Thermokarst Lakes No. 6 and 5

During the mid-1960s, crevasses formed at the upper margin of dead-ice occurrences within the creeping permafrost area (cf. Fig. 2). Two corresponding depressions progressively enlarged, forming the thermokarst Lakes No. 6 and 5.

Lake No. 6 fills up only during winter and spring. When the lake is full of water, the frozen lake bottom starts leaking and the lake slowly empties through the permeable permafrost active layer, through remaining dead-ice layers or through the underlying permafrost. Freezing water seems to seal the bottom of the lake depression in late autumn. Subsequently, the lake fills with water, ice and finally snow until melting starts in spring. At its deepest point, the bottom of Lake No. 6 has lowered continuously by about 1 m a⁻¹ since 1970; the related depression had an area of roughly 10 000 m² in the late 1990s (cf. Fig. 2). Although the enlargement of the lake was accompanied by various manifestations of thermokarst development such as undercutting of ice walls by circulating lake water, melting and down-wasting of exposed massive ice, small landslides in morainic material along the oversteepened lake shores, rock falls, or collapse phenomena in the permafrost active layer, signs of catastrophic emptying were never observed, or corresponding runoff changes were damped below critical levels by the retention function of the artificially regulated Lakes No. 3 and 1 lower in the drainage.

The surface of Lake No. 5 grew at an accelerating rate and asymmetrically towards the south, exposing there an increasingly high ice front. In 1994, the lake reached an area of 10 000 m² and a volume of some 50 000 m³ (Fig. 3). During the summer, the lake surface was more or less free of ice, and the water warmed to temperatures of about 2°C under the influence of solar radiation and high air temperatures. As shown by strong undercutting of the ice front and drifting of small icebergs away from it, the density increase due to warming triggered a process of thermal convection: warmer and denser surface water sinks to the lake bottom, flows towards the ice front, cools by ice melting and undercutting the ice front, thereby reducing its density, and rises again to the lake surface where it flows away from the ice front and warms up (Fig. 4). Such convection is strengthened by a positive feedback, in that melting of ice leads to a larger surface area of the lake, increased energy input and more melting, etc. This positive feedback together with the existence of massive ground ice represents the main cause for the accelerating lake growth.

Fig. 3. Growth of thermokarst Lake No. 5. (a) Planimetrie representation of the lake outline in selected years of the time period 1967–95 (note asymmetric lake growth towards the massive ground ice in the south; the sudden displacement of the northern lake shore is a consequence of the lake-level lowering in 1984). (b) Lake level and area over the period 1968–95. Lake level smoothly dropped some meters in 1985 and1988; assessment of future lake growth suggested an area of about 20 000 m² or a volume of100 000 m³ in the year 2000 (dashed line with arrow). In 1995, the lake was pumped out due to its increasing hazard potential.

Fig. 4. Schematic representation of energy input by solar radiation, convective heat transport to the damming front of buried massive ice, subsequent ice melt and water cooling of the lake water leading to a positive feedback which drives the non-linear growth of thermokarst Lake No. 5. After drainage of the lake, strong undercutting and vertical pear-like convexities of about 1 m height were found in the exposed ice front, presumably related to local thermic convection cells.

