- Email: [email protected]
Long-term ground penetrating radar monitoring of a small volume DNAPL release in a natural groundwater flow field
Available online at www.sciencedirect.com
Journal of Contaminant Hydrology 97 (2008) 1 – 12 www.elsevier.com/locate/jconhyd
Long-term ground penetrating radar monitoring of a small volume DNAPL release in a natural groundwater flow field Yong Keun Hwang 1 , Anthony L. Endres ⁎, Scott D. Piggott, Beth L. Parker Department of Earth and Environmental Sciences, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 Received 2 March 2007; received in revised form 13 November 2007; accepted 22 November 2007 Available online 15 December 2007
Abstract An earlier field experiment at Canadian Forces Base Borden by Brewster and Annan [Geophysics 59 (1994) 1211] clearly demonstrated the capability of ground penetrating radar (GPR) reflection profiling to detect and monitor the formation of DNAPL layers in the subsurface. Their experiment involved a large volume release (770 L) of tetrachloroethylene into a portion of the sand aquifer that was hydraulically isolated from groundwater flow by sheet pile walls. In this study, we evaluated the ability of GPR profiling to detect and monitor much smaller volume releases (50 L). No subsurface confining structure was used in this experiment; hence, the DNAPL impacted zone was subjected to the natural groundwater flow regime. This condition allowed us to geophysically monitor the DNAPL mass loss over a 66 month period. Reflectivity variations on the GPR profiles were used to infer the presence and evolution of the solvent layers. GPR imaging found significant reflectivity increases due to solvent layer formation during the two week period immediately after the release. These results demonstrated the capacity of GPR profiling for the detection and monitoring of lesser volume DNAPL releases that are more representative of small-scale industrial spills. The GPR imaged solvent layers subsequently reduced in both areal extent and reflectivity after 29 months and almost completely disappeared by the end of the 66 month monitoring period. Total DNAPL mass estimates based on GPR profiling data indicated that the solvent mass was reduced to 34%–36% of its maximum value after 29 months; only 4%–9% of the solvent mass remained in the study area after 66 months. These results are consistent with independent hydrogeological estimates of remaining DNAPL mass based on the downgradient monitoring of the dissolved solvent phase. Hence, we have concluded that the long-term GPR reflectivity changes of the DNAPL layers are likely the result from the dissolution of chlorinated solvents residing in those layers. The long-term monitoring results demonstrated that GPR profiling is a promising non-invasive method for use at DNAPL contaminated sites in sandy aquifers where temporal information about immiscible contaminant mass depletion due to either natural flow or remediation is needed. However, our results also indicated that the GPR signature of older DNAPL impacted zones may not differ greatly from the uncontaminated background if significant mass reduction due to dissolution has occurred. © 2007 Elsevier B.V. All rights reserved. Keywords: Chlorinated solvents; DNAPL; Ground penetrating radar; Hydrogeophysics
⁎ Corresponding author. E-mail addresses: [email protected] (Y.K. Hwang), [email protected] (A.L. Endres), [email protected] (S.D. Piggott), [email protected] (B.L. Parker). 1 Presently Department of Geological Sciences, University of Michigan, Ann Arbor, Michigan, USA. 0169-7722/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jconhyd.2007.11.004
2
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
1. Introduction Contamination by immiscible organic liquids is widely recognized as a serious groundwater quality issue (Domenico and Schwartz, 1998). These liquids are divided into two categories on the basis of their densities. Liquids with specific gravities less than water, such as gasoline and jet fuel, are referred to as LNAPLs (light non-aqueous phase liquids). These immiscible contaminants are largely confined to the region above the water table due to their buoyancy. Conversely, DNAPLs (dense non-aqueous phase liquids) have specific gravities greater than water. Hence, they migrate downwards through the saturated zone. As they descend, DNAPLs form immiscible pools at the boundaries between permeability contrasts. These pools accumulate until there is sufficient pressure differential to initiate downward drainage of the pool through this permeability boundary. The pool empties until a state of residual saturation occurs, at which point the remaining DNAPL in the pool becomes immobile (Kueper and Frind, 1991). The residual saturation of DNAPL in sands commonly ranges between 1% and 15% (Kueper et al., 1993). Chlorinated solvents are a class of DNAPLs that have been recognized as a major group of groundwater contaminant since the early 1980s (Feenstra and Cherry, 1988; Mackay and Cherry, 1989). The immiscible phase pools or layers of these solvents may remain in an aquifer for many decades as sources of dissolved phase contamination due to their low solubility values in water (Feenstra et al., 1996). In spite of its low solubility, the dissolved phase of a chlorinated solvent is a significant threat to the drinking water quality. For example, the maximum contaminant level (MCL) of trichloroethylene (TCE) for drinking water is 5 parts per billion (ppb) according to the U. S. Environmental Protection Agency (2006) while its solubility is 1100 ppb (Domenico and Schwartz, 1998). As a result, the characterization of the immiscible solvent phase within the subsurface is an important issue with regard to groundwater resources and protection. Because the physical properties of immiscible organic liquids significantly differ from those of the groundwater they displace, geophysical methods have proven useful for the detection and monitoring of these contaminants. In particular, the relative dielectric permittivities of immiscible organic liquids (κ = 2–4) are very low compared to the relative permittivity of water (κ = 80) (Lucius et al., 1992). Ground penetration radar (GPR) profiling has been found to be an effective geophysical method for imaging immiscible organic liquid layers in the subsurface because electromagnetic (EM) wave propagation is strongly de-
pendent on the subsurface dielectric properties. Furthermore, GPR profiling is a non-invasive technique that is capable of providing high-resolution images of the subsurface over a significantly large survey area. A sizable body of literature currently exists on the application of GPR profiling to the detection and monitoring of LNAPL contaminants. These studies report the results of controlled release experiments (e.g., Kim et al., 2000), field studies of actual impacted sites (e.g., Lopes de Castro and Castelo Branco, 2003) and advanced data analysis techniques for detecting and quantifying contaminant pools (e.g., Jordan et al., 2004). In comparison, the amount of literature on the application of GPR profiling to DNAPL contaminants is much smaller. The most notable paper is the field experiment conducted by Brewster and Annan (1994) that clearly demonstrated the effectiveness of GPR profiling for detecting the presence of tetrachloroethylene (PCE) in the subsurface and monitoring its movement in a natural aquifer. Their experiment released 770 L of solvent into a portion (9 × 9 × 3.3 m depth) of the Borden aquifer that was hydraulically isolated by sealable-joint sheet pile walls that were fixed into the underlying aquitard. They geophysically monitored the solvent movement within the cell for approximately two months after the release. A subsequent paper by Brewster et al. (1995) described how their GPR profiling data could be used to obtained quantitative estimates of the immiscible phase DNAPL mass. The study we present in this paper further evaluates the capacity of GPR profiling as a geophysical method for characterizing DNAPL impacted sites. In contrast to the large solvent volume released by the earlier Brewster and Annan study, we injected only 50 L of chlorinated solvent, an amount that is more representative of small-scale industrial spills. In addition, no containment structures were used in our experiment to isolate the solvent impacted zone from the natural groundwater flow regime. Hence, we were able to examine the sensitivity of GPR profiling to DNAPL mass reductions resulting from dissolution and its ability to provide quantitative information about these changes. Finally, our much longer 66 month monitoring period permitted us to evaluate the potential of GPR profiling as a geophysical tool for characterizing older impacted sites that have been subjected to the effects of groundwater flow. 2. Experiment description Our experiment site was located in the unconfined aquifer at Canadian Force Base (CFB) Borden which is situated approximately 90 km northwest of Toronto, Ontario, Canada. The unconfined aquifer is composed
