sec6_GLACIORISK_D4 - GLACIORISK

GLACIORISK
EVG1 2000 00512
Deliverables
Report Period : 01.01.2001 – 31.12.2003
D4: Monitoring of the most representative glaciers
SURVEY AND PREVENTION OF EXTREME GLACIOLOGICAL HAZARDS
IN EUROPEAN MOUNTAINOUS REGIONS
http://glaciorisk.grenoble.cemagref.fr
Compiled by Didier Richard and Michel Gay

Glacier lake outburst floods (GLOF)
THE EMERGENCY CAUSED BY THE “EFFIMERO” LAKE ON THE BELVEDERE
GLACIER (Macugnaga, Monte Rosa Group, Italian Alps)
Birth, growth and evolution of a supra-glacial lake
Giovanni Mortara, Marta Chiarle, Andrea Tamburini
CONSIGLIO NAZIONALE DELLE RICERCHE
Istituto di Ricerca per la Protezione Idrogeologica
Sezione di Torino
Deviverable 4 1

Introduction and background
The Belvedere Glacier (Fig. 1) is a humid-temperate one, with a flat, heavily debris-covered
tongue, fed by several steep glaciers (the main one being the Monte Rosa Glacier) and by ice
and snow avalanches (impressive those coming from the Tre Amici Peak). In recent years an
important contribution to debris cover has been related to rockfalls coming from the Monte
Rosa east face.
The Belvedere Glacier is known as a classic example of a glacier with an elevated sediment
bed (Monterin, 1923; VAW, 1985).
Figure 1- Location map of the Belvedere Glacier.
The Belvedere glacial basin is well known to glaciologists because of a long history of
instability events which have been repeatedly reported since 19 th century and which
threatened the village of Macugnaga and affected the original shape of the moraines (Dutto &
Mortara, 1992; Chiarle & Mortara, 2001; Haeberli et al., 2002):
- August 1868: the strong pressure exerted by water accumulated inside the glacier, as a
consequence of prolonged rainfall, caused a sudden collapse of the right lateral
moraine of the left glacier lobe. The 60-70 m wide grass plain in front of the breach
was covered by boulders for about 1 km 2 (Stoppani, 1871);
- August 1896: water cut two ways through the moraine and devastated meadows near
Pecetto and Macugnaga (“La Voce” magazine, 1896);
- 1904: a water pocket inside the glacier provoked progressive saturation of the right
lateral moraine and its breaking down near the Alpe Pedriola (Somigliana, 1917);
- September 1922: Following several days of rain, a huge mass of water was expelled
from the glacier, destroying a 100 m long wall constructed to defend Macugnaga. Big
ice blocks were carried for 6 km down to Borca (Monterin, 1926);
- 13 August 1970, 2 August 1978, July 1979: outburst floods issuing from the moraine-
and ice-dammed Lago delle Locce at one of the tributaries (Ghiacciaio delle Locce),
progressively widened the breach initially cut in 1904 through the right lateral moraine
near Alpe Pedriola. The 1979 event seriously damaged the Belvedere chair-lift and
flooded the valley bottom for a length of 1 km and a mean with of 150 m, almost
reaching the Pecetto hamlet near Macugnaga (Tropeano et al., 1999).
2 Deliverable 4

Recent developments
After the last Lago delle Locce outburst in 1979, for a twenty-years period no unusual
situation was observed at the Belvedere Glacier, periodically monitored by Italian
Glaciological Committee operators.
Since autumn 1999 (according to Mazza, 2003) a change in surface morphology started
becoming apparent, suggesting that an acceleration of glacier dynamics was taking place.
Around the same period, a growing rock fall activity at medium height, immediately south of
the Imseng Ridge, was noticed from an extended detachment zone located at the base of an
hanging glacier. For several years, a continuous noise could be heard, caused by incessant
block rolling along the rock face. Rock fall activity was not limited to summertime, when
melt-water contribution could change rock fall in debris flow processes, but continued all the
winter season through, pointing to a change in rock wall thermal regime and not only to
instability related to seasonal melting phenomena (Fig. 2).
Figure 2 - Monte Rosa east face: rock-fall path during winter 2002/03; the red circle indicates the
starting point of the mass-wasting process.
The most striking changes developed in spring 2001: the ablation tongue, which used to show
a continuous debris-cover and a smooth surface, was almost completely perturbed by a
chaotic system of crevasses, which radically modified its characteristic aspect of glacier noir.
The compressive strain appeared to be directed mainly towards the orographic right glacier
margin and in correspondence of the left and right tongue divergence at the location
“Belvedere”, accounting for an ice-mass bulking which, in a few months, elevated the glacier
surface of 20 m or more above the LIA moraine, reaching its maximum at “Belvedere”
location (Fig. 3). On this right margin, glacier ice started filling the breach cut by the repeated
outburst of Lago delle Locce (Fig. 4), and pushing against the moraine in a few points, so that
Deliverable 4 3

further moraine breaches were expected to form. Changes in glacier surface geometry and
morphology caused problems to the access trail to the Zamboni Hut, which cross the glacier at
“Belvedere”, while surface rise caused ice and rock blocks to fall outside the moraines of the
right tongue, affecting the trail to the Zamboni Hut and the sky run along the left margin of
the right tongue.
Figure 3 – Right side of the glacier in 1996 (left) and in September 2001 (right).
Figure 4 – Glacier-ice filling in the morainic breach near Alpe Pedriola (left: 1996; right: September
2001)
At the same time that the ablation tongue was rising its surface, a depression showed up in the
upper part of the Belvedere Glacier, right at the foot of the Monte Rosa east face, at an
elevation of about 2150 m a.s.l. Photogrammetric studies revealed that already in 1995-1999 a
20 m loss in ice thickness occurred at the location of the depression (Kääb et al., 2003a).
During 2001, this depression was repeatedly partially filled by a half-moon shaped
supraglacial lake (named “Effimero”, which means “short-lived”), which reached the
maximum extent of about 20,000 m 2 in October (Figg. 5-6). High water pressure in the glacier
was testified by small dirty water pounds occurrence between ice and moraine, along the right
glacier margin, observed in summer 2001.
In the first phase, events at the Belvedere Glacier had mainly a scientific interest, being a
“surge-type movement” (as was defined by Haeberli et al., 2002), at present, a unique
phenomenon in the whole Alps.
4 Deliverable 4

Figure 5 – Aerial view of the Belvedere Glacier before (left, 1985) and after (right, October 2001) the
formation of the new supraglacial lake. North is directed towards images’ bottom.
Deliverable 4 5

Figure 6 – Early images of the «Effimero » Lake, taken
in late spring 2001 from the left lateral moraine (left) and
in the up-ward direction (right): in this late image, please
notice the glacier surface dirtied by rock-fall
accumulation (www.mandalafoto.it).
Photogrammetric investigations for the period 1999-2001 assessed average speeds of up to
110 m/y, with a peak of up to 200 m/y during the autumn 2001. These values by far exceeded
average surface speeds measured in the mid-1980s and during 1995-1999 on the lower part of
the Glacier, which were respectively in the order of up to 40-45 m/y and 35 m/y (VAW, 1985;
Kääb et al., 2003a). Afterwards, flow acceleration slowed down to up to 80 m/y in summer
2002, while in spring 2003 a marked shear band at the right glacial margin testified the
continuation of the exceptional speeds (Kääb et al., in press).
During winter 2001/2002 the depression at the foot of the Monte Rosa east face enlarged, and
consequently the volume of water trapped got increased so that lake surface reached by the
end of May 2002 an area of about 20,000-40,000 m 2 , as reconstructed from ASTER satellite
imagery by Kääb et al. (2003b). In mid-June 2002, a persisting period of anomalously high
temperatures made the 0°C isotherm to lay at an elevation higher than 4,000 m (with frequent
peaks at 4,600 m a.s.l.), causing the entire Monte Rosa Massif to be subject to melting
processes. The huge and rapid melting of winter snow, avalanche snow and ice in the
Belvedere basin caused an enhanced input of water to the lake, which grew, by the end of
June, to a volume of 3 million m 3 , extending on a surface of about 150,000 m 2 , with a
maximum depth of 58 m and a mean one of about 20 m (Tamburini et al., 2003). At that time
lake surface was rising up to 1 m/day and just a few meters freeboard was left with respect to
glacier surface (Fig. 7).
6 Deliverable 4

October 2001
N
19 July 2002
Figure 7 – Photographic comparison highlights
the striking growth of the supraglacial lake
By the beginning of July 2002, a cold spell, together with pumping performed by Italian Civil
Protection Department and naturally occurring subglacial drainage, allowed a progressive lake
level lowering, which restored, by the end of October, the lake size of autumn 2001.
In the second half of May 2003, again, an exceptional, prolonged period of high air
temperatures caused a new, rapid lake growth, quite comparable in volume and shape to that
of the previous year.
Nevertheless, water level was about 10 m a.s.l. deeper compared to 2002, probably due to a
continuation of the accelerated glacier flow and to processes of melting for heat exchange
along icy lake margins (Fig. 8).
Deliverable 4 7

Figure 8 – “Effimero” Lake maximum growth, in June 2003.
However, very soon a natural flow path opened inside the glacier and partially in surface, at
the contact between glacier margin and the left moraine. This flow path became suddenly
active, allowing a strong lake level lowering starting from June 18, 2003. Correspondently,
increasing discharge of very dirty water from the left glacial tongue (Torrente Anza) was
observed. The outburst got exhausted by June 20, 2003, showing the maximum discharge (15-
20 m 3 /s) on June 19 th . The total volume released by the event was 2.3 million m 3 , which
corresponded to a lake level lowering of about 20 m. No significant morphological effects nor
relevant damage occurred (Fig. 9).
8 Deliverable 4
a
b

c
18 June 2003
5 August 2003
Figure 8 – Supraglacial lake outburst on June 18-20, 2003. In a) flood waters reach the surface
between glacier and left lateral moraine, close to “Fillar”. Outburst waters swelled the Anza Torrent,
which overtopped the bridge to the “Alpe Burki” (b). Lateral erosion was produced in the inner flank
of the left moraine (c). The event caused an overall lake lowering of about 20 m; by the beginning of
August, almost no water was left inside the depression (d).
After sub-glacial drainage path was established, lake level remained stable all the summer
long.
In the meantime, in summer 2003 the Belvedere glacier started loosing its swelling in most
parts, moving back from the lateral moraines which used to push in the previous two years.
The detachment area on the east Monte Rosa face significantly enlarged, involving also part
of the rock wall, ice-free, located above the original instability area, and the “Canalone
Marinelli” incision, on the orographic left, as a consequence of the dramatic loss in ice cover,
both in thickness and width.
Phenomenon interpretation
Mechanism of the surge-type movement is still unclear. For sure, the accelerated Belvedere
glacier flow is not related to climatic conditions favourable to glacier advance, as testified by
marked ice depletion shown up by all the steep glaciers feeding the Belvedere tongue.
The depression formed at the foot of the Monte Rosa east face seems to indicate in that area
the upper limit of the surge-type movement, while the shear band observed in March 2003
along the right Belvedere margin points to a movement of the glacier as a whole (Kääb et al.,
in press). A critical role could have been played by water in pressure underneath the ice body:
as already recalled, the presence of strongly pressurized water was testified by dirty water
ponds observed along the right Belvedere margin as well as by the formation of the same
supraglacial lake. The cause of this increased water availability might perhaps be identified in
the growth of air temperature registered in the last summer seasons (since 1998, Mortara &
Mercalli, 2002).
Following this interpretation, the intense crevassing shown by the Monte Rosa Glacier has to
be considered more as a consequence of Belvedere Glacier surge, which left a lack of contrast
a the foot of the Monte Rosa ice body, than as a cause of the Belvedere Glacier bulking.
Finally, a marginal role seems to have been played in the change of glacier dynamics by the
persistent rock fall activity interesting, since a few years, the Monte Rosa east face.
Deliverable 4 9
d

Hazards raised and mitigating actions undertaken
Recent developments in the Belvedere basin posed important hazard issues to the Macugnaga
area, starting in summer 2001. Initially, problems were limited to the access trail to the
Zamboni hut and to the sky-run located just outside the right lobe left moraine, in relation to
glacier surface elevation and crevassing. The sky-run was subject to rock/ice block fall from
glacier surface exceeding lateral moraine height, so that in autumn 2002 a new sky-run was
realized, in order to avoid the most direct danger zone, and a retention wall and a net were
realized to hold back falling blocks. The access to the Zamboni hut, which during the summer
season is daily frequented by even hundreds of people, mostly family groups, besides being
threatened at some location by the same hazard, became arduous in the stretch crossing the
right glacier tongue, close to the “Belvedere” location. Instead of a flat, smooth, debris
covered crossing, in 2002 tourists found a crevassed, irregular passage, threatened by ice
lamella fall; a detour had to be constructed and progressively adapted to glacier geometry
changes. Up to this stage, risks concerned only skiers and excursionists, so that just local
authorities and sky-run managers were considered charged with tourists’ safety.
a
Figure 9 – Ice blocks fall (a) and glacier
surface raising (b) made hazardous the trail
to the Zamboni Hut. The sky-run close to the
“Belvedere” location had to be protected
from ice falls by a retention wall (c).
b
c
wall
Sky-run
The scientific board (among which CNR-IRPI) following glacier evolution, pointed out that
attention had to be paid also to other possible phenomena relating to the ongoing situation.
First, an advance of the right lateral tongue could have proceed to the point of damming the
10 Deliverable 4

Pedriola torrent, creating a water retention basin which could subsequently generate an
outburst flood. Second, the opening of new moraine breaches could represent new escapes for
glacial waters, as occurred in the 1970s events. Third, a large rock/ice fall event from the
instable Monte Rosa east face could propagate down to “Belvedere” and eventually Pecetto,
not anymore confined by lateral moraines, already exceeded by glacier surface, especially in
case of snow-covered glacier. Finally, already in autumn 2001 the attention of local
authorities was driven to the small supra-glacial lake of new formation at the foot of the east
wall.
Nevertheless, the major risk raised at the end of June 2002, when the rapid, uncontrolled
growth of the “Effimero” supra-glacial lake endangered not only the glacier area and its
environs, but all the upper Anzasca Valley, with particular regard to the Macugnaga inhabited
area. Infact, in case of a sudden outburst of the 3 million m 3 of water stored in the lake, a huge
flood wave propagating for a long distance and taking in charge large amounts of debris could
be expected (mobile material is particularly abundant along the Torrente Pedriola, the
orographic right glacier stream).
Several outburst scenarios were identified. If the lake had continued to grow, an overflow
flood could have occurred, running both on the glacier surface, or at glacier/moraine contact,
and eventually escaping through moraine breaches. The outburst could also occur through
en/sub-glacial paths, and important discharges could have been expected in case of hydraulic
ice dam break (Clague and Mathews, 1973; Walder & Costa, 1996). The most catastrophic
scenarios was envisaged in a potential outburst flow triggered by a large rock/ice fall from the
overhanging rock face.
At this point, the Italian National Department for Civil Protection took over by decree the
responsibility at beginning of July 2002, assuming the command of all the parts involved
(local and regional authorities, armed forces, police departments, fire departments, alpine
rescue volunteers). Emergency actions were undertaken, including (Regione Piemonte, 2002):
− flood hazard mapping of the Macugnaga inhabited area and the Anzasca Valley
bottom down to the Ceppo Morelli municipality, taking into account the discharge
values of the July 1979 Lago delle Locce outburst flood (drawn by Regional Technical
Services);
− restriction on the access to the Belvedere glacier area, upstream of the Pecetto hamlet;
− closure of the Pecetto-Burki chair-lift, heavily damaged by the July 1979 Lago delle
Locce outburst;
− establishment of a warning and evacuation procedure, based on the visual monitoring
of the lake level and of the glacier area in general;
− hourly manual measurement of lake level, afterwards replaced by an automatic
hydrometric station;
− visual moraine stability survey (twice a day);
− a video camera pointing to the lake, in order to perform a real-time visual monitoring
of lake stage;
− 3 photo cameras pointing to the main glacier outlets (Torrente Anza, Torrente
Pedriola, Fontanone), to visually assess changes of stream discharge and torbidity (a
photo each 5 minutes);
− installation of a pump system at the lake (with a discharge of about 200 l/s);
− installation of a meteorological station at the Locce Lake;
− creation of a monitoring centre in Macugnaga, where data and images were collected
and displayed;
− installation of a hydrometric station on the Torrente Anza at Pecetto;
Deliverable 4 11

− for the purpose meteorological forecasting service (performed by Regional Technical
Services), with a twice a day bulletin.
Pumping activities, in particular, were affected by the severe environmental conditions:
besides technical difficulties, relating in particular to the continuous, rapid change of glacier
and lake geometry, workers were continuously exposed to the risk of rock/ice falls, snow
avalanches, unexpected lake waves (Fig. 10).
By mid- July 2002, as lake stage was significantly lowered, so that the main emergency phase
could be considered over, responsibility for interventions in the area went back to Regional
Technical Services.
Over winter 2002/03 mitigation plans were elaborated in case of a new formation of
“Effimero” Lake in 2003. Discussed measures included pumps mounted on a cable crossing
the glacier and artificial outlet channels through ice and moraines. Due to uncertainties in the
lake development and in technical and financial issues, it was decided to prepare but not
perform emergency actions and to put into service again the monitoring and alarm system.
During the lake outburst in June 2003, warning thresholds for river discharge and changes in
lake-level were set up, and, as a precaution, forbidden the entry to the glacier’s environs.
The monitoring system remained active all the summer long, even after the lake emptied, and
was deactivated at the end of October (Fig. 11).
12 Deliverable 4
Figure 10 – Pump installation on the
supra-glacial lake. Activities were
heavily hampered by icebergs and glacier
movements.

Figure 11 – Hazard scenarios opened by recent developments in the Belvedere glacial basin.
Investigations carried out (Contribution of A.Tamburini, Enel.Hydro)
In the first stage of the surge-type movement (year 2001), investigations were limited to
periodical field surveys and to aerophotographic analysis. In particular, the CNR-IRPI, in the
contest of Glaciorisk Project, ordered a special flight for the purpose, which was realized on
11 October 2001 (see Fig. 5) and which came out to be of critical importance for assessing
surge-type movement velocities and glacier surface geometry changes (Kääb et al., 2003a).
However, when the Effimero Lake grew to its maximum extent in June 2002, a number of
investigations were carried out, in order to gather those data necessary to undertake the
appropriate mitigation measures.
In particular, it was necessary to quickly get the following information:
− lake’s overall volume and the related level-volumes curve to determine the capacity of
the pumping plant and assess its efficiency;
− the morphology of the lake bottom, to evaluate the optimum position for the pumping
station and the presence of possible areas of critical stability within the reservoir;
− the ice thickness underneath the lake and its immediate areas to evaluate the stability
of the damming ice and the opportunity to undertake a possible drilling of the basin’s
bottom to promote the subglacial drainage;
Deliverable 4 13

