The Measurement of DNAPL in Low-Permeability Lenses within Alluvial Aquifers by Partitioning Tracers R. E. JACKSON INTERA Inc., P.O. Box 818, Niwot, CO 80544
M. JIN INTERA Inc., 9111A Research Blvd., Austin, TX 78758
Key Terms: DNAPL, Alluvial Aquifers, Partitioning Tracers
ABSTRACT The partitioning interwell tracer test (PITT) was introduced into the practice of contaminant hydrogeology in 1994, since which time about 50 PITTs have been conducted in the field. Confirmation of results of vadose-zone and ground-water zone PITTs by subsequent field testing of aquifers in Utah and New Mexico have demonstrated the reliability of the technique for operational purposes, i.e., the measurement of average inter-well NAPL (non-aqueous phase liquid) saturations that provide meaningful values upon which to base remedial decisions. However, the use of the PITT has become increasingly infrequent, partly because it has attracted criticism concerning its reliability to quantify DNAPL (dense nonaqueous-phase liquid) either in low-permeability units surrounded by high-permeability materials or present in depressions at the base of aquifers. In the vadose zone, where gaseous partitioning tracers are used, the criticism maintains that the partitioning tracers will necessarily follow cleaned pathways created during soil vapor extraction and will thus bypass DNAPL zones. In the ground-water zone, the criticism is that it fails to detect DNAPL in lowpermeability lenses either in the middle or at the base of aquifers, again because of bypassing. This paper presents numerical simulations of two laboratory experiments that demonstrate that the criticisms are incorrect and PITT will accurately detect residual NAPL volume when proper design and field implementation are undertaken. INTRODUCTION In contaminant hydrogeology, the measure of nonaqueous-phase liquid or NAPL concentration in aquifer materials is given by the NAPL saturation, i.e., the volume
of NAPL per unit volume of pore space, which is reported as a fraction of the porosity (Bear, 1972). Any measurement of NAPL saturation assumes that the technique used quantifies the NAPL concentration in a representative elementary volume (REV). In this article, we shall confine our discussion to denser-than-water NAPLs (DNAPLs), such as chlorinated degreasing (e.g., trichloroethene, or TCE, and 1,1,1-trichloroethane, or TCA) and dry-cleaning solvents (e.g., tetrachloroethene, or PCE), and to the use of the partitioning interwell tracer test (PITT) for measuring DNAPL saturations in heterogeneous alluvium. Although both Annable and others (1998) and Meinardus and others (2002) have presented compelling evidence of its reliability in the field measurement of NAPL concentrations in alluvium at two sites at Hill Air Force Base in Utah, the PITT has become discredited as a practical tool for measuring NAPL volumes. This dismissal of the accuracy of the PITT is clearly demonstrated in recent review documents by expert panels (Everett et al., 2000; Kavanaugh and Rao, 2003; and ITRC, 2004). The quantitative measurement of DNAPL saturations has typically been achieved by soil coring. Mayer and Miller (1992) concluded that the non-uniformity of typical porous media means that the REV for alluvial materials is probably larger (10–104 cm3) than most soil samples. Mariner and others (1997) described an algorithm for computing the NAPL saturation for chemicals in a typical soil sample analysis and have provided a free software program to undertake the analysis (see review in McCray and Cohen, 2003). A group of EPA and Rice University researchers concluded that soil-core-derived estimates of NAPL saturation are unreliable (Bedient et al., 1999). Likewise, Meinardus and others (2002) pointed out that it is not possible using soil cores to economically estimate reliable average interwell DNAPL saturations—i.e., large-scale values for aquifer volumes of 100 m3, such as pumping tests yield of interwell hydraulic conductivity—because of the expense of conducting the numerous preservation, separation, chemical analysis, and data interpretation steps. It was concerns like these that led to the introduction of the PITT (Jin et al., 1994, 1995, 1997). The first PITT was
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
405
Jackson and Jin