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
primarily of medium-grained sand of glacio-deltaic or glacio-fluvial origin; it can be locally heterogeneous due to lenses and beds of fine-, medium-, and coarse-grained sand (Macfarlane et al., 1983). A low degree of spatial variability in hydraulic conductivity exists among the discontinuous beds and lenses that are up to several meters in length; the measured conductivity values range between 2 × 10− 4 and 6 × 10− 6 m/s (Sudicky, 1986). The aquifer has a porosity of 0.33 ± 0.017 (Mackay et al., 1986). The aquifer is approximately 3.5 m thick at our site; it is underlain by a clayey silt aquitard. The water table depth seasonally varies between 0.5 and 1.5 m below the ground surface. A previous experiment at this site by Laukonen et al. (2000) found groundwater velocities between 0.08 and 0.16 m/day. Our site was located within 25 m of two previous experimental DNAPL releases (Brewster and Annan, 1994; Kueper et al., 1993) that were both conducted inside confining sheet pile structures. Our experiment was initiated on April 8, 1999 with the release of a 50 L mixture of chlorinated solvents (tetrachloroethylene (PCE), trichloroethylene (TCE) and chloroform (TCM)) through an open-ended pipe at a depth of 1.8 m below the ground surface (approximately 0.5 m below the watertable at the release time). GPR profiling was conducted before and after the solvent injection at the following six times: April 7–8, 1999 (before the injection); April 8, 1999 (1/2 day after); April 9, 1999 (1 day after); April 22, 1999 (2 weeks after); September 19, 2001 (29 months after); and September 28–October 14, 2004 (66 months after). The GPR monitoring was performed over a
3
4.2 m by 4.25 m area (Fig. 1) that was covered by eighteen profile lines (Line #1–Line #18) oriented in a NW–SE direction. The spacing between profile lines in the SW–NE direction was 0.25 m. The solvent injection point was located near the middle of the survey area between Lines #11 and #12 (2.12 m SE, 2.58 m NE). GPR profiling data was acquired using a Sensors and Software Inc. pulseEKKO IV system during the 1999 and 2001 surveys; a pulseEKKO 100 system was used for the 2004 survey. Both systems were equipped with 200 megahertz (MHz) antennas and a 400 V transmitter. The station spacing along the 4.2 m long profile lines was 0.1 m. The time sampling rate was 0.8 ns, and 64 stack was performed at each station position. The antenna separation used during profiling was 0.5 m; the standard perpendicular-broadside antenna orientation was employed to acquire the GPR data. In addition to reflection profiling, common midpoint (CMP) data were acquired to determine the EM wave velocities of the aquifer material; this value was used to estimate depth to the reflecting interface from the traveltime information. These data were analyzed using the normal moveout (NMO) technique (Yilmaz, 2001). The average EM wave velocity obtained was 0.062 meters/nanosecond (m/ns); this value is consistent with the EM wave velocity of 0.063 m/ns for the Borden aquifer given by Brewster and Annan (1994). 3. Data processing GPR profiling data processing was done using the Sensors and Software pulseEKKO (v. 4.22) and EKKO
Fig. 1. Plan view of the experiment area showing the location of the injection well and the GPR profile lines.
4
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
View Deluxe packages. The following procedures were applied to the profiling data in the sequence listed below; detailed information about these geophysical data processing techniques is found in Yilmaz (2001) and Sensors and Software (2003). A zero-time correction was applied to compensate for instrument drift. The low frequency ‘wow’ superimposed on the higher frequency reflections was removed by applying a signal saturation correction. Horizontal spatial (two trace average) and vertical time (two time sample average) filters were applied to enhance reflections and reduce random noise. A spherical and exponential compensation (SEC) gain was applied to compensate for amplitude changes due to geometrical spreading and ohmic dissipation while preserving the relative reflectivity of the subsurface boundaries. The SEC gain parameters were selected such that consistent amplitude information was obtained for all datasets. This objective was achieved by performing an amplitude balancing analysis on aquifer reflections at locations not impacted by the solvent release. The value of attenuation coefficient in the SEC gain was adjusted for each monitoring time such that temporal variability in the reflectivity of these events was minimized. A second set of GPR profile images was obtained by determining the instantaneous amplitude (i.e., envelope) of the individual GPR traces. The instantaneous amplitude is a complex trace attribute that removes the effects of wavelet polarity and phase, allowing a more accurate determination of reflection strength. A discussion of complex trace theory and its application to seismic reflection profiling is given by Sheriff and Geldart (1995). Previously, Orlando (2002) had successfully used complex trace attributes for the analysis of GPR profiling data from LNAPL impacted sites. 4. Temporal changes in GPR response The GPR profiles acquired along Line #11, which is near the middle of the grid and the injection point, are presented in Fig. 2 for the six survey times. The GPR profile before the injection (Fig. 2A) shows the aquifer stratigraphy at our site. The reflection from the underlying clay aquitard occurs at approximately 118 ns twoway traveltime. Although small-scale heterogeneities are evident in the pre-release profile, the stratigraphic boundary reflectivity within the aquifer is low.
5
The formation of the immiscible chlorinated solvent layers and their subsequent dissipation are clearly observed over time on the post-release profiles (Fig. 2B– F). The aquifer reflectivity increased during the first day as the solvent moved downward and formed layers (Fig. 2B and C). During this time, two closely spaced immiscible solvent pools were identified between 78 ns and 100 ns. The positions of these DNAPL layers coincide with subtle stratigraphic boundaries imaged on the pre-release profile (Fig. 2A), demonstrating the importance of stratigraphy and the associated permeability contrast in controlling the DNAPL movement. The solvent layer reflectivity is largely unchanged two weeks after the release (Fig. 2D). During the later stages of the GPR monitoring, the solvent layer reflectivities have significantly diminished after 29 months (Fig. 2E) and eventually reached levels that are comparable to that of the stratigraphic interfaces on the preinjection after 66 months (Fig. 2F). The complex trace instantaneous amplitude profiles along the Line #11 are presented in Fig. 3. The instantaneous amplitude changes are consistent with the reflectivity variations seen in Fig. 2. There is a low level of instantaneous amplitude in the pool locations before the injection (Fig. 3A). The instantaneous amplitude of the DNAPL layers increases as the solvent accumulates in them during the first day (Fig. 3B and C) and maintains its high level for the two weeks after the injection (Fig. 3D). The instantaneous amplitude of the solvent layers subsequently diminishes during the longer-term monitoring (Fig. 3E and F). The spatial distribution of the upper and lower DNAPL layers was inferred by comparing the background and post-release profiles over the entire GPR profiling grid. Both the maximum reflection amplitude and complex trace instantaneous amplitude for the reflection event along the stratigraphic horizon corresponding to each solvent layer were manually determined for each trace along the GPR profiles using the Sensors and Software Picker software. The reflectivity variation of the DNAPL layers during the experiment were quantified for each monitoring time by a simple differencing of the background and post-release maximum amplitude values at each trace position. Let us first consider the GPR imaged spatial changes for the upper DNAPL layer. Fig. 4 shows a plan view of the upper layer evolution using the reflection amplitude
Fig. 2. GPR profiling along the survey Line #11 for pre-release (a) and post-release (b–f) times. Position axis units are meters; time axis units are nanoseconds. Solid box indicates region of solvent layer development. The injection well is located adjacent to the profile at 2.12 m. Post-release times: (b) 1/2 day, (c) 1 day, (d) 2 weeks, (e) 29 months and (f) 66 months.
