Bedrock Fracture by Ice Segregation in Cold Regions

See allHide authors and affiliations

Science  17 Nov 2006:
Vol. 314, Issue 5802, pp. 1127-1129
DOI: 10.1126/science.1132127


The volumetric expansion of freezing pore water is widely assumed to be a major cause of rock fracture in cold humid regions. Data from experiments simulating natural freezing regimes indicate that bedrock fracture results instead from ice segregation. Fracture depth and timing are also numerically simulated by coupling heat and mass transfer with a fracture model. The depth and geometry of fractures match those in Arctic permafrost and ice-age weathering profiles. This agreement supports a conceptual model in which ice segregation in near-surface permafrost leads progressively to rock fracture and heave, whereas permafrost degradation leads episodically to melt of segregated ice and rock settlement.

The fracture of bedrock is fundamental to debris production and landscape development. Rock fracture in polar and alpine regions has often been attributed to the freezing and volumetric expansion of water trapped within pores and cracks (1). An alternative process of bedrock fracture, involving ice segregation, remains poorly characterized despite a number of theoretical and experimental studies over the past two decades (28). Ice segregation occurs when temperature gradient–induced suction in freezing or frozen ground drives unfrozen water—held in capillaries and adsorbed on the surfaces of mineral particles—through a porous medium, such as soil, toward freezing sites where lenses or layers of ice grow. If ice segregation fractures bedrock permafrost, the fractures and ice lenses are expected to concentrate in wet, porous rock just beneath the topofthepermafrostandinthebaseoftheactive layer (3, 57), as they do in porous silty soils (9, 10). Here, we report results from experiments that test this hypothesis and elucidate the ice-segregation process in bedrock.

We developed an experimental methodology in a pilot study (5) that instrumented a block of chalk; the block's lower half was maintained at temperatures below 0°C (simulating permafrost) and its upper half cycled above and below 0°C (simulating seasonal thawing and freezing of the overlying active layer). We then performed systematic experiments with 10 substantially larger chalk blocks that had different moisture contents in order to simulate and monitor bidirectional freezing of a bedrock active layer above permafrost and unidirectional freezing of seasonally frozen bedrock (11). Bidirectional freezing refers to the combination of upward freezing from the permafrost table and downward freezing from the rock surface, whereas unidirectional freezing refers simply to the latter. The results validated those of the pilot study (6) and revealed discrete stages of micro- and macrocrack development. Based on numerical simulations of the depth and timing of rock fracture, we propose a conceptual model of ice segregation, fracture development, and surface stability in porous bedrock permafrost.

The experiments demonstrate that ice segregation fractures wet chalk. We discounted volumetric expansion as a cause of fracture because measured rock saturation levels at the depths of fracture in the bidirectional experiments did not exceed ∼65% when macrofracture growth commenced, indicating that fracturing occurred at moisture levels substantially below the predicted threshold saturation level of ∼91% (12). In addition, sustained periods of rock-surface heave during thawing cycles (Fig. 1) are inconsistent with bursts of crack growth predicted by volumetric expansion during freezing cycles, whereas they are expected with sustained periods of ice segregation (12). Thaw-related rock heave (Fig. 2) is analogous to summer frost heave in Arctic permafrost soils (13), when unfrozen water migrates down into underlying frozen ground, permitting ice-lens growth in still-frozen soil beneath the thawing front. Fractures formed by ice segregation should generally be parallel to the cooling surfaces (14), as we observed. All but the thinnest fractures in the artificial permafrost contained visible segregated ice (Fig. 3, A and C).

Fig. 1.

Rock-surface heave and air temperature during 21 cycles of bidirectional freezing of blocks B1 and B2 (11). Air temperatures above 0°C correspond with thawing cycles, and those below 0°C correspond with freezing cycles. The data are 4-hour running means. Thresholds in the heave data at day 150 for block B1 and day 370 for block B2 mark the change from micro- to macrocracking, and the onset of development of an ice-rich fractured layer of permafrost. Thaw settlement of nearly 10 mm at the surface of B1 (day 283) coincided with the European heat wave of July and August 2003.

Fig. 2.

Rock temperature in the lower part of the active layer (150- and 200-mm depths) and the upper part of the permafrost (250- and 300-mm depths), and rock-surface heave during thaw cycle 10, block B1. The active layer attained a depth of 241 mm on day 167.

Fig. 3.

Fractures and segregated ice within 450-mm-high blocks of chalk subjected to bidirectional freezing (A to D) and unidirectional freezing (E to H). The photographs show vertical sections through frozen blocks at the end of the experiments, with lenses or layers of segregated ice visible in many cracks. The box plots show fracture depths and, for the bidirectional experiments, the base of the active layer (the depth of maximum seasonal penetration of the 0°C isotherm into the rock). The red boxes show the interquartile range, the horizontal line within the boxes is the median, and the whiskers extend to the highest and lowest values. In the unidirectional experiments, the depth of maximum penetration of the 0°C isotherm exceeded 350 mm during each freezing cycle.

The freezing regime determines the depth of rock fracture. Bidirectional freezing results in fracture of the upper layer of permafrost and the base of the active layer, whereas unidirectional freezing results in fracture close to the surface (Fig. 3). This distinction reflects seasonal differences in the directions of heat and water transfer. Unidirectional (downward) freezing produces an upward transfer, favoring ice segregation near the rock surface. Bidirectional freezing causes transfer from the center of the active layer down toward the permafrost and up toward the rock surface. The concentration of segregated ice in the upper layer of permafrost and the base of the active layer results in part from downward transfers associated with the slower rate of upward freezing, allowing more time for water migration. But the major factor, as indicated by the seasonal timing of rock heave (Fig. 1), is summer ice segregation beneath the permafrost table.

