Research Article

Transportation Research Record

1–12
Ó National Academy of Sciences:
Analysis of the Stability of Thawing Transportation Research Board 2019
Article reuse guidelines:
Slopes by Random Finite Element sagepub.com/journals-permissions
DOI: 10.1177/0361198119850805

Method journals.sagepub.com/home/trr

Shaoyang Dong1 and Xiong (Bill) Yu1

Abstract
A significant number of landslides occur in cold regions because of freezing and thawing cycles. The instability of thawing
slopes can cause serious damage to transportation infrastructure and property, and even loss of human life. This type of land-
slide is difficult to analyze by the traditional limit-equilibrium methods, however, because of the complicated multi-physics
processes involved. This paper describes a holistic microstructure-based random finite element model (RFEM) to simulate
the stability of a thawing slope. The RFEM model is developed to simulate the bulk behaviors of frozen and unfrozen soils
based on the behaviors of individual phases. The phase coded image of a frozen silty clay is first custom built and then con-
verted into a finite element model. The mechanical behaviors of individual phases of the frozen soil are calibrated by uniaxial
compressive test. The triaxial test is then simulated by RFEM to obtain the shear strength parameters of frozen and unfrozen
soils. Coupled thermal-mechanical REFM models are developed to simulate the effects of temperature on the displacement
field and stress field in the slope. From the results, the local factor of safety field can be determined. The development of local
factor of safety and potential failure surface associated with the thawing process over a typical year are simulated by this new
model. The variations in the stability of thawing slopes predicted by this model are consistent with field observations as well
as the global-wise slope stability analysis.

In recent decades, both permafrost regions and season- model to simulate the change of soil strength during the
ally frozen ground areas in the northern hemisphere have freezing and thawing process.
experienced significant warming trends because of cli- Slope stability analysis is generally performed by con-
mate change (1). High altitude and high latitude areas ventional limit-equilibrium methods (15–18). These
such as the Alps in Europe, the Qinghai-Tibet Plateau in methods have been applied extensively for the analysis of
China, northern Canada, and Alaska have experienced general slope stability in engineering practice. The limit-
accelerating rates of degradation of permafrost and sea- equilibrium methods may not be effective on the stability
sonally frozen soil (2–5). The melting of frozen soil has of thawing slopes, however, because, from field observa-
triggered landslides, mudflows, and rock falls in cold tions, the failure surface of a thawing slope is parallel to
regions. These geological disasters have reshaped the the slope surface, which is different from that of land-
landscape and the environment, destroyed local trans- slides which generally have a curved sliding surface (8,
portation infrastructure and property, and even caused 19, 20). The thaw slumping is relatively less thick com-
loss of human life (6–10). pared with the slumping observed in other slopes, with
The mechanical behaviors of frozen and unfrozen soils the thickness determined by the soil temperature and the
have been studied in the past few decades. The stability thawing depth. Therefore, limit-equilibrium methods
problems of frozen/unfrozen soils are often controlled by provide little insight into the location of the failure sur-
shear strength parameters (11). Graham and Au (12) and face of a thawing slope.
Leroueil et al. (13) reported that the undrained shear
strength of natural soil samples decreased significantly
1
because of the freeze–thaw effect. Other researchers con- Department of Civil Engineering, Case Western Reserve University,
Cleveland, OH
ducted numerical simulations to study the mechanical
behaviors of frozen and unfrozen soils. Zhang and Corresponding Author:
Michalowski (14) developed a thermal-hydro-mechanical Address correspondence to Xiong (Bill) Yu: [email protected]
2 Transportation Research Record 00(0)

interesting phenomena of the stability of slopes in cold

regions.

