You are on page 1of 15

Near Surface Geophysics, 2020 doi: 10.1002/nsg.

12101

Reprocessing of high-resolution seismic data for imaging of shallow


groundwater resources in glacial deposits, SE Sweden
Ruixue Sun∗ , Ayse Kaslilar and Christopher Juhlin
Earth Sciences Department, Uppsala University, Villavagen 16, Uppsala 75236, Sweden

Received November 2019, revision accepted March 2020

ABSTRACT
Reprocessing of high-resolution seismic reflection data over groundwater-bearing
glacial deposits near Heby, southeastern Sweden, improved the images of near-
surface structure at this site. Post-stack time migration and pre-stack depth migration
were tested and compared to determine the improvements on imaging an undulating
bedrock surface. The pre-stack depth migration image displays better continuity of
the dipping structures within the glacial sediments and provides a more detailed to-
pography of the bedrock reflector. First-arrival picks were used to define an initial
model for input into tomographic inversion. The tomography result then formed the
basis for building the migration velocity model. The final pre-stack depth migration
image shows a strong reflection at around 35 m elevation (about 9 m below the
surface) that can be correlated to a thin (0.2 m) hard silt layer. The upper 20 m of
overburden is interpreted to consist of clay, and the seismic images show weaker
sub-horizontal reflections within this unit, except for the strong silt reflection, con-
sistent with our modelling results. Below 20 m, sand/gravel sediments are present
and overlay the bedrock. Forward modelling based on the pre-stack depth migration
image and subsequent processing shows that pre-stack depth migration provides a
higher resolution image compared with post-stack time migration. Our study shows
that pre-stack depth migration is preferable to post-stack time migration even for the
shallow near-surface seismic data acquired at Heby, and that integrating tomography,
migration, modelling and geological information provides a better understanding of
the structure which these groundwater resources are contained in.

Key words: Groundwater, Migration, Modelling, Seismic, Tomography.

INTRODUCTION ages of structure than refraction methods (Bachrach and Nur,


1998; Baker et al., 1998), but these methods have not been
Seismic refraction methods and ground penetration tests (e.g.
used to the same extent in environmental and engineering ap-
Voyiadjis and Song, 2003; Butler, 2005; Dietrich and Leven,
plications as in the oil and gas industry. However, some early
2006) have been developed for groundwater exploration since
studies showed that the potential of these methods existed to
the 1940s (e.g. Hasselström, 1969; Haeni, 1986). The applica-
image the upper 100 m of the subsurface (e.g. Hunter et al.,
tion of seismic reflection methods for shallow hydrogeological
1984; Birkelo et al., 1987; Steeples et al., 1997; Ghose et al.,
problems started in the 1980s (e.g. Geissler, 1989; Bruno and
1998; Juhlin et al., 2000; Bradford and Sawyer, 2002). In the
Godio, 1997; Buker et al., 2000; Giustiniani et al., 2008).
last few decades, the popularity of reflection seismic meth-
Seismic reflection methods may provide higher-resolution im-
ods has increased significantly in near-surface studies (e.g.
Schmelzbach et al., 2005; Pugin et al., 2009; Sloan et al., 2010;
∗ E-mail: ruixue.sun@geo.uu.se Maries et al., 2016; Burschil et al., 2018). A key aim in many

C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association 1
of Geoscientists and Engineers.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction
in any medium, provided the original work is properly cited.
2 R. Sun, A. Kaslilar and C. Juhlin

