SurfSeis©

Active and Passive MASW

Seismic Processing Software

for use with
Microsoft® Windows™

(Supplementary)
User’s Manual
v 2.0
written by
Choon B. Park
and edited by
Marla Adkins-Heljeson
November 2006

modified by
Julian Ivanov and
Mary Brohammer
January 2007

Kansas Geological Survey

1930 Constant Avenue
Lawrence, Kansas 66047-3724 USA
Phone: 785-864-3965/2176, Fax: 785-864-5317
E-mail: [email protected]
www.kgs.ku.edu/software/surfseis/index.html
 Table of Contents

Chapter 1: Introduction

Chapter 2: Active MASW

2.1 Data Acquisition
2.2 Data Processing

Chapter 3: Passive Remote MASW

3.1 Receiver Array
3.2 Data Acquisition
3.3 Data Processing

Chapter 4: Passive Roadside MASW

4.1 Receiver Array
4.2 Data Acquisition
4.3 Data Processing

Chapter 5: Data Processing − Dispersion Analysis

5.1 Step 1–Processing for Overtone (OT) Image Data
5.2 Step 2–Extraction of Dispersion Curves from OT Image

Chapter 6: Data Processing − Inversion Analysis

6.1 Inversion of Dispersion Curve
6.2 Inversion of Dispersion Image

Chapter 7: Combining Active and Passive Overtone (OT) Records

7.1 Combining One OT to Multiple OT’s
7.2 Combining Multiple OT’s of Passive Roadside Data

References
 Chapter 1 Introduction

Since the release of SurfSeis version 1.0 in 2000, the surface-wave method in general has gone through a
tremendous amount of applications and research at many places, including the Kansas Geological Survey
(KGS). In consequence, SurfSeis 2.0 has new features that take into account those new methodological
developments as well as users’ demands for enhanced user-friendliness. Although the manual
(“Manual20.pdf”) has been prepared mainly to explain the new features, such as those related to the
passive part of the multichannel analysis of surface waves (MASW) method, it can serve as a stand-alone
manual for both active and passive MASW methods for those who have previous experience in seismic
data acquisition and data processing in either body- or surface-wave methods. Other information dealing
with previous versions of SurfSeis can be found in the previous manual (v 1.5) and README files stored
in “…\Manual” and application folders, respectively.

In this manual, the MASW method has been divided into three categories by survey mode: active, passive
remote, and passive roadside MASW methods (Figure 1.1). The active method (Park et al., 1999) was
introduced in previous versions of SurfSeis. It is the conventional mode of survey using an active seismic
source (e.g., a sledge hammer) and a linear receiver array, collecting data in a roll-along mode. The other
two methods utilize surface waves generated passively from ambient cultural (and natural) activities such
as traffic (and thunder, tidal motion, atmospheric pressure change, etc.). The passive remote method
(Park et al., 2004; 2005) employs a two-dimensional (2-D) receiver array such as a cross or circular lay-
out to record passive surface waves. This results in the most accurate evaluation of shear-wave velocity
(Vs) at the expense of more intensive field operation and the burden of securing a wide-open space for the
array. This can be a good choice if a relatively regional one-dimensional (1-D) Vs profiling is needed.
The passive roadside MASW method (Park and Miller, 2006) adopts the conventional linear receiver
array and tries mainly to utilize those surface waves generated from local traffic. It tries to overcome
limitations with the passive remote method such as difficulty in securing a spacious area and inconven-
ience in field operations by sacrificing the accuracy (usually less than 10%) of the Vs evaluation. With
this method, the array can be set along the sidewalk or the shoulder of a road and the survey can continue
in a roll-along mode for the purpose of 2-D Vs profiling. Using a land streamer for the array can improve
the survey speed by as much as a few orders of magnitude. In addition, an active impact (e.g., by using a
sledge hammer) can be applied at one end of the array to trigger the long (e.g., 30 sec) recording of data.
This can result in the active-passive combined analysis of surface waves for the purpose of obtaining both
shallow (e.g., 1-20 m) and deep (e.g., 20-100 m) Vs information simultaneously. A more detailed
description of each method can be found in previously cited references. However, field procedures for
these methods (both passive and active) have been summarized in this version of manual. Data-processing
steps are explained by using the sample data sets stored in the “…\SurfSeis20\SampleData\” folder. All
acquisition parameters for the sample data sets are listed in Table 1.1.

A summary of the entire procedure with a MASW method (active or passive) is displayed in the flowchart
in Figure 1.2. Major changes and new features with this version are summarized as follows:

1
1. Modules to process passive surface waves
have been added in addition to the previously
existing active module (Figure 1.1). Two
different types of passive surveys are avail-
able: one, called the passive remote MASW
method, uses a two-dimensional (2-D) re-
ceiver array and the other, called the passive
roadside MASW method, uses the conven-
tional 1-D linear array.
2. The way dispersion analysis is executed has
been changed so that the previous sequence
of ‘Preprocess → Overtone → Run → Save’
has been divided into two separate steps:
(1) generation of dispersion image (called
overtone, OT) data and (2) mouse-aided
extraction of the dispersion curve from the
image. The previous sequence, however, can
still be accessed by right-clicking (instead of
normal clicking) the ‘Dispersion’ button in
the analysis menu when importing an input
seismic file.
3. A new mode of inversion has been added.
This is a general Monte-Carlo method
applied directly to the dispersion image
(instead of the dispersion curve) seeking the
best-matching solution through a random
search. With this mode, up to four modes of
dispersion can be accounted for and all the
parameters in a five-layer earth model can be
manually changed, if desired, to compare
theoretical curve(s) with dispersion trend(s)
in the background image.

Most (if not all) of the bugs existing in previous

versions of SurfSeis have been fixed, thanks to
the comments and reports from many practi-
tioners who used SurfSeis to conduct the MASW
method and who were very patient and always
willing to help this quite new geophysical Figure 1.1
method still in its infancy evolve into a better
form. We researchers at KGS sincerely appre-
ciate your patience and assistance.

2
Figure 1.2

3
Table 1.1 Summary of Sample Data Parameters

Active Passive Remote Passive Remote Passive Roadside

Survey Type
MASW MASW MASW MASW
“1011.dat” ⎯ “Passive- “4000.dat” ⎯
File Name(s) “Passive-Cross.dat”
“1022.dat” Circular.dat” “4009.dat”
Folder “…\Active\” “…\PassiveRemote\” “…\PassiveRemote\” “…\PassiveRoadside\”
Survey
2-D Vs Profiling 1-D Vs Profiling 1-D Vs Profiling 2-D Vs Profiling
Purpose
Data Format SEG-2 KGS KGS SEG-2
Acquisition 24 channel 48 channel 24 channel 48 channel
Source 20-lb Hammer Traffic Traffic 12-lb Hammer/Traffic
Receivers 10-Hz 4.5-Hz 4.5-Hz 4.5-Hz (land streamer
(Geophones) (spike coupling) (spike coupling) (spike coupling) with 30 takeouts)
Linear Linear
Receiver Array Cross (x-y) Circular
(roll along) (roll along )
Array
23 m 115 m 115 m 35 m
Dimension (D)
Receiver
1.22 m 5m 15 m 1.2 m
Spacing (dx)
Source
1.22 m N/A N/A 4.8 m
Offset (x1)
Receiver Array
1 dx (1.22 m) 0 0 4 dx (4.8 m)
Move
Sampling
1 ms 4 ms 4 ms 4 ms
Interval (dt)
Recording
2 sec 20 sec 120 sec 120 sec
Time (T)
Record
1011 ⎯ 1022 2000 ⎯ 2009 3000 ⎯ 3009 4000 ⎯ 4009
Numbers

Chapter 2 Active MASW Survey

This is the most common type of MASW survey that can produce a 2-D Vs profile, an example
of which is displayed in Figure 2.1. Maximum depth of investigation (zmax) that can be achieved
is usually in 10-30m range. However, it can vary with different sites and different types of
active sources used. Field procedures and data-processing steps are briefly explained below.
Some of the field parameters, for example, source offset (x1) and receiver spacing (dx), are
described based on the most recent research results and therefore may be different from those
previously reported. Description of data-processing steps is presented by using a sample field
data set stored in “…\SurfSeis20\SampleData\” folder. Recording parameters are listed in Table
1.1. The description is focused on those new features with this particular version (2.0) of
SurfSeis and it is recommended to refer to previous (v 1.5) manual and README files for the
description of all those features not explained here.

Surface waves are best generated over a ‘flat’ ground at least within one receiver-spread length
(D) (Figure 2.2). Then, overall topographic variation within an entire survey line should not
matter. However, any surface relief whose dimension is greater than, say, 10 % of D will cause a
significant hindrance to the surface-wave generation (Figure 2.2).