6
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Fig. 3. Complex trace instantaneous amplitude profiles of the GPR data along the survey Line #11 for pre-release (a) and post-release (b–f) times. Position axis units are meters; time axis units are nanoseconds. Solid box indicates region of solvent layer development. The injection well is located adjacent to the profile at 2.12 m. Post-release times: (b) 1/2 day, (c) 1 day, (d) 2 weeks, (e) 29 months and (f) 66 months.
differencing; the result of complex trace instantaneous amplitude differencing is shown in Fig. 5. The results obtained from these two reflectivity measures are
mutually consistent, displaying the same large-scale features and temporal changes. Both Figs. 4A–C and 5A–C show the development of a laterally extensive
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
7
Fig. 4. Plan view of the upper solvent layer based on reflection amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
upper layer during the first two weeks after the release. The apparent shift in the center of mass of this layer during the initial two week period is likely due to spatial differences in downward migration rates resulting from localized stratigraphic features or the presence of old root zones. The upper layer subsequently reduced in size and reflectivity after 29 months (Figs. 4D and 5D) and almost completely disappeared by the end of the GPR monitoring period at 66 months (Figs. 4E and 5E). The GPR imaged evolution of the lower DNAPL layer is illustrated in Figs. 6 and 7 which give plan views of the reflection amplitude and the complex trace instantaneous amplitude differencing results, respectively. The results for the lower layer are analogous to those obtained for the upper layer. During the first
two weeks, the development of a laterally extensive lower layer was observed (Figs. 6A–C and 7A–C). The lower layer showed significant reductions in extent and reflectivity after 29 months (Figs. 6D and 7D) and nearly vanished after 66 months (Figs. 6E and 7E). 5. GPR estimate of solvent pool mass In principal, the DNAPL layer reflection amplitude can be used to quantitatively estimate the local immiscible phase saturation and the total DNAPL mass. Brewster et al. (1995) developed a linear relationship between reflection amplitude and the local solvent saturation-layer thickness product for individual pools
8
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Fig. 5. Plan view of the upper solvent layer based on complex trace instantaneous amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
at CFB Borden and used it to estimate the total DNAPL mass from their GPR profiling data. A similar method was applied to our GPR profiling data to estimate the DNAPL mass changes over the 66 month monitoring period. Our study area was divided into a uniform rectangular grid with all elements having horizontal dimensions Δx and Δy. The mass of a DNAPL layer in the i, jth grid element (mij) is related to solvent density (ρDNAPL), aquifer porosity (ϕ), local DNAPL saturation (Sij) and local layer thickness (Tij) by the following expression:
On the basis of a GPR modeling study, Brewster et al. (1995) proposed that the local amplitude of the GPR reflection from a solvent layer (Aij) was related to elemental saturation-layer thickness product (SijTij) by
mij ¼ qDNAPL /Sij Tij DxDy:
Sij Tij ¼ bAij
ð1Þ
The total DNAPL mass (MDNAPL) within our study area can be determined by summing the elemental DNAPL mass over the area of all solvent layers: MDNAPL ¼ qDNAPL /DxDy
X XX layers
i
Sij Tij :
ð2Þ
j
ð3Þ
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
9
Fig. 6. Plan view of the lower solvent layer based on reflection amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
where b is a coefficient determined from a comparison between DNAPL saturation profiles obtained from core samples and coincident GPR profiling data. Hence, the total DNAPL mass can be obtained from the GPR reflection amplitude data using X XX MDNAPL ¼ qDNAPL /bDxDy Aij : ð4Þ layers
i
total DNAPL mass MMax, we obtain the following expression for the relative total DNAPL mass (MRel): PP P i j Aij M layers : MRel ¼ ¼ ð5Þ PP P MMax A ij i j layers
j
The While no calibration cores were obtained during our experiment that coincided with the GPR profiling times, our GPR data can be used to determine changes in relative DNAPL mass in the survey area. It can be seen from Eq. (4) MDNAPL is directly proportional to the P that P P quantity i i Aij. If our total DNAPL mass layers
computation is normalized by the maximum observed
PP P layers
i
i
Max
Aij term was evaluated for all five
surveys performed after the injection using both the reflection amplitude and complex trace instantaneous amplitude data. To minimize the potential contribution of stratigraphic contrasts to the computations, differencing between the background and post-release profiles was used in this analysis. We found that MMax occurred one day after injection for both measures of reflectivity
10
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Fig. 7. Plan view of the lower solvent layer based on complex trace instantaneous amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
change; hence, these values were used to normalize the relative mass estimates. The relative total mass of the chlorinated solvent layers remaining in the experiment site estimated from both the reflection amplitude and the complex trace instantaneous amplitude data are given in Table 1. The GPR-estimated relative total DNAPL mass increased rapidly during the first day and only decreased slightly after two weeks. The longer-term GPR monitoring data indicated significant reductions in the total DNAPL mass. The remaining total solvent mass at 29 months post-release was estimated to be 34%–36% of its maximum amount. After 66 months, the total DNAPL mass remaining in the layers was further diminished to 4%–9% of its maximum value. We compared our results with an independent estimate of total DNAPL mass determined from an analysis of groundwater samples. This groundwater sampling was
performed along a cross-sectional, multi-level piezometer network located immediately downgradient of our DNAPL release. The water sampling started one month after the injection and continued for 18 sampling events over the subsequent five year period. The analysis of the groundwater samples estimated that the DNAPL mass remaining in the experiment area was approximately 40% of the initial DNAPL mass two years after the release and 10% of the initial DNAPL mass five years post-release. The similarity of these two independent measures of DNAPL mass strongly implies that the observed GPR reflectivity changes of the DNAPL layers were primarily due to dissolution occurring in the groundwater flow field. It should be noted that while both the GPR profiling and groundwater sampling gave similar results, the GPR profiling required a much smaller amount of effort and expense than the groundwater sampling program which
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12 Table 1 Relative total DNAPL mass remaining in the experiment site inferred from the GPR profiling data X XX Post-release time Aij Relative total DNAPL j layers i mass (MRel) a.) Reflection amplitude 1/2 Day 643,188 1 Day 1,790,680 2 Weeks 1,716,053 29 Months 638,908 66 Months 79,524
35.9% 100.0% 95.8% 35.7% 4.4%
b.) Instantaneous amplitude 1/2 Day 290,603 1 Day 853,513 2 Weeks 797,043 29 Months 286,360 66 Months 76,451
34.0% 100.0% 93.4% 33.6% 9.0%
Reflectivity measure used: (a) reflection amplitude and (b) complex trace instantaneous amplitude (i.e., envelope).