of these studies has been to map the top of the groundwater shallow high-resolution seismic reflection profile (Juhlin et al.,
table and the thickness of the groundwater-bearing sediments. 2002) acquired nearly 20 years ago over groundwater-bearing
In seismic groundwater studies, the P-wave velocity in glacial deposits in the Heby area, southeastern Sweden. We
unconsolidated sediments is significantly affected by the tran- chose this data set since the data are of very high quality over
sition from dry to partially saturated to fully saturated con- much of the profile, with usable frequencies up to 500 Hz and
ditions (Bachrach and Nur, 1998). Water capillary forces in clear reflections in the shot gathers. First, we apply first-arrival
the capillary fringe push the actual water level higher than tomography based on an initial model derived from the first-
the water table, resulting in partial saturation. Hasselström arrival picks with subsequent refraction statics calculations.
(1969) suggested P-wave velocities of 1200 to 1800 m/s in Second, we compare images from PSDM and post-stack time
water-bearing porous materials. Later, Haeni (1986) showed migration (PSTM). Lastly, we numerically model the Heby
that saturated glacial sediments do not have a unique veloc- subsurface to generate synthetic seismograms that we process
ity, but that it varies depending upon the composition of the to compare seismic images from both PSDM and PSTM in
sediments. Velocities as low as 1000 to 1200 m/s have been order to test the validity of PSDM on imaging and locating
reported in some cases (Grelle and Guadagno, 2009). How- the shallow reflecting interfaces better.
ever, the P-wave velocity contrast between the unsaturated
and saturated zones is nearly always high.
GEOLOGICAL SETTING
More recently, 2D first-arrival tomography has been used
to produce P-wave velocity models of the subsurface in order Sweden is a part of the Fennoscandian Shield with mostly
to better map lateral variations in the velocity (e.g. Bishop Precambrian igneous and metamorphic rocks comprising the
et al., 1985; Lankston, 1990; Lanz et al., 1998; Morey and basement. The most common overburden soil type is till that
Schuster, 1999; Yordkayhun, 2011; Maries et al., 2016; Liu was deposited through several glaciation and deglaciation pe-
and Wu, 2018). It is a relatively cost-efficient and effective riods. The Heby field site (Fig. 1a) is located in the N–S
method to obtain information on the uppermost layers, but, trending Hårsbäcksdalen valley (Juhlin et al., 2002) with the
as with standard refraction interpretations, the rays do not 528 m long Heby 1 profile running nearly perpendicular to
penetrate very deep because of the large velocity contrast be- the strike of the valley. The elevation of the westernmost part
tween sediment and bedrock. To increase penetration depth, of the profile is 52 m with the ground consisting of sandy till,
long offsets can be used, but at the cost of resolution. changing to clay east of 60 m along the profile. The elevation
An additional complicating factor in refraction analysis of the easternmost end of the profile is 43 m.a.s.l. Three bore-
can be the significant topography of the interfaces of interest, holes had been drilled along the profile that provide important
such as the bedrock surface. If severe, the interfaces cannot be information concerning the interpretation of the seismic data
mapped with refraction methods if the source–receiver offsets (Fig. 1b). BH1 (R0007 in Juhlin et al., 2002) consists of clay
are limited due to logistics, e.g. forests on the sides of fields. down to 10.5 m with a hard 0.2 m thick silt layer present
Reflection seismic methods may then be attractive since the at 9.5 m. Below 10.5 m, sand is present down to the final
normally strong elastic properties contrast between the sedi- drilled depth of 15 m. Clay is present in BH2 (S0003 in Juhlin
ments and the crystalline bedrock should be a good reflector. et al., 2002) down to 21 m with a 0.2 m thick hard silt layer at
One challenge in reflection seismic processing and interpreta- 10.1 m. Below 21 m, sand and coarse sand is present to 39.1 m.
tion is determination of the velocity to use for depth conver- From 39.1 to 48 m sandy and cobbley gravel are present. The
sion. Some velocity information is obtained from the normal interval 48 to 59 m consists of course sand and the lowermost
moveout corrections, but these velocities may not represent metre is moraine. An even thicker sequence of clay is found
the best velocity function to use for depth conversion (Dix, in BH3 (S0004 in Juhlin et al., 2002), extending down to
1955; Mayne, 1962; Bradford, 2002; Bradford and Sawyer, 25.6 m and containing a 0.2 m thick course silt layer 10.8 m.
2002; Perz et al., 2019). This is especially the case when dip- A 1.3 m thick silt layer is also present just below the clay in
ping structures are present. For such situations, a better strat- this borehole. From 26.9 to 38 m, medium sand occurs with
egy to improve the image may be to determine the near-surface sandy gravel in the interval 38 to 40.6 m. The lowermost two
velocity structure from tomography and then apply pre-stack metres of the borehole were drilled into crystalline basement.
depth migration (PSDM) using a layer-stripping approach to The groundwater reservoir (confined aquifer) at the Heby
update the velocity model. We test this approach to improve site is in the sand-gravel units below the clay layers. At the
the seismic image in this paper by reprocessing data from a time of the survey, the pressure in this reservoir corresponded


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 3

Figure 1 (a) Location of the study area from Google Earth (two maps on the right) and the geological map from the Geological Survey of
Sweden (main map) (http://apps.sgu.se/kartgenerator/maporder_en.html) with the location of the seismic profile and boreholes BH1, BH2 and
BH3 (right top corner). (b) The profile with the geological units identified in the boreholes plotted at their respective locations.


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
4 R. Sun, A. Kaslilar and C. Juhlin

Figure 2 Data acquisition geometry. (a) Elevation of sources and receivers, left side of blue arrows have 2 m shot spacing and 1 m receiver
spacing, conversely, right side of green arrows have shot spacing 4 m and receiver spacing 2 m. Shots 10, 50, 100 and 150 represent the locations
of the shot gathers shown in Fig. 3. (b) CDP fold resulting from the employed acquisition geometry.

to a water level of c. 35 m (potentiometric surface) above along the profile given in 2(b). Considering the parameters
sea level, corresponding to a depth of about 9 m below the given in Table 1, the nominal fold varies from 24 in the west
central part of the seismic profile (Müllern, 2008). The top of to 12 in the east.
the saturated zone (groundwater level) was much shallower For illustrative purposes, we show four shot gathers (see
than 9 m as evidenced by the presence of a brook just east Fig. 2a for locations) acquired along the profile with the first
of the profile and calculations that can be made from first arrivals and subsurface reflections marked (Fig. 3a–d). Note
arrivals (Juhlin et al., 2002) giving a depth of a few metres that ground is very weak in most shot gathers, especially for
to the top of this zone. Therefore, the groundwater reservoir those fired in the clay. This is probably due to that shear
and the surface waters are decoupled from one another by the wave velocities in the uppermost sediments are very low and
clay layer at this location. that the shots were fired at a depth of 1.4 m. Likewise, direct
shear wave arrivals are almost absent. Velocities based on the

