Research Paper

Earthquake Spectra
Tunnel performance during 2020, Vol. 36(S2) 288–313
Ó The Author(s) 2021
the Puebla-Mexico 19 Article reuse guidelines:
sagepub.com/journals-permissions

September 2017 earthquake DOI: 10.1177/8755293020976140

journals.sagepub.com/home/eqs

Juan M Mayoral, M.EERI1, Gilberto Mosqueda, M.EERI2,

Daniel De La Rosa1, and Mauricio Alcaraz1

Abstract
Seismic performance of tunnels during earthquakes in densely populated areas
requires assessing complex interactions with existing infrastructure such as bridges,
urban overpasses, and metro stations, including low- to medium-rise buildings. This
article presents the numerical study of an instrumented tunnel, currently under con-
struction on stiff soils, located in the western part of Mexico City, during the Puebla-
Mexico 19 September 2017 earthquake. Three-dimensional finite difference models
were developed using the software FLAC3D. Initially, the static response of the tunnel
was evaluated accounting for the excavation technique. Then, the seismic perfor-
mance evaluation of the tunnel was carried out, computing ground deformations and
factors of safety, considering soil nonlinearities. Good agreement was observed
between predicted and observed damage during post-event site observations. Once
the soundness of the numerical model was established, a numerical study was under-
taken to investigate the effect of frequency content in tunnel-induced ground motion
incoherence for tunnels built in cemented stiff soils. A series of strong ground
motions recorded during normal and subduction events were used in the simula-
tions, considering a return period of 250 years, as recommended in the Mexico City
building code. From the results, it was concluded that the tunnel presence leads to
important frequency content modification in the tunnel surroundings which can
affect low- to mid-rise stiff structures located nearby. This important finding must be
taken into account when assessing the seismic risk in highly populated urban areas,
such as Mexico City.

Keywords
Tunnel, seismic, interaction, resilience, urban areas, risk, damage
Date received: 2 November 2020; accepted: 3 November 2020

1
Institute of Engineering, National Autonomous University of Mexico, Mexico City, Mexico
2
Jacobs School of Engineering, University of California, San Diego, CA, USA

Corresponding author:
Juan M Mayoral, Institute of Engineering, National Autonomous University of Mexico, Mexico City 04510, Mexico.
Email: [email protected]
Mayoral et al. 289

Introduction
Although underground structures such as tunnels, open shafts, and underground parking
lots historically have exhibited lower damage rates during earthquakes than on ground
structures, a large number of cases compiling tunnel damage are documented in the tech-
nical literature (Sharma and Judd, 1991). Some benchmark examples are major cracking
in tunnel lining during the 2004 Niigata earthquake in Japan (Figure 1) and the collapse
of the Bolu tunnels during the Kocaeli earthquake in Turkey (Federal Highway
Administration (FHWA), 2009). The National Institute of Standards and Technology
(NIST, 1996) reported the damage of a cut and cover tunnel during the 1995 Kobe earth-
quake. Yu et al. (2016) describe the seismic performance of the Longxi tunnel, one of the
most severely damaged tunnels during the 2008 Wenchuan earthquake in China. To date,
however, there is still a lack of proper understanding of the response of tunnels in heavily
populated urban areas and their interactions with other surrounding structures.
Understanding the complex interplay among tunnels, open shafts, underground structures,
and their surroundings in urban environments (Figure 2) requires proper calibration of
numerical models through comparing estimations with actual measured responses.
Key issues to account for in tunnel seismic performance evaluations in urban environ-
ments include (1) tunnel–soil–structure interaction at connections with open shafts, metro
or train stations, or adjacent buildings and bridges; (2) abrupt changes in structural or soil/
rock stiffness due to geologic faults, soil layering, or lining thickness, and as built-shape;
(3) soil nonlinearities generated at both free and near field and those associated with pore
pressure developed at the tunnel–soil interface; (4) three-dimensional (3D) state of stresses
prevailing after construction around the tunnel lining, and its evolution during the eco-
nomic life of the structure, including ground subsidence effects; and (5) 3D seismic wave
propagations (Hashash et al., 2001). The interplay of these factors can lead to differential
deformations along the tunnel in both longitudinal and transversal directions and in turn
generate stress concentrations, which will induce static or seismic failure, in both the tunnel
lining and the surrounding soil. Three main deformation modes can be observed during

Figure 1. Highway tunnel lining falling from Tunnel Crown—2004 Niigata Earthquake, Japan (FHWA,
2009).
290 Earthquake Spectra 36(S2)

Figure 2. Dynamic interaction between underground and on-ground structures.

seismic tunnel loading: (1) axial compression and tension, (2) longitudinal bending, and (3)
transversal distortion (Tsinidis et al., 2020).
This article reviews the seismic performance of a tunnel under construction in the stiff
soils found in the western part of Mexico City, during the Puebla-Mexico, 2017 earth-
quake. 3D finite difference models were developed using the software FLAC3D and cali-
brated based on observed tunnel behavior in the post-earthquake reconnaissance. This type
of studies also help to explain some typical failures of engineered structures observed in the
field, to have better and resilient designs (Elnashai et al., 2012). The damage predicted by
the numerical simulation through computed ground deformations, factors of safety, forces,
and bending moments developed in the primary lining are in good agreement with the
observed damage during the UNAM-GEER reconnaissance conducted after the earth-
quake (Mayoral et al., 2017b). Once the soundness of the numerical model was established,
a series of simulations were undertaken to evaluate tunnel-induced free field ground motion
modification, which potentially can affect low- and mid-rise adjacent buildings.

