Introduction

A hydroelectric power station in south-western Norway (Fig.l) receives part of its water from Bondhusbreen, an outlet glacier from Folgefonni. Bondhusbreen extends down to 450 m a.s.l., but the collecting tunnel system for the power station is above 850 m a.s.l. Therefore meltwater from Bondhusbreen is collected subglacially. Tunnels for this purpose were driven through the rock beneath the glacier, with shafts extending up to the ice-rock interface under ice 160 m thick. A sedimentation chamber to remove coarse material from the diverted water was constructed in the tunnel system (Fig. 2) (Wold and Østrem 1979[b]). The total volume of water diverted annually from the subglacial intake is about 60x10⁶ ra³.

Fig. 1. Location map showing a generalized outline of Folgefonni and part of the system of water collection for the hydroelectricl power station

Fig. 2. The tunnel system at Bondhusbreen. The subglacial terrain and streams are indicated, (a): helicopter platform and living quarters, (b): tunnel for water diversion, (c): sedimentation chamber. Intake shafts are represented by (1), (2) and (3).

Previous glaciological investigations at Bondhusbreen, made for planning purposes, have been reported by Pytte (1963, 1967, 1969), Pytte and Østrem (1965), Pytte and Liestøl (1966), Østrem and Pytte (I960), Reference TvedeTvede (1973), Reference ZieglerZiegler (1974) Reference HaakensenHaakensen (1975), and Kjeldsen (1975). Studies were continued during the period of construction of the tunnel system (Reference HagenHagen 1977, Wold and Haakensen 1978 Wold and Østrem 1979[a] and [b]). Since 1978, water discharge and sediment transport have been recorded at the point where meltwater enters the collecting tunnel under the glacier.

The tunnel system in the rock beneath Bondhusbreen provided access to the glacier bed and, thus, an exciting opportunity to make detailed measurements at the ice-rock interface. In order to make observations and to install instruments, ice tunnels were melted from the shafts along the glacier sole with the use of a hot-water system. Electrical cables from the instruments were run through bore holes from the ice-rock interface to the tunnel system in the rock, and thence to recorders. Once the instruments were installed, the ice tunnels closed by plastic deformation. The closing rate was 120 to 150 mm d^-1 in tunnels with diameters of 1 to 3 m.

The objectives of this paper are to describe methods used in this study and to present preliminary results. Further results will be presented in later papers (Hagen and others in preparation Footnote ^* ). The studies were initiated by B Wold and the field work was performed during the period 1980 to 1982, mainly by J 0 Hagen and B Wold.

Previous observations and measurements in sub-glacial cavities (Reference CarolCarol 1947, Boulton and others 1979, Reference TheakstoneTheakstone 1979, Reference VivianVivian 1980, Anderson and others 1982, among others) have yielded valuable information on subglacial processes. The experiments at Bondhusbreen extend these measurements to regions where the ice pressure is much greater and which were inaccessible up to now.

Hydrology

Melting on Bondhusbreen starts normally in the middle of Hay, and 85% of the total discharge follows during the summer period from June to September. A definite decrease in discharge occurs between October and December. Thereafter a stable discharge of 50 to 250 1 s⁻¹ is observed during the winter period from January to April. The decreasing discharge during the autumn, after surface melting has ceased, is inferred to be largely from water stored in the glacier during the summer. The stable winter flow is interpreted to be water released from storage in the glacier and in the lake, Holmavatn. Melting caused by frictional and geotnermal heat is at least an order of magnitude less than the observed flow.

The volume of the mean annual discharge during the years 1978 to 1981 was 60x10⁶ m³. Of this, 10% originates from Holmavatn (Fig. 2), Tracer experiments showed that the total run-off from this lake drains into the diverted tunnel system at intake (3) in Figure 2. This intake shaft catches about 90% of the water, whereas about 10% runs through shaft (1) and a minor amount, often under high pressure, comes through small drill holes in the bedrock. Some of the latter are relatively far from the main intakes. The entire winter discharge drains through the main intake at (3) in one single concentrated stream.

During the first days of the melt season, water appears to flow in multiple small intra- and sub-glacial channels. Later, the discharge gradually stabilizes in the main channels. Measurements at the glacier front indicated that almost all available water above the intake area was caught by the sub-glacial intakes.

