Fig. 1 Location map and site plan for the Ludington Pumped Storage Plant site (LPSP).
Marine SP Surveys for Dam Seepage
Investigations:
Evaluation of Array Geometries Through Modeling and Field Trials
By
John Jansen, Ned Billington, Fred Snider and Patrick Jurcek
| ABSTRACT
Marine spontaneous potential (SP) surveys have been used on reservoirs for several decades. Most surveys have used gradient arrays towed at the surface or at some depth in the water column. Water column gradient arrays are efficient, but at the cost of some sensitivity. Other surveys have used a single fixed electrode and a roving electrode dragged across the reservoir bottom. These surveys have typically been labor intensive and logistically limited. Finite difference models and field trials were used to evaluate the sensitivity and field productivity of gradient arrays and roving electrode arrays, in the water column and on the reservoir floor, for detecting narrow seepage pathways in a large reservoir. The pre-survey finite-difference models suggested the greatest sensitivity would be achieved with the electrodes at the sediment-water interface. The gradient array models produced a bipolar anomaly with the spacing between the spikes equal to the electrode spacing. The single roving electrode models produced the sharpest, most easily interpreted anomalies. Single roving electrode and gradient arrays were field tested. Several electrode coupling modes were tested, including electrodes suspended at different heights in the water column, mechanically planted in the sediment, and dragged across the sediment (benthic arrays). The field trials demonstrated that physical coupling between the electrodes and the sediment significantly improved survey results. Benthic gradient arrays were most suitable for reconnaissance surveys. Benthic single roving electrode arrays were most suitable for detailed surveys in areas of suspected seepage features. The field trials were followed by two production surveys using a benthic single roving electrode array. The surveys successfully mapped seepage paths under the clay liner of the reservoir. The detected anomalies were generally less than 3 meters wide, which made high spatial sampling density and accurate position control critically important. |
Introduction
Many dams and reservoirs suffer from seepage problems that can cause piping or related problems. Under extreme conditions, seepage pathways and related features can undermine the stability of the structure. Spontaneous potential (SP) surveys have been used for decades for dam seepage studies on the downstream side of earthen dam embankments. The ability of the method to detect seepage pathways is well documented (Ogilvy et al., 1969; Bogoslovsky and Ogilvy, 1970; Sill, 1983; Corwin, 1989; 1990). Marine SP surveys have also been used on the upstream side of dams to detect seepage through the reservoir floor. While marine SP surveys are less common and not as well known, they have been used for various purposes for at least 60 years (Corwin, 1973).
Marine SP survey methods have been tried in seepage studies with varying degrees of success. In this investigation we have conducted limited pre-survey model studies of the expected sensitivity of several survey methods. The model results were used to design field trials to confirm the model predictions and directly evaluate survey methods for application at the Ludington Pumped Storage Plant (LPSP), a reservoir with a known seepage problem near Ludington, Michigan (fig. 1). Seepage conduits related to trench features (trench-like openings in the reservoir clay liner) have been documented at the LPSP by dye tests and diver inspections. Production surveys were performed over known seepage areas of the reservoir using survey parameters based on model studies and field trials. This paper presents the results of these investigations.
Discussion of Previous Methods
Many marine SP surveys have been conducted but few have been published. Most of the published surveys have used gradient arrays towed at the surface of the water column or within the water column some distance above the sediment-water interface (Corwin, 1990). Figures 2a and 2b are schematic representations of these arrays. Measuring potential in the water column is known to decrease the sensitivity of the method (Corwin, 1973). For practical reasons, however, to avoid snagged and fouled lines, this method is often employed as a compromise between sensitivity and mobility. Water column gradient array surveys have been successful at finding offshore ore bodies (Corwin, 1973) and measuring the corrosion state of revetments on the Mississippi River (Sjostrom et al., 1993).
