Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

GeoRisk 2011 © ASCE 2011 845

Surface Wave Benchmarking Exercise: Methodologies, Results and Uncertainties

B.R. Cox1, A.M. ASCE and C.M. Wood2, S.M. ASCE


1
University of Arkansas, Assistant Professor, Dept of Civil Engineering, 4190 Bell
Engineering Center Fayetteville, AR 72701, USA, tel: 479-575-6027, brcox@uark.edu
2
University of Arkansas, Graduate Student, Dept of Civil Engineering, 4190 Bell
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

Engineering Center Fayetteville, AR 72701, USA, tel: 479-879-5297, cmwood@uark.edu

ABSTRACT

This paper details the methodologies, results, and uncertainties involved in


analyzing three sets of data for the Surface Wave Benchmarking Symposium. Each set of
surface wave data was analyzed independently (blindly) and without prior knowledge of
subsurface conditions. Both intra-method uncertainty and inter-method variability (bias)
are discussed. SASW dispersion data were obtained using the phase unwrapping method,
and then a full wave-field (effective mode) forward modeling algorithm available in the
software WinSASW was used to determine the shear wave velocity (Vs) profile. MASW
dispersion data were obtained using the frequency domain beamformer method (FDBM),
wherein data from each available source offset distance were combined to form a
composite dispersion curve with estimated uncertainty. A fundamental mode, least
squares inversion was then used to determine the MASW Vs profiles. ReMi dispersion
data were obtained using a slowness-frequency (p-f) transform. A fundamental mode
forward modeling algorithm in the SeisOpt ReMi software was used to determine the Vs
profiles. In general, the Vs profiles from all methods were most similar near the surface
and varied by as much as 30% at depth, yielding Vs30 values that were within 25%.

INTRODUCTION

Surface wave methods have exponentially increased in use over the past decade.
Generally speaking, the three most popular methods are Spectral Analysis of Surface
Waves (SASW), Multi-channel Analysis of Surface Waves (MASW), and Refraction
Microtremor (ReMi). However, there are many different data collection procedures,
dispersion analysis schemes (e.g. phase unwrapping, linear regression of phase angles, f-k
transform, p-f transforms, Park transform, beamformers, weighted linear regression of
phase offsets, etc.), and inversion/forward modeling methods (e.g. stiffness matrix,
propagator matrix, full wavefield, least squares, simulated annealing, neural networks,
etc.) that can be used to obtain a shear wave velocity (Vs) profile for a site (Nazarian
2010). At simple, normally dispersive sites, these varied methodologies often yield
similar dispersion curves and Vs profiles. However, occasionally these methods can have
significant inter-method variability, typically depending on the complexity of the site.
This paper details the methodologies, results, and uncertainties involved in
analyzing three sets of data (SASW, MASW, and ReMi) for the Surface Wave
Benchmarking Symposium. For this exercise, a common researcher (Hiltunen) collected
surface wave data with variable receiver locations from 0 to 400 ft (122 m) along a linear
array at the National Geotechnical Experimental site at Texas A & M University. The
data was then distributed to interested parties in the surface wave community to analyze
using any desired methodology.

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 846

The SASW data was collected from two separate arrays called SASW61 and
SASW128 using the common receiver midpoint geometry. Six receiver spacings from
1.22 m (4 ft) to 37.2 m (122 ft) were used with both forward and reverse source locations.
Both a sledgehammer and a vibroseis were used as sources during SASW testing. The
MASW data was collected from two separate arrays called MASW0-122 and MASW98-
220 using 62 receivers equally spaced at 0.61 m (2 ft). A sledgehammer was used to
generate surface waves at source-to-first receiver offsets of 3.05, 6.10, 9.15, 12.2, 15.25
m (10, 20, 30, 40, and 50 ft). The ReMi data was collected from one array called ReMi0-
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

310 using 32 receivers equally spaced at 3.05 m (10 ft). Twenty six (26) separate time
windows were recorded for ReMi analysis: the first 20 were passive noise records, and
the last 6 were records with a person jogging next to the array providing active energy.