4
 Figure 2.1

Figure 2.2

2.1 Data Acquisition

The following sections describe most of the parameters related to data acquisition. They are the
ones most commonly used at KGS, and by no means represent a required set of values. Slight
variation in any parameter can always be expected.

Source
A fairly heavy sledge hammer (e.g., 10 lb) will be a good choice. Although any other type (e.g.,
a weight drop) that can deliver more impact power into the ground can be an advantage over a
sledge hammer because of its capacity of generating lower (longer) frequencies (wavelengths) of
surface waves, the gain is often not enough to warrant the cost of equipment and inconvenience
in field operation. Using an impact plate (also called base plate) will help the source impact
point become less intrusive into soil. Detailed study on the role of the base plate in surface wave
generation has not been reported yet and needs to be undertaken in the near future. Recently, it
has been reported that a non-metallic plate (e.g., a firm rubber plate) can generate noticeably
stronger energy at the lower-frequency (e.g., < 10 Hz) part of surface waves than the conven-
tional metallic one.

5
Vertical stacking with multiple impacts can suppress ambient noise significantly and is therefore
always recommended, especially if the survey takes place in an urban area. Optimum number of
stacking can be determined when there is little change in signal-to-noise ratio (S/N) in the dis-
played seismic record during the stacking. We often use three-five vertical stacks at the Kansas
Geological Survey (KGS). This number, however, should increase as ambient noise level in-
creases and/or total receiver array length (D) increases.

Receivers
Low-frequency (e.g., 4.5-Hz) geophones are always recommended. Effectiveness of somewhat
high-frequency (e.g., 10-20 Hz) phones, however, is often comparable to that of much lower-
frequency ones. Vertical (instead of horizontal) phones have to be used. Although spike-coupled
geophones always give the highest sensitivity, the coupling provided by a land streamer can be
equally well suited with a significant convenience in field operation.

Field Geometry
Length of the receiver spread (D) (Figure 1.1) is directly related to the longest wavelength (λmax)
that can be analyzed, which in turn determines the maximum depth of investigation (zmax):

D ≈ λmax ≈ zmax. (2-1)

On the other hand, receiver spacing (dx) is related to the shortest wavelength (λmin) and therefore
the shallowest resolvable depth of investigation (zmin).

dx ≈ λmin ≈ zmin. (2-2)

In practice, however, λmax (therefore D) in an active survey is usually limited by the seismic
source because it is the primary governing factor. It is usually in a few tens of meters (e.g., < 30
m) in most cases.

The source offset (x1) controls the degree of contamination by the near-field effects. Its optimum
value has been a debatable subject. We, however, suggest a value of about 20 % of D (e.g., x1=5
m when D=25 m). A large value of x1 (e.g., > 10 m) and a large D (e.g., > 100 m) will increase
the risk of higher-modes domination and reduce S/N for the fundamental mode. The following
values are most recommended:

x1 = 5 m, dx = 1 m (or less), and D ≤ 30 m. (2-3)

Interval (dSRC) of Source-Receivers Configuration (SRC) Movement (e.i. Roll)

An interval between 1dx-12dx is recommended. 4dx is most commonly used in the case of
24-channel acquisition.

Recording Parameters
One millisecond (dt = 1 ms) of sampling interval is most common with a 2-sec total recording
time (T = 2 sec). Use a smaller dt (e.g., 0.5 ms) if any body-wave processing (e.g., refraction and
reflection) is planned as by-product(s). In the case of extremely low velocities (e.g., Vs < 100

6
m/sec), a longer T (e.g., 4 sec) will be a better choice. In any case, an excessively long T (e.g., T
≥ 5 sec) is discouraged because it can increase the chance of recording ambient noise (e.g.,
traffic). Usually, 24-channel acquisition will be optimal. If 48-channel acquisition is available,
it is recommended to shorten dx rather than to increase D.

2.2 Data Processing

Procedure of processing an active data set is explained by using a sample data set stored in
“…\SurfSeis20\SampleData\Active” folder. Twelve field records of 24-channel acquisition were
acquired over a soil site near a highway and saved with file names of 1011.dat – 1022.dat (Figure
2.3). Source offset of 1.22 m (x1=1.22 m) along with a receiver spacing of 1.22 m (dx=1.22 m)
were used and the source-receivers configuration (SRC) moved by five-station intervals
(dSRC=1.22 dx=1.22 m). A short sampling interval of 1 ms (dt=1 ms) was used because
refraction analysis was also planned. A fairly short recording time of 1 sec (T=1 sec) was used
to avoid inclusion of fairly strong ambient noise as much as possible. To suppress the ambient
traffic noise, three impacts (with about 2-sec interval) were applied at each shot location to save

Figure 2.3 Figure 2.4

one field record of three-vertical-stack.

Step 1−Formatting
Go to ‘Utility → Format’ and then select all 12 records (1011.dat – 1022.dat). Specify output
file name (“Line1.dat”) and then click ‘Run Format’ button.

Step 2−Field Setup

Go to ‘Utility → Field Setup’ and then open the formatted data (“Line1.dat”). Make sure
‘Active’ button is selected in survey type dialog box (Figure 2.4). A graphical dialog box will
appear (Figure 2.5). Type in station numbers for the first two traces of the field record displayed
in the background. Stations numbers are arbitrary fictitious numbers used as reference surface
coordinates. Assign ‘1001’ and ‘1002’ for the first two receivers in the graphics window (Figure
2.5). Type in ‘1.22’ and ‘1.22’ for the source offset (x1) and receiver spacing (dx), respectively.
Choose ‘meter’ for the distance unit. Make sure ‘Source Location’ has been correctly detected
by the program. If not, override it manually. Then, select ‘Source/Receiver Move’ tab and type
in station number (‘1001’) for the next impact location to reflect the interval (‘1.22’) between
current and next impact locations used during the survey. Also, make sure ‘Moving Direction’
has been properly selected. This direction will determine if the acquisition moved to a direction
of higher or lower station numbers. Range of records to which current setup applies can be

7
 Figure 2.5

Figure 2.6 Figure 2.7

specifically selected if necessary in the two drop-down boxes at the bottom of the dialog. If
multiple (rather than one) records were saved at the same place before the source-receivers
configuration (SRC) was moved to the next location, then the number in the ‘Record(s) / Move’
edit box should reflect if appropriate with a number greater than one. Click ‘OK’ button to close
the dialog box and then click ‘Run Field Setup’ to launch the process. At the end of the process,
another window will appear where encoded field geometry is marked (Figure 2.6). Check if all
the records have been correctly encoded. The active data set has now been field-geometry
encoded. This table can be accessed later whenever necessary by clicking the ‘Table of
source/receiver stations’ button on top of the seismic display window (Figure 2.7).

Flexibilities with Field Setup: Although it is recommended to acquire data in a roll-along mode
by maintaining the same source-receiver configuration throughout the entire portion of a survey
line, such deviations as redundant or skipping of recordings at certain places, changing source
offset (x1) and/or receiver spacing (dx), etc., may always occur. Or, sometimes one may attempt
to process surface waves recorded during a body-wave survey (e.g., refraction or reflection
survey). Accounting for most (if not all) of these unusual circumstances can be made by a
proper manipulation of the ‘Record Range To Apply’ (Figure 2.8) and ‘Fixed Receiver Array’
(Figure 2.5 and 2.8) options. The latter option does not move, when checked, the location of
receivers although the source location may do so within a specified record range. Whenever an

Figure 2.8
8
inconsistent field setup is applied, it is always recommended to specify a separate output file on
first setup and then to append subsequent records of different setups to the specified file.

Step 3−Generation of Overtone (OT) Records

See section 5.1 to produce OT records from a field-geometry-encoded seismic file. Note that
once the output file of OT records is ready, you have two options for the remaining steps of
processing to produce a 2-D Vs profile: (1) a sequence of extracting dispersion curves and then
inverting them by following those two steps (Steps 4 and 5) explained below, and (2) inversion
of OT records without extraction of dispersion curves explained in section 6.2.

Step 4−Extraction of Dispersion Curves from Overtone Image

See section 5.2.

Step 5−Inversion for 2-D Vs Profile

See section 6.1.

Chapter 3 Passive Remote MASW

A passive surface wave survey with a 2-D receiver array will give the most accurate evaluation
of dispersion trend (Park and Miller, 2006). This mode of survey, however, requires a wide area
for the 2-D array, which is to be deployed some distance away (remote) from points of surface-
wave generation to meet the plane-wave-propagation assumption. In the case of a roadside
survey, this distance can be a fraction (e.g., 20 %) of the dimension (D) of the receiver array.
Procedures in data acquisition and processing are explained below.

