SWAN (Surface waves analysis)

This document contains detailed instructions to interpret appended MASW-SASW data through the
software SWAN. Two examples are processed here and through this the user gets acquainted with
most of the tools that SWAN includes.
Together with this document you find two set of sg2 files:
• The folder Sample1 contains the files File1.sg2, File2.sg2, File3.sg2, File4.sg2 and
File5.sg2. The data contained in these files needs to be cleaned before the spectrum
calculation and inversion steps; in this some of the pre-processing tools available in SWAN
can be used.
• The folder Sample2 contains only the file FileTot.sg2. This file is the result from proper
cleaning of the 5 in Sample1. By choosing this file to create a SWAN project the initial pre-
processing step is avoided.

Raw data are related to MASW-SASW acquisition, and they are saved in SEG-2 format.
Acquisitions are done using an array made by 24 geophones, located with geophone step of 2
meters. The source is located at a distance of 3 meters from the first geophone, externally to the
array.

Source
Geophones
1 2 3 4 22 23 24

3m 2m 2m 2m

The sampling interval is 1 millisecond (i.e. the sampling frequency is 1000Hz or 1KHz), and the
length of the acquisition is 2 seconds.
Was acquired 5 shots, with the same configuration, every one saved on a different file.
The files are altered to simulate some of the possible problems that are connected to the acquisition
step:

1
 • In the file File1.sg2 the geophones number 18 and 19 are connected to the instrument with
inverted polarity.
• In the file File1.sg2 the geophone number 23 is not working properly, e.g. due to a bad
connection.
• In the file File2.sg2 the pre-trigger time smaller than in the other files.
• In the file File3.sg2 the geophone array has been reversed.
• In the file File4.sg2 on the geophones number 9 to number 14 a continuous DC component
is added to the signal.
• In the file File5.sg2 there is a strong noise event added to the signal.

Elaboration steps:
To execute the processing of the data using SWAN the following operations should be executed in
the given sequence:
• Create a MASW-SASW project.
• Link the files that contains the raw data to be processed to the created project.
• Pre-process the data for better quality of the following spectrum calculations, dispersion
curve extraction and inversion.
• Transformation of the seismogram to the spectral domain through FK or FV (fVph)
transform.
• Extraction of the experimental dispersion curve from the transformed spectrum.
• Inversion, generation of a Vs model through fitting of this models respons (i.e. the
synthetical dispersion curve) to the experimental dispersion curve.

Create a new MASW-SASW project:

• After that SWAN is installed, open it using the icon on the desktop or through the link in
the start menu of Windows.
• Create a new project through the command Menu>File>New, or through the button on the
toolbar.

• Select the right type of project, in this case press the button: rayleigh MASW-SASW.

Link the data-files to the project:

2
In the following you can find the description of the different procedures in the two different cases

Project using the raw data contained in the folder Sample1

• After that the project type is selected, SWAN will open a dialog window. The five files
File1.sg2 ..File5.sg2 in the folder Sample1 should be selected. The selection is done using
the multi-selection tools in Microsoft, e.g. draw a rectangle with the mouse over the files or
select the files separately while pressing the CTRL key.

• After this a dialog window is opened, where the acquisition parameters should be inserted:
the initial position is the distance from the first geophone to the source, in this case it is 3
meters; the step is the distance between two geophones, in this case 2 meters; and the
sampling interval is in this case 1 millisecond. Then press the OK button to complete the
data import step.

• SWAN automatically detects some of the possible problems the files that has been
generated during acquisition. The file File3.sg2 has been acquired with the channels
numbered in the reversed order and automatically a window appears where the user is asked
if it should be flipped or not. Press the button Yes to specify that is desired that SWAN
corrects the problem.

3
 • SWAN also detects if there is a trig-shift between the separate files and if so an option to
correct this is displayed. The file File2.sg2 it is not well synchronized with the other data,
with an offset about a tenth of second and it is a good idea to correct this. Press the button
Yes present on the shown window.

Project using the raw data contained in the folder Sample2

• In this case you use the same project type (Rayleigh MASW-SASW) and when SWAN
opens the file select dialog window the file FileTot.sg2 from folder Sample2 should be
selected.

