USGS Groundwater Information: Branch of Geophysics
by Marc L. Buursink and John W. Lane, Jr. **
U.S. Geological Survey, 11 Sherman Place, U-5015, Storrs, CT
A study incorporating numerical modeling, physical modeling, and field surveys at the U.S. Geological Survey Fractured Rock Research Site at Mirror Lake, Grafton County, New Hampshire, was conducted to test the use of ground-penetrating radar (GPR) surface reflection methods to delineate fractures in heterogeneous bedrock. Results of one- and 2.5-dimensional numerical modeling correlate with results of laboratory-scale physical modeling and establish different GPR reflection characteristics for saturated and unsaturated (dry) fractures. Saturated fractures generate higher amplitude reflections than unsaturated fractures and have an opposite phase. GPR reflection data collected over a highway bedrock outcrop near Mirror Lake were processed to reduce noise and clutter, to correct geometric and topographic distortions, and to enhance weak reflections from structures more than 15 meters (m) deep.
Guided by the results of numerical modeling, 18 reflectors with lengths ranging from 4 to 32 m and dips ranging from 5 to 40 degrees were interpreted as fractures in the processed GPR field records. All of the interpreted fractures that project above land surface correlated with fractures recorded by detailed outcrop mapping. Interpretation of the processed field data was limited to reflectors dipping less than 45 degrees, although more steeply dipping fractures exist at the field site. Spatial aliasing effects and other limitations constrain the range of dip-angles that GPR reflection methods can usefully image from the surface. The results of the field study together with the results of the numerical and physical modeling indicate that surface GPR reflection methods can be used to help characterize fractures in heterogeneous bedrock.
This paper presents the results of a ground-penetrating radar (GPR) field study conducted at the U.S. Geological Survey (USGS) Fractured Rock Research Site at Mirror Lake, Grafton County, New Hampshire, to test the use of this surface reflection method to delineate fractures in heterogeneous bedrock. The study also used numerical and physical modeling to establish GPR reflection characteristics in fractured bedrock and to identify reflection characteristics useful for understanding and interpreting the processed GPR field data.
GPR has been used in many studies to identify and characterize fractures and faults in rock (Ulriksen, 1982; Holloway, 1992; Benson, 1995; Stevens and others, 1995; Toshioka and others, 1995; Grasmueck, 1996; Serzu and others, 1996). In general, these studies have been conducted at field sites where interpretation of GPR data to delineate fractures or faults is constrained by relatively simple geologic conditions.
In contrast, the geology at the Mirror Lake research site is extremely heterogeneous. Bedrock at the site consists of gneiss and schist intruded by granite and pegmatite dikes (Hsieh and others, 1993; Barton, 1996). Bedrock fractures at the Mirror Lake site have highly variable orientations, are poorly connected, and occur at length scales from centimeters to tens of meters.
NUMERICAL AND PHYSICAL MODELING
Numerical and physical modeling of the GPR signal response to fractures in bedrock was conducted to predict radar reflection characteristics in fractured rock. A one-dimensional (1-D) forward modeling program (Powers and Olhoeft, 1995) and a 2.5-dimensional (2.5-D) forward modeling program (Powers, 1995) were used to simulate saturated and unsaturated (dry) fractures in bedrock.
Results from 1-D GPR models containing a single fracture show the response of a 100-megahertz (MHz) radar signal to fractures with a range of apertures. The 1-D full waveform modeling results for 1-, 4-, and 16-millimeter (mm) aperture saturated and unsaturated fractures are shown in figure 1. The modeled radar reflections from the fractures are shown adjacent to the reflection from the overburden-bedrock interface to provide a relative scale for assessing the differences in normalized reflection amplitude and phase. The modeled radar reflections from saturated fractures have the largest normalized peak-to-peak amplitudes and are out of phase with the transmitted pulse and the overburden-bedrock reflection. The reflection amplitude increases and the waveform changes as the modeled fracture aperture increases from 1 to 16 mm. Modeled radar reflections from unsaturated fractures have amplitudes that are slightly larger than the overburden-bedrock reflection and are in-phase with it and the transmitted pulse.
