IIHHIWIHIIHHWIWUHWWIUlltlHllHllHWHl 132 517 THS Ti'iFSIiP. SLIITY 111111111111111111111111 3 1293 016914 This is to certify that the thesis entitled THREE-DIMENSIONAL GROUNDWATER FLOW AND CONTAMINANT TRANSPORT IN MEDIUM SCALE HIGHLY HETEROGENEOUS ENVIRONMENTS presented by Ken A. Ewers has been accepted towards fulfillment of the requirements for M.S. degree in Geological Science L; 2‘7?) /r/:fl WM Major prdfessor Date ////[//C(7 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE 1/98 c'JClHCIDatoDuo.p85-p.14 THREE-DIMENSIONAL GROUNDWATER FLOW AND CONTAMINANT TRANSPORT IN MEDIUM SCALE HIGHLY HETEROGENEOUS ENVIRONMENTS By Ken A. Ewers A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTERS OF GEOLOGICAL SCIENCE Department of Natural Science 1997 ABSTRACT THREE-DIMENSIONAL GROUNDWATER FLOW AND CONTAMINANT TRANSPORT IN MEDIUM SCALE HIGHLY HETEROGENEOUS ENVIRONMENTS By Ken Ewers Laboratory hydrology studies are often limited to one of two dimensions and often suffer from edge flow, which results in data that is not characteristic of the porous media. A large, minimally disturbed, heterogeneous aquifer sample in a laboratory allows the Study of small-scale processes in a controlled setting. We present the design of this collection apparatus and uSe it to collect and study a large sample, called a mesocosm, from a glacial debris-flow deposit. A Simulation regression algorithm was used to estimate effective aquifer properties and produce a numerical model which is used to illustrate the it is crucial to understand small scale processes that result from aquifer heterogeneity in order to accurately describe data collected over larger scales. Additionally, high-resolution tracer data is used to investigate effects of diffusion on flow and transport of two conservative tracers with different molecular weights. The data is used to demonstrate that fluorescein travels faster than tritium in highly heterogeneous environments as a result of its higher molecular weight which caused it to also be less diffusive. ACKNOWLEDGEMENTS A very special thanks to Dave Hyndman, Grahame Larson, Kaz Fujita, Dave Boutt, Robin Sutka, Sandy Treccani, Nathaniel Ostrom, and Peggy Ostrom. Without your help this study could not have succeeded. iii TABLE OF CONTENTS Background...... .. . Mesocosm Design... Numerical Model... Darcy’s Law (Equation 1....1) . Darcy’s Law (Equation 12.) Tracer Tests... Dimensionless Time Equaiton (Equation 2.1)... .. First Fluorescein Tracer Test... .. Tritium Tracer Test... Second Fluorescein Tracer Test... ..... . ..Vi vii 12 19 .19 .21 .23 24 .26 .27 3O 32 .33 .36 .38 ..38 Slug Tests... .. .. .. OBSERVATIONS/RESULTS... .. Saturated Porous Media Diffusion Coefficient Equation (Equation 2.2) . .. .. SUMMARYAND CONCLUSIONS... .. APPENDIXA - Slug Test Data... APPENDIX B - Tracer Concentration Histories... .39 .42 .50 53 ..58 .72 LIST OF TABLES 1.1 Well depths, measured head drops, and tracer arrival times... 1.2 General tracertest information... .. .. 1.3 Estimated parameters of the homogeneous model...... .. 2.1 Well depths, measured head drops, tracer arrival times, and estimated hydraulic conductivity... 2.2 General information about tracer tests... .... .... 2.3 Percent change of fluorescein arrival... vi 11 13 18 .34 36 .46 LIST OF FIGURES 1.1Mesocosm Design... .. 1.2 Fluorescein Concentration Histories (sample ports 1-5)... .. . 1.3 Fluorescein Concentration Histories (sample ports 6-10)... 1.4 Fluorescein Concentration Histories (sample ports 11-15)... . 1.5 Measured vs. Simulated Concentration History of Effluent... 2.1 Diffusive Retardation... .... 2.2 Fluorescein Molecular Structure... 2.3 Well Attachment to Mesocosm... .... .... .. 2.4 Peak Concentrations... . 2.5 Center ofMass Arrival Times... .. .... 2.6 Correlation of Hydraulic Conductivity and Percent Change of Tracer Arrival... vii .. 15 .16 .. 17 ...18 ..31 37 .. .41 .. .43 .. 45 ..47 CHAPTER 1 INTRODUCTION Contaminant and tracer transport studies in heterogeneous environments have been limited by inadequate definition of aquifer boundaries. One approach to overcome these difficulties is to collect a large intact sample of a heterogeneous aquifer that can be studied in a controlled laboratory environment. This paper presents a laboratory apparatus that allows for the analysis of three-dimensional groundwater flow, and solute transport in minimally disturbed aquifer material under controlled stresses and boundary conditions. A large minimally-disturbed sediment sample, which we call a mesocosm, can be used to explore the effects of medium scale heterogeneities on groundwater flow and contaminant transport. Small scale laboratory and field scale studies are both unable to address these intermediate scale heterogeneities. Laboratory studies Often use homogeneous media or hand packed sediments (Skibitzke and Robetson, 1963; Silliman and Simpson, 1987; and Silliman et al., 1987), and fail to explore three-dimensional transport (in a heterogeneous environment. Conversely, field experiments are generally incapable of addressing the processes that influence sub-meter scale transport because of an inability to densely sample the system (Knopman et al., 1991) and subsequent uncertainty in the flow field. Additionally, field studies typically have poor conservation of tracer mass in heterogeneous environments where plumes may be elusive (Garabedian et al., 1991). A large minimally disturbed mesocosm allows for a more rigorous study of solute transport and sorption in these environments. One of the most significant advantages of the mesocosm approach is the ability to verify estimates of aquifer properties. Geophysical and hydrogeological studies Often estimate subsurface structure or properties but can not quantify errors because measurements are sparse and local. The mesocosm could offer an inexpensive means of characterizing aquifer properties over an entire domain. At the completion of an experiment, the mesocosm could be dismantled and physically characterized with any desired sampling density. High-resolution sampling would provide direct measurements of the properties of interest, which could be compared to the estimated values to provide true estimation errors. The tracer data illustrates the need for high-resolution parameter estimates in order to accurately model solute transport. A numerical model based on reasonable aquifer parameters can not adequately reproduce observed concentration histories at discrete locations without knowledge of the geologic structure of the aquifer, even with accurately defined boundary conditions and aquifer stresses. Background Small-scale column experiments are commonly used to estimate biodegradation rates and sorption parameters for contaminants (Siegrist and McCarty, 1987; Lanzarone and McCarty, 1990). Barone et al. (1992) used small (1.6centimeter thick by 2.6-centimeter diameter), undisturbed clay plugs to estimate diffusion and adsorption coefficients for a suite of volatile organic species. Myrand et al. (1992) studied the diffusion of several volatile organic compounds in a slightly larger (10-centimeter long by 5-centimeter diameter) diffusion-cell apparatus. These studies have provided useful information about physical and chemical processes at small (centimeter) scales. A variety of studies have explored transport through medium scale systems. Vanclooster et al. (1995) monitored vertical solute transport through a 1m tall by 0.8 m diameter multi-layered sandy lysimeter using Time Domain Reflectometry (TDR). Silliman and Simpson (1987) and Silliman et al. (1987) used a 1.07 m tall by 2.13 m long by 0.1 m wide sandbox with various sand-packing arrangements to study the scale effect Of heterogeneities on solute transport and longitudinal transport in two dimensions. Silliman et al. (1992) used this two-dimensional sandbox design to study vertical and longitudinal dispersion. Wood et al. (1994) studied two-dimensional contaminant transport and biodegradation in layered media with a 1m long by 0.2 m tall 0.1 m wide hand-packed sandbox. These studies have provided hydrologists with significant information on the effects of heterogeneities on flow and transport. Few studies however, have characterized three-dimensional transport in a controlled setting. Tucker and Kuper (1996) used a very dense sampling network to characterize a section of the Borden aquifer and concluded that hydraulic conductivity can not be characterized by a single spatially invariant statistical distribution and illustrated the need for characterization of heterogeneities. Corwin and LeMert (1994) studied vertical contaminant transport with very large, but disturbed soil lysimeters. In contrast to these studies, a large minimally disturbed mesocosm offers a highly controlled environment to study the effects of three-dimensional heterogeneities on contaminant transport, and is more applicable to phenomenon observed at the field scale. Mesocosm Design A large sediment sample was collected and transported to the laboratory with minimal disturbance to the physical structure. The sediment sample was collected from a shallow glacial debris flow deposit (Grahame Larson, personal communication) in East Lansing, Michigan. Numerous lithologic