Before the floods

Prior to mid-July 1987, run-off from Bas Glacier d’Arolla had been weak (Fig. 2.). The transient snow line was at a lower elevation than is normal for the time of year, and supraglacial and englacial drainage channels were poorly developed. Water levels in moulins were observed to be close to the glacier surface, suggesting some throttling in the drainage system. High ablation rates (average 6 mm d₋₁, maximum 10 mm d^-1) were measured during the 2 weeks before the flood events. High-intensity rainfall (maximum 4.4 mm h⁻¹) commenced at 1700 h on 15 July, petering out by midnight.

The Overflow event of 15 July 1987

The first flood event of July 1987 — the “overflow event” — occurred at 2150 h on 15 July, when a loud crash was heard both in the village of Arolla, about 2 km away, as well as in the field camp, sited 500 m from the glacier. From 2150 to 2220 h, large volumes of water were observed flowing over the snout of the glacier. Flow emerging from the portal subglacial stream fell during this half-hour period, being small in relation to that flowing over the snout (ratio approximately 3:1 overflow to underflow), and then returned to its former level. The peak flow of the event was over the 5 m³s⁻¹ design capacity of the Grande Dixence flow gauge (to prevent damage to underground installations, flow is automatically diverted away from the stage recorder when flow reaches 5 m³s^-1). Flow prior to the event was around 3.75 m³s^-1.

The event caused significant geomorphological changes in the proglacial zone. Supraglacial moraine carried over the snout of the glacier by the overflow was initially dumped in a large cone in front of the snout (this probably caused the loud crash at the start of the event). This debris, together with large quantities of ice from the margins of the glacier, was then transported downstream. Boulders (500-600 mm diameter) set in a cobbly, gravel-sand matrix were purged from the Grande Dixence sediment trap in an almost continuous sequence during the night of 15-16 July (and during the following 5d). Suspended-sediment concentrations, determined from hand samples collected from the main channel during the flood were very high (31 108-39 476 mg l^-1) in comparison to the mean concentration for the ablation season (1200 mg l^-1 concentration estimated from a calibrated turbidity record). The ratio of bed load to suspended-sediment load transported during the overflow event was approximately 1:1. Electrical conductivity rose from around 13 μS^-1 before the event to around 18 μm^-1 during the event, then fell abruptly to around 10 μm^-1 immediately after the event (Fig. 3.).

The high flow period of 16-18 July 1987

Rain continued to fall during 16 and 17 July but at much reduced intensities. Flow remained high but variable during the 4 d following the overflow event (Fig. 3.). Discharge on 16 July (approximately 3.6-4.6 m³s^-1) was generally much less than the overflow peak, but surges of flow around 5 min duration occurred throughout the day. Throughout 16 July, flow was observed surging out from the glacier along new outflow channels on either side of the central portal stream (Figs 4 and 5). During the afternoon of 16 July, several discharge surges (e.g. increases from 4.11 to 4.44 m³s^-1 in 5 min) and sediment pulses (e.g. increases in suspended-sediment concentration from 950 to 1600 mg 1^-1) occurred. These events involved flow along an increasing number of marginal channels (especially on the western side of the glacier). The marginal streams were seen to be dry at 0730 h on 17 July, but were carrying flow by 0940 h. By the afternoon of 17 July, the duration of the flow surges had increased to around 15 min.

Throughout this period, the main proglacial stream switched from braiding to down-cutting modes in response to the surges in discharge. Considerable erosion of channels close to the glacier occurred, mainly as a result of the collapse of banks of unconsolidated material. Electrical conductivity remained low (at 10 μS cm⁻¹) during the first part of 16 July then rose to around 15 μS cm^-1.

The subglacial event of 18 July 1987

The subglacial flood which occurred at around 1000 h on 18 July 1987 was the largest of the sequence, and caused the greatest difficulties for Grande Dixence S.A. (not least because it occurred on a Saturday). The peak flow during the event is estimated (by timing floats in the Grande Dixance S.A. weir) to have been around 10 m³S^-1. This is about three to four times greater than the 2.45 m³s^-1 flow listed as a “minimal outburst” from Bas Glacier d’Arolla in Reference HaeberliHaeberli’s (1983) review of floods from Swiss glaciers. Heavy rain storms occurred throughout 18 July then declined during 19 July. Flows remained high throughout the weekend and thereafter declined to pre-flood levels (Fig.3).