conducted at Hill Air Force Base (AFB) in Utah in 1994 (Annable et al., 1998). Their results show good comparison between soil-core-derived and PITT estimates of NAPL saturation and volume. Later, Meinardus and others (2002) demonstrated in greater detail how properly designed and executed PITTs would measure DNAPL saturations in heterogeneous alluvium as accurately as the soil-corederived values. The twin tracer peaks in the breakthrough curve (Meinardus et al., 2002, Figure 9, p. 188) clearly indicate that the PITT swept the whole pore volume within the DNAPL source zone, including the relatively lowpermeability unit at the base of the DNAPL zone. A second study that also affirmed the reliability of PITTs to provide information necessary for remedial decision making was the first vadose-zone PITT described by Mariner and others (1999) at Sandia National Laboratories in New Mexico. This PITT indicated that TCE disposed at the Chemical Waste Landfill at Sandia did not migrate any deeper than 10 m below ground surface. This result was subsequently confirmed during the excavation of the landfill and reported by Sandia National Laboratories (2003). Thus, the reliability of PITTs to map the spatial distribution and volume of DNAPL has been tested and confirmed in the field. In light of these findings, it is therefore noteworthy that both federal and state expert panels have identified the relative accuracy of PITTs as (ITRC, 2004) ‘‘low to intermediate, potential limitations based upon DNAPL architecture.’’ These findings of fault with the PITT by ITRC (2004) and the EPA Expert Panel on DNAPL Remediation (Kavanaugh and Rao, 2003) are best addressed by numerical simulation of laboratory experiments in which the location of the DNAPL zone is clearly identified and the properties of the surrounding soils are well documented. Using the UTCHEM simulator (Delshad et al., 1996; Pope et al., 1999), it is possible to reproduce the flow field in the sand-tank experiments and then show whether different designs of PITTs can overcome the permeability contrasts and yield the true estimates of DNAPL contained in the sand tanks, which of course are known. This process of analysis is one of ‘‘falsification’’ in the sense of the term introduced by the philosopher of science Karl Popper, who believed that scientific theories could not themselves be verified but only falsified by conducting experiments that might cause the rejection of the theory (see Kuhn, 1996). The theory in question to be tested is the opinion held by these expert panels that the PITT is incapable of accurately measuring known volumes of DNAPL in low permeability lenses within a higher permeability matrix of alluvium. That is, if the numerical experiments using UTCHEM reproduce the known volume of DNAPL introduced into the two sand tanks under the known experimental conditions, then it follows that the expert panels were incorrect in their statements
406
about the PITT’s accuracy. Furthermore, if the criticisms of the expert panels are falsified by these numerical experiments, then the field evidence cited above from Hill AFB and Sandia National Laboratories supports the contention that the PITT is a viable technique for measuring DNAPL saturation in a heterogeneous alluvial geosystem with an accuracy sufficient to make informed remedial decisions at DNAPL sites. For this purpose, we consider two sand-tank experiments. The first sand-tank experiment, that of Nelson and others (1999), is used to examine the bypassing problem in the ground-water zone and is referred to as the ‘‘Nelson sand-tank experiment.’’ The second experiment, that of Yonge and others (1996), tested the penetration of a tracer into a low-permeability unit containing DNAPL that was part of the vadose zone. This experiment is referred to as the ‘‘Yonge sand tank’’ and was conducted with the fate of carbon tetrachloride beneath the U.S. Department of Energy site at Hanford, Washington, in mind. The discussion of the assessment of the accuracy and capability of the PITT in this article is restricted to the issue of bypassing that has been raised by various expert panels (Everett et al., 2000; Kavanaugh and Rao, 2003; and ITRC, 2004) and others (e.g., Parker et al., 2003; Holbert et al., 2004; and Moreno-Barbero, 2004). The issues related to the mass-transfer limitations (Piepenbrink et al., 2002; Bohy et al., 2004) and compositional changes during remediation (Lee et al., 1998) are not discussed. These issues are unlikely to affect remedial decision making in the field. ANALYSIS OF THE NELSON SAND-TANK EXPERIMENT The ITRC attributes its conclusion of poor accuracy to the results of a laboratory experiment conducted by Nelson and others (1999) in which the PITT failed to detect a zone of emplaced TCE DNAPL in a lowpermeability sand lens (70 mesh) contained within a matrix of coarse sand (20/30 mesh). It is important in the context of the present analysis to understand that DNAPL would not normally penetrate the low-permeability, watersaturated sand shown in Figure 1 by natural DNAPL migration from above. We have simulated with UTCHEM the case in which the geosystem presented by the Nelson sand tank might occur due to a large-scale TCE spill and mobilization and observed that the DNAPL will not penetrate the lowpermeability lens. The results are shown in Figure 1 and clearly demonstrate that no TCE enters the lowpermeability sand lens. Increasing the head of TCE on the 70 mesh sand (h . 40 cm) will not cause DNAPL entry of the 70 mesh sand but rather further lateral bypassing. However, the TCE was not spilled in Nelson’s experiment as in the manner shown in Figure 1, but rather it was
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
DNAPL Measurement in Alluvial Aquifers
Figure 1. UTCHEM simulated TCE spill.
