Water Stewardship

Ground Water Resources of British Columbia

Chapter 13 — Case Histories

13.8 MONITORING, MIGRATION AND CONTROL OF AN ETHYLENE DICHLORIDE CONTAMINANT PLUME IN A GRAVEL AQUIFER

SYNOPSIS

On February 15, 1986, derailment of a Canadian National Rail freight train near Fort Langley, B.C. resulted in a 247,500 litre spill of ethylene dichloride (EDC). Within a few hours, the spilled chemical had seeped into the ground.

Piteau Associates were retained to supervise a drilling program that was carried out to delineate the extent of the contamination and to establish a monitoring network. A monitoring program was instituted to determine temporal changes and to determine the direction and velocity of contaminant migration. Initial monitoring indicated that a plume in a gravel aquifer was travelling towards the Fraser River, about 350 m away, at a maximum rate of about 2.5 m/day.

Many plastics, (such as PVC) which are normally used for ground water monitoring, could not be used due to the corrosive nature of the EDC. Readily available steel and non-corrosive plastic materials were used to avoid delays in implementing a monitoring system.

The hydraulics of the aquifer were analyzed, and the pumping rate required to control migration of the plume was estimated. Subsequent to this study, a recovery program was instituted by others, and recovery of the EDC is still ongoing.

INTRODUCTION

On the afternoon of February 15, 1986, a Canadian National freight train derailed at a point approximately 2 km east of Fort Langley, B.C. (see Figure 13.5). The derailment site is in a rural area, approximately 350 m from the Fraser River.

Figure 13.5 Plan showing location of EDC spill site

Ten tank cars carrying liquid chemicals were among those derailed. Eight of these cars contained ethylene dichloride (EDC) and two contained sodium hydroxide. Some of the derailed cars remained intact, however, an estimated 247,500 litres of EDC and 60,000 litres of sodium hydroxide spilled or leaked from damaged cars into a ditch, and within a few hours had seeped into the soil.

PHYSICAL PROPERTIES OF SPILLED CHEMICALS

Ethylene dichloride (or 1, 2-dichloroethane) is used in the manufacture of vinyl chloride monomer, in antiknock compounds for gasoline and in the processing of adhesives and coatings. It is a heavy, colourless and flammable liquid, with a moderately low viscosity, moderate solubility in water, a pleasant odour and a sweet taste. The ambient ground water temperature is 10° C at the Fort Langley site. At this temperature, the solubility in water and viscosity of EDC are 8,500 mg/L and 1.267, respectively (see Figure 13.6).

Figure 13.6 Physical properties of ethylene dichloride

EDC will evaporate quite rapidly from the surface of streams, especially under turbulent conditions. It is readily broken down in the lower atmosphere, ultimately to hydrogen chloride, carbon dioxide and water. The estimated half life of EDC in the atmosphere is about one month (Dow Canada, 1986). In the ground, it is very slowly broken down by bacterial action (Dow Canada, 1986).

Ethylene dichloride is slightly toxic to fish, the static acute LC50 for rainbow trout ranging between 155 and 175 mg/L (Watts, 1982). Although fish eggs are fairly resistant to EDC, newly hatched fish are very sensitive. For example, the no-effect level for fathead minnows is 20 mg/L (Dow Canada, 1986).

Based on its relatively low octanol/water partition coefficient (Log Kow 1.48, Jackson et al 1984), EDC is not expected to bioaccumulate to any significant extent (Dow Canada, 1986).

Sodium hydroxide is one of the strongest alkalies known. It is readily soluble in water and is usually transported as a 50% solution in tankers. Below 30° C, a concentrated solution becomes very viscous, and at temperatures below 12° C, it is essentially frozen. At Fort Langley this solution was virtually immobile once it had seeped into the ground, and the majority of it was recovered with simple ditching and a sump pump. No anomalous pH vales were recorded for ground water sampled from the aquifer beneath the spill site. Thus, the unrecovered sodium hydroxide had very little apparent impact on ground water flow under the site.

