Site Description

The 92-acre Coakley Landfill Superfund site is located in the towns of Greenland and North Hampton, New Hampshire. The Coakley Landfill covers about 27 acres in the southern part of the site and is entirely within the town of North Hampton (figs. 1, 2B; CES, Inc., 2018a). The town of North Hampton operated the landfill between 1972 and 1985 (EPA, 2021). It accepted municipal and industrial wastes from local municipalities and federal entities. Coincident with landfill operations, rock quarrying took place at the site from 1973 through 1977. Site operators placed some of the disposed wastes in open (some liquid-filled) trenches created by rock quarrying and sand and gravel mining. Site operators also accepted incinerator residue from nearby incineration plants that operated from 1982 to 1985. Disposal activities resulted in the contamination of soil, groundwater, and surface water with metals and organic contaminants. Prior to the introduction of public water supply to the area in the 1980s, these contaminants were found in the private water-supply wells in the vicinity of the Coakley Landfill (Hoffman, 2025).

The landfill forms a prominent raised plateau with moderately steep slopes except on the northern side, which has a gentler slope. As part of site remedial design and construction activities implemented in the mid- and late 1990s, stormwater runoff from the landfill surface is conveyed by a series of perimeter drainage ditches and other structures to two unlined stormwater retention basins, one near the northeast corner of the landfill and one near the northwest corner of the landfill. Stormwater retained in the basins is subsequently discharged to adjacent wetland areas and ultimately to Berrys Brook through infiltration and an outlet structure in each basin. Several discrete groundwater discharge zones, or seeps, are located where the steep northwestern landfill slopes intersect natural ground. The seeps likely have some combination of interflow (subsurface flow above the water table) and shallow groundwater flow.

The Coakley Landfill site borders undeveloped woodlands and wetlands to the west, and commercial and scattered residential properties to the north, east, and south. Some residential properties in the area use private wells for water supply. A nearby golf course to the north has several wells in both the surficial aquifer and the bedrock aquifer for irrigation and potable water, respectively. Much of the area is also unsewered; septic fields return a part of pumped water from the domestic wells to the underlying aquifers. The commercial development to the east is serviced by the town water supply.

Climate in the model area is characterized by wet continental conditions with cold winters and warm summers (Köppen, 1884). Average annual precipitation for southeast New Hampshire ranges from 35 to 53 inches per year (Flynn and Tasker, 2004). Based on analysis of streamflow regression characteristics, about 40 percent of the annual precipitation of the area recharges the underlying aquifers (Flynn and Tasker, 2004).

The landfill sits atop a local topographic high and is underlain by a complex geologic structure that includes surficial sediments (0–88 feet [ft] thick; fig. 4A) and the underlying Rye Complex, including metasedimentary schists and gneisses (Lyons and others, 1997) and the informal Breakfast Hill granite (figs. 4B; Hussey and others, 2016). Transmissivity of surficial sediments is indicated in figure 4A. Mapped transmissivity values are representative of permeable glaciofluvial stratified drift deposits, generally Quaternary in age (Stekl and Flanagan, 1992). Areas without mapped transmissivity values consist primarily of less permeable sediments: Quaternary glacial till (unsorted mixture of various sediment sizes), glaciomarine clays and silts, and Holocene wetland deposits.

The underlying rock of the Breakfast Hill granite and schists and gneisses of the Rye Complex have been mapped by Lyons and others (1997) and Hussey and others (2016). Multiple faults at the contact between the Breakfast Hill granite and other rocks of the Rye Complex have been identified. The fracture network in these rocks is strongly anisotropic. The predominant fracture orientation is northeast-southwest, roughly parallel to foliation, with a median fracture dip of 64 degrees west-northwest (Haley Ward, Inc., 2021).