− a map of the intra/sub-glacial discharge routes, with the aim of searching the possible
connections with streams and springs below the glacier, defining thus the most likely
flow directions in case of an intra/sub-glacial outburst and the efficiency of the
intra/sub-glacial drainage system.
During the emergency phase, first the Civil Defence and then the Regione Piemonte have
charged Enel.Hydro with the responsibility of carrying out the necessary surveys. The
selection of the survey techniques had to consider both the urgency to collect the requested
data in the shortest time possible and the need to minimize the number of operators and their
stay in such a dangerous situation. For this reason, Enel.Hydro decided to use an advanced
data gathering technique based on a Mobile GIS station integrated with DGPS.
Bathymetric surveys
These surveys were carried out with an innovative technique based on the use of a boat
equipped with a GPS positioning system connected with an echo sounder for the
measurement of the bottom’s depth. The positioning and depth surveys, obtained at regular
time intervals and accurately matched, could determine the bottom’s plano-altimetric
coordinates along the trajectories followed during the survey. The echo sounding and the GPS
antenna were fixed in a co-axial position to a small catamaran (Fig. 12). The positioning
system used during the surveys, later used for the geo-radar surveys, was made of a GPS CSI
WIRELESS receiver, DGPS MAX model, set for the reception of the differential corrections
in real time transmitted via satellite by the OMNISTAR service. With this technique it is
possible to achieve a “sub-metric” precision without a local master station set in a position of
known coordinates.
Figure 12 – The catamaran used during the
bathymetric survey. The GPS aerial was fixed on
top of the vertical bar, while the echo sounder’s
sensor was positioned 15 cm below the water
surface at the other end. The catamaran was pulled
by a motorboat belonging to the Milan Fire Depts’
Diving Team.
The adopted technique, already well tested (Mercalli et al., 2002a), does not need a radio or
GSM link between the reference station and the mobile station which would otherwise be
necessary if one wanted to produce the differential corrections locally. A single-frequency
SonarLite Ohmex depth finder was used to measure the bottom’s depth; equipped with a 200
kHz transducer and with a maximum range of 80 metres, it was particularly appreciated for its
small size and ease-of-use.
The tracking of the route followed during the surveys was ensured by a Mobile GIS station,
made of a Compaq iPAQ Palm PC, equipped with ESRI ArcPad 5.0.1. software. The boat
position was therefore displayed in real time on a cartographic background represented by the
Carta Tecnica Regionale map integrated with the coastline derived from aerial photos. In this
way, it was possible to obtain an immediate overview of the surveyed areas and, therefore, to
plan a route minimizing the period of stay in the areas yet to be surveyed. The bathymetric
14 Deliverable 4

survey was almost completed in one day (3 July 2002); the lake’s total volume, a map of the
contour bottom lines and the level-volume relation were available the same evening.
On the basis of these results, some days later (8 July 2002) a more detailed survey of a “bay”
located in the lake’s right sector was carried out; the area was identified by the first survey as
a suitable location for the installation of the first pumping plant. On the whole, more than
6000 points were identified.
Lacking a topographic survey of the coastline and having only one, non-geo-referenced
picture taken from the helicopter, 13 points were identified on the 30 June with the same
differential sub-metric GPS technique used for the bathymetric survey. They were arranged
along the lake’s accessible edges and then used for an approximate geo-referencing of the
picture itself, which produced the estimated coastline used in the preliminary papers. The
final documents were based on the aerial photogrammetry obtained by G. Viazzo; however
the final results showed no significant deviation from the preliminary ones. The measured
depth figures along the surveyed trajectories were turned into altitude levels for the lake’s
bottom, by subtracting the depth value from the corresponding surface level, and filed under
ArcView; it was then possible to build a DEM of the basin’s bottom. In particular, such
digital pattern was created by the ArcView TIN module, as well as the contour lines map of
Fig. 13 and the area-volume relation for various levels of the basin (Fig. 14).
Coast-line elevation: 2163.78 m
(data from G.Viazzo)
Figure 13 – Bathymetric survey:
bottom’s contour lines map. The depth
was more than 57 m in the central part.
Deliverable 4 15

altitude (m a.s.l.)
altitude m a.s.l.)
“Effimero” Lake – Area Curve
area (m 2 )
“Effimero” Lake – volume curve
volume (10 6 m 3 )
Figure 14 – Bathymetric survey: area curve (above) and volume curve (below).
The overall volume was calculated at 3 million cubic metres with a measured maximum
depth of 57 metres.
Later surveys showed a gradual deepening of the bottom due to the action of thermo-karst
phenomena (Kääb & Haeberli, 2001; Mercalli et al., 2002a); under such conditions the
stability of the glacial sill depends upon the relationship between the basin’s depth and the ice
thickness; knowing both is therefore extremely important (Huggel et al., 2002). For this
reason, it was decided to run a geo-radar survey of the glacial area of the lake.
Geo-radar surveys
On 2 and 3 August 2002, in order to determine the ice thickness and thus the depth of the
glacial bed in the area around the lake, some geophysical surveys were carried out using the
16 Deliverable 4

geo-radar technique on 8 profiles across the glacial tongue between 2100 and 2260 m a.s.l.
(Fig. 15). The following equipment was used:
- GSSI SIR-10 radar system;
- RADARTEAM antenna, SUBECHO 40 model, with 35 Mhz base frequency;
- GPS differential positioning system (identical to the one used in the bathymetric
survey);
- Mobile GIS station for route control.
Figure 15 – Geo-radar survey: location of transversal profiles (left) and example of interpreted profile (n.3,
right). One can notice the processed radar recording in the upper part, while the two horizons related
respectively to lake’s bottom and to glacial bed are indicated in the lower part.
The radar and GPS antennas were installed into a custom-made aerodynamic wooden
structure which could be flown with a helicopter, as it was impossible to pull the equipment
directly across the glacial surface.
The interpretation of the data was extremely difficult, mainly because of the glacial till cover
and of water and debris inside the ice body, which considerably hindered the usefulness of
the technique.
Moreover, other inside reflections, often difficult to sort out, caused a further weakening of
the signal. Hereafter some comments on the outcome of the investigation:
- it was not possible to reach the glacier bed because of the signal loss due to the strong
reflection of the till covering the glacial surface;
- it was, instead, possible to clearly identify the course of the lateral moraine up to
maximum depths ranging between 120 and 140 m and then to extrapolate the glacier’s
approximate thickness. The estimated thickness of the ice in the surveyed area varied
between a minimum of 120 m along profile 1 to a maximum of about 220 m along
profile 4;
- it was assumed that the ice thickness under the lake’s bottom ranged between 120 and
150 m (profile 3 and 9), while beyond the lake, at the breach in the right moraine
(profile 7 and 8), the estimated thickness was about 150 m;
Deliverable 4 17

- particularly strong reflections seemed to originate from areas of remarkable shearing
stress induced by the confluence of glacial sectors with different flow-speed. The
phenomenon was visible also from the morphological point of view, highlighted by
the presence of both a longitudinal sinking of the glacial surface and of short crevasses
set at a 45° angle from the contact line between the two sectors with different speed;
- other particularly intense inside reflections could be linked to the presence of debris of
metric dimensions (the minimum size of the blocks detectable by the antenna) coming
from the landslide affecting the rocky walls above;
- in profile 4, at a depth between 40 and 70 metres, one could notice a transition
between the upper zone, characterised by a lack of reflection, and a lower area where
the reflection was more evident. This could represent the transition between ice sheets
with different water content. The presence of a basal area full of water had already
been assumed after the geophysical surveys carried out in 1984 (VAW, 1985). The
phenomenon was repeated on profiles 5, 6, 7 and 8, located beyond the lake, and
should be further studied to better understand the endo/subglacial hydrology.
The activity carried out has led to some important results; however, from a methodological
point of view, it does not seem possible to investigate depths greater than those surveyed by
using the geo-radar equipment mounted on the helicopter. The experience collected so far
indicates that far better results could be achieved if the radar antenna was directly on top of
the glacial surface (Frassoni et al., 2001; Mercalli et al., 2002b).
Tests with tracers
In order to complete the study of the evolution of the “Effimero” Lake and the Belvedere
Glacier, three tests with tracers were carried out in order to qualitatively evaluate the englacial
and subglacial drainage system of the lake’s water and to verify possible connections between
the lake and the waters emerging downstream.
Three tests have been performed up to now, in October 2002, June 2003 and October 2003
respectively.
The tracing was performed by the adding sodium-fluorescein (C20H10Na2O5) to the lake’s
water (Fig. 16). Considering the lake’s capacity at that time of each test and the fact that not
all the substance would dissolve in the water, the estimated initial concentration of fluorescein
was slightly above the visibility level.
18 Deliverable 4

Figure 16 - Position of tracer injection
point and locations of the checkpoints
for the tracer tests performed in June
and October 2003 (left). Tracer’s dropin
point (right; P.Federici).
The powdered fluorescein, dissolved in a large container three times over, was poured into the
“Effimero” Lake using a pump connected with a 15 m-long rubber hose. The solution was not
poured on the surface, but released at a depth of about 10 m in at least three different points.
At the checkpoints, fluorescence detectors made by activated carbon granules set into a closemesh
net were inserted at regular intervals and several times over. The positioning and the
removal of the detectors were programmed at the following intervals:
first test (October 2002): 6, 12 hours and 1, 2, 4, 8, 16 and 32 days from the tracer’s
addition;
second test (June 2003): 30’, 1, 2, 4, 8 hours, 1, 2, 4, 8 days from the tracer’s addition;
third test (October 2003): 1, 2, 4, 8 hours, 1, 2, 4, 8, 16 days from the tracer’s addition;
A set of detectors was installed and removed before the fluorescein addition, in order to
identify the possible presence of other organic substances that might influence the results of
the follow-on analyses. Another series of detectors was left at each point for the whole
duration of the test (32 days).
During the third test (October 2003) a continuous-recording fluorometer was installed at A
point.
The following diagrams (Figg.17-18) represent the results obtained after the above tests. Only
the continuous-recording fluorometer diagram is available for the third test, as the lab analysis
of the detector is still in progress.
Deliverable 4 19

Figure 17 - Normalised tracer load vs time since injection during June 2003 test.
Figure 18 - Normalised tracer load vs time since injection: June 2003 vs October 2002.
20 Deliverable 4

Data interpretation is still in progress; anyway the results obtained provided good information
about the preferential drainage routes of the lake waters. Some preliminary comments are
listed below:
all tests showed the presence of tracers only in the checkpoints located on the left side
of the valley (Alpe Fillar, Torrente Anza, Fontanone);
in October 2002 test, the analyses highlighted the first comparison of dye tracer at the
A and F checkpoints (Torrente Anza and Fontanone) after a 24-hour delay and peak
concentrations after about 2 days;
in June 2003 the peak concentration of tracer is about 3 to 8 hours after injection,
showing a higher efficiency of the englacial/subglacial drainage system, as expected;
the results of October 2003 test are more similar to June 2003 ones than to the first
test; this appears rather surprising and can only be explained by the formation of a
stable englacial passage after the June 2003 outburst.
Future perspectives
If global climatic trend is going to be confirmed in the next years, the dramatic changes in the
Monte Rosa east face are destined to continue. The enhanced glacial cover shrinking has led
to diffuse bedrock instability phenomena. Up to now, rock and ice mass-wasting occurred just
as individual block falls, but the risk of large events must not be underestimated, as warned by
rock/ice avalanches occurred on the Brenva Glacier, Mont Blanc Massif, in 1997 (Barla et al.,
2000) and more recently (2002) on the Kolka/Karmadon Glacier in Caucasus (Haeberli et al.,
2003).
Lake’s evolution is expected to go towards a growth in depression’s volume, as ice flux from
the Monte Rosa east face is decreasing while extending glacier flow at the foot of the rock
wall is continuing. This could entail a greater volume of water stored in the Effimero Lake in
spring/summer 2004 (even if tests with tracers seem to point towards a quite well established
en/sub-glacial drainage system), but, on the other side, a large depression at the foot of the
Monte Rosa wall could act as a retaining dam for ice/rock falls.
Anyhow, attention has to be maintained high in the area, as the characteristics of the ice body
and of glacier bed beneath and downstream the lake are only partially known, and could
change in agreement with the evolution of the overall glacial system.
Acknowledgements
Authors wish to thanks all the persons and authorities involved in the studies carried out on
the Belvedere Glacier who have contributed with data and discussion to the present report, in
particular the Department of Geography of the Zürich University, the Regione Piemonte
Technical Services, the Italian Civil Protection Department, Dr. F. Epifani, Ing. G. Viazzo,
the SMS, Macugnaga Major T. Valsesia, Macugnaga Alpine Guides.
References
Barla, G., Dutto, F. and Mortara, G. (2000): Brenva Glacier rock avalanche of 18 January 1997 on the
Mont Blanc range, NW Italy. Landslide News, 13, 2-5.
Chiarle M. & Mortara G. (2001). Esempi di rimodellamento di apparati morenici nell’arco alpino
italiano. Atti dell’VIII Convegno Glaciologico Italiano, “Risposta dei ghiacciai alpini ai
Deliverable 4 21

cambiamenti climatici”, Bormio 9-12 Settembre 1999. Suppl. Geogr. Fis. Dinam. Quat., V, 41-
54.
Clague, J.J. and Mathews, W.H. (1973): The magnitude of jökulhlaups. Journal of Glaciology,
33(113), 501-504.
Dutto F. & Mortara G. (1992) – Rischi connessi con la dinamica glaciale nelle Alpi Italiane. Geogr.
Fis. Dinam. Quat., 13, 85-99.
Frassoni A., Rossi G.C. and Tamburini A. (2001): Studio del Ghiacciaio dell’Adamello mediante
indagini georadar. Suppl. Geogr. Fis. Dinam. Quat., V, 77-84.
Haeberli, W., Kääb, A., Paul, F., Chiarle, M., Mortara, G., Mazza, A. and Richardson, S. (2002): A
surge-type movement at Ghiacciaio del Belvedere and a developing slope instability in the east
face of Monte Rosa, Macugnaga, Italian Alps. Norwegian Journal of Geography, 56(2), 104-
111.
Haeberli, W., Huggel, C., Kääb, A., Polkvoj, A., Zotikov, I. and Osokin, N. (2003): Permafrost
conditions in the starting zone of the Kolka-Karmadon rock/ice slide of 20th September 2002 in
North Osetia (Russian Caucasus). Extended Abstracts, Eighth International Conference on
Permafrost, 49-50.
Huggel C., Kääb A., Haeberli W., Mortara G., Chiarle M., Epifani F., Viazzo G. and Tamburini A.
(2002): An integrative view on glacier hazard related to the surge-type development at
Ghiacciaio di Belvedere. Poster. Proceedings of the Conference “I ghiacciai, le montagne,
l’uomo. Le variazioni dei ghiacciai montani e le modificazioni dei sistemi naturali ed antropici”,
Bormio 13-14 September 2002.
Kääb, A. and Haeberli, W. (2001): Evolution of a high-mountain thermokarst lake in the Swiss Alps.
Arctic, Antarctic, and Alpine Research, 33(4), 385-390.
Kääb, A., Huggel, C., Haeberli, W., Mortara, G., Chiarle, M. and Epifani, F. (2003a): Studio sui
problemi connessi alla recente evoluzione dei fenomeni di instabilità riguardanti il Ghiacciaio
del Belvedere e la parete orientale del Monte Rosa, Department of Geography, University of
Zurich, and CNR-IRPI, Torino.
Kääb, A., Wessels, R., Haeberli, W., Huggel, C., Kargel, J. and Khalsa, S.J.S. (2003b): Rapid ASTER
imaging facilitates timely assessment of glacier hazards and disasters. EOS Transactions,
American Geophysical Union, 84(13), 117,121.
Kääb A., Huggel C., Barbero S., Chiarle M., Cordola M., Epifani F., Haeberli W., Mortara G., Semino
P., Tamburini A. and Viazzo G. (in press): Glacier Hazards At Belvedere Glacier And The
Monte Rosa East Face, Italian Alps: Processes And Mitigation. Intrepraevent 2004.
Mazza, A. (2003): The kinematics wave theory: a possible application to "Ghiacciaio del Belvedere"
(Valle Anzasca, Italian Alps). Preliminary hypothesis. Terra glacialis, 6, 23-36.
Mercalli L., Cat Berro D., Mortara G. and Tamburini A. (2002a): Un lago sul Ghiacciaio del
Rocciamelone, Alpi Occidentali: caratteristiche e rischio potenziale. Nimbus, 23-24, A.VII, 3-9.
Mercalli L., Mortara G. and Tamburini A. (2002b): Il ghiacciaio sospeso della Croce Rossa, Valli di
Lanzo: misure ed evoluzione recente. Nimbus, 23-24, A.VII, 18-26.
Monterin, U. (1923): Il Ghiacciaio di Macugnaga dal 1870 al 1922. Bolletino del Comitato
Glaciologico Italiano 5: 12-40.
Monterin, U. (1926): La fine della fase progressiva e l’inizio della nuova fase di ritiro dei ghiacciai
italiani nel Monte Rosa 1922-1925. Zeitschrift für Gletscherkunde 15 (1): 31-54.
Mortara G. and Mercalli L. (2002): Il lago epiglaciale «Effimero» sul ghiacciaio del Belvedere,
Macugnaga, Monte Rosa. Nimbus, 23-24, 10-17.
Regione Piemonte, Direzione Servizi Tecnici di Prevenzione (2002): Il lago epiglaciale del Ghiacciaio
del Belvedere a Macugnaga (VB). Versione 2, 18 luglio 2002.
Somigliana, C. (1917): Primi rilievi del Ghiacciaio di Macugnaga. Rivista Club Alpino Italiano, 36 (3-
4): 65-67.
Stoppani, A. (1871): Il Bel Paese. Ed. Agnelli, Milano.
Tamburini, A., Mortara, G., Belotti, M. and Federici, P. (2003): The emergency caused by the "Shortlived
Lake" of the Belevdere Glacier in the summer 2002 (Macugnaga, Monte Rosa, Italy).
Studies, survey techniques and main results. Terra glacialis, 6, 37-54.
22 Deliverable 4

Tropeano, D. Govi, M., Mortara, G., Turitto, O., Sorzana, P., Negrini, G. and Arattano, M. (1999):
Eventi alluvionali e frane nell’Italia Settentrionale (1975-1981). Consiglio Nazionale delle
Ricerche, Publ. N. 1927 del GNDCI.
VAW (1985): Studi sul comportamento del Ghiacciaio del Belvedere, Macugnaga, Italia. Relazione,
Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie der ETH Zürich, 97(3), 157.
Walder, J.S. and Costa, J.E. (1996): Outburst floods from glacier-dammed lakes: the effect of mode of
lake drainage on flood magnitude. Earth Surface Processes and Landforms, 21, 701-723.
Study of the Arsine pro-glacial lake.
C. Vincent et E. Le Meur,
Laboratoire de Glaciologie et de Géophysique de l’Environnement, CNRS, BP 96 38402 Saint Martin d’Hères
Cedex, France
1 General objectives.
Within the framework of WP2, the LGGE is involved in a field study relative to the proglacial lake
hazard of Arsine glacier.
N
Grenoble
Ecrins
Durance
50 km
Geneva
Switzerland
Chamonix
Vanoise
ARSINE
France
Mont-Blanc
Rhône
Swiss Alps
TACONNAZ
Nice
Turin
Italy
Mediterranean
Figure 1: French Alps. Locations of the 2 glaciers studied in the framework of the Glaciorisk project.
Taconnaz glacier is located in the Mont Blanc range. Arsine glacier is located in the Ecrins area.
The Arsine study is relative to proglacial lakes formation: this study is based on regular surveys which
began in 1985/1986 when a large proglacial lake started to threat the valley of La Guisane At that
time, overflow was already expected to occur in the near future. In order to prevent this flood hazard, a
channel was dug in 1986, and since then, many field observations have been conducted to ensure this
proglacial lake would not overflow in the future. However, the mechanisms relative to ice flow and
glacier fluctuations which result in the lakes formation generally following glacier retreat remain
poorly known. Therefore, the Glaciorisk programme was an excellent opportunity to make extensive
measurements to describe the glacier fluctuations in the past (from photogrammetric measurements
with past aerial photographs) and to understand the mechanisms which led to the lakes evolution
Deliverable 4 23

2 Previous observations and results.
Lake of Arsine first appeared during the forties and has been growing steadily until 1986. The
evolution of this proglacial lake until 1986 has been thoroughly described by Vallon (1989).
Following this extensive study and consequent mitigating works under the form of the construction of
an artificial spillway, the Arsine glacier-lake system has always been the subject of special care by
LGGE since then. Since 1986, a network of observations has been carried out in order to measure 1°)
the surface mass balance of this glacier from ablation stakes, 2°) the surface ice flow velocities from
ablation stakes and from painted stones, using topographic measurements, 3°) the extension of the
lakes. From the data collected over the last 3 decades, it emerged that the glacier has undergone
noticeable changes in its surface velocity field especially in its lower part which was calving into the
main pro-glacial lake until 1991. A strong steadily decrease in these velocity values since 1985 added
to the fact that ice thickness changes between 1969 and 2000 (inferred from comparison of the
respective maps) are small over the same area leads to the conclusion that these velocities must have
been much higher in the 1969-1985 period. Moreover, the still-going decrease associated with minor
thickness changes is an indicator of fast-evolving dynamics for the glacier. This is corroborated by the
rapid evolution of the snout-lake system where the right tongue receeded before stopping calving
(1991) and finally disconnecting from the main lake in 1994. In the mean time, more to the west,
melting of downstream dead ice gave birth to several small lakes that rapidely merged into a
secondary main lake which developped a calving front but with the left snout of the glacier this time.
Although the present-day situation seems pretty safe (also as a result of the works carried out in 1986),
this fast-evolving context due to changing dynamics for the glacier can provide favourable conditions
leading back to a risky situation. Indeed, an other calving front is likely to develop in case of the onset
of a new lake from a possible further retreat of the glacier. Although less probable in the global
context, a glacier readvance could reactivate these calving fronts since the lakes are still there and the
spillway has been damaged since its construction.
3 Measurements carried out in the framework of the Glaciorisk project.
On top of the classical survey that has been going on since the mid eighties, some more insight into the
glacier dynamics have been undertaken in the framework of the GLACIORISK project. Available
aerial photographs from 1967, 1981,1986 and 1991 have been used for photogrammetric
measurements in order to better understand the past glacier dynamics from thickness changes. These
data are not only useful as such, but can also serve as constraints for an ice flow model available at
LGGE that could reasonably be applied to the Arsine glacier in order to predict its future behavior. For
this purpose, available aerial photographs taken by the National Geographic Institute and by the
Sintegra company have been used. The scale of these photographs ranges from 1/15000 to 1/40000.
Moreover field geodetic measurements have been performed in order to determine accurate
coordinates of ground control points which are visible on the photographs. Thanks to an analytical
instrument (Leica DSR ), digital elevation models have been made from these aerial photographs.
4 Results
4.1 Thickness variations from the photogrammetric measurements (Figure 10)
The thickness variations obtained from the digital elevation models using photogrammetric
measurements (aerial photographs of 1981, 1986 and 1991) have been reported on Figure 10.
Additional photogrammetric measurements obtained from aerial photographs taken in 1967 have been
used to determine 5 transect profiles. Between 1967 and 1981 the surface has been swelling (5 to 10
m) in the upper part of the glacier (>2600 m) whereas the measurements carried out in the lower part
do not show a clear pattern. One must notice that the glacier topographic surface in this area below
2600 m is very chaotic and is covered by rock debris whose thicknesses range from 0 to 1 m thick.
Between 1981 and 1986 (Figure 10), the left stream evolution is clearly different from that of the right
one: the surface has been swelling on the left western half of the glacier (0 to 1.6 m/yr) wheras the
right stream has been decreasing (0 m to -2 m/yr). Between 1986 and 1991, the thickness variation is
24 Deliverable 4

strongly negative on the right stream (0 to –3 m/yr) although the left stream surface variation is close
to zero. Finally, the topographic measurements performed in 2002 on 5 cross profiles show a large
thinning everywhere (-0.5 to –2 m/yr).
4.2 The lakes extensions (Figure 11).
Figure 10: Thickness variations of the Arsine glacier.
The limits of the lakes are reported on Figure... The eastern lake appeared during the forties and has
been steadily growing until 1986. By the end of 1985, it was obvious that the lake would overspill
within the next year and an articial channel was dug in the moraine in Spring 1986 in order to stabilize
the level of the lake and to prevent the water from overflowing (Vallon, 1989). Between 1986 and
1991, this eastern lake has been extending toward the south as the snout of the glacier has been
retreating and calving in the lake. In 1991, the glacier stopped calving in the eastern lake and in 1994,
the snout disconnected from it
The western lake appeared during the eighties. In fact, several water pockets of water first appeared
and merged to form a lake after 1990. The extension of this lake is fed by the melting of the “dead ice”
(stagnant ice) which is not visible from the surface because it is covered by a thick layer of rock debris
(more than 50 cm). The melting of this stagnant ice has lowered the surface thereby allowing the lake
to extend. The process was supposed to last as long as this dead ice had not entirely melt. Since 1994,
the glacier has begun to calve in this western lake. In 1999, from bathymetric measurements, the
volume of this lake was estimated to 82300 m 3 with a surface of 24200 m 2 and a mean depth of 3.60
m. In 2003, new bathymetric measurements have been performed: the volume was estimated to 90500
m 3 with a surface of 24500 m 2 and a mean depth of 3.70 m, which is not significantly different from
the value obtained in 1999. That means the stagnant ice located at the bottom of the lake has now
entirely melted. However, the lake surface area is still extending and the glacier continues to calve
into the lake.
Deliverable 4 25