4
 • The dialog window for acquisition parameters is displayed. The parameters in this case are
the same as in the first case: initial position (the distance from the first geophone to the
source) is 3 meters; the step (the distance between two geophones) is 2 meters; and the
sampling interval is 1 millisecond. Then press the OK button to complete the data import
step.

Independent of which of the 2 project that is created, SWAN opens a set of windows: one figure to
show the seismogram of every linked file; one figure to show the mean seismogram; and one
window that contains some text that describes the project history i.e. the operations that have been
performed.

Pre-Processing:
This step needs to be done only in the Sample1 project.
• Observing the mean seismogram it is possible to see that some channels are showing a
signal that has got a mean value that is not exactly null, this can disturb the spectral analysis
that need to be done.

5
 If the separate seismograms are viewed it can be seen that this problem is related to the file
File4.sg2. Go on by activating the window with the seismogram related to File4.sg2. Select
the disturbed channels (i.e. from the geophone number 9 to the geophone 14) through the
command Menu>Edit>Select Traces. Then select Menu>PreProcessing>DC Remove, to
specify this operation, and press OK in the window shown to confirm the area selection.

• The strong noise event present late in the last eight traces is related to the file File5.sg2. To
delete this noise select the area of the seismogram between the channel 17 and 34 and
between 800 and 1700 milliseconds by drawing a rectangle with the mouse while pressing
the right mouse button. After the area is selected press the Del key on the keyboard, or use
the command Menu>Edit>Delete.

6
• In the file File1.sg2 it is possible to see that the geophone number 23 hasn’t recorded a good
signal. This channel should be deleted from this seismogram so that the noise is not left to
disturb the analysis. Activate the window relative to that file and select the command
Menu>Pre Processing>Kill Trace. In the dialog window that SWAN opens, specify trace
number 23 and then press OK.

• To magnify a region of the seismogram draw a rectangle using the mouse while pressing the
right mouse button, then select the command Menu>View>Zoom Area or press the
magnifying glass button on the toolbar. To againg view the entire seismogram select the
command Menu>View>Zoom Out or double click with the right mouse button on the
window.
• An interesting area to view is on the seismogram related to the file File1.sg2, between the
geophone 16 and 21 and between 400 and 600 ms. While zooming this area it becomes clear
that geophones 18 and 19 have reversed polarity, they have probably been connected to the
instrument with the connectors in the wrong way.

• To correct this problem activate the seismogram related to File1.sg2, select the command
Menu>Edit>Select Traces and specify geophones 18 to 19. After this select the command
Menu>Pre-Processing>Polarity, and press the OK button to confirm the area to be
processed.

7
 • Finally, the spectral analysis should be limited only to the part of the seismogram containing
useful data. Activate the window related to the mean seismogram and select the area
between 100 and 700 milliseconds, from the first to the last geophone (as displayed in
picture), then select the command Menu>Pre-Processing>Resize, and press OK. This
reduces the active part of the window to that selected which in turn reduces the time it takes
to perform spectral calculations. This operation also removes all noise present outside the
portion of the signal generated by the source.

FK spectrum:
When the mean seismogram is correctly cleaned it is necessary to perform a Fourier transform,
bringing the data to the FK domain. In the FK domain the energy distribution between different
modes of propagation can then be viewed. Before executing this operation the bandwidth in the
signal can be observed:
• Activate the window with the mean seismogram. When this window is activated SWAN
enables the command Menu>Processing>Spectrum (and at the same time it enables the
corresponding buttons on the toolbar). When selecting that command a dialog window
appears where it is possible to indicate the value of the maximum frequency that limits the
spectrum. In this example, and most often in geological conditions, the wanted information
is present below 100Hz. In this case select this value.

8
 • Press the FK Spectrum button to start the processing. After a few seconds the FK spectrum
appears.

Extraction of the experimental dispersion curve:

The MASW-SASW theory teaches that the fundamental dispersion curve must be search on the
maxima of the FK spectrum. In SWAN this can be done automatically or manually. However,
before this step it is necessary to do some simple observations:
Geophones used in surface wave seismic acquisition normally has a cut off frequency of 4.5 Hz,
therefore it is not advisable to work with lower frequency than this. It can be seen here that the
value of the maximum frequency, in this example, where the dispersion curve can be searched is 80
Hz.
Observing the spectrum finally it is possible to see that, in this case, the main part of the energy (red
and yellow color in the seismogram) is limited between 200 and 400m/s (the oblique black lines
that are exiting from the origin of the FK axes each represent a velocity starting with 100 m/s in the
bottom and the increasing with 100 m/s for each line). This area is where the dispersion curve
should be searched.
• Activating the window of the FK spectrum, SWAN will enable the command
Menu>Processing>Dispersion Curve the corresponding buttons on the tool bar.