Theoretical Background
Local Factor of Safety
The local factor of safety is defined based on the com-
parison of the local stress condition to the soil strength
(Figure 1). The factor of safety is defined as the ratio of
the distance of stress centroid to the strength curve (CB)
to the radius of the Mohr circle (CD). The mathematical
relationship is given in Equation 1. The value of local
factor of safety is a scalar function dependent upon the
stress conditions at a local point. The definition of local
Figure 1. Illustration of local factor of safety from Mohr circles factor of safety utilizes a scalar function to indicate the
of a given point P (x, y). stability of each location in the slope. The contour of
local factor of safety gives the areas sharing similar likeli-
To overcome this limitation, the local factor of safety hood of slope failure. Figure 1 and Equation 1 illustrate
and random finite element methods are combined to the concept and derivation of local factor of safety at a
simulate the stability of thawing slopes (21–24). For given point P (x, y):
example, Griffiths and Fenton (25) treated shear strength
as a single random variable to analyze the stability of a t  ðx, yÞ jBC j jBEj + jEC j
LFS(x, y) = = =
slope. The local factor of safety of a slope is defined as tðx, yÞ jDC j jDC j
the ratio of the local shear strength to the local shear s1 + s3
sin u + c  cos u
2
stress of a particular element in the slope. Local factor of = s1 s3 ð1Þ
2
safety is a scalar field, a function of position in the slope.
The random finite element method adds randomness to where c and u are cohesion and the angle of internal fric-
characterize the inherent variabilities of soil and geologi- tion, respectively, of frozen or unfrozen soils (depending
cal features, which is an improvement over the traditional upon the thermal condition). They are also called shear
finite element method (26–29). As is generally known, strength parameters of soils. s1 and s3 are the maximum
frozen soil is a four-phase material containing solid parti- and minimum principal stress of soils, respectively, which
cles, ice, water, and air. These phases have their corre- is obtained from the finite element analysis.
sponding mechanical and thermal properties (such as The stress state at any point in the slope has a shear
Young’s modulus, Poisson’s ratio, density, thermal con- strength t * and shear stress t. The ratio of shear strength
ductivity, heat capacity, and coefficient of thermal expan- to shear stress is defined as local factor of safety, which
sion) and are randomly distributed in the thawing slope. indicates how far the current shear stress state is from
The local factor of safety determined by the random failure. As illustrated in Figure 1, if Mohr circle is sepa-
finite element method makes it possible to trace the stress rate from the Mohr-Coulomb failure envelope, t*/ t is
state and development process of failure surface along greater than 1, which indicates that the point in the slope
the thawing slope. is currently stable. If the Mohr-Coulomb failure envelope
This paper describes the development and implemen- is tangent or intersects with Mohr circle, t */ t is less than
tation of a random finite element model (RFEM) for or equal to 1, and the point in the slope is unstable. The
analyzing the stability of sequentially thawing slopes. local factor of safety in thawing slopes is not only a scalar
The results are compared with the traditional limit- function of position, but also a function of time and tem-
equilibrium method. The spatial randomness of soil para- perature. For example, during the thawing process, the
meters is described with this new model. The mechanical shear strength of the soil would decrease and it is affected
parameters of frozen soil and unfrozen soil are simulated by the temperature (30–34). On the other hand, the inter-
by the RFEM model and are calibrated by the experi- nal stress in the slope would increase during the freezing
mental data. The seasonal responses of the slope to tem- process because of matric suction and ice expansion. The
perature variations, including the slope deformation and local factor of safety has the advantages of providing the
the local factor of safety distribution of the slope are location of initial failure and allows to describe how the
calculated for a whole year period. The results unveil region of instability evolves with time and temperature.
Dong and Yu 3

Table 1. Physical Parameters and Uniaxial Compressive Testing Results of Silty Clays

Constant Value Units Description

R 1.65 cm Radius of soil specimen
H 7.2 cm Height of soil specimen
w0 0.15 1 Initial water content
r 1708 kg/m3 Dry density of soil specimen
Gs 2.65 1 Specific gravity of soil specimen
Efrozen 20 MPa Young’s modulus of frozen soil specimen
Eunfrozen 10.2 MPa Young’s modulus of unfrozen soil specimen
sc,frozen 775 kPa Unconfined compressive strength of frozen soil specimen
sc,unfrozen 375 kPa Unconfined compressive strength of unfrozen soil specimen

