http://www.scar.ac.

cn

Sciences in Cold and Arid Regions

2012, 4(1): 00130020
DOI: 10.3724/SP.J.1226.2012.00013

Seasonally frozen soil effects on the dynamic behavior

of highway bridges
ZhaoHui "Joey" Yang 1*, Qiang Li 1, Elmer E. Marx 2, JinChi Lu 3
1. Department of Civil Engineering, University of AlaskaAnchorage, Anchorage, Alaska, USA
2. Bridge Design Engineer, State of Alaska DOT&PF, 3132 Channel Drive, Juneau, AK 99801, USA
3. Assistant Project Scientist, Department of Structural Engineering, University of California, San Diego, La Jolla, CA 92093,
USA
*Correspondence to: Prof. ZhaoHui "Joey" Yang, Department of Civil Engineering, University of AlaskaAnchorage, Anchorage,
Alaska, USA. Email: afzy@uaa.alaska.edu
Received: June 20, 2011

Accepted: August 19, 2011

ABSTRACT
Frozen ground is significantly stiffer than unfrozen ground. For bridges supported on deep foundations, bridge stiffness is also
measurably higher in winter months. Significant changes due to seasonal freezing in bridge pier boundary conditions require additional detailing in order to ensure a ductile performance of the bridge during a design earthquake event. This paper reports the latest results obtained from a project that systematically investigated the effects of seasonally frozen soil on the seismic behavior of
highway bridges in cold regions. A bridge was chosen and was monitored to study its seismic performance and assess the impact
of seasonally frozen soil on its dynamic properties. A Finite Element (FE) model was created for this bridge to analyze the impact
of seasonal frost. It was found that when frost depth reaches 1.2 m, the first transverse modal frequency increases about 200%
when compared with the no-frost case. The results show that seasonal frost has a significant impact on the overall dynamic behavior of bridges supported by pile foundations in cold regions, and that these effects should be accounted for in seismic design.
Keywords: dynamic behavior; seasonally frozen soil; pile foundation; earthquake; Finite Element modeling

1. Introduction
Seasonally frozen ground is significantly stiffer than unfrozen ground due to the ice matrix that is formed in the
ground during freezing. Stevens (1973) found that the stiffness of soils can change by as much as two orders of magnitude when they are frozen. By conducting dynamic pile
tests in the field, Vaziri and Han (1991) found that the presence of a frozen soil layer, even at a modest thickness of less
than 0.5 m, could significantly influence the dynamic response of piles. For bridges supported on deep-pile group
foundations, overall bridge stiffness was found to increase
measurably in winter months (Yang et al., 2007). As bridge
stiffness increases, seismic demand increases significantly as
well (Suleiman et al., 2006; Sritharan et al., 2007; Xiong
and Yang et al., 2008). However, there has been no systematic study completed regarding lateral pile behavior that

takes into consideration seasonal freezing, nor do there exist

any design guidelines that address these effects.
This paper reports the findings on the dynamic response
of highway bridges under the influence of deep seasonal
freezing. A newly built highway bridge was selected and
several instruments, including accelerometers and thermistors, were installed in order to monitor dynamic responses
and frost penetration in the field. Bridge dynamic response
data and seasonal frost penetration data are presented and
analyzed. In addition, a Finite Element (FE) model was built
to investigate the impact of seasonally frozen soil on the
bridge dynamic behavior.
2. Description of the test bridge and monitoring facilities
A prestressed, reinforced-concrete girder bridge in Anchorage, Alaska was outfitted for field monitoring. This

14

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

concrete girder bridge is supported by two center piers

founded on 10 0.9-m-diameter concrete-filled steel pipe
piles, with abutments founded on 11 0.6-m-diameter concrete-filled steel pipe piles at both ends. This bridge was
selected for the following reasons: (1) it is supported by an
extended steel pipe pile-shaft system filled with reinforced
concrete; (2) the bridge site consists of original sandy gravelly soils overlaid with 2 m of organic soils; and (3) the water table is within 0.5 m of the ground surface. This bridge
provides an excellent test bed for detecting upper-bound
seasonal freezing effects.
In order to collect dynamic as well as seismic response
data, a network consisting of 11 accelerometers (ES-U2
EpiSensor; www.kinemetrics.com) was installed on the
bridge superstructure in November 2008, according to the

plan shown in Figure 1. A 12-Channel Granite Data Recorder, also from Kinemetrics, was used to record the dynamic
responses of the bridge. A 1.5-m-deep borehole was prepared beneath the bridge for ground temperature/frost penetration monitoring. Air temperature, frost penetration, ambient
vibration data, and earthquake-induced vibration data were
collected during the field monitoring process.
A temperature acquisition cable (TAC) with sensors arranged at a predefined spacing, together with a handheld
data collector, were used to collect ground temperature/frost
penetration data, as illustrated in Figure 2. The TAC was
brought to the project site specifically for data collection.
The TAC was lowered into the borehole and left uninterrupted for at least 60 min to ensure that the sensors were
fully acclimated before data collection commenced.