Coakley Landfill PFAS Plume Background

Water-quality monitoring of sediment, surface water, seeps, and groundwater including monitoring wells at the Coakley Landfill site and adjacent water supply wells have been ongoing since 1998 (CES, Inc., 2018a). PFAS contamination in soil and groundwater is a growing concern internationally (Simon and others, 2019); six PFAS were added to the list of contaminants of concern at the Coakley Landfill beginning in 2016 (EPA, 2016). Four of these are regulated compounds in New Hampshire; perfluorooctanesulfonic acid (PFOS), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), and perfluorohexanesulfonic acid (PFHxS). The two unregulated compounds are perfluoroheptanesulfonic acid (PFHpS), and perfluorobutanesulfonic acid (PFBS).

At the Coakley Landfill site, PFOS and PFOA are the compounds most commonly found above the New Hampshire ambient groundwater quality standards in surface and groundwater (12 and 15 ng/L, respectively [NHDES, 2023]). Perfluorohexanesulfonic acid (PFHxS) and perfluorononanoic acid (PFNA) do not routinely exceed their respective state standards of 18 ng/L and 11 ng/L (CES, Inc., 2020b; NHDES, 2023).

Historically, the highest PFAS concentrations have been found in groundwater seeps at the embankments of the landfill and from surface runoff from the Coakley Landfill site. In 2019, a maximum combined PFOA and PFOS concentration of 516 ng/L was found in water from a seep on the western side of the landfill and a concentration of 1,525 ng/L was observed in samples of surface water near the Coakley Landfill (CES, Inc., 2020a). Further, in 2019, a maximum combined PFOA and PFOS concentration of PFAS compounds in sediment was found to be 0.02798 milligram per kilogram (CES, Inc., 2020a).

In 2019, PFOA and PFOS were found in samples of groundwater from monitoring wells near the landfill with maximum combined concentrations of 1,398 ng/L in the fall and 1,404 ng/L in the spring (CES, Inc., 2020a). Notably, combined concentrations of PFOA and PFOS exceeded 1,000 ng/L in multiple onsite bedrock monitoring wells, demonstrating penetration of contaminants from the overburden into the bedrock aquifer. Offsite, the maximum combined PFOA and PFOS concentration in samples of water-supply wells was 16.57 ng/L (CES, Inc., 2020a).

Based on 2019 sampling results (CES, Inc., 2020a) in surface water and soil near the landfill, concentrations of PFOS generally exceed concentrations of PFOA. Farther from the landfill, concentrations of PFOA are generally higher than concentrations of PFOS in surface water and soil. This is consistent with PFOS’s relatively lower solubility and mobility (EPA, 2017; Xing and others, 2021). Groundwater concentrations of PFOA exceed concentrations of PFOS at almost every sample location (CES, Inc., 2020a). The distribution of total dissolved PFOA concentrations in groundwater of the bedrock near the Coakley Landfill site is shown in figure 5. The PFOA plume initiates near the landfill and extends to the north, beneath the headwaters of Berrys Brook.

Orientation and Extent

The model extent includes most of North Hampton, the majority of Greenland and Rye, and part of Portsmouth and Stratham, N.H. (fig. 1). The model is rotated 29 degrees from true north to align with the strike of the major rock contacts as reported by Lyons and others (1997). The model is about 7.4 miles (mi) wide in the X direction, and about 5.5 mi in the Y direction.

Boundary Conditions

The western boundary of the model is a no-flow boundary for the full thickness of the model and is located near the western bank of the Winnicut River (fig. 6). The Winnicut River and its tributaries primarily act as a discharge boundary within a low-lying valley (Mack, 2009). To accommodate groundwater discharge from the western bank of the Winnicut River to the river, a recharge rate equivalent to the area runoff of the upland areas on the western side of the Winnicut River was assigned to cells along the western boundary. The rate of recharge was checked against groundwater flow computed by the southeast regional model for the same area. The line recharge was assigned to the surficial aquifer (layers 1 and 2 of the model).