DATA ANALYSIS

The surface wave data from all five array locations (2 for SASW, 2 for MASW,
and 1 for ReMi) were analyzed to obtain dispersion curves and Vs profiles for the site.
Each set of surface wave data was analyzed independently (blindly) and without prior
knowledge of the subsurface conditions. When fitting the experimental dispersion curves,
Poisson’s ratio was assumed to be 0.3 for the entire soil profile, expect for the active
MASW results, where the depth to the water table was identified through P-wave
refraction analysis of the data. The density of individual soil layers was assumed to
increase as subsurface stiffness increased. The maximum depth of the Vs profiles was
always set equal to ½ the maximum resolvable experimental wavelength, and the
thickness of the first layer in each Vs profile was always greater than or equal to ½ the
minimum experimental wavelength resolved for each method.

SASW Results. The SASW data from each array was analyzed using the phase
unwrapping method. Individual experimental dispersion curves for each receiver spacing
were generated using both the forward and reverse phase data. Portions of the phase data
with low coherence, irregular phase (i.e. reverse phase, false cycles), or near-field waves
that had propagated less than ½ a wavelength were “masked” (removed). Composite
dispersion curves (Figures 1a and 2a) were then generated from the phase information
from all receiver spacings for a given array. The experimental dispersion curves from
both SASW arrays have a minimum wavelength of approximately 1 m and a maximum
wavelength of approximately 75 m, with SASW61 having a slightly shorter maximum
wavelength. Overall, each dispersion curve shows normal scatter and good agreement
between individual receiver spacings.
Phase velocities calculated using the phase unwrapping method represent
apparent velocities, which are the superposition of both fundamental and higher mode
surface waves along with body waves (Gucunski and Woods 1992). Consequently, a
fundamental mode solution is not ideal to develop the forward model for SASW data. A
more appropriate solution is the “3-D” forward model in the software WinSASW, which
uses the superposed-mode dynamic stiffness matrix method (Joh 1996). It considers the
location of each receiver relative to the source and uses an “effective mode” solution that
is a superposition of multiple surface wave modes and body waves (Roesset et al. 1991)
to solve for the Vs profile. The Vs profiles for both array locations are very similar
(Figures 1b and 2b), and yield identical Vs30 values of 245 m/s.

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 847

Shear Wave Velocity (m/s)


Wavelength (m)
1 10 100 0 200 400 600
4 6 8 2 4 6 8 2 4 6 8 0 0
Rayleigh Wave Phase Velocity (ft/sec) 1400
400

Rayleigh Wave Phase Velocity (m/s)


1200 5
350 20
Experimental
Theoretical 10
1000 300 40

Depth (m)
Depth (ft)
800 250 15
60
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

200 20
600
150 80 25
400
100
100 30
200 50
35
0 2 4 6 8 2 4 6 8 2 4
0 120
1 10 100 0 1000 2000

Wavelength (ft) Shear Wave Velocity (ft/sec)

Figure 1. Dispersion curve and shear wave velocity profile for SASW61.

Shear Wave Velocity (m/s)


Wavelength (m)
1 10 100 0 200 400 600
4 6 8 2 4 6 8 2 4 6 8
1400 0 0
Rayleigh Wave Phase Velocity (ft/sec)

400
Rayleigh Wave Phase Velocity (m/s)

1200 20
Experimental 350
Theoretical 10
1000 300 40

Depth (m)
Depth (ft)

800 250 60
20
200
600 80
150
400 100 30
100
200 50 120
40
0 2 4 6 8 2 4 6 8 2 4
0 140
1 10 100 0 1000 2000

Wavelength (ft) Shear Wave Velocity (ft/sec)

Figure 2. Dispersion curve and shear wave velocity profile for SASW128.

MASW Results. The active source MASW data from each array was analyzed using the
frequency domain beamformer method (Zywicki 1999). A 3-D dispersion curve was
generated for each source offset location (see Figure 3a) and the maximum spectral peak
for each frequency/wavelength was picked automatically to reduce user bias. The
individual dispersion curves from each source offset were then combined to form a
composite dispersion curve (see Figure 3b) as a means to: (1) identify potential near-field
effects, (2) aid in selecting the fundamental mode of surface wave propagation, and (3)
provide a robust means for estimating dispersion uncertainty. Dispersion points clearly
displaying near field effects, higher modes, or obvious inconsistencies were removed
from the data. However, much of the “normal” dispersion scatter was left intact to
provide a better estimate of the uncertainty in the dispersion data with increasing

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 848

Wavelength (m)
1 10 100
6 8 2 4 6 8 2 4 6 8
1400

Rayleigh Wave Phase Velocity (ft/sec)


400

Rayleigh Wave Phase Velocity (m/s)


