You are on page 1of 18

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/291262199

Integration of controlled-source and radio magnetotellurics electric


resistivity tomography, and reflection seismics to delineate 3D structures of a
quick-clay landslide site in s...

Research · January 2016


DOI: 10.13140/RG.2.1.1388.1367

CITATIONS READS

3 284

5 authors, including:

Chunling Shan Mehrdad Bastani


Institute of Geophysics and Geomatics Geological Survey of Sweden
15 PUBLICATIONS   50 CITATIONS    107 PUBLICATIONS   827 CITATIONS   

SEE PROFILE SEE PROFILE

Alireza Malehmir Lena Persson


Uppsala University Geological Survey of Sweden
271 PUBLICATIONS   2,105 CITATIONS    54 PUBLICATIONS   329 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

TRUST 4.2 Integrated use and interpretation of geophysical and non-geophysical data in preinvestigations for underground infrastructure facilities. View project

SEG-GWB project (2011-2013 and continuation): Integration of geophysical, hydrogeological and geotechnical methods to aid monitoring landslide in Nordic countries:
A 4D approach for landslide risk assessment View project

All content following this page was uploaded by Chunling Shan on 20 January 2016.

The user has requested enhancement of the downloaded file.


GEOPHYSICS, VOL. 81, NO. 1 (JANUARY-FEBRUARY 2016); P. B13–B29, 16 FIGS., 1 TABLE.
10.1190/GEO2014-0386.1
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Case Histroy

Integration of controlled-source and radio magnetotellurics,


electric resistivity tomography, and reflection seismics
to delineate 3D structures of a quick-clay landslide
site in southwest of Sweden

Chunling Shan1, Alireza Malehmir2, Mehrdad Bastani3, Lena Persson3, and Emil Lundberg2

ABSTRACT resistive crystalline bedrock underlying the marine conductive


clays and showed considerable correlations with the 3D reflection
Quick clay, which is the main cause of landslides that occur in seismic data in resolving a coarse-grained layer that was inter-
the northern countries, liquefies easily, and its presence implies an preted to act as a conduit directing freshwater into the clays under
increased risk of landslide. Geophysical methods have been in- a confined pressure, leaching their salt and forming quick clays.
creasingly used in landslide investigations. Three-dimensional The 3D CSRMT resistivity model and 3D reflection seismic data
electric resistivity tomography, radio magnetotelluric (RMT), showed that the coarse-grained layer has a varying thickness. At
controlled-source RMT (CSRMT), and high-resolution reflection some locations, it was too thin to be resolved by the methods used
seismic data were acquired at a quick-clay landslide site in the here. Combination of the CSRMT model, reflection seismic data,
southwest of Sweden. The main objectives were to evaluate and the borehole data suggested that a layer with thickness of
the capability of each method in delineating different subsurface approximately 5 m and resistivity between 20 and 30 Ωm was
geologic structures that controlled a peculiar and hazardous retro- potentially quick clay, which probably extended laterally in
gressive-type landslide in the study area. A 3D resistivity model the entire study area. These observations suggested that future
from the inversion of CSRMT data showed the best correlation developments should focus on joint inversion of such 3D data
with the reflection seismic data and borehole information, thanks sets incorporating sharp boundaries as constraints in the inversion
to the broad frequency range of the data set. It better imaged the and particularly when quick clays were studied.

INTRODUCTION cause 1–2 billion dollars damage and on average more than 25 fatal-
ities annually (USGS, 2013). The most recent landslide that occurred
A landslide is the movement of rock mass, sediments, or debris in Oso, Washington, during 22 March 2014 claimed 43 lives (Keaton
down a slope (Cruden, 1991), which is an important landscape-form- et al., 2014). Petley (2012) reports that 2620 fatal landslides were
ing process, providing the main mechanism for sediment release from recorded worldwide during a seven-year study period from 2004
slopes to permit rock and/or soil transportation through the fluvial to 2010, causing a total of 32,322 fatalities. Landslide hazard pro-
system (Petley, 2010). Landslides constitute a major part of geologic grams have been established in many countries (Eeckhaut and Her-
hazard because they can affect large areas in a short period of time. vas, 2012) to reduce long-term losses by improved understanding of
They have been reported for example in all 50 states in the USA and the causes of ground failure and suggesting mitigation strategies.

Manuscript received by the Editor 16 August 2014; revised manuscript received 7 September 2015; published online 12 January 2016.
1
China University of Geosciences, Hubei Subsurface Multi-scale Imaging Key Laboratory, Institute of Geophysics and Geomatics, Wuhan, China. E-mail:
chunling.shan@cug.edu.cn.
2
Uppsala University, Department of Earth Sciences, Uppsala, Sweden. E-mail: alireza.malehmir@geo.uu.se; emil.lundberg@geo.uu.se.
3
Geological Survey of Sweden, Uppsala, Sweden. E-mail: mehrdad.bastani@sgu.se; lena.persson@sgu.se.
© 2016 Society of Exploration Geophysicists. All rights reserved.

B13
B14 Shan et al.

Landslides also occur in Sweden and other northern countries. properties of the normal clay considerably. Bjerrum (1954) studies
The Geological Survey of Sweden (SGU) has reported nine known the differences in geotechnical properties of quick and non-quick
major landslides in Sweden during the past century. One of the clays, such as the remolded shear strength Sr , which is less than
largest landslides with no casualties occurred at Tuve (Figure 1) 0.4 kPa for quick clays (Table 1). The sensitivity of clays is defined
in November 1977, damaged seven electric cables, and destroyed as the ratio between undisturbed shear strength Su and the remolded
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

65 single-family houses (Nadim et al., 2008). Landslides in Sweden shear strength Sr . The sensitivity is greater than 50 for quick clays.
have a slightly different nature. What makes them different is a par- Low-, medium-, and high-sensitivity clays have sensitivities of <8,
ticular type of clay, the so-called quick clay, that is the main cause 8–30, and >30, respectively (Rankka et al., 2004). Shan et al.
for most landslides in Sweden. During the last glaciation (approx- (2014) also summarize the physical properties of quick clays in-
imately 10,000 years ago), the earth’s crust, in Scandinavia, was cluding their electric properties as well as their host soils (see also
compressed by the weight of a nearly 3 km thick ice sheet (Freden, Berger, 1980; Solberg, 2007). Table 1 shows the resistivity and sen-
1984). After the retreat of the ice sheet, large boulders of rocks, sitivity variations among different soil types. More information
gravels, silts, and clays transported by the glaciers to the surround- and physical properties of quick clays are introduced in Shan et al.
ing ocean under a marine condition were then deposited (Rankka (2014).
et al., 2004). The deposition took place at locations where the water Changes in the subsurface physical (and chemical) properties due
flow was rather low and clam (i.e., at great distances from the ice to a landslide are likely affecting the geophysical responses, which
front and mainly in sea and lake bays). Rankka et al. (2004) point in turn can not only be used to map the landslide body but also to
out that the formation of these sediments with different structures monitor its motion and physical or chemical changes (Jongmans
was highly dependent on the concentration of particles and the salt and Garambois, 2007). After the pioneering work of Bogoslovsky
content in the water, among other factors. The postglacial sediments and Ogilvy (1977), geophysical techniques have been increasingly
were further affected by later sedimentation of younger deposits and used in landslide investigations. Geophysical techniques, as pointed
other subsequent processes. A common characteristic of the clays in out by Mccann and Forster (1990) and Hack (2000), have the fol-
Sweden formed in this way is that they consist of aggregates con- lowing attributes: They (1) are flexible, fast, and deployable on
nected by links made of smaller particles. Due to the isostatic re- slopes; (2) provide information about the internal structure of the
bound, the landmass was (and is) uplifted, and the clays deposited soil or rock mass; and (3) allow a large volume to be investigated
in the marine conditions started to gradually be exposed above sea in a relatively short time. Moreover, reliable and portable geophysi-
level. The exposed marine clays can be found in Scandinavia, North cal equipment as well as advanced 2D and 3D geophysical imaging/
America, and northern parts of Russia (see Malehmir et al., 2013a; modeling techniques have dramatically attracted geologic engineers
Shan et al., 2014). During the postglacial period, fresh-/rainwater to use geophysical techniques for landslide investigations and mon-
resulted in the infiltration of water in the subsurface, which in turn itoring (e.g., Fell et al., 2000). An example of using multigeophys-
caused extensive leaching of salt from the highly porous, open ical methods is presented by Bichler et al. (2004) in which a 3D
structures with a high void ratio of clays. This process leads to model of the Quesnel Forks landslide in Canada was generated from
the formation of quick clays. These quick clays liquefy easily, different-type data sets including ground-penetrating radar, direct
and areas with a presence of quick clays have a higher risk of land- current resistivity (DCR), reflection, and refraction seismic. The
slides. The leaching process affects the physical and mechanical DCR data were effective for resolving stratigraphic relationships