Determination of Shear Strength

Parameters for Frozen and Unfrozen Soils with black, dark gray, light gray, and white to represent
soil particles, ice, water, and air, respectively. At a partic-
To determine the local factor of safety and analyze the ular pixel element of the matrix, the value of the element
stability of thawing slope, the information of shear (or the phase of the element) is determined by a value
strength parameters (cohesion and the angle of internal representing the phase times—a random number whose
friction) of frozen and unfrozen soils are needed. For this probability of occurrence is equal to the volumetric con-
purpose, RFEM model simulations are conducted on tent of each phase. For the four-phase soil specimen, the
special cylindrical shaped soil sample. The soil is assumed phase value protocol is set 0, 1/3, 2/3, 1 for soil particle,
to be silty clay. The general procedures used to obtain these ice, water, and air respectively. This process was repeated
parameters include: Firstly, the uniaxial compressive test for each pixel element of the m 3 n matrix. Consequently,
was conducted to obtain the Young’s modulus and com- the m 3 n matrix contains the phase coding that repre-
pressive strength of the frozen or unfrozen silty clays. The sents the volume proportion of different phases, which
RFEM was then generated based on the phase composi- can be visually represented as a grayscale image. With
tion and was applied to simulate the responses of the digi- the procedure described, the probability of occurrence
tal specimen under uniaxial compressive test. The results or the percentage of each phase is approximately equal to
are compared with the experimental curve. From the the volumetric content of each phase when the m 3 n
stress–strain curve, the mechanical parameters of each matrix is sufficiently large. The phases in the soil speci-
phase in frozen/unfrozen soils were calibrated. The exam- men are assumed to be uniformly random distribution
ple results are summarized in Table 1. Finally, random because, in the experiment, the soil specimen is prepared
finite element simulation was conducted to simulate the by Harvard miniature compactor, and soil particles and
triaxial test of frozen/unfrozen soils under different confin- water are uniformly mixed to form the soil specimen.
ing pressures. The Mohr-Coulomb failure envelope and Comsol is a finite element software which supports
the shear strength parameters were obtained from stress importing images and assigning material properties based
conditions that failed based on the simulation results. on the color scale of the image. The phase is color coded
The RFEM is implemented to simulate the behaviors in Comsol with dark blue, light blue, yellow, and red to
of the frozen/unfrozen soils. The volume portions of dif- represent soil particles, ice, water, and air, respectively.
ferent phases are incorporated in producing the RFEM The two-dimensional phase coded frozen and unfrozen
model based on the physical properties of the soil speci- soil models generated from Matlab are converted into
men prepared by Harvard miniature compactor. To pro- Comsol to simulate the uniaxial test (Figure 3b) and
duce the microstructure of frozen/unfrozen soils, the triaxial test (Figure 3c). As shown in Figure 3b, the
volumetric content of each phase is firstly calculated RFEM is applied to simulate the uniaxial test, the fixed
from its dry density, initial gravimetric water content, constraint is applied at the bottom of the phase coded
and specific gravity of the soil specimen. This informa- frozen soil model, and the pressure load s1 is exerted on
tion is used to produce a matrix of the soil image. The top of the model. The mechanical parameters are
image is characterized by a m 3 n matrix grayscale image assigned to each pixel based on the color scale of the
in Matlab. In the matrix, m is set equal to the height of image. By matching the simulated stress–strain curve
the soil specimen divided by the size of the soil particle, with the one obtained from the experiment (Figure 2),
while n is set equal to the radius of the soil specimen the mechanical parameters of each phase are calibrated
divided by the diameter of the soil particle. Thus, m 3 n and the calibration results are listed in Table 1.
equals the total number of pixel elements in the image of The triaxial test was simulated in Comsol with the
two-dimensional soil model. The phase is color coded RFEM phase coded model. The calibrated parameters
4 Transportation Research Record 00(0)

Figure 2. Comparison of stress–strain relationship between

experimental measurement and numerical simulation of a totally
frozen soil.