Seismic event
A 7.1-Mw earthquake struck the central region of Mexico on 19 September 2017. The zone
of energy released was located at 18.5838N and 98.3993E, down to a depth of 51 km,
according to the United States Geological Survey (USGS). This earthquake occurred in a
complex region of normal and reverse faults with a regional tectonic mechanism associated
with the subduction of the Cocos plate under the North American plate. The focal
mechanism was normal faulting. The epicenter was located at about 122.3 km southeast of
the tunnel construction site. The closest seismological station to the tunnel site, located in
firm soil, that recorded the ground motions during the earthquake is identified as TACY.
This station is located about 1.8 km from the tunnel location (Figure 3). Both the
Mayoral et al. 291

Figure 3. Tunnel location, active seismological stations (II-UNAM), and geotechnical zoning.

acceleration time histories and their corresponding response spectra for the three compo-
nents recorded at TACY during the 2017 earthquake are depicted in Figure 4. Most of the
energy is concentrated in the low period range, which agrees with the nature of the earth-
quake (i.e. normal event). It has been observed by several researchers (e.g. Garcia et al.,
2005; Singh et al., 1985) that normal-faulting inslab earthquakes of Central Mexico decay
faster than interplate events, most likely due to the larger geometrical spreading of these
events. The normal faulting inslab events have considerably larger stress drops than the
interplate ones, which leads to higher levels of the source spectra of inslab earthquakes at
high frequencies, in comparison with thrust events. The characteristic response spectra of
normal events of Central Mexico had most of their energy concentrated in the low period
range. Several months after the earthquake, a seismological station identified as VCRA
was installed at the studied site. Through mid-July 2019, this station recorded the strong
ground motions compiled in Table 1 and the corresponding response spectra shown in
Figure 5.

Case study description

Figure 3 shows the tunnel location and the Mexico City geotechnical zonation. The tunnel
was under construction when the earthquake occurred. The tunnel geometry is shown in
Figure 6a. It was projected to have an external height of 8.6 m and external width of
11 m, and primary and secondary linings. The primary lining is 0.2 m thick, and it is com-
posed of shotcrete reinforced with steel fibers (Figure 6a); the secondary lining is 0.4 m
thick and made of reinforced concrete (Figure 6b). The compression strength of the pri-
mary lining concrete at 28 days, fc0 , is about 25 and 30 MPa for the secondary lining. The
292 Earthquake Spectra 36(S2)

Table 1. Earthquakes registered in VCRA seismological station

Date Epicenter Magnitude (Mw) Epicentral distance (km)

15 July 2018 Magdalena Contreras, Mexico City 2.2 9.2

13 October 2018 Alvaro Obregon, Mexico City 1.8 1.4
12 July 2019 Miguel Hidalgo, Mexico City 2.1 5.2
12 July 2019 Miguel Hidalgo, Mexico City 2.5 4.6
13 July 2019 Miguel Hidalgo, Mexico City 2.3 4.2
13 July 2019 Miguel Hidalgo, Mexico City 2.1 5.7

(a) (b)
Figure 4. (a) Strong ground motion recorder at TACY during the Puebla-Mexico earthquake and (b)
5% damping response spectra.

tunnel runs through the so-called hill zone. From a geological standpoint, this zone falls
within the Tarango formation. Mostly cemented silty sands and sandy silts, dense to very
dense, are found at the area. Often these soils exhibit very large shear strength and low
compressibility.
The tunnel is constructed following four steps, as depicted in Figure 7. Step 1:
Excavation of the upper middle section up to a length of 1.5 m, stabilizing the tunnel walls
with shotcrete reinforced with fibers (i.e. primary lining), and leaving without excavating a
Mayoral et al. 293

Figure 5. Response spectra of earthquakes registered in VCRA.

Figure 6. (a) Tunnel cross section and (b) cross-section of secondary lining.

central volume of soil (i.e. central bench) in order to reduce the risk of tunnel face failure.
Step 2: Once an excavation length of 19.5 m has been reached following the process
described in Step 1, a 9-m-long ramp is built, and the central bench is excavated. Step 3: The
full section of the tunnel walls is excavated alternately in 3 m advances, stabilizing them
again with shotcrete reinforced with fibers. Step 4: Secondary lining, composed of reinforced
concrete, is placed, along with the foundation mat, closing the full section of the tunnel peri-
meter, and Steps 1 to 4 are repeated sequentially, while the excavation is ongoing.

Tunnel instrumentation
Five control points are placed within the tunnel to measure tunnel cross-section distortion
(i.e. convergences and divergences) during the excavation. Sets of 0.0124 m diameter steel
bars, inserted into the ground at a depth of 0.5 m, and embedded in concrete, were used
as deformation control points. The control points are depicted in Figure 8a. These control
294 Earthquake Spectra 36(S2)

Figure 7. Construction procedure: (a) Step 1, (b) Step 2, (c) Step 3, and (d) Step 4.

Figure 8. Tunnel instrumentation: (a) convergences and divergences and (b) topographical elevation
survey references.

points are installed in two phases. Initially, when the upper section of the tunnel is exca-
vated, one control point is installed at the center of the tunnel crown and two others in the
tunnel walls of the upper-middle section (Phase I). Later, when the lower-middle section
of the tunnel is excavated, the control points 4 and 5 are installed on the lateral walls of
the lower-middle section (Phase II). These measurements are complemented with reference
surface points to monitor the settlements along the tunnel axis as depicted in Figure 8b.
Mayoral et al. 295

Table 2. Properties of the idealized soil profile

Initial depth (m) Final depth (m) Material Properties
NSPT c (MPa) f (°) E (MPa)

0 1.9 Fill 12 0.01 20 3–18

1.9 9.1 Sandy clay 50/13 0.06 35 27–41
9.1 14.5 Clayey sand 50/23 0.03 37 15–34
14.5 19.5 Sandy clay 30 0.04 38 7–37
19.5 24.1 Clayey sand 50/12 0.03 40 30–36
24.1 25.9 Sandy clay 25 0.03 37 6–34
25.9 50.4 Clayey sand 50/15 0.03 37 24–34

