Multichannel analysis of surface waves (MASW) for pavement: Feasibility test

Choon B. Park,† Julian Ivanov,† Richard D. Miller,† Jianghai Xia,† and Nils Ryden‡
†
Kansas Geological Survey (park@kgs.ukans.edu), Lawrence, Kansas 66047, USA
‡
Department of Geotechnology, University of Lund, Box 118, S-22 Lund, Sweden

SUMMARY The traditional seismic method for this application, how-

ever, has been rooted on the simplicity of recording and
A feasibility test of the multichannel approach to seismic processing seismic data. Usually only a pair of receivers
investigation of a pavement system is described. This test were used, and it had to be assumed that recorded seismic
followed the procedure normally taken in the multichannel energy would be dominated by one specific type of seismic
analysis of surface waves (MASW) method by using wave such as fundamental-mode Rayleigh wave or direct
geophones and a light (8-oz) hammer source. A wavefield P-wave, depending on the specific method. In general, how-
transformation of recorded multichannel data shows a ever, seismic phenomena are usually more complicated than
strong fundamental-mode dispersion curve image in the usually thought. For example, there are many different
frequency range of 30-600 Hz with normal (30-50 Hz) and types of seismic waves being generated simultaneously at
reverse (50-600 Hz) trends. However, the transformation the time of impact. They can be grouped into body (e.g.,
shows that this fundamental mode disappears quite abruptly direct, refracted, reflected, scattered, guided, and air waves)
and higher modes start to dominate in the higher frequen- and surface (e.g., fundamental, higher modes, scattered)
cies up to 2000 Hz. Simultaneous recording of both vertical waves. Relative dominance of a specific type is a compli-
and horizontal components of seismic wavefields facilitates cated function of many parameters such as seismic source,
identification of seismic events. In order to record the hori- source point condition, impact strength, distance from
zontally travelling direct (or possibly guided) P-wave event source, and the layer model that includes elastic parameters
in the uppermost layer, it seems critical to use horizontal of the medium. Among these, the layer model is a most
phones with longitudinal orientation. Results of test indicate determining factor. It is not only difficult to theoretically
that for an investigation focused into the uppermost layers predict the seismic response based on a given layer model,
of a pavement system it is essential to use a different acqui- but it is also usually the layer model that is to be obtained
sition system that can handle much higher (> 2000 Hz) fre- from a seismic test and therefore is not known prior. Con-
quencies. In addition, complicated and unique elastic prop- sidering all these perturbing factors, any traditional method
erties of pavement systems call for an inter-disciplinary that treats recorded seismic energy being dominated by one
study to develop an effective multichannel seismic method desired type can often become extremely dubious.
for this area of application.
The multichannel recording method is a pattern recognition
KEY WORDS: MASW, pavement, surface wave, body method that enables the identification of different types of
wave seismic waves from their arrival and attenuation patterns.
With its effectiveness being proven by oil exploration
INTRODUCTION industries over the last several decades, it is a method that
assures the most optimum data-acquisition and data-
The main objective of non-destructive tests on a pavement processing parameters. It also enables utilization of the
system is to estimate the modulus and Poisson’s ratio of various multichannel processing techniques available to
each layer. These material parameters, together with thick- make the recognition more effective under diverse situa-
ness of the layers, are the main factors determining status tions.
and elastic response of the construction. With known
external axle loads, fatigue and rutting can be calculated A pavement system is quite challenging for a seismic sur-
(i.e., remaining life of the construction). Both laboratory vey not only because of the shallow (< 5 ft) target depth,
and field tests are usually used to determine these material but also because of the unusual seismic velocity structure.
parameters. Seismic methods have been proven very useful The uppermost layer has both S and P velocities signifi-
for field testing of pavements (Nazarian et al., 1999). These cantly higher than those of the underlying layers and each
days, seismic methods have drawn even more attention layer has a significant velocity contrast to the overlying (or
because of their potential in measuring accurate physical underlying) layer. Total thickness of a pavement system
material properties. From the seismic velocity, the modulus usually does not exceed several feet. Earlier, it was theo-
can be calculated without any empirical relationship. This is retically predicted (Jones, 1962) that this unique layering of
essential for analytical pavement design. elasticity may result in unusual seismic phenomena in phase
velocity and attenuation of surface waves. The unique
velocity structure may complicate surface-wave character-
 Pavement MASW