During the early 1990s, the last years of its existence, Lake No. 5 grew annually by about 1500 m² in surface area and 7500 m³ in volume. On two occasions in the 1980s, the lake level smoothly dropped by several meters without leaving any traces of surface runoff. These changes in lake level pointed to the existence of near-surface crevasses in the surrounding ice and permafrost, but the corresponding water losses did not significantly slow down the overall growth rate of the lake (Fig. 3). Photogrammetrically measured differential melt, subsidence and flow rates (Fig. 2; Reference Kääb, Haeberli and GudmundssonKääb and others, 1997) indicated that massive ground ice at the southern lake border continued right into the steeper slope towards Lake No. 3 where creep is accelerating (extending flow). Extrapolated lake growth would have reached this critical area with potential crevasse formation and fast lowering of the southern lake shore in the years 1998–2000 at the latest (cf. Fig. 3b). It was therefore decided to determine the geometry of the lake bottom from a rubber boat, using a microdigitizer echo-sounder in combination with a tachymeter and later a differential real-time global positioning system receiver. A total of 4200 points were measured with an estimated accuracy of ±0.1 m. Based on this information and the photogrammetric data, the area and volume of the lake at the expected outburst time were estimated at 20000 m² and 100 000 m³, respectively. A sudden outburst of such volumes could have produced peak discharges of several m³ s⁻¹ (cf. Reference HaeberliHaeberli, 1983), potentially triggering outbursts of Lakes No. 3 and 1 with corresponding debris-flow hazards at Lake No. 1, at the rock glacier front or on steep slopes lower down. The recognizable hazard potential from the quickly enlarging Lake No. 5 was therefore considered to be unacceptable (cf. further comments in the section on hazard-mitigation works).

Ice-marginal Lakes No. 4, 3 and 2

The topographic depression at the right glacier margin contains the ice-dammed Lakes No. 4, 3 and 2 (Fig. 1), the development and safety of which are strongly affected by changes in mass and geometry of the glacier.

Lake No. 4 started to grow in the late 1960s up to an area of some 5000 m² in the mid-1970s. During the following years of general glacier growth, the related depression was mostly filled year-round by snow and ice. From the late 1980s on, the lake enlarged again by successively melting towards the glacier, reaching an area of 6000 m² in the mid-1990s. Almost every year, Lake No. 4 slowly emptied in late fall. The normal lake drainage towards Lake No. 3 is partially along and partially beneath the glacier margin. Presently, Lake No. 4 is growing from its flat sediment bottom towards thicker ice and steeper, rocky parts of the glacier bed. Formation of slush avalanches in snow covering the outflow can cause suddenly increased discharge towards Lakes No. 3 and 1 during spring and early summer. However, the lake has remained empty during the past few years.

The outbursts of Lake No. 3 in 1968 and 1970 (Reference RöthlisbergerRöthlisberger, 1979) had occurred when the lake water rose to the flotation threshold of the damming ice and snowdrift, triggering subglacial drainage and progressive enlargement of a subglacial channel during early summer (cf. Reference NyeNye, 1976, for the principle of the outburst mechanism). After artificial water-level regulation via a subglacial drainage pipe, the lake was stabilized until recent years when the surface of the ice dam started to sink and approach the flotation threshold from above. Such a potentially dangerous situation had been anticipated on the basis of careful monitoring by regular visual inspection and photogrammetric analysis of the specially flown annual air photos. Between 1976 and 1989, the slightly advancing glacier had reduced the area of the lake surface (Fig. 5). More recently, glacier shrinkage enabled renewed increase in lake surface area by melting and undercutting of the calving ice front (cf. Reference Haeberli and RöthlisbergerHaeberli and Röthlisberger, 1976; Reference IkenIken, 1977). In 1996, the lake surface area had reached 18 000 m². A collapse structure in the ice, visible on air photos since 1994, indicated that the lake water had begun to melt a large cavity underneath the glacier, probably in connection with water from Lake No. 4 flowing over morainic material and entering the eastern part of Lake No. 3. The total lake volume could, therefore, only roughly be estimated at about 100 000 m³ (for 1996). The lake had not always been completely full during the air-photo flights in the fall, a fact which must be taken into account when interpreting the evolution of the photogrammetrically mapped lake surface.

Fig. 5. Evolution of ice-dammed Lake No. 3 during the time period 1968–95. (a) Planimetric representation of the lake outline at selected times of photography, 1967–94. (b) Lake level and area as a function of time. Note that lake level is varying over time. The presented values only reflect the situation at the time of aerial photography in the fall.

