Lateral resolution and lithological interpretation of surface wave profi ling

1330 The Leading Edge November 2008

Lateral resolution and lithological interpretation of surface wave profi ling

In civil engineering, geological hazards can be in the form of both soft and hard zones. Soft zones, notably sinkholes,

can lead to collapse structures reaching the surface and expensive damages during or after construction. Hard zones might be an impediment to excavation or, if they are to be used as foundations, it is desirable to know the thickness and strength.

Surface waves are becoming increasingly popular for geological mapping in severely contrasting geology. One is-sue with active-source surface-wave profi ling is the trade-off between lateral and vertical resolution. Longer arrays are re-quired to record lower frequencies, to image deeper layers, but at the expense of increased lateral smearing. Conversely, short arrays can achieve higher lateral resolution but with poor depth penetration.

A second issue is the interpretability of surface-wave data, and the worthwhile contribution they add to a project. Shear-wave velocity is probably more indicative of material strength than P-wave velocity in unconsolidated sediments, especially if near full saturation. It is also a vital parameter in site ampli-fi cation modeling, but apparently quantitative interpretation of shear-wave velocity to known lithologies in general engi-neering is not practiced.

We try to shed light on these two issues through fi eld and synthetic data sets. Numerical modeling of a “sinkhole” of various widths in a buried hard layer is used to simulate stan-dard surface-wave profi ling, and the strong dependency of the fi nal images on the acquisition parameters. Field data are then used to illustrate the added lithological discrimination when both surface-wave and resistivity methods are combined.

ADAM O’NEILL, DownUnder GeoSolutions, Perth, AustraliaTRISTAN CAMPBELL, Geoforce, Perth, AustraliaTOSHIFUMI MATSUOKA, Kyoto University, Japan

SPECIAL SECTION: N e a r - s u r f a c e g e o p h y s i c s

Surface wavesRecently, seismic landstreamers have become the method of choice for recording active-source surface-wave data, both on- and off -road. Good quality data from landstreamers are possible in comparison to planted geophones. Geophone fre-quencies of 2-100 Hz may be used, but usually 4.5-28 Hz are favored for MASW (multichannel analysis of surface waves). Sources are generally sledgehammer on plate (metal or PVC), or accelerated weight drop mounted on a towing vehicle.

We prefer to use close near off sets, usually 1-2 m, because we use a full-wavefi eld model for inversion and desire a more impulse-response type record. Acquisition geometry, near-fi eld eff ects (e.g., cylindrical spreading) and P-wave “noise” are all included in the inversion. An added advantage of close near off sets is that the same records can be used for refraction interpretation. Geophone spacing of 1–2 m is usually used, so spread lengths using a 24-channel receiver are usually 23 or 46 m. Maximum resolvable wavelength (and thus depth penetration) is about 0.4 of the spread length.

Th e standard active-source surface-wave workfl ow in-volves three stages: (1) acquisition of (usually off -end) shot gathers (Figure 1a); (2) processing by plane-wave transform to extract the surface-wave dispersion, by picking the ridge of the (usually f-k or tau-p) lobe maxima (Figure 1b); and (3) inversion of the dispersion curve, either fundamental, multi-ple or “eff ective” modes, to a layered shear-wave velocity (VS) model (Figure 1c).

Electrical resistivity imaging (ERI)ERI combines electrical sounding and profi ling through

Figure 1. (a) Acquisition of surface wave (ground-roll) data by off -end shot. (b) Processing (by plane-wave transform) to a dispersion image, showing the fi eld and synthetic dispersion curves. (c) Inverted shear-wave velocity model. Th e “true” model is a nearby downhole VS log.

November 2008 The Leading Edge 1331

N e a r - s u r f a c e g e o p h y s i c s

computer-controlled data acquisition from multichannel electrode arrays. A number of electrode confi gurations (e.g. dipole-dipole, Schlumberger, or Wenner) are simultaneous-ly used to optimize lateral and vertical resolu-tion. Stainless steel electrodes are placed at typically 1-10 m spacing and up to 72 can be concurrently linked to the control unit. A voltage of up to 800 V is applied across the current electrodes to inject up to 2.5 A, and separate potential electrode pairs measure and stack the resulting electrical fi eld at the surface.

Apparent resistivity is calculated for the multiple sets of quadripoles, and the data inverted using a 2D model. Depth penetra-tion of up to 100 m can be achieved, with vertical resolution typically 10–15% of depth (e.g., 4–6 m accuracy at 40 m depth). Lateral resolution is typically similar to the depth of burial — e.g., at a depth of 20 m, minimum resolvable target size is 20 m laterally.