Acknowledgements
305100
305000
304900
304800
304700
Top of the cliff
in 2001
in 2003
304600
304500
304400
304300
Lake sides
in 2002
in 2003
N
Front
1994
Lake
1986
Western lake
1999
Front
1994
Front
2001
Lake
1986
Eastern lake
(1986 edges)
304200
921200 921300 921400 921500 921600 921700 921800
0 100 200 300 400
Figure 11: Arsine lake extensions since 1986.
We would like to thank all those who collected data from field measurements. A part of this study was
also supported by Le SIVOM de la Haute Vallée de l’Arve, by La commune de Le Monestier les
Bains, et le service du RTM des Hautes Alpes.
References :
Paterson, W.S.B. 1994. The Physics of glaciers. Third edition. Oxford, Elsevier.
Vallon, M. Evolution, water balance, potential hazards, and control of a pro-glacial lake in the French
Alps, Annals of Glaciology, 13, 273-278, 1989.
Vincent, C. 2002. Influence of climate change over the 20 th Century on four French glacier mass
balances. J. Geophys. Res., 107(D19), 4375,doi:10.1029/2001JD00832.
26 Deliverable 4

GLOFs in Norway.
1 Miriam Jackson
Glacier and Environmental Hydrology Section, Norwegian Water Resources and Energy Directorate,
P.O. Box 5091 Majorstua, N-0301 Oslo, Norway.
Glacier outburst floods can also be a serious problem in Norway. Two glaciers that have been studied
within Glaciorisk in relation to outburst floods are Blåmannsisen in northern Norway and Folgefonn in
western Norway. Mass balance measurements have been performed previously on Folgefonn by
NVE. In 2003, these measurements began anew. During one fieldwork period, evidence of an
outburst flood was noted. As well as observing that the lake was empty, large blocks of ice were also
seen that were below the level of a visible waterline. The water level was then 28 m below this level,
which gives a volume of water that had been released of between one and one and a half million cubic
metres. At the west end of the lake there were signs that there could have been a river outlet from the
glacier. The lake was not completely empty. It appears that the level of the water was approximately
as high as the outflow for the lake. The glacier area about 50 m upstream was highly crevassed, which
may be an indication of drainage in this area. However, that contradicts the possibility of a discharge
outlet from the glacier at the western end of the lake. There were no signs of glacial sediment in the
emptied lake, which may imply little subglacial input. Because of new snow, it was not possible to
see if there were traces of flooding in the gully downstream from the outlet (Figures 1 and 2). The
figures show the glacier-dammed lake in 1997, and the almost-emptied lake in 2002, with common
features (marked A and B) in both pictures for comparison.
Figure 1. Glacier-dammed lake before jøkulhlaup, Folgefonn. Photo: Hallgeir Elvehøy, August 1997.
1 Contact Miriam Jackson at mja@nve.no.
Deliverable 4 27
B
A

Figure 2. Glacier-dammed lake after jøkulhlaup, Folgefonn. Photo: Hallgeir Elvehøy, October 2002.
A GLOF from the glacier Blåmannsisen in September 2001 has also been studied (Figure 5). This was
the first known occurrence of a jøkulhlaup from this glacier. It began on 6 September 2001 and lasted
35 hours. The entire lake, ~40 x 10 6 m 3 of water, drained into Norway; previously the lake drained
over a rock spillway into Sweden. There were no casualties or material damage; on the contrary, it
increased the volume of water in the hydropower reservoir Sisovatn, so was financially beneficial.
Map comparison and mass balance modelling show evidence of glacier retreat during the last four
decades. Ice thickness decreased in the ablation zone, reducing ice barrier stability. The lake drained at
a water level 40 m below what was required to equalise the ice-overburden pressure. The shape of the
flood hydrograph indicates drainage through a subglacial channel. Measurements show an ice barrier
thinning of 3.5 m since the jökulhlaup. The lake volume is 82 % full 2 ½ years after the event, giving a
likely repeat frequency of three years. Future events may be triggered at lower water levels and
produce lower water volumes. The company responsible for the reservoir has instigated a fieldwork
programme that is being carried out by NVE investigating the glacier mass balance and the likelihood
of a future jøkulhlaup, and has supported the work financially and logistically. Additional analysis
and presentation of results has been carried out through Glaciorisk. Several summaries of this work
are included in previous Glaciorisk reports and have been published in reports by NVE.
28 Deliverable 4
B
A

Figures 5a (above) and b (below). Photographs of Øvre Messingmalvatn and Rundvassbreen. showing (a) the
emptied lake photographed one week after the event, and (b) the partially filled lake one year after the jökulhlaup
(on 4 September 2002). Both photos are taken in direction West from a helicopter by Hans Martin Hjemaas,
Elkem Energi Siso.
Acknowledgements
Ole Magnus Tønsberg, Bjarne Kjøllmoen, Rune Engeset and Hallgeir Elvehøy all participated in the
work in Work Package 2 of Glaciorisk at NVE.
Deliverable 4 29

Jökulhlaups
Case by case scientific studies of jökulhlaups.
Helgi Björnsson, Finnur Pálsson and Alexandra Mahlmann
Science Institute, University of Iceland, Dunhagi 3, 101 Reykjavik
I. Subglacial lakes and jökulhlaups in Iceland.
1 Introduction
Jökulhlaups can be traced to subglacial lakes at geothermal areas and to meltwater drained during
volcanic eruptions. Iceland is a unique and valuable study site for glaciovolcanic interactions. At
present 10% (11.200 km 2 ) of Iceland is ice-covered and 60% of the ice overlies the active volcanic
zone (Fig. 1). During the 20 th century 15 subglacial volcanic eruptions (10 major and 5 minor events)
took place, about one-third of all eruptions in Iceland during that century.
Jökulhlaups, both those draining meltwater stored in subglacial lakes and meltwater produced during a
volcanic eruption, have significant landscaping potential: they erode large canyons and transport and
deposit enormous quantities of sediment and icebergs over vast outwash plains and sandur deltas.
Pleistocene glacial river canyons were formed in such catastrophic floods from subglacial lakes.
Jökulhlaups have threatened human populations, farms and hydroelectric power plants on glacier-fed
rivers. They have damaged cultivated and vegetation areas, disrupted roads on the outwash plains and
have even generated flood waves in coastal waters. Knowledge of the sources and behaviour of
jökulhlaups is essential for advanced warnings and civil defence in Iceland. Monitored seismic activity
announces volcanic eruptions, and maps of bedrock and glacier surface topography can be used to
delineate jökulhlaup hazard zones [14].
Fig. 1 Location map of Iceland,
showing icecaps, the volcanic zone and
the central volcanoes.
2 Ice caps, volcanic centres and drainage basins
Subglacial topography of all the major ice caps in Iceland has been mapped by radio echo sounding,
revealing bedrock elevations from -300 to 1800-2000 m relative to sea level. The volcanic zone
underlies the four largest ice caps and the most active volcanoes are located under the ice cap
Mýrdalsjökull (600 km 2 ) and the western part of Vatnajökull (8100 km 2 ), which is positioned over the
centre of the Icelandic mantle plume [66]. No eruptions have taken place in Hofsjökull (900 km 2 ) or
Langjökull (925 km 2 ) ice caps during the last millennium.
30 Deliverable 4

Five volcanic systems have been identified under the ice cap Vatnajökull (Fig. 1), each containing
central volcanoes and fissure swarms that stretch for tens of kilometres [9,31,14]. The most active are
the Grímsvötn and Bárdarbunga volcanic systems, both of which have relief of up to 1000 m and
contain calderas 600-700 m deep and 15-20 km in diameter. In Grímsvötn about 30 eruptions have
been reported over the last 400 years [60,14]. More than 80 volcanic eruptions occurred during the last
800 years in Vatnajökull [37]. Since the settlement of Iceland around 870 A.D., twenty volcanic
eruptions have been traced to the Katla volcanic system in Mýrdalsjökull ice cap [36]. This volcanic
system contains a 600-750 m deep caldera that is 10-15 km in diameter [16].
At several of the central volcanoes under Vatnajökull and Mýrdalsjökull hydrothermal activity results
from the interaction of water with magmatic intrusions at shallow depths in the crust [61,9,16,17]. Ice
is continuously melted at the glacier bed creating permanent depressions in the glacier surface that
reveal this hydrothermal activity. The meltwater, however, may be trapped in a lake at the bed due to
relatively low basal fluid potential under the depression. High ice overburden pressure at the rim
around the depression seals the lake, the best-known examples of such subglacial lakes are Grímsvötn
in the interior of Vatnajökull and the Skafta cauldrons (Fig.2). Smaller lakes of this kind are located at
Pálsfjall and Kverkfjöll. The current hydrothermal activity in Mýrdalsjökull is concentrated just inside
the caldera rims, where faults allow rapid vertical transport of hydrothermal fluid.
2.1 Drainage basins
On the basis of the glacier surface and bedrock maps, the ice and water drainage basins on all the
major ice caps in Iceland have been delineated and the location and geometry of subglacial lakes
identified [7,8,9,16]. The glacierized portion of catchment basins for the major glacial rivers has been
drawn as a continuation of the watershed outside the glacier. In principle the watershed at the glacier
base is situated along ridge crests in the fluid potential
φb = ρw g zb + pw , (1)
which is the sum of terms expressing the gravitational potential and the water pressure, pw. The
symbol ρw = 1,000 kg m -3 represents the density of water, g = 9.81 m s -2 is the acceleration due to
gravity and zb is the elevation of the glacier substrate relative to sea level. Water in an isotropic basal
layer would flow perpendicularly to lines of equal potential. The basal water pressure is assumed to be
static,
pw = k pi , (2)
where pi = ρi g H is the ice overburden pressure, k is a constant, ρi = 916 kg m -3 is the density of ice,
H = zs - zb is the thickness of the glacier and zs is the elevation of the ice surface relative to the sea
level [51]. In our experience k is close to one [7,9] and thus the gradient driving the water is
∇φ b = (ρ w - ρ i ) g ∇z b + g ρ i ∇z s . (3)
This expression predicts that the glacier
surface slope is about ten times more
effective than the bed slope in
controlling basal water flow. This static
approximation of water pressure (pw ≈ pi)
is plausible at low values of discharge
and is therefore useful for delineating
water divides.
Fig. 2 The location of subglacial geothermal systems in
Vatnajökull. The internal drainage basins of Grímsvötn and
Skaftá cauldrons drain in jökulhlaups.
Vatnajökull and Mýrdalsjökull have been divided into 15 and 3 major drainage basins, respectively
(Fig. 2). In six basins in Vatnajökull water accumulates into subglacial lakes, which are not connected
Deliverable 4 31

to the glacier margin. Water may accumulate under 12 small ice cauldrons in Mýrdalsjökull (Fig 19).
During volcanic eruptions depressions are created in the glacier surface above the eruption site and
the location of the watersheds may change. Nevertheless, if our model of a static fluid potential
applies at the beginning of an eruption we consider it likely that meltwater from the eruption site will
continue to drain through existing conduits.
2.2 Subglacial lakes
Subglacial lakes can be situated where there is no gradient in the fluid potential that drives water
along the glacier bed. The condition ∇φb = 0 has been used to define the location and geometry of
subglacial lakes. Hence
∇z b = - ( ρ i /(ρ w - ρ i )) ∇z s , (4)
describes a relationship between the slope of the ice/water boundary of the lake and the upper glacier
surface. The shape of the lake results from equilibrium of vertical forces as the overlying glacier
floats in static equilibrium. The roof of the subglacial lake slopes approximately ten times more
steeply than the glacier surface, and in the opposite direction (Fig. 3).
In Icelandic ice caps several lakes are known to exist beneath surface depressions created above
hydrothermal systems. The lakes rise as a dome above the bed, even capping mountains [4]. Theory
combined with topographic maps suggests that the slopes of glacier-bed depressions beneath ice caps
in Iceland are not sufficient to accumulate water without accompanying depressions in the glacier
surface. However, there could be lakes, not treated here, with equal in- and outflow that don’t meet
the condition of zero gradient in the fluid potential.
2.3 Anatomy of a jökulhlaup
Subglacial lakes beneath ice-surface depressions drain periodically in outburst floods. A subglacial
lake gradually expands as water flows toward the depression, the basal water pressure increases and
the overlying glacier is lifted. Before the surface depression is completely flattened the hydraulic seal
is broken and water begins draining out of the lake at the base under the ice dam. The water escapes
through narrow passages at the ice-bed interface but the pressure head maintained by a voluminous
lake drops slowly. Although the pressure of the ice squeezes the tunnel draining water from the lake,
water flow is primarily controlled by tunnel enlargement. In most cases, tunnel enlargement can be
explained as melting of the ice walls by frictional heat generated by the flowing water and thermal
energy stored in the lake [39,22,11]. The lake may become sealed again before it is empty and
accumulation of water begins until a new jökulhlaup occurs.
Nye [39] presented a general theory of flow in water-filled tunnels based on the principles of mass
continuity, energy conservation, heat transfer and water discharge for a given fluid potential gradient.
He derived an analytical solution predicting discharge Q to rise asymptotically with time as Q(t) =
k(1/t) 4 if the overburden closure is neglected and the expansion of the ice tunnel is solely due to the
instantaneous transfer of frictional heat from the flowing water to the ice walls (loss of potential
energy). Under the assumption of instantaneous heat transfer water would emerge at the river outlet at
the melting point, as is generally observed in jökulhlaups in Iceland, whereas the heat transfer
equation would predict higher temperatures.
Occasionally the discharge hydrograph for jökulhlaups increases faster than can be explained by the
expansion of conduits by melting [12]. The water pressure exceeds the ice overburden and the glacier
is lifted to make space for the water [11,12]. This event cannot be explained by the classical theory of
32 Deliverable 4

jökulhlaups. Jóhannesson [32] has described these floods as characterised by a propagation of a
subglacial pressure wave. A preliminary model study suggests that the idea of a sheet flood preceding
conduit drainage is plausible [17].
3 Subglacial lakes, volcanic eruptions and jökulhlaups in Vatnajökull
Six subglacial lakes have been discovered in Vatnajökull (Fig. 2). For two of these we will describe
their geometry and growth, and the characteristics of their jökulhlaups.
Fig. 3 Schematic drawing of an
unstable lake that drains into
jökulhlaups.
3.1 Grímsvötn
3.1. 1 Geometry and drainage of ice and water
Fig. 4 An oblique aerial photograph of Grímsvötn after the
10-km-wide and 270-m-thick ice cover had subsided 175 m
during the 1996 jökulhlaup.
Grímsvötn, the largest subglacial lake in Iceland, is located in the western part of Vatnajökull (Fig. 2,
Table 1). The glacier covers a hydrothermal area and a 10 km wide and 300 m deep depression has
been created in the ice surface (Fig. 4) [3,9]. The extent of the subglacial lake is identified by the flat
ice shelf floating on the lake and the abrupt change in surface slope at the margins. Subglacial
topography is known from radio echo-soundings and seismic profiling [9,23,24]. The lake is situated
within the caldera floor of the Grímsvötn volcano and to the south and the west caldera walls (Mt.
Grímsfjall) protrude through the glacier surface and confine the lake, but the lake can expand to the
north and northeast as the water level rises. The lowest breach of the caldera rims is 1150 m but due
to the glacier surface depression the lake level at the slopes of Mt. Grímsfjall can rise more than 300
m higher. The ice-shelf thickness has been measured regularly by radio echo-soundings and computed
from measured surface elevations, assuming the shelf is floating in hydrostatic equilibrium. The ice
shelf gradually thickened from 150 m in the 1950's to 230-260 m in the 1980's due to reduced melting
by the hydrothermal system [15]. Hence, the extent and volume of the subglacial lake has gradually
been reduced. The total volume of water drained out of Grímsvötn in jökulhlaups has been derived
using known variations in lake level during jökulhlaups, the thickness of the floating ice cover on the
lake and the bedrock topography in the Grímsvötn area [9,11].
In recent years the lake level has risen 10-15 m a -1 and a jökulhlaup occurrs when it rises 80-110 m
and reaches a particular threshold (Fig. 5). The onset of lake drainage is marked by ice-quakes and
subsidence of the lake level, and the arrival time of lake water to the glacier margin is identified by a
sulphurous odour in the glacial river. Jökulhlaups from Grímsvötn flow beneath Skeiðarárjökull a
distance of 50 km to the terminus at Skeiðarársandur. The flow path reaches a depth of 200 m below
sea level before it emerges on the outwash plain. In the most violent jökulhlaups from Grímsvötn, the
entire outwash plain, Skeidarársandur, has been flooded. Jökulhlaups from Grímsvötn have occurred
at 1 to 10 year intervals, with peak discharges of 600 to 4-5 10 4 m 3 s -1 at the glacier margin, duration of
Deliverable 4 33

2 days to 4 weeks and a total volume of 0.5 to 4.0 km 3 [15,26,12,54]. In general the frequency of
jökulhlaups and the volume of water released depends upon the thickness of the ice barrier [60].
The typical threshold lake-level for triggering a jökulhlaup is 60-70 m lower than required for simple
flotation of the ice dam [9,11]. This suggests that some process other than lifting finally breaks the
seal and permits the water to enter conduits beneath the ice dam, resulting in a jökulhlaup. Jökulhlaups
from Grímsvötn occur at any time of the year so sudden changes in subglacial drainage due to surface
melting do not, in general, trigger jökulhlaups. A few jökulhlaups have occurred from Grímsvötn at
lake levels far below the usual threshold. These premature jökulhlaups may have been triggered by
the opening of waterways from the lake along the
Lake: Grímsvötn
Lake Type: Subglacial
Lat: 64°25´ N
Lon: 17°21´ W
Area: 5 - 40 km 2
Catchment Basin: 160 km 2
Ice shelf thickness: 150 - 300 m
Lake level elevation: 1270 - 1450 m
Deepest point at lake bottom: 1060 m
Max. Depth: 390 m
Failiure Mechanism:
hydraulic, subglacial drainage
Geothermal output: 2x10 3 – 4x10 3 MW
Basal melting rate: 6 – 12 m 3 /s
Meltwater accumulation rate: 0.2-0.5
km 3 /a
Lake level rising rate: 10-15 m/a
Lake level fluctuations between
jökulhlaups: 80-150 m
Downstream Topography: ice, sandur
Legnth of floodpath: subglacial: 50 km,
over sandur: 20 km
Average grade to glacier margin: 1.3°
Flood River: Skeidará, (Sandgígjukvísl)
Flood Interval: between 1 and 10 years
Peak discharge: 600-50,000 m 3 /s
Flood Volume: 0.5-5 10 6 m 3 /s
Sediment load: 30-150 10 6 ton
Duration: 2 to 30 days
Lake Level Change: 40 - 160 m
Mortality: At least one in 19 th century
Damages: road, bridges, fertile land
Last event: July 2003
3.1.2 Characterisitics of jökulhlaups from Grímsvötn
north-eastern slopes of Grímsfjall (higher than the
lowest crest at the caldera rim) facilitated by
increased localised melting due to hydrothermal or
volcanic activity.
Jökulhlaups from Grímsvötn terminate abruptly,
often within a few hours, once the water level has
dropped about 100 m. This occurs before the lake is
empty, when the ice overburden pressure at the rock
rim exceeds the hydrostatic water pressure by 10-15
bars. The water level in the lake does not drop to
1,150 m, the level of the subglacial caldera ridge.
Fig 5 The lake level of Grímsvötn, 1930-2000. The lake
level ascends until a jökulhlaup takes place. In 1996 the
lake rose to the level required for flotation of the ice dam.
Table 1 Data on the subglacial lake Grímsvötn and
typical jökulhlaups from it [9,15].
Typically jökulhlaups from Grímsvötn increase approximately exponentially to the peak, and falls
more rapidly afterward (Fig. 6a). The duration of large floods tends to be shorter than that of small
floods. The smaller jökulhlaups reach their peak in 2-3 weeks and terminate about one week later.
This discharge pattern can be explained by enlargement of a single basal ice tunnel due to melting by
frictional heat generated in the flowing water. The drainage affects glacier sliding only over small
localized areas. Outburst floods drain through one main ice tunnel, which feeds river Skeiðará. If
34 Deliverable 4

discharge exceeds about 3,000 m 3 s -1 water starts to drain from other, smaller tunnels at the central
part of the terminus, collecting into the river Gígja. In the most voluminous floods, ten to fifteen highcapacity
ice tunnels develop. The rapid injection of water creates high water pressures, as evidenced
by water forced up to the glacier surface through crevasses that formed in 300 m thick ice.
To date, modelling based on the theory of Nye [39] and Clarke [22], has succeeded in simulating the
increasing discharge of several jökulhlaups from Grímsvötn but fails to adequately represent the sharp
transition from increasing to decreasing dicharge [11]. A cylindrical tunnel apparently cannot describe
the fast closure of the conduit. Nye’s [39] prediction that the discharge Q rises asymptotically with
time as Q = (1/t) 4 successfully described the rising limb of the jökulhlaup from Grímsvötn in 1972.
Spring and Hutter [55] applied the general equations, which include thermal energy stored in the lake,
to simulate the 1976 hydrograph and concluded the lake water would have been 4 o C. The lake
temperature, however, was measured close to the melting point [11]. This supports the observation
that the actual heat transfer by flowing water during jökulhlaups is more effective than accounted for
by the empirical heat transfer equation described by Nye [39] and Spring and Hutter[55].
Occasionally, jökulhlaup discharge has increased at a rate suggesting that thermal energy stored in the
lake contributed to tunnel expansion. The rise of the jökulhlaup in 1934 can be simulated by lake
water of 1 o C, and in 1938 by water of 4 o C when a peak of 3x10 5 m 3 s -1 was reached in 4 days (Fig.
7). Lake temperature values should not be taken seriously, however, as the heat transfer theory is
questionable.
An empirically based regression relation between the peak discharge Qmax (in terms of m 3 s -1 ) and the
total volume Vt (in 10 6 m 3 ) of water passing through subglacial tunnels in jökulhlaups was formulated
by Clague and Mathews [21]:
Qmax = K Vt b
Referring to eleven jökulhlaups from the Icelandic lake Grímsvötn, Björnsson [11] found K = 4.15 x
10 -3 s -1 m -2.52 and b = 1.84.
Fig. 6 a.Typical shape of a hydrograph when a
single basal ice tunnel enlarges due to melting.
b.Rapidly rising hydrograph that is not explained
by the classical theory of jökulhlaups (typical for
jökulhlaups from the Skaftá cauldrons, see 3.2)
3.1.3 Extraordinary jökulhlaups from Grímsvötn
Fig. 7 Typical hydrographs of jökulhlaups from
Grímsvötn (1934, 1938, 1954, 1976, 1982, 1986 and
1996).
The jökulhlaup from Grímsvötn in November 1996 was of an extraordinary type, draining so rapidly
that melting alone could not have expanded the conduits (Fig. 7and Fig 8). For the first time in the
observational history of Grímsvötn the ice dam was floated off the bed (Fig. 10). Downstream from
the dam, water pressure exceeded the ice overburden and the glacier was lifted off the bed along the
water flowpath [12,48]. Crevasses were observed along the main flowpath and icebergs were broken
off the margin and spread over Skeidarársandur outwash plain. As the first jökulhlaup of this kind to
take place after scientific observations began, it has provided new insight into jökulhlaup hydrology.
Deliverable 4 35