DESCRIPTION OF SPILL SITE

The area in which the spill occurred is low, undulating land used primarily for pasture. The actual spill site was the edge of an aero club landing strip (see Figure 13.7). Prior to construction of the landing strip, the area between the Canadian National tracks and the Fraser River had been raised with wood waste fill to provide a level surface free from flooding.

Average annual precipitation at the spill site is about 1740 mm, with December typically being the wettest month. Average annual air temperature is 9.8 ° C, ranging from mean monthly values of 1.8° C in January to 17.5° C in July.

Natural drainage in the area south of the Canadian National track is not well developed, but a network of ditches which parallels the railway and some of the main roads serves to drain the area fairly well. One of these ditches runs westward, from the site of the spill along the north side of the CN track (Figure 13.7).

Figure 13.7 Plan of EDC spill site

GEOLOGY

The study area is underlain by more than 100 m of unconsolidated sediments, of which only the near surface material will be discussed herein. Fraser River sediments, typically silty, fine to medium sand, are exposed on surface at lower elevations (less than 8 m-asl) in the study area. In the lowland area south of the derailment site, these sediments are overlain by peat, but generally, wherever these sediments are present, they are exposed on surface (Halstead, 1986).

In the general area of the spill, Fraser River sediments are known to overlie Sumas Drift sand and gravel, Fort Langley glacial, deltaic and glacial marine sediments, and Capilano marine and glaciomarine sediments.

The sequence of surficial sediments at the spill site is shown in the accompanying section (Figure 13.8). From ground surface, the sequence consists of 0.5 to 1.0 m of silty find sand surface cover and 4.5 to 7 m of wood waste, over 3 to 5.5 m of silty sand. The silty sand is underlain by 3 to 9 m of sand and gravel, which is in turn underlain by clayey silt. None of the monitoring holes drilled for the study penetrated more than about 1.2 m into the lowest unit. However, based on available information from holes drilled in the surrounding area, this unit is likely to be in excess of 100 m thick.

Figure 13.8 Regional hydrogeological section

Early in the hydrogeological field investigation, it became apparent that the wood waste had been backfilled onto an irregular surface, hence a varying thickness of wood waste was encountered. A set of pre-fill air photographs were located and a topographic plan of the pre-fill ground surface was prepared. This plan indicated the presence of a shallow channel or slough, parallel to and immediately north of the Canadian National track. As is discussed later, this channel had a significant effect on the direction of migration of EDC from the spill site.

FIELD INVESTIGATION

Immediately following the spill, some test pits were dug in an attempt to located the EDC which had leaked into the ground. Although these test pits were dug between four and five metres into the wood waste, no sign of EDC was found. It was at this point hat a decision was made to carry out a hydrogeological investigation to locate the spilled chemical.

DRILLING AND PIEZOMETER INSTALLATION

The initial hydrogeological field investigation was designed to delineate the local geology, the local hydraulic gradients and the extent of contamination. Between February 19 and 25, 1986, a total of 12 test holes were drilled around the spill site and piezometers were installed. Access to sites near the damaged tank cars was somewhat restricted, due to activities of work crews who were clearing the cars and repairing the railway track.

Due to the presence of both gravel and wood waste at the site, and the need to implement the monitoring quickly, a Becker Hammer rig was used to drill the holes. Penetration rates were fast and the reverse air flush method allowed continuous sampling of sediments and formation water during drilling. A photograph of the rigs set up near the derailed tank cars is shown in Figure 13.9.

All site staff were provided with gas masks, protective clothing and instruction on safety. While the fumes issuing from some drillholes were highly concentrated, explosive mixtures in the air rarely occurred. Only one explosion was reported where an acetelyene torch ignited EDC vapour in a dewatering well.

