Brazilian Journal of Geophysics (2021) 39(4): 551-563

© 2021 Brazilian Geophysical Society

ISSN 0102-261X
DOI: 10.22564/rbgf.v38i4.2115

SPATIAL AUTOCORRELATION OF PASSIVE SURFACE WAVE DATA

FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

Victor José C. B. Guedes 1,2*, Welitom Rodrigues Borges 1,

Luciano Soares da Cunha 1 and Susanne Taina Ramalho Maciel 3

ABSTRACT. Surface wave methods are commonly applied to engineering problems for S-wave velocity estimations.
Conventional active Multichannel Analysis of Surface Waves (MASW) surveys for earth dam assessment suffer from
limitations mainly associated to restrict depth of investigation and negative influences of near noise sources. In Brazil, the
need to image around noisy sites over 30 m deep, which are non-ideal contexts for traditional active seismic data campaigns,
is commonly in demand. We acquired ten minutes of ambient vibration data at the crest of a large earth dam in Brasília,
Federal District. The Spatial Autocorrelation (SPAC) method was applied to develop a 2D S-wave model velocity using
surface wave generated from passing vehicles on the adjacent road. A maximum depth of 42 m was achieved, and the
model presented a S-wave velocity range from 274 m/s up to 713 m/s. The water level, foundation ground and possible low
and high anomalous compaction zones were interpreted. Vs30 was found to vary from stiff to very dense soil along the
profile, with higher values observed towards the left abutment.

Keywords: earth dam; SPAC; MAM; S-wave; ambient noise.

RESUMO. Métodos de onda de superfície são rotineiramente aplicados a problemas de engenharia para obtenção de
estimativas de velocidade da onda S. A realização de levantamentos de Análise Multicanal de Ondas de Superfície (MASW)
ativos convencionais para avaliação de barragens de terra sofrem limitações principalmente associadas à profundidade de
investigação restrita e às influências negativas de fontes de ruído próximas. No Brasil, ocorre a necessidade de obter
resultados em locais ruidosos com mais de 30 m de profundidade, que são contextos não ideais para as campanhas de
sísmica ativa tradicionais. Dez minutos de dados de vibração ambiental foram registrados na crista de uma grande barragem
de terra em Brasília, Distrito Federal. O método de Autocorrelação Espacial (SPAC) foi aplicado para desenvolver um
modelo 2D da velocidade de onda S considerando ondas de superfície geradas a partir do tráfego de veículos na rodovia
adjacente. Foi atingida uma profundidade máxima de 42 m e o modelo apresentou uma faixa de velocidade da onda S de
274 m/s a 713 m/s. Foram interpretados o nível freático, terreno de fundação e possíveis zonas anômalas de compactação
baixa e mais elevada. Verificou-se que o Vs30 varia de solo rígido a muito denso ao longo do perfil, com valores mais altos
observados em direção à ombreira esquerda.

Palavras-chave: barragem de terra; SPAC; MAM; onda S; ruído ambiental.

*Corresponding author: Victor José Cavalcanti Bezerra Guedes

1Universidade de Brasília - UnB, Instituto de Geociências, Brasília, DF, Brazil – E-mails: [email protected], [email protected],
[email protected]
2Neogeo Geotecnologia, Belo Horizonte, MG, Brazil
3Universidade de Brasília - UnB, Faculdade UnB Planaltina, Planaltina, DF, Brazil – E-mail: [email protected]
552 SPAC OF PASSIVE SURFACE WAVE DATA FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