Similar jökulhlaups may have occurred before, although descriptions do not provide unquestionable
evidence for as rapid a rise in discharge (e.g. in 1861 and 1892) [60,9].
The 1996 jökulhlaup was preceded and indirectly triggered by the Gjálp eruption, which took place
inside the drainage basin of Grímsvötn (Fig. 2). Meltwater was accumulated for a month until it
drained in the catastrophic jökulhlaup. The eruption broke through the ice cover at one place after 30
hours eruption, but continued subglacially for two weeks on a 6 km long fissure. Melting by the
eruption was measured both from the volume of the surface depressions above the eruptive fissure,
and from the volume of meltwater accumulated in Grímsvötn [27,19]. During the first four days
meltwater was produced at a rate of 5000 m 3 s -1 and the heat output at the peak of the eruption was
10 12 W, (more than 100 times all the power stations producing electricity in Iceland at that time). The
high rate of melting can only be due to fragmentation of the lava into glass (hyaloclastites) and rapid
cooling of the fragments by quenching. The total volume of ice melted during the first six weeks after
the beginning of the eruption was 4.0 km 3 , equivalent to 1.1x10 12 kg of magma cooling from 1,000 o C
to 0 o C if all the heat were used for melting. After one year (January 1998) the melted volume was 4.7
km 3 .
The meltwater accumulated in the Grímsvötn subglacial lake for a month until it drained in a
catastrophic jökulhlaup from November 4-7 1996, in which 3.2 km 3 of water drained from the lake
within a period of 40 hours. On November 4th, the lake had risen to the level required for flotation of
the ice dam, 1510 m, and ice-quakes marked the onset of lake drainage. About 10.5 hours later water
emerged from the margin of Skeiðarárjökull as a flood wave inundating Skeiðarársandur (at 100 m
elevation), in the most rapid jökulhlaup ever recorded from Grímsvötn (Fig. 8). The discharge out of
the lake during this jökulhlaup was derived directly from lake level observations and the known lake
hypsometry. Discharge increased linearly with time and reached a peak value of 4 x10 4 m 3 s -1 in 16
hours and dropped to zero 27 hours later. The total volume of water released from the glacier was
estimated at 3.2 km 3 . During the drainage the lake surface subsided by 175 m (Fig. 5), and the floating
ice cover was reduced from 40 to less than 5 km 2 . While the discharge hydrograph for the lake was
derived, the shape of the hydrograph for the outburst from Skeiðarárjökull is unknown and likely to
be different since the flood was not drained through a single subglacial conduit. Four points on the
hydrograph, two on the rising limb and two during the recession were estimated but their accuracy is
uncertain [54].
36 Deliverable 4
Fig. 8 Discharge hydrograph
of the 1996 jökulhlaup from
Grímsvötn, the cumulative
volume of water drained and
the subsidence of the lake
level (measured by precision
barometric altimetry).

Fig. 9 The margin of
Skeiðarárjökull during the
jökulhlaup of 1996. The flood
water emerged from the base
through crevasses 200 m
higher than the front. (Photo:
Helgi Björnsson, November
1996)
Fig. 10 A cross-section from
Grímsvötn to Skeiðarársandur
along the flowpath of the
jökulhlaup in 1996.
Subglacial hydraulic
conditions at the beginning of
the 1996 jökulhlaup: Pi/(ρwg)
is the ice overburden pressure
in metres of hydraulic head;
Pw/(ρwg) is the hydraulic head
maintained by the lake. The
ice dam was floated at the
onset of the 1996 jökulhlaup.
During the lake drainage a 6 km long, 1 km wide and 100 m deep depression was created by collapse
of the jökulhlaup flowpath across the ice dam (Fig 12 and Fig 13). The volume of the depression was
0.3 km 3 . Assuming that all of the thermal energy in the lake was used in the formation of the
depression, we calculate the average temperature of the 3.2 km 3 of water released to be 8 o C. In
addition to this we calculate that 0.1 km 3 of water was produced by frictional melting during the
descent of the water from Grímsvötn to Skeiðarársandur.
Grímsfja
ll
500 m
Grímsvötn
100
Deliverable 4 37

Grímsfja
ll
500 m
Grímsvötn
Fig. 11 (left) The map shows the subglacial flood path of the
3
When floodwater started to drain from the jökulhlaup glacier in margin 1996. a volume of 0.6 km of water had
accumulated under the glacier (Fig. 8). Melting enlargement of the conduit by the frictional heat of
the flowing water can only account for a portion Figs. 12 (0.01 and km 13. Oblique air photos show part of the 6 km
long, 2 km wide and 100 m deep depression formed above the
flowpath of the 1996 jökulhlaup across the ice dam of
Grímsvötn.
3 ) of the required conduit volume. Hence,
lifting of the ice by water pressure in excess of the overburden took place during the beginning of the
flood, prior to conduit formation. Longitudinal crevasses were observed above the entire flowpath.
Approaching the glacier terminus, basal water burst out on the glacier surface through several
hundred metres of ice (Fig 9). Icebergs were broken off the margin and spread over the outwash plain.
3.1.4 Modelling jökulhlaup discharges
There has neither been a theory nor sufficient empirical data yet available for describing how
jökulhlaups drain subglacially at a faster rate than can be explained by the expansion of conduits
through melting. This quicker drainage involves the ice overburden being exceeded by water pressure,
which lifts the glacier and allows space for the water. Initially propagated as a pressure wave, the
water then moves forward in a sheet flow, prior to draining through conduits [12,17,18,32].
The 1996 flood propagation mechanism was fundamentally different than that of previously observed
floods from Grímsvötn and has been successfully described by classical jökulhlaup theory. Flowers et
al. [20, submitted] have advanced a new model whereby floodwater initially propagates in a turbulent
subglacial sheet, which feed a nascent system of conduits. The model combines two interactive
components: a turbulent subglacial water sheet and a system of conduits. This model suggests that
unusually rapid conduit growth was facilitated by the distribution of source water along the length of
the flood path (Fig. 14). This model is able to explain the key observations made of the 1996
jökulhlaup and provides a hopeful starting point for understanding other floods that have eluded
classical jökulhlaup theory.
38 Deliverable 4
100

3.1.5 Newest research data on Grímsvötn
Fig. 14 Schematic diagram
explaining the model combining two
interactive components: a turbulent
subglacial water sheet and a system
of conduits.
Fig. 15 Discharge from, Grímsvötn
QL computed from lake level
measurements and known
hypsometry, shown with simulated
outlet discharge Qout. Qout is the sum
of sheet ( wsQ s , dashed line) and
conduit (wsQ c /dc, dotted line)
discharges at the glacier terminus.
Although not visible on this scale,
Qout increases from its background
value at ~10.5 h as required.
Vertical dashes indicate dusk and
dawn, between which peak outlet
discharge occurred [20].
The present status in lake Grímsvötn (November 2003) differs significantly from that before 1996.
The ice dam has recovered after having been damaged during the jökulhlaup in November 1996, but is
35 m lower than before. In spite of the restored ice dam the lake level has not risen high enough to
lead to significant jökulhlaups. The reason is that increased geothermal activity along Grímsfjall
mountain since the 1998 eruption has created a drainage channel along the slopes of the mountain,
where the lake water escapes continously (Fig.17). Thus the lake water level has never reached more
than 1410 m a.s.l. during the past 5 years, allowing only volume of less than 0.6 km 3 to be
accumulated in the lake. Four small jökulhlaups (with lake level changes of 40-50 m, 0.3 – 0.6 km 3 ,
150–500 m 3 s -1 ) have occurred since 1999, the last one in October 2003 when the ice-shelf floating on
Grímsvötn lowered about 20 m (Fig. 16). Furthermore mineature jökulhlaups, with less than 10 m
lowering of the ice-shelf have ocurred, for example in December 2001. On the other hand eruptions
north of Grímsvötn producing meltwater draining to the lake could cause larger jökulhlaups.
Systematic measurements of the lake level of Grímsvötn have been collected once or twice a year
since about 1950, using precison barometric altimetry and standard geodaetic survey methods. In 1990
a system to continuously monitor the lake level with precision barometric altimetry was set up. A data
logger recorded the air temperature and barometric pressure (hourly values) at a mast on the floating
iceshelf and in a hut on Grímsfjall, at a known elevation (1723 m a.s.l.). The system monitored the
start of the 1991 jökulhlaup successfully and operated only for year and a half due to technical
problems due to the harsh conditions. At the end of the 1996 Gjálp eruption a similar system (with
upto date technology) was set up on ice shelf. Occasional submeter differential GPS (accuracy ca. 1m)
measurements were also done. The drainage of out of Grímsvötn was recorded during the 1996
jökulhlaup (Fig. 8).
Deliverable 4 39

In June 1998 a pressure transducer (swinging wire transducer, accuracy of decimeter scale) was placed
on the lake floor connected to a cable through a 300 m deep hole drilled by a hot water drill through
the ice shelf (together with an 150 m extra length of the cable to allow for the ice-shelf lifting). A data
logger recorded the water pressure, water temperature, air temperature and barometric air pressure. A
radio modem link was set up between the lake datalogger and a similar datalogger system run in the
hut on Grímsfjall where an NMT cellular telephone was also installed (a direct radio connection to a
station on the ice shelf in Grímsvötn is impossible except via a satilite link). Data was read from both
data loggers via the NMT telephone and a standard modem.
This system ran successfully until destroyed by lightning and mudflow in the Grímsvötn volcanic
eruption in December 1998. Two other water pressure transducers were also installed in 1998 in the
flowpath of jökulhlaups from Grímsvötn (at the glacier bed of 350 and 450 m thick ice on the ice
dam). These recorded the basal water pressure for about 2 years before they broke (probable the
cables were damaged due to ice deformation). In June 1999 a new monitoring system of the same type
was set up in Grímsvötn. The whole system functioned perfectly until November 2000, when the
water pressure sensor was crushed between the iceshelf and the lake floor, - when due to almost
continuous drainage the icecover on Grímsvötn did not float anymore. Since then data to calculate the
lake level with precision barometric altmetry has however been collected successfully and DGPS
surveys of the lake elevation are done every time the station is vistited. In June 2002 a standard GPS
was added to the system and the elevation is logged every 10 minutes. This yields an accuracy of
about 2 m after some filtering.
At present the Grímsvötn lake contains less than 1 km 3 of water and during jökulhlaups only a peak
discharge of less than 1,500 m 3 /s are expected; not endangering people or constructions. The radio link
to Grímsfjall is not run, but data read from the monitoring station a few times a year. Changes in
Grímsvötn and vicinity are monitored by visual inspection, ice penetrating radar, DGPS, remote
sensing and InSar and seismometers. If the condition change a monitoring system with radio
communication will be set up immediately based on our experience of the past few years.
Fig. 16 The lake level of Grímsvötn 1999-2003 (in m. a.s.l), showing three small jökulhlaups and several minor
drainage events.
40 Deliverable 4

3.2. The Skatfá ice cauldrons and jökulhlaups
The two Skaftá cauldrons are situated over
geothermal systems, 10-15 km to the
northwest of Grímsvötn (Figs. 2 and 18).
Since 1955, at least thirty jökulhlaups have
drained from these cauldrons to the river
Skaftá. The period between these drainage
events is about 2 to 3 years. Jökulhlaup
discharge from the eastern Skaftá Cauldron
typically rises rapidly and recedes slowly [6].
The form of the hydrograph is a mirror
image of the typical Grímsvötn hydrograph
(Fig. 6). The peak discharge from the eastern
cauldron is 1,000-1,500 m 3 s -1 and is reached
in 1-3 days. These jökulhlaups recede slowly
in 1-2 weeks and the total volume of water
expelled is 200-350x10 6 m 3 , so far
Theoretical simulations fail to describe these
hydrographs [11].
Fig. 17 Two maps showing the different
location of the runoff-path of the meltwater
from lake Grímsvötn in the years 1997 and
2002.
Increased geothermal activity along
Grímsfjall mountain since the 1998
eruption created a channel along the
slopesof the mountain, where the
meltwater escapes continiously.
Fig. 18 The 150 m deep and 3 km eastern Skaftá
cauldron, just after a jökulhlaup in January 1982. The
crevasse trending from the upper left of the cauldron
suggests lifting of the ice dam. Further cauldrons are
seen in the background.
Deliverable 4 41

The steep rise in discharge can be
simulated with lake water well above the
melting point (10-20 °C) but these values
are derived from a flawed theory and
should not be taken seriously. Water at
the melting point drains from the glacier.
The rapid Skaftá jökulhlaups fall in the
category that is not explained by the
classical theory of jökulhlaups. Crevasses
observed across the ice dam of the eastern
cauldron after jökulhlaups suggest that
the jökulhlaups start with flotation of the
glacier (Fig. 10). At the time of the peak
discharge only 25% of the flood volume
has drained out of the lake. The slow
recession after the peak suggests that
these floods do not drain through a single
tunnel but spread out beneath the glacier
and later slowly collect into the river
outlet.
Table 2 Data on the subglacial Eastern Skaftá
Cauldron and typical jökulhlaups from it (The
basal melting rate may be underestimated as it
is calculated from the drained meltwater
during jökulhlaups and does not include
meltwater drained into groundwater). [9,15].
Lake: Eastern Skaftá Cauldron
Lake Type: Subglacial
Lat: 64°´ N
Lon: 17°´ W
Area:

negligible. However, jökulhlaups may triggere eruptions by the pressure release on top of the volcano
subsequent to the lake level falling about 80-100 m [60,26]. Seismic records suggest that this
mechanism may have triggered several small but invisible eruptions under the Skaftá cauldrons after
1985. Pressure release during a jökulhlaup may also cause explosive boiling in a subglacial
hydrothermal area.
The largest and most catastrophic jökulhlaups in Iceland may be caused by eruptions in the
voluminous, ice-filled calderas of Bárdarbunga and Kverkfjöll in northern Vatnajökull [9,14]. These
calderas may be the source of prehistoric jökulhlaups in Jökulsá á Fjöllum, with estimated peak
discharges of up to 4x10 5 to 1x10 6 m 3 s -1 [62,35] that swept down Jökulsá á Fjöllum and carved a
scablands and deep canyons (Jökulsárgljúfur).
Jökulhlaups during eruptions in steep ice and snow-covered stratovolcanoes are swift and dangerous
and may become lahars and debris-laden floods. The first contemporary description of a jökulhlaup
during a volcanic eruption in Iceland dates from the 1362 eruption of the ice capped stratovolcano
Öræfajökull [58]. The flood was over in less than one day and the peak may have been greater than
10 5 m 3 s -1 [60, p.36]. The meltwater originated in the 500 m deep caldera summit area at 2,000 m
elevation and drained beneath the outlet glaciers. Sediment, hummocks (2-4 m high), and dead ice
were spread over the lowland and into the sea, and a plain was formed where there had been thirty
fathoms water (Skálholt Annal). Several farmsteads were washed away. A contemporary record
described “several floods of water that gushed out, the last of which was the greatest. When these
floods were over the glacier itself slid forwards over the plane ground, just like melted metal poured
out of a crucible…. The water now rushed down on the earth side without intermission, and destroyed
what little of the pasture grounds remained…. Things now assumed quite a different appearance. The
glacier itself burst and many icebergs were run down quite to the sea, but the thickest remained on the
plain at a short distance from the foot of the mountain…we could only proceed with the utmost
danger, as there was no other way except between the ice-mountain and the glacier that had slid
forwards over the plain, where the water was so hot that the horses almost got unmanageable” [29, 58-
p. 31]. Obviously, the water conduits could not adjust to this rapid injection of meltwater, and
increased water pressure reduced basal friction and facilitated sliding. A less devastating eruption took
place in 1727 A.D. [58]. The sediment and icebergs that were transported down to the lowland were
later named glaciers and these placenames are still used centuries after all the ice has melted away.
Similar rapid floods carrying volcanic material were produced from Eyjafjallajökull (in 1612 and
1821-23), and during the eruption at Mt. Hekla in 1845, 1947 [33], 1970, and 1981 in the river Rangá.
4. Subglacial lakes, volcanic eruptions and jökulhlaups in Mýrdalsjökull
Mýrdalsjökull can be divided into three major catchment basins that supply water to the outwash
plains Mýrdalssandur, Sólheimasandur and Markarfljótsaurar, respectively (Fig. 19). The glacier
surface has 12 small depressions that have been created by subglacial hydrothermal activity. These ice
cauldrons are typically 20 to 50 m deep and 500 to 1000 m wide. A persistent hydrogen sulphide
odour from Jökulsá á Sólheimasandi indicates that meltwater is drained continuously. Meltwater
accumulates beneath some of the cauldrons and frequently drains in small jökulhlaups [16].
Katla eruptions rapidly melt large volumes of ice, triggering enormous jökulhlaups and breaking off
large blocks of ice from the glacier margin. Meltwater may also drain out supraglacially through
crevasses and moulins. During the jökulhlaups, a mixture of water, ice, volcanic products and
sediment, surges over the outwash plain. Water-transported volcanic debris has been estimated from
0.7 to 1.6 km 3 or on the order of 2x10 9 tons per event [64,36].
Deliverable 4 43

Fig. 19 Mýrdalsjökull. Cauldrons produced by
subglacial hydrothermal activity.
Vatnajökull Mýrdalsjöku
ll
Period (year) 5 - 30 40 - 80
Subglacial
flowpath (km)
Peak discharge
(m 3 /s)
Duration
(days)
Volume of
meltwater
(km 3 )
Sediment load
(10 6 ton)
Peak thermal
output (W)
Total thermal
energy (J)
50 20
5x10 3 - 1x10 6
1x10 5
3x10 5
2 - 30 2 - 10
3 - 5 1 - 8
100 - 300 2000
1.7x10 12
10 18
3x10 13 - 10 14
10 18 - 10 19
Table 3. Data on typical jökulhlaups triggered by
volcanic eruptions. Thermal output of subglacial
volcanic eruptions derived from production of
meltwater [9,15]
During the most violent jökulhlaups such large quantities of icebergs and sediment were clustered
together that some places on the outwash plain were named for glaciers. The peak discharge of 1-
3x10 5 m 3 s -1 is reached in a few hours, and total volumes of 1-8 km 3 draining over 3-5 days have been
suggested [61,64,36]. The rate of increase of discharge and peak discharges are an order of magnitude
higher than for any known jökulhlaup from subglacial lakes. These jökulhlaups, along with heavy
fallout of tephra, make Mýrdalsjökull the most hazardous volcano in Iceland.
Fascinatingly, the Katla eruptions break through an ice cover of 400 m in one or two hours.
Nontheless, the melting rate during the most violent Katla eruptions is an order of magnitude higher
than during the Gjálp eruption in 1996 (Table 3), [27]. The heat flux of Katla may be overestimated
because the peak jökulhlaup discharges are higher than the production rate of meltwater. Water may
be stored at the eruption site, either at the beginning of the eruption because the conduits cannot
accommodate the flow, or over a longer period due to increased heat flux from the volcano and
subsequent formation of a depression in the surface. Eruptions of magma into such a subglacial water
body would make possible rapid heat transfer during the eruption and ensure rapid cooling by
fragmentation of the lava into glass.
Mýrdalssandur and the adjacent Sólheimasandur and Skógasandur have to a large extent been built in
floods caused by volcanic eruptions of the Katla volcano. During 18 of the 20 documented eruptions
the associated jökulhlaups flowed southeast down to the Mýrdalssandur outwash plain, but in two
cases jökulhlaups flowed southwest to the Sólheimasandur outwash plain. During the last eruption in
1918, the jökulhlaup moved the coastline seawards by 3 km [64], and the coastline now lies 2.2 - 2.5
km further south than it did in 1660. Marine sediments found several hundred km south of Iceland,
contain debris transported from the eruption sites [65,53]. The most rapid jökulhlaups have produced
flood waves in coastal waters. Jökulhlaups from the northern part of Mýrdalsjökull have eroded
canyons in Markarfljót River [52]. The most recent jökulhlaup taking that route was 1600 years ago
[28,29].
44 Deliverable 4
-