heterogeneities are present in this sample, which contains approximately 60% clay, 20% silt, 20% sand (based on visual inspection), with some larger cobbles. This deposit was chosen because of its high degree of sediment variability, and its proximity to the laboratory. A section of commercially available PVC (polyvinyl chloride) sewer pipe was machined to be the collection tube for the aquifer mesocosm. The tube has a 44.5- centimeter inner diameter, 109.0—centimeter length, and a wall thickness of 1.4 centimeters. This wall thickness is strong enough to maintain shape under heavy stress yet thin enough to minimize the force required to press the mesocosm apparatus into the sediment for collection. This PVC pipe makes a good collection apparatus because it is fairly inert, and it is easy to Obtain. Rigid, thick walled PVC pipe was used as the collection and housing apparatus, to allow for sampling in a variety of soil types varying from coarse sand to clay. Larger grained sediment types, such as gravel sized grains, can be collected but, this may result in a pipe rupture or difficulties creating watertight seals due to damaged pipe ends. A common limitation of small column studies is conduit flow between the sample and the collection tube walls, termed edge flow. Edge flow is reduced with our design by: 1) collection of a sample with no annular gap between the walls of the housing and the aquifer sample, 2) collection of the sample at depth (approximately 3.5 m below ground surface), and 3) reduction of flow to the edges through the design of the end plates. Collection of the aquifer sample was aided with the cutting tip formed on the end of the PVC pipe (Figure 1.1). The ends of the pipe were beveled with a router to a 45-degree edge to produce this cutting tip. This edge trims Off excess aquifer material as the pipe is pressed horizontally into the aquifer. These machined ends also facilitate sealing the mesocosm collection tube. The mesocosm was collected at approximately four meters depth to provide a sample that is more representative of a deep aquifer than one collected from a lesser depth, due to significant overburden pressure and the location near the water table. Ideally, the sample would have been collected from a de-watered region, however this would have been prohibitively expensive and the residual saturation would likely have caused slumping during collection. The sample was collected from a depth such that the bottom of the sample was approximately 20 centimeters above the dry season capillary fringe. A construction site was selected as the site for collection of the mesocosm because a large Caterpillar 375L excavator could push the collection tube into the side of an excavation. The mesocosm was then carefully excavated from the wall by hand to ensure that soil would not slump away from the ends. With the ends exposed, a large steel blade was used to shave the excess sediments flush with the ends of the tube so that the end plates could be installed in the field. Cables were looped around the mesocosm so that a front-end loader, could lift it out of the ground and place it on a push cart, which was used to move the sample to the lab. Edge flow was further controlled through the design of the PVC end plates that were machined to exclude flow within three centimeters of the pipe wall (Figure 1.1). These end plates were held in place by four 5/8-inch Steel tie-rods and sealed with thin (0.16-cm) medium-hardness Buna-n rubber, which was placed between the end plates and the PVC pipe. The end plates were used to generate constant head boundaries at the mesocosm sample ends by preferentially directing flow to enter and exit the central 75% of the faces (Figure 1.1). This was accomplished by hollowing out a portion of the inflow and outflow plate to create a thin reservoir at the mesocosm ends that would create constant head boundaries and allow fluids to follow the paths of least resistance as they entered and exited the mesocosm. Stainless steel mesh was placed between the PVC end plates and the sediments in the mesocosm to minimize particulate loss from the aquifer sample. Two layers of type 316 stainless steel mesh were used, a lighter gage mesh to reduce sediment loss, and a heavier gage screen to prevent the lighter from tearing. The heavy gage mesh (25 holes per square inch) was placed nearest the sediment, in front of the much finer gage mesh (62,500 holes per square inch). These sheets of mesh were attached to the end plates with Stainless steel screws prior to sample collection. During the study, no fine-grained sediments were Observed in the effluent demonstrating the effectiveness of this mesh. V Direction of flow Figure 1.1 Mesocosm Design. This figure shows the PVC tube in the center, the well locations at the top and the end plates at far left and right. The hatched areas of these endplates indicate recessed portions of the plates used to create constant head boundaries. These recessed areas have an outer diameter that is 4 cm less than the inner diameter of the PVC mesocosm apparatus which causes flow to enter and exit only through the inner 75% of the sediment faces at the mesocosm ends. The end plates are held firmly in place with four tie rods (not shown) attached to the end plates through the four outer holes on each. As the tie rods are tightened the ends seal to the PVC pipe. The gray lines indicate the rubber seals used to prevent leakage at the ends. Stainless steel screen is placed between the mesocosm end plates and the sediment face inside the PVC tube, these are not shown for clarity. End plates were constructed from 1.9-cm (3/4-inch) thick PVC plates in a manner that allows flow enter and exit the mesocosm along preferential flow paths. As shown in Figure 1.1, a portion of the inflow plate was recessed to disperse liquid evenly over the end of the mesocosm. The circular recessed portion is 0.64-cm (1/4-inch) deep with a 19.25- centirneter radius, 3 centimeters less than the inner radius of the PVC tube. Several non- recessed ridges of PVC were left radiating inward to hold the screen against the sediments in the sample and to prevent sediment slumping into the recessed portions. The outflow plate was also composed of a 1.9-cm (3/4-inch) thick PVC plate, with five equal-area (125 cmz), 0.64-cm (1/4-inch) recessed portions as shown in Figure 1.1. Each of these recessed areas provides an areally averaged sample. Each of these areas serves a three-way connector with a sampling valve, then flow proceeds to a six-way coupler which merges the five outflow lines into one effluent line to provide samples representative of the entire mesocosm outflow. A peristaltic pump was set at a constant flow of 5 ml/ min to induce the groundwater gradient (0.18) through Size 14 platinum-cured silicone pump tubing attached to the inflow plate. This flow rate was chosen to achieve a realistic average groundwater velocity and was maintained for 30 days prior to the tracer experiment to saturate the mesocosm under a steady state flow field. A peristaltic pump was selected because it can deliver a low flow rate and functions well with moderate operating pressures. The pump was connected to the mesocosm with 0.32 cm (1/ 8 inch), type 316, stainless steel tubing. This tubing was selected for its resistance to corrosion, but if necessary could have been substituted with other materials. Compression fittings, constructed of 316 stainless Steel, were used to connect the 10 pump to the inflow plate. Additionally, temperature was controlled in the laboratory to reduce any density effects caused by density effects caused by temperature fluctuations. Miniature wells were installed to allow tracer sampling from within the mesocosm. The-sampling wells were made from 0.16-cm (1/ 16-inch) outer diameter type 316 stainless steel tubing. The tip of each well was crimped to aid installation and prevent clogging, and well screens were made by drilling small holes across one-centimeter intervals near the base of each well using a jewelers drill. Each well was marked for the desired depth (Table 1.1) and installed by pressing it into the mesocosm through the pre-drilled holes in the PVC tube (Figure 1.1). Each of these wells was located within one of the two planes shown in Figure 1.1. The predrilled holes were protected during collection of the mesocosm with flush- mount setscrews. The wells were sealed in place after installation with high-pressure PEEK (poly-etheretherkeytone) fingertight fittings. 