1. Introduction
Societal demands for flood control, water resources, and domestic/industrial energy helped initiate a prolific period of dam design and construction throughout the United States during the early- and mid-twentieth century
[1]. During much of this time, engineering efficiency and economic practicality allowed civil projects, particularly earthfill embankment dams, to utilize suitable
in situ earth materials (i.e., clay, sand, and gravel) for impervious cutoffs, filter drains, and other fundamental structural elements. It was not until the near-catastrophic dam failures associated with the 1971 San Fernando earthquake in southern California that more stringent consideration was given to the practice of using
in situ materials as part of a dam’s composition
[1]. Subsequently, most regulatory agencies required the excavation and replacement of unlithified natural materials with engineered fill in all aspects of high-hazard dam design and construction. However, the
in situ natural materials used in the earlier structures retained the potential for having undiscovered zones of adverse geotechnical properties (e.g., soft or low seismic-velocity zones, etc.), or acting to mask the presence of unfavorable geologic features (i.e., faults, karst, etc.) in the underlying bedrock foundation. Depending on the particular site conditions, resultant problems associated with these unknown geotechnical flaws may take decades to manifest.
Change orders or other post-design alterations often occur during the construction phase of any project, but can be especially numerous and significant for large complex structures such as flood-control dams
[2]. Although modern digital tracking systems and databases provide a more complete engineering archive, it should not be surprising that many construction modification records made during the early period of prolific dam building have been lost or poorly documented due to the large number of amendments and record-keeping practices of the time. Nevertheless, these previous standards of practice can make assessing
in situ geological or as-built engineered features within or beneath older earthen embankment dams a common source of geotechnical uncertainty.
All of these issues are of particular concern for large embankment dams and other high-hazard structures, because a large part of this infrastructure is passing into the latter stages of design life where minute flaws, which were overlooked or thought to be insignificant in the original design and/or construction phase, are proving problematic. Moreover, these geotechnical conditions can be especially consequential for structures exposed to high-hydraulic heads such as flood-control dams, where degradation can be accelerated and long-term integrity compromised.
Traditionally, solutions to geotechnical dam problems have largely relied on engineering measurements derived from various invasive drilling methods
[1]; however, this approach can be ineffective (e.g., lack of lateral resolution) and/or cost exorbitant at this size and scale. In addition, the boreholes are generally distributed evenly across the project “footprint,” thus problem areas can easily go undetected. However, a target-based drilling strategy can be technically and economically more effective by integrating the broad utility of noninvasive near-surface geophysics. Having said that, the selected geophysical method must be physically capable of resolving the geotechnical targets and objectives. The horizontally polarized shear-wave seismic-reflection technique can often provide the required resolution and optimal imaging for large embankment dams, including engineered structural features and
in situ geological foundation conditions. This supposition was documented at two earthfill flood-control dams in the central United States. Notwithstanding the high-noise conditions at both dams, the SH-wave seismic-reflection method proved successful for imaging low-impedance intra-embankment boundaries separating
in situ geologic sediment and compacted engineered borrow-fill of equivalent material (i.e., impervious cutoff), as well as buried as-built concrete elements (i.e., bulkhead cutoff and outlet works conduit). These results provide corroborative evidence for utilizing noninvasive cost-effective near-surface geophysics as a supplement to traditional drilling programs at large-scale civil works projects.
2. Background
SH-wave seismic-reflection profiles collected at two earthfill flood-control dams in the central United States were used in order to evaluate their geophysical effectiveness for imaging and locating the following: ① low-impedance intra-embankment boundaries separating engineered fill and in situ geologic materials incorporated into the dam design as impermeable cutoffs and drain filters; and ② relatively small, fundamental elements required for a properly functioning dam (i.e., interior concrete outlet conduit), as well as features added during the construction phase for the abatement of a foundation hazard discovered during excavation at the dam abutment (i.e., concrete bulkhead cutoff).
