Permafrost, Phillips, Springman & Arenson (eds)
© 2003 Swets & Zeitlinger, Lisse, ISBN 90 5809 582 7
The experimental simulation of ice-wedge casting
J.B. Murton
Department of Geography, University of Sussex, Brighton, UK
C. Harris
Department of Earth Sciences, University of Cardiff, UK
ABSTRACT: This paper describes a new methodology to simulate ice-wedge casting by scaled physical modelling
experiments in a geotechnical centrifuge. The purpose of the experiments is to determine the effect of host sediment
granulometry and ice content on (a) the mechanisms of casting and (b) the size and structure of casts. Six models
were constructed with host sediments comprising medium sand, loessic silt and silt-clay mixtures, and gravimetric
moisture contents ranging from ⬃15% to 60%. The models thawed from the surface downwards. Model ice wedges
were 150 mm high and tapered uniformly downward from a maximum true width of 50 mm, equivalent to prototypescale wedges 4.5 m high and 1.5 m wide. Model 6 data on temperature, pore-water pressure and drainage during
thaw, and on cast geometry after thaw demonstrate the potential of this approach to the study of ice-wedge casting.
1 INTRODUCTION
described in an unpublished study by T. Rudzinska
(1972, cited in Jahn 1975, pp. 72–73). Alternating layers
of gravel, sand and silt (⬃10–40 mm thick) were placed
in a small vessel and the sediment frozen. A wedgeshaped hole was cut in the middle of the frozen sediment, filled with water and the vessel again chilled,
producing an artificial ice wedge ⬃80 mm high and
with a maximum width of ⬃40 mm. Above the wedge
and host sediment was laid a sand layer ⬃10 mm thick.
The sediment was thawed and then refrozen, before cutting the sediment block in half to compare the resultant
ice-wedge cast with the size and shape of the initial ice
wedge. The experiment was repeated using two freezethaw cycles to determine the effect of repetitive freezethaw on the cast structure.
The experiments caused downturning of sediment
layers into the cast due to subsidence or collapse, producing a “cone-in-cone” structure. In addition, a sediment lining (⬃3 mm thick) formed around the lower
sides of one cast, the sediment deriving from within the
ice wedge. The cast was reported to be “larger and
deeper” than the former ice wedge, and Rudzinska concluded that the degree of sediment subsidence depends
on texture and the number of freeze-thaw cycles.
The approach taken in Rudzinska’s experiment
applies only at the small scale at which it was conducted. To apply the results to the field scale it is
essential to reproduce the correct self-weight stresses
associated with thaw of the upper few metres of soil.
This can be achieved either by carrying out full-scale
laboratory experiments or, more simply, by centrifuge
modeling in which 1/N scale models are thawed at
N times gravity. The centrifuge scaling laws applicable
to simulating the degradation of the upper metres of
permafrost are described by Harris et al. (2001a, b).
Ice-wedge casts are key stratigraphic indicators of former permafrost and are often used to reconstruct
palaeotemperatures (Huijzer and Vandenberghe 1998,
Murton & Kolstrup 2003). However, the mechanisms
of casting and their controlling factors have rarely
been studied and are difficult to monitor in the field
(Dylik 1966, Jahn 1975, Harry & Gozdzik 1988,
Murton & French 1993). Two key factors that that are
believed to influence casting are the particle size and
ice content of the host sediment (Black 1976). A new
approach to verify the influence of these factors is to
simulate casting by laboratory modelling experiments
and to compare the artificial casts with natural ones.
This approach permits (1) control of particle size and
ice content; (2) monitoring of temperature, pore pressure and cast development during thaw; and (3) measurement of the initial ice wedge and the final cast.
The objectives of this paper are to (i) describe the
new experimental methodology using scaled centrifuge
modelling, and (ii) discuss the results from one test in
order to demonstrate the potential of this new approach
to studies of ice-wedge casting. A companion paper
will describe the results from all six tests and evaluate
their significance to the cryostratigraphic record
(Harris and Murton in prep.). First we describe a previous experiment to simulate ice-wedge casting in order
to highlight the necessity of correct scaling of stresses
in thawing soil.