involved collection and analysis of thousands of individual water samples. Furthermore, GPR profiling is a non-invasive method whereas groundwater sampling requires the installation of monitoring wells; hence, GPR profiling has the capacity to quantitatively characterize DNAPL contamination at sandy aquifer sites where the emplacement of monitoring wells is restricted or prohibited. Because of the inherent limitations of traditional hydrogeological sampling techniques, DNAPL zone depletion due to groundwater dissolution under natural flow conditions has been determined at only a few of the many thousands of actual DNAPL sites existing in sandy aquifers around the world; GPR profiling offers a practical and cost-effective means of closing this large information gap. 6. Discussion and conclusions Our experimental results clearly demonstrate that GPR profiling has a greater potential for characterizing DNAPL impacted sites than previously indicated in the landmark papers by Brewster and Annan (1994) and Brewster et al. (1995). We have shown that GPR profiling has the ability to detect DNAPL releases having lesser volumes that are more representative of small-scale industrial spills. Even though our injected DNAPL volume was only 50 L, GPR profiling detected significant reflectivity increases due to the formation of laterally extensive solvent layers along stratigraphic boundaries during the two week period after the release. Since the DNAPL layers were subjected to the natural groundwater flow regime, we were able to assess
11
the sensitivity of GPR profiling to DNAPL mass reductions resulting from dissolution. The long-term GPR monitoring (29 and 66 months after the injection) observed the significant reduction and almost complete dissipation in the solvent layer reflectivity. The spatial extent of the DNAPL layers determined from GPR profiling significantly reduced in size after 29 months and nearly vanished by the end of 66 months. Using a method similar to that proposed by Brewster et al. (1995), reflection amplitude and the complex trace attribute instantaneous amplitude (i.e., envelope) were used to estimate total DNAPL mass variations from the DNAPL layer reflectivity observed on the GPR profiles; both of these approaches gave mutually consistent results. Due to the lack of appropriate calibration information, relative total DNAPL mass changes were determined for the survey area. The GPR-estimated total DNAPL mass reached its maximum value during the first day and remained high over the first two weeks following the release. Longer-term GPR monitoring shows a significant reduction in total DNAPL mass. It was estimated that only 34%–36% of the total DNAPL mass remained in the survey area after 29 months. After 66 months, the total DNAPL mass further diminished to 4%–9% of its maximum amount. The GPR estimates of total DNAPL mass are consistent with an independent hydrogeological estimate of remaining solvent mass based on the monitoring of the dissolved phase downgradient of the release. Hence, we conclude that the long-term changes in GPR reflectivity of the DNAPL layers most probably result from the dissolution of those chlorinated solvents due to groundwater flow. Furthermore, the corresponding dissipation of the DNAPL layer reflectivity indicates that the GPR signature of older DNAPL impacted zones may not differ greatly from the uncontaminated background if significant mass reduction has occurred. One question that has been raised recently in the geophysical monitoring of immiscible contaminant layers is the potential for dramatic geophysical changes due to biodegradation. Bermejo et al. (1997) and Sauck et al. (1998) found zones of GPR signal attenuation (shadow zones) associated with aged light non-aqueous phase liquid (LNAPL) pools. These shadow zones result from enhanced electrical conductivity coincident with hydrocarbon degradation through microbial reactions (Atekwana et al., 2000, 2004). We did not observe any “shadow zone” of the type reported by these studies. In addition, groundwater monitoring of dissolved volatile organic compounds (VOC) constituents emanating from the source over time does not show degradation products, supporting the conclusion that significant biodegradation activity did not
12
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
occur in the source zone over the five year time period of our experiment. In conclusion, our results clearly demonstrate the ability of GPR profiling to detect and monitor a small volume DNAPL release into a sandy aquifer and to quantify longterm depletion of the DNAPL mass due to dissolution. At many actual DNAPL sites, there is a need to determine the DNAPL mass depletion rate due to dissolution in natural groundwater flow or enhanced mass removal through remediation; this information is important for assessment of long-term impacts and remediation potential. GPR profiling has the potential to obtain this goal in a non-invasive and more efficient manner in comparison to traditional hydrogeological sampling techniques. Acknowledgments The GPR data acquisition and analysis was supported by a Collaborative Research and Development Grant from the Natural Sciences and Engineering Research Council of Canada and a grant from the University Consortium Solvents-in-Groundwater Research Program awarded to Dr. Parker. Mr. Hwang's studies were partially supported by an Individual Research Grant to Dr. Endres from the Natural Sciences and Engineering Research Council of Canada. The authors wish to thank the Editor and two anonymous reviewers for comments that have significantly improved the quality of this paper. References Atekwana, E.A., Sauck, W.A., Werkema, D.D., 2000. Investigations of geoelectrical signatures at a hydrocarbon contaminated site. Journal of Applied Geophysics 44, 167–180. Atekwana, E.A., Werkema, D.D., Duris, J.W., Rossbach, S., Atekwana, E.A., Sauck, W.A., Cassidy, D.P., Means, J., Legall, F.D., 2004. Insitu apparent conductivity measurements and microbial population distribution at a hydrocarbon-contaminated site. Geophysics 69, 56–63. Bermejo, J.L., Sauck, W.A., Atekwana, E.A., 1997. Geophysical discovery of a new LNAPL plume at the former Wurtsmith AFB, Oscoda, Michigan. Ground Water Monitoring & Remediation 17, 131–137. Brewster, M.L., Annan, A.P., 1994. Ground-penetrating radar monitoring of a controlled DNAPL release: 200 MHz radar. Geophysics 59, 1211–1221. Brewster, M.L., Annan, A.P., Greenhouse, J.P., Kueper, B.H., Olhoeft, G.R., Redman, J.D., Sander, K.A., 1995. Observed migration of a controlled DNAPL release by geophysical methods. Ground Water 33, 977–987. Domenico, P.A., Schwartz, F.W., 1998. Physical and Chemical Hydrogeology. John Wiley & Sons, New York. Feenstra, S., Cherry, J.A., 1988. Subsurface contamination by dense nonaqueous phase liquid (DNAPL) chemicals. International Groundwater Symposium. International Association of Hydrogeologists 61–69.
Feenstra, S., Cherry, J.A., Parker, B.L., 1996. Conceptual models for the behavior of DNAPLs in the subsurface. In: Pankow, J.F., Cherry, J.A. (Eds.), In Dense Chlorinated Solvents and Other DNAPLs in Groundwater: History, Behavior and Remediation. Waterloo Educational Services, Rockwood, Ontario, pp. 53–88. Jordan, T.E., Baker, G.S., Henn, K., Messier, J.P., 2004. Using amplitude variation with offset and normalized residual polarization analysis of ground penetrating radar data to differentiate an NAPL release from stratigraphic changes. Journal of Applied Geophysics 56, 41–58. Kim, C., Daniels, J.J., Guy, E.D., Radzevicius, S.J., Holt, J., 2000. Residual hydrocarbons in a water saturated medium: a detection strategy using ground penetrating radar. Environmental Geosciences 7, 169–176. Kueper, B.H., Frind, E.O., 1991. Two-phase flow in heterogeneous porous media, 2. Model application. Water Resources Research 27, 1059–1070. Kueper, B.H., Redman, J.D., Starr, R.C., Reitsma, S., Mah, M., 1993. A field experiment to study the behavior of tetrachloroethylene below the water table: Spatial distribution of residual and pool DNAPL. Ground Water 31, 756–766. Laukonen, K.A., Parker, B.L., Cherry, J.A., 2000. Internal characteristics of a bromide tracer zone during natural flow in the Borden aquifer, Ontario, Canada. Tracers and Modelling in Hydrogeology. Proceedings of TraM'2000 Conference. IAHS Publication, Liege, Belgium, pp. 227–233. No. 262. Lopes de Castro, D., Castelo Branco, R.M.G., 2003. 4-D ground penetrating radar monitoring of a hydrocarbon leakage site in Fortaleza (Brazil) during its remediation process: a case history. Journal of Applied Geophysics 54, 127–144. Lucius, J.E., Olhoeft, G.R., Hill, P.L., Duke, S.K., 1992. Properties and Hazards of 108 Selected Substances. U. S. Geological Survey Open File 92-0527. Macfarlane, D.S., Cheery, J.A., Gillham, R.W., Sudicky, E.A., 1983. Migration of contaminants in groundwater at a landfill: a case study. Journal of Hydrology 63, 1–29. Mackay, D.M., Cherry, J.A., 1989. Groundwater contamination: limits of pump and treat remediation. Environmental Science & Technology 23, 630–636. Mackay, D.M., Freyberg, D.L., Roberts, P.V., Cherry, J.A., 1986. A natural gradient experiment on solute transport in a sand aquifer 1. Approach and overview of plume movement. Water Resources Research 22, 2017–2029. Orlando, L., 2002. Detection and analysis of LNAPL using the instantaneous amplitude and frequency of ground-penetrating radar data. Geophysical Prospecting 50, 27–41. Sauck, W.A., Atekwana, E.A., Nash, M.S., 1998. High conductivities associated with LNAPL plume imaged by integrated geophysical techniques. Journal of Environmental and Engineering Geophysics 2, 203–212. Sheriff, R.E., Geldart, L.P., 1995. Exploration Seismology. Cambridge University Press, Cambridge. Sensors and Software, Inc., 2003. EKKO View Enhanced & EKKO View Deluxe Users Guide. Sudicky, E.A., 1986. A natural gradient experiment on solute transport in a sand aquifer: spatial variability of hydraulic conductivity and its role in the dispersion process. Water Resources Research 22, 2069–2082. U. S. Environmental Protection Agency, 2006. Consumer Factsheet on: Trichloroethylene. http://www.epa.gov/safewater/dwh/c-voc/ trichlor.html, Accessed on November 29, 2006. Yilmaz, Ö., 2001. Seismic Data Analysis: Processing, Inversion and Interpretation of Seismic Data, Volume 1. Society of Exploration Geophysicists, Tulsa.