Sediment Transport

Sediment transport in the diverted water has been measured since 1978 (Reference KjeldsenKjeldsen 1979, 1980, 1981). Annual transport of coarse material ranged from 2.8 × 10³ to 5.2 × 10³ tonnes. Annual transport of suspended material is about 4 × 10³ tonnes, so the total mean annual transport amounts to about 7.6 × 10³ tonnes. The amount of solid matter transported varies with water discharge and frequency of flash floods. The concentration of suspended sediment ranges normally from 0.015 to 0.070 kg m⁻³ increasing to a maximum of 0.785 kg m⁻³ during flash floods.

The content of solid material incorporated in the ice was determined by collecting ice samples, each of about 0.01 m3 from different places in the tunnels. These samples contained, in general, much more debris than the water did, the greatest concentration being about 60 kg m⁻³ (Fig. 3). The greatest concentrations occurred in a layer 1 to 2 m thick at the sole; the amount decreased rapidly upwards. However, the cleanest basal ice was located near what is inferred to be the normal course of a subglaeial water channel; this ice had a total debris content of only 0.030 kg m⁻³. In order to estimate the concentration of material likely to be suspended, the coarser (diameter >0.5 mm particles were excluded and the concentration data replotted in Figure 4. The maximum concentration of this material was about 15 kg m⁻³.

Fig. 3. Distribution of material load in the basal ice. Samples 6, 11 and 12 represent clean ice close to the subglacial stream shown in Figure 5.

Fig. 4. Fine material (diameter <0.5 mm) load in the basal ice compared to the amount of suspended load (diameter <0.5 mm) in the subglacial stream during (A) normal discharge and (B) flood discharge.

It is inferred that the ice near the subglaeial river is clean because the running water melts the surrounding dirty glacial ice, resulting in inflow of cleaner ice from above. Large increases in sediment transport which are observed during floods are partly due to melting of debris-filled ice outside the ordinary channel.

If a subglacial stream crosses more or less diagonally over the glacier bed, ice will move across the stream course. Melting of this basal ice will then release debris to the stream. Such conditions will affect the mode of transport of the debris, i.e. whether it is transported englacially or by the subglaeial water.

The amount of material transported englacially can be calculated approximately by multiplying the debris content of the basal ice by a mean annual velocity of 30 m a⁻¹ and by the width of the glacier. Comparison of this value with the measured transport in the subglacial stream suggests that only 10% of the transport occurs englacially while 90% is transported in the river. Movement of subglacial moraine, if any, is not included in this calculation.

Pressure Measurements

Ice pressures were measured by ten sensors (P₁ to P₁₀) in three different installations (Figs.5 and 6). P₁, P₂ and P₃ were mounted in an artificial roche moutonnée in December 1980 (Fig. 7). This construction was made of steel bolted to the bedrock and filled with reinforced concrete. A teflon sheet 2 mm thick covered the top of the construction in order to protect the sensors. This protection was removed from P₁ and P₂ during the summer of 1981, probably by scouring. The sensors were oriented with P₂ parallel to the general rock slope (30°) in the area, P₁ inclined 4° down-glacier, and P₃ inclined 65° down-glacier.

Fig. 5. The tunnel and ice cavities melted out in the vicinity of intake (3) (Fig. 2), showing the terrain and the position of the instruments. R: artificial roche moutonnée, L: dome-shaped construction, P₄ and P₅: pressure gauges in the bedrock, S: strain gauges, K₁ and K₂: small spheres with strain gauges and thermistors (S and K are not described in this paper). Contours are on the glacier bed. Contour interval is is 0.5 m.

Fig. 6. Longitudinal profiles on the bedrock. Locations shown in Figure 5.

Fig. 7. Construction of roche moutonnée with pressure gauges P₁, P₂ and P₃.

Sensors P₄ and P₅ were mounted at inclinations of 38° and 86 , respectively, in shallow cavities hollowed out of the bedrock surface slightly up glacier from the roche moutonnée in May 1981 (Figs.5 and 8).

Fig. 8. Cross-section along profile EF in Figure 5.

