Professional Documents
Culture Documents
Horizontal Resolution of Multichannel Analysis of Surface Waves Geo2016-0202.1
Horizontal Resolution of Multichannel Analysis of Surface Waves Geo2016-0202.1
10.1190/GEO2016-0202.1
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Binbin Mi1, Jianghai Xia2, Chao Shen1, Limin Wang1, Yue Hu1, and Feng Cheng1
Manuscript received by the Editor 18 April 2016; revised manuscript received 11 January 2017; published online 20 March 2017; corrected version published
online 27 April 2017.
1
China University of Geosciences, Institute of Geophysics and Geomatics, Subsurface Imaging and Sensing Laboratory, Wuhan, China. E-mail: mibinbin1991@
126.com; geosc@126.com; liminwang_1983@126.com; huyue0716@163.com; mars_cfeng@126.com.
2
Zhejiang University, School of Earth Sciences, Hangzhou, China. E-mail: jianghai_xia@yahoo.com.
© 2017 Society of Exploration Geophysicists. All rights reserved.
EN51
EN52 Mi et al.
method represents the minimum horizontal length of the recogniz- encing factors can be classified into two categories: parameters of
able geologic anomalous bodies on a pseudo-2D S-wave velocity the observation system and properties of the lateral anomalous
section. The increasing popularity of the MASW method has led body. The former include the nearest source-receiver offset, receiver
to significant methodological research and improvements on its res- spacing, receiver spread length (the distance between the first and
olution and accuracy in recent years, with the aim of supplying the last receivers), and source interval, all of which can be adjusted ar-
S-wave velocity distribution in complex structures (e.g., Beaty et al., tificially for the optimum according to the needs of practical inves-
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
2002; Foti and Strobbia, 2002; Beaty and Schmitt, 2003; Forbriger, tigation (e.g., Xia et al., 2004, 2006, 2009; Zhang et al., 2004; Xu
2003a, 2003b; Luke et al., 2003; Socco and Strobbia, 2003; Xia et al., 2006, 2009). The latter include the shape (the extent of the
et al., 2003, 2005, 2006, 2010; O’Neill, 2004; Ryden et al., 2004; detail of vertical and lateral changes), burial depth, and velocity
Zhang et al., 2004; Ivanov et al., 2006a, 2006b; Xu et al., 2006, contrast of the anomalous body, which are determined by a practical
2009; Calderón-Macías and Luke, 2007; Luo et al., 2007, 2008a, subsurface V S model.
2008b, 2009b, 2009c; O’Neill et al., 2008; Boiero and Socco, 2010; Processing parameters influencing the vertical resolution such as
Wang et al., 2012, 2015; Yilmaz and Kocaoglu, 2012; Zeng the model thickness and number of layers (Rix and Leipski, 1991;
et al., 2012). Xia et al., 1999, 2005, 2010) may not be relevant to the horizontal
The traditional surface-wave analysis method is a 1D approach resolution. Park (2005) shows that the receiver spread length used
because the inverted S-wave velocity profiles from surface waves during acquisition of multichannel records most influences the hori-
are based on the assumption of a horizontally layered earth model zontal resolution on the 2D V S map generated by the MASW
(Xia et al., 1999). In 2D environments, this 1D approach usually method. O’Neill et al. (2008) also highlight that the spatial resolution
neglects the presence of lateral variations (Semblat et al., 2005). The of the surface-wave analysis technique is related to the ratio between
resulting model is a simplified description of the site because the the width of the heterogeneity and the receiver spread length. The
surface-wave path crosses different materials (Socco et al., 2010). receiver spread length sets the theoretical lower limit of horizontal
Strobbia and Foti (2006) show that it can cause perturbations on resolution (Park, 2005; Xia et al., 2005; Luo et al., 2008a, 2008b).