The eastern boundary of the model is the Atlantic Ocean and is simulated as a constant head boundary at 1.0-ft altitude to compensate for salinity and related density effects on an equivalent freshwater head. An approximate calculation of equivalent freshwater head is provided in Kuniansky (2018). The ocean acts as a discharge boundary and was assigned to layer 1 of the model.

The southern and northern boundaries are constant-head boundaries with heads assigned from a simulation of the southeast regional model (Mack, 2023) for that location to the full thickness of the model. The southern and northern boundary heads were adjusted to help match observed heads (Mack, 2009) near that boundary. The northwestern area of the model by Great Bay was assigned a constant head of 3.0-ft altitude based on marshland elevation data reported in Stevens and others (2022). The constant heads at Great Bay were assigned to layer 1 of the model.

Discretization and Layering

The model cell sizes range from 50 by 50 ft at the landfill area to 250 by 250 ft in distal parts of the model (app. 1, fig. 1.1). There are 334 rows by 347 columns or 115,898 cells per layer. The total model thickness ranges from 1,000 to 1,100 ft thick. Most wells in the area are less than 850 ft deep (NHDES, 2020a).

The model is subdivided vertically into six layers (fig. 7). Model grid-cell elevations were interpreted with respect to 30-meter digital elevation model (DEM) point elevations (Mack, 2009). The upper surface of the model (layer 1) corresponds to the nearest DEM 30-meter grid-cell elevation. The elevations of all other model layer surfaces were subtracted from the surface elevation. Layer thicknesses were chosen to incorporate both hydrogeologic characteristics and numerical simulation considerations. Layers 1 and 2 represent primarily surficial deposits (overburden) or water bodies. The thickness of layer 1 in marine areas (Great Bay and the Atlantic Ocean) is equal to the water depth indicated by bathymetric data (Mack, 2009). On land, the thickness of layer 1 (8–88 ft) was determined from contours representing thickness of the stratified-drift aquifer (Moore, 1990; Stekl and Flanagan, 1992), bedrock well casing lengths and boring data, and thicknesses inferred from surficial mapping. The thickness of model layer 2 was assigned at 6.6 ft to represent an average till thickness over the bedrock surface. Model layer 3 was assigned at 10 ft thick to represent the zone where bedrock water-supply wells are typically cased off during well installation. Model layers 4 and 5 were assigned thicknesses of 290 ft and 300 ft, respectively. Model layer 6 was assigned a thickness of 400 ft (fig. 7).

Groundwater Withdrawals and Return Flows

Major groundwater withdrawals (generally greater than 19,300 cubic feet per day [ft³/d] or about 100 gal/min) representing municipal or commercial water usage and selected private (residential) usage of lesser withdrawal amounts (generally less than 19,300 ft³/d) were simulated with the Well package (WEL) of MODFLOW-2005 (Harbaugh, 2005). Figure 6 shows the location of withdrawal wells as specified with the Well package. Wells are assigned to the layer containing the interval at which each well is screened. Table 2 lists additional information on these wells including assigned withdrawal rates, model row and column location, and layer information. Appendix 1 contains a map with a cross-reference number based on the well number in table 2 (fig. 1.2).

Only withdrawal wells where the location and well log could be identified were simulated with the Well package. Information on well logs and location were obtained from the New Hampshire Department of Environmental Service well inventory records (NHDES, 2020a).

To account for wells not simulated with the Well package, the Flow and Head Boundary (FHB) package (Leake and Lilly, 1997) was used to simulate distributed withdrawals (Mack, 2009). Distributed withdrawals were assigned to model layers 4 and 5. The location of distributed withdrawals is shown in appendix 1, fig. 1.3A.

Return flows from predominantly nonsewered areas of the Coakley Landfill model, including septic fields, were also simulated as distributed return flow with the Flow and Head Boundary package. Return flow was assigned to model layer 2 and occurred in coincident locations as distributed withdrawals (app. 1, fig. 1.3A; NHDES, 2025). A return rate of 85 percent of water consumption was used (Horn and others, 2008).

