Journal of Contaminant Hydrology

Transcription

1 Journal of Contaminant Hydrology 147 (213) Contents lists available at SciVerse ScienceDirect Journal of Contaminant Hydrology journal homepage: Laboratory and numerical investigation of transport processes occurring above and within a saltwater wedge Sun Woo Chang 1, T. Prabhakar Clement Department of Civil Engineering, Auburn University, Auburn, AL 36849, United States article info abstract Article history: Received 27 July 212 Received in revised form 5 January 213 Accepted 1 February 213 Available online 16 February 213 Keywords: Saltwater intrusion Laboratory experiments Benchmark problems Contaminant transport Density-coupled flow Submarine groundwater discharge SGD Salt wedges divide coastal groundwater flow regime into two distinct regions that include a freshwater region above the saltwater freshwater interface and a saltwater region below the interface. Several recent studies have investigated saltwater transport in coastal aquifers and the associated flow and mixing processes. Most of these studies, however, have either focused on studying the movement of salt wedge itself or on studying contaminant transport processes occurring above the wedge. As per our knowledge, so far no one has completed laboratory experiments to study contaminant transport processes occurring within a saltwater wedge. In this study, we completed laboratory experiments to understand contaminant transport dynamics occurring within a saltwater wedge. We used a novel experimental approach that employed multiple neutral-density tracers to map and compare the mixing and transport processes occurring above and within a saltwater wedge. The experimental data were simulated using SEAWAT, and the model was used to further investigate the saltwater flow and transport dynamics within a wedge. The laboratory data show that the transport rates active within the wedge are almost two orders of magnitude slower than the transport rates active above the wedge for the small-scale experimental system which is characterized by very low level of mixing. The numerical results, however, postulate that for large-scale systems involving higher levels of mixing (or dispersion) the transport rate active within the wedge could be comparable or even higher than the rates active above the wedge. More field or laboratory studies completed under high dispersion conditions are needed to further test this hypothesis. 213 Elsevier B.V. All rights reserved. 1. Introduction Due to rapid urban development, shallow unconfined aquifers in several coastal regions have become highly vulnerable contamination by pollutants inadvertently discharged from various sources including leaking underground storage tanks and landfills. These anthropogenic sources have the potential to contaminate large volume of coastal groundwater and the associated shoreline beach environment. Presence of a saltwater wedge, a common feature of coastal aquifers, would influence the transport pathways of these anthropogenic contaminants. Corresponding author. Tel.: ; fax: address: clement@auburn.edu (T.P. Clement). 1 First author is now at: Geologic Environment Division, Korea Institute of Geoscience and Mineral Resources, 35-35, Republic of Korea. For example, Westbrook et al. (25) completed a field study near Swan-Canning Estuary which documented the transport pattern of a dissolved BTEXN (benzene, toluene, ethybenzene, the xylene isomers and naphthalene) plume in a shallow aquifer located near a tidally and seasonally forced saline water body in Perth, Western Australia. Their field data show that the contaminant plume discharged along beach-face sediments within a narrow zone of less than 1 m from the high tide mark. Groundwater flow occurring within a saltwater wedge can also actively transport various contaminants and nutrients. For example, another field study completed at a site located near the Swan-Caning Estuary investigated nutrient recirculation patters within an unconfined aquifer (Smith and Turner, 21). The study simulated nutrient circulation between the saline river-water and fresh groundwater systems using a numerical model; the model results show that nutrients are primarily /$ see front matter 213 Elsevier B.V. All rights reserved.

2 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) transported along a flow pathway induced by the densitydependent flow dynamics. Field observations supported these numerical simulation results. Smith (24) further demonstrated that coastal groundwater systems can circulate considerable amount of nutrients or other contaminants of marine origin within a saltwater wedge. Fig. 1 shows an idealized conceptual diagram that illustrates how contaminants are transported above and below a saltwater wedge present in coastal unconfined aquifer. As shown in the figure, contaminants discharged within the freshwater region (located above the wedge) are transported towards the ocean boundary and discharged along the beach face. On the other hand, contaminants present within the saltwater region would follow a circulatory flow pattern. The density contrast between the fresh groundwater and saline ocean water induces this circulatory flow within a saltwater wedge. The saline water flux induced by this motion has been identified to be an important component of the net coastal groundwater budget known as the submarine groundwater discharge (SGD) (Destouni and Prieto, 23; Li et al., 1999; Smith, 24). It has been well established that nutrients and chemicals transported by SGD have considerable impact on marine ecosystems (Johannes, 198; Moore, 1996; Moore and Church, 1996; Robinson et al., 27b; Simmons, 1992). These studies have highlighted the need for understanding the groundwater flow processes occurring within the wedge to better predict SGD and the associated contaminant exchange between terrestrial and marine waters. Researchers have either used model data or field data to understand the circulatory groundwater flow cell present within a saltwater wedge. Kohout (196) completed a field study in Florida and estimated that the saline groundwater flux circulated within the wedge was about 1% of the fresh groundwater flux which was directly discharging towards the sea. Smith and Turner (21) investigated the role of density-driven convection induced by the seasonal variations in density gradients between the pore water in the sediments of a saline river (Swan River in Perth, Australia) and the groundwater in an adjacent aquifer. Both two-dimensional (cross section model) and three-dimensional modeling efforts were completed to develop an understanding of the convective flow cycle which was responsible for transporting high levels of ammonium in the estuarine river bed to groundwater and eventually back to the estuary. Destouni and Prieto (23) completed a modeling study to develop empirical correlations between groundwater fluxes present above and below a saltwater wedge. Their correlations show that the recirculated saltwater flux is roughly an order of magnitude smaller than the freshwater flux. Smith (24) completed a comprehensive modeling study and concluded that the amount of saltwater