Journal of Applied Geophysics 198 (2022) 104572

Contents lists available at ScienceDirect

Journal of Applied Geophysics

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

High resolution seismic reflection PP and PS imaging of the bedrock surface

below glacial deposits in Marsta, Sweden
Ruixue Sun , Ayse Kaslilar , Christopher Juhlin *
Department of Earth Sciences, Uppsala University, Villavagen 16, Uppsala 75236, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: Multi-component high resolution seismic reflection data were acquired at the Uppsala University field test site in
Groundwater Marsta, Sweden, in March 2019 with the aim of obtaining an improved understanding of the subsurface structure
Seismic reflection in the area. An advantage of the site is that a number of boreholes have been drilled there for both hydro­
Converted waves
geological and geophysical purposes, allowing surface geophysics to be compared with downhole information.
Reflection data processing
Imaging
The presence of a low velocity layer above the water table generates significant processing challenges due to
Modeling trapped waves. A further complication with the data is that the upper part of this layer was frozen at the time of
the survey, resulting in the sediments just below the surface having a significantly higher velocity than those
below. By analyzing and processing both the vertical and radial component data it was possible to build a ve­
locity model that is consistent with the observed data. A strong PS converted reflection allows the bedrock to be
imaged on the radial component stacks to much higher resolution than on the vertical component stacks. Both
common conversion point binning and pre-stack depth migration were used to process the radial component
data. We confirmed that our processing strategy was effective by generating synthetic data that were processed in
a similar manner as the real data. The PS images indicate a step in the bedrock of about 2 m, with depth to it
increasing from about 15 m to 17 m close to one of the boreholes. This step is not observed on the PP stacks, due
to their lower resolution.

1. Introduction surface sediments. Therefore, the seismic reflection method in particular

can be utilized when high resolution images of the subsurface are
Shallow groundwater aquifers serve as a vital resource, but some are desired.
under threat of over-exploitation due to increased agricultural devel­ Compared to other geophysical methods, acquisition and processing
opment, industrialization and urbanization. In Sweden, accurate of seismic reflection data is costly. To reduce the cost of acquiring land-
methods for estimating quantitative information about the hydro­ based seismic data, a landstreamer was initially patented in 1976
geologic properties of groundwater aquifers is becoming increasingly (Kruppenbach and Bedenbender, 1976). Since then, landstreamer sur­
important for groundwater exploitation (https://www.sgu.se/en/grou veys have gradually increased in popularity for acquiring high quality
ndwater2/). In particular, geoelectric and magnetic geophysical seismic data over shallow sediments (Van der Veen and Green, 1998;
methods have been widely used to complement drilling and hydraulic Van Der Veen et al., 1999; Miller et al., 2005; Dolena et al., 2005; Pugin
testing (Perttu, 2011; Hongen et al., 2019; Wattanasen and Elming, et al., 2004; Suarez and Stewart, 2007; Malehmir et al., 2017; Brodic,
2008; Pedersen et al., 2005; Vizheh et al., 2020; Ismail et al., 2011; 2018). The landstreamer developed at Uppsala University (Brodic,
Persson and Erlström, 2015; Olofsson et al., 2006; Bastani et al., 2017; 2018) consists 3-component sensors, allowing improved analysis of
Jørgensen et al., 2018; Juhojuntti and Kamm, 2015; Turesson and Lind, shear wave arrivals, including PS converted wave reflections.
2005; Juhojuntti et al., 2011; Juhlin et al., 2000; Malehmir et al., 2015). PS data processing and interpretation has a number of benefits
However, resistivity and magnetic methods cannot generally map (Larson, 1996; Stewart et al., 2002, 2003). Focus has been on converted
geologic boundaries with high resolution at depth. Seismic refraction wavefield binning since the converted wave reflection point is displaced
and reflection methods will generally have better success in identifying from the CDP (Lawton, 1993; Yilmaz, 2001; Prasad et al., 2013), as well
geological and structural boundaries when characterizing the near as on converted-wave velocity analysis (eg., Alkhalifah and Tsvankin,

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

https://doi.org/10.1016/j.jappgeo.2022.104572
Received 11 August 2020; Received in revised form 18 January 2022; Accepted 10 February 2022
Available online 19 February 2022
0926-9851/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 1. Receiver (blue triangle) and borehole (B, red dots) locations in the study area Marsta, Sweden (red star in the inset). Stations 1–120 correspond to the
landstreamer with 0.5 m spacing, and stations 121–157 are 1C units with 1 m spacing. Sources were activated at 1 m distance to the west of each receiver location.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2. Quaternary map Marsta, Sweden. Blue rectangle shows the location of the study area. Modified from SGU https://apps.sgu.se/kartvisare/kartvisare-jordarter
-25-100.html. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

2
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 3. Groundwater reservoir properties near the Marsta test site. Modified from SGU https://apps.sgu.se/kartvisare/kartvisare-grundvattenmagasin.html.

Fig. 4. Depth to bedrock near Marsta, Sweden. Map modified from SGU; (https://www.sgu.se/produkter/kartor/kartgeneratorn/).

1995; Li, 2003). PS images provide information on polarization of be extended. There are several differences between PS and PP data. (1)
seismic events (Guevara and Stewart, 1998), contribute to investigating The conversion of P to S moves the energy to larger wave numbers. Due
vp/vs ratios to identify lithology (Miller, 1996; Ogiesoba and Stewart, to the higher wave numbers in the PS reflection, absorption affects the
2005; Li and Zhang, 2011) and help to delineate particle displacement at PS data more significantly than the P data (Vermeer, 2002), thus the
specific azimuths using different wave components for fracture detec­ bedrock can be characterized more completely when PS information is
tion (Li, 1997; Brodic, 2018). The PS converted wave has a different path included. (2) For dense receiver geometries, the illumination range by
from the P-wave, converting to an upcoming S-wave at different points each common conversion point (CCP) gather in the PS data is broader
on an interface, allowing the range that the interface can be mapped to than that by each common midpoint (CDP) gather in the PP data, leading