Samples of sediments from above the water table were collected every 1.2 m, placed in plastic bottles and analyzed with an organic vapour analyzer (OVA) which could detect EDC vapour down to a concentration of about 10 ppm. Samples from below the water table were collected every 1.2 m for analysis with the OVA to detect presence of EDC, and with a pH meter to detect presence of sodium hydroxide. Water samples were also collected, and stored in glass bottles with sealed teflon caps for analysis with a gas chromatograph (GC). The GC was on loan from DOW Chemicals Ltd. and was set up for convenience, and a fast turn around time, at the nearby Fort Langley firehall.

Selection of materials for piezometer construction posed a problem, as PVC is dissolved when in contact with high concentrations of EDC. The only readily available alternative to PVC was a combination of black steel pipe and stainless steel well points. In the initial drilling program, 7 of the holes were completed with 31 mm diameter, steel piezometers, and the 5 holes up gradient or most distant from the spill site were constructed of 37 mm, flush coupled, threaded PVC.

Figure 13.9 Reverse circulation Becker Hammer Drill set up on test hole near derailed tank cars	Figure 13.10  Bundle type          multiple point piezometer

All piezometers were steam cleaned prior to installation. They were installed through the Becker Hammer drill steel at the bottom of each hole. A clean Monterey sand was used as a pack around the piezometer tip. The drill steel was then withdrawn to the base of the silty sand, and the sand and gravel below this point was allowed to cave. Zones that did not cave were backfilled with Monterey sand. In the first few holes, attempts were made to place bentonite seals in the gravel, but this turned out to be a very time consuming and unnecessary exercise, as the sand and gravel was very permeable (hydraulic conductivity of 10-2 m/s). Bentonite seals, generally between 2 and 5 m in length, were placed at the base of the silty sand layer.

Twenty eight well points, constructed of one 6.1 m length of 31 mm steel pipe with a stainless steel well point on the end, were also installed. The installation procedure involved digging a hole with a backhoe to a depth of about 4 m, placing the well point assembly in the hole and then backfilling around the pipe. Once the hole was backfilled, there was sufficient lateral support on the pipe to allow it to be pushed, without buckling, to full depth or refusal with the bucket of the backhoe. Depths varied between 4 and 6.5 m.

Following completion of the first phase of drilling and installation of well points, an initial assessment of the extent and migration rate of the contaminant plumes in both the wood waste and sand and gravel units was made. A second phase of drilling was then carried out to further delineate the plumes.

All piezometers installed in the second phase drilling program were of the bundle type, as illustrated in Figure 13.10. A centre steel piezometer, constructed from 12 mm diameter (except TH-20, 31 mm diameter) steel pipe with a slotted bottom was used for the deep piezometer in each hole. Polyethylene tubes, with a 9 mm I.D. and a 150 mm long stainless steel woven screen tip, were taped around the steel pipe before it was lowered into the hole. The stainless steel tips were placed at different locations along the pipe so that profile sampling could be carried out. While pulling the drill rods out of the hole, small amounts of bentonite slurry were pumped into the hole in the interval between each piezometer tip, to provide a seal against vertical migration of ground water in the hole annulus. Uniform sand was placed in the intervals where the piezometer tips were located.

Following installation, all piezometers and driven well points were developed by either blowing water out of the piezometers with a compressor, or by pumping water out of the piezometers with a peristaltic pump.

SAMPLING

As piezometric levels were typically within 6 m of ground surface, it was possible to use a peristaltic pump to withdraw samples. The samples were drawn through dedicated sampling tubes, which consisted of 3 mm diameter polyethylene tubes, with a short length of C-Flex tubing on the upper end which was compatible with the peristatic pump head. The tubing materials were selected after reviewing manufacturers recommendations and testing selected tubing at the site. Both polyethylene and C-Flex were not noticeably affected by the EDC, although after repeated sampling of highly concentrated EDC, the C-Flex tubing did rupture on two occasions, and was simply replaced each time.

Some purging of each piezometer tube was carried out before water samples were obtained. Extensive purging was not warranted, as the sampling tubes were of small diameter, were located in the piezometer tips, and were pumped at rates which do not induce any significant drawdown in the piezometers. Thus, minimal pumping time was required to get new formation water into the piezometers and up through the 3 mm diameter sampling tubes. It was found that prolonged pumping tended to draw in water preferentially to non-aqueous phase EDC.