introduced in a completely artificial way by adding it with the fine sand as part of the packing of the cell. The behavior of the TCE in Figure 1 is similar to that demonstrated by Kueper and others (1989), who showed that a sand lens (#50 silica; k ¼ 5.26 3 1011 m2, entry pressure ¼ 13.5 cm water) was not penetrated by tetrachloroethene (PCE) DNAPL that was migrating in a coarser sand (#16 silica, k ¼ 5.04 3 1010 m2) above the #50 sand. In a geosystem of such high permeability contrasts, the DNAPL spreads laterally before the entry pressure of the lower permeability unit is exceeded. The Kueper sandbox has a lower permeability contrast (i.e., 10:1) than that of Nelson and others (53:1), therefore, the existence of such a DNAPL-contaminated lens represented by Zone 1 is most improbable. As Kueper and others commented: ‘‘. . . significant lateral flow of a nonwetting fluid can occur above a finer-grained lens even once that lens has been penetrated.’’ ‘‘The effect is due in part to the fact that the finergrained lens has a lower overall permeability, but also to the fact that the higher saturation of the nonwetting fluid above the lens gives rise to a higher relative permeability to the nonwetting fluid above the lens as compared with that within’’ (Kueper et al., 1989, p. 93). Therefore, the presence of the DNAPL in such a low-permeability lens in water-saturated porous media defies multiphase flow principles and indicates that the Nelson sand-tank experiment was conducted under unrealistic conditions that will not be encountered in the field. The only way that DNAPL might penetrate such a low-permeability lens in water-saturated porous media would be when the lens is at the base of the aquifer in a depression formed into a capillary barrier, such as a clay aquiclude, not in the center of the center of the aquifer as in the Nelson sand tank. However, it was precisely this case of a lowpermeability lens in a depression that Meinardus and others (2002) demonstrated was quantitatively estimated by the PITT, although the permeability contrast in the Hill AFB aquifer was not as great as the 53:1 used in the Nelson sand-tank experiment.
Figure 2. Comparison of the simulated tracer response and the actual tracer test results of the Nelson experiment.
With this condition in mind, we have used UTCHEM to simulate the Nelson sand-tank experiment and the results are shown in Figure 2. The swept pore volume computed from the simulated bromide (i.e., conservative tracer) concentrations is 27,500 ml, which is in excellent agreement with the value computed from the dimensions of the cell (pore volume ¼ length 3 height 3 width 3 porosity ¼ 220 cm 3 71 cm 3 5.3 cm 3 0.338 ¼ 28,000 ml). Using the travel time for bromide of 14.8 hours (888 minutes) as given in Table 3 of Nelson and others (1999), the swept pore volume is simply the flow rate times the travel time, which gives a swept pore volume of (50 ml/ min)(888 minutes)¼44,400 ml. This discrepancy between the values of Nelson and others and the one estimated from the dimension of the sand pack indicate that there are major problems with the data quality in this experiment. The semi-log plot shown in Figure 2 indicates that the measured tracer concentrations are higher than the simulated values and are very noisy. Accurate measurements of the tracer concentrations in the tracer tail are required for valid PITT analysis. Very often, the tracer concentration data are biased on the high side in such experiments unless very careful calibration of the data is made below 10 mg/L. Accurate measurements below 10 mg/L are indeed feasible and best made with in-line gaschromatography sampling of the effluent stream (Silva et al., 2003). It has also been demonstrated for various laboratory and field PITTs (Jin et al., 1995, 2000; Deeds et al., 1999; Young et al., 1999; Meinardus et al., 2002; and Silva et al., 2003) that the tracer tail can be extrapolated to yield even better NAPL estimation results when the data are sufficiently accurate and complete to give a straight line trend on the semi-log plot. Figure 3 presents simulations for tracers with TCE partition coefficients K ¼ 10 (not used by Nelson et al., 1999) and K ¼ 46.5 (DMP or dimethyl pentanol, as used by Nelson et al.). The simulated curves are shown as
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
407
Jackson and Jin
Figure 3. Simulated tracer response curves of the Nelson experiment for the case with additional tracer.
semi-logarithmic plots, which is the preferred form to identify small saturations of NAPL (Jin et al., 1995, 1997). Because the DMP tracer had a reported detection limit of 1.4 mg/L, it would have been undetectable at the extraction well after 2 days of the 6-day PITT. As Figure 4 shows, this would result in a PITT estimate of approximately 50 percent of the true TCE saturation (0.01415). Nelson and others show a NAPL detection of 30 percent. This design problem, which resulted in underestimation of TCE volume in the Nelson sand-tank experiment, could have been corrected by increasing the concentration of the partitioning tracer DMP and by adding more partitioning tracers. For the variety of reasons presented in Jin and others (1995), it is typical to use at least three partitioning tracers. The incremental cost of the additional tracers is very small compared with the total cost of the PITT. The above simulation results show that PITT could have measured the actual TCE saturation in the low permeability unit if the experiments had been conducted more carefully. In summary, the particular distribution of permeability and DNAPL that was created in the Nelson sand-tank experiment does not appear realistic. Nevertheless, by careful design and implementation, it would have been possible for Nelson and others to have measured the actual TCE saturation had the PITT been implemented as shown. Therefore, the citation of Nelson and others (1999) by the ITRC expert panel (ITRC, 2004) to conclude that the relative accuracy of PITTs is ‘‘low to intermediate’’ is not justifiable.