Dam Site 1, located in northern Indiana, was constructed during the early to mid-1960s (Fig. 1(a)). It is a compacted earthfill embankment approximately 2.4 km in length with a maximum height of nearly 53 m above bedrock. The embankment is founded on approximately 25 m of glacio-alluvial sediment overlying a Silurian limestone bedrock. The 788 m long SH-wave seismic-reflection profile was acquired along the downstream edge of the dam’s crest, approximately 12 m downstream of the axial center line. It originated at the right abutment and terminated at the onset of the centrally located outlet works. An earlier interpretation of these seismic-reflection data assessed existing geologic hazards associated with mature karst development in the limestone foundation beneath the embankment, which were encountered during both design and construction
[3]. The previous study successfully applied high-resolution SH-wave seismic-reflection surveying of the bedrock surface beneath the dam in order to approximate the depth to bedrock, and to identify signal anomalies indicative of unfavorable geological foundation conditions (i.e., karst). The data are used here to evaluate the SH-wave seismic-reflection method for imaging intra-embankment boundaries separating engineered fill (compacted glacial till borrow) and
in situ glacial till and alluvial sand used for an impermeable cutoff and horizontal drain filter, respectively. The data were also used in order to assess the geophysical ability to resolve the internal basal concrete conduit associated with the dam’s outlet works.
Dam Site 2, located in southern Kentucky, was constructed during the mid- and late 1960s, and consists of a rock-fill dam, earthen dike, gate-controlled outlet works, and an open-cut spillway between the dike and dam. The dike is the structure of interest for this study (Fig. 1(b)). It is an appurtenant impoundment dam 595 m in length and approximately 31 m in height, founded directly on Mississippian limestone. The SH-wave seismic-reflection transect was collected orthogonal to the centerline and atop the near-vertical zonated backfill in an excavated notch just above the base of the right dike abutment (Fig. 1(b)). These data were used to evaluate the effectiveness for geophysically imaging a concrete bulkhead cutoff, the excavated notch rock surface, and the engineered material zonation used in the backfill.
3. Methods
Pioneering research associated with near-surface high-resolution seismic-reflection imaging primarily incorporated primary wave (P-wave), or compressional wave, energy, because of the higher frequency characteristics and operational control offered by the various weight-drop, projectile, and vibratory sources (e.g., Refs. [
4–
8]). However, a groundwater masking effect can be problematic for P-wave energy that is used to image the solid matrix in subsurface environments where relatively low-velocity geologic media and their elastic impedance are water-saturated, such as the unlithified sediment generally found in the near surface of active fluvial and alluvial environments. The horizontally polarized shear-wave (SH-mode) common-midpoint (CMP) method has subsequently proven an amenable alternative for imaging these low-velocity, water-saturated, near-surface conditions (≪100 m) (e.g., Refs. [
2,
9–
18]). Shear waves can be more responsive for near-surface imaging because they are “framework waves” (i.e., not subjected to groundwater masking), and thus propagate with the solid media relative to the fluid-sensitive P-wave. The SH phase is theoretically easier to identify, unlike vertically-polarized shear-wave (SV-mode) signals, because of the lack of mode conversion at ideal impedance boundaries
[19]. Relative to P-wave energy, the lower velocity SH-wave has an optimal spatial window at the nearer offset; it also expands the temporal separation between the reflecting boundaries
[20]. An increased temporal window permits more distinct, accurate identification of the reflecting boundaries that result from the larger separation between the signal and coherent noise events, while the nearer offset optimal recording window decreases the potential for introducing adverse wide-angle-reflection effects. More importantly, however, experience shows that although SH-waves commonly have frequency content only one-half to one-third that of P-waves, the velocities of P-waves are 5 to 10 times higher than those of SH-waves; therefore, resolution commonly improves by a factor of two to three when using SH-waves [
14,
20]. This is a very important point when considering relatively small subsurface targets. For example, the major reflection horizons observed for the data presented herein have an average velocity range of 220–400 m·s
−1, and a dominant frequency of approximately 50 Hz. This yields a temporal resolvable limit (i.e., calculated by the one-quarter wavelength criteria
[19]) ranging between 1.1 and 2.0 m. The detectable limits are considerably smaller (i.e.,
λ/8–
λ/20). The spatial resolution of the reflecting horizons is constrained between approximately two and four shot points, based on the radius of the first Fresnel zone
[19].