DATA ACQUISITION Table 1 Data acquisition parameters for the Heby 1 profile

Data were acquired in September 2000 using two 24-channel Parameter


ABEM Terraloc seismographs. To ensure the highest possible
signal quality a 50 g dynamite source was used with shot Source Dynamite
Geophone Single 60 Hz
holes being drilled with a motorized auger down to 1.4 m.
Receiver spacing 1 m/2 m
This source strategy was based on the successful imaging of Source depth 1.4 m
the bedrock interface at another site in an earlier test (Juhlin Profile length 528 m
et al., 2000). However, for the Heby 1 profile, the acquisition Number of channels 48
geometry was changed along the profile to test two different Shot spacing 2 m/4 m
Sample interval 0.1 ms
source and receiver spacings (Fig. 2a). Additional acquisition
Record length 409 ms
parameters are shown in Table 1 with the CDP fold variation


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 5

Figure 3 Examples of raw shot gathers from the dynamite source at different locations along the profile as shown in Fig. 2a. (a) Shot 10; (b)
shot 50; (c) shot 100; (d) shot 150; (e) amplitude spectra in the interval 0–80 ms from the respective shots. Blue lines mark first breaks (FB), red
curves mark the upper silt layer, purple curves represent the top of the cobbley gravel (CG) and green curves mark the bedrock reflection (BR).

first arrivals for corresponding shots are given in Table 2. On DATA PROCESSING AND RESULTS
some of these shot gathers, the silt reflection (red curve) and
In order to compare the post-stack time migration (PSTM)
bedrock (BR) reflection (green curve) can be clearly distin-
to produce pre-stack depth (PSDM) processing it was neces-
guished. Reflections from within the sedimentary section can
sary to first apply standard processing methods to the Heby
also be identified in some of the raw shot gathers. Amplitude
1 profile. To do this, a similar processing strategy as outlined
spectra from the time window 0–80 ms show a dominant fre-
in Juhlin et al. (2002) was applied to generate the stacked
quency of about 125 Hz. However, useful frequencies up to
500 Hz are present in three of the shots (shot 50, shot 100 and
shot 150), demonstrating the broadband nature of the dyna-
Table 3 PSDM and PSTM data processing flow
mite source. It is only the westernmost shot (shot 10) that has
a 200 Hz upper bound for the useful frequency range, prob- Data processing flow
ably due to the shot being fired in dry sandy till. Given the
Step Process
first-arrival velocities of 460 m/s and 6000 m/s and a crossover
distance of about 20 m, the thickness of the sandy till layer at 1 Geometry installation and trace editing
this location can be estimated to about 8 m. 2 Geometrical spreading correction
3 Trace balance: 0–100 ms
Table 2 Measured first-arrival velocities for corresponding shot 4 Refraction statics
gathers in Fig. 3 5 Spectral whitening and Bandpass filtering: 100–150–400–
600 Hz
Shot First-arrival velocity (m/s) 6 Fixed datum correction: datum = 45 m, velocity = 1600 m/s
7 NMO
10 460/6000 8 Residual statics (iterations before and after velocity analysis)
50 1388 9 PSDM and RMO (Fig. 8b)
100 1412 10 Stack
150 1382 11 PSTM and time-depth conversion (Fig. 8a)


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
6 R. Sun, A. Kaslilar and C. Juhlin

Figure 4 Velocities from refraction statics that were used as the initial model for tomography. Text boxes mark velocities at certain distances
along the profile.

section. The work flow shown in Table 3 provides a gen- each iteration of the inversion and after six iterations, it had
eral overview of the processing flow. PSDM and PSTM pro- dropped from about 0.04 s to about 0.0025 s. The resulting
cessing steps are identical until the normal moveout (NMO) tomographic velocity model is shown in Fig. 6(b). The velocity
step. model obtained from refraction statics (Fig. 4) indicates that
Proper migration depends heavily on a good velocity the saturated clay has a velocity on the order of 1400–1500
model. This is true for both PSTM and PSDM, but especially m/s, consistent with other observations (Haeni, 1986; Grelle
for PSDM. Reflection traveltime tomography can provide and Guadagno, 2009). For our tomographic image, we define
subsurface velocity structure which can be used as a back- the 1400 m/s contour as the top of the saturated zone. Thus,
ground or initial model for PSDM or full waveform inversion the tomographic inversion indicates the top of the saturated
(Li et al., 2019). We find that first arrivals provide some zone to be at a few metres depth over much of the profile at
immediate information on the velocities along the profile the time of the survey. There is a tendency for this velocity
(Fig. 3a–d). Picked arrivals were input into a refraction statics to be reached at a greater depth in the eastern part of the
module to produce a two-layer model (Fig. 4) which served as profile.
the starting model for first-arrival tomography (Tryggvason Different 1D velocity functions were used for PSTM and
et al., 2002). Scatter plots (Fig. 5) show that this initial model its depth conversion. For migration, a velocity of 1400 m/s
(Fig. 6a) did not match the observed data well. However, was fixed down to 50 ms and then increasing linearly to
the root mean square (RMS) misfit decreased rapidly for 3000 m/s at 200 ms. For time-depth conversion, a function
based on the borehole lithology was used that differs some-
what from the migration velocity. The model was 1100 m/s
at 10 s, 1650 m/s at 40 ms, and 1800 m/s at 80 m. The final
PSTM depth-converted image is shown in Fig. 7(a).
For PSDM, we used for the shallower parts an initial
velocity model based on the resulting tomographic velocity
model given in Fig. 6(b). Velocities at greater time in the model
were based initially on the NMO velocities from the PSTM
processing. Different velocities were then tested to produce the
best image of the strong reflection observed between distances
of 100 and 300 m at around 0 m elevation (Fig. 7b). Then, fol-
lowing a layer-stripping approach (Al Yahya at greater time
in the model were based initially on the NMO velocities from
the PSTM processing. Different velocities were then tested
Figure 5 Travel time residuals comparison between the initial model to produce the best image of the strong reflection observed
based on the refraction statics solution and the sixth iteration of between distances of 100 and 300 m at around 0 m ele-
the inversion. The blue scatter points represent the RMS from the vation (Fig. 7b). Then, following a layer-stripping approach
initial velocity and the black points represent the RMS from the sixth
(Al Yahya, 1989), the migration velocity was updated. Trial
iteration of the inversion.