Journal of Contaminant Hydrology 97 (2008) 1 – 12 www.elsevier.com/locate/jconhyd
Long-term ground penetrating radar monitoring of a small volume DNAPL release in a natural groundwater flow field Yong Keun Hwang 1 , Anthony L. Endres ⁎, Scott D. Piggott, Beth L. Parker Department of Earth and Environmental Sciences, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 Received 2 March 2007; received in revised form 13 November 2007; accepted 22 November 2007 Available online 15 December 2007
Abstract An earlier field experiment at Canadian Forces Base Borden by Brewster and Annan [Geophysics 59 (1994) 1211] clearly demonstrated the capability of ground penetrating radar (GPR) reflection profiling to detect and monitor the formation of DNAPL layers in the subsurface. Their experiment involved a large volume release (770 L) of tetrachloroethylene into a portion of the sand aquifer that was hydraulically isolated from groundwater flow by sheet pile walls. In this study, we evaluated the ability of GPR profiling to detect and monitor much smaller volume releases (50 L). No subsurface confining structure was used in this experiment; hence, the DNAPL impacted zone was subjected to the natural groundwater flow regime. This condition allowed us to geophysically monitor the DNAPL mass loss over a 66 month period. Reflectivity variations on the GPR profiles were used to infer the presence and evolution of the solvent layers. GPR imaging found significant reflectivity increases due to solvent layer formation during the two week period immediately after the release. These results demonstrated the capacity of GPR profiling for the detection and monitoring of lesser volume DNAPL releases that are more representative of small-scale industrial spills. The GPR imaged solvent layers subsequently reduced in both areal extent and reflectivity after 29 months and almost completely disappeared by the end of the 66 month monitoring period. Total DNAPL mass estimates based on GPR profiling data indicated that the solvent mass was reduced to 34%–36% of its maximum value after 29 months; only 4%–9% of the solvent mass remained in the study area after 66 months. These results are consistent with independent hydrogeological estimates of remaining DNAPL mass based on the downgradient monitoring of the dissolved solvent phase. Hence, we have concluded that the long-term GPR reflectivity changes of the DNAPL layers are likely the result from the dissolution of chlorinated solvents residing in those layers. The long-term monitoring results demonstrated that GPR profiling is a promising non-invasive method for use at DNAPL contaminated sites in sandy aquifers where temporal information about immiscible contaminant mass depletion due to either natural flow or remediation is needed. However, our results also indicated that the GPR signature of older DNAPL impacted zones may not differ greatly from the uncontaminated background if significant mass reduction due to dissolution has occurred. © 2007 Elsevier B.V. All rights reserved. Keywords: Chlorinated solvents; DNAPL; Ground penetrating radar; Hydrogeophysics
⁎ Corresponding author. E-mail addresses: [email protected] (Y.K. Hwang), [email protected] (A.L. Endres), [email protected] (S.D. Piggott), [email protected] (B.L. Parker). 1 Presently Department of Geological Sciences, University of Michigan, Ann Arbor, Michigan, USA. 0169-7722/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jconhyd.2007.11.004
2
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
1. Introduction Contamination by immiscible organic liquids is widely recognized as a serious groundwater quality issue (Domenico and Schwartz, 1998). These liquids are divided into two categories on the basis of their densities. Liquids with specific gravities less than water, such as gasoline and jet fuel, are referred to as LNAPLs (light non-aqueous phase liquids). These immiscible contaminants are largely confined to the region above the water table due to their buoyancy. Conversely, DNAPLs (dense non-aqueous phase liquids) have specific gravities greater than water. Hence, they migrate downwards through the saturated zone. As they descend, DNAPLs form immiscible pools at the boundaries between permeability contrasts. These pools accumulate until there is sufficient pressure differential to initiate downward drainage of the pool through this permeability boundary. The pool empties until a state of residual saturation occurs, at which point the remaining DNAPL in the pool becomes immobile (Kueper and Frind, 1991). The residual saturation of DNAPL in sands commonly ranges between 1% and 15% (Kueper et al., 1993). Chlorinated solvents are a class of DNAPLs that have been recognized as a major group of groundwater contaminant since the early 1980s (Feenstra and Cherry, 1988; Mackay and Cherry, 1989). The immiscible phase pools or layers of these solvents may remain in an aquifer for many decades as sources of dissolved phase contamination due to their low solubility values in water (Feenstra et al., 1996). In spite of its low solubility, the dissolved phase of a chlorinated solvent is a significant threat to the drinking water quality. For example, the maximum contaminant level (MCL) of trichloroethylene (TCE) for drinking water is 5 parts per billion (ppb) according to the U. S. Environmental Protection Agency (2006) while its solubility is 1100 ppb (Domenico and Schwartz, 1998). As a result, the characterization of the immiscible solvent phase within the subsurface is an important issue with regard to groundwater resources and protection. Because the physical properties of immiscible organic liquids significantly differ from those of the groundwater they displace, geophysical methods have proven useful for the detection and monitoring of these contaminants. In particular, the relative dielectric permittivities of immiscible organic liquids (κ = 2–4) are very low compared to the relative permittivity of water (κ = 80) (Lucius et al., 1992). Ground penetration radar (GPR) profiling has been found to be an effective geophysical method for imaging immiscible organic liquid layers in the subsurface because electromagnetic (EM) wave propagation is strongly de-
pendent on the subsurface dielectric properties. Furthermore, GPR profiling is a non-invasive technique that is capable of providing high-resolution images of the subsurface over a significantly large survey area. A sizable body of literature currently exists on the application of GPR profiling to the detection and monitoring of LNAPL contaminants. These studies report the results of controlled release experiments (e.g., Kim et al., 2000), field studies of actual impacted sites (e.g., Lopes de Castro and Castelo Branco, 2003) and advanced data analysis techniques for detecting and quantifying contaminant pools (e.g., Jordan et al., 2004). In comparison, the amount of literature on the application of GPR profiling to DNAPL contaminants is much smaller. The most notable paper is the field experiment conducted by Brewster and Annan (1994) that clearly demonstrated the effectiveness of GPR profiling for detecting the presence of tetrachloroethylene (PCE) in the subsurface and monitoring its movement in a natural aquifer. Their experiment released 770 L of solvent into a portion (9 × 9 × 3.3 m depth) of the Borden aquifer that was hydraulically isolated by sealable-joint sheet pile walls that were fixed into the underlying aquitard. They geophysically monitored the solvent movement within the cell for approximately two months after the release. A subsequent paper by Brewster et al. (1995) described how their GPR profiling data could be used to obtained quantitative estimates of the immiscible phase DNAPL mass. The study we present in this paper further evaluates the capacity of GPR profiling as a geophysical method for characterizing DNAPL impacted sites. In contrast to the large solvent volume released by the earlier Brewster and Annan study, we injected only 50 L of chlorinated solvent, an amount that is more representative of small-scale industrial spills. In addition, no containment structures were used in our experiment to isolate the solvent impacted zone from the natural groundwater flow regime. Hence, we were able to examine the sensitivity of GPR profiling to DNAPL mass reductions resulting from dissolution and its ability to provide quantitative information about these changes. Finally, our much longer 66 month monitoring period permitted us to evaluate the potential of GPR profiling as a geophysical tool for characterizing older impacted sites that have been subjected to the effects of groundwater flow. 2. Experiment description Our experiment site was located in the unconfined aquifer at Canadian Force Base (CFB) Borden which is situated approximately 90 km northwest of Toronto, Ontario, Canada. The unconfined aquifer is composed