3
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

to a better signal-to-noise ratio of each image (Vermeer, 2002). (3) The

amplitude of the PS converted waves is usually greater than that of the P
wave (Schwind et al., 1960).
Three component seismic data recorded by a P-wave source can
provide shear wave velocity (Vs) information and can reduce ambigu­
ities in interpreting lithology (Miller, 1996). Since P- and PS-energy are
generally observed within distinct frequency bands, a bandpass filter
can often be used for their separation (Hardage et al., 2011). When a
low-velocity near-surface layer is present Eaton and Stewart (1989)
pointed out that nearly all incident P- and PS-wave energy are recorded
by the vertical and radial geophones, respectively. Therefore, we can
expect the vertical component to mainly contain P-wave reflections and
the radial component the converted shear wave reflections.
Our study investigates the P- and PS-wave modes to delineate
structure in the glacial deposits and depth to bedrock at the Marsta test
site. After CDP and CCP sorting for P- (vertical component) and PS-
(radial component) reflections and stacking, the final obtained P- and
Fig. 5. Geological units in the boreholes shown in Fig. 1 (B, red dots). Numbers PS-images show a clear reflection from the bedrock, consistent with
in the columns refer to the thickness of the units. (For interpretation of the borehole data and first break analysis. We compare the PS converted
references to colour in this figure legend, the reader is referred to the web wave image after CCP binning with the same data without CCP binning
version of this article.) after pre-stack depth migration, attempting to delineate horizontally
varying structure. Our preferred interpretation is supported by pro­
cessing of synthetic data.
Table 1
Borehole information in Marsta (see Fig. 1 and Fig. 5 for location and layout) 2. Geological setting
from the drilling in August 2012 and the water table level as measured in March
of 2019.
Sweden's landscape has been shaped by earthquakes, volcanism and
System Lithology Depth Hydrostratigraphy glaciation. Sweden forms part of the Fennoscandian (Baltic) Shield,
(m) which is an area of old Precambrian crystalline and metamorphic rocks.
Quaternary Clay (above water 0–3.5 Unsaturated Unconsolidated Common rocks that constitute the shield are gneiss, granite, granodio­
table) sedimentary rite, sandstone and marble. The Precambrian shield area is overlain by
Clay (below water 3.5–9 Unconfined
younger sedimentary rock cover and the Caledonian meta-sedimentary
table) aquifer
Clay/Moraine 9–10 Aquifer and meta-volcanic rocks. The Quaternary period, characterized by the
transition latest glaciations and inter-glaciations, contains the youngest deposits of
Moraine 10–15 Confined Sweden. These deposits are important not only for groundwater re­
aquifer sources, but also for agriculture, forestry, natural resources such as
Bedrock (Acidic 15–118 Basement Consolidated
intrusive rock
minerals, aggregates and peat (SGU: https://www.sgu.se/en/geology-o
including granite, f-sweden/).
granitoid and The main aquifers of Sweden are found in glacial-fluvial sand and
monzonite) gravel deposits. However, these are present over limited areas. In
southwest Sweden, aquifers are found in porous sedimentary rocks, but
these also cover only small regions as well. The most common soil type,
till, is a porous aquifer, and covers 75% of Sweden (SGU: https://www.
Table 2
Data acquisition parameters. sgu.se/en/geology-of-sweden/) and often contains groundwater. Swe­
den's sand and gravel deposits in the form of eskers provide fresh and
Attribute Parameters
easily extracted groundwater. Half of the drinking water in the country
Acquisition system SERCEL Lite 428 (3C SERCEL RAU-ex (1C UNIT) comes from groundwater and surface water (rivers and lakes) that are
landstreamer)
filtered through gravel deposits.
Sensor type DSU3 (mounted on sleds) Geophone (planted, wireless
recorders) Our survey site, Marsta, is located in southeastern Sweden, about 10
Sensor spacing 0.5 m 1m km north of Uppsala (Fig. 1, red star in the inset). The study area is
Resonant frequency – 10 Hz covered by post-glacial/glacial clay and moraine overburden overlying
Recording 0-800 Hz 4.5-400 Hz granitoid and subordinate granitoid metamorphic rocks of the Sveco­
Frequency
karelian orogeny of Fennoscandia (Fig. 2). In the region there are two
No. of channels 120 37
Sample interval 1 ms main groundwater reservoirs: The eastern reservoir, with a more than
Source type 45 kg accelerated weight drop 125 l/s flow rate (purple area in Fig. 3), is filled with glacial sediments
Source spacing/ 0.5 m 1m with a N-S trend, and the western reservoir, also filled with glacial
interval
sediments, (our study area, blue rectangle in Fig. 3) has a low flow rate
Source depth 0m
No. of source 118 37
of 1–5 l/s. At the test site, the water table extends into the low perme­
positions ability soil strata, generally with the same water capacity. To the south
Survey line [X(m)] 0–59.5 m 60-96 m and north of the profile erosion has been more severe, resulting in
Data format SEGD SEGD exposed bedrock. The bedrock depth at the site ranges from 5 to 10 m to
10-20 m (Fig. 4).

4
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 6. Example seismograms with first

break velocities indicated in the (a) vertical
component and (b) radial component and (c)
amplitude spectrum for the vertical compo­
nent time window [0-100 ms] (green curve)
and for the radial component time window
[140-250 ms] (blue curve).The blue arrow in
(c) indicates the high frequency air wave
arrival. No traces are plotted in the radial
component section past 60 m since only
vertical geophones were used in this range.
Note the strong surface waves propagating
at about 80 m/s. (For interpretation of the
references to colour in this figure legend, the
reader is referred to the web version of this
article.)