Subsoil mechanical properties

The tunnel was built in the so-called tobaceus formation found at the north–east region of
the Mexico City valley, which is mostly composed of cemented sandy silts or silty sands,
with variable degrees of cementation, fine content, and grain size distribution. The varying
degrees of cementation and fine content exhibited in these soils lead to some variability in
the strength parameters (i.e. cohesion, c, and friction angle, f), which are also very sensi-
tive to the test type used in its determination. The soil found at the site is a cemented silty
sand and sandy silt. Depending on the amount of cementation, which is directly related to
the cohesive component, the number of standard penetration test (SPT) and blow counts
(N) increase significantly. In those cases, when N values were larger than 50, the values
used in the correlation were extrapolated according to the LRFD Bridge Design
Specifications (2012) AASHTO, as follows in Equation 1:

ð50 + N2 Þ330
SPT = ð1Þ
15 + N3

where N2 and N3 are second and third advances of the SPT. Typical geotechnical proper-
ties of the materials found at the site are summarized in Table 2; they were determined
based on data gathered though geotechnical exploration conducted in the study area dur-
ing the tunnel design. Initially, for the cemented silty sands and sandy silts found at the
site, the elasticity modulus was estimated using Equation 2, which was determined by a
multiple linear regression between the mechanical strength parameters c and f, and the
deformability parameter, E50, developed in previous research (Vital and Mayoral, 2014)
for the stiff tobaceous soils found at the north west Mexico City area.

E50 = 277:3c + 579:25f + 3902:8 ð2Þ

where E50 = elasticity modulus at 50% strain, kN/m2; c = cohesion, kN/m2; and f =
friction angle, in degrees. These values were compared with those obtained from Equation
3, based on SPT blow counts, proposed by Kulhawy and Mayne (1990), respectively.

E = Pa35N60 ð3Þ

where Pa = atmospheric pressure in kN/m2, and N60 is the number of SPT blows, cor-
rected by energy at 60%. Figure 9 shows the soil profile characterization. The tunnel cover
is 27 m at the analyzed section (Figure 9).
296 Earthquake Spectra 36(S2)

Figure 9. Soil profile characterization.

Figure 10. Cracking in the primary lining.

Observed damage
Collapse and major structural damage was observed in on-site structures in some areas
within the so-called lake, transition, and hill zones during the Puebla-Mexico 19 September
2017 earthquake (Asimaki et al., 2020; Franke et al., 2019; Mayoral et al., 2019a, 2019b,
2019c, 2019d; Montgomery et al., 2020; Roman et al., 2019; Wood et al., 2019). Overall,
tunnel seismic performance was considered satisfactory; however, the tunnel discussed
herein presented minor cracking in the primary lining (Figure 10). Stress concentration at
the tunnel-open shaft contact lead to a 1.5-m-long crack, 0.05 mm wide, in the connector
beam between the open shaft and the tunnel (Figure 11). The measured vertical displace-
ment at the tunnel crown and ground surface before and after the earthquake are presented
in Figure 12. As can be noticed, due to the ground shaking, a vertical displacement of
1.3 cm was observed with respect to the static deformation at the tunnel crown and about
0.4 cm at the ground surface.
Mayoral et al. 297

Figure 11. Crack in the beam connector.

Figure 12. Displacement measurements before and after the earthquake in (a) surface and (b) tunnel
crown.
298 Earthquake Spectra 36(S2)

Table 3. Correlations between shear wave velocity and SPT

Author Correlation

Imai and Yoshimura (1970) Vs = 80.6 N0.331

Seed and Idriss (1981) Vs = 61.4 N0.5
Pitilakis et al. (1999) Vs = 145 N0.178
Dikmen (2009) Vs = 73 N0.33
Vs: the shear wave velocity in m/s; N: the number of SPT blow counts.

Shear wave velocity profile

Due to the lack of in situ shear wave velocity profiles, Vs, several well-established empirical
equations were used to estimate Vs in terms of SPT blow counts, N. These criteria are sum-
marized in Table 3. Figure 13 shows the shear wave velocity profiles obtained with the cor-
relations and the idealized Vs profile selected for analyses.
Once the Vs distribution with depth at the studied site was estimated, the compressional
wave velocity, Vp, profile was obtained based on Vs, and Poisson’s ratio, using Equation 4.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2  2n
Vp = Vs ð4Þ
1  2n

where Vs is the shear wave velocity in m/s, N is the number of SPT blow counts, and n is
Poisson’s ratio.

Normalized modulus degradation and damping curves

Due to the practical difficulty associated with sampling the sandy silt and silty sands, the
upper bounds proposed by Seed and Idriss (1970), for normalized modulus degradation
and damping curves, were used for the analyses (Figure 14). These curves have been used
in one-dimensional (1D) wave propagation analyses, where similar materials were found,
to predict the measured response during the 1985 Michoacan earthquake, and other key
seismic events (Mayoral et al., 2008, 2016; Seed et al., 1988) measured at seismological sta-
tions and instrumented sites with vertical arrays. To further establish the effect of the effec-
tive confining stress, these curves were compared with those proposed by Darendeli and
Stokoe (2001), as can be clearly seen in Figure 14, a small effect of this parameter was
observed for the case studied herein.

Numerical model
To assess the tunnel lining performance, tridimensional finite difference models were devel-
oped in FLAC3D (Itasca, 2009), as depicted in Figure 15. An elasto-plastic Mohr-Coulomb
model was used to simulate the stress–strain relationship for the soil for monotonic loading
(i.e. static conditions). Due to the lack of data to calibrate a more robust and complex con-
stitutive model for cyclic loading, the Sig3 model available in FLAC3D was used. This
model followed the Masing rules to match the shear stiffness degradation curves of clays,
silts, and sands found at the site. It was deemed appropriate considering the low to medium
strain level expected to occur during the Puebla-Mexico 19 September 2017 earthquake
due to the soundness of the soils through which the tunnel crosses. To properly
Mayoral et al. 299

Figure 13. Shear wave velocity profile.

