2D Reflection Seismic Imaging at a Quick-clay Landslide Site, in Southwest Sweden

(1)

i

Examensarbete vid Institutionen för geovetenskaper ISSN 1650-6553 Nr 249

2D Reflection Seismic Imaging at a Quick-clay Landslide Site, in

Southwest Sweden

Muhammad Umar Saleem

Supervisor: Alireza Malehmir

(2)

i

(3)

i

Abstract

Results from a series of 2D high resolution reflection seismic profiles collected at a quick-clay landslide site are presented. This study is a part of the “Integration of geophysical, hydrogeological and geotechnical methods to aid monitoring landslide in Nordic countries”

project, sponsored by the (Geoscientists Without Borders) program of Society of Exploration Geophysicists (SEG). The study area is located on the shoreline of the Göta River, about 60 km north of Göteborg. The Göta River is the largest river in Sweden which runs from Lake Vänern to Göteborg, it follows the Götaälv Zone, which is an approximately 4 km wide fault zone dipping towards the west. The site is known for its quick-clay formation and landslides. The aim of this study is to image bedrock topography and the overburden layerings above it, within the overburden layers of specific interest are a coarse-grained layer and a quick-clay layer which is responsible for “quick-clay landslides”. The area was recently studied by the Swedish Geotechnical Institutes in a nation-wide project that dealt with investigating areas along the Göta River that are prone to landslides. The study area was investigated by various geotechnical (CPTU, CPTU-R soundings, laboratory measurements) and electrical resistivity investigations.

Results from this investigation are used to interpret shallow seismic reflections and better understand the near surface geology. The seismic reflection data presented in this thesis were acquired along two profiles, 484 m and 384 m long (lines 2 and 3), in September 2011. A weight- drop and sledgehammer source was used to generate the seismic signal. A receiver spacing of 4 m and source spacing of 4-8 m was used. Power spectras of various sources and raw shots (before and after vertical stack) are also discussed. Processing of high resolution data was challenging in this area where the bedrock is very shallow. Therefore, stacked sections were very sensitive to stacking velocity; a great effort was made to obtain the velocity model. Pre- stack spectral whitening and band pass filtering after stack were the key steps for successful imaging. After carefully processing the data, we were able to image very shallow reflections and bedrock topography. A coarse-grained layer interpreted at 25-50 m depth, may be playing an important role in the formation of quick-clay and hence provide the triggering mechanism for landslides. The coarse layer potential to form quick-clays and its role in landslides, however, requires further investigation using other types of hydrogeological data, geotechnical data and geophysical well logging.

(4)

ii

Acknowledgements

First and foremost, my gratitude to my supervisor Dr. Alireza Malehmir for his guidance and endless support through my graduate dissertation at Uppsala University. Without his guidance, motivation and encouragement it was impossible to remain sustainable in the crucial situation. I had full freedom to work and come up with new ideas during my project. He provided the seismic data in SEGY format for processing and technical insight for Claritas software.

I am also thankful to my friend Jawwad Ashraf for his guidance and the reflection seismic group for their moral support. I gratefully acknowledge Daniel Sopher for his comments. I would like to thank people who constructed social welfare society of Sweden, which provided me free education and health care.

At the end but not the least, I would like to express my love and gratitude to my beloved family for their support and endless love, through whole duration of my studies.

GLOBE Claritas (Seismic Processing Software) under the license from the Institute of Geological and Nuclear Sciences limited, Lower Hutt, New Zealand was used to process the seismic data.

(5)

iii

“I do not know what I may appear to the world, but to myself I seem to have been only like a boy playing on the sea-shore, and diverting myself in now and then finding a smoother pebble or a prettier shell than ordinary, whilst the great ocean of truth lay all undiscovered before me”

“It is the perfection of God's works that they are all done with the greatest simplicity. He is the God of order and not of confusion”

(Sir Isaac Newton)

Dedicated To

All the people who have been a positive influence in my life

(6)

iv

(7)

v

List of Figures

Figure 1.1: Map of landslide risks in Sweden (provided by the Geological Survey of Sweden) showing the location of the study area and recent quick-clay landslides in southwest Sweden (e.g., Surte, 1950, Göta, 1957, 445 Tuve, 1977, and Småröd, 2006). Figure is modified from Malehmir and Bastani (2012)...……….……….2 Figure 1.2: Locations of the previous geotechnical boreholes, resistivity profiles (modified from Löfroth et al., 2011) and our seismic reflection lines 2 and 3. Seismic data along lines 2 and 3 are the focus of this study (also see Fig. 2.1)……….………..……….………...4 Figure 2.1: Lidar (light detection and ranging) elevation map overlaid on airphoto from the study area, showing also the location of the seismic lines (2 and 3), available geotechnical boreholes, and landslide scar. Lidar data and airphoto are provided by Lantmäteriet (2011). Figure is modified from Malehmir et al. (2012)...…...8 Figure 2.2:(a) and (b) show comparisons between power spectras before and after vertical stacked shots. (c) and (d) show comparisons between sledgehammer and weight-drop before and after vertical stack. The raw shot power spectras show higher frequency content than the vertical stacked shots. Note that an average power spectra containing more than ten shots is shown……….…….12 Figure 2.3: Field photos showing various types of seismic sources used in the project. (a) Weight- drop, (b) dynamite and (c) sledgehammer……….……….14 Figure 2.4: Comparisons of sledgehammer, weight-drop and dynamite sources used along different lines. The dynamite source produces the highest frequencies and best signal-to-noise ratio. On the other hand weight-drop has a broader frequency spectra. Note that an average power spectra containing more than ten shots is shown for every line or source type...………..….15 Figure 2.5: (a) An example raw shot gather from line 2, (b) after vertical stacking of the repeated shots and (c) processed shot. Bed rock reflection B1 is clearly visible. Note the power spectra in the lower right corner after each processing step…….……….………...20 Figure 2.6: (a) An example raw shot gather from line 3, (b) after vertical stacking of the repeated shots and (c) processed shot. Note the power spectra in the lower right corner of each figure

resulting from different processing steps………..…..………..………21 Figure 2.7: Six fold CMP gather. S: sources locations, G: receiver locations (from Yilmaz, 2001)…….22

Figure 2.8: Illustrating the NMO geometry of a horizontal reflector and the NMO correction (from Yilmaz, 2001)……….………….………...………..………..23