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
primarily of medium-grained sand of glacio-deltaic or glacio-fluvial origin; it can be locally heterogeneous due to lenses and beds of fine-, medium-, and coarse-grained sand (Macfarlane et al., 1983). A low degree of spatial variability in hydraulic conductivity exists among the discontinuous beds and lenses that are up to several meters in length; the measured conductivity values range between 2 × 10− 4 and 6 × 10− 6 m/s (Sudicky, 1986). The aquifer has a porosity of 0.33 ± 0.017 (Mackay et al., 1986). The aquifer is approximately 3.5 m thick at our site; it is underlain by a clayey silt aquitard. The water table depth seasonally varies between 0.5 and 1.5 m below the ground surface. A previous experiment at this site by Laukonen et al. (2000) found groundwater velocities between 0.08 and 0.16 m/day. Our site was located within 25 m of two previous experimental DNAPL releases (Brewster and Annan, 1994; Kueper et al., 1993) that were both conducted inside confining sheet pile structures. Our experiment was initiated on April 8, 1999 with the release of a 50 L mixture of chlorinated solvents (tetrachloroethylene (PCE), trichloroethylene (TCE) and chloroform (TCM)) through an open-ended pipe at a depth of 1.8 m below the ground surface (approximately 0.5 m below the watertable at the release time). GPR profiling was conducted before and after the solvent injection at the following six times: April 7–8, 1999 (before the injection); April 8, 1999 (1/2 day after); April 9, 1999 (1 day after); April 22, 1999 (2 weeks after); September 19, 2001 (29 months after); and September 28–October 14, 2004 (66 months after). The GPR monitoring was performed over a
3
4.2 m by 4.25 m area (Fig. 1) that was covered by eighteen profile lines (Line #1–Line #18) oriented in a NW–SE direction. The spacing between profile lines in the SW–NE direction was 0.25 m. The solvent injection point was located near the middle of the survey area between Lines #11 and #12 (2.12 m SE, 2.58 m NE). GPR profiling data was acquired using a Sensors and Software Inc. pulseEKKO IV system during the 1999 and 2001 surveys; a pulseEKKO 100 system was used for the 2004 survey. Both systems were equipped with 200 megahertz (MHz) antennas and a 400 V transmitter. The station spacing along the 4.2 m long profile lines was 0.1 m. The time sampling rate was 0.8 ns, and 64 stack was performed at each station position. The antenna separation used during profiling was 0.5 m; the standard perpendicular-broadside antenna orientation was employed to acquire the GPR data. In addition to reflection profiling, common midpoint (CMP) data were acquired to determine the EM wave velocities of the aquifer material; this value was used to estimate depth to the reflecting interface from the traveltime information. These data were analyzed using the normal moveout (NMO) technique (Yilmaz, 2001). The average EM wave velocity obtained was 0.062 meters/nanosecond (m/ns); this value is consistent with the EM wave velocity of 0.063 m/ns for the Borden aquifer given by Brewster and Annan (1994). 3. Data processing GPR profiling data processing was done using the Sensors and Software pulseEKKO (v. 4.22) and EKKO
Fig. 1. Plan view of the experiment area showing the location of the injection well and the GPR profile lines.
4
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
View Deluxe packages. The following procedures were applied to the profiling data in the sequence listed below; detailed information about these geophysical data processing techniques is found in Yilmaz (2001) and Sensors and Software (2003). A zero-time correction was applied to compensate for instrument drift. The low frequency ‘wow’ superimposed on the higher frequency reflections was removed by applying a signal saturation correction. Horizontal spatial (two trace average) and vertical time (two time sample average) filters were applied to enhance reflections and reduce random noise. A spherical and exponential compensation (SEC) gain was applied to compensate for amplitude changes due to geometrical spreading and ohmic dissipation while preserving the relative reflectivity of the subsurface boundaries. The SEC gain parameters were selected such that consistent amplitude information was obtained for all datasets. This objective was achieved by performing an amplitude balancing analysis on aquifer reflections at locations not impacted by the solvent release. The value of attenuation coefficient in the SEC gain was adjusted for each monitoring time such that temporal variability in the reflectivity of these events was minimized. A second set of GPR profile images was obtained by determining the instantaneous amplitude (i.e., envelope) of the individual GPR traces. The instantaneous amplitude is a complex trace attribute that removes the effects of wavelet polarity and phase, allowing a more accurate determination of reflection strength. A discussion of complex trace theory and its application to seismic reflection profiling is given by Sheriff and Geldart (1995). Previously, Orlando (2002) had successfully used complex trace attributes for the analysis of GPR profiling data from LNAPL impacted sites. 4. Temporal changes in GPR response The GPR profiles acquired along Line #11, which is near the middle of the grid and the injection point, are presented in Fig. 2 for the six survey times. The GPR profile before the injection (Fig. 2A) shows the aquifer stratigraphy at our site. The reflection from the underlying clay aquitard occurs at approximately 118 ns twoway traveltime. Although small-scale heterogeneities are evident in the pre-release profile, the stratigraphic boundary reflectivity within the aquifer is low.
5
The formation of the immiscible chlorinated solvent layers and their subsequent dissipation are clearly observed over time on the post-release profiles (Fig. 2B– F). The aquifer reflectivity increased during the first day as the solvent moved downward and formed layers (Fig. 2B and C). During this time, two closely spaced immiscible solvent pools were identified between 78 ns and 100 ns. The positions of these DNAPL layers coincide with subtle stratigraphic boundaries imaged on the pre-release profile (Fig. 2A), demonstrating the importance of stratigraphy and the associated permeability contrast in controlling the DNAPL movement. The solvent layer reflectivity is largely unchanged two weeks after the release (Fig. 2D). During the later stages of the GPR monitoring, the solvent layer reflectivities have significantly diminished after 29 months (Fig. 2E) and eventually reached levels that are comparable to that of the stratigraphic interfaces on the preinjection after 66 months (Fig. 2F). The complex trace instantaneous amplitude profiles along the Line #11 are presented in Fig. 3. The instantaneous amplitude changes are consistent with the reflectivity variations seen in Fig. 2. There is a low level of instantaneous amplitude in the pool locations before the injection (Fig. 3A). The instantaneous amplitude of the DNAPL layers increases as the solvent accumulates in them during the first day (Fig. 3B and C) and maintains its high level for the two weeks after the injection (Fig. 3D). The instantaneous amplitude of the solvent layers subsequently diminishes during the longer-term monitoring (Fig. 3E and F). The spatial distribution of the upper and lower DNAPL layers was inferred by comparing the background and post-release profiles over the entire GPR profiling grid. Both the maximum reflection amplitude and complex trace instantaneous amplitude for the reflection event along the stratigraphic horizon corresponding to each solvent layer were manually determined for each trace along the GPR profiles using the Sensors and Software Picker software. The reflectivity variation of the DNAPL layers during the experiment were quantified for each monitoring time by a simple differencing of the background and post-release maximum amplitude values at each trace position. Let us first consider the GPR imaged spatial changes for the upper DNAPL layer. Fig. 4 shows a plan view of the upper layer evolution using the reflection amplitude
Fig. 2. GPR profiling along the survey Line #11 for pre-release (a) and post-release (b–f) times. Position axis units are meters; time axis units are nanoseconds. Solid box indicates region of solvent layer development. The injection well is located adjacent to the profile at 2.12 m. Post-release times: (b) 1/2 day, (c) 1 day, (d) 2 weeks, (e) 29 months and (f) 66 months.