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 7

Figure 6 First-arrival tomography. (a) Ray coverage; (b) tomographic result after six iterations; (c) the same as model (b), but with contour lines
added.

velocities were tested below an elevation of 0 m to gener- poorer due to the reduced spatial sampling geometry over this
ate the best image of the undulating deeper reflection (see section (Fig. 2).
bedrock reflection (BR) in Fig. 8). In order to refine the ve- An overlay of the tomographic section on the PSDM im-
locity model, common image gathers (CIG) were subjected age (Fig. 8a) shows that the velocity contour of about 1400
to residual moveout (RMO) analysis to further flatten criti- m/s is very shallow, c. 2 m, along much of the profile. How-
cal reflections. The final PSDM image (Fig. 7b) has several ever, east of a distance of 400 m this contour deepens, as
enhancements compared with the PSTM image (Fig. 7a). Ar- mentioned earlier. The interpreted silt layer at about 9 m
rows in Fig. 7 emphasize the most prominent improvements depth stands out as a strong reflection in the PSDM image,
achieved by PSDM, in comparison to the PSTM processing. but has a tendency to deepen east of 400 m, in conjunction
Even though the PSDM image is improved comparatively with the 1400 m/s contour. This deepening in our tomogra-
to the PSTM image, there are still some sections within it phy results is consistent with the resistivity results obtained
which are problematic. For the distances of 0–50 m, the data in the same area by Ismail et al. (2011). They observed a
quality is poor, probably due to the fact that the explosive decrease in the resistivity values on the eastern half of the pro-
source with high energy was blasted in dry sandy till. Unsat- file in their inversion of electric currents flowing parallel and
urated and unconsolidated sediments are known to attenuate perpendicular to the structural strike (transverse electric (TE)
seismic energy significantly (Jefferson et al., 1998). Between and transverse magnetic (TM) inversion) and related this to
50 and 528 m shots were blasted in or close to the top of an increasing clay content in the region. Borehole data also
the saturated clay, resulting in better source–ground coupling support the thickening of the clay layer in the east. Conse-
than in the west. However, at 420–528 m, the image becomes quently, the deepening of the 1400 m/s velocity contour in the


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
8 R. Sun, A. Kaslilar and C. Juhlin

Figure 7 (a) PSTM depth-converted image with borehole information. (b) PSDM image with borehole information. Thin horizontal red short
lines on the boreholes represent thin hard silt layers. Blue arrows show the most prominent improvements of the PSDM image over the PSTM
image.

eastern part of our profile is supported by the borehole data For modelling purposes, a finite-difference acoustic ap-
and the resistivity results of Ismail et al. (2011). This suggests proach (Thorbecke and Draganov, 2011) was used. A Ricker
that there is a real change in subsurface conditions east of wavelet with a maximum frequency of 800 Hz and a dom-
400 m. The top of the cobbley gravel (CG) and the bedrock inant frequency of 240 Hz was used as the source function.
(BR) are also clearly delineated, with the cobbley gravel inter- Spatial sampling interval along both the x and z directions was
preted reflection matching well with BH2 data. Several pos- 0.25 m. Time sampling was 0.5 ms and the record length was
sible faults are marked along the interpreted bedrock surface set to 145 ms. Following the employed real acquisition geom-
(Fig. 8b). etry (Fig. 2a), the source depth was selected as 1.4 m and a
total of 134 shots along the profile were simulated with a shot
spacing of 2 m to the west and a spacing of 4 m to the east of
Synthetic Data Test
200 m, as shown in Fig. 2(a). Receivers were considered ver-
In order to verify that produce pre-stack depth (PSDM) does tical sensors on the surface and seismic waves were recorded
indeed improve the image of the Heby 1 profile, synthetic by 48 channels. Receiver coordinates were obtained from the
data were generated based on the interpreted section and a field data. Four selected synthetic shot gathers are shown in
generalized velocity model based on the processing (Fig. 9). In Fig. 10(a–d) with their locations marked along the profile in
terms of the forward velocity model definition, the correlation Fig. 10(e). Corresponding time-offset reflection events from
of borehole data with the PSDM image was used, including key interfaces are marked. A shot gather comparison between
clay1 and clay2 (different types of clay due to different sat- the real data and the synthetic data is shown in Fig. 11. The
urations). Given previous results (Juhlin et al., 2002; Ismail silt layer, modelled as only 0.25 m thick, can be seen in both
et al., 2011) as a reference, the bedrock surface undulates at the real and synthetic data. Also, the clay2-sand reflection
a depth between 40 and 80 m along the Heby 1 profile. is clear in both seismograms. Overall, the character of the