11 Table 1.1: Well depths, measured head drops, and tracer arrival times. Well or Well Measured Head Center of Mass Sample Depth (cm) Arrival Time Port (cm) (days) Inflow — 100 - 2 9.6 cm. 87.5 5.985 3 24.1 cm. 85.8 9.056 4 9.6 cm. 86.9 2.694 5 9.6 cm. 80.8 7.807 6 9.6 cm. 85.6 3.412 7 9.6 cm. 84.0 12.752 8 23.7 cm. 85.7 1.390 9 9.6 cm. 85.7 14.083 10 9.6 cm. 81.3 7.150 11 - 62.0 2.520 12 - 62.0 3.746 13 — 62.0 3.930 14 - 62.0 2.902 15 - 62.0 3.323 Outflow - 62.0 2.850 METHODS Tracer concentration data were used to evaluate the performance of the mesocosm design and explore the impact of small-scale heterogeneities (Figure 1.2) on flow and transport. First, the fluorescein data were used to produce concentration histories and illustrate the impact of lithology heterogeneity on groundwater flow and solute transport at small to medium scales. The data were also used to illustrate that the mesocosm design successfully eliminated edge flow. Finally, this data was used to estimate effective aquifer parameters for a numerical flow and transport model. Fluorescein dye was added to the inflow of the mesocosm as the tracer for this study. A volume of two liters of this tracer was injected under a steady state flow gradient that was maintained for the duration of the study. This provides approximately six hours of fluorescein tracer pulse at the inlet. Details on groundwater flow rate and injection concentrations can be found in Table 1.2. 12 13 Table 1.2: General tracer test information Fluorescein Initial Concentration (Co) 100 ppm Injection Period 400 minutes Average Pump Rate 5 ml/min Tracer Volume 2 liters Average Tracer Velocity 0.6 m/ day Mass Conserved 74.1% Tracer concentrations were monitored at observation wells and the mesocosm outflow to infer transport processes in the mesocosm sediments. A spectrophotometer was used to analyze 2-mL samples drawn from each well for the fluorescein concentration. An hourly sampling schedule was selected to optimize temporal resolution without adding significant stresses to the hydrodynamics of the system. Initially, samples were to be taken from each of 15 sampling locations, but due to the extremely low production at well 1 due to low conductivity material around the well screen, it could not be sampled. Two ml samples were collected from the remaining 14 locations for 15 days, at which time the scheduled sampling frequency was progressively reduced for 60 additional days. The first 20 days of this data set are illustrated as tracer concentration arrival histories for the 14 sampling locations in Figures 1.2, 1.3, and 1.4. Tracer data were used to estimate aquifer properties such as average tracer and groundwater velocity, effective porosity, and hydraulic conductivity. Average hydraulic 14 conductivity and effective porosity were first estimated using Darcy’s Law, assuming the mesocosm to be homogeneous. Estimates were based on the rate of effluent from the column, solute arrival times in the column effluent, and measured heads in the mesocosm (Table 1.1). The volumetric flow rate was maintained at 5 mL/ min, the gradient was calculated from the head measurements in Table 1.1, and the cross-sectional area was measured. Finally, the arrival time of the center of fluorescein mass in the effluent (Figure 1.5) was used to calculate the average groundwater velocity and, therefore, estimate effective porosity. This method assumes a non-skewed arrival history, and therefore the estimated porosity was later updated. Three-dimensional groundwater flow andsolute transport simulations were used to estimate aquifer parameters for the mesocosm. Simulations were completed using a 27 by 27 by 59-cell, evenly Spaced, rectangular grid system in MODFLOW (McDonald and Harbaugh, 1988) and MT'3D (Zheng, 1992) with no flow boundaries set at the tube edges and constant head cells placed at the tube ends. Table 1.1 contains the measured heads used at these boundaries. Estimates of the Langmuir sorption constants, porosity, and dispersivity (Table 1.3) were developed by iteratively minimizing squared residuals, or differences, between the simulated and measured effluent tracer concentration histories using a simulation regression approach Wagner and Gorelick, 1987). Hydraulic conductivity was estimated previously with the center of mass arrival time in the effluent and Darcy’s law. Figure 1.5 shows the modeled concentration data compared to the measured effluent concentration data. It should be noted that this sorption model was used to fit the tracer data, and was not based on typical or expected sorption constants for this type of media. 0'2 ; Well 4 C/C0 ; 0.0 0 Time (days) TTTT117711jTT—TTT1YTI 20 0.0005 0.01 . C/C0 : 0.00 ' 0.01."" j C/C0 _‘ 15 0.00 C/C0 0.00‘ 0 Time (days) 20 0.01. C/C0} 0.00 IT.VVTTI'VVII WellZ ‘ a b It > d f '4 0 Time (days) 20 Figure 1.2 Fluorescein Concentration Histories (sample ports 1-5). The vertical axes are exaggerates to indicate most Significant concentrations, and the concentrations are normalized by the injection concentration (100 ppm). Well 5 has concentrations near at or below the sensitivity threshold of the spectrophotometer used to analyze concentrations. 0.1 . 0.0 0.02111llT1TIIIIIIIIIII 1.0' C/C0 l 0.00 ”U 0 Time (days) 20 0.01 , C/C0 T 0.0 16 C/C0 ’. « C/C0: llllllllLLllllllll Time (days) 20 Time (days) 20 Figure 1.3 Fluorescein Concentration Histories (sample ports 6-10). The vertical axes are exaggerates to indicate most significant concentrations, and the concentrations are normalized by the injection concentration (100 ppm). 0.12 3' ' CICO' 0.00 ’ ITITITIIIIIIIIITII 0.12_ Outflow l4 C/C0 0.00 ' 0 Time (days) 20 0.12 . C/C0 _" 0.00 17 [ITIIITITITTTI Outflow 11 lTjI lellllllllll‘l Outflow 13 TFTWITT TTrTjleTTTT Outflow 15 Time (days) 20 TTTiTr lllllllll Outflow 12 “...... Time (days) 20 Figure 1,4 Fluorescein Concentration Histories (sample ports 11-15). The vertical axes are exaggerates to indicate most significant concentrations, and the concentrations are normalized by the injection concentration (100 ppm). These samples were collected over 125 cm2 areas at the outflow end. 18 12 . - Simulated "‘ Measured Concentration (ppm) 0 Time (days) 30 Figu_re 1.5 Measured and Simulated concentration history for the mesocosm effluent. This simulation uses the homogeneous effective aquifer properties found in Table 1.3. Table 1.3 Estimated parameters of the homogeneous model. Parameter Va_lue Effective Hydraulic Conductivity 1.3e-4 m/s Effective Average Porosity 6.8 °/o Longitudinal Dispersivity 9 cm Langmuir Sorption Site Concentration 0.99 (dimensionless) Langmuir Sorption Equilibrium 0.83 m3 / day Concentration RESULTS Leg Heterogeneous properties within the mesocosm have a strong influence on the measured concentration data (Figures 1.2, and 1.3). The locations with the most significant concentrations in Figures 1.2, 1.3, and 1.4 have an asymmetric shape characterized by a rapid rise to peak and a long tailing period. Asymmetry of the concentration histories, or tailing, is often considered to be due to sorption, but since fluorescein is fairly conservative (Zahn and Behrens, 1992), this asymmetry could result from heterogeneities in lithology. The observed asymmetry is likely indicative of a heterogeneous environment where high conductivity regions control the peak arrival time and diffusion into and out of lower conductivity regions, cause tailing after the pass of the conservative tracer peak. The importance of heterogeneity is also evident at small scales by comparing the measured concentration histories (Figures 1.2, 1.3, and 1.4). Figure 1.2 shows the first four wells in the mesocosm, referred to as the first plane of wells. In this first plane, significant concentrations were only encountered at well 4. Similarly, in the second plane (Figure 1.3) significant concentrations were encountered in only two locations (wells 6 and 8). Finally, in the last plane, where areally averaged dataare presented, all five concentration histories have different peak heights. These observations indicate that over a separation as small as 5 centimeters, the separation between screens in wells 1 and 3, concentrations vary radically 19 20 depending on the degree of heterogeneity present and the hydraulic conductivity of the material penetrated by the well screen. Groundwater flow and solute transport is primarilly controlled by the highest conductivity paths through the mesocosm. The concentration history from well 8 (Figure 1.3) Shows that the bulk of tracer transport occurs through a region that was only encountered at one well location. This demonstrates the importance of heterogeneities and shows that a great deal of transport can take place through a region small enough that it may go undetected even in a dense sampling network. The tracer data also shows that edge flow was minimized in the mesocosm. If edge flow had occurred during the experiment, a sharp peak in concentration histories and the center of mass would have arrived much earlier at sampling points 11, 12, 14, and 15 than at port 13, indicating preferential transport of solute in a conduit along the walls of the pipe. However, peak arrivals in the concentration histories at these locations did not occur at times appreciably different from that for sample port 13. Additionally, these concentration histories (11, 12, 14, and 15) were similar in shape to concentration histories associated with the wells. This indicates that the properties encountered along the flow paths for 11, 12, 14, and 15 were not very different than those encountered for port 13, or at wells with high concentrations. Tracer data from the wells also indicate that edge flow was minimized in the mesocosm. Tracer traveled along some flow path to arrive at well 8 at a higher velocity than any other flow path intercepted by a well or sampling point, including those on the outflow 21 plate. Recall that well 8 is screened in the