The safety of the ice dam at Lake No. 3 depends on the height of the ice surface above lake level. This geometry critically depends on ice advection by glacier flow which, in turn, is a function of the mass-balance evolution of the entire glacier. Recent mass changes of the entire glacier up to the summit of Fletschhorn were determined by photogrammetric analysis of high-altitude flights (Reference KääbKääb, 2000) carried out by the Swiss Federal Office of Topography at time intervals of several years for the standard revision of the Swiss National Topographic Map. As compared to the time interval 1975–85 with predominant volume increase especially in the lower glacier part, the time interval 1985–91 is characterized by average thickness losses of −0.8 m a⁻¹ (Fig. 6), a value which is in good agreement with similarly strong mass losses measured directly at other Alpine glaciers (Reference Haeberli, Hoelzle and FrauenfelderHaeberli and others, 1999b).

Fig. 6. Spatial pattern of mean annual changes in surface elevation of Grubengletscher during the time intervals 1975–85 (left) and 1985–91 (right). The results were derived from regular spaced photogrammetrically measured digital terrain models (DTMs) (50 m grid) of 1975, 1985 and 1991. Note the strong increase in surface elevation affecting the lower part of the glacier tongue and indicating an advance between 1975 and 1985 followed by decrease in ice thickness from 1985 to 1991.

Information from the annually flown low-altitude air photos (Reference KääbKääb, 2000) showed that the velocity of ice flow on the glacier tongue at Lake No. 3 (Figs 7 and 8) had reached a maximum around 1980 and then dropped by about 50% until 1994. This drastic reduction of flow velocity is only partially correlated with changes in ice thickness (cf. Fig. 8), probably due to changes in the subglacial water pressure. It caused the ice dam at Lake No. 3 to lower after 1985, i.e. 5 years after the velocity maximum in 1980, at a rate of −1.5 m a⁻¹ on average with an extreme value of −3 m a⁻¹ in 1990. In 1994, the ice surface was lower than in the years of the lake outbursts in 1968 and 1970 and at the time of artificial lake-level regulation. In view of such developments, imminent reversal of the ice-thinning trend near Lake No. 3 could not be assumed.

Fig. 7. Field of mean annual velocities on the tongue of Grubengletscher during the time intervals fall 1973–fall 1974, fall 1979–fall 1980 and fall 1991–fall 1992. The average surface velocity dropped drastically between the latter two stages, presumably due to marked changes in subglacial water pressure.

Fig. 8. Change in flow velocity and surface elevation on the tongue of Grubengletscher during the time interval 1973–92. (a) Cumulative changes in mean annual velocity. Bold data points indicate high-precision results from redundant and independent photogrammetric measurements; unfilled ones indicate single measurement procedures, (b) Cumulative changes in average elevation of the entire Grubengletscher (filled area) and its tongue (dotted line), 1967–91 (1967–95 for the tongue). Due to the limited availability of high-altitude aerial photographs covering the entire glacier, DTMs could be collected only in selected years, whereas a DTM of the tongue region could be measured almost every year.

Ice depth near Lake No. 3 is known from radio-echo soundings and hot-water drillings which were combined with subglacial d.c. resistivity soundings using borehole electrodes (Reference Haeberli and FischHaeberli and Fisch, 1984). Based on this information and the above-mentioned photogrammetric data, the safety against dam failure was calculated as the excess thickness above the critical flotation threshold (Fig. 9). In the fall of 1995, this safety reached a minimum of 8 m at the thinnest point of the ice dam. By extrapolating the observed subsidence rates (−1.5 to −3 m a⁻¹) of the ice surface, a potential early outburst time was estimated as 1998–2000. It was therefore considered advisable to undertake the necessary steps for avoiding another uncontrolled lake drainage (cf. further comments in the section on hazard-mitigation works).

Fig. 9. Safety of ice-dammed Lake No 3 in 1995: ice thickness above flotation threshold was calculated from glacier bed topography determined by radio-echo sounding, photogrammetry-derived ice-surface topography and extrapolated lake level applying the “iceberg model” of Reference NyeNye (1976).

Lake No. 2 was situated at the right margin of the heavily debris-covered part of the glacier tongue (Fig. 1). Its area was continuously reduced by the growing glacier (cf. Fig. 6) from some 2000 m² in 1975 to its complete disappearance in 1981.

