ENGEO-03274; No of Pages 13

Engineering Geology xxx (2011) xxxxxx

Contents lists available at SciVerse ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic

transfer functions
Edwin A. Obando a, b,, Nils Ryden a, Peter Ulriksen a
a
b

Department of Engineering Geology, Lund Institute of Technology, P.O.B. 118, SE-221 00 Lund, Sweden
Centro de Investigaciones Geocienticas de la Universidad Nacional Autonoma de Nicaragua, (CIGEO-UNAN), P.O. B. A-131 Managua, Nicaragua

a r t i c l e

i n f o

Article history:
Received 25 August 2010
Received in revised form 16 August 2011
Accepted 4 September 2011
Available online xxxx
Keywords:
Vertical arrays
Borehole transfer function
Site response
Site amplication
Site effect
MASW

a b s t r a c t
The best method to evaluate the seismic site response is by means of borehole vertical arrays that use earthquake records from different depths. In this paper we introduce the implementation of a single borehole sensor system (synchronized to a sensor on the surface) that is xed at variable depths within a single well. This
system is used for recording small amplitude earthquake signals at variable stiffness conditions in depth to
compute empirical borehole transfer functions. The computed average empirical borehole transfer functions
allow the estimation of an S-wave velocity model that is constrained using the frequency peak observed in
the H/V ratio curve.
Pairs of surface and borehole earthquake records were obtained with the borehole sensor placed at 10, 20, 50,
and 100 m depth in a test site in Managua, Nicaragua. The average velocity of the nal model down to 100 m
appeared to be in good agreement with the average velocity computed via cross-correlation using the surface and
borehole signals. Likewise, an inverted MASW prole and H/V ratio at the same site agree with the S-wave velocity
model obtained.
2011 Elsevier B.V. All rights reserved.

1. Introduction
When earthquakes occur one of the factors that increases the level of
damage is the amplication of the ground motion of the soft surface
layers overlying hard rock material (Kramer, 1996). In order to minimize the damage during the ground shaking, it is required to evaluate
the site response, which is strongly dependent on the local soil characteristics at the site. Several researchers have proposed different techniques to predict the site response (Borcherdt, 1970; Archuleta et al.,
1992; Lermo and Chavez-Garcia, 1994) and help to understand the effect of the most inuential factor in the ground motion amplication
such as S-wave velocity distribution and dynamic parameters. The
most accurate method to evaluate the site response is by means of borehole vertical arrays (Archuleta et al., 1992; Steidl et al., 1996; Zeghal and
Elgamal, 2000; Bonilla et al., 2002). This system consists of several borehole tri-axial sensors installed in various cased boreholes at various
depths and coupled to tri-axial sensors placed on the free surface
(Archuleta et al., 1992; Elgamal et al., 1996; Kokusho and Sato, 2008).
Vertical array records have been used to evaluate not only the actual
ground amplication accurately, but also a number of aspects related
to dynamic parameters in a wide range of strain levels (Elgamal et al.,

Corresponding author at: Department of Engineering Geology, Lund Institute of

Technology, P.O.B. 118, SE-221 00 Lund, Sweden. Tel.: +505 22703983; fax: + 505
22770613.
E-mail addresses: edwin.obando@tg.lth.se, edwinobando1981@hotmail.com
(E.A. Obando).

1998; Gunturi et al., 1998; Pavlenko and Irikura, 2002; Kwok et al.,
2008). More recently vertical array records have provided the opportunity to estimate the incident wave, as well as the attenuation characteristics of the site (Assimaki et al., 2006; Mehta et al., 2007; Bindi
et al., 2010; Parolai et al., 2010). The S-wave velocity distribution can
also be estimated via the waveform deconvolution method (Mehta et
al., 2007) that computes of the travel time of the incident wave from
one station to another. This technique is also similar to the crosscorrelation technique widely implemented in vertical array data analysis (Elgamal et al., 1996; Assimaki et al., 2006, 2008; Assimaki and
Steidl, 2007). S-wave velocity conguration and attenuation characteristics have been evaluated using an inversion technique that uses the
input (borehole) and output (surface) ground motions (Assimaki et
al., 2006, 2008; Assimaki and Steidl, 2007; Parolai et al., 2010). However, these methods are normally used with data collected in permanent vertical arrays which are not available everywhere. The reason is
that vertical arrays are very expensive mainly due to e.g. drilling and
equipment acquisition which makes its implementation, in many countries prone to earthquakes, prohibitive.
It is clear that the value of borehole records is that both the input
(bottom) and output (free surface) ground motions contain the information of the local site characteristics between the two sensors.
In order to provide an alternative to evaluate the site response
from the surface relative to different depths in earthquake prone
areas and where a high budget is not available, in this paper we propose the use of a multi-depth single borehole sensor set-up. This system allows obtaining pairs of surface and depth ground motion

0013-7952/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.enggeo.2011.09.001

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

records with a single borehole sensor that can be placed at different

depths within the same well. The collected data are used to compute
the borehole transfer functions at multiple depth positions and identify the different resonance modes within a soil column. Based on the
fact that in the linear range the observed resonant modes are dependent on the local shear wave velocity conguration, it is possible to
approximate a shear wave velocity model using the response of the
surface layer relative to multiple depths. The shear wave velocity
model is assumed 1D for the computation of the seismic transfer
function.
Thus, a simple methodology was developed to estimate an S-wave
velocity model whose theoretical transfer function produces the frequency peaks observed on the empirical transfer function from the surface relative to the different depths. For the specic purpose of
computing the 1-D linear seismic transfer function, in this paper the
use of small amplitude earthquakes is required. A limitation of this
set-up is that it does not allow recording simultaneous records at various depths; as a result it will not be possible to evaluate some of the features that vertical arrays normally deal with such as the evaluation of
shear stressstrain histories (Elgamal et al., 1996; Gunturi et al., 1998).
The proposed set-up is tested in Managua, the capital city of
Nicaragua, where small earthquakes are frequent. Thus, from existing
geotechnical information it is well known that throughout Managua
city in general soft layers are in the rst 10 to 15 m depth (Faccioli et
al., 1973). Available SPT (Standard Penetration Test) records at the installation site show that the stiff material starts to appear down to the rst
10 m depth. Since at the site no S-wave velocity information is available, the maximum depth of the borehole was chosen to be 100 m.
At this depth it is expected that a very stiff material exists. To evaluate the stiffness variation along the entire 100 m prole a number of
records are obtained with the borehole sensor placed at 10, 20, 50,
and 100 m depth. A nal velocity model for the site can be estimated
by initially tting the theoretical surface-to-base multiple depth
transfer functions to the empirical borehole transfer functions from
the surface to each corresponding depth. With this initial velocity
model the maximum velocity contrast in identied. The average velocity of the estimated S-wave velocity model is then compared to the
average velocity computed using the travel time estimated using cross-