Figure 1. Landslide hazard map (right side figure)


of Sweden and location of the study area near the
Göta River. A historical quick-clay landslide, the
Lilla Edet landslide, 1957, is approximately 5 km
south of the study area. Modified from Malehmir
et al. (2013b).
3D delineation of quick clays B15

between geologic units to a maximum depth of 40 m. The seismic man et al. (2003) apply a 3D MT inversion algorithm to a scalar
surveys obtained unit boundaries up to a depth of more than 80 m. RMT data set acquired over a buried waste site in Germany. They
Electric resistivity is one of the most popular geophysical meth- find that the applied 3D analysis could overcome previous limita-
ods used for near-surface investigations (Telford et al., 1990; Reyn- tions in the RMT method using 2D data analysis and inversion. Bas-
olds, 1997). By providing 2D or 3D models of electric resistivity, it tani et al. (2012) carry out 2D and 3D inversions of full tensor RMT
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

has widely been used in landslide studies, such as for delineation of to study an aquifer underlying a contaminated site at Trecate in
slip-surface geometry and depth, bedrock depth, and various geo- Italy. They find that the resistivity models from 3D inversion of
logic boundaries (Havenith et al., 2000; Batayneh and Al-Diabat, RMT data modeled the depth to the top and thickness of the
2002; Demoulin et al., 2003; Wisén et al., 2003; Bichler et al., water-saturated zone more accurately than do the models from
2004; Lapenna et al., 2005; Lebourg et al., 2005; Meric et al., the 2D independent inversions.
2005; Dahlin et al., 2013). Malehmir et al. (2013a) present a preliminary assessment of po-
Electromagnetic (EM) methods are also used for landslide inves- tential utility of various geophysical measurements carried out in a
tigations to determine the geometric boundaries of an unstable mass quick-clay landslide area in Sweden (Figure 2, labeled lines 2–5),
(Jongmans and Garambois, 2007). Controlled-source tensor magne- which is the same area of this study. Their results showed that geo-
totelluric (CSTMT, 1–12 kHz) and radio magnetotelluric (RMT, physical data are capable of delineating major subsurface structures
14–250 kHz) methods are two EM methods that have not been associated with quick-clay landslides. Malehmir et al. (2013b)
widely used for landslide investigations. The CSTMT method is present high-resolution reflection seismic data from the same four
proposed by Li and Pedersen (1991). They prove that the imped- 2D profiles crossing the quick-clay landslide scar (Figure 2, lines 2–
ance and a tipper tensor could uniquely be defined for any earth 5). They find that a layer containing coarse-grained materials ap-
structure and any source and receiver positions; these tensors are pears to be spatially linked to the presence of quick clays. Shan et al.
independent of the orientation and strength of the electric or mag- (2014) present the results of 2D inversion of electric resistivity
netic dipole sources. We combine the data sets from the RMT and tomography (ERT) and RMT data along the same profiles shown
CSTMT methods, and we call the combined method the CSRMT in Malehmir et al. (2013b) and make an integrated interpretation of
method. The CSRMT method has been successfully applied for the collected data together with the reflection seismic and borehole
groundwater exploration and other types of near-surface studies data. They find that the integration of ERT and RMT data with re-
(e.g., Pedersen et al., 2005; Bastani et al., 2009, 2012; Kalscheuer flection seismic data is superior for quick-clay landslide studies than
et al., 2013). In most cases, the data were collected along single or their individual implementation. They also underline that a 2D
sparsely situated profiles, and as a result, the resistivity models de- assumption may not be accurate enough to explain the subsurface
rived from the RMT and CSTMT data were only modeled in 2D geology. Lundberg et al. (2014) study the 3D seismic data collected
even though the geologic evidences suggested 3D structures. A in the same study area (Figure 2), and they also note that the 3D
landslide is a 3D phenomenon; 2D models of 3D structures may seismic data better resolved the smaller scale structures than the 2D
be highly biased and difficult to interpret (Wisén et al., 2005). Be- seismic data; two sets of coarse-grained layers, one at approxi-
cause of increasing computational performance, algorithms of 3D mately 20 m depth and another one located deeper, were found.
magnetotelluric (MT) inversion have now been developed that The bedrock topography was clearly imaged. The 2D traveltime
make 3D modeling and interpretation of MT data possible (e.g., and full-waveform tomography data along the seismic profiles pre-
Newman and Alumbaugh, 2000; Zhdanov et al., 2000; Sasaki, sented by Adamczyk et al. (2013, 2014) provided complementary
2001; Farquharson et al., 2002; Siripunvaraporn et al., 2005). New- results to the reflection seismic data from the study area. Salas-Ro-

Table 1. Resistivity classifications of different soils in a quick-clay area from Berger (1980) and Solberg (2007). Modified from
Shan et al. (2014).

Sensitivity Remolded shear strength


Classification of soil material from resistivity (Berger, 1980) (Lundström et al., 2009) (Rankka et al., 2004)

Soil material Resistivity interval kPa


Clay, salt 1 − 20 Ωm <15 >2
Leached, possible quick clay 20 − 90 Ωm >50 <0.4
Clay, dry crust 70 − 300 Ωm <8 >2
Silt, wet 50 − 200 Ωm <15 >2
Sand, saturated 200 − 1000 Ωm <15 >2
Classification of soil material from resistivity (Solberg, 2007)
Soil material Resistivity interval
Salt/intact marine clay 1 − 10 Ωm <15 >2
Leached, possible quick clay 10 − 80 Ωm >50 <0.4
Dry crust clay, slide deposits, coarser >80 Ωm <8 >2
B16 Shan et al.

mero et al. (2015) use surface geophysical, core sample, and down- due to the isostatic adjustments. The most common sediments along
hole property to identify landslide preconditions in the same area. the west coast are marine clays (often silty with different proportions),
As part of a detailed study over the landslide scar (Figure 2), we sand, and gravel (Lindström et al., 2000). The sediments originally
collected 3D ERT, RMT, and CSRMT data over the same 3D reflec- deposited in a marine environment are now exposed approximately
tion seismic area as presented by Lundberg et al. (2014). The moti- 20–30 m above sea level in the study area. Some of the clays in
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

vation of incorporating CSTMT measurements in this study was to the GötaÄlv valley are classified as quick clay, which are mainly close
improve the EM signal strength and to obtain deeper information not to the surface and are on top of or below permeable layers.
resolved just by using the RMT method. We present the results from The Swedish Geotechnical Institute (SGI) in a nationwide project
3D inversions of these data sets with the main objective of evaluating carried out several geotechnical investigations in the study area in-
the performance of each method in delineating the geometry and cluding cone penetration tests (CPTs) of various types (e.g., CPT,
physical properties of subsurface structures that are crucial to under- piezo-CPT, and resistivity CPT) and laboratory (mainly geotechni-
stand the formation of quick clays and their role in occurrence of cal) measurements (Löfroth et al., 2011). The collected data (the
landslides. The resistivity models are compared with the migrated locations are shown in Figure 2) suggested the presence of
reflection seismic sections extracted from the 3D seismic volume coarse-grained materials (often sandy) at varying depths and thick-
to improve the interpretation of the geometry of the different layers nesses. The SGI’s measurements showed a clear increase in the fric-
and their physical properties. These models are also compared with tional resistance and electric resistivity where the coarse-grained
the geotechnical data for a better understanding of the geologic struc- materials were intersected (Löfroth et al., 2011). Quick clays (or
tures that give rise to retrogressive landslides in the area as well as for highly sensitive clays) were found above the coarse-grained mate-
checking the validity of the models. rials and were not found in boreholes toward the west. Based on the
results from the processing of high-resolution reflection seismic
STUDY AREA data, Malehmir et al. (2013b) show that the coarse-grained materials
deepen from the east toward the west along an east–west-directed
The study area is located in the Göta älv (or Göta River) valley profile, a profile that crosses the 3D survey area (Figure 2, line 4),
(Figures 1 and 2). The valley stretches between Lake Vänern in further supporting the reason why these quick clays were missing
the north to Gothenburg in the south. The bedrock in this area is do- toward the west of the study area.
minated by crystalline gneissic, partly banded granitoids aged at ap-
proximately 1.1–1.8 Ga, partly intruded by other rock types (Freden,
2002). The Göta älv valley is intersected at several locations by the GEOPHYSICAL DATA ACQUISITION AND
Göta älv shear zone, a tectonic zone stretching from Lake Vänern to QUALITY
Kungsbacka. The depth to the bedrock in the Göta älv and in our study ERT data
area varies from exposed to more than 80 m deep (Dahlin et al., 2001).
The postglacial sediments are dominant at the surface in the western The ERT data were collected along 11 profiles (Figure 2) using a
part of Sweden. They are comprised of clay, silt, and sand to gravel. As multielectrode system, ABEM Terrameter LS, and a Wenner elec-
the ice retreated from the west coast approximately 10,000– trode configuration with a minimum dipole length of 5 m. In the
14,000 years ago, the compressed crust gradually started to uplift, ERT method, direct current voltage is maintained to a pair of elec-