Sampling was carried out on a schedule which became less frequent as time progressed. During the drilling program, when there were only a few monitoring points, samples were collected almost daily, and were delivered to the temporary field laboratory twice a day. This was done so that results of samples obtained from the drilling program would be available about 4 hours after drilling. Once the drilling program was completed, an entire suite of sampling required about 2 1/2 man days. Starting at about one month after the spill, sampling frequency was reduced to twice per week, and six weeks after the spill it was reduced to once per week. Contour plans showing EDC concentrations were prepared for submission to the Waste Management Division of B.C. Environment, on a weekly basis, starting about 6 weeks after the spill.

The reasons for reducing the monitoring frequency were three-fold:

1. Migration rate and direction of movement became reasonably predictable;
2. The data generated by the frequent sampling of approximately 100 monitoring points was expensive to collect and evaluate quickly and was not considered worthwhile;
3. The GC analyzer in the field laboratory was returned to DOW Chemicals Ltd. six weeks after the spill occurred. This resulted in an increase in the costs and time required to analyze samples.

Weekly sampling of surface water for EDC analysis and Bioassay was also carried out at a number of points along the Fraser River bank.

HYDRAULIC CONDUCTIVITY TESTING

Hydraulic conductivity of the sand and gravel unit was estimated by back analysis of an aquifer test run on a nearby production well, and by applying the Hazen formula to grain size analyses. No aquifer pump tests were run at the site. Based on these data analyses, the estimated hydraulic conductivity of the sand and gravel was about 10^-2m/s.

Rising head tests performed on well points in the silty sand, and the results of grain size analyses, indicated the hydraulic conductivity of the silty sand ranged from about 10-6 to 10-4 m/s. A short pump test performed in a shallow recovery well in the wood waste indicated that the hydraulic conductivity of this fill was about 10^-3m/s.

HYDROGEOLOGY

The principal hydrogeologic units at the site, starting with the upper-most, are:

Natural ground water flow at the site is comprised of two main components:

1. Vertical flow from the wood waste and silty sand down into the sand and gravel (see Figure 13.8);
2. Horizontal flow in the sand and gravel aquifer towards the northwest (see Figure 13.8).

The vertical flow between the wood waste and the aquifer is a function of hydraulic head differential. The hydraulic head differential is a function of the quantity of precipitation which infiltrates to recharge the water table in the wood waste, and the piezometric head in the aquifer. When the spill occurred in mid February, the ground water levels were near the annual high, following a period of heavy precipitation. Water levels in the wood waste peaked on February 25, 1986, and continually declined after that date. Water levels in the aquifer and in the silty sand aquiclude generally declined from February 25, but fluctuated a great deal due to tidal activity in the Fraser River. The relationship between the water levels measured in the aquifer, the silty sand, the wood waste and the Fraser River, for a three day period in March, 1986, is shown in Figure 13.11. As illustrated in this figure, there is a varying but persistent downward hydraulic gradient between the wood waste and the sand and gravel aquifer.

Flow from the wood waste into the aquifer, over the spill area, was estimated to be 0.54 L/s. The natural flow in the aquifer directly under the spill site was estimated at 1.1 L/s, therefore dilution was expected to be minimal (Piteau Associates, April, 1986).

Ground water flow in the aquifer is generally in the northwest direction, although for brief periods during flooding of the Fraser River ground water flow may be to the southwest. The average velocity of ground water flow in the sand and gravel aquifer was estimated to range between 0.7 and 1.3 m/day (Piteau Associates, April, 1986 and J. F. Sykes, June, 1986).

POTENTIAL IMPACT OF EDC

There were two major concerns regarding the impact of the EDC on the local environment:

1. EDC contaminated ground water would discharge into the Fraser River, producing a contaminant plume of sufficient concentration to harm migrating salmon or other indigenous aquatic species.
2. The EDC contaminant plume could reach domestic wells located on McMillan Island (Figure 13.8) or a municipal well field located about 7 km southwest of the site.