Fractures within permafrost develop in two stages. We attributed gradual progressive heave during initial seasonal temperature cycles to microcrack development (Fig. 1). A threshold is then crossed, after which rapid heave commences during thawing cycles (summers), recording the rapid growth of segregated ice in macrocracks. During these subsequent seasonal temperature cycles, ice segregation forms an ice-rich layer of fractured rock just below the permafrost table (Fig. 3A). The icy layer is sensitively dependent on summer air temperatures, rapidly melting during the European heat wave of summer 2003 (∼day 283 on Fig. 1), but re-forming afterward.

We performed numerical simulations of the experiments by coupling a description of transient heat and mass transfer in porous freezing media with the segregation ice–induced fracture model of Walder and Hallet (2). Assuming an initially uniform distribution of very small (5-mm) defects throughout the block (11), the simulations successfully predict the approximate depth of maximum cracking for both the unidirectional (Fig. 4A) and bidirectional (Fig. 4B) freezing experiments. Notably smaller cracks are predicted during unidirectional freezing. The simulations also provide insight into the interdependence of temperature, temperature gradient, and pressure within a growing microcrack, and the conditions necessary for eventual fracture. The three conditions necessary for crack growth due to ice segregation are a subzero (°C) temperature gradient, to cause a thermomolecular pressure gradient; a warm (slightly subzero) temperature, for sufficient water permeability; and an intracrack pressure moderately above the stress-corrosion limit (2). When all three conditions are met simultaneously, cracks that begin small can lengthen almost exponentially (Fig. 4C). During cyclic freeze-thaw events, the pressure within a crack also cycles, increasing when water is drawn inside by the thermomolecular pressure gradient and frozen, and decreasing when ice thaws and pressurized water is forced out (Fig. 4D). As a crack grows, the pressure required to lengthen it further decreases as a result of the increasing stress intensity factor (the same phenomenon causes small windshield cracks to grow quickly). The rapid increase in rock-surface heave around 150 days in Fig. 1 correlates well with the predicted time that a crack at 300-mm depth would reach 300 mm in length, effectively fracturing the block in two and allowing for relatively unrestrained ice accumulation within the fracture. Although the exact time for complete fracture depends on the initial defect size (5 mm), order-of-magnitude variations in initial size (0.5 to 50 mm) still lead to complete fracture within the 125- to 175-day range. The model results agree with the theory of Walder and Hallet (2) and also provide insight into the effects of a cyclic thermal regime (supporting online material text).

Fig. 4.

Numerical model predictions showing the final crack length at 150 days as a function of vertical location of the initial crack for (A) unidirectional freezing and (B) bidirectional freezing. During bidirectional freezing, an initial crack at a depth of 300 mm begins to rapidly lengthen around day 75 (C). Complete fracture denotes the time when the crack length becomes equal to the block width (300 mm). The intracrack pressure during bidirectional freezing fluctuates between a maximum during ice accumulation and a minimum after thaw (D).

The layer of ice-rich fractured chalk near the top of the artificial permafrost (Fig. 3A) resembles ice-rich fractured limestones, shales, and sandstones in Arctic permafrost. The natural fractures have been attributed to both thermal contraction cracking (15) and ice segregation (14, 16), and the ice has been variably interpreted as hoar frost (15) or segregated ice (14, 16). We discounted thermal contraction cracking because, in frozen soils, it produces large widely spaced vertical cracks and ice wedges (17), whereas the majority of fractures in the artificial and natural permafrost are small, closely spaced, and horizontal, consistent with horizontal cooling surfaces during ice segregation. The experimental fracture patterns also resemble those in weathering profiles from regions where ice-age permafrost previously existed. Fractures of the upper ∼0.5 to several meters of bedrock are abundant in fine-grained porous rocks such as the chalk in northern France (18) and southern England (19), the muschelkalk (shelly limestone) of central Germany (15), and various limestones, mudstones, and clays in the United Kingdom (19). In all cases, the fractures are mostly horizontal to subhorizontal, and they decrease in number density as depth increases. We could not discount the possibility, however, that some horizontal cracks result from other weathering processes or preexisting weaknesses, such as bedding planes.

The agreement between modeling and field observations supports the hypothesis that ice segregation preferentially fractures porous bedrock near the permafrost table and contributes substantially to a more precise understanding of the ice-segregation process in bedrock. The wider importance of this process to geomorphology, Quaternary science, engineering, and climate change concerns the dynamics of the ice-rich fractured layer in the upper meters of permafrost. Ice segregation leads progressively to rock fracture, accumulation of ground ice, and rock heave, whereas active-layer deepening leads episodically to melt of ground ice and rock settlement. The interplay between the buildup and decay of the ice-rich layer operates on a number of time scales. During unusually hot summers, the development of this layer is interrupted by thaw settlement (Fig. 1), just as in ice-rich permafrost soils (20), when the top of the permafrost thaws and supplies rock fragments to the active layer. The fragments are then available for sorting into patterned ground (21), comminution by weathering, or downslope transport by mass wasting, fluvial activity, or glaciers. During climate warming at the end of Quaternary glacial or stadial periods, the ice-rich layer in mid-latitude regions such as the chalklands of England and France melted completely, triggering thaw settlement of the ground surface, deformation of soil and rock, and enhanced mass wasting (22). With recent climate warming predicted to continue during the next century, amplified in Arctic and sub-Arctic regions (23), the warming and thawing of ice-rich bedrock permafrost is likely to destabilize many rock substrates. Such issues will undoubtedly become important to scientists, engineers, and inhabitants of permafrost regions.

Supporting Online Material

Materials and Methods

SOM Text

Fig. S1

Table S1


References and Notes

View Abstract

Navigate This Article