center of the mesocosm and therefore is far from the edges and will yield data representative of the properties of the porous media. This Shows that the highest tracer velocities, and most conductive passageways for the tracer are in the porous media and not at its edges (edge flow). If edge flow was occurring to a large extent, we would expect higher tracer velocities to the outermost outflow ports (11, 12, 14, and 15) than to any well. Numerical Model A simulation-regression algorithm (Wagner and Gorelick, 1987) was used to estimate flow and transport properties for the mesocosm sediments. Hydraulic conductivity was estimated using flow rate (Table 1.2) and hydraulic gradient (calculated across the entire mesocosm from the values in Table 1.1) in Equation 1.1 (Darcy’s Law). Porosity was then estimated based on fluorescein data and Equation 1.2 (another form of Darcy’s Law). MODFLOW (McDonald and Harbaugh, 1988) and MT3D (Zheng, 1992) were used to iteratively improve these estimates as well as sorption parameter estimates. These programs were used to simulate an effluent concentration history based on head measurements, groundwater flow rate, and tracer data. Simulated effluent concentrations were then compared to actual data for every tracer sampling time, and differences (residuals) were calculated. Then the estimates were modified using a golden search algorithm in effort to better match the effluent concentration history, by reducing residuals with the next simulation. This procedure was then iterated many times to closely reproduce effluent concentrations with a homogeneous model. A homogeneous model was used to Show the impact of lithologic heterogeneity on groundwater flow and contaminant transport. As expected, the model provided poor estimates of concentration histories measured at discrete 22 locations within the mesocosm. Therefore, if a homogeneous model is used at a heterogeneous site, the estimated parameters will be effective averages and thus concentration predictions at most discrete locations in the domain will be inaccurate. Understanding processes resulting from multiple scales of heterogeneities in geology is necessary to develop more accurate predictions of tracer concentrations at discrete locations. Detailed knowledge of heterogeneities and lithology structures would lead to lower values of dispersivity because there would be less unexplained changes in groundwater velocity in a more highly defined conductivity field. The sorption portion of the model would likely be unnecessary to describe the measured concentration history if the heterogeneous hydraulic conductivity field was properly described. Because we have limited geological information about this system until the mesocosm is sub-sampled, our ability to predict solute transport is limited even in the presence of accurately defined boundaries and SU’CSSCS. Equation 1.1 Darcy’s Law Q = -K r i * A Where: Q = Volumetric Flow Rate (ml/ day) i = Hydraulic Gradient (dimensionless) A = Cross-sectional Area (cmz) K = Hydraulic Conductivity (cm/ day) 23 Equation 1.2 Darcy’s Law V = —K * i / nc Where: V = Average Linear Velocity (cm/ day) i = Hydraulic Gradient (dimensionless) n, = Effective Porosity (dimensionless) K = Hydraulic Conductivity (cm/ day) CONCLUSIONS The mesocosm apparatus can be used to collect and study three-dimensional, heterogeneous, minimally disturbed aquifer materials. This apparatus allows for the study of flow and transport of many solutes or contaminants at longitudinal scales of 1.09 meter and less and transverse scales of 0.44 meters and less. In addition, the collected sample can be sub-sampled and the previously estimated parameters can be physically measured. Although this procedure is possible in the field, it is not yet economically feasible outside of the laboratory. Average outflow concentrations from a heterogeneous column can be modeled using large dispersivity and sorption coefficients in a simulation with homogeneous aquifer properties. However, effective properties will not adequately represent concentrations measured at individual wells. In such homogeneous simulations large values of sorption and dispersivity constants must be used to fit measured data. The use of the large dispersivity accounts for the large variations in groundwater velocity due to heterogeneities in hydraulic conductivity that are unrealized by a homogeneous model. Therefore, high-resolution hydraulic conductivity estimates are necessary in heterogeneous environments to reduce the uncertainty in the flow field and improve contaminant transport predictions. The estimated porosity value for the mesocosm is very low because the estimated parameters are estimated based on the effluent concentration data which is representative of 24 25 the average properties in the mesocosm that have the greatest effect on the flow system. Therefore, the effective average porosity in a domain with clay lenses (60% by volume) will be lower one that does not, because these lenses do not contribute to the effective transport region in the mesocosm. Further, if one divides this porosity value by the percentage of material contributing to transport in the mesocosm (0.4), a value results that would be in the range for silty sand. The mesocosm design could be used for many other scientific and engineering studies. The presented experiment could be augmented through the use of multiple tracers to infer processes leading to the asymmetrical shape of the concentration histories. Transport of bacteria and bioremediation of contaminants could also be examined in this type of system to test new strategies for enhanced remediation. Finally, geophysical analysis such as very high frequency RADAR might help estimate heterogeneities in aquifer properties with this mesocosm design if the signal could be propagated through the PVC pipe. CHAPTER 2 26 INTRODUCTION Diffusion is an often-ignored chemical process that causes mixing of solutes in groundwater in the presence of a concentration gradient due to molecular vibrations (Freeze and Cherry, 1979). This process is generally ignored because mechanical dispersion processes are usually considered to control the spreading of a tracer. Several studies have presented evidence that diffusion can significantly effect average linear velocity of some solutes. Zahn and Behrens (1992) concluded that fluorescein, when occurring in high concentration moves faster than chloride over flow distances between 1 and 575 meters. Maloszewski and Zuber (1989) found that diffusion of solutes into and out of dead end pores might slow transport. Garnier et al (1985) observed fluorescein moving faster than deuterium (ZHZO), iodide (I), and carbon 13 (HZBCO3), which are all considered as conservative tracers. However, few studies have addressed either the relationship between molecular weight and diffusion, or the relationship between diffusion and transport velocity. Several studies have also explored diffusion of solutes from fractures into surrounding matrix of dead end pores. Maloszewski and Zuber (1993) used two tracers with different coefficients of molecular diffusion to explore matrix diffusion, and found that their model, which incorporated matrix diffusion, was as good or better at fitting measured data in fractured rocks than models that did not. Neretnieks et al (1982) and Moreno et al. (1985) used four and five tracers respectively in a laboratory column to study tracer movement in fractured granite cores and found that diffusion is a key component of any model used to 27 28 predict solute movement through fractured media. Lever et al. (1985) studied effects of dead-end pores with a laboratory diffusion ecu and developed a dual porosity model to account for diffusion into diffusion into rock matrix. Grisak et al. (1980) studied a large intact sample of fractured clayey glacial till and determined that diffusion of tracer species into the porous matrix significantly retarded Ca and Cl. Additionally Grisak et al. (1980) concluded that Ca traveled faster than Cl as a result of a smaller effective diffusion coefficient. In this Study, we investigate the effects of diffusion on transport in a highly heterogeneous, minimally disturbed, glacial sediment mesocosm (Chapter 1). This mesocosm can be used to study the influence of diffusion on solute transport under controlled stresses and boundary conditions. The purpose of this paper is to Show that diffusion influences solute transport in heterogeneous environments. In these environments, low molecular weight solutes diffuse more, and are retarded relative to, higher weight solutes. The process increases the flux of low molecular weight solutes, relative to high molecular weight solutes, out of the predominant flow paths in an aquifer into less conductive media, and possibly into dead-end saturated pore space. A lower average linear velocity, which we call diffusive retardation, results for these lighter, more diffusive molecules as they become delayed with respect to heavier tracer molecules. Diffusive retardation can be explained with a thin layer of high hydraulic conductivity sediment between two layers with much lower conductivity (Figure 2.1). As a slug of solute is introduced