Field exampleAn iron ore railway route in the Pilbara of Western Australia was planned to pass over known areas of calcrete dissolution and po-tential cavities. ERI data were collected with a Syscal Pro receiver using 2-m electrode spac-ing. MASW data were collected with a Geo-metrics Geode and 24-channel landstreamer with 2-m geophone spacing. Shot off set and shot spacing were both 2 m.

Th e coincident ERI and MASW results from one line are shown in Figure 2. Th e ERI data show a clear, broad, sym-metrical low-resistivity zone (<20 ohm-m) at depth, at least 20-m wide, within a more resistive calcrete basement >30 ohm-m). Th ere is a coincident MASW anomaly, though not as pronounced, and a “halo” of higher velocity material al-most appears to drape over the low VS core. Moreover, there is less symmetry, with the “basement” showing a highly undu-lating topography compared to the ERI. Th e shotpoint was to the left of the array, “pushing” from left to right. Figure 3 shows individual MASW results from three spread mid-points, either side of and over the middle of the low resistiv-ity/low VS anomaly. Figure 3b shows the clear low-frequency anomaly due to the soft material at depth. Figures 3a and 3b show the stiff er material at depth, but note the asymmetry in the shallow layering.

In the overburden at a depth of about 5–15 m, the ERI shows a low-resistivity zone with little vertical or lateral character, thought to be clay-rich sediments. However, the MASW data over this range show clear vertical stratifi cation and strongly laterally varying “blocky” character. Th e near surface (0–5 m) is gravelly sediments, of low VS and mostly very high resistivity.

So it can be said there are two main challenges to in-

terpreting these sections, in particular the usefulness of the MASW: the lateral resolution ability of the MASW and any acquisition geometry artifacts, and the lithological interpreta-tion of the combined ERI and MASW responses.

Th ese two issues are discussed in detail in the following two sections.

Lateral resolutionWe use a full-wavefi eld, elastic 2D fi nite-diff erence (2DFD) code to generate synthetic P-SV shot gathers in roll-along mode. Th e model, shown in Figure 4, is an analogous struc-ture to the fi eld case, intended to represent a break in a fer-ricrete or calcarenite layer, covered by unconsolidated soil. Th e elastic and simulation parameters are shown in Table 1 and Table 2.

A reference shot gather, dispersion image, and inversion result at a distance far from the sinkhole are shown in Figure 5. Th e dispersion curve observed from the 2D fi nite-diff er-ence shot gather is inverted using the 1D full-wavefi eld meth-od of O’Neill et al. (2003). Th e purpose is to illustrate the robustness of the inversion at an “ideal” location, away from any lateral subsurface variations. Th e stiff layer is imaged well, but smearing at depth arises due to loss of vertical resolution. Closer inspection of the observed dispersion curves from the 2DFD seismic and the 1D refl ectivity synthetics is shown in

Figure 2. Geophysical data sections from the Hope Downs survey. (a) Resistivity from ERI, and (b) VS from multichannel seismic surface-wave inversion.



Figure 6. Th ere is good agreement between the two methods, well within the noise level, so they can be considered system-atically compatible. Note how many multiple higher modes are generated for this layering, due to the “eff ective” disper-sion curve (black lines) crossing over the “modal” dispersion curves (grey bands), so it is impossible to invert using just a fundamental-mode method.

Roll-along 48-channel shot gathers were then simulated by 2DFD, with 1-m geophone spacing and 2.5 m near-off set. Dispersion curves were measured from each shot gather and the eff ective surface-wave dispersion inverted to shear wave velocity using the 1D refl ectivity method. Each “sounding” is then positioned at the spread midpoint, similar to general fi eld practice. Th e VS images for each “sinkhole” width are

Figure 3. Shot gathers, dispersion images, and inverted models from the Hope Downs surface-wave line, at positions (a) 60 m (left of deep anomaly), (b) 90 m (over anomaly) and (c) 120 m (right of anomaly).



shown in Figure 7.Th e 10-m wide sinkhole is apparently not resolved at all,

with a continuous stiff layer imaged across the section. Th e 20-m sinkhole appears partially resolved, but some artifacts of the stiff layer remain within the soft zone. Note too the asymmetry in the image, where the stiff layer appears thinner at the left side of the sinkhole and thicker to the right. Th e shot is located at the lower position end of the recording array (i.e., pushing from the left). Th e soft zone of the 50-m sink-hole is well resolved, but the asymmetry of the image still ex-ists. For all models, the very shallow soil is imaged accurately, but the deeper zone shows lateral variations due to noise and reduced resolution at lower frequency.