Mean with
1200
+/- 1 σ 350
10 ft offset
1000 20 ft offset 300
30 ft offset
800 40 ft offset 250
50 ft offset
200
600
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

150
400
100
200 50
02 4 6 8 2 4 6 8 2 4
0
10 100

Wavelength (ft)

Figure 3. Example MASW 3-D experimental dispersion curve with peak values (left)
and composite dispersion curve showing mean dispersion points with
uncertainty (right).

wavelength. The composite experimental dispersion curve was then divided into 30
wavelength bins using a log distribution. The mean phase velocity and associated
standard deviation was then calculated for each bin, resulting in an experimental
dispersion curve with associated uncertainty for each array (Figures 4a and 5a). Both
curves have a minimum wavelength of just less than 2 m. The standard deviation for the
dispersion data starts to increase significantly for both curves at a wavelength of 30 m.
One may choose to limit the maximum useable experimental wavelength based on
uncertainty, or conversely, choose to accept the increasing uncertainty and account for it
in the forward analysis with a range of Vs profiles.
A fundamental mode, least squares inversion was used to fit the MASW
experimental dispersion curves. In addition to the mean Vs profile, an upper and lower
bound Vs profile were generated (Figures 4b and 5b) by fitting the +/- 1 σ dispersion
curves as a way to account for uncertainty in the Vs profile. The MASW Vs profiles for
both arrays are very similar to a depth of 15 m, and indicate little variability between the
mean and +/- 1 σ profiles. At depths greater than 15 meters, the +/- 1 σ profiles for
MASW0-122 are within 70 m/s (19%) of the mean profile, and indicate the depth to the
half space within 1 m of the mean profile. However, the depth to the half space layer for
MASW98-220 occurs at a greater depth, and varies by approximately +/- 6 m from the
mean profile. Below a depth of 15 m, the velocities of the +/- 1 σ profiles for MASW98-
220 generally fall within 24% of the mean. The Vs30 values for the mean and +/- 1 σ Vs
profiles for MASW0-122 are 258 m/s, 282m/s, and 232 m/s, respectively. The Vs30
values for the mean and +/- 1 σ Vs profiles for MASW98-220 are 241 m/s, 275 m/s, and
201 m/s, respectively.

ReMi Results. The ReMi (termed passive source MASW in the benchmarking study)
data was analyzed using a two dimensional slowness-frequency (p-f) transform in the
software SeisOpt ReMi (Optim 2006). Individual p-f images from each time window
were examined and combined to form an average p-f image. The peak spectral ratio and

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 849

Wavelength (m) Shear Wave Velocity (m/s)


10 100
2 3 4 5 6 2 3 4 5 6 0 200 400 600
1600 0 0
Rayleigh Wave Phase Velocity (ft/sec)

Rayleigh Wave Phase Velocity (m/s)


450
1400 Experimental with +/- 1 σ 20 Median
Theoretical 400
+/- 1 σ 10
1200 +/- 1 σ Theoretical 40
350
1000 300 60

Depth (m)
Depth (ft)
20
800 250 80
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

200 30
600 100
150
400 120
100 40
200 50 140

04 5 6 2 3 4 5 6 2 3 4 5
0 160
10 100 0 1000 2000

Wavelength (ft) Shear Wave Velocity (ft/sec)

Figure 4. Dispersion curve and shear wave velocity profile for MASW0-122.

Shear Wave Velocity (m/s)


Wavelength (m)
10 100 0 200 400 600
2 3 4 5 6 2 3 4 5 6
1600 0 0
Rayleigh Wave Phase Velocity (ft/sec)

Rayleigh Wave Phase Velocity (m/s)

450 20
1400 Experimental with +/- 1σ Median
Theoretical 400 +/- 1 σ 10
1200 +/- 1 σ Theoretical 40
350
60
1000 300 20

Depth (m)
Depth (ft)

250 80
800
200 100 30
600
150 120
400 40
100 140
200 50
160 50
04 5 6 2 3 4 5 6 2 3 4 5
0
10 100 0 1000 2000

Wavelength (ft) Shear Wave Velocity (ft/sec)

Figure 5. Dispersion curve and shear wave velocity profile for MASW98-220.

