Soil Dynamics and Earthquake Engineering 190 (2025) 109215

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Coupling effect of seepage and densification on soil

reliquefaction resistance
Chi-Chin Tsai * , Shao-Chi Yang
Department of Civil Engineering, National Chung Hsing University, Taiwan

A B S T R A C T

Case histories have demonstrated that once a site experiences liquefaction during an earthquake, it may be susceptible to repeated liquefaction in future events.
However, the mechanisms behind this increased susceptibility are not fully understood. This study investigated this phenomenon using a series of shaking table tests
on loose and medium dense soil specimens subjected to repeated shaking events. A unique aspect of these tests was the inclusion of prior upward, no, and downward
static seepage conditions before shaking. The results revealed that prior seepage can indeed influence soil liquefaction resistance. Upward seepage reduces resistance,
while downward seepage increases it. However, even with upward seepage due to dissipation of excess pore water pressure (i.e. post-liquefaction consolidation)
during shaking, the tests exhibited decreased excess pore water pressure and settlement in the shaking sequence. This indicates an overall increase in liquefaction
resistance due to densification from repeated shaking. The combined effects of these factors (densification and seepage) on reliquefaction resistance depend on the
initial relative density of the soil and the duration of seepage during consolidation. This complex interplay can lead to either an increase or decrease in resistance.
Finally, the study proposes a preliminary method to modify liquefaction resistance curves, incorporating the seepage effect. An example demonstrates how this
modified curve can be used to quantitatively evaluate reliquefaction resistance.

1. Introduction mechanisms like secondary compression and cementation. These

mechanisms enhance particle contact interlocking and, consequently,
Case histories by Quigley et al. [1], Tsai et al. [2], Wakamatsu [3], increase resistance to liquefaction. Numerous element tests have
and Youd [4] have shown that sites that have experienced liquefaction demonstrated the benefit of aging [12–14]. However, physical model
have a high potential for liquefaction in the future. For instance, studies suggest that this aging effect is “reset” by liquefaction [8,15].
following the 2010 Mw 7.1 Darfield earthquake in New Zealand, ten The large shear strains associated with liquefaction disrupt the previ­
liquefaction episodes occurred within a year in eastern Christchurch [1]. ously established soil structure. In simpler terms, the large strains
These events were triggered by a sequence of aftershocks (Mw 5.2–7.1) “erase” the aging effect, leaving the soil with a reduced resistance to
with peak ground acceleration (PGA) as low as 0.057 g, which is typi­ subsequent liquefaction events, as observed in repeat liquefaction tests
cally insufficient to induce initial liquefaction. Similar reliquefaction [16,17].
phenomena have been observed during the 1983 Nihonkai-Chubu Researchers believe unstable sand mesostructures, created by prior
earthquake and the 2011 Tohoku earthquake in Japan [5,6]. Likewise, strong shaking events, act as a weakening factor, reducing the sand
the site in Xinhua District, Tainan City, Taiwan, has experienced lique­ resistance to reliquefaction. Oda et al. [9] attributed the reduction to the
faction multiple times (1946, 2010, and 2016) [2,7]. These occurrences induced anisotropy, where a column-like structure with connected voids
have motivated significant interest in the geotechnical community forms in the soil as it liquefied. This anisotropic structure is unstable and
regarding the unclear mechanism of reliquefaction. Interestingly, prone to collapse under renewed loading. In another study, Yamada
research suggests that previously liquefied sand deposits exhibit a et al. [10] investigated the development of anisotropy during liquefac­
considerably lower resistance to subsequent liquefaction events, even tion and its effects on granular soil’s reliquefaction behavior through
though they become denser due to post-liquefaction reconsolidation undrained cyclic triaxial tests. They suggested that a higher level of
following the initial event [6,8–11]. developed anisotropy decreases liquefaction resistance because the sand
One possible explanation for the increased vulnerability to relique­ exhibits looser behavior when sheared in a certain direction. Centrifuge
faction at previously liquefied sites is the loss of the “aging effect.” This model tests by Ye et al. [18] also support the notion that changes in sand
refers to the pre-existing soil structure that develops over time through fabric during a previous liquefaction event could lead to lower resistance

* Corresponding author.
E-mail address: [email protected] (C.-C. Tsai).

