HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Parker, B. L. Articles by Guilbeault, M. A. Search for Related Content PubMed Articles by Parker, B. L. Articles by Guilbeault, M. A. GeoRef GeoRef Citation Agricola Articles by Parker, B. L. Articles by Guilbeault, M. A. Related Collections Field-Scale Studies Chlorinated Hydrocarbons Organic Compounds

SPECIAL SUBMISSIONS: Contaminant Characterization, Transport, and Remediation in Complex Multiphase Systems

Review and Analysis of Chlorinated Solvent Dense Nonaqueous Phase Liquid Distributions in Five Sandy Aquifers

Department of Earth Sciences, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1
^* Corresponding author (blparker{at}uwaterloo.ca)

Received 14 November 2002.

	ABSTRACT

TOP ABSTRACT INTRODUCTION SITE USES AND GENERAL... APPROACH AND METHODS RESULTS DISCUSSION AND IMPLICATIONS SUMMARY OF CONCLUSIONS REFERENCES

 
To select and design effective remedial measures for dense, nonaqueous phase liquid (DNAPL) source zones, better understanding of the architecture of these zones is needed. In this study, a suite of investigative techniques was applied to perform detailed vertical delineation of chlorinated-solvent source zones in sand aquifers at five contaminated industrial sites (two in Connecticut, and one each in Florida, New Hampshire, and Ontario). The DNAPL occurs in the middle of the aquifers at three of the sites and at or near the bottom at the other two. The DNAPL entered the subsurface at these sites decades ago, and therefore the DNAPL zones have aged due to groundwater dissolution. The suite of investigative techniques was used to perform profile sampling using direct-push methods, in which depth-discrete soil and groundwater samples were taken with extremely close vertical spacing. The sampling included methods to distinguish between free-product and residual DNAPL at two of the sites. At each location where DNAPL was found, the DNAPL occurred in one or a few thin layers, generally between 1 and 30 cm thick. These layers were positioned within distinct grain-size zones, or at contacts between sedimentological layers. In some cases, the DNAPL layers have no apparent textural association. For any particular sampling hole to have a high probability of finding such layers, continuous cores must be collected and sampling of these cores must be done at very close vertical spacing (5 cm or less). Free-product DNAPL occurrences in conventional wells at three of the sites indicated, misleadingly, much greater DNAPL layer thicknesses than actual, and in one case, the conventional well may have caused short-circuiting of DNAPL from the middle to the bottom of the aquifer. Although all of the DNAPL source zones are comprised of only sporadic, thin DNAPL layers representing little total mass, these source zones are the cause of high-concentration dissolved plumes down gradient.

Abbreviations: bgs, below ground surface • DCT, drainable core technique • DNAPL, dense nonaqueous phase liquid • MIP, membrane interface probe • PCE, tetrachloroethylene • PITT, partitioning interwell tracer test • TCE, trichloroethylene • VOA, volatile organic analysis • VOC, volatile organic compound

	INTRODUCTION

TOP ABSTRACT INTRODUCTION SITE USES AND GENERAL... APPROACH AND METHODS RESULTS DISCUSSION AND IMPLICATIONS SUMMARY OF CONCLUSIONS REFERENCES

 
PERSISTENT CHLORINATED SOLVENT contamination is common in unconfined sandy aquifers in industrial areas, normally as a result of organic solvents present as DNAPLs residing below the water table in zones known as source zones (U.S. EPA, 1992; Feenstra et al., 1996). Natural groundwater flow through the source zones causes the formation of plumes of aqueous-phase contamination, which typically evolve to occupy much larger aquifer volumes than the source zones, and pose much more risk to receptors and the environment than the source zones. The common remedial action taken to reduce these risks is to control the plume or the mass flux from the source zone by pump-and-treat. However, this approach requires pumping for many decades or even longer, and therefore it is not a permanent solution to the problem. A permanent solution requires removal or destruction of the DNAPL mass from the source zone. In recent years, several in situ technologies have been proposed for remedial restoration of chlorinated solvent DNAPL source zones (Interstate Technology and Regulatory Council DNAPL Team, 2002), and numerous site trials have been conducted, but no complete successes have been documented to date.

One of the limiting factors in the design and application of these in situ technologies is the paucity of information about the architecture of DNAPL source zones. Dense nonaqueous phase liquid source zones at industrial sites typically formed decades ago, when free-product DNAPL was spilled or leaked, causing infiltration of free product into the groundwater zone. Nearly all published information about the nature of DNAPL source zones in sand deposits comes from laboratory and field experiments and simulations using numerical models. Field experiments conducted in the sand aquifer at the Borden site showed that infiltration of free-product tetrachloroethylene (PCE) resulted in complex source-zone architecture (Kueper et al., 1993; Brewster et al., 1995). Shortly after the DNAPL marked with red dye was released in these experiments, the DNAPL achieved a stable distribution comprised of two types of DNAPL subzones: (i) horizontal layers or thin pools containing most of the DNAPL mass and (ii) vertical residual DNAPL pathways connecting the layers. The layers, which represent discontinuous patches of DNAPL, comprise both residual and free-product DNAPL. The layers typically form on top of interfaces between small sedimentological units exhibiting different texture. Kueper et al. (1989) and Illangasekare et al. (1995) used laboratory experiments to show that layers form in sand deposits even when the grain-size contrasts between beds in the sand are small, such as a change from coarse to somewhat finer sand. In the Borden experiments cited above, extremely small-scale sampling of cores and visual inspection for small-scale core features were needed to locate the DNAPL layers. The DNAPL layers formed preferentially in the coarser-grained horizons, which was expected given the nonwetting nature of the DNAPL.

Although laboratory and field experiments, as well as mathematical modeling, have provided considerable insight into the nature of DNAPL source zones in sandy aquifers at a few research sites, little is known about the nature of DNAPL source zones at actual contaminated industrial sites where the DNAPL entered the subsurface decades ago. The DNAPL source zones at industrial sites are commonly between two and five decades old. Old DNAPL source zones in sandy aquifers may have DNAPL volumes and distributions that are much different from what existed at the time of formation. The dissolution and mass export of DNAPLs from the source zones for decades would have removed a portion of the DNAPL mass, and this may have caused considerable change in some aspects of the source-zone architecture. At most DNAPL sites, the actual DNAPL has never been found (Feenstra and Cherry, 1996), and the field evidence used to designate a site as a DNAPL site is typically indirect (U.S. EPA, 1992). There is a need for detailed information about the occurrence and distribution of DNAPL at selected industrial sites for improvement of conceptual models of DNAPL source zones.