In some reservoirs, the bottom consists of an engineered soil layer that is relatively smooth. In most reservoirs, a layer of fine sediment builds up which covers irregularities and smooths the reservoir floor. As a result, snagging towed lines may be less of a problem in many reservoirs than for typical open marine environments. Some published surveys have used conventional well logging equipment to physically drag an electrode across the bottom of a reservoir while logging the potential difference with respect to a fixed electrode planted in the reservoir sediment (Ogilvy et al. 1969) as illustrated on fig. 2c. This approach is analogous to the familiar roving electrode method used for land surveying (Burr, 1982).
Each of these survey methods presents a compromise between survey efficiency and survey sensitivity. Towing electrodes in the water column provides fast, relatively easy survey operations, but at a cost of lower sensitivity to small anomalies. Towing electrodes along the floor of a reservoir significantly complicates field procedures, but increases the sensitivity of the survey.
Gradient arrays, in which both electrodes are towed at a constant spacing, do not require the survey vessel to be tethered to a fixed electrode. This makes the field procedures faster, but produces a bipolar anomaly which can cause interference between multiple anomalies in theory, plots of the summation of gradient array data should be equivalent to the results of a single roving electrode array (Milsom, 1989). Since there is no fixed point of reference, however, uncorrected electrode biases for the gradient array method tend to be cumulative. Single roving electrode arrays require the vessel to be physically tethered to a fixed electrode which limits maneuverability. However, these arrays produce sharp mono-polar anomalies which do not interfere with other anomalies.
While each of these methods has particular advantages, each has a drawback in terms of data quality or survey efficiency. Fortunately the relative cost versus benefit for each of these methods can be evaluated before starting a survey to optimize the survey design.
Corwin (1973; 1976) provides a solution, modified from Van Nostrand and Cook (I 966), for the expected attenuation of a SP point source placed at a given depth below the water sediment interface. The calculation of the solution is somewhat tedious, but lends itself well to a spreadsheet. This solution predicts that in general, the higher the resistivity of the water column, the lower the attenuation of the anomaly in the water column.
While the approximation described by Corwin is useful, the point source assumption is not truly valid for actual SP sources. For small sources at great depth, the difference should be negligible. However, for shallow SP sources with nonzero dimensions, the point source approximation is not valid, and the potential distribution in the near-field is different.
Analytical solutions for potential fields from sources with simple geometric shapes are available, but the solutions become complex as the number of variables increase. Corwin (1973) points out it is unlikely that the physical parameters of most SP sources will be known well enough to make the calculations worthwhile. While analytical solutions are difficult to apply in all but the simplest cases, approximate solutions can be easily obtained using finite-difference methods and reasonable assumptions as to the properties of the SP source.
Before starting survey operations at the LPSP site, the estimated site conditions were used to conduct limited computer simulations of the expected sensitivity of several survey geometries. The theoretical anomalies from a given SP source were calculated for gradient and single roving electrode arrays towed at various depths in the water column and at the sediment-water interface. The results of these simulations were used to design a field testing program to confirm the model results and design the production survey.
Most of the electrical properties of the reservoir were estimated from the construction details. The reservoir was constructed with a base of approximately 1 meter of compacted silty sand and clay over fine to medium sand. The reservoir contains fresh water with a resistivity of approximately 100 ohmmeters. The water level in the reservoir fluctuates daily in response to power generation. At the deepest point of the reservoir the water column ranges from about 10 to 30 meters high, depending on the pool level. Resistivity measurements made by previous marine resistivity surveys in the reservoir found the resistivity of the clay liner to be approximately 33 ohmmeters. The sand beneath the clay liner was assumed to have a resistivity of approximately 100 ohmmeters.
The SP source was arbitrarily assumed to be a 50 mV prism of infinite lateral extent. The source was simulated as a rectangle one meter high and two meters wide with infinite lateral extent. The position of the source was simulated (1) at the sediment-water interface (in the clay liner) and (2) one meter below the interface (beneath the clay liner). These source locations were chosen to simulate the case of an active seepage face in the clay liner and an active seepage path just beneath the intact liner.