Figure 14. (a) Normalized shear modulus and (b) damping curves.
300 Earthquake Spectra 36(S2)

Figure 15. Tunnel numerical model.

Figure 16. Calibration of the numerical model.

characterize the state of stresses before the earthquake, the simulation included the con-
struction process according to the steps previously described, considering excavation
advances lengths of 1.5 m. The primary lining was simulated using shell elements and the
soil with solid elements. The model has a total of 269,456 elements and 285,747 nodes. The
main geotechnical units are a 1.9 m thick fill, underlaid by intercalations of sandy clay,
and clayey sand, down to a depth of 50 m as depicted in Figure 16. From a seismic analysis
standpoint, the thickness of the element was selected based on the geometry and sizes of
both structural elements and soil layers. However, numerical distortion of the propagating
wave can occur in a dynamic analysis as a function of the modeling conditions (Itasca,
2009). Therefore, both the frequency content of the input wave and the wave speed charac-
teristics of the system will affect the numerical accuracy of wave transmission.
For this study, the criterion provided by Kuhlemeyer and Lysmer (1973), regarding the
spatial element size, Dl, was used to accurately represent wave transmission through the
Mayoral et al. 301

Table 4. Optimization process used in the back analysis

Step Process

1 Generation of the three-dimensional finite difference model of the tunnel, using the
properties obtained prior excavation as initial estimates
2 Estimation of displacements at the locations where the measurement stations are located
by means of the three-dimensional finite difference model
3 Estimation of the actual displacements from the convergences and divergences
measurements and survey reference points
4 Comparison of the displacements obtained in Steps 2 and 3. In case there is a considerable
discrepancy, Young’s modulus was adjusted to account for changes associated with the
excavation process at the soil adjacent to the excavation
5 Repeat Step 4, until the error between computed and measure displacements is minimized

numerical models employed. Therefore, Dl was kept smaller than one-fifth of the wave-
length associated with the highest frequency component of the input wave that contains
appreciable energy, fmax (i.e. Dl < l/5). The shortest wavelength l is obtained from
l = Vs/fmax. The smallest average shear wave velocity Vs of the studied site in the upper
less stiff soils (i.e. upper 1 m of the fill) was about 200 m/s, as can be seen in Figure 13, and
the highest significant frequency of the excitation where the energy is concentrated is
around 2-4 Hz. Thus, l ranged between 100 and 50 m. Moreover, the maximum spectral
responses of the excitation are presented at higher frequencies (i.e. 1.0 and 5 Hz), as
depicted in Figure 4. Hence, a Dl of 1.5 m was deemed appropriate. The damping consid-
ered for the shotcrete reinforced with fibers was 5% (Ahmed, 2012) and for steel-reinforced
concrete was 3%. The primary and secondary lining were modeled with shell elements.

Static tunnel performance

During the excavation process, the elastic Young modulus, E, varies according to the
changes in the state of stresses of the soil adjacent to the excavation. In tobaceus materials
(i.e. cemented sandy silts and silty sands), these changes are mostly a function of the degree
of cementation. Back analyses were carried out based on the deformations measured in the
tunnel during construction, slightly adjusting the Young modulus, E, of the soil adjacent
to the excavation within the range of values determined from a statistical analyses of the
mechanical and deformability parameters of the soils found at the site (Mayoral, 2014),
until the error between computed and measured deformations was minimized. The steps
followed, in this iterative optimization process, are summarized in Table 4.
Table 5 summarizes the final parameters considered in the simulations. A comparison
between the displacements predicted by the simulation and those obtained through in situ
measurements are presented in Figure 16 at the tunnel crown. The vertical displacement
contours for the static condition are shown in Figure 17. Under static loading and horizon-
tal ground conditions, the maximum displacement in tunnels occurs at their center, in the
tunnel crown (Golshani et al., 2015; Maji and Adugna, 2016). At the base of the tunnel,
due to the vertical stress reduction associated with the excavation process, there is an elas-
tic expansion of the soil that led to the displacement distribution depicted in Figure 17.
The factors of safety, FS, distribution prevailing at the tunnel prior to the earthquake
were established from the 3D finite difference model developed with the program FLAC3D
(Figure 18). FS was evaluated with the standard definition of failure of safety as the ratio
302 Earthquake Spectra 36(S2)

Table 5. Final elastic modulus, E, obtained by back analysis along the tunnel trace
Initial depth (m) Final depth (m) 0–25 m 25–35 m 35–60 m 60–80 m 80–130 m
E (MPa) E (MPa) E (MPa) E (MPa) E (MPa)

0 1.9 5 5 5 5 5
1.9 9.1 174 174 174 174 174
9.1 14.5 71 44 57 89 81
14.5 19.5 93 41 38 108 257
19.5 24.1 226 167 177 324 623
24.1 25.9 226 167 177 324 623
25.9 50.4 226 167 177 324 623

Figure 17. Vertical displacement contour in static conditions (units in meters).

Figure 18. Static factor of safety computed with t cap/t dem.

of capacity over demand (i.e. t cap/t dem), which allows identifying local instability zones in
the soil mass. Mayoral (2014) proposed using the traditional concept of capacity over
demand to better establish potential local failure zones in tunnels built in brittle stiff silty
sands or sandy silts, with varying degrees of cementation. In this approach, t cap is obtained
Mayoral et al. 303

assuming a Mohr-Coulumb failure criteria as c + soct tan f, where soct is the octahedric
stress defined as (s1 + s2 + s3)/3, and c and f are the soil cohesion and angle of shear
strength, respectively. The acting shear stress in the soil mass, tact, is expressed in terms of
the deviator stress as 1/3((s1 2 s2)2 + (s2 2 s3)2 + (s3 2 s1)2)0.5, where s1, s2, and
s3 are the principal stresses. This approach allows defining overstressed zones within the
soil mass, which may reach a failure state when local stresses are equal to or larger than
local capacity.