https://doi.org/10.1016/j.soildyn.2025.109215
Received 19 September 2024; Received in revised form 22 November 2024; Accepted 8 January 2025
Available online 13 January 2025
0267-7261/© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

during subsequent shaking events. 2. Test model and program

Building on these observations, recent research has indeed shown a
difference in sand mesostructures before and after liquefaction. Suzuki 2.1. Model configuration and instrumentation
and Suzuki [19] first reported that triaxial liquefaction tests with sand
specimen freezing resulted in the long axes of liquefied sand particles A torsional shaking table developed by Tsai et al. [22] was used to
being oriented vertically. Using stereomicroscopic imaging during conduct 1-g model tests with a capacity of 105 kg. The table can
shaking table tests, Ye et al. [11] also observed a vertical distribution of generate torsional amplitudes up to 0.5◦ at frequencies ranging from 1 to
sand particle long-axis directions after liquefaction in centrifuge tests. 50 Hz As shown in Fig. 1, to induce liquefaction, this shaking table
Xie et al. [20] further corroborated these findings by capturing meso­ applies torsional vibrations that twist the soil sample along the cylin­
scopic images of sand particles during repeat shaking (liquefaction) der’s central axis. This specific vibration pattern minimizes the influence
events using a microimage acquisition system in centrifuge model tests. of the cylinder’s edges on the shaking behavior. However, this testing
Their images revealed that pore fluid seepage (upward direction) device does not aim to reproduce to the field conditions but induce soil
following the initial liquefaction event caused previously connected liquefaction. Four pore pressure transducers were installed along the
particles to separate and form large voids during the reconsolidation transparent acrylic cylinder (inner diameter of 14.8 cm) at 10 cm in­
stages. The sand with these large voids then contracted rapidly during tervals and a 20 cm interval. An accelerometer monitored base accel­
the subsequent shaking event, demonstrating a decrease in sand reli­ eration, and a laser displacement transducer continuously measured
quefaction resistance. Yang and Huang [21] further explored this specimen settlement. For more details on the test equipment, please
concept using a 3D discrete element method (DEM) to quantify fabric refer to Tsai et al. [22] and Chu and Tsai [23].
evolution in granular soils throughout the liquefaction, reconsolidation, Additionally, the inflow/outflow openings (Fig. 1a) modified from
and reliquefaction processes. Their findings support the notion that the Tsai et al. [22] enable seepage tests. Plastic balls with a diameter of 2 cm
effects of strain histories on reliquefaction resistance are intrinsically were placed at the bottom chamber to ensure uniform flow during these
linked to changes in soil fabrics before and after reconsolidation. tests. An important advantage of this setup is that all sensors are
Therefore, it is reasonable to deduce that seismic history can signifi­ pre-installed before sample preparation (unlike typical 1-g shaking table
cantly impact sand’s susceptibility to reliquefaction by intrinsically or centrifuge tests, which require time for sensor installation within the
changing the sand mesostructured. specimen). This allows shaking tests to be performed immediately after
This study investigated the reliquefaction phenomenon using a series specimen preparation, eliminating the aging effect.
of shaking table tests on loose and medium-dense soil specimens sub­
jected to repeated shaking events. These tests included prior upward, no,
and downward seepage conditions before shaking. Since all tests could 2.2. Test program
be performed immediately after specimen preparation, this study elim­
inated the aging effect and focused on the effects of structural changes 2.2.1. Sample preparation
during reliquefaction by applying prior seepage. The combined effects of The basic properties of the used C306 sand are listed in Table 1. A
densification and seepage on reliquefaction resistance are discussed, wet sedimentation method was employed to achieve a high degree of
along with a preliminary method to modify liquefaction resistance saturation (approximately 99.5 %) [24]. After filling the cylinder with
curves incorporating the seepage effect. An example demonstrates how water, uniform silica sand was sieved (#10 and #20 mesh) and carefully
this modified curve can be used to quantitatively evaluate reliquefaction poured while maintaining a constant drop height to achieve consistent
resistance. density. This pouring continued until the cylinder reached the target
weight of dry sand corresponding to the desired relative density (Dr =
40 % or 60 %). However, the initial Dr achieved through this method
was only around 10 %. Therefore, gentle shaking was applied to densify
the specimen to the target Dr when the designed specimen height (i.e.
69.3 cm) was achieved given the poured dry soil weight.