3.1 Receiver Array

Any type of 2-D array of fairly symmetric shape can be used (Figure 3.1). An array of
significant asymmetric shape, for example an elliptical or elongated rectangular shape, is not
recommended due to bias toward a specific direction of incoming surface waves that do not
necessarily coincide with actual direction of major surface-wave energy. Common types may
include circular, cross, square, triangular, random, etc., arrays (Figure 3.1). A detailed study
comparing each different type of array in its effect on the dispersion analysis has not been

Figure 3.1
reported yet, as far as systematic and scientific perspectives are concerned. Intensive modeling
tests performed at the Kansas Geological Survey (KGS), however, indicated an insignificant

9
difference between different types insofar as the symmetry of the array is maintained. It is,
therefore, the convenience of field operation that determines a specific type to be used. We
include sample data sets acquired with two different types of array: cross (“Passive-Cross.dat”)
and circular (“Passive-Circular.dat”) types. Both sets were collected at the same site (but at
different times about one year apart) and are in KGS format (no need to format). Each set
contains ten (10) records of identical recording parameters.

Array Dimension (D) and Receiver Spacing (dx)

Array dimension (D) is directly related to the longest wavelength (λmax) that can be analyzed,
which in turn determines the maximum depth of investigation (zmax):

D ≈ λmax ≈ zmax (3-1)

On the other hand, (minimum if uneven) receiver spacing (dx) is related to the shortest
wavelength (λmin) and therefore the shallowest resolvable depth of investigation (zmin).

dx ≈ λmin ≈ zmin (3-2)

3.2 Data Acquisition

Parameters described here are the ones most commonly used at KGS, and by no means represent
a required set of values. Slight variation in any parameter can always be expected.

Recording Parameters
Please note that the maximum number of samples per trace that can be processed with SurfSeis is
limited to 32000, which is more than enough in most cases as explained below. Sampling inter-
val of 4 ms (dt = 4 ms) and total recording time of 30 sec (T = 30 sec) are most commonly used
at KGS when surveys are performed inside the city of Lawrence. Total recording time (T) is
determined in such a way that there is at least one occurrence of passive surface-wave generation
during recording. Therefore, it can be reduced or increased depending on local situations related
to the surface-wave generation. In case of surveys of nearby roads, for example, there should be
a vehicle passing near the survey area at least once during recording.

The longer T is not always better. This is because the chance of recording surface waves gener-
ated at different locations on the road increases as well and it will in general degrade the data-
processing resolution unless those locations are well apart in azimuth (for example, 90° or more).
If the main source point is fixed in location, however, the longer T will be better. This case of
fixed source point of major surface waves can be noticed when you hear a loud jolting sound
coming from fairly the same spot on the road as a vehicle passes over.

Vertical stacking during recording is strongly discouraged (unless T is significantly limited),

because it will increase the change of recording multi-azimuth surface waves explained pre-
viously. Instead, it is recommended to save individual records separately. Then, the dispersion-
imaging process can be applied to an individual record and then these image records can be
vertically stacked together to enhance the definition of an image. However, if T is limited to
such a short time (e.g., < 30 sec) by the recording device, then a few times of vertical stacking

10
(e.g., stacking 10-sec recording three times) may be necessary either during recording or post-
acquisition stage.

D and dx will approximately determine minimum number of channels needed. In any case, how-
ever, the more channels will be an advantage that can increase the resolution of dispersion pro-
cessing to a certain degree. Forty-eight-channel is most desirable for a survey aiming at zmax ≈
100 m. When only a smaller number of channels are available (e.g., 24), data acquisition can
proceed with an array of smaller dimension (e.g., D = 25 m) and then with progressively larger
dimensions (e.g., D = 50 m, 75 m, 100 m, etc.) to cover a broader range of wavelengths (λ’s). In
this case, dispersion-image data sets processed separately can be combined together to construct
a broader-band dispersion image (see Chapter 7).

If dynamic range is a controllable option with a recording device, choosing the highest value,
along with the highest gain if available also, will always be a benefit.

Necessity of Separate Active Survey

Because a passive survey usually operates with a much larger receiver spacing than normally
used in an active survey, a processed dispersion image usually lacks information at shallower
(higher) depths (frequencies). Although in theory this missing information can be filled through
multiple surveys with progressively smaller dimensions (D’s), higher frequency components of
passive surface waves may not be recorded effectively because of their relatively rapid attenu-
ation properties. The best way would be to perform a separate active survey, preferably at the
array center, by using a sledge hammer. Then, two separate dispersion-image data sets can be
combined together to form a broader-band image.

An alternative approach to replace the need of a separate active survey would be to apply active
impact(s) at a place close to (but outside) the receiver array during the recording of a passive
record. Following constraints, however, are to be considered in this case:

• Analyzed shear-velocity (Vs) information at shallow depths can only be associated with
those near-surface materials near the impact point.
• Analyzed dispersion trend for those high frequencies generated by the active source can
be slightly distorted (at most 10 %) because of violation of the plane-wave assumption.
• Dispersion image for those low frequencies of passive waves can be adversely influenced
resulting in a degraded definition or slightly distorted trend or both.

A more detailed study on this issue has been planned at KGS and will be reported soon. Reports
of experimental results from practitioners are always welcome.

11
 3.3 Data Processing
Overall procedure of processing passive data acquired with a 2D receiver array is explained by
using two sample sets stored in “…\SurfSeis20\SampleData\Passive-Remote\” folder: “Passive-
Circular.dat” and “Passive-Cross.dat.” Both were
acquired at the same site (Figure 3.2) using a 48-
channel cross and a 24-channel circular array,
respectively. Recording parameters are listed in
Table 1.1.

Step 1−Formatting
Ten records were recorded on each survey with
original file names of 2000.dat – 2009.dat for the
cross array and of 3000.dat – 3009.dat for the circu-
lar array, all saved in SEG-2 format. Then, each set
of 10 records was formatted and grouped together
to form a single file of 10 records in KGS format
with the above names. Therefore, both data sets are
in KGS format and don’t need further formatting.

Step 2−Field Setup

Please note that it is the relative coordinate of each
receiver that needs to be specified in the field setup
procedure with a passive data set and that absolute
values of coordinates have no meaning. Go to
‘Utility’ in SurfSeis main menu and then choose
‘Field Setup’ tab. Import either of the two sample
data sets: “Passive-Circular.dat” or “Passive-
Cross.dat.” A dialog box asking for the survey type Figure 3.2
will show up. Choose ‘Passive’ and ‘Remote’
(Figure 3.3).

Then, another dialog box to set up receiver array parameters will

appear (Figure 3.4). Number of channels will show the correct
value detected by the program. Dimension is the approximate
diameter of the array. In the case of circular array, however, type
in the correct value of the actual diameter. Choose ‘circular’
(Passive-Circular.dat) or ‘Non-circular’ (Passive-Cross.dat) for
each data set. There will be some additional parameters to be Figure 3.3
specified in the case of circular array to specify relative location
of each individual receiver along the circle. This specification becomes critical only for the
azimuth information of analyzed passive surface waves (telling which direction passive surface
waves came from).

12
 Once all array-related parameters, except for the location infor-
mation, are specified, then a x-y grid chart window will appear
(Figure 3.5) where receivers of all channels are ready to be man-
ually edited for their correct locations. All receivers will have
been aligned in the upper half of the chart window with ‘Edit
Array’ button pressed (Figure 3.5). To move a receiver of a par-
ticular channel (a numbered circle), click and drag it to the proper
location in the chart. In the case of circular array (Figure 3.6),
they will have been already positioned at the correct locations
according to the previously specified parameters. ‘New Array’
button in the top tool panel will always enable you to go one step
back so that you can change some of the previous parameters.
Figure 3.4 To change chart axis scale and attributes of grid lines, click
‘Controls’ button in the tool panel. If you have an already saved
array, you can import it by clicking ‘Import Array’ button.

To change a receiver location, click and drag it to the proper coordinate.

Once all the receivers are correctly positioned on the chart, you can close the chart and then it
will ask you to save the array. Next step is to specify the reference station number for the center
of the receiver array (Figure 3.7). It can be an arbitrary number of your own invention. It is
used to assign a unique surface coordinate to the inverted 1-D Vs profile that will be obtained at
the end of data processing. ‘Increment (per record)’ applies only when input file contains mul-
tiple records acquired by using exactly the same array but at different sites. In this case, it will
assign a different station num-
ber to each record by incre-
menting a certain number
from the one used for the
previous record. If all those
multiple records in the input
file were acquired at the same
site, it should be zero (‘0’) to
assign the same station num-
ber to all the records. Instead,
if they were acquired at differ-
ent locations by moving the
receiver array, then the incre-
ment should be a non-zero
value indicating different
representative locations. In
this case, another edit box
(‘Record(s) / Move’) will
appear to specify the number
of records saved with the same Figure 3.5
array at one location.