In December 1981, another subglacial tunnel was excavated up-glacier from the roche moutonnée. During this tunnelling, an almost vertical bedrock wall was found only a few metres upstream from sensors P₁ to P₅ (Figs.5 and 6). In the upper part of the tunnel a dome-shaped construction of steel and reinforced concrete, containing five pressure sensors P₆ to P₁₀ was placed on the bedrock (Fig. 9). The bedrock in this area had a general slope of 35°. The sensors were placed symmetrically to simplify the interpretation of pressure and ice flow. The top sensor (P₇) was parallel to the general bedrock slope, the others dipped 15° relative to P₇.

Fig. 9. Dome-shaped construction with pressure gauges P₆ to P₁₀

The pressure sensors used were standard instruments employing a vibrating wire system (Fig. 10). They were calibrated at 0°C and had an accuracy of ± 1%. Although the signals were recorded some distance from the sensors, the vibrating wire system is so designed that resistance in the wires and interference from spurious electrical fields do not influence the accuracy. Pressures were recorded at intervals of 1 h, using a digital data-logger. This recording frequency is considered to be adequate for present purposes. As the study area (Figs.5 and 6) was situated under the upper part of a major ice fall, where surface velocities reach Q.30 to 0.35 m d^_1, relatively rapid movement and high dynamic pressures were expected at the bottom of the glacier. The hydrostatic pressure under 160 m of ice is about 14 bar, so the pressure sensors were designed to span the range from 0 to about 50 bar.

Fig. 10. Cross-section through a pressure gauge. (1): steel plate, (2): recessed bolt, (3): binding ring with recessed screw, (4): steel ball, holding the membrane, (5): thermistors drilled into membrane, (6): vibrating wire, (7): membrane with steel rods, (8): magnetic system, (9): rubber gaskets, (10): steel box, (11): water-tight seal around electric wires. The vibrating wire is excited by the magnetic system. The change in frequency is directly related to the change in pressure applied to the membrane. Pressure P is then given by P = K(F² -F0² where K is a calibration constant and F0 and F are the frequencies before and after the pressure is applied.

It was intended that the main axis of each group of sensors should be parallel to the direction of movement of the basal ice. To this end, glacial striae were sought in the areas of the installations, and, upon re-opening of the tunnels, care was taken to observe the directions of striae on the various constructions. Directions on the same rock face showed local variations of about 30°, and some striae differed from the direction of ice flow on the surface by 50°. Data presented below suggest that such variability was also a natural feature of the flow field at this location.

The subglacial stream shown in Figure 5 plays an important role in our interpretation of the pressure measurements. This stream appears to remain in approximately the location shown throughout the year, as evidence for its existence has been found on a number of occasions. In December 1981, for example, the stream could be heard from the artificial tunnel, and in two places water from it leaked into this tunnel.

During the winter period from January to April 1981, sensor Pi recorded a stable pressure of 18 to 19 bar. Sensor P₁ normally showed a pressure between 17 and 18 bar, but there were several sudden decreases of about 2 bar over a period of an hour or so, followed by a slow increase during the following 25 to 30 h. Sensor P3 recorded a pressure of 16 to 17 bar for several weeks, and then slowly decreased to 13 to 14 bar. This sensor showed a number of abrupt decreases of about 1 bar, lasting for 1 to over 30 h, followed by equally abrupt increases. Figure 11, covering the period from 9 to 16 March 1981, illustrates these features.

Fig. 11. Pressure recordings from the roche moutonnée in March 1981.

During the summer season of May to October 1931, relatively even pressures of 16 to 17 bar were recorded on all three sensors on the roche moutonnée. However, during long periods P3 had a higher pressure than Pi (Fig. 12, 10-11 July 1981). This suggests slow ice motion towards the subglacial water channel where the continuously melting ice is presumably replaced by plastic inflow. Alternatively, it could possibly represent some sort of large-scale lee-side effect in the lee of the cliff up-glacier from the installation (Figs.5 and 6).

Fig. 12. Pressure recordings from a period in July 1982 when a cavity was present.

