Journal of Applied Geophysics 103 (2014) 31–42

Contents lists available at ScienceDirect

Journal of Applied Geophysics

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

Using all seismic arrivals in shallow seismic investigations

Gabriel Fabien-Ouellet ⁎, Richard Fortier
Département de géologie et de génie géologique, Université Laval, Québec, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Near surface seismic investigations are expensive and time-consuming. Moreover, seismic processing usually fo-
Received 5 August 2013 cuses on one particular type of wave and wastes much of the information contained in seismic records that could
Accepted 16 December 2013 be used to make near surface seismic surveys more valuable and cost effective. A workflow is proposed herein
Available online 7 January 2014
that combines seismic refraction tomography, multichannel analysis of surface waves (MASW), and seismic re-
flection using P-waves and SV-waves, which takes advantage of P-wave first arrivals, Rayleigh waves, and P-wave
Keywords:
Seismic reflection
and SV-wave reflections, respectively. The use of the proposed methodology is shown through three case studies
SV waves carried out in the Outaouais region, Quebec, Canada, using a 24-channel seismograph, vertical geophones, and a
Multichannel analysis of surface wave sledgehammer. The results show that it is possible to acquire SV-reflections at sites where a strong velocity rever-
Seismic refraction tomography sal is present at the surface using only vertical geophones. Under that condition, or more generally when two
Surface wave filter component geophones are used, the proposed workflow leads to two complementary stacked sections: 1) an
Hydrogeophysics SV-wave section that has a high resolution even at shallow depths but can lack coherency and 2) a P-wave section
that has better coherency but is blind at shallow depths. Two velocity models are also produced: an SV-wave
model that combines the results from MASW and SV-wave reflections and a P-wave model that combines the re-
sults from seismic refraction and P-wave reflections. The workflow uses the frequency variant linear move-out
(FV-LMO) surface wave filter, which is much more efficient than band pass or f-k filters to process SV waves.
The value of many near surface seismic surveys can thus be enhanced by processing all propagation modes,
especially when SV-wave reflections are present due to their high resolution.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction geophones are recommended to successfully acquire SV-wave reflections

in all terrain conditions, Pugin et al. (2013) found that their polarization is
Seismic energy propagates in the earth through different types of more vertical in soft clayey soils and horizontal when the medium is hard,
waves (P, SH, SV, Rayleigh, Love, Stoneley) that can undergo reflections, such as a sand deposit or outcropping rock. This indicates that SV-wave
refractions, conversions from one type to another and diffractions (Aki reflections may be recorded with vertical geophones over soft soils at cer-
and Richards, 2002). Conventional seismic investigations focus on a par- tain sites.
ticular seismic arrival. For instance, one of the most popular seismic Rayleigh waves can be processed by multichannel analysis of surface
methods to map the depth to bedrock, the seismic refraction method, waves (MASW). This method has proven to be a reliable technique to
uses direct and refracted P-waves (Hagedoorn, 1959; Palmer, 1981). assess SV-wave velocities (Park et al., 1998; Xia et al., 2002, 2003)
First arrivals can also be processed using tomographic inversion tech- and, in certain cases, their quality factor (Lai et al., 2002; Xia et al.,
niques that require far less input from the interpreter, can take into 2012). The spread configuration of the geophones and seismic shots
account velocity inversions and support one-direction shot spreads used in a typical MASW survey is similar to common seismic reflection
(Sheenan et al. 2005; White, 1989; Zang and Toksöz, 1998). In many field geometries (Park et al., 2002b).
cases, the geophones and shot spacings used for refraction tomography As it was previously mentioned, these methods use a similar geom-
are similar to those used in seismic reflection surveys, with shots at etry in the field and multiple propagation modes carrying useful infor-
every two or three geophone intervals (Lanz et al., 1998). mation are usually generated and recorded. Moreover, the unwanted
Seismic reflection surveys normally use only reflections from modes of propagation are often regarded as noise and much effort is
P-waves or SH waves. Recent work by Pugin et al. (2008, 2009) showed spent removing them. Therefore, only a fraction of the available infor-
the power of using SV-wave reflections, which exhibit high resolution at mation in seismic gathers is used and processed. By obtaining more
very shallow depths, similar to SH reflections. Although three component information out of the same dataset using a more complete signal pro-
cessing approach, the seismic acquisition can become more valuable
and cost efficient. A workflow is presented herein that integrates the
⁎ Corresponding author at: Département de géologie et de génie géologique, Université
Laval, Québec (QC), G1V 0A6, Canada. Tel.: +1 418 656 2746.
processing techniques specific to MASW, seismic refraction tomogra-
E-mail addresses: gabriel.fabien-ouellet.1@ulaval.ca (G. Fabien-Ouellet), phy, P-wave seismic reflection and SV-wave seismic reflection. The
Richard.Fortier@ggl.ulaval.ca (R. Fortier). objective is to obtain useful information from all the seismic arrivals.

0926-9851/$ – see front matter © 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jappgeo.2013.12.009
32 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42

The workflow puts an emphasis on SV-wave reflections due to their

higher resolution than P-wave reflections.
The use of this processing workflow is illustrated through three case
studies. The first case illustrates the full workflow for a survey designed
for the acquisition of SV reflections. The second case describes the appli-
cation of a surface wave filter, the frequency variant linear move-out
(FV-LMO) filter proposed by Park et al. (2002a) which is a logical exten-
sion of the proposed workflow. The third case shows how the workflow
can help identify the presence or absence of SV-wave reflections. The
conditions necessary to acquire SV-wave reflections using only vertical
geophones are discussed first.

