory results may be due to the orientation of crystals in ice or impurities (Benn & Evans, 1998).

3.3 Ice thickness

In order to measure ice thickness, radio-echo sounding (RES) was used. The concept of RES is based on the fact that electromagnetic waves can travel through ice and will be reflected from bedrock (Appendix 12). Comprehensive reviews of the principles of radar signal propagation and reflection can be found in Robin et al. (1969), Smith & Evans (1972) and Bogorodsky et al. (1985). Knowledge about the ice thickness was required for three reasons:

Given that electromagnetic waves travel through ice at a specific rate, ice thickness can be calculated from the travel time, i.e. the difference between the time of transmission and reception. The plotting of a sub-glacial bed topography requires multiple depth soundings which should follow profiles across the glacier. The accuracy of this constructed topography heavily depends on the number of readings taken, and thus, on the degree of interpolation needed. Furthermore, the quality of the derived visualisation is influenced by the radar frequency and the accuracy of field methods and data processing.

To understand the way in which RES works, the first thing to know is that radio waves travel through ice at a rate of 1,69 x 10⁸ ms^-1. The travel time is recorded, i.e. the time which elapses from the transmission of the radio pulse to the reception of the electromagnetic signal bouncing-back from the glacier bed. Once the readings are taken, the data is down-loaded into a computer. Thereafter, ice thickness can be calculated by applying the equation:

z = ut₂ / 2

where z is the ice thickness, u is wave velocity through ice, and t₂ is two-way travel time.

To position the measurement location precisely, the Magellan GPS were used and the records linked with the ice radar readings. Hereby, the GPS was able to locate the position by taking an average from positions logged over a three minute period at each site. This record was corrected using differential GPS from the GPS constantly operating from the fixed station, the Community Centre of Heimaland (exact position known). The accuracy of the GPS data mainly depends on the selective availability and variability of the number of satellites available to the system.

The correct interpretation of the reflected signal is limited by characteristic properties of ice (e.g. dielectric properties), especially in temperate glaciers such as Gígjökull (Welch et al., 1998). In regard to these properties, it has to be mentioned that temperate glaciers are a heterogeneous substance and water may be present due to ice at the pressure melting point and surface melt-water in various forms such as supra-glacial channels at the ice-bedrock interface and in englacial lenses. Since radio waves are reflected from water (as it is from bedrock) it is difficult to determine the origin of the received signal. Furthermore, water can scatter or absorb electromagnetic waves which makes the interpretation of signals more difficult (Robin et al., 1969; Smith & Evans, 1972).

Another important property of temperate glaciers is debris. Debris is incorporated sub-, en- or supra-glacially in the glacier. Yet its influence at the site of interest, Gígjökull, might be of minor importance with the exception of tephra, an acidic ash from eruptions from the neighbouring volcanoes, which is, once deposited on the glacier surface, reworked and embedded in the form of layers within the glacier. Just as water scatters and reflects the radio waves, so does debris, and it may therefore attenuate the radar signal. Although changes in wave velocity will cause incorrect calculations of ice depth, the effect will be negligible when compared to the resolution of radio-echo sounding (Welch, 1996). However, knowing about the effect of englacial debris or water lenses to the readings can be of great importance to the interpretation of glacier dynamics.

Further limitations to successful data collection are: the most appropriate placing of the radar equipment, which is subjective, and the fact that accuracy is reduced in highly crevassed areas due to the complex ice topography and the possible proximity to the ice margin in some crevassed regions. This proximity, owing to the narrow valley shape of Gígjökull, causes the radar pulses to be reflected off the rock walls and the bed. Distinguishing between the reflected signals in order to determine the actual bed surface is therefore no easy task (Appendix 13). If bedrock slopes are very steep the signal will not even be bounced back directly to the receiver. The accuracy of the surface location therefore determines the validity of the plotted bed topography, and the final resolution of a subglacial bedrock map does not only depend highly on the wavelength of the radar system, but also on the data spacing in the field and the migration routines (Welch et al., 1998).

Having established the ice thickness, shear stress and strain rate can be calculated according to equation 3.1. and 3.2 (see Chapter 3.2).

In June/ July 1997, transects and profiles were spread over the glacier. Particular attention was paid to the creation of a long profile down the glacier whereby, for the sake of efficiency, it was desirable that the long profiles passed through the strain nets in order to analyse the topography above and below the strain nets and to simultaneously obtain data from which the strain rate could be calculated. Figure 3.1 shows the orientation of the two longitudinal profiles A and B, whereas Figure 3.2 gives the exact position of the ice depth profiles in relation to the position of the strain nets, and Figure 4.1 the exact GPS position of all ice radar profiles.

One longitudinal profile was therefore established along the assumed main flow line, visible as a ridge in the middle of the glacier terminal lobe which passed through strain net 1, 2 and 4 (Profile A) (Fig. 3.1); and another profile parallel to the first, but shorter, on the western flank of the glacier, passing strain net 3 where the tephra outcrop showed the "Christmas tree-shape" (Profile B) (Fig.3.1). This latter profile was specifically set up to observe flow in order to give an explanation for the folding of the tephra layers through which it passed. In co-operation with I. Barnes, M. Downey and S. Wood, eight transverse ice radar profiles were produced at distances of 50-100 m across the width of the glacier.

In 1998, ice radar records were taken at every transect point and at two markers of each strain net, as well as at additional points which were located by using the E.D.M. Although, due to glacier movement, these stakes were not at fixed positions during the observation period, their readings taken on Day e can be taken to be representative since electromagnetic waves propagate spherically and the resolution of the ice radar therefore reflects ice thickness over an area on the bed (~ 10 m²), not from a point (Appendix 12).

This procedure allows the plotting of a rough (interpolated) underlying bed topography, which can be compared with the surface topography and its patterns of ice flow. Nevertheless, to get reasonable results from the radio-echo sounding, the data should be compared with other data in order to infer the correct interpretations (e.g. borehole depth data, surface morphology such as crevasses, and the surrounding topography).

In terms of the measuring error which has to be taken into account, the ice radar works relatively precisely. The fluke (receiver) resolves travel time distances of about 0,04 m s * 168 * 10⁶ ms^-1 = 6,72 m if we consider two-way traveltimes. This means that the best feasible vertical resolution is about half of this, i.e. 3,5 m. This accuracy is based on the assumption that we are able to pick ‘exactly’ the right point on the digital waveforms when processing the data. The slightest deviation from this point results in higher measuring errors (e.g. about 7 m). If we therefore take 10 m as vertical resolution we lie well within the maximum range of the possible error. In addition to the ice radar error, we have to consider the error that occurs when using the GPS. The best resolution we can get with differential GPS in regard to elevation is about 10 m. The total error can therefore be estimated to be less than 20 m whereby the best achievable value from the instruments is 13,5 m.

3.4 Surface slope & ablation

Slope angles are required in the above equations in order to calculate strain rate. In order to get the most accurate measurements, slope angles were calculated in 12 positions; 3 positions spanning each strain net. It is sufficient to calculate the approximate slope angle over the area of interest. One person therefore walked 50 m down slope from each marked position, whilst another person walked 50 m up slope. The steepest slope angle at each point was roughly estimated approximately in the direction of flow. A bearing was taken from the lower person to the upper person, and 23 degrees subtracted in each case in order to correct for true north. To guarantee a representative result, both an Abney level and a clinometer were used. Alternatively, trigonometry can also be used to give surface angles from plotted co-ordinates for the strain nets.

Further measurements of the surface slope were taken in 1998. Nevertheless, it is important to note here that this data can only be correlated with other data sets of 1998 since the surface has critically changed from 1997 to 1998. Nevertheless, in 1998 the glacier surface slope was calculated in the direction of flow and in the direction of maximum ice surface slope where the two were not equal. Main data was recorded at the poles 1-35 of the Flow Group. In general, the quality of measurements of slope angles carried out in this way depends on the (subjective) interpretation of the flow direction and the consistency or error in the use of the Abney level.

As an aside to the investigation on ice flow, measurements of ablation around the stakes^² were taken. The procedure for recording ablation was straightforward: When the poles were drilled into the glacier, the level at which the pole met the ice surface was marked in pen on the pole. In 1997, the first mark was placed on the 29^th June and the final mark five days later on the 4^th July. Measuring the difference between the previously placed mark and the new level at which the pole met the ice surface gave the total amount of melting around the poles within five days. The E.D.M. measurements change in elevation of the aforementioned points, but true elevation change of the ice surface can only be given once it has been understood how much is being removed by ablation. Therefore the vertical component of ice flow could be calculated by subtracting the ablation records from the change in elevation recorded by using the E.D.M.

3.5 Crevasses and debris distribution

In regard to the most visible features reflecting glacier movement, crevasses, I have produced sketches in the field, as well as a maps on the basis of aerial photographs, showing the crevasse pattern (see Chapter 3.6) in order to correlate the latter to the underlying topography. However, I will include some more quantitative records I have collected in co-operations with Coppins, Foy & Kemp in July 1998. Additional radio-echo sounding was carried out in the most crevassed regions of the glacier. However, the results will not be presented here.

Crevasses are one of the most eye-catching features of the surface morphology of glaciers. They can be regarded as vertical cracks of different size and shape. Usually their width ranges from a few millimetres to several metres. Crevasses can be expected to occur in areas of high strain rate where the ice can not adjust to the tensional stresses which exceed the strength of the ice by creep (Sugden & John, 1967). They are usually oriented at right angles to the main direction of stress within the glacier. Crevasses are therefore useful indicators of stress patterns in a glacier. Nye (1952) distinguishes three types of crevasses that are illustrated in Appendix 17 and modified according to Sugden & John (1976) and Paterson (1994).

Since crevassing is expected to occur in areas of high strain rate (cf. Vaughan, 1993; Hubbard et al., 1998), crevasse properties such as width, depth, orientation and type will be analysed in order to determine average zone values. The glacier terminus has been divided into Zone A to F anti-clock-wise, i.e. from the eastern side around the snout to the western side of the glacier. I will restrict the presentation of the data to a table in which the most important properties of each zone are displayed, and give a rather qualitative description of the crevasse patterns (see Chapter 4.3.5).

In addition to ice flow parameters, debris will be considered as far as previous work (Spedding, 1997; Moore, Porter & Thornton, 1998) about origin of debris, shape and site of deposition show comprehensive results and as far it could be observed on the glacier by myself.