Radar reflection characteristics were also simulated using a 2.5-D modeling program. The 2.5-D modeling program generates data that resemble field GPR data. For example, individual fracture or fracture-zone reflections with horizontal or sub-horizontal dips simulated in 2.5-D provide a visual image of what may be observed in the field data.
Figure 1. Results of 1-D radar modeling using a 100 MHz starting wavelet, showing the overburden-bedrock interface reflection and showing changes in the reflection amplitude and waveform for 1-, 4-, and 16-mm aperture saturated and unsaturated fractures. Model values for the dielectric permittivity (er) and electrical conductivity (s) are indicated.
Figure 2. The fractured granite block used for the laboratory-scale physical modeling of GPR reflection characteristics. A GPR reflection survey in progress is shown.
The results from the 1- and 2.5-D numerical modeling were compared to GPR data acquired over a laboratory-scale physical model. The physical model comprised a relatively homogeneous, feldspathic granite block about 1.9 meters (m) long, 1.1 m wide and 1.4 m high (fig. 2). A sub-horizontal fracture was induced through the center of the block about 0.8 m from the top. The edges of the fracture were sealed to prevent leakage, and plastic valves were installed in the upper and lower edges of the block permit water injection and withdrawal.
Radar data were acquired with a commercial radar field-instrument (manufactured by Geophysical Survey Systems Inc.) using a 500 MHz center-frequency antenna. Thirty traces of radar reflection data were acquired at discrete 5-centimeter (cm) intervals by stepping the antenna across the length the block. Figure 3 shows the radar reflection data for the saturated (fig. 3a) and unsaturated (fig. 3b) fracture experiments. The reflection from the fracture appears at about 14 nanoseconds (ns) in the GPR records. The amplitude of the radar reflection from the saturated fracture is greater than the amplitude from the unsaturated fracture. The phase of the radar reflection from the saturated fracture is opposite the phase of the reflection from the unsaturated fracture. The amplitude of the reflections varies across the face of the block, increasing in amplitude from the left (higher end of the block) to the right side (lower end of the block). The differences in the reflection amplitude and waveform are caused by the increase in the fracture aperture at the lower end of the block.
Comparing the results of both the 1-D and 2.5-D numerical modeling to the results of the physical modeling indicates that the numerical models accurately predict radar reflection amplitude and phase characteristics for simple geologic models. Both modeling methods indicate that saturated fractures generate high-amplitude reflections with a phase opposite that of the transmitted pulse. Based on the correlation between the numerical and the physical model data, 2.5-D modeling results were used to interpret field GPR data to identify bedrock fractures.
Figure 3. 500 MHz GPR reflection record from the physical model: (a) saturated fracture, (b) unsaturated fracture. The reflection from the fracture at 14 ns is labeled. The depth scale was calculated assuming a radar propagation velocity of 117 m/ms.
GPR reflection surveys were conducted at the I-93 highway outcrop near Mirror Lake (fig. 4). The outcrop, along the center median of the highway, is more than 200 m long and 20 m wide. The average height of the outcrop is about 6 m above land surface, with the lowest point at land surface, and the highest at 12 m. Fractured bedrock exposed by the outcrop is representative of that underlying most of the study area (Barton, 1996). The outcrop is particularly suitable to GPR reflection surveys because (1) the bedrock is electrically resistive, a physical property favorable for penetration of the radar; (2) the large area of exposed bedrock is not covered with overburden, therefore minimizing sources of near surface noise and clutter; and (3) the fractures exposed in the bedrock outcrop walls have been extensively mapped (Barton, 1997), providing excellent ground-truth.
Figure 4. Site diagram showing the Mirror Lake Fractured Rock Research Site, Grafton County, New Hampshire. The bedrock outcrop that was used as the location for the GPR surveys is located at the eastern end of the study area.
The GPR surveys were conducted along the centerline of the median outcrop, about 10 m from the outcrop wall. GPR data were collected with a portable battery-operated, commercial radar field-instrument (manufactured by Malå GeoScience, A.B.) using 200 MHz center-frequency antennas. For this study, two types of GPR surveys were conducted on the bedrock outcrop: (1) continuous profiling common-offset (CO) reflection surveys (fig. 5a), and (2) common mid-point (CMP) reflection surveys (fig. 5b). The continuous profiling surveys were conducted to image the subsurface; the CMP surveys were conducted to establish the radar propagation velocity in bedrock, a value required for data processing.