Seismic soil structure interaction

Seismic performance evaluation of the tunnel during the Puebla-Mexico 19 September
2017 earthquake was carried out using sets of tridimensional models developed with the
program FLAC3D. The outcropping motion recorded at station TACY, located at about
1.8 km away from the studied site (Figure 3), was deconvolved at the base of the finite dif-
ference model with the computer code SHAKE (Schnabel et al., 1972). The soil type at the
seismological station is composed of very stiff cemented silty sand and sandy silt that exhi-
bit high shear wave velocity values (larger than 650 m/s). Therefore, for the deconvolution
process, the expected soil nonlinearities are negligible from a practical stand point. A fully
nonlinear site response analysis was carried out using the program FLAC3D, to further
study soil nonlinearities at the studied site. The importance of accounting for nonlinear
effects on ground motion attenuation as well as frequency content modification during
strong shaking has been established in the past by several authors for both site response
(Borja et al., 1999; Lee et al., 2006; Yee et al., 2013) and seismic soil–structure interaction
analyses (Borja et al., 1993; Paolucci et al., 2013; Wang et al., 1998). The finite differences
model of the free field is 50 m deep. The free field boundaries implemented in FLAC3D
were used along the edges of the model, and a flexible base was considered at the bottom.
For the dynamic analysis, the three ground motion components shown in Figure 4 were
considered. Although several constitutive models have been developed to account for non-
linearities, there is insufficient experimental data to develop and calibrate a reliable consti-
tutive model. Thus, as previously mentioned, the practical oriented hysteretic model
available in FLAC3D known as ‘‘Sig3’’ was used to approximately deal with both modulus
stiffness degradation and damping variation during the seismic event. This model consid-
ers an ideal soil, in which the stress depends only on the deformation and not on the num-
ber of cycles. With these assumptions, an incremental constitutive relationship of the
degradation curve can be described by t n/g = G/Gmax, where tn is the normalized shear
stress, g is the shear strain, and G/Gmax is the normalized secant modulus. The Sig3 model
is defined in Equation 5:

G a
=  Lx0  ð5Þ
Gmax 1 + exp  b

where L is the logarithmic strain defined as L = log10(g), and the parameters a, b, and
x0, used by the Sig3 model, were obtained by an iterative approach, in which the modulus
degradation curves were fitted with the model equations. The corresponding damping is
given directly by the hysteresis loop during cyclic loading. For this case, the parameters
‘‘a,’’ ‘‘b,’’ and x0 are 1.014, 20.4792, and 21.249, respectively (Figure 14).
Initially, in order to validate the finite difference model proposed for analyses, the free
field response computed with the program FLAC3D considering equivalent linear
304 Earthquake Spectra 36(S2)

Figure 19. Computed response spectra at control points.

properties was compared with that obtained with SHAKE, considering the shear wave
velocity distribution depicted in Figure 13. The program SHAKE has been extensively
calibrated in the past by several researchers (Seed et al., 1988; Seed and Idriss, 1970).
Nonlinear soil behavior is a function of the shaking level, which, if high, leads to shear
stiffness degradation and damping increase. The fact that FLAC3D generates larger damp-
ing at high strains than experimentally derived curves is due to the very well-known limita-
tion of hysteretic type models, which are not able to fully capture simultaneously both
shear stiffness degradation and damping curves developed under steady-state conditions.
However, nonlinear analysis attempts to characterize the transient ground response in each
loading cycle as a function of the evolution of shear strains during ground shaking, rather
than the steady-state response established in the resonant column and cyclic triaxial test
from which modulus degradation and damping curves, such as the Seed and Idriss model
(1970), were developed. The control points used in the 3D numerical model are shown in
Figure 15. The response spectra obtained using SHAKE and FLAC3D are compared in
Figure 19; here the finite element model is able to reproduce both free field conditions and
tunnel–soil interaction effects. The presence of the tunnel, coupled with soil nonlinearities,
modifies the frequency content, leading to migration of the predominant period to the low
period range (i.e. 0.25–0.08 s).

Lining performance
The interaction diagram was used to evaluate the seismic performance of the primary and
secondary linings of the tunnel during the Puebla-Mexico 19 September 2017 earthquake.
The interaction diagram provides the domain in which the combination of bending
moments (M) and normal forces (Nx, Ny) acting on the tunnel lining are below or equal to
the boundary of maximum flexo-compression strength of the lining cross section. This
boundary is obtained from the elastic and geometric parameters of the cross section, con-
sidering the shotcrete reinforced with fibers, for the primary lining, and the concrete and
steel reinforcement, for the secondary lining. As depicted in Figure 20, there are four key
points: Point A, which corresponds to only axial compression, Point B that corresponds
to maximum flexural strength with a given axial strength, Point C is determined as the
flexural strength of the section, and Point D represents the tensile strength of the section,
Mayoral et al. 305

(a) (b)
Figure 20. Interaction diagram of (a) primary lining and (b) secondary lining.

Figure 21. Computed factor of safety at (a) the time when the maximum acceleration is reached
(t = 36 s) and (b) at the end of ground motion.

considering the cross section and material properties depicted in Figure 6. The computed
factor of safety distribution at the time when the maximum acceleration is reached
(t = 36 s) is depicted in Figure 21a and at the end of ground motion (Figure 21b). The
corresponding lining displacements before and after the earthquake are presented in
Figure 22. As can be noticed, the factors of safety during the dynamic event reach a value
close to one. This leads to a maximum seismic tunnel deformation of 2.8 cm in the trans-
versal direction and 4.5 cm of vertical deformation. The predicted permanent seismic-
induced vertical displacements are very close to observed displacements of 5 cm from the
practical stand point, considering that none of the constitutive models employed are able
to fully capture the brittle behavior of the cemented silty sands and sandy silts found at
the studied site. The deformation pattern is also in agreement with the observed crack
location and orientation.
306 Earthquake Spectra 36(S2)

Figure 22. Comparison between the lining displacements before and during the earthquake in the
(a) transversal direction, (b) longitudinal direction, and (c) vertical direction.