High rates of sediment transport and channel change occurred during this final flood event. The sluices on the Grande Dixence S.A. sediment trap were kept open throughout the event. Large ice slabs (the largest having dimensions of a = 1.5 m, b = 0.7 m, c = 0.3 m) were imbricated in the valley train. The pattern of channel change involved initial widening and braiding followed by down-cutting and a transition to a single thread of dominant flow (Fig. 5.). Electrical conductivity readings during the event were sparse (due to equipment break-age) but were around a relatively high 16 μS cm^-1.

Morphologically altered moulins

Measurements of the internal geometry of the major moulins on Bas Glacier d’Arolla had been made in the 3 weeks prior to the flood events. Measurements made on 16 July showed that the two moulins located in the central valley of the lower tongue of the glacier were both considerably altered. The more westerly of the two had been cleared of the debris which hitherto had choked its shaft. The more easterly moulin had developed a new surface outlet 3 m downstream of its previous stream entry point (Fig. 6.); this could have been the entry point to the moulin in the previous ablation season. Both moulins had clear wash lines on their walk well above supraglacial stream entry levels.

Fig. 6.

The internal morphology of the eastern moulin. Measurements were made during a controlled abseil descent.

A new ice-cut channel leading from the moulins to the snout

A new channel cut into the surface of the glacier, running from the eastern moulin through the moraine of the lower tongue of the glacier to and over the snout had also been created (see Fig. 4.). An ice-cut channel running from the lower lip of the western moulin joined this main channel. The channel was measured as 139 m in length, from 2.5 to 13.8 m in width, from 0.3 to 0.9 m in depth and from 4° to 25° in slope.

Fissuring of the glacier surface

A set of arcuate fractures running across the glacier (concave to the snout) had also been created. The fractures were spaced 4-10 m apart and were up to 10 cm wide. One fracture, 25 m downstream of the eastern moulin, ran across the full width of the glacier. The fractures tended to focus on the two large moulins described above. The largest concentration of fractures occurred on the eastern side of the glacier. The front face of the glacier also appeared to have bulged out during the event.

All of the above features were observed on the day following the overflow event. Observations made before, during and after the subglacial event of 18 July suggested that the separation distance between the glacier sole and the underlying rock bed increased during the 18 July flood (Fig. 7.).

Fig. 7.

Photographs of the snout of Bas Glacier d’Arolla before and after the subglacial outburst of 18 July 1987. Upper photograph taken at 0930 h, 18 July 1987; lower photograph taken at 1030 h, 19 July 1987. Note the increase in distance between the glacier sole and the debris bed.

Top-down or bottom-up surcharging?

Three lines of evidence support the hypothesis that the surcharging of the moulins during the overflow event was due to a bottom-up rather than a top-down filling process. The first of these lies in the discharge capacity of the supraglacial channel system above and below the moulins. Measurements of the slope and channel capacity of the supraglacial streams running into (from upstream) and out of (downstream from) the moulins were made after the floods had subsided. The capacity of the channel downstream of the moulins is much greater than would be expected were the waters responsible for carving the downstream channel to have passed through the up-stream channel network (Fig. 8.). Abney level surveying of the slope of the channel upstream of the moulins indicates slopes in the range 4-7°, while the channel below the moulins (but above the front face of the glacier) has slopes ranging from 4° to 25°. The conveyance potential of the downstream channel is thus much greater than that of the upstream channel, to a degree which is not consistent with the supraglacial conveyance of supraglacial waters into and beyond the moulins. A flow input from the moulins is implied.

Fig. 8.

Channel capacity (cross-section area) of the supraglacial stream system upstream of the moulins and the overflow channel system downstream of the moulins.

The observed increases in both suspended-sediment concentration and electrical conductivity also support a bottom-up surcharge hypothesis. A top-down supply would require the suspended-sediment load to be derived from the proglacial zone (given the paucity of fines on the glacier surface). While it is known that some fines are derived from the proglacial zone (especially during extreme flood events), the overwhelming majority are known to be derived from subglacial sources (Reference Gurnell and WarburtonGurnell and Warburton, 1990). Whilst unremarkable in itself, the observed rise in electrical conductivity during the overflow event (from 13 to 18 μS cm₋₁) is also consistent with the release of subglacial as opposed to supraglacial waters.