Results at a few single points along the 20-m wide sink-hole model are shown in Figure 8. When the cen-ter of the spread is over the edges of the hard layer (Figures 8a and 8c), the asymmetry is clearly seen in the VS profi le. Th is pattern is also seen in the fi eld data in Figures 3a and 3b. Over the middle of the sinkhole, an artifact remains because over 1/4 of the front and back of the recording spread is still lying over the hard layer. Th e fi eld results in Figure 2b show a similar artifact when the sur-rounding hard rock is being imaged across an ex-pected soft zone.

A possible solution to improve the symmetry of the imaging is nonlinear geophone spacing. By using 1-m spacing at near off sets, 2-m at mid

Layer Lithology Th ickness (m) VS (m/s) VP (m/s)

1 Loose soil 2 350 8572 Compacted soil /

sinkhole2 400 980

3 Hard rock 8 600 14704 Weathered rock 38 400 9805 Basement Inf 1000 2449

Parameter Value

nx (cells) 3001nz (cells) 1001

Δx and Δz (m) 0.1nt (steps) 51200Δt (μs) 10

Time window (ms) 512Source wavelet Berlage (impact)Peak frequency 40 Hz

Phase -90°

Table 1. Structure and elastic parameters used for the 2DFD modelling. Poisson’s ratio and density of all layers is 0.4 and 1.8 g/cc respectively. Quality factors of all layers for P- and S-waves are 100 and 45 respectively.

Table 2. Parameters used for the 2DFD modeling.

Figure 4. Sinkhole model used for 2DFD synthetic seismogram calculations and surface-wave “para-section’” imaging.

Figure 5. (left to right) Shot gather, dispersion image, and inverted model for a point far from the soft sinkhole.



off sets, and 4-m at far off sets, increased resolution of both high and low frequen-cies can be achieved with a 24-channel receiver. Moreover, it has been previously noted that the bulk of the surface-wave response appeared to be dominated by the soil closer to the shot, so a higher-channel density at near off sets is desirable.

Th e inversion image for the same sinkhole model of Figure 7b, using the same off set ranges (2.5-49.5 m) but with a nonlinearly spaced array of 24-channels is shown in Figure 9. Th e sinkhole is still not clearly resolved, but the high-velocity layer is imaged with slightly more clar-ity and the asymmetry of the image is reduced. Results at the same individual midpoints (left edge, middle and right edge of sinkhole) are shown in Figure 10, where the VS of the stiff layer is more ac-curately estimated than with an array of constant-spaced geophones.

Lithological interpretationSoil and rock types are generally only in-ferred qualitatively from shallow resistiv-ity and seismic surveys. Th e Hope Downs geophysical results were used to derive a data-driven classifi cation of local rock types to resistivity and VS measured at the surface. Figure 11a shows the clear clus-tering of points for diff erent rock groups intersected in several boreholes along the ERI and MASW lines and the best-fi t boundaries. Some points of note are:

Shear velocity is essential to discrimi-• nate between gravel/clayey-gravel and clayey-sandy-gravel/ferricrete.Ferricrete and clayey-sandy-gravel can-• not be readily discriminated with either resistivity or VS.Th e clayey-gravel overburden has the • highest resistivity and softest VS, possi-bly due to very dry and porous nature.Th e resistivity-V• S data points for voids lie entirely within the calcrete ellipse. Th is is probably because voids will mainly occur within these calcrete units and are below the geophysical resolution limits, thus are eff ectively not discriminated from calcrete.Sandy-clay will give the most estimated • lithological “noise” due to its occur-rence over much larger resistivity and VS ranges.

Figure 6. Comparison between the dispersion curves measured from full-wavefi eld 1D refl ectivity and 2DFD shot gathers, far away from the sinkhole. Th e grey bands in the background are the plane wave “modal” solutions. Th e observed “eff ective’’ mode dispersion can be seen to comprise several dominant higher modes, due to the “jumping” between modal solutions.

Figure 7. Shear-wave velocity models from the 48-channel spread data at 1-m geophone spacing, for sinkhole widths: (a) 10 m, (b) 20 m, and (c) 50 m.



MASW line. Th e fi nal section (Figure 11b) shows how the results in the shallow zone, depth less than 10 m, are driven by the coarser VS model structure. Th e shallow ferricrete and deeper calcrete units tie well with the borehole intersections.

A probabilistic method, similar to that used by Lamont et al. (2008) for estimating rock types and hydrocarbon oc-currence from deep seismic inversion, was used to produce an automatic lithology image along the coincident ERI and

Figure 8. (left to right) Shot gathers, dispersion images and inverted models from the 2DFD modeling results with 48-channel recording over the 20-m wide “sinkhole” at spread midpoint positions (a) -10 m (left edge of sinkhole), (b) 0 m (middle of sinkhole) and (c) +10 m (right edge of sinkhole).