Recharge

The Recharge package (RCH) of MODFLOW-2005 (Harbaugh, 2005) was used to specify groundwater recharge to the Coakley Landfill model. Rates of recharge were informed by rates specified in Mack (2009, 2017, 2023) from baseflow separation techniques using continuous streamflow data (Flynn and Tasker, 2004). Three initial rates were used, based on surficial sediments: (1) 0.0025 foot per day (ft/d) in high infiltration areas associated with high permeability surficial sediments, (2) 0.0013 ft/d in low infiltration areas associated with low permeability surficial sediments, and (3) 0 ft/d over large surface-water bodies. Most of the areas of mapped transmissivity values on fig. 4A are coincident with high infiltration areas. Areas without mapped transmissivity values are primarily areas of low infiltration.

Stream-Aquifer Interactions

The Stream package (STR) was used to simulate stream-aquifer interactions within the groundwater-flow model (Prudic, 1989). The Stream package accounts for streamflows along designated reaches from groundwater discharge and recharge to the aquifer. Streamflow along reaches is instantaneously routed to downstream reaches. Altitude data for streambed and stream stage, and thicknesses for streambed, were assigned to each stream segment. The streambed conductance term, which is also assigned to each segment, is used to relate the rate of model-computed flow to the difference in head between the stream stage and model-computed heads. Streambed conductance was kept spatially uniform in the model.

Stream surface and stream bed altitude data were incorporated from Mack (2023). Elevations were checked to ensure streamflow was monotonic from upstream to downstream locations. Simulated streams and basin locations, which are used to assess measured and model-computed stream leakages, are shown in fig. 3.

Hydraulic Conductivity

The initial hydraulic conductivity values of hydrogeologic units were derived from the calibrated model by Mack (2009). Estimates of the horizontal hydraulic conductivity for the overburden hydrogeologic units, consisting of the stratified drift, till, and marsh, were 10 ft/d, 1 ft/d, and 0.1 ft/d, respectively. The vertical hydraulic conductivity was initially estimated by multiplying the horizontal conductivity by 0.1, 1, and 0.1, respectively.

Hydraulic conductivity of the rock units was estimated to be about 0.1 to 1.0 ft/d, with a regional horizontal anisotropy of 2.5:1 to 5:1 (Y direction:X direction) following the regional structural trend and predominant northeast-southwest fracture orientation (Mack, 2009). Model cells representing the Rye Complex (except the informal Breakfast Hill granite) and the Kittery Formation of the Merrimack Group were assigned higher estimated hydraulic conductivities (0.2 to 1 ft/d) than areas representing the Breakfast Hill granite (0.1 ft/d). The Rye Complex was further broken into two informal units—gneiss and schist—based on primary lithology (figs. 4B, 7), with the Rye Complex gneiss being assigned an initial hydraulic conductivity of 0.5 ft/d and the Rye Complex schist being assigned an initial hydraulic conductivity of 0.2 ft/d. Vertical hydraulic conductivity was set equal to horizontal hydraulic conductivity for the rock units. The model hydraulic conductivity field was further refined using data collected from purge records of low-flow sampling, drilling logs, and surface geophysical surveys (CES, Inc., 2017, 2018a, 2018b5, 2019, 2020b) and analyzed with the Purge Analyzer Tool (PAT) software (Harte and others, 2019; Harte and Huffman, 2020).

Mack (2009, 2017, 2023)2627 incorporated multiple lineament features set at 10 times the hydraulic conductivity of the respective rock unit in which each lineament was located. The number of lineament features was reduced during the calibration of the Coakley Landfill model. Three lineament type features that were incorporated in the Coakley Landfill model include the following: (1) a linear zone of low electrical resistivity on the west side of the Coakley Landfill mapped by CES, Inc. (2017); (2) a fault contact between the Kittery Formation and the Rye Complex (fig 4B; Mack, 2009), and (3) a northeast trending lineament just northeast of the Coakley Landfill (Mack, 2009).