3.6 Mapping of the ice surface morphology and aerial photographs of 1947 and 1989

Eventually, a qualitative interpretation of the ‘visible’ properties on the glacier shall be provided. Direct observations and sketches in the field are very important tools and essential methods for geographic examination and worth the effort. They allow us to give precise statements about features and processes in the landscape which are characteristic for a specific study site, and can help us to explain anomalies, as well as giving evidence for phenomene we cannot proof directly, either due to the lack of appropriate equipment or natural restrictions (e.g. observations of/in englacial conduits). Therefore aerial photographs from 1947, 1989 have been analysed, and crevasse patterns, supra-glacial melt-water channels, closed-up conduits, moulins, debris bands and tephra outcrop have been mapped. The result is presented in Figure 4.3. Their location on the glacier as well as their characteristics will be evaluated (see Chapter 4.).

Chapter 4. Results

4.1 Glacier flow

4.1.1 Strain nets

The comparison between the four strain nets set in 1997 reveals a glacier movement similar to that which had been expected with only a few exceptions. These will be shown in the following section. The movement of the poles of each strain net over a period of six days is presented in form of graphs in Appendix 7a and b as well as in tabular form in Appendix 4a and 5a. Table 4.1 shows the total movement of the poles in each direction as well as the flow rate per day. Since strain net 1 is located the furthest up-glacier, and therefore the closest to the equilibrium line (EL), the highest flow rates can be expected to be there. As Table 4.1 shows, this is the case. The total movement of site 1 downslope over the period of 6 days was considerably greater than a metre, although the individual movement of the stakes was not always as expected. Thus, pole 2 and 4 did not spread laterally as had been expected for a longitudinal compression alongside with transverse extension (Appendix 8). However this could be due either to measurement in accuracy (since the difference in the total movement of pole 2 and 4 was less than 8 cm), or, as discussed below to bedrock undulations which are reflected on the ice surface, or even to the beginning of the overdeepening coupled with bedrock steepening and thus increasing tensile stresses and hence acceleration in ice flow. Only further observations over a longer period of time can resolve this problem. In summary, strain net 1 exhibits a downhill movement towards the eastern side of the glacier.

Site 2, as site 1, was located near the centre-flow-line. It also shows high flow velocities. Since strain net 2 was set up further downglacier, and thus farther from the EL, flow could be expected to be less than at site 1. Table 4.1 and Appendix 9 exhibit a total movement of each pole of just under a meter. However, it can be noted here that neither extension nor compression were significant. Only the sideways shift as observed at site 1 was significantly high with a value of ~ 0,5 m (cf. Appendix 9).

The reason for the choice of site 3 proved to be justified since it revealed an interesting flow pattern. At first sight, the greatest longitudinal movement of site 3 occurred in the eastern triangle of the strain net (~ 90 cm/ pole 15) (Appendix 10d) and at the top pole 9 (~ 80 cm) (Appendix 10a). Whilst most stakes have moved eastwards, i.e. towards the centre line of the glacier, two poles, namely 10 and 12, have moved westwards, i.e. towards the margin of the glacier (Appendix 10b, c). In general it can be said that the transverse shift was less significant than that of site 1 and 2. Viewing the individual triangles it becomes obvious that strain net 3a exhibits extensional flow in longitudinal as well as transverse direction (Appendix 10a) whilst strain net 3b shows compression in both dimensions (Appendix 10b) which could be explained by the proximity of the margin and thus the valley sides which enhance the basal friction. The influence of the vicinity of the margin is also revealed in the flow pattern of strain net 3c which seems to have rotated travelling towards the north-east, i.e. the middle of the glacier (Appendix 10c). Triangle 3d, the one closest to the centre line, demonstrates slight longitudinal compression and shift towards the east, both being of little extent (Appendix 10d). In comparison to ice depth data (Table 4.3) we can also question if there is not a large hollow in the bedrock extending towards the centre-line, which would account for the stretching of strain net 3.

The flow and deformation of site 3 sketched above and shown in Table 4.1 and Appendix 10 demonstrate the complicated pattern of ice flow at the western side of the glacier. However, the complex folding of the tephra outcrop across which strain net 3 was placed cannot be explained by this flow pattern. Rather, it indicates that this significant deformation must have occurred further up-glacier and is not the result of a present deformation in situ (cf. 4.3.3 Hekla-1947-tephra layer).

Finally, at site 4, the flow velocity was the lowest of all strain nets to the north-east owing to the fact that it was located the farthest from the ELA, almost at the terminus of the glacier. Looking at Appendix 11 hardly any movement can be seen, despite the high resolution of the graph. Therefore, little can be concluded in terms of flow patterns.

4.1.2 Transverse transects

Courtesy of Barnes et al. (1997) an additional data set could be analysed. This consists of four transverse transects spread across the glacier in order to measure variations in surface flow velocity across the width of the glacier. The initial position is plotted in Fig. 3.2 and Appendix 6a. As expected all stakes moved markedly downglacier, i.e. to the North East (cf. Fig. 3.3b and Appendix 6b). An exception was the movement of stake 9 which moved to the north west. This is probably due to measuring errors. The connecting lines in Fig. 3.3b represent the total movement (distance and direction) of each stake which is also displayed in Table 4.2 and Appendix 4b. As Table 4.2 shows the average flow rate per day was about 0.19 m. The most even movement could be observed in the third transect from the top. If we compare this finding with the ice radar recordings of 4.2, we notice that there is a close relationship between high surface velocity and bedrock steepening though this will not be strongly perceptible in Fig. 4.7 due to the low resolution.

As described in Chapter 1.1, strain may occur either by deformation of the ice, the underlying bed or sliding at the ice-bed interface. The deformation of the strain nets I have outlined in Chapter 4.1.1 and the flow measurements of the transverse transects reflect the cumulative effect of all these processes.

In general, we can note a significant decrease in flow velocities as we move donwglacier, i.e. the farther the reading was taken from the equilibrium line. The high rate at stake 8 (Table 4.2) seems to reflect a measuring error and has therefore not been plotted in Fig. 4.1.

Figure 4.1 Total movement of all transect stakes 1-21 (28^th June to 4^th July 1997).

Figure 4.2 GPS position of all ice radar profiles.

Table 4.1 Total glacier movement at strain net 1 to 4 (28^th June to 4^th July 1997).

Table 4.2 Total glacier movement at transverse transects 1-4 (28^th June- 4^th July 1997).

4.2 Radio-echo sounding survey

4.2.1 Longitudinal Profile A and B

In order to explain the influence of the underlying bed topography and to establish the ice thickness of Gígjökull two longitudinal profiles (A and B) and eight transverse profiles (a to h) have been examined and graphed in Fig. 4.3 and 4.5. The exact position of all ice radar profiles can be seen in Figure 4.2. Additionally, the position of the two long-profiles and cross-profile e is given in relation to the strain nets 1-4 in Figure 3.2. In order to facilitate the comparison between the two long-profiles A and B and the transverse profiles, the position of the cross-profiles labelled a to h relative to the long-profile points is marked in Figure 4.5. This cross-correlation allows us to back-up the results. However, it has to be noted here, that these highly depend on the interpretation of the wave forms which revealed itself to be very difficult (cf. Appendix 13 and text below).

Viewing Figure 4.3, the first thing we notice is a relatively uniform and even ice surface whilst the glacier bed shows a striking drop in the middle section of both long-profiles.

Long Profile B along the western margin of the glacier, crossing strain net 3 demonstrates the kind of correlation between glacier surface and bed rock one would normally only expect in its upper-section: the ice shows moderate thickness below the icefall and thins where the bedrock rises before droping away into the overdeepened area. The latter could be explained by increasing tensile stresses downglacier where the bedrock steepens dramatically. These stresses could cause surface flow velocities to be higher than usually. As strain net 1 (Chapter 4.1.1) has shown and Figure 4.7 displays, mean surface velocities are highest at this point of the Long profile A. This could also go for Long-Profile B since the rise in the bedrock displayed in Figure 4.3b correlates with the slight rise in Figure 4.3a just before the bed slope steepens extremely. Generally, every ridge or other large obstacle on the glacier bed must be reflected on the glacier surface as a large undulation. We must then observe diverging and converging flow around these areas revealed by different deformation rates and flow velocities. In fact, the slight ridge revealed by radio-echo sounding can be related to observations made on the ice surface just below (Long-Profile A):

Figure 4.3a

Figure 4.3b

Figure 4.4
Gígjökull: map derived from aerial photographs. Left map showing the glacier snout in 1947, right map in 1989. Note: Gígjökull had retreated several hundred meters in 1947 before it advanced again until 1997. Presently, the glacier retreats again. The influence of the huge lateral-terminal moraine can be estimated from the maps. Legend overleaf.

Legend:

Explanation (Fig. 4.4):

On the left hand map the small extend of the glacier can be seen and big amounts of debris-covered ice. Note: Only few crevasses occur near the snout. Most of the meltwater appears to drain supra-glacially into the lake.

The right hand map reflects the situation in 1989 when the glacier had advanced. Note: The glacier expands into the valley and exhibits many compressional radial crevasses at the snout. Near the centre-line the longitudinal crevasses upglacier of the suspected ‘bump’ are visible. The ‘Christmas-tree-shape’ 1947-tephra outcrop can be seen at the eastern side of the glacier.

The glacier surface exhibits a slight rise of its slope and the surface morphology clearly demonstrated the typical ‘flow-around-an-obstacle’-pattern (Fig. 4.4). Thus, at the up-glacier side of the bump, longitudinal crevasses indicated transverse extension which could be due to diverging flow around the obstacle. This would mean albeit that the ridge does not run across the whole width of the glacier. Features visible on the glacier surface as well as the deformation pattern of strain net 1 match well with the radio-echo sounding data and support the latter.

Nevertheless, if we correlate Long Profile A and B with cross-profile b we can see that the bedrock is relatively even and could well reflect a ridge since ice depth changes markedly from cross-profile b to c. However, viewing 4.5b in comparison with 4.5c we note that the rise in the bedrock in Long Profile A and B is also reflected in 4.5c (left and right) which could be interpreted as two large bedrock obstacles with surface flow pattern as described in section 4.1.1. Yet, this could not be backed-up for Long Profile B with the mere help of ice surface morphology as it was the case for Long Profile A.

Finally, at both Long Profile A and B along the centre line of the glacier, some interesting observations can be made (Fig. 4.3). Both long-profiles display an extreme steepening towards the snout which almost seems unrealistically high though the radio-echo sounding data provides us with this result. Ice depth along Long-Profile A and B is confirmed by ice depth of the eight cross-profiles which show the same depth. Nevertheless, from the difficulties of radio-echo sounding data collection and processing as discussed in Chapter 5.1 we have to assume that the ice thickness in the near-snout area is less high, but reflecting the same bedrock undulations.