INTRODUCTION date, MAM has been applied in many studies for

Surface wave methods are known for a significant investigation over 30 m deep, such as
demand in engineering problems for S-wave geotechnical, environmental, and earthquake
velocity (Vs) estimations. Vs can correlate with engineering (e.g., Eker et al., 2012; Hayashi et
shear modulus (Sheriff & Geldart, 1995), thus al., 2018; Moon et al., 2019; Zhang et al., 2019;
being generally used as a satisfactory stiffness Ku et al., 2021).
indicator. Park et al. (1999) developed the For Vs estimations, active surveys are
multichannel analysis of surface waves (MASW), generally associated with better resolution near
probably the most popular surface wave method surface (Asten & Hayashi, 2018; Foti et al., 2018).
nowadays for near surface Vs estimation. The However, there are still few evaluations about the
MASW method is applied along active seismic resolution of dispersion image to this date
datasets, and a traditional survey relies on a (Baglari et al., 2018). The main distinct advantage
controlled seismic source for wavefield of passive methods is the potential to achieve
generation registered by a linear spread of penetration depths over 30 m with significantly
geophones in line with the energy source. shorter acquisition time and field effort.
The field operation, level of cultural noise and Therefore, there is a growing preference for the
depth of investigation are the most addressed passive method over the active method if only
aspects for consideration around active surveys. one of the surveys can be executed (Hayashi et
Many earth dams in Brazil present near cultural al., 2016; Asten & Hayashi, 2018). Considering
noise sources (e.g., heavy machinery work and this context, we acquired only seismic ambient
vehicle traffic) and require over 30 m of vibration data and applied the SPAC method to
investigation depth. Despite being less sensitive obtain dispersion curves at the crest of the
to noise sources when compared to methods that Paranoá dam, the largest dam of Brasília, the
rely on body wave propagation, the accuracy of a federal capital of Brazil. The goal of this study is
dispersion curve obtained with the MASW to calculate a representative pseudo-2D S-wave
method enhances with the removal of noise on velocity model of the dam, interpret the
ground roll data (Park et al., 1999). With active observable internal features from velocity
surveys, sledgehammer or weight drops rarely contrasts, compute Vs30 values across the
achieve penetration depths greater than 30 m massif and evaluate general aspects of the
(Foti et al., 2018), which suggests that it is not a passive analysis as a geophysical assessment
self-sufficient approach for adequate Vs30 methodology for earth dams.
estimation (e.g. Hayashi et al., 2016).
Due to the limitations around active MASW, Study area
the use of ambient vibrations originating from The Paranoá dam (Fig. 1) is located east of
natural or cultural sources for Vs estimation has Brasília, in the Federal District of Brazil. The area
gained a great deal of attention over the last is composed of slates and quartzites from the
years. The most traditional analysis for Vs Paranoá Group (Campos et al., 2013) and red-
estimation using ambient vibrations has been yellow Latosol and haplic soil (Reatto et al.,
introduced by Aki (1957), which proposed the 2004). The structure is a rockfill earth dam with a
spatial autocorrelation (SPAC) method. This 600 m crest length. The massif is composed by
approach measures seismic phase velocity from clay soils, clean natural sands, upstream rockfills
ambient vibration data based on the variation with with quartzite rocks and sandstones, and
frequency of the autocorrelation coefficient vegetation covering downstream. The main
(coherence) between two signals. From the access road to the site is the Estrada Parque do
SPAC method, Okada (2003) presented the Contorno (EPCT - DF-001), which is partially built
microtremor array measurement (MAM) over the crest of the dam, resulting in a constant
technique to estimate deep Vs variation. To this vehicle flow parallel to the study area.

Braz. J. Geophys., 39(4), 2021

 GUEDES ET AL. 553

Figure 1 - Location map of the Paranoá dam with the acquired passive data profile (black line).
The arrow indicates the direction of acquisition, pointing to the end of the profiles.

MATERIAL AND METHODS where r is the distance between two receivers of a

Spatial autocorrelation (SPAC) 2D isotropic array (e.g., a circle or a triangle), φ is
their direction in relation to a central receiver, COH
The MAM technique uses the SPAC method to
is the complex coherency of observed data, c(ω) is
analyze the signal complex coherency between
the phase velocity at an angular frequency ω, and
multiple observations in an array of receivers,
independently of source location. Coherency is the J0 is the Bessel function. The left term in Eq. (1) can
similarity between all possible pairs of geophones, be calculated from observed ambient vibration and
and generally, as receiver separation increases, relates to calculating coherency for two receivers
coherency decreases. with a separation r and direction φ and averaging
The SPAC function represents the variation of the complex coherency. Phase velocity is
the coherency with frequency between two calculated by the comparison of the left term and
signals, and is given by the Bessel function, the right term in Eq. (1), by
changing the phase velocity c(ω). The velocity that
1 minimizes the error can be considered as the phase
SPAC(r,ω)=
2π velocity at ω. Equation (1) is valid for isotropic
φ=2π (1) arrays, but also appliable for an anisotropic array
ω
� Re�COH(r,φ,ω)�dφ = J0 � r� observing ambient vibration propagating equally
φ=0 c(ω)
from all directions (Hayashi, 2008).