lower-bound spectral ratio were chosen by eye from two separate, average p-f images.
The first p-f image was the average of time windows 2, 11, and 12, which provided good
low frequency energy. The second was an average of time windows 2,11,12,22,23,24,25,
and 26, which provided good high frequency energy. The dispersion picks from the high
frequency and low frequency images were combined to form composite dispersion curves
for the peak and lower-bound spectral ratios (see Figure 6a). Wavelengths from
approximately 6 - 400 m were resolved for both the peak and lower-bound spectral ratios.
The dispersion data for the peak and lower-bound spectral ratios are very similar at short
wavelengths, but begin to diverge at a wavelength of approximately 40 m.
Each spectral ratio curve was fit using the fundamental mode forward modeling
algorithm in the software SeisOpt ReMi (Figure 6b). The two profiles are similar to a

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 850

Wavelength (m) Shear Wave Velocity (m/s)

1 10 100 0 400 800 1200


4000 0 0
Rayleigh Wave Phase Velocity (ft/sec) 1200

Rayleigh Wave Phase Velocity (m/s)


Lower Bound
Experimental Lower Bound Peak
Spectral Ratio 1000 50
Experimental Peak
3000 Spectral Ratio
20
Theoretical Lower Bound 800
Spectral Ratio

Depth (m)
100

Depth (ft)
Theoretical Peak
Spectral Ratio
2000 600 40
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

150
400
1000
200 60
200

0 0 250
1 10 100 1000 0 2500 5000

Wavelength (ft) Shear Wave Velocity (ft/sec)

Figure 6. Dispersion curve and shear wave velocity profile for ReMi0-310.

depth of 22 m (approximately ½ the wavelength where the dispersion curves begin to


deviate). Both profiles identify a stiffer layer at approximately 22 m, however, the
velocities of this layer vary from 430 m/s to 700 m/s (a difference of 63%) for the lower-
bound and peak spectral ratios, respectively. Additionally, a final “rock” layer was
identified at approximately 60 m deep on both profiles, with velocities ranging from 1006
m/s to 1250 m/s for the lower-bound and peak spectral ratios, respectively. Theoretically,
the Vs profiles could have been extended deeper based on the maximum wavelength
divided by two rule-of-thumb. However, this did not seem prudent for comparison with
active source methods, and furthermore the final layer shown in Figure 6b is the half-
space velocity. The lower-bound and peak spectral ratio Vs profiles have Vs30 values of
259 m/s and 302 m/s, respectively.

INTER-METHOD COMPARISONS

Comparisons between the dispersion curves and shear wave velocity profiles for
each array/method are shown in Figures 7a and 7b, respectively (only the mean MASW
data is presented). In general, all six dispersion curves are very similar in the region of
mutual overlap, with differences in phase velocity and increased uncertainty at
wavelengths greater than 40 m.
The shear wave velocity profiles for each array/method are within 12% of each
other for the top 5 m of the subsurface, expect for the ReMi profile, which lacked short
wavelength data and failed to resolve the 1-m thick surface layer identified on the SASW
and MASW profiles. At depths greater than 5 m, the Vs profiles are generally within 20-
30% of each other (neglecting the peak spectral ratio ReMi profile, which is likely the
result of high apparent phase velocities from oblique wave arrivals). The SASW and
lower-bound ReMi results are most similar, while the greatest differences are between the
SASW and MASW profiles.
In Figure 8a, the +/- 1 σ Vs profiles for MASW98-220 are plotted with the
profiles for all other arrays/methods. For depths greater than 15 m, the MASW standard

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 851

Shear Wave Velocity (m/s)


Wavelength (m)
1 10 100 0 200 400 600 800
6 2 3 4 56 2 3 4 5 6
Rayleigh Wave Phase Velocity (ft/sec) 1600 0 0

Rayleigh Wave Phase Velocity (m/s)


450 SASW61
SASW128
1400 SASW61 MASW0-122
SASW128 400 MASW98-220 10
MASW0-122 ReMi0-310
1200 50 lower bound
MASW98-220 350 ReMi0-310 Peak
ReMi0-310 Lower Bound
1000 ReMi0-310 Peak 300 20

Depth (m)
Depth (ft)
800 250
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

100 30
200
600
150
400 40
100
150
200 50 50
02 3 4 5 6 2 3 4 56 2 3 4 5
0
10 100 0 1000 2000 3000

Wavelength (ft) Shear Wave Velocity (ft/sec)

Figure 7. Dispersion curve and shear wave velocity profile comparison between
methods.

error profiles virtually bracket the other shear wave velocity profiles. For this site, the
MASW +/- 1 σ Vs profiles (intra-method uncertainty) seem to be a good estimate of the
inter-method variability. Modified SASW and ReMi Vs profiles, which were obtained
from forward analyses using the water table determined from P-wave refraction analysis
of the MASW data, are presented in Figure 8b. By using a common (and presumably
correct) water table at the site, the inter-method variability between all mean Vs profiles
is reduced from a maximum of 30% to approximately 10%. These findings reassert the
point that knowledge of the water table can be important to the accuracy of surface wave
methods.