Fig. 1. Diagram and photo of the seepage and shaking test model and close view of filled specimen.

2
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

Table 1
Basic properties of C306 silica sand.
d(Δu) 1
i= × (1)
dz γf
Specific gravity of solids 2.65

Median size D50 (mm) 0.193 where z denotes the streamline length (calculated between 10 cm and
Effecive size D10 (mm) 0.147 30 cm positions in this study), and γf is the unit weight of fluid. As shown
Coefficient of uniformity Cu 1.36
Coefficient of curvature Cd 1.1
in Fig. 1c, the obtained i is around 0.8, which is close to critical hydraulic
Hydraulic conductivity k(cm/s) 0.034 gradient of 0.88.
Maximum dry unit weight 16.3
γd,max (kN/m3)
Minimum dry unit weight 13.9 3.2. Repeat shaking
γd,min (kN/m3)
Unified soil classification SP Fig. 3 shows the settlement measured in Test 1. As shaking com­
mences, the specimens under all shaking events exhibit settlement,
indicating a densification process. However, the induced settlement
2.2.2. Seepage and shaking test conditions
decreases with each subsequent shaking event despite the constant
After completing specimen preparation to the target Dr, this study
shaking duration and frequency. This is likely because the specimens
manipulates water flow direction through the specimens to alter their
become progressively denser with each shaking cycle. The first and
soil structures before shaking. Three flow conditions were established:
second shaking events induce similar settlement, but the magnitude
static state (no seepage), downward seepage, and upward seepage. The
decreases thereafter. By the fifth shaking event, no further settlement is
seepage process was controlled by an external pump connected to the
observed. The stabilized settlement values after each shaking are used to
inflow outlet that regulates inflow water pressure. The duration of this
re-calculate the corresponding Dr.
applied pressure lasted for 80 s. For the upward seepage condition, the
Fig. 4 presents the ru values at different depths for Test 1 (no prior
inflow water pressure was adjusted to a value that nearly triggers
seepage). During the first shaking event, significant liquefaction is
liquefaction (i.e. effective stress is close to zero), essentially loosening
evident throughout most of the specimen (i.e. 10, 20, and 30 cm), with ru
the soil structure without causing complete liquefaction (i.e., no sand
reaching values close to 1. However, ru remains lower (around 0.6) at
boils). The same pressure and duration were applied for the downward
the 50 cm depth, indicating no liquefaction in that zone. This pattern of
seepage. Note that the prior seepage history doesn’t cause volumetric
near-complete liquefaction throughout most of the specimen is observed
change of the specimen.
followed by a decrease in ru in the subsequent shaking events. Notably,
Following the seepage tests, repeat shaking tests were performed to
by the fifth shaking event, no ru is accumulated for all depths, suggesting
induce liquefaction. The sand specimens with Dr = 40 % and Dr = 60 %
no further liquefaction occurs. This decrease in ru implies an increase in
were shaken at frequencies of 10 Hz and 15 Hz (with the same amplitude
liquefaction resistance probably due to densification caused by the
of rotation), respectively, for a fixed duration of 20 s each in all test
shaking. This continuous increase of liquefaction resistance differs from
sequences. After each shaking event, shaking was paused until excess
the findings of Ha et al. [8] and Xie et al. [20] who reported the second
pore water pressure (EPWP, Δu) from the previous shaking cycle had
shaking to have the lowest liquefaction resistance. The mechanism
fully dissipated. An additional 10-s grace period was then provided
behind this discrepancy will be discussed in detail later.
before applying the next shaking cycle. This rest period allows for any
Fig. 5b shows the hydraulic gradient (i) calculated from the
residual pore pressure effects to diminish entirely before the subsequent
measured Δu for different shaking sequences. During the first shaking,
shaking event. A total of five shaking cycles were applied to each
the maximum i reaches values as high as 0.9 similar to that induced by
specimen. Table 2 listed total 6 test series performed in this study.
upward seepage (Fig. 2c). However, the duration of this high hydraulic
gradient (~20 s) is shorter compared to that induced by upward seepage
3. Test results
(~90 s). The hydraulic gradient (i) continuously decreases in subsequent
shaking events, reflecting the lower EPWP observed in Fig. 4.
3.1. Static seepage
4. Discussions
Fig. 2 shows the result of an upward seepage test on loose sand (Dr =
40 %) prior to shaking. Fig. 2a shows the time history of Δu, defined as
4.1. Seepage effect
the difference between the actual water pressure and the initial water
pressure before flow is initiated (i.e. static condition). The applied
Fig. 6 compares the shaking-induced ru at a depth of 10 cm for tests
seepage pressure lasts for more than 80 s, as indicated by the duration of
with prior upward seepage, no seepage, and downward seepage. All tests
the plateau. Higher EPWP at the bottom confirms upward seepage.
have the same initial Dr (i.e. ~40 %). During the first shaking event, the
Fig. 2b shows the time history of ru, which is defined as Δu normalized
test with prior upward seepage exhibits the longest duration of ru, fol­
by effective vertical stress. The ru values at different depths are around
lowed by the no seepage case. The downward seepage case shows the
0.9, indicating a condition close to liquefaction. Based on the measured
shortest ru duration. This suggests that even with identical Dr, prior
Δu, the average hydraulic gradient (i) is calculated by
seepage significantly influences EPWP generation. Similar trends are
observed in the subsequent shaking events. By the fourth shaking, the
downward seepage case no longer generates ru, while the other two
Table 2 cases still exhibit some ru. Finally, during the fifth shaking, the no
Test program. seepage case shows no ru, but the upward seepage case still maintains
some level of ru. These results demonstrate that prior seepage can alter
Test series Dr (%) before Seepage Dr (%) after No. of shaking
No. seepage seepage events liquefaction resistance even at the same Dr.
The observed influence of prior seepage on liquefaction resistance
1 40 no 40 5
2 40 upward 39 5 aligns with the findings of Xie et al. [20]. In their study, upward seepage
3 40 downward 40 5 caused by excess pore water pressure during liquefaction mobilized fine
4 60 no 60 5 particles, leading to the formation of larger voids (i.e. unstable struc­
5 60 upward 58 5 tures) and consequently, lower liquefaction resistance. While the
6 60 downward 60 5
mechanism of seepage induction differs between this study and Xie et al.