4.2.2 Transverse Ice Radar Profiles a to h

To relate subglacial topography and glacier surface and to evaluate the influence of an overdeepening on ice flow more precise records of the bedrock were required in addition to the two long-profiles. Therefore, in co-operation with Barnes et al. (1997), most of the glacier terminus was covered by ice radar survey in the form of eight transverse profiles of approximately 550 m length across the glacier. Figure 4.2 shows the exact GPS position of these cross-profiles and the relative position to the Long Profile A and B. The dense spreading would normally guarantee a reasonable resolution, although, as will be explained at the end of this section, other limitations restricted the accuracy of resolution. In Figure 4.5 a to h the glacier geometry, i.e. ice surface and bedrock of all cross-profiles is plotted against the width of each profile, whereby all of them begin on the western side of the glacier and end on the eastern side (View downglacier!).

As we move downglacier we can see that the ice thickness ranges from 70 to 170 m in the upper-section of the glacier snout (Cross-profile a-d) getting thicker as we reach cross-profile d. Viewing cross-profile e-h we an extreme increase in ice depth can be observed with mean values of 300 m. This abrupt steep slope of the bedrock is also reflected in Long Profile A and B (Figure 4.3). On the grounds of the thoroughly interpreted radio-echo sounding data, we can therefore confirm a marked overdeepening as we reach the snout area. Whether the absolute values can be taken for true however remains doubtful, since the radio-echo sounding puts certain limitations on precise results (cf. Chapter 3.3 and 5.1). In particular, the fact that Gígjökull seems to overdeepen below sea level raises further questions. Furthermore, in cross-profile d and c, we can observe marked irregularities in the bedrock, which could be due to measuring errors (see Chapter 5.1) or the low resolution of the radio-echo sounding survey.

In summary, the mean ice thickness below the icefall is about 120 m and in the near-snout area about 300-350 m. From the few ice radar readings taken, it does not seem sensible to draw too many conclusions from the underlying bed topography in regard to the ice surface morphology. Nevertheless, it is in most profiles observable that the bedrock topography is reflected on the ice surface. This goes for cross-profile b (western section), e, f, g and h though the data density decreases toward the bottom profile (h).

Fig. 4.5a Fig. 4.5b Fig. 4.5c

Fig. 4.5d

Fig. 4.5e

Fig. 4.5f

Fig. 4.5g

Fig. 4.5h

Figure 4.5 Transverse ice radar profiles a to h.

As an example, cross-profile e was chosen for the cross-glacier geometry since it lies within the area of extremely high ice thickness. As we can see from Figure 4.5e, there is some correlation between ice surface and bedrock topography, both being relatively even and showing slight undulations in the same direction. The ice was found to be thickest in the middle of the glacier, albeit a sudden rise was recorded at the western margin (left) which is not reflected on the surface. The shallow ice depth there (cf. Fig. 3.2, site 3/ tephra-loop) is either due to the general thinning of the ice towards the glacier margins or, and this is most likely, it reflects a measuring error. The latter can be assumed if we compare Figure 4.5e and Figure 4.3b in addition to Figure 3.2 which show ice thickness of more than 300 m above this point. Nevertheless, it is possible that adjacent to Long Profile B at the western tephra loop, the rock-walls begin. This is reflected in the sudden thinning displayed in Figure 4.5e. If this is the case, we can see the shape of the valley trough. However, this is not displayed by cross-profile f and hence remains unconfirmed.

Further, we can see the slight rise in the middle section where the glacier surface is highest. If we compare this with Figure 3.2 we can observe that it shows the middle-ridge (ice lobe) where the highest flow velocities had been expected and strain net 1 had therefore been set up. Long Profile A passes along this ridge (indicated by ‘A’ in Fig. 4.8a). From field observations we have to assume that the bedrock rises again towards the snout where the glacier eventually calves into the lake.

In contrast to the assumption that irregularities in the bedrock are reflected on the glacier surface, cross-profile a, (b to some extent), c and d display rising glacier surfaces above hollows in the bedrock (particularly cross-profile c, middle section).

There are two possible explanations for this: first, and particularly in the case of top-profile a, the divergence of bedrock and ice surface is accounted for by an increasing amount of ice having just left the icefall. This results in local compression and hence thickening. A second possibility is that it could indicate an increasing overdeepening as we move downglacier, in which case it would explain the increasingly high compressive flow as we approach the glacier terminus. As will be confirmed by the crevasse pattern, flow is highly compressive at the snout (Chapter 4.3.5). A combination of both models is also plausible.

4.2.3 Basal shear stress, strain rate and surface flow velocity

Basal shear stress

Having established ice thickness, shear stress and strain rate could be calculated according to equation 3.1 and 3.2. The results are presented in Table 4.3. We observe that highest values for both basal shear stress and strain rate occur where the ice is thickest, i.e. at strain net 3, and that both are lowest where the ice is thinnest, i.e. at site 1 near the centre-line. It might, however, at first sight seem surprising that, though the ice depths at site 2 and 3 are almost the same, both shear stress and strain rate differ significantly. This can be explained if we look at equation 3.1 and 3.2: In the first, the value for basal shear stress varies with the sinus of the slope angle and is therefore higher at site 3 (8,4° compared with only 5,7° at site 2). Also, owing to the exponent n=3 in Equation 3.2, the strain rate e increases exponentially. If the shear stress is doubled, (by a factor of 1,5 if compared between site 2 and 3), the strain rate increases eightfold (by a factor of 3,4 respectively). Although all results are consistent within themselves, the low depths at site 1 (centre-line) and high depth at site 3 (margin) raise further questions. They have to be explained in the light of the topographical context and in view of the limitations imposed by RES (Chapter 5.1).

Table 4.3 Basal shear stress, strain rate and surface flow velocity.

In conclusion, we can observe that shear stresses at the four sites are not within the range Paterson suggests: "Values are normally between 50 and 150 kPa." (1994, p. 240). However highest strain rates occur as expected where the ice is thickest and where the surfaces slope most steeply, namely at site 3. Also, the expected decrease towards the ice margin and the glacier terminus could not be proved.

Strain rate

In addition to the calculations for shear stress and strain rate attempted in table 4.3, strain rates have been graphed against ice thickness along the longitudinal ice radar profiles A and B in order to examine the relationship between ice depth and strain rate visually. This is done mathematically using equation 3.1 and 3.2. This allows us to make some general observations: viewing Figure 4.6a,b the first thing apparent is that, with a few exemptions to be discussed below, strain rates are highest where the ice is thickest and vice-versa.

Figure 4.6a Longitudinal ice geometry and strain rates at Long-profile A. Left axis ice surface and bed elevation data. Right axis: calculated strain rate. See Fig. 4.2 for location of profile.

At Long-profile A we can see at immediately that strain rates vary significantly showing lower rates in the top section (left), and highest rates in the middle section. However, if we take account of the underlying bed topography, and hence highly variable ice thickness, we can explain the variations in strain rate.

On the top left margin of the profile, strain rates increase as the glacier steepens towards the first topographical ridge in the bedrock (discussed in 4.2.1). This seems reasonable as it is the point still closest to the icefall (i.e. where highly compressive flow occurs) and is upglacier of a large bedrock ridge, or even the edge of the overdeepened area. Thus, the ice there thickens slightly, resulting in high shear stress and hence high strain rate. As we move downglacier towards the first ridge, strain rates increase slightly in proportion ice thickness and slope angle. The maximum strain rates are displayed in the middle section of Long Profile A. They are due to the thickening of the ice and the steep slope downglacier. These high strain rates in the bottom area could be additionally enhanced by a ice thickening of the ice and overdeepening in this section of the glacier whereby the overdeepened area is limited towards the lake by a dipping ‘riegel’ (Spedding, 1997) (see also Fig. 5.1).

Figure 4.6b Longitudinal ice geometry and strain rates at Long-profile B. Left axis ice surface and bed elevation data. Right axis: calculated strain rate. See Fig. 4.2 for location of profile.

Strain rates along Long-profile B show higher variations. Again, they can be explained in the context of ice thickness and bedrock topography. The first thing we observe is a remarkable decrease in strain rate as we move from the top-left margin of the profile downglacier to the edge of the overdeepening, just before the glacier bed drops away. Here strain rate reaches its minimum values. We can explain this low value if we correlate the strain rate to the ice thickness above this point: the ice is extremely thin above the bedrock ridge owing to tensile stresses downglacier. These increase as the bedrock steepens and cause ice flow to accelerate and resulting in a longitudinal stretching; ice thickness is below 70 m at this point. As the ice thins and the slope angles drop, so does the strain rate.

In the middle section, strain rates are moderate although ice thickness increases downglacier of the bedrock ridge (i.e. where the slope angle is relatively high). Nevertheless, strain rates eventually increase again toward the end of the profile where they reach maximum values.

In conclusion, we can say that the strain rates plotted against ice depth at Long Profile A and B show the expected correlation, mathematically given by equation 3.1 and 3.2, and fit into the strain pattern described in Chapter 4.1, though values in the near-snout area are higher than one would generally expect them to be from field experience, due to the marked overdeepening and ice thickening.

Centre-line flow speed

In order to correlate surface ice flow to the underlying topography, strain net data and transect data has been linked to ice radar points and graphed as mean surface flow speed against the course of Long Profile A (Fig. 4.7) and cross-profile e, representative for mean cross-glacial flow speed on the glacier snout (Fig. 4.8a,b).

Figure 4.7 Mean longitudinal ice surface velocity near the centre-line.

Viewing Figure 4.7, the normal ice flow pattern of a valley glacier can be seen: the overall flow speed decreases towards the glacier terminus. The fact that the graph stems from strain net data and interpolation of surface speed linked to Long Profile A explains the few records displayed there. However, we can observe a marked decrease in ice flow velocities as we move towards the end of the glacier, reaching minimum values in the bottom region. This decrease in surface flow speed can be explained by the increasing distance from the EL (cf. turnover at the equilibrium line, Appendix 15) and the decreasing slope angle, in particular as we reach the near-snout area which we had expected to be overdeepened. This could further be an indication of highly compressive flow at the terminus constrained by a dipping ‘riegel’ (cf. Fig. 5.1; Spedding, 1997) and large moraine ramparts. Along most of Long-profile A flow speeds are constant (at high level) though ice thickness increases dramatically in the middle section. Flow velocities are generally high and within the expected range of maritime-regime glaciers. In particular in the upper-section of Long Profile A, subglacial hydrology could also contribute substantially to high flow speed since there is no indication for a steep slope (3^d and 4^th measuring point).