A 10 by 21 single layer grid was used to model the reservoir and water column in cross-section with the horizontal direction extending along rows and the vertical dimension (depth) extending along columns. Nodes were spaced two meters along columns and one meter along rows. Row 10 was set at a constant voltage of 0 mV with a single element in column 11 set to a constant voltage of -50 mV to simulate the SP source. Figure 3 illustrates the finite-difference grid and source position.
The model was run at steady-state under confined conditions so that no flow occurs into or out of the plane of the model. The remaining boundaries were treated as flowthrough boundaries. The simulated depth of the SP source was changed by modifying the conductivity of the appropriate rows of the model above the source to represent either sediment or water.
Figures 4 and 5 present the model predictions for a single roving electrode array survey with the SP source in the clay liner and just beneath the clay liner. The model results indicate that the SP anomaly is attenuated in the water column. With the SP source at the sediment-water interface, the potential decreased by 36% at a distance of one meter off the bottom and by 86% at a distance of 2 meters off the bottom. At ten meters off the bottom, the anomaly is reduced to less than 2% of its initial intensity. For the SP source beneath the clay liner (fig. 5), the potential drops only 20% at the sediment-water interface, but is noticeably broadened. Two meters off the bottom, the potential is reduced by approximately 72% and the anomaly is broader. At ten meters, the anomaly is not detectable under typical field conditions. A vertical planar source was also simulated by using a finite-difference grid with additional rows to extend the SP source 20 meters into the reservoir floor. The magnitude of the planar source anomaly was found to diminish more slowly (about 1/2 as fast than the anomaly for the horizontal prism source) as the depth to the top of the planar source was increased.
The expected responses for gradient array surveys were also calculated from the model data. Figures 6a and 6b show the expected anomaly of a gradient array dragged across a 20 mV SP point source at the sediment-water interface using 10 meter and 2 meter electrode spacings, respectively. The gradient arrays produce distinctly bipolar anomalies caused by the positive and negative electrodes crossing the source independently. Figure 6b illustrates the sharp bipolar anomaly produced when the electrode spacing is approximately the same as the width as the SP source. When the electrode spacing is much greater than the width of the source (fig. 6a), the two poles of the anomaly are offset. This creates the possibility that response from the second electrode may be incorrectly interpreted as an independent positive anomaly. If the source is wider than the electrode spacing, the responses of the two electrodes interfere with each other, causing a complex anomaly that is difficult to recognize and which could be mistaken for noise.
The models, incorporating several assumptions and simplifications, were useful in determining the most significant parameters for the survey design. The model results suggest that strong SP sources near the sediment-water interface probably could be detected several meters into the water column. Weaker sources, or strong sources more than a few meters below the clay liner, should be detectable on the bottom but not at any appreciable distance into the water column.
The model results also suggest that the single roving electrode array would provide the simplest data to interpret. The modelling study also pointed out that some assumptions must be made about the width of the expected anomalies when selecting the electrode spacing for gradient arrays. Whatever spacing is used, a compromise must be struck between the need to accommodate wider anomalies and the need to keep the electrodes from straddling multiple anomalies in the reservoir was designed to confirm conclusions of the model studies and set the parameters for the production survey. The field trials were specifically designed to evaluate the effects of towing depth on sensitivity, directly compare the gradient and single roving electrode arrays, and determine the importance of electrode coupling on the survey results.
The field trials were carried out in a 5 meter survey vessel with assistance from a more mobile inflatable support boat. The data was collected using jellied copper sulfate electrodes and a high impedance voltmeter with digital output. The voltmeter output was combined with a triangulation navigation system. SP measurements were recorded at one second intervals and the boat position was updated every 3 seconds. The horizontal position of the boat was measured to within one meter and the electrode positions were calculated from the length of the tow line and its angle from the horizontal.