3
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

Fig. 2. Test 2 result of static upward seepage. (a) EPWP, (b) ru, (c) i.

As the shaking events progress, the specimens become denser

(densification is evident from the increase in Dr in Fig. 9, where Dr is
updated by the settlement measurement), leading to a decrease in ru and
a corresponding increase in liquefaction resistance. Interestingly, the Dr
increase after each shaking event is approximately 5 % for all tests,
regardless of their initial conditions before the aforementioned stop of ru
generation (Fig. 9).

4.3. Relative density effect

Fig. 10 shows the ru at a depth of 10 cm for tests of Dr = 60 % with

and without prior seepage, across different shaking sequences. Similar to
the Dr = 40 % case, ru values also decrease with each shaking event for
the Dr = 60 % condition. However, the influence of prior seepage on ru
appears to be less significant for the Dr = 60 % case. This is evident from
the similar ru curves of upward, no and downward seepage cases at the
same shaking event for Dr = 60 % (Fig. 10), compared to the distinct ru
curves observed for Dr = 40 % (Fig. 8). As the Dr increases, improved
inter-particle contact likely mitigates seepage-induced void formation
and structural instability. Consequently, the liquefaction resistance be­
Fig. 3. Settlement time history of shaking sequence (Test 1). comes less dependent on prior seepage for denser sands. While the up­
ward and no seepage cases exhibit similar liquefaction resistance due to
[20], the overall influence on soil structure and liquefaction behavior is their high ru values at the 5th shaking, the downward seepage case
consistent. Moreover, the observed low liquefaction resistance due to shows a slightly higher resistance due to its lower ru at the same shaking
upward seepage also indirectly confirms the hypothesis that unstable event. Based on this observation, it is concluded that the influence of
sand mesostructures formed after previous shaking events contribute to prior seepage on liquefaction resistance diminishes as the initial Dr
the decrease in sand reliquefaction resistance. increases.
Fig. 7 illustrates the maximum ru distribution along the depth for
tests with different prior seepage conditions. In the upward seepage 4.4. Coupling effect
case, ru remains high throughout the shaking events and exhibits a
relatively uniform distribution across the depth. In contrast, for the no Shaking triggers two competing mechanisms: densification
seepage and downward seepage cases, high ru values are primarily (strengthening soil) and upward seepage (potentially weakening soil
concentrated near the top of the specimen, with significantly lower ru in structure) during dissipation of EPWP. Densification increases lique­
the lower half, especially during the later shaking events. This suggests faction resistance, while seepage may counteract this effect. The relative
that upward seepage may induce more severe liquefaction throughout influence of these mechanisms determines the soil’s ability to resist
the entire soil layer, whereas the other two cases tend to cause repeated liquefaction after repeated shaking events.
liquefaction primarily in the shallow layers during subsequent shaking Our test results show a consistent increase in liquefaction resistance
events. Overall, upward seepage appears to exacerbate the severity of during repeated shaking, evidenced by the decreasing ru values. This
soil liquefaction. Similar patterns are observed for the Dr = 60 % and Dr contradicts the findings of Ha et al. [8] and Xie et al. [20], who reported
= 40 % case. the second shaking event to have the lowest resistance. One possible
reason is this study didn’t include aging effect while Ha et al. [8] and Xie
4.2. Consequent shaking effect et al. [20] unintentionally induce aging effect due a long time in pre­
paring specimen. However, even excluding this influence, prior upward
Fig. 8 compares the ru at a depth of 10 cm across the shaking se­ seepage in our tests significantly reduced liquefaction resistance. Based
quences, for tests with and without prior seepage. Consistent with the on these observations, we would also expect a decrease in resistance
