Professional Documents
Culture Documents
www.elsevier.com/locate/soildyn
Abstract
Recent ®eld tests illustrate the accuracy and consistency of calculating near-surface shear (S)-wave velocities using multichannel analysis
of surface waves (MASW). S-wave velocity pro®les (S-wave velocity vs. depth) derived from MASW compared favorably to direct borehole
measurements at sites in Kansas, British Columbia, and Wyoming. Effects of changing the total number of recording channels, sampling
interval, source offset, and receiver spacing on the inverted S-wave velocity were studied at a test site in Lawrence, Kansas. On the average,
the difference between MASW calculated Vs and borehole measured Vs in eight wells along the Fraser River in Vancouver, Canada was less
than 15%. One of the eight wells was a blind test well with the calculated overall difference between MASW and borehole measurements less
than 9%. No systematic differences were observed in derived Vs values from any of the eight test sites. Surface wave analysis performed on
surface data from Wyoming provided S-wave velocities in near-surface materials. Velocity pro®les from MASW were con®rmed by
measurements based on suspension log analysis. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Shear-wave velocity; Surface wave; Borehole measurements; Near-surface materials
relationship is responsible for the dispersive nature of was successfully demonstrated to produce, record, and
surface waves. consistently analyze broad-band ground roll (Fig. 1)
Ground roll (often referred to as pseudo-Rayleigh wave) [16,17]. Multichannel recording proved ideal for effective
is a particular type of surface wave that travels along or identi®cation and isolation of noise/events (body waves,
near the ground surface and is usually characterized by scattered and non-source generated surface waves, and
relatively low velocity, low frequency, and high amplitude, higher-mode Rayleigh waves) using trace-to-trace pattern
(p. 143; [7]). Stokoe and Nazarian [8] and Nazarian et comparison and recognition of coherent waveform arrivals
al. [9] introduce a surface-wave method, called spectral based on amplitude frequency and phase. This inherent
analysis of surface waves (SASW), which focuses on advantage proved instrumental in accomplishing phase
analyzing the ground roll dispersion relationship to pro- two of this research program. Data processing algorithms
duce near-surface S-wave velocity pro®les. SASW has developed as part of phase two allow for ef®cient extraction
been widely applied to many engineering problems over of accurate phase velocities as a function of frequency from
the last couple decades [5,10±14]. SASW has been an fundamental mode Rayleigh waves [18].
effective method of estimating S-wave velocity. However, A robust method for inverting high frequency Rayleigh
a few weaknesses of the SASW method become obvious in wave dispersion data completed as phase three of this
practice. research program provides reliable near-surface S-wave
With a single pair of receivers the SASW method needs to velocities with assumptions limited to the dependence of
con®gure and recon®gure as many times as necessary to surface wave propagation on shear-wave velocity [19,20].
sample the desired frequency range and to reduce body Critical to establishing a relationship between S-wave
wave and/or scattered noise, etc. Empirical criteria [5,13] velocity and the Rayleigh wave dispersion curve is to de®ne
can be used to control the inclusion of noise during SASW quantitatively the sensitivity of Rayleigh wave dispersion
measurements. Because of complexity of the near-surface data to earth properties. Unique to the MASW approach
geology, however, it is dif®cult to optimize these criteria. was effectively accomplishing the relationship between
Besides that, it is also dif®cult sometime impossible to Rayleigh wave phase velocities and earth properties through
assess and distinguish signals (the fundamental and higher analysis of the ®rst partial derivative matrix. High frequency
modes) from noise with only a pair of receivers. Because Rayleigh wave dispersion data is de®ned for a layered
of con®guration of ®eld data acquisition parameters and earth model by the S-wave velocity, P-wave velocity,
multiple shots at each measurement point, it is common to density, and thickness. It has been shown empirically that
take around 1 h to complete measurements at one station. P-wave velocities and densities can be assigned as known
A four-phase research project designed to develop an constants to each layer if the relative error in these selected
ef®cient and accurate method to estimate near-surface S- values is 25% or less [19]. Since the S-wave velocity is
wave velocity from ground roll using multichannel seismic the dominant property of the fundamental mode, S-wave
data was undertaken by the Kansas Geological Survey. An velocities can be quickly estimated from Rayleigh wave
initial objective of the multiphase research project was to dispersion data [19]. In a compressible Gibson half-space,
better utilize the full wave®eld recorded during common S-wave velocity can be more accurately determined by
midpoint (CMP)-style re¯ection pro®ling [15]. The four inverting the Rayleigh wave phase velocity than any other
phases break down as follows: methods [21].