circulated within the system can vary widely and would depend on the level of dispersion in the system. A few published studies have completed laboratory and numerical experiments to investigate contaminant transport processes occurring near a saltwater wedge. Zhang et al. (22) conducted laboratory experiments to visualize contaminant transport above a wedge. They conducted a set of tracer transport experiments that tracked the movement of various types of dense groundwater plumes above the salt wedge. Their experimental observations indicated that the presence of the saltwater interface influenced the sinking and lateral transport patterns of these dense plumes. Brovelli et al. (27) simulated Zhang et al. (22) experimental data using a numerical model and used the model to further investigate the interactions between a dense contaminant plume and a dense saltwater wedge. Their results show that the two dense water phases would never completely mix; instead a slice of plume would be continuously removed from the plume front by the freshwater flow moving through the system. Illangasekare et al. (26) used multi-colored tracers to study the movement of dense saltwater plumes in a tsunami impacted coastal aquifer. Their laboratory data showed that the dense tsunami waters infiltrated via the unsaturated zone floated over the regional saltwater wedge and the two waters never fully mixed. Boufadel (2) studied the transport pathways of nutrients used for remediating beach-face contaminants resulting from the past oil spills. Laboratory experiments were completed using two types of density gradients to investigate the effects of tides and buoyancy on beach hydraulics. In both experiments, the authors observed that the groundwater flow from the aquifer moved seaward via the intertidal zone and pinched out near the low tide mark. The experimental data was modeled using a numerical simulator, which was then used to design nutrient delivery strategies for bioremediating crude oil contaminated beaches. Volker et al. (22) completed a numerical modeling study to investigate the transport of a dense contaminant plume in an unconfined aquifer. Their simulation results were compared against experimental data. The study modeled the sea-side boundary Sea Level Freshwater level Contaminant/ Nutrient? Contaminant/ Nutrient Salt wedge toe, X T Fig. 1. Conceptual model illustrating various types of contaminant transport processes occurring in a coastal groundwater system.

3 16 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) using two types of conditions: a) constant head freshwater boundary which neglected density effects, and b) later as constant head saltwater boundary which included density effects. Their results indicated that when the plume is away from the boundary, predictions based on both approaches matched reasonably well with experimental results. However, as the plume approached the coastal boundary, the experimental data indicated that the plume will be uplifted along the saltwater interface and exited along the shoreline. This observed uplifting and shore-focused discharge patterns could only be recreated by the model that accounted for density effects. The model that ignored density effects predicted an incorrect travel pathway, where the plume travelled further into the sea and exited under the seabed. The study concluded that ignoring seawater intrusion effects would not only underestimate the amount of contaminant mass exiting along the beach face, but it would also result in simulating unrealistic transport pathways. Recently, Kuan et al. (212) used a physical sand tank model to study tide-induced circulation processes that lead to the formation of an upper saline plume within the intertidal beach zone. They found that the presence of the upper saline plume moved the fresh groundwater discharge zone towards the low tide mark and restricted the movement of the salt wedge. A systematic review of the published literature indicated that past experimental efforts have either focused on studying the bulk movement/dispersion of transient and steady-state saltwater wedges [e.g., Abarca and Clement (29), Goswami and Clement (27)], or on studying the contaminant transport occurring above a steady-state wedge (Illangasekare et al., 26; Kuan et al., 212; Zhang et al., 22). To the best of our knowledge, so far, no one has conducted laboratory experiments to visualize the transport processes occurring within a saltwater wedge. The objective of this effort is to conduct well-controlled contaminant transport experiments to study the dynamics of groundwater flow and transport processes occurring within a saltwater wedge, and relate it to the processes occurring above the wedge. We used a novel experimental approach that employed a variety of dyed waters to map and compare the transport patterns occurring above and within a saltwater wedge. The experiments were simulated using a numerical model. The model was then used to analyze a field-scale problem. H s = 25.1 cm Saltwater Pump Injection point for Test cm Freshwater Pump Injection point for Test 1 & Test 2 Fig. 2. Schematic diagram of the physical model used in this study. was set at 26.3 cm. The seawater level was set to 25.1 cm by adjusting the saltwater head at the left outlet chamber. The hydraulic conductivity (K) of the tank was measured using the in-situ method described in Simpson et al. (23). The average value of conductivity was 1.3 cm/s. The measured values of saltwater and freshwater densities were 1.25 g/cm 3 and 1. g/cm 3, respectively. All the experimental parameters used in this study are summarized in Table 1. The transport experiments were recorded using a video camera (Panasonic HDC-HS25). Other experimental methods used were similar to the ones reported in our previous studies (Abarca and Clement, 29; Chang and Clement, 212; Goswami and Clement, 27; Goswami et al., 212). One of the differences, however, was that the porous media section was slightly modified to accommodate two fine needles of lengths 3 mm and 5 mm (see Fig. 2). These needles were used to inject tracer slugs above and below the wedge. Both needles were buried in the tank when the porous medium was wet packed. In addition to this physical modification, in this study, we employed multiple colored tracers to track various waters. The scheme used three different types of waters that used different colors: freshwater (dyed red), dense tracer (dyed green), and H f = 26. cm 2. Methods 2.1. Experimental methods Fig. 2 shows the experimental flow tank used in this study. The dimension of the flow tank used was 5.5 cm 28 cm 2.2 cm. The origin of the Cartesian coordinate system used to record experimental observations was set at the left bottom corner of the tank. The tank was fitted with three distinct flow chambers. The right inlet chamber was used to supply freshwater flow using a peristaltic pump. The central porous media chamber was filled with uniform silica beads of average diameter of 1.1 mm. The beads were packed under saturated conditions to prevent air bubbles being trapped within the tank. Thin stainless-steel meshes were used to separate the side chambers (boundary conditions) from the porous media chamber. Freshwater level at the right boundary Table 1 Model parameters used for simulating the laboratory problem. Parameter Symbol Value Horizontal length L 5.5 cm Saltwater elevation H s 25.1 cm Freshwater elevation H f 26.3 cm Hydraulic conductivity K 1.3 cm/s Specific storage, S s cm 1 Longitudinal dispersivity α L.5 cm Transverse dispersivity α T.5 cm Saltwater concentration Cs.35 g/cm 3 Saltwater density ρ s 1.25 g/cm 3 Freshwater density ρ f 1. g/cm 3 Density slope E.714 Porosity n.385