This review presents results of field investigations of the nature of DNAPL source zones at five industrial sites on sandy aquifers, where trichloroethylene (TCE) or PCE DNAPL has caused groundwater contamination. One of the sites is located in the Province of Ontario, two are in Connecticut (Connecticut A and B), and one each in New Hampshire and Florida. Each of the source zones is the cause of a distinct and persistent dissolved-phase plume that has been subjected to intensive plume monitoring. At the time of our studies, no subsurface remedial actions had been undertaken in the study areas, except for source zone containment and minimal free-product DNAPL recovery at the Connecticut A site.

Given what was known about the geology of each site and historical information on solvent use, four conceptual cases for DNAPL source-zone architecture (Fig. 1) were considered as reasonable paradigms for the formation of the source zones decades ago.

1. In the first type of source zone (Fig. 1a), the DNAPL penetrated to the bottom of the relatively homogeneous sand aquifer, where it accumulated in a free-product layer or pool on top of the aquitard. As the DNAPL traveled downward, it left a trail of residual DNAPL.
   
2. In the second case (Fig. 1b), much of the DNAPL did not reach the top of the aquitard, because of a stratified transition between the aquifer and aquitard. The DNAPL penetrated into the coarser-grained layers in this transition zone, where horizontal DNAPL accumulations formed. In some situations, the DNAPL would not reach the aquitard surface because of large retention capacity in the transition zone compared with the volume of release, or a lack of stratigraphic discontinuities or pathways downward through the transition zone.
   
3. In another case (Fig. 1c) the DNAPL descends into the sand aquifer, where it forms multiple DNAPL layers due to effects of subtle permeability contrasts in the sand or presence of finer-grained silty or clayey layers. This DNAPL layering, comprised of both residual and free-product DNAPL, provides greatly enhanced retention capacity in the aquifer, which limits DNAPL penetration.
   
4. On the other hand, if sufficient DNAPL was released to exceed the bulk retention capacity of the aquifer, DNAPL may also reach and accumulate at the aquifer bottom (Fig. 1d).
   

View larger version (38K): [in this window] [in a new window]   Fig. 1. Schematic representations of four scenarios for dense nonaqueous phase liquid (DNAPL) source zones in sandy aquifers: (a) DNAPL penetrates to bottom of homogeneous sand aquifer to form a bottom pool, (b) DNAPL penetrates through homogeneous sand and accumulates in a layered transition zone, (c) DNAPL forms layers of residual and free-product DNAPL suspended in the sand aquifer, (d) DNAPL forms multiple layers distributed throughout the aquifer thickness. Note that all scenarios have residual trails because the source zones are represented at early time.

	SITE USES AND GENERAL FEATURES

TOP ABSTRACT INTRODUCTION SITE USES AND GENERAL... APPROACH AND METHODS RESULTS DISCUSSION AND IMPLICATIONS SUMMARY OF CONCLUSIONS REFERENCES

 
The five sites have several characteristics in common, but also show considerable variety in geologic origin and site use (Table 1). The features in common include: single-component DNAPL (either TCE or PCE) that has been in the ground for decades, shallow water table (<5 m below ground surface [bgs]), shallow maximum depth of DNAPL occurrence (<20 m) and geological conditions suitable for effective use of direct-push drilling equipment (e.g., lack of obstructions such as cobbles or boulders). The DNAPL at each site is the result of routine solvent use and storage at facilities engaged in production of metal products. Each of the sites has several locations where DNAPL may have entered the subsurface, which contributes to considerable spatial variability of the source zones. Information on site use (Table 1) indicates that DNAPL probably first entered the subsurface at these sites in the 1950s or 1960s, and the releases ceased by the early 1970s to 1980s. Although historical information on the use of chlorinated solvents is available for all but the Ontario site, the volume of DNAPL lost to the subsurface and the exact locations of the release points are generally unknown. The prospect for source zone remediation using in situ technologies is an issue at all of the sites. At each of the sites, groundwater monitoring was conducted by consultants for the site owners for at least several years before the initiation of our studies in 1996. At three of the sites, free-product DNAPL was encountered during previous investigations in conventional monitoring wells (Connecticut A and B, Ontario). At the Connecticut A site, the area in which wells showed free-product DNAPL was isolated in 1994 with a steel sheet piling enclosure keyed into the underlying aquitard. At the other sites, no subsurface remedial measures were deemed needed or implemented until after our source zone studies were completed.

View larger version (49K): [in this window] [in a new window]   Fig. 2. Columns representing the geology at each of the five study sites. The depth zone of DNAPL source zone investigations at each site are indicated by the dashed lines.

	APPROACH AND METHODS

TOP ABSTRACT INTRODUCTION SITE USES AND GENERAL... APPROACH AND METHODS RESULTS DISCUSSION AND IMPLICATIONS SUMMARY OF CONCLUSIONS REFERENCES

Each of the investigative techniques filled a particular role. Whenever a hole is sampled, DNAPL depth and thickness are unknown until DNAPL is encountered. For each hole, there was the expectation that the DNAPL would occur in thin layers, and therefore it was necessary that at least a portion of the hole be sampled at very close vertical spacing. Therefore, to focus this intensive core sampling on zones with the highest probability of DNAPL occurrence, the Waterloo Profiler was used first at all but the Ontario site, where the flow rate into the sampling tool was too slow due to the fine-grained nature of the aquifer deposits. Therefore, at four of the five sites, the general depth intervals in which the coring techniques were applied most intensively were those identified first using the Waterloo Profiler, in combination with rapid on-site VOC analyses so sampling decisions could be made in the field as the investigations proceeded. Each of the investigative techniques applied at these sites is summarized below.