Figure 1 Schematic of the bridge selected for study and the seismic instrumentation plan

Figure 2 The temperature acquisition cable used for frost penetration depth monitoring

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

3. Field monitoring results and analyses

This section discusses the results of the field monitoring
in regards to frost penetration and bridge dynamic performance under seismic excitation.
3.1. Frost penetration
Binonwangan (2009) reported the field monitoring results from December 2008 to March 2009, as shown in Figure 3. The results show that the deepest frost penetration was
1.45 m, while the depth remained at approximately 1.2 m for
the entire month of March.
The Modified Berggrens formula (Andersland and
Ladanyi, 2004) was utilized to provide a continuous assessment of the frost penetration depth. This formula is the
Stefan formula corrected for the effects of temperature
changes in soil mass (University of Washington, 2008). The
Modified Berggren formula is:
X

48kave nFI
L

(1)

where X is the depth of freeze or thaw (m); is a dimensionless coefficient which takes into account the effect of
temperature changes in the soil mass and sensible heat
changes; kave is the thermal conductivity of soil (the average
of frozen and unfrozen) in units of kJ/(hrmC); n in units
of Cdays, including nf and nt, is a dimensionless conversion factor for the air freezing and thawing index to surface
freezing and thawing index, and L is the latent heat in kJ/m3.
For comparison purposes, the estimated frost depth is also

15

shown in Figure 3. This calibrated formula was used to

evaluate the frost depth during the 20092010 winter and the
result is presented in Figure 4.
It is noted that values of 1.3 and 0.6 have been used for
the dimensionless factors nt and nf, respectively, to produce a
better match between the Berggrens formula estimation and
the field observations. The correlation coefficient between
the estimated and observed frost penetration depths is
0.89. The difference between the field data and the estimation can be attributed to the presence of the nearby creek.
The creek, which did not remain entirely frozen, became a
heat source that affected the measured freezing depths. In
the following analysis, the actual data were used when
available; otherwise, estimated data were used.
3.2. Bridge dynamic response
Strong Motion Analyst (SMA), Matlab, and ARTeMIS
Extractor were used to process the recorded earthquake
data. SMA, a program for reading and processing data
from EVT files and saving the data as USGS format V2
files (www.kinemetrics.com), was used to preprocess the
strong-motion accelerogram records. Matlab was used to
convert seismic data from USGS format V2 files to individual files for acceleration, velocity, displacement, and
response spectrum (ASC files). The acceleration data from
the ASC files were used to generate an acceleration input
file for the ARTeMIS program (structural system identification software; www.svibs.com) to analyze the bridges
dynamic properties, including modal frequencies and mode
shapes. The steps for data processing are summarized in
Figure 5.

Figure 3 Comparison of frost depths estimated using the Modified Berggrens formula with field monitoring data

16

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

Figure 4 Frost depth evaluated using the Modified Berggrens formula from October 2009 to March 2010

Figure 5 Data processing flowchart

During the data collection period, bridge vibration data

were collected from a total of 22 earthquakes that triggered
the accelerometers, with magnitudes ranging from ML 3.3 to
5.7. The raw vibration data were processed and analyzed
using ARTeMIS to obtain the bridges dynamic properties,
which include modal frequency and mode shape. Table 1
summarizes the source information of earthquake events, frost
depths, and first transverse modal frequencies of the bridge.
The modal frequencies of the first lateral mode of each
earthquake event and the corresponding frost thickness (frost
depth minus thaw depth) are presented in Figure 6. From
Figure 6 it is clear that modal frequency increases with frost
penetration. In fact, there is a strong correlation between
frost depth and structural modal frequency and the correlation coefficient between these two variables is 0.98. Specifically, when there is no frozen ground, the first transverse

modal frequency is 2.5 Hz. As frost thickness increases to

1.2 m, the modal frequency increases to around 7.5 Hz, representing a 200% increase.
4. Finite element modeling
The Open System for Earthquake Engineering Simulation (OpenSees) is an FE software framework developed by
the Pacific Earthquake Engineering Research Center (PEER;
University of CaliforniaBerkeley) for modeling and simulating the performance of structural and geotechnical systems that are subjected to earthquakes and other loading
conditions (Mazzoni et al., 2006). This study used the graphic
user interface OpenSeesPL (http://cyclic.ucsd.edu/openseespl),
a program based on OpenSees specifically developed for
conducting pilesoil interaction analysis.