4 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) dense seawater (colorless). In Illangasekare et al. (26) study we provided an experimental dataset which was also collected in our laboratory using a variety of dyes to map the mixing of waters in a tsunami-impacted coastal aquifer. In the current effort, we adapt this approach to study the transport processes occurring above and below a saltwater freshwater boundary. However, it is important to note that in all our previous experimental datasets, the saltwater wedge was always marked using solid red or green dye. In this effort, we reversed the coloring scheme and created a colorless, steady-state saltwater wedge first and then used a green-colored saltwater tracer slug to visualize the transport patterns occurring within the colorless wedge. The dense saltwater tracer slug was marked using 4 ml/l of green food color. To differentiate the regional saltwater wedge from the freshwater region, we marked the freshwater using 1 ml/l of commercial red food color. Goswami and Clement (27) observed no measurable sorption of these food dyes on the glass beads used in this study. They also found that the difference between diffusion coefficients for salt constituents and dye are negligible and hence the dye concentration levels can be used to delineate the saltwater region within the system Details of tracer transport experiments Three types of tracer solutions with density values of 1., and 1.25 g/cm 3 were used to design three different transport experiments identified as Test-1, Test-2 and Test-3. Test-1 and Test-2 are baseline reference transport experiments that involved injection of 22 ml of tracer slug with relative density values of 1. (tracer) and (dense tracer) above the saltwater wedge close to the freshwater boundary (see Fig. 2 for details). The total volume of the tracer slug used was 22 ml and it was injected within 8 s, yielding an effective injection rate of 2.75 ml/s. The x z coordinates of the injection point are: x=45 cm and z=22.5 cm. The tracer slugs were allowed to transport with the freshwater flow and the images were recorded at different times until the plume reached the exit boundary. The result of Test-1 was used to calibrate a numerical model by adjusting the dispersivity values, and Test-2 data was used to verify the model performance. Test-3 was the most important study that recorded the transport patterns occurring within the wedge. The coordinates of the injection point are: x=4. cm and z=.5 cm. About 11 ml of green-colored saltwater solution with a density value of 1.25 g/cm 3 (note the density is identical to the seawater density in the left tank) was injected beneath the wedge. The total amount of time take for injection was 4 s, yielding an effective injection rate of.275 ml/s. Since the ambient density of water saltwater within the wedge was indeed 1.25 g/cm 3, the green-colored dense saltwater slug was expected to behave like a neutral tracer within the wedge Numerical modeling methods used for simulating the experiments The MODFLOW-family variable density flow code SEAWAT (Guo and Langevin, 22) was used to simulate the experimental results. A uniform two-dimensional grid of dimension.5 cm (12 cells in the x-direction and 5 cells in the z-direction) was used to discretize the flow tank. This grid resolution was similar to the one used in our previous studies (Chang and Clement, 212; Goswami and Clement, 27). As a part of our previous modeling efforts, we have completed test simulations to verify model convergence and have found.5 cm to be an optimal grid size for simulating this experimental flow tank. Lateral thickness of the model cells were assumed to be 2.2 cm, which is identical to the thickness of the physical model. The bottom of the model was set as no-flow boundary. Similar to the Goswami and Clement (27) study, we used confined-flow approximation to model the top boundary as no-flow boundary. Goswami and Clement (27) have shown that the confined-flow approximation yielded good quality solutions that accurately reproduced all observed saltwater wedge transport processes. The saltwater head at the left boundary was set to 25.1 cm for the entire simulation. Freshwater head was set at 26.3 cm and the freshwater flowed from right to left. Vertically, the flow tank was discretized into 5 confined layers of thickness.5 cm and the boundary heads (i.e., the water levels at both saltwater and freshwater boundaries) always remained slightly above the upper level of the top layer. The value of specific storage (S s )was set to 1 6 cm 1. The values of longitudinal (α L )andtransversal (α T ) dispersivity were initially selected based on the values suggested by Abarca and Clement (29), and were slightly adjusted to fit the neutral tracer transport (Test-1) data; these calibrated values were later verified using dense tracer transport (Test-2) data. A combination of total variance diminishing (TVD) solver for advection and fully implicit solver for dispersion was used to simulate the experimental datasets. The maximum time step size used in these simulations was 2 s. A Courant number constraint of Cr.5 was used in all the simulations. 3. Results and discussions 3.1. Experimental data for characterizing transport above the wedge Fig. 3 shows the data for neutral tracer slug experiment (Test-1). The figure shows photographs of the tracer slug as it travelled from the injection point, located near the freshwater boundary, to the saltwater boundary. The photographs were taken at, 5, 1 and 15 min after injecting the slug. The experimental data indicated that the plume required about 2 min for traversing the entire flow domain. The figure also shows the model simulated plume at different times. In the figure, the salt wedge is identified by white solid line which represents the 5% isochlor. As a part of the simulation study, the dispersion parameters α L and α T were adjusted within an expected range to reproduce the observed plume spreading levels. Based on qualitative comparison, dispersivity values of α L =.5 cm and α T =.5 cm were able to recreate the observed plume spreading patterns. These values are within the range of values reported in a previous study (Abarca and Clement, 29) which used a similar type of flow tank. All the model parameters used for simulating this baseline experiment are summarized in Table 1. Both laboratory and modelsimulated results (see Fig. 3) indicate that the shape and spatial extent of the plume lifted upward as it approached towards the discharge boundary. Overall, the model simulated tracer transport patterns agreed well with the experimental data.