13
Start encoding by clicking ‘Run
Field Setup’ button. The map
of the receiver array encoded in
a data set can be displayed
whenever necessary by clicking
the ‘Show receiver-array map’
button on top of the seismic
display window (Figure 3.8).

Step 3−Generation of Overtone

(OT) Records
See section 5.1 to produce OT
records from a field-geometry-
encoded seismic file. Note that
once the output file of OT
records is ready, you have two
options for the remaining steps
of processing to produce a 2-D Figure 3.6
Vs profile: (1) a sequence of
extracting dispersion curves and
then invert them by following those two steps (Steps 4 and 5)
explained below, and (2) inversion of OT records without extraction
of dispersion curves explained in section 6.2.

Step 4−Extraction of Dispersion Curves from Overtone Image

See section 5.2.

Step 5−Inversion for 2-D Vs Profile

See section 6.1.

Figure 3.8

Figure 3.7

14
 Chapter 4 Passive Roadside MASW
This is a passive mode of survey that can be implemented with a conventional linear receiver
array (Figure 4.1). Although it can result in a certain degree (usually less than 10%) of over-
estimated Vs values in comparison to the 2-D receiver array (see Park and Miller, 2006), this
survey mode can result in a
great utility because of the
convenience in field opera-
tion that does not require a
large open area for receiver
deployment. In fact, the
survey can be repeated by
progressively moving the
receiver array by a certain
(consistent) distance along
the road (the roll-along
survey mode) so that a 2-D
Vs profile can be obtained.
Procedures of data acquisi-
tion and processing are Figure 4.1
explained below.

4.1 Receiver Array

A linear array deployed along the roadside will suffice. Although it does not have to be close to
the road, it is recommended to maintain the offline distance (between the array and the road
center) fairly constant (for example, within ±30 %) throughout the entire survey.

Array Length (D) and Receiver Spacing (dx)

Array dimension (D) is directly related to the longest wavelength (λmax) that can be analyzed,
which in turn determines the maximum depth of investigation (zmax):

D ≈ λmax ≈ zmax (4-1)

On the other hand, (minimum if uneven) receiver spacing (dx) is related to the shortest
wavelength (λmin) and therefore the shallowest resolvable depth of investigation (zmin).

dx ≈ λmin ≈ zmin (4-2)

4.2 Data Acquisition

The following explanation covers most of parameters related to data acquisition. These
parameters are the ones most commonly used at KGS, and by no means represent a required set
of values. Slight variation in any parameter can always be expected.

15
Recording Parameters
Please note that the maximum number of samples per trace that can be processed with SurfSeis is
limited to 32000, which is more than enough in most cases as explained below. Sampling inter-
val of 4 ms (dt = 4 ms) and total recording time of 30 sec (T = 30 sec) are most commonly used
at KGS when surveys are performed inside the city of Lawrence. Total recording time (T) is
determined in such a way that there is at least one occurrence of passive surface-wave generation
during recording. Therefore, it can be reduced or increased depending on local situations related
to the surface-wave generation. In case of surveys near roads, for example, there should be a
vehicle passing near the survey area at least once during recording.

The longer T is not always better. This is because the chance of recording surface waves gener-
ated at different locations on the road increases as well, and it will in general degrade the data-
processing resolution unless those locations are well apart in azimuth (for example, 90° or more).
If the main source point is fixed in location, however, the longer T will be better. This case of
fixed source point of major surface waves can be observed when you hear a loud jolting sound
coming from fairly the same spot on the road as a vehicle passes over.

Vertical stacking during recording is strongly discouraged (unless T is significantly limited)

because it will increase the chance of recording multi-azimuth surface waves explained pre-
viously. Instead, it is recommended to save individual records separately. Then, the dispersion-
imaging process can be applied to individual records and these image records can be vertically
stacked together to enhance the definition of an image. However, if T is limited to such a short
time (e.g., < 30 sec) by the recording device, then a few times of vertical stacking (e.g., stacking
10-sec recording three times) may be necessary either during recording or post-acquisition stage.

D and dx will approximately determine the minimum number of channels needed. In any case,
however, more channels will be an advantage that can increase the resolution of dispersion
processing to a certain degree. Forty-eight-channel is most desirable for a survey aiming at zmax
≈ 100 m. When only a smaller number of channels are available (e.g., 24), data acquisition can
proceed with an array of smaller dimension (e.g., D = 25 m) and then with progressively larger
dimensions (e.g., D = 50 m, 75 m, 100 m, etc.) to cover a broader range of wavelengths (λ’s). In
this case, dispersion-image data sets processed separately can be combined together to construct
a broader-band dispersion image (see Chapter 7). Because of the 1-D nature of the array, usually
24-channel acquisition may be used to aim at zmax=100 m with dx=5 m. A 48-channel acquisi-
tion, however, will be advantageous with an increased resolution in data processing especially at
higher (shallower) frequencies (depths).

If dynamic range is a controllable option with a recording device, choosing the highest value,
along with the highest gain if available also, will always be a benefit.

Necessity of Separate Active Survey

Instead of a separate active survey, use a sledge hammer to apply an impact at a certain (inline)
distance away from one end of the array to trigger a recording of, for example, a 30-sec-long
record (Figure 4.1). Then, take the earliest few seconds (for example, 0-2 seconds) of the record
at the beginning of data processing to make a separate active data set, whereas take the remain-
ing portion (for example, 2-30 seconds) as the passive data set as explained in the Data Process-

16
ing section following. In this case, analyzed Vs information at shallow depth will be biased
towards the impact location.

4.3 Data Processing

The overall procedure of data processing is explained by using a set of sample data stored in
“…\SurfSeis20\SampleData\Passive-Roadside\” folder. There are ten (10) field files (4000.dat –
4009.dat) in SEG-2 format. They were acquired along the Clinton Parkway (Figure 4.1) in
Lawrence, Kansas, by using a land streamer consisting of thirty (30) 4.5-Hz geophones installed
with 1.2-m separations. A 48-channel acquisition system was used to record field records with
4-ms sampling interval. Eighteen (18) traces (out of 48 traces) will therefore have to be dis-
carded at the beginning of processing as explained later. After recording one 120-sec (2 min-
utes) field record triggered by an active hammer impact (Figure 4.1), the streamer was moved by
four stations (4dx) towards the east to be ready for acquiring the next field record. The impact
was applied at four stations (4dx) away inline from the west end of the streamer. Therefore, only
one field record was obtained at a given location of the array and the advantages of acquiring
multiple (for example, ten) field records at the same place, explained in section 3.2, were not
tried at the time. Obviously, the fairly short receiver spacing (dx=1.2 m) with 30-channel acqui-
sition (D≈35 m) was not optimal for a passive survey aiming at deeper (for example, zmax=
100 m) depth of investigation as the land streamer was designed mainly for active surveys. A
longer spacing (for example, dx=3 m) would have been highly preferable.

Step 1– Formatting
Go to ‘Utility → Format’ and then select all ten (10) records (“4000.dat – 4009.dat”) in SEG-2
format. Specify output file name (e.g., “Roadside.dat”) and then click ‘Run Format’ button.

Step 2– Splitting Data into Active and Passive Sets

Because an active impact was applied at the beginning of the
120-sec recording, the early portion of data contains active
surface waves that can be processed by the normal active pro-
cessing procedure explained in 2.2, whereas the later portion
contains the passive surface waves. As processing schemes are
different for active and passive cases, they have to be separated
to avoid adverse influence on each other. Usually the first
2-sec portion is enough to take most of the active surface
waves. In addition, because of 48-channel acquisition was
used along with 30-receiver land streamer, eighteen (18) traces
in each record have to be discarded. These procedures are
explained below.

Go to ‘Display → Seismic’ and then open the file formatted in

the previous step (e.g., “Roadside.dat”). Display first few
seconds portion of data by using the zoom button in the top
tool panel, to estimate possible record length for the active part
(Figure 4.2). Then, click ‘Cut/Append Records’ button in the
tool panel (Figure 4.2). A dialog box will show up. Check ‘All Figure 4.2

17
Records’ box in ‘Record’ tab and then choose ‘Trace’ tab to type in ‘1’ and ‘30’ for beginning
and ending traces to cut (Figure 4.3). Then, choose ‘Time’
tab to type in ‘0’ and ‘2000’ for beginning and ending
times to cut in milliseconds. Click ‘Save Output As’
button at the bottom to type in output file name (e.g.,
“Roadside-Active.dat”). Click OK in the dialog box to
launch the cutting process. Repeat the same procedure
with ‘2000’ and ‘120000’ for the cutting time range of the
passive data set to be saved with another unique file name
(e.g., “Roadside-Passive.dat”).