ANALYSIS OF THE YONGE SAND-TANK EXPERIMENT In the vadose-zone case, in which a three-phase system (air, water, DNAPL) exists, the DNAPL will penetrate a low-permeability lens, such as shown in Figure 1,
408
Figure 4. Estimated average TCE saturation as a function of time based on the numerical simulation presented in Figure 2. (The average TCE saturation and swept volume increase with time as low-velocity stream tubes from the 70-mesh sand are captured.)
because DNAPL will preferentially wet the lens and displace air from it. Therefore, in this case, it is entirely reasonable for the DNAPL to be predominantly trapped in a low-permeability lens, which is quite the contrary to the water-saturated Nelson sand-tank experiment. The conceptualization of air flow in the vadose zone reflects what is believed to occur during soil-vapor extraction (SVE). The streamlines of air flow take the path of least resistance and will bypass zones of low relative permeability caused by the presence of DNAPL. In low-permeability regions, the air flow will tend to pass through a zone created by fluid instabilities that cause viscous fingering of the air because the air is much less viscous than either soil water or DNAPL. Therefore, very little of the low-permeability region will be subjected to vapor extraction because there will be paths of lower resistance available to the air flow than through the low-permeability region. Consequently, much of the DNAPL in the low-permeability region will be bypassed by SVE streamlines. Unfortunately, this behavior of air flow in the vadose zone has also been attributed to PITTs and led to the conclusion that partitioning tracer flow would necessarily behave similarly (Everett et al., 2000). However, this will not be the case when tracer is injected simultaneously with clean air to force the tracer in a particular direction. By using ‘‘pneumatic control wells’’ that inject tracer-free air and a tracer injection well, the tracer pulse is constrained in its flow pattern. Pneumatic control wells behind the tracer injection well act as pneumatic mirrors to force the tracer forward and prevent it from dispersing radially backward. Other pneumatic control wells may be placed above and below the tracer-injection and -extraction interval (i.e., the swept pore volume) and are used to inject tracer-free air to cause uniform flow in the tracer streamlines.
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
DNAPL Measurement in Alluvial Aquifers
Figure 5. Schematic of tracer test configuration of Yonge experiment.
Yonge and others (1996) undertook a sand-tank experiment of the Hanford and Cold Creek units and demonstrated that their tracer did not bypass the Hanford silt layer. Several sand-tank experiments were set up to examine SVE effectiveness of the sand-over-silt layering seen at Hanford, including a careful tracer study using SF6 that had excellent tracer sweep of both the overlying sand and lower silt layers. The test was set up with injection and extraction only in the overlying sand unit, but predicted the pore volume to within 97 percent and porosity nearly exactly. The authors noted that hydraulic residence time was 145 minutes, which corresponded to an actual pore volume of 31.6 L and a porosity of 38.8 percent, and indicated that the 4-ft. flow-through cell configuration exhibited minimal short circuiting or dead volume. Therefore, the concern that a PITT would necessarily bypass the DNAPL zone because it would follow the same path as soil vapor during SVE was not substantiated in this test and, in fact, appears wrong. The reason for the success of the Yonge sand-tank experiment is that the tracer was injected in such a manner that its flow path was controlled by the flow field established by both injection and extraction wells, which differs substantially from SVE where only vapor extraction occurs. The great advantage of a tracer test is that tracer and clean air can be applied to a target zone so that the tracer is forced through DNAPL zones that would otherwise be bypassed without pneumatic control of the tracer pulse. Figure 5 shows the sand tank of Yonge and others that confirms this argument.
Figure 6. Simulated tracer response curve of the Yonge experiment.
Figure 7. Schematic of modified tracer test configuration of Yonge experiment.
Figure 6 illustrates the UTCHEM-simulated tracer response curves for the case in which the tracer test was conducted with injection and extraction only in the overlying sand unit. NAPL was only trapped in the silt unit with 2 percent in saturation. The results from tracer data analysis using the method of first moment indicate only approximately 10 percent of the NAPL trapped in the silt unit was detected by the tracers. To increase the tracer detection accuracy, a second simulation was conducted using a different tracer injection and extraction configuration. In this case, clean air was injected in the overlying sand unit and tracer slug was injected in the silt unit by using a separate screen interval (see Figure 7). Similarly, at the extraction end, air was extracted separately from the overlying sand unit and the underlying silt unit. Figures 8 and 9 show the tracer response of the sand unit and the silt unit, respectively. Tracer data analysis for this case indicates 100 percent of detection of NAPL trapped in the silt unit with good tracer separation. This comparative simulation study clearly indicates that tracer can be forced into a particular low-permeability zone or flow direction by using pneumatic control wells and will sweep completely different zones from that of soil vapor during SVE.
Figure 8. Tracer response curve of the extraction well in the overlying sand unit.
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
409
Jackson and Jin
Figure 9. Tracer response curve of the extraction well in the underlying silt unit.