5 18 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) Time = min Time = 5 min Time = 1 min Time = 15 min Fig. 3. Comparison of observed data with numerical predictions for the neutral plume transport experiment (Top figure: laboratory data; Bottom figure: SEAWAT predictions with color filled contour red>2%; yellow 2 14%; green 14 8%; and turquoise 8 1%; also, white line is the 5% saltwater contour of the wedge). The second tracer experiment (Test-2) was completed to primarily assemble a dataset to test model performance using the calibrated parameter values. In addition, the second tracer test also allowed us to test model performance for simulating a dense plume as it sunk and interacted with the saltwater freshwater boundary. All the model parameters including the dispersion coefficients used were identical to the values used for simulating the Test-1 dataset. Fig. 4 shows the experimental and numerical data for the dense tracer plume as it travelled from the injection point to the saltwater boundary. The photographs were taken at, 5, 1 and 15 min after injecting the slug. The data show that the dense plume also took about 2 min to traverse the entire flow tank. As expected, the dense plume initially started to sink and the sinking process continued for about 1 min; during this time, the core of the plume sank by about 1 cm. The numerical model was able to reproduce this sinking pattern. It is interesting to note that as the dense plume approached the salt wedge the flow direction changed and the plume started to rise due to the presence of saltwater wedge that almost acted as a flow barrier. Since the freshwater flow regime was characterized by strong advection, the dense plume remained stable throughout the experiment. This is supported by Kanel et al. (28) work, who studied the transport of a dense plume of nano-particles under strong advection-dominated flow conditions, and their data also showed a stable dense plume transport pattern. Goswami et al. (212) recently published a comprehensive study to investigate the dynamics of sinking and buoyant plumes and demonstrated the role of advection on plume stability Experimental data for characterizing transport within the saltwater wedge The two experiments described above (Test-1 and Test-2) allowed us to estimate the transport parameters for the experimental flow tank; they also provided independent datasets for characterizing the transport patterns occurring above the wedge. The third experiment, Test-3, was used to investigate the transport patterns occurring within the wedge. Before starting Test-3, a steady-state saltwater wedge was established. Typically, it took about 2 h to establish a steady state saltwater wedge within this experimental system (Chang and Clement, 212; Goswami and Clement, 27). However, as pointed out in Chang and Clement (212), adjusting the initial state to simulate receding-wedge conditions allowed us to develop steady-state conditions more rapidly than simulating steady states using intruding-wedge conditions. After obtaining a steady-state wedge, we conducted several screening transport experiments by injecting neutral tracer slugs with salt concentration similar to the ambient saltwater concentration within the wedge. These experiments indicated that even a small density difference between the ambient salt water and the tracer slug would lead to buoyancy effects that forced the injected slug to either sink or rise. Therefore, we carefully matched the density of the green tracer solution to be identical to the ambient saltwater density to form the dense, yet locally neutral, tracer plume. Fig. 5 shows the transport patterns of the injected neutral tracer slug (of about 11 ml dense green water) within the wedge. The digital pictures were recorded at, 6, 12 and Time = min Time = 5 min Time = 1 min Time = 15 min Fig. 4. Comparison of observed data with numerical predictions for the dense plume transport experiment (Top figure: laboratory data; Bottom figure: SEAWAT predictions with color filled contour red>2%; yellow 2 14%; green 14 8%; and turquoise 8 1%; also, white line is the 5% saltwater contour of the wedge).