Waterloo Groundwater Profiler
The Waterloo Profiler (Pitkin et al., 1999) is a direct-push device for collecting depth-discrete groundwater samples in unconsolidated granular deposits. This tool allows rapid collection of samples at multiple depths in the same hole without retrieving, decontaminating, and redriving the tool between samples, with no drill cuttings and only minimal purge water generated. The device comprises a profiler head, consisting of a 4.4-cm-diam. stainless-steel drive-point with open ports fitted with stainless-steel screens. The ports convey water into a common internal fitting, which is then connected to 3-mm-o.d. stainless-steel tubing. The head screws into conventional AW drill rods of the same diameter, with the sample tube running inside the rods to convey groundwater from the head to ground surface, using a peristaltic pump placed downstream from a stainless-steel sampling manifold, containing 25- or 40-mL volatile organic analysis (VOA) sample vials. This setup avoids contact between the groundwater sample and the pump tubing and exposure to air.

In summary, to collect a groundwater sample, the Profiler tip is advanced using direct-push equipment, while contaminant-free water (e.g., distilled water) is pumped down the sample tube and out the ports to prevent ports from clogging with fine sediment, and to purge formation water from the previous sampling depth. When the desired sampling depth is reached, the pump is reversed to begin pumping water to the surface, and sufficient volume purged (typically 200 mL) to remove any distilled water remaining in the tubing before collecting the sample. If necessary, additional water can be purged to "develop" the zone around the sampling ports if turbidity is a problem. After collection of the sample, the tip is advanced to the next depth, again injecting clean water, and the process repeated until the maximum desired depth is reached. Samples can be collected at vertical spacings as close as 15 cm without causing overlap of sample zones in the aquifer. In finer-grained silty or clayey zones, the time required to obtain sufficient sample volume may be large, so that sampling in such zones is not practical. Purge times can be used to provide a relative indication of permeability for each sample depth. By monitoring pressures with a gauge positioned before the sampling manifold, the depths of low permeability zones (e.g., aquitard interface or lower permeability layers) can be accurately determined. An adaptation of the profiler head is the use of a disposable tip, so that the hole can be grouted during removal in cases where collapse is not expected and contaminant cross-connection is of concern (e.g., when sampling within DNAPL source zones).

Direct-Push Piston Sampler
Cores in the aquifers, and in some cases from the upper part of the underlying aquitards, were collected using the piston core barrel described by Zapico et al. (1987). This coring method provides excellent recovery of relatively undisturbed samples of cohesionless sandy deposits from below the water table. The major components of the system consist of an aluminum sample tube (5.1-cm o.d., 1.52 m long, 1.3-mm wall thickness) inside a steel casing that forms the core barrel and a piston inside the sample tube. The sample tube contains the sample, and a new tube is used for each core run. Steel casing is connected to the top of the core barrel and a drive head is attached to the top of the steel casing. The core barrel was driven into the ground using the air-hammer and scaffolding method of Starr and Ingleton (1992) or the direct-push rig described by Einarson (1995). The bottom of the piston extends beyond the sample tube and is pointed to facilitate driving the sampler through the soil and to prevent material from entering the sample tube before the sampling depth is reached. The sampler is driven to the desired core start depth with the piston fixed in place. Then the piston is tied off and the core barrel advanced through the sampling interval. The moving piston creates suction that prevents the sand or gravel from falling out of the core barrel as the sampler is brought to the surface and also helps retain the original pore fluids in the core. In DNAPL zones, the core was handled with extra care to minimize the potential for fluids (water and DNAPL) to redistribute within the core before subsampling. The ability to collect relatively undisturbed cores using the piston corer, which provided excellent recovery (typically >95%), was essential for the subsequent core examination and subsampling procedures described below.

Drainable Core Technique
At two of the field sites (Connecticut A and B), cores were subjected to a draining procedure in the field before soil subsampling, termed the drainable core technique (DCT), to determine the depths where high DNAPL phase saturations were present and free product would drain from depth-specific zones in the core. The DCT is used in combination with the Sudan IV screening test, described below, which identifies in the field the presence of any DNAPL, but does not distinguish between free-product (i.e., potentially mobile) and residual DNAPL. Therefore, the DCT can be used to determine depths at which free-product DNAPL occurs, and by default the depths of residual product are inferred where the Sudan IV screening test indicates a positive result for nonaqueous phase liquid (NAPL) presence, but free-product DNAPL was absent during the draining. Because this procedure is time-consuming, it was not conducted during initial investigations of DNAPL accumulation zones, but only after the depths of DNAPL zones were first investigated by detailed core subsampling (described below).

The standard version of the DCT is mostly applicable to coarser-grained sand aquifers where fluids drain readily under gravity (e.g., Connecticut A Site). The DCT requires cores collected using the piston core barrel, and it is applied when the cores are still contained in the aluminum sample tube. At locations where DNAPL accumulations are at the bottom of aquifers, the core run is selected to extend a short distance into the underlying aquitard to allow the aquitard material to form a plug in the bottom of the core; this ensures containment of DNAPL in the core while the core is raised to the surface. During and after removal from the core barrel, the core tube is maintained in a vertical position throughout the draining procedure. The core is clamped to a vertical board or wall, and the upper core cap removed. The core is then sequentially drained from top to bottom at closely spaced vertical intervals (typically 2.5–5.0 cm) by drilling small holes (3 mm) into the core tube, and collecting fluids that drain from each depth into a 25- or 40-mL glass vial. Where DNAPL drained from an interval, it typically flowed first followed by water. After fluid drainage ceased from an interval, usually after about 10 min, the volumes of fluids (water/DNAPL) that drained from the interval were recorded and draining of the next interval was started. Draining of the core continued until the aquitard interface was reached and no fluids would drain from the core. The core was then subsampled in the manner described below. Application of the DCT typically requires several hours for each core, although several cores can be drained concurrently to increase efficiency. For sites with finer-grained sands where pore fluids do not drain readily by gravity, such as the Connecticut B site, the technique can be modified. In this case, fluid extraction was enhanced by applying pressure to sections of the core to expel the pore fluids. The modified DCT was applied to specific sections of cores collected adjacent to an initial continuous core that was screened for DNAPL using the Sudan IV method described below to target DNAPL zones identified by the rapid screening technique.