Figure 2.9: Travel-time curves recorded for different syncline and anticlines before (a) and after (b) migration (from Yilmaz, 2001)……….….……..………25 Figure 2.10: Stacked and migrated sections along line 2. (a) Stacked time section. (b) and (c) finite- difference and Kirchhoff migrated (interpreted) depth sections with geotechnical borehole (U07208),

(9)

vii

respectively. B1 represents the bedrock topography. It is not clear at this stage what is the origin of the S1 reflection package…….…………..……….….………..…..………...26 Figure 2.11: Stacked and migrated sections along line 3. (a) and (b) Stacked sections with different stacking velocities. (c) Kirchhoff migrated and interpreted depth section with available geotechnical boreholes. B1 represents bedrock topography, and S1 and S2 two sets of sedimentary packages above it.…….……….……….………..………..……….……27 Figure 3.1: Comparison of CPTU, CPTU-R and laboratory measurements from borehole U7208 (from Swedish Geotechnical Institute, 2011)……….……….……….…….…..…29 Figure 3.2: Shows a 3D view from all the processed 2D lines (including lines 2 and 3) and correlation between the reflections. Note the coarse-grained layer (represented with black arrows) and bedrock reflections at various depths along different lines……..……….………31 Figure 3.3: Hypothetical scenarios (a-c) explaining the formation and potential triggering

mechanism of quick-clay landslides in the study area (from Malehmir et al., 2012)……..……….……...…32

(10)

viii

List of tables

Table 2.1: Main reflection seismic acquisition parameters, September 2011…..10 Table 2.2: Main characteristics of the seismic sources used in the study……...…13 Table 2.3: Principle processing steps applied to both lines……….17

(11)

ix

List of Abbreviations

CPTU-R Cone penetration test with resistivity measurement and measurement of the total penetration resistance

SEG Society of Exploration Geophysicists GWB Geoscientists Without Borders PAN Polish Academy of Sciences

ICG International Centre for Geohazards LIAG Leibniz Institute for Applied Geophysics UU Uppsala University

SGI Swedish Geotechnical Institute SGU Geological Survey of Sweden

CDP Common Depth Point

CMP Common Mid-point

AGC Automatic Gain Control

NMO Normal Move-out

GPR Ground Penetration Radar

ERT Electrical Resistivity Tomography CVS Constant Velocity Stack

(12)

x

(13)

1

Chapter 1 1.1 Introduction

My thesis work is a part of the project “Integration of geophysical, hydrogeological and geotechnical methods to aid monitoring landslide in Nordic countries”, sponsored by the Geoscientists Without Borders (GWB) program of SEG. The aim of this project is to understand one particular type of landslide, the quick-clay landslide, which causes rapid earth flow landslides. Several international groups are involved from various organizations and research institutes including Leibniz Institute for Applied Geophysics (LIAG), Polish Academy of Sciences (PAN), International Centre for Geohazards (ICG), Uppsala University (UU), Swedish Geotechnical Institute (SGI) and Geological Survey of Sweden (SGU), to better understand the structural, geometry and conditions for quick-clay liquefaction. Geophysical investigations were carried out with different geophysical methods such as gravity, magnetics, geoelectrics, controlled source/radio magnetotellurics, and 2D, 3D P- and S-wave source seismic methods. In this thesis, I present results from high- resolution 2D reflection seismic data and their correlation with previous geotechnical investigations along two seismic reflection lines.

1.2 Study area

The study area is located on the shoreline of the Göta River, 7 km north of the municipality of Lilla Edet and 60 km north of Göteborg (Fig. 1.1). The Göta River is the largest river in Sweden that runs from Lake Vänern to Göteborg, following the Götaälv Zone, which is an approximately 4 km wide fault zone dipping towards the west. The Göta River is the source of drinking water for about 700,000 people. Göta landslide occurred in 1957. The Göta landslide caused three causalities, large infrastructural damage as well as a significant environmental problem which included stopping the flow of the Göta River for some time (see Göransson et al., 2009). The study area (Fig 1.1) is located near a quick-clay landslide scar that occurred about 30-40 years ago. Therefore, the site is known for its quick-clay formations and landslides (Löfroth et al., 2011).

(14)

2

Figure 1.1: Map of landslide risks in Sweden (provided by the Geological Survey of Sweden) showing the location of the study area and recent quick-clay landslides in southwest Sweden (e.g., Surte, 1950, Göta, 1957, 445 Tuve, 1977, and Småröd, 2006). Figure is modified from Malehmir and bastani (2012).

1.3 The problem statement

Landslides are one of the more frequent natural disasters that cause widespread destruction and loss of life. Quick-clay sliding is one particular kind of landslide that is characterized by rapid earth flow sliding. Usually quick-clay is only found in northern countries such as Russia, Canada, Norway, Sweden, Finland and in the US state of Alaska, which were glaciated during the Pleistocene epoch. A clay is defined as quick based on its sensitivity. Quick-clay is a unique form of highly sensitivity marine clay, with the tendency to change from a relatively stiff condition to a liquid mass when it is disturbed. Undisturbed quick-clay resembles a water-saturated gel (Wikipedia.org).

(15)

3

When a mass of quick-clay undergoes sufficient stress, however, it instantly turns into flowing ooze, a process known as liquefaction (Solberg, 2007). A small block of quick-clay can liquefy from a stress as small as a modest blow from a human hand, while a larger deposit is mainly vulnerable to greater stresses such as earthquake vibrations or saturated by excess rainwater (Malehmir et al., 2012).

The clay in the Göta valley has mainly been deposited in a marine environment at the end of the latest ice age. In marine clay, salt content is the key factor to maintain the stability of the clay structure. Other chemical substances (notably magnesium), also play a role to stabilize the clay structure. When the salt is leached out by upward flowing ground water (artesian water), through fissures in the bedrock or coarse-grained materials, the clay can become “quick” and sensitive to excess stress, which can result in a quick-clay landslide (Solberg, 2007).

Quick-clay landslides are retrogressive. They usually start from close to a river, and grow upwards at a slow walking speed. They have been known to penetrate kilometers inland (Wikipedia.org). A well-known example of this retrogressive nature is the Rissa landslide in Norway (Gregersen, 1981).

Our geophysical understanding to delimit quick-clay geometry and its physical properties, are limited and requires a better understanding to minimize future landslide disasters.