Neither of the potential impacts above were very immediate, as travel distances to the Fraser River and the nearest domestic well were 350 m and 2000 m, respectively. However, the potential long term threat to aquatic life and human health required that the EDC eventually be recovered.

CONTAMINANT MIGRATION

General

There is currently very little information available which relates specifically to EDC migration into the ground, other than a few case histories dealing with the migration of other dense, immiscible chemicals in ground water. some laboratory work has been carried out in Germany (Schwille, 1981) to investigate the movement of dense, non-aqueous phase liquid (DNAPL) chemical in saturated, porous media. Based on this work, and on published observations at a number of contamination sites in North America, Stan Feenstra (1986) described three general concepts for the development of contaminated zones in sandy aquifers by DNAPL chemical. The three general concepts are illustrated in Figure 13.12. All concepts consider two phase movement of the chemical (i.e. both movement of the pure, heavier than water liquid chemical (non-aqueous phase), and movement of dissolved chemical in water).

If the quantity of DNAPL chemical spilled is small, it is likely to penetrate into the unsaturated zone where some residual amounts will be retained in the pore spaces by capillary tension. The limited data available indicate that these residual contents of DNAPL chemical in sandy soils are generally equivalent to 10% to 30% of the pore space (Feenstra, 1986). Rain which infilitrates and then percolates down through the soil pores in which the residual DNAPL chemical is retained will dissolved some of the contaminant and carry it down to the water table where a contaminant plume will originate. This concept is illustrated in Figure 13.12a.

If the quantity of spilled DNAPL chemical is larger, it will exceed the retention capacity of the unsaturated soil. Thus, some of the non-aqueous phase chemical will reach the water table in the aquifer where it will remain as a residual in pore spaces in the aquifer until it is dissolved in ground water flowing under the site, thus initiating a contaminant plume. This concept is illustrated in Figure 13.12b.

The third concept, illustrated in Figure 13.12c, is very similar to the observed distribution of EDC at the Fort Langley spill site. This situation occurs when a spilled volume of DNAPL chemical exceeds the retention capacity of both the unsaturated sediments above the water table and the saturated aquifer material. A portion of the non-aqueous phase chemical will eventually settle on the bottom of the aquifer to form a pool. If the bottom of the aquifer is sloped, the DNAPL can then migrate in the downslope direction. In some circumstances this could even occur in the opposite direction to ground water flow. While a contaminant plume would develop over the entire thickness of the aquifer (see Figure 13.12c), the contaminant concentration a short distance from the spill site would tend to be greatest near the base of the aquifer.

Once in the aquifer, DNAPL chemicals migrate in two different phase: (1) dissolved in water (aqueous) and (2) liquid (non-aqueous phase). Movement of dissolved DNAPL is primarily a function of advection and dispersion. The DNAPL could also migrate through the aquifer, usually at a slower rate than the aqueous plume. It is expected that EDC, which has a viscosity similar to water, would flow readily through the aquifer in a non-aqueous phase. The rate of movement of any non-aqueous phase EDC would depend on many factors, such as surface tension, vapour pressure, relative densities, percentage of EDC in the pore spaces, etc.

Figure 13.11 Hydrograph showing relationship between Fraser River level and piezometer levels under the spill site	Figure 13.12  Three concepts    for development of DNAPL    contaminant plumes

Drilling with the Becker Hammer started four days after the spill occurred. Twelve test holes and nineteen well points were installed over a six day period. Based on monitoring results from these holes, it became apparent that a large quantity of EDC was ponding in the wood waste which had been placed on the surface of the silty sand sediments. Initially, high concentrations of EDC were present in the base of the wood waste along the trend of an old buried slough.

The second phase of drilling, which was initiated about nine days later (19 days from the time of the spill), augmented the monitoring network both within the immediate spill area, and in the area northwest and down gradient of the spill site. More well points were also installed along the trend of the old slough to delineate migration of EDC in the wood waste.