2. Acquisition of SV-waves

The most common way to record S-waves is by using cross-line

horizontal geophones with a cross-line polarized source (Haines
and Ellefsen, 2010; Hunter et al., 2002). Using that configuration,
SH-waves can be generated and recorded. In contrast, vertically polar-
ized shear waves can be generated with any conventional sources and
can be recorded by inline vertical and horizontal geophones (Helbig
and Mesdag, 1982).
The direction of polarization of SV reflections depends on the angle
of incidence and the velocity distribution and usually varies with offset.
For that reason, vertical and horizontal geophones are usually required.
However, in some geological settings, a significant amount of SV-wave
energy can be recorded with vertical geophones, even at short offsets. Fig. 2. Processing workflow combining MASW, SV-wave reflections, seismic refraction
tomography, and P-wave reflections.
One such setting is the presence of a strong velocity inversion close to
the surface (Fig. 1). According to Snell's law, an incoming ray is horizon-
tally shifted if a high velocity layer lies on top of a much slower layer. In reflection velocity model. The P-wave processing begins by picking
that case, the particle motion becomes mostly vertical. Such a situation first breaks, which are used for seismic refraction tomography and for
is quite common in clay deposits that are affected by freeze-thaw and surface static corrections. The P-wave reflections are then processed
wetting-drying cycles, which cause over-consolidation of the surficial using the tomographic velocity model as a first estimate for the stacking
layer that significantly increases its shear wave velocity (Motazedian velocities. Similar to the SV-wave processing, a P-wave stacked section
and Hunter, 2008). Another common case is a paved or gravel road and a combined refraction/reflection velocity model are produced.
constructed on soft soils; the pavement then acts as a high velocity This is a general methodology, and the specific procedures of each
layer. In such conditions, a significant amount of SV-wave energy can inversion method could vary. The details of the present study are
be recorded using only vertical geophones as shown in Section 6. How- given in Section 6.
ever, careful planning and testing are required to use only vertical Even though 2-C geophones were not used in any of the case studies
geophones to acquire SV-waves and two‐component (2-C) geophones presented herein, this methodology is particularly well-suited for 2-C
are preferable in all circumstances. processing because all four seismic arrivals should be present in those
records. In contrast, SV-wave reflection processing must be skipped
3. Processing workflow for sites where SV-waves cannot be observed in the vertical component,
but the rest of the workflow is still valid because surface waves and P-
The processing workflow that combines MASW, seismic refraction waves should usually be present.
tomography and SV- and P-wave reflection inversion is shown in
Fig. 2. The SV- and P-waves are processed separately. The starting 4. The frequency variant linear move-out filter
point of both processing flows is the spatially referenced seismic data.
For the SV-wave processing, MASW is first performed to build an It is critical to remove Rayleigh waves before performing the CMP in-
S-wave velocity model. The dispersion curves produced during this version of the SV reflections. The velocities of P-waves and Rayleigh
analysis are used to filter the surface waves. The SV-wave reflections waves are so different that most of the surface wave energy is removed
are then inverted to obtain a stacked section and a combined MASW/ by CMP stacking. However, because Rayleigh waves travel at approxi-
mately 90% of the velocity of SV-waves, surface waves often overlap
and even hide SV reflections. For this reason, performing the CMP inver-
sion of SV reflections without previously removing surface waves can
lead to spurious reflections on the stacked section.
Surface waves are normally removed using band-pass filters, f-k
filters or muting. However, the frequency spectrums of surface waves
and SV-waves often overlap and band-pass filters cannot effectively
remove surface waves without removing a significant part of the de-
sired signal. f-k filters are not well adapted for this filtering because
the energy of surface waves is difficult to identify in the f-k domain.
Moreover, the use of a simple velocity fan is problematic because of
the multimodal and dispersive nature of surface waves. One of the
Fig. 1. Due to the presence of a strong velocity inversion close to the ground surface, the
SV-waves particle motion is deflected to vertical. The velocities for this example are
most common techniques to remove surface waves is muting. However,
taken from a seismic piezocone penetration test (SCPTu) performed along the survey muting can be subjective and a compromise must be made between
line of case study 1 (Fig. 3b) (Fabien-Ouellet et al., 2014). removing surface waves and preserving SV reflections. It is often
 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42 33

impossible to remove surface waves without removing a significant part 5. Study sites
of the reflection energy, especially for very shallow reflections.
Park et al. (2002a) proposed an effective method for filtering surface The seismic surveys presented herein were performed in the
waves. Their method is based on the same transformation as the MASW Outaouais region, Quebec, Canada. The locations of the study sites are
method. Given the observed modes of surface waves Ci(ω), the frequen- shown in Fig. 3. The general stratigraphy of the region consists of a
cy variant linear move-out (FV-LMO) φ(ω,x) is obtained according to thin layer of till above bedrock that is covered in places by glaciofluvial
the following equation: sediments. This sequence is usually overlain by a thick glaciomarine unit
that was deposited by the Champlain Sea. As shown in Fig. 3, two of the
iωx=C i ðωÞ
φðω; xÞ ¼ e : ð1Þ seismic lines were performed over the soft glaciomarine deposits,
namely the Buckingham (Section 6.1) and the Shawville (Section 6.2)
Given observed data d(x,t), the Fourier transform D(ω,x) is calculated. case studies. However, a glaciofluvial sand deposit is present in the
The inverse FV-LMO correction is then applied: lake Green case study (Section 6.3).
The three surveys were acquired under the Programme d'acquisition
−iωx=C i ðωÞ
DFV‐LMO ðω; xÞ ¼ e Dðω; xÞ: ð2Þ des connaissances sur les eaux souterraines en Outaouais (PACES), which
was a regional study of groundwater in the Outaouais region. One of the
This correction horizontally aligns each frequency of the surface main objectives of the PACES project was to acquire new information for
waves but anything that does not share the same dispersion relation is areas where existing databases failed to provide a comprehensive
not aligned. All of the horizontal energy is then removed using an f-k hydrogeological overview. To achieve this objective, seismic methods
filter. To obtain the filtered data, the FV-LMO correction is applied and were used to spatially extend information from nearby wells (as for
the inverse Fourier transform is performed. To filter all of the energy the Shawville case study) and to obtain the stratigraphy for sites
of the Rayleigh waves, the dispersion curve of each mode is identified where drilling could not be performed (as for the Buckingham and
and the FV-LMO filter is performed separately for each mode. Lake Green case studies).