Recognition of the geometry of clay layers and the behavior of their physical properties during and after a disturbance (e.g., excess rainfall) is essential to understanding the mechanism of a landslide. Thus, this research project conducted by multi-disciplinary organizations and institutes is a part of the struggle to better understand the behavior of quick-clay. Known examples of quick-clay landslides in Nordic countries are Trögstad, 1967 (1 million-m³, 4 casualties) (see Nadim et al., 2008; Lundström et al., 2009), Rissa, 1978 (33 hectares, one casualty), and Verdal, 1893 (55 million- m³, 116 casualties) in Norway, and the Surte, 1950 (Caldenius et al., 1956; 22 hectares, 1 casualty), Göta, 1957 (15 hectares, 3 casualties), Tuve, 1977 (27 hectares, 9 casualty), and Småröd, 2006 (10 hectares, no casualty) in Sweden (www.sgu.se/sgu/eng/samhalle/risker/skred_e.htm; Malehmir et al., 2012).

(16)

4

1.4 Previous geophysical and geotechnical investigations

The study area has recently been studied by SGI in a nation-wide project that dealt with investigating areas along the Göta River that are prone to landslides (Löfroth et al., 2011).

The purpose of that study was to understand, how this geophysical and geotechnical investigations contribute to better understanding of quick-clay detection and its sensitivity.

The study area was investigated by various geotechnical methods such as CPTU, CPTU-R soundings, laboratory measurements, hydrogeological studies including pore-pressure monitoring (near the river), and geophysical investigations including geoelectrical and IP measurements. Surface resistivity measurements were acquired along five profiles perpendicular to the Göta River (Fig. 1.2), before the CPTU-R and soil sampling. CPTU-R and sampling were later carried out and suggested the presence of clays with both medium and high sensitivity.

Figure 1.2: Locations of the previous geotechnical boreholes, resistivity profiles (modified from Löfroth et al., 2011) and our seismic reflection lines. Seismic data along lines 2 and 3 are the focus of this study (also see Fig. 2.1).

The CPTU-R measurements were carried out with the measurement of total penetration resistance of the soil. Soil chemical analysis was carried out on samples collected from selected positions to analyze pore water and solid clay particles, and elements in the pore

(17)

5

water including ions, chloride, sulphate, and carbonate. In addition, organic matter, carbonates, pH, electric conductivity and mineralogy were also analyzed (see Löfroth et al., 2011).

The results from all these investigations showed high resistivity (>5 Ωm) indicating leached clay in the upper 15 to 25 m and less leached clay from 25 to 38 m in most part of the study area. Quick-clay has been detected to a depth of more than 15 m. In the western part of the study area, CPTU-R measurements show leached clay, down to a depth of 17 m and less leached clay from a depth of 17 to 38 m. However, in the eastern part, more leached or possibly quick-clay is present than in the western part. In addition, coarse-grained materials, such as sand and silt have been detected underlying the leached clay at shallow depth, especially in the central and eastern part of the study area (Löfroth et al., 2011). Outcrops of crystalline rocks are observed in the southern part of the surveyed area.

1.5 Objectives

To my knowledge, no one has used the reflection seismic method to study quick-clay landslide areas before this project, in Sweden. In addition, refraction seismic methods have only been used for landslide studies (Carroll et al., 1972; Bekler et al., 2011); however, the seismic reflection method has been widely used in countries outside Sweden (e.g., Eichkitz et al., 2009; Polom et al., 2010). Thus, our main aim was to use this traditional geophysical method for imaging quick-clay landsliding areas, so we can compare these high-resolution images with other geophysical and geotechnical investigations. However, in this thesis, the main objectives are:

 Testing the ability of high-resolution reflection seismic method to image the subsurface across quick-clay landslides, where thick marine clay that are highly conductive do not allow deep signal penetration for some geophysical methods such as GPR and ERT.

 Testing different seismic processing methods to produce high resolution images of the subsurface especially refraction statics and high frequency filtering, particularly in this area where the targets are shallow bedrock and layering above it.

(18)

6

 Testing the efficiency of different seismic sources for high resolution imaging by comparing their power spectras.

 To improve the geometrical understanding of the clay and the bedrock topography, and their potential relationships with quick-clays and quick-clay landslides.

(19)

7

Chapter 2 2.1 Reflection seismic method

Reflection seismic method is the most extensively used geophysical exploration method.

It has been successfully used for many decades to delineate subsurface features related to petroleum exploration and engineering/environmental studies. It has also been used to detect faults and interpret stratigraphic relationships, and to resolve the problem related to various geological settings (Steeples and Miller, 1990). In the reflection method, we usually analyze the time-amplitude of an acoustic wave that is generated by a seismic source and reflected back to the surface after reflecting from an acoustic interface. The reflection of acoustic waves from a subsurface interface is similar to echoes which travel through air if one were shouting to close to a cliff or canyon wall. Once these time series are recorded, then processing is required to attenuate noise and boost the strength of the reflected events, so that it can be displayed in a suitable form for interpretation.

Recent developments in the use of high resolution seismic reflection imaging have enhanced its potential and effectiveness in near surface studies. The shallow seismic reflection technique is inexpensive compared to drilling, and can be used successfully to image structure (>100 m) using small sources such as a sledgehammer or weight-drop.

Successful near surface imaging depends on several factors. Firstly, how much acoustic velocity or density contrasts vary in the subsurface. Secondly, the attenuation or near surface layers and the water table which can reduce high-frequencies in the data. Finally, the acquisition parameters and recording equipment must be chosen carefully. They must be compatible with the purposed target and its resolution, as well as environmental issues (Miller and steeples, 1994). Other important considerations related to this method are to record reflections with broader band-width, which also depend on the selection of suitable source and its seismic and technical parameters (Feroci and Orlando et al., 2000).

(20)

8

2.2 Data acquisition

The 2D seismic reflection data along lines 2 and 3 (Fig. 2.1) were acquired in September 2011. The layout for the lines was chosen with regard to a road and after previous geotechnical investigations conducted by SGI. An overall assessment of different geophysical methods used in the study area is recently presented by (Malehmir et al., 2012).