5 Conclusions
Of our main findings of the glacier volcano interactions in Iceland we can summarise the following
conclusions.
Volcanic eruptions to the glacier bed are the most violent expressions of glaciovolcanism, rapidly
melting ice and producing hazardous floods outside the glacier. Subglacial hydrothermal systems are
quieter expressions, generated by interaction of glacial meltwater with magma intrusions in the Earth’s
crust. The hydrothermal activity continuously melts ice at the glacier bed, creating a depression in the
glacier surface. The meltwater drains either continuously to the glacier margin or accumulates in
subglacial lakes that are situated beneath the glacier surface depressions. Some of the lakes are located
in caldera depressions but they rise as a dome above the caldera rims due to the relatively low fluid
pressure under the glacier depression. The release of water from the subglacial lakes can take place by
two different conduit initiation mechanisms. Drainage begins either through conduits at pressures
lower than the ice overburden at the ice dam (in Grímsvötn, typically 6-7 bar lower) or the lake level
rises until the ice dam is floated. The subsequent drainage from the lake can take place by two
different modes. The conduits may enlarge over days or weeks due to melting of the ice walls by
frictional heat in the flowing water and stored lake heat. This is typical for the jökulhlaups initiated at
lake levels lower than that required for flotation. These processs have been explained by classical
jökulhlaup theories. Alternatively, the discharge rises faster than can be accommodated by melting of
the conduits and the glacier is lifted along the flowpath to make space for the water. A preliminary
modeling study suggests that the idea of a sheet flood preceding conduit drainage is plausible.
During rapidly rising jökulhlaups, water may drain through braided watercourses at high pessures.
However, these passageways quickly develop into high-capacity ice tunnels, and jökulhlaups are not
known to have led to surges of glaciers. Icebergs may be broken off the glacier margins and
transported by water over the river courses. During eruptions in steep ice-capped stratovolcanoes (like
Öræfajökull) distributed drainage of meltwater has led to sliding of the glacier down to the lowland.
Processes and products of glacier volcano interaction are displayed on a broad rage of dimensions.
Hydrothermal systems affect drainage areas of 5 to 200 km 2 and the area of the subglacial lakes is
from one to 40 km 2 in area. The production rate of basal glacial meltwater spans over four to five
orders of magnitude, from 2-6 m 3 s -1 in hydrothermal systems to 5x10 3 to 10 5 m 3 s -1 in volcanic
eruptions (requiring a heat flux of 10 3 to 10 7 MW). The lakes vary in volume by three orders of
magnitude (lakes beneath cauldrons of 2x10 9 m 3 to 4x10 12 m 3 in Grímsvötn). The lakes drain
periodically with an interval of 1 to 10 years. The duartion of the jökulhlaups may be from 2-3 days to
2-3 weeks. The peak discharge in jökulhlaups can be from 200 to 10 6 m 3 s -1 . Ordinary jökulhlaups
from subglacial lakes may transport of the order of 10 7 tons of sediment but during the most violent
volcanic eruptions the sediment load has been 10 8 tons.
II. Marginal lakes and jökulhlaups in Iceland.
1 Introduction
The meltwater from outlet glaciers may drain into lakes at the glacier margins, those in front of the ice
are called proglacial, but those situated at the sides, lateral lakes. Sudden floods occur from lateral
lakes when the water reaches a critical level. The drainage typically starts before the water pressure
maintained by the lake becomes equal to the ice overburden at the ice dam. At present, jökulhlaups
originate from some fifteen marginal ice-dammed lakes in Iceland, most of them located at the outlet
glaciers from Vatnajökull. Typical values for peak discharges are 1,000 - 3,000 m 3 /s, duration 2-5 days
and total volumes of 2,000 x 10 6 m 3 .
Deliverable 4 45

2 Marginal lakes at Vatnajökull
At present, about 14 marginal lakes are situated at the margin of the ice cap Vatnajökull (Fig. 20)
[56,5]. Most of them drain regularily in jökulhlaups, which have become more frequently but less
voluminous during the last decades. Some lakes, as Gjávatn for instance have totally disappeared.
The thinning of glaciers, which began around the turn of the last century, initiated jökulhlaups from
marginal lakes, which during the 19th century had drained continuously over a col as for instance at
Grænalón and Vatnsdalslón (Lakes 1 and 7, Fig. 20). These floods caused much damage to farms and
fertile land. They were unexpected and drained lakes which were filled to their maximum capacity.
During the later glacier recession jökulhlaups have become more frequent (occurring once, even twice
a year) and gradually smaller in volume due to the thinning of the ice dams. This trend has been
manifested in successive lowering of shore lines in marginal lakes. Also, the number of dumping
glacier lakes has been greatly reduced. Now, the damage done by them is largely limited to roads and
bridges. In a few instances, this trend has been interrupted by the thickening of ice dams during surges
and temporary formation of new ice-dammed lakes (e.g. Hamarslón, lake 14, Fig. 20, after surges of
the outlet Köldukvíslarjökull in NW Vatnajökull, draining to Kaldakvísl,). The advance of some steep
and active glacier outlets during a cold spell in the 1970´s has dammed ravines at the glacier margin
which have dumped small jökulhlaups, e.g. by Sólheimajökull, a southfacing outlet of Mýrdalsjökull
(Fig.19).
Grænalón
1898 – 1940 after 1950
Vatnsdalslón
1898 1974
Maximum lake area (km 2 ) 22.5 16 1.9 1
Fig. 20 The location of 14
marginal, ice-dammed lakes
situated at the margin of the ice cap
Vatnajökull, most of which
regulerily drain into outburst
floods.
1. Grænalón / 2. Langagilslón / 3.
Nordurdalslón / 4. Unnamed / 5.
Breidamerkurfjallslón / 6.
Vedurárdalslón / 7. Vatnsdalslón /
8. Unnamed / 9. Gjávatn / 10.
Hnútulón / 11. unnamed / 12.
Thorbergsvatn / 13. Hvítalón / 14.
Hamarslón
Gjánúpsvatn
1951
Flood interval (years) 0.5 – 1 0.5 - 1 0.5 - 1
Duration (days) 7 7 1 6 7
Peak discharge (10 m 3 s 1 ) 4 – 5 1.5 – 2 3 0.7 0.37
Volume (10 9 m 3 ) 1.5 – 2 0.2 – 0.5 120 x 10 -3
37 x 10 -3
Initial lake level elevation 640 590 464 350 227
Final lake level elevation 450 560 350 300 200
Length of drainage tunnel 20 20 7 7 4.5
Glacier-surface elevation at seal 700 ~620 370 370
Glacier–bed elevation at seal ~580 230 230
Distance from lake to seal (km) 2.5 0 1 1
Elevation of tunnel outlet 90 90 90 100 60
Lake geometry [10]
20 x 10 -3
Flood Rivers Súla Kólgríma Hornafjardarfljót
References [56] [43,45,46,47] [56] [43,45,46,47] [1]
Table 4. Selected marginal jökulhlaups from ice-marginal lakes in Iceland and values for their
jökulhlaups (all elevations given in m above sea level).
46 Deliverable 4

In general, the impact of the present day jökulhlaups from ice dammed marginal lakes is small
compared with those, which occurred at the end of the last glacial and eroded several large canyons in
Iceland (Tómasson, 1973, 1991). Table 4 gives some values on the geometry of three marginal ice
dammed lakes and typical values for the jökulhlaups, which originate from them. The two columns for
the lakes Grænalón and Vatnsdalslón show changes in the size of jökulhlaups in this century.
Gjánúpsvatn disappeared some decades ago, because of the retreat of the outlet glacier. Hypsometric
data for Grænalón were published by Björnsson and Pálsson [10].
Hydrographs for jökulhlaups from marginal lakes have a shape similar to those of the typical
Grímsvötn jökulhlaup (Fig 6a). Simulations give reasonable ascent of the hydrographs for constant
lake temperature of about 1°C but fail to show the recession. Some floods from marginal lakes,
however, have reached their peaks exceptionally rapidly, in one day. That ascent could be simulated
by drainage of lake water of 4-8°C.
2.1 Grænalón
Grænalón is the largest marginal ice dammed
lake in Iceland, situ situated west of the
outlet glacier Skeidarájökull and dammed by
part of it. Its area has been up to 18 km 2 and
floods have reached peak discharges of 5,000
m 3 s -1 . During the 20th century the ice dam
has become thinner, and the lake and the
jökulhlaups been reduced to typical peak
discharge of 2,000 m 3 /s. The last jökulhlaup
occurred in September 2003.
During most of the 19 th century the outlet
glacier Skeidarárjökll was so thick that the
highest possible lake level did not cause the
ice to lift, hence there were no floods.
Instead there was a steady outflow out of the
lake over a ridge (ca. 650 m a.s.l) and into
river Núpsá.
As the glacier became thinner the water
pressure became sufficient to lift the ice dam
of the outlet glacier and drain beneath and
trigger a flood. The largest ones occurred in
1935 and 1939 with numerous icebergs
reaching the sandur several km south of
Skeidarárjökull [60].
Lake: Grænalón
Lake Type: Ice-marginal, lateral
Lat: 64°11´ N
Lon: 17°09´ W
Area: 5 - 16 km 2
Lake level elevation: 550 - 580 m
Max. Depth:

2.1.1 New research data on Grænalón.
In the latest years, jökulhlaups from Grænalón
occurred at least every year. This period of
abnormally frequent jökulhlaups started with one,
quiet large flood, where the water level lowered
by more than 20m. Then, from September 2001
until September 2003 at least 10 small
jökulhlaups occured, which were hardly
noticeable by discharge increase only. The last
jökulhlaup (September 2003) was said to be the
largest one since 1986. It drained from beneath
the glacier and flooded river Súla. Part of the
water found its path over a low ridge and into
river Sandgígjukvísl. There was no damage done
to dams or dikes nor bridges or roads.
Fig. 21 The marginal lakes 1 to 4, all situated at the
outlet glacier Skeidarárjökull and the area that could
be affected by a large outburst flood.
ACKNOWLEDGEMENT
The work has been supported by the Public Roads Administration, The National Power Company of
Iceland, and by the European Union (Framework V – Energy, Environment and Sustainable
Development, contract EGV1-2000-00512, GLACIORISK).
REFERENCES
1. Arnborg, L. 1955. Hydrology of the glacial river Austurfljót. Geografiska Annaler, Årg. 37, Ht. 3-4, 185-201.
2. Bemmelen, R. W., van Rutten, M. G., 1955. Tablemountains of Northern Iceland. E. J. Brill, Leiden. 217 pp.
3. Björnsson, H. 1974. Explanation of jökulhlaups from Grímsvötn, Vatnajökull, Iceland. Jökull 24, 1-26.
4. Björnsson, H. 1975. Subglacial water reservoirs, jökulhlaups and volcanic eruptions. Jökull 25,1-14.
5. Björnsson, H. 1976. Marginal and supraglacial lakes in Iceland. Jökull, 26, 40-51.
6. Björnsson, H. 1977. The cause of jökulhlaups in the Skaftá river, Vatnajökull. Jökull 27, 71-78.
7. Björnsson, H. 1982. Drainage basins on Vatnajökull mapped by radio echo soundings. Nordic Hydrology, 1982, p. 213-
232.
8. Björnsson, H. 1986: Surface and bedrock topography of ice caps in Iceland mapped by radio echo soundings. Annals of
Glaciology 8, 11-18.
9. Björnsson, H. 1988. Hydrology of ice caps in volcanic regions. Soc. Sci. Isl., 45, Reykjavík. 139 pp.
10. Björnsson, H. and F. Pálsson. 1989. Rúmmál Grænalóns og breytingar á stærd og tídni jökulhlaupa. Jökull, 39, 90-95.
11. Björnsson, H. 1992. Jökulhlaups in Iceland: prediction, characteristics and simulation. Annals of Glaciology 16, 95-
106.
12. Björnsson, H. 1997. Grímsvatnahlaup fyrr og nú. In: Vatanjökull. Gos og hlaup 1996 (editor Haraldsson, H.), 61-77.
Icelandic Public Roads Administration, Reykjavík, Iceland.
13. Björnsson, H. 1998. Hydrological characteristic of the drainage system beneath a surging glacier. Nature 395, 771-774.
14. Björnsson, H., Einarsson, P. 1991. Volcanoes beneath Vatnajökull, Iceland: Evidence from radio echo-sounding,
earthquakes and jökulhlaups. Jökull 40, 147-168.
15. Björnsson, H., Guðmundsson, M. T., 1993. Variations in the thermal output of the subglacial Grímsvötn Caldera,
Iceland. Geophysical Research Letters 20, 2127-2130.
16. Björnsson. H., Pálsson, F., Gudmunsdsson, M. T. 2000. Surface and bedrock topography of the Mýrdalsjökull ice cap,
Iceland: The Katla caldera, eruption sites and routes of jökulhlaups. Jökull 49, 29-46.
17. Björnsson, H., Rott, H., Gudmundsson, S., Fischer, A., Siegel, A., Gudmundsson, M. T. 2001. Glacier-volcano
interactions deduced by SAR interferometry. Journal of Glaciology, Vol. 47, No. 156, 58-70.
18. Björnsson, H. 2002. Subglacial lakes and jökulhlaups in Iceland. Global Planet. Change, 35, 225-271
48 Deliverable 4

19. Einarsson, P. Brandsdóttir, B., Gudmundsson, M. T., Björnsson, H., Sigmundsson, F., Grönvold, K. 1997. Center of the
Iceland hotspot experiences volcanic unrest. Eos 78, No. 35, 374-375.
20. Flowers, G. E., Björnsson, H. and Pálsson, F. 2003. New insights into the subglacial and periglacial hydrology of
Vatnajökull, Iceland, from a distributed phisical model. Journal of Glaciology, in press.
21. Clague, J.J. and Mathews, W.H. 1973. The magnitude of jökulhlaups. J. Glaciology 12 (66), 501-504.
22. Clarke, G. K.C. 1982.Glacier outburst flood from "Hazard Lake", Yukon Territory, and the problem of flood magnitude
prediction. Journal of Glaciology, 28, 3-21.
23. Gudmundsson, M.T. 1989. The Grímsvötn Caldera, Iceland, subglacial topography and structure of caldera infill.
Jökull 39, 1-19.
24. Gudmundsson, M.T. 1992. The crustal structure of the subglacial Grímsvötn volcano, Vatnajökull, Iceland, from
multiparameter geophysical surveys. (Ph.D. thesis, University of London).
25. Gudmundsson, M.T., Björnsson, H.. 1991. Eruptions in Grímsvötn, Vatnajökull, Iceland, 1934-1991. Jökull 41, 21-
45.
26. Guðmundsson, M. T., Björnsson, H., Pálsson, F. 1995. Changes in jökulhlaup sizes in Grímsvötn, Vatnajökull, Iceland,
1934-1991, deduced from in situ measurements of subglacial lake volume. Journal of Glaciology, Vol. 41, No. 138, 263-
272.
27. Guðmundsson, M T., Sigmundsson, F., Björnsson, H. 1997. Ice- volcano interaction of the 1996 Gjálp subglacial
eruption, Vatnajökull, Iceland. Nature 389, 954-957.
28. Haraldsson, H., 1981. The Markarfljót sandur area, southern Iceland. Sedimentological, pertographical and
stratigraphical studies. Striae, Vol. 15, 1-65.
29. Henderson, E. 1818. Iceland; or the Journal of a Residence in that Island during the Years 1814 and 1815, Vol. I.
Oliphant, Waugh and Innes, Edinburgh. 412 pp.
30. Jakobsson, S. P., 1979. Petrology of recent basalts of the Eastern Volcanic Zone, Iceland. Acta Naturalia Islandica 26,
103 pp.
31. Jóhannesson, H., Sæmundsson, K., Jakobsson, S. P. 1990. Geological map of Iceland, sheet 6, South-Iceland. Icelandic
Museum of Natural History and Iceland Geodetic Survey, Reykjavík.
32. Jóhannesson, T. Propagation of a subgacial flood wave during the initiation of a jökulhlaup. 2002. Hydrolog. Sci. J., 47,
417-434.
33. Kjartansson, G., 1943. Yfirlit og Jarðsaga. In: Náttúrulýsing Árnessýslu I. (Ed. G. Jónsson). Árnesingafélagið í
Reykjavík. Reykjavík. pp. 1-249.
34. Kjartansson, G. 1967. The Steinholtshlaup, Central-South Iceland on January 15th, 1967. Jökull, 17, 249-262.
35. Knudsen, Ó., Russell, A. R. 2002. Jökulhlaup deposits at Ásbyrgi, northern Iceland: sedimentology and implications or
flow type. In: The Extremes of the Extremes. Special Publication of the International Association of Hydrology. Publ. no.
271.
36. Larsen, G., 2000. Holocene eruptions on the Katla volcanic system, Iceland: Notes on characteristics and environmental
impact. Jökull 50, 1-28.
37. Larsen, G., Guðmundsson, M. T., Björnsson, H. 1998. Eight centuries of periodic volcanism at the center of the Iceland
hot spot revealed by glacier tephrastratigraphy. Geology, Vol. 26, No. 10, 943-946.
38. Mathews, W. H., 1947: “Tuyas”. Flat-topped volcanoes in northern British Columbia. American Journal of Science
245, 560-570.
39. Nye, J. F. 1976. Water flow in glaciers: jökulhlaups, tunnels and veins. Journal of Glaciology, Vol. 17, No. 76, 181-
207.
40. Rist, S. 1955. Skeidarárhlaup. Jökull, 20, 89-90.
41. Rist, S. 1967. Jökulhlaups from the Ice Cover of Mýrdalsjökull on June 25, 1955 and January 20, 1956. Jökull, 17,
243-248.
42. Rist, Sigurjón. 1970. Annáll um jökulhlaup. Jökull 20, 89 – 90
43. Rist, S. 1973. Jökulhlaupaannáll 1971, 1972 og 1973. Jökull, 23, 55-60.
44. Rist, S. 1976a. Grímsvatnahlaupid 1976. Jökull, 26, 80-90.
45. Rist, S. 1976b. Jökulhlaupaannáll 1974, 1975 og 1976. Jökull, 26, 75-79.
46. Rist, S. 1981. Jökulhlaupaannáll 1977, 1978 og 1980. Jökull, 30, 31-35.
47. Rist, S. 1984. Jökulhlaupaannáll 1981, 1982 og 1983. Jökull, 34, 165-179.
48. Roberts, M. J., Russel, A. J., Tweed, F. S., Knudsen, Ó. 2000. Ice fracturing during jökulhlaups: implications fror
englacial floodwater routing and outlet development. Earth Surface Processes and Landforms 25, 1-18.
49. Russel, A. J., Knudsen, Ó, Maizels, J. K., Marren, P. M. 1999. Channels cross-sectional area change and peak discharge
calculations in the Gígjukvísl river during the November 1996 jökulhlaup, Skeiðarársandur, Iceland. Jökull 47, 45-70.
50. Russel, A. J., Tweed, F. S., Knudsen, Ó. 2000. Flash flood at Sólheimajökull heralds the reawakening of an Icelandic
subglacial volcano. Geology Today 16, 102-106.
51. Shreve, R., 1972. Movement of water in glaciers. Journal of Glaciology 62, 205-214.
52. Sigurðsson, F., 1988. Fold og vötn að Fjallabaki. Árbók Ferðafélags Íslands 1988, pp.181-202.
53. Sigurdsson, H., Lacasse, C. and Jóhannesson, H. 1995. Katla seilist suður í Atlantshaf. Eyjar í Eldhafi. Gott mál.
Reykjavík, pp. 101-108.
54. Snorrason, A., Jónsson, P., Pálsson, S., Árnason, S., Sigurdsson, O, Víkingsson, S., Sigurdsson, Á., Zópóníasson, S.
1997. Hlaupið á Skeiðarársandi haustið 1996. Útbreiðsla, rennsli og aurburður. In: Vatnajökull. Gos og hlaup 1996 (ed.
Haraldsson, H.), 79-137. Icelandic Public Roads Administration, Reykjavík.
55. Spring, U., Hutter, K. 1981. Numerical studies of jökulhlaups. Cold Regions Science and Technology 4 (3), 221-244.
56. Thorarinsson, S. 1939. The ice dammed lakes of Iceland with particular reference to their values as indicators of glacier
oscillations. Geografiska Annaler, 21, 216-242.
Deliverable 4 49