It might seem surprising however that Figure 4.7 fails to reveal a significant increase in surface flow velocities at the ‘icefall-overdeepening-edge’ since we would expect tensile stresses downglacier of the latter to be particularly high, and ice flow hence to accelerate. There are two possible explanations for this: first, it is possible that the attempt to link flow speed records of transect 3 has not succeeded, due to a decrease in resolution (cf. 4.1.2); secondly, the remarkable steepening in the bedrock towards the overdeepened area is afflicted by radio-echo sounding weaknesses and is in reality, less pronounced than displayed in the long profiles. Although there is strong evidence for the latter (see Chapter 5.1), we get a higher resolution and a more precise record of flow acceleration at this point by considering transect readings (4.2.1); furthermore, strain net 1 exhibits increasing flow speed on its downglacier side, indicating longitudinal stretching. The high flow velocities at this site can best be explained by tensile longitudinal stresses downglacier of the bump in the bedrock and enhanced creep around the ‘obstacle’ (see Chapter 1.1 and 4.2.1).

Lateral flow speed

Figure 4.8a,b can be interpreted similarly. They were produced by linking ice depth data of transverse profile e with flow measurements along the third transverse transect from the top (Stake 13-18; Fig. 3.3 and Table 4.2).

Surface flow speed decreases towards the ice margins (East and West) and increases towards the centre-line. It might seem at first sight surprising that highest values for flow speed are reached on the western side (Marker 13; Fig. 4.8b) close to the ice margin where values drop again to 0,06 m d^-1 (Marker 13; Fig. 4.8b). This high surface flow speed might be due to a measuring error. However, interpreted in the wider topographical context, the

measurements may also indicate increased flow velocities at the western flank (hence the bending of the terminal lobe to north-west and the ‘gap’ in the tephra outcrop where the glacier calves into the lake (cf. Chapter 4.3.3).

Figure 4.8a Lateral flow speed at cross-profile e; long-profile A and B marked.

The generally low values for surface flow speed in comparison to the other velocities (Table 4.1 and 4.2) are due to the position of the profile near the glacier terminus and thus decreasing flow velocities (see Chapter 4.1.2).

Figure 4.8b Lateral flow speed plotted against marker stakes of transect 3 from the top.

4.3 Ice surface morphology

In the following, the visible properties on the glacier surface will be examined. There are the ice fall, the tephra outcrop of the Hekla eruption in 1947, debris bands and crevasses as well as moulins, closed-up conduits, and supra-glacial meltwater channels. However, the results of slope measurements and ablation rates will be considered first.

4.3.1 Slope measurements and ablation rates

For practical reasons, both surface slope angles and ablation rates were measured. They were required in order to calculate stress, and hence strain rates, and also in order to adjust EDM readings to the true elevation change. As Table 4.3 and 4.4 show, only site 1 had slope angles of more than 10°. The other sites had slope angles well below 10° which was of importance when calculating shear stress and strain rate (cf. 4.2.3)

Table 4.4 Slope measurements.

However, we can infer from Table 4.5 some general observations: Ablation was high, both in 1997 and 1998. This can be explained by the general climatic conditions as described in Chapter 1.2.3 and confirms the maritime glacier type of Gígjökull.

The high deviation of rates measured by Barnes et al. (1997) seems to be due to measuring errors at their relatively unstable bamboo stakes, whilst my own records and those of 1998 correspond well. The result of the slope measurements and the ablation rates are given in Table 4.4 and 4.5.

Table 4.5 Ablation rates over a sequence of 6 days and in average.

4.3.2 The icefall

Figure 4.9 The icefall. Clear view on séracs and cliff-rocks (background left).

Over its course, the glacier flows over a steep cliff-rock where the ice-cap descends into the Markarfljót-Valley. This results in an icefall where the ice flows extremely rapidly. Due to the acceleration of the glacier as it enters the icefall stresses there are tensile and flow is extending. The result is that numerous crevasses open up and unstable séracs break (Fig. 4.9). Along the length of ~ 3000 m, the icefall exhibits a height drop of ~ 1000 m which means that its slope is very steep. Below the icefall, a gentle terminal lobe follows and flows into the Markarfljót valley where it ends in Lake Lónið. The zone just below the icefall shows extremely compressive flow, which is reflected in the closed-up crevasses (visible on the surface as blue crevasse traces formed by the freezing of meltwater in the crevasse prior to closure, or white if the crevasse had been filled with snow) and the increasing ice depth. This correlates with observations of Paterson who notes: "At the foot of an icefall there is a large longitudinal compressive stress in the ice." (1994, p.93)

During the course of a day, the dynamics of ice flow through the icefall can be observed in a spectacular way as it takes the form of collapsing séracs which, once broken out of the ice, fall with a sometimes terrifying rumble. The importance of the presence of an icefall at Gígjökull lies in its influence on englacial processes of ice deformation. Thus, originally almost horizontally accumulated fallout from the atmosphere such as ash from volcanic eruptions will be disturbed and deformed in some way and might explain some anomalies in the foliation of certain areas of the terminal lobe. I will try to explain these in a further step.

4.3.3 The Hekla-1947-tephra layer

Hekla, a nearby volcano located about 40 km to the North of Eyjafjallajökull, erupted for the last time in 1947, resulting in a fallout of ash on the ice-cap above the equilibrium line. This tephra layer was subsequently buried by snowfall and reworked into the glacier. The most likely way in which this process might have progressed is described in more detail in Appendix 16. Being part of the glacier, just like any other ice layer, the tephra follows the same passage-way as other ice layers, and its outcrop tells us as much as ogives and seasonal ice layers of different colour. The general throughput of a glacier along its length is demonstrated in Appendix 15 according to the ‘wedge’-concept of Sugden & John (1976). From theses considerations it can therefore be expected that the tephra outcrop at the terminus of Gígjökull exhibits the result of stress and strain within the glacier uphill of the present location. Since the tephra layer is black, it provides a rewarding marker horizon which exhibits the deformation pattern of processes that take place within the glacier where they can hardly be examined.

The foliation of this layer is of metamorphic structure produced by alteration under high stress. It shows the typical pattern of transverse foliation (Benn & Evans, 1998) near the middle of the terminal lobe where it reflects the previous vicinity of transverse crevasses and the icefall. Towards the glacier edges, longitudinal foliation predominates, owing to the drag of the valley walls which causes the layer to rotate during ice flow. At the margins therefore, foliation is aligned parallel to glacier flow. The pattern of foliation can best be described in three dimensions as a set of "nested spoons" (Allen et al., 1960). However, as the case of Gígjökull shows, deviation from this normal condition can occur and the layer can exhibit complex folding (cf. Fig. 3.1 and 4.4).

On the eastern side of the glacier; the tephra outcrop shows a fairly regular shape, following the contours of the terminal lobe with its arc stretching towards the snout. However, as we move further towards the middle of the snout, the foliation disappears abruptly from the ice surface. This can be seen in the ‘gap’ on the rampart calving into the lake (Fig. 4.4). This sudden disruption in the tephra layer right at the arc of the outcrop could be explained by the high flow velocity in the middle of the glacier which has caused the tephra in this area to be washed out more quickly. This is a reasonable explanation since ice flow must be expected to be highest where there is no terminal moraine or other feature that could constrain the glacier movement but a small lake into which it calves.

This reflection of high flow velocities can not be observed on the eastern side of the glacier where a generally uniform and slightly convex curved foliation can be seen stretching towards the glacier terminus though overall reflecting the contours of the glacier (Fig. 4.4). Only lower flow velocities in this area could explain this retardation, and, looking at the surrounding topography, the most sensible explanation is the constraining effect of the huge lateral-terminal moraine on the eastern flank and to the north of the glacier snout. Further evidence for this effect can be derived from the crevasse pattern in this part of the glacier (radial crevasses) (cf. Chapter 4.3.5 The crevasse pattern) which demonstrates compressive flow. The tephra of the near-snout area (small tephra-cones) represents a redistribution of the 1947-tephra which originally only occurs in discreet layers within some cones. The respective tephra layers can only be found by carefully examining the tephra-cones suspected to incorporate true foliation.

However, the situation on the western side of the glacier is far more complicated. The foliation of the Hekla-1947-tephra at this side exhibits a complex pattern and does not follow the typical transverse or longitudinal foliation pattern at all. On the contrary, the arc of the tephra outcrop points up-glacier (concave in shape) and leads us to the assumption that flow is slowed down or even inhibited there. The ‘Christmas-tree’-like shape (Fig. 4.4) could be due to a big obstacle on the glacier bed which would cause the ice to flow around on each side whilst inhibiting it on the upstream side of the obstacle. This could explain why the layer stretches out longitudinally on each side due to higher flow velocities. However, if we examine the surroundings, no evidence can be found for such an obstacle or any other feature which could hinder ice flow in this way. Neither is the glacier strongly influenced by the lateral moraine nor does radio-echo sounding reveal particular roughness of the bedrock underlying this area (cf. Fig. 4a). Even the ice movement down-glacier measured by using EDM and differential GPS of strain net 3 does not show any evidence which could explain the complex foliation of the tephra outcrop there. On the contrary, flow measurements at site 3 show the expected increase of ice movement from the glacier margin towards the centre of the glacier (Tab. 4.1). Again, as was the case for the eastern part of the terminal lobe, the crevasse pattern can help us to understand the processes that do go on. Thus, if the complex folding of the tephra layer were due to an active process, extensional and compressional crevasses would necessarily be found in the vicinity of the layer. However only the common type of Chevron crevasses are found (cf. Chapter 4.3.5 The crevasse pattern). It can therefore be concluded that the present pattern must represent a relict of processes which have taken place further upglacier, maybe even above the icefall. This will be discussed in the final chapter.