These recordings suggest that the flow of ice along the sole at this location must be very slow. Measurements of sliding velocity here, using a segmented wooden rod inserted into the ice through a bore hole in the rock and later re-excavated, showed a velocity of less than 2 mm d^-1 during spring. However, there must also be a dynamic effect since the pressure was consistently 3 to 4 bar above the hydrostatic pressure during long periods.

During the summer, disturbances were also observed during periods of increased water discharge which were presumably accompanied by widening of the subglacial stream. An example of this is shown in Figure 12. The pressure at all three sensors increased on 12 July 1981 and then fell suddenly to zero. It is supposed that the increase in pressure was due to transfer of pressure to the roche moutonnée as an open tunnel or cavity (at atmosphere pressure due to the proximity of the intake at (3)) migrated laterally toward the roche moutonnée. The decrease in pressure would then represent uncovering of the sensors as the cavity passed successively over them from Pi to P3. Such a pressure variation would be consistent with Weertman' (1972) theoretical solution for the pressure distribution around a subglacial channel. This cavity could have been formed by lateral migration of the stream channel. The water discharge increased from 3 to .5 m³ s⁻¹ in a period of seven days before the event shown in Figure 12.

It is interesting to note that as the ice re-covered the roche moutonnée it moved up-glacier, reaching P₁ first and P₂ last.

The pressure stabilized during the fall and winter of 1981-82. P₃ initially showed a somewhat higher pressure than P₁ but this pressure distribution changed slowly, so that by the beginning of March 1982, the situation was as follows: 16 bar at P₁, 16.5 bar at P₃, P₂ bar at P3, Pg being the highest. Sensors P4 and P5 (Fig. 8) were more stable than P₁, P₂ and P₃. P₄ showed variations between 14 and 18 bar, and P₅ between 8 and 14 bar. The difference between these two sensors was initially 4 to 5 bar, reflecting the leeward position of P₅. This difference gradually decreased during the fall and winter, finally reaching zero near the end of March 1982.

Although the pressure at P5 was never below about 8 bar, separation of the ice from the bed apparently occurs here, as a light-brown iron-manganese(?) precipitate was observed in leeward positions on the adjacent rock. It is suggested that these precipitates form in subglacial waterfilled cavities (Andersen and Sollid 1971, Sol lid 1980). Separation is possible in such situations if the water pressure is high enough (e.g. Reference KambKamb 1970: 720).

Quite different pressure conditions were observed at the upper dome-shaped construction (Figs.5 and 9). During the period from January to April 1982 relatively stable readings were obtained, 26 to 29 bar at P₆, 7 to 11 bar at P₇, zero at P₈, 4 to 7 bar at Pg, and 19 to 21 bar at P₁₀ (Fig. 13). There was an obvious "stoss" effect on P₈ and Pin, and a "lee" effect on Pg and P9. All leeward sensors indicated more variable pressures than the stoss-side ones.

Fig. 13. Pressure recordings from the dome-shaped construction in March 1982.

This period of stability ended on 12 April 1982, when the upper sensor Pg experienced a dramatic increase in pressure to 90 bar before it broke (Fig. 14). It is presumed that a boulder of about 0.3 m³, which was observed about 4.3 m up-glacier from the sensor in December 1981, moved across the sensor. In connection with this event, disturbances were observed on the other sensors. A study of the pressure variations (Fig. 14) suggests that the boulder hit P5 and destroyed it, and may have then moved over the construction between P7 and Pg (Fig. 9).

Fig. 14. Pressure recordings in the period when Pg was destroyed by a boulder.

The mean velocity of the boulder was calculated to be 40 mm d^-1, based on its initial position (Fig. 5) and on the interval of time required for it to move to the sensor. The ice velocity, measured by a wire running from a block of wood fixed in the ice, down through a bore hole, and thence to a recorder in the tunnels, was 80 mm d^-1 in June 1981. Obviously boulders at the sole of the glacier should move more slowly than the glacier ice itself, due to friction against the bedrock. This results in a leeward effect in front of the boulders, which can be clearly seen by the pressure drop at Pg approximately two weeks before the boulder reached the sensor. Apparently the the lee effect extended roughly 0.5 m, or somewhat less than the diameter of the boulder (0.7 m), in front of the boulder.