57. Thorarinsson, S. 1957. The jökulhlaup from the Katla area in 1955 compared with other jökulhlaups in Iceland. Jökull
7, 21-25.
58. Thorarinsson, S. 1958. Acta Naturalia Islandica.
59. Thorarinsson, S. 1965. Changes of the water-firn level in the Grímsvötn caldera 1954-1965. Jökull 15, 109-119.
60. Thorarinsson, S. 1974. Vötnin Stríd. Saga Skeidarárhlaupa og Grímsvatnagosa [The swift flowing rivers. The history
of Grímsvötn jökulhlaups and eruptions]. Menningarsjódur, Reykjavík. 254 pp.
61. Thorarinsson, S. 1974. Katla og annáll Kötlugosa. (Katla and annals of Katla eruptions). Árbók Ferðafélags Íslands,
Reykjavík. pp. 125-149.
62. Tómasson, H., 1973. Hamfarahlaup í Jökulsá á Fjöllum. Náttúrufræðingurinn 43: 12-34.
63. Tómasson, H. 1990. Glaciofluvial sediment transport and erosion. In: (eds.Y. Gjessing, J.O. Hagen, K. A. Hassel, K.
Sand and B. Wold) Arctic Hydrology. Present and future tasks. Hydrology of Svalbard - Hydrological problems in cold
climate. Norwegian National Commmittee for Hydrology. Report No. 23. Oslo. p. 27-36.
64. Tómasson, H. 1996. The jökulhlaup from Katla in 1918. Annals of Glaciology 22, 249-254.
65. Walker, G. P. L. 1972. Petrology of the Volcanic Hyaloclastic Sandstone. Initial Reports of the Deep Sea Drilling
Project, p.365-369. Volume Xii, Washington.
66. Wolfe, C. J., Bjarnason, I. Th., VanDecar, J. C., Solomon, S. C. 1997. Seismic structure of the Iceland mantle plume.
Nature 385, 245-247.
50 Deliverable 4

Stability of steep glaciers
Ice avalanche from Bockkarkees on 26th June 2003
Michael Krobath a and Heinz Slupetzky b
a Institute for Geography and Regional Science, University of Graz, Heinrichstraße 36, 8010 Graz, Austria
b Institute for Geography and Applied Geoinformatics, University of Salzburg, Hellbrunnerstraße 34, 5020
Salzburg, Austria
1. Introduction
The Bockkarkees – “Kees” is the local name for “glacier” in the Hohe Tauern range - is the most
interesting glacier for the project Glaciorisk Austria. It is situated on the northern side of the Hohe
Tauern range in the upper part of the Fuscher valley – the small Käfer valley - and belongs to the
Großglockner mountains. The border of the counties Carinthia and Salzburg is crossing the glacier in
the middle. Table 1 shows the basis data of the glacier.
Region Salzburg
Massif Großglockner mountains
Municipality Fusch
WGI Id A4J143SA051
lat (°, cent.) 44,58
lon (°, cent.) 7,12
orientation (°) 130
average slope (°) 22
minimal altitude (m) 2700
maximal altitude (m) 3360
Lenght (km) 4,1
large (km) 1,4
surface (km²) 3,4
Table 1. Main data of Bockkarkees
Figure 1 shows the position of the glacier north of “Pasterze” (Austrians biggest glacier) and
Großglockner (Austrians highest summit). The red road in the east is the “Großglockner
Hochalpenstraße”, one of Austrians most famous touristic attractions.
Figure 1. Position of Bockkarkees (in the grey rectangle)
Deliverable 4 51

In the north-east of the map we find a famous viewpoint “Edelweißspitze”. From this point the
Bockkarkees can be observed quite well. Figure 2 shows the view from this spot to the glacier before
and after an avalanche.
Figure 2. View from Edelweißspitze to the glacier (on the right) before an avalanche in 1980 and after an
avalanche with the deposit area in 1993
The glacier encloses an area of about 3 km² and is exposed to the east. About 70% of the glacier are
located between an altitude from 2900 to 3100 m. The small tongue reaches from 2900 down to 2750
m. In the first half of the 20 th century the tongue even reached down above a cliff to 2100 m.
2. Former ice avalanches
Since the melting off of the tongue over that cliff the glacier lost it´s support and ice avalanches had
the possibility to break away from the new tongue. Since that time about 50 bigger and smaller
avalanches have been documented and can reach a radius of action over an elevation of 1500 m
downhill. The last one happened on June the 26 th 2003 year with a volume of about 500.000 m³ .
The avalanches in the last decades had ice volumes between 10.000 and 5 million m³. This 5 million
m³ event was caused by the biggest documented ice avalanche in Austria that came down on 26 th July
1945. Unfortunately there are no fotos of this event respectively of the ice masses fallen down.
Nevertheless Figure 3 shows the reconstructed deposit area of this big avalanche in comparison to the
one of the last avalanche in June 2003.
Figure 3. Reconstructed deposit area from 1945 and deposit area from 26 th June 2003
52 Deliverable 4

The area covered by ice after the event in 1945 can also be seen as the area still endangered by the
Bockkarkees. The event in 1945 didn´t harm people but destroyed a building and killed about 70
cattles and sheeps which stayed on this pasture ground.
In the area affected by avalanches there was set up a building to collect the melting water of the glacier
to send it to the storage lake of the power station “Kaprun” in the neighboured valley through a tunnel
and of course the building and the workers who sometimes stay there are threatened by the ice
avalanches. Sometimes the deposit of ice after an avalanche also causes a damming up of the melting
water and the possible following outbreak of the lake can be powerful as well, so happened after the
event in 1945 when 10 more cattles where killed by such an outbreak. The most important aim is of
course to predict such events, which is done by terrestrial and areal observation. The ice avalanches
are preceded by a stage of sliding of the whole glacier tongue which can show a rate of one to two
meters a day (Figure 4).
Figure 4. Ice during sliding, chaotic crevasses are formed, the supporting ice masses
at the borders get lost ( 5 th October 1985)
During
that time the tongue is teared up in a chaotic way, as this picture shows, but is still supported
by the ice parts at the border of the glacier. When the ice pressure from the upper parts gets too strong
these supports break and the ice avalanche goes down as shown in Figure 5 when only half moon
shaped ice walls stay back.
Deliverable 4 53

Figure 5. The glacier after the big avalanche on 5 th October 1979. (17 th October 1979)
Table 2 shows all documented ice avalanches from Bockkarkees so far.
1933-1944 ?? smaller ice avalanches
1945 26th June 4 – 5 Mio. m³
1955 July some 10 000 m³
1956 15th and 19th August some 10 000 m³
1957 27/28th July ?
1958 28th August ?
1959 7th August 400 – 600.000 m³
1960 17th August, 1st October ?
1961 2nd,16th,28th July ?
3rd August ?
1962 20th August ?
9th October ?
1963 26th August ?
1964 21thJuly (?) 2-3 Mio m³
29/30th July ?
1965 ---------- ---------
1966 17th August ?
1967 21th August 1,5 – 2 Mio m³
1968 --------- -----------
1969 2nd September 0,7-1,2 Mio m³
12th October ?
1970 --------- ----------
1971 2nd July 2-3 Mio m³
15th July (some 100.000 m³?)
6./10th August (0,5 Mio m³?)
1972 31th July ???
1973 7 th August 0,7-1 Mio m³
1974 --------- ---------
54 Deliverable 4

1975 September 150.000 m³
7/8th October 1,1-1,8 Mio m³
1976 Before 24th August 0,5-1,0 Mio m³
1977 middle of august 200.000 m³
august/september 200.000 m³
1978 about 10th October 400.000 m³
1979 September 100.000 m³
30th September some 10.000 m³
5/6th October 2,6 Mio m³
1980 before 30th June some 10.000 m³
end July 100.000 m³
middle of September 200-300.000 m³
1981 ----------- -----------
1982 1st half of September 10-20.000 m³
autumn sliding
1983 15th September 500-600.000 m³ and 2,5-3,5
Mio m³
sliding
Between 17th and 20th 100-200.00 m³
September
1984 September Some 10.000 m³
1985 1st-3rd October ½-1 Mio (sliding)
1986 September/October Some 10.000 m³
September/October (sliding)
1987 5th October Some 10.000 m³
(sliding)
1988 16th-18th August Small avalanches
2nd half of August – 1st Sliding
half of September
1990 July Small avalanches
September Small avalanches
End October Small avalanches
1991 15th September Small avalanches
19 th September max. 0,5 Mio m³
23th September 200-300.000 m³
September - October Sliding
1992 30th September 1 Mio m³
1993 19. (/20.) July 1,5-2,5 Mio m³
27./28th July 200-300.000 m³
August/ September small avalanches
1994 August small avalanches
8th-11th September 50.000 m³
26th September some 1.000 m³
September sliding
1995 31th July 400-700.000 m³
4 th August 300-500.000 m³
8 th August ???
August/September sliding
4th September 200-300.000 m³
20th October 350-400.000 m³ sliding
Deliverable 4 55

3. Avalanche in 2003
1996 end Aug./Sept. (before
5.9.)
200-300.000 m³
6.-9th September 70-120.000 m³
14th October 50-80.000 m³
middle
October/November
sliding
18th October 70-100.000 m³
24th October 60.-80.000 m³
28. October 250-350.000 m³
2002 end June small avalanches
Before 5th August small avalanches
6th September 20 – 30 000 m³
2003 26th June 500.000 m³
Table 2. List of all documented ice avalanches from Bockkarkees so far.
The first working step after the avalanche on 26 th June 2003 was the mapping of the accumulated ice
masses by laser scanning and a repetition of the measurements afterwards the melting of the ice.
Figure 6 shows the accumulation area and the measurements on 4 th July 2003.
Figure 6. Deposit area of the avalanche from 26 th June 2003 and laser scanning on 4 th July 2003
The second step was the mapping of the whole glacier an the area affected by avalanches with
ArcView. The following coverages have been generated: terrain, glacier, water runs, wood, buildings
(huts for agricultural use and buildings for collecting water and streets). In this map all documented
ice avalanches of the last 60 years are documented.
The next aim is to generate a map of possible permafrost distribution near the ice front at the breaking
off zone (Figure 7).
56 Deliverable 4

Figure 7. Breaking off zone of Bockkarkees
The half moon shaped rupture zone with the about 50 m high ice wall was left back. Observation of
the contact zone between ice and rockbed should show whether the glacier is frozen to the ground in
this area, which would affect the kind of movement of the ice. The results so far are not very valid
now and can´t be published yet. The aim is a long time monitoring with temperature loggers, which is
a very cheap method and therefore can be carried on without any problems.
Quite interesting in the zone just beneath the rupture is the channel that was left back by the avalanche
(Figure 8).
Figure 8. Ice dam on the right and avalanche channel on the left
Obviously the high velocity of the ice masses in the middle of the stream didn´t leaf back lots of ice
but at the borders of the channel we find ice dams some meters high caused by the high frictional
resistance in this area between ice and rockbed.
Deliverable 4 57

Figure 9. Accumulation area just underneath the breaking off zone
There have obviously been more generations of ice avalanches (Figure 9), a dirty bigger one and one
or more smaller ones with clean ice. This fact has also been documented after avalanches in the past
that one big event is put together by some smaller ones.
Moreover there has been carried out a flight for taking areal pictures on 17 th August 2003. Results of
this working step will be available at the end of 2003.
References
1. G. Kettner. Die Eisstürze vom Nördlichen Bockkarkees und ihre Auswirkungen. Thesis, Univ. Salzburg, 1982.
2. H.Slupetzky "Der Eissturz vom Nördlichen Bockkarkees (Hohe Tauern, Glocknergruppe, Käfertal) im Jahr 1945",
Grazer Schriften der Geographie und Raumforschung Universität Graz, 38, pp 211-226, 2002.
3. H.Slupetzky, R.Puruckherr & Ch.Hoberg " Zur Karte "Nördliches Bockkarkees 1979", 1:10000", Zschr. f.
Gletscherkunde und Glazialgeologie, 19/2, pp 163-171, 1983.
4. K. Hofmannn & J.Stüdl "Wanderungen in der Glocknergruppe", Zschr. d. Dt. Alpenvereins, 2, pp 333-544, 1871.
5. L. Krasser. Gutachten über die Lawinengefährdung der Bau- und Betriebsstellen der Tauernkraftwerke A.G. im
Käfertal und im Mölltal vom 18.9.1959, Salzburg, 1959.
6. P. Ganahl. Hohe Tauern-Glocknergebiet-Bockkarkees-Käfertal. Abt. Hydrologie, Tauernkraftwerke AG, Salzburg,
1984.
7. S. Rohrbacher. Volumsänderungen von Gletschern an ausgewählten Beispielen. Thesis, Univ. Salzburg, 1983.
58 Deliverable 4

Snow and ice avalanches triggered by serac falls
Taconnaz ice-mass breaking off
C. Vincent et E. Le Meur,
Laboratoire de Glaciologie et de Géophysique de l’Environnement, CNRS, BP 96 38402 Saint Martin d’Hères
Cedex, France
1. General objectives
Within the framework of WP2, the LGGE is involved in a field study relative to the ice avalanche
hazard of Taconnaz glacier.
The Taconnaz study, concerned by serac avalanche, is a new approach which began in June 2001 in
the framework of this project. No similar study about this seracs fall had been conducted in the past
(except a one-day survey carried out in 2000), and the Glaciorisk programme appeared as a very good
opportunity to start this extensive research which is connected with a real risk down the valley of
Chamonix. Many field observations have been collected since June 2001 and are reported below.
N
Grenoble
Ecrins
Durance
50 km
Geneva
Switzerland
Chamonix
Vanoise
ARSINE
France
Mont-Blanc
Rhône
Swiss Alps
TACONNAZ
Nice
Turin
Italy
Mediterranean
Figure 1: French Alps. Locations of the 2 glaciers studied in the framework of the Glaciorisk project.
Taconnaz glacier is located in the Mont Blanc range. Arsine glacier is located in the Ecrins area.
The glacier of Taconnaz located in the Mont-Blanc range starts from Dôme du Gôuter at 4300 m asl
and flows down to the altitude of about 1700 m above the Taconnaz village in a small valley
orthogonal to the main valley of Chamonix. The glacier is globally North orientated . Owing to the
bedrock topography, a roughly 100m -high ice cliff develops over most of the glacier width (500 -600
m wide) from about 3200m asl (bottom) to 3300m asl (top). This ice cliff separates the glacier into an
upper accumulation area and a lower ice tongue partly fed by regenerated ice from the ice fall.
Deliverable 4 59

Figure n°2: Picture of the Taconnaz serac fall
The large ice falls breaking off the glacier at 3300 m asl are responsible for large avalanches of snow
and ice in winter. These large avalanches can cause serious damage in the valley of Chamonix. During
winters with large amounts of snow, the ice mass breaking off triggers off large snow avalanches
downstream and these avalanches formed by a mixture of ice and snow can reach the inhabited areas
as was the case in 1988. Although large dams have been built in the valley to prevent avalanche from
reaching the inhabited areas, this ice fall continues to threaten the valley.
In the framework of Glaciorisk project, the LGGE proposal consists of carrying out a fundamental
study in order to understand the ice flow regime of this glacier in the vicinity and upstream of the
breaking off zone. The objectives of this study are the following:
- The first topic is to estimate the discharge of the glacier at the elevation of the ice breaking
off.
- The second objective is to estimate the frequency of the ice mass breaking off.
- Last, the third goal is to determine the maximum amount of ice which could break off at once
from the cliff.
2 Observations carried out in the framework of the Glaciorisk project
At the beginning of the project, an observation network has been set in order to determine mass
balance and surface flow velocities in the upper area of the glacier. For theses purposes, ten stakes
have been set up on the 12 th June 2001 in the upper part of the glacier. These stakes consist of large 5meter-long
pieces of wood wih a diameter of 10 cm weighting 25 kg each. These stakes were installed
thanks to helicopter transportation by dropping them at the altitude of 4000 m asl. They were then
carried to each site by fieldworkers on ski. These stakes have been replaced on the 17 th of January
60 Deliverable 4

2002 and on the 17 th of September 2002. They allow for mass-balance and horizontal velocities to be
monitored.
The emergences of these stakes have been measured several times during the year. The density of the
snow has been determined from two nearby ice core drilling. From these measurements, annual massbalance
have been calculated between October 2001 and September 2002, and are reported on Figure
3.
From the same stakes, surface ice flow velocities have been measured using topographic instruments.
For safety reasons, and because the access to the glacier needs helicopter transportation, most of these
surveys have been done from geodetic sites outside the glacier without anybody staying on the
glacier. The topographical method used for this purpose is based on intersections from 2 or 3 geodetic
stations. The coordinates of these stations have been obtained from GPS geodetic receivers (Leica
510) and the coordinates are known with an accuracy of 5 cm in the coordinates network of the
National Geographic Institute. From these geodetic stations, horizontal and vertical angles have been
measured with T2 theodolite in order to determine the coordinates of each stake. The topograpic
stations are located close to Aiguille du Midi (3800 m) and to Plan de l’Aiguille (2300 m). The access
to these geodetic points is easy thanks to the cable car of Aiguille du Midi. The distances between
stakes and geodetic stations remain however very large (4.5 km and 5 km) and deteriorate the
accuracy of the stakes coordinates (± 0.3 m on each measured position). In this way, annual surface ice
flow velocities (2001/2002) have been obtained and are reported on Figure 4. As can be seen on this
figure, the annual surface velocity measured close to the ice cliff is around 80 m/y.
Another topic of this study is to estimate the frequency of the ice mass breaking off and which
maximum amount of ice could break off at once from the cliff. For this purpose, regular topographic
measurements have been performed (every 3 weeks in average, since February 2002) in order to
measure the position of the ice cliff edge. For safety reasons, these surveys have also been done from
geodetic sites outside the glacier without any operator on the glacier. It is indeed impossible to access
the seracs cliff. Measurements have been carried out using an intersection method: horizontal and
vertical angles have been measured at some 100 points from each topographic station (Aiguille du
Midi, 3800 m asl, and Plan de l’Aiguille, 2300m asl). Since each measured point from the first station
is not recognizable from the second station, the sights cannot be associated directly and the
calculations cannot be done with usual methods. An indirect method must be used: for each sight of
the first station, we have to calculate the intersection XY with every sights of the second station and to
compare the altitudes Z1 and Z2 obtained from these calculations. When the difference between both
altitudes is minimum and is less than 1 meter, the coordinates are stored.
However, this method is valid only if the altitudes of the stations and of the sighted points are very
different, in such a way that it is possible to discriminate the intersections from the altitudes
comparison. Using this method, it is then possible to survey the top of the ice cliff without any
recognizable points on the characteristic feature on the edge. Unfortunately, these measurements
cannot be done using automatic topographic station and thus require an operator. The weather
conditions however remain the largest limitations of these surveys.The topographic surveys have been
carried out at the dates shown in Table 1.
From these surveys, the fluctuations of the front (edge) through time have been determined. Finally,
two photogrammetric campaigns from helicopter have been performed on 10 th March 2003 and on 23 rd
May 2003. The topic of these measurements was to determine the volume variations of the seracs fall
before and after a large ice mass breaking off, in order to link the lentgh fluctuations (observed from
terrestrial topographic measurements) with the corresponding volume variations. Two digital elevation
model have been performed from these aerial photographs thanks to numerous and detailed
photogrammetric measurements obtained from Leica DSR instruments).
Deliverable 4 61

3 Results
Date Field work
12.06.2001 First set up of stakes – Topographic survey
11.10.2001 Topographic survey
17.01.2002 Second set up of stakes
13.02.2002 Topographic survey
05.03.2002 Topographic survey
26.03.2002 “
19.04.2002 “
17.05.2002 “
13.06.2002 “
08.07.2002 “
29.07.2002 “
30.08.2002 “
17.09.2002 Third set up of stakes
01.10.2002 “
20.10.2002 “
06.11.2002 “
19.12.2002 “
17.01.2003 “
13.02.2003 “
21.02.2003 “
05.03.2003 “
10.03.2003 Photogrammetric measurements
20.03.2003 Topographic measurements
14.04.2003 “
01.05.2003 “
23.05.2003 Photogrammetric measurements
20.06.2003 Topographic measurements
14.08.2003 “
15.10.2003 “
Table 1: Measurements dates and types
3.1 Annual discharge from mass balance measurements
The mass balance measurements have been performed for the hydrological year 2001-2002.
In order to be able to compute balance flow, net accumulation data (measured along the central line)
had to be generalized all over the upper accumulation area. Mass balance was thus calculated by
applying two altitudinal gradients so as to fit the existing data. The best match was obtained with a
first positive gradient varying linearly from 0.27 m w. e. per 100 m at 3300 m a.s.l. to 0.1 m w.e. per
100 m at 4300 m and a second wind –induced gradient varying linearly from 0 at 3750 m to –0.45 m
w.e. per 100 m at 4300 m. Details about this parametrization are given in Le Meur and Vincent
(submitted ). Moreover, as mass balance on two neighbouring glaciers are strongly negative for the
year 2001-2002, the mass balance data have been increased by 0.6 m w.e. in order to obtain a mass
balance distribution that is more in accordance with the current state of the glacier. The corresponding
mass balance map is depicted on Fig.3
62 Deliverable 4