Braz. J. Geophys., 39(4), 2021

554 SPAC OF PASSIVE SURFACE WAVE DATA FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

Comparison of SPAC with other passive Ambient vibrations

array methods Surface waves suffer much less energy decay with
Surface wave methods based on ambient propagation distance than body waves. This means
vibrations for Vs estimation consist in array that, far away from the source position, most of the
analysis with multiple receivers. Besides SPAC, seismic energy is carried out through surface
two other popular array analyses are the waves, and far-field ambient vibrations are mainly
frequency-wavenumber (f-k) beamforming and composed of surface waves (Foti et al., 2018).
seismic interferometry (SI). SPAC uses ambient vibrations originated from
With the f-k method (Horike, 1985), ambient natural (e.g., ocean wave action at coastlines and
vibrations are processed in the frequency- microseisms) or cultural sources (e.g., vehicle
wavenumber domain using spectral estimation traffic, vibrations from construction or other
methods. Dominant source direction is favorable machineries). Usually, low frequencies are created
for f–k, but a dominant ambient vibration by large-scale events, while high frequencies come
direction can add bias into SPAC estimates (Foti from local sources, generally related to human
et al., 2018). Flores-Estrella et al. (2001) activities. The term “ambient noise” in passive
obtained more consistent results with SPAC surveys is judged to be inappropriate by some
regarding the expectations from geological authors. Foti et al. (2018) relate “noise” with effects
conditions in comparison with the f-k method. that are not directly associated to wave propagation
Claprood & Asten (2008) concluded that SPAC (e.g., instrumental self-noise, weather actions on
gives information over a wider range of the receiver and bad coupling with the ground) and
frequencies than f-k, which enhances the wave propagation features that are not usable for
interpretation at higher frequencies, allowing a analysis (e.g., body wave components), while
better characterization of shallow layers. As a “signal” is surface waves originating from distant
general understanding, f-k tends to overestimate sources.
phase velocity, a potentially hazardous aspect, The ideal vibration sources for SPAC are
while SPAC tends to underestimate it (Claprood steady signals without strong changes in amplitude.
& Asten, 2008; Asten & Hayashi, 2018). The fundamental assumption is that the vibration
SI is a relatively new approach to retrieve wavefront is planar and isotropic (comes from all
the Green’s functions from the crosscorrelation directions), making it independent of source
of ambient vibrations (Wapenaar, 2004). The positions. Passive methods in general may face
method has been vastly used in global difficulties in areas where the level of ambient
seismology to obtain velocity models at a crustal vibrations is very low, while a higher success
scale from passive data and measurement of chance occurs in environments that have a good
group velocity (often addressed as ambient level of ambient vibrations with a reasonable
noise tomography). Although the use of SI for degree of isotropy (Hayashi & Craig, 2017; Foti et
shallow phase velocity calculation for al., 2018). In such cases, coherent vibrations
engineering purposes is gaining more attention dominated by surface wave can be recorded, and
over the last few years (e.g., Cheng et al., 2015; reliable results can be obtained with a limited
Olivier et al., 2018), it can be considered a number of receivers and a relatively short recording
relatively new topic of research still in time window (Foti et al., 2018).
development phase (Asten & Hayashi, 2018).
Tsai & Moschetti (2010) presented an explicit
Data acquisition
comparison of the two approaches and pointed
that the SPAC theory in the frequency domain is A passive survey was executed along a profile
equivalent to the crosscorrelation theory used in downstream of the Paranoá dam in September
SI in time domain if vibrations are assumed to 2020 at the crest of the massif. Using four Geode
come from all directions equally. seismographs (Geometrics) of 24 channels each,

Braz. J. Geophys., 39(4), 2021

 GUEDES ET AL. 555

ambient vibrations were registered by a 475 m For inversion of an observed dispersion curve
long linear array of 96 vertical 14 Hz geophones (Fig. 2C), an initial model based only around the
(Geospace), spaced every 5 m. We acquired 20 fundamental mode of vibration was constructed by
continuous SEG2 waveform files with a time one-third wavelength transformation (e.g.,
length of 32 seconds each, totaling 640 seconds Hayashi, 2008) in terms of apparent depth and
(approximately 10 minutes) of data acquisition Rayleigh wave velocity. The non-linear least
period, a sufficient interval pointed by Hayashi squares method was used for model fitting (Xia et
(2008) for SPAC analysis. A sampling frequency al., 1999). The number of layers was fixed as 15
of 500 Hz was used, summing 16,000 samples and only Vs were modified throughout the
per trace (one sample every 2 ms). inversion iterations, while density and P-wave
velocity were changed based on empirical
relations (Ludwig et al., 1970; Kitsunezaki at al.,
Data processing 1990). The theoretical dispersion curves were
For editing the waveform files, we used Pickwin calculated with a matrix method (Saito &
from the SeisImager/SW package. From all 20 raw Kabasawa, 1993). The iterative process
passive datafiles with 96 traces each, 24 sequential recalculated Vs until a best fit with low RMS error
traces were extracted from the original vibration was obtained between the observed and
sections every 10 m and saved into new SEG2 files. calculated phase velocities (Fig. 2D). Finally, a 1D
After the “roll along” trimming processes, a total of Vs model in depth was plotted (Fig. 2E).
740 waveform files were generated (20 files of 32 s The average Vs down to 30 m (Vs30) is a
of observed ambient vibration by a 115 m long popular parameter of geotechnical interest. Vs
linear array of receivers). Park et al. (2001) and Xia measurements with surface wave methods are
et al. (2004) reported that the longer the geophone generally presented as a layered format. From
spread, the higher the resolution of the dispersion the obtained 1D Vs models, all respective Vs30
image. The goal of this approach is to obtain a values were calculated after the International
pseudo-2D Vs model from the interpolation of Building Code IBC-2000 (Paz & Leigh, 2004), as
horizontal aligned 1D models at every 10 m along ∑ di
the survey line which was addressed as Two- Vs30= (2)
d
dimensional Linear Array Microtremor Survey (2D- ∑ i
Vsi
LAMS) by Kita et al. (2011).
where di and Vsi are the thickness and the S-wave
As for calculation and inversion of dispersion velocity of the ith layer of the model, respectively.
curves obtained from ambient vibration data, we
used WaveEq, also from the SeisImager/SW
package. For each set of passive data at every 10 RESULTS AND DISCUSSION
m along the survey profile (Fig. 2A), complex Observable frequency range
coherencies were calculated for every receiver Figure 3 shows the frequency spectrum for every
pair. The real parts with the same spacing were 32 seconds of ambient vibration records
averaged in frequency domain, and the separation observed by the 96 in line receivers spread. The
between each pair of receivers was plotted against registered cultural vibrations at the site present a
their coherency as a function of frequency (Fig. frequency content mainly between 8 Hz and 30
2B). Coherencies were finally compared with the Hz. Vehicle-related vibrations are generally
Bessel function, where the match between dominant at 2-30 Hz (Coward et al., 2003). As the
coherencies and the Bessel function provided survey happened during daytime, it is reasonable
phase velocity information, used to develop the to consider that the main sources of signal were
dispersion image (Fig. 2C). moving vehicles.

