SUMMARY: In 1972 in the town of Woburn, Massachusetts, families of 13 different children who had contracted a rare form of childhood leukemia sued two large companies for the contamination of their community water supply wells. But the companies denied any responsibility, arguing that water originating from their plants may never be able to reach the community wells given the local hydrogeologic conditions. Instead, the companies blamed the heavily polluted Aberjona River as the source of contamination to the wells, given it's very close proximity to the community wells. You be the judge... who is responsible?!? Revisit the famous trial as modern-day ‘expert consultants’, addressing the debate issues based on the interpretation of groundwater flow and contaminant transport simulations.

Figure 1: Woburn Superfund site. Right: Location of the known VOC sources during the trial. Adapted from: https://serc.carleton.edu/woburn/resources/Woburn_maps.html

Background & Arguments

In 1972 in the town of Woburn, Massachusetts, 13 children living in the same neighborhood contracted a rare form of childhood cancer (acute lymphocytic leukemia). The families wondered: is there something wrong with the water we drink or the air we breathe? That seemed to be the only thing they shared in common. The families complained to the local environmental health department, requesting that their community well water be tested. But nobody really listened. The response was that the water quality was "perfectly fine" based on their past well tests.

Until one day … a developer who was excavating a parcel of vacant land nearby accidentally discovered numerous 55-gallon drums containing toxic chemicals, causing significant subsurface contamination. The story immediately made the news the next day in the local paper. The shocked citizens thought “no wonder!” this place is so unfortunate! The inflicted families immediately hired a law firm and experts to get to the bottom of it. They identified a number of potential responsible parties, the biggest being the WR Grace company, a chemical manufacturing giant located less than half a mile up hill from the community wells, and Beatrice Food Inc., owners of the former J.J. Riley tannery properties.

A nasty legal fight ensued – one of the first major class-action, environmental lawsuits in US history. The citizens accused WR Grace of polluting their exclusive source of drinking water, because the chemical plant is uphill from their wells and groundwater moves from the plant towards their wells located in the valley below. They pointed out that the company uses the same chemicals that were found their wells. But the both the chemical company and the food processing plant denied any responsibility.

The argument from WR Grace:

It couldn’t possibly be us. The chemical plant had only been in operation since 1964. Groundwater moves very slowly and the soils underneath the plant are very tight – they don’t drain well. So, it may take decades before the groundwater originating from here to reach the valley. Even if there was an ‘infinite’ amount of time available, the hydrology is such that the water originating from our plant may never be able to reach both wells. And if you look around, there are other industries or places nearby that could be responsible:  Unifirst Corporation, a dry-cleaning facility; and the Olympia and Wildwood properties, formerly part of a tannery operation, which are much closer to the wells than our chemical plant is. Plus, Woburn has a legacy of industrial production since its founding in 1642. The Aberjona river was so polluted, collecting pollution effectively from everywhere - the entire watershed - and you were pumping right next to the heavily polluted river!

The argument from Beatrice Foods Inc.:

Although our properties are relatively close to your wells, we never used the chemicals found in the drinking wells in our operations. Most importantly, we are located on the other side of the Aberjona river, so we are hydraulically separated from the community wells – the Aberjona River is a hydraulic divide!  Water originating from our premises can never reach the other side of the river. That is just common sense and basic hydrogeology!

The plaintiff families countered and insisted that both WR Grace and Beatrice foods were contaminating their wells. To WR Grace they responded:

Our experts calculated that it only took months or – at most – a few years for the contaminated groundwater to flow from WR Grace to the community wells. You have been operating since 1964!

To Beatrice foods they responded:

Your property was formerly a tannery that used the same toxic chemicals found in the community wells. You inherited the liability when you purchased the property. And according to our experts, the river is a hydraulic divide only if the river is well connected to the aquifer. But the bottom of the Aberjona River is so full of fine sediments, muds and clays … that it is almost impervious, effectively sealing the river from the aquifer. The water pumped was not from the river!

Objective & Deliverable

In this project, you will be the judge... who is responsible?!?

Revisit the famous trial as modern-day ‘expert consultants’, addressing the debate issues based on the interpretation of groundwater flow and contaminant transport simulations. The key scientific issues to address include:

• Where does the pumped water come from?
• What is the area of contribution to wells G and H?
• Where do the spills from the various potential responsible parties go? - WR GRACE, the Beatrice Foods property, the dry cleaner, etc. - can they reach the wells?
• How long will it take for the spills to reach the wells from the various sources?