In addition to overall varying targets and objectives requiring varying acquisition parameters, the seismic data collection at the two dam sites occurred several years apart; as a result, there was a deployment of different recording systems. The Site 1 survey used a Geometrics StrataView® 48-channel engineering seismograph and two inline spreads of 24 and 30 Hz, with horizontally polarized Mark Products® geophones with 7.5 cm ground spikes. The seismograph was a 24-bit system with an instantaneous dynamic range of 115 dB. The sampling interval was 0.5 ms, and a 25 Hz low-cut acquisition filter and a 250 Hz high-cut anti-alias filter were used. Due to several nearby electrical power sources, a 60 Hz notch filter suppressed potential unshielded noise. The shear-wave energy source was a modified section of steel H-pile (∼11 kg) struck horizontally with a 4.5 kg sledge hammer. The hold-down weight of the H-pile was between approximately 70 and 80 kg, including the weight of the hammer swinger and the H-pile section. The H-pile flanges and hammer swings were perpendicular to the geophone spread for SH-mode generation. The flanges were in prepared slit trenches in order to resist movement and maximize the energy couple with the ground. The geophones in the two inline spreads were spaced at 2 m intervals, for a total spread length of 96 m. The production survey maintained a 2 m near offset between the end-on shot point and the first active geophone. Example processing steps for the pre-stack field gathers are given in Fig. 2. These seismograms consist of two sets of three consecutive data field files collected near the first and last parts of the survey. The primary reflection events had a two-way travel time (TWTT) of 70–300 ms.
The Site 2 SH-wave seismic-reflection data were collected with two 24-channel Geometrics Geode® seismographs, and with a geophone array similar to that used in Site 1 but with 1 m spacing. The instrumental fidelity was equivalent to the seismograph used at Site 1, but was controlled with a separate laptop computer. Data sampling at Site 2 occurred at 0.25 ms intervals. The digital acquisition filters included 15 Hz low-cut and 60 Hz notch filters. The shear-wave energy source was similar to that used in Site 1, but involved a smaller H-pile (3 kg) and hammer (0.9 kg). Fig. 3 shows processing steps for the example pre-stack field gathers at Site 2. These are three single gathers that were selected from near the beginning, middle, and end of the survey.
Each SH-wave survey used instrumental polarity reversals and correlative reverse hammer impacts on both sides of the energy source for each shot point. This ensured correct identification of the SH-wave energy and decimated P-wave contamination. In addition, hammer blows were stacked (or linearly superimposed) between four and six times per shot point. Both sites realized minimal acquisition downtime; however, work was suspended during prolonged wind gusts and/or passing traffic.
Data acquisition parameters for the two sites are in Table 1. Although it is generally preferable not to apply acquisition filters, the local conditions provided justification for the partial use at the two sites. The processing for both data sets was on a personal computer using the commercially available VISTA seismic data processing software, versions 7.0 and 13.0, along with a conventional shallow CMP processing sequence. The primary signal processing steps focused on coherent noise muting, digital filtering, trace editing, appropriate trace balancing, and careful correlation statics for improving the pre-stack quality of the events seen on the raw field file. This is a minimal, but acceptable, approach for processing shallow seismic-reflection profiles [
21,
22]. These standard near-surface data-processing procedures are similar to those used in the petroleum industry, but appropriately scaled and conservatively applied. No other shallow-reflection processing (i.e., deconvolution, migration, etc.) were applied because of the insignificant signal quality improvement.