Fig. 3. Quaternary geologic maps: a) locations of the three study sites along the Ottawa River, b) Buckingham case study, c) Shawville case study, d) Lake Green case study.
Modified from Hurtubise et al. (2012) and Leduc et al. (2013).
34 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42

6. Case studies geophones was dictated by budget restrictions, but a 2-C survey
would have been preferable.
6.1. Buckingham case study: application of the workflow An example of such a test conducted for the Buckingham case study
in a farm field near the survey line is given in Fig. 5, where a coherent
The goal of the seismic survey in the Buckingham case study was to SV-wave reflection is clearly present at approximately 600 ms. P-wave
characterize an aquifer that is confined by a surficial sensitive clay reflections are also present at times less than 400 ms. This result was
deposit. According to boreholes located in the survey area, a clay unit judged to be good enough to perform the seismic survey. It is worth
more than 30 m thick overlies a glaciofluvial sand and gravel unit, and mentioning that a PS reflection is present at 400 ms, which is indicated
the bedrock is 40 to more than 100 m deep. However, no wells were in Fig. 5. PS reflections can be observed on most of the records in the first
located directly along the line. One of the objectives of this investigation two case studies. It is possible to process them, and they can provide an
was to confirm and clarify the general stratigraphy described above. independent stacked section (Pugin et al., 2013). However, they were
not processed in this study.
6.1.1. Acquisition
This seismic survey was designed for the acquisition of SV- and
P-wave reflections, but the configuration also allowed for the acquisi- 6.1.3. SV-wave processing
tion and processing of Rayleigh waves and P-wave first arrivals. The The first step for processing SV-waves is the MASW method. This
acquisition parameters described here are essentially the same for all was performed as described in Park et al. (1998, 1999). The phase-
of the surveys; any differences will be described when introducing velocity transformation was performed on CMPs binned at twice the
each case study. shot interval (6 m) to improve the signal-to-noise ratio at low frequen-
The seismic acquisition was performed using a 24-channel engineer- cies. The fundamental mode, as well as the higher modes, was identified
ing seismograph (Stratavisor NZ-24) and 40-Hz vertical geophones. The manually during the dispersion analysis, but only the fundamental
impact of an 8-kg sledgehammer on a steel plate lying horizontally on mode was considered for the inversion. The S‐wave velocities at this
the ground surface was used as a seismic source. This source was very site were expected to vary continuously with depth (Fabien-Ouellet
efficient in the field and only two shots were performed and stacked et al., 2014), so a 15-layer initial model reaching a maximum depth of
for each shot point. Moreover, strong SV-waves were generated with 15 m was used for the inversion. Each dispersion curve was separately
this seismic source. Each survey line was located on the shoulder of inverted until convergence, and the P-wave velocities and the thickness
a gravel road, which acted as the hard layer that refracts SV-waves of each layer were kept constant. All of the curves were then smoothed
vertically (Section 2) and facilitated the execution of the survey. horizontally with a moving median filter that was three curves wide,
The configuration used in this hydrogeophysical investigation is and they were inverted separately again until convergence. The inver-
similar to the SH-wave reflection surveys carried out by Guy et al. sion was performed with the software Seisimager/SW. Several algo-
(2003) and Pugin et al. (2004). The geophones were located 0.75 m rithms are available for the inversion of Rayleigh waves that takes into
apart to avoid aliasing while the seismic shots were located every 3 m account higher modes (Xia et al. 2000), effective dispersion curves
along the survey line. In the Buckingham case study, long offsets were (Lai and Rix, 1998) or that uses full waveforms (O'Neill et al., 2003).
required to reach the bedrock at a depth of 100 m, and each shot posi- The use of these algorithms could provide better S-wave velocity
tion was used four times as the geophone spreads were moved forward models, but the inversion of the fundamental mode alone gave satisfac-
along the line. The configuration when the records are sorted by seismic tory results. Dispersion curves from the fundamental mode and higher
shot location is shown in Fig. 4. modes were also used to filter the surface waves with the FV-LMO filter.
This configuration is not only adequate to process SV- and P-wave
reflections, but it is also very similar to the recommendations for 2D
MASW surveys for targets at depths exceeding 50 m (Park et al.,
2002b). The only difference is the natural frequency of the geophones;
40 Hz geophones were used in this study, while 4.5 Hz geophones are
normally recommended.

6.1.2. Presence of SV-wave reflections

Before performing an SV seismic reflection survey using vertical geo-
phones, it is necessary to check that SV-wave reflections are indeed re-
corded by the vertical geophones. This can be achieved by walkaway
tests (Steeples and Miller, 1998) performed at several locations along
the planned seismic line. In our case, the choice of vertical over 2-C

Fig. 5. A walkaway test acquired near the survey line of the Buckingham case study
(see Fig. 3 for its location). The blue (nos. 1 and 2), green (no. 3) and red (no. 4) arrows
Fig. 4. Configuration of geophones and seismic shots after sorting the records relative to indicate SV, PS and P reflections, respectively, while the yellow arrow (no.5) indicates
their seismic shot locations. Rayleigh waves.
 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42 35

Table 1 was performed at every 6 m using the MASW velocity model at times
CMP processing of SV-wave reflections. The steps identified by an asterisk were performed where there were no reflections. The traces were then corrected for
with in-house MATLAB codes based on the CREWES algorithms.
the normal move-out and stacked. Depth conversion was performed
Processing step Description by f-k migration, and the topographic correction was performed by
1. Surface wave filter* Described in Section 4 shifting each trace to the datum level.
2. CMP binning Bin length: 0.75 m The SV-wave processing steps are illustrated in Fig. 6. Clear SV-wave
3. Bad trace removal* Automatic based on S/N ratio reflections can be observed from 200 ms to the end of the gather at
4. Scaling AGC with a window of 1 s
every offset (Fig. 6a). The reflections are more coherent than in the
5. Semblance analysis Includes MASW model
6. NMO correction Stretch mute: 75% walkaway test shown in Fig. 5. One potential explanation is that the
7. Alpha trim stacking Alpha percentage: 75% velocity inversion caused by the road material was stronger than that
8. AGC Window: 400 ms of the over-consolidated clay layer in the farm field where the walkway
9. Band-pass filter 10, 15, 125, and 250 Hz trapezoid frequencies test was performed.
10. Median filter Length: 5 traces
The phase-velocity transformation for the CMP gather is shown in
11. Depth conversion Using f-k migration and the smoothed velocity model
12. Topographic shift Fig. 6b. The energy of the Rayleigh waves is separated into several
modes of propagation, and each mode was considered for surface
wave filtering. The fundamental mode shown in the inset in Fig. 6b
The processing flow for SV-wave reflections is summarized in was inverted to obtain the S-wave velocity model in Fig. 6c. To evaluate
Table 1. The P- and S-wave processing was performed with the the errors caused by picking the fundamental mode, the higher and
CREWEWS MATLAB package (Margrave, 2003) in combination with lower envelopes of the dispersion curves of the fundamental mode
in-house MATLAB codes and GEDCO Vista. After filtering the surface (inset in Fig. 6b) were picked and inverted (Fig. 6c). CMP gather no.
waves, Common Middle Point gathers (CMP) were defined with a 43 after application of the surface wave filter is shown in Fig. 6d. Most
0.75 m bin. Bad traces were then removed based on their signal-to- of the energy of the surface waves is removed. The semblance analysis,
noise (S/N) ratios. The traces were scaled, and a semblance analysis including the MASW model, and the NMO corrected gather are shown