Figure 5. Diagram showing (a) the antenna geometry for a common-offset (CO) ground-penetrating radar reflection survey and (b) the antenna geometry for a common mid-point (CMP) survey used to derive the average subsurface radar propagation velocity. In both, Tx and Rx indicate the locations of the transmitting and receiving antennas, respectively.
For the CO profiling surveys, the receiver and transmitter antennas were separated by 40 cm in a fiberglass sled, which was pulled across the outcrop surface. A GPR trace was collected every 10 cm along the profile, with a recording window of 350 ns. The use of these acquisition parameters ensured that reflections from structures dipping less than 45 degrees were accurately imaged. For dips exceeding 45 degrees, however, geometric distortion from spatial aliasing hinders the interpretation of steeply dipping fractures when using CO reflection methods. The CMP survey was collected by placing the antennas on the rock surface and increasing the transmitter-receiver antenna offset from 0.5 to 25 m, in 0.5-m increments.
Data processing methods can reduce the presence of noise and can correct the geometric and topographic effects that severely limit the usefulness of the unprocessed field data. The GPR reflection data from the outcrop were processed and displayed in a manner that facilitated the interpretation and identification of continuous reflectors (fig. 6). Details of the methods used to process the field data can be found in Yilmaz (1987) and Telford and others (1990). The data processing methods used in this study included (1) filtering to reduce random noise, instrument noise, and clutter, (2) application of a time-range gain to enhance deep reflections, (3) depth conversion and topographic correction to compensate for changes in surface elevation, and (4) migration to focus reflection energy and to correct for geometric distortion.
De-wow, frequency, and horizontal filtering methods were applied to the field data to remove noise and to preserve true GPR reflections. Instrument noise (DC-shift) was removed with a very low frequency low-cut or de-wow filter. High frequency clutter and noise, and the remaining low frequency noise, both outside the useful 200 MHz antenna frequency range, were removed with a band-pass filter. Horizontal filtering reduced background noise that was consistent from trace to trace thereby accentuating dipping reflectors.
A time-range gain was applied to the data after filtering. Time-range gain compensates for geometric spreading losses and material absorption losses of the radar signal.
Converting radar arrival-times to depth (depth-conversion) requires the radar propagation velocity. The CMP data were analyzed using a common velocity analysis method to determine radar propagation velocity (Telford and others, 1990). Analysis of CMP data from the outcrop yielded a radar propagation velocity in the bedrock outcrop of 117 meters per microsecond (m/ms). This is comparable to the propagation velocity calculated from borehole-radar experiments at Mirror Lake (118 m/ms) and to the propagation velocity for granite (100 m/ms) reported by Annan (1992). The calculated radar propagation velocity (117 m/ms) was used to establish the vertical scale of the GPR records. Following depth conversion, the elevation of each GPR trace was shifted to account for topography along the survey line.
Figure 6. (a) 200 MHz unprocessed GPR field record from the bedrock outcrop. (b) Data shown in (a) after processing and topographic correction. Interpreted reflections are labeled 1 through 3. (c) Photograph of the section of outcrop surveyed with GPR. Fractures 1 and 2 are annotated in the photograph. The depth scale was calculated assuming a radar propagation velocity of 117 m/ms. The vertical exaggeration is 1.7 times.
After the topographic correction was applied to the data, the data were migrated. The migration processing method uses the radar propagation velocity and the antenna geometry in relation to the target, and correctly locates reflectors in the subsurface.
Figure 6 shows a 40 m section of the field data from the outcrop before (fig. 6a) and after (fig. 6b) data processing. Processing improved the field data in several ways. First, true dipping reflectors were differentiated from background clutter caused by small-scale bedrock heterogeneities. Second, deep reflectors were enhanced, increasing the apparent depth of penetration of the GPR survey to at least 17 m below the top of the outcrop. Third, reflectors were correctly positioned and oriented.
Interpretation of the processed field data was based on the results of the numerical and physical modeling. Reflectors interpreted as fractures (1) displayed the reflection amplitude and phase characteristics identified in the numerical and physical modeling and (2) were linear and were continuous from trace to trace for at least 40 traces. In figure 6b, for example, three continuous reflectors (labeled 1 to 3) are interpreted as fractures. The depth and orientation of reflectors labeled 1 and 2 in the figure correlate with fractures observed on the walls of the outcrop (fig. 6c). The reflector labeled 3 cannot be confirmed by ground-truth because it projects below the land surface.
