Hydrological characteristics of the drainage system beneath a surging glacier

Abstract

A rare combination of natural circumstances permits assessment of current theories on water flow beneath glaciers. Outburst floods from the subglacial lake Grı´msvöGrı´msvö tn in Iceland took place before, during and after surging of Skeiðará rjö kull, the glacier beneath which the outburst floods drain. The observable drainage patterns associated with these floods show the different nature of the basal water conduit system of the glacier during surge and non-surge phases. During surge conditions, basal water is dispersed slowly across the bed in a distributed drainage system; but when the glacier is not surging, water is transported rapidly through a system of tunnels. Water plays an important role in the dynamics of temperate and subpolar glaciers: by lubricating the glacier base, water facilitates the sliding of ice over its bed. Some water is produced at the bed by frictional and geothermal heat, but the main water supply is melt water from the glacial surface which descends through the glacier to the bed through various internal conduits. Occasionally, water is injected directly at the bed from ice-dammed lakes or during subglacial volcanic eruptions. Drainage along the bed itself is through a system of passageways controlled by the composition and topography of the substratum, the thickness and slope of the overburden ice, and the water supply. Theoretical analyses and field observations suggest the existence of two main types of subglacial drainage systems which may, however, represent the opposite poles of a continuum. In the first type of system-that which is normally present according to current thinking-water is conducted through tunnels that have been incised upwards into the ice, forming a river-like system of widely spaced tributary branches. These join down-glacier to form a relatively small number of outlet rivers, which emerge from the glacier terminus 1,2. In this case, the steady-state water pressure of the passageways varies inversely with water flux 1 because the size of the tunnels adjusts to the flux of water by enlargement of the tunnel walls as a result of frictional melting and contraction caused by inflow of ice. The tunnels transporting large amounts of water grow at the expense of smaller tunnels. However, such ice tunnels cannot adjust to the injection of water over a short time span of hours to days, so there may be a transient increase in water pressure, which reduces basal friction and facilitates sliding. Input of water to this type of drainage system affects sliding of the glacier only within small, local areas. In the second type of drainage system, water is dispersed across the bed through a distributed network of passageways of variable width. The largest of these may be visualized as water-filled cavities which form behind protuberances in the glacier bed and are hydraulically interconnected, either by small channels cut in the bed, or by narrow ice tunnels located on the downslope side of small steps in the bed. Such chambers lie perpendicular to the ice flow 3-7. In this kind of system, water pressure varies directly with the water flux 5,6. There is no tendency for larger passageways to collect water and grow at the expense of smaller ones, and water emerges from the glacier margin in many small outlets. Discharge through the system is controlled by the throttling action of the narrow passageways. High pressures is required to drive water through such a drainage system; this lubricates the bed and facilitates sliding over large areas. We refer to a system of this type as a distributed (or linked-cavity) system. Both types of drainage system may exist simultaneously beneath glaciers, and may change with seasonal and diurnal inputs of melt water. For instance, a linked-cavity system, conducive to rapid sliding but carrying only small amounts of the melt water, may form temporarily under some parts of a glacier which is otherwise drained by a tunnel system; this formation may be in response to sudden injections of water to which the tunnels cannot immediately adjust. Glacial surges may occur when the normal, rapid subglacial drainage is disrupted and a linked-cavity system spreads out beneath the glacier and persists for some time 8-11. On the basis of observations of long travel times of dye tracers during a surge of the Variegated glacier in Alaska, and of theoretical analyses of a linked-cavity system, Kamb et al. 9 and Kamb 6 suggested that surging was initiated by a switch from a tunnel system to a distributed system that facilitated sliding. Further, they suggested that the surge continued as long as high water pressure sustained rapid sliding, the sliding itself counteracting the enlargement (by melting) of the orifices connecting the water-filled cavities. Moreover, the observed rapid and concentrated drainage of water from the glacier following the surge was believed to indicate that a tunnel system had been re-established and that this conversion in fact terminated the surge. Kamb's 6 model predicted that for sufficiently large perturbations, in the form of increased water pressure or decreased sliding, the narrow passageways linking the cavities would grow unstably, and an ice tunnel system would be restored. Because subglacial drainage systems themselves are inaccessible, and surges infrequent, normal field observations can provide only limited support for models such as Kamb's. However, the massive discharge of water from an ice-dammed lake (a glacier outburst flood or jökulhlaup), draining along the glacier bed, provides a unique opportunity to test current hypotheses by comparing outburst floods which occur during glacial surges with those that occur during normal periods. The former are relatively rare; but a series of outburst floods from the subglacial ice-dammed lake Grímsvötn in the interior of the ice cap Vatnajökull (8,100 km 2) in Iceland drained beneath the glacier Skeiðarárjökull both before, during and after its 1991 surge (Fig. 1). Observations of these events are consistent with theories of basal drainage through tunnels during the non-surge state and distributed drainage during the surge, as discussed below. They are likewise in accord with the hypothesis that surges are caused by a temporary transition from tunnel to distributed drainage.