Data acquisition parameters are given in Table 2. The seismic profile

Table 3 used a dense receiver sampling and is divided into two parts due to that
First break velocities offset range based on the vertical and the profile consisted of two components with two different receiver
radial component source gathers as shown in Fig. 6a and b. types. For the northernmost part the 3C landstreamer was laid out with a
Offset range (m) Velocity [m/s] station spacing of 0.5 m with a total length of 59.5 m (stations numbered
0–20 800 1 to 120 in Fig. 1b). The 1C wireless recorders extended the land­
20–55 1600 streamer profile to the south and were used with a spacing of 1 m
55–100 5500 starting from 60 m and ending at 96 m (stations after 120 to 157 in
Fig. 1b) for a total length of 36 m. The combined length of the survey is,
thus, 96 m. The purpose of the 1C units was to ensure that bedrock
Boreholes (see Fig. 1) were drilled in 2012 at Marsta with the
refracted waves could be clearly observed from the northernmost source
composition of the lithology (Fig. 5) in them determined by the drillers.
locations. The source was a 45 kg accelerated weight drop source. The
We select the deepest borehole (depth of 118 m, B6 in Fig. 5) as a
source line was located parallel to the receiver line and was activated at
reference for this study. The borehole information shows that the
every station (except at stations 43 and 44 due to obstacles in the field)
thickness of the overburden above the bedrock is about 15 m. During
with an offset of 1 m to the west of the receiver line for a total of 155
data acquisition in March 2019, the actual water table level was
source points being activated along the profile. The amplitude spectrum
measured to be 3.5 m below the surface. Details on the sediment
of a shot-gather located at the beginning of the seismic profile (Fig. 6c)
composition are specified in Table 1. An earlier seismic experiment at
shows that the dominant frequency is about 30-70 Hz. Every shot in the
the site that tested different sources provided images of the bedrock
vertical component contains 157 traces as a result of merging the 120
surface using both P-wave reflection seismic methods and P-wave to­
channels from the landstreamer and the 37 channels from the 1C
mography (Brodic et al., 2017). However, the detailed structure of the
wireless data.
site near B6 was not imaged in that study.

3. Data acquisition 4. Data processing

High resolution seismic data were acquired in March 2019 along a N- An example source gather (Fig. 6a, located at point 1 in Fig. 1b)
S striking profile (Fig. 1). A fertilizer tank and a building are located at displays three visible velocities, estimated at around 800 m/s, 1600 m/s
the north and south ends of the seismic profile, respectively (as shown in and 5500 m/s, respectively (Table 3). In the radial component record
Fig. 1). The fertilizer tank is about 60 m away from the northernmost only the 800 m/s direct wave can be observed (Fig. 6b). This arrival is
geophone (Fig. 1, receiver 1) and the building is about 20 m away from poorly seen in the vertical component source gather, but becomes visible
the southernmost geophone (Fig. 1, receiver 157). A gravel road east of in the vertical component after sorting the data into the common
the seismic profile runs parallel to it. These structures give rise to un­ receiver gather domain. Fig. 6c shows the amplitude spectra of the
wanted reflections in our seismic records. vertical and radial component source gathers. The time window [0-100
ms] of the vertical component shows a dominant frequency of 60 Hz

5
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 7. Synthetic seismograms for the models given in Table 4 and comparison with the field data. Column 1 shows the vertical and radial components for a raw
source-gather at 0 m along the survey line. Column 2 shows the corresponding synthetic gathers for Model 1 in Table 4. Column 3 shows the corresponding synthetic
gathers for Model 2 in Table 4. In the figure ‘R', ‘M' and ‘S' denote the low velocity (Vrms ≈ 215 m/s) reflection, multiples and surface waves respectively. The red
curve in the vertical component field data shows the expected arrival time for a reflection off the top of the groundwater level for Model 1. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 4
Estimated velocities from the field data. Model 2 assumes that the first break
velocity of 800 m/s represents the velocity in the entire unsaturated zone while
Model 1 assumes that the normal velocity in the unsaturated zone is 250 m/s and
only the upper 1 m has a velocity of 800 m/s due to the ground being frozen.
Model Depth P velocity S velocity Quality Material
no. (m) (m/s) (m/s) factor Q

Unsaturated
0–1 800 100 50
frozen clay
Unsaturated
1–4 250 100 5
1 clay
Saturated clay &
4–15 1600 100 25
Moraine
15–105 5500 3000 200 Granitoid
Unsaturated
0–1 800 100 50
frozen clay
Unsaturated
1–4 800 100 5
2 frozen clay
Saturated clay &
4–15 1600 100 25
Moraine
15–105 5500 3000 200 Granitoid

(green line in Fig. 6c), while the time window [140-250 ms] of the radial
component has a dominant frequency of 30 Hz (blue line in Fig. 6c). The Fig. 8. The velocity model used for ray tracing and seismic field
blue arrow on Fig. 6c shows high frequencies in the range 160-400 Hz data processing.
caused by a strong air wave (also visible in the source gather).
processing. Below we describe how we obtained the starting Vp and Vs
velocity models for the seismic processing.
4.1. Velocity estimation
Velocity model building is an important step in seismic imaging,
including data processing, static corrections and migration (Jones,
In this study we are mainly interested in obtaining initial Vp and Vs
2010, 2012; Opara et al., 2018; Law and Trad, 2018.). There are many
models for the reflections seismic imaging. These models are then
different methods to obtain near-surface velocity models, including:
updated during the velocity analysis so that the velocities which give the
surface wave analysis (Socco et al., 2010), first arrival tomography (Shi
clearest images are used. Note that these processing velocities most
et al., 2015; Wo et al., 2020), waveform inversion (Plessix, 2008; Virieux
likely do not represent the best estimate of the subsurface velocities,
and Operto, 2009), layer-by-layer inversion (Yilmaz, 2001), and multi-
they are simply the velocities which give the best images in the

6
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 9. Vertical component (a-b) and radial component source gathers (c-d). The calculated traveltimes based on the model in Fig. 8 are overlain in b and d. The
colors and corresponding reflections are explained in the legend below the figure. The reflections are coded as trefp1 is the PP reflection off the water table, trefp2 is
the PP reflection off the bedrock, trefc1 is the PS reflection off the water table, trefc2 is the PS reflection off the bedrock, trefs1 is the-SS reflection off the water table,
trefs2 is the SS reflection off the bedrock. AW represents the air wave, and GR the ground roll.