Figure 2 Surface geologic map of the study area


(from SGU) showing the locations of the geo-
physical lines (black lines, surveyed during
2011, Malehmir et al., 2013a, 2013b) and avail-
able geotechnical boreholes (black dots). The dark
red circle shows the location of the CSTMT
source. The black square marks the boundaries
of the 3D survey area (this study) that covers an
approximately 40-year-old landslide scar. The
landslide scar within the 3D survey area is shown
by the closed blue region. The inset map shows the
data acquisition setup; black lines are the reflec-
tion seismic and ERT profiles, green dots are
RMT stations, and green dots with red squares
are the CSRMT stations. Note that the CSRMT
data were acquired along every second RMT pro-
file. The profiles are numbered as P1, P2, etc.
3D delineation of quick clays B17

trodes (current electrodes), and as a result, an electric current flows most of the receiver stations and with each shot location repeated
into the ground. At the same time, the potential difference between up to five times using an accelerated weight-drop source. In some
two other points is measured using another pair of electrodes (po- locations in which the weight drop was not possible for logistical
tential electrodes). The apparent resistivity of the ground is then reasons (i.e., the northern margins of the scar close to the river), a
estimated, which is a function of the applied voltage, the injected sledgehammer was used instead. Standard 3D CMP stacking was
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

current, and the electrode configuration (Daily et al., 2005). The applied using a bin size of 2 m in the inline direction and 10 m in the
ERT data are of good quality along all the profiles. crossline direction. Strong coherent noise remained in the inline
data after stacking was removed using an f-k filter. The maximum
RMT and CSRMT data frequency (containing data) was approximately 140 Hz in the final
stack and with a main frequency of approximately 100 Hz. Details
In RMT and CSTMT methods, three components of the magnetic of the 3D seismic survey and processing results can be found in
field (H x ; H y ; Hz ) and two horizontal components of the electric field Lundberg et al. (2014).
(Ex ; Ey ) of the EM signals are recorded simultaneously. Using the
measured components, the EM transfer functions that contain infor- Near-field effect of CSTMT data
mation about the variation in the electric resistivity of the ground can
be estimated (Chave and Jones, 2012) at each frequency. In the fre- To illustrate the quality of the acquired CSRMT data, we present
quency domain, the impedance tensor Z is defined as the measured apparent resistivities, phases from an example profile,
     and induction arrows from real parts of T(A, B) (Parkinson, 1962;
Ex Zxx Zxy Hx Schmucker, 1964) in Figures 3 and 4, respectively. The resistivity
¼ ; (1)
Ey Zyx Zyy Hy and phases are shown as XY and YX modes, where X is used for the
magnetic north direction and Y is used for magnetic east direction.
and the magnetic transfer function or tipper T(A, B) is defined as Prior to modeling, we have to make sure that the plane-wave
  assumption (Chave and Jones, 2012) that is used in the modeling
Hx programs is valid. That is to say that the CSTMT data are not biased
H z ¼ TðA; BÞT ; (2)
Hy by the near-field effect. Li and Pedersen (1991) study the near-field
effects on the collected data generated from various types of nearby
where x and y denote the magnetic north and east, respectively: Z, the EM sources (transmitters). Near-field effects are different depend-
complex impedance tensor, relates the two horizontal magnetic field ing on the type of source used. For the magnetic dipole source used
components to the two horizontal electric field components via the in this study, near-field effects on the apparent resistivity would ap-
earth’s electric resistivity structure. When an EM wave travels within pear as sudden drops of resistivities and on the apparent phase as
a conductive medium, the energy is dissipated as heat, and the wave sudden increases in phases (Li and Pedersen, 1991). For electric
amplitude is attenuated faster than when traveling in a resistive dipole source, the apparent resistivity and phase behave oppositely
medium. The penetration depth of an EM wave is inversely related (Li and Pedersen, 1991; Zonge and Hughes, 1991). The near-field
to its frequency and the medium’s electric conductivity (Bazinet and effects increase with decreasing frequencies. Moreover, the real part
Legault, 1985). of the induction arrow points toward the source/transmitter location
In the RMT method, the sources of the signal are the distant very (Li and Pedersen, 1991).
low frequency and low-frequency transmitters in the frequency The RMT signals are considered as plane waves, and the first data
range 10–250 kHz. CSTMT makes use of a remotely controlled quality control to study is the continuity of the data in the transition
source that is either a grounded double-electric or double-magnetic from RMT to CSTMT frequencies. The resistivities and phases in
dipole. We used the latter (see Figure 2 for the CSTMT source lo- both directions at the transition frequencies seem to be stable, and
cation), and generated source frequencies of 2, 4, 8, and 12.5 kHz. no abrupt changes are observed (Figure 3e and 3f). As mentioned
Hereafter, we use the term CSTMT when we present the results above, the near-field bias is stronger and closer to the EM sources,
from the combination of both methods. The RMT and CSTMT data and in the case of a magnetic dipole source, it corresponds to sudden
were acquired by the EnviroMT system from Uppsala University decrease in resistivity values and sudden increase in phase values
(for details about the instrument and CSTMT source, see Bastani closer to the source location or at the lower frequencies. In the ap-
et al., 2009) in the 3D study area. The RMT data were collected parent resistivities of CSRMT data (Figure 3a and 3b), the southern
over 11 profiles with station spacing of 10 m and profile spacing part of the profile shows decreasing resistivities and the phases (Fig-
of 20 m (Figure 2). Profile number 1 is located in the westernmost ure 3c and 3d) become slightly higher. However, as is shown in
side of the 3D study area. Totally, 259 RMT data points were col- Figure 3f, the phase at the lowest CSTMT frequency changes its
lected along the 11 profiles. A total of 35 CSRMT stations with trend and decreases. We can therefore conclude that the dropping
spacing of 40 m were measured along every second RMT profile resistivities and rising phases in the southern part of the profile
(Figure 2). The data acquisition took two days for the RMT and one might be due to the conductive layers at depth instead of effects
day for the CSRMT. from the near-field source. Furthermore, the induction arrows (Fig-
ure 4) for the lowest frequency (2 kHz) do not indicate any regular
Reflection seismic data pattern caused by the source. The arrow from the closest profile to
the source does not have the largest amplitude. The arrows closer to
The 3D seismic data were acquired using a SERCEL 428 record- the source do not point toward the source implying that the near
ing system. Single 28 Hz vertical component geophones were used field even at lower frequencies does not affect the CSTMT data,
with a spacing of 4 m in the inline direction and with receiver line and the variations observed in the data are related to the underlying
spacing of 20 m. Shots were recorded along the receiver lines at geology.
B18 Shan et al.