Other advantages of the MASW method include both
1. development of a portable, repeatable, and high energy speed of data acquisition (a few minutes at a measurement
acquisition system and technique designed around point) and resulting time necessary to produce a single 1D
recording broad band (2±100 Hz) ground roll, pro®le as well as the redundancy in measuring dispersive
2. creation of ef®cient, robust, ¯exible, user friendly, and characteristics of surface waves across short distances
accurate set of algorithms organized in a straightforward [22,23]. With a standard CMP roll-along acquisition method
data processing sequence designed to extract and analyze [15] routinely used for acquiring conventional 2D petro-
1D Rayleigh wave dispersion curves, leum exploration seismic re¯ection data, the MASW
3. development of stable and ef®cient inversion algorithms method was successfully used to image the near-surface.
incorporating the very minimum number of assumptions Combining these two methods allows high con®dence,
necessary to obtain 1D near-surface S-wave velocity non-invasive delineation of horizontal and vertical varia-
pro®les, and tions in near-surface material properties [22,23]. Extending
4. construction and validation of the accuracy of the 1D this imaging technology to include lateral variations in
shear-wave velocity pro®les extending the application lithology as well as tunnel and fracture detection, bedrock
into 2D where the subsurface could be imaged using mapping, and delineation of subsidence in karst terrains has
the S-wave velocity ®eld. required a unique approach that incorporates MASW and
CMP methods.
As part of phase one, multichannel analysis of surface MASW is a noninvasive and environmental-friendly
waves (MASW) using Vibroseis and/or an impulsive source method, keenly suited for estimations of shear-wave
J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190 183
Fig. 1. A diagram of the MASW method. Multichannel raw ®eld data, which contain enhanced ground roll signals, are acquired. Rayleigh wave phase velocity
is extracted from the ®eld data in the F±K domain. The phase velocity, ®nally, is inverted for a shear-wave velocity pro®le (Vs vs. depth).
velocity with depth at construction or hazardous material speci®cally for this experiment [19]. An IVI minivib was
sites. MASW can also used in practice in assessing the used to produce a four-octave seismic energy sweep,
liquefaction potential of gravelly soils when penetration recorded with a spread of forty groups of 10 Hz geophones
methods such as standard penetration testing and cone on 1 m intervals starting 2 m from the test well. The source
penetration testing are not feasible. Any penetration of the was offset 27 m from the geophone spread and delivered
earth surface or subsequent layers risks dispersal of soil 10 s linear up-sweeps of seismic energy within a frequency
contaminates and creates a potential new migration path band from 10 to 200 Hz. Raw shot gathers were rich in
for contaminants. ground roll energy (Fig. 2). The range of phase velocities
Economical advantages of the MASW method to present in the dispersive ground roll arrivals extends from
measure S-wave velocity in near-surface material over 220 to 780 m/s. A dispersion curve (Fig. 3a) (Rayleigh wave
other approaches are signi®cant. Time needed in data acqui- phase velocities as a function of frequency) with a frequency
sition at each station for the MASW method is approxi- range from 15 to 80 Hz was extracted from ®eld data
mately a few minutes while time required for the SASW (Fig. 2) using the MASW method.
method is around 1 h. The cost of acquiring borehole veloci- Both P- and S-wave velocity vertical pro®les were co-
ties for a `non-permit' site can be an order of magnitude incidentally acquired using a three-component borehole
greater than using MASW to estimate the S-wave velocity receiver. Con®dent determination of S-wave ®rst break arri-
(several thousand dollars vs. several hundred dollars). vals on the recorded three-component borehole data was
Health, safety, and environmental concerns routinely play accomplished with the aid of a cross-correlation technique
a major role when drilling and run the cost of a borehole [24]. Most error present on the S-wave velocity pro®le (the
measure up signi®cantly. solid line in Fig. 3b) is a result of digital undersampling. An
In this paper, S-wave velocities calculated using surface overall error in S-wave borehole velocity from this survey
wave data acquired and processed using the MASW method has been estimated to be approximately 10%.
will be presented and discussed. Results of this approach Inverting the Rayleigh wave dispersion curve to deter-
will be compared with direct borehole measurements. mine S-wave velocities requires de®nition of densities and
P-wave velocities. Density estimates can be assigned with
relative con®dence based on geology and the assumption
2. Validation of MASW technique
that they increase approximately linearly with depth. In
practice, P-wave velocity can be estimated from surface
MASW-derived S-wave velocity pro®les (S-wave vel-
seismic methods such as refraction with suf®cient accuracy
ocity vs. depth) were compared to direct borehole measure-
for most applications. In this particular case, P-wave veloci-
ments at three North American sites in an attempt to
ties were obtained from borehole data and therefore possess
thoroughly evaluate the technique in a variety of near-
a bit more accuracy than possible when measured from
surface conditions, a wide range of velocities, and through
surface seismic methods.
a relatively diverse set of applications.