istics due to interference from strong higher modes and hammer was used as a vertical source. Six different source
body waves (Stokoe et al., 1994; Sheu et al., 1988). The offsets were used to produce a simulated 120-channel shot
dominance of higher mode surface waves has been not only gather per each geophone orientation. A sampling interval
predicted from a theoretical perspective (e.g., Herrmann, of 0.062 ms was used to generate 125-ms seismic records
2000), but also speculated about during the surface wave by using a Geometrics Strataview (60 channel) seismo-
measurement by traditional methods. It can also limit any graph. Horizontal and vertical phones were connected to the
body-wave method in data acquisition and processing. The 1st-20th and 31st-50th channels of the seismograph, respec-
nature of a pavement system being a well-defined layer tively. Figure 2 shows the simulated 120-channel shot
model with a significant velocity contrast indicates a pos- gathers for three different geophone orientations: vertical,
sible dominance of body-wave energy from reflection, longitudinal, and transverse. Horizontal phones were first
refraction, and channel waves. Necessity of unusually high laid out in a longitudinal direction with respect to the re-
(> 2000 Hz) frequency generation and recording requires a ceiver line, and then changed into the transverse direction
special seismic source, receivers, and field logistics. later.

At the Kansas Geological Survey (KGS) we developed a The main purpose of the multicomponent recording was to
unique seismic method called multichannel analysis of sur- ensure an accurate identification of seismic events. Further-
face waves (MASW) (Park et al., 1999) that investigates the more, the thinness and shallowness of the target medium
shallow (< 150 ft) part of the earth by utilizing Rayleigh- raised speculation that a certain type of body-wave event
type surface waves. We recently extend this as a co-project (e.g., direct P-wave event along an asphalt layer) may con-
between KGS and the geotechnical department of Lund sist predominantly of horizontal component in its vibration,
University, Lund, Sweden, to further develop similar whereas others may consist of vertical only (e.g., refraction)
technique to investigate the pavement system. Contents or both (e.g., surface wave) components, depending on the
described on this abstract represent preliminary results from distance from source.
our recent experiments as a feasibility test. The goals are to
describe both seismic surface- and body-wave phenomena The strong vertical (major axis) component of Rayleigh
in the pavement system through the multichannel method wave is best identified on the vertical shot gather in Figure
and to assess limitations with the conventional multichannel 2a obtained from the vertical phones along with the reverse
method. trend of its dispersion. The reverse dispersion is identified
by the curved-down (decreasing apparent velocity) trend of
the surface-wave envelope as offset (apparent frequency)
increases (decreases). At far (> 20 ft) offsets an event is
seen merging out from the surface-wave envelope that has
an apparent velocity of 4500 ft/sec. According to the gen-
eral velocity structure of a pavement system, occurrence of
this refraction is possible only when the seismic P-waves
penetrate below the base layer and merge out to the surface
only at far offsets.