The third line of evidence in support of a bottom-up over a top-down filling hypothesis lies in the higher likelihood for water being stored within the glacier (in the conduit system, in bed cavities or in water pockets) than on top of it (the snowpack would have been the only possible water source or store).

Our conclusion is therefore that the overflow event was caused by an upwelling, and subsequent overspill, of water in the portal-area moulins. Similar rises of water in moulins have been reported for Storglaciären, Sweden, where water levels reaching 80% of moulin depth have been observed on numerous occasions, and one case of a rise to 97% of moulin depth has been reported (Reference Holmhund and HookeHolmhund and Hooke, 1983). Geyser-like spouts of water issuing from vertical fissures on glaciers in Spitsbergen, Greenland and Switzerland have also been documented (Reference BaranowskiBaranowski, 1973).

We can evaluate the water pressure needed to drive flow up the Bas Glacier d’Arolla moulins as follows. Treated as a problem in hydrostatics, and ignoring viscosity effects, the water pressure (P _w) at the base of a steeply inclined shaft of constant diameter and height h is given by:

(1)

where w is the density of water (1000 kg m^-3) and g is acceleration due to gravity (9.81 m^-1 s^-2). On the evidence of valley and glacier geometries, and data from seismic surveys (Reference Knecht and SusstrunkKnecht and Susstrunk, unpublished) and thermal probing studies (personal communication from A. Bezinge), 60 m would appear to be an appropriate value to take for glacier thickness in the vicinity of the moulins. On being filled, the water pressure at the base of the moulins would hence be 0.59 MPa. A water pressure of over 0.6 MPa is thus implied for the overflow event.

Source of the excess water

For such a short-lived (30 min), high water-pressure (> 0.6 MPa) episode to occur, there must have been a sudden arrival of water in large quantities. This could have resulted from the rupture of a stored water pocket, or could have marked the arrival of a travelling wave of high water pressure, of the type monitored on Stor-glaciären by Reference Holmhund and HookeHolmhund and Hooke (1983). Both hypotheses appear plausible. The 1974 minor outburst from Bas Glacier d’Arolla was attributed to the rupture of an internal water pocket. Geophysical surveys of the tongue of the glacier (Reference Knecht and SusstrunkKnecht and Susstrunk, unpublished) revealed a topographic low near the snout. The relatively low conductivities recorded during the event are, however, inconsistent with the release of water held long term in a subglacial store. The travelling-wave hypothesis appeals in view of the high melt rates of the previous fortnight, the heavy rainfalls of the previous 5 h and the retarded state of the englacial drainage network. Given that the rainfall was able to make its way to the bed of the glacier in the upper area of the glacier tongue, it could have augmented water held in the conduit-cavity system, raising water pressures to supercritical levels, causing local bed separation, further integrating the drainage system and thence initiating a growing wave of high water pressure moving down-glacier (cf. Reference Holmhund and HookeHolmhund and Hooke, 1983). On balance, we prefer to think of the source of water for the overflow event in terms of the addition of recent rain and snowmelt waters to accumulated meltwaters held in an hitherto poorly integrated subglacial drainage system.

Why was bottom outlet not achieved during the overflow event,yet was thereafter?

The rise of water in the moulin shafts must have been a response to surcharging of the subglacial conduit system. The maintenance of some flow from the main portal stream during the overflow event, and the subsequent reversion to subglacial flow about 30 min later, implies that the subglacial outlet was partially and temporarily blocked (e.g. by localized roof collapse, by general ice collapse or by accumulation of debris). Were blockage not to have occurred, the head-discharge effect would have driven higher than normal flows through the subglacial stream. This was not the case (subglacial flows fell during the overflow event). The restriction responsible for causing water to be forced up the moulins rather than out through the portal conduit was thus probably a combination of limited hydraulic capacity and partial conduit blockage. The rise of waters in the moulins and the subsequent overspill would have coped with the initial peak of the flood, and could have forced the removal of any temporary blockage in the subglacial conduit system. An alternative explanation for the short-lived nature of the overflow event is that the high water pressures generated were able to cause ice-bed separation (hydraulic jacking), which facilitated the resumption of subglacial flow. The fracturing of the ice in the snout zone is good evidence for displacement at the snout but does not necessarily mean that the glacier was moved upwards and forwards. However, the question as to whether or not hydraulic jacking could have taken place at Bas Glacier d’Arolla merits discussion.

Could hydraulic jacking have occurred?