The ‘Record(s) / Move’ edit box in ‘Record’ tab indicates

the number of records repeatedly saved at the same loca-
tion before the source-receivers configuration (SRC) was
moved to the next location. This option will take effect
only when the ‘Stack Vertically before Output’ option is
checked. In that case, these repeated records will be
stacked together before written into the output file.

Step 3– Field Setup

Go to ‘Utility → Field Setup’ and then open the active
data set first (e.g., “Roadside-Active.dat”). Make sure
‘Active’ button is selected in the survey-type dialog box.
A graphical dialog box will appear (Figure 4.4). Type in
station numbers (e.g., ‘4000’ and ‘4001’) for the first two
traces of the field record displayed in the background.
Stations numbers are arbitrary fictitious numbers used as
reference surface coordinates. However, station numbers
should match those that will be used in the passive data set
and therefore need to be remembered. Type in ‘4.8’ and
‘1.2’ for the source offset (x1) and receiver spacing (dx),
respectively. Choose ‘meter’ for the distance unit. Make
sure ‘Source Location’ has been correctly detected by the
program. If not, override it manually. Then, select
‘Source/Receiver Move’ tab and type in station number
(‘4000’) for the next impact location. Make sure the inter-
val between current and next impact locations matches
actual value (‘4’) used during the survey. Also, make sure
‘Moving Direction’ has been properly selected. This Figure 4.3
direction will determine if the acquisition moved to a
direction of higher or lower station numbers. Range of records to which current setup applies
can be specifically selected if necessary in the two drop-down boxes at the bottom of the dialog.
If multiple (rather than one) records were saved at the same location before the source-receivers
configuration (SRC) was moved to next location, then the number in ‘Record(s) / Move’ edit box
should reflect it appropriately with a number greater than one. Click ‘OK’ button to close the
dialog box and then click ‘Run Field Setup’ to launch the process. At the end of the process,

18
 Figure 4.4
another window will appear where encoded field geometry is marked. Check if all the records
have been correctly encoded. Active data set has now been field-geometry encoded.

Now, open the passive data set (e.g., “Roadside-Passive.dat”).

Make sure to choose ‘Passive’ and ‘Roadside’ in the survey-
type dialog box (Figure 4.5). A different graphical dialog box
will appear (Figure 4.6). Type in station numbers for the first
two traces of the field record displayed in the background.
These numbers, however, should be those values used for the
active data set (Figure 4.4). Type in the channel number of
the first trace (usually ‘1’), receiver spacing (‘1.2’), and then Figure 4.5
select a range in offline offset (‘20-30’), which is an
approximate distance between the array and road center (see ‘Searching Angle’ in section 5.1).
‘Receiver Array Move’ indicates the distance in number of stations that the array moved after
acquiring data at a given location. It can be + or – depending on the direction of its move and the
way stations are defined (for example, whether it increases toward left or right direction) in the
first two yellow edit boxes. This should match ‘Source/Receiver Move’ interval used in the field
setup for the active data set previously encoded. In the case being demonstrated here, it has to be
‘4’. The ‘Record(s) / Move’ edit box indicates the number of records repeatedly saved at the
same location before the source-receivers
configuration (SRC) was moved to the next location.
Choose ‘meter’ for the distance unit.

It is always assumed during the field-setup processing

that a road runs west-east (W-E) direction and that the
receiver array was located at the northern side of the
road. This convention should be noted when interpret-
ing the azimuth information of processed surface
waves as explained in section 5.1.

Once all parameters are set correctly in the graphical

setup dialog, then close it and click ‘Run Field Setup’
Figure 4.6
button at the lower-right corner of the main window to
start encoding field geometry.

19
If multiple records (for example, ten) were acquired at the same array location, then specify the
corresponding record range (e.g., 5001-5010) to apply the same field geometry with zero (‘0’)
interval of array move and then repeat the encoding process with the next set(s) of records (e.g.,
5011-5020) by specifying proper station numbers to reflect the new location for the array.

Step 4 – Generation of Overtone Records

Process active and passive data sets (e.g., “Roadside-Active(OT).dat” and “Roadside-
Passive(OT).dat”) separately by following the procedure described in section 5.1.

Step 5 – Combining Active and Passive Overtone Data Sets

See section 7.2.

Step 6 – Extraction of Dispersion Curves from Overtone Image

See 5.2.

Step 7 – Inversion for 2-D Vs Profile

See Chapter 6.

20
 Chapter 5 Data Processing ⎯ Dispersion Analysis

Once field geometry is encoded, the next step is to process for dispersion information. This
process consists of two steps: (1) generation of dispersion-image, called overtone, data set, and
(2) extraction of dispersion curves from it.

5.1 Step 1−Processing for Overtone (OT) Image Data

Go to ‘Analysis’ in the main menu and then ‘Dispersion’ to open a field-geometry-encoded

seismic file (*.dat). Then, a dialog box will appear (Figure 5.1). In the case of active data, results
of pre-scanning input data will be displayed in a separate box (Figure 5.2). It will show qualita-
tive evaluation of data on possible ranges in surface-wave velocities, frequencies, and on the

Figure 5.1

21
relative degree of higher mode domination. It will be,
however, a rough evaluation and by no means a
complete or definitive one.

The OT generator dialog box (Figure 5.1) will show

input and output information with output file name
assigned by the program. Because most processing
parameters will usually have been set properly by the
program from the pre-scanning process, all you need
to do at this stage will be simply to launch the wave- Figure 5.2
field transformation process by clicking ‘Start
Processing’ button. After the process is finished, you can now move to the next step of either
‘Step 2-Extraction of Dispersion Curves from OT Image’ (section 5.2) or ‘Inversion of Disper-
sion Image’ (section 6.2).

‘Stack Same Mid-Station OT’s’ option (Figure 5.1) will, if checked, vertically stack those
overtone (OT) records of the same mid-station numbers and then the stacked record will be
saved. This option is useful when multiple records were acquired at the same location and saved
separately with different record numbers (e.g., 5001-5010). On importing the input file, the
program scans the input records to see if it contains repeated records and checks the option if
detected. ‘Append Output’ option will, if checked, allow the output to be appended to an exist-
ing file. In this case, it is critical to ensure the same ranges and intervals in frequency and phase
velocity between the current processing and the processing applied to the existing file to be
appended.

Processing parameters influ-

encing the OT data quality can
be accessed by clicking ‘Show
Parameters…’ button in the
dialog box (Figure 5.1). These
parameters are grouped into
four tabbed pages: ‘Frequency
and Phase Velocity,’
‘Searching Angle,’ ‘Contrast,’
‘Records,’ and ‘Algorithm.’
In the case of active data,
‘Searching Angle’ tab will be
missing, whereas the ‘Algo-
rithm’ tab will be missing in
the case of passive data. Each
page is explained below.

Figure 5.3

22
Frequency and Phase Velocity
Ranges and intervals of phase velocity and frequency of the output OT data to be generated can
be specified in this page of dialog. Computation time will directly increase as ranges increase
and/or intervals decrease. In the case of active data, there will be an additional box called ‘Offset
Range to Process.’ Specifying non-zero values for minimum and maximum offset edit boxes
will make the wavefield transformation offset-dependent. For example, MIN = 5 m and MAX =
50 m will include only those data within the specified offset range in the transformation. On the
other hand, MIN = 100% and MAX = 500% will select only those offset ranges within 1-5 times
the wavelengths (λ’s) of calculation in the transformed space. For example, calculation of
transformed energy at 20 Hz for a phase velocity of 200 m/sec (i.e., λ = 10 m) will select offsets
in the 10m – 50m range.

Searching Angle
This page will appear only with passive data sets: passive remote and passive roadside ones.
Searching quadrants can specify the angle range (0-360°, positive clockwise from north) of in-
coming surface waves. Specifying limited quadrants, rather than all four, will make the transfor-
mation focus into certain directions. This can often enhance the image resolution obtained from
the transformation as far as those quadrants are properly selected. For example, if the survey
took place on the west side of a road running south-north (S-N) where major portion of surface
waves were generated, specifying only the two quadrants of NE and SE will be a good choice.
Also, using a smaller number of quadrants will save computation time.