2 PREVIOUS EXPERIMENT
To our knowledge, only one previous laboratory experiment has attempted to simulate ice-wedge casting, as
807
Model ice wedges were 150 mm high and tapered
uniformly downward from a maximum true width of
50 mm. During thaw in the centrifuge at 30 gravities,
these were equivalent to prototype-scale wedges 4.5 m
high and 1.5 m wide. Six models were constructed to
1/30 scale (Table 1). Host sediments comprised medium
sand (models 1 and 2), natural loessic silt (models 3
and 4) and silt-clay mixtures (models 5 and 6).
Gravimetric moisture contents ranged from ⬃15%
(model 1) to 60% (model 6). Models 3 and 5 had moisture contents between the Plastic Limit and Liquid
Limit, whereas in models 4 and 6, moisture contents
exceeded the Liquid Limit.
Centrifuge tests lasted between approximately
8 hours (model 1) and 22.5 hours (model 6). Soil
drainage was allowed during thaw, except in model 2.
Temperature and pore-water pressure were recorded
at 10-second intervals on a data logger. Time-lapse
photographs were taken through the clear Perspex
front of the strongbox, and videos through the front
and top of the model in order to document the formation of ice-wedge casts. After complete thawing of the
wedge and host soil, the models were allowed to drain
for 2–4 weeks before they were vertically sectioned
and structures within them were measured, sketched
and photographed.
3 EXPERIMENTAL METHODOLOGY
The present experiments were carried out in the
Cardiff Geotechnical Centrifuge Facility. This comprises a beam centrifuge of radius 2.8 m capable of
accelerating a model package with a mass of up to
1000 kg at an acceleration of 100 gravities. The test
package is contained within a strongbox of external
dimensions ⬃750 mm 550 mm 500 mm built of
steel with Perspex sides.
Six ice-wedge models were constructed in a
polypropylene test box of internal dimensions 750 mm
long by 450 mm wide by 500 mm deep that fitted
closely inside the centrifuge strongbox. Freezing of the
model soil took place within a chest freezer, and the
models were thawed from the surface downwards in
the geotechnical centrifuge under an acceleration of
30 gravities (g). Thus, dimensional scaling during
thaw between model and prototype was 30.
A layer of sand ⬃20 mm thick formed the base of
each model. Above it were laid successive layers of
soil ⬃20–35 mm thick. The top of each layer was
carefully flattened and the layer was frozen before the
next was added. Marker layers 2–3 mm thick and of
different colour and/or texture to the soil layers were
placed between them in order to highlight any deformation structures formed during thaw. The soil layers
and markers were laid on both sides of an aluminium
mould with external dimensions identical to the internal dimensions of a similar mould used to form the
model ice wedges.
After the uppermost soil layer had been laid and
frozen, the mould was removed from the soil, leaving
a wedge-shaped trough into which a previously-frozen
model ice wedge was inserted. Any gaps between the
sides of the wedge and trough, usually no more than
1 to 2 mm wide, were filled with chilled water and
frozen prior to commencing thaw of the ice wedge.
Finally, a model active layer 25 mm thick was placed
above the top of the wedge and moistened with warm
water in order to expedite thaw of the model.
Instrumentation buried in the soil during model
construction comprised six miniature pore pressure
transducers (Druck PDCR-81, 350 mbar), and up to
14 Type K thermocouples. The transducers were
placed at 50 mm and 100 mm distances from the edge
of the ice wedge in order to determine if (1) excess
pore pressures resulted during thaw of the ice wedge
and host soil (2) whether these were associated with
lateral or vertical hydraulic gradients. Monitoring of
pore pressure also facilitates interpretation of soil
rheology and hence casting mechanisms during thaw.
The thermocouples permitted monitoring of the rate
of thaw and the pattern of isotherms adjacent to the
thawing ice wedge. The positioning of transducers
and thermocouples in model 6 is shown in Figure 1.
Figure 1. Vertical section through ice-wedge model 6,
showing location of pore-water pressure transducers and
thermocouples adjacent to the ice wedge, and of marker
horizons in the host silt-clay soil.