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 9

Figure 8 (a) Tomographic result overlain


on the PSDM image. (b) Interpretation of
the PSDM image based on the borehole in-
formation. Red, purple and green curves
represent the silt, cobbley gravel (CG) and
bedrock (BR) interfaces. Possible faults in
the bedrock surface are shown by blue lines.

synthetic data matches the real data reasonably well. Note Once the source gathers were generated, they were
that the back-scattered events starting at about 25 m offset in subjected to a similar processing flow as the field data
the synthetic data are due to changes in the surface topography (Table 3), but without any filtering. Velocity functions used
of the model at these locations. for post-stack time migration (PSTM) with depth conversion

Figure 9 Numerical model. (a) Velocity


model overlain onto the PSDM image in
Fig. 7(b) for comparison with field data.
Red, purple and green curves represent the
silt, cobbley gravel (CG) and bedrock (BR)
interfaces, respectively; (b) velocity model
with boreholes marked, velocity bars show
the velocities used in the modelling for the
corresponding rock materials. Note that
the upper 1.25 m is modelled as clay1,
the silt is modelled as being 0.25 m thick
and clay2 (1450 m/s) is present below
and above the silt layer. Bedrock surface
matches with the PSDM image interpreta-
tion.


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
10 R. Sun, A. Kaslilar and C. Juhlin

Figure 10 Four shot gathers located at the red stars in (e). (a) Shot 10, distance = 20 m; (b) shot 49, distance = 98 m; (c) shot 109, distance =
248 m; (d) shot 159, distance = 448 m. The uppermost dry sand in the western part of the profile was not included in the modelling. Introduction
of this sand layer generated seismograms that were difficult tointerpret. Therefore, shot 10 in the synthetic data differs significantly from shot
10 in the field data (Fig. 3a).

and PSDM are given in Table 4. Note that the uppermost both show relatively good agreement with the model, but the
unsaturated clay velocity is not used in the field data since undulating bedrock surface is better imaged with PSDM. The
this low velocity zone was compensated for in the refraction uppermost silt layer is also sharper and can be traced to shal-
statics. The resulting PSTM and PSDM images (Fig. 12b,c) lower levels.

Figure 11 Comparison of the shot gather at 376 m distance, (a) raw field data with refraction statics applied, (b) processed data, after refraction
statics correction and bandpass filter [100, 150, 400, and 600 Hz] (c) synthetic data. The zero offset times of clay2-silt, clay2-sand and bedrock
are about 0.015, 0.025 and 0.06 s, respectively.


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 11

Table 4 Velocity definition for field data and synthetic data. Plots in Fig. 13 are marked with Silt, cobbley gravel (CG) and bedrock (BR) depths
to compare various velocities with interpretation. Reflection coefficients for the synthetic modelling are based on the layer that the value is listed
for and the layer below

Field data

PSTM velocity profile PSDM velocity profile

Time (ms) Velocity (m/s) Time (ms) Time-depth V (m/s) Depth (m) Velocity (m/s)

0–35 1140 0–10 1100 0–8 1090


36–60 1445 11–40 1650 9–12 1400
61–108 1500 41–120 1800 13–40 1650
109–250 3000 41–120 1800

Synthetic data (Fig. 13e–h)

CIG: CDP = 225 m and CDP = 250 m

Thickness (m) Velocity (m/s) Density (kg/m3 ) Reflection coefficient (RC)

1.25 1090 1900 0.14


9 1450 1900 0.17
0.25 1700 2300 −0.17
14.75 1450 1900 0.07
20 1600 2000 0.13
35 2050 2200