‘Angle Increment’ specifies interval of azimuth scanning and can sometimes influence the image
resolution depending on the relative location of possible multiple points of surface-wave genera-
tion. A smaller interval will increase computation time. The azimuth information (Figure 5.3) of
the surface waves will be saved with a file name of “…(Azimuth).dat” if the option is checked
accordingly.

In the case of passive roadside data, there will be an additional parameter to specify: the offline
distance between the linear receiver array and center of the road from which the major energy of
surface waves came. This value can be precise only within a certain fraction, such as within ±
50%. Zero value (0) will make the transformation be based on the assumption that all the surface
waves recorded came in perfect inline with respect to the array without any azimuthal deviation.
This will therefore make the transformation skip the azimuthal scanning process, and the compu-
tation can proceed at a significantly faster speed to generate a fairly clean image in most cases.
However, the dispersion trend will be overestimated often by 30% or more (see Park and Miller,
2006). It is recommended, therefore, to use this option (zero offline offset) only as a recon-
naissance tool in the field.

Contrast
This controls relative emphasis of one dominant mode of dispersion in comparison to other
modes possibly existing together. The higher contrast will make one dominant mode at each
frequency prominent, whereas the smaller one will enhance other modes if they exist as well.

Records
Specific record(s), rather than all, can be selected for the image data generation.

23
Once overtone data set is generated, you can move either to (1) extraction of dispersion curves
(explained in the next section 5.2) or directly to (2) overtone-image-based (Monte Carlo) inver-
sion (explained in 6.2) to come up with a Vs profile. The extracted dispersion curves will be
used for the linearized-gradient-based inversion to come up with a Vs profile explained in
section 6.1.

Algorithm
This option is available only for the active data. ‘Normal’ algorithm accounts for the relative
phase relationship between seismic data from different receiver locations, and this is the most
reliable method of dispersion imaging with the least number of computational artifacts. The
‘Advanced’ algorithm can achieve a higher resolution in the constructed dispersion image by
accounting for the possible initial phase of surface waves at the time of generation. This method,
however, may result in significant computational artifacts. ‘Evaluation Integrity’ that appears
when the ‘Advanced’ algorithm is selected will determine the searching degree for the phase
relationship of surface waves and will directly increase computational time.

5.2 Step 2−Extraction of Dispersion Curves from OT Image

A fundamental mode (M0) dispersion curve is extracted from each overtone image in the output
file saved in the previous step. The extraction first requires identification of the M0 trend among
multiple modes possibly existing together. It is usually the one starting at lowest frequency with
lowest phase velocity for a given frequency. The lowest (fmin) and highest (fmax) frequencies of
the curve will determine the range of investigation depth (zmax and zmin, respectively) through the
following relation:

zmax ≈ λmax = cmax / fmin and zmin ≈ λmin = cmin / fmax (5-1)

where cmax and cmin are phase velocities at fmin and fmax, respectively. Each extracted curve will
usually have different frequency (depth) range, and the final 2-D Vs profile will show an average
range with a proper spatial interpolation applied.

Go to ‘Analysis → Dispersion’ and then select the overtone (OT) image data previously saved in
Step 1. After display of data attributes, the first overtone record in the input file will be displayed
and a new set of buttons will appear on the left side of the window (Figure 5.3). The task of each
button is explained below in actual order of implementation.

Bounds
This button will enable you to establish lower and upper limits of phase velocities for the disper-
sion curve to be extracted. While this button is down, click several (5-10) reference points on
top of the dispersion trend that you identify as the fundamental mode (M0) (Figure 5.). The first
and last points will determine the frequency range of the dispersion curve. A pair of solid curves
will be drawn as reference points are marked, and they represent lower and upper limits (bounds)
of the phase velocities to be extracted for the curve. If bounds setting was made previously
during a previous running of the program, you can import the saved setting upon clicking the
button.

24
You can edit reference points either by click-and-dragging to move one point or simply clicking
to delete. Bounds can be changed by click-and-dragging one nodal point (marked with ‘*’) in a
bound curve. You can only move it along the vertical phase velocity axis. It is advised not to
make excessively rapid mouse moves during the editing procedure because the mouse response
time varies and the program can sometimes skip certain mouse commands if they are delivered
too fast.

Figure 5.4

Extract
Clicking this button will extract a dispersion curve most likely within the bounds specified in the
previous step (Figure 5.4). You can edit the curve if necessary by clicking the ‘Edit Curve’
button in the top tool panel. In a short summary of how to edit the curve, click a point to delete,
click anywhere between two consecutive points to add a point, click and drag to move, and draw
a rectangle to delete multiple points inside the rectangle. Whenever you find it is necessary to
modify the bounds for a better extraction, you can do so by de-pressing and pressing again the
‘Bounds’ button (clicking twice normally).

Save
This will enable you to save the extracted curve with its own name. Default name will be the
input overtone file name with corresponding record number appended at the end (for example,
“Active(1000).dc”). Once the extracted curve is saved, then the program will show the next
overtone record being ready to repeat the previous steps. In most cases, the bound curves pre-
viously established will be reusable without any modification and you can start from the extrac-
tion step. In the case of data acquired through the roll-along mode (active and passive roadside

25
 Figure 5.5

cases), the station (mid-station of the original field record) will be marked as processed in the
upper station chart after saving. Once all the overtone records are processed and their dispersion
curves are saved, you can move to the next step of inversion analysis by clicking the ‘Inversion’
button. Inversion analysis is explained in Chapter 6.

Controls
Clicking this button will show a separate dialog box where
parameters related to dispersion extraction and some of the
display attributes can be controlled (Figure 5.5). ‘Number of
data points’ determines the total number of points constituting
the extracted curve. It is to be noted that the greater number of
data points does not always lead directly to an increased accur-
acy in the subsequent inversion process. Instead, it will simply
increase the computation time of the inversion process. Any
number between 20 and 30 should be most optimal. ‘Curve
smoothing’ applies a moving-window linear regression to the
preliminary set of data points extracted from every frequency
sample in the overtone data to come up with the most accurate
dispersion trend. The extracted curve displayed is then obtained
through a down-sampling (decimation) of the preliminary curve.
The higher value of curve smoothing will use a larger moving
window and will generate a more smoothed curve. Clicking
‘Equal-wavelength frequency interval’ option will make the
frequency interval vary with frequency (dense at low and coarse Figure 5.6
at high frequencies).

26
‘Save Overtone Record’ and ‘Image’ options in ‘Save and Display’ page will enable you to have
both types of data (one in KGS and the other in JPEG format, respectively) saved separately at
the time of saving dispersion curve. Checking ‘Show All Energy Peaks’ option will display all
energy peaks at each frequency and can be useful to visualize details of hidden dispersion trends.
‘Max Display of Saved Curves’ option specifies the number of the concurrent display of pre-
viously saved curves as the extraction process continues.

Illustration of Curve Extraction with Sample Data

Some practical issues related to the curve extraction are illustrated through a display (Figure 5.7)
of overtone images processed from a sample data set with extracted curves overlapped.

Figure 5.7

27
 2.44 m
Figure 5.7 (continued)

28
 Chapter 6 Data Processing ⎯ Inversion Analysis

Inversion is the last step in data processing with SurfSeis. Either a 1-D or 2-D shear-velocity
(Vs) profile is obtained at the end of the inversion process depending on whether single or mul-
tiple dispersion curves (or overtone records) are processed. Two different types of inversion are
available: the one working on extracted (and saved) dispersion curve(s) and the other working on
the overtone (OT) image. The former type is a gradient-based iterative method by Xia et al.
(1999) and is the one included in previous versions of SurfSeis. On the other hand, the latter
type is a Monte-Carlo method that randomly searches for a layer model whose dispersion curves
best match those trends in the overtone image. This type can therefore perform a multi-modal
inversion and all the parameters (Vs, Vp, thickness, etc.) of five-layer earth model can be man-
ually manipulated. Although the former type has been explained in the previous version (v 1.5)
of the manual (Chapter 4), it is briefly explained again here.

6.1 Inversion of Dispersion Curve

Go to ‘Analysis → Inversion’ and select dispersion-

curve files previously extracted and saved by follow-
ing procedures (Steps 1 and 2) described in Chapter 5.
Once the files are imported, the program will sort
them by an increasing order of mid-station numbers
encoded in each file. Then, click the ‘Run’ button to
launch the inversion process with all default param-
eters of inversion set by the program. Description of
controllable parameters at this stage can be found in
the previous version of the manual (section 4-4).
However, changes made into this version of SurfSeis
are described below.