6 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) min after injecting the slug. The data show that the saltwater slug travelled about 2 cm in 18 min, yielding an average transport velocity of.1 cm/min. On the other hand, the freshwater slug travelled about 5 cm in 2 min, yielding a transport velocity of 2.5 cm/min. The average rate of tracer transport within the wedge is only about 4% of the transport rate observed above the wedge (the difference was about an order of magnitude). The shape of the tracer slug was almost circular at the time of injection (at t= min); however, the circular shape started to degenerate rapidly with time and the slug evolved into an elongated plume of irregular geometry. At later times, a large portion of the plume mass remained at the bottom of the tank with a finger like structure evolving upward. The finger extended towards the saltwater-freshwater interface and it was rapidly flushed out of the system by the strong advective flow field active near the interface. After about 4 to 5 h transport, it was almost impossible to delineate a distinct green plume from digital pictures. However, it took approximately 6 h (estimated based on visual observation of the remnant green residues in the tank) for the tracer slug to fully flush out of the system. Several interesting observations can be inferred from the laboratory dataset. First, the residence time of the tracer released within the saltwater region was estimated to be about 6 h, which is much higher than the residence time of 2 min estimated for the tracer released in the freshwater region. Several previous studies have postulated that the transport rate within a saltwater wedge could be much slower than the transport rate above the wedge (Chang and Clement, 212; Robinson et al., 27a; Smith and Turner, 21; Xin et al., 21); however, this is the first time a laboratory dataset has been provided to quantify this phenomenon. Secondly, the tracer plume was not tracking an idealized circular flow path commonly depicted in textbooks (Fetter, 21). Due to higher level of mixing active near the interface, the plume attained an elongated shape as it approached the saltwater-freshwater interface. It is interesting to note that once the plume reached the interface a diffused fraction the plume was transported rather rapidly along the mixing zone and the plume eventually discharged at the sea-side boundary as a narrow finger instead of a circular slug. Fig. 5 also shows the model predicted results for the experimental system. The overall plume patterns simulated by the numerical model are in good agreement with observed data. The model results also showed an elongated plume that fingered along the saltwater-freshwater interface. In terms of time scales, the numerical model also required about 6 h to fully flush the tracer slug out of the system. It is important note that this is the first attempt to study the fate and transport of a neutral tracer released within a saltwater wedge; as per our knowledge, no one has studied this problem using experimental approaches. Also, while numerical models have been used to study saltwater flow pathways within a steady-state saltwater wedge (e.g., Chang and Clement, 212; Smith, 24) so far no one has simulated transport patterns within a saltwater wedge which are mediated by advection-dispersion processes. The above numerical exercise had a limited scope as it only considered transport patterns occurring with our small-scale experimental flow tank. To further understand the transport patterns within a saltwater wedge occurring at larger scales, we completed additional numerical simulations for a field-scale test problem and the results are summarized in the section below Numerical data for characterizing transport within the saltwater wedge in large-scale systems The field-scale problem considered is a modified version a test problem discussed in Chang et al. (211). The problem considers a two-dimensional confined aquifer which is 1 m long, 3 m deep and 1 m wide. Dispersivity values assumed are: α L =.4 m and α T =.4 m. Porosity was assumed to be.35. The sea level was fixed at 3 m along the left side of the model boundary. A constant amount of regional freshwater flow was allowed to flow from the inland boundary and areal-recharge rate was set to zero. The fully implicit option available in SEAWAT was employed in these simulation experiments. Use of the fully implicit option allowed us to be consistent with previously published studies which also employed this option (Chang and Clement, 212; Chang et al., 211). Also, other more robust advection solvers, such as the TVD or MOC schemes, would require significant amount of computational effort. These advanced schemes could also introduce spurious oscillations in coarse-mesh simulations (Goswami et al., 212) and these oscillations could lead to errors in flux estimates. Therefore, we employed the fully implicit scheme. Uniform finite difference grid with Δx=4 m and Δz=.4 m was used, and unit width was assumed. The time step sizes used in this exercise ranged from.5 to 5 days. Time = min Time = 6 min Time = 12 min Time = 18 min Fig. 5. Comparison of observed data with numerical predictions for the transport experiment completed within the wedge (Top figure: lab data; Bottom figure: SEAWAT predictions with color filled contour red>2%; yellow 2 14%; green 14 8%; and turquoise 8 1%; also, white line is the 5% saltwater contour of the wedge).

7 2 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) Other numerical model parameters used were identical to the ones used in Chang et al. (211) study and the details are summarized in Table 2. Simulations were completed for two different flow conditions that simulated low and high regional groundwater flow scenarios and used the flow rates of.15 m 3 /day and.21 m 3 /day, respectively. These flow rates were selected to simulate relatively lengthy salt wedges that intruded deep into the formation with steady-state toe extending up to 73 m and 39 m from the sea-side boundary under low and high flow conditions, respectively. To simulate tracer transport scenarios, two neutral density tracer slugs were simultaneously released within freshwater and saltwater zones (above and within the wedge) by setting the initial tracer concentration in four grid cells to 1 g/m 3. The multi-species option available in SEAWAT/MT3DMS tool was used to track these two tracer slugs. Fig. 6 shows the tracer transport scenarios predicted under both low and high flow conditions. In the figure, the regional salt wedge is delineated by black solid line which represents the 5% isochlor. The color filled plumes were generated by the contouring software using following color coding scheme: red>.8 g/m 3 ; yellow.8.7 g/m 3 ; green.7.3 g/m 3 ; turquoise.3.5 g/m 3 ; and blueb.5 g/m 3. The figures show the model predicted plumes after 36, 2,, 4,, and 6, days of transport. A closer analysis of the freshwater tracer slug predicted after 2, days of transport show that the center of the slug traversed about 2 m and 4 m under low and high flux conditions, respectively. These travel distances agree well with the