Both lines 2 & 3 with a total length of 480 m and 384 m, respectively, run from south- east to north-west. The difference between the lowest (in the NW) and highest (in the SE) elevations is about 9 m. Asymmetric shooting was carried out with a shooting interval varying from 4 to 8 m depending on the local surface topography along the lines. A 110 kg weight-drop and 15 kg iron sledgehammer were used to generate the seismic signal along these two lines. A SERCEL 428 acquisition system was used for the data recording. A differential global positioning system (DGPS) registered accurately the locations of the shot Figure 2.1: Lidar (light detection and ranging) elevation map overlaid on airphoto from the study area, showing also the location of the seismic lines (2 and 3), available geotechnical boreholes, and landslide scar. Lidar data and airphoto are provided by Lantmäteriet (2011). Figure is modified from Malehmir et al. (2012).

(21)

9

and receiver positions. For lines 2 and 3, 120 and 96 geophones were used respectively, with 4 m spacing. The geophone array remained fixed during the survey. The data were recorded in SEG-D format. These data were collected in two days. Shots were repeated five times at each shot position to enhance the signal-to-noise ratio. The data contain about 105 and 64 (weight-drop), and 2 and 14 (sledgehammer) shots along line 2, 3 respectively. No recording filter and gain were applied in order to observe the signal characteristics as accurately as possible. Seismic data quality is generally good which increased our hope for imaging very shallow near surface structure. First and refracted arrivals are also clearly observed on raw shot gathers as shown in section 2.4 (examples of raw shot gathers from lines 2 and 3). Table 2.1 summarizes the main acquisition parameters used in this study.

(22)

10

Table 2.1: Main reflection seismic acquisition parameters, September 2011

Survey parameters Line 2 Line 3

Survey type 2D 2D

Recording system Sercel 428 Sercel 428

No. of receivers 120 96

No. of shots 108 78

Receiver interval 4 m 4 m

Shot interval 4-8 m 4-8 m

Max source-receiver offset 480 m 384 m

Source /1-117

/118-12

/1-64 /64-78

CDP size 2 m 2 m

CDP range 100-340 100-290

Spread parameters

Record length 6 s

Sampling rate 0.5 ms

Receiver and source parameters

Geophone frequency 28 Hz

No. of geophone per set Single

Source pattern Five shots at each position 1:

2:

(23)

11

2.3 Power spectra study

All average power spectras which are shown here, were calculated by averaging shot power spectras at 5 to 10 shot spacing along both lines. The power spectras were produced in MATLAB; however, the data were extracted from the processing software Claritas.

2.3.1 Comparison between raw and vertical stacked shots

Comparison between the average power spectras are shown in Fig. 2.2(a, c) and (b, d) along lines 2 and 3 respectively. The raw shot power spectras show higher frequency content than the vertical stacked shots regardless of the sources type. Most of these high frequencies are likely noise but it is also possible that some signals were also removed as the results of inaccurate picking of the impact time. This is particularly noticeable at frequencies higher than 200 Hz. We later show the effects of vertical stacking on real shot gathers and the increase in the signal-to-noise ratio due to this process.

2.3.2 Comparison of various seismic sources

Selection of a suitable seismic source is very important to acquire good quality data during a shallow high-resolution seismic survey. Selection of source is also dependent upon the exploration targets and required resolution (Herbst et al., 1998). The compromise between cost and effectiveness is always hard to balance, environment issues and permission limitations are also another issue. A sledgehammer is always cheaper and easier to operate and maintain in a near subsurface study. On the other hand weight-drop and dynamite sources produce a higher frequency signal with higher amplitude, cost more money and involve more safety and permission issues (Rashed, 2009). Table 2.2 shows major characteristics of the various seismic sources used in the study.

(24)

12

Figure 2.2: (a) and (b) show comparisons between power spectras before and after vertical stacked shots. (c) and (d) show comparisons between sledgehammer and weight-drop before and after vertical stack. The raw shot power spectras show higher frequency content than the vertical stacked shots. Note that an average power spectra containing more than ten shots is shown.

(25)

13

Table 2.2 Main characteristics of the seismic sources used in the study

Characteristics Sledgehammer Accelerated weight-drop Dynamite Manufacturer Local hardware store

< 100 $ GISCO USA

(model no: ESS100HM) Specification 15 kg with steel plate

Moderate energy impact 110 kg

Operate with 12 V battery Moderate energy impact

10-20 gr

Shot-depth 0.5 m High energy impact Frequency

content Variable Variable Variable

Repeatability Five times at each shot-peg Five time at each shot-peg No

Field record Need stack and shift Need stack and shift Ready for processing

Site preparation Easy Fairly complex Complex

Mobility Very easy Hard-to-move

Approximate

production rate 30 sec/shot 1 min/shot 5 min/shot

Safety Low Moderate High

Permission issues Less Less High

Fig. 2.2c and d show comparison between the sledgehammer and weight-drop. Weight- drop shows higher frequency content with broader band-width spectra as compared with the sledgehammer. The notably broader frequency band of the weight-drop indicates better resolution than the sledgehammer.

For a better comparison, power spectras of reflection seismic data using a dynamite source were also calculated. These dynamite data were obtained from lines 4 and 5 near to our lines (Fig. 2.1). Power spectras of dynamite, weight-drop and sledgehammer are shown in Fig. 2.4. The aim is to compare and study the frequency contents and band-width of these sources. All sources contain dominant signal frequencies around 30-150 Hz (>2 octave). Since no field filtering or preprocessing was applied to the raw data, amplitudes outside of this range could be considered to be noise. Moreover, this is the range, where most of the energy exists and providing enough band-width for resolving our shallow target. It also might be difficult to retain higher frequencies than this as our sources are weak. The high frequency portions of the power spectra are different for weight-drop and dynamite.

(26)

14

Figure 2.3: Field photos showing various types of seismic sources used in the project. (a) Weight-drop, (b) dynamite and (c) sledgehammer.Figure 2.3: Field photos showing various types of seismic sources used in the project. (a) Weight-drop, (b) dynamite and (c) sledgehammer.

(27)

15

Overall dynamite shows higher high frequencies than the others. The lost of amplitude at higher frequencies represents the earth filtering effect at higher frequencies.

Figure 2.4: Comparisons of sledgehammer, weight-drop and dynamite sources used along different lines. The dynamite source produces the highest frequencies and best signal-to-noise ratio. On the other hand weight-drop has a broadened frequency spectra. Note that an average power spectra containing more than ten shots is shown for every line or source type.