‘Update (with Vs)’ option in ‘Iteration’ tab of the

control dialog box (Figure 6.1) can specify either P-
velocity (Vp) (default) or Poisson’s ratio (POS) that is
updated whenever Vs is updated on each iteration.
Previously, it was fixed to POS. ‘Update depth (Z)
model with each curve’ option in ‘Initial Vs Layer’
tab will update the thickness-depth part of the layer
model according to the wavelength range (λmin−λmax)
determined from each curve being inverted. Pre-
viously, it was fixed to the average value determined
from all the input curves. Checking ‘Excel File (*.xl)’
option in ‘Output Files’ tab will make an additional
output file (with “XL” extension) where shear
velocities (Vs’s) are displayed along each column
representing a specific surface location. This option
is available only in the case of a fixed-depth model Figure 6.1
(second and third options in ‘Initial Vs Layer’ tab).

29
6.2 Inversion of Dispersion Image (OT Record)

Go to ‘Analysis → Inversion’ and then select the file type

of overtone record (“*.dat”) at the bottom of the open-file
dialog (Figure 6.2). Then, select an overtone file generated
from a seismic data set by following the procedure ex-
plained in Section 5.1. A separate window will show up
where the first overtone record in the file is displayed and
all the inversion controls are displayed in a separate panel
on the right (Figure 6.3). The window type is basically
identical to the one displayed during the step to extract
dispersion curves (section 5.2) only with the inversion panel Figure 6.2
added. Controls in this panel are explained in a grouping in
three tabbed pages: ‘Modeling,’ ‘Inversion,’ and ‘Output’
(Figure 6.3).

Figure 6.3

Modeling
All parameters related to the layered earth model are grouped here in another set of tabbed pages:
thickness (H), shear-velocity (Vs), P-velocity (Vp), and Poisson’s ratio (POS) (Figure 6.4).
Number of layers is limited to five (5) at most (and two at least) and can be changed in ‘# of
layer’ edit box (Figure 6.3). The last layer always represents the layer of infinite thickness (the
half space). All these parameters are used to generated theoretical dispersion curves through the

30
forward modeling scheme
by Schwab and Knopoff
(1972). Dispersion-curve
parameters are grouped in a
separate page of ‘Disper-
sion’ where frequency
range and interval can be
specified for multiple (up to
four) modes (M0, M1, M2,
and M3) (Figure 6.4).
‘Wavelength dependent
increment’ option, when
checked, will make the fre-
quency interval vary in
such a manner that the
interval of a corresponding Figure 6.4
wavelength becomes fairly
constant. It will make the frequency interval denser at lower frequencies and coarser at higher
frequencies. ‘Save dispersion curve files’ option will save, if checked, individual curves sepa-
rately. Thickness-depth of the model can be specified in ‘H’ tab by moving slide bars assigned
to each layer (Figure 6.4). Thickness (H) and depth (Z) of each layer are automatically updated
in the corresponding edit boxes as a slide bar changes its position. An exact value can also be
typed in any edit box. Velocities (Vs and Vp) and Poisson’s ratio (POS) of each layer can also
be specified in the same manner by using slide bars in its own tabbed page. Because all these
three parameters are linked together, their mutual dependency can be objectively set by choosing
either ‘Update Vp’ or ‘Update POS’ radio button (Figure 6.4). Vp will be updated, if ‘Update
Vp’ is chosen, whenever Vs or POS changes, whereas POS will be updated, if ‘Update POS’ is
chosen, whenever Vs or Vp changes. ‘Open’ and ‘Save’ buttons in the top tool panel will enable
you to save the current set of layer parameters in a text file (“*.LYR”) and to import a previously
saved file (“*.LYR”), respectively. A previously saved dispersion curve file (*.dc) can also be
chosen by clicking the ‘Open’ button so that a Vs model can be constructed based on the phase
velocity-wavelength relationship depicted by the dispersion curve.

Clicking the ‘Run Modeling’ button will generate dispersion curves to be displayed on top of the
overtone image (Figure 6.5). Fitting degree of a particular curve with respect to any possible
trend in the background image can be judged from the corresponding signal-to-noise ratio (S/N)
curve in the level of individual frequency. Its average over the entire frequency range is dis-
played in the corresponding colored bar graph at the bottom of the control panel. ‘Save Current
Layer Model’ will enable you to save values of Vs, Vp, and POS parameters into separate text
files (“*.txt”) in the same format used to construct a 2-D Vs map. Specific type(s) of the output
file (e.g., any or all of Vs, Vp, and POS) to be saved will be determined by the checked option(s)
of ‘Output type(s)’ explained in the ‘Output’ section described later. Saving Vs values deter-
mined through a series of forward modeling will provide a manual inversion with complete
control over all the constituent inversion parameters.

31
 Figure 6.5

Inversion
Parameters related to an automatic, iterative, Monte-Carlo inversion method are grouped here
(Figure 6.6). This method of inversion searches for a five-layer earth model whose dispersion
curves best match those trends in the image. It is automatic because the searching is performed
on its own without user’s intervention, iterative because the searching is repeated to update the
model, and a Monte-Carlo method because the searching is a random process. This page of

Figure 6.6

32
controls has another set of two tabbed pages of ‘General Parameters’ and ‘2-D Vs Mapping.’ If
the input overtone file has only a single record, the ‘2-D Vs Mapping’ tab will be missing.

General Parameters: Those parameters related to general aspects of the Monte-Carlo inversion
are grouped into three different tabbed pages: ‘Search Option,’ ‘Layer Model,’ and ‘Matching.’

‘Open Search’ option in ‘Search Options’ tab sets, if selected, allows the searching range of layer
parameters to be wide open, whereas the other option (‘Search from current model in modeling
tab’) uses the parameters specified in ‘Modeling’ tab as an initial model to start from (Figure
6.4). When the latter option is chosen, another set of parameters will be shown. ‘Limit velocity
(Vs) change (%)’ sets bounds (in ±%) in all layer parameters of the current model within which
new values are to be searched. Depth parameters (H and Z) are set to current values and not
updated if ‘Fix depth (Z) model’ option is checked. Maximum number of random searches can
be set in ‘Search Intensity (%)’ box. 100% corresponds to 500 times of trials.

Searching ranges in depth (Z) and velocities (Vs and Vp) can be specified in ‘Layer Model’ tab.
Depth (Z) here means the depth to the top of the half space, which is normally referred to as
maximum depth of investigation (zmax). This option becomes effective only when ‘Fix depth (Z)
model’ option is not selected. Searching range of Vs is specified in the two (MIN and MAX)
edit boxes on the right side of the ‘Velocity (Vs)’ label. Vp and POS are updated accordingly
through the relation set from ‘Update Vp’ or ‘Update POS’ option in ‘Modeling’ tab. ‘Velocity
(Vs) increase with depth (%)’ specifies a relative increasing degree in Vs change with depth.
Zero (‘0’) % will remove such bias to make the Vs changing trend completely random, whereas
100% will make it an absolute increase with depth.

‘Matching’ tab shows options used to evaluate the matching degree between modeled curves and
background image (Figure 6.5). Because a trend of dispersion in the image has relative energy
usually changing with frequencies and phase velocities with a higher energy level associated
with a higher S/N, this relative energy can be accounted for on evaluating the match (if ‘Disper-
sion and relative Energy’ option is selected). On the other hand, the relative energy variation can
be ignored and only the overall trend of dispersion can be considered (if ‘Dispersion only’ option
is selected). In this case, the energy of a particular point on the image is normalized first along
the phase-velocity axis during the evaluation. In the case of multi-modal inversion (specified in
‘Dispersion’ tab in ‘Modeling,’ Figure 6.4), the relative weighting of individual mode can be
specified in ‘Weighting of Dispersion Curves (%)’ box.

2-D Vs Mapping: If the input overtone file contains more than one record, the inversion can be
applied to all records, if ‘All Records’ option is checked, to come up with a 2-D Vs map. A
specific range of records to process can also be set. ‘Random’ option in ‘Starting Layer Model’
radio box will make the inversion of each overtone record completely independent of each other.
On the other hand, ‘Previous model inverted’ will make the subsequent inversion of the next
record start with the inverted layer model of current record as an initial layer model.

Clicking ‘Run Inversion’ button will launch the inversion process. Dispersion curve(s) best
matching (as of current) the trends in the overtone image will be displayed on top of the image
and will be updated if a better match is found. Layer parameters responsible for displayed

33
curves are also reflected in the corresponding tabbed pages in the ‘Modeling’ page whenever a
better match is found.