7
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Table 5 Model 1 contains a low velocity (250 m/s) layer (unfrozen dry clay)
Field data processing flows for vertical and radial component data. below the frozen surface layer (v = 800 m/s). At the time of the survey in
Step Process March the study area had experienced low temperatures during the
winter months and the uppermost ground was frozen. Comparison of
Vertical component Radial component
source gathers from Model 1 (Fig. 7b), Model 2 (Fig. 7c) and the raw
1 Convert SEGD/SEG2 to SEGY format Read SEGD data to SEGY format data shows that Model 1 has characteristics that are more similar to the
2 Trace edit
3 Add geometry/CDP binning Add geometry/CCP binning
real data than Model 2. The Model 2 seismogram does not agree well
4 Trace balance [full trace] with the field data in terms of the reflection time and the interference by
5 Deconvolution the first arrivals with later P-wave reflections. Red curve on Fig. 7 dis­
6 Band-pass filter [50–80–180-200 Hz] Band-pass filter [20–40–120- plays the ray tracing traveltime for the reflector at depth of 4 m, based
200 Hz]
on Model 1. Ignoring the air wave impact on the field data (Fig. 7a), the
7 Trace balance [full trace]
8 Geometrical spreading amplitude compensation synthetic source gather in Fig. 7b matches quite well with the field data
9 Median filter [velocity = 330 m/s] in in terms of reflection interference, both including the reflection with
10 AGC 50 ms low RMS velocity and some multiples. The main difference between the
11 Surgical mute [front/bottom] Surgical bottom mute observed and modeled data is the lack of a clear first arrival at far offsets
12 CDP sort bin spacing 0.25 CCP sort bin spacing 0.175
13 Residual statics Residual statics
on the radial component field data. This is likely due to the background
14 Median filter [− 170 m/s] N/A surface wave noise impacting the radial component more than the ver­
15 Velocity analysis [0 ms-400 m/s, 40 ms- Velocity analysis [140 ms-700 tical component, thus masking the first arrivals.
1100 m/s] m/s]
16 NMO NMO
4.1.2. Shear wave velocity
17 Stack Stack
18 FX-Deconvolution FX-Deconvolution Observed surface wave velocities (VR) range between 70 m/s and
19 N/A FK mute 150 m/s while the P-wave velocity for the unfrozen unsaturated up­
20 Time-depth conversion [0 ms-400 m/s, Time-depth conversion [140 permost layer is estimated to VP = 250 m/s. These values are consistent
40 ms-1100 m/s] ms-210 m/s] with a shear wave velocity of about 100 m/s according to the empirical
formula vs = vR/0.9 and vs = 0.4vp for unconsolidated sediments (Burger,
1992). These estimates provide shear wave velocity constraints for the
upper unsaturated zone at the site (Fig. 8). An average shear wave ve­
locity of 100 m/s for the sediments can also be found by modeling the
traveltimes from the strong reflection in the field radial component
(Fig. 9c). Assuming it is a converted wave off the top of the bedrock, then
the calculation of its traveltime is consistent with a shear wave velocity
of 100 m/s. A shear wave velocity of 100 m/s in the saturated clay and
moraine implies a Vp/Vs ratio of 16 for this layer, a very high value.
High values have been reported in the literature (e.g. Bailey et al., 2013),
even values approaching 16 (Salas-Romero et al., 2021). The high values
can be explained by the very low shear strength of sediments even when
they are water saturated. It may be that we are overestimating the shear
wave velocity in the unsaturated zone and underestimating it in the
saturated zone. Regardless, we need very high Vp/Vs ratios in the sed­
Fig. 10. Conversion point of PS converted wave. ACP: Approximate Conversion
Point, CMP: Common-Mid-Point, CCP: Common Conversion Point. iments to explain the arrival time of the PS converted reflection off the
top of the bedrock. If we were to use a higher S-wave velocity to model
the PS converted wave arrival we would need to reduce the P-wave
geophysical methods to control velocity building (Mantovani, 2016).
velocity considerably to match the PS converted wave traveltimes. Such
When clear branches in the apparent velocities are present in the first
a reduced P-wave velocity model would be inconsistent with the P-wave
breaks it can be better to use crossover distances as a guide in the model
first arrival times which we consider to be robust. Future borehole
building rather than more advanced tomographic inversion. This is the
seismic experiments will allow more direct measurement of the shear
case for the P-wave arrivals in our data. For the shear wave velocity
wave velocity and potential verification of our assumptions. Fig. 8 shows
estimation we rely on analysis of the surface wave velocity and the
the preferred model at B6 at the site based on all available information.
arrival times of the PS converted reflection off the bedrock.
This model is taken as the reference in our data processing. The P-wave
velocity profile is based on the first breaks and the modeling shown in
4.1.1. P-wave velocity
Fig. 7 (along with Model 1 in Table 4). The S-wave velocity profile is
First arrival information (Fig. 7 and Table 3) allows us to estimate a
based on the surface wave velocities and the traveltimes for the PS
starting 3-layer model for forward modeling. We then iteratively adjust
converted wave reflection along with an assumption that the Vp/Vs ratio
the velocities layer by layer to better match the field records. Observa­
in the bedrock is on the order 1.8–1.85.
tions are primarily from the upper 50 ms in the vertical component
source gathers. A potential shallow reflection with an RMS velocity of
4.1.3. Ray tracing
215 m/s can be seen at a traveltime of about 21 ms (at the location
We are aware that the Vs estimation outlined above is rather ad hoc,
marked by ‘R' in Fig. 7a). However, the measured direct wave has a
but it serves as a starting point for the PS processing. There is significant
higher velocity of 800 m/s. In order to determine an appropriate ve­
uncertainty in using a constant value of 100 m/s for all of the sediments.
locity model for the shallow layers, two velocity models are compared
It may well be that there is a gradient in Vs with values less than 100 m/s
(Table 4). Model 2 is based on the first break arrival times, and model 1
at the shallowest levels and values greater than 100 m/s deeper down.
is an adjustment of model 2, by combining the first break information
Therefore we have performed ray tracing (crewes.org, ray tracing tools)
and the potential reflection with a 215 m/s RMS velocity. Therefore

8
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 11. Stacked time domain images; (a) PP image; (b) PS image.