DISCUSSION: OPERATIONAL APPLICATIONS AND REMEDIAL DECISION MAKING In this article, we have presented a defense of the PITT on the basis of numerical simulations of two sand-tank experiments. These demonstrate that for the PITT to be operationally effective in the field, i.e., to allow informed remedial decision making to occur, it must be carefully designed along the lines that were originally presented in Jin and others (1995). Sufficient numbers of tracers must be used, with differing partition coefficients and adequately low analytical detection limits, so that accurate estimates of DNAPL saturation and its spatial distribution can be obtained through the placement of multilevel samplers (e.g., Meinardus et al., 2002). We have also shown that the bypassing problem can be minimized by using injection wells that introduce tracer-free air or water so that the tracers are controlled and directed in their flow path. These are not standard techniques in contaminant hydrogeology but are borrowed from petroleum engineering and should be adopted for the characterization of hazardous waste sites. Nevertheless, the bypassing issue is a persistent one in discussions of the PITT (e.g., Parker et al., 2003; Holbert et al., 2004; and Moreno-Barbero et al., 2004), as are those of cost and complexity. The recent EPA Expert Panel report (Kavanaugh and Rao, 2003) makes the same point of the potential inaccuracies inherent in using PITTs as does the ITRC (2004) report. Kavanaugh and Rao (2003, p. 22) argue that ‘‘for heterogeneous DNAPL distribution (especially pools), [the PITT] underestimates DNAPL volume.’’ The issue of pooled DNAPL was addressed by Jin and others (1997), who showed that the errors in estimating the volume of pooled DNAPL in the extreme case of a sumplike depression at the base of an aquifer were of the order of 50–60 percent. It is unlikely, however, that errors of this magnitude in estimating DNAPL saturation and DNAPL volume, even in this extreme case, will cause
410
faulty remedial decisions to be reached. The only advantage of using the PITT that the Kavanaugh and Rao report identified was that it could estimate DNAPL saturations; none of the other 10 ‘‘tools’’ identified in the Kavanaugh and Rao report allow the measurement of this parameter. Consequently, the Kavanaugh and Rao report implicitly indicates that the PITT is the only tool available that will provide a quantitative estimation of the average DNAPL concentration in soil. The other 10 ‘‘tools’’ cited are DNAPL detection methods, not DNAPL quantification methods like the PITT. Meinardus and others (2002) showed that PITTs conducted before and after an experimental surfactant-foam flood yielded very similar results to soil cores. The two estimates of DNAPL volumes computed by the PITTs were 79 6 27 L and 9.8 6 7.6 L. The equivalent volume estimates by soil coring, followed by sample preservation and analysis by NAPLANAL (Mariner et al., 1997) and, finally, geostatistical averaging were 74 6 48 L and 5.2 6 3.4 L, respectively. As Meinardus and others pointed out, this result was achieved only by collecting soil samples in a spatial pattern of boreholes each separated from the previous borehole by only 1.5 m distance, and with vertical sub-coring of the continuous soil cores conducted every 5–10 cm. This meant that approximately 0.1 percent of the soil mass in the DNAPL zone was sampled and analyzed. For operational purposes, such as remedial decision making, one would not arrive at a different decision based on the soil core results than one would based on the PITT results. If the residual DNAPL zone was approximately 100 m3 in volume, these results suggest that, to have confidence in soil coring results, one would have to sample and analyze 170 kg of soil—an impractical task. In fact, there is no reason to believe that the soil-core-derived results are more reliable than the PITT results because the tracers swept a much larger volume of the DNAPL zone. Meinardus and others (2002) showed that the PITTs not only gave similar volumes as intensive soil coring but also that the PITTs effectively sampled a low-permeability zone at the base of the alluvial aquifer overlain by higher permeability sands. Clearly, DNAPL was present at the base of the aquifer and was sampled by the PITT as the twin peaks in the tracer breakthrough curves illustrate. The low-permeability zone was ‘‘sampled’’ by the second slower peak, demonstrating that the PITT would measure DNAPL in such zones. Similarly, PITTs conducted in the vadose zone using gas tracers (e.g., Mariner et al., 1999) beneath the Chemical Waste Landfill at Sandia National Laboratories produced results that were later confirmed. The PITTs indicated that the TCE DNAPL had penetrated no deeper than 10 m beneath the landfill, which fact was later confirmed during the remedial excavation operations (Sandia National Laboratories, 2003).