Proglacial Lake No. 1

Lake No. 1 grew on top of a very thick sediment bed or morainic apron. Such thick morainic beds form where glaciers receive much more debris from the surrounding rock walls than the meltwater stream can evacuate (Reference HaeberliHaeberli, 1996; Reference Maisch, Haeberli, Hoelzle and WenzelMaisch and others, 1999). As a positive feedback, they reduce the transport capacity of meltwater streams through percolation of surface water into the subsurface or through lake formation in closed topographic depressions. Lakes thus formed represent an obvious potential for moraine-dam breaching and the formation of large debris flows. At Gruben, the flood events of 1968 and 1970 had initiated the breaching process in the morainic apron, and an artificial dam consisting of coarse blocks with a concrete outlet structure had to be installed in order to prevent further incision and debris-flow formation in the eroded stream channel.

During recent years, hydrogeological conditions at Lake No. 1 have undergone remarkable changes. Lake level more and more remained below the artificial dam and only reached the concrete outlet during time periods of strong ice melt combined with events of heavy precipitation. Water leaving Lake No. 3 drained into coarse, highly permeable morainic deposits before entering Lake No. 1. Moreover, the southern part of the lake (Fig. 1) quickly enlarged into dead ice buried within material of the 1890/1920 moraines but only filled when water from the northern lake part was flowing over a threshold between the two parts. This development was favourable in that it increased the potential retention volume of Lake No. 1 with respect to unexpected floods from upstream lakes and ponds. However, it also created concern among the local population that the stability of the moraine dam could degrade due to internal erosion or ice melt. This concern was later substantiated by the observation made during one event of heavy precipitation and fast-rising lake level, that percolating water was flowing underneath the outlet structure to reappear as strong springs in the moraine breach on the valley side. The stability of the moraine dam is, indeed, a crucial component of the hazard potential in the entire drainage basin, because the retention capacity of Lake No. 1 essentially helps protect the village of Saas Balen from the formation of dangerous debris flows in the eroded stream channel but could most seriously aggravate the danger in case of dam failure. Geophysical, hydrological and photogrammetric investigations were therefore carried out on the huge morainic apron in order to investigate the stability aspects and to redesign the protective measures.

Characteristics and stability of morainic materials at Lake No. 1

Breaching of moraine dams may involve (a) piping (progressive ground-water flow) within the morainic material, (b) liquefied flow/slippage on steep slopes, and (c) overtopping with retrogressive incision (Reference HaeberliHaeberli, 1992, Reference Haeberli1996; cf. Reference Youd, Wilson and SchusterYoud and others, 1981). In all cases, high water levels are critical and the internal structure, the porosity and the water saturation within the morainic material constitute the main factors influencing the stability. The primary questions to be answered by the investigations, therefore, related to trajectories of subsurface water flow and to the presence or absence of ground ice, of loose erodible material and of water-saturated sediments.

Three seismic refraction profiles and 13 geoelectrical d.c. resistivity soundings were measured in 1993 by GEOTEST Consulting Company (Geotest, unpublished; Figs 1 and 10). Geoelectrical d.c. resistivity of (unfrozen) bedrock was determined to be about 9 kΩm at outcrops near the foot of the moraine slope. The internal moraine structure (Fig. 10) was analyzed using characteristic values for seismic velocities and d.c. resistivities usually observed in starting zones of debris flows (Reference Haeberli, Rickenmann, Zimmermann and RösliHaeberli and others, 1990). In addition, the ratio ρ (Ωm)/V _p (m s⁻¹) between the electrical resistivity ρ and the seismic velocity V _p, multiplied by a factor R reflecting the vertical sequence of resistivities, was also considered (Reference Vonder Mühll, Haeberli and KlingeléVonder Mühll and others, 1996). Gravimetric observations were made in 1994. Determination of the regional gravity field requires measurements on outcropping bedrock on each side of the expected anomaly, here the morainic layers. This task was made difficult by the fact that nearby rock outcrops hardly exist in the eastern part of the investigated area. Hydrological investigations were carried out in 1993 and 1994.