Fig. 6. An example of the SV-wave processing: a) CMP gather no. 43 along the survey line of the Buckingham case study, where SV-waves are located in the blue solid line frame, Rayleigh
waves in the yellow dash-dotted line frame and P-waves in the red dashed line frame, b) dispersion curves showing the fundamental and higher modes of propagation (inset: detail of the
fundamental mode; uncertainties are identified with red lines), c) SV-wave velocity model in purple obtained from the inversion of the fundamental mode (the red and green lines are the
estimated errors), d) the same CMP gather after filtering of surface waves using all modes, e) semblance analysis of SV-waves using the MASW results (green dots), and f) NMO-corrected
CMP gather with a stretch mute of 100%. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
36 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42

in Fig. 6e and f, respectively. A more thorough discussion of the integra- 6.1.5. Interpretation
tion of MASW and SV-wave reflections can be found in Fabien-Ouellet The SV and P-wave time sections for the Buckingham case study are
et al. (2014). Following the muting of the samples in which the stretch shown in Fig. 8, and the depth sections overlain by the velocity models
in frequency exceeds 60%, a considerable amount of the energy at far are shown in Fig. 9. A very coherent reflection extends from 550 ms at
offsets is muted on the NMO corrected gather. This can be problematic the northern end (left side) to 850 ms at the southern end (right side)
for this dataset and can cause some reflections to be strongly attenuated in the S-wave section (Fig. 8a) and from 75 to 125 ms in the P-wave sec-
on the stacked section (see Section 6.1.5). tion (Fig. 8b). When converted to depth (Fig. 9), this reflection deepens
Because the stacking velocities of the S-waves and P-waves are very from 45 m at the northern end to 90 m in the middle of both sections to
different, the stretch mute removes most of the P-wave energy on the as deep as 120 m at the southern end. This is consistent with the depth
CMP gather after the NMO correction. This, along with the fact that to bedrock provided by boreholes in the area. This reflection is therefore
S-waves are much more energetic at short offsets on the CMP gathers, interpreted as the bedrock contact. Every reflection below this reflection
ensures that no P-wave energy remains on the S-wave stacked section. is a multiple. On the SV-wave stacked section, this interpretation is sup-
ported by the fact that the multiples have the same stacking velocity as
6.1.4. P-wave processing the bedrock reflection. In the P-wave section, greater offsets would have
The first step for the P-wave processing is seismic refraction tomog- been required to obtain a confident velocity estimate below the bedrock
raphy using wavepath eikonal traveltime (WET) inversion with the surface, so those reflections cannot be reliably interpreted; however,
software Rayfract (Schuster and Quintus-Bosz, 1993). This algorithm they are most likely multiples from the bedrock reflection. In both
supports the inversion of seismic refraction tomography with an end- cases, a strong multiple from the bedrock is present at twice the
on spread and one-way shots. First breaks were picked for every shot traveltime of the primary and is visible for the first 400 m on both
without any previous processing. The final velocity model was obtained profiles. The presence of multiples for such deep reflections illustrates
after twenty iterations of the WET inversion algorithm. the depth of investigation that can be achieved using a simple sledge-
The first breaks were also used to apply surface static corrections as hammer source in this environmental setting.
described in Pugin and Pullan (2000). An average model that included Both sections also show a reflection along the survey line at times
three layers was assigned to the entire survey, and each trace was between 250 and 400 ms for S-waves and between 48 and 100 ms for
shifted in time to align the picked first breaks with the synthetic arrivals P-waves (Fig. 8). On the depth sections (Fig. 9), this reflection is present
from the average model. Only one model was used because the P-wave at depths between 14 and 31 m and corresponds to the contact
velocity model obtained by seismic refraction tomography was constant between the near surface clay unit and the sand and gravel unit. This
along the entire survey. confirms and clarifies the information found in nearby boreholes that
The processing flow for P-wave reflections is summarized in Table 2 was described above.
and is similar to the SV-wave processing. The main difference is the use In the S-wave section, the clay–sand reflection is coherent between
of predictive deconvolution to remove multiples. Moreover, a band-pass 0 and 500 m (see point 1 in Fig. 8). At distances greater than 500 m,
filter was used to remove the low frequency noise shown in Fig. 7a. This the coherence is lost and the reflections appear to thicken (point 2).
enhanced lower energy reflections on the CMP gathers (Fig. 7d). At least two factors may explain this loss of coherence. First, according
The workflow for P-wave processing is presented in Fig. 7. The first to Aylsworth et al. (2000), the clay sediments over the bedrock at
breaks were picked on a raw shot gather (Fig. 7a). After WET inversion depths greater than 50 m were disturbed by a Holocene earthquake in
of the first arrivals (Fig. 7c), surface static corrections were applied. The the Outaouais region, which is prone to high seismicity. Pullan et al.
sudden time shift of the first breaks and of the bedrock reflection at an (2011) attributed this loss of coherence to this earthquake event.
offset of approximately 35 m is clearly removed after these corrections Second, the presence of coarser material, such as a gravel deposit, can
(Fig. 7b). A super-gather that combines four CMP gathers and its induce the scattering of the S-wave energy due to its small wavelength
semblance analysis are shown in Fig. 7d and e, respectively. Several of approximately 2 m. Gravely layers over a finer mix of sand and gravel
reflections can be observed on the gather with clear semblance peaks. were encountered below the clay unit in several boreholes in the area.
The velocity analysis includes velocities from reflections in black and Several reflections are present in the sand and gravel unit in the
velocities from refractions in green. The velocities from the refractions P-wave section that are not clearly present in the S-wave section
are necessary because no reflections can be seen before 60 ms. Finally, (point 3 in Fig. 8b). This is the result of two phenomena. First, the stack-
the NMO corrected gather is shown in Fig. 7f. For early times, the severe ing velocity for the reflections between the clay–sand interface and the
stretch mute is caused by the velocity variations in the first few meters. bedrock reflections varies from 150 m/s to more than 250 m/s. This
causes severe stretching after the NMO correction that had to be
muted before stacking. As a result, a significant amount of the energy
Table 2 was removed, which resulted in poor signal-to-noise ratios for some
CMP processing of P-wave reflections. The steps identified by an asterisk were performed reflections.
with in-house MATLAB codes based on the CREWES algorithms.
Second, some reflections are still present in the S-wave section, but
Processing step Description they cannot be distinguished because of the higher impedance contrasts
1. Surface static corrections* Model used: 285, 547, and 1351 m/s with between the clay–sand and sand–bedrock contacts than the layering in
intersection times at 0, 3.8, and 9.8 ms, respectively the sand and gravel unit. With such large velocity variations, the density
2. CMP binning Bin length: 1.5 m does not contribute significantly to the impedance contrast. For
3. Bad trace removal* Automatic based on S/N ratio instance, the S-wave velocity varies from 150 m/s in the clay unit to
4. Scaling AGC with a window of 250 ms
more than 250 m/s in the sand and gravel unit. The reflections within
5. Predictive deconvolution Lag: 5 ms. Window: 50 ms
6. Ormsby band-pass filter 60, 120, 650, and 800 Hz trapezoid frequencies the sand and gravel unit do not have sufficient energy to appear in the
7. First break muting S-wave section with the gain settings that were used. They could be
8. Semblance analysis Includes the refraction model revealed using an AGC gain with a short window, but much of the
9. NMO correction Stretch mute: 75%
amplitude information would be lost. On the other hand, the P-wave
10. Alpha trim stacking Alpha trim percentage: 75%
11. AGC Window: 15 ms velocity difference between these units is much smaller and varies
12. Band-pass filter 60, 120, 650, and 800 Hz trapezoid frequencies from 1500 m/s in the clay unit to approximately 1700 m/s in the sand
13. Median filter Length: 5 traces and gravel unit. This makes the energy of the reflections more equal
14. Depth conversion f-k migration using the smoothed velocity model and they have similar amplitudes on the stacked section. The sections
15. Topographic shift
are then mutually complementary, and their combined interpretation
 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42 37