Core Subsampling Procedure
Core samples contained within the aluminum core tubes were subsampled using both a screening technique using a hydrophobic dye (Sudan IV) for visual identification of DNAPL and for quantitative laboratory analysis of VOCs. In this procedure, the core tube is placed horizontally on a firm surface in a wooden holder, and split from end to end, first using a circular saw to cut through the aluminum, and then using a wire to separate the soil material. The half of the core intended for subsampling is immediately covered with aluminum foil to minimize volatilization of VOCs and moisture loss, and the other half used for core photographs, detailed geologic logging, selection of depths for subsampling, and for subsequent physical parameter tests (e.g., f_oc, grain size analysis, permeameter tests). Numerous small cylindrical soil samples are collected along the core using stainless-steel sampler and plunger devices. The potential for cross-contamination is minimized by scraping off the upper few millimeters of soil and taking care not to advance the sampler to the outside wall of the core tube. Subsamples are taken at close vertical spacings, typically 2.5 to 10 cm, so that detailed concentration profiles are obtained. The samplers and plungers are decontaminated between each sample depth using a three-part wash and rinse sequence with soapy water, methanol, and deionized water.

Two subsamples are collected from each depth, one for visual detection of NAPL phase using the hydrophobic dye (Sudan IV) test, and the other for later laboratory VOC analyses. Cohen et al. (1992) described the use of Sudan IV in DNAPL investigations. The sample for DNAPL screening was extruded into a 25-mL glass vial containing a small amount of Sudan IV powder (Fisher Scientific, Hampton, NH), a hydrophobic dye that solubilizes into immiscible-phase chlorinated organic liquids, causing a bright red color, and a few milliliters of deionized water. The sample for VOC analysis was extruded into a preweighed 25-mL glass VOA bottle, containing a known volume of HPLC-grade methanol (approximately 15 mL) for preservation and VOC extraction. Small variations in analytical procedures were followed for the different sites; however, the following is a general summary of the procedure. Soil samples preserved in methanol were shaken on a vortex mixer after arriving at the lab to break up the soil. Samples were then stored at 4°C for at least 2 wk to allow adequate time for extraction. The samples were then centrifuged to separate the soil and methanol in the sample vial, and a small aliquot of methanol extract from the sample vial diluted into pentane (capillary GC grade) containing an internal standard. Further dilutions of samples into methanol were performed as necessary based on initial analysis of the undiluted sample. Standards were laboratory-prepared mixtures of the target analytes (typically PCE, TCE, DCE isomers) in methanol, which were spiked into pentane as described above for the samples. The pentane was then analyzed by direct injection on a gas chromatograph equipped with a micro-electron capture detector and a liquid autosampler. This technique provides the total TCE content per mass of wet soil, and therefore does not distinguish between TCE present in the aqueous, sorbed, and NAPL phases.

Calculation Procedure Applied to Volatile Organic Compound Analyses
The total soil concentration (C_t) obtained directly from the lab analysis represents the total analyte (e.g., TCE) mass per unit mass of bulk wet soil sample, and therefore includes dissolved mass in the pore water and sorbed mass on the solids, as well as any DNAPL phase in the sample. This total analyte concentration can be converted to a pore water concentration (C_w) assuming equilibrium chemical partitioning between the solid and water phases, and no DNAPL phase present using

[1]

where C_t is the total analyte concentration in the bulk sample from the lab analyses (µg g^-1 wet soil), _bwet is the wet bulk soil density (g cm^-3), is the soil porosity, and R is the estimated retardation factor due to sorption that assumes rapid, linear, and reversible partitioning between the VOC and solid-phase organic C in the aquifer sediments. Presence of DNAPL is inferred for samples where the estimated porewater concentration exceeds the aqueous solubility of the contaminant. Feenstra et al. (1991) described such partitioning calculations, including more general cases involving multicomponent DNAPL and unsaturated conditions. The retardation factor is estimated using the relation:

[2]

where K_d is the distribution coefficient and _b is the dry bulk soil density. Where possible, site-specific parameter values were applied, while in some cases estimated values were used. The distribution coefficient can be obtained from laboratory batch or column experiments, or estimated using the correlation K_d = K_ocf_oc, using literature values for K_oc (e.g., Table A.1 in Pankow and Cherry, 1996) and measured f_oc values (when available). Equation [1] can also be used to estimate the expected minimum soil concentration in a sample containing DNAPL, by setting C_w to the solubility limit and calculating the expected total soil concentration. Thus the presence of NAPL can be inferred where the measured total soil concentration exceeds this minimum calculated value, providing a check on the Sudan IV screening test for NAPL presence. Site-specific f_oc measurements were performed on sediments from the Connecticut A, Florida, and Ontario sites using the method described by Churcher and Dickhout (1987), and were subsequently used for estimates of retardation factors and concentrations indicative of DNAPL presence (Table 3). For the Connecticut B and New Hampshire sites, f_oc measurements were not performed, so for these sites the aquifer f_oc was assumed to be the same as the Connecticut A site, given the similarity in depositional environments.

[3]

where S_nw is the nonwetting phase (NAPL) saturation (unitless), _{b wet} is the soil wet bulk density (g cm^-3), C_t is the total soil concentration (µg VOC g^-1 wet soil), _nw is the NAPL density (g cm^-3), and is the aquifer porosity (unitless). This calculation assumes that all of the contaminant mass occurs in the DNAPL phase and thus neglects mass in the sorbed and aqueous phases. This assumption is reasonable because, in the types of soil materials involved in this study, the mass in the sorbed and aqueous phases is small compared with the DNAPL mass. For example, using the minimum C_t values representative of DNAPL presence (i.e., maximum aqueous and sorbed mass present), the NAPL saturation values would be less than about 0.1 and 0.05% for TCE and PCE, respectively, using typical site parameters. In most cases such low values are well below estimated DNAPL saturations, and therefore neglecting the sorbed and aqueous mass is reasonable unless the DNAPL saturation is very low. At the Connecticut A site, cores subjected to the DCT were subsequently subsampled for VOC analyses. For depths where DNAPL drained, this calculation of NAPL saturation based on VOC analyses alone could greatly underestimate the percentage of NAPL saturation. Therefore, the DNAPL volume drained from free-phase zones before core subsampling was also included in the NAPL saturation estimates.