Power spectras for the weight-drop source show a different pattern along lines 2 and 3.

Power spectra along line 2 shows higher frequencies than line 3. Similar differences in spectral behavior are observed along lines 4 and 5 for the dynamite. Near-surface variations or topography may explain this behavior. As (Pullan and Macaulay, 1987) explain a similar effect while comparing weight-drop (75 kg) with shotgun in hole. When the surface was wet and contain fine-grained materials the shotgun produced a stronger signal than the weight- drop. But in the case of dry and coarse material, the shotgun was weaker than the weight- drop. Weather conditions were variable during the seismic data acquisition from sunny days with almost no wind to rainy and stormy days. This could as well explain the difference between the power spectras from one line to another.

(28)

16

In summary, the explosive source produces the highest frequencies and best signal-to- noise ratio; because it has higher energy related to higher burn/blast velocity and source containment than the others (Praeg, 2003; Yordkayhun et al., 2009). As we know, the higher the source energy the greater the attenuation of the higher frequencies. The weight-drop source has a broader frequency band compared with dynamite and the sledgehammer, if the power spectras are described in terms of band-width. Therefore, the weight-drop source might be better suited if for example a combination of shallow and deeper targets were considered. Moreover, it’s cheaper and involves less permission and safety issues as compared to Dynamite.

2.4 Data processing

Applying different methods of signal processing is necessary after data acquisition in order to produce a seismic image that can be interpreted. Data processing sequences depend on the geological scenario, acquisition parameters and target depth (Yilmaz, 1989).

However, our target is very shallow so processing parameters were carefully chosen to produce a high resolution image. For the data processing, a conventional post-stack migration method was used with the main focus being the key processing steps to remove the coherent noise and enhance the image quality. Table 2.2 summarizes the processing steps that were used to process both lines 2 and 3. Globe Claritas was used to process these lines. Different job files were created to apply different processing process.

Firstly, the seismic data were converted from SEG-D to SEG-Y and then all raw data were shifted up to zero time as the time differs from shot to shot for a given location. We used a nearby geophone to estimate the delay. Vertical stacking was applied for all raw shots that were repeated at the same shot location. Figs. 2.5 and 2.6 show example raw shot gathers from lines 2 and 3, respectively before and after vertical stacking with their amplitude spectrum. We can clearly see the overall quality and continuity of the data increase, especially at far-offset traces.

(29)

17

Table 2.3: Principle processing steps applied to both lines

Step Parameters

1. Read 0.2 s SEG-D data

2. Extract and apply geometry (CDP bin size 2 m) 3. Elevation correction: datum 25 m

4. Trace muting

5. Spectral whitening 40-60-220-240 Hz with a window length of 30 Hz 6. Deconvolution

Application window 0-200 ms:

Filter length = 100 ms, Gap length = 5 ms , 0.1% noise added 7. AGC, window 100 ms

8. Common depth point sorting 9. Velocity analysis

10. Normal moveout corrections (NMO): 100% stretch mute

11. Stack

12 FX-deconvolution 13. Trace balance

14. Band-pass filtering: 20-30-110-120 15. F-K filtering

16. Migration: finite-difference (line 2), Kirchhoff (lines 2 and 3) 17. Time-to-depth conversion: using constant velocity of 1300 m/s

2.4.1 Apply geometry and elevation statics

The geometry information of the shots and receivers was extracted from the header of the SEG-Y file and applied to the data after SHOTID, CPD, offset information were introduced to the data header. In total, 241 and 191 CDPs with a rectangular bin size of 2 m along the line were generated for lines 2 and 3, respectively. The land was non planar and irregular along both lines, therefore sources and receivers do not share the common elevation and require traveltime adjustment to shift them to a common plane surface. This process is called elevation statics (Yilmaz, 1989). To compensate for surface irregularities, elevation corrections were applied using a 25 m seismic datum with replacement velocities of 700 m/s

(30)

18

and 1200 m/s for lines 2 and 3, respectively. The choice of different velocity was due to variation in the bedrock geometry from lines 2 to 3.

Refraction static estimation was performed several times using a single-layer model after picking first breaks for both lines. Each shot was carefully checked before and after refraction static corrections. Unfortunately, refraction statics did not work well on these data. We lost the continuity and quality of the reflections observed in raw shot gathers, by removing the effect of the low velocity layer from the data. Two major problems were encountered with refraction static corrections. Firstly, bedrock is very shallow in this area that is why refraction static solution was not accurate to account for this large bedrock variation. Bedrock outcrops are seen near the lines. In this scenario, it’s quite difficult to figure out the velocity of uppermost undulating topography layer, where one side of the line has very low velocity and the other side has very high velocity due to the presence of the bedrock. Secondly, our target in this study is shallow quick-clay layer at 10-20 m according to previous geophysical and geo-technical investigations. So, refraction static step was skipped and later proved to be not significant for these two lines.

2.4.2 Trace editing, spectral whitening and deconvolution

Trace editing is actually a quality control step, performed either during the fieldwork or during the data processing. In trace editing, we can remove bad traces and noisy shots, and mute (zeros) first break, air waves and stretching effect after applying NMO corrections. Bad traces were killed; these may have been due to dead geophones or bad couplings of the geophone with the earth (Figs. 2.5 and 2.6).

Band-pass filtering and deconvolution are the most utilized, and effective processing steps for removing coherent noise from the shot gathers. Filtering is used to suppress the ground roll and for removing high frequency ambient noise. After filtering, the output contains only desired frequencies with no phase change (Khan, 2010). It can be done in time domain or frequency domain. Filtering in the frequency-domain involves multiplying the amplitude spectrum of input seismic trace by a filter operator. On the other filtering in the time-domain involves convolving the filter operator with the input time series (Yilmaz, 2001).

Additionally, the filtering process in the time or frequency domain, are based on the

(31)

19

following important concepts: “Convolution in the time domain is equivalent to multiplication in the frequency domain. Similarly, convolution in the frequency domain is equivalent to multiplication in the time domain (Bracewell, 1965)”. When seismic signals travel through the earth, high frequencies are lost due to convolution with earth materials.

Deconvolution is usually used to recover these absorbed high frequencies and for removing the multiples. As deconvolution remove the effect of earth convolution on recorded data.