Note how the interbedded gravel and clayey-gravel zones are clearly identi-fi ed in the generally smooth-looking resistivity image. Th e shallow zone, where most borehole core is lost, is dominated by clayey gravel. Th e grav-elly-clay, sandy-clay and calcrete zones show a high degree of mixing, which was expected due to the overlap in the rock classifi cation boundaries.

In the deep zone, depth more than 10 m within the coincident central low-resistivity and VS anomaly, the

Figure 10. Shot gathers, dispersion images, and inverted models from the 2DFD modeling results with 48-channel recording over the 20-m wide “sinkhole” at spread midpoint positions (a) -10 m, (b) 0 m, and (c) +10 m.

Figure 9. Shear-wave velocity model from the 24-channel spread data at nonlinear geophone spacing, with spread length of 47 m for the 20-m sinkhole width.



Figure 11. (a) Crossplot of resistivity and shear velocity for various lithologies intersected in boreholes along the Hope Downs ERI and MASW lines, with solid lines showing the fi tted Gaussian distributions, and (b) automatic lithology classifi cation results by Bayesian probabilities.

estimated lithologies are dominated by calcrete and voids. It is expected that the calcrete is most susceptible to dissolu-tion and infi ltration by saline groundwater. Th e reduced VS suggests that competency is also lost, possibly as a result of interconnected fractures and voids. Th e surrounding rock is

estimated as sandy-clay and siderite, due to the increase in VS and/or resistivity away from the anomalous zone. Essentially, the mapping of anomalous zones within the calcrete bedrock is suffi cient to warrant further geotechnical investigation for cavity risk.



SummaryIn active-source multichannel analysis of surface wave (MASW) profi ling, a soft sinkhole in a shallow, hard layer can only be qualitatively interpreted with confi dence if the lateral extent is more than half of the recording array length. If the extent is less than half the spread length, the surround-ing rock properties are “smeared” into the anomalous zone. In general, a 48-channel array at 1-m geophone spacing will not resolve anomalous zones of 20 m or less in lateral extent. A nonlinear geophone spacing, with close spacing near to the shot and far spacing away from the shot, may improve the clarity of sharp lateral geological variations and symmetry in end-on profi ling results.

Automatic rock classifi cation using combined ERI and MASW allows discrimination between cemented (calcrete/ferricrete) and uncemented (clays/sands/gravels) sediments in an overall low-resistivity overburden. In particular, it al-lows the diff ering resolutions of ERI (smooth) and MASW (blocky) to be probabilistically combined, and raises the risk assessment of void-susceptible calcretes for further geotechni-cal investigation. Such a scheme could be applied in mineral exploration for supergene deposits in regolith hosted by ce-mented ferricretes, which are desired to be distinguished from uncemented barren gravels and sands.

Suggested reading. “Surface-wave inversion limitations from laser-Doppler physical modeling” by Bodet et al. (Journal of En-vironmental and Engineering Geophysics, 2005). “1.5D inversion of lateral variation of Scholte-wave dispersion” by Bohlen et al.

(Geophysics, 2004). “Geophysical applications to detect sink-holes and ground subsidence” by Dobecki and Upchurch (TLE, 2006). “Drilling success as a result of probabilistic lithology and fl uid prediction—A case study in the Canarvon basin, WA” by Lamont et al. (APPEA Journal, 2008). “Genetic Algorithm inver-sion of Rayleigh wave dispersion from CMPCC gathers over a shallow fault model,” by Nagai et al. (Journal of Environmental and Engineering Geophysics, 2005). “Shear velocity model ap-praisal in shallow surface wave inversion,” by O’Neill, (Proc. ISC-2 on Geotechnical and Geophysical Site Characterization, 2004.) “Full-waveform P-SV refl ectivity inversion of surface waves for shallow engineering applications,” by O’Neill et al. (Exploration Geophysics, 2003). “Rapid shear wave velocity im-aging with seismic landstreamers and surface wave inversion,” by O’Neill et al. (Exploration Geophysics, 2006).“Delineation of a collapse feature in a noisy environment using a multichannel surface wave technique,” by Xia et al. (Geotechnique, 2004.) “A case study of seismic zonation in municipal areas,” by Yilmaz et al. (TLE, 2006).

Acknowledgements. Kyoto University Engineering Geology and Ex-ploration Geophysics laboratories are thanked for fi eld and comput-ing assistance. AO was funded with a postdoctoral fellowship from the Japan Society for the Promotion of Science (JSPS) from 2003–2005. Geoforce is thanked for fi eld data acquisition and permission to publish these data.

Corresponding author: [email protected]

Lateral resolution and lithological interpretation of surface wave profi ling

Technology

surfacewave dispersion

dard surfacewave proling

pwave velocity

surface andchoice

interpretability of

eri data

field data

masw data