CONCLUSIONS

• Without a-priori information about the site, the Vs profiles from each of the three
methods generally agreed within approximately 20-30% (inter-method variability).
• The inter-method variability was reduced to less than 10% when a common water
table, obtained from P-wave refraction analysis of the MASW data, was used in the
SASW and ReMi forward analyses.
• The subsurface profile of this benchmarking site is one of the simplest types of
profiles (i.e. normally dispersive) to resolve with surface wave methods. Therefore,
the inter-method variability noted herein is more than likely the minimum value one
might expect to obtain. More complex sites with sharp velocity contrasts (i.e. soft-
over-stiff profiles) or inversions will likely yield greater intra-method uncertainty and
hence larger inter-method variability.
• The wavelength/depth range where the MASW data begins to exhibit significant
intra-method uncertainty seems to be a good indicator of the wavelength/depth range
where the results from all methods begin to deviate from one another.
• It is important for surface wave researchers to develop meaningful methods to
quantify intra-method uncertainty in phase velocity estimates. Then, one may choose

Geo-Risk 2011
GeoRisk 2011 © ASCE 2011 852

Shear Wave Velocity (m/s) Shear Wave Velocity (m/s)

0 200 400 600 800 0 200 400 600 800


0 0 0 0
SASW61 SASW61
SASW128 with water table
MASW0-122 10 SASW128 10
MASW220 with water table
50 MASW220 +/- 1σ 50 MASW0-122
ReMiDepth MASW98-220
20 ReMi0-310 20

Depth (m)

Depth (m)
Depth (ft)

Depth (ft)
with water table
Downloaded from ascelibrary.org by DEAKIN UNIVERSITY on 08/09/15. Copyright ASCE. For personal use only; all rights reserved.

100 30 100 30

40 40
150 150
50 50

0 1000 2000 3000 0 1000 2000 3000

Shear Wave Velocity (ft/sec) Shear Wave Velocity (ft/sec)

Figure 8. a) Shear wave velocity profile with +/- 1 σ profiles for MASW220 b) Shear
wave velocity profiles with water table included in SASW and ReMi
forward modeling analysis.

to limit the maximum useable experimental wavelength based on a predetermined


uncertainty; or conversely, choose to accept the increasing uncertainty and account for it
in the forward analysis with a range of Vs profiles.

REFERENCES

Gucunski N., and Woods R.D. (1992). “Numerical simulation of SASW test.” Soil
Dynamics and Earthquake Eng., 11 (4), 213-227.
Joh, S. H., (1996). Advances in interpretation and analysis techniques for spectral-
analysis-of-surface-waves (SASW) measurements. Ph.D. Dissertation, Dept. of
Civil, Architectural, and Environmental Engineering, University of Texas, Austin,
TX, 240 p.
Louie, J. N. (2001). “Faster, better shear wave velocity to 100 meters depth from
refraction microtremor arrays.” Bulletin of Seismological Society of America, 91
(2), 347-364.
Nazarian, S, Hiltunen, D.R., and Rix, G. L. (2010). “Surface Wave Benchmarking
Symposium,” Geo-Institute ASCE. http://saswbench.ce.ufl.edu>(Sep. 29, 2009).
Optim (2006). “User’s Manual: SeisOpt® ReMi™ Version 4.0.” Optim, Inc., Reno,
NV. 85p.
Roesset J.M., Chang D.W., Stokoe K.H. (1991). “Comparison of 2-D and 3-D models for
analysis of surface wave tests.” Proc. 5th Int. Conf. on Soil Dynamic and
Earthquake Engineering, 1, Kalsruhe, 111-126.
Zywicki, D.J. (1999). Advanced signal processing methods applied to engineering
analysis of seismic surface waves. Ph.D. Dissertation, School of Civil and
Environmental Engineering, Georgia Institute of Technology, Atlanta, GA, 357 p.

Geo-Risk 2011

You might also like