trends observed earlier, ru values continue decreasing with the shaking similar to the findings of Ha et al. [8] and Xie et al. [20]. The observed
sequences regardless of prior seepage conditions. However, the rate of discrepancy likely arises from the relative influence of two opposing
decrease is influenced by the initial conditions, as previously discussed. mechanisms: upward seepage and consolidation (densification). In our
The upward seepage case still exhibits significant ru at the fifth shaking, tests, the densification effect appears to be dominant compared to the
even though the no seepage and downward seepage cases no longer upward seepage effect. This is likely because the duration of
show any ru by the fourth and fifth shaking events, respectively. liquefaction-induced seepage is insufficient to significantly change soil

4
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

Fig. 4. ru of Test 1 series (No seepage) at different depths.

lower Cv exhibits a lower reliquefaction resistance. This is because soil

with lower Cv requires longer time to EPWP, thus, allowing a longer
duration of upward seepage to change soil structure.
Real-world scenarios, such as centrifuge tests (e.g., Xie et al. [20]) or
field conditions, often involve longer drainage paths and may have a top
cap (unsaturated or non-liquefiable layer) that can extend the time for
EPWP dissipation and induce longer seepage durations. The extended
duration of liquefaction-induced seepage observed in field cases,
sometimes leading to sand boils several minutes or longer after shaking
[25–27], likely contributes to the lower liquefaction resistance.

5. Reliquefaction resistance assessment

5.1. Conceptual development

According to Fig. 9, the downward seepage case reaches a Dr of 55 %

by the third shaking and remains constant thereafter. Similarly, the no
seepage case reaches a Dr of 59 % by the fourth shaking and stabilizes.
This plateau in Dr indicates no further settlement or volume change, a
state known as the critical state. Importantly, even though their final Dr
values differ (55 %, 59 %, and assumed 67 % for upward seepage after
Fig. 5. Hydraulic gradient of Test 1 series (No seepage).
the fifth shaking), all cases likely exhibit similar liquefaction resistance
under these stable conditions (as illustrated in Fig. 11). Each distinct
structure and reduce resistance. Although prior seepage and liquefaction
seepage condition corresponds to a unique liquefaction resistance curve
seepage generate similar hydraulic gradients (~0.9), their durations
represented by cyclic resistance ratio (CRR). The observed deviations
differ. Prior seepage in our tests lasts for 80 s (Fig. 2c), whereas lique­
between these curves can be attributed to the duration of seepage.
faction seepage might only last for 20 s (Fig. 5).
Notably, the curves converge as Dr increases, suggesting that the influ­
A longer duration of liquefaction-induced seepage could potentially
ence of seepage becomes less significant in denser soils as discussed
amplify its negative impact on resistance. To test this hypothesis, one
earlier.
approach would be to increase shaking duration in our model tests to
Following Fig. 11, for a given Dr, the downward seepage case is likely
extend the seepage period. However, this approach has limitations in
to have the highest liquefaction resistance (CRR), followed by the no
our test. While a longer shaking duration would increase seepage time, it
seepage and upward seepage cases. This chart can be a valuable tool for
would also induce further densification, potentially negating the
evaluating reliquefaction resistance by constructing these curves quan­
seepage effect. An alternative approach is to perform tests using soil
titatively. For example, consider a point A on a curve representing the
with a low consolidation coefficient (Cv), which can prolong seepage
initial Dr and associated CRR. After shaking, Dr increases, and CRR
without additional densification. Ha et al. [8] also found that soil with

5
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

Fig. 6. Effect of seepage direction on shaking-induced ru at 10 cm during shaking sequence (Dr = 40 %).