An initial S-wave model (labeled `initial B' on Fig. 3)
2.1. Lawrence, Kansas must be created to begin the iterative inversion process. For
most cases a good initial model can be derived using the
Surface wave data, acquired on the campus of the Univer- inverse program (based on Eq. (6), [19]). For this case,
sity of Kansas in Lawrence, Kansas, were analyzed and the root-mean-square error between measured data and the
compared to data collected in a borehole drilled and logged initial model dropped from 70 to 30 m/s with two iterations.
184 J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190
Fig. 2. The ®eld data acquired near the Kansas Geological Survey. Forty groups of 10 Hz geophones were spread 1 m apart. An IVI minivib was the energy
source and located 27 m from the right side of the geophone spread. Two linear events are velocities (220 and 780 m/s) of dispersive ground roll at frequencies
approximately 15 and 50 Hz.
When comparing inverted S-wave velocity and borehole to borehole values from two different directions (Fig. 3b).
measured S-wave velocity, some consideration must be Overall accuracy for all the inverted models were visually
given to the averaging effects associated with the receiver the same. Con®dence in the inverted model can be obtained
spread. The inverted S-wave velocity pro®le is horizontally from the observation that both initial models approach to
averaged across the length of the source-geophone spread borehole results, especially for the shallower part of the
(in this case, 66 m), while in the borehole case the sampling vertical pro®le (,20 m).
is along a vertical segment. Theoretically, considering all Initial values provide critical constraints necessary for
the effects of this averaging there should be only small inversion to converge on a meaningful result. As yet another
differences between inverted and borehole measured S- off-shoot of this study, we tested the robustness of the inver-
wave velocities. From the comparison at this Lawrence, sion technique used for the MASW method. An initial
Kansas, test hole site, the average relative difference model assumed to be a uniform half-space with a constant
between inverted S-wave velocities and borehole measured velocity was inverted using different constant initial values
S-wave velocities is 18%. By excluding the ®rst layer, the that ranged from 100 to 1800 m/s (Fig. 4). The inverted
difference drops to only 9%. results from all these randomly selected (yet realistic) initial
As an ancillary test to analyze the sensitivity of the values converge to borehole results. The total number of
inverted model to initial values, initial values for Vs were iterations necessary to accomplish convergence is less
manually selected to be uniformly greater than or less than than 20 for each case. The limits of the inversion's ability
borehole values (Fig. 3). `Initial A' and `Initial B' are to converge are exceeded for initial values of 50 or
symmetrical to the borehole values and therefore converge 1900 m/s. For these inversions convergence was not
J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190 185
Fig. 3. Inverse results of the real example. Labels on dispersion curves (a) and S-wave velocity pro®les (b) have the same meaning as in labels in Fig. 4 except
that the dispersion curve labeled `measured' (a) is real data extracted from ®led data (Fig. 2) by the MASW method. `Initial B' model (b) was calculated from
the `measured' data in (a). `Borehole' (b) was S-wave velocities derived from the three-component seismic borehole survey. `Initial A' and `initial B' models
(b) are symmetrical to the borehole values. Both initial models converge to the model determined by borehole data. One of every two phase velocities from the
inverted models is shown by diamonds and dots (a).
achieved. In the real world, however, values as extreme as thoroughly sampled for all layers (depths) of interest the
50 or 1900 m/s would not be potential candidates for initial inversion will produce a representative Vs pro®le. Compar-
values when using meaningful constraints on the range of ing the inversion results from data points from 15 to 47 Hz
possible Vs values. (33 points), 15 to 31 Hz (17 points), and 15 to 47 Hz with
Sampling frequency of the dispersion curve (phase vel- interval of 2 Hz (17 points) with inverted results from 66
ocity vs. frequency) can have a signi®cant in¯uence on the data points, no obvious decrease in accuracy can be
performance and quality of inversion results. As long as observed as a result of reducing the number of data points
each unique wavelength component of the surface wave is (Fig. 5). Our testing, as well as the experience of other
186 J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190
Fig. 5. Effect of the number of data points on inversion results. See text for
details.