the observed phase velocity of surface waves if the wave path is hori- The spatial source interval between two successive records can be a
zontally heterogeneous. As a consequence, artifacts may be intro- multiple-station distance, but it should not be greater than the receiver
duced in spatially 2D S-wave imaging when not accounting for spread length. The smaller interval will be beneficial because redun-
the effects of lateral heterogeneity (Lin and Lin, 2007). In the real dant measurements obtained can increase the resolution through a
world of MASW applications, however, the 1D approach is still statistical principle of the random-noise reduction (Park, 2005).
adopted to investigate lateral variations for processing and inversion, Xia et al. (2005) discuss the resolving power of the MASW technique
and 1D velocity profiles are eventually merged to reconstruct 2D and resolution of the S-wave velocity, and present a lateral unblurring
velocity structures to display lateral variations (e.g., Miller et al., processing technique by generalized inversion to increase the hori-
1999; Xia et al., 2004). In other words, data are processed and in- zontal resolution of a pseudo-2D S-wave velocity section.
verted, disregarding the effect of lateral variations, but the lateral var- Although researchers methodically studied the MASW method
iations are then retrieved and considered in the final interpretation in the late 1990s (Miller et al., 1999; Park et al., 1999; Xia et al.,
(Socco et al., 2010). In this context, it is very important to assess 1999), the horizontal resolution of this technique has not been con-
the errors that could be introduced because of the presence of un- firmed explicitly. In many real near-surface applications, S-wave
known lateral variations. Without understanding the horizontally velocities could vary dramatically along a survey line. It is common
resolving power of MASW techniques, our ability to solve geologic that there exist V S structures whose lateral dimension is shorter than
problems would not be clearly defined (Xia et al., 2005). Therefore, the receiver spread length. Under these circumstances, assessing the
assessing the achievable lateral resolution is one of the main issues in horizontal resolution capability and accuracy of the MASW method
lateral variation reconstruction using the MASW method. is very important. Great effort should be made to evaluate the val-
It is difficult to quantitatively determine the horizontal resolution idity to distinguish events that are laterally displaced from each
of the MASW method because so many factors play a part in lateral other. This work is a continuation of the previous study (Miller et al.,
variation reconstruction on a final 2D V S section. The major influ- 1999; Park et al., 1999; Xia et al., 1999; Park, 2005), and it intends
to further improve the MASW technique. More specifically, differ-
ent from the research on vertical resolution in which we can directly
analyze dispersion curves of Rayleigh waves in theory, effects of the
influencing factors on horizontal resolution of the 2D V S map are
explained quantitatively based on the numerical modeling experi-
ments reported in this paper.
METHODOLOGY
The horizontal resolution of the MASW method is quantitatively
analyzed by a numerical simulation technique using synthetic mod-
els containing geologic anomalous bodies, which is described as
follows (Figures 1 and 2):
1) Establish different laterally heterogeneous models and observa-
Figure 1. A general procedure of the numerical investigation for the tion systems according to different influencing factors for the
research of horizontal resolution of the MASW method. horizontal resolution of the MASW method. There are several
Horizontal resolution of surface waves EN53
papers discussing selections of data acquisition parameters of data acquisition parameters: The nearest source-receiver offset
surface waves (e.g., Forbriger, 2003a; O’Neill, 2004; Xia et al., (represented by A in Figure 2b) is approximately equal to the
2004, 2006, 2009; Zhang et al., 2004; Xu et al., 2006, 2009). maximum investigation depth, a receiver spacing (represented
Xia et al. (2004, 2009) make summaries of selections of optimal by B) is selected as the thinnest layer of the layered model,
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Figure 2. (a) An established lateral-heterogeneous model. (b) Illustration of the observation system for the roll-along acquisition mode in MASW
surveys. (c) One of the synthetic multichannel records computed by a finite-difference method. (d) An image of dispersive energy in the f-v
domain generated from (c) by the phase-shift method (Park et al., 1998) and the picked fundamental dispersion curve marked with dots. The color
scale of the images represents the distribution of the normalized wavefield energy in the f-v domain. (e) One of the inverted S-wave velocity
versus depth profiles. (f) A pseudo-2D S-wave velocity section generated by the MASW method for the established lateral-heterogeneous model.