The evolution of the seismic demand, in terms of bending moments and axial forces,
acting on the primary lining, was compared against the interaction diagram of the primary
lining, as depicted in Figure 20. At the time when the maximum acceleration is acting in
the tunnel, the seismic demand exceeds the actual lining strength. Thus, the primary lining
fails to withstand the seismic loads induced by the Puebla-Mexico 19 September 2017
earthquake and tends to crack. This agrees with the observed cracking pattern during the
site reconnaissance. These forces, however, fall within the secondary lining interaction dia-
gram. Thus, a good performance of the tunnel is expected during the life time of the
Mayoral et al. 307

Table 6. Earthquakes considered in the analyses

Seismogenic zone Earthquake name Year Moment PGA Duration (s)
magnitude (g)
(Mw)

Normal Montenegro (former Yugoslavia), 1979 6.9 0.251 47.9

MONTE
Umbria Marche (Gubbio-Piana, Italy), 1998 4.8 0.223 40
UMBRIA
Puebla-Mexico City (CU, Mexico), 2010 7.1 0.059 412
CU17
Subduction Michoacan (CU, Mexico), CU85 1985 8.1 0.033 178
Maule (Concepcion SanPedro, Chile), 2010 8.8 0.638 120
CHILE
Honshu (Haga, Japan), JAPAN 2011 9 0.939 300
PGA: peak ground acceleration.

structure, once the secondary lining is in place. The numerical analysis suggests that to
optimize the lining design, it is better to have a more robust primary lining, and a thinner
secondary lining, to decrease the risk during the construction stage in case of a seismic
event.

Parametric study
To further study the effect of period migration toward the low frequency range, and
tunnel-induced ground motion incoherence, the seismic response of the tunnel–soil system
during a series of subduction and normal earthquakes was investigated. Uniform hazard
spectra, UHS, for return periods of 250 years, as recommended in the Mexico City build-
ing code, were used in the simulations to characterize the seismic environment. Details of
the process to generate the UHS can be found in Mayoral et al. (2017a). Table 6 compiles
the ground motions considered in the parametric study. According to the ASCE/SEI
Standard 7-10, when the required number of recorded ground motion is not available,
appropriate simulated ground motion may be used to make up the total number required
(ASCE, 2010). To develop an acceleration time history for which response spectrum rea-
sonably matches the design response spectrum for the return period of analysis (i.e.
T = 250 years), the selected time history, usually called seed ground motion, was modified
using the method proposed by Lilhanand and Tseng (1988) as modified by Abrahamson
(2000). This approach is based on a modification of an acceleration time history to make it
compatible with a user-specified target spectrum. The modification of the time history can
be performed with a variety of different modification models. In doing so, the long period
non-stationary phasing of the original time history is preserved. The 5% damped response
spectra calculated for the modified time histories are compared with the target UHS in
Figures 23 and 24. It can be seen that the response spectra calculated from the modified
time histories reasonably match the target spectra.
The synthetic ground motions are shown in Figures 23 and 24. The 3D numerical
model shown in Figure 15 was used to simulate the response of the tunnel–soil system,
considering the synthetic ground motions shown in Figures 23 and 24. Again, the results
obtained with the 3D finite difference model at the free field, assuming an equivalent lin-
ear analysis, were compared with those computed with SHAKE for model calibration
308 Earthquake Spectra 36(S2)

Figure 23. Synthetic time histories, uniform hazard spectra, and adjusted ground motion response
spectra for normal events: (a) MONTENEGRO, (b) UMBRIA, and (c) CU2017.

purposes. Then, fully nonlinear analyses were carried out with FLAC3D, using the Sig3
model. The response spectra computed using SHAKE and FLAC3D are compared in
Figure 25. As can be noticed, the presence of the tunnel modifies the frequency content in
the spectral accelerations computed in the control points A and B, located at zero and one
diameter away from the tunnel axis, respectively, leading to migration of the predominant
period to the low period range (i.e. 0.25–0.3 s to 0.05–0.08 s), and ground motion incoher-
ence and frequency content modification. Thus, a clear modification of the free field
motions is induced by the presence of the tunnel. This can potentially impact the seismic
performance of low- to medium-rise stiff structures adjacent to the tunnel. These observa-
tions agree with research conducted by other researchers (e.g. Baziar et al., 2014; Wang
et al., 2018), who have studied the effect of the input motion on seismic tunnel–soil inter-
action using shake table and centrifuge tests. They concluded the presence of the under-
ground structure modifies the frequency content and spectral amplitudes of the free field
ground motion.
Mayoral et al. 309

Figure 24. Synthetic time histories, uniform hazard spectra, and adjusted ground motion response
spectra for subduction events: (a) CU1985, (b) CHILE, and (c) JAPAN.

Conclusion
This article presents the seismic performance evaluation of an instrumented tunnel under
construction on rigid soils, located in the western part of Mexico City, during the Puebla-
Mexico 19 September 2017 earthquake. The seismic performance evaluation of the tunnel
was carried out through computed ground deformations, factors of safety, and interaction
diagrams. From the numerical model, the factors of safety during the dynamic event reach
a value close to one. This leads to a maximum seismic tunnel deformation of 2.8 and
4.5 cm in the transversal and vertical directions, respectively. From a practical standpoint,
these results agree with the tunnel deformation of 5 cm observed after the earthquake,
and crack location and orientation observed during the post-earthquake reconnaissance.
The evolution of the seismic demand, in terms of bending moments and axial forces, act-
ing on the primary lining, was compared against the interaction diagram of the primary
lining. At the time when the maximum acceleration is acting in the tunnel, the seismic
demand exceeds the actual lining strength. This agrees with the observed tunnel cracking
along both the longitudinal and transversal directions due to the limited tension strength
capacity of the shotcrete and the compression forces generated during the earthquake in
310 Earthquake Spectra 36(S2)

Figure 25. Comparison between the responses obtained in control points with FLAC3D and SHAKE
for (a) normal events and (b) subduction events.

the transversal direction. Once the soundness of the numerical model was established, a
parametric study was undertaken to assess the effect of the frequency content modification
due to the presence of the tunnel. From this study, it was concluded that the tunnel pres-
ence leads to important frequency content modification in the tunnel surroundings, which
can affect low- to mid-rise stiff structures located nearby. This important finding must be
taken into account when assessing the seismic risk in highly populated urban areas such as
Mexico City.