Prepare a 2-3 page report that summarizes your approach and findings. You should discuss your findings with regards to responsibility for the contamination. Include any detailed model results / graphics in support of your conclusions in an appendix. 

Specific steps required for analysis:

• Model Set up
• Model Calibration
• Model Application (Particle Tracking)  

You are provided with the following information collected that was collected during and after the lawsuit. 

Site Hydrogeology

Wells G and H are located east-central Woburn, MA - about 10 miles (16 km) northwest of Boston - in a gently sloped valley with a large stream, the Aberjona River, flowing through it from roughly north to south. A number of small tributaries contribute to upstream flow of the Aberjona River. Flanking either side of the Aberjona River are a collection of wetlands, throughout much of which peat deposits occur near the surface. Much of the land outside of the wetland areas is developed, an intermix of industrial, commercial and residential properties.

The river and its surrounding lowlands are underlain by 0.5- to 1.0 mile-wide stratified (layered) glacial deposits primarily consisting of sand and gravel, but also clays and silts.  These deposits fill a basin-shaped, buried bedrock channel (see below).  The uplands are underlain by thin layers of till over low permeability bedrock, and thus, negligible flow in this area is expected.

Recharge to the stratified drift is from precipitation. 

Figure 2: Conceptual 3D view of the groundwater at the Superfund site.

Field Data / Information

The top boundary of surficial/unconfined aquifers follows the land surface, which can be represented with detailed 10m Digital Elevation Models (DEMs).

The depth to the bedrock surface ranges from close to zero to 240 ft near the central part of the valley. Wells completed in the bedrock typically yield relatively small amounts of water (e.g., a few GPM). 

The lithology of the stratified drift can be separated generally into four layers on the basis of the principal lithology in each layer:

• In the deep, central part of the aquifer, coarse sands of about 15m thickness sit on top of fine-grained sands and silts.  
• Above the layer of coarse sands and gravel is a layer of fine-to-coarse sand with a thickness of about 20m
• The uppermost layer, occupying the top 27m (or so) of the aquifer is a mix of sand, silt, clay, and – where the wetlands are present – deposits of peat.

Long-term average recharge to the stratified drift aquifer is unknown, but long-term average recharge in this part of the country varies between 16 and 28 in./yr. 

Wells G and H (see below) are screened in the drift aquifer. Until they were closed down in 1979, they pumped at an average rate of 700GPM (well G) and 400 GPM (well H).

Two industrial wells southwest of the Aberjona River have long extracted groundwater from the drift aquifer to support the tannery operations associated with John J. Riley Tannery. The first well (industrial well 1) began operating in 1954 while the 2^nd well (industrial well 2) began operating in 1958. Industrial wells 1 and 2 were estimated to operate at 70 GPM and 200 GPM, respectively.

Also available are  measured Aberjona River flows and observed heads collected around the site during a real-world investigation first published in the USGS Water Resource Investigations report by de Lima and Olimpio (1989). This includes scattered head measurements collected prior to and after a 30 day "pumping test" (wells G and H pumping at 'operational' rates). 

Figure 3: Study area extent, with locations of Wells G and H and the industrial wells shown. Also shown are the approximate wetland areas.  This image is available for overlay in MAGNET (see top of problem posting). 

Model Setup:

Figure 4: Conceptual model to follow during model setup. 

Steps for model setup:

• Navigate to Woburn, Massachusetts using the navigation tools near the top of the MAGNET Modeling Environment.
• Conceptualize the model with 3 aquifer layers – one for each of the glacial drift layers described above. The approximate extents of each of the three layers is shown in Figure 5 below.
• First, overlay the site map and delineate the model domain for the first layer. The extent of the first layer follows the boundary between the uplands and the glacial drift valley. The top of the 1st layer is the land surface (DEM), and the bottom is the bedrock top surface.  (The shape of the bedrock surface is available from the the global bottom elevation data layer available on the MAGNET server.)
  •  A georeferenced image file of the site map is available in the problem posting (file available at top of page). Access the overlay tool with: ‘Other Tools’ > ‘Utilities’ > ‘Overlay myImage’
  • Also included is the ‘base_map_wetlands_extent.txt’ file which includes the image spatial extents needed for overlaying in the modeling environment.
• Then, add two (2) more geologic layers of appropriate thickness ('Conceptual Model Tools' > 'Layer' > 'Add a New Layer'). Use 'Zone' features to map the spatial extent of each layer.
  •  In Zone attributes menu:  'Flow properties' tab > 'Zone Type' > 'Active'
  • Then in Domain Attributes menu: 'Aquifer Attributes' tab > check box next to 'Domain as an Inactive Zone'
• Assign an initial estimate of  hydraulic conductivity for each layers based on dominant lithology (see above). These values will be 'fine-tuned' during calibration.
• Similarly, assign an initial value of recharge which will later be calibrated. 
• Assume a porosity of 0.2. a specific yield of 0.25, and a specific storage of 5x10^-3 m^-1 throughout the aquifer system.
• Conceptualize the Aberjona River as a head-dependent boundary condition in the 1st layer, where the stage follows the aquifer top (DEM from land surface). The river leakance is unknown and needs to be fine-tuned during calibration. You may assume an average river depth  of 1.7m
• Treat the wetland areas as a one-way drain, allowing water to leave the aquifer in places where the head exceeds the land surface elevation.
  • Typically, a leakance of 1 d^-1 is appropriate (some fine-tuning during calibration is expected).
  • The approximate extents of the wetland areas is shown in the site map (Figure 3).
• Add the industrial wells and Wells G and H in the locations indicated in the site map (Figure 3). Assign steady pumping rates during the model calibration exercise (see next section). All wells should be added to the 2nd model layer (middle glacial drift).
• Utilize a grid size of NX=100 (to effectively capture position of the wells).
  

Figure 5: Model layer extents.

Model Calibration

The following situation, based on a real-world investigation first published in the USGS Water Resource Investigations report by de Lima and Olimpio (1989), will be used to calibrate the model (i.e., determine appropriate values of hydraulic conductivity, leakances and long-term average recharge):

• First, simulate steady groundwater flow conditions to represent long-term 'natural' conditions. During this time, both wells G and H were turned off. Groundwater withdrawal rates of tannery wells 1 and 2 should be set to 70 and 200 gal/min.
• Next, for 30 days, simulate 'operational' conditions when wells G and H are turned on and pumping at their actual rates of 700 gal/min and 400 gal/min, respectively. (Industrial wells still pumping during this time.)
  • This transient modeling requires an initial head distribution. Use the steady-state head solved for pre-pumping, natural conditions as the initial condition. ( 'DomainAttr' > 'Simulation Settings' > 'Initial and Boundary Condition for Head' > 'Parent' option).
  • Make sure to "turn on" transient flow modeling ( 'DomainAttr' > 'Simulation Settings' > check box next to 'Transient Flow Modeling').  Use a Simulation Length of 30 days and a timestep of 1 day.
• You will compare your simulated heads to observed heads at scatted locations both before and after the pumping test. You will also compare your change in streamflow to an observed streamflow depletion measured after the 30-day pumping test of 1.26 cubic feet per second between Olympia Avenue and Salem Street.

 The calibration parameters (i.e., the inputs to be fine-tuned through trial-and-error) are:

• hydraulic conductivities of the different layers;
•  river and wetland leakances; and
• steady-state (natural) recharge to the aquifer system.

Head Comparisons:

The following calibration target files have been provided in the problem posting:

•  pre-pumping scatter-point data: calib_Dec4_1986_USE_THIS_ONE.txt  
•  post-pumping scatter-point data: calib_Jan3_1986_USE_THIS_ONE.txt

Use the 'Calibration' tool to load the observed data and compare them to simulated heads ('Analysis Tools' > 'Calibration' ...choose to compare points across entire model domain). Be sure to use '3D Weighted' as the Interpolation Option (because you are working with a 3D model).

Section 6.3 of the MAGNET User Manual provides information related to loading, displaying and comparing scatter-point data with simulated head. Please refer to this section and make use of the provided observational data to evaluate the model performance. 

Streamflow Depletion Analysis: 

The goal is to compare simulated net water gained from the river (as a source of water to the groundwater model) to the observed streamflow depletion (observed loss to the aquifer).  This can be done by 1) creating a ‘mass balance’ zone along the Aberjona River and 2) analyzing the fluxes in/out of the zone for pre- and post-pumping test conditions:

• Add a zone that traces and fully encompasses the Aberjona River from just north of Olympia Avenue to Salem Street.
• After finalizing the zone with the ‘SaveShape’ button, navigate to the Flow Properties tab  and check the box next ‘Zone Budget’. Save the changes and exit the menu.
• Run the transient simulation for the full 30-days (time-step of 1 day or less). When the simulation is finished, use the ‘ViewResults’ button and the ‘Display Charts’ option to display the Mass Balance chart.
• Select the ‘Zone’ option from the Chart options, then select the zone along the river from the corresponding drop-down menu. Click ‘Redraw’ to update the chart.
• Compare the simulated net water flow gained from the river to the observed streamflow depletion of 1.26 ft³/s between Olympia Avenue and Salem Street measured after the 30-day pumping test. Is the simulated result reasonable? If not, you can improve the model by fine-tuning the calibration parameters!