Thus, deconvolution can be considered as an inverse filter. Many types of deconvolutions are available, depending upon the different assumptions, used and established for different data sets. Typically, deconvolution techniques are based on the optimum wiener filtering (Khan, 2010).

In order to remove unwanted signal especially the strong ground roll observed in the data and to increase the temporal resolution and to flatten the amplitude spectrum, deconvolution and zero-phase spectral equalization with a window length of 30 Hz were applied. Figs. 2.5c and 2.6c present the processed shots of those shown in Figs. 2.5b and 2.6b. Both shots now show the improved signal to noise ratio, and flattened amplitude spectrum. All these processing steps helped to prepare the shots for final processing which includes stacking and migration. As it is clear from Fig. 2.5c, a reflection (B1) presumably from the bedrock is identifiable in the processed shot gather increasing the hope that proper stacking and migration would bring out numerous reflections in the data. Before sorting the data to CDP domain, all the energy above the first arrivals were muted.

(32)

20

Figure 2.5: (a) An example raw shot gather from line 2, (b) after vertical stacking of the repeated shots and (c) processed shot. Bedrock reflection B1 is clearly visible. Note the power spectra in the lower right corner after each processing step.

(33)

21

Figure 2.6: (a) An example raw shot gather from line 3, (b) after vertical stacking of the repeated shots and (c) processed shot. Note the power spectra in the lower right corner of each figure resulting from different processing steps.

(34)

22

2.4.3 CDP sorting

After performing initial signal processing, we transferred the data from source-receiver order to common depth point-offset order using geometry information stored in the header of SEG-Y file, so that we can stack after considering the NMO corrections. In this process, the traces from different source-receiver pair which image the same CDP are assigned into a CDP gather. Note that, CMP is often called CDP; however this is valid when reflectors are horizontal and velocities are varying horizontally (Fig. 2.7).

Figure 2.7: Six fold CMP gather. S: sources locations, G: receiver locations (from Yilmaz, 2001).

2.4.4 Velocity analysis and NMO corrections

Velocity analysis is an important interactive process in seismic data processing, where the resulting velocity field is used for the NMO and stacking processes. The purpose of the velocity analysis is to build a velocity model that can flatten the reflected events using the NMO correction, in order to stack constructively (Yilmaz, 2001).

The Earth is composed of stratified layers with different lithologies, having different velocities. Typically, velocity increases with depth due to the compaction process. Our main objective in velocity analysis is to find a correct velocity to align our reflections horizontally.

Correct velocity analysis is the only way to reveal a truly geologic picture of the subsurface.

There are varieties of methods available to perform velocity analysis, i.e. constant velocity

(35)

23

stack (CVS), constant velocity gather, semblance analysis. Velocity analysis was performed manually and iteratively using CVS (constant velocity stack). We stacked selected data panels several times, after applying NMO using a range of velocities from 600 m/s to 2500 m/s with increment of 25 m/s. This made it possible to define a correct velocity function to image shallow events and bedrock topography.

The NMO correction is a function of time, offset and velocity (Fig. 2.8). We need NMO corrections to remove the offset effect from travel times.

Figure 2.8: Illustrating the NMO geometry of a horizontal reflector and the NMO correction (from Yilmaz, 2001).

Actually NMO is not a static time shift like static shifts. We often need stretch mute to mute these shallow stretched events produced as a result of different time shifts (Yilmaz, 2001).

The NMO correction was applied using 100% stretch mute in order to preserve wide-angle reflections from the bedrock (see Fig. 2.5).

2.4.5 Stacking

Finally CDP sorted data were stacked together to enhance signal-to-noise ratio and remove incoherent noise. The stack adds all the same depth-point traces together and makes a single trace for each ensemble. To process lines 2 and 3, conventional stack method was used for the stacking. Figs. 2.10a and 2.11a, b show time sections along lines 2 and 3,

(36)

24

respectively. The time section along line 2 was stacked with stacking velocities ranging from 800 to 1350 m/s. Fig. 2.11a, b shows two time sections for the stack with different stacking velocities around 1350 m/s and 800 m/s, respectively. As mentioned earlier, stacked sections were very sensitive to the stacking velocities, that’s why two time sections were generated to enable us to image the bedrock and shallow events. We were not able to image these events simultaneously because the reflections are too close and require very different velocities to be imaged. Our conventional based processing method is not able to handle such a large velocity variations within a very short time window.

2.4.6 FX-deconvolution and FK-mute

Post-stack FX-deconvolution and FK-mute were used to attenuate random noise. FX- deconvolution is used as a post-stack filter. Complex wiener deconvolution is performed in the X direction for each frequency after converting the whole section into FX space. FX- deconvolution usually gives better result than other post-stack coherency filters. FX- deconvolution process works one dip at a time which makes it less superior than other methods; events with secondary dip are attenuated (Claritas dictionary).

FK-mute can also be used before and after stack. It is used to eliminate coherent linear noise and side-scattered energy from the reflection energy in (FK) space. Coherent noise is removed based on their dips in the collected data (Yilmaz, 1989). We applied FK-filtering on the stacked section in order to attenuate steeply-dipping events, believed to be the remaining of ground roll and shear-wave energy.

2.4.7 Migration

Migration is used to shift the dipping reflectors into their true subsurface positions (Fig.

2.9) and also to increase their spatial resolution. Migration is also used to collapse the diffractions. The goal of migration is to make the stacked section appear similar to the geologic cross section along the seismic line (Yilmaz, 1989). Different migration methods were tested. Kirchhoff poststack migration results were more reliable as shown in Figs. 2.10

(37)

25

and 2.11. Seismic data were migrated and depth converted using smoothed version of the stacking velocities (see Figs. 2.10 and 2.11).

Figure 2.9: Travel-time curves recorded for different syncline and anticlines before (a) and after (b) migration (from Yilmaz, 2001).

(38)

26

Figure 2.10: Stacked and migrated sections along line 2. (a) Stacked time section. (b) and (c) finite-difference and Kirchhoff migrated (interpreted) depth sections with geotechnical borehole (U07208), respectively. B1 represents the bedrock topography. It is not clear at this stage what is the origin of the S1 reflection package.

(39)

27

Figure 2.11: Stacked and migrated sections along line 3. (a) and (b) stacked sections with different stacking velocities. (c) Kirchhoff migrated and interpreted depth section with available geotechnical boreholes. B1 represents bedrock topography, and S1 and S2 two sets of sedimentary packages above it.