Although steeply dipping fractures are present in the outcrop, only reflectors dipping less than 45 degrees were considered in the interpretation process. This constraint, based on calculations using trace spacing, radar signal frequency and propagation velocity (Yilmaz, 1987), was imposed to prevent interpretation of reflectors whose orientation could be affected by spatial aliasing distortions. In the processed data, 18 reflectors with lengths ranging from 4 to 32 m and dips ranging from 5 to 40 degrees were interpreted as fractures. All of the fractures interpreted from the processed GPR records and projecting above land surface correlated with the ground-truth fractures provided by detailed mapping.
Radar data from 2.5-D modeling simulating unsaturated and saturated fractures were used to guide the interpretations of reflections in the GPR field data. For example, figure 7 shows a reflector interpreted as a fracture in the field data next to the results of a 2.5-D model of a fracture pair. Two fractures dipping approximately 5 degrees with apertures of 1 mm are shown; one is unsaturated, the other is saturated (fig. 7a). The amplitude, continuity, and phase of the reflector in the field data (fig. 7b) and in the modeled data for the saturated fracture are similar. The reflector in the field data is, therefore, interpreted as a saturated fracture. Most reflectors interpreted as fractures were consistent with numerical models of saturated fractures.
Figure 7. (a) Radar data modeled in 2.5-D using a 200 MHz starting wavelet. The reflectors are modeled as fractures with an aperture of 1 mm dipping 5 degrees. The results show the difference in the reflection amplitude and phase for an unsaturated (upper) and saturate (lower) fracture. (b) A reflector extracted from the GPR field record of 200 MHz data. The depth scale was calculated assuming a radar propagation velocity of 117 m/ms. The vertical exaggeration is 2 times.
A study incorporating numerical modeling, physical modeling, and field surveys at the USGS Fractured Rock Research Site at Mirror Lake, Grafton County, New Hampshire, was conducted to test the use of GPR surface reflection methods to delineate fractures in heterogeneous bedrock. Results of 1-D and 2.5-D numerical modeling correlate with results of laboratory-scale physical modeling and establish different GPR reflection characteristics for saturated and unsaturated fractures. Saturated fractures generate higher amplitude reflections than unsaturated fractures and have an opposite phase.
GPR data collected over a highway outcrop near Mirror Lake were processed to reduce noise and clutter, correct geometric and topographic distortions, and enhance weak reflections from structures more than 15 m deep. Guided by the results of numerical modeling, 18 reflectors with lengths ranging from 4 to 32 m and dips ranging from 5 to 40 degrees were interpreted as fractures in the processed field data. All of the interpreted fractures that project above land surface correlated with fractures recorded by detailed mapping. Interpretation of processed field data was limited to reflectors dipping less than 45 degrees, although more steeply dipping fractures exist at the field site. Spatial aliasing effects and other limitations constrain the range of dip-angles that GPR reflection methods can usefully image from the surface.
The results of the field study and the results of the numerical and physical modeling indicate that surface GPR methods can be used, within limits, to characterize fracture location and orientation in heterogeneous bedrock.
The authors thank Prof. Lanbo Liu at the Department of Geology and Geophysics, University of Connecticut, for his input into the data processing of the GPR field data. We especially thank Wayne Martin of the U.S. Forest Service for permitting the experiment to be conducted at the Hubbard Brook Experimental Forest, Grafton County, New Hampshire. The Hubbard Brook Experimental Forest is operated and maintained by the Northeastern Forest Experiment Station, USDA Forest Service, Radnor, Pennsylvania.
Annan, A.P., 1992, Ground penetrating radar workshop notes: Mississauga, Ontario, Sensors & Software, Inc., 128 p.
Barton, C.C., 1997, Bedrock geologic map of Hubbard Brook Experimental Forest and maps of fractures and geology in roadcuts along Interstate 93, Grafton County, New Hampshire: U.S. Geological Survey Miscellaneous Investigations Series Map I-2562, scale 1:12,000, 2 sheets.