Fig. 6. (a) Field shot gather with 60 traces at location of borehole FD92-11, Vancouver, Canada; (b) Rayleigh wave phase velocities measured from ®eld data
shown in (a) and calculated based on inverted Vs model shown in (c); (c) inverted Vs model compared to borehole measured velocities.
vertically from source to downhole receiver. S-wave veloci- to relative differences observed between S-wave velocities
ties calculated from MASW, however, are estimated using from borehole and crosshole measurements. This leaves
horizontally propagating energy. Due to this distinctive some question as to what is the best measure of true S-
difference in travel paths of measured energy differences wave velocities in the subsurface.
between MASW results and borehole measurements are The MASW velocity pro®le at the `blind' test site
expected and will be particularly noticeable at hetero- matches borehole measurements extremely well (Fig. 7).
geneous sites. S-wave velocities measured in boreholes The relative difference between the results was only 9%.
can easily vary from measurement technique to measure- This almost insigni®cant difference is primarily the result
ment technique. The sites tested for this study observed of random noise and/or re¯ected ground roll contaminating
differences that are not likely related to heterogeneity portions of the Rayleigh wave signals. Re¯ected ground roll
with the shallow subsurface. In the following paragraphs, has been observed in areas with voids, near vertical faults, or
we will show surface wave data and results from two bore- vertical surface structures extending into the subsurface.
hole sites. Noise of this nature (re¯ected ground roll) suf®ciently
Sixty-channel surface wave data were acquired at well smears the dispersion curve in the F±K domain to reduce
FD92-11 on a 0.6-m geophone interval (Fig. 6a). A disper- linearity of the phase velocity in the time domain. When this
sion curve (Fig. 6b) was extracted from that data (Fig. 6a). smearing occurs, both the accuracy of phase velocity deter-
MASW-derived S-wave velocities and direct borehole minations and the signal-to-noise ratio decrease.
measurements were shown in Fig. 6c. An S-wave velocity An overall difference between the direct borehole
pro®le calculated using crosshole techniques at the site of measurements and inverted S-wave velocities of approxi-
well FD92-11 (Fig. 6c) produces results quite similar to mately 15% or less was observed at the seven well locations.
MASW results at shallower subsurface depths (,15 m) Using blind testing as the ultimate veri®cation of computa-
while MASW matches borehole measurements much better tional accuracy, the 9% differences between S-wave veloci-
at the greater depths (.15 m). It is interesting to note that ties using MASW and velocities measured in the blind
relative differences between S-wave velocities determined test borehole con®rms the legitimacy and accuracy of this
using MASW vs. borehole measurements are quite similar methodology (Fig. 7).
188 J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190
Fig. 7. (a) Field shot gather with 60 traces at location of an unknown borehole at the Vancouver site; (b) Rayleigh wave phase velocities measured from ®eld
data shown in (a) and calculated based on inverted Vs model shown in (c); (c) Inverted Vs model compared to borehole measured velocities. The seismograph
was manually triggered due to interference of an electromagnetic ®eld nearby. First arrivals are not correctly related to source-geophone offsets.
2.3. A mining site, Wyoming multichannel records from opposite ends of the pro®le
verify accuracy of inverted S-wave velocities. Data were
Surface wave data were collected to determine S-wave acquired with forty-eight 8 Hz vertical-component geo-
velocities of near-surface materials (upper 7 m) at a mining phones separated by 0.9 m, with a 1.8 m source-to-nearest-
site in Wyoming (Fig. 8) [28]. Consistency in results using geophone spacing. The source was a 6.3 kg hammer vertically
impacting a metal plate.
Dispersion curves were extracted from surface wave data
(Fig. 8) using the MASW method. Dispersion curves were
inverted using a 10-layer model with layer thickness gradu-
ally increasing from the shallowest at 0.9 m to the deepest at
2.7 m (Fig. 9a). Density of each layer was ®xed at 2.0 g/cm 3
with P-wave velocities of each layer estimated from ®rst
arrival analysis (Fig. 8). Initial S-wave velocities were esti-
mated from the dispersion curve [19]. Inverted S-wave
velocities (Fig. 9a) for these two data sets (collected with
the same receiver stations but source locations on opposite
ends of the spread) (Fig. 8b) possess a relative difference of
about 8%.
To check the accuracy of the MASW method to estimate
S-wave velocity by inverting the dispersion curve, a bore-
hole was installed and suspension logged (Fig. 10). The
inverted S-wave velocities within the top 3 m are very
Fig. 8. Forty-eight-channel surface wave data acquired off both ends of the close to the velocities from suspension logs (Fig. 10). The
W±E line at a mining site in Wyoming. (a) A shot off the east end of the line inverted S-wave velocity linearly increases with depth
and (b) a shot off the west end of the line. below 7 m. In the range of 0±6 m, the average difference
J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190 189
3. Conclusions
Fig. 10. S-wave velocities from a suspension log and MASW inverted S-wave velocities labeled as Area-A W±E (E) and Area-A W±E (W). Labels (E) and
(W) have the same meaning as labels in Fig. 9.
190 J. Xia et al. / Soil Dynamics and Earthquake Engineering 22 (2002) 181±190