(40)

28

Chapter 3 3.1 Interpretation

Fig. 2.10 shows unmigrated and depth converted migrated stacked sections along line 2.

Two packages of reflections (S1, B1) are seen on the stacked sections. The characteristics of both reflections are consistent in both the migrated and unmigrated sections. S1 reflection, dips from the middle of line 2 to the NW. Other reflection, B1 dips in both directions from the center of line 2 i.e., SE and NW. B1 reaches close to the surface in the central part of the line. So, it most likely represents the bedrock. S1 reflection package on laps the bedrock at the center of the line and may also represent part of the bedrock or a contact between the clay and coarse-grained material.

Results from a geotechnical borehole U07208 (Fig. 3.1) located at the NW end of line 2 are presented to better understand the near surface geology. The borehole measurements were obtained from previous investigations carried out by the Swedish Geotechnical Institute. Conventional CPTU with resistivity (CPTU-R) and laboratory measurements were carried out at point U7208. The CPTU measurements based on total resistance against the penetration was compared with CPTU-R and lab measurements. Five ohm-m resistivity was chosen as a starting point for leached and possibly quick-clay. Borehole U7208 shows a decrease in the resistivity from about 100 ohm-m at 3 m depth to 1-2 ohm-m at 28 to 38 m depth. The rod friction in the sounding indicates about one meter quick-clay at 7 to 8 m depth. Moreover, laboratory tests did not confirm the presence of quick-clay but rather sensitivities higher than 50 between 5 to 13 m depth. So, results from borehole measurements suggest the presence of leached clay down to about 17 m depth and less leached clay from a depth of 17 to 38 m (Löfroth et al., 2011). Seismic data along line 2, however, show no reflection down to this depth implying that no quick-clay or coarse-grain layer could be detected or simply that it is too thin. No other materials are intersected down to 38 m depth which confirms the reflection seismic results (see Fig. 2.10).

(41)

29

Figure 3.1: Comparison of CPTU, CPTU-R and laboratory measurements from borehole U7208 (from Swedish Geotechnical Institute, 2011).

Fig. 2.11a clearly shows the B1 reflection, dipping to the NW direction along line 3, identified as the bedrock. The B1 reflection starts at about 40 ms in the SE and extends down to about 120 ms in the NW direction. Three reflection packages (S1, S2 & B1) are observed in a different stacked section shown in Fig. 2.11b. As mentioned earlier, to obtain this section we used a much lower stacking velocity. This section shows, S1 as flat-lying at about 20-40 ms, S2 as gently dipping from the center of the section towards the end of line 3. The migrated and time-to-depth converted section shows a combination of the two sections.

However, the reflections are not as continuous and clear as observed on the individual sections of line 3. Results from geotechnical boreholes 07072, 07071, 07070 and 07069 along line 3 suggest a top layer of fluvial deposit with thickness of about 3-6 m which overlay normal clays that extend down to at least 35 m depth. This implies that the reflection S1 may be an artifact from the processing. We are certain that the S2 reflection is not an artifact from the processing since we observe it in the processed shot gathers. Reflection S2 is likely to represent the contact between clay materials and a coarse-grained layer which was also

identified in a perpendicular seismic line presented by Malehmir et al., 2012 (Fig. 3.2).

(42)

30

3.2 Discussion

A coarse-grained layer was successfully imaged and interpreted by other seismic reflection lines (Malehmir et al., 2012) near to these two lines (Fig. 3.2). Previous geotechnical investigations also confirmed the presence of quick-clay at various depths in the central and eastern parts of the study area. Thus, the S2 reflection could be from the contact between clay and underlying coarse-grained layer along line 3. This highlights the importance of monitoring this site for the formation of quick-clay and future landslides in the study area.

(Malehmir and Bastani, 2012) presented three scenarios which may explain the formation of quick-clays in the study area and their triggering mechanism. In their scenarios, the coarse- grained layer plays an important role in bringing excess rain/snow water from the highland areas, potentially leaching the salt from the clay and also occasionally increasing the pore- water pressure which then leads to quick-clay landslides. Fig. 3.3 represents theses scenarios.

3.3 Conclusions

Results from processed shots and stacked sections demonstrate the high quality of seismic data. Good surface conditions and near surface saturated clay may be a reason for the good quality data. Vertical resolution is mainly depend on the band-width and on the frequency content. So, selection of suitable source is necessary which can produce the required high frequencies and broader band spectra. In fact, an inappropriate source not only affects the frequency contents but also the quantity of energy generated and signal-to- noise ratio. In our seismic data, dynamite is characterized by relatively high frequency content or higher energy as compared with the other sources as shown by the power spectras. Vertical stacking of weight-drop and sledgehammer shots proved to be crucial to increase the signal-to-noise ratio and helped to image very shallow reflections. Nevertheless, the source comparisons show that all these sources produce seismic signals with sufficient energy for shallow subsurface imaging down to about 100 m depth.

(43)

31

Figure 3.2: 3D view from all the processed 2D lines (including lines 2 and 3) and correlation between the reflections. Note the coarse-grained layer (represented with black arrows) and bedrock reflections at various depths along different lines.

(44)

32

Figure 3.3: Hypothetical scenarios (a-c) explaining the formation and potential triggering mechanism of quick-clay landslides in the study area (from Malehmir and Bastani., 2012).

(45)

33

The weight-drop shows a broader frequency spectra. Weight-drop may be more suitable than the dynamite especially in this area where our aim is to image very shallow reflections and bedrock topography. Weight-drop is cheap and also involves less permission issues and safety problems compared with dynamite. Moreover, it is also possible to increase the signal-to-noise ratio through repeated shots and if seismic data is processed carefully with suitable processes and processing parameters. Some processing operations (static corrections and frequency filtering) needed special attention. Deconvolution and spectral equalization are also important. Spectral equalization worked very well in our data. It successfully shifted the frequency spectra toward the high frequencies. Overall, prestack spectral equalization and frequency filtering after stack as well as velocity analysis were the key processing steps in our data to produce these high resolution seismic images. In summary, to produce a good seismic image it is necessary to choose the correct source, geophones, acquisition geometry and processing parameters. In addition, seismic and technical parameters of the data acquisition mainly depend on the surface conditions and geology of the area. Finally, we are successfully able to image the shallow reflections and bedrock topography. The reflection S2 (25-50 m) along line 3 interpreted as a coarse-grained layer, may be playing an important role in the formation of quick-clay making the area more acceptable to landslides. Future studies especially geotechnical and geophysical well logging investigations are strongly recommended to better understand quick-clay landslide mechanism and to avoid further landslide disaster in this area.