Fig. 10. Two sections through the moraine dam near Lake No. 1 as obtained from seismic refraction and d.c. resistivity soundings. Numbers refer to an index combining the results from the two sounding methods mentioned in the text. The light-dotted area shows loosely compacted layers, the darker-dotted area represents possible occurrences of ground water, and the remaining parts are non-saturated and well-compacted morainic material.

The seismic refraction soundings indicated bedrock as a marked refractor with P-wave velocities of 3900–4900 m s⁻¹ (Fig. 10). Bedrock depth is only a few meters at the foot of the moraine dam but increases continuously towards Lake No. 1 where it reaches 100 m and more. Within the uppermost 5–20 m of the moraine, seismic velocities are 400–900 m s⁻¹ and resistivities are 2 to > 15 kΩm in places. Deeper morainic layers are characterized by seismic velocities of 1000–1700 m s⁻¹ and resistivities of 2–8 kΩm. The highest resistivity values are correlated with very low seismic velocities, indicating large porosity rather than ground ice (which would have high seismic velocities). Particularly loose parts of the moraine therefore occur relatively close to the surface, whereas indications for ground-water-bearing layers are sparse and concentrated at the interface between bedrock and the moraine.

With total thickness and internal structure of the moraine being known from the seismic and geoelectrical soundings, gravimetry was applied to estimate density and porosity of the morainic material. As the gravity map of Switzerland (Reference Klingelé and OlivierKlingelé and Olivier, 1980) gives small gradients in the investigated area, the interpretation of the long profiles is based on the assumption that the regional gravity field is horizontal (−156.0 mgal). The resulting Bouguer anomaly for the moraine area is −1.2 mgal (Fig. 11), and the corresponding density contrast about 0.3 Mg m⁻³. Bedrock density was measured, using local samples, as 2.76 Mg m⁻³, giving an average density (ρ) of the moraine of roughly 2.4–2.5 Mg m⁻³, and a mean porosity (n) of 11% in air/rock mixtures and of 17% in water/rock mixtures. Such a low mean porosity points to a good overall consolidation with large amounts of fine sand, a phenomenon indeed observed in the subglacial channel excavated for the artificial drainage of Lake No. 3.

Fig. 11. Regional and residual gravity field of a profile across the moraine dam along profiles P₁ and P₂ in Figure 1. (Lake No. 1 is to the left, and the bottom of the moraine to the right of plot) assuming a constant regional field of −156.0 mgal. The residual anomaly is up to −1.2 mgal.

Concerning the internal structure of the moraine as indicated by the results from the seismic and resistivity soundings, the entire deposit was treated in a second step as consisting of three main layers: the deepest layer was assumed to be water-saturated and to have average density, the intermediate layer to have air-filled pores and average density, and the uppermost layer to have air-filled pores but unknown density. A corresponding sensitivity study using variable thicknesses for the deepest and uppermost layer showed that near-surface layers of the moraine can have high to extremely high porosities if not real caverns, possibly from melting of buried ice or permafrost (Reference Vonder Mühll, Haeberli and KlingeléVonder Mühll and others, 1996). In fact, such a cavern with a diameter in the meter range had already been encountered during original construction work for the dam consisting of coarse blocks at Lake No. 1 where ice lenses had previously been found in an exploratory core drilling (cf. Reference HaeberliHaeberli, 1992). Thaw settlement from dead-ice melting is known to occur to the south of Lake No. 1, where photogrammetric analysis for the time interval 1976–94 indicates terrain subsidence at rates up to 0.3 m a⁻¹ (Reference KääbKääb, 1996). Near the dam, however, the applied spacing for measurements of 25 m showed no significant changes in surface altitude.