correlation technique. A nal velocity model is obtained by constraining with the frequency peak observed in the average H/V ratio curve
at the site. The response of the nal S-wave velocity model is compared to the one obtained from surface wave MASW at the same site.
2. Multi-depth single borehole sensor concept
The mobile multi-depth single borehole sensor set-up consists of
one surface tri-axial sensor placed on the free surface and one tri-axial
borehole sensor, installed in a single cased well, which are connected
to a surface digital recorder. The borehole sensor is installed in such a
way that it can be moved to different depths and xed to record earthquake signals at the surface and different depths in a similar manner as
in vertical arrays and evaluate the surface-to-base seismic transfer function at different depths (Figure 1).
The borehole sensor can be moved and xed at different depths
(BH1, BH2, BH3 BHn) using an inatable (or air-bladder) coupling device similar to the one used in conventional refraction S-wave downhole or cross-hole surveys. By computing the seismic response of the
surface relative to variable depths the identication of different resonant modes within the soil prole can be obtained. Then, with this information it is possible to estimate an S-wave velocity model required
for the seismic site response evaluation. This is especially useful when
a geotechnical S-wave velocity model is not available.
A difference compared to normal vertical arrays is that with the
multi-depth set-up the evaluation of the transfer function at the different depths is made with earthquakes of different characteristics.
The possible effect on the subsequent transfer function computations
due to differences in source distances can be minimized when averaging different transfer functions. For the specic purpose of the linear
site response all the earthquake records selected should be of small
amplitude, since including large amplitude events (strong motion records) would introduce higher strain levels outside the linear elastic
range. However, the systems can also be used to record strong motion
records and evaluate nonlinear soil characteristics. A limitation of the
data collected with this system is that features such as the shear
stressstrain histories and the identication of resonant site characteristics (Elgamal et al., 1996; Zeghal et al., 1996; Gunturi et al., 1998) cannot

Fig. 1. Conceptual scheme of borehole vertical arrays and mobile multi-depth single borehole sensor set-up.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

be evaluated; because, it would require simultaneous records from multiple depths which is possible only with vertical arrays fully equipped.
Considering the relative simplicity of this system, it can also be temporarily installed at different sites where the site response evaluation is
required. Thus, in seismic areas a number of earthquake records can be
collected at different sites and evaluate the spatial variation of the site
response in a reasonable time. This is, as mentioned before, especially
benecial in earthquake prone countries where high budgets are not
available.

3. Site description and experimental set-up

Managua, the capital of Nicaragua, has historically been hit by
earthquakes that have devastated the city as e.g. the one occurred
on December 1972 (Algermissen et al., 1974). The seismic environment of the area is dominated by the activity of the subduction area
in the Pacic Ocean region at approximately 150 km from Managua
city and also by the active local fault systems across the city. Within
the subduction activity three seismic sources have been previously
dened (Walther et al., 2000; Castrillo, 2011) where normally small
and moderate magnitude earthquakes between 2 and 6 ML are dominant. The rst one is created by the seismic activity of the volcanic
chain which produces small magnitude. A second earthquake source
is occurring offshore in front of the Pacic coast which is characterized by small to moderate events down to 25 km depth. The third
seismicity cluster occurs in the deeper part of the subduction slab
ranging between 40 and 220 km depth. Fig. 2ab shows the distribution of the earthquakes along the subduction trench in the Pacic
Ocean and a cross section with the depth distribution for the three

seismic areas aforementioned. In the cross-section AB (Figure 2b)

the areas are related to the concentration of events at the different
depths.
This seismic environment is suitable for recording small earthquakes that can be used to evaluate the multi-depth single borehole
sensor set-up. Earthquakes capable of producing strong motion are
likely to occur in the area as well. As mentioned, the local fault system
of the urban area of Managua can release energy enough to cause important damage, as occurred in December 1972 (Algermissen et al.,
1974; Langer et al., 1974).
The soil characteristics of the Managua area in general consist of
quaternary deposits of volcanic origin. The soils in the area are classied
into: Loose supercial soils that mostly consisting of sands. Intermediate and hard soils layers are compounded of tuff, medium dense sand,
gravel, pumice and conglomerate. Based on SPT records (Faccioli et al.,
1973) throughout the city the thicknesses of these materials are variable. Loose supercial soil thicknesses vary between 1 and 5 m. Intermediate and dense soils between 3 and 7 m thickness and very hard
soils are found below 15 m depth. Downhole S-wave velocity proles
have also been measured (Faccioli et al., 1973) in 4 representative
sites within the city and the highest velocity found was around 580
600 m/s which appears between 10 and 15 m depth. The multi-depth
single borehole sensor set-up described is installed in the campus of
the National Autonomous University of Nicaragua (UNAN), in Managua.
The site is located at about 3 km south from the Tiscapa fault that triggered the earthquake in 1972. At the site no reference S-wave velocity
information is available, but from existing geotechnical and geological
information it is known that after the rst 10 m depth stiff soils start
to appear. For the installation of the borehole sensor a well of 21.6 cm
diameter was drilled (cable tool method) down to 100 m depth
where very stiff formations are present. A limitation is that after the
drilling no down-hole or cross-hole S-wave surveys could be made.
But from the collected samples it was observed that a marked transition
occurs close to the 50 m depth (Obando, 2009). Thus, for the 050 m
interval there is a predominance of sandy soils with variable clay content, while for 50100 m interval there is a predominance of coarse
sand that was correlated to Cantera material (local geological name)
which is a stiff volcanic tuff related to Las sierras formation (Shah et
al., 1975). At the site the water table is at 150 m depth.
3.1. Experimental set-up

Fig. 2. (a) Earthquake distribution along the subduction area on the Pacic Ocean and
(b) cross-section of the subduction area (Castrillo, 2011).