Acknowledgments
I would like to acknowledge Direccion General de Apoyo al Personal Academico, DGAPA, for their
partial support during this research, through the scholarship program PASPA.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/
or publication of this article.
Mayoral et al. 311

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this
article.

References
Abrahamson N (2000) State of the practice of seismic hazard evaluation. In: Proceedings of GeoEng
2000: An International Conference on Geotechnical & Geological Engineering, Melbourne, VIC,
Australia, 19–24 November, pp.659–685. Lancaster: Technomic.
Ahmed L (2012) Models for analysis of shotcrete on rock exposed to blasting. Licentiate Thesis, Royal
Institute of Technology (KTH), Stockholm.
ASCE (2010) Minimum design loads for buildings and other structures. ASCE/SEI Standard 7-10, 1
May. Reston, VA: American Society of Civil Engineers.
Asimaki D, Mohammadi K, Ayoubi P, Mayora JM and Montalva G (2020) Investigating the
spatial variability of ground motions during the 2017 Mw7.1 Puebla-Mexico City earthquake via
idealized simulations of basin effects. Soil Dynamics and Earthquake Engineering 132: 106073.
Baziar MH, Moghadam MR, Kim DS and Choo YW (2014) Effect of underground tunnel on the
ground surface acceleration. Tunnelling and Underground Space Technology 44: 10–22.
Borja R, Chao H, Montáns F and Lin C (1999) Nonlinear ground response at Lotung LSST site.
Journal of Geotechnical and Geoenvironmental Engineering 125(3): 187–197.
Borja R, Wu W and Smith H (1993) Nonlinear response of vertically oscillating rigid foundations.
Journal of Geotechnical Engineering 119(5): 893–911.
Darendeli M and Stokoe I (2001) Development of a new family of normalized modulus reduction and
material damping curves. Geotechnical engineering report, GD01-12001, 27 August. Austin, TX:
University of Texas at Austin.
Dikmen U (2009) Statistical correlations of shear wave velocity and penetration resistance for soils.
Journal of Geophysics and Engineering 6: 61–72.
Elnashai A, Gencturk B, Kwon O, Hashash Y, Kim S, Jeong S and Dukes J (2012) The Maule
(Chile) earthquake of February 27, 2010: Development of hazard, site specific ground motions
and back-analysis of structures. Soil Dynamics and Earthquake Engineering 42: 229–245.
Federal Highway Administration (FHWA) (2009) Technical manual for design and construction of
road tunnels—Civil elements. Report FHWA-NHI-10-034, January. Washington, DC: FHWA.
Franke K, Candia G, Mayoral JM, Wood C, Montgomery J, Hutchinson T and Morales-Velez A
(2019) Observed building damage patterns and foundation performance in Mexico City following
the 2017 M7.1 Puebla-Mexico City earthquake. Soil Dynamics and Earthquake Engineering 125:
105708.
Garcia D, Singh S, Herraiz M, Ordaz M and Pacheco J (2005) Inslab earthquakes of Central
Mexico: Peak ground motion parameters and response spectra. Bulletin of the Seismological
Society of America 95(6): 2272–2282.
Golshani A, Joneidi M and Majidian S (2015) 3D numerical modeling for construction of tunnels
intersections—Case study of Hakim tunnel. In: The 15th Asian regional conference on soil
mechanics and geotechnical engineering, Fukuoka, Kyushu, Japan, 9–13 November 2015,
pp.1523–1527. Japanese Geotechnical Society Special Publication.
Hashash Y, Hook J, Schmidt B and Yao J (2001) Seismic design and analysis of underground
structures. Tunnelling and Underground Space Technology 16: 247–293.
Imai T and Yoshimura Y (1970) Elastic wave velocity and soil properties in soft soil. Tsuchito-Kiso
18(1): 17–22 (in Japanese).
Itasca Consulting Group. FLAC3D (2009) Fast Lagrangian Analysis of Continua in 3 Dimensions.
Minneapolis, MN: User’s Guide.
Kuhlemeyer R and Lysmer J (1973) Finite element method accuracy for wave propagation problems.
Journal of the Soil Dynamics Division 99: 421–427.
Kulhawy F and Mayne P (1990) Manual on Estimating Soil Properties for Foundation Design. Palo
Alto, CA: Electric Power Research Institute.
312 Earthquake Spectra 36(S2)