4.3.4 Debris distribution and nature

Viewing the total ablation zone and thus, the terminal lobe of Gígjökull, the overall ‘clean’ ice surfaces -apart from the tephra layer- might seem surprising. In fact, a closer look at the different areas on the glacier does not give the impression of much debris being part of the glacier system –at least not on the surface. Several spots of accumulated debris (£ 3m in diameter) could be found in summer 1998 consisting of fluvio-glacial gravels and till (mostly Plagiat). The only explanation I can imagine for this phenomenon is that it represents material from a rockfall which has taken place at the rock-cliffs flanking both sides of the icefall. This rock might have subsequently been buried by snow and ice and incorporated into the ice where it has been transported downglacier in englacial meltwater channels. As the channels were twisted upwards and subsequently cut off, the debris ablated on the ice surface, forming these circular spots of accumulated material. In the near-snout parts of Gígjökull, the ice around these debris deposits can be dated as being from the 1950s since they are located in the vicinity of the 1947-tephra loop. In connection with the supraglacial meltwater drainage system, the fact that very little debris can be found on the ice tells us a lot about the general drainage system within the glacier and leads to the assumption that most of the debris must be transported through the glacier in englacial water channels (whose discharge must be very high) and must eventually be washed out.

Debris found at the end of the snout near the portal was of well-rounded shape and different size which indicates glacio-fluvial transport. On the shore behind the ice-cored barrier (which seemed to ‘swim’ in Lake Lónið) (Fig. 3.1 and 4.14) some really fine debris could be found but there was also coarser debris of angular shape. This range of debris could be an indication for most of the debris being transported by meltwater and most likely rather englacially than subglacially.

4.3.5 The crevasse pattern

On Gígjökull, all three kinds of crevasse types can be found. In co-operation with Coppins, Foy and Kemp, I have mapped the location of the principal areas where crevasses occur on the glacier, and have measured their width, depth and orientation as indicated in chapter 3.5. The crevasses examined were between 4,3 m and 7,5 m wide and ranged from 3,5 m to 10 m in depth (Table 4.6). The average slope angle of the areas selected for investigation were at the minimum 7° but never exceeded 13°. From the seven areas distinguished by Coppins et al. (1998), I will only describe the most comprehensive results here (Tab.4.6 and Fig. 4.3)

First, in the zone of the icefall, numerous deep transverse crevasses can be found. These stretch from east to west and resemble a flow of faulted ice blocks. This extensive crevassing is due to the high extending ice flow which reflects the underlying bed topography, namely the cliff-rock over which the glacier flows into the rock-walled trough.

Just below the icefall, the crevasses close up again indicating highly compressive flow as the ice flows into the valley. This is supported by the fact that the ice thickens there (cf. 4.2.1 Fig. 4.4). The ice surface is therefore fairly smooth and gently shaped and forms the top-area of the terminal lobe. However, on both sides of the centre-line at approximately the height at which the lateral moraines start, deep and wide crevasses stretch towards the ice edge. These crevasses can be associated with extending flow which could be due to the widening of the valley but also to the increasing flow velocities (longitudinal stresses!). They resemble in fact the transverse crevasses but are closer to the Chevron type which are characterised by the fairly regular alignment of the crevasses at 45° to the ice edge. Also near the centre-line, just a few metres upglacier of site 1, a locally constrained region of spreading crevasses can be seen. The most comprehensive explanation for their occurrence is the large bedrock obstacle revealed upglacier of site 1 using the ice radar (Fig.4.3 and 4.5).

The crevasses found downslope of site 3, and thus the tephra loop at the western side of the glacier, show similar patterns, although they are even closer to the Chevron type (Sugden & John, 1976) which results from frictional drag of the valley walls. The bearings taken at this section of the glacier show the expected 45° although sometimes even more. The orientation of these crevasses therefore indicates the main direction of the stresses operating in this area since the crevasses stretch perpendicularly to the main flow line. This flow line is directed northwards describing a gentle curve towards the lake as it flows into the ‘rampart’ (cf. Fig. 3.1).

Table 4.6 Width, depth, orientation and average slope angle of crevasses

On the eastern side of the glacier snout the situation seems to be completely different and does not fit into the expected pattern of regularly aligned crevasses at the terminus of a glacier which flows into a widening valley. On the contrary, the crevasses there can be referred to as being of the splaying crevasse type, since they meet the ice margin at roughly 45° whilst curving markedly upstream near the centre-line. Near the middle of the glacier they are roughly parallel to the flow direction (Fig.4.4). Splaying crevasses usually form where flow is compressive (Benn & Evans, 1998). In the case of Gígjökull, compressional radial crevasses stretch along several hundred-metres from the tephra layer on the north-eastern side of the glacier snout to the snout margins cutting deep into the ice. The crevasses there are mostly due to high ablation rates caused by the redistributed tephra mentioned above which results in an opening and widening of crevasses (4.3.3). It can be expected that most of these crevasses will close up again as the 1947-tephra is progressively washed out. Whilst indicating highly compressive flow, they also give evidence for an evasive lateral movement of the ice towards the ‘gap’ where the glacier can flow unhindered into the lake. Their orientation could be explained as in the case of the tephra foliation by the surrounding topography. Thus, the huge lateral-terminal moraine is likely to constrain ice flow in the near-snout area resulting in compressional radial crevasses whilst flow velocities near the lake reach a relative maximum. Respectively, the few transverse crevasses that can be found on the ‘rampart’ must be due to extensive flow.

The two or three large and deep crevasses upglacier of the ‘gap’ cannot be explained by the general stress and strain patterns but could be due to englacial meltwater channels which became partly supra-glacial and cut into the ice. They have thus to be seen in the context of the overall surface morphology of this section of the glacier as described above.

In conclusion, the results in general fit well into the expected pattern of strain distribution on a small valley glacier as has been confirmed by the deformation rate at the four strain net sites (4.1.1). However, the concentration of radial crevasses at the snout of the glacier could also give evidence for the influence of an overdeepening which exhibits a dipping slope beneath the terminus in front of the lake. The outcropping of the 1947-tephra layer at an angle of almost 76° (Dugmore, 1998; personal communication) at the terminus of the glacier gives further evidence for this concept (cf. Appendix 16).

  Figure 4.10 Ice marginal meltwater channel (left) and radial crevasse running north (right). Note: a meltwater channel has carved in draining into the lake (right).

4.3.6 Meltwater channels, moulins and conduits

Meltwater channels

Where melting exceeds refreezing rates water will accumulate and then drain following the steepest slope. Supra-glacial water channels usually only occur where discharge is high enough (Benn & Evans, 1998). This is in general the case in the ablation areas of a glacier where, in addition to high ablation rates, the primary permeability of the ice is low. The drainage network of the supra-glacial meltwater exhibits a dendritic shape in this area. As could be observed on Gígjökull, channels are typically a few millimetres to a few meters in depth and have smooth channel walls. Real meltwater channels (persisting from one season to the next) can be verified by absolute downslope orientation and relief inversion which occurs where tephra has been transported in water channels during summer. There, as rising stops in autumn and the ablation period begins, the tephra deposited in the channels insulates the latter and the water channel will thus rise resulting in a relief inversion.

The fact that most of the water channels meandered with great regularity was intriguing since I had expected them to follow cracks or crystal planes within the ice. Yet, it was only there where cracks, foliation or crevasses strongly controlled the structure of the surface that straight channels predominated as had been expected (Sugden & John, 1976). Although these drainage systems share many characteristics with those on rock or sediment (Sugden & John, 1976), there are some important differences which can also be observed on Gígjökull (cf. Fig. 3.1). The first is that the density is very high and well-developed trunk streams are rare. Albeit, I did observe two main meltwater channels on each side of the glacier just below the ice fall. This could be explained contextually as the high compressive flow in this area squeezes water out of the ice and, due to the convex shaped lobe, the water drains laterally. The two big meltwater channels found on both sides of the glacier at approximately the height of half way between the weather station and the lake could well be a following-up of these supra-glacial meltwater streams. They flow out of the ice-cored moraine through a little portal whereby the channel behind the eastern lateral ice-cored moraine is huge in comparison to the western one. It looks in fact like a little canyon.

Another difference between supra-glacial meltwater channels and normal river systems is that the density of net works decreases upglacier because the amount of melting decreases with altitude. Furthermore, it has to be mentioned that the water channels near the glacier terminus, in particular on the ‘rampart’, are less deep and less frequent. This could be due to the general ‘flattening’ of this area as the glacier retreats. However, it also indicates that most of the meltwater must be drained further upglacier, and, as pointed out above, mainly towards the ice margins. The wider implications of all this will be discussed in my final discussion.

The most important difference in this context however might be that glacial channel systems are highly changeable and unstable. This is because high ablation constantly alters the surface morphology of a glacier. As on most small valley glaciers with markedly convex snouts commonly develop (Benn & Evans, 1998). This explains why meltwater predominately drains towards the margins at Gígjökull where, in general, supra-glacial meltwater could be found in bigger volumes at the lateral parts of the centre flow line.

Moulins

Most water which is transported in englacial drainage systems originates from surface melting and is conducted to the highly efficient drainage network by supra-glacial meltwater streams and big holes in the ice surface where theses water channels often end abruptly. These almost vertical and deep holes, called moulins, resemble sink holes in karst and commonly arise where the ice structure is weak, as is the case with crevasses. Moulins are formed where water flows into a crevasse which links to presumed horizontal englacial tunnels (Sugden & John, 1976; Paterson, 1994). Their diameter ranges between 0,5 m and 1 m and they can be 25-30 m deep (Sugden & John, 1976, p.289). The structural influence of crevasses on the distribution of moulins is remarkable (Benn & Evans, 1998). Thus, changes in crevasse patterns are followed by changes in the supraglacial catchments of the englacial drainage system. The best example for the consequence of this phenomenon is the "moulin of death" in the north-eastern section of the glacier snout of Gígjökull (Fig. 3.1). Where there had been a large and deep moulin in summer 1997 spanned by an arc of ice, by summer 1998 a wide and deep canyon-like trench had been cut into the glacier and had crossed most of the radially crevassed section with its tephra bands oriented towards the north. The previous arc collapsed at its convex point and forms a spectacular ‘claw’ today.