into the water flowing into one end of the high conductivity region, a chemical gradient is induced in two directions. The first is oriented in the direction 29 of flow and does not change the center of mass arrival time but does smear the front and rear of the slug arrival (divergence from square wave arrival). However, there is a second concentration gradient oriented perpendicular to flow and in this case into the low conductivity regions. This gradient will cause particles in the solute to be diffused into the low conductivity regions and slowed in the direction of flow as a result (Figure 2.1). With a solute particle now trapped in the low conductivity material, only diffusive transport could return it to the more conductive media. However, diffusion only occurs in the direction of a chemical gradient, and this gradient will be oriented away from the high conductivity material until the tracer in the high conductivity region passes. After the plume has passed, the concentration gradient is reversed because no concentration of particles exists in the high conductivity region near the diffused particles. As particles are stripped off the front of the plug flow and re-introduced at the rear, the arrival of the center of mass is delayed or retarded. We propose that this occurs in porous media and is one of several possible processes responsible for differences in arrival times between different tracers. METHODS Large-scale field studies and small-scale laboratory studies both fail to address some processes that influence transport of contaminants. laboratory studies typically use homogeneous media or artificial sediment heterogeneities and thus fail to explore three- dirnensional transport in a naturally heterogeneous environment. Field studies are generally incapable of studying the small-scale processes because of the assumptions that must be made to deal with boundary conditions. The inability to densely sample these large-scale systems often results in poor conservation of mass in heterogeneous environments where plumes may be elusive. Tracer studies through a large undisturbed sediment sample, called a mesocosm, would bridge the gap between small-scale laboratory and field scale studies. A large aquifer sample was collected and transported to the laboratory with minimal disturbance to the physical structure. The aquifer sample was collected from a shallow glacial debris-flow deposit (Grahame Larson, personal communication) in East Lansing, Michigan, on August 7, 1996, by pushing a rigid tube horizontally into the formation with a heavy-duty excavator. The sample was then sealed in the field and transported back to the laboratory (Chapter 1). Numerous lithologic heterogeneities are present in this sample, which contains approximately 60% clay, 20% silt, 20% sand, with some larger cobbles based on visual inspection of sediments adjacent to the collection location. This deposit was chosen because of its high degree of heterogeneity, and its proximity to the laboratory. 30 31 *— Low K region —> 4— High K region —> 4.— Low K region __> Time Step 1 . White Tlme particle diffused Step2 outofhigh -I-I-I-I-I-I-I-I-I xii-PEEP??? K region . Black :-:-: -:-:-:-:-:- Tlme partiCle . .9 ...... diffused ... ..... . Step3 outofhigh .:.:.’:.:.:.:.:. .................... K region Tlme Step 4 Gradient reversal allows particles to re-enter ......... ......... IIIIIIIIII CCCCCCCCC IIIIIIIII IIIIIIIII Tim HighK , , Step: region. ..... f , M ............ Diffusive Solute Non-diffusive Solute M Diffusive Retardation. Three identical particles are shown on the left and three different identical particles are shown on the right. The media on both sides is identical with a highly conductive layer between two less conductive layers. From top to bottom 5 progressive time steps are represented. Notice in time step 2 on the left side that the concentration gradient caused the leading particle to exit the highly conductive layer and become ‘dormant’ in the less conductive surroundings. Only in time step 5 is this particle released back into the highly conductive layer. Also notice that the non-diffusive tracer on the right is exiting the aquifer in time step 5, while the diffusive tracer is still at least 2 steps behind. 32 The mesocosm collection tube was designed to collect a large undisturbed aquifer sample and transport it to the laboratory as discussed in Chapter 1. PVC (polyvinyl chloride) was used to minimize sorption of solutes to the apparatus. Sampling points were added by installing miniature wells and effluent sampling ports at the locations shown in Figure 1.1. Additionally, edge flow, a common flaw of laboratory columns, was minimized so that the predominant flow is through porous media. Edge flow was controlled by collecting a sample that has no annular gap between the walls of the housing and the aquifer sample, and through the design of the mesocosm endplates. A cutting tip (Figure 1.1) eliminates the annular gap by trimming off excess aquifer material as the collection tube was pressed horizontally into the aquifer. Edge flow was further controlled by the mesocosm end pates (Figure 1.1) which permit flow to enter and exit the mesocosm only at the ends and not along the edges. More details on this mesocosm design can be found in Ewers et al. (in review). Tracer Tests The mesocosm offers a highly controlled system for a comparison of tracers in a highly heterogeneous undisturbed aquifer system. The boundaries for the system are no- flow at the tube walls and constant head at the endplates. The flow rate in the system was controlled by a peristaltic pump, and the flow rate was checked hourly with a graduated cylinder and timer and all data was converted to dimensionless time using these flow rate measurements. Instead of reporting data in terms of time, which could lead to erroneous conclusions if pump rate fluctuations occur, the data is reported in terms of tube volumes. One tube volume is defined as the amount of time required to pump a volume of water equivalent to the volume of the mesocosm tube with no sediment. Effluent volume is 33 tracked temporally and flow rates are therefore monitored. These outflow rates are then used to calculate the total volume of water pumped through the mesocosm (since the start of the tracer test) at each sampling time, then this volume is normalized to the entire mesocosm volume (Equation 2.1). This method of time keeping will not be sensitive to pump rate fluctuations because they are included in the volume calculation. Equation 2.1 The dimensionless time (Tube Volume) equation. T = Vt/Vm Where: T -- The dimensionless time of a sampling event in tube volumes (dimensionless) Vt = The total volume of liquid that has exited the mesocosm since the start of the test (cm3) Vm = The total volume of the empty mesocosm apparatus (169,440 cm’) Fluorescein and tritium were selected for this study based on their properties. Both are generally considered conservative, or non-reactive with the aquifer matrix, and both are typically considered to be nearly ideal tracers by hydrogeologists. However, they have very different molecular weights, making them excellent tracers for studying diffusion processes. Fluorescein has a high molecular weight causing it to be less diffusive than tritium, which has a much lower molecular weight. 34 Table 2.1 Measured head, tracer arrival times and hydraulic conductivity based on slug tests. A dash indicates that there was not sufficient data collected for center of mass calculation. Fluorescein Tritium Fluorescein Center of Center of Center of Well or Well Head Hydraulic Mass Anival Mass Mass Arrival Sample Depth (cm) Conductivity Time Arrival Time Port (cm) (cm/ 5) (Tube Time (Tube (Tube Volumes) Volumes) Volumes) (First Test) (Second Test) inflow 100 2 9.6 87.5 0.0025 0.21774 0.72818 - 3 24.1 85.8 0.0027 0.33878 0.31075 0.15299 4 9.6 86.9 0.028 0.098862 0.10671 0.068411 5 9.6 80.8 0.0013 0.29000 0.62410 0.086568 6 9.6 85.6 0.0013 0. 12608 0.11672 0.081186 7 9.6 84.0 0.0017 0.47880 0.45499 0.41324 8 23.7 85.7 0.69 0.052708 0.088947 0.075933 9 9.6 85.7 0.0025 0.53519 0.31673 - 0.24360 10 9.6 81.3 0.00074 0.26587 0.58358 - 11 62.0 0.093490 0.10993 0.094159 12 62.0 0.14033 0.14502 0.10506 13 62.0 0.14702 0.12208 0.10078 14 62.0 0.10844 0.11871 0.10392 15 62.0 0.12426 0.12222 0.10478 Outflow 62.0 0.10649 0.1 1708 0.097067 35 The tracer tests were conducted separately to insure no interference occurred between concentration measurement methods used for tritium and fluorescein. The analysis of each tritium sample requires mixing a fluorescent cocktail with the groundwater sample to cause the tritium decays to emit light that is' detected in the scintillation counter. The tracer tests were conducted using identical methods for tritium and fluorescein. Each tracer was mixed to form two liters of groundwater solution with a resulting density approximately the same as groundwater, and injected into the laboratory mesocosm in a manner that allowed tracer to be distributed evenly over the end of the column. Each tracer was injected under a steady state gradient that was maintained for the duration of the study, including the time between tracer tests. Details on flow rate and injection concentrations can be found in Table 2.2. The data collected from these tests were used to estimate tracer velocities and groundwater velocity. The effluent concentration data from the first fluorescein tracer test were used to estimate effective porosity (6.8%) and hydraulic conductivity (1.3 e-2 cm/s) as , described in Chapter 1. This value of conductivity is representative of a clean sand, illustrating that a relatively small component of the aquifer has a large impact on effective properties (Chapter 1). This porosity value is low because it is an effective value for the entire mesocosm and only about 6.8% of the volume of the mesocosm is effective pore space for advective transport. A detailed description of the estimation procedures and estimated parameters can be found in Chapter 1. 