Table 1. Ice-wedge model parameters.
Model Grain size
Moisture
content* (%) Cover soil
1
2
3
4
5
6
15
20
40
20
30
60
medium sand
medium sand
Pegwell silt
Pegwell silt
2/3 silt, 1/3 clay†
2/3 silt, 1/3 clay†
sand
sand
Prawle silt
Prawle silt
2/3 silt, 1/3 clay†
2/3 silt, 1/3 clay†
*Approximate gravimetric moisture content. †Kaolinite.
808
4 RESULTS
4.2
To illustrate the results of these casting experiments, we
present some data on temperature, pore pressures and
cast geometry from model 6. This model simulated
thawing of ice wedges in ice-rich silt-clay host materials
(Table 1) with a silt-clay active layer. The ice content
in the host sediment was high, so the latent heat component of soil thawing was greater than in the other models, and the time for thaw was correspondingly long
(22.5 hours). Since time for seepage force similarity
and time for heat transfer both scale in the centrifuge as
1/N2, (where N is the number of gravities at which the
test is run), no scaling conflicts arise in modelling the
thaw-consolidation process, and model time is reduced
by a factor of N2 compared with prototype time (Harris
et al. 2000, 2001a, b). Thus at 30 g, the thaw period
scales to 843.75 days, or 2.3 years for the prototype.
Pore pressures generated during thaw consolidation
are illustrated in Figure 3 for transducers located at
frozen depths of 62.5 mm, 100 mm and 137.5 mm
below the surface (Fig. 2), scaling at 30 g to 1.875 m,
3.0 m and 4.125 m, respectively. During thaw, the
model surface lowered by 87.5 mm, equivalent to
some 41% of the model frozen thickness; there was no
ice within the cover soil, and much less settlement in
this layer than in the frozen host sediments. Thus,
although pore pressures approximated hydrostatic relative to the frozen model depth, at the point of thaw
and thereafter, the pore pressures recorded were
nearer geostatic than hydrostatic as a result of (a) low
initial soil density and (b) the reduction in overburden
thickness as consolidation progressed. Near-surface
pore pressures fell only slightly through the thaw
phase, reflecting continued upward migration of water
from below. Pore pressures at greater depths remained
high during the period when the adjacent ice wedge
thawed. Thus, it is likely that the frictional strength of
the soil was low, facilitating deformation as the void
left by the thawing wedge filled with soil from above.
4.1
Temperature
Since ice content of the host soil was 60% dry weight,
but the wedge consisted only of ice, the latent heat
required to thaw the wedge exceeded that of the soil,
slowing penetration of the thaw front above the ice
wedge compared with the adjacent host soil (Fig. 2).
Thaw of the model ice wedge was delayed by approximately 1.8 hours (scaling to approximately 67 days)
compared with soil 50 mm away from it (1.5 m at prototype scale). Thaw-consolidation processes would
therefore be expected to have at least partially dissipated excess pore pressures in the thawed soil by the
time the adjacent wedge had melted, stiffening the
soil, and reducing the potential for deformation around
the wedge void.
4.3
Pore pressures
Geometry of ice-wedge cast
The model ice-wedge cast from test 6 is shown in
Figure 4. The mean maximum true width was 42 mm
(std dev. 7 mm; n 14) and the mean height was 84 mm
(std dev. 4 mm; n 14), equivalent to prototype-scale casts
with a mean maximum true width of 1.26 m wide and a
mean height of 2.32 m. The model values represent a
narrowing of 17% relative to the initial maximum true
width of the ice wedge, and a shortening of 44% relative
60
A
Active Layer
B
C
40
Pore Pressure (kPa)
2 hours
A
B
14 hours
C
20
0
-20
-40
-60
18 hours
-80
0
Thaw front
TC
5
10
15
20
Time in hours (model scale)
25
Figure 3. Pore pressures recorded by three transducers
during test 6. Arrows indicate time of thaw. Transducer
locations are shown in Figure 2. Transducers A and B were
50 mm from the sides of the ice wedge, transducer C
100 mm.