are assigned to each pixel based on the color scale of the

image. As shown in Figure 3c, the fixed constraint is
applied at the bottom of the unfrozen soil model. The
confining pressure s3 and the compressive pressure s1
are exerted on the sides and the top of the soil model,
respectively. The deviatory stress increases until it reaches Figure 3. Uniaxial testing: (a) experimental setup; (b) RFEM
a state where the numerical solution is unstable, which is phase coded model for uniaxial test of frozen soil; (c) RFEM phase
used to indicate failure. The stress condition at failure is coded model for triaxial test of unfrozen soil. The grayscale image
recorded and analyzed. The simulations were repeated is generated from Matlab and the color image is the converted
version from Comsol.
for different magnitudes of confining pressures.
The simulated Mohr circles and the Mohr-Coulomb
failure envelopes for frozen soil and unfrozen soil are
plotted in Figure 4. From the Mohr-Coulomb failure circular sliding surface. The interslice shear force is not
envelopes, the shear strength parameters can be deter- considered by this method.
mined from the linear portion of the curve. For the fro- The height of the slope is assumed to be 4.5 m with
zen soil, cohesion and the angle of internal friction are slope gradient of 1:2. The soil in the slope is assumed to
determined to be 150 kPa and 30 degrees, respectively. be the same soil used in the experiment and the RFEM
For the unfrozen soil, cohesion and the angle of internal numerical simulation. Its physical and mechanical prop-
friction are 37 kPa and 20 degrees, respectively. The erties are listed in Table 1. It is assumed that the slope is
simulated shear strength parameters of frozen/unfrozen initially totally frozen, and the stability of the permafrost
soils were compared with the tested results, and good slope is calculated by applying the shear strength para-
agreement was obtained (34). The simulation results indi- meters of frozen soil. After that, the slope is assumed to
cate that the shear strength of the frozen soil decreases begin to thaw. To simplify the analyses, it is assumed that
after thawing, which is consistent with observations from thawing is in the vertical direction only. The slope is hori-
prior research and engineering practice (18, 35). zontally divided into five layers of equal dimensions for
the convenience of describing different thawing depth.
The slope is thawed gradually from the top layer to the
Stability of Thawing Slope Analyzed by bottom layer by sequentially replacing the strength para-
meters of the corresponding layer of frozen soil with that
Simplified Bishop Method of the unfrozen soil. The calculation results of the global
The simplified Bishop method is one of the popular limit- factor of safety at different extent of thawing by the simpli-
equilibrium methods for slope stability analysis. This fied Bishop method are listed in Figure 5. As shown in
method satisfies the vertical force equilibrium for each Figure 5, the slope is most stable under the complete frozen
slice and moment equilibrium about the center of the condition with a safety factor of 1.671. The factor of safety
Dong and Yu 5

Figure 4. Mohr circles and the Mohr-Coulomb failure evenlopes for RFEM simulations for (a) frozen soil and (b) unfrozen soil.

Figure 5. The stability of thawing slope calculated by simplified Bishop method.

continuously decreases during the thawing process; at the slope and the failure surface are affected by the tempera-
totally thawed state it is 0.9 which indicates slope failure. ture distribution inside the slope, which is not considered
This simplified case indicates that shear strength para- in the conventional slope stability analyses. The distribu-
meter plays an important role in determining the safety tion of stress and shear strength inside the thawing slope
factor and the stability of the slope. The thawing slope evolve with time and temperature. These characteristics
gradually lost its stability because of the reduction of the are not incorporated by the simplified Bishop method or
shear strength when frozen soil melted to unfrozen soil. other limit-equilibrium methods. New methods of analy-
This method, however, has some drawbacks in analyzing sis and simulation tools are necessary to simulate the sta-
the stability of thawing slopes. First, it is assumed that bility of thawing slopes holistically. A new RFEM
the failure surface of the thawing slope is circular with simulation method based on the the local factor of safety
the location of the circular surface predetermined. For is introduced to simulate stability conditions of thawing
most thawing slopes in reality, however, the failure sur- slope considering the temperature distribution associated
face is not circular. The deformation of the thawing with weather conditions.
6 Transportation Research Record 00(0)