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

A model for the entire bridge-foundation-soil system

would have been ideal for investigating frozen soil effects.
However, such a model would demand a great amount of
computing power and time to analyze, and, as such, these
resources are not readily available. Therefore, it was decided
to model the soil-pile system and the bridge superstructure
separately. First, the equivalent cantilever concept was
adapted to evaluate the equivalent depth-to-fixity of the
soil-pile system. Then, a three-dimensional FE model for the
superstructure was constructed to conduct modal analysis.
The FE model was validated by experimental data and used
to predict extreme condition scenarios, i.e., the frost depths
of 2.0 m or more.
The equivalent cantilever method assumes that the
soil-pile system can be replaced by an equivalent cantilever
that is fully restrained against lateral translation and rotation
(Dowrick, 1987; Chai, 2002). In this approach, the equivalent depth-to-fixity is used to account for the flexibility of
the embedded pile (Chai, 2002). In Figure 7, the sketch on
the left is a soil-pile system and the one on the right is the
corresponding equivalent cantilever. La is the height of pile
above ground and Lf is the equivalent depth-to-fixity. The

17

equivalent depth-to-fixity depends on the relative stiffness

between the pile and surrounding soil and may be determined by equating the lateral stiffness of the soil-pile system
to that of an equivalent cantilever (Chai, 2002).
Three-dimensional FE models (Figure 8a) of both the
seasonally frozen soil with different frost depths and the
concrete-filled steel pipe pile were constructed in order to
conduct "pushover analysis" using OpenSeesPL. Meanwhile,
corresponding cantilever models were also created for the
pushover analysis, and the lateral stiffness at the elastic stage
of both cantilever beam and soil-pile system were compared
to obtain the depth-to-fixity. The pile cross section was discretized into an array of fibers, as shown in Figure 8b, with
three different regions to represent the confined concrete, the
longitudinal steel reinforcement, and the steel pipe. Both the
confining effect on core concrete (Mander et al., 2008a, b)
and the temperature effects on concrete and steel (Lee et al.,
1988a, b; Filiatrault and Holleran, 2001; Sritharan et al.,
2007; Montejo et al., 2008) were taken into consideration in
this analysis. The pile heights above ground surface for the
0.9-m pile and 0.6-m pile are 2.74 m and 1.83 m, respectively. The details of this model can be found in Li (2011).

Table 1 Earthquake events, frost depth, and first transverse modal frequencies
Air temperature
(C)

Frost depth
(m)

First transverse
modal frequency
(Hz)

101

7.8

1.16

7.065

47

50

16.7

1.17

7.458

4.4

499

35

13.9

1.19

7.774

4.3

48

40

1.1

1.19

7.550

Earthquake
event date

Earthquake
Magnitude, ML

Epicentral
distance
(km)

2009-1-24

5.7

322

2009-2-1

3.3

2009-2-09
2009-2-15

Earthquake
depth (km)

2009-2-16

3.7

80

69

1.1

1.19

7.505

2009-3-12

4.0

225

192

1.7

1.24

7.593

2009-4-7

4.6

18

24

0.0

1.22

7.579

2009-4-26

3.7

27

6.7

0.91

6.450

2009-10-4

4.0

117

98

6.1

0.00

2.847

2009-11-11

3.9

225

158

6.1

0.19

3.110

2009-12-15

3.9

193

119

2.2

0.63

5.020

2009-12-16

3.0

26

51

2.2

0.65

5.120

2010-1-5

3.2

169

72

6.1

0.79

5.300

2010-1-21

3.7

50

43

14.4

0.88

5.747

2010-1-30

3.5

109

95

8.9

0.94

6.500

2010-1-31

3.0

40

71

11.1

0.95

7.020

2010-2-2

3.5

121

79

4.4

0.95

6.994

2010-2-11

4.2

185

106

3.3

1.00

7.029

2010-2-20

3.6

272

98

1.1

1.00

7.235

2010-2-21

3.0

56

58

0.0

1.00

6.830

2010-2-27

3.7

90

60

10.0

1.02

6.956

2010-3-2

3.7

129

47

0.6

1.02

6.690

18

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

Figure 6 Frozen ground thickness vs. first lateral modal frequency

Figure 7 Representation of a soil-pile system by an equivalent fixed-base cantilever