9
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 12. (a) PP depth section; (b) PS depth section based on CCP binning; (c) PS image after pre-stack depth migration.

10
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

multiples and significantly broadens the frequency bandwidth, resulting

in enhanced data resolution.
After preprocessing, PP and PS modes need to be sorted under
different geometries due to the asymmetric PS wave ray-paths (Fig. 10).
We calculated the CCP position and fold by using the relations given in
Yilmaz (2001) (eqs. (1) and (2) below where x is the source-receiver
offset and and xp is the conversion point location) for a bedrock depth
of 15 m. The xp positions then replace the CDP values when binning the
radial component data to utilize the PS converted reflections. It is clear
that the PS wave illumination range is wider than the PP conventional
coverage.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅2̅
x
γ 2 + (γ2 − 1) z2p
xp = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅2̅ x (1)
x
1 + γ2 + (γ 2 − 1) z2p
/
γ = vp vs (2)

Lastly, velocity analysis and the previous velocity model estimation

provide the basis for the NMO correction. The time-velocity pairs [0 ms-
400 m/s, 40 ms-800 m/s] and [140 ms-1050 m/s] were the NMO ve­
locities used for the P- and PS-reflections, respectively. Stacked sections
in the time domain are shown in Fig. 11a-b. Seismic images of the PP, PS
and PS after pre-stack depth migration sections are shown in Fig. 12a-c,
respectively.

5. Numerical study

Fig. 13. 2D P-wave velocity model for synthetic modeling (see Table 6 for To verify our processed images, we implement 2D elastic modeling
details). The model is based on the information obtained from the seismic field by using a 2D elastic finite difference code (Juhlin, 1995) from the
data (Fig. 12). Seismic Unix Software Package (www.cwp.mines.edu). We use the ve­
locity model given in Fig. 13 based on the processing information for the
final images (Fig. 12). A 2 m vertical step is set at 30 m lateral location.
Table 6
The increase in depth from 15 m down to 17 m is similar to what is
2D velocity model used in synthetic modeling.
observed in the bedrock interpretation in the PS sections (Fig. 12b-c).
Depth P-velocity (m/s) Depth P-velocity (m/s) S-velocity (m/s) However, the main purpose of the modeling is to test our processing flow
(m) (distance 0-30 (m) (distance 30-60 (distance 0-60
and determine if the step can be better imaged using the PS reflections.
m) m) m)
That is, the objective is to examine how much improvement can be
1 800 1 800 100 expected with PS processing compared to PP processing with a 2 m step
2–4 250 2–4 250 100
4–15 1600 4–17 1600 100
at about 15 m depth. The details of the model are given in Table 6.
15–105 5500 17–105 5500 3000 A Gaussian wavelet derivative having a dominant frequency of 90 Hz
was used as a source along with a source spacing of 1 m. In total, 120
records were generated and the wavefield recorded on 120 channels
on the preferred model (Fig. 8) to see how well we can match the PP and spaced at 0.5 m. The general appearance of the synthetic gathers is
PS reflections from the water table (at 4 m depth) and bedrock (at 15 m similar to the field data, both in the radial and vertical components
depth). In Fig. 9a-b we see that the P-wave reflections from the water (Fig. 14). The PS converted wave reflection in Fig. 14a is clearly visible,
table (red curve, Fig. 9b) and bedrock (green curve, Fig. 9b) are while the PP reflection from the bedrock is less obvious. Note again that
consistent, while these events in the radial component (Fig. 9c-d) are not the poor visibility of the P-wave first arrival in the field data radial
clearly visible and only the assumed PS converted wave arrival (purple component compared to the synthetic data. We attribute this to the
curve, Fig. 9d) is matching with the strong event in the field data. The greater level of surface wave noise on the radial component compared to
good match for the PS reflection off the bedrock gives us some confi­ the vertical component. By applying a similar processing flow as given in
dence that using a Vs of 100 m/s is a reasonable starting assumption for Table 5, we obtain the time domain images shown in Fig. 15a-b and use
the seismic processing. them for comparison with the field data.

4.2. Multi-component seismic data processing 6. Discussion

We process the multi-component high-resolution seismic data to We interpret the high amplitude reflection at 42 ms (PP) and 150 ms
obtain both PP and PS reflection images. The processing flow for the -170 ms (PS) in Fig. 11 to mark the transition from unconsolidated
vertical and radial components are shown in Table 5. Strong surface sediments to consolidated bedrock. A near-vertical step can be inter­
waves at near offsets are suppressed by bandpass filtering and a surgical preted in the PS image at a distance of about 30 m along the profile, but
cone mute (steps 6 and 11). Frequency filters of 50–80-180-200 Hz for it is difficult to see any small corresponding increase in time in the PP
enhancing the PP reflections and 20–40-120-200 Hz for enhancing PS image. The synthetic modeling shows similar results (compare Fig. 15
reflections were chosen. Then an optimum window (Hunter et al., 1982) with Fig. 11), no clear increase in the traveltime of the bedrock reflec­
was chosen for a suitable surgical mute. The muted regions were tion is observed on the PP section while a clear increase is observed on
determined by overlaying the curves obtained via ray tracing on the the PS section. This is likely due to the P-wave's lower resolution. At 60
source gathers (Fig. 9). A dip dependent median filter with a velocity of Hz the wavelength for P-waves in the sediments above the bedrock is
330 m/s helped to suppress the air waves. Deconvolution reduces

11
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 14. Comparison of seismic field data and synthetic data. (a) Radial component field data; (b) Vertical component field data; (c) Radial component synthetic
data; (d) Vertical component synthetic data. The dashed lines mark the assumed PS (a and c) and PP (b and d) reflections from the bedrock layer.