3D INVERSION ALGORITHMS in data space, see Constable et al. (1987), de Groot-Hedlin and Con-
stable (1990), Parker (1994), and Siripunvaraporn and Egbert
ERT (2000). Data space inversion significantly reduces the computa-
The inversion program RES3DINV (Loke et al., 1996) was used tional costs in memory and CPU time, and it is therefore practical
to invert the acquired ERT data in 3D. Topography information is to run it on ordinary PCs and workstations (Siripunvaraporn
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

included as part of the inversion process. The program divides the et al., 2005).
subsurface into many small rectangular blocks and attempts to de- The WSINV3DMT uses all the elements of the impedance
termine the resistivity values of the prisms by minimizing the differ- tensor Z (equation 1) including the real and imaginary parts. When
ence between the calculated and observed apparent resistivity applying 3D inversion, it is no longer necessary to meet any dimen-
values in an iterative manner. The inversion routine used by the pro- sionality assumptions, and small-scale inhomogeneities can be ac-
gram is based on the smoothness-constrained least-squares method counted for in the model by proper discretization (Siripunvaraporn
(de Groot-Hedlin and Constable, 1990; Sasaki, 1992; Loke et al., et al., 2005).
1996). We use two nodes between adjacent electrodes for the inver- One important point to emphasize here is that the inversion code
sion. This corresponds to 2.5 m along the profiles and 10 m assumes a quasistatic approximation (see Nabighian, 1991) to cal-
perpendicular to the profiles. The minimum root-mean square culate the model responses. Because it is stated by several studies
(rms) value of 1.9% is reached after 10 iterations. (Huang and Fraser, 2001, 2002; Hatch et al., 2013; Kalscheuer et al.,
2013) at higher frequencies (>100 kHz) where the ground resistiv-
RMT and CSRMT ity is considerably high (>5000 Ωm), the displacement currents
have a large impact on the measured field components, which
The 3D inversions of the RMT and the CSRMT data were carried has to be taken into account in the inversion process. In this study,
out using WSINV3DMT (Siripunvaraporn et al., 2005), which is a the shallower parts of the subsurface are mainly postglacial sedi-
full 3D inversion program for MT data or other data that are as- ments that are composed of conductive marine clays or silts that
sumed to be in the far field. The program is extended and imple- have resistivities of less than 100 Ωm. The data at the highest
mented from the 2D data space Occam’s inversion (Siripunvaraporn RMT frequencies are used to model the shallowest parts of the sub-
and Egbert, 2000). The inversion algorithm seeks the smoothest surface, which implies that compared to the conduction currents, the
minimum structure model subject to an appropriate predefined fit displacement currents are low enough to be neglected in the inver-
to the data. The inversion scheme is based on a data space approach, sion and the quasistatic approximation is valid for the RMT and
in which matrix dimensions depend on the size of data set (N), CSRMT data sets.
rather than the number of model parameters (M). Generally, N ≪ In the 3D RMT and CSRMT inversion, the model blocks had
M for the MT data. For further details on the Occam-type inversion horizontal dimensions (x and y) of 10 m and a thickness that log-

Figure 3. Measured CSRMT data along profile 7


(see Figure 2). (a) Apparent resistivity, XY mode,
(b) apparent resistivity, YX mode, (c) phase, XY
mode, and (d) phase, YX mode. (e) Resistivity ver-
sus period plot of the station closest to the source
from the same profile. The dashed line shows the
transition from the RMT to the controlled-source
frequency band. (f) Phase versus period plot of the
same station.
3D delineation of quick clays B19

arithmically increased with depth. Every second, a data point was weighting the data points with unrealistic low error values. Data
taken from the 259 RMT points along all the profiles to reduce the points with unrealistic large error bars were excluded before carry-
calculation time and CPU memory demanding for the RMT data. In ing out the inversion. We tested inversions for RMT and CSRMT
total, 130 stations were inverted for the RMT data and 35 stations data with a homogeneous half-space as initial models. We inverted
for the CSRMT data and nine frequencies for both data sets. An one data set from RMT data (because the whole data set was divided
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

error floor of 2.5% was applied on the impedances to avoid over- into two data sets, each data set consists of data points from every
second station along profiles in the north–south), and we used the
final model as initial model in the inversion of the other data sets to
save time. We also carried out a test to study the effect of the re-
sistivity values in the starting model. It turned out that the a priori
model had more influence on the final model than the starting model
because the final model tries to deviate less from the a priori model
during the inversion process. We finally chose an a priori model that
is a homogeneous half-space with resistivity of 100 Ωm for RMT
and CSRMT inversions. It took approximately one full day to finish
eight iterations. The smallest overall rms reached after three itera-
tions for RMT (1.474) data and seven iterations for CSRMT (1.30)
data. The inversions were carried out assuming a flat topography,
and real topography was introduced to the final models afterward.
We anticipate some geometric artifacts from this inclusion and take
this into account when interpreting the results.

RESULTS
ERT data
A layered structure is observed in the 3D ERT resistivity model
(Figure 5). The differences between the observed data and the for-
ward response from inversion along profile 1 (south–north) are
shown in Figure 6 suggesting a good data fit. To study the resolution
of the models, we present model sections in two perpendicular di-
rections (Figure 7), along the profiles (north–south) and across the
profiles (east–west). The data coverage is mainly semi-3D than in a
real full 3D case (Loke and Dahlin, 2010) because we have coarser
Figure 4. Induction arrows from the real part of tippers for trans- sampling in the east–west direction than the north–south direction
mitter frequency of 2 kHz (CSRMT data). The magnetic source lo- and also the electric current runs mainly in the north–south direction
cation is indicated at approximately 250 m from the beginning of and the variations in the perpendicular direction (east–west) are less
profile 7.
pronounced in the collected data. The model is shown down to 60 m

Figure 5. The 3D ERT resistivity model with


topography (Lidar data) included during inversion
process. The origin is at the center of the model in
horizontal plane with coordinate (0,0). Minus co-
ordinates in the east–west and north–south direc-
tions are to the west or south from the origin, and
positive coordinates are to the east or north from
the origin. The region marked by the black line
marks the location of the landslide scar (Figure 2).
Transparent triangles cover the model blocks esti-
mated from no data coverage due to the electrode
configuration.
B20 Shan et al.

Figure 6. Difference between measured ERT data


and forward response shown as log resistivity
along a south–north-running section in the west-
ernmost part of the study area. White areas
represent no data coverage. Note the big difference
in data coverage and density from electrode spac-
ing of 20 m and higher than that.
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure 7. Resistivity sections extracted from the 3D inversion model of ERT data: (a) the east–west-running sections and (b) north–south
sections. (c) The locations of the sections in panels (a and b) in the 3D area. The central point (0, 0) is in the middle of the 3D model. The black
arrow in panel (c) points to an example location of the north–south section at 60 m. The landslide scar is clearly visible in some of the sections
(e.g., the high-resistivity feature on the east–west section 60 m from 25 m toward the east and the south–north section 60 m from 50 m toward
the north). Evidence of resistive bedrock is only evident in the north–south sections at 20 and 60 m. Translucent triangles cover the model
blocks estimated from extrapolations for no data coverage in the ERT data acquisition.
3D delineation of quick clays B21

below sea level because there are no considerable resistivity varia- the thickness of this layer varies, but due to the limited penetration
tions observed below this elevation. For the purpose of presenting depth, the bottom of this layer is not resolved by the 3D inversion.
more details on the structural variations, a few north–south sections However, the more resistive feature in the southeast corner of the 3D
were extracted from the 3D ERT model (Figure 7). Shan et al. model may be an indication of the resistive crystalline bedrock.
(2014) interpret the geologic structures in their 2D resistivity mod-
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

els based on the resistivity classification suggested by Berger (1980) RMT data
and Solberg (2007) as shown in Table 1. We use the same classi-
fication scheme to interpret the 3D resistivity model, the topmost Figure 8 shows the 3D resistivity RMT model. To show the data
resistive layer (often oxidized and weathered clay) represents the fit, the observed data and the forward response of all the impedance
resistivity between 300 and 1000 Ωm, with a thickness of 2 to elements at a given station (station 1 in the south from profile 1) are
3 m. The second layer has a relatively low resistivity of approxi- shown in Figure 9. Figure 10 shows a selection of resistivity
mately 20 Ωm and thickness of 5 m, which is interpreted to be sections from the 3D model in east–west and north–south direc-
a mixture of marine and quick clays. The third layer represents tions. The forward responses fit reasonably well to the observed
coarser-grained sediments mainly composed of sand and gravel (ap- data, which is almost valid for most of the stations used in the in-
proximately 100 Ωm, 10 m thick). A very low resistivity zone/layer version. All the sections generally display a five-layer structure. In
of approximately 1–10 Ωm represents marine clays. It seems that the sections shown in Figure 10a, we observe a resistive top

Figure 8. A 3D inversion model from the RMT


data. Elevation data were added to the final inver-
sion model afterward and were not included
during the inversion process. The final rms is
1.47. The region marked by the black line marks
the location of the landslide scar (see Figure 2).