The biggest ‘active’ moulins are mapped in Figure 3.1. Taking their location into account, their symmetric position on the ice might seem surprising: On each side of the centre-line lobe at approximately the height of the upper-end of the lateral moraines (cf. conduits), a moulin can be found. On the one hand, this will certainly be due to the two big meltwater channels discharging along each side of the centre-lobe just below the icefall and the general convex shape of the ablation area. On the other hand, they provide further evidence for a specific, subglacial drainage network which seems to be strongly influenced by the overdeepened area. In Figure 3.1, I have sketched the most likely englacial passage beneath site 3 based on observations made upglacier of the Hekla-1947-tephra outcrop adjacent to the western moulin, and downglacier, where parts of an englacial water channel were open and allowed an observation of the englacial drainage system. The depth and width of this tunnel was remarkable and gives proof for the high efficiency of englacial meltwater discharge. From this stems my assumption that meltwater derived from supraglacial meltwater channels and transported in englacial tunnels is squeezed out from the centre-line towards the margins of the glacier. This in turn could be the result of several processes. First, as mentioned above, it could be due to the overall convex shape of the surface of small valley glaciers. Secondly, it could be owing to the fact that the ice thickens markedly below the icefall (compression!) and that water pressure increases rapidly as the ice thickens at the beginning of the overdeepening which squeezes the water out towards the margins. Finally –and most likely in combination with the thickening- the uplift of englacial water channels owing to the process of overdeepening, could result in a discharge of water following the steepest slope and the shortest way through the ice
(cf. "equipotential surfaces and predicted drainage system in a glacier" Paterson, 1994, p.113) which is towards the lateral margins.

Figure 4.11 "Moulin of death" (left) and one of the active moulins adjacent to the centre-flow line (right).

Conduits

Conduits could be found on Gígjökull on both sides of the centre line below approximately the height of the uphill end of the two lateral moraines where the rock cliffs begin. They did however occur significantly more often on the eastern part of the glacier and were much bigger there (indicated by ð ï in Fig. 3.1). They all seem to be relicts of former englacial conduits which linked to the sub-glacial drainage network. As the glacier advances and the ice flows towards the snout leading to compression, englacial conduits are twisted upwards (cf. ‘wedge’-concept Sugden & John, 1976, p. 41) (cf. App. 15) and appear as cut-off englacial water channels at the surface. Their general shape and orientation in the near-snout area was irregular and sometimes even ‘bizarre’. In the middle section of the eastern flank their shape and size varied between circular and drop-like shapes and diameters of 30 cm to 80 cm (Fig. 4.12). Their depth ranges from 30 cm to approximately 6 m (guessed by using a measuring tape!) and sometimes even the twist upglacier can be seen. It should be stressed that all conduits were debris-free, which gives further evidence for the glacier being ‘clean’. This fact raises the question of whether there is no debris at all or, if there is some, where the debris is transported to. It could be that most of the debris is transported within englacial conduits and is flushed out of the ice at high pressure as switches from distributed to channelised networks occur (Willis, 1995), and that no debris is left in the conduits as they close up. The fact that the conduits increasingly occur as one moves down from approximately the height of the lateral moraines just below the icefall gives strong evidence for the concept of a lifting in the subglacial meltwater system from this point downglacier.

Figure 4.12

Shape of closed-up conduits. Note: Closed-up conduits found on Gígjökull were of different shapes which varied from circular to drop-like. Their average diameter was 30-80 cm. Most of the conduits were squeezed longitudinally, and all of them were debris-free.

4.3.7 Other observations

During the course of the year which followed the summer of 1997 many surface features had changed giving evidence for the high dynamics which the glacier undergoes year by year. As such, the most striking observation was a ‘flattening’ of the snout where the glacier calves into Lake Lónið (Fig. 4.13). Also, the area which had been supposed to be overdeepened seemed to have dropped since the summer of 1997. Obviously, the glacier is at present retreating. The small lake in front of the glacier has enlarged correspondingly. Further, the portal of the glacier has shifted towards the north-eastern side of the glacier terminus, and in front of it, a long narrow barrier of debris with small particle size (~ 0,2 to 20 m m) has built up. This barrier lies perpendicularly to the main flow direction into the lake (Fig. 4.13 and 4.14).

From the relicts of the aeroplane that crashed on Gígjökull in 1948, only a few pieces were left on the ice this year. Most has been reworked and pushed towards the ice margin since last year. However, some pieces might have been taken away by visitors. In conclusion, glacier dynamics were clearly perceptible.

Finally, and even more striking were the little fountains which burst like geysers just below the icefall and upglacier of the western tephra loop (Fig. 15). Water was pressed out of the ice and burst 0,5–2 m high. In 1996 even more of these fountains could be observed and seemed to occur periodically during the course of a day (Dugmore, 1998, personal communication). This phenomenon gives proof of the obviously high hydrostatic pressure in this section of the glacier which forces englacial water to rise up to the ice surface.

Figure 4.13 Lateral view on the glacier snout. Note: the flat ‘rampart’ which indicates that the glacier retreats at present. Also to be seen: ice-cored moraines and debris barrier.

Figure 4.14 Terminal portal at the glacier snout (NE) in 1998. Note: debris cover.		Figure 4.15 Fountain, bursting out of the ice due to high hydrostatic pressure. (Next to it: Thomas Robb and Robert Bingham).

Chapter 5. Discussion and Conclusion

5.1 Discussion

The flow pattern of Gígjökull is typical for outlet glaciers and is exemplary of a glacier terminating in soft sediment such as the Markarfljót-Valley. By and large, the glacier shows the behaviour of a typical maritime glacier though some unusual patterns were found.

The question as to whether the influence of near-snout overdeepening on ice flow can be quantified is difficult to answer. Fundamental to a response to this question is proof for an overdeepened area as suggested by Spedding (1997). In this dissertation I have tried to test his hypothesis by examining all sets of data available, and will evaluate his idea in the following section. Despite radio-echo sounding’s many strengths, the results here nevertheless have to be considered as provisional owing to significant weaknesses in the method. What we can state is that there is indeed strong evidence for overdeepening beneath the terminus of Gígjökull.

Ice thickness calculated from the traveltimes of radio waves pulses propagating through ice, as explained in Chapter 3.3 (Appendix 12), were found to vary extremely owing to the fact that it was sometimes hardly possible to pick out the true signal of the reflected radio wave, so that in some cases, the true signal had to be chosen from two peaks, resulting in two different traveltimes with one mostly ranging from 1,4-1,8 m s and the other from 3,9-4,1 m s. This required a thorough examination and evaluation of the most plausible result. As in cross-profile f and h, there were in some cases two options for ice depth. Although the shallower depth (resulting from shorter traveltimes) sometimes seemed to be more reasonable, in the general context of glacier shape, size and location there was stronger evidence for the longer traveltime and hence the higher ice thickness being the correct one when compared with the other radio-echo sounding records. Thus, the radio-echo sounding survey revealed the following sequence of ice thickness: just below the icefall the glacier exhibited about 130 m of ice thickness. This was followed by a shallower area of less than 100 m. This area is revealed by the ice radar records as a ridge (‘icefall-overdeepening-rim’?) or large-scale obstacle at the bedrock just below the ice-fall and is followed by a sharp dip of the bedrock into the near-snout area where ice thickness increases markedly reaching more than 300 m (Figure 4.1). In conclusion, there is strong evidence for an overdeepening, although its exact extent requires further investigation. From the data presented in Figure 4.3 and 4.5 we can note that the glacier bed even deepens below sea-level which in could well indeed be possible since Gígjökull lies at a very low altitude. However, the ambiguity in waveform interpretation rather casts doubt on the true ice thickness. Viewing the aerial photograph of 1947 (cf. Fig. 4.4) it appears increasingly unlikely that the area exposed at that time would have been carved out that much in only 50 years.

Although the results presented in Chapter 4.2 are gleaned from only a few records, they do allow us to get a rough idea of the underlying bed topography which is highly dominated by the overdeepening. We can only really benefit from the strengths of radio-echo sounding, as outlined in Chapter 3.3 (e.g. evidence for ice lenses or debris bands), if we know exactly how to interpret the signal of the radio pulse, which implies that we know how this signal changes as the wave is scattered from water, debris, ice crystal boundaries and faults or bedrock, and that we apply the equipment appropriately in the field being in particular aware not just of the depth and orientation of crevasses and the ice margin, but also the frequency of the fluke (receiver). This would however require exact trimming of the RES equipment by correlating waveforms with borehole data in situ. It is only then–on condition that radio-echo sounding readings are taken systematically (!)- that we can give precise information on ice thickness and ice properties.

In summary, we can state therefore that the accuracy of the radio-echo sounding data presented in Figure 4.3 and 4.5 is limited by the severe weaknesses of the interpretation of radio waves which are sometimes base upon subjective common sense. This does not however seriously affect the evidence for an overdeepening provided by radio-echo sounding survey.

The deformation of four strain nets was recorded over the period of 6 days, and stress and strain rates have been calculated with the help of ice thickness. Glacier movement results from permanent strain of the ice and the glacier bed in response to stress. It was possible to calculate stress and strain rates using the relationship (Glen’s Flow Law) provided by Nye (1955) and Paterson (1994). The deformation of the strain nets revealed the expected pattern: In general they were squeezed longitudinally whilst being stretched laterally, although there are exceptions. Overall, ice flow followed the main glacier channel axis, so that it can be said that the shape of the bed directly affects the direction of flow on the glacier surface. Whilst at site 1, 2 and 3 a correlation between subglacial bed topography and surface flow direction could be observed, it could not be proved at site 4.

The deformation of ice, represented by strain rate, results from creep or fracture. Strain rates calculated according to the relationship given by Nye (1957) revealed to be highest rates where the ice was thickest which correlates with the expected pattern. Thus the highest strain rates occurred downglacier of the icefall-overdeepening edge (Long-profile A and B) and in the near-snout region as well as just below the icefall (Long-profile B). They varied between 1,06 and 3,14*10^-9 s^-1. Since strain rate calculations are based on theoretically developed relationships all values have to be backed-up and compared with field measurements, such as data gained from strain nets. Since the strain nets could only reveal surface ice deformation and not ice deformation at a greater depth, further work remains to be done in order to obtain more precise records of ice deformation below the surface-layer. Theoretically, this can be done by calculating strain rates at different ice depths resulting in a two-dimensional profile.

To answer the question as to what extent bedrock topography, and thus near-snout overdeepening, influences ice flow is no easy task. The influence was found to be relative and to vary throughout the glacier. Although the correlation between surface flow velocities and ice radar recordings allow a confirmation of a close relationship between bed topography and glacier movement in some parts of the glacier (Chapter 4.2. Fig. 4.7 and Fig. 4.8), this is not the case everywhere. The same goes for the influence of the bed topography on ice surface morphology: whilst the correlation between bedrock and ice surface is in some cases, strong (Fig. 4.5 b, f, g and h) it is hardly noticeable in others (Fig. 4.5 a, c, d and e). The precise extent of the effect cannot be derived from the findings described above.