6
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Fig. 3. Complex trace instantaneous amplitude profiles of the GPR data along the survey Line #11 for pre-release (a) and post-release (b–f) times. Position axis units are meters; time axis units are nanoseconds. Solid box indicates region of solvent layer development. The injection well is located adjacent to the profile at 2.12 m. Post-release times: (b) 1/2 day, (c) 1 day, (d) 2 weeks, (e) 29 months and (f) 66 months.
differencing; the result of complex trace instantaneous amplitude differencing is shown in Fig. 5. The results obtained from these two reflectivity measures are
mutually consistent, displaying the same large-scale features and temporal changes. Both Figs. 4A–C and 5A–C show the development of a laterally extensive
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
7
Fig. 4. Plan view of the upper solvent layer based on reflection amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
upper layer during the first two weeks after the release. The apparent shift in the center of mass of this layer during the initial two week period is likely due to spatial differences in downward migration rates resulting from localized stratigraphic features or the presence of old root zones. The upper layer subsequently reduced in size and reflectivity after 29 months (Figs. 4D and 5D) and almost completely disappeared by the end of the GPR monitoring period at 66 months (Figs. 4E and 5E). The GPR imaged evolution of the lower DNAPL layer is illustrated in Figs. 6 and 7 which give plan views of the reflection amplitude and the complex trace instantaneous amplitude differencing results, respectively. The results for the lower layer are analogous to those obtained for the upper layer. During the first
two weeks, the development of a laterally extensive lower layer was observed (Figs. 6A–C and 7A–C). The lower layer showed significant reductions in extent and reflectivity after 29 months (Figs. 6D and 7D) and nearly vanished after 66 months (Figs. 6E and 7E). 5. GPR estimate of solvent pool mass In principal, the DNAPL layer reflection amplitude can be used to quantitatively estimate the local immiscible phase saturation and the total DNAPL mass. Brewster et al. (1995) developed a linear relationship between reflection amplitude and the local solvent saturation-layer thickness product for individual pools
8
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Fig. 5. Plan view of the upper solvent layer based on complex trace instantaneous amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
at CFB Borden and used it to estimate the total DNAPL mass from their GPR profiling data. A similar method was applied to our GPR profiling data to estimate the DNAPL mass changes over the 66 month monitoring period. Our study area was divided into a uniform rectangular grid with all elements having horizontal dimensions Δx and Δy. The mass of a DNAPL layer in the i, jth grid element (mij) is related to solvent density (ρDNAPL), aquifer porosity (ϕ), local DNAPL saturation (Sij) and local layer thickness (Tij) by the following expression:
On the basis of a GPR modeling study, Brewster et al. (1995) proposed that the local amplitude of the GPR reflection from a solvent layer (Aij) was related to elemental saturation-layer thickness product (SijTij) by
mij ¼ qDNAPL /Sij Tij DxDy:
Sij Tij ¼ bAij
ð1Þ
The total DNAPL mass (MDNAPL) within our study area can be determined by summing the elemental DNAPL mass over the area of all solvent layers: MDNAPL ¼ qDNAPL /DxDy
X XX layers
i
Sij Tij :
ð2Þ
j
ð3Þ
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
9
Fig. 6. Plan view of the lower solvent layer based on reflection amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
where b is a coefficient determined from a comparison between DNAPL saturation profiles obtained from core samples and coincident GPR profiling data. Hence, the total DNAPL mass can be obtained from the GPR reflection amplitude data using X XX MDNAPL ¼ qDNAPL /bDxDy Aij : ð4Þ layers
i
total DNAPL mass MMax, we obtain the following expression for the relative total DNAPL mass (MRel): PP P i j Aij M layers : MRel ¼ ¼ ð5Þ PP P MMax A ij i j layers
j
The While no calibration cores were obtained during our experiment that coincided with the GPR profiling times, our GPR data can be used to determine changes in relative DNAPL mass in the survey area. It can be seen from Eq. (4) MDNAPL is directly proportional to the P that P P quantity i i Aij. If our total DNAPL mass layers
computation is normalized by the maximum observed
PP P layers
i
i
Max
Aij term was evaluated for all five
surveys performed after the injection using both the reflection amplitude and complex trace instantaneous amplitude data. To minimize the potential contribution of stratigraphic contrasts to the computations, differencing between the background and post-release profiles was used in this analysis. We found that MMax occurred one day after injection for both measures of reflectivity
10
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
Fig. 7. Plan view of the lower solvent layer based on complex trace instantaneous amplitude differencing. The scale indicates the value of the amplitude difference. The injection well is located at (2.12 m SE, 2.58 m NE). Post-release times: (a) 1/2 day, (b) 1 day, (c) 2 weeks, (d) 29 months and (e) 66 months.
change; hence, these values were used to normalize the relative mass estimates. The relative total mass of the chlorinated solvent layers remaining in the experiment site estimated from both the reflection amplitude and the complex trace instantaneous amplitude data are given in Table 1. The GPR-estimated relative total DNAPL mass increased rapidly during the first day and only decreased slightly after two weeks. The longer-term GPR monitoring data indicated significant reductions in the total DNAPL mass. The remaining total solvent mass at 29 months post-release was estimated to be 34%–36% of its maximum amount. After 66 months, the total DNAPL mass remaining in the layers was further diminished to 4%–9% of its maximum value. We compared our results with an independent estimate of total DNAPL mass determined from an analysis of groundwater samples. This groundwater sampling was
performed along a cross-sectional, multi-level piezometer network located immediately downgradient of our DNAPL release. The water sampling started one month after the injection and continued for 18 sampling events over the subsequent five year period. The analysis of the groundwater samples estimated that the DNAPL mass remaining in the experiment area was approximately 40% of the initial DNAPL mass two years after the release and 10% of the initial DNAPL mass five years post-release. The similarity of these two independent measures of DNAPL mass strongly implies that the observed GPR reflectivity changes of the DNAPL layers were primarily due to dissolution occurring in the groundwater flow field. It should be noted that while both the GPR profiling and groundwater sampling gave similar results, the GPR profiling required a much smaller amount of effort and expense than the groundwater sampling program which
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12 Table 1 Relative total DNAPL mass remaining in the experiment site inferred from the GPR profiling data X XX Post-release time Aij Relative total DNAPL j layers i mass (MRel) a.) Reflection amplitude 1/2 Day 643,188 1 Day 1,790,680 2 Weeks 1,716,053 29 Months 638,908 66 Months 79,524
35.9% 100.0% 95.8% 35.7% 4.4%
b.) Instantaneous amplitude 1/2 Day 290,603 1 Day 853,513 2 Weeks 797,043 29 Months 286,360 66 Months 76,451
34.0% 100.0% 93.4% 33.6% 9.0%
Reflectivity measure used: (a) reflection amplitude and (b) complex trace instantaneous amplitude (i.e., envelope).