Table 2. Constant Parameters for RFEM Simulation of Thawing Slopes

Constant Value Unit Description

W 15 m Horizontal distance of the study region

H 4.5 m Height of the slope
SR 1:2 1 Slope ratio
w0 0.15 1 Initial water content
r 1708 kg/m3 Dry density of soil specimen
d 50 mm Diameter of soil particle
Gs 2.65 1 Specific gravity of soil specimen
Es 10.2 GPa Young’s modulus of soil particle
Ei 9 GPa Young’s modulus of ice
Ew 3.9310-5 Pa Young’s modulus of water
Ea 0 Pa Young’s modulus of air
ms 0.3 1 Poisson’s ratio of soil particle
mi 0.3 1 Poisson’s ratio of ice
mw 0.5 1 Poisson’s ratio of water
ma 0 1 Poisson’s ratio of air
rs 2600 kg/m3 Density of soil particle
ri 917 kg/m3 Density of ice
rw 1000 kg/m3 Density of water
ra 1.29 kg/m3 Density of air
ks 2 W/mK Thermal conductivity of soil particle
ki 2.2 W/mK Thermal conductivity of ice
kw 0.58 W/mK Thermal conductivity of water
ka 0.025 W/mK Thermal conductivity of air
Cps 835 J/kgK Heat capacity of soil particle
Cpi 1960 J/kgK Heat capacity of ice
Cpw 4181.3 J/kgK Heat capacity of water
Cpa 1005 J/kgK Heat capacity of air
Lf 334 kJ/kg Latent heat fusion of water
as 9310-6 1/K Volumetric coefficient of thermal expansion of soil particle
ai 1.9310-4 1/K Volumetric coefficient of thermal expansion of ice
aw 2.07310-4 1/K Volumetric coefficient of thermal expansion of water
aa 3.43310-3 1/K Volumetric coefficient of thermal expansion of air

Note: RFEM = random finite element model.

Stability of Thawing Slope by RFEM

The slope is assumed to have the same dimensions and
properties as that analyzed by the simplified Bishop
method. The phase coded image of the frozen slope is
generated in Matlab based on the physical constituents
of the soil and the properties of individual phases. The
image is then converted into Comsol. The corresponding
physical, thermal, and mechanical properties (Table 2) of
each phase are assigned based on the phase coding. For Figure 6. Computational domains of the slope with finite
the thermal boundary conditions, thermal insulations are element model mesh and annotated probe locations.
applied on the right side and the bottom of the slope.
The initial temperature of the study region is assumed to the temperature amplitude and its value is 16°C. B is the
be –17°C. The boundary temperature is applied on top initial phase and its value is  p=2, and t is time in
of the slope according to the following trigonometric for- the unit of day. For the mechanical boundary condi-
mula which describes the air temperature variations over tions, the fixed constraints are applied at the bottom of
a one-year period (36): the slope and roller constraints are applied on the right
side of the slope. Figure 6 illustrates the computational
2p domains with finite element model mesh.
T(t) = T0 + A0 sin( t + B) ð2Þ
365 The computational simulations are conducted over
where T0 is the average temperature of the slope surface 365-day period. The temperature field and mechanical
and its value is 1°C based on the climate data (36). A0 is field of the slope are coupled together by setting the
Dong and Yu 7

Figure 7. Contour plot of horizontal displacement (mm) in the slope during the thawing/freezing process.

thermal and mechanical parameters of the soil as func- of frozen soil). If the temperature of a soil element is
tions of temperature. Latent heat is applied as equivalent above 0°C, then parameters of unfrozen soil are assigned.
to heat capacity at a small temperature transition zone. Examples of the simulation results are summarized in
When the temperature of a particular soil element is less Figures 7 to 10.
than or equal to 0°C, phase transition is assumed to Figure 7 shows the development of horizontal displa-
occur. The corresponding parameters of frozen soil are cement in the slope. As seen from this figure, the hori-
applied to the soil element (i.e., thermal and mechanical zontal displacement in the slope continues increasing
parameters for the ice phase; shear strength parameters from February 1 to June 1 during the thawing process.
8 Transportation Research Record 00(0)

Figure 8. Temperature distribution in the slope during the thawing/freezing process.