TC and PWP
Figure 2. Thaw-front geometry interpolated from thermocouple data, test 6. TC thermocouples, PWP porepressure transducers. Ice wedge is shaded.
809
coefficient. The resulting cast geometry also differed
markedly, reflecting contrasting degrees of host soil
deformation.
Clearly, we are unable to represent precisely the
regime of permafrost degradation associated with icewedge casting, since this process may occur over many
years, with seasonal surface freezing punctuating progressive downward thaw or with complex histories
involving, for example, partial ice-wedge thaw and
renewed permafrost aggradation (Murton and Kolstrup
2003). However, the approach taken here does allow us
to investigate the effects of host soil geotechnical properties, and provides insights into the factors influencing
the geometry of Pleistocene ice-wedge casts.
ACKNOWLEDGEMENT
The research was funded by a grant from the Royal
Society.
REFERENCES
Black, R.F. 1976. Periglacial features indicative of permafrost: ice and soil wedges, Quaternary Research 6: 3–26.
Dylik, J. 1966. Problems of ice-wedge structures and frostfissure polygons, Biuletyn Peryglacjalny 15: 241–291.
Harris, C. & Murton, J.B. in prep. Scaled centifuge modeling of ice-wedge casting.
Harris, C., Murton, J.B. & Davies, M.C.R. 2000. Softsediment deformation during thawing of ice-rich
frozen soils: results of scaled centrifuge modelling
experiments. Sedimentology 47: 687–700.
Harris, C., Rea, B. & Davies, M.C.R. 2001a. Geotechnical
centrifuge modelling of gelifluction processes: validation of a new approach to periglacial slope studies.
Annals of Glaciology 31: 263–269.
Harris, C., Rea, B. & Davies, M.C.R. 2001b. Scaled physical modelling of mass movement processes on thawing slopes. Permafrost and Periglacial Processes 12:
125–136.
Harry, D.G. & Gozdzik, J.S. 1988. Ice wedges: growth,
thaw transformation, and palaeoenvironmental significance, Journal of Quaternary Science 3: 39–55.
Huijzer, B. & Vandenberghe, J. 1998. Climate reconstructions of the Weichselian Pleniglacial in northwestern
and central Europe. Journal of Quaternary Science
13: 391–418.
Jahn, A. 1975. Problems of the Periglacial Zone (Zagadnienia strefy peryglacjalnef). Warsaw: Panstwowe
wydawnictwo Naukowe.
Murton, J.B. & French, H.M. 1993. Thaw modification of
frost-fissure wedges, Richards Island, Pleistocene
Mackenzie Delta, western Canadian Arctic, Journal of
Quaternary Science 8: 185–196.
Murton, J.B. & Kolstrup, E. (2003, in press). Ice-wedge
casts as indicators of palaeotemperature: precise
proxy or wishful thinking? Progress in Physical
Geography 27.
Figure 4. Vertical sections at 20 mm (a) and 100 mm (b)
through the ice-wedge cast in model 6. The central part of
the cast comprises grey cover soil that descended into the
void left by the thawing ice wedge. Downturned marker
horizons extend into the cast.
to its initial height. Above the cast was a trough whose
mean maximum width was 214 mm (std dev. 29; n 14)
and mean maximum depth was 16 mm (std dev. 1 mm; n
16). Adjacent to the cast was a crack whose mean width
was 18 mm (std dev. 4 mm; n 12) and whose mean depth
was 113 mm (std dev. 7 mm; n 10).
5 CONCLUSIONS
Scaled centrifuge modelling facilitates detailed investigation of the behaviour of thawing soils. In this
example we have simulated the thawing of an ice
wedge, and the morphology of the resulting ice-wedge
cast. In the case described here, pore pressures generated by thaw-consolidation processes were high, and
reduced the strength of the thawed host soil sufficiently for significant deformation during casting. In
earlier tests with host soils of differing ice content and
granulometry, thaw-consolidation ratios differed significantly, reflecting different rates of thaw-front penetration and different values of the consolidation
810