36 Table 2. 2 General information about tracer tests conducted on a glacial aquifer mesocosm acquired form E. Lansing, MI on Aug 7,1996. Injection concentrations are in decays per minute per milliliter for tritium and parts per million for fluorescein. Fluorescein Tritium (3H) Fluorescein First Test Second Test Date of Test Start Aug. 23, 1996 Jan. 23, 1997 Sept. 12, 1997 Initial Concentration 100 PPM 950,000 DPM/rnl 100 PPM Average Pump Rate 5.3 ml/ min. 5.2 ml/ min. 5.3 ml/ min. Tracer Volume 2 liters 2 liters 2 liters Mass Conserved 74.1% 87.8% 80.5% First Fluorescein Tracer Test Fluorescein dye is a nearly ideal tracer that is easily detectable with a fluorimeter, or spectrophotometer. Fluorescein is stable at normal groundwater temperatures, and decay occurs at pH greater than 6.5 and only in the presence of sunlight (Behrens and Demuth, 1992), therefore decay is minimized in an aquifer setting. Fluorescein has a molecular mass of 376 grams per mole and has a complicated structure (Figure 2.2) which can be chemically modified by changes in water chemistry, therefore it is expected that some sorption will occur (Zahn and Behrens 1992), especially in a heterogeneous environment. If these structure-modifying reactions occur, this green tracer molecule will become colorless and be undetectable. This process can be described with a decay reaction. Zahn and Behrens (1992) found that a slight retardation, a result of sorption reactions, occurs with fluorescein in glacial sediments but using injection concentrations above 5 ppm can minimize it. 37 / 0 / \O Ho \ O \ 0 Figure 2.2 Fluorescein Molecular Structure. Molecular weight 2376 grams per mole. tritium decays and determine concentrations for this tracer test. Much higher resolution concentration data could be achieved with tritium, so smaller samples were taken. Ultima Gold scintillation cocktail (8 ml) was added to 0.7-ml samples for analysis in a Packard scintillation counter. Samples were counted until a 2-sigma reproducibility was reached to insure accurate concentration measurement. Concentrations were monitored hourly at 14 sampling locations beginning August 23, 1996 during the first of two fluorescein tracer tests. Each concentration was measured using a 2-ml sample drawn from the mesocosm, centrifuged and run through a spectrophotometer. The hourly schedule was selected to optimize temporal resolution without adding significant changes in hydraulic gradient within the mesocosm. Initially, 38 ' samples were to be taken from each of 15 sampling locations, but well 1 could not be sampled because it was screened in material with extremely low hydraulic conductivity. Data was collected from the remaining 14 locations (Figure 1.1) on an hourly basis for 15 days at which time the scheduled sampling frequency was progressively reduced for 60 more days. This data set was used to construct concentration arrival histories for the 14 sampling locations. These concentration histories can be found in Appendix B. Tritium Tracer Test Tritium does not behave differently from other water molecules in an aquifer (Athavale et al., 1980) and is easily detected with a scintillation counter making it an ideal tracer for monitoring groundwater movement. However, this tracer can rarely be used outside of the laboratory due to the biological hazard it poses. This unstable isotope emits beta particles with a maximum energy of 18 keV as it radioactivity decays with a half-life of . 12.26 years. Tritium decay is very predictable, and a decay correction can be applied to the data to conserve the minimal mass that could be lost through this reaction. The tritium tracer pulse was introduced on January 23, 1997, and monitored with the same sampling schedule used for fluorescein. The only differences were sample size and the method of sample analysis. A computer-automated scintillation counter was used to count Second Fluorescein Tracer Test The Second fluorescein tracer test began on September 12, 1997, (after the tritium tracer test) to check reproducibility of the data. The same methods as the first fluorescein tracer test were used in this test except the sampling frequency was reduced. During this test 39 samples were collected and analyzed every two hours until the peaks in concentration had passed. Then the sampling schedule was reduced in the same manner as the first fluorescein tracer test after seven days. Slug Tests A slug test can be used to estimate the hydraulic conductivity of an aquifer by monitoring the rate of recovery of the water level in a well after a certain volume or ‘slug’ has been instantaneously added (or removed). When water is removed the test is usually referred to as a ‘bail test’, and when water is added the term ‘slug test’ is applied. In either case the recovery time is of importance and estimating procedures are the same (Freeze and Cherry, 1979). The method of Hvorslev (1951) was used in this study to estimate local hydraulic conductivity values. This method was designed around a point pieziometer, and relies on the following assumptions: Homogeneous, isotropic, and infinite medium in which both soil and water are incompressible. This is rarely the case in any aquifer, but if the assumption is made that the volume of the slug is relatively small, and a finite portion of the aquifer is being affected, then the assumption dealing with infinite extent of the aquifer may be relaxed. It then becomes necessary to assume that this finite portion of the aquifer is isotropic and homogeneous, and therefore estimates from slug tests must be treated as effective hydraulic conductivity (or transmissivity) over some discrete interval around each well. 40 The slug tests were performed for all nine wells using the same procedure. Each well was analyzed at different times to insure that there was no interference between tests. The procedure is as follows: 1.) Remove a volume of water from a well. 2.) Monitor the recovery rate of water level in the well being tested. 3.) Analyze the data using the standard Hvorslev method (1951). The instantaneous removal of water was accomplished by closing the Teflon valve in Figure 4, removing the desired amount of water, then reopening the valve. The water level was then monitored in the clear flexible tubing (Figure 2.3) and recovery was timed with a chronograph. Estimated values for local hydraulic conductivity can be referenced in Table 2.1. 41 Flexible clear tubing (used as Barbed hose fitting (threaded to teflon Teflon valve with compression fitting at base and threaded fitting at Well casing type 316 stainless steel, 1.6-mm diameter 30-cm length Finger tight fitting i 5 Finger tight receiver machined 1K in PVC collection tube Figgre 2.3 Well Attachment to Mesocosm. OB SERVATIONS/ RESULTS Several trends are apparent in the data sets compiled from the three tracer tests (Figures 2.4 and 2.5). First, peak heights get progressively larger with successive tracer tests at each sampling location, except well 8. Second, physical properties of the aquifer, such as hydraulic conductivity, may be changing with time. Third, the center of tritium mass arrives later than fluorescein at several locations. And finally, tritium mass is better conserved than fluorescein. A tendency of increasing peak heights with successive tracer tests is observed in the tracer data. This observation is most apparent in Figure 2.4, which shows this trend occurring at most sampling locations (excluding wells 2, 5 and 10 which did not have discemable peal-s). For example, in the first fluorescein tracer test, all observed peaks were smallest, except well 8, relative to peaks observed in later tracer tests. The next tallest peaks are associated with the second tracer test (Tritium) in all sampling points except well 8. Two unlikely explanations for this behavior are trapped air in the mesocosm that effectively cut off flow paths, and dissolution of carbonates, both could change the hydraulic conductivity distribution in the mesocosm. However these are unlikely because an air bleed valve was installed in place of well one at the start of the first fluorescein tracer test, and ground water from a similar aquifer was used for this study. Therefore, air would have escaped before it could have effected the tracer results, and we would not expect dissolution of carbonates to occur because the water would be in equilibrium with these sediments. This may lead to the 42 43 conclusion that some fine grained sediments were mobilized in the aquifer causing an increase in hydraulic conductivity and less dilution of tracers to occur along paths intersecting many of the sampling locations. Additionally, in Figure 2.4 it is apparent when most of the sediment redistribution occurred. At sampling locations 4, 11, 13, and 15 most of the conductivity change occurred between the first fluorescein tracer test and the tritium tracer test. At locations 6, 7, 14, and the effluent, most of the conductivity change occurred between the tritium and second fluorescein tracer test. Finally, at wells 3, and 9 not much change in hydraulic conductivity is apparent. 