Fig. 7. An example of the processing of P-wave reflections: a) first breaks picked for shot gather no. 217 along the survey line of the Buckingham case study, b) the same shot gather after
surface static corrections, c) P-wave velocity model obtained by WET inversion of the arrival times of the first breaks, d) CMP gather no. 772, e) semblance analysis of P-waves using the
refraction results, and f) NMO-corrected CMP gather with a stretch mute of 75%.

provides a far better interpretative cross-section than by considering Because fluids do not sustain S-waves, soil saturation does not affect
only one of these sections (Fig. 9c). the S-wave velocity. However, the P-waves are very sensitive to water
The P-wave section is blind within the first few meters of soil. How- content and can indicate the depth to the capillary fringe. A change in
ever, the S-wave section provides great detail at shallow depths; in fact, P-wave velocity from 350 to 1500 m/s can occur in a clay deposit
many reflections are present in the shallow clay layer. For instance, from unsaturated to fully saturated conditions. In the stacked P-wave
parallel horizontal reflections are present in the clay unit (see point 4 section (Fig. 9b), no reflection can be associated with this abrupt imped-
in Fig. 8a) and are attributed to thin sand beds (Fabien-Ouellet et al., ance change because the water table is too shallow (approximately
2014). Dipping structures are also present (point 5 in Fig. 8a) and 6 m). This portion of the velocity model is obtained from the refraction
could be interpreted as landslide scars or ancient riverbeds. analysis and shows the strength of combining refraction and reflection
Both the P- and SV-wave reflections are affected by strong statics analyses.
due to topographic variations at distances of 500 and 1300 m along In summary, the processing workflow used in this seismic survey
the survey line (points 6 and 7 in Fig. 8a). This was addressed by the was successful in characterizing a confined aquifer at depths as great
static corrections for the P-wave section, but no easy solution was as 90 m at a resolution on the order of 2 m (Fig. 9c). Only the seismic re-
found for the S-wave section. The absence of static corrections for flection method using SV-waves can achieve such results. This method-
S-waves can be observed clearly at point 6 in Fig. 8a; a bump is present ology leads to velocity models and stacked sections of SV-waves and
in the S-wave section that does not appear in the P-wave section. With- P-waves, and provides much more information than the processing of
out static corrections (not shown herein), the P-wave section would P-waves alone. The combination of all of this information leads to a
also exhibit this pattern due to the topography. The presence of statics robust interpretation that confirms and clarifies the general interpreta-
must be taken into account during interpretation. tion provided by boreholes in the area.
The combined velocity models (Fig. 9) are good interpretation tools.
Because the S-wave velocities of unconsolidated sediments can vary 6.2. Shawville case study: a survey over a silt deposit
by as much as four folds, this seismic property is very useful for
distinguishing soft from hard materials. The difference in S-wave veloc- To illustrate the performance of the FV-LMO filter, a short seismic
ity between the clay unit (150 m/s) and the sand and gravel unit (200 to survey over a silt deposit is presented. Although this filter was used
350 m/s) is clear (Fig. 9a). The variation in P-wave velocity is much for the Buckingham case study (compare Fig. 6d and a), the smaller
more subtle for the same units (Fig. 9b). length of the Shawville line makes the effects of the FV-LMO filter easier
38 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42

Fig. 8. a) SV-wave stacked section in time and b) P-wave stacked section in time for the Buckingham case study. The very coherent SV-wave reflections detected only with the vertical
geophones are shown in this example.