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
DNAPL Measurement in Alluvial Aquifers
Another case of reliable quantification by PITTs was reported by Annable and others (1998). They showed reliable NAPL saturation measurement when the measurements were compared with soil core data in the first PITT ever conducted. Finally, Hatfield (1998) reported that the Universities of Florida and Waterloo conducted a test of the accuracy of the PITT at Canadian Forces Base Borden in which PCE was injected into ‘‘a steel isolation cylinder that was driven into the sandy aquifer.’’ Hatfield reported that ‘‘99% of the injected PCE mass was estimated through tracers.’’ Having argued for the accuracy of the PITT, it remains to present a summary of average interwell saturations from five chlorinated-solvent DNAPL sites in Table 1. These saturations measured by PITTs in DNAPL zones are typically much less than 10 percent. At Camp Lejeune, North Carolina, the average interwell saturation in the fine sands was 1.8 percent (Holzmer et al., 2005), while at Hill AFB, following free-phase DNAPL removal by pumping from the basal sands (Oolman et al., 1995), it was 2 percent, with a range of 4.7 to 0.4 percent (Meinardus et al., 2000). In the basal sand and gravel at Portsmouth, Ohio, it was only 0.3 percent (Young et al., 1999). Such interwell saturations are much lower than the experimental values listed by Mercer and Cohen (1990) for water-saturated sands, i.e., 15–50 percent. While we have measured higher point saturations in soil samples, including one of 77 percent at Hill AFB in contact with the clay aquiclude that prevented DNAPL drainage, we have typically found much smaller saturations even when the soil texture is fine-grained. In the sandy silts beneath a Gulf Coast chemical plant, only a few samples contained DNAPL at saturations approaching 20 percent (Mariner et al., 1997), while most were less than 5 percent. Similarly, at Camp Lejeune, the highest point saturation was 13.7 percent PCE (Holzmer et al., 2005).
Table 1. Geosystem conditions in alluvium at five chlorinated-solvent DNAPL sites in the USA where PITTs have been conducted to determine the average interwell DNAPL saturation. All values represent residual saturations. At Hill AFB, the average value of 2% represents the geosystem following free-phase removal of some 40,000 gallons of DNAPL.
Site and Location (Reference)
Depositional Environment
Average Hydraulic Conductivity (m/s) [k in darcies]
Average Interwell DNAPL Saturation (%)
Gulf Coast Chemical Plant, Lake Charles, LA (Mariner et al., 1997) Hill AFB, Ogden, UT (Meinardus et al., 2000)
Deltaic silty sands
2 3 108 [0.002]
1.4
Fluvial paleochannel
;1 3 104 [6–18]
Portsmouth Gaseous Diffusion Plant, Piketon, OH (Young et al., 1999) USMCB Camp Lejeune, Camp Lejeune, NC (Holzmer et al., 2005) Savage Well Superfund Site, Milford, NH (DE&S and UT-Austin, 1999)
Fluvial paleochannel
2.1 3 104 [21]
Panel Panel Panel Panel Panel 0.3
Near-shore marine or deltaic
5 3 106 [0.5]
1.8
Glacial outwash
Upper: 5 3 104 [54] Middle: 4 3 104 [35] Lower: 6 3 104 [61]
Upper: 0.56
1: 2: 3: 4: 5:
4.7 2.8 0.4 1.4 1.2
Middle: 0.36
Lower: 0.70
CONCLUSIONS REFERENCES Contrary to the opinions of a number of expert panels (Everett et al., 2000; Kavanaugh and Rao, 2003; and ITRC, 2004), the PITT—when properly designed and implemented with injection and extraction that are appropriately placed—will detect DNAPL in lowpermeability materials surrounded by higher permeability materials or in a depression at the base of an aquifer, irrespective of whether the PITT is conducted in the vadose or ground-water zones. Given the need for informed remedial decision making and the impracticability of using vast numbers of soil cores that have to be field preserved, chemically analyzed, and geostatistically interpreted to obtain the same result, the PITT remains a practical and accurate means of measuring average DNAPL concentrations over large volumes of the subsurface, although a costly and complex one.