Lee C, Tsai Y and Wen K (2006) Analysis if nonlinear site response using the LSST downhole
accelerometer array data. Soil Dynamics and Earthquake Engineering 26: 435–460.
Lilhanand K and Tseng W (1988) Development and application of realistic earthquake time histories
compatible with multiple damping response spectra. In: Proceedings of the 9th world conference on
earthquake engineering, Tokyo, Japan, 2–9 August 1988, Vol. II, pp.819–824. 9WCEE Org.
Committee.
LRFD Bridge Design Specifications (2012) American Association of State Highway and
Transportation Officials (AASHTO).
Maji V and Adugna A (2016) Numerical modelling of tunnelling induced ground deformation and
its control. International Journal of Mining and Geo-Engineering 50(2): 183–188.
Mayoral JM (2014) Performance evaluation of tunnels built in rigid soils. Tunnelling and
Underground Space Technology 43: 1–10.
Mayoral JM, Asimaki D, Tepalcapa S, Wood C, Roman A, Hutchinson T, Franke K and Montalva
G (2019a) Site effects in Mexico City Basin: Past and present. Soil Dynamics and Earthquake
Engineering 121: 369–382.
Mayoral JM, Castañon E, Alcantara L and Tepalcapa S (2016) Seismic response characterization of
high plasticity clays. Journal of Soil Dynamics and Earthquake Engineering 84: 174–189.
Mayoral JM, Castañon E and Albarran J (2017a) Regional subsidence effects on seismic soil-
structure interaction in soft clay. Soil Dynamics and Earthquake Engineering 140: 123–140.
Mayoral JM, De la Rosa D and Alcaraz M (2019b) Seismic performance evaluation of tunnel linings
in stiff soils. In: International conference on earthquake geotechnical engineering, Roma, 17–20
June.
Mayoral JM, De la Rosa D and Tepalcapa S (2019c) Topographic effects during the September 19,
2017 Mexico City earthquake. Soil Dynamics and Earthquake Engineering 125: 105732.
Mayoral JM, Hutchinson T and Franke K (2017b) Geotechnical engineering reconnaissance of the
19 September 2017 mw 7.1 Puebla-Mexico City earthquake, Version 1.0. Geotechnical Extreme
Events Reconnaissance Association. Available at: https://doi.org/10.18118/G6JD46. (Editors
www.geerassociation.org).
Mayoral JM, Romo M and Osorio L (2008) Seismic parameters characterization at Texcoco Lake,
Mexico. Soil Dynamics and Earthquake Engineering 28: 507–521.
Mayoral JM, Tepalcapa S, Roman A, Mohtar C and Rivas R (2019d) Ground subsidence and its
implication on building seismic performance. Soil Dynamics and Earthquake Engineering 126:
105766.
Montgomery J, Candia G, Lemnitzer A and Martinez A (2020) The September 19, 2017 Mw 7.1
Puebla-Mexico City earthquake: Observed rockfall and landslide activity. Soil Dynamics and
Earthquake Engineering 130: 105972.
National Institute of Standards and Technology (NIST) (1996) The January 17, 1995 Hyogoken-
Nanbu (Kobe) Earthquake (NIST Special Publication 901). Gaithersburg, MD: NIST.
Paolucci R, Figini R and Petrinia L (2013) Introducing dynamic nonlinear soil-foundation structure
interaction effects in displacement-based seismic design. Earthquake Spectra 29(2): 475–496.
Pitilakis K, Raptakis D, Lontzetidis KT, Vassilikou T and Jongmans D (1999) Geotechnical and
geophysical description of Euro-Seistests, using field and laboratory tests, and moderate strong
ground motions. Journal of Earthquake Engineering 3: 381–409.
Roman A, Mayoral JM, Hutchinson TC, Candia G, Montgomery J and Tepalcapa S (2019)
Assessment of fragility models based on the Sept 19th, 2017 earthquake observed damage. Soil
Dynamics and Earthquake Engineering 125: 105707.
Schnabel PB, Lysmer J and Seed HB (1972) SHAKE: A computer program for earthquake responser
analysis of horizontally layered sites. Report no. EERC 72–12, December, Berkeley, CA: College
of Engineering, University of California.
Seed HB and Idriss IM (1970) Soil moduli and damping factors for dynamic response analysis. Report
no. UCB/EERC-70/10, December. Berkeley, CA: University of California.
Seed HB and Idriss IM (1981) Evaluation of liquefaction potential sand deposits based on
observation of performance in previous earthquakes. Reston, VA: ASCE National Convention,
pp. 481–544.
Mayoral et al. 313

Seed HB, Romo MP, Sun J, Jaime A and Lysmer J (1988) Relationships between soil conditions and
earthquake ground motions. Journal of Earthquake Spectra 4(2): 687–730.
Sharma S and Judd WR (1991) Underground opening damage from earthquakes. Engineering
Geology 30: 263–276.
Singh SK, Suarez G and Dominguez T (1985) The Oaxaca, Mexico, earthquake of 1931:
Lithospheric normal faulting in the subducted Cocos plate. Nature 317(5): 56–58.
Tsinidis G, de Silva F, Anastasopoulos I, Bilotta E, Bobet A, Hashash Y, He C, Kampas G,
Knappett J, Madabhushi G, Nikitas N, Pitilakis K, Silvestri F, Viggiani G and Fuentes R (2020)
Seismic behaviour of tunnels: From experiments to analysis. Tunnelling and Underground Space
Technology 99: 103334.
U.S. Geological Survey (USGS). Available at: https://www.usgs.gov (accessed September 2017).
Vital D and Mayoral JM (2014) Performance analysis of excavations in rigid soils. In: XXVII
National Meeting of Geotechnical Engineering, Puerto Vallarta, Jalisco, Mexico. 19–21 November
2014 (in Spanish).
Wang G, Yuan M, Miao Y, Wu J and Wang Y (2018) Experimental study on seismic response of
underground tunnel-soil-surface structure interaction system. Tunnelling and Underground Space
Technology 76: 145–159.
Wang S, Kutter BL, Chacko MJ, Wilson DW, Boulanger RW and Abghari A (1998) Nonlinear
seismic soil-pile structure interaction. Earthquake Spectra 14(2): 377–396.
Wood C, Deschenes M, Ledezma C, Meneses J, Montalva G and Morales A (2019) Dynamic site
characterization of areas affected by the 2017 Puebla-Mexico city earthquake. Soil Dynamics and
Earthquake Engineering 125: 105704.
Yee E, Stewart JP and Tokimatsu K (2013) Elastic and large-strain nonlinear seismic site response
from analysis of vertical array recordings. Journal of Geotechnical and Geoenvironmental
Engineering 139: 1789–1801.
Yu H, Chen J, Bobet A and Yuan Y (2016) Damage observation and assessment of the Longxi
tunnel during the Wenchuan earthquake. Tunnelling and Underground Space Technology 54:
102–116.