A series of plans and sections are presented in Figures 13.13 and 13.14, showing contours of EDC concentration at various times after the spill occurred. The distribution of the EDC in the wood waste and upper silty sand, approximately four weeks after the spill, is shown in Figure 13.13a. As evident in this figure, initial migration of the EDC was in the east-west direction, along the orientation of the old slough. The high concentration plume actually extends a few metres south of this slough, but this was likely due to the ditch which parallels the Canadian National tracks, which would have concentrated infiltration of the spilled chemical a few metres south of the old slough.

Within the area of high concentration, some samples of pure or very highly concentrated EDC were obtained, indicating that in some areas, pods of EDC were accumulating at the base of the wood waste. However, as illustrated in monitoring data plotted in Figure 13.15, these pods existed for only relatively short periods. After a few weeks the plume in the wood waste and silty sand expanded slightly to the east, but the major changes in the shape of the plume included pronounced migration to the northwest of a low concentration front, represented by the 5 mg/L contour, and a slight advanced of the same contour to the south. The area in which EDC must still have existed as two phases (roughly the area inside the 10,000 mg/L contours) continued to contract throughout the monitoring period. Eleven weeks after the spill, concentrations greater than the solubility limit for EDC had all but disappeared in samples from the wood waste (Figure 13.13).

Figure 13.13 Distribution of EDC concentrations at
4 and 11 weeks after the spill

The vertical migration of the EDC is illustrated in Figure 13.14. Monitoring data for the sand and gravel aquifer was not available for the spill area until week four. The first available profile (week four) showing the vertical distribution of EDC concentration under the site indicates extremely high concentrations of EDC in the wood waste in the base of the old slough, and at the base of the aquifer. Concentrations of EDC at the base of the aquifer almost directly below the spill remained greater than 70% from week four, when the first sample was obtained, until the end of week seven (see Figure 13.15). Concentrations fell rapidly to less than 1% by the end of week nine, possibly due to the installation of a low capacity recovery well (10 L/min) nearby. The series of three profiles in Figure 13.14, which show the distribution of EDC concentrations at weeks four, seven and eleven, all show the same pattern of highest concentrations at the base of the wood waste and the base of the aquifer immediately under the spill site, with a plume of lower concentrations migrating to the northwest (left end of section). As illustrated in Figure 13.14, the highest concentrations in the expanding plume were generally near the base of the two aquifers. EDC concentrations in the overlying silty sand were much lower and less evenly distributed, possibly due to the presence of permeable bedding planes in the sand which would allow relatively high lateral velocities of ground water flow.

Horizontal migration of EDC in the aquifer is depicted in Figure 13.13b. By week four, a zone of very high EDC concentrations existed immediately under the spill site, and migration was predominately to the northwest. Lower concentrations (represented by the 5 mg/L contour) migrated in all directions, but primarily to the west and north. Dispersion was probably the mechanism which controlled the advance of this low concentration front, as there was significant migration in the southeast direction, which was against the natural gradient of ground water flow through the area.

Between weeks four and eleven, the contaminant plume spread out mainly towards the north and northwest, with little or no up gradient migration. The 5 mg/L contour contracted slightly on the west side of the plume, apparently due to the flushing action of the natural ground water flow. Concentrations of EDC in the plume were also much reduced, due primarily to the dilution by natural ground water flow. Although sampled concentrations were generally below the solubility limit, it was not conclusive evidence that very little EDC remained in the non aqueous phase. General experience on site was that new holes would often encounter high concentrations of non-aqueous phase EDC, but that after repeated sampling, concentrations would drop off, due to lack of mobility of the DNAPL in the formation around the monitoring point. A short distance away from the monitoring points, there was likely to be non aqueous phase EDC retained in the pore spaces.

Figure 13.14 EDC concentrations along Section B-B through spill site

Figure 13.15 Time base plotting of EDC concentrations directly under spill site

Based on the movement of EDC concentrations through the aquifer, the advective rate of transport of the plume was estimated to be about 1m/day. The low concentration front (5 mg/L), which as apparently controlled by dispersion and advection, was advancing at a rate of about 2.5 m/day. Calculations based on these rates indicated that it would take about 2 years before the low concentration front would reach McMillan Island, where the closest domestic wells were located.