Recent (2021) aquifer pump test results from bedrock wells show a range in hydraulic conductivity values of 0.75 ft/d to 178 ft/d. The upper end of the range was calculated from a shallow rock well that is likely influenced by rock weathering (Haley Ward, Inc., 2021). A previous pump test in 1994 estimated that hydraulic conductivity for the bedrock ranges from 0.99 to 3.69 ft/d (Haley Ward, Inc., 2021). Reported ranges of bedrock hydraulic conductivity imply that there are localized high hydraulic conductivity zones within low hydraulic conductivity zones. The high hydraulic conductivity values reported by Haley Ward, Inc. (2021) are more likely to occur along lineaments or other structural features.

Parameterization

Parameterization is used to define spatiotemporal properties of the model to facilitate representation of the aquifer and groundwater flow system and to assist in model calibration. The initial parameters were taken from Mack (2009, 2017, 2023). There are 21 parameters to represent spatial variability in horizontal and vertical hydraulic conductivity, 3 parameters to represent spatial variability in recharge, and 1 parameter to represent streambed conductance.

Model Solution

MODFLOW-NWT requires use of the Upstream Weighting (UPW) package to specify properties controlling flow between cells (Niswonger and others, 2011). The Upstream Weighting package treats nonlinearities of cell drying and rewetting by use of a continuous function of groundwater head, rather than the discrete approach of drying and rewetting that is used by the other flow packages in MODFLOW-2005 (Harbaugh, 2005). This facilitates application of the Newton method for unconfined groundwater flow problems and complex nonlinear solutions (Niswonger and others, 2011).

The Coakley Landfill model features nonlinear stress packages and multiple unconfined layers representing complex hydrogeologic dynamics. Therefore, MODFLOW-NWT’s “complex” solver keyword was invoked to facilitate model convergence. Default solver parameters were retained (Niswonger and others, 2011, table 2), with the following exceptions: (1) a head closure criterion of 0.25 ft was used; (2) a flux tolerance criterion of 5,000 ft³/d was applied, and (3) a maximum of 5,000 outer iterations was permitted to achieve the closure criteria.

UCODE_2014 software (Poeter and others, 2014) was used for sensitivity analysis and parameter estimation. The software graphical interface ModelMate (Banta, 2011) was used to visualize UCODE_2014 input and output. The sensitivity analysis was used as a screening tool to identify the most sensitive parameters for calibration. Parameter estimation was done to bracket parameter values for further iterative model calibration. For parameter estimation, a maximum of 35 iterations was permitted, with a default parameter change of 0.1 and a sum of squares closure criterion of 0.01.

Numerical Model Calibration

As part of the calibration, boundary flows from the Coakley Landfill model were checked against the larger southeast regional model by Mack (2023). Flows from the calibrated Coakley Landfill model at boundary conditions (fig. 6) are similar to flows from the southeast regional model at the same locations (table 3). For the Coakley Landfill model, inflow predominantly occurs at the northern boundary and outflow or water loss occurs along the southern boundary. There is also inflow from the west boundary and outflow to the Atlantic Ocean.

In general, the model-computed leakage shows greater amounts of groundwater discharge to streams than measured leakage as reported by Mack (2009). The average relative percent difference is −28.8 percent for all basins (table 4). Basin locations are shown in fig. 3 and listed in table 1, and individual streamflow information is provided in appendix 1, table 1.1. The measured leakages represent groundwater discharges collected during a two-day synoptic measurement event in October 2004 by Mack (2009). These measured flows are lower than long-term averages (table 1.1); therefore, model-computed stream leakages for the Coakley Landfill model are likely more representative of average conditions than the reported measured streamflow from Mack (2009).