Braz. J. Geophys., 39(4), 2021

556 SPAC OF PASSIVE SURFACE WAVE DATA FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

Figure 2 - Processing flow with the SPAC method. A) 32-second record out of the 10-minute
continuous data acquisition; B) Plot of the separation between each pair of geophones against their
coherence as a function of frequency; C) the obtained phase velocity image; D) the observed and
calculated dispersion curves; E) the resulting 1D Vs profile after a non-linear least squares inversion.

Figure 3 - Frequency spectrogram of all 32-second records of the 10-minute continuous data
acquisition. The arrow indicates the direction of acquisition, pointing to the end of the profiles.

Braz. J. Geophys., 39(4), 2021

 GUEDES ET AL. 557

For MASW, Park et al. (2002) found that the zone between the clay soil and the quartzite
lower-frequency limits of higher-frequency foundation.
geophones of 10 Hz and 40 Hz were not limited by Figure 5B shows the Vs30 distribution across
their natural frequencies for dispersion imaging. the profile. After UBC (1997) site classification, the
Similarly here, using 14 Hz geophones, we blue circles mark Vs30 related to stiff soil (180 m/s
observe reliable phase velocities down to about 4 < Vs ≤ 360 m/s), and red circles mark Vs30 related
Hz, far below the instrument natural frequency. to very dense soil (360 m/s < Vs ≤ 760 m/s). It is
Figure 4 shows the observed dispersion curves clear from the profile that higher Vs30 values are
during data processing, with phase velocity data found towards the end of the acquisition line,
from 4 Hz up to 18 Hz. closer to the dam spillway (left abutment).
The obtained velocity values are within the
S-wave velocity model and Vs30 range of values found in other similar structures.
Table 1 presents ranges of S-wave velocities
Figure 5A shows the obtained pseudo-2D velocity
obtained in other studies that used seismic
model. Low velocity values are presented in
methods to characterize earth dams.
reddish color tones, while higher velocities are in
blueish color tones. Vs ranges from 274 m/s up to
713 m/s. The maximum depth of 42 m is observed Assumptions around the SPAC method
around the center of the section. At the edges, Vs Some aspects must be considered when applying
changes are only imaged down to approximately the SPAC method. The key assumptions, as
20 m. This is likely due to a smaller content of described in many studies (e.g., Asten, 2006;
observable phase velocities at lower frequency Asten & Hayashi, 2018; Baglari et al., 2018; Foti et
intervals around these positions. The black al., 2018), are: a) the study area can be sufficiently
triangles show the horizontal position of each 1D represented as a layered earth model; b) far-field
Vs profile obtained after data inversion, separated Rayleigh waves are the main content of the
every 10 m (a total of 37 velocity profiles). The first vertical-component recorded ambient vibration
and last 1D Vs profiles are at 57.5 m and 417.5 m, data; and c) there is a spatial averaging of
respectively. sources.
The white dashed line marks the interpreted According to CEB (2020), the Paranoá dam is
water level. It is found to vary from 6 m down to 13 mostly composed of compacted clay soil, followed
m deep. A water saturated soil usually presents an by a quarzitic foundation. Despite not being a
increase in P-wave velocity and a decrease in S- layered earth, the studied site can be considered
wave velocity (Baechle et al., 2009; Kassab & well represented with the SPAC method, since
Weller, 2015; Konstantaki et al., 2016; Foti et al., there are no expected sharp variations and the
2018). This is caused due to the decrease in the massif composition is considerably homogeneous.
shear modulus of materials when water is present Considering the short distance between the
(Baechle et al., 2009). Low Vs near the surface receiver spread and the road (approximately 6 m)
between 57.5-100 m and 320-370 m may be and the wavelengths of Rayleigh waves at 5 Hz
related to lower soil compaction, marked as light and 10 Hz (around 100 m and 30 m, respectively),
pink. Likewise, high velocity anomalies near it is unlikely that a pair of receivers could properly
surface are pointed as higher compaction zones. record surface waves generated by a very near
At the center of the dam crest, the foundation passing vehicle, specially at lower frequencies
ground is expected at a depth of approximately 48 (larger wavelengths). This short distance may
m (CEB, 2020). The obtained velocities can be cause a distortion in phase velocity estimation for
correlated to the stiff clay soil, which usually low frequency, known as the near-field effect. On
ranges approximately from 200 m/s up to 600 m/s the other hand, increasing the distance between
(Foti et al., 2018). The black dashed line marks the source and receiver can raise attention to far-field
Vs contrast of 600 m/s as a possible transition effects. Considering the attenuation property of

Braz. J. Geophys., 39(4), 2021

558 SPAC OF PASSIVE SURFACE WAVE DATA FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

Figure 4 - The observed dispersion curves used for data inversion,

obtained from the SPAC phase velocity images.