A relatively weak horizontal (minor axis) component of

Rayleigh wave is identified on the longitudinal shot gather
in Figure 2b that was obtained from the longitudinally
oriented horizontal phones. Contamination by traffic noise
Figure 1. Field setup to acquire both vertical and horizon-
from a nearby (about 50 ft away) road is obvious, especially
tal (longitudinal and transverse) components of seismic
at far offsets, due to the relative weakness of the signal.
waves on top of an asphalt layer in a KGS parking lot. [Noise contamination was still obvious even after a low-cut
(< 100 Hz) filter was applied.] The refraction event iden-
MULTICHANNEL MULTI-COMPONENT tified on the vertical shot gather is not seen on this hori-
RECORDING ON ASPHALT zontal shot gather as anticipated. Instead, high-frequency
(> 1000 Hz) first arrivals are seen at near (< 20 ft) offsets
One line of twenty 100-Hz vertical phones and another line with an apparent velocity of about 9000 ft/sec. This event is
of twenty 14-Hz horizontal phones were laid out side by interpreted as a direct P-wave event traveling horizontally
side with 0.3 ft geophone spacing on an asphalt parking lot within the asphalt layer. This event is vaguely seen in the
at KGS (Figure 1). Geophone spikes were replaced by vertical shot gather and the apparent velocity estimation
conical base plates 0.3-ft in diameter. Each line of geo- performed on a high-gain display of this record confirms
phones was topped by three sand bags weighing about 40 lb that they are identical events.
each to enhance the coupling. A small (7 ounce) carpentry
 Pavement MASW

(a) (a)

(b) (b)

(c) (c)

Figure 2. Walkaway records of 120 traces obtained Figure 3. Phase velocity images obtained from near-
by using (a) vertical, (b) longitudinal, and (c) offset (first 60) traces of corresponding walkaway
transverse geophones. records in Figure 3.
 Pavement MASW

The transverse shot gather in Figure 2c shows the weakest event (Park et al., 1998). However, the constant phase
surface wave energy. This confirms the identification of velocity of about 9000 ft/sec can be extracted from this
surface waves on the previous two shot gathers discussed image without a significant difficulty.
above. The weak first-arrival event seen at near-offset
(< 15 ft) traces has an apparent velocity of 9000 ft/sec and The transverse image vaguely displays the surface wave
is interpreted as direct P-wave event. Although motion of dispersion image at lower (< 500 Hz) frequencies. No other
this event should be perpendicular to the geophone orienta- image that can be associated with a coherent event on the
tion, its horizontal nature seems to have been effective in x-t domain is found. Although the first arrivals of 9000 ft/
inducing the transverse component (if minor) inside the sec apparent velocity are seen on the shot gather, the
horizontal phones. corresponding image in the f-Cf domain seems to be lost
due to the strong noise.
PHASE VELOCITY ANALYSIS
BY THE IMAGING METHOD

Each shot gather in Figure 2 was transformed from offset- (a)

time (x-t) domain into frequency-phase velocity (f-Cf)
domain by using a modified version (Park et al., 2000) of
the wavefield transformation method by Park et al. (1998).
This transformation method makes it possible to observe
the frequency-phase velocity relationship of both surface-
and body-wave events. This method does not make any
assumption on the nature of any seismic event in associa-
tion with its phase velocity, and therefore the image con-
struction is performed through an objective pattern-recogni-
tion skill. Because of the severe noise contamination at far
offsets (> 20 ft), only the near-offset traces (first 60 traces)
were used for the transformation and the obtained images
are displayed in Figure 3. Images were obtained for the
frequency and phase velocity ranges up to 2500 Hz and
12000 ft/sec, respectively. Comprehensive interpretation of
x-t records and their corresponding f-Cf images is displayed
in Figure 4.

The f-Cf image (Figure 3a) for the vertical shot gather (b)
shows the dispersion of surface waves most prominently. It
is noticeable that the quality of the image is directly propor-
tional to the quality (i.e., signal-to-noise ratio) of the shot
gather. It is shown in the figure that the fundamental mode
of surface waves takes most of energy up to the frequency
of about 600 Hz, and then higher modes dominate at the
higher frequencies (600-2000 Hz). The refraction event at
far offsets is imaged only when the corresponding far-offset
traces were included in the transformation (Figure 5).