theoretical estimates, as explained here. The average pore-scale transport velocity in the freshwater zone can be estimated using the expression: Q f v ¼ nbw where Q f [L 3 T 1 ] is the rate of freshwater flow, n is the value of porosity, B [L] is the thickness of the aquifer, and W is aquifer width [L]. Using Eq. (1), the pore-scale transport velocities under low and high flow conditions are estimated to be 1 cm/day and 2 cm/day, respectively. Therefore, after 2, days of transport the plumes must travel about 2 m and 4 m under low and high flow conditions, respectively. These estimates agree well with the simulated transport data shown in Fig. 6. The figure also Table 2 Model parameters used for simulating the field problem (modified from Chang et al., 211). Parameter Symbol Value Horizontal aquifer length L 1 m Vertical aquifer thickness B 3 m Inland groundwater flow rates Q 1 (low flux).15 m 3 /day Q 2 (High flux).21 m 3 /day Hydraulic conductivity K 1 m/day Specific storage, S s m 1 Longitudinal dispersivity α L.4 m Transverse dispersivity α T.4 m Saltwater concentration C s 35 kg/m 3 Saltwater density ρ s 125 kg/m 3 Freshwater density ρ f 1 kg/m 3 Density slope E.714 Porosity n.35 ð1þ shows that, at later times, the salt wedge acted almost like a no-flow boundary to freshwater flux which was pinched to flow through a narrower region above the wedge. Due to this converging flow condition, the freshwater slug started to lift and travelled slightly faster above the wedge. Comparison of the two tracer slugs shows that the transport rate within the wedge was considerably slower than the transport rate above the wedge. The saltwater plume released within the wedge travelled only by about 15 m and 23 m under low and high flux conditions, respectively, after 4, days of transport. This yields effective transport velocities of.375 and.575 m/day, which are about 38 and 28% of the freshwater flow velocities estimated for the two systems (note the theoretical estimates of freshwater velocities are 1 and 2 m/day). Once the saltwater tracer plume reached the saltwater-freshwater interface, it fingered along the interface and was removed rapidly from the system. The transport patterns predicted by the numerical model for this field-scale problem are quite similar to those observed in our small-scale experimental flow tank. The simulation results also show that while the center of tracer slug appears to follow a circular flow path at earlier times, once the slug reached the saltwater freshwater interface most of the contaminant mass was transported rapidly along the interface forming an elongated finger-like structure. Both low and high flow simulation results indicated elongated plume transport patterns. To further understand the plume elongation process, we plotted model-simulated plume contours along with modelsimulated velocity vectors occurring above and below the saltwater wedge under high flow conditions after 4, days of transport (see Fig. 7). The velocity vectors shown in Fig. 7 clearly demonstrate that the transport rates active near the saltwater freshwater interface are much higher than the transport rates active within the wedge. As the plume approached the interface it encountered more active advection and dispersion processes that rapidly mixed the plume and formed an elongated finger-like structure along the interface. Finally, we used the numerical simulation results to analyze the amount of saltwater flow recirculated within the wedge. The amount of recirculated saltwater flow is an important parameter that will influence contaminant transport rates (both advection and dispersion processes) occurring within the wedge. Analysis of the steady-state flow budgets for the field problem indicated that the saltwater flow recirculated within the low and high flux systems are.69 and.96 m 3 /day, respectively. The recirculated saltwater flow rate is about 65 to 45% of the fresh water flow rate (note, the assumed freshwater flow rates are:.15 m 3 /day and.21 m 3 /day). The flow budget for the small-scale experimental problem indicated that freshwater and recirculated saltwater flow rates are m 3 /day, m 3 /day, respectively; this data show that the recirculated saltwater flow in the experimental flow tank is extremely low, about 1.4% of the freshwater flow. It is interesting to note that the percentage of recirculated saltwater flow estimated for the field-scale problem was considerably higher than the recirculated flow estimated for the small-scale experimental problem. One of the major differences between the small-scale and large-scale problems is the assumed level of mixing (or dispersivity values). Smith

8 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) (a) 3 (b) day , days , days , days x (m) x (m) Fig. 6. Model simulated results illustrating the transport processes occurring above and below the saltwater wedge for the field-scale test problem: (a) simulations for low regional flow of.15 m 3 /day and (b) simulations for high regional flow of.21 m 3 /day. Color filled contours represent concentration in g/m 3. (24) pointed out that the saltwater flux circulated within the system under steady-state conditions is a strong function of the level of mixing active in the system. Furthermore, the saltwater intrusion problems are also sensitive to the type of boundary condition used to deliver freshwater into the system (Chang and Clement, 212); note, typical options include areal-recharge-flux and regional-flux boundary conditions. Therefore, we performed sensitivity simulations to quantify the sensitivity of the recirculated saltwater flows to variations in mixing conditions (dispersivity values) and boundary conditions, and the results are summarized below Sensitivity of recirculated saltwater flux to mixing conditions and boundary conditions We first completed numerical experiments to characterize the impacts of mixing (dispersion) on the field-scale problem for different boundary flux values. Simulations were completed by varying the regional freshwater inflow rate from.15 to 1.8 m 3 /day; the models were run until steady state conditions were reached. Two sets of simulations were completed using longitudinal dispersivity values of.4 m and 4 m; the ratio of the transverse to longitudinal dispersivity value was set to.1. Furthermore, additional simulations were completed using identical model parameters and freshwater fluxes, but with the boundary condition modified to simulate areal recharge conditions. Essentially, instead of injecting the freshwater flow from the right regional boundary, we distributed the same amount of flow from the top of the model to simulate the areal-recharge-flux boundary condition. The regional flux was set to zero in these simulations. Chang and Clement (212) completed laboratory experiments to compare saltwater intrusion problems involving regional and areal recharge boundary conditions. Their laboratory 3 z (m) x (m) Fig. 7. Model simulated results illustrating the velocity vectors at 4, days for the high-regional flow test problem. Color contours vary from red at the concentration of 1.4 g/m 3 to blue at the concentration of.15 g/m 3.