Figure 9. Impedance comparisons of observed


data and forward response from 3D inversion of
one example RMT station: (a) Zxx, (b) Zxy,
(c) Zyx, and (d) Zyy. The plus symbols and circles
are the real and imaginary parts of each imped-
ance, and the solid and dashed lines are the corre-
sponding forward response.
B22 Shan et al.

thin layer (100–1000 Ωm) overlying a relatively conductive layer nal, it is difficult to judge if it can be an indication of the resistive
(20–30 Ωm), which is followed by a more resistive layer crystalline bedrock.
(30–300 Ωm). The latter overlies a very conductive (less than
10 Ωm) and considerably thicker layer. The deepest layer is rela- CSRMT data
tively more resistive (approximately 100 Ωm) than the overlying
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

conductive layer, which may correspond to the bedrock. The bed- The 3D resistivity model from the inversion of CSRMT data is
rock has an undulating shape in the south and becomes more flat shown in Figure 11. Figure 12 shows the data fit between the ob-
toward the north (see the east–west sections at 60 and 100 m). The served impedance tensor elements and calculated responses from
deepest resistive feature has a more undulating top surface (see Fig- the 3D inversion. The off-diagonal elements Zxy and Zyx are much
ure 10b). However, due to the limited penetration of the RMT sig- larger than the diagonal elements Zxx and Zyy. The error bars of the

Figure 10. Resistivity sections from 3D RMT


model. (a) East–west and (b) north–south sections.
The numbers on top of the sections are the loca-
tions as described in Figure 7. The landslide scar is
clearly visible in some of the sections. The resis-
tive bedrock is more evident than those from the
ERT model (Figure 7).
3D delineation of quick clays B23

observed data are small, implying good-quality data. The 3D tive layer becomes more resistive toward the north especially at the
CSRMT model, similar to the RMT model, also mainly pictures landslide scar area. The second layer is more conductive than the
five different layers with changing resistivities and depths. In the top layer and the underlying layer below. At some parts of the pro-
results from 2D inversion of RMT data along profile 5 (Figure 2) file, this layer is not distinguishable from the top and bottom layers
presented by Shan et al. (2014), the resistivity models show a re- (i.e., in the north–south section −60 m, at 0 and at 60 m along the
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

sistive feature at depth (probably the bedrock) that deepens toward section). The third layer, which is likely the coarse-grained layer,
the Göta River. The same resistive feature at depth is also observed becomes more resistive in the northeastern part of the study area
here in the north–south sections at −20 and 20 m (Figure 13b). The (i.e., in the north–south section −60 m, from 0 to 50 m along
top of the resistive feature is approximately 40 m and becomes even the section). The conductive layer becomes gradually thicker as
deeper toward the north. The top surface of the resistive feature the bedrock deepens toward the north (see the east–west sections
(probably bedrock) is much shallower in the south and deepens to- in Figure 13a). In the southernmost part of the section at
ward the river in the north. The same feature also occurs at a shal- −100 m, the shallowest top surface of the deepest resistor/bedrock
lower level in the northeastern corner close to the boundary of the imaged by the model is approximately −40 m elevation and has an
3D study area where the surface elevation is higher. The top resis- undulating shape. The depth to the bedrock is still uncertain in this

Figure 11. The resistivity model from 3D inver-


sion of CSRMT data projected on high-resolution
topography data. The region marked by the black
line marks the location of the landslide scar (Fig-
ure 2).

Figure 12. Impedance comparisons of observed


data and forward response from 3D inversion of
one example CSRMT station: (a) Zxx, (b) Zxy,
(c) Zyx, and (d) Zyy. The plus marks and the
circles are the real and imaginary parts of each
impedance, and the solid and dashed lines are
the corresponding forward response.
B24 Shan et al.

model even though lower frequency CSTMT data are incorporated terials such as silt or dry clay may have resistivities in the same
in the inversion. It is, therefore, vital to use other independent in- range (depending on their pore content), we have used the available
formation such as those from the reflection seismic survey and geo- geotechnical data to validate and refine our interpretations. In the
technical data to verify the results from the 3D inversions. previous studies conducted by SGI (Löfroth et al., 2011), quick
clays have often been associated with an underlying layer made
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Comparison with CPT data up of coarse grains. Two of the boreholes (7202 and 7203) are lo-
cated next to one of south–north profile 6 in the middle of the 3D
Leached marine clays have electric resistivities ranging between survey area. Variations in the measured CPT-R (CPT with resistivity
20 and 90 Ωm (Berger, 1980; Solberg, 2007). Because other ma- measurements) inside different boreholes are presented by Mal-

Figure 13. Resistivity sections extracted from the


3D CSRMT model. (a) East–west running sections
and (b) north–south running sections. The loca-
tions are shown on top of each subplot.
3D delineation of quick clays B25

ehmir et al. (2013b). The observed depth and thicknesses of the ducted by Adamczyk et al. (2013). The origin of the S2 reflection
coarse-grained and quick-clay layers inside the two boreholes is uncertain, although it can likely be attributed to a similar coarse-
are superimposed on the closest resistivity sections from the grained layer as for S1. The S1, S2, and B1 reflections can be traced
ERT, RMT, and CSRMT models and are shown in Figure 14. across most of the 3D volume. Some reflections occur in between
The resistivity variations in the second (from 15 to 20 m eleva- the main horizons in some inlines (e.g., between S1 and S2 in Fig-
tion) and third layers (from −5 to 15 m elevation) shown in the three
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

ure 15a), but these cannot be traced across the volume to the same
sections (Figure 14) correlate relatively well with those of the extent as the main picked horizons. These reflections are interpreted
coarse-grained and quick-clay layers measured in the geotechnical also as similar coarse-grained layers, but less extended, although it
borehole. This correlation is somewhat better for RMT and CSRMT could also represent reflections from the bottom or top boundaries
data than for the ERT data (Figure 14). From the geotechnical in- of the coarse-grained layer in places where the layers are thicker.
formation, at borehole 7202 (Figure 2), the coarse-grained layer With a main frequency of 100 Hz and a velocity of approximately
starts at 11 m depth with 1 m thick quick clay on top. At borehole 2000 m∕s in the sediments (Adamczyk et al., 2013), the seismic
7203, the coarse-grained layer starts at 21 m depth with 1.5 m thick wavelength is approximately 20 m. This gives a vertical resolution
quick clay on top. The resistivities of the second layer in the three of between 5 and 10 m for the seismic data. Because the coarse-
sections generally vary from 20 to 30 Ωm and approximately grained layer thicknesses are expected to be within this range most
200 Ωm for the third layer. But the two layers observed from
the three sections are somewhat thicker than those determined in
the two boreholes. The thickness of the coarse-grained layers from
the borehole data is unknown due to the limited drilling depth and
technical difficulties of these types of geotechnical soundings.
However, there is a good indication in the resistivity models from
all the inversions that a very conductive layer (probably intact post-
glacial marine clays) is located below the coarse-grained layer at
elevation of approximately 20 m below the surface.

Comparison with the reflection seismic data


The resistivity overlap of quick clay with other types of materials
cannot provide clear boundaries, whereas reflection seismic meth-
ods can provide some information about the geometry of the con-
tacts’ internal structures within the normally consolidated materials
(Jongmans and Garambois, 2007). The conductive marine clays
(clays with high salt content either as solid or salt water in the pores)
in the area strongly attenuate the EM signals and thus limit their
penetration depth. So we compare the 3D CSRMT resistivity model
with the reflection seismic data collected along the same profiles
(Lundberg et al., 2014). Reflections at approximately 0.5 s (corre-
sponding to a maximum depth of 750 m) could be detected in the
raw shot gathers, indicating that the signal was strong enough to
penetrate the sediments and down to the bedrock contact. The
processing was, however, geared toward the upper 0.2 s and aimed
at attaining a high-resolution 3D seismic volume of the layering and
the bedrock topography. Glacial sediments exhibit velocities on the
range of 1300–1800 m∕s, but generally, they are approximately
1500 m∕s at our site (Adamczyk et al., 2013, 2014; Salas-Romero
et al., 2015). Thus, a 0.2-s two-way traveltime corresponds to a
depth of approximately 150 m at the site. Three sections are selected
from the 3D seismic volume and compared with sections extracted
from the 3D CSRMT resistivity model along the same lines (Fig-
ure 15). The south–north-directed sections −80 m are selected from
the western part of the study area, and two sections at 20 and 40 m
are selected from the central part of the area. These two sections (20
and 40 m) cross the landslide scars in their northern part. The upper
Figure 14. North–south sections extracted from the 3D resistivity
reflection (S1) is interpreted to be from the coarse-grained layer in- models. (a) ERT, (b) RMT, and (c) CSRMT sections at 0 m in the
tersected by several boreholes in the area. The bedrock reflection middle of the 3D study area close to the geotechnical boreholes. The
(B1) was interpreted after correlation with a longer 2D seismic pro- purple and black boxes represent the interpreted coarse-grained
file (using a dynamite source) on which the reflection could be layer and quick clay from the CPT measurements (Löfroth et al.,
2011). The thickness of the coarse-grained layer is unknown at
traced to the near surface further south, close to where outcrops the two boreholes due to the limited drill depth. The models, how-
are visible (Malehmir et al., 2013b). The interpretation was also ever, indicate a thickness on the order of 10 m. Interpretations for
aided by the results of the 2D traveltime tomography study con- different layers are shown in panel (c).
B26 Shan et al.