9
• First the approximate position of the experimental dispersion curve has to be selected. This
is done by drawing a line connected by a few points on the spectrum. Usually it is enough
with 2 or 3 clicks on the image of the spectrum to obtain a result similar to what is shown in
the picture below. Press the button on the toolbar again or press Enter on the keyboard to let
SWAN extract the experimental dispersion curve. A similar image to what is show in the
picture below will appear.

• The points describes the experimental dispersion curve and correspond to the energy
maxima in the spectrum. This curve now need a fast clean before the inversion. Draw a
rectangle, as shown in the picture, to select the points that are not following the dispersion
curve, then press Del on the keyboard to remove this data.

10
• The dispersion curve can be interpolate in the frequency range below 15 Hz where a lot of
scatter is evident. Activate the window with the experimental dispersion curve (in the
domain frequency-velocity), SWAN will then enable the command Menu>Edit>Add
Points and the corresponding button on the toolbar. Select it and add some points to the
dispersion curve with some clicks with the left mouse button. When end disable this tool
pressing again the button on the toolbar.

Inversion:
The inversion aims to find a shear wave velocity profile of the ground that is linked to the
experimental dispersion curve. Basically this is done by comparing the theoretical dispersion
curve from a Vs model with the experimental one. After this the Vs model is updated based on
the difference between the two curves. Experimental data is what you have retrieved from the
processing, i.e. the measured data. Theoretical data is numerically calculated based on the
assumed Vs model.
Initially SWAN creates a start model based on the experimental data. As mentioned above this
model must now be updated so that the experimental and theoretical curves superimpose. The
Vs model can be changed through automatic inversion or through simple manual operations.
The window with the Vs model has to be activated. Then, the velocity and thickness of each
layer in the Vs model can be easily changed by dragging the corresponding line with the mouse

11
pointer while the left mouse button is pressed. Immediately when this operation is done the
theoretical dispersion curve is updated.
Based on scaling properties between the dispersion curve and the Vs model it is also possible to
move the theoretical dispersion curve to search a better fit with the experimental dispersion
curve. This is done by dragging the curve with the mouse pointer keeping the left mouse button
pressed. When letting go of the left mouse button the Vs model is immediately updated. During
these operations the data error (i.e. some measure of the difference between the experimental
and theoretical dispersion curves) can be viewed in the project history window.

• An inversion can be started with the command Menu>Processing>Forward Model>4

Layers or the corresponding button on the toolbar, that is enabled by SWAN when the
experimental dispersion curve window is activated. When doing this a group of windows, as
in the figure below, appears. (Depending on how the dispersion curve is cleaned during the
execution of this sample data processing, the start model can appear a little bit different from
what is showed here. Follow the procedures described before to reach the same start point).

• Use the mouse to reduce the thickness of the third layer to 5 meters (using the procedure
described before) by moving the depth of the fourth layer from 10 meters to about 17
meters.

12
• After that you have created a model similar to the one above you should have a shape on
your theoretical dispersion curve that is similar to the experimental dispersion curve. Now
try to drag the theoretical curve to get a better fit with the experimental curve.

• Finally, increase the fitting between the curves through the execution of an automatic
inversion on the velocity of the layers. This is done through the command
Menu>Processing>Velocity Inversion or by pressing the corresponding button on the
toolbar.
13
• If after this step the data fit is sufficient (normally it can be considered sufficient when you
can’t easily improve it further) the inversion step is finished. Through the command
Menu>Processing>Mean Propagation Velocity (enabled if the window that show the Vs
profile is active), you can now compute the value of the parameter Vs30 (useful index to
classify the site according to the design and engineering standards and regulations). Insert
the values as shown in the picture: the Thickness of the part of the profile to be considered is
30 meters, and the depth of the foundations (Start depth) is located at 0 meters (i.e. on the
ground surface). Press the OK button to retrieve the value of the parameter Vs30.

GeoStudi Astier s.r.l.

eng. Matteo Russo

14