The purpose of these models was to obtain the equivalent depth-to-fixity of the pile using the concept of an
equivalent cantilever. Table 2 summarizes the equivalent
lengths for the soil-pile systems with different frost depth
penetrations. When the frost depth reached 0.45 m, the
0.9-m piles equivalent length decreased 27% when compared to the unfrozen condition. With increased frost depth,
the equivalent length continued to decrease; 33% and 35%
decreases were recorded when the depth of frozen ground

reached 1.20 m and 1.85 m, respectively. However, the

effect of frozen ground on the depth-to-fixity diminishes
with increasing frost depth. At 1.85-m frost depth, the
equivalent depth of fixity was 0.75D, compared to 2.8D for
the 0.9-m pile at an unfrozen condition, where D is the pile
diameter. Since the 0.6-m piles were covered by loose
gravel above the ground surface, the 0.6-m pile-soil system
was not as sensitive to the frost depth as the 0.9-m pile-soil
system.

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

19

other conditions, by varying only the depth-to-fixity of the

piles. The results are presented in Table 3. It can be seen that
the first transverse modal frequency of the bridge predicted
by the calibrated FE model agrees well with the field test
results in both the no-frost and 1.2-m frost cases. For the
2.0-m frost case, the predicted first transverse modal frequency is 7.9 Hz, representing a 230% increase compared
with the no-frost case.

With the equivalent cantilever lengths at different soil

conditions for each soil-pile system, a simplified lumped
mass model for the bridge superstructure was built using
OpenSeesPL. Superstructure mass and stiffness were first
estimated based on the design drawings, and refined by
matching the first transverse modal frequency and shape of
the 0.45-m frozen soil case. The calibrated model was then
used to predict the dynamic characteristics of the bridge for

Figure 8 Finite element model of the soil-pile system: (a) isotropic view of the entire model; (b) fiber section of the 36-inch pile

cally, when the frost depth reaches 1.2 m deep, the first
transverse modal frequency increases more than 200%
when compared with the no-frost case; the equivalent
depth-to-fixty of the pile-soil system decreases from about
three pile diameter when there is no frost to less than one
pile diameter when the frost depth reaches 1.2 m. The
equivalent depth-to-fixity of the pile-soil system based on
FE modeling can be used to analyze the dynamic behavior
of other deep foundations in seasonally frozen soils. These
findings serve as an important step in understanding the
dynamic behavior of bridges under seasonally frozen soil
conditions and also shed light on the structural health monitoring of bridges in cold regions.

5. Conclusions
The dynamic performance of a selected bridge under
the excitation of many small to moderate earthquakes in a
full calendar year were recorded and analyzed. In particular,
the frequency of the first transverse mode was identified
for each earthquake occurred in different seasons. A Finite
Element model was created to analyze the impact of seasonally frozen soil on the dynamic behavior of the bridge.
Both field experimental data and numerical simulation
results show that seasonally frozen soil has a great impact
on the lateral behavior of both the soil-pile system and the
overall behavior of the entire bridge system. More specifi-

Table 2 Equivalent cantilever length for the soil-pile system

Frost depth (m)
Equivalent length (m)
Equivalent depth of fixity
(m/ D)
D = pile diameter.

0.9-m pile
0.6-m pile
0.9-m pile
0.6-m pile

0
5.30
2.80
2.56 / 2.80D
0.98 / 1.60D

0.45
3.87
2.53
1.13 / 1.23D
0.70 / 1.15D

1.2
3.54
2.41
0.79 / 0.87D
0.58 / 0.95D

2.0
3.44
2.29
0.70 / 0.77D
0.46 / 0.75D

Table 3 Comparison between identified and calculated results and the first transverse frequencies calculated by FE model
Frost depth (m)
0 (Unfrozen soil)
0.45
1.20
2.00

Calculated
The first transverse
frequency (Hz)
2.416
4.960
7.336
7.938

Percentage
increase
0.0%
105.3%
203.6%
228.6%

Identified
2.528
5.020
7.579

Difference between
calculated and
identified data
4.6%
1.2%
3.3%

20

ZhaoHui "Joey" Yang et al., 2012 / Sciences in Cold and Arid Regions, 4(1): 00130020