ANNABLE, M. D.; RAO, P. S. C.; HATFIELD, K.; GRAHAM, W. D.; WOOD, A. L.; AND ENFIELD, C. G., 1998, Partitioning tracers for measuring residual NAPL: Field-scale test results: Journal Environmental Engineering, Vol. 124, pp. 498–503. BEAR, J., 1972, Dynamics of Fluids in Porous Media: Elsevier, New York, 764 p. BEDIENT, P. B.; HOLDER, A. W.; ENFIELD, C. G.; AND WOOD, A. L., 1999, Enhanced remediation demonstrations at Hill Air Force Base: Introduction. In Innovative Subsurface Remediation: Field Testing of Physical, Chemical and Characterization Technologies, ACS Symposium Series 725, American Chemical Society, Washington DC, pp. 36–58. BOHY, M.; SCHAFER, G.; AND RAZAKARISOA, O., 2004, Caracte´risation de zones sources de DNAPL a` l’aide de traceurs bisolubles: mise en evidence d’une cine´tique de partage: Comptes Rendus Geoscience, Vol. 336, pp. 799–806. DEEDS, N.; POPE, G. A.; AND MCKINNEY, D. C., 1999, Vadose zone characterization at a contaminated site using partitioning interwell
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
411
Jackson and Jin tracer technology: Environmental Science Technology, Vol. 33, pp. 2745–2751. DELSHAD, M.; POPE, G. A.; AND SEPEHRNOORI, K., 1996, A compositional simulator for modeling surfactant enhanced aquifer remediation: Journal Contaminant Hydrology, Vol. 23, pp. 303– 327. DE&S AND UT-AUSTIN, 1999, Determination of Spatial Distribution and Volume of Chlorinated Solvent Contamination: Report to U.S. Environmental Protection Agency, Region I, Boston, MA, and Idaho National Engineering & Environmental Laboratory, Idaho Falls, ID: Duke Engineering & Services and The University of Texas at Austin, Austin, TX. EVERETT, L.; KAVANAUGH, M.; AND MCWHORTER, D., 2000, Technical Review of Partitioning Interwell Tracer Test (PITT) at Hanford: Prepared for the Groundwater/Vadose Zone Integration Project, U.S. Department of Energy, Richland, Washington. HATFIELD, K., 1998, Newsletter of the American Institute of Hydrology, June issue, p. 7. HOLBERT, C.; ARENS, S.; HIMLE, C.; AND GORDER, K., 2004, Field observations four years after aggressive source removal using SEAR. In Gavaskar, A. R. and Chen, A. C. S. (Editors), Remediation of Chlorinated and Recalcitrant Compounds 2004: Battelle Press, Columbus, OH., 8 p. (CD distribution) HOLZMER, F. J.; DWARAKANATH, V.; EWING, J. E.; JIN, M.; AND LONDERGAN, J. T., 2005, Characterization of tetrachloroethene DNAPL in lowpermeability coastal plain sediments, North Carolina: Environmental Engineering Geoscience, Vol. 11, No. 3, pp. 197–208. ITRC, 2004, Strategies for Monitoring the Performance of DNAPL Source Zone Remedies: Interstate Technology & Regulatory Council, Washington, DC. JIN, M.; DELSHAD, M.; MCKINNEY, D. C.; POPE, G. A.; SEPEHRNOORI, K.; TILBURG, C.; AND JACKSON, R. E., 1994, Subsurface NAPL contamination: Partitioning tracer test for detection, estimation and remediation performance assessment: In Dutton, A. R. (Editor), Toxic Substances and the Hydrologic Sciences: American Institute of Hydrology, Minneapolis, MN, pp. 131–159. JIN, M.; DELSHAD, M.; MCKINNEY, D. C.; POPE, G. A.; SEPEHRNOORI, K.; TILBURG, C.; AND JACKSON, R. E., 1995, Partitioning tracer test for detection, estimation and remediation performance assessment of subsurface nonaqueous phase liquids: Water Resources Research, Vol. 31, No. 5, pp. 1201–1211. JIN, M.; BUTLER, G. W.; JACKSON, R. E.; MARINER, P. E.; PICKENS, J. F.; POPE, G. A.; BROWN, C. L.; AND MCKINNEY, D. C., 1997, Sensitivity models and design protocol for partitioning tracer tests in alluvial aquifer: Ground Water, Vol. 35, No. 6, pp. 964–972. JIN, M.; JACKSON, R. E.; AND POPE, G.A., 2000, The interpretation and error analysis of PITT data. In Wickramanayake, G. D.; Gavaskar, A. R.; and Gupta, N. (Editors), Treating Dense Nonaqueous-Phase Liquids (DNAPLs): Battelle Press, Columbus, OH, pp. 85–92. KAVANAUGH, M. C. AND RAO, P. S. C., 2003, The DNAPL Remediation Challenge: Is There a Case for Source Depletion: Report EPA/ 600/R-03/143, U.S. Environmental Protection Agency, 129 p. KUEPER, B. H.; ABBOTT, W.; AND FARQUHAR, G., 1989, Experimental observations of multiphase flow in heterogeneous porous media: Journal Contaminant Hydrology, Vol. 5, pp. 83–95. KUHN, T. S., 1996, The Structure of Scientific Revolutions, 3rd edition: University of Chicago Press, Chicago, 212 p. LEE, C. M.; MEYERS, S. L.; WRIGHT, JR., C. L.; COATES, J. T.; HASKELL, P. A.; AND FALTA, JR., R. W., 1998. NAPL compositional changes influence partitioning coefficients: Environmental Science Technology, Vol. 32, No. 22, pp. 3574–3578. MARINER, P. E.; JIN, M.; AND JACKSON, R. E., 1997, An algorithm for the estimation of NAPL saturation and composition from typical soil chemical analyses: Ground Water Monitoring Remediation, Vol. 17, No. 2, pp. 122–129. MARINER, P. E.; JIN, M.; STUDER, J. E.; AND POPE, G. A., 1999, The first