about 25 m (assuming Vp = 1600 m/s), while the wavelength of S-waves arrival of the direct wave through the uppermost 800 m/s layer is most
(assuming Vs = 100 m/s) is about 1.5 m. This explains why the PS clearly seen on the radial component data (Fig. 14a). It is not easily seen
reflection shows the bedrock interface and its structure much better than in the vertical component (Fig. 14b). This provides an additional
the PP section, both on the field and synthetic data. If only vertical incentive to record multi-component data for near surface studies. If the
component sensors had been used in the survey then an image of this survey had been carried out in the summer we would expect that the
step in the bedrock would not have been obtained. The step is also not sediments above the groundwater table (0-4 m depth) to all have a P-
visible on the refraction data in the first break arrivals so it would not be wave velocity of 250 m/s.
possible to interpret it from tomography, nor first break processing.
The PS image shows that the interpreted step in the bedrock is close 7. Conclusion
to the location of the B6 borehole. Bedrock is found in the borehole at 15
m depth, suggesting that the borehole penetrates the bedrock on the The present study along with that of Brodic et al. (2017) are initial
northern side of the step. This interpretation is supported by B9 (Figs. 1 steps to a better understanding of the near-surface structure in the
and 5), located just 10 m south of B6 and which shows a bedrock depth Marsta area and can form the basis for integrated studies with hydro­
of 18 m. At present, the borehole measurements are consistent with the geology. In the current study, we processed high resolution seismic 3C
bedrock topography in Fig. 12b-c, although changing the S-wave ve­ data with two major goals: 1) velocity estimation based on the field data
locity by 10–15% could change the depths by a few meters. Future and 2) imaging the bedrock surface. By means of first break analysis and
borehole measurements of the S-wave velocity and surface wave anal­ forward modeling, an initial model was obtained. Ray tracing according
ysis will provide better constraints on the S-wave structure at the site to the estimated model helped understand PP and PS events on the field
and the best velocity model to use for seismic processing. We hope with recorded vertical and radial component source gathers. Acquisition of
these future studies to be able to verify the high Vp/Vs ratios that our the radial component data allowed more detailed velocity estimation
interpretation implies to be present at the site. than if only vertical component data had been obtained. Observed sur­
It is interesting to note that in order to simulate some of the char­ face wave velocities and traveltimes of the PS converted wave imply
acteristics of the field data it is necessary to include a low velocity layer very high Vp/Vs ratios at the site. We aim to verify these high ratios by
below the uppermost layer (Fig. 7). Setting the upper 1 m to a P-wave future borehole seismic measurements.
velocity of 800 m/s, as observed in the first breaks, and the 3 m thick The processed stacked PS images display a small step (2 m) in the
layer below to 250 m/s shows the importance of carefully analyzing the bedrock surface at a lateral location of about 30 m along the profile,
recorded seismograms. The change in properties in the near surface deepening from 15 m to 17 m, close to the location of borehole B6. The
during winter conditions would need to be accounted for in any PS image is supported by processing of data generated by 2D forward
refraction survey carried out during this season. Note also that the elastic modeling that includes a step discontinuity at 30 m along the

12
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

Fig. 15. Stacked synthetic images in the time domain (a) PP image; (b) PS image.

profile. Without the radial component field data this step would not References
have been observed due to the lower resolution of the P-waves and
interference of surface waves with the PP reflections from the bedrock. Alkhalifah, T., Tsvankin, I., 1995. Velocity analysis for transversely isotropic media.
Geophysics 60, 1550–1566. https://doi.org/10.1190/1.1443888.
Bailey, B.L., Miller, R.D., Peterie, S., Ivanov, J., Markiewicz, R., 2013. Implications of
CRediT authorship contribution statement Vp/Vs ratio on shallow P and S reflection correlation and lithology discrimination.
SEG Techn. Progr. Expand. Abstr. 1944–1949. https://doi.org/10.1190/segam2013-
1107.1.
Ruixue Sun: Conceptualization, Methodology, Visualization, Bastani, M., Persson, L., Löfroth, H., Smith, C.A., Schälin, D., 2017. Analysis of ground
Writing – original draft. Ayse Kaslilar: Writing – review & editing, geophysical, airborne TEM, and geotechnical data for mapping quick clays in
Visualization. Christopher Juhlin: Writing – review & editing, Super­ Sweden. In: Thakur L’Heureux, J.S., Locat, A.V. (Eds.), Landslides in Sensitive Clays:
From Research to Implementation. Springer, Swedish Geotechnical Institute,
vision, Resources. pp. 463–474. https://doi.org/10.1007/978-3-319-56487-6_41. Chapter 41.
Brodic, B., 2018. Three-Component Digital-Based Seismic Landstreamer: Methodologies
Declaration of Competing Interest for Infrastructure Planning Applications. PhD thesis. Uppsala University, p. 82.
Brodic, B., Malehmir, A., Juhlin, C., 2017. Bedrock and fracture zone delineation using
different near-surface seismic sources. In: Conference Proceedings, 23rd European
The authors declare that they have no known competing financial Meeting of Environmental and Engineering Geophysics (EAGE), 2017, pp. 1–5.
interests or personal relationships that could have appeared to influence https://doi.org/10.3997/2214-4609.201702068.
the work reported in this paper. Burger, H.R., 1992. Exploration Geophysics of the Shallow Subsurface. Prentice Hall,
Englewood Cliffs, N.J., pp. 18–19
Dolena, T.M., Link, C.A., Miller, P.F., Duaime, T.E., 2005. A land streamer aided, Three-
Acknowledgements dimensional (3-D) seismic reflection survey, Belt, Montana. In: Proc. Symposium on
the Applications of Geophysics to Engineering and Environmental Problems,
pp. 971–978. https://doi.org/10.4133/1.2923555.
We thank two anonymous reviewers for their valuable suggestions Eaton, D.W.S., Stewart, 1989. Aspects of seismic imaging using P-SV converted waves.
for improving the manuscript. RS thanks CSC (China Scholarship CREWES Res. Rep. 1 (6), 25.
Council) for financial support. Research data are not shared. GLOBE Guevara, S.E., Stewart, R.R., 1998. Multicomponent Seismic Polarization Analysis.
CREWES Res. Rep. 109 (7), 19.
Claritas™ under license from the Institute of Geological and Nuclear Hardage, B.A., DeAngelo, Michael V., Murray, P.E., Sava, D., 2011. Multicomponent
Sciences Limited, Lower Hutt, New Zealand was used for the reflection seismic technology. In: Geophysical References Series. Society of Exploration
seismic processing. Geophysicists. https://doi.org/10.1190/1.9781560802891.
Hongen, L., Zheng, L., Haifeng, X., Ning, Z., Yongjun, H., 2019. Embankment dam
leakage detection by joint use of magnetic resonance sounding and electrical