The system consists of a tri-axial borehole sensor model FBA ESDH (with a built-in magnetic compass) connected to a K2 recorder
(Kinemetrics, Inc) with an internal tri-axial sensor. The K2 digital recorder is a strong motion recorder of 24 bits and can support up to 6
channels and it is equipped with an external GPS mainly used as a
time reference. The sampling rates of the recorder are 50, 100, 200
and 250 sps (samples per second) with a bandwidth from DC to
200 Hz which meet the different engineering requirements. The
FBA ES-DH borehole sensor can record a maximum acceleration of
4.0 g with a full-scale output of 2.5 V giving a sensitivity factor of
0.625 V/g. The internal sensor of the recorder can record a maximum
acceleration of 0.5 g with a full-scale of 2.5 V and a sensitivity factor
of 5.0 V/g. The borehole sensor includes a cable of 105 m length that
is connected to the K2 recorder via a transient protection box which
provides access to the internal sensor and magnetic compass of the
sensor package. Prior to the installation the wiring and connections
were checked and tested following the instructions in the manual
provided by the manufacturer. To record small amplitude earthquakes the system was set to be only triggered by the borehole channels by using the STA/LTA (short-time average/long-time average).
For the data collection we used 200 sps (some were recorded at
100 sps) sampling rate.
Before the installation of the borehole sensor the well has to be prepared. Thus, the 100 m drilled borehole was cased with PVC tubes of

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

10.16 cm diameter (following the installation requirements provided

by Kinemetrics) and grouted (backlled with sand pack to meet the
local soil conditions). The bottom of the casing was sealed with a PVC
slip cap. To handle the borehole sensor and the coupling mechanism
easily, the well was not lled with water completely as in normal borehole vertical array installations. A portion of 20 m of water was gradually poured during the location of the grouting material once the entire
casing was installed to avoid buoyancy between the bottom of the
well and the casing. To prevent a possible collapse of the well the type
of casing used was strong enough to resist lateral external pressures
exerted by the surrounding ground. It is assumed that possible vibrations due to tube waves or the casing itself does not affect the collected
records signicantly.
The external cable of the borehole sensor of 105 m length (with a
Kevlar core) is handled with a manual winch which is installed permanently. A coupling mechanism to x the borehole sensor within
the well consists of an inatable device that is attached to the borehole sensor package (Figure 3ab). Initially, the borehole sensor is
lowered, using the manual winch to a desired position and oriented
to the North using its calibrated internal magnetic compass. Then
the inatable device is lled with air providing pressure of about 3
bars to x the sensor against one side of the casing (this pressure
was tested before lowering the sensor package). The pressure provided appeared to be enough to x the sensor and capture the actual
ground vibration. Although similar coupling systems are available in
the market, an efcient coupling system can be conveniently assembled at a probably lower cost.
The surface recorder with a tri-axial sensor built-in is attached on
top of a concrete plate of square shape located on top of the borehole
and large enough to place the K2 recorder on. The concrete plate provides adequate coupling to the ground to capture the ground motion
properly. To provide protection to all the components a small container was placed on top of the well. The oor of the container was
isolated from the concrete plate and the borehole by opening a square
hole slightly larger to avoid any contact and in that way to minimize
potential structure borne noise from the container.
Finally when the system was installed the external GPS antenna of
the K2 was synchronized to the local time (GMT-6 h). This would
allow retrieving the focal parameters of the recorded earthquakes
that are reported by the Nicaraguan seismic network operated by

INETER (Instituto Nicaraguense de Estudios Territoriales). The manual winch, attached to the oor of the container, is isolated from the
concrete plate to avoid possible external vibrations introduced to
the surface recorder.
4. Collected data
During the years 2008 and 2009 a number of 20 small amplitude
earthquakes were recorded with the borehole sensor at different
depths at different periods of time. The data recorded consist of 6channel records with the sensors located at the surface, 10, 20,
50, and 100 m depth. More levels are planned to be surveyed in
the future. The focal parameters presented in Table 1 were provided
by INETER in its daily report published online (http://www.ineter.
gob.ni/geosica/sis/monitor.html). Note that for depths of 10,
20, and 50 m at least 3 earthquake events were recorded. The
magnitude of the events reported by INETER ranged between 2.4
and 5.9 ML with epicentral distance between 4 and 323 km. Local
events are attributed to the local faulting system and to the volcanic
activity nearby the city, while distant earthquakes are related to the
subduction activity in the Pacic Ocean. For our analyses we are interested in records with amplitude accelerations low enough to obtain
elastic behavior. A parameter to obtain linear elastic behavior is by
selecting events with small PGA (peak ground acceleration). Thus,
the PGA should not exceed 0.1 g (Assimaki et al., 2006). However,
nonlinear behavior could be triggered with the combination of the
input ground motion level and soil column strength. The overall
PGA value of the collected records is around 10 cm/s/s (0.1 m/s/s) almost 10 times lower than 0.1 g (0.981 m/s/s) and linear elastic behavior is assumed.
Examples of the records collected with the borehole sensor located at the different depths are illustrated in Fig. 4ad. On visual inspection each recorded waveform with the borehole sensor located at
10, 20 and 50 m appear to be consistent to the ones recorded
at the free surface, but with smaller amplitude accelerations as
expected. Also, the waveform along the entire time record appears
to be similar in P and S-wave arrivals and duration at both surface
and borehole records. This, a priori, indicates that the borehole sensor
seems to properly capture the actual ground vibration at the different
depths.

Fig. 3. (a) Fixing coupling device to the FBA ES-DH triaxial borehole sensor, (b) coupling device fed with air.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Table 1
Recorded earthquakes at the surface and with the borehole sensor down to 10, 20, 50 and 100 m.
Eq

Epicenter

Date

Local time

(dd/mm/yy)

(GMT-6)

Coordinates

(km)

(ML)

Depth

Magnitude

Surface/10 m
1
2
3

108 km, West

17 km, NW
100 km, west

31/08/2009
29/09/2009
11/12/2009

03:31:00
20:08:00
08:01:39

11.80
12.24
11.79

87.18
86.38
87.0

7.0
4.0
34.4

5.0
3.1
3.8

Surface/20 m
4
5
6

14 km, West
113 km, SW
116 km, West

09/01/2009
18/01/2009
18/01/2009

23:09:00
08:18:00
21:05:00

12.19
11.56
12.37

86.38
87.10
87.76

5.0
53.0
60.0

2.8
4.6
4.8

Surface/50 m
7
8
9
10

53 km, South
144 km, West
123 km, West
323 km, NW

12/10/2008
13/11/2008
14/11/2008
15/11/2008

09:06:00
03:49:00
01:13:00
17:03:00

11.87
10.85
12.33
13.24

86.65
86.26
87.37
89.01

75.0
68.0
72.0
1.0

3.3
5.1
4.5
5.9

Surface/100 m
11
12
13
14
15
16
17
18
19
20

108 km, SW
211 km, West
54 km, SW
87 km, SW
117 km, West
48 km NW
47 km, SW
111 km SE
9 km, NW
4 km, NW

02/02/2008
12/02/2008
13/02/2008
20/02/2008
19/03/2008
24/03/2008
27/04/2008
28/04/2008
10/05/2008
16/05/2008

01:58:00
21:02:00
14:52:31
02:00:12
05:58:00
15:53:00
17:32:00
18:03:00
20:40:00
18:57:00

11.29
12.48
11.69
11.33
12.22
12.48
11.80
11.42
12.22
12.46

86.71
88.17
86.54
86.26
87.33
86.54
86.50
85.55
86.31
86.53

56.0
1.0
66.4
63.7
57.0
4.0
101.0
177.0
10.0
6.0

3.6
4.0
2.4
3.1
4.2
3.7
4.0
3.4
2.4
3.3

Focal parameters are provided by INETER.