EN54 Mi et al.
and the receiver spread length (represented by C) is about twice sion. We use an iterative solution to a weighted least-squared
the maximum investigation depth. The spatial source interval be- inversion problem of Rayleigh-wave phase velocities (Xia et al.,
tween two successive records (represented by D) can be a multi- 1999) for S-wave velocity versus depth profiles. Steps 2–4 are
ple-station distance, but the smaller interval will be beneficial for repeated to generate several S-wave velocity profiles when the
complex V S structures (Park, 2005). For generating synthetic seismic data acquisition system is moved along a line.
5) Generate pseudo-2D contour maps of the S-wave velocity in the
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
32, and 40 m receiver spread lengths, respectively, which means that receiver spread should be long enough for the sake of accuracy
the V S differences between the anomalous body and surrounding of the extracted dispersion curves. As the selection principle for op-
rock only reach 27.5%, 24%, 20%, and 16.5% of the V S differences timum receiver spread length (about twice the maximum investiga-
in the true model, respectively. The longer the receiver spread, the tion depth) (e.g., Xia et al., 2004), we choose 32 m receiver spread
smaller the V S differences between the anomalous body and the length for the investigation depth of 15 m and 40 m receiver spread
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
surrounding rock on the 2D V S map, and the lower the horizontal length for the investigation depth of 20 m in the following numerical
resolution of the MASW method. In addition to the decreasing investigations.
V S values of the anomalous body, the anomalous regions on 2D
V S maps become larger with the longer receiver spread. All of the Lateral anomaly lengths
anomalous regions (Figure 4a–4d) are bigger than the real dimen-
sions of the anomalous body. A two-layer earth model with an anomalous bulge is established
We also notice that there are some spurious V S anomalies outside in Figure 6, and the lateral length L of the bulge is set to be 6, 8, 10,
the anomalous regions of the bulge on the pseudo-2D S-wave veloc- and 12 m, respectively. Parameters of the observation system are
ity sections (Figure 4a and 4b). Because surface
waves could be backpropagated due to lateral
velocity variations or backscattered by short-
wavelength heterogeneities (e.g., Yilmaz and
Kocaoglu, 2012; Schwenk et al., 2016), if the
receiver spread is not long enough, dispersion
curves of Rayleigh waves extracted from syn-
thetic shot gathers will be disturbed and inaccu-
rate, which leads to spurious V S anomalies
inverted on the 2D V S maps. Figure 5 shows im-
ages of dispersive energy in the f-v domain with
different receiver spread lengths when the mid-
points of the receiver spreads are moving right
above the center of the anomalous body. Com-
paring the dispersion images in Figure 5, we
can conclude that resolution of the dispersion
image in the f-v domain will increase as the geo-
phone spread length increases, which is consistent
with previous research (e.g., Park et al., 1998;
Forbriger, 2003a; Xia et al., 2006). Hence, the ac-
curacy of the picked dispersion curves and in-
verted V S profiles will improve and there is no
spurious V S anomaly in Figure 4c and 4d. On the
other hand, because of the existence of the anoma-
lous body, the picked dispersion curves (Figure 5)
based on dispersion energy (marked by blue dots)
are different from the theoretical fundamental-
mode dispersion curves of the two-layer earth
model without the anomalous body (marked by
black diamonds, calculated by the Knopoff
method, Schwab and Knopoff, 1972). With the
receiver spread length increasing, the differences
between the picked and theoretical two-layer-
model dispersion curves decline gradually. The
average relative differences drop from 7.3% to
6.0%, 5.5%, and 4.5% for the 16, 24, 32, and Figure 4. Pseudo-2D S-wave velocity sections generated from the earth model in Figure 3
40 m receiver spread lengths, respectively. This using the MASW method with different receiver spread lengths. (a) The parameters of the
drop causes a decrease in the differences between observation system are selected as the nearest offset A ¼ 30 m, receiver spacing B ¼ 1 m,
inverted V S of the anomalous body and the sur- receiver spread length C ¼ 16 m, and source interval D ¼ 2 m. (b) Parameters of the
observation system are selected as the nearest offset A ¼ 30 m, receiver spacing
rounding rock on the 2D V S maps directly, which B ¼ 1 m, receiver spread length C ¼ 24 m, and source interval D ¼ 2 m. (c) Parameters
also means a decrease in the horizontal resolution of the observation system are selected as the nearest offset A ¼ 30 m, receiver spacing
capability. B ¼ 1 m, receiver spread length C ¼ 32 m, and source interval D ¼ 2 m. (d) Parameters
Based on the analysis given above, it is con- of the observation system are selected as the nearest offset A ¼ 30 m, receiver spacing
B ¼ 1 m, receiver spread length C ¼ 40 m, and source interval D ¼ 2 m. Data are col-
cluded that anomalous bodies on pseudo-2D S- lected with sources at lower station numbers, pushing the geophone array forward from
wave velocity sections become blurrier with a left to right. The red rectangles in (a-d) represent the real position of the anomalous body,
longer receiver spread. On the other hand, the respectively.