SIMULATION OF PLUME MIGRATION

Preliminary computer modelling was carried out (Sykes June, 1986) to simulate the probable impact of the spill upon the local ground water resource, and to predict the time required for natural flow of ground water through the area to remove the EDC. Two models were used in the analyses. A two-dimensional model was used to simulate the movement of dissolved and non-aqueous phase EDC, and the dissolution rate of the non-aqueous phase EDC. A three dimensional model was used to model the ground water flow domain.

The results of the 3-D hydraulic modelling indicated that the estimated transport velocities were reasonable, and that the migration of the plume in the sand and gravel aquifer could be controlled by pumping. The results of the 2-D transport modelling indicated that flushing of non-aqueous phase EDC from the wood waste and silty sand would require about 10 years, unless volatization or recovery methods were instituted. Approximately 500 days would be required for the non-aqueous phase EDC in the sand and gravel aquifer to be entirely dissolved by natural ground water flow through the area. Based on the modelling work, it was recommended that pumping of the aquifer be continued until all EDC at the site had been dissolved, and acceptable concentrations were reached. It was also concluded that unless measures taken to remove EDC from the wood waste were successfully implemented, prolonged pumping of the aquifer would be required.

It was estimated that a pumping rate of between 4 and 6 L/s was required to control migration of the contaminant plume, as it existed in mid April, 1986.

MITIGATIVE MEASURES

Canadian National Rail and their consultants have been handling the spill clean up work and aquifer decontamination, and Piteau Associates' involvement with the project ceased as of the end of April, 1986. We understand that the decontamination measures implemented at the site include:

1. Pumping the aquifer and air stripping the discharge water and / or decanting concentrated EDC from the bottom of large temporary storage tanks;
2. Pumping with low capacity, positive displacement pumps installed at the base of the aquifer in small ID wells which were drilled into pools of undissolved EDC (typically at concentrations of > 10%). The usual trend with these pumps was that high concentrations of EDC would be pumped for periods of a few hours to a few days, but then concentrations would quickly drop to levels below the solubility limit;
3. An air injection / venting system in the wood waste to promote volatization of the EDC and to provide lateral containment.

CONCLUSIONS AND RECOMMENDATIONS

The movement of the DNAPL-EDC at the Fort Langley spill site was very similar to that which would have been expected, based on the generally accepted concepts for development of DNAPL contaminant plumes in saturated, granular aquifers. Due to density effects, vertical migration of the EDC was very prevalent at the spill site. Lateral migration of EDC in the unconfined wood waste aquifer unit was very limited. In the granular aquifer unit, located beneath the wood waste and the silty sand units, lateral migration was very significant due to the high permeability and significant natural ground water flow. The density effect resulted in the highest EDC concentrations occurring near the base of this aquifer. While recovery of the spilled EDC is possible, it will be a very slow and expensive process. This is prolonged by the difficulty of removing residual DNAPL from pore spaces into which it migrates.

Based on the experience at the Fort Langley spill site, the following two points should be emphasized, which may be useful in the event of other, similar spills:

1. The importance of having a team of specialists capable of implementing a hydrogeological investigation, monitoring, and recovery program at the site is very important. This team must include personnel familiar with the various methods of on site treatment, and disposal of contaminated wastes so that the appropriate systems can be installed soon after sufficient information is available to design a recovery program.
2. If the initial investigation indicates that a large proportion of the contaminant is present at relatively shallow depth, serious consideration should be given to excavating contaminated sediments. As initial migration will have a significant vertical component, excavating directly under the spill site should provide positive results.

ACKNOWLEDGEMENTS

We gratefully acknowledge Canadian National Rail for allowing us to present this summary of the spill, and Grant Potoliki of Canadian National Rail for providing us with invaluable assistance throughout the drilling and monitoring program.

Back to Table of Contents