5. Methods

The single spectral ratio (SSR) method using two accelerograms

can be represented by:

5.1. Surface-to-borehole response

To compute the empirical borehole response (transfer function) we
use the single spectral ratio (SSR) and the cross-power spectral ratio
(CPSR) methods. Both methods use the Welch's averaged, modied
periodogram method of spectral estimation (Proakis and Manolakis,
1996). The averaging is applied in time domain dividing the discretesignal in sub-windows.

SSR

q
Pii f =Pjj f

Where Pii(f) and Pjj(f ) are the auto-power spectral density function for the input j and output i signals respectively. A more accurate
estimate of the surface-to-base transfer function is by means of the
cross-power spectral density ratio (CPSR) method (Assimaki et al.,

Fig. 4. 6-channel earthquake records with the borehole sensor at (a) 10 m, (b) 20 m, (c) 50 m and (d) 100 m.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Fig. 5. Theoretical idealization of a 1-D multi-layer soil system for computing the theoretical transfer function.

2008). The CPSR method computes the cross-power spectral density

Pij of an input signal j and output signal i and the auto-power spectral
density Pjj of the input signal j.
CPSR Pij f =Pjj f

Eq. (2) can be dened as the product of the SSR by the magnitude
square of coherence (Steidl et al., 1996; Safak, 1999):

2
Cij Pij f  =Pii f Pjj f

Since the coherence (Eq. (3)) varies from 0 to 1, decay in coherence will cause the amplitude of the CPSR to drop. This decaying indicates the destructive interference of the downgoing waves which

is normally present in shallow boreholes and can be considered by

using Eq. (2) as suggested by Steidl et.al, (1996).
To properly capture the total horizontal ground motion, for all the
computed transfer functions in this paper, we use the formulation
implemented by Steidl et al., (1996) and more recently by Assimaki
et al., (2008). The assumption is that the total ground motion can
be successfully captured by using the total acceleration from both
components without the need for orientation considerations. So,
we write:
aT t aEW t iaNS t

aEW(t) and aNS(t) are the Eastwest and Northsouth components of

the ground motion. This complex representation of the total ground

Fig. 6. (a) Selected portions for the signal-to-noise ratio analysis, (b) smoothed vs. unsmoothed FAS, (c) SNR calculation using smoothed FAS, and (d) SNR for selected surface and
borehole records.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

motion maximizes the total amplitude at a given frequency and preserves the phase information between both components.

of the layers between sensors (Elgamal et al., 1996). Alternatively the

computation can be made in time domain. However, it can be time consuming when using large records.

5.2. Cross-correlation
5.3. S-wave velocity inversion via seismic transfer function constrain
The cross-correlation between the signals recorded at the surface
and the different depths has been widely implemented using vertical
array records (Elgamal et al., 1998; Assimaki et al., 2006, 2008;
Assimaki and Steidl, 2007). For our calculations we implement the
cross-correlation in the frequency domain that uses the discrete
Fourier Transform (Davis, 2000) which provides faster time for computation. Thus the cross-correlation can be represented as the product of the complex representation of both input and output signals as
in Eq. (5).


tk Jk Ik

Where Jk represents the Fourier Transform of the input signal j and I*k
is the complex conjugate of the output signal i and the separation (difference in depths) between both sensors is dened asH. is computed
from the inverse Fourier transform of the complex product of Eq. (5).
The peak occurring in the positive axis in the cross-correlation occurs at a time lag td and corresponds to the incident wave of vertical
propagation. In knowing the distance, H, between both sensors, the
average apparent velocity, va, is calculated as:
va H=td

The soil media can be idealized as a stack of nite number of horizontal layers of variable thickness (h), S-wave velocity (Vs), density
() and damping () overlying an elastic half-space (Figure 5). The incoming incident wave is assumed to be propagating vertically through
the layer boundaries causing transmitted and reected wave components along the direction of propagation. Soil material is approximated
as a KelvinVoigt solid of visco-elastic behavior. After using some
boundary conditions the theoretical borehole transfer function, F(f),
can be dened as:

The apparent velocity between the sensors computed by Eq. (6) can
be assumed to be equivalent to the actual average shear wave velocity

F f 2 As =Ab Bb

Eq. (7) denes the theoretical transfer function as the ratio of the
up-going wave component of the surface As to the down-going Bb and
up-going Ab of the elastic bedrock. The surface amplitude As is multiplied by 2 due to the free surface condition. Notice that for the case of
an outcrop rock site both Ab and Bb are the same. For a detailed description on the derivation of the transfer function for multiple number of layers the reader can refer to (Kramer, 1996; Robinson et al.,
2006).
Eq. (7) can be used to relate the surface layer to any point within
the layer model. Since the 1-D transfer function is linear it can be used

Fig. 7. Spectrograms for pair of earthquake records (surface and 50 m) using (a) local event and (b) distant event.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Fig. 8. Computed transfer functions and coherence for (a) surface/10 m, (b) surface/20 m, (c) surface/50 m and (d) surface/100 m.

for example to relate the surface ground motion to all the layers of the
prole by means of deconvolution (Kramer, 1996).
In Fig. 5 borehole measurements (dotted lines rectangle) can be associated to the layering distribution of the idealized 1-D model. This
means there should be a layering distribution whose theoretical 1-D
transfer function corresponds to the peak frequencies observed in the
empirical 1-D transfer functions. This is based on the fact that the
shape of the theoretical linear borehole transfer function is dominated
by two factors. The frequencies where the larger amplications appear
are dependent on the S-wave velocity conguration of the soil media
(thickness and velocities). On the other hand, the level of amplication
depends mostly on the damping. Thus, considering small amplitude
earthquakes the obtained 1-D borehole transfer function is assumed
to be linear and a derivation of a 1-D S-wave velocity model can be
done.
The derivation of an S-wave velocity model using borehole response
from the surface-to- the different borehole depths is as follows:

6. Data processing and analysis

An important step is to evaluate the quality of the different small amplitude earthquake records with the borehole sensor at different depths.
Thus, we evaluate the signal-to-noise ratio of individual borehole and
surface signals. Initially, we remove the DC component in the collected
data, by applying a baseline correction (using a low order polynomial tting). The signal-to-noise ratio of each signal was evaluated by comparing the Fourier amplitude spectra (FAS) of the noise and signal.
The noise and signal amplitude spectra are computed from 8 s selected from the pre-event and S-wave portion of the signal respectively (Figure 6a). An acceptable lower limit is of SNR N 3. Note that
previous to the FAS computation the portions selected are multiplied
by a Hanning-tapered window to minimize the effect on transients in
the end of the signals (Figure 6c). Then, before the SNR computation
the resulting FAS are smoothed using the movavg (moving average) function built-in Matlab using the linear moving average option. Enough level
of smoothing was implemented using a 10 points window. A comparison

a. An initial S-wave velocity prole is obtained by using the empirical

transfer function from the surface-to-multiple depth earthquake records. The initial velocity model in general should reproduce approximately the frequency peaks observed in the empirical transfer
function from the surface down to the different depths. The average
S-wave velocity (Eq. 8) of the initial model should be comparable to
the apparent velocity using the time lag obtained via cross-correlation
(Eq. 6).


v s hi = hi =vsi

b. This initial velocity model obtained in (a) is used to calculate its corresponding dispersion curve using an available forward model (Kausel
and Roesset, 1981). The dispersion curve is computed from a layer
model with a maximum depth equal to the depth where the maximum velocity contrast appears.
c. The computed dispersion curve is used as a target dispersion curve
in the inversion which is constrained using the frequency peak observed in the H/V ratio curve.
The inversion is performed using dinver tool from geopsy (http://
www.geopsy.org/).

Fig. 9. Average empirical borehole transfer functions for (a) surface/20 m, (b)
surface/50 m and (c) surface/100 m.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Fig. 10. (a) Estimated velocity model for the UNAN-Managua site, (b)(d) tted transfer functions. In gures bd, thick black lines are theoretical transfer functions and thin lines
correspond to empirical ones.

between the smoothed and unsmoothed spectrum of the EW component of the May 10, 2008 event (2.4 ML) is presented in Fig. 6b. The
SNR presented correspond to the best ratios at each borehole sensor position and the surface.
In Fig. 6d the SNR for the EW component of individual events at all
borehole depths and the surface are displayed together. Note that in
most of the cases the SNR in the frequency range of 1.210 Hz is
higher than 3. A marked decay in SNR occurs in a record at 10 m.
This may be related to possible complications as waves approach the
surface.
As a simple test we examine the consistency of the surface-borehole
pair of signals by comparing the frequency content variation with time.
This means we compare earthquakes of different characteristics where
the spectral content varies according to the epicentral distance which is
a very well known issue. Thus, for earthquakes that occurred at very
long distances the spectral content would dominate low frequencies
while local events should enhance higher frequencies. This feature
should be observed in the pair of surface and borehole signals. In
Fig. 7, we display the signal recorded at the surface and at 50 m
depth that correspond to two earthquakes with marked differences in
magnitude, distance and azimuth of the source.
To describe the frequency content variation along the time series
we use the spectrogram method using a sliding window of 1.28 s
that is equal to a 256 point of FFT for a sampling rate of 200 sps. In
Fig. 7a, the pairs of surface and borehole ( 50 m depth) signals are
plotted together with their respective spectrograms for an event
recorded at 53 km from Managua. The spectrograms show that
along the record the major energy correspond to the main event
that starts at about 20 s. For both signals, at the surface and borehole,
the same energy trend appears to be consistent in time and frequency
domain between 1 and 15 Hz. In Fig. 7b the records correspond to a
distant earthquake that occurred at 323 km north and within the subduction zone. Here the spectrograms show a dominance of the energy
of low frequency below 10 Hz which is in accordance with the long
wavelengths enhanced by the long distance to the source. In both

Table 2
Modeling parameters for the initial S-wave velocity estimation at the UNAN-Managua
site.
Depth interval

Maximum depth (m)

Number of layer

(ton/m3)

(%)

020
2050
50100

20
50
100

4
2
2

1.5
2.2
2.5

7
7
7

surface and borehole signals the major spectral amplitudes occurs at

the S-wave portion that starts at about 38 s (as highlighted by a dotted line-square) and the frequency range appears to be similar between 1 and 10 Hz.
7. Results and discussion
7.1. Empirical transfer functions at the UNAN-Managua site
The empirical transfer functions computed using both the SSR
(Single Spectral Ratio) and CPSR (Cross-Power Spectral Ratio) are
compared to the coherence function calculated using the same portion
of the selected records. This allows evaluating the occurrence of the destructive interference of the downgoing wave effect, normally present
on borehole records (Steidl et al., 1996; Safak, 1999). With this comparison we intend to evaluate the downgoing wave effect using the collected data especially at intermediate depths and the empirical transfer
functions at all depths under study. For computing the empirical transfer function relative to each borehole depth, at least 3 records with high
signal-to-noise ratio were selected. Initially each record is rst band
pass ltered (0.820 Hz) using a causal Butterworth lter of degree 4.
Thus we select a 10 s window starting from approximately 0.5 s before
the S-wave onset. The power density spectra are computed using 1024
(512 when sampling rate is 100 sps) number of points in the FFT with
50% overlapping. Prior to the computation of the transfer functions
and the magnitude square coherence each spectrum is smoothed
using the movavg function (selecting the linear moving average option)
using a 10 points window. This window length appears to provide adequate smoothing producing clear and consistent peaks in the empirical
transfer functions. Note that for using the same level of smoothing (for
100 and 200 sps) the spectra were interpolated to a common number of
samples.
Fig. 8 shows the empirical transfer functions for a single pair of records for the four depths under study. Fig. 8a shows the surface and
10 m depth and apparently no resonance occurs. Considering that
the coherence between both signals is very high, the absence of resonance may indicate that the rst 10 m of soil is probably uniform.
But in the case of the surface/20 m transfer function (Figure 8b) a
prominent frequency peak appears close to 5.0 Hz. It is observed that
the decay in coherence causes that the amplitude appears lower with
CPSR compared to the SSR. The decay in coherence also coincides with
the resonance more close to 5.0 Hz.
In the cases of surface/50 m and surface/100 m (Figure 8cd)
the situation becomes complicated. The coherence between the two