Fig. 7. Maximum ru distribution along the depth for tests with different prior seepage conditions. Top row is Dr = 40 % and bottom row is Dr = 60 %.

initially increases due to densification (movement from A to B). How­ 5.2. Application
ever, due to upward seepage at the same Dr, CRR can then decrease
(movement from B to C). The final point (C or C′) depends on the The developed concept can be applied to assess the liquefaction
duration of liquefaction-induced seepage. A shorter duration would resistance of sand that has already liquefied (re-liquefied sand). Bases on
result in a final point like C, indicating an overall increase in CRR after cyclic triaxial test results, the CRR for C306 sand at 15 cycles (CRR15)
liquefaction. Conversely, a longer duration would lead to a final point for Dr = 40 %, 60 %, and 78 % are 0.21, 0.24, and 0.27, respectively. As
like C′, signifying a decrease in CRR compared to that at A. shown in Fig. 12, the relationship between CRR and Dr can be

6
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

Fig. 8. Effect of consequent shaking on ru at 10 cm (Dr = 40 %). (a) upward, (b) no, (c) downward seepage.

Fig. 9. Variation of Dr during shaking sequence (a) Initial Dr = 40 %, (b) Initial Dr = 60 %.

Fig. 10. Effect of consequent shaking on ru at 10 cm (Dr = 60 %). (a) upward, (b) no, (c) downward seepage.

approximated by a hyperbolic function [28,29]. Based on our experi­ during post-liquefaction consolidation, the CRR actually decreases to
mental results, the CRR for the no-seepage case at Dr = 60 % (repre­ 0.2149 at Dr = 52 % according to the repeated liquefaction curve. In
senting the first liquefaction event) is equivalent to the CRR for the essence, the CRR slightly decreases from 0.2211 to 0.2149 after the first
prior-seepage case at Dr = 67 % (representing the second or subse­ shaking (meaning the CRR for the second event is lower). This aligns
quent liquefaction event). This suggests that the CRR curve for repeated with observations from previous studies. However, the influence of
liquefaction can be constructed using the same hyperbolic function, but shaking on CRR can vary depending on the initial Dr. If the initial Dr is
passing through the point Dr = 67 % and CRR = 0.24. This curve will higher (e.g., Dr = 67 % in Test 4, representing a medium-dense condi­
also asymptotically approach the virgin liquefaction resistance curve (no tion), following the same analysis, the CRR remains at 0.244 before and
prior seepage) at high densities, as illustrated in the figure. In essence, after shaking, even with a 5 % increase in Dr. This suggests that for
these two distinct CRR curves can be used to evaluate the liquefaction denser soils (Dr>67 %), the CRR can actually increase after shaking,
resistance of the sand for the first and subsequent shaking events. even when considering the seepage effect. Therefore, the critical point is
An example is provided to illustrate the influence of initial Dr and that the change in CRR during subsequent shaking events (increase or
upward seepage on CRR. Assume a soil with an initial Dr of 47 % (Test decrease) is highly dependent on the initial Dr of the soil.
1). Based on the virgin liquefaction resistance curve, the CRR for this
condition is 0.2211. After the first shaking event, our tests show that Dr 6. Conclusions
increases by 5 %–52 %. According to the virgin curve again, a CRR of
0.2252 is expected at this new Dr. However, due to upward seepage This study investigated this soil reliquefaction resistance using a