Figure 5 - A) The 2D velocity profile obtained after 1D Vs profile interpolation; B) Vs30

distribution across the profile, according to the reference UBC (1997) site classification.

Table 1 - Comparison of S-wave velocity values obtained in other works using seismic methods at earth dams.

Approximate maximum
Reference Country Methods Vs (m/s)
depth of investigation (m)

Kim et al. (2011) South Korea MASW 100-1480 30

Cardarelli et al. (2014) Italy SRT 120-300 9

Hayashi et al. (2014) USA MASW 120-350 16

Rahimi et al. (2019) USA MASW and FWI 100-2100 25

Guireli Netto et al. (2020) Brazil SRT and MASW 150-700 16

This study (2021) Brazil MAM (SPAC) 274-713 42

Note: MASW = Multichannel Analysis of Surface Waves; SRT = Seismic Refraction Tomography; FWI = Full Waveform Inversion;
MAM (SPAC) = Microtremor Array Measurements (Spatial Autocorrelation).

Braz. J. Geophys., 39(4), 2021

 GUEDES ET AL. 559

higher-frequency components of Rayleigh complex when using traditional recording systems

waves, the recorded passive data at far receivers that still rely on long spread cables. A linear array is
may contain wavefields generated from local the most practical alternative and can be considered
sources, such as interference from dominant high as an option where logistical efficiency is required.
frequency body waves (Baglari et al., 2018). The assumption of homogeneous and
In general, as the random source positions isotropic distribution of the ambient vibration
are not known, the closeness of sources are sources around the surveyed area or in-line with the
frequently neglected in passive surveys, with still array direction is needed when using a linear
no well stablished consensus to avoid near and spread of receivers for SPAC (Foti et al., 2018).
far-field effects (Baglari et al., 2018; Foti et al., Hayashi & Kita (2010) showed, through a field
2018). This is frequently the case for urbanized experiment, that the linear array configuration can
areas, like the Paranoá dam, where localized provide reliable phase velocities and almost
microtremor sources, passing vehicles and other identical dispersion curves as 2D array spreads
general human activities can occur. However, when the propagation direction of the ambient
there was a significant level of vehicle traffic vibration is distributed at a minimum range of 120º.
Kita et al. (2011) employed a linear roll along
along all the extension of the road during data
acquisition of passive recording of ambient
acquisition, which makes the recorded ambient
vibration to obtain a pseudo-2D Vs model. Hayashi
wave-field not related solely to close sources, but
et al. (2018) performed a comparison of dispersion
to far-field sources as well. Despite the nearness
curves calculated from passive records acquired
of sources (vehicle traffic) to receivers placed on
using a linear array and a L-shaped array, obtaining
sidewalks, Stephenson et al. (2009) presented
similar dispersion trends.
credible Vs values with the SPAC method for a
For the survey line adjacent to the road of the
site characterization in lower Manhattan, New
Paranoá dam, the SPAC method assumes that
York City. Roberts & Asten (2008) reported that passing vehicles generate Rayleigh wave signals
significant near source effects are unlikely to over a wide azimuth angle. This means that
occur in real field scenarios, where a large passing vehicles on the south half of the road would
number of sources, source directions and source generate seismic energy almost in line with
distances are present. receivers on the north half, removing most of the
An axiomatic assumption is that the SPAC bias associated with wavefronts parallel to the
method depends on spatial averaging of sources, linear array. In general, roadside passive surveys
either by multiple orientations of station separations have been pointed as a practical alternative around
or by an azimuthal distribution. As the ambient the use of a conventional linear receiver array to
vibration wavefield might propagate from different obtain results with low overestimation of Vs values
and unknown directions, theoretically, a 2D isotropic in comparison with conventional 2D arrays (usually
receiver array, such as a circle or an equilateral less than 10%; Park et al., 2007).
triangle, is preferable for passive surveys (Foti et al.,
2018). An isotropic array provides the same CONCLUSIONS
response regardless of the direction of the incoming We acquired ten minutes of ambient vibration data
wavefield and better ensures that velocities will be at the crest of the Paranoá dam in Brasília, Brazil.
well estimated, even in the case of anisotropic The SPAC method was applied for the
vibrations (Hayashi & Craig 2017). development of 1D velocity models, and a
In practical terms, perfectly isotropic array subsequent 2D interpolated velocity model.
configurations may be difficult to set up in the field. The main sources of signal were moving
They may require an extensive open area, which is vehicles. The cultural vibrations at the site
often not the case for many survey sites, and can be presented a frequency content mainly between 8