A further verification that PSDM is more appropriate for Furthermore, it shows how PSDM improves the continuity of
this data set can be found in common image gathers (CIGs) sub-horizontal features and provides better imaging of dipping
as shown in Fig. 13. Where the structure is most complex, at interfaces. Velocities used for the pre-stack depth migration
distances of about 225–250 m, the PSDM CIGs show more were based on the first-arrival tomography and the normal
horizontal reflections and a sharper image, especially in the moveout velocities from the post-stack migration processing.
field data. Potentially, the image could be further improved by using ve-
For the synthetic data, the reflections also match bet- locity measurements in the boreholes, especially in BH2 below
ter in depth when the CIGs are compared with the model 10 ma.s.l. in the deep syncline structure (Fig. 8) where dip-
(Fig. 12). Note that it is not possible to plot CIGs for the ping interfaces are present. For the shallower more horizontal
PSTM depth-converted image since the data are migrated af- structures, borehole velocities would be useful to have, but it
ter stack; therefore, we show CIGs for the normal moveout is not likely the image would be improved significantly since
depth-converted data. These will inherently show a poorer the velocities used in the processing were chosen to optimize
match if dip is present. the image.
The thin silt layer at about 9 m depth has a thickness
of about 0.2 m along most of the profile. For a velocity of
DISCUSSION
1800 m/s, the tuning frequency ( 41 wavelength) for such a
This study focuses on how the use of tomography and pre- layer is 1600 Hz. This is far higher than the maximum use-
stack depth migration (PSDM) may improve imaging and the ful frequency recorded. At a frequency of 450 Hz, a 0.2 m
interpretation of reflection seismic data acquired over glacial thickness corresponds to 1/20 of the seismic wavelength. The
sediments. The Heby area is of particular importance since Heby data set provides an example of how layers much thin-
the deep sands and gravels are an important groundwater ner than the 14 wavelength resolution criteria (Widess, 1973)
reservoir. The new processing does not change the origi- can be detected. In fact, if it is assumed that it is the dominant
nal interpretation in Juhlin et al. (2002); however, it does frequency that determines the resolution or detection limit,
provide additional information concerning the near-surface then the silt layer is about 1/32 of the wavelength of the dom-
velocity structure and on potential faults in the basement. inant frequency. The reason we can map this layer at Heby is


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
12 R. Sun, A. Kaslilar and C. Juhlin

Figure 12 Synthetic images (a–c) overlain by the velocity model. (a) Image based on synthetic data after NMO, stack and depth conversion; (b)
PSTM image after depth conversion; (c) PSDM image.

probably that it is very laterally continuous with little change for the silt layer should be about 0.085. Such a layer should
in properties and that the data quality are excellent. We would generate reflections that are easily detectable on high quality
like to point out that it is important to differentiate between data.
resolution and detection when discussing the amplitudes of The potentiometric pressure in the groundwater reser-
reflected waves. Widess (1973) shows that the amplitude of voir (sands and gravels below 10 m elevation) was reported
the reflected wave from a thin layer that has a thickness 1/8 as corresponding to a pressure head of 35 m at the time of the
of the dominant seismic wavelength is the same as for a single survey. This is approximately the same elevation as the thin
interface. In principle, a layer with a thickness of 1/40 of the silt layer. It is therefore important to differentiate between the
seismic wavelength can be detected if ambient noise is very groundwater level and the reservoir head in the interpretation
low (reflection amplitude will be about 0.2 times the single (Sun et al., 2019). The best velocity model shows that satu-
interface amplitude). For the 1/20 wavelength criteria, then rated clay is present already at depths of 2–3 m, consistent
the reflection amplitude is expected to be about half of that with surface waters being present just east of the profile, and
of a single interface. Given that the single interface reflection implying that the surface groundwater table is at an elevation
coefficient is 0.17, then the thin layer reflection coefficient of 42–43 m in the area. The clay layers decouple the surface


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 13

Figure 13 Common image gathers from field data (a–d) at lateral distances of 225 m and 250 m. (a) CDP = 225 m, after NMO and time-depth
conversion; (b) CDP = 250 m, after NMO and time-depth conversion; (c) CDP = 225 m, after PSDM; (d) CDP = 250 m, after PSDM. Common
image gathers from synthetic data (e–h). (e) CDP = 225 m, after NMO and time-depth conversion; (f) CDP = 250 m, after NMO and time-depth
conversion; (g) CDP = 225 m, after PSDM; (h) CDP = 250 m, after PSDM. Red arrows mark silt, clay, cobbley gravel (CG) and bedrock (BR)
following the interpretation in Fig. 8(b).

waters from the deeper reservoir. Our study is an example of eastern Sweden. Our aim was to improve the seismic images
how several sources of information need to be combined to to better map groundwater-bearing structures and the top of
make a proper interpretation. the bedrock. Picked first arrivals on shot gathers for refrac-
Even though we see reflections that can be interpreted tion statics were used in first-arrival travel time tomography
as being generated at the top of the bedrock in the western to estimate the near-surface velocity model. Subsequently, the
part of the profile fired in the dry sandy till (Fig. 3a), it has tomographic results were used as a starting point in the pro-
not been possible to image these in the stacked sections, even cessing to produce pre-stack depth (PSDM) and post-stack
with PSDM. This may be due to the structure changing so time migration (PSTM) images.
rapidly in the lateral direction that it is not possible to stack The upper 20 m of overburden is represented by clay
signals coherently. Closer spacing between sources and re- and silt. These layers are underlain by sand/gravel deposits
ceivers may be necessary in this area to image the bedrock that overlie an undulating bedrock surface. Migrated sections
surface. show three clear reflections, one at about 35 m elevation that
is interpreted to represent a thin silt layer, a deeper one at
about 10 m elevation that is interpreted as the top of the
CONCLUSIONS
cobbley gravel, and a deeper one corresponding to the top
We reprocessed high-resolution shallow seismic reflection of the bedrock. The depths to all three are consistent with
data acquired over glacial deposits in the Heby area, south- borehole data along the profile. The surface water table is