7
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

due to densification from repeated shaking. The combined effects of

these factors (densification and seepage) on reliquefaction resistance
depend on the initial relative density of soil and the duration of seepage
during post-liquefaction consolidation. This complex interplay can lead
to either an increase or decrease in resistance. As Dr increases, the in­
fluence of upward seepage of post-liquefaction consolidation become
minor. Therefore, for the in-situ condition, a thick loose deposit un­
derlaying an impermeable or less permeable (e.g. clay or unsaturated
layer) could be most vulnerable to reliquefaction due to its low Dr and
long upward seepage duration.
Finally, the study proposes a preliminary method to modify typical
liquefaction resistance curves obtained from cyclic traxial test, incor­
porating the post liquefaction upward seepage effect. An example
demonstrates how this modified curve can be used to quantitatively
evaluate reliquefaction resistance. Given the initial Dr and the it’s
increment after shaking, the liquefaction resistance could be reduced,
remained, or increased, consistent with the shaking table test result and
previous study. This study provides valuable insights into reliquefaction
behavior and a procedure to quantify its resistance. To enhance this
method, additional basic tests are required to quantify the duration ef­
Fig. 11. Illustration of coupling effect of seepage and densification on CRR. fect of upward seepage on liquefaction resistance and the centrifuge
model test to mimic the real condition (i.e. the over burden pressure and
duration of seepage).

CRediT authorship contribution statement

Chi-Chin Tsai: Writing – review & editing, Writing – original draft,

Validation, Supervision, Methodology, Funding acquisition, Formal
analysis, Conceptualization. Shao-Chi Yang: Writing – original draft,
Visualization, Investigation, Formal analysis.

Data availability statement

Some of experimental data that support the findings of this study are
available from the corresponding author upon reasonable request.

Declaration of competing interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence
the work reported in this paper.

Acknowledgment

Assistance in performing the tests was provided by students Hui

Xuan Tang and Tang Hui Xuan at NCHU. The author expresses gratitude
Fig. 12. Example of modified CRR curve accounting for upward seepage effect. for their help. This work was supported by the National Science and
Technology Council Taiwan under Award No. 113-2625-M-005 -010 -
series of shaking table tests on loose and medium dense soil specimens The authors gratefully acknowledge such support.
subjected to repeated shaking events. Moreover, a prior upward, no, and
downward seepage were applied before shaking. Porewater pressure at Data availability
four depths of specimen and settlement at the surface are monitored
during the seepage and sequent shaking tests. Based on these measure­ Data will be made available on request.
ments, the change of specimen condition (e.g. EPWP accumulation/
dissipation and relative density) was documented to assess relique­ References
faction resistance during shaking.
The results revealed that prior seepage can indeed influence lique­ [1] Quigley M, Bastin S Barrett n, Bradley B. In: Recurrent liquefaction in Christchurch,
New Zealand, during the Canterbury earthquake sequence; 2013.
faction resistance through changing soil structure. Upward seepage re­ [2] Tsai C-C, Hsu S-Y, Wang K-L, Yang H, Chang W-K, Chen C-H, Hwang Y-W.
duces resistance, while downward seepage increases it as distinguishing Geotechnical reconnaissance of the 2016 M L 6.6 meinong earthquake in taiwan.
by the maximum ru and the number of shaking events to reach the sta­ J Earthq Eng 2017;22:1–27.
[3] Wakamatsu K. Recurrent liquefaction induced by the 2011 great east Japan
bilized condition (no settlement or PWP accumulation). However, even earthquake. Journal of Japan Association for Earthquake Engineering 2012;12:5_
with upward seepage due to dissipation of excess pore water pressure 69–65_88.
during shaking (i.e. post liquefaction consolidation), the specimens still [4] Youd TL. In: Recurrence of liquefaction at the same site; 1984.
[5] Yasuda S, Tohno I. Sites of reliquefaction caused by the 1983 nihonkai-chubu
exhibited decreased excess pore pressure and settlement in the shaking earthquake. Soils Found 1988;28:61–72.
sequence. This indicates an overall increase in liquefaction resistance [6] Wahyudi S, Koseki J, Sato T, Chiaro G. Multiple-liquefaction behavior of sand in
cyclic simple stacked-ring shear tests. Int J GeoMech 2016;16:C4015001.