to appraise. Along with other geophysical investigations, this seismic SV-wave reflections are confirmed by the CPTu that was carried out a
line was part of an effort to extend information laterally from a nearby few meters away from the survey line. According to the CPTu logs
piezocone penetration test (CPTu) (Fig. 3). The general objective was (Fig. 11e), a superficial 5 m thick sandy silt layer overlies a deep clayey
to build a hydrostratigraphic model to use for transport modeling. The silt unit. At depths greater than 13 m, the friction ratio becomes noisier,
acquisition parameters are the same as for the Buckingham survey indicating the presence of a sandier clay deposit. This correlates very
except that each shot position was revisited only twice instead of four well with the seismic section, on which two main reflections appear
times due to the shallower bedrock. above the bedrock: one at a depth of 5.3 m at the contact between the
The surface waves in the Shawville case study are very energetic sandy silt unit and the clayey silt unit and the other at a depth of
(Fig. 10a). They hide some reflections and overlap the bedrock reflec- 12.5 m due to traces of fine sand. Penetration refusal on bedrock oc-
tion at 300 ms. After the FV-LMO filtering (Fig. 10c), nearly of all the curred at a depth of 30 m, which is consistent with the seismic sections.
surface waves are gone, and new reflections appear, as shown by the The results of the seismic survey confirm the lateral extension of the units
blue arrows. In addition, a PS reflection that was nearly hidden by sur- obtained from the CPTu interpretation; therefore this information can be
face waves appears after applying the FV-LMO filter. In comparison, extended spatially along the line.
the application of a velocity fan reject filter in the f-k domain
(Fig. 10b) leaves most of the surface wave energy intact. 6.3. Lake Green case study: a survey over a sand deposit
The depth stacked sections for the Shawville survey are shown in
Fig. 11. They were produced using a processing flow similar that in the To highlight the fact that vertical geophones cannot be used in all con-
first case study. On the SV-wave section after FV-LMO filtering ditions for the acquisition of SV-wave reflections, a seismic survey over a
(Fig. 11c), the bedrock reflection at a depth of approximately 25 m is sand deposit is presented in the last case study. The same acquisition
more pronounced, the reflection at a depth of approximately 12.5 m is parameters that were used in the Buckingham case study were used for
clearer and several near surface reflections appear much more clearly this survey. A shot gather is shown in Fig. 12. No SV-wave reflections ap-
than on the SV-wave section after only pass-band filtering (Fig. 11a). pear in this shot gather, and only back-scattered surface wave patterns
Contamination by surface waves is evident in Fig. 11a; the stacking of are present where SV-wave reflections should be found. Moreover, sur-
the linear arrivals of surface waves causes the dipping features at depths face waves are also difficult to distinguish because other arrivals are su-
less than 15 m. Fig. 11b shows the stacked section where the surface perposed over them. To determine if reflections are hidden by Rayleigh
waves were muted by the top mute shown in Fig. 10a. All of the reflec- waves, the shot gather in Fig. 12a was processed using the FV-LMO filter
tions above the bedrock are removed by the top mute because the sur- applied in both directions to remove back-scattered surface waves
face waves and SV-reflections strongly overlap in this survey. Hence, (Fig. 12c). Using the MASW velocity model and the depth to bedrock ob-
muting is not appropriate in this case. Those results show the strength tained by processing the P-wave reflections, the expected SV-wave re-
of the FV-LMO filter. More generally, this profile highlights the impor- flections were calculated using the NMO equation. This region is shaded
tance of filtering surface waves for SV-wave reflection processing; in red in Fig. 12c. There is still no coherent reflection because the particle
new features become apparent and false reflections are avoided. motion is more or less horizontal in this survey. This is supported by the
Only the bedrock reflection appears in the P-wave stacked section velocity profile found from the inversion of the fundamental mode of the
(Fig. 11d). The discontinuity of the bedrock reflection at a distance Rayleigh waves (inset in Fig. 12b). There is no S-wave velocity inversion,
of 50 m in the SV-wave section also appears in the P-wave section but and the S-wave velocities increase regularly with depth from 200 m/s to
is smoothed due to the longer wavelength of the P-waves. Early more than 300 m/s in the first 8 m of the sand deposit. In such conditions,
 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42 39

Fig. 9. Stacked sections in depth overlaid by the combined velocity model for the Buckingham case study: a) SV-wave section and b) P-wave section. c) Hydrostratigraphic interpretation
overlaid by both P- and SV-wave sections.

Fig. 10. Shot gather no. 3 along the Shawville survey line: a) without filtering; the top mute used in Fig. 11b is indicated by the orange dashed line, b) f-k filtering for velocities between 130
and 250 m/s, and c) FV-LMO filtering as proposed by Park et al. (2002b), which allows a better identification of the reflections than without filtering; the blue arrows (nos. 1, 2 and 3) show
SV-wave reflections, and the green arrow (no. 4) shows PS-wave reflections.
40 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42

Fig. 11. Depth stacked sections for the Shawville case study: a) SV-wave section after an Ormsby band-pass filter with corner frequencies of 30, 50, 200, and 300 Hz, b) SV-wave section
obtained by muting the surface waves, c) SV-wave section after FV-LMO filtering, and d) P-wave section. e) Results from a piezocone penetration test performed close to the survey line ; qc
is the cone resistance, Rf is the friction ratio, ρ is the electrical resistivity, and u2 is the pore pressure. Modified from Comeau et al. (2012).

the seismic ray paths at the surface are almost vertical for short offsets Even though SV-wave reflections were not recorded in the Lake
and the particle motion due to the propagation of S-waves is in the hor- Green survey, the proposed workflow can still be used. As described in
izontal plane. Thus, in contrast to the Buckingham and Shawville case the previous paragraph, MASW can be performed. This leads to an SV-
studies over clay and silt deposits, horizontal geophones would have wave velocity model (Fig. 12b) that provides the shear wave velocity
been necessary to record S-wave reflections in the Lake Green survey. for shallow depths. Filtering surface waves is helpful to assess the

Fig. 12. A shot gather along the survey line of the Lake Green case study: a) without filtering, b) dispersion analysis of surface waves, where the fundamental mode is identified by white
dots, and the superior modes are identified by the red dots (the inset shows the velocity model obtained by the inversion of the fundamental mode), and c) the filtered shot gather, overlaid
by the expected SV-wave reflections. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42 41

Fig. 13. Seismic refraction P-wave velocity model overlaid by the P-wave stacked section for the Lake Green survey.