13
R. Sun et al. Journal of Applied Geophysics 198 (2022) 104572

resistivity imaging. IOP Conf. Ser. Earth Environ. Sci. 304, 042005 https://doi.org/ Opara, C., Adizua, O.F., Ebeniro, J.O., 2018. Near-surface seismic velocity model
10.1088/1755-1315/304/4/042005. building from first arrival travel-times - a case study from an onshore, Niger Delta
Hunter, J.A., Burns, R.A., Gané, R.M., Good, R.L., MacAulay, H.A., 1982. Field Field. Univ. J. Phys. Applic. 12, 1–10. https://doi.org/10.13189/ujpa.2018.120101.
experience with the “Optimum Window” hammer seismic reflection technique. SEG Pedersen, L.B., Bastani, M., Dynesius, L., 2005. Groundwater exploration using combined
Techn. Progr. Expand. Abstr. 1982, 466–468. https://doi.org/10.1190/1.1827094. controlled-source and radiomagnetotelluric techniques. Geophysics 70. https://doi.
Ismail, N., Schwarz, G., Pedersen, L.B., 2011. Investigation of Groundwater Resources org/10.1190/1.1852774, 12JF-Z26.
using Controlled-source Radio Magnetotellurics (CSRMT) in Glacial Deposits in Persson, L., Erlström, M., 2015. Geophysical imaging of silurian carbonates by using of
Heby, Sweden. J. Appl. Geophys. 73, 74–83. https://doi.org/10.1016/j. ground and airborne electromagnetic and radiometric methods on the Island of
jappgeo.2010.11.008. Gotland, Sweden. Interpretation 3, SY1–SY11. https://doi.org/10.1190/INT-2014-
Jones, I.F., 2010. An Introduction to Velocity Model Building. European Association of 0186.1.
Geoscientists & Engineers (EAGE). Perttu, N., 2011. Magnetic Resonance Sounding (MRS) in Groundwater Exploration, with
Jones, I.F., 2012. Tutorial: incorporating near-surface velocity anomalies in pre-stack Applications in Laos and Sweden. Ph.D. Thesis. Luleå University of Technology,
depth migration models. First Break 30, 47–58. p. 134.
Jørgensen, F., Erlström, M., Persson, L., Bastani, M., Sopher, D., Gulbrandsen, M.L., Plessix, R.-E., 2008. Introduction: towards a full waveform inversion. Geophys. Prospect.
Dahlqvist, P., 2018. A 3D geological model of the Island of Gotland based on 56, 761–763. https://doi.org/10.1111/j.1365-2478.2008.00736.x.
extensive airborne EM mapping, seismic data and log stratigraphy. In: 7th Prasad, T.K., Barve, B.K., Sarvesam, G., Khanna, R.K., Baskaran, K., 2013. Land
International Workshop on Electromagnetics (AEM2018). multicomponent seismic survey: PS binning approach–a case study. In: 10th Biennial
Juhlin, C., 1995. Finite difference elastic wave propagation in 2-D heterogeneous International Conference & Exposition.
transversely isotropic media. Geophys. Prospect. 43, 843–858. https://doi.org/ Pugin, A.J.M., Larson, T.H., Sargent, S.L., McBride, J.H., Bexfield, C.E., 2004. Near-
10.1111/j.1365-2478.1995.tb00284.x. surface mapping using SH-wave and P-wave seismic land-streamer data acquisition
Juhlin, C., Palm, H., Müllern, C.-F., Wållberg, B., 2000. High-resolution reflection in Illinois, U.S. Lead. Edge 23, 677–682. https://doi.org/10.1190/1.1776740.
seismics applied to detection of groundwater resources in glacial deposits, Sweden. Salas-Romero, S., Malehmir, A., Snowball, I., Brodic, B., 2021. Geotechnical site
Geophys. Res. Lett. 27, 1575–1578. https://doi.org/10.1029/1999GL011295. characterization using multichannel analysis of surface waves: a case study of an
Juhojuntti, N., Kamm, J., 2015. Joint inversion of seismic refraction and resistivity data area prone to quick-clay landslides in Southwest Sweden. Near Surf. Geophys.
using layered models — applications to groundwater investigation. Geophysics 80, https://doi.org/10.1002/nsg.12173. In press.
EN43–EN55. https://doi.org/10.1190/geo2013-0476.1. Schwind, J.J., Berg Jr., J.W., Cook, K.L., 1960. PS converted wave from large explosions.
Juhojuntti, N., Aaro, S., Jönberger, J., Larsson, O., 2011. Combined gravity and seismic J. Geophys. Res. 65, 3817–3824. https://doi.org/10.1029/JZ065i011p03817.
measurements for mapping a buried tectonic valley in Western Sweden. In: European Shi, T., Zhang, J., Huang, Z., Jin, C., 2015. A layer-stripping method for 3D near-surface
Association of Geoscientists & Engineers (EAGE) Near Surface 2011 - 17th EAGE velocity model building using seismic first-arrival times. J. Earth Sci. 26, 502–507.
European Meeting of Environmental and Engineering Geophysics, cp-253-00012. https://doi.org/10.1007/s12583-015-0569-0.
https://doi.org/10.3997/2214-4609.20144376. Socco, L.V., Foti, S., Boiero, D., 2010. Surface-wave analysis for building near-surface
Kruppenbach, J.A., Bedenbender, J.W., 1976. Towed Land Cable: US Patent no. 3 954 velocity models — established approaches and new perspectives. Geophysics 75,
154. 1SO–Z116. https://doi.org/10.1190/1.3479491.