100 - 0 First Fluorescein Tracer Test h 121 Tritium Tracer Test P 0 Second Fluorescein Tracer Test C 80 *- 2 - g h 4» - c ,_ . OJ 8 60 — — O o - _ 'D a _ .9 P - E 40 — - E :— III a) .. -t 8 a, _ . CL 0 E 20 — $1 iii <> <, <> 7 L o 121 4 :1 11 I1 a U - ; 4» 9 i 0 4 s . o = ' o Lug i 11 3 i i 3 4 5 6 7 8 9 1112131415 Effluent Sampling Location Egg 6 2.4 Peak Concentrations. Plot of peak concentrations for all tracers (center of mass) at each sampling location. Well 2 and Well 10 did not have sufficient data to perform center of mass analysis. 44 A trend is observed between the arrival of three conservative tracers further indicating that fine-grained sediments may have been transported within the mesocosm (Figure 2.5). With each successive tracer test, the tracer centers of mass arrive progressively earlier at locations 3, 6, 7, 13, and 15 (Figure 2.5) indicating the possibility of sediment mobility. This sediment mobility may be causing a change in hydraulic conductivity to by flushing fine-grained sediments in the direction of induced gradient during sampling resulting in an increase of hydraulic conductivity where fines flushed from an da decrease where fines flushed to. Fluorescein centers of mass arrive at different times in successive tracer tests for all sampling locations indicating how physical properties may have changed in the mesocosm. This difference in arrival times is probably best illustrated in Table 2.3 that shows the percent change in fluorescein center of mass arrival times between the two tests at each location. A positive value indicates that the second test had delayed breakthrough with respect to the first (decrease in conductivity) and a negative value indicates that the second fluorescein test resulted in earlier breakthrough of solute and therefore conductivity increased over time. The results of these tests were expected to be similar, and values of percent difference in fluorescein arrival times near zero. However, during second fluorescein test the center of mass arrived earlier than for the first fluorescein test at all sampling locations except 8 and 11 (well 2 and 10 did not have sufficient data to compute center of mass). This indicates that flow paths to most sampling locations experienced an increase in hydraulic conductivity between the two fluorescein tracer tests. Conversely, a decrease in hydraulic conductivity (K), or increase in tracer arrival time in successive tests was observed at sampling locations 8 and 11. In either case (increasing or decreasing K) we 45 would expect a continuous shift in hydraulic conductivity values. Therefore, sediment mobility does not explain the late arrival of tritium at locations 4, 8, 11, 12, 14, and in the effluent with respect to both fluorescein tracer tests. This observed behavior is likely the result of differences between the transport properties of tracers themselves rather than changes in the media. 0.7 _ - 0 First Fluorescein Tracer Test : [1 121 Tritium Tracer Test 0.6 :- 0 Second Fluorescein Tracer Test E 4» : 0.5 - - : 0 Z 8 _ 121 q E 0.4 - 8 — 2 ' I g : 0 1 j 3 0.3 — 11 0 l _ E : 2 . <> _ 0.2 E- j : <> 5 I; <5 2 .. r . 0.1 — U 0 f U <> <> U g E .3 I <> 0 El 3 : i 1‘ 0 l 3 4 5 6 7 8 9 1112131415Effluent Sampling Location Figyre 2.5 Center of Mass Arrival Times. Plot of arrival time for all tracers (center of mass) at each sampling location. Well 2 and Well 10 did not have sufficient data to perform center of mass analysis. Sediment mobility within the mesocosm may be correlated with the logarithm of hydraulic conductivity (log(K)) as shown in Figure 2.6. This figure shows a plot of log(K) verses the percent differences in fluorescein center of mass arrival times which are presumed to result from sediment redistribution within the mesocosm. The correlation coefficient for 46 this relationship is approximately 0.83 indicating a fairly strong correlation. This may not be a fundamental relationship but it does make sense that areas of high conductivity would become less conductive and areas of low conductivity would become more conductive as a global average value of conductivity is reached. Additionally, the two points that do not fall near the correlation line (well 6 and 7) may be low conductivity regions that are not as susceptible to change because they are not in close proximity to higher conductivity regions (where fines could be flushed). Table 2.3 Percent change of fluorescein arrival between the first fluorescein test and the second fluorescein tracer test. Sampling Location Percent Change of Fluorescein Center of Mass Arrival From Fluorescein Test One to Fluorescein Test Two 3 ~55 4 -31 5 -70 6 -35 7 -13 8 44 9 -54 11 7 12 ' -25 13 -3 1 14 -4 15 -15 effluent -8 47 Tritium arrives unexpectedly later than fluorescein (in both fluorescein tracer tests) at several sampling locations. In Figure 2.5 it is apparent that nitium arrives later than fluorescein at sampling locations 4, 8, 11, 12, 14, and in the effluent. This delay of tritium is most likely related to different diffusion coefficients for these two solutes. Other possible explanations for this behavior may be sediment mobility, or variable sorption. However, both are unreasonable because this experiment was duplicated and fluorescein arrived earlier than tritium on both occasions and sorption reactions would only act to slow the breakthrough of fluorescein. Fluorescein has a much lower diffusion coefficient than tritium that may cause it to not be diffused into low conductivity regions which in turn may cause this earlier fluorescein arrival. o I I I I I I I I I I I I I I I I I I I l I I r I I I I y = -1.5186 + 2.2256x R= 0.83338 ’ IIIIIIIII'IIIIIIIIIIIIIIIIIII riiilriirliiriliiiiliiiiliiii g an -1.5 O _l -2 -2.5 O _31.1liirlriilmrilrriliiiliii -0.8 -0.6 -0.4 -0.2 O 0.2 0.4 0.6 Percent Change (Center of Mass Arrival of Fluorescein) Figgge 2.6 Correlation of Hydraulic Conductivity and Percent Change of Tracer Arrival. Scatter plot of Log (hydraulic conductivity) verses percent change in center of mass arrival time between the first and second fluorescein tracer tests. Well 2 and Well 10 did not have sufficient data to perform center of mass analysis and slug tests could not be performed at sampling locations 1 1 -1 5. 48 Effects of diffusion can be best explained with the saturated porous media diffusion equation (Equation 2.2). Diffusion is inversely proportional to the square root of molecular weight of the solute in this equation. Given that the average molecular weights of tritium and fluorescein are approximately 19 and 376 grams per mole respectively, we see that tritium has about a five times greater chance of being diffused out of a predominant flow path and into less conductive matrix than fluorescein. Because of its larger molecular mass, fluorescein should have more inertia to resist diffusion than tritium. This reduced molecular energy will lessen the tendency of a solute to diffuse into dead end saturated pore space or less conductive regions. Those particles that undergo less diffusion will also tend to spend less time in clays where sorption potential would be expected to be greatest. These particles would then breakthrough earlier than a sorbed particle or a particle that spent some time in dead end pore space or less conductive regions. This process explains the early breakthrough of fluorescein relative to tritium. Equation 2.2 The Saturated Porous Media Diffusion Coefficient Equation D= 1.728*10“ mz/day * «1(32 g / M) Where: D= Diffusion Coefficient (mz/ days) M= Molecular Weight (grams) 49 The mass conservation difference between the two solutes is likely best accounted for by a combination of chemical reactions. Fluorescein is likely taking part in irreversible sorption reactions causing some mass to be loSt. Zahn and Behrens (1992) concluded that significant irreversible sorption was occurring in a laboratory experiment they were conducting, and resulted in losses of fluorescein mass. Other issues to consider include the radioactive decay of tritium, but due to the short duration of this tracer test (75 days) and the half life of tritium (over 12 years), radioactive decay is minimal. Fluorescein also takes part in photolytic decay reactions, but it is doubtful that this occurred to any measurable extent because the system was closed from light. Samples were only exposed to light during collection and were moved to a dark storage place after immediate concentration analysis. Mass conservation differences between the two fluorescein tracer tests