These pressure observations suggest that flow in the basal ice is unsteady on an annual time scale, with natural cavities opening and closing, and with direction of flow changing radically. The latter is supported by observations of striae on the glacier bed exposed in the artificial tunnels. Striae on the same rock face frequently differed in orientation by 30°, and from one face to another; over a distance of only a few metres, they could vary by 120°. For example, the roche moutonnée was originally aligned parallel to striae on the rock surface on which it was placed, but striae that developed on the roche moutonnée itself suggested flow, which actually had an up-glacier component, at an angle of 120° with respect to the axis of the construction. These could have been made during closure either of the artificial cavity or of cavities formed during the summer. Striae up-glacier from the dome construction suggested that the boulder (Fig. 5) would pass a couple of metres to the south of the installation, but instead the boulder appears to have hit the construction directly.

Temperature Measurements

Each pressure sensor also contained one or two thermistors placed about 2 mm under the surface of the steel membrane. The thermistors were Fenwall UNI-curve, UUD 23JI with a standard resistance of 760 Ω at 0°C, and R/ΔT = 30 Ω per °C.

Unfortunately, for technical reasons, the temperature measurements were never completely successful. Thermistors subsequently placed in the dome construction were carefully calibrated, but water gained access to the connection boxes, so consistent results were obtained for only a few of the thermistors, and for only part of the duration of the experiments. Measurements of temperature differences AT were more stable and reliable than the measurements of temperature T itself.

Some of the results from thermistors in sensors P₄ and P₅ (Fig. 8) appear to be reliable. The temperature difference ΔT (P₅ - P₄), quickly stablized at about 0.05°C, Pa being 0.05 C colder than P5. This is consistent with expectations, as the difference in pressure between these two positions was then about 5 bar, with P4 reading about 15 bar and P5 about 10 bar. In the winter of 1982, the pressures at Pa and Pc were nearly equal at about 14 bar, and ΔT (P₅ - P₄) was nearly 0 (< 0.015 deg).

In the dome construction, we intended to measure temperature differences between P₆ and P₈ as well as between P₉ and P₁₀. and absolute temperatures at P₆, P₇, and P₈. The temperature difference ΔT (P₆ - P₈) seems to have been stable for several weeks in the winter, P₆ being about 0.025 deg lower than P₈. The pressure difference Δ P (P₁₀ - P₉) was almost constant at 28 bar. Similarly the difference AT (P₁₀ - P₉) was 0.01 deg, P₁₀ being colder, while ΔP (P₁₀ - P₉) was 10 to 12 bar. These results suggest that large differences in pressure do not produce as large differences in temperature as expected. This may be because water is squeezed out of the ice on the stoss side (Reference RobinRobin 1976), so that ice in the lee of the obstacle lacks water for re-freezing, and thus cannot wann up.

Conclusions

Since the main water intake under Bondhusbreen came into operation, almost all of the available water has been captured. In the spring, meltwater first enters the tunnel system through numerous small drill holes and in small streams through the larger intakes. Within a few days, however, it concentrates in one or two main streams at intakes (1) and (3), with minor drainage elsewhere. About 90% of the eroded material passing this cross-section of the glacier is transported by the meltwater while the rest is carried englacially in the basal ice.

There are large local variations in ice movement along the glacier sole. In leeward positions, ice masses can be stagnant with almost equal pressure in all directions for large parts of the year. As water discharge increases in the summer, however, ice in these locations becomes more active. Elsewhere a general ice velocity of 80 mm d^-1 is observed, with consistent differences in pressure between the upper and lower sides of obstacles. Pressure of about 28 bar occurred on the up-glacier side; very little pressure was recorded in leeward positions of an artificial obstacle.

During the course of one experiment a large boulder, approximately 0.7 m in diameter, is inferred to have moved across one of the pressure sensors. The boulder moved with a speed of about 40 mm d^-1, and was preceded by a low-pressure zone about 0.5 m long. The boulder exerted a pressure greater than 90 bar on the sensor.

Temperature measurements, unfortunately, were less successful. The temperature differences ΔT were one order of magnitude less than expected, in relation to the observed differences in pressure ΔP.