Viewing the four strain net sites and surface velocities along Long-profile A, a gradual reduction in longitudinal flow velocity was noticeable which fits into the general pattern of an outlet glacier terminating on land (Sugden & John, 1976, p. 43). However, the longitudinal flow velocities do not only decrease towards the glacier terminus, but also towards the western flank of the glacier (cf. differences between strain net 1 and 3, as well as between strain net 3a, b, c and d). Nevertheless, the lateral stretching of the strain nets, in particular of strain net 3, and thus the lateral reduction of flow speeds, fit well into the theoretical model of Nye (1965) and the corresponding marginal friction found by Raymond (1971) on the Athabasca Glacier (Appendix 18). However, it has to be stressed here that all results are based on records taken during the summer and do have to be compared to winter data if a complete picture of the mean values is to be attempted (e.g. Spedding’s winter data, 1998, p.158). Additionally, the records are likely to be affected by daily and seasonal cycles (such as ablation cycle, weather cycle, etc.). Also, we have to take into account that the period of observation was short in the time-scale of glaciers.

Spatial variations in flow velocities were observed near site 1 (centre-line), and could be correlated with the underlying bed topography. Here, enhanced creep could be observed where the bedrock shows a large rise (Figure 4.3). Crevasses on the ice surface above this area give further evidence for such a subglacial obstacle which, however, could also be a ridge stretching across the width of the glacier, as indicated by ice radar surveys (Fig. 4.3 and 4.5). The anomalous behaviour observed at site 1, which is in general exhibiting compressive flow, could be explained by Weertmann’s theory of enhanced creep (1957) and Lliboutry’s demonstrations (1993). Although site 1 is located at a point of the glacier where we would expect the ice to splay out laterally into the Markarfljót-Valley, the ice does not follow the expected pattern here. The most sensible explanation for this fact is that the ice flows around the big bedrock obstacle (or ridge), confirmed by ice radar survey, resulting in increased flow velocities on the lee-side of the obstacle. It can be assumed that this rise in bedrock also explains the relatively smooth flow in the centre of the ice-fall and the compressional flow just below it (cf. crevasse traces). However, the extent of this bedrock bulge can only be defined exactly by further radio echo-sounding.

In addition to the above mentioned observations, flow velocities above the icefall-overdeepening-rim as well as below the ice-fall on the upglacier and downglacier side of the latter, revealed themselves to be highest of all. Ice was found to flow faster over shallow areas, and slower over deeper sections which corresponds with the observations of others (e.g. Benn & Evans, 1998) that ice thins above bedrock obstacles due to longitudinal tension on the lee-side (i.e. steeper slope gradient) and hence an acceleration in flow velocities. Above the area expected to be overdeepened, surface flow speeds decreased significantly towards the terminus. Since the glacier cannot expand due to the constraints put on it by the moraine rampart and the riegel mentioned above, this could indicate longitudinal compression, and a respective thickening of the ice. This confirms the outlined expectation drawn out at the beginning, that subglacial bedrock topography usually directly influences surface flow velocities. From these observations we can conclude that the position on the glacier has a large influence on the relationship between ice thickness and flow rate, the latter being highest where either the slope gradient was highest, or where the highest turnover in ice masses has to be expected, as is the case near the centre-line (= main flow line) (Fig. 4.8). However, areas of high bedrock gradient did not always correlate with high surface velocity and vice-versa, which could be due to the low resolution of the data, but also to variations in flow velocity due to changes in the subglacial drainage system which could account for unusually high flow rates in areas of filled-cavity networks or ‘switches’ (cf. Willis, 1995). The latter means that periods of highly efficient drainage (i.e. high rates of debris transport), alternate with periods of stagnation. The above mentioned anomalies could be explained by the concept of Iken and Bindschadler (1986). Thereafter, an elevated water pressure results in the spatial extent of cavities filled with water which enhances basal sliding rates and such the total glacier movement. Nevertheless, in order to explain the driving forces of this process completely, further investigations remain to be done. It would also be necessary to produce a more detailed 3-dimensional terrain-model of the complete bedrock, as well as of the averaged (over one hydrological year) ice surface topography. A thorough survey of flow velocities on a large scale could then be linked with this model.

In conclusion, we can note that the glacier flows faster in the centre and exhibits longitudinal compression and transverse extension in the snout-region whilst flow velocity decreases towards the terminus (Chapter 4.2.3). The highly compressive flow at the snout indicates a concave bedrock (i.e. overdeepening), constrained by some sort of barrier to the Þórsmórk-Valley. Thus, the question raised at the beginning of this work as to whether differences in flow velocity between near-snout areas and mid-glacier areas can be recorded can be answered satisfyingly: Flow differences can be quantified on Gígjökull. Nevertheless, flow rates are likely to be the result of the accumulated effect of the whole ice-fall overdeepening-regime, rather than solely being affected by subglacial topography.

In order to describe stress fields on the glacier main crevasse zones were mapped and examined. In fact, stress is best exhibited on the ice surface by crevasses which revealed themselves to be one of the most reliable indicators, giving strong evidence for a highly compressive flow regime in the snout-region. Whilst indicating highly compressive flow, they also give evidence for an evasive lateral movement of the ice towards the ‘gap’, where the glacier can flow unhindered into the lake. Few transverse crevasses on the latter confirmed extensive flow, as the glacier calves into the lake. Further evidence for the compression is given by the foliage of the ice and the outcrop of the Hekla-1947-tephra which is twisted upwards in the snout-region at an angle of almost 75°. Taking all the crevasse zones together as described in Chapter 4.3.5, we can distinguish between 3 main ‘stress-and-strain’-fields indicated by the crevasse pattern: the first lies just below the icefall and is marked by highly compressive flow indicated by a closing-up of crevasses (transverse crevasses, Chapter 4.3.5) and a thickening of the ice (Fig. 4.3), the second in the middle section of the glacier where flow revealed itself to be extensive and crevasses were almost of the Chevron-type, and the third at the snout of the glacier where radial compressional crevasses indicate highly compressive flow. These findings give strong evidence for an overdeepening, since an icefall-overdeepening-regime would be the best explanation for this sequence of stress and strain fields. Nevertheless, to confirm this quantitatively further measurements would have to be carried out and the individual crevasse areas would have to be examined in more detail.

The Hekla-1947-tephra proved to be the second most revealing feature visible on the glacier surface, since its outcrop provides a good manifestation of stress variations throughout the glacier. Corresponding to the general flow pattern described above the tephra outcrop exhibits a convex shape towards the glacier snout, except at the ‘gap’ on the rampart calving into the lake. Here the tephra has already been washed out, which gives further evidence for flow velocities being highest at the centre-line and for a slight westward twist of the terminal lobe, confirmed by rotation in strain net 3. Our hope that the ‘Christmas-tree’-like outcrop of the Hekla-1947-tephra layer would tell us more about the deformation processes working at the western side of the glacier was not fulfilled. The complex outcrop pattern does not correlate with the deformation observed at site 3 and is not due to an bedrock obstacle further upglacier as ice radar survey revealed (Fig. 4.5). The observed deformation of strain nets 3a, b, c and d rather indicates low deformation rates, and we can assume that the deformation of the tephra layer is due to a complex folding further up the glacier and most probably even above the ice-fall. This is the most likely since flow in an ice-fall is generally very fast and abrupt. However, the incorporation of the tephra into the ice could also have happened within the ice-fall. To determine the exact area where the main distortion of the 1947-layer took place, further research would have to be done to understand the processes working in ice-falls. Certainly, the ice-fall can be considered to be a driving force for glacier movement. Both the crevasse pattern and the tephra-outcrop give evidence for a large hollow, i.e. overdeepening, underneath the snout, being constrained by an upward-dipping riegel.

In order to explain anomalies in surface flow, such as areas of higher flow speed, bed deformation was considered to accomplish a significant part of ice flow by facilitating sliding over the bedrock. The debris content of the ice is however so low that it can reasonably be expected to have negligible influence on strain rates (cf. Nickling & Benett, 1984) and flow velocities. (If the glacier were flowing over a water-saturated deformable bed, friction, were reduced so much that this had a markedly effect on ice flow which however could not be proved.) Although spatial variations in flow velocity can be observed on Gígjökull, there is no further evidence for bedrock deformation. On the contrary, the bedrock exposed on the aerial photograph of 1947 rather indicates hard bedrock, and the overdeepening revealed by radio echo-sounding suggests that the glacier lies on firm bedrock. The main reason most likely lies in the process of overdeepening. This makes it highly improbable that any debris remains on the valley floor which has not been scoured out of the glacier bed. This does not exclude debris transport within the ice which, rather, has to be very efficient if we consider the large amounts of debris that are accumulated as sandur in the Markárfljót-Valley and on the large moraine ramparts. With respect to the latter, there is good evidence for most of the debris being flushed out of the glacier through a channelised drainage network (Chapter 4.3.6). The high amounts of debris deposited on the moraine ramparts therefore give evidence for weak bedrock facilitating sliding. In summary, data collected concerning the subglacial topography implies that the bed is only to some extent deformable and that an increase in flow velocity caused by bed deformation is likely to be of only relatively low importance. The relative importance of the englacial debris, however, needs further research.

Since many anomalies found on the glacier could not be explained with the help of theoretical models or calculations, subglacial meltwater may exert a significant influence on glacier sliding being critically controlled by the overdeepening. From field observations we can infer that the meltwater drainage of Gígjökull does not follow the general (glacial) pattern. It rather shows an anomalous drainage system as far as the investigations allowed us to predict. Two main meltwater channels could be observed just below the ice-fall on each side of the centre-line flowing towards the ice edge (see Chapter 4.3.6). They indicate (contextually) the general squeezing-out of englacial water towards the ice margin as the glacier overdeepens, and the ice consequently thickens. Fountains blowing out of the ice in the same section of the glacier can be taken as further evidence for this phenomenon (Fig. 4.15). They must result from high water pressure within the ice. The latter can only be explained by a marked increase in water pressure as the ice leaves the icefall and flows into the noticeably thickened overdeepened area. Due to the overlying weight of the ice, water pressure must rise there. The two big meltwater channels found on both sides of the glacier at approximately the height of the half-way point between the weather station and the lake, could well be a continuation of the two supra-glacial meltwater streams (see above), flowing out of the ice-cored moraines through a little portal. Only a small amount of the total meltwater seems to drain supraglacially into the lake, which gives further evidence for the assumption that most water is squeezed out laterally towards the margins, and only some meltwater flows through the little portal at the snout. Most of the meltwater, seems in fact to drain laterally, parallel to the lateral moraines. It can therefore be assumed that the drainage system of Gígjökull is strongly controlled by the process of overdeepening (Spedding, 1997). A switch in the routing of basal and englacial meltwater, resulting from the process of overdeepening, could increase the amount of debris pushed towards the ice margin where it would accumulate subsequently. This method of transporting debris seems to work extremely efficiently and could explain the large lateral-terminal moraine (74 m high) (cf. Spedding, 1997).