9 22 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) data indicated that the wedge in the regional-flux experiment shaped like a liner function, whereas the wedge in the arealflux experiment was relatively thin and its shape looked like an exponential curve. However, the impacts of these physical differences in salt wedge patterns (which were influenced by boundary conditions) on saltwater recirculation flows have not been investigated in Chang and Clement (212) study. Fig. 8 summarizes the results of the sensitivity simulations. Firstly, as expected, the amount of saltwater flow entering system was proportional to the amount of freshwater flow entering the system. Furthermore, when the dispersivity value was increased, the amount of saltwater flow circulated within the system also increased. This is because, higher dispersion resulted in more mixing and hence flushed out relatively large amount of saltwater from the system; this resulted in driving a cyclic process that forced more saltwater to enter the system. Interestingly, the relationship between saltwater and freshwater flows was not sensitive to the type of boundary condition used for modeling the freshwater inlet boundary (i.e., the location where the freshwater flow was allowed to enter the system). The results show that both regional- and areal-flux boundary conditions resulted in recirculating similar amounts of saltwater flow. We completed a non-dimensional analysis to generalize some of these results using an approach proposed by Smith (24). In order to normalize the relationship between freshwater and saltwater inflows, we used a following version of a dimensionless freshwater flux parameter proposed by Smith (24): V ¼ βkbw Q f Where β is the fluid density ratio, K [LT 1 ] is the hydraulic conductivity, B [L] is the saturated aquifer depth, Q f [L 3 T 1 ] is the rate of freshwater inflow, and W is the aquifer width of the aquifer [L]. As pointed out by Smith (24), the parameter V* is the ratio of free convection to forced convection active within the system. Also, V* is simply the inverse of the ð2þ non-dimensional parameter a commonly used in Henry problems (Henry, 1964; Simpson and Clement, 24). The amount of saltwater circulated within the system was quantified using the following dimensionless saltwater to freshwater flow fraction [defined as the percent saltwater circulation (PSC) in Smith (24) study]: PSC ¼ 1 Q s Q f ðorþ 1 V s V f where Q s [L 3 T 1 ] is the average rate of saltwater flow entering from the sea-side boundary. Because all results were steady state, Q S can be obtained from the flow budget as the total flow that entered from the sea-side boundary over a fixed time period. Also note that PSC is simply the fraction of the total volume of saltwater (V s ) to the total volume of freshwater (V f ) transmitted within the system over a given time period. Fig. 9 presents the flow data reported in Fig. 8 using the non-dimensionless format described by Eqs. (2) and (3). Visualizing the recirculated flows in this form allowed us to compare the current data with other literature-derived data (which are marked using various symbols in Fig. 9). The results indicate that the percentage of saltwater flow circulated within the system initially scales almost linearly with the non-dimensional freshwater flux parameter V*. However, similar to Smith (24) study, the relationship between PSC and V* shows a non-linear trend at higher V* values; also, the relationship is a strong function of the mixing level (or assumed dispersivity values). Furthermore, the values V* and PSC predicted for the current problem are within the range of values observed in other published problems that involved different scales. One the interesting observations inferred from this flow budget analysis is that while the amount of recirculated saltwater flow in our experimental flow tank was estimated to be about 1% of the freshwater flow, the amount of recirculated flow predicted for the field-scale problem ranged from 2% to 12% of the freshwater flow; also, these recirculated flow percentages varied considerably with the assumed level of ð3þ Saltwater Flow Q s (m 3 d -1 m -1 ) Regional flux - Dispersivity.4 Regional flux - Dispersivity 4 Areal flux - Dispersivity.4 Areal flux - Dispersivity Inland Freshwater Flow Q f (m 3 d -1 m -1 ) Fig. 8. Relationship between freshwater flow and recirculate-saltwater flow for regional and areal recharge problems with different dispersivity values.

10 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) mixing. Further field and/or laboratory studies should be completed at high dispersion conditions to verify whether one could realistically sustain high levels recirculation within a saltwater wedge. 4. Conclusions In this effort, we have completed both laboratory and numerical experiments to study contaminant transport patterns occurring above and within a saltwater wedge. This study is novel in two important ways: firstly, as per our knowledge, this is the first time tracer plume transport within a saltwater wedge has been investigated using laboratory experiments. Secondly, this is the first time the fate and transport patterns of a tracer plume within a wedge, mediated by advection and dispersion processes, have been analyzed in detail using numerical simulation. The experimental data show that the time scales associated with the transport processes occurring below the wedge are much slower than the transport processes occurring above the wedge. The experimental dataset indicated that it took approximately 6 h to fully flush the tracer slug when it was released within the wedge; on the other hand, it took only about 2 min to flush the tracer slug when it was released above the wedge. The experimental data also show that the tracer slugs released below the wedge did not follow an idealized circular flow path. Due to dispersion effects, the tracer slug attained an elongated shape as it approached the saltwater freshwater interface. Higher velocities active near the interface rapidly transported a portion of the plume mass along the mixing zone forming this elongated fingering pattern. The modeling results for both small and large scale systems also predicted similar type of elongated transport patterns. The modeling data also confirmed that dispersion plays a significant role in controlling the amount of saltwater flow recirculated within a wedge. Model simulations show that when mixing across the wedge was allowed to increase (using larger dispersivity values) the amount saltwater flow circulated within the wedge also increased. The percentage of recirculated saltwater flow in a system appears to scale well with freshwater flux for variety of problems of different scales when the results are plotted in a dimensionless format proposed by Smith (24). The insights gained from this study help us better understand the solute exchange processes occurring between terrestrial and marine waters. The current experimental study and Smith (24) numerical study, however, focused only on understanding steady-state systems with a constant sea level boundary condition. Further studies are needed to understand transient systems influenced by dynamic sea-side boundary conditions impacted by tidal cycles and other long term changes such as sea level rise. It will also be interesting to quantify transport within the wedge under different mixing conditions that are influenced by fluctuations in boundary conditions, changes in recharge fluxes, and other perturbations arising from spatial heterogeneities. The experimental data collected from our small scale experimental flow tank show that the saltwater transport rates and the associated recirculated flows are almost two orders of magnitude smaller than the rates observed within the freshwater region. On the other hand, the theoretical modeling results presented here, as well as in Smith (24), postulate that the differences in the flow and transport rates would depend on the level of dispersion. When dispersion is high the saltwater flow could be comparable to (or even be higher than) the freshwater flow, which implies the transport rate within thewedgecouldbeashighorevenmuchhigherthanthe transport rate occurring above the wedge. It will be interesting and worthy to test this theoretical hypothesis using laboratory or field-scale observational datasets. 12% Regional flux - Dispersivity.4 Percent Saltwater Circulation PSC (%) 1% 8% 6% 4% 2% Regional flux - Dispersivity 4 Areal flux - Dispersivity.4 Areal flux - Dispersivity 4 % Dimensionless flux parameter V * Fig. 9. Non-dimensional analysis of the relationship between freshwater and saltwater fluxes [The symbols indicate the following literature data points: from the tank experiment of this study, + from Chang et al. (211) study, regional flux experimental problem from Chang and Clement (212), areal flux experimental problem from Chang and Clement (212), Cyprus case study from Destouni and Prieto (23), Israel case study from Destouni and Prieto (23)].