8
C.-C. Tsai and S.-C. Yang Soil Dynamics and Earthquake Engineering 190 (2025) 109215

[7] Huang F-K, Tsai C-C, Ge L, Lu C-W, Chi C-C. Strength variations due to re- [18] Ye B, Zhang L, Wang H, Zhang X, Lu P, Ren F. Centrifuge model testing on
liquefaction—indication from cyclic tests on undisturbed and remold samples of a reliquefaction characteristics of sand. Bull Earthq Eng 2019;17.
liquefaction-recurring site. Bull Eng Geol Environ 2022:81. [19] Suzuki T, Suzuki T. Effects of density and fablic change on reliquefaction resistance
[8] Ha I-S, Olson S, Seo M-W, Kim M-M. Evaluation of reliquefaction resistance using of saturated sand. Soils Found 1988;28:187–95.
shaking table tests. Soil Dynam Earthq Eng 2011;31:682–91. [20] Xie X, Ye B, Zhao T, Feng X, Zhang F. Changes in sand mesostructure under
[9] Oda M, Kawamoto K, Suzuki K, Fujimori H, Sato M. Microstructural interpretation repeated seismic liquefaction events during centrifuge tests. Soil Dynam Earthq
on reliquefaction of saturated granular soils under cyclic loading. Journal of Eng 2021;150:106940.
Geotechnical and Geoenvironmental Engineering - J GEOTECH GEOENVIRON [21] Yang S, Huang D. Fabric evolution and liquefaction resistance in multiple
ENG 2001;127. liquefaction process:a micromechanical study using DEM clumps. Acta
[10] Yamada S, Takamori T, Sato K. Effects on reliquefaction resistance produced by Geotechnica 2022.
changes in anisotropy during liquefaction. Soils Found 2010;50:9–25. [22] Tsai C-C, Lin W-C, Chu M-C, Chi C-C. Experimental study on the mechanism of sand
[11] Ye B, Hu H, Bao X, Lu P. Reliquefaction behavior of sand and its mesoscopic boils and associated settlements due to soil liquefaction in loose sand. Eng Geol
mechanism. Soil Dynam Earthq Eng 2018;114:12–21. 2022;306:106708.
[12] Arango I, Migues RE. Investigation on the seismic liquefaction of old sand deposits. [23] Chu M-C, Tsai C-C. Experimental investigations on the effect of density, grain size,
Bechtel Corporation; 1996. and fines content on ejecta-induced settlement. Soil Dynam Earthq Eng 2024;177:
[13] Bwambale B, Andrus R, Cubrinovski M. Influence of age on liquefaction resistance 108366.
of Holocene alluvial and marine soils in Christchurch and Kaiapoi. 2017. New [24] Suits L, Sheahan TC, Ueng TS, Wang MH, Chen MH, Chen C-H, Peng LH. A large
Zealand. biaxial shear box for shaking table test on saturated sand. Geotechnical Testing
[14] Chen Y-C, You P-S. Evaluation of liquefaction potential by the test results of in-situ Journal - GEOTECH TESTING 2006;J:29.
frozen samples. In: ISOPE international ocean and polar engineering conference. [25] Housner GW. The mechanism of sandblows. Bull Seismol Soc Am 1958;48:155–61.
ISOPE; 2004. ISOPE-I-04-228. [26] Ambraseys N, Sarma S. Liquefaction of soils induced by earthquakes. Bull Seismol
[15] Dobry R, Abdoun T. Cyclic shear strain needed for liquefaction triggering and Soc Am 1969;59:651–64.
assessment of overburden pressure factor kσ. J Geotech Geoenviron Eng 2015;141: [27] Ansary M, Yamazaki F, Towhata I. Application of a dynamic effective stress model
04015047. at a reclaimed site during the Great Hanshin earthquake, vol. 1995; 1995.
[16] Darby K, Boulanger R, Dejong J, Bronner J. Progressive changes in liquefaction and [28] Hwang J-H, Khoshnevisan S, Juang CH, Lu C-C. Soil liquefaction potential
cone penetration resistance across multiple shaking events in centrifuge tests. evaluation – an update of the HBF method focusing on research and practice in
J Geotech Geoenviron Eng 2019;145. Taiwan. Eng Geol 2021;280:105926.
[17] Fardad Amini P, Huang D, Wang G, Jin F. Effects of strain history and induced [29] Tsai C-C, Kishida T, Chen S-C, Lu C-W. Empirical evaluation of the effect of fines
anisotropy on reliquefaction resistance of toyoura sand. J Geotech Geoenviron Eng content on spt-N liquefaction triggering curves: insights from experimental and
2021;147:04021094. field databases. 2024.