	RESULTS

TOP ABSTRACT INTRODUCTION SITE USES AND GENERAL... APPROACH AND METHODS RESULTS DISCUSSION AND IMPLICATIONS SUMMARY OF CONCLUSIONS REFERENCES

 
Connecticut A Site
The TCE DNAPL source zone at the Connecticut A site occurs in a former tank farm area at the east side of a large manufacturing facility. The geology in this area (Fig. 2) consists of a 9-m-thick medium- to coarse-grained sand aquifer with exceptional uniformity overlying a thick clayey silt aquitard, with an abrupt contact separating the aquifer and aquitard. The elevation of the aquitard interface varies by <1 m within the source area. Three conventional wells in the source zone indicated an apparent DNAPL thickness at the base of the aquifer ranging from about 0.3 to 0.7 m. Therefore, the initial conceptual model for the DNAPL distribution was a continuous pool of free-product DNAPL at the base of the very permeable sand aquifer. The majority of the source zone area was isolated by a steel sheet-pile enclosure installed in late 1994, which was keyed into the underlying aquitard.

In this study, the nature of the DNAPL accumulation zone was investigated using the detailed core subsampling procedures and DCT described above during field episodes from 1996 to 1997, where a total of 36 cores were collected throughout the source area within and outside the enclosure, in an approximate area of 30 by 40 m. Cores were typically collected over a 1.52-m interval at the base of the aquifer, extending a short distance into the underlying aquitard. Dense nonaqueous phase liquid was present at 23 of the 36 locations, typically occurring within the bottom 0.3 m of the aquifer, with 13 of the locations indicating both free-phase and residual DNAPL and 10 locations indicating residual DNAPL only. At three locations where the aquitard surface was found at a lower elevation, DNAPL was found in the bottom 1 m of the aquifer. Where DNAPL was present, it was distributed in thin layers typically ranging from two to several centimeters thick. Figures 3a to 3c show three profiles collected from this site. Site-specific measurements of f_oc of the aquifer sediments were used for estimates of the TCE retardation factor and soil concentration indicative of DNAPL (Table 3), shown on the example profiles.

View larger version (21K): [in this window] [in a new window]   Fig. 3a. Profile from the Connecticut A Site WCP-9, next to a conventional well with DNAPL.

View larger version (21K): [in this window] [in a new window]   Fig. 3b. Profile from the Connecticut A Site WCP-7, next to a conventional well with DNAPL.

View larger version (19K): [in this window] [in a new window]   Fig. 3c. Profile from the Connecticut A Site WCP-11, within an area of a localized depression in the aquitard interface showing the thickest DNAPL accumulation zone.

In cores where the DNAPL accumulation zones were thicker, typically within depressions on the aquitard surface, the DNAPL was stratified vertically in discontinuous layers of residual or free-product DNAPL (Fig. 3b and 3c). This vertical stratification suggests that the textural variations controlling DNAPL distribution are so subtle that they are not evident during visual core examination. A conventional well next to one of these locations (Fig. 3b) suggests a thick continuous DNAPL column with the top of the column coinciding with the uppermost DNAPL layer in the core. No conventional well was positioned next to the other core location (Fig. 3c). However, it's likely that the DNAPL level in such a well would be controlled by the upper free-phase layer and therefore would indicate a thick continuous zone of free-phase DNAPL in the aquifer. In both cases where cores were collected adjacent to conventional wells, the wells provided a misleading interpretation of actual DNAPL conditions, giving the impression of a large free-product pool in the bottom of the aquifer.

Using the measured total soil concentrations, estimates of the DNAPL phase saturations were performed using Eq. [3]. The DNAPL volumes that drained from the cores during application of the DCT were also included in these estimates. The DNAPL saturations (S_nw) ranged from about 10% or less in residual zones to more than 50% in free-phase zones at the base of the aquifer. These values indicating residual and free-product are consistent with the ranges reported by Feenstra et al. (1996).

Connecticut B Site
The Connecticut B facility is situated on a layer of fill with a thickness of 1.5 to 3.0 m, which overlies a 6-m-thick medium to fine sand aquifer. The aquifer sediments become finer and change from brown to gray at a depth of 10 to 12 m bgs, marking the top of a transition zone consisting primarily of layered silt and fine sand. Below the transition zone is a layered silt and clay aquitard (Fig. 2). At this site, two DNAPL source areas were investigated in 1999 using the Waterloo Profiler and detailed core subsampling and modified DCT procedures described above. In each source area, a monitoring well indicated the presence of free-product PCE. In the first stage of investigation at this site, the Profiler was used to collect depth-discrete groundwater samples from a hole at each location. Then, at each location, two continuous cores were collected, with one core subjected to subsampling and the other subjected to the modified DCT with screening for DNAPL using Sudan IV in both the drained fluids and soil core subsamples. Site-specific measurements of f_oc were not performed. Therefore the average value for aquifer sediments at the Connecticut A site was assumed for estimating the PCE retardation factor and DNAPL indicative concentration (Table 3). These estimates are subject to more uncertainty, particularly given the degree of geological layering at this site, but compare well to Sudan IV test results.