108000
107000
106000
105000
104000
N
Refuge
du Goûter
3817 m
Aiguille
du Goûter
3863 m
Mass-balance (m.we) 2001/2002
3154 m
Ice cliff
Le gros Béchar
2846 m
>1.50
>1.50
0.50
0.45
0.95
0.25
0.50
Dôme du Goûter
4304 m
0 500 1000
Montagne de
la Côte
103000
948500 949500 950500
2642 m
3330 m
Figure 3: Annual mass balance measurements from stakes emergence.
The glacier discharge has been determined from this mass balance map, using a balance fluxes model.
The method first assumes a steady state for the glacier. From long-term mass balance measurements
carried out in Mont Blanc area, we know that the mass balance are strongly negative over the last 20
years. Nevertheless, glaciers in the Mont Blanc area have only experienced minor changes (less than
10 m) above 3000 m a.s.l over the twentieth century. It was therefore assumed that deviations from
steady state are small enough for the method to give a realistic picture. Under steady state, the
continuity equation for ice reduces to:
ρ(A-M)=∇.q
where A is net accumulation (m w.e.a -1 ), M basal melting (m w.e.a -1 neglected here), ρ the water to ice
density ratio (1.09) and q the ice flux vector per unit width (m 2 .a -1 ). For every grid point, two upward
Deliverable 4 63

flow lines are computed from the two neighbouring half-grid points along the line orthogonal to the
main direction of flow. Points along these flow lines are computed step-wise along the line of steepest
slope until the maximum altitude (the upper dome) is reached or until the two flow lines merge. This
delineates the surface over which the accumulation is integrated from the previously deduced mass
balance map. Balance fluxes are reported on Fig.4 where as an illustration, all the computed flowlines
on an east-west transect passing around the cliff have also been reported.
Figure 4: Computed yearly balance fluxes per unit width (m 2 .a -1 ) over the accumulation area
When considering the total flux along the two branches of the cliff as depicted on Fig.4 various levels
of smoothing gave about the same results with an annual flow of 0.456 10 6 m 3 on the north-west
section and 0.858 10 6 m 3 on the north one, that is an estimated total flux accross the cliff of about 1.31
10 6 m 3 By accounting for uncertainties when reading stake emergence and for inacuracies in the mass
balance map parameterization as well as in the outlining of the drainage area, the overall error in mass
balance has been estimated to 0.25 m w.e. a-1, which represents 15% of the average net accumulation
over the drainage area leading to the cliff. Adding now an estimated 10% error due to the method
assumptions (steady state glacier and flow along the steepest slope) one comes to a discharge value to
within 25 %, in other words 1.3 +/- 0.3 10 6 m 3 a -1 . Finally, it should be noted that the cliff as outlined
on the Figure 4 is less than the actual cliff (about 600 m wide) because it was restricted to what is
considered as the active part (the one succeptible of triggering off avalanches)
3. 2 Annual discharge from ice flow velocities measurements
Surface velocities obtained from remote topographic surveys of the stakes are reported on Fig. 5. From
the respective positions of the stakes along the principal direction of flow (North axis) velocity
64 Deliverable 4

gradients in between the stakes have been computed. As can be seen, except between the last two
stakes, similar gradients reveal a rather steady increase in velocity as the flow goes on, the slight
variations being probably indicative of local flow variations. Conversely, this velocity gradient
significantly increases between the last two stakes without any major change in surface slope. This
could indicate that this last stake, which was initially placed already close to the cliff edge, gives an
abnormally high velocity. It is well known that prior to collapsing, calving blocks are subject to an
acceleration indicative of a destabilization process by which equilibrium is broken, a fact confirmed by
several studies (Flotron, 1977; Iken, 1977). As a consequence, this last value of 91 m.a-1 might be
misleading and not representative of a steady flow above the cliff. It was therefore chosen to compute
the flow across the gray east-west section between stakes 5 (76 m.a-1) and 6 (80 m.a-1) with an
average velocity of 78 m.a-1.
108000
107000
106000
105000
104000
N
Refuge
du Goûter
3817 m
Aiguille
du Goûter
3863 m
Mean surface velocities 2001/2002
Le gros Béchar
3154 m
2846 m
91 m/y
73 m/y
71 m/y
66 m/y
66 m/y
55 m/y
38 m/y
Ice cliff
Dôme du Goûter
4304 m
0 500 1000
Montagne de
la Côte
103000
948500 949500 950500
2642 m
3330 m
Figure 5: Ice surface velocities from topographic measurements.
This section (depicted in gray on Fig. 5) has an equivalent width with regards to the flow direction of
400 m, which is the same as that of the two-branch featured white cliff. Given the flow line pattern
over the area (Fig. 4), it is very reasonable to assume similar fluxes between the two. Because ice
essentially deforms in shear, ice particles velocities increase from the bedrock up to the surface. Given
the flow properties of ice as expressed by a viscous power flow law (Glen, 1955), and with some
assumptions about the flow, it is possible to express the ratio of the velocity averaged over the ice
thickness to the surface velocity. In the case of a cold glacier frozen to its bed as is certainly the case
Deliverable 4 65

for the Taconnaz glacier (mainly because of its high altitude), this ratio only depends upon n the
exponent factor in the flow law and amounts to 0.8 with the traditionally accepted value n=3
(Paterson, 1994).We will therefore use an average velocity of 62 m.a -1 over the ice column. Having a
velocity value and a width over which it is assumed to apply, the last unknown in order to compute a
flux is the ice thickness along the section. Unfortunately, this kind of data is presently unavailable and
would require lengthy radar or seismic sessions or borehole down to the bedrock. The only assumption
at our disposal consists of extrapolating the actual ice cliff height inward underneath the section.
Thanks to the accurate photogrammetric measurements of the cliff upper edge, numerous
measurements can be used to assess the ice thickness at the cliff edge. An average value of 70 m has
been assessed and extrapolated back all over the theoretical 400 m width of the section. Under these
assumptions, the net annual flux is obtained as 1.736 10 6 m 3 .a -1 , a value that agrees fairly well with
that of 1.3 10 6 m 3 a -1 previously obtained from mass balance measurements. Considering uncertainties
of 10 m.a -1 and 15m for the surface velocity and cliff thickness respectively, the total uncertainty on
this flow value is now around 0.45 10 6 m 3 a -1 .
3. 3 Frequency of the ice mass breaking off.
From accurate topographic measurements of the cliff edge every 3 to 4 weeks, average positions of the
calving front ahead of each of the two segments as featured in Fig. 4 have been computed and
represented on Fig. 6. This averaging procedure excludes the central part of the cliff (the two inner
parts of the two base lines) where the geometry is too complex as a result of frequent falls of smallsized
blocks. Because of the discontinuous nature of the measurements, interpretation of Fig. 6 is
subject to caution, first because small events can be missed and secondly because extreme points are
minimized (a collapse does not systematically follows right after the measurement of a maximum and
similarly, neither does it systematically strictly precede a measured minimum). However, several
major collapses remain noticeable and have the particularity to occur once the cliff edge reaches a
treshold distance, which yields an interesting pseudo-cyclic pattern. Because our time series are still
short, it is difficult to know how meaningful the corresponding pseudo-period can be. However, if it is
assumed that calving volumes associated with these large collapses of entire lamellas are similar from
one event to the other, and that the flow rate through the cliff is steady, it then becomes possible to
infer a characteristic period.
66 Deliverable 4

Mean length (m)
Mean length (m)
60
50
40
30
20
80
70
60
50
40
30
Left stream
0 120 240 360 480 600 720
Julian days
Right stream
0 120 240 360 480 600 720
Julian days
Figure 6: Stepwise average position of the cliff upper lip as a fonction of time. The first julian day corresponds
to the 1 st of January 2002.
Calving volumes associated with these events can only be roughly assessed for several reasons. First,
because extreme points are minimized (see above), the associated front average retreat is similarly
minimized. Secondly, deriving a calving volume from the observed retreat supposes to know the
geometry of the breaking surface in depth (the surfacial area on the other hand can be rather precisely
derived as the difference between the two appropriate cliff edge profiles ). However, from one cliff
survey to the other, it appears that the bottom of the cliff (where observable) undergoes much smaller
displacements than the top. This is in line with a glacier frozen to the bedrock as already suggested. It
then suggests a breaking surface more of the wedge type than one orthogonal to the ice upper surface
or simply vertical. Moreover, if one assumes similar breaking geometries from one collapse to the
other, the lamella cross section must more or less match the area featured by the velocity profile. As a
consequence, the volume obtained by multiplying the dark triangle area by the appropriate cliff length
should give a realistic volume.
For the north west cliff, where two similar retreats of about 25 m have been measured over a width of
140 m and for an average thickness of 80 m (Blanc, 2003), the volume of collapse can be
approximated by 140m*25m* 60 m = 0.210 10 6 m 3 . When divided by the yearly flux of 0.456 10 6 m 3 ,
a characteristic period of some 168 days emerges which is of a similar order as that of 180-190 days
that can be directly deduced from Fig.6.
Deliverable 4 67

2106200
2106000
2105800
2105600
2105400
2105200
2105000
3. 4 Maximum amount of ice which could break off at once from the cliff.
For this purpose, the volume variation between the maximum and the minimum limit of front ( Figure
6) has been estimated. For that, we assume that the maximum amount of ice which could
simultaneously break off from the cliff corresponds to the volume variation between the 2 extreme
situations measured from the front fluctuations (the shortest and the longest front). For this estimation,
photographs from helicopter have been taken before and after a large event of ice mass breaking off.
These aerial photographs have been taken on the 10 th of March 2003 and on the 23 rd of May 2003.
From these photographs, photogrammetric measurements have been carried out (using a
photogrammetric instrument Leica DSR) to build 2 Digital Elevation Model (DEM). Geodetic
positionning measurements using differential GPS instruments were required to determine the
coordinates of points (with cm-accuracy ) which are visible on photographs (ground control points
required for geometric transformation between photographs coordinates and geodetic coordinates).
The difference between these 2 DEM allows to provide volume variation between the 2 campaigns.
Although the dates of the campaigns were chosen according to weather condiditons and instrument
availability and therefore do not correspond to the lowest and the highest limits which might have
been reached by the seracs, it is still possible to link the length fluctuations to the volume variations.
Therefore, the maximum amount of ice which could break off at once from the cliff inferred from
these calculations (extrapolation) is about 350 000 m3. These calculations have been performed for the
left stream of the ice fall and not for the right one. Nevertheless, the left part of the cliff concerrned by
this estimation as outlined on Figure 4 is less than the actual cliff (about 600 m wide) and could miss a
large active part succeptible of triggering off avalanches). From photogrammetric measurements
carried out on 10 th March and 23 rd May 2003, and from photographs taken from Aiguille du Midi, we
know that a large single block of 90 000 m3 fell from the right stream of the ice fall. Consequently, the
assumption about the active part (restricted to the left stream in a first step) can be questionned and
would require further observations.
N
Ice cliff
10 March 2003
949000 949200 949400 949600 949800 950000 950200 950400
3.5 Ice flow velocities measurements.
2106200
2106000
2105800
2105600
2105400
2105200
2105000
Figure 7: Digital elevation models.
N
Ice cliff
23 May 2003
949000 949200 949400 949600 949800 950000 950200 950400
The ice flow velocities measurements have been performed from stakes which have been set up in the
upper area of the glacier of Taconnaz. Some of these stakes have been set up in the vicinity of the
breaking zone, about 60-80 m upstream the edge of the seracs (Figure 8). The positions of these stakes
have been computed from topographic measurements. Many stakes have been lost because of the
avalanches coming from the upper area of the glacier. However, some of these stakes have been
measured until the seracs breaking off (stakes 10, 18, 19 and 20) and are reported on Figure 8. The
topographic positions accuracy is about 0.5 m. The surface velocity results can also be affected by the
stakes inclination. The surface velocity accuracy depends on the time intervals which have been used
68 Deliverable 4

for the velocities calculations. On Figure 9, the velocities calculations have been performed over a 2
months interval approximatively. The overall error in surface velocities have been reported on Figure
9 and have been estimated to +/- 4 m/year, in average. The origin of the distance on Figure 9 has been
fixed at 162 m from the limit of the cliff edge. One must notice that the stake 10 has been set up on the
12th of June 2001 whereas the stakes 18, 19 and 20 have been set up on the 17th of September 2002.
Moreover, the flowlines of the stake 10 is very different from those of the stakes 18, 19 and 20 (Figure
8). The small velocity variations shown on Figure 9 between 60 and 110 m from the origin are
significant because these spatial variations are the same for the stakes 18 and 19 which passed through
this area at different time. On Figure 9, the breaking off limits have been reported thanks to the
detailed observations relative to the cliff edge. From the stakes 10 and 20 measurements, one can be
seen that a large acceleration is visible in the breaking off zone, that is to say about 25-30 m before the
breaking off limit.
105500
105480
105460
105440
105420
105400
105380
105360
105340
105320
105300
105280
105260
105240
Stakes set up on 12/06/01
Stakes set up on 17/01/02
Stakes set up on 17/9/02
19
18
20
7
105220
105200
Ice flow
13
949500 949550 949600 949650 949700
10
14
16
6
17
15
9
8
14.04.2003
17.09.2002
15.10.2003
5.03.2003
14.08.2003
17.01.2003
06.11.2002
20.06.2003
Figure 8: Map of the breaking off zone. The stakes
set up since June 2001 have been reported. Some
positions of the front in 2003 have been drawn and
show the limits of the cliff edge.
Distance (m)
Surface velocity (m/year)
220
200
180
160
140
120
100
80
60
40
20
180
160
140
120
100
80
60
a)
10
20
19
18
100 200 300 400 500 600 700
Days
b)
18
Breaking off
zone
20
19
10
Break off
0 20 40 60 80 100 120 140 160 180 200
Distance (m)
Figure 9: a)Moving stakes as function of time
(top). The origin of time is the 1 st January of 2001
for the stake 10 and the 1 st January of 2002 for the
stakes 18, 19 and 20, b) Surface velocities as
function of distance. The origin of the distance has
been fixed at 162 m from the maximum limit of the
front.
Deliverable 4 69

4 Conclusions and prospects about Taconnaz study.
This study shows how from 3 different methods based on 3 independent data sets, it is possible to infer
consistent characteristics for the Taconnaz ice fall. It therefore gives a large credit to such an
essentially phenomenological approach, in a domain where more physically-based accurate results
like those from numerical models for instance are still presently lacking. In front of a growing and
urgent need for a better understanding of the processes underlying glacial hazards, such approaches
should be fully used with as many field data as required, especially when some possibilities of
forecasting the associated risk arise. For instance, in the present case, it is possible to detect a risky
situation whenever one of the two cliff portions advances up to a characteristic limit whose position is
already detectable from our time series. Should this limit be confirmed by the forthcoming
observations, the method could rapidely become operational. Indeed, in case of a local avalanche risky
situation (following heavy snow falls for instance) and a simultaneous serac advance close to this
limit, a warning could be issued with ensuing closing of the nearby ski track (which was swept by the
february 1999 avalanche, but fortunately enough during the night). Moreover, from the same series
of cliff position through time, similar calving retreats from one event to the other as well as steady
advances in between are clearly visible (at least on the north-west cliff) which unavoidably result in
cyclic occurences of collapses. The two obvious similar inferrred periods of 180-190 days can be
criticized on the grounds that they may just be coincidental, except that two consistent flow values
determined independently, associated with calving volume estimates do confirm this cycle with a
similar characteristic period. This approach however suffers from some uncertainties on both the exact
times of collapse and the associated falling volumes. This is why the survey protocol is meant to be
improved with more frequent observations (daily remote photographs from one of the geodetic
stations). From figure 6, it seems that the seracs grow up to a limit and break when they reach this
limit. Therefore, regular surveys would allow time of occurrence to be predicted.
In the future, and probably in the framework of a regional contract, we will propose 1°) to improve the
survey protocol of the front fluctuations, 2°) to perform temperature measurements in deep hole
drilling, in order to know the basal conditions of the glacier.
The local authorities are involved in this project . A first meeting has been held on the 15th of January
2003 in order to explain the scientific purposes of this study and to show the first results. Another
meeting is planned at the beginning of the year 2004 in order to show the last results and to decide the
possible observations to perform in the future.
Acknowledgements
We would like to thank all those who collected data from field measurements. A part of this study was
also supported by Le SIVOM de la Haute Vallée de l’Arve, by La commune de Le Monestier les
Bains, et le service du RTM des Hautes Alpes. We are deeply indebted to Renaud Blanc for the
extensive field measurements carried out on the glacier of Taconnaz.
References :
Blanc, R. 2003. Etude de la barre de séracs du glacier de Taconnaz. Mémoire de fin d’étude, Ecole d’Ingénieur
ESGT, présenté le 28 novembre 2003 au Mans, France.
Flotron, A. 1977. Movement studies on a hanging glacier in relation with an ice avalanche. J. Glaciol 19 (81),
671-672.
Glen, J.W.1955. The creep of polycrystalline ice. Proc. R. Soc. London, Ser. A 228 (1175), 519-538.
Iken, A. 1977. Movement of a large ice mass before breaking off. J. Glaciol. 19 (81), 595-605.
Lemeur, E. and C. Vincent. Monitoring of the Taconnaz ice fall from mass balance, surface velocities and ice
cliff positions. Submitted to the Journal of Glaciology.
70 Deliverable 4

Paterson, W.S.B. 1994. The Physics of glaciers. Third edition. Oxford, Elsevier.
Vallon, M. Evolution, water balance, potential hazards, and control of a pro-glacial lake in the French Alps,
Annals of Glaciology, 13, 273-278, 1989.
Vincent, C. 2002. Influence of climate change over the 20 th Century on four French glacier mass balances. J.
Geophys. Res., 107(D19), 4375,doi:10.1029/2001JD00832.
Deliverable 4 71

Icefalls in Norway.
2 Miriam Jackson
Glacier and Environmental Hydrology Section, Norwegian Water Resources and Energy Directorate,
P.O. Box 5091 Majorstua, N-0301 Oslo, Norway.
In Norway, icefalls from glaciers are a minor, but recurring problem. NVE has been involved in the
study of several icefalls in Norway and compiled a description and list of the main known icefall
events in both English and Norwegian. The full report has been submitted previously. The list of
events is as follows:
1693 Loen in Bødal and Nesdal, Sogn and Fjordane fylke. Falling glacier ice and flood, lots of
damage.
1723 Engabreen, Svartisen. The glacier destroyed the Storsteinøyra farm, and caused a lot of
damage to the Fonnøyra farm. Glacier advance.
1734 Olden at Tungøen, Sogn and Fjordane. Flood. Water and ice from the glacier. Grazing areas
seriously damaged.
1741 Tuftebreen, Jostedalen. Bergseter farm at the end of Krundalen was seriously damaged by the
glacier. Glacier advance.
1742 Nigardsbreen, Jostedalen. Nigard farm was totally destroyed because of a glacier advance.
1743 Brenndalsbreen, Jostedalen. Glacier advance.
1743 12 th December. Olden at Tungøen, Sogn and Fjordane. Glacier ice, flood, gravel and rockslide.
The farm was totally wiped out by an icefall from Åbrekkebreen. All but two occupants of the
farm died. The farm, which had previously had 40 dairy cows, horses and large fields and
grazing areas was removed from the land register the day of the icefall (Eide, 1955).
1756 Ulvundeid in Nordmøre, More and Romsdal. Icefall from glacier.
1773 Ulvundeid in Nordmøre, More and Romsdal. Icefall from glacier.
1819 Ulvundeid in Nordmøre, More and Romsdal. Icefall from glacier.
1850 Vinnufjellet in Sunndalen, More and Romsdal. Icefall from glacier.
1900 Frostisen in Skjomen. Many icefalls were registered here in about 1900. Six icefalls were
registered in the course of only one day. The biggest icefall was in 1906 where at least 7000
m 3 of ice fell. The danger was over some years later when the ice core had melted. An outlet
of the glacier causes icefalls on a steep hillside (Hoel & Werenskiold, 1962).
1966 20 th March. Møsvikfjorden in Sørfold. Calving ice resulted in an enormous wave that swept
over the land on the other side of the water. Many farms were seriously damaged. Parts of
Reinvikisen slid out and down into Dalavann. The wave destroyed settlements in Reinvika
(Jørstad, 1968). Ref: Unmarked memorandum from the folder "Isras", HB, NVE.
2 Contact Miriam Jackson at mja@nve.no.
72 Deliverable 4

1986 22 nd July. Nigardsbreen, Jostedalsbreen. The tunnel opening at the front of the glacier
collapsed and dammed the rived for a while. When the dam broke open large quantities of ice
blocks and water ran down from the glacier. Twelve people were caught in the water and ice
blocks and two of them died – drowned or injured by the ice blocks. Ref: Finnmark dagblad
23.7.1986.
1987 Fonndalsbreen – outlet glacier from Svartisen. An icefall from an elevation of about 800 m
a.s.l. occurred in the spring or early summer. Ice blocks were found down to an elevation of
215 m a.s.l. That is, the icefalls vertical fall was 585 m, and the horizontal reach was 1300 m.
The valley is covered by a regenerated glacier between 600 and 280 m a.s.l. There are no
given figures for the icefalls volume but subsequent calculations suggest a volume of about
300,000 m 3 – not an unreasonable figure for a glacier of this size.
1987 27 th July, Baklibreen – outlet glacier of Jostedalsbreen. Three people were killed when an ice
mass of about 250 000 m 3 in size became loose from the eastern part of the glacier front at an
elevation of 1200 m a.s.l. and fell into the valley underneath at an elevation of 500 m a.s.l.
1994 12 th August. Nigardsbreen, Jostedalsbreen. Calving. Some blocks from the glacier front fell
and slid down from the glacier. A Polish woman was seriously injured when she was struck
by one of the ice blocks and thrown into the water. Ref. Stavanger aftenblad 13.8.1994.
2002 1st April. Tuftebreen/Jostedalsbreen. Icefall.
The main glacier at risk for icefalls is Baklibreen. This is in a popular area with tourists and an icefall
occurring in 1986 killed three people who were walking on a path below the glacier. An observation
programme was set up in 1987 to study the risk of future icefalls, and was in operation until 1999.
The programme began again in a more limited capacity and was carried out as part of Glaciorisk from
2001 to 2003. Measurements on Baklibreen since 1993 have been made from a survey point
established on a nearby prominent rock outcrop, and sightings are then made with a GDM to different
points on the glacier. These points are visited by helicopter, and prisms are used for sighting. Previous
work showed that the glacier surface became higher as ice built up between 1984 and 1994 by 12 m to
19 m, but that there was little change between 1994 and 1999. Surveys of the surface of the glacier
between 2001 and 2003 show that there has been consistent lowering, probably due mainly to
increased melting during the warm summers of those three years (Figures 1 and 2).
Deliverable 4 73