"Sawhorse" Electrode Plant Tests
Initial tests were conducted to determine the importance of electrode coupling with the sediment. Figure 7a shows a weighted "sawhorse" apparatus that was used to keep a constant spacing of 4.25 meters between electrodes. The sawhorse served as a semirigid frame that could be suspended in the water column or lowered to the bottom to "plant" the electrodes into the sediment using the weight of the sawhorse. When suspended in the water column, SP measurements were observed to be erratic, with rapid variations of several mV, even with the boat anchored at a fixed location and the electrodes suspended only a few meters off the bottom. Periodic noise bursts of over 10 mV were observed with the electrodes in the water column. The SP measurements became stable almost immediately when the electrodes were planted in the sediment.
The high noise levels in the water column were surprising. Other workers (Corwin 1973; 1976) have reported typical noise levels of less than a few tenths of a mV. We do not fully understand the reason for the erratic measurements. Readings did not improve after shutting down all electrical systems on the boat and making the SP measurements with a battery powered voltmeter. Surface conditions in the reservoir were poor on both test days, with heavy chop and waves of 1 to 1.5 meters. Corwin (1973) reports that SP signals of up to several millivolts can be created by waves. Possibly more significant than wave noise was the turbid quality of the water caused by the waves. It is possible that the variations in the water column SP measurements were due to a combination of the wave generated SP signals and temporal variations in the chemistry and conductivity of the turbid water. Whatever the cause, SP measurements made in the water column were found to be too noisy for practical use on this survey.
Stable readings were obtained whenever the electrodes were in contact with the sediment. The measured potential between PI and P2 with both electrodes planted into the sediment by the weight of the sawhorse was compared to the potential between two electrodes directly planted into the sediment by a diver immediately adjacent to the sawhorse electrodes. The measurements were generally found to agree to within one mV, suggesting the gravity plants from the sawhorse produced satisfactory coupling with the sediment.
Single roving electrode and gradient array readings were made at several depths in the water column and on the bottom (benthic arrays) at discrete points along a 180 meter test line. The depth of the electrodes was changed by raising and lowering the sawhorse at each measurement location. The lateral position of the measurements were changed by winching the boat along forward and aft anchor lines between measurement points. Single roving electrode measurements were referenced to a fixed electrode planted in the sediment and tethered to the ship with a cable.
The water column readings for both the gradient and single roving electrode arrays suffered from high noise levels. Readings were averaged over a few minutes to obtain a representative reading for each station. The averaged water column readings were found to have poor repeatability and little correlation with the benthic readings. The single roving electrode benthic readings detected an anomaly of about -6 mV on the test line. The anomaly was difficult to detect in the benthic gradient array data because of higher noise levels in the data.
The test line required several hours of physically demanding work to obtain less than 200 meters of data. The sawhorse tests suggested that water column measurements are subject to high noise levels on this reservoir and are unlikely to provide usable data. This finding was consistent with the disappointing results of a water column gradient array survey attempted several years earlier by other workers. The test line also showed that usable SP measurements could be made with weighted electrode plants from the surface. Unfortunately, the production rate of this method was far too low to be practical.
To increase production, towed benthic gradient and single roving electrode arrays were tested using weighted tow eels with jellied copper sulfate electrodes. Figures 7b and 7c show the configuration of the tested arrays. Side-scan sonar surveys and numerous dives had shown the reservoir floor to be relatively smooth and covered with a few inches to a few feet of soft sediment. The soft sediment was expected to provide a uniform media for direct electrode coupling for the towed benthic arrays.
Benthic Gradient Array Tests
The towed benthic gradient array was relatively easy to deploy and production rates of approximately 10 km/hr were maintained. Data quality was generally good with typical noise levels of less than 2 mV, except for one area of the reservoir which had a rough bottom composed of cobbles. Noise levels in this area were over 3 mV, probably due to decoupling of the electrodes with the bottom.
Figure 8 is a plot of a test line of the benthic gradient array. The plot shows a clear bipolar anomaly with an amplitude of about ± 14 mV centered at about 125 meters north. Several smaller bipolar anomalies are also present, some of which are fairly distinct. However, many of the smaller anomalies are difficult to distinguish due to noise and interference between multiple anomalies.