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

10

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Fig. 11. (a) Surface and 100 m signals, (b) corresponding cross-correlation, and (c) average cross-correlation using EW and NS components from 4 earthquake events.

signals appears to be much lower. This causes that the two (Figure 8c)
and three modes (Figure 8d) (110 Hz) observed appear with lower
amplication with the CPSR compared to the SSR method. This is something expected because the level of coherence decays with a larger distance between sensors.
It is observed in Fig. 8cd that at the low frequencies the level of
amplication is lower compared to higher frequencies. This aspect is
in part related to the frequency dependent-damping of the transfer
functions which appears to decrease with frequency as discussed in
previous works (Kokusho and Sato, 2008). In this paper the attenuation characteristics are not addressed. Rather, considering its inherent
complexity we believe it has to be treated separately.
7.2. Inverted S-wave velocity prole at the site
An estimation of an S-wave velocity model for the site is made as
described in section 5.3. The estimation of the S-wave velocity model

is made using the average empirical transfer functions obtained with

the CPSR method. The average empirical transfer functions for the
surface relative to 20, 50, and 100 m depth are displayed in
Fig. 9. For the calculation at least 3 records were selected at each
depth. The three averaged transfer functions show marked vibration
modes. In Fig. 9a, a mode appears at about 5.2 Hz with an amplication of 5. This may indicate that a rst velocity transition occurs within the rst 20 m, but more precisely between 10 and 20 m
considering that at the surface/10 m no amplication was observed.
In Fig. 9b, two modes are observed at 2.6 Hz and a frequency band between 5 and 6 Hz approximately. The two modes also indicates that
probably a second velocity transition that is related to the 2.6 Hz
start to appear between 20 and 50 m depth. Finally, in Fig. 9c
the deepest borehole position three modes occur at about 6.5, 4.5,
and 1.8 Hz (1.82.5 Hz could also be considered).
As explained the inversion of a velocity model for the site is
started by initially approximating a velocity model down to the rst

Fig. 12. (a) Initial S-wave velocity model for UNAN-Managua site down to 100 m, (b) experimental dispersion curve computed for the upper 50 m thickness of the initial model
displayed in (a), (c) H/V ratio curves collected at UNAN-Managua site (thick black line corresponds to the average H/V ratio curve).

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Fig. 13. (a) Inverted velocity model using dispersion curve of Fig. 12b, (b) corresponding
dispersion curves. Target dispersion curve in (b) is represented by circular shapes.

100 m (Figure 10a) and the maximum velocity contrast is identied. A reasonable velocity model, at least, should contain the most
important velocity contrasts within the prole. In Table 2, the parameters used to obtain the theoretical transfer functions are described. It
is motivated to use a ner layer interval in the shallower depths to
help make the searching more exible. For that reason 4 layers are
used for the rst 20 m and 2 for the two other intervals. The densities were approximated based on existing information in the area of
Managua (Faccioli et al., 1973). The level of damping was selected
for better comparison in amplication with the empirical transfer
functions. The theoretical computation of the transfer function is
made using the SUA software (Robinson et al., 2006) based on Matlab.
The code can be downloaded on the web (http://www.iamg.org/
CGEditor/index.htm). We set the target depth as 100 m which is the
maximum depth of the borehole.
The parameters selected are considered as appropriate to approximate a velocity model capable of reproducing the same frequency
peaks observed in the empirical transfer function down to 100 m
depth (Figure 10bd). In Fig. 10a the maximum velocity variation
seems to appear down to 50 m depth. This seems to be consistent
with the existing geological and geotechnical information which suggest the occurrence of the very stiff material close to 50 m depth.
The resulting velocity model is compared to the average velocity
estimated via cross-correlation analysis. Thus, for the evaluation of
the overall velocity model we take the surface and 100 m pair of records. We use 4 earthquake records with the highest SNR ratio.
In Fig. 11a, the cross-correlation is computed for the EW component
of one of the selected earthquakes. For the cross-correlation computation
(between pairs of surface and borehole signals) a window of 5 s that contain the S-wave portion of the ltered signal were selected. The corresponding time lag where the maximum correlation occurs is at 0.18 s
that for a 100 m travel distance corresponds to a velocity of 555.56 m/s.
The average cross-correlation displayed in Fig. 11c is calculated by averaging EW and NS components of the 4 records. Two out of eight components were omitted due to a poor correlation. In general the time lags
appear to be consistent in spite of having used earthquakes of different
characteristics. The average velocity of the initial S-wave velocity model
obtained via transfer functions (Figure 10a) is about 548.5 m/s. By taking
555.56 m/s as the true average, the corresponding error of the initial velocity model with respect to the average velocity obtained via crosscorrelation is 1.2%. This difference is almost insignicant.
From the initial velocity model the maximum impedance contrasts
correspond to the maximum velocity variation down to 50 m depth
(Figure 12a). The layer distribution of the rst 50 m thickness is used
to obtain an experimental or target dispersion curve (Figure 12b).
This dispersion curve is used to look for, through an inversion procedure

11

(dinver/geopsy), possible equivalent velocity models that describe similar empirical transfer functions.
At the same place ambient noise records were also collected. This
will allow the evaluation of the H/V ratio technique which provides information of the fundamental frequency attributed to the maximum
impedance contrast at the site (Bonnefoy-Claudet et al., 2006). Noise
measurements were obtained at 8 points in the surroundings of the station and the duration of each record was of 60 s at 100 sps. The H/V ratios were computed using SSR technique (but using the horizontal and
the vertical components) explained in the method section. For each record two windows of 20 s with no spikes were selected. Then, the number of FFT used was 512 which is equivalent to 5.12 s with 50% overlap.
The mode observed at 2.53 Hz seems to correspond to the fundamental frequency showed in the H/V ratio curve (Figure 12c) that seems to
be related to the upper 50 m thickness.
In Fig. 13a the inverted velocity model prole is displayed and the
corresponding computed dispersion curves are displayed in Fig. 13b.
The colorful images are the mist value for the different velocity models
and dispersion curves. The lower mist values correspond to the best
models. For the computed dispersion curves in Fig. 13b the lowest mist
appears to capture the shape of the experimental dispersion curve
(circular marks) accurately. In the corresponding inverted velocity proles (Figure 13a) the velocity distribution appears to be rather consistent. Therefore, a rst velocity contrast (from 200 m/s to 580 m/s)
seems to consistently occur down to 810 m depth. Then a velocity reduction from 580 to 440 m/s occurs between 20 and 50 m depth. Finally
the very stiff or rock material with VsN 760 m/s occurs between 48
and 52 m depth.
A second inversion can be made by including the frequency peak
observed in the H/V ratio curve displayed in Fig. 12c. So, the inversion
is constrained using a frequency of 2.5 Hz that is the frequency where
a maximum amplication in the H/V ratio curve appears (Figure 14).
This time the inversion uses very narrow velocity and thickness limits
close to the solution observed in Fig. 13a. As a result a nal inverted
prole, displayed in Fig. 14a, shows very small variation respect to
the velocity model obtained in Fig. 13a and the corresponding computed dispersion curves follow the experimental one accurately
(Figure 14b). The constrained inversion implies that the depth to
rock material is indeed close to 50 m interface. Notice that the rst
peak of the theoretical ellipticity curve in Fig. 14c reects the constrained frequency used in the inversion and it is attributed to the
rst 50 m thickness.