Barton, C.C., 1996, Characterizing bedrock fractures in outcrop for studies of ground-water hydrology – an example from Mirror Lake, Grafton County, New Hampshire, in Morganwalp, D.W., and Aronson, D.A., eds., U.S. Geological Survey Toxics Substances Hydrology Program -- Proceedings of the Technical Meeting, Colorado Springs, Colorado, September 20-24, 1993: U.S. Geological Survey Water-Resources Investigations Report 94-4015, v. 1, p. 81-87.
Benson, A.K., 1995, Applications of ground-penetrating radar in assessing some geologic hazards--examples of groundwater contamination, faults, and cavities: Journal of Applied Geophysics, v. 33, p. 177-193.
Grasmueck, M., 1996, 3-D ground-penetrating radar applied to fracture imaging in gneiss: Geophysics, v. 61, no. 4, p. 1050-1064.
Holloway, A.L., 1992, Fracture mapping in granite rock using ground probing radar in Pilon, J., ed., Ground Penetrating Radar: Geological Survey of Canada Paper 90-4, p. 85-100.
Hsieh, P.A., Shapiro, A.M., Barton, C.C., Haeni, F.P., Johnson, C.D., Martin, C.W., Paillet, F.L., Winter, T.C., and Wright, D.L., 1993, Methods of characterizing fluid movement and chemical transport in fractured rock in Field Trip Guidebook for the Northeastern United States: Geological Society of America, Field Trip Guidebook for the Northeastern United States, v. 2, chap. R, 30 p.
Powers, M.H., 1995, Dispersive ground penetrating radar modeling in 2D: Golden, Colorado, Colorado School of Mines, Ph.D. Thesis T-4820.
Powers, M.H., and Olhoeft, G.R., 1995, GPRMODV2--One dimensional full waveform forward modeling of dispersive ground penetrating radar data, Version 2.0: U.S. Geological Survey Open File Report 95-58, 41 p. and floppy diskette.
Serzu, M.H., Street, P.J., Lodha, G.S., and Stevens, K.M., 1996, Characterization of a moderately fractured granitic rock using single-hole radar reflection, crosshole radar tomography, and ground penetrating radar at AECL's underground research laboratory, Pinawa, Manitoba, Canada in Sixth International Conference on Ground Penetrating Radar (GPR’96), Sendai, Japan, September 30-October 3, 1996, Proceedings: Sendai, Japan, Tohoku University, Department of Geoscience and Technology, p. 173-177.
Stevens, K.M., Lodha, G.S., Holloway, A.L., and Soonawala, N.M., 1995, The application of ground penetrating radar for mapping fractures in plutonic rocks within the Whiteshell Research Area, Pinawa, Manitoba, Canada: Journal of Applied Geophysics, v. 33, p. 125-141.
Telford, W.M., Geldart, L.P., and Sheriff, R.E., 1990, Applied Geophysics (2d ed.): New York, N. Y., Cambridge University Press, 770 p.
Toshioka, T., Tsuchida, T., and Sasahara, K., 1995, Application of GPR to detecting and mapping cracks in rock slopes: Journal of Applied Geophysics, v. 33, p. 119-124.
Ulriksen, C.P.F., 1982, Application of impulse radar to civil engineering: Lund, Sweden, Lund University of Technology, Ph.D. dissertation, 179 p., published by Geophysical Survey Systems, Inc., North Salem, N.H.
Yilmaz, O., 1987, Seismic Data Processing, ed. by Doherty, S.M., of the series Neitzel, E.B., ed., Investigations in Geophysics Series, Vol. 2: Tulsa, Okla., Society of Exploration Geophysicists, 526 p.
 The use of trade, product, or firm names in this report is for identification purposes only and does not constitute endorsement by the U.S. Geological Survey.
*Citation:Buursink, Marc L. and Lane John W., Jr., 1999, Characterizing fractures in a bedrock outcrop using ground-penetrating radar at Mirror Lake, Grafton County, New Hampshire, in Morganwalp, D.W. and Buxton, H.T., eds., U.S. Geological Survey Toxic Substances Hydrology Program -- Proceedings of the Technical Meeting, Charleston, South Carolina, March 8-12, 1999: USGS Water-Resources Investigations Report 99-4018C, v. 3, p. 769-776.