Braz. J. Geophys., 39(4), 2021

560 SPAC OF PASSIVE SURFACE WAVE DATA FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

Hz and 30 Hz, a coherent range based on already ACKNOWLEDGMENTS

reported values. We found that the lower- The authors would like to thank: a) UFPR, for
frequency limits of geophones of 14 Hz were not providing the seismographs and supplementary
limited by their natural frequencies for dispersion equipment for the seismic data acquisition; b) CEB
imaging. Phase velocities down to 4 Hz and up to Geração, for providing reference documentation
18 Hz were observed. A maximum depth of 42 m about the study area; c) The AINOA project, for the
was achieved, which suggests that higher research grant provided during the first months of
frequency geophones, such as 14 Hz, can be development of this work; d) Diogo Olivetti, for kindly
used to obtain deeper results, suitable for Vs30 providing the orthoimage of the dam massif; and e)
site classification. The anonymous reviewers, whose comments
The developed Vs model presented a greatly helped to increase the quality of the paper.
velocity range from 274 m/s up to 713 m/s. The
possible water level could be interpreted as a low
velocity horizon from 6 m down to 13 m deep REFERENCES
across the crest of the dam. Low and high Vs
Aki, K., 1957. Space and time spectra of stationary
anomalies near the surface were marked as stochastic waves, with special reference to
possible zones of lower and higher soil microtremors, Bull. Earthq. Res. Inst., 35: 415–457.
compaction, respectively. A Vs contrast of 600 Asten, M.W. 2006. On bias and noise in passive
m/s was interpreted as a possible transition zone seismic data from finite circular array data
from clay soil to the quartzite foundation. Vs30 processed using SPAC methods. Geophysics,
was found to vary from stiff to very dense soil, 71(6): V153-V162.
with higher Vs30 values found towards the Asten, M.W.; Hayashi, K. 2018. Application of the
abutments. Spatial Auto-Correlation Method for Shear-Wave
The calculated S-wave velocities were found Velocity Studies Using Ambient Noise. Surveys in
to be within the interval of already reported values Geophysics, 39: 633–659.
in other similar structures, also obtained from Baechle, G. T., Eberli, G. P., Weger, R. J.,
seismic data analysis. However, we must point Massaferro, J. L., 2009. Changes in dynamic
possible unknown effects around the use of the shear moduli of carbonate rocks with fluid
SPAC method, such as the dam geometry (not a substitution. Geophysics, 74(3): E135-E147. DOI:
10.1190/1.3111063
layered earth), the possible biases associated
with the close approach of sources, and the use Baglari, D.; Dey, A.; Taipodia, J. 2018. A state-of-
of a linear array of receivers. There is still no the-art review of passive MASW survey for
subsurface profiling. Innovative Infrastructure
forward guide to predict, without failure, passive
Solutions, 3: 1–13.
survey parameters, such as the number of
geophones, minimum and maximum offset Campos, J.E.G., Dardenne, M.A., Freitas-Silva,
F.H., Martins-Ferreira, M.A.C., 2013. Geologia do
distances or which array geometry is sufficient.
Grupo Paranoá na porção externa da Faixa
Most of these variables are likely to be site
Brasília. Brazilian J. Geol., 43: 461–476. DOI:
specific when executing a field survey. Despite 10.5327/Z2317-48892013000300004
the uncertainties regarding the use of the SPAC
Cardarelli, E.; Cercato, M.; De Donno, G. 2014.
method, the obtained results show that this Characterization of an earth-filled dam through the
passive surface wave analysis is a promising and combined use of electrical resistivity tomography,
time saving approach for investigating greater P- and SH-wave seismic tomography and surface
depths at noisier sites, such as large earth dams wave data. Journal of Applied Geophysics, 106:
with a significant presence of human activities. 87–95.