Nevertheless, the asymmetry in the distribution of supra-glacial meltwater raises further questions. Why do we find so many more conduits on the eastern side of Gígjökull? Why is there so much more water on the surface? And why is the surface there either deeply crevassed or rough and covered by little ice bumps whereas the western side of the glacier, in particular above the overdeepened area, consists of bigger, extended, ice lobes with a fairly smooth surface only cut by a few but deep open crevasses (approx. below the tephra loop)?

In regard to Gígjökull’s drainage pattern, we also have to consider the location of the active moulins and conduits. The two biggest ones were assumed either to be a result of the two big meltwater channels discharging along each side of the centre-line-lobe and the general convex shape of the ablation area, or, and this is more likely, they give particular evidence for an anomalous drainage pattern. In this context particularly, the englacial meltwater channel exposed upglacier and downglacier of site 3, could give evidence for the highly efficient drainage of meltwater in large englacial conduits. These however result from switches in the subglacial drainage network from distributed to channelised drainage as described by Willis (1995) (cf. Appendix 14). From this englacial channel also stems the assumption that water is squeezed-out from the centre-line to the ice margins of the glacier. The increasing occurrence of conduits as one moves downglacier from approximately the height of the lateral moraines (just below the icefall) gives strong evidence for the concept of a lifting in the subglacial meltwater system downglacier from this point. Switches in the subglacial drainage system could also cause debris to accumulate in form of ‘pockets’ abandoned within the ice, which could be rapidly carried towards the ice margins under the influence of the compressive flow regime at the glacier snout (cf. Spedding, 1997). From the above observations we can conclude that meltwater, and thus the subglacial drainage system in general, might highly influence ice flow and result in areas of increased sliding and hence flow velocities. In addition, it would account for a substantial part of the moraine build-up (cf. Spedding, 1997). However, since Gígjökull’s special drainage system is to be strongly determined by the overdeepening, we cannot think of two separate processes but have to link both.

Wider implications

As discussed in the above section, meltwater seems to have a critical effect on the glacier dynamics of Gígjökull being strongly controlled by the overdeepening.

In order to give an idea of the possible passage way of the meltwater and an explanation of the process of overdeepening I propose a hypothesis for outlet-glaciers flowing from ice-caps into relatively soft sediment, which certainly needs further research in order to be confirmed. To gain a better understanding of the process of overdeepening we can simplify it by splitting it into three stages (Figure 5.1):

1. At the beginning, the small outlet glacier flows from the ice-cap into the relatively soft though still very thin sediment of the valley. Meltwater drains subglacially and abrades the bedrock. Debris from the ground as well as from rockfalls onto the ice is therefore transported beneath the glacier and increases its erosional potential. Big amounts of sediment will be deposited pro-glacially on the valley-floor, which in turn thickens respectively.
2. In the next stage, the glacier begins to carve out a valley trough. The ice thickens as the sediment of the valley-floor inhibits longitudinal extension. The glacier thus begins to overdeepen. This process will be accompanied by an uplift of the formerly subglacial meltwater channels since water always takes the shortest route to a lower level. Less debris will now be transported subglacially, but will rather be transported within the englacial conduits whose drainage is highly efficient. At the same stage, as the ice thickens in the overdeepened area, the water pressure within the ice increases markedly and meltwater is squeezed out towards the ice margins of the glacier. [This is exhibited by the two supra-glacial meltwater channels (see 4.3.6). Debris transported towards the margins indicates glacio-fluvial transport (well-rounded) see 4.3.4.] The debris supply towards the ice margins must by now be very efficient and results in the deposition and accumulation of large amounts of debris in front of the glacier. A large terminal moraine builds up and begins to constrain the extension of the glacier which in turn results in a even faster thickening.
3. Eventually, about ¾ of the debris will be transported towards the moraine and only ¼ will reach the valley (estimation of Dugmore and Spedding, 1997), i.e. the moraine gets higher faster than the sediment of the valley-floor thickens. The latter has thickened noticeably by a few hundred meters.

Along the length of the Markarfljót-Valley it can be assumed that sediment of glaciers terminating in it further upstream will be washed out or at least transported further downstream. This leads to an even faster thickening of the valley-floor towards its end. Since Gígjökull is located downstream of other glaciers, the sediment in front of its snout is even thicker which promotes the process of overdeepening. This concept could explain the complex process and the large terminal moraine, and shows that process and form have to be linked.

Additionally, I would like to include another consideration, a comparison of Gígjökull with the adjacent Steinhóltsjökull. The comparison between the two can give further evidence to the hypothesis suggested above. Whilst Gígjökull flows straight into the Markarfljót-Valley, Steinhóltsjökull is ‘protected’ at the eastern side (i.e. upstream) by a riegel. This means that no further sediment, except the debris washed out into the valley by the glacier itself, can accumulate in front of the glacier terminus (cf. Spedding, 1997). The valley-floor here does not therefore thicken as it does pro-glacially of Gígjökull, and Steinhóltsjökull does not overdeepen. In contrary, its sediment is flushed into the valley and is gradually deposited in front of the snout of Gígjökull.

As we can see from the examination of various parameters which influence ice flow the relationship between process and form is mutual. On the one hand, processes result in certain landforms, such as subglacial erosional forms or as pro-glacial accumulation forms, on the other hand, these features can in turn influence processes such as glacier movement by constraining their area of efficiency. This relationship can be positive or negative. If we imagine Gígjökull, for example, as it might have flown out of the little ice-cap of Eyjafjallajökull thousands of years ago, we can assume that it markedly shaped the valley it flowed through. The subglacial erosion resulted in a valley trough which became more and more carved out. This in turn resulted in a steepening of the bedrock which increased surface slope and thus ice flow velocities, whereby the erosional potential of the glacier increased and the bedrock could be abraded even further.

Figure 5.1 Simplified model of the icefall-overdeepening concept. (a) Initial situation. (b) Increasing thickening of the snout as the valley floor thickens; hence a relative overdeepening in addition to a natural carving out, accompanied by marginal sediment deposition. (c) Lift of the formerly subglacial meltwater channels and change to an englacial drainage system. Further explanation: see text above.

This common positive feedback was probably only interrupted by some periods of glacier retreat. As stated above, the relationship can also be negative, which can be exemplified by the influence of the terminal moraines in front of the lake (and thus the glacier terminus) on ice flow. Such a large feature definitely constrains the glacier significantly. As suggested in Chapter 2.2, a positive feedback could exist in which more debris would be transported towards the ice margin, resulting in a increase of the moraine. The latter would dam the lake and constrain the glacier in such a way that, given an ice advance, a certain threshold would be reached at which the system had to collapse due to a thickening of the glacier terminus. To reach a new steady-state the glacier had to burst out in the form of Jökullhaup. However, dated evidence on Gígjökull’s moraines of (1740 and 1890) shows that the glacier was much further forward than the present position, (i.e. over the present terminal moraine) (Dugmore, 1987). The large terminal moraine, which is dated as being less then 2,000 years old, could withstand the overlying ice pressure during these advances, which makes this scenario unlikely.

The application of the gained insight into the processes of modern glaciers to historical implications seems straightforward since it allow us to reconstruct past events. It could help us to determine the circumstance in which old developments have taken place, but also, to predict the behaviour of glaciers in similar topographic situations (i.e. plateau/ valley-terminating glaciers in soft sediment) as at Gígjökull, i.e. in the Þórsmórk-Valley. The only glacier known to behave in a similar way to Gígjökull is Kvíarjökull (Iceland). However, (viewing the wider scientific context), it seems likely that we can apply the gained insight to other places, such as many outlet-glaciers on Greenland. This gives special importance to further investigations.

With regard to Gígjökull’s sensitivity to change this work has shown the highly variable processes within the glacier, but also that we have to regard all of them with respect to the surrounding topography as well as past changes in the landscape (caused, for example, by climate change or by short-term events such as rockfalls, jökullhaups etc.).

5.2 Conclusion

The investigation of the flow pattern on Gígjökull has, in general, shown the expected situation though many anomalies occur on this small outlet glacier. The measuring of deformation on the ice surface, the recording of ice thickness over most of the glacier snout, and the mapping of visible properties on the glacier surface, have provided insight into the glacier dynamics. Many questions remain open about the precise conditions which control stress and strain rates and hence glacier flow. However, as could be shown in this dissertation, it is possible to quantify the influence of a near-snout overdeepening. In order to get a better idea of the subglacial topography, and its influence on ice flow, it would be extremely useful to establish a three-dimensional terrain model of the glacier and a three-dimensional model for glacier flow as has been done most recently by Hubbard et al. (1998). Such a model could then be compared with field data from Gígjökull and help us to determine the glacier’s dynamics.

The presence of a terminal overdeepening seems to exert a significant influence on various components of the glacier: ice thickness (and hence stress and strain rates), flow regime (i.e. highly compressive), subglacial and englacial drainage patterns, as well as debris transport through the glacier and deposition on the moraine (and hence the build-up of the large lateral-terminal moraine at Gígjökull, which in turn influences ice flow).

The features observed on Gígjökull today can only be explained by considering present and past variations with respect to the surrounding topography, i.e. by changes through space and time. This links processes and landforms and helps us to understand the wider implications of those mutual relationships. (Here: terminal moraine = landform; overdeepening/ ice flow = process). Further research remains to be done in order to give more detailed insight into the mechanism and the implications of near-snout overdeepening. Much has also to be done in order to give precise information about the extent of the overdeepening as well as the nature and influence of the subglacial drainage system. Subglacial hydrology remains a ‘black hole’ in glacier research and should gain special consideration. Although data on actual conditions below and within the glacier is extremely difficult to obtain, we have to concentrate our strengths on the investigation of the subglacial hydrology. To focus on the relationships between the process of ice, water and debris transfer seems to be increasingly important. Despite this lack of insight into the dynamics of subglacial hydrology, at present, in particular Hubbard et al. (1998) and Sharp et al. (1993) have provided good work with regard to glaciers’ hydrology.

Personally, I would certainly like to see an expansion on the research behind in order to expand our knowledge about icefall-overdeepening-systems and, respectively, the thickening of the valley floor. Further work remains to be done in order to enlighten the issues faced in this thesis in particular with regard to the subglacial hydrology.