Tracer experiments together with discharge and temperature measurements at the main springs of the area (Fig. 1) were carried out by members of ETH Zürich in June 1993 and by the University of Fribourg in June–August 1994 (Reference Gardaz and LugonGardaz and Lugon, unpublished; Reference SpeckSpeck, unpublished). The tracer experiments were difficult to interpret: fluorescine dye was injected where the meltwater stream between Lakes No. 3 and 1 disappeared into the ground, but reached two springs at the foot of the moraine apron in weak concentrations and only after several days. It must be assumed that the meltwater stream enters deeper layers of Lake No. 1 and that the dye had been dissolved by the lake volume and adsorbed by fine sediments at the lake bottom. Because discharge in the meltwater stream as measured by the salt dilution method corresponds to the discharge of the main spring at the foot of the moraine (A in Fig. 1), water from the northern part of Lake No. 1 must be connected with this main spring but percolates and flows through the moraine with velocities of 10⁻³ to 10⁻² m s⁻¹.

Water from the southern part of Lake No. 1 is connected to a number of smaller springs (B in Fig. 1) in the southern part of the moraine apron, which provide water with a delay of several days after the onset of overflow from the northern into the southern part of Lake No. 1 and the filling of the latter. This observation confirms the slow subsurface flow velocities estimated by the tracer experiments, and thus the generally low porosity of deeper layers in the moraine as indicated by the geophysical soundings. The bottom of the southern lake part is covered by a silty layer about 0.1–0.2 m thick. This layer is eroded where water from the northern lake part enters the depression of the southern lake part (Fig. 12). Because a discharge of 0.04 m³ s⁻¹ filled the southern section of the lake within about 1 day, percolation into the underlying moraine via the gullies eroded into the silty lake bottom must be considerably less. Comparison with the total volume of the southern lake section (1–2 × 10³ m³) indicates a percolation rate of about 0.02 m³ s⁻¹. Characteristic water temperatures of springs A and B were around 3°C. Since the water probably flows at depths between several tens of metres and 100 m below the surface, average temperatures inside the moraine apron are assumed to be 0–3°C. Given considerably colder temperatures during the Little Ice Age and maximum glacier advance around 1850 (Reference Haeberli and BenistonHaeberli and Beniston, 1998; Reference Maisch, Wipf, Denneler, Battaglia and BenzMaisch and others, 2000), it is highly probable that upper layers of the moraine apron had, at that time, been frozen underneath the cold glacier margins of the more extended polythermal glacier. Degradation of former subglacial permafrost with melting of ground ice could explain the weakly consolidated near-surface layers detected by the geophysical soundings. Permafrost still exists in an inactive, south-exposed rock glacier east of the moraine apron (Reference Haeberli, King and FlotronHaeberli and others, 1979). The corresponding spring (C in Fig. 1) has a temperature of about 1.5°C, a value which is typical for permafrost drainage.

Fig. 12. Silty bottom of empty southern part of Lake No. 1; the view is towards the threshold between the southern and northern lake parts. Note small channels eroded to the gravelly bed (photo W. Haeberli, September 1992).

The combined consideration of all collected evidence shows that the moraine apron at Lake No. 1 (a) is up to > 100 m thick, (b) today contains no large bodies of buried ice or perennially frozen ground, but (c) could have been frozen in surface layers covered by the polythermal and more extended glacier tongue during the maximum stage of the Little Ice Age, (d) now has extremely loose parts near the surface, probably as a consequence of glacier retreat and permafrost degradation during the 20th century, (e) is well consolidated at greater depth, (f) only contains relatively thin water-saturated layers at the contact between the moraine and bedrock, and (g) discharges water from Lake No. 1 at great depths and slow rates. With respect to the safety of Lake No. 1, it was concluded that the 25–35° steep outer slope of the moraine dam as a whole could be considered stable against deep-seated sliding because of the low water saturation and good consolidation. It was, however, also clear that the loose material near its surface could be easily eroded by piping, shallow slope instability, backward incision of the breach or debris flows, possibly even in a combination of such processes with positive feedbacks between them. Hence, use could be made of the retention capacity represented by the lake basin, but uncontrolled flooding at the lake outlet and discharge of large water masses across the open moraine slope would have to be avoided.