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
14 R. Sun, A. Kaslilar and C. Juhlin

at an elevation of about 42–43 m along the profile based on Bruno, P.P.G. and Godio, A. (1997) Environmental risk assessment of
analysis of first breaks and surface water observations. a shallow aquifer in Piana Campana (Italy): a field comparison be-
tween seismic refraction and reflection methods. European Journal
We observe that PSDM improves the seismic image, es-
of Environmental and Engineering Geophysics, 2, 61–76.
pecially reflections from the silt layer and from dipping struc-
Buker, F., Green, A. and Horstmeyer, H. (1998) Shallow 3-D seis-
tures. These reflections became more prominent and continu- mic reflection survey: data acquisition and preliminary processing
ous with PSDM. Synthetic model tests provide support for strategies. Geophysics, 63, 1434–1450.
that PSDM is a meaningful alternative and that it is fea- Buker, F., Green, A.G. and Horstmeyer, H. (2000) 3-D high-
sible to apply it on the shallow aquifer structure in Heby. resolution reflection seismic imaging of unconsolidated glacial and
glaciolacustrine sediments. Geophysics, 65(1), 18–34.
High-resolution seismic reflection data and PSDM can con-
Burschil, T., Buness, H., Tanner, D.C., Wielandt-Schuster, U., Ell-
tribute to reliable and accurate imaging of groundwater re- wanger, D. and Gabriel, G. (2018) High-resolution reflection seis-
sources. A comprehensive link of tomography, PSDM and mics reveal the structure and the evolution of the Quaternary
seismic modelling is capable of improving images and in- glacial Tannwald basin. Near Surface Geophysics, 16(6), 593–
terpretations of reflection seismic data acquired over glacial 610.
Butler, J.J. Jr. (2005) Hydrogeological methods for estimation of
sediments.
spatial variations in hydraulic conductivity. In: Singh, V.P. (Ed.)
Hydrogeopyhsics (Water Science and Technology Library 50).
ACKNOWLEDGEMENTS Springer, pp. 23–58.
Dietrich, P. and Leven, C. (2006). Direct push technologies. In: Kirsch,
We sincerely thank the editors and the reviewers João Car- R., Groundwater Geophysics a Tool for Hydrogeology, 1st edition.
valho and Raymond Durrheim for their constructive com- Berlin: Springer, pp. 321–337.
Dix, C.H. (1955) Seismic velocities from surface measurements. Geo-
ments. RS thanks CSC (China Scholarship Council) for finan-
physics, 20, 68–86.
cial support. Data acquisition was funded by the Geological Geissler, P.E. (1989) Seismic profiling for groundwater studies in Vic-
Survey of Sweden (SGU). Research data are not shared. toria, Australia. Geophysics, 54, 31–37.
Ghose, R., Nijhof, V., Brouwer, J., Matubara, Y., Kaida, Y. and Taka-
hashi, T. (1998). Shallow to very shallow, high-resolution reflection
ORCID seismic using a portable vibrator system. Geophysics, 63(4), 1295–
1309.
Ruixue Sun https://orcid.org/0000-0002-5359-7985
Giustiniani, M., Accaino, F., Picotti, S. and Tinivella, U. (2008) Char-
acterization of the shallow aquifers by high-resolution seismic data.
Geophysical Prospecting, 56, 655–666.
REFERENCES
Grelle, G. and Guadagno, F.M. (2009) Seismic refraction methodol-
Al Yahya, K.M. (1989) Velocity analysis by iterative profile migra- ogy for groundwater level determination: “water seismic index”.
tion. Geophysics, 54, 718–729. Journal of Applied Geophysics, 68, 301–320.
Bachrach, R. and Nur, A. (1998) High-resolution shallow-seismic Hasselström, B. (1969) Water prospecting and rock-investigation by
experiments in sand, Part 1: Water table, fluid flow, and saturation. the seismic refraction method. Geoexploration, 7(2), 113–132.
Geophysics, 63(4), 1225–1233. Haeni, F.P. (1986) Application of seismic refraction methods in
Baker, G.S., Steeple D.W. and Drake M. (1998) Muting the noise groundwater modelling studies in New England. Geophysics, 51(2),
cone in near-surface reflection data: an example from southeastern 236–249.
Kansas. Geophysics, 63, 1332–1338. Hunter, J.A., Pullan, S.E., Burns, R.A., Gagne, R.M. and Good,
Birkelo, B.A., Steeples, D.W., Miller, R.D. and Sophocleous, M.A. R.S. (1984) Shallow seismic reflection mapping of the overburden
(1987) Seismic-reflection study of a shallow aquifer during a pump- bedrock interface with the engineering seismograph—some simple
ing test. Ground Water, 25, 703–709. techniques. Geophysics, 49, 1381–1385.
Bishop, T., Bube, P., Cutler, R., Langan, R., Love, P., Resnick, J., Ismail, N., Schwarz, G. and Pedersen, L. (2011) Investigation of
Shuey, R., Spindler, D. and Wyld, H. (1985) Tomographic determi- groundwater resources using controlled-source radio magnetotel-
nation of velocity and depth in laterally varying media. Geophysics, lurics (CSRMT) in glacial deposits in Heby, Sweden. Journal of
50(6), 903–923. Applied Geophysics, 73(1), 74–83.
Bradford, J.H. (2002) Depth characterization of shallow aquifers with Jefferson, R., Steeples, D., Black, R. and Carr, T. (1998) Effects of
seismic reflection, Part I, The failure of NMO velocity analysis and soil-moisture content on shallow-seismic data. Geophysics, 63(4),
quantitative error prediction. Geophysics, 67, 89–97. 1357–1362.
Bradford, J.H. and Sawyer, D.S. (2002) Depth characterization Juhlin C., Palm H., Müllern C.F. and Wållberg B. (2000) High-
of shallow aquifers with seismic reflection – Part II: Pre- resolution reflection seismics applied to detection of groundwater
stack depth migration and field examples. Geophysics, 67, 98– resources in glacial deposits, Sweden. Geophysical Research Let-
109. ters, 27, 1575–1578.