Braz. J. Geophys., 39(4), 2021

 GUEDES ET AL. 561

CEB - Companhia Energética de Brasília. 2020. Methods and its Application to Site Investigations.
Relatório técnico de atividade: Estudo da Dissertation, Kyoto University, Japan, 278 pp.
estabilidade da Barragem Paranoá. DF, Brazil. Hayashi, K.; Kita, T. 2010. Applicability of a spatial
Technical Report, 96 pp. autocorrelation method (SPAC) using a linear
Cheng, F.; Xia, J.; Xu, Y.; Xu, Z.; Pan, Y. 2015. A array in comparison with triangular and L-shaped
new passive seismic method based on seismic arrays. Proceedings of the Symposium on the
interferometry and multichannel analysis of Application of Geophysics to Engineering and
surface waves. Journal of Applied Geophysics, Environmental Problems, SAGEEP 2: 543–551.
117: 126–135. Hayashi, K.; Cakir, R.; Walsh, T.J.; LaVassar, J.
Claprood, M.; Asten, M.W. 2008. Comparison of 2014. A safety evaluation of dams using integrated
array microtremor survey methods for estimation geophysical method: A case study in Washington
of dispersion curves in Launceston, Australia. In: state. In: 27th Symposium on the Application of
Environmental and Engineering Geophysical Geophysics to Engineering and Environmental
Society - 21st Symposium on the Application of Problems. 2014, SAGEEP 2014: 46–54.
Geophysics to Engineering and Environmental Hayashi, K.; Cakir, R.; Walsh, T.J. 2016.
Problems 2008, Philadelphia, United States of Comparison of dispersion curves and velocity
America. 2: 1377–1384. models obtained by active and passive surface
Coward, D.; Blair, D.; Burman, R.; Zhao, C. 2003. wave methods. SEG Technical Program
Vehicle-induced seismic effects at a gravitational Expanded Abstracts, 35: 4983–4988.
wave observatory. Review of Scientific Hayashi, K., Craig, M. 2017, S-wave velocity
Instruments, 74: 4846–4854. measurement and the effect of basin geometry on
Eker, A.M.; Akgün, H.; Koçkar, M.K. 2012. Local site response, east San Francisco Bay area,
site characterization and seismic zonation study by California, USA: Physics and Chemistry of the
utilizing active and passive surface wave methods: Earth, 98: 49–61. DOI: 10 .1016/j.pce.2016.07.001.
A case study for the northern side of Ankara, Hayashi, K.; Lorenzo, J.M.; Gostic, A. 2018.
Turkey. Engineering Geology, 151: 64–81. Application of 2D ambient noise tomography to
Flores-Estrella, H.; Aguirre, J.; Boore, D.; Yussim, levee safety assessment in New Orleans. Leading
S. 2001. Estimation of Velocity Structure Using Edge, 37: 740–745.
Microtremor Recordings from Arrays: Comparison Horike, M., 1985, Inversion of phase velocity of
of Results from the SPAC and the F-K Analysis long-period microtremors to the S-wave-velocity
Methods. AGU Fall Meeting Abstracts, 21: 7212. structure down to the basement in urbanized areas:
Foti, S., Hollender, F., Garofalo, F., Albarello, D., Journal of the Physics of the Earth 33, 59-96.
Asten, M., Bard, P. Y., Comina, C., Cornou, C., Kassab, M.A.; Weller, A. 2015. Study on P-wave
Cox, B., Di Giulio, G., Forbriger, T., Hayashi, K., and S-wave velocity in dry and wet sandstones of
Lunedei, E., Martin, A., Mercerat, D., Ohrnberger, Tushka region, Egypt. Egyptian Journal of
M., Poggi, V., Renalier, F., Sicilia, D., Socco, V., Petroleum, 24: 1–11.
2018. Guidelines for the good practice of surface
Kim, K.Y.; Jeon, K.M.; Hong, M.H.; Park, Y.G.
wave analysis: a product of the InterPACIFIC
2011. Detection of anomalous features in an
project. Bulletin of Earthquake Engineering, 16(6):
earthen dam using inversion of P-wave first-arrival
2367-2420. DOI: 10.1007/s10518-017-0206-7
times and surface-wave dispersion curves.
Guireli Netto, L. G., Gandolfo, O. C. B., Malagutti Exploration Geophysics, 42: 42–49.
Filho, W. M., Dourado, J. C., 2020. Non-
Kita, T.; Hayashi, K.; Bingöl, H.; Karaoglu, H.;
destructive investigation on small earth dams
Duran, K.; Gümü, S. 2011. The development of a
using geophysical methods: seismic surface wave
2-dimensional microtremor survey method based
multichannel analysis (MASW) and S-wave
on SPAC method using sequential linear arrays.
refraction seismic methods. Brazilian Journal of
Proceedings of the Symposium on the Application
Geophysics, 38(1): 5-19.
of Geophyics to Engineering and Environmental
Hayashi, K., 2008. Development of Surface-Wave Problems – SAGEEP. 30: 115–120.

Braz. J. Geophys., 39(4), 2021

562 SPAC OF PASSIVE SURFACE WAVE DATA FOR ASSESSMENT OF AN EARTH DAM IN BRASÍLIA, BRAZIL

Kitsunezaki, C., N. Goto, Y. Kobayashi., T. Ikawa, M. DOI: 10.1007/978-1-4615-0481-8_25.