The first source area investigated (AT) as part of this study site was the location of a former plating-shop degreaser, where a conventional well (MW-18M) showed free-product DNAPL with a thickness ranging from 0.5 to 0.7 m during 1997 to 1999 (Fig. 4a). The detailed core sampling done at this location found seven separate thin DNAPL layers. Six of these DNAPL layers occurred in the depth range of 4.9 to 7.0 m bgs, and one at 10.7 m bgs. Seven DNAPL layers were also observed in the cores subjected to the modified DCT at location AT-2 collected about 1 m away. The shallowest DNAPL zone (Zones 1 and 2) were located from 4.9 to 5.5 m bgs and were about 0.25 and 0.45 m thick, respectively, with only a 2.5-cm-thick layer separating them. Free-phase DNAPL was not observed in the drained fluids in AT-2 from this depth, indicating residual DNAPL only. Three thinner DNAPL layers (Zones 3, 4, and 5) were observed in AT-1 between 5.8 to 6.1 m bgs, with a thickness of about 10, 5, and 5 cm, respectively. These were not observed in the adjacent AT-2 core, suggesting the DNAPL layers may be quite discontinuous laterally. The sixth DNAPL zone (Zone 6) was observed from 6.5 to 7.0 m, in both AT-1 and AT-2, with DNAPL also observed in the drained fluids from AT-2, suggesting the presence of free-phase DNAPL. The seventh DNAPL zone (Zone 7) was a thin layer (approximately 5 cm) at a depth of 10.7 m bgs observed in AT-1 and in the soil at AT-2. A final and eighth DNAPL zone was observed from 11.0 to 11.3 m bgs in AT-2 drained fluids and soil, but this zone was not observed in the AT-1 cores and is not shown in Fig. 4a. The positions of the DNAPL zones indicated by the Sudan IV method coincided closely with DNAPL presence indicated by the total PCE concentrations, suggesting that the parameters used to estimate the DNAPL indicative concentrations were appropriate. The product levels in the nearby conventional well at this location (MW-18M) were misleading and provided no useful measure of the actual DNAPL distribution in the subsurface. At this location, it is quite possible that DNAPL entered the well from the shallow layers (possibly Zone 6, which intersects the top of the well screen, and contained free-phase DNAPL) to cause accumulation at the bottom of the well. This cross-connection may have caused the deeper DNAPL zones observed in the formation near the bottom of the well (Zone 7). It is conceivable, however, that the DNAPL migrated downward from the upper DNAPL layers through stratigraphic discontinuities, and then laterally at depths where the deepest DNAPL layers were observed in the cores. Such vertical pathways would be difficult to observe in a limited number of vertical cores.

View larger version (22K): [in this window] [in a new window]   Fig. 4a. Profile from Connecticut B Site AT-1, where conventional wells contain DNAPL, at the location of a former plating shop degreaser.

View larger version (23K): [in this window] [in a new window]   Fig. 4b. Profile from Connecticut B site at two locations where conventional wells contain DNAPL, BT-1 and BT-2, at the location of a former PCE bulk storage tank.

Ontario Site
At the Ontario Site, a former manufacturing facility situated within a primarily residential area, the DNAPL zone comprised of TCE occurs in a small area near the northeast corner of the property with a portion of the source occurring beneath the building and an adjacent roadway. No records of TCE use at the site were found in previous studies. It is surmised that releases occurred from a storage tank in this area, or by disposal of spent solvents just outside the building. The general geology in the source area (Fig. 2) consists of an upper perched aquifer comprised of about 4 m of fine sand, overlying a 1-m-thick finer-grained and layered transition zone that abruptly changes to a uniform clayey silt till aquitard. The transition zone over the bottom meter or so of the aquifer exhibits much more textural variability, with thin layers or lenses of silty and clayey material separating distinct sandy layers. Four conventional monitoring wells in the source zone showed free-product thickness ranging from 20 to 100 cm in previous investigations. Contours of the surface of the underlying clayey till aquitard suggested that the DNAPL was trapped in a stratigraphic depression in the aquitard surface.

Source zone investigations, conducted between 1998 to 2000, included collection of continuous cores from 19 locations, spanning the zone from the water table (approximately 2.5 m bgs) to slightly into the underlying aquitard. In a few cases, cores were advanced deeper into the aquitard, so that the full TCE distribution in the aquitard could be investigated. Given the fine-grained and layered nature of this aquifer, the DCT was not applied. Table 3 provides estimated TCE retardation factors and soil concentrations indicative of DNAPL, using separate average f_oc values for the silty sand aquifer and sandy layers in the transition zone, vs. the silt–clay layers in the transition zone.

Figure 5 shows a representative profile at one core location adjacent to a conventional well containing about 100 cm of DNAPL. At this location, the detailed coring and subsampling indicated the DNAPL occurs entirely within the transition zone, within two thin vertically discrete sandy layers separated by silty or clayey beds. These results are consistent with results from other locations, where DNAPL also occurs in one or two thin layers with a maximum layer thickness of 10 cm. No DNAPL was found in the aquifer overlying the transition zone, suggesting there has been sufficient time for any DNAPL within the aquifer to dissolve away. The conventional wells indicated much thicker DNAPL accumulations. Where cores were collected adjacent to wells with DNAPL, the top of the DNAPL zone in each well coincided with the highest DNAPL level in the cores, suggesting that DNAPL from the shallow thin layers flowed into the wells to provide the apparent large product column. The full extent of the TCE contamination in the underlying aquitard was determined to nondetect concentrations at one location (Fig. 5), with the shape of the concentration profile below the aquitard interface showing diffusion-dominated transport in the aquitard, confirming the maximum depth of DNAPL penetration and lack of DNAPL entry into the aquitard. The abrupt shift in the total soil TCE concentration across the aquitard interface (Fig. 5) can be attributed to the difference in sorption for these lithologies. Based on an average f_oc of the aquitard of 0.24% (average from four samples), the TCE retardation factor is estimated to be 2.2 for the aquitard, compared with about 1.2 for the aquifer sediments. The difference in retardation factors, along with differences in porosity and bulk density, adequately accounts for the observed shift at this depth.

View larger version (24K): [in this window] [in a new window]   Fig. 5. Profile from the Ontario Site AC-6 at the edge of a roadway adjacent to the site and next to a conventional well containing free-product DNAPL. Note that TCE concentrations for three points that fall above the concentration scale are posted.

The difference in the actual DNAPL distribution vs. the initial impression based on large product thickness in wells has important implications. First, the actual volume of DNAPL present within the thin layers is much lower than expected if DNAPL formed a continuous free-product pool. Second, because the DNAPL is suspended within the transition zone and not present at the aquitard interface, there is no driving force for DNAPL entry into and migration through the aquitard, which is relatively thin, and may have fractures or other preferential pathways for DNAPL flow. This is a favorable condition because the aquitard is underlain by an aquifer used for municipal water supply. Finally, as the profiles indicate, considerable TCE mass occurs within low permeability layers within the transition zone and in the underlying aquitard. Removal or in situ destruction of such mass will be difficult due to diffusion control, which has implications for selection and design of source zone remedies.