The critical water pressure (P _c) needed to achieve uplift under model bed conditions has been defined by Reference IkenIken (1981), and applied by Reference Röthlisberger and IkenRöthlisberger and Iken (1981) and Reference Iken, Röthlisberger,, Hotron and HaeberliIken and others (1983). The critical water pressure for uplift becomes smaller as the slope of the bed (a) increases and the reverse slope of undulations or steps on the bed (ψ) decreases (Reference Röthlisberger and IkenRöthlisberger and Iken, 1981). Ice-overburden pressure (P ₀) is given by:

(2)

where i is the density of ice, g is acceleration due to gravity, h is mean thickness of glacier ice and a is the mean surface slope. The critical water pressure (P _c) for translatory displacement is then given by:

(3)

where τ_b is basal shear stress τ_b = igh sin α, or, taking account of valley-wall effects, τ_b = figh sin α; f is a shape factor (Reference PatersonPaterson, 1981)) and ψ is the backward tilt angle of steps forming an undulatory hard bed. To apply the model to the snout zone of Bas Glacier d’Arolla, we need to provide values for h, α and ψ We have previously determined 60 m to be an appropriate value for h. Table 2 shows evaluations of overburden pressure (P_o), basal shear stress (τ_b) and critical water pressure (P _c) for a range of realistic bed-slope angles, assuming the presence of bed steps or undulations with a backward slope angle (ψ) of 2° (implying a gently undulating hard bed). The upper set of values neglects valley-side effects upon τ_b; the lower set incorporates this effect, assuming a value of 0.75 for f. It would seem that a water pressure of around 0.25 MPa would be sufficient to cause hydraulic jacking of the glacier, were the assumptions of the model satisfied. The model defines a rigid, undulating bed with free water access to all points on the bed. The properties of the bed at Bas Glacier d’Arolla, as far as they are known, do not conform to these assumptions especially well. The rock bed is covered by a layer of unconsolidated glacial sediments (Reference Knecht and SusstrunkKnecht and Susstrunk, unpublished). Were this to be in a frozen or semi-frozen state, as is thought to be the case for many glaciers in the Valais region of Switzerland (Reference HaeberliHaeberli and others, 1989), or were the till/rock bed to be hydraulically sealed and saturated, the necessary water pressures for uplift of the snout may still be attainable under circumstances of rapidly increased water supply. We have already noted that for water to be driven 60 m up a vertical shaft implies the occurrence of a water pressure of over 0.6 MPa. Were bed conditions favourable, it would seem that hydraulic jacking would have been possible under the basal water regime existing during the July 1987 flood phase. Despite the fact that the bed at Bas Glacier d’Arolla does not correspond closely to the hard-bed type, it would seem, given the high water pressures, that hydraulic jacking cannot be dismissed as a possible generating mechanism for the resumption of the subglacial drainage after the overflow event (and for the outburst event of 18 July).

Table 2.

Overburden pressure (P₀), based shear stress (τ_b) and critical water pressure (P_c) calculations for bed slopes (α) of 2°, 5° and 10°, under an ice thickness of 60 m and a bed step/undulation dip (ψ) of 2°

The observed fissuring of the lower tongue of the glacier—cause or product of the overflow event?

As noted above, the observed fissuring of the lower tongue of the glacier after the overflow event lends support to the hydraulic jacking hypothesis. Whatever the precise cause of the surcharging of the basal and englacial conduit system during the overflow event, there would have been a rapid rise then fall in water pressures. This would have exerted pronounced loading then unloading stresses on the adjacent ice mass, whether or not uplift occurred (Reference RöthlisbergerRöthlisberger and Lang, 1987). Such circumstances have been noted to induce icequake activity and surface fracturing on Storglaciaren (Reference Holmhund and HookeHolmhund and Hooke, 1983). Were the fissuring of the glacier to have occurred at an early stage during the flood, fracturing of ice could have caused temporary blockage of subglacial conduits. Thus, the overflow event could have been a product of the fissuring. If ice blockage of subglacial channels did occur, high concentrations of ice blocks would have been expelled into the proglacial stream on the resumption of subglacial flow. This was not the case, at this time. On balance, our preference is to regard the fissuring as a product of the high water pressures exerted during the overflow event. Local ice fracturing and blockage of conduits may have occurred, but we do not regard the fracturing as a necessary precursor of the overflow.