presence or absence of SV-reflections and provides a better understand- arrivals. This could be achieved by full waveform inversion based on the
ing of the seismic data. Seismic refraction tomography and P-wave seis- elastic wave equation. However, the proposed workflow has the advan-
mic reflection were also possible with this dataset (Fig. 13). This figure tage of being based on readily available inversion algorithms and can be
shows the very good fit between both methods. Three main reflections implemented at low cost.
are present in this section. The shallowest reflection is interpreted as the
water table because the P-wave velocity increases from 500 to 1500 m/s
at this interface. This is very consistent with the water level of Lake Acknowledgments
Green, which is located 50 m north of the line. At the northern end of
the profile, an intermediate reflection is created by a unit with a mean Funding for this research came in part from the Ministère du
velocity of 2000 m/s. This is consistent with a till deposit. The last reflec- développement durable, de l'environnement, de la faune et des parcs
tion corresponds to bedrock with a mean velocity of 5000 m/s. (MDDEFP) of the province of Québec, Canada, the Natural Sciences
and Engineering Research Council (NSERC) of Canada and from the
7. Conclusions Fonds de recherche du Québec - Nature et technologies. We also thank
Simon Bérubé, Majid El Baroudi, Yan Vincent and Alexandra Roberge
The conversion between SV- and P-waves at interfaces produces a for their help in the field. We acknowledge the helpful advice of
more complex system than SH-waves. For this reason, most research Christian Dupuis in writing and editing the paper.
has focused on SH-waves (Helbig and Mesdag, 1982). However, this
complexity can be used to increase the value of the survey because References
information can be obtained from many seismic arrivals, including
Rayleigh waves, first arrivals, and P-wave, SV-wave and PS-wave reflec- Aki, K., Richards, P.G., 2002. Quantitative Seismology 2nd edition. Univ Science Books.
Leduc, E., Daigneault, R.A., Roy, M., et Lamothe, M., 2013. Géologie des formations
tions. In contrast to SH-waves, no special sources are required to gener- superficielles de la région de Cobden (31 F10); carte présentée au Ministère des
ate SV-waves; a simple sledgehammer is sufficient. Ressources Naturelles du Québec, scale 1:50000.
In general, vertical and horizontal geophones (2-C) are required Hurtubise, M.A., Milette, S., Daigneault, R.A., Roy, M. et Lamothe M., 2012. Géologie des
formations superficielles de la région de Thurso (31G11); carte présentée au
to perform an SV-wave reflection survey. However, in some cases, Ministère des ressources naturelles et de la Faune du Québec, scale 1:50 000.
SV-wave reflections can be recorded by vertical geophones, as Aylsworth, J.M., Lawrence, D.E., Guertin, J., 2000. Did two massive earthquakes in the
shown in this study. For this reason, the presence of SV-wave Holocene induce widespread landsliding and near-surface deformation in part of
the Ottawa Valley, Canada? Geology 28 (10), 903–906. http://dx.doi.org/10.1130/
reflections should be evaluated during data processing even if only ver-
0091-7613(2000)28b903:DTMEITN2.0.CO;2.
tical geophones are used. In a general seismic survey that is not designed Comeau, G., Molson, J., Lemieux, J.M., Talbot, M.C., Therrien, R., Montcoudiol, N., Ayotte, S.,
for SV-wave acquisition, small geophone spacings and long record Bérubé, S., Fabien-Ouellet, G., Fiset, P., Fortier, R., Graf, T., Moreno, A., Sterckx, A., 2012.
lengths should be used to maximize the chances of detecting SV-waves Projet d'acquisition de connaissances sur les eaux souterraines en Outaouais, Rapport
d'étape II. Département de géologie et de génie géologique, Université Laval 54
and successfully processing them. (mars 2012).
The processing of SV-wave reflections can be more complex and Fabien-Ouellet, G., Fortier, R., Giroux, B., 2014. Joint acquisition and processing of seismic
time consuming than P-wave reflections due to their high sensitivity reflections and surface waves in a sensitive clay deposit in the Outaouais Region
(Québec), Canada. Landslides in Sensitive Clays. Springer Netherlands, pp. 241–252.
to the stacking velocity. A careful velocity analysis must be performed http://dx.doi.org/10.1007/978-94-007-7079-9_19 0.
because small errors in stacking velocity can lead to a severe loss of Guy, E.D., Nolen-Hoeksema, R.C., Daniels, J.J., Lefchik, T., 2003. High-resolution SH-wave
coherence in the stacked section. However, the processing of Rayleigh seismic reflection investigations near a coal mine-related roadway collapse feature.
J. Appl. Geophys. 54 (1), 51–70. http://dx.doi.org/10.1016/S0926-9851(03)00055-7.
waves with MASW can help to obtain a suitable velocity model. Hagedoorn, J.G., 1959. The plus–minus method of interpreting seismic refraction
Surface waves interference during stacking can also be problematic. sections. Geophys. Prospect. 7 (2), 158–182. http://dx.doi.org/10.1111/j.1365-
The FV-LMO filter is much more efficient at removing surface waves 2478.1959.tb01460.x.
Haines, S.S., Ellefsen, K.J., 2010. Shear-wave seismic reflection studies of unconsolidated
than a simple f-k filter. Surface waves should not be considered noise. sediments in the near surface. Geophysics 75 (2), B59–B66. http://dx.doi.org/
On the contrary, their processing can be helpful during SV-wave reflec- 10.1190/1.3340969.
tion processing as long as the reflections contain sufficient energy. Fur- Helbig, K., Mesdag, C.S., 1982. The potential of shear wave observations. Geophys.
Prospect. 30 (4), 413–431. http://dx.doi.org/10.1111/j.1365-2478.1982.tb01314.x.
thermore, Rayleigh waves are generally produced in all seismic surveys
Hunter, J.A., Benjumea, B., Harris, J.B., Miller, R.D., Pullan, S.E., Burns, R.A., Good, R.L., 2002.
and can be used to estimate the variation in SV-wave velocity in the first Surface and downhole shear wave seismic methods for thick soil site investigations.
few meters of soil. Soil Dyn. Earthq. Eng. 22 (9), 931–941. http://dx.doi.org/10.1016/S0267-7261(02)
Finally, the proposed combined workflow is only a first step in using 00117-3.
Lai, C.G., Rix, G.J., 1998. Simultaneous inversion of Rayleigh phase velocity and attenuation
all of the information contained in the P-SV system. A more rigorous for near-surface site characterization. School of Civil and Environmental Engineering,
method has to be developed as a single inversion scheme for all seismic Georgia Institute of Technology 258.
42 G. Fabien-Ouellet, R. Fortier / Journal of Applied Geophysics 103 (2014) 31–42