EN56 Mi et al.
selected as follows: the nearest offset A ¼ 30 m, receiver spacing higher than that of the surrounding rock for the 6, 8, 10, and
B ¼ 1 m, receiver spread length C ¼ 32 m, and source interval 12 m anomaly lengths, respectively, which means that the V S
D ¼ 2 m. The data acquisition and inversion procedure is the same differences between the anomalous body and the surrounding
as the data acquired and inverted in previous investigations with rock reach 15%, 20%, 25%, and 30% of the V S differences in
different receiver spread lengths. Four pseudo-2D S-wave velocity the true model, respectively. The longer the lateral anomalous body,
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
sections are generated (Figure 7). the greater the V S differences between the anomalous body and the
Comparing the 2D V S maps in Figure 7, we notice that both surrounding rock on the 2D V S maps, which indicates that the
of the anomalous regions and maximum V S values of the anoma- anomalous body is resolved with greater accuracy as its length-
lous bodies increase gradually with the increasing lateral anomaly to-depth ratio increases. In addition, all the anomalous regions
lengths in the models. On the generated 2D V S maps, V S values of (Figure 7a–7d) are bigger than the real dimensions of the anomalous
the anomalous body are approximately 30, 40, 50, and 60 m∕s body.
V S contrast
A two-layer earth model with an anomalous bulge is established
in Figure 12 and the V S of the bulge increases from 300 to 400, and
500 m∕s, respectively. Meanwhile, the V P of the bulge increases
from 1000 to 1200, and 1400 m∕s, respectively. Parameters of
the observation system are selected as follows: the nearest offset
A ¼ 30 m, receiver spacing B ¼ 1 m, receiver spread length C ¼
32 m, and source interval D ¼ 2 m. With the same data acquisition
and inversion procedure as discussed in previous investigations with
different receiver spread lengths, three pseudo-2D S-wave velocity Figure 8. Illustration of a two-layer earth model containing an
sections are generated (Figure 13). anomalous bulge with various thicknesses. The V P , V S , and ρ of
Comparing the 2D V S maps in Figure 13, we notice that the the first and second layers are the same as the model in Figure 3.
anomalous regions and maximum V S values of the anomalous body The thickness h of the first layer is 10 m, and the second layer is the
half-space. There is a rectangular bulge in the first layer with V P ,
increase with the increasing V S in the model. On the generated 2D V S , and ρ being the same as the second layer. The lateral length L of
V S maps, the V S values of the anomalous body are approximately the bulge is 10 m. The thickness H of the bulge is set to be 3, 5, and
30, 50, and 60 m∕s higher than that of the surrounding rock for 300, 7 m, respectively.
EN58 Mi et al.