Larson, G.A., 1996. Acquisition, Processing, and Interpretation of P-P and P-S 3D Seismic Stewart, R.R., Gaiser, J.E., Brown, R.J., Lawton, D.C., 2002. Converted-wave seismic
Data. M.Sc. Thesis. University of Calgary, p. 128. exploration: methods. Geophysics 67, 1348–1363. https://doi.org/10.1190/
Law, B.K., Trad, D., 2018. Robust refraction statics solution and near-surface velocity 1.1512781.
model building using feedback from reflection data. Geophysics 83, U63–U77. Stewart, R.R., Gaiser, J.E., Brown, R.J., Lawton, D.C., 2003. Converted-wave seismic
https://doi.org/10.1190/geo2018-0060.1. exploration: applications. Geophysics 68, 40–57. https://doi.org/10.1190/
Lawton, D.C., 1993. Optimum bin size for converted-wave 3-D asymptotic mapping. 1.1543193.
CREWES Res. Rep. 5 (28), 14. Suarez, B.G.M., Stewart, R.R., 2007. Acquisition and analysis of 3C land streamer data.
Li, X.Y., 1997. Fractured reservoir delineation using multicomponent seismic data. CREWES Res. Rep. 19 (19), 12.
Geophys. Prospect. 45, 39–64. https://doi.org/10.1046/j.1365-2478.1997.3200262. Turesson, A., Lind, G., 2005. Evaluation of electrical methods, seismic refraction and
x. ground-penetrating radar to identify clays below sands - two case studies in SW
Li, X.Y., 2003. Converted-Wave Moveout Analysis Revisited: The Search for a Standard Sweden. Near Surf. Geophys. 3, 59–70. https://doi.org/10.3997/1873-
Approach: 73rd SEG Annual Meeting, pp. 805–808. 0604.2005001.
Li, X.Y., Zhang, Y.G., 2011. Seismic reservoir characterization: How can multicomponent Van der Veen, M., Green, A.G., 1998. Land streamer for shallow seismic data acquisition:
data help? J. Geophys. Eng. 8, 123–141. https://doi.org/10.1088/1742-2132/8/2/ evaluation of gimbal-mounted geophones. Geophysics 63, 1408–1413. https://doi.
001. org/10.1190/1.1444442.
Malehmir, A., Zhang, F., Dehghannejad, M., Lundberg, E., Döse, C., Friberg, O., Van Der Veen, M., Wild, P., Spitzer, R., Green, A., 1999. Design characteristics of a
Brodic, B., Place, J., Svensson, M., Möller, H., 2015. Planning of Urban underground seismic land streamer for shallow data acquisition. In: 61st EAGE Conference and
infrastructure using a broadband seismic landstreamer — tomography results and Exhibition. European Association of Geoscientists & Engineers. https://doi.org/
uncertainty quantifications from a case study in Southwestern Sweden. Geophysics 10.3997/2214-4609.201407807 cp-132-00210.
80. https://doi.org/10.1190/geo2015-0052.1, 1ND-Z124. Vermeer, Gijs J.O., 2002. 3-D seismic survey design: Society of Exploration
Malehmir, A., Maries, G., Bäckström, E., Schön, M., Marsden, P., 2017. Developing cost- Geophysicists. ISBN-13: 978–1560801139 ISBN-10: 1560801131, chapter 6.
effective seismic mineral exploration methods using a land-streamer and a drop Virieux, J., Operto, S., 2009. An overview of full-waveform inversion in exploration
hammer. Sci. Rep. 7, 10325. https://doi.org/10.1038/s41598-017-10451-6. geophysics. Geophysics 74. https://doi.org/10.1190/1.3238367, 1ND-Z107.
Mantovani, M., 2016. Near-surface modelling: a multiphysics approach. Lead. Edge 35, Vizheh, M.M., Oskooi, B., Bastani, M., Kalscheuer, T., 2020. Using GPR data as
926–1008. https://doi.org/10.1190/tle35110968.1. constraints in RMT data inversion for water content estimation: a case study in Heby,
Miller, S.L.M., 1996. Multicomponent Seismic Data Interpretation. M.Sc. Thesis. Sweden. Pure Appl. Geophys. 177, 2903–2929. https://doi.org/10.1007/s00024-
University of Calgary, p. 96. 019-02391-1.
Miller, C.R., Allen, A.L., Speece, M.A., El-Werr, A.-K., Link, C.A., 2005. Land streamer Wattanasen, K., Elming, S., 2008. Direct and indirect methods for groundwater
aided geophysical studies at Saqqara, Egypt. J. Environ. Eng. Geophys. 10, 371–380. investigations: a case of MRS and VES in the Southern part of Sweden. J. Appl.
https://doi.org/10.2113/JEEG10.4.371. Geophys. 66, 104–117. https://doi.org/10.1016/j.jappgeo.2008.04.005.
Ogiesoba, O.C., Stewart, R.R., 2005. VP/VS from prestack multicomponent seismic data Wo, Y., Zhou, H., Hu, H., Zong, J., Ding, Y., 2020. A layer-cell tomography method for
and automatic PS to PP time mapping. CREWES Res. Rep. 17 (37), 15. near-surface velocity model building using first arrivals. Pure Appl. Geophys. 177,
Olofsson, B., Jernberg, H., Rosenqvist, A., 2006. Tracing leachates at waste sites using 4161–4175. https://doi.org/10.1007/s00024-020-02466-4.
geophysical and geochemical modelling. Environ. Geol. 49, 720–732. https://doi. Yilmaz, O., 2001. Seismic Data Analysis Society of Exploration Geophysicists. https://
org/10.1007/s00254-005-0117-9. doi.org/10.1190/1.9781560801580.

14