Florida Site
The Florida site is an operating metal fabricating and cleaning facility situated on an aquifer formed of beach sand and bioclasts, with a thin (5–15 cm) continuous clayey layer at a depth of approximately 8 to 10 m bgs, and a few discontinuous silt and clay lenses at greater depths (Fig. 2). Accidental TCE releases may have occurred in three general areas on the property between 1964 and 1976, when records show TCE was used at the site. Trichloroethylene contamination in groundwater was discovered in 1966, which prompted sampling of shallow domestic wells in the adjacent residential area in 1968. Several monitoring wells were installed in 1984, and an extensive network of monitoring wells was installed between 1992 and 1994, with wells located on and off the property.

In 1996, detailed sampling along a vertical section across the plume was conducted at a location immediately down gradient of areas of the suspected DNAPL release (Guilbeault, 1999). The TCE concentration distribution along this cross section indicated, by projection up gradient, the presence of three areas where DNAPL must reside in the sand aquifer near and beneath the main industrial building. A total of 64 cores, each 1.52 m long, were collected from 14 holes during a search for DNAPL in these source areas. The Waterloo Profiler was also used in this search. One of these locations (B207), within the aquifer of a documented DNAPL release in 1971, showed TCE concentrations at aqueous solubility (about 1100 mg L^-1, Table 2) at one sampling depth. Based on this indirect evidence of DNAPL, a core hole (PM-18) was drilled within 0.3 m of this location. Figure 6 shows the results of the detailed sampling at PM-18, along with groundwater sampling at B207. The Sudan IV method showed DNAPL presence in one thin zone where the sample spacing was 2.5 cm. In this depth zone, two adjacent core subsamples clearly showed DNAPL presence, and therefore the Sudan IV results indicate a DNAPL zone less than 5 cm thick. The measured TCE concentrations from soil VOC analyses showed a peak in one sample at this same depth, but at a value slightly less than the estimated TCE DNAPL indicative value for the aquifer. This thin DNAPL zone at the shallower peak occurs in sand lying on top of a distinct clayey layer. Additional high VOC concentrations occur in samples below where the Sudan IV results were positive. However, these occur within the fine-grained clayey layer that has a higher affinity for TCE sorption, and therefore do not necessarily indicate the presence of DNAPL. The DNAPL indicative value for the clayey layers is much higher (Table 3) than the total TCE concentrations, which supports the lack of DNAPL determined from the Sudan IV tests in the clayey zone.

View larger version (22K): [in this window] [in a new window]   Fig. 6. Profile from Florida Site PM-18 at the only location where DNAPL was found. Groundwater concentrations are also shown for an adjacent profiler location (B207).

New Hampshire Site
The New Hampshire site is located in a valley where a tool and dye factory used PCE for degreasing operations between 1957 and 1983. The site lies on an unconfined sandy aquifer of valley fill of glaciofluvial origin composed mainly of stratified sand and gravel on top of a sandy glacial till deposited directly on bedrock (Fig. 2). A boulder and cobble layer extends 3 to 6 m below the ground surface, and bedrock occurs at about 30 m bgs. Industrial activities at the site ceased in 1983, after testing of a municipal well located approximately 1 km down gradient from the factory showed contamination. At this time, inspection of site infrastructure and mapping of the contaminant plume using conventional monitoring wells, combined with soil sampling, indicated that part of the plume was caused by leaks through a floor drain from an aboveground PCE storage tank in the building.

A small area centered on the location of the former PCE storage tank was selected for detailed investigation as part of this study after the demolition of the building in 1998. In an area with a diameter of 5 m, groundwater sampling at five locations was done using the Waterloo Profiler. Target zones for coring to search for DNAPL zones were then selected based on these groundwater PCE concentration profiles, with collection of eight cores, each 1.52 m long, using the piston core barrel from six locations.

Using the Sudan IV method at 5-cm spacing, DNAPL was found at two of these core holes (SM-2 and SM-5), with two layers in each hole (Fig. 7a and 7b). In one core, SM-5 (Fig. 7a), the two layers were barely distinct within a 30-cm interval, with only a 2.5-cm layer with no DNAPL (based on the Sudan IV test) separating them. However, the total soil PCE concentration at this depth falls above the DNAPL indicative value, suggesting a single DNAPL layer, as opposed to two distinct layers. Site-specific measurements of f_oc were not performed on these cores, so the same average value determined for sediments of the Connecticut A site was applied for estimation of the PCE retardation factor and DNAPL indicative concentration for these aquifer sediments (Table 3). In the second core, SM-2 (Fig. 7b), the two DNAPL layers were separated by a 70-cm-thick zone. Soil PCE concentrations in the upper 15 cm of the coarser-grained sandy layer, below the upper DNAPL zone, are also above the DNAPL indicative value, indicating DNAPL may also occur in this layer, but at extremely low residual saturations that may not have been observed by the Sudan IV screening technique. However, the lack of site-specific parameters for estimating sorption leaves this technique unreliable compared with the Sudan IV test. The top of the upper DNAPL layer and bottom of the lower DNAPL layer were not determined at this location.

View larger version (19K): [in this window] [in a new window]   Fig. 7a. Profile from New Hampshire Site SM-5 where DNAPL was found.

View larger version (20K): [in this window] [in a new window]   Fig. 7b. Profile from New Hampshire Site SM-2 where DNAPL was found.

	DISCUSSION AND IMPLICATIONS

TOP ABSTRACT INTRODUCTION SITE USES AND GENERAL... APPROACH AND METHODS RESULTS DISCUSSION AND IMPLICATIONS SUMMARY OF CONCLUSIONS REFERENCES

 
Continuous cores subjected to Sudan IV sampling at extremely close spacing, 5 cm and sometimes closer, were required to find DNAPL and define layer thicknesses at each of the five sites. The DNAPL was not visible in the cores without using the Sudan IV dye test. The core samples analyzed for VOCs, which were generally taken at the same vertical spacing as the Sudan IV samples, corroborated the Sudan IV results. However, if the VOC results had been used alone, the DNAPL identifications would have been uncertain because the calculation procedure for DNAPL identification depends on estimates of porosity and sorption (i.e., mass partitioning). This uncertainty is particularly relevant to those DNAPL layers where the percentage of DNAPL saturation (S_nw) is very low. The depth-discrete groundwater sampling done using the Waterloo Profiler served, for screening purposes, to identify high-concentration zones where DNAPL occurrence was most probable. This sampling only rarely identified actual DNAPL layers, because the sample spacing was too large, and because of dilution caused by the depth integration of each sample, given the thinness of the DNAPL layers typical of these source zones.