4. Results
4.1. Dam Site 1
The 12-fold, 788 m long, east–west SH-wave seismic-reflection image was parallel with the dam’s centerline and along the downstream edge of the embankment crest (Fig. 4). The profile provides subsurface coverage across the right half of the dam (i.e., between the right abutment and outlet works). The uninterpreted and interpreted images, along with a borehole-derived geologic profile for correlative assistance, are provided in Fig. 5. Trace numbers along the top of the reflection profiles are at 100 m intervals. In general, the data quality is good along the entire length of the profile. This was somewhat surprising, because significant noise was associated with the reservoir water discharge from the downstream outlet works. The relatively high-frequency shear-wave geophones may have acted as a pre-emphasis filter, minimizing the record contamination. The continuous, relatively coherent, large-amplitude reflection (doublet) visible between 150 and 300 ms TWTT (i.e., ∼24 and 53 m depth) across the profile is the top-of-bedrock unconformity
[3]. This reflection event provides a clear image of the terraces and former river channel. The previous interpretation for the abrupt loss of coherency at two locations in the “bright” bedrock reflector (Fig. 5) was karst solutioned joints and/or cavity collapse
[3]. The associated seismic-derived depths to bedrock were within 5% of those expressed in the geologic profile. A borehole targeted the karst interpretation at approximately trace number 410, and found 6.5 m lower-than-expected top-of-rock elevation from the geologic profile
[3]. The borehole observations found a 6.5 m interval below the expected top-of-rock consisting of very soft foundation material (i.e.,
N-values < 5), a primary indicator of karstic cavity fill likely formed from embankment “piping.” In addition, during excavation operations for the final remedial positive cut-off wall constructed through the limestone foundation and keyed into a lower shale unit, mature karst solution features were found in both anomalous areas identified in the initial interpretation.
Numerous intra-embankment reflectors are also apparent; these horizons agree reasonably well with the areas and elevations of engineered fill and in situ foundation soils described in the borehole-derived profile (Fig. 5). Of particular note is the steeply dipping reflector observed between traces 225 and 600, which is visible between TWTTs 70 and 150 ms. This event correlates with the boundary separating the engineered fill and glacial till interpreted in the borehole-derived geologic profile. The “onlap” characteristic of the shallow intra-embankment reflection just above the in situ till boundary near trace 250 (and ∼100 ms TWTT) is interpreted as emplaced engineered backfill abutted with the in situ till (Fig. 5). Although the borehole logs mention weathering being observed in the shallow till, the final geologic profile for the design document does not explicitly interpret a weathered zone for the till. A slightly steeper dipping reflector just beneath the interface is interpreted here as the base of the weathered zone. This zone may indicate a colluvial wedge. Other discontinuous reflectors appear within the till, and may represent the scattered sand and gravel bodies described in the geologic profile; however, this is speculative, and additional corroboration is required to verify these interpretations. The interpreted package of reflectors within the old river channel is the in situ alluvium that was left to act as a horizontal drain. The far-west side of the channel has a steeply rising reflector, and correlates with the onset of the outlet works in the as-built reports. A relatively strong shallow reflector is present between ∼80 and 90 ms TWTT, but appears to be discontinuous, likely due to the top muting. This horizon may correlate with the top of the vertical “chimney” drain constructed on the downstream side of the crest (Fig. 6). A reduced aperture for the acquisition array would be necessary to better image this feature (i.e., contract the shallow muted zone) and corroborate the interpretation. Nevertheless, the overall interpretive evidence suggests that the original as-built geologic profile for the project was well constrained, and is a reliable source of geotechnical detail for most subsequent remedial design actions.