Acknowledgments:
This project was jointly funded by the Alaska University
Transportation Center (AUTC) and the State of Alaska Department of Transportation and Public Facilities (AK
DOT&PF), under AUTC Project No. 107014. The authors
gratefully acknowledge this support. The authors are thankful to Dr. Utpal Dutta of the University of Alaska-Anchorage
School of Engineering, who provided software guidance on
strong-motion data processing and former graduate students
Ruel Binonwangon and Gang Xu, who took part in the collection and processing of field monitoring data.
REFERENCES
Andersland OB, Ladanyi B, 2004. Frozen Ground Engineering. 2nd Edition.
John Wiley & Sons, Inc., New York.
Binonwangan CR, 2009. Seasonal freezing effects on the dynamic response of bridge structures on pile foundation. Graduate project, Dept.
of Civil Engineering, University of AlaskaAnchorage, Anchorage,
AK.
Chai YH, 2002. Flexural strength and ductility of extended pile-shafts. I:
Analytical model. Journal of Structural Engineering, 128(5): 589594.
Dowrick DJ, 1987. Earthquake Resistant Design. 2nd Edition.
Wiley-Interscience, New York.
Filiatrault A, Holleran M, 2001. Stress-strain behavior of reinforcing steel
and concrete under seismic strain rates and low temperatures. Material
Structure, 34(5): 235239.
Lee GC, Shih TS, Chang KC, 1988a. Mechanical properties of concrete at
low temperature. Journal of Cold Regions Engineering, ASCE, 2(1):
1324.
Lee GC, Shih TS, Chang KC, 1988b. Mechanical properties of high-strength
concrete at low temperature. Journal of Cold Regions Engineering,
ASCE, 2(4): 169178.
Li Q, 2011. Effects of frozen soils on the seismic behavior of highway bridge

foundation. Masters thesis, Dept. of Civil Engineering, University of

AlaskaAnchorage, Anchorage, AK.
Mander JB, Priestley MJN, Park R, 1988a. Theoretical stress-strain model
for confined concrete. Journal of the Structural Division, ASCE, 114:
18041826.
Mander JB, Priestley MJN, Park R, 1988b. Observed stress-strain behavior
of confined concrete. Journal of the Structural Division, ASCE, 114:
18271849.
Mazzoni S, McKenna F, Scott MH, Fenves GL, 2006. Open System for
Earthquake Engineering Simulation User Command-Language Manual.
University of California, Berkeley, CA. http://opensees.berkeley.edu/
OpenSees/manuals/usermanual/ index.html.
Montejo LA, Sloan JE, Kowalsky MJ, Hassan T, 2008. Cyclic response of
reinforced concrete members at low temperatures. Journal of Cold Regions Engineering, ASCE, 22: 79102.
Sritharan S, Suleiman MT, White DJ, 2007. Effects of seasonal freezing on
bridge column-foundation-soil interaction and their implications. Earthquake Spectra, 23(1): 199222.
Stevens HW, 1973. Viscoelastic properties of frozen soils under vibratory
loads. Proceedings of the North American Contribution to the Second
International Conference on Permafrost, Yakutsk, USSR. 400409.
Suleiman MT, Sritharan S, White DJ, 2006. Cyclic lateral load response of
bridge column-foundation- soil systems in freezing conditions. Journal of
the Structural Division, ASCE, 132(11): 17451754.
University of Washington, 2008. Modified Berggren Formula for Estimating
Freeze/Thaw Depth. http://training.ce.washington.edu/pgi/Modules/04_
design_parameters /berggren.htm.
Xiong F, Yang Z, 2008. Effects of seasonally frozen soil on the seismic
behavior of a Bridge Bent-Foundation-Soil System. Cold Regions Science and Technology, 54:4453.
Vaziri H, Han Y, 1991. Full-scale field studies of the dynamic response of
piles embedded in partially frozen soils. Canadian Geotechnical Journal,
28: 708718.
Yang Z, Dutta U, Zhu D, Marx E, Biswas N, 2007. Seasonal frost effects on
the soil-foundation-structure interaction system. Journal of Cold Regions
Engineering, ASCE, 21(4): 108120.
Yang Z, Xiong F, Xu G, Hulsey JL, Marx E, 2008. Seasonal freezing effects
on the lateral behavior of steel pipe piles. Proceedings of the 14th World
Conference on Earthquake Engineering. Beijing, China.