Fig. 14. (a) Inverted velocity model, (b) corresponding dispersion curves, and (c) theoretical ellipticity H/V curve with frequency peak of 2.5 Hz.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

12

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

Fig. 15. (a) Inverted S-wave velocity models using MASW dispersion curve, (b) extracted dispersion curve; (c) computed and MASW dispersion curves at UNAN-Managua site.

7.3. Comparing computed S-wave velocity model to inverted MASW surface

wave prole
At UNAN-Managua site low frequency surface wave measurements
are also available and a velocity model down to 50 m depth can be estimated. The surface wave method used is the Multichannel Analysis of
Surface Waves (MASW) method. This method is implemented using a
road-bump low frequency source to generate focused low frequency
energy produced by the trafc and appears to be an alternative to retrieve long wavelengths (Obando et al., 2009). For the selection of the
maximum wavelength we took into consideration the criterion that
suggests that the maximum wavelength resolved is equivalent to the

length of the receiver spread and that the maximum depth can be
about half the maximum wavelength (Park et al., 1999). This criterion
has often been used in site response applications (Mahajan et al.,
2007; Mahajan, 2009) and seems to provide a general picture of the
soil stiffness characteristics required for site response evaluation. Although more accurate results can be obtained if using a more conservative criterion e.g. (Bodet et al., 2005), we selected longer wavelengths in
order to compare down to 50 m.
The experimental dispersion curve for the site was retrieved using
a receiver spread of 115 m length (24 receivers with 5 m separation).
The MASW survey line was placed a few meters from the borehole station. So, the retrieved velocity model is considered to be representative

Fig. 16. (a) Initial, inverted MASW and constrained velocity models, (b) corresponding theoretical and empirical (surface/50 m) transfer functions.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001

E.A. Obando et al. / Engineering Geology xxx (2011) xxxxxx

for the site. From the dispersion image (Figure 15b), it is observed that
at 5.0 Hz the phase velocity is 550 m/s approximately that is equivalent
to 110 m wavelength which fulll the criterion by Park et al., (1999).
Then, it is feasible to resolve a maximum depth of about 55 m. The inversion of the velocity model displayed in Fig. 15a and c is made using
dinvert tool (http://www.geopsy.org/).
In Fig. 16a the initial velocity model obtained via transfer function,
inverted velocity (constrained) and the inverted MASW are displayed. The three velocity models show the occurrence of the rock
material down to 48 to 50 m depth which conrms the observation
made with the previous geological information. In Fig. 16b the theoretical borehole transfer functions of the two models (constrained
and inverted MASW) compare well with the empirical transfer function of the surface/50 m depth. Both MASW and inverted model
predict a very similar frequency peak at 2.5 Hz, although some difference in the second mode close to 6.0 Hz appears more noticeable
mainly due to the variation in the rst 20 m in Fig. 16a. This variation
in the rst 20 m depth may be explained by the fact that the inversion
of the MASW dispersion curve was made using the fundamental
mode only. This implies that a multi-modal inversion could provide
better results especially at shallower depths.
8. Conclusions
Results presented in this paper show that the collected records at
the surface and different depths contain valuable information about
the local soil conditions. Thus, these results suggest that the multidepth single borehole sensor set-up could be used in sites with similar seismic environments to evaluate the local site characteristics.
This system can be a suitable alternative when a high budget is not
available. Considering that the installation of this system is relatively
simple, it could be used as a temporary station for collecting records
at different depths and at different sites where the seismic site response evaluation is required. Although in this paper only small amplitude earthquakes were used, the implementation of this system
may be extended for strong motion recording to analyze non-linear
characteristics of a site of interest.
Acknowledgment
This research was sponsored by the Swedish International Development Agency (SIDA) through a cooperation grant with the National
Autonomous University of Nicaragua (UNAN-Managua) and the Lund
Institute of Technology at Lund University in Sweden. Focal distribution information was provided by INETER. We want to thank Dr. Fabian Bonilla and the anonymous reviewers for their valuable comments
and suggestions.
References
Algermissen, S.T., Dewey, J.W., Langer, C.J., Dillinger, W.H., 1974. The Managua, Nicaragua,
earthquake of December 23, 1972: location, focal mechanism, and intensity distribution. Bulletin of the Seismological Society of America 64, 9931004.
Archuleta, R.J., Seale, S.H., Sangas, P.V., Baker, L.M., Swain, S.T., 1992. Garner Valley
downhole array of accelerometers; instrumentation and preliminary data analysis.
Bulletin of the Seismological Society of America 82, 15921621.
Assimaki, D., Steidl, J., 2007. Inverse analysis of weak and strong motion downhole array
data from the M b sub N b/subN w7.0 Sanriku-Minami earthquake. Soil Dynamics and
Earthquake Engineering 27, 7392.
Assimaki, D., Steidl, J., Liu, P.C., 2006. Attenuation and velocity structure for site response
analyses via downhole seismogram inversion. Pure and applied geophysics PAGEOPH
163, 81118.
Assimaki, D., Li, W., Steidl, J.H., Tsuda, K., 2008. Site amplication and attenuation via
downhole array seismogram inversion; a comparative study of the 2003 Miyagi-Oki
aftershock sequence. Bulletin of the Seismological Society of America 98, 301330.
Bindi, D., Parolai, S., Picozzi, M., Ansal, A., 2010. Seismic input motion determined from a
surface-downhole pair of sensors: a constrained deconvolution approach. Bulletin of
the Seismological Society of America 100, 13751380.