Comparison of offsite measured water levels (Mack, 2009) and onsite measured water levels (Haley Ward, Inc., 2022) with model-computed heads (unweighted) shows an overall acceptable model fit (fig. 8). Scatter of head residuals about the 1:1 line is larger for offsite wells—indicating a larger mean absolute residual—than for onsite wells. For onsite wells, the simulated heads are, on average, slightly higher than measured heads. For offsite wells, the simulated heads are slightly lower than measured heads on average (table 5).

The spatial distribution of the head residuals shows a tendency for the model to underestimate heads in some areas to the west of the Coakley Landfill (darker purples in fig. 9). Several regional wells that are situated along the fault contact between the Rye Complex and the Kittery Formation show low model-computed heads. For these wells, there may be preferential flow along and across the contact that is not represented by the model.

Several parameters were adjusted during the model calibration process. The largest change from initial parameter values to final calibrated parameter values is in the horizontal hydraulic conductivity of the stratified-drift material (parameter name Ksd), and the Breakfast Hill granite (Rx3; table 6). An increase in the horizontal hydraulic conductivity of the parameter Ksd from 10 to 60 ft/d was driven by dewatering of model cells in areas where large groundwater withdrawals are simulated. Increasing Ksd to 60 ft/d prevented dewatering of model cells and decreased head residuals. A decrease of horizontal hydraulic conductivity in the Breakfast Hill granite unit from 0.1 to 0.003 ft/d was primarily the result of rezoning the unit relative to its extent in Mack (2009) based on bedrock geologic mapping by Hussey and others (2016) and hydraulic conductivity estimates from wells that were positioned in the Breakfast Hill Granite zone. Changes to the parameter zonation and associated parameter values of the formation also led to a decrease in head residuals. The inclusion of several lineaments of preferred horizontal hydraulic conductivity in the Coakley Landfill model, which resulted in an upper value of horizontal hydraulic conductivity of 6 ft/d in the bedrock, also decreased residuals in a few wells near lineaments including the fault contact between the Rye Complex and the Kittery Formation.

Sensitivity of Parameters

Composite scaled sensitivities for heads show that the groundwater recharge parameter associated with high infiltration areas of the model, Rech1, is the most sensitive model parameter (table 7). The horizontal hydraulic conductivity parameters of the three rock units with large extents in the model area, Rye Complex gneiss (Rxk1), Rye Complex schist (Rxk4), and Kittery Formation (Rxk2), are the next most sensitive parameters. The least sensitive parameters relate to surface waterbodies (hydraulic conductivity and recharge; Ksw and Rech0). Several of the hydraulic conductivity parameters for the surficial sediments are also insensitive because of the limited areal extent of those deposits.

The relatively large number of model parameters (table 7) precludes the ability to bracket all parameter values with computed confidence intervals. Therefore, selected parameter confidence intervals were computed for the horizontal hydraulic conductivity of the Rye Complex gneiss, Rye Complex schist, and Kittery Formation (table 6). Final parameter values from model calibration were within the range of model-computed parameter confidence intervals. The Rye Complex (gneiss and schist) computed maximum potential values of horizontal hydraulic conductivity were much larger than the values for the Kittery Formation.

Recharge and Flow Paths

Recharge to the aquifer accounts for 80 percent of total inflow in the Coakley Landfill model compared to 24 percent for the southeast regional model (table 8; Mack, 2023). A much smaller amount of inflow for the Coakley Landfill model comes from constant-head boundaries. The western boundary is the primary inflow boundary to the Coakley Landfill model with a smaller amount of flow coming from the north and south (table 3). This means that most flow in the Coakley Landfill model originates from the model area itself and not from the southeast New Hampshire region beyond the Coakley Landfill model area.

The Coakley Landfill is slightly northwest of a simulated groundwater high (fig. 10; 100-ft contour). This positioning next to an area with high groundwater heads induces groundwater flow to the northwest with negligible flow to the east. Maximum groundwater gradients from the landfill are to the north, facilitating contaminant transport in that direction.