often, the seismic sections show these layers as a single reflection various sections (Figures 7, 10, and 13) have not been able to image
(Lundberg et al., 2014). the boundaries of the shallowest layer (the dry crust), which is
mainly because of the frequency content of the signal. Instead,
DISCUSSION the seismic data are superior in finding the sharper layer boundaries
and they demonstrate the best correlation with the borehole infor-
The results presented in this study indicate that there are certain
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

mation. CSRMT method is superior to the RMT method because it


features that are commonly observed in almost all of the models
uses higher and lower frequencies. In the CSRMT model (Fig-
from the methods used in this study, including the following: layers
ures 11 and 13), better resolution at depth is observed because it
closer to the surface that include the dry soil/sediments at the sur-
better delineates the coarse-grained layer and the bedrock as higher
face, which is mostly more resistive than the underlying layer; quick
resistivity features. According to Berger (1980), sand (or saturated
clay; a coarse-grained layer; and marine clay that is very conduc-
sand) in a quick-clay area can show resistivities between
tive. As is evident from Figures 5, 8, and 11, all the resistivity sec-
tions generally show the same second layer with resistivity from 200 and 1000 Ωm. The fact that the coarse-grained materials in
20 − 30 Ωm. Among these methods due to a limitation with data our study area appear as resistive is an indication of the flow of
coverage (usually at two ends of one profile), the 3D ERT inversion fresh ground water through them. These materials then act as a con-
models cannot image the structures close to the two ends of the lines duit (Malehmir et al., 2013a, 2013b) directing freshwater from the
(Figure 6). This is more problematic where one end of the line south where the bedrock is closer to the surface (Figure 16) toward
reaches an obstacle, as is the case in our study area of the Göta the Göta River. The continuous or nearly continuous flow of fresh-
River. The 3D ERT models lose their resolutions below a certain water then can leach the salt from the clays above and leads to the
depth especially at the boundaries of the 3D study area (Figures 5 formation of quick clays above the coarse-grained layer. An excess
and 7). The 3D RMT model (Figure 8) is better at delineating the of rain and meltwater in the layer would then increase the pore water
resistivity variations than the ERT model and reveals sharper var- pressure and destabilize the system and as a result triggers a land-
iations at the layer boundaries especially from a conductive to a slide (combined with other natural and human activities such as ero-
resistive layer at depth. In Figure 10, a conductive layer with resis- sion of the river embankments or excavation). A layer with
tivities of less than 10 Ωm is observed in all of the sections. The resistivity of 20–30 Ωm between 5 to 10 m elevation is well re-
CSRMT model shows similar structures at shallower parts of the solved in the RMT and ERT models. The borehole data indicate
model down to the elevation of −40 m; however, due to the coarser that this layer contains quick clays (Figure 14).
spacing between the profiles and stations, it shows smother lateral The layering structures imaged by the CSRMT model show good
resistivity variations. The ERT, RMT, and CSRMT data as shown in correlations with the interpreted horizons from the reflection seis-

Figure 15. (a, c, e) Selected sections extracted


from the 3D migrated seismic volume and (b, d,
f) their corresponding resistivity sections from
the 3D CSRMT model. S1 and S2 reflections
are interpreted to be generated from coarse-grained
materials, and B1 is interpreted to be a bedrock re-
flection (Lundberg et al., 2014). The interpreted
coarse-grained and bedrock horizons are marked
with red and pink arrows, respectively, in panels
(b, d, and f). The depth to the coarse-grained layer
from the geotechnical boreholes (Löfroth et al.,
2011) is indicated by red (on seismic sections)
and white vertical lines (on the resistivity sections).
Note that the coarse-grained layer is modeled to be
resistive, which may be an indication of freshwater
flow in the layer (Malehmir et al., 2013b).
3D delineation of quick clays B27

mic data (Figure 15), and they support the use of integrated EM and at approximately −25 m along the south–north direction (without
seismic methods for quick-clay landslide studies. These two meth- the topography incorporated). The depth variations in the bedrock
ods are often not routinely used for these purposes. Figure 16 shows modeled by the CSRMT data are well supported by the seismic
the iso-resistivity surfaces of the 3D CSRMT model with four re- horizons (Figures 15 and 16b); the elevation of B1 (Figure 16b)
sistivity values and elevations of the B1 and S1 horizons. Our in- shows that the shallowest top surface of the bedrock is in the south
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

terpretation of each surface is labeled with a text in the figure. As and deepens toward the north. In the northeast (Figure 16b), the
pointed out by Lundberg et al. (2014), the S1 horizon (Figure 16c), elevation of the bedrock is approximately −65 m, which correlates
which represents the top surface of the coarse-grained layer, be- with the CSRMT sections with the topography incorporated in Fig-
comes shallower with a northeast direction toward the river. We ure 13a (section 100 m).
have found the same tendency in the resistivity model (e.g., section
20 m in Figure 10a) in which the depth to the top of the third layer
CONCLUSIONS
(with a resistivity of approximately 25 Ωm) is smaller in the east
with a northeast direction. We also observed discontinuities of this The collected 3D ERT, RMT, CSRMT, and reflection seismic data
layer in the middle of the 3D study area (Figure 16a), which cor- sets have reasonably high quality to be used for delineating detailed
relates well with the discontinuities in the S1 horizon (Figure 16c). subsurface structures that control the formation of quick clays and
An example discontinuity can be observed at approximately 0 m in their associated landslides. The 3D resistivity models resolved four
the eastmost part of the model (Figure 16a). The red line in Fig- relatively distinct layers on top of the resistive crystalline bedrock.
ure 16a outlines the top surface of the bedrock. The surface of The 3D resistivity model from the inversion of CSRMT data better
the bedrock reaches its deepest point in the easternmost direction delineates the bedrock underlying the conductive marine clays.

Figure 16. (a) Isosurfaces of the 3D CSRMT


model without considering the topography. Four
resistivities are chosen to image the approximate
boundaries of structures, and they represent very
conductive marine clays (5 Ωm), quick clays
(25 Ωm), coarse grains (80 Ωm), and bedrock
(300 Ωm). (b) Elevation of the B1 bedrock reflec-
tion (also shown in Figure 15) from Lundberg et al.
(2014). (c) Elevation of the S1 reflection horizon.
The region marked by the black line marks the lo-
cation of the main landslide scar (study area).
Black dots in panels (b and c) represent where
the seismic horizons could be picked. The red line
in panel (c) indicates discontinuities observed in
seismic horizon S1.
B28 Shan et al.