(46)

34

References

Bekler T., Ekinci YL., Demirci A., Erginal AE., and Ertekin C (2011). Characterization of a landslide using seismic refraction, electrical resistivity and 370 hydrometer methods, Adatepe – Çanakkale, NW Turkey. Journal of Environmental and Engineering Geophysics 16 (115-126).

Bracewell R N (1965). The Fourier transform and Its applications. McGraw-Hill, 2nd edition 1978, revised 1986.

Caldenius C and Lundström R (1956). The landslide at surte on the river Göta Älv. A geologicgeotechnical study. Swedish Geological Survey Ca27 (p 88).

Carroll RD., Scott JH and Lee FT (1972). Seismic refraction studies, in geological, geophysical and engineering investigations of the Loveland Basin landslide, Clear 380 Creek county, Colorado 1963-65. U.S. Geological Survey Professional Paper 673 C (17-19).

Eichkitz CG., Schreilechner MG., Amtmann J and Schmid C (2009). Shallow seismic reflection study of the Gschliefgraben landslide deposition area – interpretation and three dimensional modeling. Austrian Journal of Earth Sciences 102 (52-60).

M., L., R., C., E and G (2000). Some considerations on shallow seismic reflection surveys. Journal of Applied Geophysics 45 (127-139).

Gregersen O (1981). The quick clay landslide in Rissa, Norway. The sliding process and discussion of failure modes. Norwegian Geotechnical Institute Publication No 135.

Göransson GI., Bendz D and Larson PM (2009). Combining landslide and contaminant risk: a preliminary assessment. Journal of Soils and Sediments (9:33–4).

Herbst R., Kapp I., Krummel H and Luck E (1998). Seismic sources for shallow

investigations: A field comparison from Northern Germany. Journal of Applied Geophysics 38 (301-317).

Khan K A (2010). Geophysical techniques for explortion of natural resources. Oil and Gas Training Institute (OGDCL), Islamabad Pakistan.

(47)

35

Löfroth H., Suer P., Dahlin T., Leroux V and Schälin D (2011). Quick clay mapping by resistivity – surface resistivity, CPTU-R and chemistry to complement other geo- technical sounding and sampling. Swedish Geotechnical Institute Report GÄU 30.

Lundström K., Larsson R and Dahlin T (2009). Mapping of quick clay form at ions using geotechnical and geophysical methods. Landslides 6 (1–15).

Miller R D., Steeples D W and Myers P B (1990). Shallow seismic reflection survey across the Meets Fault, Oklahoma. Geol. Soc. Am. Bull 102 (18-25).

Miller R D., Pullan S E., Steeples D W and James A Hunters (1992). Field comparison of shallow seismic sources near Chino. Geophysics Vol 57 (693-709).

Miller R D and Steeples D W (1994). Applications of shallow high-resolution seismic reflection to various environmental problems. Journal of Applied Geophysics 31 (65-72).

Malehmir A and Bastani M (2012). Ultra high-resolution reflection seismic imaging of quick-clay and associated formations at a landslide scar in southwest Sweden. Submitted to Journal of Applied Geophysics.

Malehmir A., Bastani M., Krawczyk C., Polom U., Juhlin C., Gurk M., Malinowski M., Persson L., Ismail N and Lecomte I (2012). Geophysical assessment and geotechnical investigation of quick-clay landslides - a case study in Sweden. In revision in Near Surface Geophysics.

Nadim F., Pedersen SAS., Schmidt-Thomé P., Sigmundsson F and Enggdahl M (2008).

Natural hazards in Nordic Countries. Episoes 31 (176–184).

Polom U., Hansen L., Sauvin G., LíHeureux J-S., Lecomte I., Krawczyk CM., Vanneste M andLongva O (2010). High-resolution SH-wave seismic reflection for characterization of onshore ground conditions in the Trondheim harbor, central Norway. Advances in near- surface seismology and ground-penetrating radar SEG (297-312).

Praeg D (2003). Seismic imaging of mid-Pleistocene tunnel-valleys in the North Sea Basin – high resolution from low frequencies. Journal of Applied Geophysics 53 (273–

298).

(48)

36

Pullan S E and MacAulay H A (1987). An in-hole shotgun source for engineering seismic surveys. Geophysics 52 (985–996).

Rankka K., Andersson-Sköld Y., Hultén C., Larsson R., Leroux V and Dahlin T (2005).

Quick clay in Sweden. Report 65. Swedish Geotechnical Institute, Linköping.

Rashed M A (2009). GDP: A new source for shallow high-resolution seismic exploration.

Journal of Applied Geophysics 68 (243-248).

Solberg I (2007). Geological, geomorphological and geophysical investigations of areas prone to clay slides: examples from Buvika, Mid Norway. Doctoral thesis. Norwegian University of Science and Technology, Faculty of Engineering Science and Technology, Department of Geology and Mineral Resources Engineering. Trondheim.

Steeples D W and Miller R D (1990). Seismic-reflection methods applied to engineering, environmental, and ground-water problems. Soc. Explor.Geophys., Geotechnical and Environmental Geophysics Vol 1: Review and Tutorial pp (1-30).

Yordkayhun S and Suwan J N (2012). A university-developed seismic source for shallow seismic surveys. Journal of Applied Geophysics 82 (110-118).

S., A., Giese R., Juhlin C and Cosma C (2009). Comparison of surface seismic sources at the SINK site, Ketzin, Germany. Geophysical Prospecting (doi:

10.1111/j.1365-2478.2008.00737.x).

Yilmaz Ö (1989). Seismic data processing. Society of Exploration Geophysics (SEG), Tulsa.

Yilmaz Ö (2001). Seismic data analysis, Vol 1. Society of Exploration Geophysics (SEG), Tulsa.

2D Reflection Seismic Imaging at a Quick-clay Landslide Site, in Southwest Sweden

Examensarbete vid Institutionen för geovetenskaper ISSN 1650-6553 Nr 249