The maximum displacement increases from 22.4 mm on evolving with the penetration of the thawing front. The
February 1 to 112 mm on June 1. During this period, the failure surface of the slope is parallel to the thawing front
temperature on the slope surface rises from –11.3°C to and generally parallel to the slope surface. This phenom-
15.2°C and the corresponding thawing depth continues enon agrees with field observations (8). It should be
penetrating below the slope surface (Figure 8). The shear noted that the contour plot of the local factor of safety
strength decreases in the melting soil elements where the and thaw depth penetration lines are not uniformly dis-
temperature rises above 0°C. The local factor of safety tributed because of the microstructure of the soil. The
drops below 1, indicating failure in the corresponding overall distribution patterns, however, are generally par-
soil element (Figure 9). The failure of the slope continues allel to the slope surface.
Dong and Yu 9

Figure 9. Local factor of safety distribution in the slope during the thawing/freezing process.

From August 1 to December 1, as the season transi- Figure 10 shows the variations of local factor of safety at
tions to the freezing season, the horizontal displacement four locations inside the slope. Probe no. 3 is located in
stops increasing (Figure 7). The temperature on the slope an area subjected to freezing/thawing processes. Probe
surface drops from 16.9°C–14.2°C (Figure 8) and the no. 4 is located in an area that is not subjected to freez-
previously thawed soil refreezes during this period. The ing/thawing processes. For Probe no. 3, the local factor
shear strength increases in the soil where the temperature of safety is above 1 when the temperature is below 0°C;
drops below 0°C. This increases the local factor of safety while it is below 1 when the temperature is above 0°C.
above 1, which prevents the slope from moving further This is partially because of the significant transition of
(Figure 9). soil strength under freezing or thawing conditions. For
The local factor of safety is not only a function of Probe no. 4, which is further inside the slope and not
position but also a function of time and temperature. subjected to the freezing/thawing transition, the local
10 Transportation Research Record 00(0)

Figure 10. Temperature and local factor of safety at different locations on the slope during the thawing/freezing process: (a) Probe no. 1;
(b) Probe no. 2; (c) Probe no. 3; (d) Probe no. 4.

factor of safety remains relatively stable. The minor tests were conducted by the RFEM method to determine
change is primarily because of the change of internal the shear strength parameters, that is, the cohesion and
stress field associated with the freezing/thawing process angle of internal friction of thawing soils and freezing
along the slope. soils. Slope stability analyses are conducted with the
RFEM model with coupled thermal field and mechanical
field to simulate the stability of thawing slopes holisti-
Conclusion cally. The simulation results show that the failure surface
of the slope is evolving with time and temperature over
This research describes a microstructure-based RFEM to
the year, which is consistent with field observations. The
analyze the stability of thawing slopes. Compared with
modeling results provide dynamic and reliable informa-
commonly used global stability analysis by the limit-
tion about the behaviors of thawing slopes. This newly
equilibrium method, this model considers the tempera-
developed method provides a reliable simulation tool to
ture variations associated with the freezing/thawing pro-
predict the stability of cold region slopes subjected to
cesses and their impacts on soil mechanical strengths. By
freezing/thawing processes and climate change, which is
considering the parameters and behaviors of individual
difficult to analyze by the traditional limit-equilibrium
phases, this model is able to simulate the bulk behaviors
methods.
of frozen and unfrozen soils. The randomly generated
soil structures are linked to the volumetric content of
each phase in frozen/unfrozen soils. The mechanical Acknowledgments
parameters of each phase in soils are first calibrated by This research is partially supported by the Ohio Department
comparison with the uniaxial compression test data on of Transportation and the U.S. National Science
thawing or frozen soils. From these, simulated triaxial Foundation.
Dong and Yu 11