The coarse-grained layer with a resistivity of 30–300 Ωm is iden- Berger, B., 1980, Rapport om bertikaleelectriskesonderinger I Verdal: Uni-
versity of Trondheim (UNIT), Norwegian Institute of Technology (NTH).
tified as the permeable layer that plays an important role in the oc- Bichler, A., P. Bobrowsky, M. Bestdouma, J. Hunter, T. Calvert, and R.
currence of landslides in the study area because it acts as a conduit Burns, 2004, Three-dimensional mapping of a landslide using a multi-
directing freshwater into the clays above, leaching their salt and geophysical approach: The Quesnel Forks landslide: Landslides, 1,
29–40, doi: 10.1007/s10346-003-0008-7.
forming quick clays. The 3D resistivity models from CSRMT data Bjerrum, L., 1954, Geotechnical properties of Norwegian marine clays:
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

and 3D reflection seismic data reveal that this layer has a varying Geotechnique, 4, 49–69, doi: 10.1680/geot.1954.4.2.49.
Bogoslovsky, V. A., and A. A. Ogilvy, 1977, Geophysical methods for the in-
thickness and at some location is very thin so that it cannot be re- vestigation of landslides: Geophysics, 42, 562–571, doi: 10.1190/1.1440727.
solved by the methods used. This can be partly caused by the Chave, A. D., and A. G. Jones, 2012, The magnetotelluric method: Theory
smoothing regularization used in the inversions and partly by the and practice: Cambridge University Press.
Constable, C. S., R. L. Parker, and C. G. Constable, 1987, Occam’s inver-
diffusing nature of the ERT, RMT, and CSRMT methods, which sion: A practical algorithm for generating smooth models from electro-
does not allow resolving the sharp layer boundaries and thus accu- magnetic sounding data: Geophysics, 52, 289–300, doi: 10.1190/1
rate estimation of the depth to the top and thickness of the layers. To .1442303.
Cruden, D. M., 1991, A simple definition of a landslide: Bulletin of the
overcome this problem, reflection seismic data and geotechnical International Association of Engineering Geology, 43, 27–29, doi: 10
data collected in the nearby boreholes were used instead to make .1007/BF02590167.
Dahlin, T., R. Larsson, V. Leroux, M. Svensson, and R. Wisén, 2001, Geo-
a more realistic use of the resistivity models and to gain a better fysik isläntstabilitetsutredningar Rapport 62: Statens Geotekniska Institut.
understanding about the physical properties and geometry of the Dahlin, T., H. Löfroth, D. Schälin, and P. Suer, 2013, Mapping of quick clay
structures close and within the landslide area. using geoelectrical imaging and CPTU-resistivity: Near Surface Geophys-
ics, 111, 659–670, doi: 10.3997/1873-0604.2013044.
Future studies should focus on joint 3D inversion of the ERT, Daily, W., A. Ramirez, A. Binley, and D. LaBrecque, 2005, Electrical re-
RMT, or CSTMT with seismic data to take the advantages of each sistance tomography — Theory and practice: Near Surface Geophysics,
individual method and to minimize their limitations. We suggest 13, 525–550, doi: 10.1190/1.9781560801719.
De Groot-Hedlin, C., and S. C. Constable, 1990, Occam’s inversion to gen-
development of an algorithm in which a joint 3D inversion of erate smooth, two dimensional models from magnetotelluric data: Geo-
ERT, RMT, and CSRMT data using constraints from the reflection physics, 55, 1613–1624, doi: 10.1190/1.1442813.
seismic and borehole data can be performed. Demoulin, A., A. Pissart, and C. Schroeder, 2003, On the origin of
late Quaternary palaeo landslides in the Liège (E Belgium) area:
International Journal of Earth Sciences, 92, 795–805, doi: 10.1007/
s00531-003-0354-7.
ACKNOWLEDGMENTS Eeckhaut, V. D. M., and J. Hervas, 2012, State of the art of national landslide
databases in Europe and their potential for assessing landslide susceptibil-
This work was initiated at Uppsala University while C. Shan was ity, hazard and risk: Geomorphology, 139–140, 545–558, doi: 10.1016/j
studying there. This study is part of a joint research collaboration .geomorph.2011.12.006.
Farquharson, C. G., D. W. Oldenburg, E. Haber, and R. Shekhtman, 2002,
between Uppsala University and SGU. SEG, through its Geoscient- An algorithm for the three-dimensional inversion of magnetotelluric data:
ists Without Borders program, and SGU, through its internal re- 72nd Annual International Meeting, SEG, Expanded Abstracts, 649–652.
search program, have supported this work. H. Löfroth and Fell, R., O. Hunger, S. Leroueil, and W. Iemer, 2000, Geotechnical engineer-
ing of the stability of natural slopes and cuts and fills in soil: Presented at
several other personnel from SGI provided critical information that the International Conference on Geotechnical and Geology Engineering,
helped us to design our survey and interpret our data. The geophysi- paper no. ISRM-IS-2000-002.
cal data acquisition would have not been possible without the par- Freden, C., 1984, Beskrivning till jordartskartan Vänersborg SO: Sveriges
Geologiska Undersökning 48.
ticipation of Ph.D. and undergraduate students from Uppsala Freden, C., 2002, Berg och jord: Sveriges Nationalatlas.
University, Leibniz Institute for Applied Geophysics, University Hack, R., 2000, Geophysics for slope stability: Surveys in Geophysics, 21,
423–448, doi: 10.1023/A:1006797126800.
of Cologne, Syiah Kuala University, Polish Academy of Sciences, Hatch, M., G. Heinson, T. Munday, S. Theil, K. Lawrie, J. D. A. Clarke, and
and the SEG Student Chapter of Uppsala University. Critical and P. Mill, 2013, The importance of including conductivity and dielectric
constructive comments from three anonymous reviewers helped permittivity information when processing low-frequency GPR and
high-frequency EMI data sets: Journal of Applied Geophysics, 96, 77–
to improve the quality of the paper, for which we are grateful. 86, doi: 10.1016/j.jappgeo.2013.06.007.
Havenith, H. B., D. Jongmans, K. Abdrakmatov, P. Trefois, D. Delvaux, and
A. Torgoev, 2000, Geophysical investigations of seismically induced
REFERENCES surface effects: Case study of a landslide in the Suusamyr valley,
Kyrgyzstan: Surveys in Geophysics, 21, 349–369, doi: 10.1023/A:
Adamczyk, A., M. Malinowski, and A. Malehmir, 2013, Application of first- 1006788808145.
arrival tomography to characterize a quick clay landslide site in Southwest Huang, H., and D. Fraser, 2001, Mapping of the resistivity, susceptibility and
Sweden: Acta Geophysica, 61, 1057–1073, doi: 10.2478/s11600-013- permittivity of the earth using a helicopter-borne electromagnetic system:
0136-y. Geophysics, 66, 148–157, doi: 10.1190/1.1444889.
Adamczyk, A., M. Malinowski, and A. Malehmir, 2014, High-resolution Huang, H., and D. Fraser, 2002, Dielectric permittivity and resistivity map-
near-surface velocity model building using full-waveform inversion — ping using high-frequency helicopter-borne EM data: Geophysics, 67,
A case study from southwest Sweden: Geophysical Journal International, 727–738, doi: 10.1190/1.1484515.
197, 1693–1704, doi: 10.1093/gji/ggu070. Jongmans, D., and S. Garamois, 2007, Geophysical investigation of land-
Bastani, M., J. Hübert, T. Kalscheuer, L. B. Pedersen, A. Godio, and J. Ber- slides: A review: Bulletin of the Geological Society of France, 178,
nard, 2012, 2D joint inversion of RMT and ERT data versus individual 3D 101–112, doi: 10.2113/gssgfbull.178.2.101.
inversion of full tensor RMT data: An example from Trecate site in Italy: Kalscheuer, T., M. Bastani, S. Donohue, L. Persson, A. A. Pfaffhuber, F.
Geophysics, 77, no. 4, WB233–WB243, doi: 10.1190/geo2011-0525.1. Reiser, and Z. Ren, 2013, Delineation of a quick clay zone at Smorgrav,
Bastani, M., A. Malehmir, N. Ismail, L. B. Pedersen, and F. Hedjazi, 2009, Norway, with electromagnetic methods under geotechnical constraints:
Delineating hydrothermal stockwork copper deposits using controlled- Journal of Applied Geophysics, 92, 121–136, doi: 10.1016/j.jappgeo
source and radio-magnetotelluric methods: A case study from northeast .2013.02.006.
Iran: Geophysics, 74, no. 5, B167–B181, doi: 10.1190/1.3174394. Keaton, J. R., J. Wartman, S. Anderson, J. Benoit, J. deLaChapelle, R. Gil-
Batayneh, A. T., and A. A. Diabat, 2002, Application of a two-dimensional bert, and D. R. Montgomery, 2014, The 22 March 2014 Oso landslide,
electrical tomography technique for investigating landslides along the Snohomish County, Washington: Geotechnical Extreme Events Recon-
Amman-Dead Sera Highway, Jordan: Environmental Geology, 42, naissance (GEER) report.
399–403, doi: 10.1007/s00254-002-0543-x. Lapenna, V., P. Lorenzo, A. Perrone, S. Piscitelli, E. Rizzo, and F. Sdao,
Bazinet, R., and J. Legault, 1985, Scalar audio-magnetotellurics: A tool in 2005, 2D electrical resistivity imaging of some complex landslides in Lu-
the evaluation of nuclear waste disposal sites: 55th Annual International canian Apennine chain, southern Italy: Geophysics, 70, no. 3, B11–B18,
Meeting, SEG, Expanded Abstracts, 149–150. doi: 10.1190/1.1926571.
3D delineation of quick clays B29