The Coakley Landfill cap affects localized groundwater flow near the Coakley Landfill. Different model scenarios calculated groundwater flow for current conditions with a cap and hypothetical conditions before the cap. Model results show that the Coakley Landfill cap reduces the amount of bedrock recharge under the cap by a small amount (figs. 11A–C). The Coakley Landfill cap reduces heads by about 0.75 ft.

Simulations with advective transport of particles in groundwater from underneath the cap, and placed in the shallow bedrock (656 particles total in layers 3 and 4 of the model), show that most groundwater flow discharges to the adjacent streams (Berrys Brook and Little River) west of the Coakley Landfill under both current capped conditions (fig. 12A; zone 1) and historical conditions before capping (fig. 12B; zone 1). Some particles are transported to the north from the Coakley Landfill to Packer Brook and distal discharge points along Berrys Brook for both cap conditions (figs. 12A, B; zone 2). There are three factors that tend to induce preferential advective transport to the north of the landfill: (1) the bedrock aquifer's hydraulic anisotropy along strike favoring flow to the northeast, (2) the location of the lineament to the west of the landfill, and (3) the location of the maximum gradient from the Coakley Landfill to the north direction (fig. 10). Ten percent of the particles discharge and are captured by withdrawal wells to the southwest and west for both conditions (figs. 12A, B; zones 3 and 4). Under both capped and uncapped conditions, areas designated as zones 3 and 4 are considered less probable discharge areas than zones 1 and 2 given the length and likely travel time associated with groundwater flow to zones 3 and 4 from the Coakley Landfill.

There is a large range in potential travel distances of tracked particles from the Coakley Landfill to distal locations. Radial horizontal travel distances of pathlines tracked for the simulation with the landfill cap (fig. 12A) range from 0.08 to 2.87 mi (table 9). These distances do not account for the circuitous nature of the bedrock fracture network, so actual distances are likely to be longer. Nevertheless, they provide some approximate indication of potential contaminant travel distances from the Coakley Landfill. Contaminants traveling along distal pathlines of several thousand feet or greater are likely to undergo dilution, mixing, and adsorption that could potentially decrease contaminant concentrations in groundwater downgradient from the landfill. Concentration gradients for PFOA shown in figure 4 indicate a one-order magnitude decrease over a distance of several thousand feet.

The depth of the initial placement of particles affects forward-tracking results and the ultimate discharge area of model-computed groundwater flow for the capped and uncapped scenarios (figs. 12A, B, 13A, B). Similarly to particles originating in shallow bedrock, advective transport of particles (323 particles in total) from underneath the cap area in the deep overburden (model layer 2) also shows that most groundwater flow discharges to the adjacent streams west of the Coakley Landfill for the capped and uncapped conditions (fig. 13A, B; zone 1). Some particles from the deep overburden are transported to the north from the Coakley Landfill to Berrys Brook and Packer Brook under both capped and uncapped conditions (fig. 13A, B; zone 2). Unlike particles originating in shallow bedrock, however, no particles tracked from the deep overburden discharge to withdrawal wells directly west from the landfill for either cap condition (figs. 13A, B, fig. 12A, B; zone 4). Under capped conditions, no particles from the deep overburden discharge to a withdrawal well to the south (fig. 13A; zone 3), and under uncapped conditions, few particles discharge in the same area (fig. 13B; zone 3).

The anisotropy of the bedrock aquifer is an important factor that controls advective transport from the Coakley Landfill. Simulation without horizontal anisotropy shows that in addition to advective transport of particles discharging to Little River and Berrys Brook, and headwaters of Packer Brook (fig. 14; zones 1 and 2, respectively), a large fraction of particles discharges farther to the northwest near where Packer Brook meets Great Bay (fig. 14; zone 4) unlike in simulations with anisotropy as shown in figs. 12A, B–14. The change in the distribution of particle discharges illustrates the importance of incorporating horizontal anisotropy in models where preferential flow occurs along fractures, faults, and other features.