Lai, C.G., Rix, G.J., Foti, S., Roma, V., 2002. Simultaneous measurement and inversion of sur- Pugin, A.J.M., Pullan, S.E., Hunter, J.A., 2009. Multicomponent high-resolution seismic
face wave dispersion and attenuation curves. Soil Dyn. Earthq. Eng. 22 (9), 923–930. reflection profiling. Lead. Edge 28 (10), 1248–1261. http://dx.doi.org/10.1190/
http://dx.doi.org/10.1016/S0267-7261(02)00116-1. 1.3249782.
Lanz, E., Maurer, H., Green, A.G., 1998. Refraction tomography over a buried waste dispos- Pugin, A.J.M., Brewer, K., Cartwright, T., Pullan, S.E., Perret, D., Crow, H., Hunter, J.A.,
al site. Geophysics 63 (4), 1414–1433. http://dx.doi.org/10.1190/1.1444443. 2013. Near surface S-wave seismic reflection profiling — new approaches and in-
Margrave, G.F., 2003. Numerical Methods of Exploration Seismology with Algorithms sights. First Break 2 (31), 49–60. http://dx.doi.org/10.3997/1365-2397.2013005
in MATLAB. Department of Geology and Geophysics, University of Calgary (225 (February 2013).
pp.). Pullan, S.E., Pugin, A.J.M., Hunter, J.A., Brooks, G.R., 2011. Mapping disturbed ground using
Motazedian, D., Hunter, J.A., 2008. Development of an NEHRP map for the Orleans suburb compressional and shear wave reflection sections. Symposium on the Application of
of Ottawa, Ontario. Can. Geotech. J. 45 (8), 1180–1188. http://dx.doi.org/10.1139/ Geophysics to Engineering and Environmental Problems 2011.
T08-051. Schuster, G.T., Quintus-Bosz, A., 1993. Wavepath eikonal traveltime inversion. Theory.
O'Neill, A., Dentith, M., List, R., 2003. Full-waveform P–SV reflectivity inversion of surface Geophys. 58 (9), 1314–1323. http://dx.doi.org/10.1190/1.1443514.
waves for shallow engineering applications. Explor. Geophys. 34 (3), 158–173. http:// Sheehan, J.R., Doll, W.E., Mandell, W.A., 2005. An evaluation of methods and available
dx.doi.org/10.1071/EG03158. software for seismic refraction tomography analysis. J Environ Eng. Geophys. 10
Palmer, D., 1981. An introduction to the generalized reciprocal method of seismic refraction (1), 21–34. http://dx.doi.org/10.2113/JEEG10.1.21.
interpretation. Geophysics 46 (11), 1508–1518. http://dx.doi.org/10.1190/1.1441157. Steeples, D.W., Miller, R.D., 1998. Avoiding pitfalls in shallow seismic reflection surveys.
Park, C.B., Miller, R.D., Xia, J., 1998. Imaging dispersion curves of surface waves on multi- Geophysics 63 (4), 1213–1224. http://dx.doi.org/10.1190/1.1444422.
channel record. SEG Expand. Abstr. 17 (1), 1377–1380. http://dx.doi.org/10.1190/ White, D.J., 1989. Two‐dimensional seismic refraction tomography. Geophys. J. Int. 97 (2),
1.1820161. 223–245. http://dx.doi.org/10.1111/j.1365-246X.1989.tb00498.x.
Park, C.B., Miller, R.D., Xia, J., 1999. Multichannel analysis of surface waves. Geophysics 64 Xia, J., Miller, R.D., Park, C.B., 2000. Advantages of calculating shear-wave velocity from
(3), 800–808. http://dx.doi.org/10.1190/1.1444590. surface waves with higher modes. Proceedings of the 70th Annual International
Park, C.B., Miller, R.D., Ivanov, J., 2002a. Filtering surface waves. Symposium on the Meeting, SEG, Expanded Abstracts, pp. 1295–1298. http://dx.doi.org/10.1190/
Application of Geophysics to Engineering and Environmental Problems 2002. Society 1.1815633.
of Exploration Geophysicists SEI9-SEI9. Xia, J., Miller, R.D., Park, C.B., Hunter, J.A., Harris, J.B., Ivanov, J., 2002. Comparing shear-
Park, C.B., Miller, R.D., Miura, H., 2002b. Optimum field parameters of an MASW survey. wave velocity profiles inverted from multichannel surface wave with borehole mea-
Exp. Abs SEG-J Tokyo 22–23 (17–18 May). surements. Soil Dyn. Earthq. Eng. 22 (3), 181–190. http://dx.doi.org/10.1016/S0267-
Pugin, A., Pullan, S.E., 2000. First-arrival alignment static corrections applied to shallow 7261(02)00008-8.
seismic reflection data. J. Environ. Eng. Geophys. 5 (1), 7–15. Xia, J., Miller, R.D., Park, C.B., Tian, G., 2003. Inversion of high frequency surface waves with
Pugin, A.J.M., Larson, T.H., Sargent, S.L., McBride, J.H., Bexfield, C.E., 2004. Near-surface fundamental and higher modes. J. Appl. Geophys. 52 (1), 45–57. http://dx.doi.org/
mapping using SH-wave and P-wave seismic land-streamer data acquisition in 10.1016/S0926-9851(02)00239-2.
Illinois, US. Lead. Edge 23 (7), 677–682. http://dx.doi.org/10.1190/1.1776740. Xia, J., Xu, Y., Miller, R.D., Ivanov, J., 2012. Estimation of near-surface quality factors by
Pugin, A.J.M., Pullan, S.E., Hunter, J.A., 2008. SV-wave and P-wave high resolution constrained inversion of Rayleigh-wave attenuation coefficients. J. Appl. Geophys.
seismic reflection using vertical impacting and vibrating sources. 21st EEGS 82 (2012), 137–144. http://dx.doi.org/10.1016/j.jappgeo.2012.03.003.
Symposium on the Application of Geophysics to Engineering and Environmental Zhang, J., Toksöz, M.N., 1998. Nonlinear refraction traveltime tomography. Geophysics 63
Problems. (5), 1726–1737. http://dx.doi.org/10.1190/1.1444468.