The longitudinal image in Figure 3b shows the same funda-

mental mode observed in the vertical image, but the higher
modes are not so obvious due to the lower signal-to-noise
ratio (S/N) of the shot gather. One thing obvious, however,
is the distinctive image for the direct P-wave first arrivals in
the 1000-1500 Hz range. The resolution of this image is not Figure 4. A comprehensive interpretation of both body-
as high as that of the surface waves because of the higher and surface-wave events (a) on the offset-time (x-t) records
velocity of the event. In order to achieve the comparable in Figure 2, and (b) on the frequency-phase velocity (f-Cf)
resolution for this part of the image, it would require the images in Figure 3.
further offset range in proportion to the ratio (about 9) of
phase velocities between this event and the surface-wave
 Pavement MASW

Figure 6. The first-arrival times of P-wave refractions

superimposed on the vertical record for the velocity (Vp)
model displayed in Figure 7 that was calculated through a
tomographic inversion.
Figure 5. Phase velocity image of far-offset traces (last where the wavelength of the fundamental mode becomes a
60 traces) of the vertical record in Figure 2a. few feet and the corresponding penetrating depth becomes
SURFACE-WAVE ANALYSIS comparable to the depth of subgrade. It seems that this can
be attributed to a theoretical prediction that the fundamental
A fundamental mode dispersion curve was extracted from mode does not exist near the asphalt (or concrete) layer
the vertical image in Figure 3a for the 30-600 Hz range. where seismic velocity decreases extremely with depth but
Because both normal and reverse dispersion trends exist, many higher modes can be generated (Herrmann, 1999;
the curve was broken into two parts: one (30-50 Hz) with Ryden, 1999). This may also be related to another phenom-
normal trend and the other (50-600Hz) with reverse trend. enon called the frequency gap (Jones, 1997). These associa-
The normal trend dispersion curve was then inverted with tions, however, need to be based on further studies carried
an automated inversion algorithm by Xia et al. (1999) to out on several more pavement sites. With the theoretical
generate the S-wave velocity (Vs) profile for a depth range aspect and the experimental observation made during this
of 5-20 ft. The reverse trend curve was then used to study, it is highly possible that the phase velocity calcula-
generate the Vs profile for the shallower (< 5 ft) part by tion by the conventional two-receiver method should yield
using the half-wavelength assumption (Stokoe et al., 1994). the values for the higher modes (not the fundamental mode)
The combined profile is displayed in Figure 7. The dotted for this high (or even higher) frequency range. However, as
portion of the profile at depths (< 3 ft) comparable to the theory tells that all the modes should converge asymptotic-
pavement system was calculated from the extrapolated part
(dotted line) of the dispersion curve indicated in Figure 4b.

BODY-WAVE ANALYSIS

A tomographic inversion of the refraction event on the

vertical shot gather was performed to generate a P-wave
velocity (Vp) profile (Figure 7) for the depth range investi-
gated from the surface wave analysis (Figure 6) (Ivanov
et al., 2000). An initial Vp model prepared from the Vs
profile by assuming a constant Poisson’s ratio (0.45) made
the inversion process converge to the final solution in a fast
and stable manner. This joint inversion also has the
advantage that it minimizes the risk of the non-uniqueness
problem inherent to all types of seismic inversion (Ivanov
et al., 2000).

DISCUSSIONS
Figure 7. S-wave and P-wave velocity profiles obtained
Disappearance of fundamental mode and then dominance of
from the inversion of surface- and body-wave events, re-
higher mode surface waves occurs at those high frequencies
spectively.
 Pavement MASW