400, and 500 m∕s S-wave velocities of the anomalous body in the maximum V S value in the merged anomalous region is higher than
model, respectively, which means that the V S differences between the V S value when there is only one anomalous body in the model.
the anomalous body and the surrounding rocks reach only 30%, With the greater distance between the two anomalous bodies (we
25%, and 20% of the V S differences in the true model, respectively. will discuss this point later), they will be distinguished from each
The larger the V S contrast in the anomalous body, the greater the V S other on the pseudo-2D S-wave velocity section (Figure 15c), and
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
differences between the anomalous body and the surrounding rocks the anomalous regions and maximum V S values for each anomaly
on the 2D V S map, and hence the easier the anomalous body may be are almost the same as those when there is only one anomalous body
resolved. With the increasing V S contrast of the anomalous body in in the model (compared with Figure 13b).
the model, the percentage of V S differences decreases gradually. In
addition, all of the anomalous regions (Figure 13a–13c) are bigger
than the real dimensions of the anomalous body.
receiver spread length C ¼ 32 m, and source interval D ¼ 2 m. with 48 4.5 Hz vertical component geophones separated by 1.2 m
The data acquisition and inversion procedure is the same as the data (the receiver spread length was 56.4 m). The nearest offset was 12 m
acquired and inverted in previous investigations with different in hopes of mapping up to 30 m of the subsurface and the source
receiver spread lengths. Four pseudo-2D S-wave velocity sections interval was 2.4 m. Data were acquired from both directions along
are generated (Figure 17).
Comparing the 2D V S maps in Figure 17, we notice that the mini-
mum V S values of the anomalous graben decrease gradually with
the increasing lateral anomaly lengths in the models. The longer the
lateral anomalous graben, the greater the V S differences between the
anomalous graben and surrounding rock on the 2D V S maps, and
hence the easier the anomalous graben may be resolved. Also, it is
noteworthy that spurious high V S anomalies appear in the upper part
of the graben on the 2D V S maps (Figure 17c and 17d) for the low
V S anomaly detection.
the dirt road (Xia et al., 2005). A total of 40 shots were collected in different receiver spread lengths, if only detecting sinkholes at a
each direction by use of the roll-along acquisition mode (Mayne, depth of 8 m, the receiver spread length can be reduced to less than
1962). Two pseudo-2D S-wave velocity sections were generated 20 m (about twice the maximum investigation depth) for the higher
by the technique of MASW (Figure 18a and 18b). The two striking horizontal resolution. Hence, the sinkholes with 12 m lateral length
low-velocity “bull’s-eyes” located around stations 3057 and 3095 at will be well-imaged on the pseudo-2D S-wave velocity section.
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
a depth of 8 m are associated with nearby sinkholes. We also notice that it shows anomalously high V S anomalies at
Xia et al. (2005) present a lateral unblurring processing technique shallow depths above the sinkholes (Figure 18). Miller and Xia
by generalized inversion to increase the horizontal resolution of the (1999) explain this phenomenon as high V S anomalies at shallow
two pseudo-2D S-wave velocity sections (Figure 18c and 18d). If depths being due to the higher stress above the voids caused by the
we consider the postunblurring V S maps as the true subsurface voids, and low V S anomalies at deep depths are due to the voids
model, the horizontal resolution capability of the traditional MASW themselves. These high V S anomalies are consistent with the numeri-
method can be analyzed by comparing the pre- and postunblurring cal investigation results that spurious high V S anomalies may appear
V S maps. On the postunblurring V S maps (Figure 18c and 18d), the in the upper part of the sinkhole on the 2D V S maps (Figure 17c and
lateral length of sinkholes located around stations 3057 and 3095 at 17d), located above the zone of detection of the low V S anomaly.
a depth of 8 m is approximately 12 m. However, on the preunblur-
ring V S maps (Figure 18a and 18b), the lateral length of the sink- Field data 2
holes is longer and sinkholes become blurred because of the average
within a horizontal distance of 56.4 m (the receiver spread length) at The second field data example was acquired beside the Guangle
each station. Like the previous numerical investigation results with Highway in the southern China, in 2015. There is a landslide nearby
the highway (Figure 19), and it was grouted with cement paste by surveys at different times. It is noticed that the anomalous regions
steel pipes to consolidate the subsoil with the aim of stopping the at locations 800–810 m (the lateral length is approximately 10 m)
landslide from sliding. Surface-wave data were collected on the site and 820–835 m (the lateral length is approximately 15 m) at the
of the landslide before (in June) and after (in August) the injection, depth of 7–10 m with V S increasing more than 40 m∕s are clearly
intending to evaluate the effects of the injection process. This can be identified using the MASW method. This is consistent with the
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
considered as a time-lapse MASW survey. Data were recorded with numerical investigation results, with different burial depths, that
twelve 4.5 Hz vertical component geophones separated by 1 m (the the minimum horizontal length of recognizable geologic anomalous
receiver spread length was only 11 m, Figure 19). The nearest offset bodies at a certain depth is approximately equal to the depth.