Model Application (Particle Tracking)

Now you will use the calibrated groundwater model to determine the potential source(s) of contamination to wells G and H. (Steady flow field during operational conditions, i.e., Wells G and H are turned on, with particle tracking applications). 

A useful first step for identifying potential source of contamination to wells G and H is to delineate well capture zones. Also called areas of contribution, well capture zones are delineated using a process of reverse particle tracking. In the process, several particles are added around each well and traced backwards in time. The direction and speed of the particles is controlled by the simulated 3D flow field the hydraulic properties of the subsurface.

Task 1: Add particles around wells G and H

1. Click on well G in the model.
2. Check the box next to ‘Add Particle’ in the bottom-right corner of the submenu, then click ‘OK’ to save the changes.
3. Repeat the process for well H.

Task 2: Perform reverse particle tracking to delineate potential well capture zones

Once particles are placed in the model, particle tracking will be performed anytime the model is submitted for simulation.

1. First, make sure wells G and H are "turned on"
2. Submit the model for simulation. A prompt will help to determine if forward or backward particle tracking will be performed. Choose backward particle tracking.
3. Particle pathlines (flow paths) will develop as the particles trace backwards with time.
   1. Note that the model may be paused at any point by clicking the ‘Pause’ button along the left-side of the MAGNET Modeling Environment.
   2. To reset particles to their initial position (“reset the particle clock”), select ‘ResetParticle’ from the particle tracking submenu. 

Next, you will visualize potential flow paths taken by contaminants from five known source properties, assuming that contaminant transport is dominated by advection and that dispersive processes or retardation can be ignored. Thus, a forward particle tracking process will be applied, in which particles are placed in zones representing contamination sites and traced forward in time along the 3D flow field. This will allow identifying delivery pathways and travel times of potential sources of contamination in wells G and H. 

Task 3: Add particle zones at the potential source properties

The approximate locations of five known sites of contamination is shown in Figure 6. Use the ParticleRect or ParticleZone tools to add particles at the site locations ('Simulation Tools' > 'ParticleTK' > 'ParticleRec' or 'ParticleZone').

Task 4: Perform forward particle tracking to estimate contaminant flow paths and particle times

1. Submit the model for simulation. A prompt will help to determine if forward or backward particle tracking will be performed. Choose forward particle tracking
2. Particle pathlines (flow paths) will develop as the particles trace forwards with time.

Finally, you will evaluate the contribution of the Aberjona River and wetlands using particle tracking.

Task 5: Add a particle line along a 'losing stretch of the Aberjona River

1. Consider, based on the flow patterns observed during operational conditions, where you expect the River to be losing water.
2. Add a particle line that traces the section of the River where you expect it to be losing water.
   • A 'general' delineation is recommended, i.e. make the polyline slightly larger/longer than the losing reach you have identified. 
3.  Once the last vertex has been placed, click the ‘SaveShape’ button to finalize the polyline.
   
   

Task 6: Perform forward particle tracking to determine if the Aberjona River is contributing water to Wells G and H

1. Submit the model for simulation. A prompt will help to determine if forward or backward particle tracking will be performed. Choose forward particle tracking.

Figure 6: Known VOC source properties and contamination sites.

Related MAGNET Tutorials:

• 2D Steady Flow
• Unsteady Flow
• Aquifer Layers
• Particle Tracking
• Water Budget Analysis
• Calibration

References:

Virginia de Lima and Julio Olimpio of the U.S. Geological Survey.  Students may to refer to the full report “Hydrogeology and simulation of ground-water flow at superfund-site wells G and H, Woburn, Massachusetts” [Water Resources Investigations Report 89-4059]

Disclaimer:

Although this project is based on a real-world story, some information has been simplified, modified, or omitted to make the problem suitable for educational application. The results and interpretations of this project are not meant to be liable assessments of the real-world contamination problem.