Output
Layer model parameters (Vs, Vp, and POS) can be saved separately by
checking ‘Save output (*.txt)’ box (Figure 6.7). A file-save dialog will
show up with a default file name (e.g., “Input(OT).txt”) assigned to be
used as a common part of the individual output file. For example, Vs
values will be saved with a file name of “Input(OT)(Vs).txt,” Vp values
with “Input(OT)(Vp).txt,” and so on. Surface coordinates of the output
data can be set in ‘Station # (surface coordinate)’ edit box that will have
been assigned with a value previously encoded in the overtone record.
Specific types of output data (e.g., Vs or Vp) can be selected in ‘Output
type(s)’ box.

Strategy of Inversion: It is recommended to first start with several (or

many) manual attempts to come up with a layer model that gives disper-
sion curve(s) of fairly good match by manipulating individual param-
eters in ‘Modeling.’ Then, choose (perhaps, after saving the layer
model) ‘Start from current model in modeling tab’ in ‘Inversion’ before Figure 6.7
launching the inversion process.

34
 Chapter 7 Combining Active and Passive Overtone (OT) Records
It is often useful or necessary to combine together overtone (OT) records processed from active
and passive data sets for two reasons: (1) to enlarge the analyzable frequency (therefore depth)
range of dispersion (Figure 7.1), and (2) to better identify modal nature of dispersion trends
(Park et al., 2005) (Figure 7.2). Procedures to combine two sets of overtone data are explained
here. Each set may have a different number of constituent records. Or, they may have the same
number of records with an identical record-number (and mid-station) sequence like the two sets
prepared from active (e.g., 0-2 sec) and passive (e.g., 2-120 sec) portions of the passive roadside
data explained in Chapter 4. For seamless combining, it is always recommended (but not
required) to prepare the two sets with the same ranges in frequency (e.g., 5-50 Hz) and phase
velocity (e.g., 10-2000 m/sec) and with the same intervals (e.g., 0.2 Hz and 10 m/sec incre-
ments).

Figure 7.1 Figure 7.2

35
Combining active and
passive overtone records
explained in Chapter 4
by using a sample data
set acquired during a
passive roadside survey
would not provide any
significant gain in
lower-frequency end of
the dispersion when
compared to the active
image. This is because
of the relatively short
receiver array (D≈35 m)
for a passive survey.
Or, it can also be said
that the active source
was effective enough to
generate surface waves
as long as the array
dimension. The impor- Figure 7.4
tance of longer array
dimension for recording
longer (lower) wavelengths (frequencies) of passive surface waves can be illustrated by using
another set of sample data stored as “…\Passive\Roadside\LongerArray.dat” in KGS format. It
consists of ten (10) 120-sec field records acquired by using a 24-channel linear receiver array
with 5-m spacing (dx=5 m) deployed about 5 m north of the survey line used for the acquisition
of passive roadside data (4000.dat – 4009.dat) (Figure 7.3). All ten (10) records were then en-
coded with proper field geometry, and then processed for the overtone records with the “Stack
Same Mid-Station OT’s” option checked (see Figure 5.1). The passive image in Figure 7.1
shows this image, whereas the active image is one of the (ten) active images processed from the
active portion (e.g., “Roadside-Active.dat”) of the data as explained in section 4.3. Users are
encouraged to process the sample data and go through the combining process explained later.

The active image in Figure

7.2 was obtained by pro-
cessing an active record
(“…\Passive\Active-at-
center.dat”) acquired at the
center of the circular array
(Figure 7.4) used during
the survey for the “Passive-
Circular.dat” illustrated in
section 3.3. The active sur-
vey used a 24-channel
receiver array with 0.6-m Figure 7.3

36
receiver spacing (dx) and source offset (x1).

A combining module is included in the top tool panel of the dispersion display window (Figure
7.4). First, open a file of overtone record(s) by using the ‘Open’ button in the tool panel, and
then click the ‘Combine’ button to import another file of overtone record(s). After display of
input data attributes, a control dialog will show up (Figure 7.4). The record number of the
overtone data displayed behind is shown in ‘Current RCD #’ box, whereas that of the data being
imported is shown in the ‘Imported RCD #’ box. Detailed and comparative description of data
attributes can be displayed by clicking the ‘Details…’ button at the lower right side of dialog.
The ‘as they are’ option in ‘Combine’ box will make two records combined as they are without
any pre-conditioning applied, whereas ‘after normalization’ option will apply a certain type of
normalization specified in the next box of ‘Norm. Options.’ The ‘Global’ option of normaliza-
tion will apply the normalization with respect to the maximum value of all the data samples in
one record, whereas ‘Frequency’ option applies with respect to one frequency. Normalization
amplitudes of the two records can be separately specified in ‘Norm. Amplitudes’ box. Mid-
station number of the combined record is specified in ‘Mid-Station #’ edit box with a default
value extracted from the current record. Any other overtone file can be imported, whenever
necessary, by clicking ‘Import Other (OT) File’ button on top of dialog. If an imported file also
has a multiple number of records, then only the record whose record number is the same as that
of the current record will be combined if ‘Combine OT’s of same record number’ option is
checked. Record numbers of current and imported records being combined together are dis-
played on top of the chart. If an imported file has one record, this option is ignored. Combined
record(s) can be saved by clicking ‘Save Combined OT As (*.dat)’ button. Output file will be
appended if ‘Append Output’ option is checked. As long as the ‘Combine’ button remains down,
combining the two records will take place every time a new record is displayed by pressing one
of the four arrow buttons in the tool panel.

7.1 Combining One OT with Multiple OT’s

It is often a practice to run a passive remote survey of a large dimension (e.g., D=100 m) at the
same site where active data have been acquired. The purpose would be to come up with a 1-D
Vs profile representative of a large volume of near-surface materials defined by the large D and
zmax. In this case, there will be ultimately one passive OT of the best quality although multiple
numbers (e.g., ten) of field records may have been acquired. On the other hand, the number of
active overtone records will be many (e.g., 10 or more) representing dispersion characteristics of
shallow materials at different locations (but, within D of the passive survey). In this case, com-
bining the same passive overtone record with all active records for the purpose of selecting a
combined record of most appropriate quality may be tried. First, open the active overtone file by
clicking ‘Open’ button on the top tool panel of dispersion display window (Figure 7.4) and then
click ‘Combine’ button to import the passive file. After options are properly set, then click ‘OK’
button in the control dialog. The imported OT record will be combined with every OT record in
the active file on display as long as the ‘Combine’ button remains down.

37
7.2 Combining Multiple OT’s of Passive Roadside Data
Open the active overtone file prepared from the procedure described in Chapter 4 by clicking
‘Open’ button in the top tool panel of dispersion display window (Figure 7.4). Then, click
‘Combine’ button to import the passive file prepared from the same procedure. A control dialog
will show up (Figure 7.4). Make sure ‘Combine OT’s of same Mid-Station #’s’ option is
checked. ‘Tolerance of station
#’s’ appearing right below this
option indicates the degree of
proximity in terms of station
numbers that can be treated as
the same mid-station. Zero (0)
value indicates an absolute
match without any tolerance in
station numbers between the
two OT records to be com-
bined. Then, click ‘Save Com-
bined OT As (*.dat)’ button to
specify an output file name. A
default file name will have
been assigned. Click ‘OK’
button in the dialog. This will
launch the process of combin-
ing and saving the combined
records as they are being dis- Figure 7.4
played automatically one by
one. If output file was not assigned, then the combining process will occur only whenever a new
record is selected by pressing any of the four arrow buttons in the top tool panel of the display
window.

References
Park, C. B., and Miller, R. D., 2006, Roadside passive MASW: Proceedings of the SAGEEP,
April 2-6, 2006, Seattle, Washington.
Park, C. B., Miller, R. D., Ryden, N., Xia, J., and Ivanov, J., 2005, Combined use of active and
passive surface waves: Journal of Engineering and Environmental Geophysics (JEEG), v. 10,
n. 3, p. 323-334.
Park, C.B., Miller, R.D., Xia, J., and Ivanov, J., 2004, Imaging dispersion curves of passive
surface waves: SEG Expanded Abstracts: Soc. Explor. Geophys., (NSG 1.6), Proceedings in
CD ROM.
Park, C. B., Miller, R. D., and Xia, J., 1999, Multichannel analysis of surface waves (MASW);
Geophysics, v. 64, p. 800-808.
Schwab, F. A., and Knopoff, L., 1972, Fast surface wave and free mode computations in B.A.
Bolt, Ed., Methods in computational physics: Academic Press, p. 87-180.
Xia, J., Miller, R. D., and Park, C. B., 1999, Estimation of near-surface shear-wave velocity by
inversion of Rayleigh waves: Geophysics, v. 64, p. 691-700.

38