4.2. Dam Site 2
Acquisition of the six-fold, 37 m long, SH-wave seismic-reflection subsurface image was orthogonal to centerline, approximately 10 m inside the Site 2 dike’s right abutment. The ground-surface elevation is approximately 229 m above mean sea level (AMSL), and the first subsurface image point was located 14 m upstream of the dike centerline, with the terminal point 23 m downstream of the centerline (Fig. 7). This location is the as-built location that was given in the change order for a concrete bulkhead to act as a positive cutoff for potential seepage through a large karst conduit discovered in the right abutment (Fig. 7). The location is set back approximately 15 m into the excavated notch from the originally planned location. The top-of-rock profile, determined from design through post-construction drilling programs, shows the irregular nature of this horizon (Fig. 7). Also shown is the original planned location for the bulkhead cutoff. Note the highly irregular or “pinnacled” top of limestone that is typical of a mature karstic environment. More importantly, the large subhorizontal solution conduit, which was discovered during design drilling investigations and confirmed during construction excavation, had the potential to provide an uncontrolled seepage path of reservoir water through the impoundment structure. The emplaced bulkhead acted as a positive cutoff; however, the as-built location was set back further into the abutment notch because of the highly weathered poor rock quality found at the original planned location. Fig. 8 is a simplified design schematic for the zoned engineered backfill in the excavated abutment notch. The elevation for the base of the notch, top of bulkhead, and abutment ground surface (at the survey location) is 214, 217, and 229 m AMSL, respectively. The final uninterpreted and interpreted Site 2 reflection profile is provided in Fig. 9. The profile consists of 74 seismic traces spaced 0.5 m apart. The notch clearly exhibits a synformal reflection event between seismic traces 7 and 61, and between 65 and 135 ms TWTT. The six horizontal relatively large-amplitude reflection wavelets visible between seismic traces 25 and 35, and at 135 ms TWTT, is the notch base. Using a shear-wave velocity of 220 m·s−1, this places the base of the notch approximately 15 m below ground surface, verifying the as-built elevation (i.e., 214 m AMSL). The bedrock surface outside the notch appears to be undulating but relatively coherent. A prominent “bright” reflection (i.e., higher relative frequency and larger amplitude) manifests within the engineered backfill material and notch interior. This is evident between seismic traces 28 and 33, and at 100 ms TWTT. This reflection is from the top of the concrete bulkhead cutoff. The 220 m·s−1 shear-wave velocity yields a depth of 11 m for this reflector, which is approximately equivalent to the as-built recorded elevation (i.e., 217 m AMSL). The ability to noninvasively verify the recorded as-built location without advancing one or more invasive boreholes minimized the potential for compromising the bulkhead’s cutoff integrity. However, it is also notable that there are no observable signal variations distinguishing between the clay core, sand filter, and rock shell. Truncation of the seismogram was below 200 ms TWTT, because of no target interest and multiple contamination (i.e. “ringing”) that was likely associated with the fill/rock interface.
5. Conclusions
Solving geotechnical problems with traditional geotechnical exploration (i.e., drilling) can be inefficient and often inadequate at high-hazard earthfill embankment dams
[16]; however, near-surface geophysical exploration using the SH-mode seismic-reflection method to supplement invasive drilling has improved the geotechnical assessment of two earthen flood-control dams in the central United States. The cost-effective, noninvasive, SH-wave seismic-reflection surveys clearly identified the bedrock foundation at both dam sites, as well as many primary intra-structural features. Prior to performing the survey at Site 1, it was uncertain whether the seismic-reflection method could physically resolve the contact separating the engineered fill and the
in situ geological sediment (i.e. glacial till) used for part of the impermeable cutoff, because the fill consisted of a compacted local “borrow” of the equivalent glacial till. However, the seismic-reflection survey showed that sufficient impedance does exist between the
in situ glacial till and the compacted fill to provide a coherent reflected signal. A more subtle impedance boundary was also observed beneath the fill-till boundary. This reflection was a weathering boundary or colluvial wedge in the
in situ glacial till. The
in situ alluvium in the former river channel also provided measurable elastic contrast. At Site 2, an image of the notch excavation in the abutment bedrock was clear; however, discriminating the vertical zonation for the engineered fill in the excavated notch was unsuccessful. In addition, the resolution of the SH-wave seismic-reflection signal permitted the imaging of relatively smaller engineered features, including a concrete bulkhead cutoff at Site 2 and part of the outlet works conduit at Site 1. These results corroborate the supposition that high-resolution SH-mode reflection surveys are a viable tool for evaluating dam safety issues, as well as other geotechnical problems in earthfill embankment dams.
PDF
(3359KB)