North (m)
North (m)
6839000
6838000
6839000
6838000
Acknowledgements
0m 250m 500m
Map constructed from aerial photography
taken on 10th August 1984
Contour interval 50 m
Coordinate system UTM Euref 89, Zone 32
-3 -6
397000 398000 399000
East (m)
Baklibreen
-4
2001 Survey Point
2002 Survey Point
Figure 1. Change in surface elevation in metres from 2001 to 2002.
0m 250m 500m
Map constructed from aerial photography
taken on 10th August 1984
Contour interval 50 m
Coordinate system UTM Euref 89, Zone 32
-8 -4-3 -3
-3
-4
397000 398000 399000
East (m)
-4
-5
Baklibreen
2002 Survey Point
2003 Survey Point
Figure 2. Change in surface elevation in metres from 2002 to 2003.
Ole Magnus Tønsberg, Bjarne Kjøllmoen, Rune Engeset and Hallgeir Elvehøy all participated in the work in
Work Package 2 of Glaciorisk at NVE.
74 Deliverable 4

Length changes of glaciers
Flow fields of glacier outlets from Jostedalsbreen
Our main work has been in connection to glacier outlets from the growing ice cap Jostedalsbreen in
western Norway (Fig. 1). Several outlets from this ice cap have advanced during the last ten years,
causing avalanches and ice block falls. Some of these events have led to dangerous situations and even
accidents. We have worked on development of new methods for monitoring the velocity and front
situation of these glaciers.
Background
Fig. 1. Location of Jostedalsbreen ice cap in western Norway
Aerial photography was planned for several glaciers on Jostedalsbreen. The idea was to cover three
different selected parts of Jostedalsbreen that has been picked out as potential risky areas. These
include the outlets in three main areas. 1) In Eastern part: Nigardsbreen, Baklibreen and Bergsetbreen;
2) In western part: Brigsdalsbreen and Kjøtabreen; and 3) in south-western part: Flatbreen and
Bøyabreen. Aerial photographs at a scale of 1: 15 000 was planned for the purposes of measuring
surface velocities, volume change, looking for potentially risky glaciers etc. The idea was to take
repeated photos at an interval of maximum 10 days. This was considered to be the maximum interval
if the photos could be used in pattern recognition and cross-correlation analysis for mapping of the
velocity field. Photos with longer intervals, months or years are useful for front change measurements
and for Digital Elevation Model constructions and volume change measurements, but not for
dynamical studies. For longer intervals than ten days melting of the ice surface could make it difficult
to recognise the same features in both images. That was a reason for conducting the photography as
late as possible in the summer season, when melting is less than in the middle of the summer.
Deliverable 4 75

The objectives for this study is to present a fully digital chain of image processing and analysing
techniques for automatic determination of surface geometry changes with time and surface
displacement (velocity field) on glacier from repeated aerial photography.
Digital Terrain Models (DTM’s) and Orthophotos
The same strips were flown and photographed at Jostedalsbreen both August 19 th and 29 th 2001. The
imagery covers 10 of the outlets of Jostedalsbreen. The scale of the images is 1:23,000 and a 14µm
scanning resolution gives a spatial resolution of around 30 cm. The temporal resolution of 10 days
should be feasible for mapping the expected displacements of 5-10 m.
An orthophoto of Nigardsbreen was generated for each day with a digital photogrammetric
workstation (Z/I Imaging). This was done by automatically constructing a DTM with a spatial
resolution of 2 m and then orthorectifying the aerial images. The aerial photos have a scale of about
1:23,000 and a 14µm scanning resolution gives a spatial resolution of approximately 30 cm for the
aerial photos. For these calculations an orthophoto resolution of 50 cm has been used. The orthophotos
was co-registered before the velocity calculations were done by using IMCORR.
The cross-correlation software IMCORR, developed at National Snow and Ice Data Center (NSIDC)
(Scambos et al., 1992), was used for the displacement calculations. IMCORR use a fast Fouriertransform
version of the normalized cross-covariance method (see Berenstein, In Manual of Remote
Sensing, 1983) to cross correlate sub scenes of the two orthophotos of the same glacier. This is done
by searching for the point of best correlation between a small reference chip of the orthophoto of the
first date within a larger search chip in the same area in the orthophoto of the second date. The point of
the best correlation, given that it is greater than a certain threshold, is taken to be the new position of
the reference chip in the orthophoto of the second date. The difference between these two positions is
caused by the horizontal displacement of the moving glaciers and it is the conservation of the crevasse
pattern that makes it possible to use this technique.
The IMCORR software matches small sub scenes (called chips) from the two orthophotos. For
every location in a specified grid a correlation index is calculated for the search chip and the smaller
reference ship. The best match of the search chip and the reference ship is taken to be the new position
for the reference ship in the new images. Hence the method is based on automatically feature tracking.
If the crevasse pattern is not disrupted during the time between the two acquisitions the new position
of a given crevasse or other identifiable feature is calculated. The vector between the old and new
position is taken to be the displacement.
In this test the grid size used was 12 m (40 pixels), the search ship size 38.4m (128 pixels) and
the reference chip size 4.8 m (16 pixels). The spatial resolution of the orthophotos were 0.3 m. The
results presented here have been interpolated and a median filter has also been applied to enhance the
results.
Test Site
The test site chosen for the digital photogrammetric investigations was the glacier Nigardsbreen,
Southern Norway (Fig. 2.) Nigarsbreen is an outlet glacier of Jostedalsbreen, with an area of
approximately 48 km 2 and an altitudinal range of 320-1950 m a.s.l. The glacier shows a complex
topography with three tributaries draining from the plateau down into the lower part of the glacier. The
highest surface slope can be found in of the tributaries. This is one of the most visited glaciers in
Norway where hundreds of tourists come every day during the summer season. It is popular for glacier
climbing, both as guided tours and in individual groups. The steep advancing front has made this
activity more difficult over the last years. Several ice blocks have also broken off at the front. Warning
signs and fences have been put up to prevent accidents.
Nigardsbreen has been extensively investigated since the second half of the twentieth century. The
glacier advanced very rapidly in the first half of the 18 th century and reached a neoglacial maximum in
1748 AD. After that a steady retreat took place which recently came to an end. The total retreat of the
snout was of more than 4 km. The glacier snout has readvanced during the last years as result of
positive mass balance over several years. Over the last ten years the front has advanced more than 200
meters.
76 Deliverable 4

NVE is conducting a mass-balance program of and have published detailed maps of the glacier. Massbalance
observations have been carried out since 1962. Early photogrammetrical and geodetical
investigations gave surface displacements of up to a maximum of 300-400 hundred meters per year in
the ice fall.
GPS-field work
Fig. 2. Nigardsbreen draining down from the Ice cap Jostedalsbreen.
In 2001 field investigations were carried out in June, August and October. In 2002 the fieldwork was
carried out in end of June. The surface velocity was measured by precise GPS surveying of points on
the glacier relative to fix point outside the glacier. The obtained accuracy can then be better than ± 5
cm in horizontal position and better than ± 10 cm in vertical position. Eight measurement points have
been established on the glacier (Figs. 3, 4 and 5)
In May 2003 Ole Magnus Tonsberg finished his master thesis on GPS-measurements of the dynamics
of Nigardsbreen. The work is directly linked to GLACIORISK as input to calibration of DTM,s.
Fig 3. A base station outside the front of the glacier Nigardsbreen in Western Norway used both for GPS-survey
and terrestrial stereo photography, June 2001 (photo by K. Melvold).
Deliverable 4 77

Fig. 4. Point GPS-surveying on the glacier surface on the tongue of Nigardsbreen in June 2001 (photo by K.
Melvold)
Fig. 5. Orthophotos of the lower part of Nigarsbreen. Superimposed on the image are both contour lines (contour
interval 20 m) and velocity vectors measurements during the summer 2001 by static GPS-survey.
GPS-results
Results of the surface velocity measurements by GPS show that the velocity decreases toward the
front (Fig.5). All the stakes undergo seasonal velocity variations and the increase in velocity during
the summer was up to a 60 %. For example the daily velocity at the tongue (stake N1), few hundred
meters from the front range from 22 to 35 cm/day. Maximum velocity was measured between
20.06.01-22.08.01 and minimum was measured between 19.09.01-20.06.02 (winter). Since the
thickness and slope change is insignificant during the melt season, the increased speed during summer
must have been due to enhanced basal motion. Results of the surface velocity measurements show that
that velocity in the lower part of the tongue has increased by the increasing thickness and advancing
glacier front
78 Deliverable 4

Results of cross correlations
The calculated horizontal velocities on Nigardsbreen are shown in figures 6 and 7. Some shadows on
the lower part of Nigardsbreen in the original air photos introduce difficulties for the matching
algorithm. Since the shadowing is different in the two scenes it is more difficult to perform the cross
correlation and hence the density of successfully matched points decrease in this area. The calculated
velocities agree only partly with the GPS measured data. Over half of the points on the glacier surface
are matched. Some shadows on the lower part of the glacier in the original air photos introduce some
errors. Since the shadowing is different in the two scenes it is more difficult to perform the cross
correlation. In some areas the correlation is beneath the acceptable threshold and the measurements at
these points are disregarded. Measurements in the vicinity of these points are believed to be of rather
poor quality. Hence the interpolation introduces or reinforces the errors in these areas. Some kind of
local image enhancement techniques will hopefully remove some of the effects caused by the
differences in shadowing. In Fig. 8. and Fig. 9 are shown results of the cross correlation on
Bergsetbreen, a hanging valley glacier from Jostdalsbreen. When a suitable technique is found it will
be applied for the other outlets of Jostedalsbreen covered by the imagery.
Figure 6. Magnitude of the calculated horizontal displacement at Nigardsbreen. The black dots show measured
velocity by static GPS between 22.08.01 and 19.09.01.
Deliverable 4 79

Nigardsbreen
0,2 m contours
Velocity [m/day]
1,50
0
Fig. 7. Velocity field of Nigardsbreen derived by cross correlation of orthophotos
Fig. 8 Displacement field at the upper part of Bergsetbreen shown as vectors
80 Deliverable 4

Fig. 9. Magnitude of the horizontal displacement at Bergsetbreen
References
Bernstein, R. (1983), Image geometry and rectification, In Manual of Remote Sensing (R. N. Colwell, ed.),
American Society of Photogrammetry, Falls Church,VA, pp.881-884.
Scambos, T. A., M. J. Dutkiewicz, J. C. Wilson, and R. A. Bindschadler, 1992.
Application of image cross-correlation to the measurement of glacier velocity using satellite image data. Remote
Sensing of Environment, Vol. 42, 177 - 186.
Annex 1 - Abstract of the EGS/AGU-presentation
THE FLOW FIELD OF GLACIER OUTLETS IN THE JOSTERDALSBREEN AREA,
NORWAY, USING DIGITAL PHOTOGRAMMETRY
B. Wangensteen, T. Eiken, K. Melvold, O. M. Tønsberg and J. O. Hagen Department of Physical
Geography, University of Oslo, POBox 1042 Blindern, 0316 Oslo, Norway.
During the last decades most glaciers in the world have retreated. However, some of the glaciers in the
western part of Norway have advance during the last 10 years. This is related to increased winter
precipitation in the late 1980’s and the beginning of the 1990’s. As a results the glacier fronts have
grown steeper, more active and thereby more dangerous. The ice cap Jostedalsbreen (468 km2) is one
of the most visited glaciated areas in Norway where hundreds of tourists come every day during the
summer season. It is popular for glacier climbing, both as guided tours and in individual groups. The
steep advancing glacier fronts have made this activity more difficult over the last years. Several ice
blocks have also broken off at the front. Although large icefalls from hanging glaciers happen very
rarely, the consequences of such events can be dramatic. In order to survey this relative large glacier
several aerial photographs was taken in August 2001. The same strips were flown and photographed at
Jostedalsbreen both August 19th and 29th 2001. The temporal resolution of ten days should be feasible
for mapping the expected displacements of a few meters. The imagery covers ten of the outlets of
Jostedalsbreen. In this presentation we will show how the flow field and extent of some important
hanging glacier can be determined from repeated aerial photographs. Overlapping digital images,
DTM’s with high spatial resolution (10 m or better) and orthophotos have been derived using a
Deliverable 4 81

photogrammetric workstation (INTERGRAPH). In order to improve the accuracy between the images
taken on different dates all images sets have been oriented and adjusted as one image segment. The
horizontal surface displacement vectors have been derived from the 19. August and 29. August
orthophotos of 0.3-m pixel size. Different techniques and work strategies will be tested in order to
carry out the matching automatically. The results have been tested with GPS measurements on the
glacier surface.
82 Deliverable 4

Glaciers survey in the french Alps
Glacier Survey in Savoy
P. MACABIES Forest Ingenier ,
ONF service RTM Chambéry – Savoie
Service RTM of SAVOY followed from 2001 to 2003 the 2 glacial lakes Chavières (commune of
Modane) and of Rochemelon (commune of Bessans). At the end of this period of observation, we can
already make some reflexions about the evolution of these lakes and the risks which they present for
the people and the goods located downstream, and trace some tracks for the continuation of the
observations.
1: Lake Chavières:
With 2800m, it is located on the terminal moraine downstream from the glacier of POLSET in the
national park of Vanoise., the rook coming from the glacier crosses at 1000m the town of Modane, in
a very short bed. Informations known to date is as follows:
Evolution
year surface m2 altitude m
2002 21900 2806,8
2003 22400 2807,6
diffrence 500 0,8
Map of the lake in 2003 :
Deliverable 4 83

total volume 107 385 m3
the evolution between the years 2002 and 2003 indicate:
- an increase in the perimeter of 2 %. Surface in 2003: 2,2 ha.
- an increase in the level of lake of 0.8 m (altitude: 2807.6 m in 2003). Traces of mud were
observed one meter above the real level of the lake, corresponding to high waters of the summer.
This evolution could be related to the significant melting of the glacier (upstream) due to the high
temperatures reached this summer. In spite of the very important melting of the glacier this year, the
moraine which frames the lake does not seem to have suffered from the increase in the volume nor of
the high level of water of lake. The volume of the lake is not very important, its evolution does not
appear likely to destabilize the terminal moraine. Moreover downstream, if a debris flow were formed,
several projecting ledges would allow its disintegration.
It does not seem necessary to continue precise measurements. Information of the personnel of the
national park would be sufficient to alert service RTM if they noted an unusual phenomenon.
figure 1 : chaviere lake in 2001 from point of view n°2 (rock) figure 2 : the lake in 2003
2: Lake Rochemelon:
This lake is located at 3200 m on the glacier of Rochemelon, (commune of Bessans). It currently flows
on the Italian slope by the rock collar of Novalaise. French side is limited by a ridge of ice which is
reduced each year, a brutal discharge by rupture seems possible. Downstream is the long valley of
Ribon (12 km), of weak slope, then the torrent emerges in the valley of the Arc. It then cuts 2 roads to
rather significant traffic especially in summer.
84 Deliverable 4

a). Evolution of the surface:
figure 3 : Rochemelon lake in june 2001
year Document , surface (m2) length (m) Width average (m)
1985 Photo : 1 000 ? 50 20
1994 Photo : 4 500 ? 150 30
2002 Mesure GPS : 43 000 634 80
2003 Mesure GPS : 52 000 689 90
From the GPS campain of 2003, we can draw this map of the lake :
Novalaise pass
Axe of the glacier cut
North
Rochemelon lake surface in 2003
Deliverable 4 85

) measurements of depth: This operation was possible thanks to the intervention of 4 divers. The
measurement of depth on the ice side is 24m. The bottom of the lake is covered with sediments, and
the opposed side, is rocky.
The estimated volume of the lake for 2003 is 400 000 m3. These values must be compared with those
measured by SMS in 2001 by bathymetry over the lake and 2002 per radar on the glacier (Italy,
Rapport 2002):
Volume of the lake: 150 to 200 000 m3. Maximum depth 18m.
c) Prospects for the future.
The profiles carried out show a thickness of ice which varies from 2m against the rock face to 70m,
moving away from the rock face. The measurements GPS carried out in 2003 with the CEMAGREF
make it possible to make a cut of the glacier and lake, and to suppose the surface of the bottom.
At this place we can then check that the hydrostatic pressure cannot reverse the mass of ice:
Hydrostatic uplift: 2.88 MN applied at the third the height
Moment of inversion: 23 MN.m, stabilizing moment: 1005 MN.m (weight of the ice: 9000N/m 3 )
This balance becomes more difficult to respect when one approaches the rock face in direction of the
collar. It is there that the ice is the least thick and the least broad. One can think that it is here that the
rupture will occur, when the melting of the ice and the increase depth of the lake will be sufficient.
This risk is more important if there is a film of water at the contact rocher/glacier which would
facilitate the rising of the ice under the pressure of water. The glaciologists could check it by
measuring the temperature of the ice on this level.
86 Deliverable 4

However another less brutal scenario is possible: The ice dominates the lake of 4 m approximately at
the lower point of the glacier. While melting, it can come to join the level of the lake, fix (3218.6m).
The draining of the lake will be done then gradually on the glacier by a channel which the flow will
come to dig little by little.
In conclusion:
It is necessary to better know the effects of a brutal rupture in the valley of Ribon. The CEMAGREF
must continue simulations, of flow started in 2003.
Without expecting from them the results, directly related to the assumptions of rupture (volume,
hydrogramme badly known), it would be advisable to set up at least information of the public in the
valley of the Arc and Ribon. Within sight of the results of simulations, we will be able to propose to
the persons in charge for the roads a system of alarm, adapted to the characteristics of the flow.
Deliverable 4 87

Glaciers survey in High Savoy
A. Evans, Geologist - V. Tairraz, Forest Technician
ONF service RTM Annecy – Haute-Savoie
When in 2001 our partnership began with the CEMAGREF around the european project “Glaciorisk”
it was an opportunity to put together a protocole to follow 7 glaciers among the 17 principal glaciers
recorded as potential risk glaciers, in the department of High Savoy.
Surface Altitude maxi. Altitude mini.
Glacier du Tour 8,5 Km 2
3700 m 1800 m
Glacier d’Argentière 15 Km 2
3800 m 1400 m
Glacier de la Mer de Glace 3,5 Km 2
2200 m 1400 m
Glacier des Bossons 9,9 Km 2
4810 m 1200 m
Glacier de Taconnaz 1,7 Km 2
3200 m 1500 m
Glacier de Tête Rousse 0,12 Km 2
3260 m 2950 m
Glacier de Bionnassay 3,7 Km 2
4200 m 1750 m
Table 1. Glaciers observed
The aim was to focus on the variations of the ice-fronts of those 7 glaciers and the evolution of a proglacial
lake of the Glacier de la Mer de Glace.
1. The photographic follow-up
During the summer of 2001, the locations from which photographs of the glacier fronts would be
taken, were determined and there positions fixed.
In 2002, the photographic survey was carried out in good conditions.
In 2003, 2 out of the 7 glaciers did not have a proper photographic cover, due to various technical
problems : Glacier de Tête Rousse and Glacier de Bionnassay.
During the period 2001 - 2003 the evidence of the retreat of the glaciers under survey was confirmed
(cf. Fig 1. to Fig 5). The most significant results came from the surveys of the pro-glacial lakes on the
Glacier of the Mer de Glace.
88 Deliverable 4

Fig 1. Glacier du Tour
Photo from station n°2. Focale 93 mm
25/06/2001
04/09/2003
Deliverable 4 89

Fig 2. Glacier d’Argentière
Photo from station n° 2. Focale 32 mm,
90 Deliverable 4
27/09/2001
04/09/2002
06/11/2003

Fig 3. Glacier des Bossons
Photo from station n° 1. Focale 32 mm.
27/09/2001
18/09/2002
04/09/2003
Deliverable 4 91

Fig 4. Glacier des Bossons
Photo from station n° 1. Focale 93 mm.
92 Deliverable 4
27/09/2001
04/09/2003

27/09/2001 18/09/2002
04/09/2003
Fig 5. Glacier de Taconnaz
Photo from station n° 1. Focale 32 mm,
Deliverable 4 93

2. The pro-glacial lakes of the Glacier of the Mer de Glace
In 2001, when we decided to do a more precise study of the little lake on the Mer de Glace, we had
already been observing it for 2 years.
In the spring of 2002, during our first annual visit, we discovered that a second lake had ponded
upstream of the initial lake (surface area = 3614 m 2). . (Fig 6a. And 6b).
In september 2002, a bathymetric survey was undertaken on the lower lake. This resulted in a an
estimated maximum depth of 4,10 m and a volume of 12 000 m 3 .
In 2003 we continued our measurements of both lakes and noted an important increase in their surface
area, due to the very hot summer resulting in an significant melting of the glacier (Fig. 7).
Each year the state of the frontal dams of these lakes are examined, and up to this day they give no
alarming sign concerning their stability. The moraines are quite permeable allowing the excess waters
to flow out slowly.
Considering the rapid change that has taken place on this particular site, it seems a priority for us to
continue observing the development of these lakes.
One of our main objectives for 2004, is to continue the measurements and observations of these lakes,
and keep the local authorities informed of our surveys and results.
In the years to come, the hazards due to glaciers could be a real threat to the settlements in some of our
alpine valleys. For that reason it is an important mission for our service to continue the follow up
surveys of these glaciers.
Fig 6a. Glacier de la Mer de Glace - lower lake
Photo from station n° 1. Focale 93 mm, 25/06/2001.
Fig 6b. Glacier de la Mer de Glace lower and upper lakes
Photo from station n° 1. Focale 93 mm, 05/09/ 2003.
94 Deliverable 4

Fig 7. Glacier de la Mer de Glace.
Topographic survey of the lakes
Deliverable 4 95