Horike, T. Saito, T. Kurota, K. Yamane, and K. Rahimi, S.; Moody, T.; Wood, C.; Kouchaki, B.M.;
Okuzumi, 1990, Estimation of P- and S-wave Barry, M.; Tran, K.; King, C. 2019. Mapping
velocities in deep soil deposits for evaluating ground Subsurface Conditions and Detecting Seepage
vibrations in earthquake: Journal of the Japan Channels for an Embankment Dam Using
Society for Natural Disaster Science, 9: 1–17. Geophysical Methods: A Case Study of the Kinion
Konstantaki, L. A., Ghose, R., Draganov, D., & Lake Dam. Journal of Environmental and
Heimovaara, T., 2016. Wet and gassy zones in a Engineering Geophysics 24: 373–386.
municipal landfill from P- and S-wave velocity Reatto, A., Martins, E.S., Farias, M.F.R., Silva,
fields. Geophysics, 81(6): EN75–EN86. DOI: A.V., Carvalho Júnior, O.A., 2004. Mapa
10.1190/GEO2015-0581.1 pedológico Digital – SIG Atualizado do Distrito
Ku, T.; Palanidoss, S.; Zhang, Y.; Moon, S.W.; Federal Escala 1:100.000 e uma síntese do texto
Wei, X.; Huang, E.S., Kumarasamy, J., Goh, K.H. explicativo. Planaltina: Embrapa Cerrados, DF,
2021. Practical configured microtremor array Brazil. 31 pp.
measurements (MAMs) for the geological Roberts, J.; Asten, M. 2008. A study of near
investigation of underground space. Underground source effects in array-based (SPAC) microtremor
Space (China) 6: 240–251. surveys. Geophysical Journal International, 174:
Ludwig, W.J., Nafe, J.E., Drake, C.L. 1970. Seismic 159–177.
refraction. In: The Sea, A. E. Maxwell (Editor), Vol. Saito, M., and H. Kabasawa, 1993, Computation
4, Wiley-Interscience, New York, 53–84. of reflectivity and surface wave dispersion curves
Moon, S.W.; Subramaniam, P.; Zhang, Y.; Vinoth, for layered media 2. Rayleigh wave calculations:
G.; Ku, T. 2019. Bedrock depth evaluation using Butsuri Tansa, Tokyo, Japan, 46, 283–298.
microtremor measurement: empirical guidelines at Sheriff, R.E; Geldart, L.P. 1995. Exploration
weathered granite formation in Singapore. Journal Seismology. Cambridge: Cambridge University
of Applied Geophysics, 171: 103866. Press, 628 pp.
Okada, H. 2003. The microtremor survey method: Stephenson, W.J.; Hartzell, S.; Frankel, A.D.;
SEG, SEG Monograph Series 12, 150 pp. Asten, M.; Carver, D.L.; Kim, W.Y. 2009. Site
Olivier, G.; De Wit, T.; Brenguier, F.; characterization for urban seismic hazards in
Bezuidenhout, L.; Kunjwa, T. 2018. Ambient noise lower Manhattan, New York City, from
Love wave tomography at a gold mine tailings microtremor array analysis. Geophysical
storage facility. Geotechnique Letters, 8: 178–182. Research Letters, 36: 1–5.
Park, C.B.; Miller, R.D.; Xia, J. 1999. Multichannel Tsai, V.C.; Moschetti, M.P. 2010. An explicit
analysis of surface waves. Geophysics, 64: 800– relationship between time-domain noise
808. correlation and spatial autocorrelation (SPAC)
Park, C. B., Miller, R. D., Xia, J., 2001, Offset and results. Geophysical Journal International, 182:
resolution of dispersion curve in multichannel 454–460.
analysis of surface waves (MASW): Proceedings of UBC - UNIFORM BUILDING CODE. 1997.
the SAGEEP 2001. Denver, Colorado, SSM-4. Structural Engineering Design Provisions.
Park, C.B.; Miller, R.D.; Xia, J.; Ivanov, J.M. 2007. International Conference of Building Officials
Multichannel Analysis of Surface Waves (MASW) (ICBO), Whittier, California, 492 pp.
– active and passive methods. The Leading Edge, Wapenaar, K., 2004, Retrieving the elastodynamic
26: 60–64. Green’s function of an arbitrary inhomogeneous
Park, C.B.; Miller, R.D.; Miura, H. 2002. Optimum medium by cross correlation: Physical Review
field parameters of an MASW survey. Letters, 93: 254301. DOI:
Proceedings of the Society of Exploration 10.1103/PhysRevLett.93.254301
Geophysicists of Japan, Tokyo. 22: 23. Xia, J.; Miller, R.D.; Park, C.B. 1999, Estimation of
Paz, M.; Leigh, W. 2004. International Building near-surface shear-wave velocity by inversion of
Code IBC-2000. Structural Dynamics, 757–781. Rayleigh wave: Geophysics, 64(3): 691–700.

Braz. J. Geophys., 39(4), 2021

 GUEDES ET AL. 563

Xia J.; Miller R.D.; Park C.B.; Ivanov J.; Tian, G.; Zhang, Y.; Li, Y.E.; Ku, T. 2019. Geotechnical site
Chen, C. 2004. Utilization of high-frequency investigation for tunneling and underground works by
Rayleigh waves in near-surface geophysics. advanced passive surface wave survey. Tunnelling
Leading Edge, 23(8): 753–759 and Underground Space Technology, 90: 319–329.

V.J.C.B.G.: conceptualization, field acquisition, data

processing and inversion, geophysical interpretation and
text writing; W.R.B.: conceptualization, field acquisition,
geophysical interpretation and text revision; L.S.C.:
conceptualization, field acquisition and text revision;
S.T.R.M.: conceptualization and text revision.

Received on December 4, 2021 / Accepted on May 19, 2022

- Creative Commons attribution-type BY Braz. J. Geophys., 39(4), 2021