Benthic Single Towed Electrode Tests
The test line was resurveyed using a single towed benthic electrode and a fixed electrode tethered to the boat with a cable. In general, the data quality was excellent, with noise levels of less than 1 mV. Due to the tether to the fixed electrode, line length was limited to about 250 meters. Figure 9 presents the results of the test line. The large dipolar anomaly seen on the gradient plot is apparent as a sharp anomaly of about -20 mV at 125 meters north on the single towed electrode plot.
Several smaller anomalies with amplitudes of less than 5 mV, such as at 165 meters north, are readily apparent. This anomaly is present on the gradient data plot, but it is more difficult to identify due to the complex shape of the anomaly. An anomaly of about - 12 mV is clearly present at 195 meters north on the single towed benthic electrode array data. This anomaly is virtually impossible to detect on the benthic gradient array data.
SP values gradually increased toward the north end of the single towed benthic electrode array line. Parallel survey lines also found long-wavelength positive anomalies that appear to be caused by electrical load centers on the sides of the reservoir installed to operate pumps and other equipment. Production rate with the single towed electrode array was limited to about 8 km/day.
Results of the Field Trials
The field trials clearly demonstrated the need to directly couple the electrodes with the reservoir floor. Fortunately, the reservoir floor was relatively smooth so dragging electrodes across the bottom was feasible. Had this not been possible, marine SP may not have been a viable method for identifying seepage features beneath this reservoir.
The negative SP anomalies detected were found to correlate to suspected seepage paths identified by side-scan sonar and sub-bottom profiler surveys. Most anomalies were 1.5 to 3 meters wide. Because the anomalies were so narrow, high spatial sampling density and accurate location control were critically important. Weak positive SP anomalies of several mV, such as at 80 meters north on fig. 8, were observed immediately adjacent to negative SP anomalies on the single towed electrode data sets. These anomalies have been correlated to bentonite patches previously installed to reduce seepage. Ogilvy et al. (1969) has observed positive SP anomalies of several mV from bentonite clays. The presence of the negative anomalies adjacent to the bentonite suggests that the seepage features are still active or have shifted laterally around the bentonite patches.
Because of its high production rate but limited sensitivity, we decided that the gradient array was best applied as a reconnaissance tool. Entire reservoirs could be surveyed on a tight grid in a matter of a few hours or days, providing no snagging problems occurred. The single towed benthic electrode array suffers from a lower production rate due to logistical problems created by the tether to the fixed electrode. However, data quality is superior to the gradient array data. As a result, we concluded that the single towed electrode array is most appropriate for detailed survey work in suspected areas of seepage.
 Fig. 8 Benthic gradient array test line data (12.2 meter elctrode spacing). |
 Fig. 9 Benthic single roving electrode array test line data. |
 Fig 10. SP anomaly from a seepage source that is approximately 1 meter wide. |
 Fig. 11 Positive Sp anomaly from corroding metal at 40m east. Seepage fractures may be present at 110m east. |
Following the field trials, production surveys were run using the benthic single towed electrode array. Approximately 17 km of data were collected in two field days. The survey lines were limited to about 250 to 300 meters by the fixed electrode tether. The survey lines were concentrated in three areas with known seepage problems.
Data quality was generally high, with anomalies as small as a few mV plainly visible. Figure 10 shows a sharp anomaly related to a narrow seepage feature in an area of good data quality. The SP results correlated well with seepage features that were known from side-scan sonar surveys, sub-bottom profiler surveys, and diver inspections. In addition, the SP survey was able to detect seepage features too small to detect with acoustic methods. Seepage features could be traced along adjacent SP survey lines toward the edge of the reservoir. Seepage paths could also be traced under paved surfaces and concrete, where acoustic methods could not penetrate.
One area of the reservoir had a hard bottom with cobbles and miscellaneous debris. Background noise in this area was higher (1-2 mV), but still acceptable. Figure 11 presents the data from a line across this area. Many of the small negative anomalies, such as at 110 meters east, are believed to represent seepage features. The sharp positive anomaly at about 40 meters east is thought to represent a corroding piece of metal debris on the reservoir floor.