412
vadose zone partitioning interwell tracer test for nonaqueous phase liquid and water residual: Environmental Science Technology, Vol. 33, No. 16, pp. 2825–2828. MAYER, A. S. AND MILLER, C. T., 1992, The influence of porous medium characteristics and measurement scale on pore-scale distributions of residual nonaqueous phase liquids: Journal Contaminant Hydrology, Vol. 11, pp. 189–213. MCCRAY, J. E. AND COHEN, R. M., 2003, Software Spotlight/ NAPLANAL: A tool for analyzing NAPL saturation and composition: Ground Water, Vol. 41, No. 3, pp. 298–299. MEINARDUS, H. W.; DWARAKANATH, V.; JACKSON, R. E.; JIN, M.; GINN, J. S.; AND STOTLER, C., 2000, Full-scale characterization of a DNAPL source zone with PITTs. In Wickramanayake, G. D.; Gavaskar, A.; and Gupta, N. (Editors), Treating Dense Nonaqueous-Phase Liquids (DNAPLs): Battelle Press, Columbus, OH, pp. 17–24. MEINARDUS, H. W.; DWARAKANATH, V.; EWING, J. E.; HIRASAKI, G. J.; JACKSON, R. E.; JIN, M.; GINN, J. S.; LONDERGAN, J. T.; MILLER, C. A.; AND POPE, G. A., 2002, Performance assessment of NAPL remediation in heterogeneous alluvium: Journal Contaminant Hydrology, Vol. 54, Nos. 3–4, pp. 173–193. MERCER, J. W. AND COHEN, R. M., 1990, A review of immiscible fluids in the subsurface: properties, models, characterization and remediation: Journal Contaminant Hydrology, Vol. 6, No. 2, pp. 107–163. MORENO-BARBERO, E.; SAENTON, S.; AND ILLANGASEKARE, T. H., 2004, Potential use of partitioning tracer and mass flux emission data to characterize DNAPL source zone architecture. In Gavaskar, A. R. and Chen, A. S. C. (Editors), Remediation of Chlorinated and Recalcitrant Compounds 2004: Battelle Press, Columbus, OH, 8 p. (CD distribution). NELSON, N. T.; OOSTROM, M.; WIETSMA, T. W.; AND BRUSSEAU, M. L., 1999, Partitioning tracer method for the in situ measurement of DNAPL saturation. Influence of heterogeneity and sampling method: Environmental Science Technology, Vol. 33, pp. 4046–4053. OOLMAN, T.; GODARD, S. T.; POPE, G. A.; JIN, M.; AND KIRCHNER, K., 1995, DNAPL flow behavior in a contaminated aquifer: evaluation of field data: Ground Water Monitoring Remediation, Vol. 15, No. 4, pp. 125–137. PARKER, B. L.; CHERRY, J. A.; CHAPMAN, S. W.; AND GUILBEAULT, M. A., 2003, Review and analysis of chlorinated solvent dense nonaqueous phase liquid distributions in five sandy aquifers: Vadose Zone Journal, Vol. 2, pp. 116–137. PIEPENBRINK, M.; PTAK, T.; AND GRATHWOHL, P., 2002, Natural gradient partitioning and interfacial tests for NAPL source zone characterization at a former coal gasification plant. In Groundwater Quality: Natural and Enhanced Restoration of Groundwater Pollution: IAHS Publication 275, Sheffield, U.K., pp. 17–23. POPE, G. A.; SEPEHRNOORI, K.; SHARMA, M. K.; MCKINNEY, D. C.; SPEITEL, G. E.; AND JACKSON, R. E., 1999, Three-Dimensional NAPL Fate and Transport Model: EPA/600/R-99/011, U.S. Environmental Protection Agency, Cincinnati, OH. SANDIA NATIONAL LABORATORIES, 2003, Chemical Waste Landfill; Landfill Excavation, Voluntary Corrective Measure. Final Report: Environmental Restoration Project, Sandia National Laboratories, New Mexico. National Nuclear Security Administration, Sandia Site Office, pp. 2–33. SILVA, J. K. A.; MARINER, P. E.; MEINARDUS, H. W.; AND GINN, J. S., 2003, Monitoring partitioning tracer testing and surfactant flooding by in-line gas chromatography techniques: Ground Water Monitoring Remediation, Vol. 23, No. 1, pp. 85–92. YONGE, D.; HOSSAIN, A.; CAMERON, R.; FORD, H.; AND STOREY, C., 1996, Hanford Soil Partitioning and Vapor Extraction Study: Bechtel Hanford Inc., BHI-00861, Rev. 0, Richland, WA. YOUNG, C. M.; JACKSON, R. E.; JIN, M.; LONDERGAN, J. T.; MARINER, P. E.; POPE, G. A.; ANDERSON, F. J.; AND HOUK, T., 1999, Characterization of a TCE DNAPL zone in alluvium by partitioning tracers: Ground Water Monitoring Remediation, Vol. 19, No. 1, pp. 84–94.
Environmental & Engineering Geoscience, Vol. XI, No. 4, November 2005, pp. 405–412