The drainable core technique is the most time-consuming of the field methods, and it was particularly useful at only one of the five sites. This technique did not always yield DNAPL from finer sand layers, even when we were certain that free product existed in these layers. Even when cores do produce DNAPL, the conditions within the core allowing DNAPL drainage are not controlled or monitored. Therefore, this method is not widely applicable. Of the three methods for finding DNAPL (DCT, soil VOC analyses and partitioning calculations, and the Sudan IV hydrophobic dye method), the Sudan IV method is the most direct and provides results rapidly in the field as sampling proceeds.

Free-product thicknesses measured in conventional monitoring wells at three of the sites indicated major DNAPL pools, but these measurements do not represent what is actually in the ground, because the wells do not indicate the layered (i.e., vertically discontinuous) nature or position of the DNAPL distribution. Therefore, estimates of the DNAPL volume in these source zones based on product thickness in conventional wells were grossly exaggerated relative to estimates based on the measured thicknesses of the individual layers. Another implication of the finding that the DNAPL zones consist of thin layers rather than thick pools is that the DNAPL has much less potential for downward remobilization, except in situations where wells cause short circuiting of free-product layers. The DNAPL occurrences at the five sites displayed a wide range in DNAPL saturations (S_nw). Some of the layers had free product, indicated by high S_nw values and by results from the drainable-core tests. Installation of a monitoring well through such layers can worsen the site contamination if the well screen or sand pack allows DNAPL drainage from a layer positioned near the top of the screen or sand pack down to the bottom of the well. The Connecticut B site provided a plausible example of this type of DNAPL cross connection.

In the conceptual model for aged DNAPL source zones in sandy aquifers supported by this study, all or nearly all DNAPL layers occur within finer-grained sand units, as layers sandwiched between silty or clayey strata in transition zones or at aquitard interfaces. In this study, the field techniques used to find the DNAPL layers performed well; however, these techniques are not the only ones that have been advocated for determining the nature of chlorinated-solvent DNAPL source zones. Two prominent alternative techniques are the partitioning interwell tracer test (PITT) method described by Jin et al. (1995) and direct-push probes that can sense chlorinated solvents, such as the membrane interface probe (MIP) described by Christy (1996) and Griffin and Watson (2002). The occurrence of DNAPL layers in less permeable aquifer zones, or in geologically layered transition zones, is disadvantageous for PITT because of the propensity for the tracer solution to bypass such zones. Probes such as MIP offer potential to locate high-concentration zones of chlorinated solvent contamination, but quantification to the degree necessary to differentiate highly sorptive zones from DNAPL, residual from free-product DNAPL, or the vertical separation of thin layers of DNAPL is unlikely. Both PITT and MIPs offer possibilities for contributing insight to the nature of DNAPL zones; however, they should be applied in conjunction with the techniques based on continuous cores, with detailed subsampling at the scale commensurate with subtle textural variability, such as those described here.

The history of industrial operations at each of the five sites indicates that these DNAPL source zones have had a long period of aging, caused by three to five decades of flushing by natural groundwater flow in the horizontal direction. The Darcy flux through the source zones at most of the sites is in the range of 5 to 10 cm d^-1, except for the Ontario site where it is lower. It is reasonable to expect that the vertical pipes or fingers of residual DNAPL with initially low DNAPL saturation, expected to range from about 2 to 15% in granular saturated media (Feenstra et al., 1996), that initially connected the DNAPL layers (Fig. 1), have disappeared due to dissolution. Therefore, the contaminant discharge from such aged source zones is much lower today than in previous decades. For hypothetical cases, Anderson et al. (1992) and Sale and McWhorter (2001) showed, using mathematical models, the rapid relative rate of finger and pipe removal due to natural groundwater flow. However, at the five study sites, it is also likely that the decades of groundwater flow have removed many DNAPL layers, in addition to the pipes and fingers, and significantly reduced the DNAPL thickness and saturation of other layers. This expectation is based on the occurrence of DNAPL layers only in finer-grained aquifer units, or sandwiched between silty or clayey layers in the transition zones at some of the sites. Initially, most DNAPL probably occurred in coarser-grained zones, as was observed in the PCE DNAPL release experiments in the Borden sand aquifer (Kueper et al., 1993; Brewster et al., 1995). The more rapid groundwater flow in these coarser-grained zones causes preferential removal of DNAPL from these zones, as illustrated in Fig. 8. Therefore, the present-day DNAPL occurrences represent only the less flushable remnants of the original DNAPL zones. In the case of the Connecticut A site, where DNAPL generally resides at the bottom of a uniform and very permeable aquifer, contact with flowing groundwater would have been limited within the DNAPL zones because of reduced permeability, with most contact and mass transfer occurring at the top of the DNAPL layers residing on the aquitard interface. It is also evident at this site, where local depressions in the aquifer–aquitard interface are present and initial DNAPL accumulations were thicker (Fig. 3c), that variable rates of groundwater flushing over a few decades has created suspended layers of free product separated by zones completely depleted of immiscible phase, even though contrasts in aquifer permeability are subtle. Although this discussion focuses only on the DNAPL source zones at the five study sites, the plumes emanating from these source zones were also studied in spatial detail using depth-discrete monitoring along transects perpendicular to groundwater flow, and consistency was found between the source zone characteristics and the down gradient plumes.

View larger version (38K): [in this window] [in a new window]   Fig. 8. Illustration of the evolution of a layered DNAPL source zone showing complete dissolution of residual trails, shrinkage of some layers, and complete removal of others due to decades of groundwater flushing.