was 6 m, and the source interval was 8 m. A total of seven shots
were collected for each survey using the roll-along acquisition
mode, and two pseudo-2D S-wave velocity sections were generated
(Figure 20a and 20b).
On the V S map generated before the injection process (Fig-
ure 20a), we can notice the low-velocity areas at locations 800–
805 and 820–830 m at the depth of 5–10 m. After the injection
process (Figure 20b), S-wave velocities of the two areas become
higher and almost correspond with that of the surrounding rocks.
According to the in situ borehole measurements, the upper part with Figure 16. Illustration of a two-layer earth model containing an
a V S of less than 230 m∕s is associated with mixtures of colluvium anomalous graben with various lateral lengths. The V P , V S , and
and clay, having a thickness of approximately 3–5 m. The weath- ρ of the first and second layers are the same as the model in Figure 3.
ered siltstone is under the colluvium with a V S of more than The thickness h of the first layer is 5 m, and the second layer is the
half-space. The rectangular graben in the second layer possesses the
250 m∕s. We calculate the changes of V S before and after the in- same V P , V S , and ρ as the first layer. The thickness H of the graben
jection process (Figure 20c), and V S changes less than 40 m∕s are is 5 m. The lateral length L of the graben is set to be 6, 8, 10, and
ignored in consideration of the 15% maximum errors in the two 12 m, respectively.
Figure 15. Pseudo-2D S-wave velocity sections generated from the earth model containing two anomalous bodies with different distances d in
Figure 14 using the MASW method, (a) d ¼ 10, (b) 15, and (c) 20 m. The red rectangles in (a-c) represent the real positions of the anomalous
bodies, respectively.
EN62 Mi et al.
REFERENCES
Babuska, V., and M. Cara, 1991, Seismic anisotropy in the earth: Academic
Publishers.
Beaty, K. S., and D. R. Schmitt, 2003, Repeatability of multimode Rayleigh-
wave dispersion studies: Geophysics, 68, 782–790, doi: 10.1190/1
.1581031.
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
O’Neill, A., T. Campbell, and T. Matsuoka, 2008, Lateral resolution and Xia, J., C. Chen, P. H. Li, and M. J. Lewis, 2004, Delineation of a collapse
lithological interpretation of surface-wave profiling: The Leading Edge, feature in a noisy environment using a multichannel surface wave tech-
27, 1550–1563, doi: 10.1190/1.3011028. nique: Geotechnique, 54, 17–27, doi: 10.1680/geot.2004.54.1.17.
Pan, Y., J. Xia, and C. Zeng, 2013, Verification of correctness of using real Xia, J., C. Chen, G. Tian, R. D. Miller, and J. Ivanov, 2005, Resolution of
part of complex root as Rayleigh-wave phase velocity by synthetic data: high-frequency Rayleigh-wave data: Journal of Environmental and Engi-
Journal of Applied Geophysics, 88, 94–100, doi: 10.1016/j.jappgeo.2012 neering Geophysics, 10, 99–110, doi: 10.2113/JEEG10.2.99.
.09.012. Xia, J., R. D. Miller, and C. B. Park, 1999, Estimation of near-surface shear-
Park, C. B., 2005, MASW horizontal resolution in 2D shear-velocity (Vs) wave velocity by inversion of Rayleigh wave: Geophysics, 64, 691–700,
Downloaded 06/27/17 to 84.220.195.227. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/