11 24 S.W. Chang, T.P. Clement / Journal of Contaminant Hydrology 147 (213) Acknowledgments This work was, in part, supported by the Samuel Ginn College of Engineering Dean Fellowship awarded to Sun Woo Chang at Auburn University. We thank the two anonymous JCH reviewers for providing constructive comments that greatly improved the quality of this manuscript. References Abarca, E., Clement, T.P., 29. A novel approach for characterizing the mixing zone of a saltwater wedge. Geophysical Research Letters 36 (6). Boufadel, M.C., 2. A mechanistic study of nonlinear solute transport in a groundwater surface water system under steady state and transient hydraulic conditions. Water Resources Research 36 (9), Brovelli, A., Mao, X., Barry, D.A., 27. Numerical modeling of tidal influence on density-dependent contaminant transport. Water Resources Research 43 (1), W1426. Chang, S.W., Clement, T.P., 212. Experimental and numerical investigation of saltwater intrusion dynamics in flux-controlled groundwater systems. Water Resources Research 48 (9), W WR Chang, S.W., Clement, T.P., Simpson, M.J., Lee, K.-K., 211. Does sea-level rise have an impact on saltwater intrusion? Advances in Water Resources 34 (1), Destouni, G., Prieto, C., 23. On the possibility for generic modeling of submarine groundwater discharge. Biogeochemistry 66 (1), Fetter, C.W., 21. Applied Hydrogeology (Prentice Hall, N.J.). Goswami, R.R., Clement, T.P., 27. Laboratory-scale investigation of saltwater intrusion dynamics. Water Resources Research 43 (4). dx.doi.org/1.129/26wr5151. Goswami, R.R., Clement, T.P., Hayworth, J.H., 212. Comparison of numerical techniques used for simulating variable-density flow and transport experiments. Journal of Hydrologic Engineering 17 (2), Guo, W., Langevin, C.D., 22. User's guide to SEWAT: a computer program for simulation of three-dimensional variable-density ground-water flow. United States Geological Survey. Henry, H.R., Effects of dispersion on salt encro achment in coastal aquifers (1613-C). Illangasekare, T., et al., 26. Impacts of the 24 tsunami on groundwater resources in Sri Lanka. Water Resources Research 42 (5), W521. Johannes, R.E., 198. The ecological significance of the submarine discharge of groundwater. Marine Ecology Progress Series 3, Kanel, S.R., Goswami, R.R., Clement, T.P., Barnett, M.O., Zhao, D., 28. Two dimensional transport characteristics of surface stabilized zero-valent iron nanoparticles in porous media. Environmental Science & Technology 42 (3), Kohout, F.A., 196. Cyclic flow of salt water in the Biscayne aquifer of Southeastern Florida. Journal of Geophysical Research 65 (7), Kuan, W.K., et al., 212. Tidal influence on seawater intrusion in unconfined coastal aquifers. Water Resources Research 48 (2), W252. Li, L., Barry, D.A., Stagnitti, F., Parlange, J.Y., Submarine groundwater discharge and associated chemical input to a coastal sea. Water Resources Research 35 (11), Moore, W.S., Large groundwater inputs to coastal waters revealed by 226Ra enrichments. Nature 38 (6575), Moore, W.S., Church, T.M., Submarine groundwater discharge, reply to Younger. Nature 382, 122. Robinson, C., Li, L., Barry, D.A., 27a. Effect of tidal forcing on a subterranean estuary. Advances in Water Resources 3 (4), Robinson, C., Li, L., Prommer, H., 27b. Tide-induced recirculation across the aquifer ocean interface. Water Resources Research 43 (7), W7428. Simmons, G.M., Importance of submarine groundwater discharge (Sgwd) and seawater cycling to material flux across sediment water interfaces in marine environments. Marine Ecology Progress Series 84, Simpson, M.J., Clement, T.P., 24. Improving the worthiness of the Henry problem as a benchmark for density-dependent groundwater flow models. Water Resources Research 4 (1), W154. Simpson, M.J., Clement, T.P., Gallop, T.A., 23. Laboratory and numerical investigation of flow and transport near a seepage-face boundary. Ground Water 41 (5), Smith, A.J., 24. Mixed convection and density-dependent seawater circulation in coastal aquifers. Water Resources Research 4 (8), W839. Smith, A.J., Turner, J.V., 21. Density-dependent surface water groundwater interaction and nutrient discharge in the Swan Canning Estuary. Hydrological Processes 15 (13), Volker, R.E., Zhang, Q., Lockington, D.A., 22. Numerical modelling of contaminant transport in coastal aquifers. Mathematics and Computers in Simulation 59, Westbrook, S.J., et al., 25. Interaction between shallow groundwater, saline surface water and contaminant discharge at a seasonally and tidally forced estuarine boundary. Journal of Hydrology 32, Xin, P., Robinson, C., Li, L., Barry, D.A., Bakhtyar, R., 21. Effects of wave forcing on a subterranean estuary. Water Resources Research 46 (12), W1255. Zhang, Q., Volker, R.E., Lockington, D.A., 22. Experimental investigation of contaminant transport in coastal groundwater. Advances in Environmental Research 6 (3),