ally in their dispersion pattern as frequency becomes very the modes converge in their dispersion pattern at this short
high (e.g., > 5 kHz), the calculated values may not deviate wavelength. It turns out that multi-component recording can
significantly. There is no doubt that a more detailed and be very useful for a reliable identification of complicated
reliable analysis of this phenomenon requires a seismic sur- seismic events in a pavement system.
vey with a frequency range (> 2000 Hz) beyond that
investigated in this study. This sets the limitation with this ACKNOWLEDGMENTS
traditional seismic method using geophones.
We thank Mary Brohammer for her kind and thorough
Although the first-arrival event observed on the horizontal preparation of this manuscript.
(longitudinal and transverse) records was interpreted as
direct P-wave arrivals, its more accurate nature is not yet REFERENCES
clear. It could be a guided P-wave event trapped inside the
uppermost layer(s) or be a combination of direct and guided Herrmann, R., SASW synthetics, retrieved on June 20,
P-wave arrivals. However, if it were the guided-wave event, 2000, from World Wide Web: http://www.eas.slu.edu
then it is not clear why it is missing on the vertical record. /People/RBHerrmann/NILS
Also, reflection from the interface between base and sub- Ivanov, J., Park, C.B., Miller, R.D., and Xia, J., 2000, Map-
grade might have interfered with the aforementioned ping Poisson’s Ratio of unconsolidated materials from a
arrivals. Again, for a more accurate analysis of these body- joint analysis of surface-wave and refraction events: Pro-
wave events it is inevitable that you will move beyond the ceedings of the Symposium on the Application of Geo-
limitation with the traditional method. physics to Engineering and Environmental Problems,
Arlington, Va., February 20-24, 2000.
During the tomographic inversion of the refraction event Jones, R., 1962, Surface wave technique for measuring the
observed on the vertical records, existence of the asphalt elastic properties and thickness of roads: theoretical
layer was ignored because the algorithm was based upon development; Brit. J. Appl. Phys., 13, 21-29.
the calculation of only the first arrivals (Stork and Clayton, Nazarian, S., Yuan, D., and Tandon, V., 1999, Structural
1991). Considering the exceedingly fast P-wave velocity field testing of flexible pavement layers with seismic
and the thinness (< 1 ft) of the asphalt layer, the ignorance methods for quality control; Transportation Research
may not have affected the inversion result of the Vp profile Record 1654, 50-60.
significantly. However, it seems that the severe ray bending Park, C.B., Miller, R. D., and Xia, J., 2000, Multichannel
at the bottom of the asphalt layer due to the excessive analysis of dispersion curves: in preparation for publica-
velocity contrast should have been accounted for during the tion in Geophysics.
calculation of offset-dependent arrival times. Park, C.B., Miller, R.D., and Xia, J., 1999, Multichannel
analysis of surface waves (MASW): Geophysics, 64,
There are a lot more features and issues to be addressed and 800-808.
discussed than those covered in this abstract. We decided, Park, C.B., Miller, R.D., and Xia, J., 1998, Imaging disper-
however, to put it off until we come up with a data set that sion curves of surface waves on multichannel record:
contains seismic events of several to a few tens of kilohertz [Exp. Abs.]: SEG, 1388-1380.
range. Until then, all the attempts would involve excessive Ryden, N., 1999, SASW as a tool for non destructive test-
speculations. We anticipate that a prototype of such an ing of pavements, M.S. Thesis, Lund Institute of Tech-
acquisition system being currently developed at the nology, University of Lund, Lund, Sweden.
geotechnical department of Lund University, Lund, Sheu, J., Stokoe, K., and Roesset, J., 1988, Effect of re-
Sweden, will be ready within the next several months. flected waves in SASW testing of pavements; Transpor-
tation Research Record 1196, 51-61.
CONCLUSIONS Stokoe II, K.H., Wright, G.W., James, A.B., and Jose,
M.R., 1994, Characterization of geotechnical sites by
Higher modes of surface waves dominate at the shallowest SASW method, in Geophysical characterization of sites,
depth range comparable to asphalt (or concrete) and base ISSMFE Technical Committee #10, edited by R.D.
layers. Calculation of phase velocities of the surface waves Woods, Oxford Publishers, New Delhi.
penetrating this superficial part of a pavement system Stork, C. and Clayton, R., 1991, An implementation of
should yield values of higher modes, not the fundamental tomographic velocity analysis: Geophysics, 56, 483-495.
mode. The deviation may not be unacceptably large as all Xia, J., Miller, R.D., and Park, C.B., 1999, Estimation of
near-surface velocity by inversion of Rayleigh wave,
Geophysics, 64, 691-700.