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15
Reprocessing of high-resolution seismic data 15

Juhlin, C., Palm, H., Müllern, C.F. and Wållberg, B. (2002) Imaging of Pugin, A., Pullan, S., Hunter, J. and Oldenborger, G. (2009) Hydro-
groundwater resources in glacial deposits using high-resolution re- geological prospecting using P-and S-wave landstreamer seismic
flection seismic, Sweden. Journal of Applied Geophysics, 51, 107– reflection methods. Near Surface Geophysics, 7, 315–327.
120. Schmelzbach, C., Green, A.G. and Horstmeyer, H. (2005) Ultra-
Lankston, R.W. (1990) High-resolution refraction seismic data ac- shallow seismic reflection imaging in a region characterized by high
quisition and interpretation. In: Ward, S.H. (Ed.) Geotechnical and source-generated noise. Near Surface Geophysics, 3, 33–46.
Environmental Geophysics. Society of Exploration Geophysicists, Sloan, D.S., Tsoflias, G.P. and Steeples, D.W. (2010) Ultra-shallow
pp. 45–74. seismic imaging of the top of the saturated zone. Geophysical Re-
Lanz, E., Maurer, H. and Green, A.G. (1998) Refraction tomography search Letters, 37(7), L07405.
over a buried waste disposal site. Geophysics, 63, 1414–1433. Steeples, D.W., Green, A.G., McEvilly, T.V., Miller, R.D., Doll,
Li, Z., Min, D.-J., Hwang, J. and Oh, J.-W. (2019) Ray-based reflec- W.E. and Rector, J.W. (1997) A workshop examination of shal-
tion traveltime tomography using approximate stationary points. low seismic reflection surveying. The Leading Edge, 16(11), 1641–
Near Surface Geophysics, 17(5), 463–477. 1647.
Liu, Y. and Wu, Z. (2018) Evaluation of traveltime tomography in es- Sun, R., Kaslilar, A. and Juhlin, C. (2019) Seismic imaging of shallow
timating near-surface velocity inversion. Near Surface Geophysics, groundwater resources in glacial deposits. Sweden. Near Surface
17(1), 85–97. Geoscience Conference & Exhibition 2019, 8–12 September 2019,
Maries, G., Ahokangas, E., Mäkinen, J., Pasanen, A. and Malehmir, The Hague, Netherlands.
A. (2016) Interlobate esker architecture and related hydrogeo- Thorbecke, J. and Draganov, D. (2011) Finite-difference modelling
logical features derived from a combination of high-resolution experiments for seismic interferometry. Geophysics, 76(6), H1–
reflection seismic and refraction tomography, Virttaankangas, H18.
southwest Finland. Hydrogeology Journal, 25, 829–845. Tryggvason, A., Rögnvaldsson, S. and Flovenz, O. (2002) Three-
Mayne, W.H. (1962) Common reflecting point horizontal data stack- dimensional imaging of the P- and S-wave velocity structure and
ing techniques. Geophysics, 27, 927–938. earthquake locations beneath Southwest Iceland. Geophysical Jour-
Morey, D. and G.T. Schuster. (1999) Paleoseismicity of the Oquirrh nal International, 151, 848–866.
fault, Utah from shallow seismic 2008, tomography. Geophysical Voyiadjis, G.Z. and Song, C.R. (2003) Determination of hydraulic
Journal International, 138, 25–35. conductivity using piezocone penetration test. International Journal
Müllern, C.F. (2008) Description to the map Groundwater bodies in of Geomechanics, 3, 217–224.
Heby municipality. Technical Report SGU 1652–8336: Geological Widess, M.B. (1973) How thin is a thin bed? Geophysics, 38, 1176–
Survey of Sweden, Uppsala, Sweden. 1180.
Perz, M., Bruins, C. and Heim, J. (2019) Shallow land PSDM ve- Yordkayhun, S. (2011) Detecting near-surface objects with seis-
locity model building for unconventional plays. SEG International mic traveltime tomography: experimentation at a test site. Songk-
Exposition and 89th Annual Meeting, pp. 4952–4956. lanakarin Journal of Science and Technology, 33(4), 477–485.


C 2020 The Authors. Near Surface Geophysics published by John Wiley & Sons Ltd on behalf of European Association
of Geoscientists and Engineers., Near Surface Geophysics, 1–15

You might also like