Lebourg, T., S. Binet, E. Tric, H. Jomard, and S. Bedoui, 2005, Geophysical Petley, D. N., 2010, Landslide hazards, in I. Alcántara-Ayala, and A. S. Gou-
survey to estimate the 3D sliding surface and the 4D evolution of the water die, eds., Geomorphological hazards and disaster prevention: Cambridge
pressure on part of a deep seated landslide: Terra Nova, 17, 399–406, doi: University Press, 63–73.
10.1111/j.1365-3121.2005.00623.x. Petley, D. N., 2012, Global patterns of loss of life from landslides: Geology,
Li, X., and L. B. Pedersen, 1991, Controlled source tensor magnetotelluric: 40, 927–930, doi: 10.1130/G33217.1.
Geophysics, 56, 1456–1461, doi: 10.1190/1.1443165. Rankka, K., Y. Andersson-Sköld, C. Hultén, R. Larsson, V. Leroux, and T.
Lindström, M., J. Lundqvist, and T. Lundqvist, 2000, Sveriges geologi från Dahlin, 2004, Quick-clay in Sweden: Swedish Geotechnical Institute,
Downloaded 01/19/16 to 202.114.202.219. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

urtid till nutid 2nd ed.: Lund studentlitteratur. Technical report number 65, 137.
Löfroth, H., P. Suer, T. Dahlin, V. Leroux, and D. Schälin, 2011, Quick-clay Reynolds, J. M., 1997, An introduction to applied and environmental geo-
mapping by resistivity-surface resistivity, CPTU-R and chemistry to com- physics: John Wiley & Sons.
plement other geotechnical sounding and sampling: Swedish Geotechni- Salas-Romero, S., A. Malehmir, I. Snowball, B. C. Lougheed, and M.
cal Institute, report GÄU 30. Hellqvist, 2015, Identifying landslide preconditions in Swedish quick
Loke, M. H., and R. D. Barker, 1996, Practical techniques for 3D resistivity clays — Insights from integration of surface geophysical, core sample-
surveys and data inversion: Geophysical Prospecting, 44, 499–523, doi: and downhole property measurements: Landslides, 1–19, doi: 10.1007/
10.1111/j.1365-2478.1996.tb00162.x. s10346-015-0633-y.
Loke, M. H., and T. Dahlin, 2010, Methods to reduce banding effects in 3-D Sasaki, Y., 1992, Resolution of resistivity tomography inferred from numeri-
resistivity inversion: Presented at the 16th European Meeting of Environ- cal simulation: Geophysical Prospecting, 40, 453–463, doi: 10.1111/j
mental and Engineering Geophysics. .1365-2478.1992.tb00536.x.
Lundberg, E., A. Malehmir, C. Juhlin, M. Bastani, and A. Andersson, 2014, Sasaki, Y., 2001, Full 3-D inversion of electrical and electromagnetic data on
High-resolution 3D reflection seismic investigation over a quick-clay PC: Journal of Applied Geophysics, 46, 45–54, doi: 10.1016/S0926-9851
landslide scar in southwest Sweden: Geophysics, 79, no. 2, B97– (00)00038-0.
B107, doi: 10.1190/geo2013-0225.1. Schmucker, U., 1964, Anomalies of geomagnetic variations in the
Lundström, K., R. Larsson, and T. Dahlin, 2009, Mapping of quick clay southwestern United States: Journal of Geomagnetism and Geoelectricity,
formations using geotechnical and geophysical methods: Landslides, 6, 15, 193–221, doi: 10.5636/jgg.15.193.
1–15, doi: 10.1007/s10346-009-0144-9. Shan, C., M. Bastani, A. Malehmir, L. Persson, and M. Engdahl, 2014, In-
Malehmir, A., M. Bastani, C. M. Krawczyk, M. Gurk, N. Ismail, U. Polom, tegrated 2D modeling and interpretation of geophysical and geotechnical
and L. Persson, 2013a, Geophysical assessment and geotechnical inves- data to delineate quick clays at a landslide site in southwest Sweden: Geo-
tigation of quick-clay landslides — A Swedish case study: Near Surface physics, 79, no. 4, EN61–EN75, doi: 10.1190/geo2013-0201.1.
Geophysics, 11, 341–350, doi: 10.3997/1873-0604.2013010. Siripunvaraporn, W., and G. Egbert, 2000, An efficient data-subspace inver-
Malehmir, A., M. U. Saleem, and M. Bastani, 2013b, High-resolution re- sion method for 2D magnetotelluric data: Geophysics, 65, 791–803, doi:
flection seismic investigations of quick-clay and associated formations 10.1190/1.1444778.
at a landslide scar in southwest Sweden: Journal of Applied Geophysics, Siripunvaraporn, W., G. Egbert, Y. Lenbury, and M. Uyeshima, 2005, Three-
92, 84–102, doi: 10.1016/j.jappgeo.2013.02.013. dimensional magnetotelluric inversion: Data-space method: Physics of the
Mccann, D. M., and A. Forster, 1990, Reconnaissance geophysical methods Earth and Planetary Interiors, 150, 3–14, doi: 10.1016/j.pepi.2004.08.023.
in landslide investigations: Engineering Geology, 29, 59–78, doi: 10 Solberg, I., 2007, Geological, geomorphologic and geophysical investiga-
.1016/0013-7952(90)90082-C. tions of areas prone to clay slides: Examples from Buvika, Mid Norway:
Meric, O., S. Garambois, D. Jongmans, M. Wathelet, J. Chatelain, and J. M. Ph.D. thesis, Norwegian University.
Vengeon, 2005, Application of geophysical methods for the investigation Telford, W. M., L. P. Geldart, R. E. Sherif, and D. A. Keys, 1990, Applied
of the large gravitational mass movement of Séchilienne, France: Cana- geophysics: Cambridge University Press.
dian Geotechnical Journal, 42, 1105–1115, doi: 10.1139/t05-034. USGS, 2013, About us, http://landslides.usgs.gov/aboutus/, accessed 30 No-
Nabighian, M. N., 1991, Electromagnetic methods in applied geophysics: vember 2015.
SEG. Wisén, R., E. Auken, and T. Dahlin, 2005, Combination of 1D laterally con-
Nadim, F., S. A. S. Pedersen, P. Schmidt-Thomé, F. Sigmundsson, and M. strained inversion and 2D smooth inversion of resistivity data with a priori
Engdahl, 2008, Natural hazards in Nordic countries: Episodes, 31, 176– data from boreholes: Near Surface Geophysics, 3, 71–79, doi: 10.3997/
184. 1873-0604.2005002.
Newman, G. A., and D. L. Alumbaugh, 2000, Three-dimensional magneto- Wisén, R., A. V. Christiansen, E. Auken, and T. Dahlin, 2003, Application of
telluric inversion using non-linear conjugate gradients: Geophysics Jour- 2D laterally constrained inversion and 2D smooth inversion of CVES re-
nal International, 140, 410–424, doi: 10.1046/j.1365-246x.2000.00007.x. sistivity data in a slope stability investigation: Proceedings of the 9th
Newman, G. A., S. Recher, B. Tezkan, and F. M. Neubauer, 2003, 3D in- Meeting Environmental Engineering Geophysics, EAGE, Extended Ab-
version of a scalar radio magnetotelluric field data set: Geophysics, 68, stracts, paper no. O-002.
791–802, doi: 10.1190/1.1581032. Zhdanov, M. S., S. Fang, and G. Hursin, 2000, Electromagnetic inversion
Parker, R. L., 1994, Geophysical inverse theory: Princeton University Press. using quasi-linear approximation: Geophysics, 65, 1501–1513, doi: 10
Parkinson, W. D., 1962, The influence of continents and oceans on geomag- .1190/1.1444839.
netic variations: Geophysical Journal of the Royal Astronomical Society, Zonge, K. L., and L. J. Hughes, 1991, Controlled source audio-frequency
6, 441–449, doi: 10.1111/j.1365-246X.1962.tb02992.x. magnetotellurics, in M. N. Nabighian, ed., Electromagnetic methods in
Pedersen, L. B., M. Bastani, and L. Dynesius, 2005, Ground water explora- applied geophysics: SEG, 713–810.
tion using combined controlled source and radio magnetotelluric tech-
niques: Geophysics, 70, no. 1, G8–G15, doi: 10.1190/1.1852774.

View publication stats

You might also like