Following the production survey, a remediation program was conducted to fill trench features in the clay liner of the reservoir. Approximately one year after the remediation program, the SP survey was repeated to evaluate the effectiveness of the rehabilitation efforts. The follow-up survey used a modified SP tow eel which was more streamlined to reduce snagging problems. This modification significantly improved survey efficiency. The post-remediation survey identified several areas where seepage was apparently still occurring. Figure 12 shows the correlation of SP anomalies between parallel survey lines and the projected extension of seepage related trench features.
Conclusions
The pre-survey models suggested that the single roving benthic electrode array would provide the highest sensitivity and the easiest data to interpret. The field trials confirmed the major conclusions of the pre-survey models. Field tests also detected unexpectedly high SP noise in the water column. Due to the noisy environment in the water column, SP measurements were made in direct contact with the sediment by dragging the electrodes across the bottom using "benthic" arrays. Other workers (Corwin, 1973; 1976) have produced good quality data from gradient arrays at the surface. The cause of the noise in our water column measurements is not well understood.
The gradient array was found to be best suited for reconnaissance surveys. The single towed benthic electrode array was found to be best suited for detailed investigations in areas of suspected seepage features. The SP method was found to be more sensitive to small seepage features than acoustic methods and was able to operate over paved surfaces, which acoustic methods could not penetrate.
References
Bogoslovsky, V.A., and Ogilvy, A.A., 1970, Natural potential anomalies as a quantitative index of the rate of seepage from reservoirs: Geophysical Prospecting, v. 18, no. 2, p. 261-268.
Burr, S.V., 1982, A guide to prospecting by the self-potential method: Miscellaneous Paper 99, Ontario Geological Survey, 15 p.
Corwin, R., 1973, Offshore application of self-potential prospecting, Ph.D. thesis, University of California, Berkeley, 303 p.
Corwin, R., 1976, Offshore use of the self-potential method: Geophysical Prospecting, v. 24, p. 79-90.
Corwin, R., 1989, Development of self-potential interpretation techniques for seepage detection: Technical Report REMR-GT-6, no. 3, U.S. Army Engineer Waterways Experiment Station, Vicksburg,Mississippi, 71 p.
Corwin, R., 1990, Applications of the self-potential method for engineering and environmental investigations: in Proceedings of the Symposium on the Application of Geophysics to Engineering and Environmental Problems, EEGS, Englewood, CO, p. 107-120.
Jansen, J., 1995, Using MODFLOW as an interpretive tool for several geophysical methods: in Proceedings of the Symposium on the Application of Geophysics to Engineering and Environmental Problems: Bell, R.S., Ed., EEGS, Englewood, CO, p. 699-708.
McDonald, M.G., and Harbaugh, A.W., 1984, A modular three-dimensional finite-difference groundwater flow model: U.S. Geological Survey Open File Report 83-875, Reston, VA, 528 P.
Milsom, J., 1989, Field Geophysics, Geological Society of London Handbook: Halsted Press, New York, 182 p.
Ogilvy, A.A., Ayed, M.A., and Bogoslovsky, V.A., 1969, Geophysical studies of water leakages from reservoirs: Geophysical Prospecting, v. 27, no. 1, p. 36-72.
Sill, W.R., 1983, Self-potential modeling from primary flows: Geophysics, v. 48,p. 76-88.
Sjostrom, K. J., Butler, D.K., and Corwin, R.F., 1993, Assessment of Mississippi River revetments using the self-potential method: in Proceedings of the Symposium on the Application of Geophysics to Engineering and Environmental Problems, Bell, R.S. and Lepper, C.M., Eds., EEGS, Englewood, CO, p. 435-448.
Van Nostrand, R.G., and Cook, K.L., 1966, Interpretation of resistivity data: U.S. Geological Survey Professional Paper 499, 3 1 0 P.
Wang, H. P., and Anderson, M.P., 1982, Introduction to groundwater modelling: W.H. Freeman, San Francisco, 237 p.