Author Contributions 13. Leroueil, S., J. Tardif, M. Roy, P. L. Rochelle, and

The authors confirm contribution to the paper as follows: study J. M. Konrad. Effects of Frost on the Mechanical Beha-
conception and design: XY, SD; data collection: SD; analysis viour of Champlain Sea Clays. Canadian Geotechnical
and interpretation of results: SD, XY; draft manuscript pre- Journal, Vol. 28, No. 5, 1991, pp. 690–697.
paration: SD, XY. Both authors reviewed the results and 14. Zhang, Y., and R. L. Michalowski. Thermal-Hydro-
approved the final version of the manuscript. Mechanical Analysis of Frost Heave and Thaw Settlement.
Journal of Geotechnical and Geoenvironmental Engineering,
Vol. 141, No. 7, 2015, p. 04015027.
References 15. Bishop, A. W. The Use of the Slip Circle in the Stability
1. Frauenfeld, O. W., T. Zhang, and J. L. Mccreight. North- Analyses of Slopes. Geotechnique, Vol. 5, No. 1, 1955,
ern Hemisphere Freezing/Thawing Index Variations over pp. 7–17.
the Twentieth Century. International Journal of Climatol- 16. Janbu, N. Slope Stability Computations. Wiley and Sons,
ogy, Vol. 27, No. 1, 2007, pp. 47–63. New York, 1973.
2. Haeberli, W., and M. Beniston. Climate Change and its 17. Morgenstern, N. U., and V. E. Price. The Analyses of the
Impacts on Glaciers and Permafrost in the Alps. Ambio, Stability of General Slip Surfaces. Geotechnique, Vol. 15,
Vol. 27, No. 4, 1998, pp. 258–265. No. 1, 1965, pp. 79–93.
3. Shaoling, W., J. Huijun, L. Shuxun, and Z. Lin. Perma- 18. Duncan, J. M. State of the Art: Limit Equilibrium
frost Degradation on the Qinghai-Tibet Plateau and its and Finite-Element Analyses of Slopes. Journal of
Environmental Impacts. Permafrost and Periglacial Pro- Geotechnical Engineering, Vol. 122, No. 7, 1996,
cesses, Vol. 11, No. 1, 2000, pp. 43–53. pp. 577–596.
4. Thibault, S., and S. Payette. Recent Permafrost Degrada- 19. Wu, T. H. Soil Movements on Permafrost Slopes near Fair-
tion in Bogs of the James Bay Area, Northern Quebec, banks, Alaska. Canadian Geotechnical Journal, Vol. 21, No.
Canada. Permafrost and Periglacial Processes, Vol. 20, No. 4, 1984, pp. 699–709.
4, 2009, pp. 383–389. 20. McRoberts, E. C., and N. R. Morgenstern. The Stability of
5. Jorgenson, M. T., C. H. Racine, J. C. Walters, and Thawing Slopes. Canadian Geotechnical Journal, Vol. 11,
T. E. Osterkamp. Permafrost Degradation and Ecological No. 4, 1974, pp. 447–469.
Changes Associated with a Warming Climate in Central 21. Lu, N., B. Sxener-Kaya, A. Wayllace, and J. W. Godt. Anal-
Alaska. Climatic Change, Vol. 48, No. 4, 2001, yses of Rainfall-Induced Slope Instability using a Field of
pp. 551–579. Local Factor of Safety. Water Resources Research, Vol. 48,
6. Davies, M. C., O. Hamza, and C. Harris. The Effect of No. 9, 2012.
Rise in Mean Annual Temperature on the Stability of 22. Dong, S., and X. Yu. Experimental Characterization and
Rock Slopes Containing Ice-Filled Discontinuities. Perma- Microstructure Based Random FEM Simulation of Freez-
frost and Periglacial Processes, Vol. 12, No. 1, 2001, ing and Thawing Effects on Soils. Geotechnical Special Pub-
pp. 137–144. lication, 2016, pp. 711–719.
7. Harris, C., M. C. Davies, and B. Etzelmüller. The Assess- 23. Dong, S., and X. Yu. Microstructure-Based Random Finite
ment of Potential Geotechnical Hazards Associated with Element Simulation of Thermal and Hydraulic Conduction
Mountain Permafrost in a Warming Global Climate. Per- Processes in Unsaturated Frozen Soils. Geotechnical Special
mafrost and Periglacial Processes, Vol. 12, No. 1, 2001, Publication, 2017, pp. 781–790.
pp. 145–156. 24. Dong, S., and X. Yu. A Microstructure-Based Finite Ele-
8. Wang, H., X. Xing, T. Li, Z. Qin, and J. Yang. Slope ment Model to Estimate Hydraulic Properties in Fine-
Instability Phenomenon in the Permafrost Region along Grained Soils. Geotechnical Special Publication, 2018,
the Qinghai-Tibetan Highway, China. In Landslides in pp. 258–265.
Cold Regions in the Context of Climate Change. Springer, 25. Griffiths, D. V., and G. A. Fenton. Probabilistic Slope Sta-
Cham, 2014, pp. 11–22. bility Analysis by Finite Elements. Journal of Geotechnical
9. Gruber, S., M. Hoelzle, and W. Haeberli. Permafrost and Geoenvironmental Engineering, Vol. 130, No. 5, 2004,
Thaw and Destabilization of Alpine Rock Walls in the pp. 507–518.
Hot Summer of 2003. Geophysical Research Letters, Vol. 26. Fenton, G. A., and D. V. Griffiths. Bearing Capacity Pre-
31, No. 13, 2004. diction of Spatially Random C-j Soils. Canadian Geotech-
10. Kim, J. W., Z. Lu, F. Qu, and X. Hu. Pre-2014 Mudslides nical Journal, Vol. 40, No. 1, 2003, pp. 54–65.
at Oso Revealed by InSAR and Multi-Source DEM Analy- 27. Griffiths, D. V., G. A. Fenton, and N. Manoharan.
ses. Geomatics, Natural Hazards and Risk, Vol. 6, No. 3, Undrained Bearing Capacity of Two-Strip Footings on
2015, pp. 184–194. Spatially Random Soil. International Journal of Geomecha-
11. Qi, J., P. A. Vermeer, and G. Cheng. A Review of the Influ- nics, Vol. 6, No. 6, 2006, pp. 421–427.
ence of Freeze–Thaw Cycles on Soil Geotechnical Proper- 28. Griffiths, D. V., J. Huang, and G. A. Fenton. Probabilistic
ties. Permafrost and Periglacial Processes, Vol. 17, No. 3, Infinite Slope Analyses. Computers and Geotechnics, Vol.
2006, pp. 245–252. 38, No. 4, 2010, pp. 577–584.
12. Graham, J., and V. C. S. Au. Effects of Freeze–Thaw and 29. Fenton, G. A., D. V. Griffiths, and M. B. Williams. Relia-
Softening on a Natural Clay at Low Stresses. Canadian bility of Traditional Retaining Wall Design. Geotechnique,
Geotechnical Journal, Vol. 22, No. 1, 1985, pp. 69–78. Vol. 55, No. 1, 2005, pp. 55–62.
12 Transportation Research Record 00(0)