13

Bodet, L., Wijk, K., Bitri, A., Abraham, O., Cte, P., Grandjean, G., Leparoux, D., 2005. Surfacewave inversion limitations from laser-Doppler physical modeling. Journal of Environmental and Engineering Geophysics 10, 151162.
Bonilla, L.F., Steidl, J.H., Gariel, J.C., Archuleta, R.J., 2002. Borehole response studies at
the Garner Valley Downhole Array, Southern California. Bulletin of the Seismological
Society of America 92, 31653179.
Bonnefoy-Claudet, S., Cornou, C., Bard, P.Y., Cotton, F., Moczo, P., Kristek, J., Fah, D.,
2006. H/V ratio: a tool for site effects evaluation. Results from 1-D noise simulations. Geophysical Journal International 167, 827837.
Borcherdt, R.D., 1970. Effects of local geology on ground motion near San Francisco Bay.
Bulletin of the Seismological Society of America 60, 2961.
Castrillo, E., 2011. Seismic hazard assessment in Nicaragua based on the spatially
smoothed seismicity approach, Engineering Geology. Engineering Geology, Lund
University, Lund.
Davis, R.O., 2000. Estimation of soil shear modulus softening during strong ground
shaking using ground surface and downhole acceleration recordings. Earthquake
Engineering and Structural Dynamics 29, 359376.
Elgamal, A.W., Zeghal, M., Parra, E., Gunturi, R., Tang, H.T., Stepp, J.C., 1996. Identication
and modeling of earthquake ground response I. Site amplication. Soil Dynamics
and Earthquake Engineering 15, 499522.
Elgamal, A.W., Ahmed, W., Zeghal, M., 1998. Characteristics and modeling of site response
using downhole seismic records. U. S. Geological Survey, Library, Reston, VA, United
States. 75.
Faccioli, E., Santoyo, E., Leon, J.L., 1973. Microzonation criteria and seismic response studies for the city of Managua. Conference proceedings, Earthquake Engineering Research Institute: Managua, Nicaragua, earthquake of December 23, 1, pp. 271291.
Gunturi, V.R., Elgamal, A.W., Tang, H.T., 1998. Hualien seismic downhole data analysis.
Engineering Geology 50, 929.
Kausel, E., Roesset, J.M., 1981. Stiffness matrices for layered soils. Bulletin of the Seismological
Society of America 71, 17431761.
Kokusho, T., Sato, K., 2008. Surface-to-base amplication evaluated from KiK-net
vertical array strong motion records. Soil Dynamics and Earthquake Engineering
28, 707716.
Kramer, S.L., 1996. Geotechnical earthquake engineering. Prentice-Hall, Englewood
Cliffs, NJ.
Kwok, A.O.L., Stewart, J.P., Hashash, Y.M.A., 2008. Nonlinear ground-response analysis of
Turkey Flat shallow stiff-soil site to strong ground motion. Bulletin of the Seismological
Society of America 98, 331343.
Langer, C.J., Hopper, M.G., Algermissen, S.T., Dewey, J.W., 1974. Aftershocks of the
Managua, Nicaragua, earthquake of December 23, 1972. Bulletin of the Seismological
Society of America 64, 10051016.
Lermo, J., Chavez-Garcia, F.J., 1994. Are microtremors useful in site response evaluation?
Bulletin of the Seismological Society of America 84, 13501364.
Mahajan, A.K., 2009. NEHRP soil classication and estimation of 1-D site effect of
Dehradun fan deposits using shear wave velocity. Engineering Geology 104,
232240.
Mahajan, A.K., Slob, S., Ranjan, R., Sporry, R., Champati ray, P.K., Westen, C.J., 2007. Seismic
microzonation of Dehradun City using geophysical and geotechnical characteristics in
the upper 30 m of soil column. Journal of Seismology 11, 355370.
Mehta, K., Snieder, R., Graizer, V., 2007. Downhole receiver function: a case study. Bulletin
of the Seismological Society of America 97, 13961403.
Obando, E.A., 2009. Simplied Vertical Array Set-Up for Seismic Amplication Evaluation,
Department of Electrical Measurements and Automation. Lund University, Lund, p. 140.
Obando, E.A., Park, C.B., Ryden, N., Ulriksen, P., 2009. Roadside MASW surveys using a
portable road bump: a case-study in Managua, Nicaragua. Symposium of Applied
Geophysics to Environmental and Engineering problems, SAGEEP 22.
Park, C.B., Miller, R.D., Xia, J., 1999. Multichannel analysis of surface waves. Geophysics
64, 800808.
Parolai, S., Bindi, D., Ansal, A., Kurtulus, A., Strollo, A., Zschau, J., 2010. Determination of
shallow S-wave attenuation by down-hole waveform deconvolution: a case study
in Istanbul (Turkey). Geophysical Journal International 181, 11471158.
Pavlenko, O., Irikura, K., 2002. Changes in shear moduli of liqueed and nonliqueed
soils during the 1995 Kobe earthquake and its aftershocks at three vertical-array
sites. Bulletin of the Seismological Society of America 95, 19521969.
Proakis, J.G., Manolakis, D.G., 1996. Digital signal processing (3rd ed.): principles,
algorithms, and applications. Prentice-Hall, Inc.
Robinson, D., Dhu, T., Schneider, J., 2006. SUA: a computer program to compute regolith
site-response and estimate uncertainty for probabilistic seismic hazard analyses.
Computers and Geosciences 32, 109123.
Safak, E., 1999. Local site effects and dynamic soil behaviour. International Union of
Geodesy and Geophysics General Assembly.
Shah, H.C., Mortgat, C.P., Kiremidjian, A.S., Zsutty, T.C., 1975. A study of seismic risk for
Nicaragua, part I. The John A. Blume Earthquake Engineering Center, Stanford.
Steidl, J.H., Tumarkin, A.G., Archuleta, R.J., 1996. What is a reference site? Bulletin of the
Seismological Society of America 86, 17331748.
Walther, C.H.E., Flueh, E.R., Ranero, C.R., Von Huene, R., Strauch, W., 2000. Crustal structure
across the Pacic margin of Nicaragua: evidence for ophiolitic basement and a shallow
mantle sliver. Geophysical Journal International 141, 759777.
Zeghal, M., Elgamal, A.W., 2000. Site response and vertical seismic arrays. Progress in
Structural Engineering and Materials 2, 92101.
Zeghal, M., Elgamal, A.W., Parra, E., 1996. Identication and modeling of earthquake
ground response II. Site liquefaction. Soil Dynamics and Earthquake Engineering
15, 523547.

Please cite this article as: Obando, E.A., et al., A mobile multi-depth borehole sensor set-up to study the surface-to-base seismic transfer
functions, Eng. Geol. (2011), doi:10.1016/j.enggeo.2011.09.001