involved collection and analysis of thousands of individual water samples. Furthermore, GPR profiling is a non-invasive method whereas groundwater sampling requires the installation of monitoring wells; hence, GPR profiling has the capacity to quantitatively characterize DNAPL contamination at sandy aquifer sites where the emplacement of monitoring wells is restricted or prohibited. Because of the inherent limitations of traditional hydrogeological sampling techniques, DNAPL zone depletion due to groundwater dissolution under natural flow conditions has been determined at only a few of the many thousands of actual DNAPL sites existing in sandy aquifers around the world; GPR profiling offers a practical and cost-effective means of closing this large information gap. 6. Discussion and conclusions Our experimental results clearly demonstrate that GPR profiling has a greater potential for characterizing DNAPL impacted sites than previously indicated in the landmark papers by Brewster and Annan (1994) and Brewster et al. (1995). We have shown that GPR profiling has the ability to detect DNAPL releases having lesser volumes that are more representative of small-scale industrial spills. Even though our injected DNAPL volume was only 50 L, GPR profiling detected significant reflectivity increases due to the formation of laterally extensive solvent layers along stratigraphic boundaries during the two week period after the release. Since the DNAPL layers were subjected to the natural groundwater flow regime, we were able to assess
11
the sensitivity of GPR profiling to DNAPL mass reductions resulting from dissolution. The long-term GPR monitoring (29 and 66 months after the injection) observed the significant reduction and almost complete dissipation in the solvent layer reflectivity. The spatial extent of the DNAPL layers determined from GPR profiling significantly reduced in size after 29 months and nearly vanished by the end of 66 months. Using a method similar to that proposed by Brewster et al. (1995), reflection amplitude and the complex trace attribute instantaneous amplitude (i.e., envelope) were used to estimate total DNAPL mass variations from the DNAPL layer reflectivity observed on the GPR profiles; both of these approaches gave mutually consistent results. Due to the lack of appropriate calibration information, relative total DNAPL mass changes were determined for the survey area. The GPR-estimated total DNAPL mass reached its maximum value during the first day and remained high over the first two weeks following the release. Longer-term GPR monitoring shows a significant reduction in total DNAPL mass. It was estimated that only 34%–36% of the total DNAPL mass remained in the survey area after 29 months. After 66 months, the total DNAPL mass further diminished to 4%–9% of its maximum amount. The GPR estimates of total DNAPL mass are consistent with an independent hydrogeological estimate of remaining solvent mass based on the monitoring of the dissolved phase downgradient of the release. Hence, we conclude that the long-term changes in GPR reflectivity of the DNAPL layers most probably result from the dissolution of those chlorinated solvents due to groundwater flow. Furthermore, the corresponding dissipation of the DNAPL layer reflectivity indicates that the GPR signature of older DNAPL impacted zones may not differ greatly from the uncontaminated background if significant mass reduction has occurred. One question that has been raised recently in the geophysical monitoring of immiscible contaminant layers is the potential for dramatic geophysical changes due to biodegradation. Bermejo et al. (1997) and Sauck et al. (1998) found zones of GPR signal attenuation (shadow zones) associated with aged light non-aqueous phase liquid (LNAPL) pools. These shadow zones result from enhanced electrical conductivity coincident with hydrocarbon degradation through microbial reactions (Atekwana et al., 2000, 2004). We did not observe any “shadow zone” of the type reported by these studies. In addition, groundwater monitoring of dissolved volatile organic compounds (VOC) constituents emanating from the source over time does not show degradation products, supporting the conclusion that significant biodegradation activity did not
12
Y.K. Hwang et al. / Journal of Contaminant Hydrology 97 (2008) 1–12
occur in the source zone over the five year time period of our experiment. In conclusion, our results clearly demonstrate the ability of GPR profiling to detect and monitor a small volume DNAPL release into a sandy aquifer and to quantify longterm depletion of the DNAPL mass due to dissolution. At many actual DNAPL sites, there is a need to determine the DNAPL mass depletion rate due to dissolution in natural groundwater flow or enhanced mass removal through remediation; this information is important for assessment of long-term impacts and remediation potential. GPR profiling has the potential to obtain this goal in a non-invasive and more efficient manner in comparison to traditional hydrogeological sampling techniques. Acknowledgments The GPR data acquisition and analysis was supported by a Collaborative Research and Development Grant from the Natural Sciences and Engineering Research Council of Canada and a grant from the University Consortium Solvents-in-Groundwater Research Program awarded to Dr. Parker. Mr. Hwang's studies were partially supported by an Individual Research Grant to Dr. Endres from the Natural Sciences and Engineering Research Council of Canada. The authors wish to thank the Editor and two anonymous reviewers for comments that have significantly improved the quality of this paper. References Atekwana, E.A., Sauck, W.A., Werkema, D.D., 2000. Investigations of geoelectrical signatures at a hydrocarbon contaminated site. Journal of Applied Geophysics 44, 167–180. Atekwana, E.A., Werkema, D.D., Duris, J.W., Rossbach, S., Atekwana, E.A., Sauck, W.A., Cassidy, D.P., Means, J., Legall, F.D., 2004. Insitu apparent conductivity measurements and microbial population distribution at a hydrocarbon-contaminated site. Geophysics 69, 56–63. Bermejo, J.L., Sauck, W.A., Atekwana, E.A., 1997. Geophysical discovery of a new LNAPL plume at the former Wurtsmith AFB, Oscoda, Michigan. Ground Water Monitoring & Remediation 17, 131–137. Brewster, M.L., Annan, A.P., 1994. Ground-penetrating radar monitoring of a controlled DNAPL release: 200 MHz radar. Geophysics 59, 1211–1221. Brewster, M.L., Annan, A.P., Greenhouse, J.P., Kueper, B.H., Olhoeft, G.R., Redman, J.D., Sander, K.A., 1995. Observed migration of a controlled DNAPL release by geophysical methods. Ground Water 33, 977–987. Domenico, P.A., Schwartz, F.W., 1998. Physical and Chemical Hydrogeology. John Wiley & Sons, New York. Feenstra, S., Cherry, J.A., 1988. Subsurface contamination by dense nonaqueous phase liquid (DNAPL) chemicals. International Groundwater Symposium. International Association of Hydrogeologists 61–69.
Feenstra, S., Cherry, J.A., Parker, B.L., 1996. Conceptual models for the behavior of DNAPLs in the subsurface. In: Pankow, J.F., Cherry, J.A. (Eds.), In Dense Chlorinated Solvents and Other DNAPLs in Groundwater: History, Behavior and Remediation. Waterloo Educational Services, Rockwood, Ontario, pp. 53–88. Jordan, T.E., Baker, G.S., Henn, K., Messier, J.P., 2004. Using amplitude variation with offset and normalized residual polarization analysis of ground penetrating radar data to differentiate an NAPL release from stratigraphic changes. Journal of Applied Geophysics 56, 41–58. Kim, C., Daniels, J.J., Guy, E.D., Radzevicius, S.J., Holt, J., 2000. Residual hydrocarbons in a water saturated medium: a detection strategy using ground penetrating radar. Environmental Geosciences 7, 169–176. Kueper, B.H., Frind, E.O., 1991. Two-phase flow in heterogeneous porous media, 2. Model application. Water Resources Research 27, 1059–1070. Kueper, B.H., Redman, J.D., Starr, R.C., Reitsma, S., Mah, M., 1993. A field experiment to study the behavior of tetrachloroethylene below the water table: Spatial distribution of residual and pool DNAPL. Ground Water 31, 756–766. Laukonen, K.A., Parker, B.L., Cherry, J.A., 2000. Internal characteristics of a bromide tracer zone during natural flow in the Borden aquifer, Ontario, Canada. Tracers and Modelling in Hydrogeology. Proceedings of TraM'2000 Conference. IAHS Publication, Liege, Belgium, pp. 227–233. No. 262. Lopes de Castro, D., Castelo Branco, R.M.G., 2003. 4-D ground penetrating radar monitoring of a hydrocarbon leakage site in Fortaleza (Brazil) during its remediation process: a case history. Journal of Applied Geophysics 54, 127–144. Lucius, J.E., Olhoeft, G.R., Hill, P.L., Duke, S.K., 1992. Properties and Hazards of 108 Selected Substances. U. S. Geological Survey Open File 92-0527. Macfarlane, D.S., Cheery, J.A., Gillham, R.W., Sudicky, E.A., 1983. Migration of contaminants in groundwater at a landfill: a case study. Journal of Hydrology 63, 1–29. Mackay, D.M., Cherry, J.A., 1989. Groundwater contamination: limits of pump and treat remediation. Environmental Science & Technology 23, 630–636. Mackay, D.M., Freyberg, D.L., Roberts, P.V., Cherry, J.A., 1986. A natural gradient experiment on solute transport in a sand aquifer 1. Approach and overview of plume movement. Water Resources Research 22, 2017–2029. Orlando, L., 2002. Detection and analysis of LNAPL using the instantaneous amplitude and frequency of ground-penetrating radar data. Geophysical Prospecting 50, 27–41. Sauck, W.A., Atekwana, E.A., Nash, M.S., 1998. High conductivities associated with LNAPL plume imaged by integrated geophysical techniques. Journal of Environmental and Engineering Geophysics 2, 203–212. Sheriff, R.E., Geldart, L.P., 1995. Exploration Seismology. Cambridge University Press, Cambridge. Sensors and Software, Inc., 2003. EKKO View Enhanced & EKKO View Deluxe Users Guide. Sudicky, E.A., 1986. A natural gradient experiment on solute transport in a sand aquifer: spatial variability of hydraulic conductivity and its role in the dispersion process. Water Resources Research 22, 2069–2082. U. S. Environmental Protection Agency, 2006. Consumer Factsheet on: Trichloroethylene. http://www.epa.gov/safewater/dwh/c-voc/ trichlor.html, Accessed on November 29, 2006. Yilmaz, Ö., 2001. Seismic Data Analysis: Processing, Inversion and Interpretation of Seismic Data, Volume 1. Society of Exploration Geophysicists, Tulsa.