30. Hivon, E. G., and D. C. Sego. Strength of Frozen Saline 34. Czurda, K. A., and M. Hohmann. Freezing Effect on Shear
Soils. Canadian Geotechnical Journal, Vol. 32, No. 2, 1995, Strength of Clayey Soils. Applied Clay Science, Vol. 12,
pp. 336–354. No. 1–2, 1997, pp. 165–187.
31. Nixon, J. F., and G. Lem. Creep and Strength 35. Wang, D. Y., W. Ma, Y. H. Niu, X. X. Chang, and
Testing of Frozen Saline Fine-Grained Soils. Canadian Z. Wen. Effects of Cyclic Freezing and Thawing on
Geotechnical Journal, Vol. 21, No. 3, 1984, Mechanical Properties of Qinghai-Tibet Clay. Cold Regions
pp. 518–529. Science and Technology, Vol. 48, No. 1, 2007, pp. 34–43.
32. Chamberlain, E., C. Groves, and R. Perham. The Mechan- 36. Li, M., Y. Song, and F. Chen. Analyses on Stability of
ical Behavior of Frozen Earth Materials under High Pres- Slope in a Typical Cold Region Based on Thermo-Mechanical
sure Triaxial Test Conditions. Geotechnique, Vol. 22, No. Coupling. Bulgarian Chemical Communications, Vol. 48, 2016,
3, 1972, pp. 469–483. pp. 96–103.
33. Arenson, L. U., and S. M. Springman. Triaxial Constant
Stress and Constant Strain Rate Tests on Ice-Rich Perma- The Standing Committee on Seasonal Climatic Effects on
frost Samples. Canadian Geotechnical Journal, Vol. 42, No. Transportation Infrastructure (AFP50) peer-reviewed this paper
2, 2005, pp. 412–430. (19-02254).