is probably best accounted for with sediment rearrangement within the mesocosm causing an overall increase in hydraulic conductivity in the mesocosm. As sediments rearrange many areas become more conductive resulting in higher tracer velocities. A tracer traveling slower, such as fluorescein in the first test, will have more opportunity to diffuse into low conductivity regions than a tracer moving faster (the second fluorescein test). This increased diffusion causes the slower tracer to be more diluted as it slowly diffuses out of the low conductivity regions. This dilution may have caused fluorescein to exit the mesocosm at a much lower and undetectable concentration for long periods after the first tracer test. Additionally, irreversible sorption reactions could account for some of this mass conservation difference. If a limited number of sorption sites became saturated with fluorescein during the first tracer test these sites would not be available to remove fluorescein from flow paths during the second fluorescein test. SUMMARY AND CONCLUSIONS Numerous small-scale processes affect solute transport in heterogeneous aquifers. The effects of two of these processes were observed with in this mesocosm. First, the effects of diffusion in a heterogeneous environment on solutes with different molecular weights were observed. Second, sediment mobilitywas observed in this system and was likely the result of this system having recently shifted from unsaturated to saturated conditions. Additionally, this mobility of fines may result from differential stresses incurred during collection of the aquifer sample by causing clays to become unstable. Collecting a saturated aquifer sample oriented in the direction of natural groundwater flow may minimize sediment migration in a mesocosm, although cost and concern over sample disturbance will generally prohibit collection of a saturated sample. As a solute flows through an aquifer it will flow preferentially along paths of least resistance such as high conductivity lithologies. As the solute flows along this path, a concentration gradient is produced between the high concentration in the high conductivity pathway and the lower concentration less conductive pathway. When two tracers are compared we observe that fluorescein spends more time in high conductivity regions because it is less diffusive and therefore will appear less retarded than the more diffusive and less massive tritium. 50 51 Scale of heterogeneity and length of flow path may also affect the magnitude of the delay resulting from diffusion. Homogeneous aquifers would be expected to have a much less pronounced delay because there would be no areas of low hydraulic conductivity present to slow diffused particles. However, the effect of diffusion may still be perceptible in homogeneous aquifers if the aquifer matrix has adequate dead end pores. Flow path length is important because at flow distances much shorter than the scale of heterogeneity, the tracer will encounter only one type of aquifer material and results will appear similar to the homogeneous aquifer case. In both tests a considerable amount of tracer mass was lost. This has implications for remediation of spills in glacial material. With a “conservative” tracer, about 75-80% of a spill can be recovered after only 45 days (about one pore volume) in a controlled system with pump and treat methods of remediation. However after this time concentrations are below 1% of the initial concentration and removal of the last 25% of mass may take decades or longer with conventional pumping. The laboratory mesocosm approach has allowed the study of the influence of small- scale processes which effect flow and transport in heterogeneous environments. Data collected with this approach indicate that more consideration should be given to diffusion processes in heterogeneous environments. This mesocosm approach could also be used to investigate other properties affecting flow and transport at small scales. Incorporation of geophysical techniques such as very high frequency seismic or RADAR could lead to estimates of lithology distribution, which could aid estimation of heterogeneous aquifer 52 parameters. Additionally, these parameter estimates could be verified by dissecting the mesocosm and characterizing the aquifer materials. This mesocosm design and methodology could be changed in several ways in the future to improve this study. First, the uSe of multiple mesocosms, each with different degrees of heterogeneity would allow the study of the effects of the scale of heterogeneity on flow and transport to be directly studied. Second, an auto sampler would reduce the human error during the tracer sampling and allow for higher temporal resolution. Third, the use of stainless steel fittings to attach the wells to the mesocosm may greatly improve this design. _ Considerable modifications to the PEEK fittings were necessary to seal the wells to the mesocosm and eliminate leaks prior to the study. Finally, the use of multiple tracers injected simultaneously would allow the effects of diffusion on two different tracers to be observed in equivalent environments. The second fluorescein tracer was injected simultaneously with, a bromide tracer, but these samples have not yet been analyzed. APPENDICES APPENDDC A Slug Test Data 53 Well 2 Time (min) H-h H-h/H-H., 2.0000 9.5000 0.95000 - 4.0000 9.0500 0.90500 7.0000 8.3500 0.83500 10.000 7.6500 0.76500 13.000 7.1000 0.71000 17.000 6.4000 0.64000 21.000 5.8000 0.58000 26.000 5.3000 0.53000 31.000 4.8000 0.48000 36.000 4.2000 0.42000 41.000 3.7500 0.37500 46.000 3.4000 0.34000 55.000 3.0000 0.30000 65.000 2.6500 0.26500 Well 3 Time (min) H—h H—h/H-I-L 4.0000 9.2500 0.92500 — 7.0000 8.7000 0.87000 10.000 8.1500 0.81500 13.000 7.6000 0.76000 17.000 7.0000 0.70000 21.000 6.4000 0.64000 25.000 5.8500 0.58500 31.000 5.1500 0.51500 36.000 4.4500 0.44500 41.000 3.8500 0.38500 50.000 3.3500 0.33500 59.000 2.6500 0.26500 69.000 2.0500 0.20500 83.000 1.5000 0.15000 54 Well 4 Time (s) H-h H-h/H-I-L 20.000 18.300 0.91500 - 40.000 16.600 0.83000 60.000 15.250 0.76250 80.000 13.850 0.69250 100.00 12.650 0.63250 120.00 11.500 0.57500 140.00 10.600 0.53000 180.00 9.9000 0.49500 220.00 7.4000 0.37000 260.00 6.2000 0.31000 300.00 5.2000 0.26000 364.00 4.0000 0.20000 420.00 3.0000 0.15000 500.00 2.1500 0.10750 580.00 1.5000 0.075000 680.00 1.0000 0.050000 Well 5 Time (min) H-h H-h/H-Hn 10.000 9.1000 0.91000 - 17.000 8.5500 0.85500 25.000 8.0500 0.80500 30.000 7.6000 0.76000 40.000 6. 8500 0.68500 50.000 6.1500 0.61500 60.000 5.5000 0.55000 70.000 4.9000 0.49000 80.000 4.3500 0.43500 90.000 3.9000 0.39000 105.00 3.2500 0.32500 120.00 2.7000 0.27000 140.00 2.0000 0.20000 160.00 1.5500 0.15500 180.00 1.3000 0.13000 55 Well 6 Time (min) H-h H-h/H-Ho 4.0000 9.1000 0.91000 _ 7.0000 8.6000 0.86000 10.000 8.0000 0.80000 13.000 7.6000 0.76000 17.000 7.0500 0.70500 21.000 6.5000 0.65000 25.000 6.0000 0.60000 31.000 5.3000 0.53000 36.000 4.7000 0.47000 41.000 4.1000 0.41000 50.000 3.6000 0.36000 59.000 3.0000 0.30000 69.000 2.4500 0.24500 83.000 1.9000 0.19000 Well 7 Time (min) H-h H-h/H-Hn 2.0000 9.5000 0.95000 - 5.0000 9.0000 0.90000 8.0000 8.5000 0.85000 12.000 7.9000 0.79000 16.000 7.3500 0.73500 21.000 6.7500 0.67500 27.000 6.0500 0.60500 31.000 5.7000 0.57000 39.000 4.9000 0.49000 45.000 4.5500 0.45500 51.000 4.0000 0.40000 62.000 3.5000 0.35000 78.000 2.8000 0.28000 94.000 2.2000 0.22000 109.00 1.8000 0.18000 55 Well 6 Time (min) H-h H-h/H-I-In 4.0000 9.1000 0.91000 _ 7.0000 8.6000 0.86000 10.000 8.0000 0.80000 13.000 7.6000 0.76000 17.000 7.0500 0.70500 21.000 6.5000 0.65000 25.000 6.0000 0.60000 31.000 5.3000 0.53000 36.000 4.7000 0.47000 41.000 4.1000 0.41000 50.000 3.6000 0.36000 59.000 3.0000 0.30000 69.000 2.4500 0.24500 83.000 1.9000 0.19000 Well 7 Time (min) H-h H-h/H-Hn 2.0000 9.5000 0.95000 _ 5.0000 9.0000 0.90000 8.0000 8.5000 0.85000 12.000 7.9000 0.79000 16.000 7.3500 0.73500 21.000 6.7500 0.67500 27.000 6.0500 0.60500 31.000 5.7000 0.57000 39.000 4.9000 0.49000 45.000 4.5500 0.45500 51.000 4.0000 0.40000 62.000 3.5000 0.35000 78.000 2.8000 0.28000 94.000 2.2000 0.22000 109.00 1.8000 0.18000 56 Well 8 Time (s) H-h H-h/H-Ho 5.0000 12.500 0.62500 — 10.000 7.5000 0.37500 15.000 5.1000 0.25500 20.000 3.1000 0.15500 25.000 1.9000 0.095000 30.000 1.2000 0.060000 40.000 0.5000 0.025000 45.000 0.3000 0.015000 Well 9 Time (min) H—h H-h/H-Ho 1.0000 19.600 0.98001 — 2.0000 19. 100 0.95500 4.0000 18.200 0.91000 6.0000 17.300 0.86500 9.0000 16.100 0.80500 11.000 15.300 0.76500 13 .000 14.500 0.72500 15.000 13.900 0.69500 18.000 12.900 0.64500 21.000 12.000 0.60000 24.000 11.100 0.55500 27.000 10.400 0.52000 31.000 9.4000 0.47000 3 5.000 8.6000 0.43000 39.000 7.7500 0.38750 43.000 6.9500 0.34750 48.000 6.1000 0.30500 53.000 . 5.4000 0.27000 57.000 4.8000 0.24000 63 .000 4.1000 0.20500 69.000 3 .4000 0. 17000 75.000 2.9000 0.14500 87.000 1.9000 0.09500 57 Well 10 Time firm) H-h H-h/H-Ho 10.000 9.5000 0.95000 - 20.000 9.1000 0.91000 30.000 8.7000 0.87000 40.000 8.3000 0.83000 50.000 7.8500 0.78500 60.000 7.4500 0.74500 70.000 7.1000 0.71000 80.000 6.7500 0.67500 90.000 6.3000 0.63000 105.00 5.7500 0.57500 120.00 5.3000 0.53000 140.00 4.6500 0.46500 160.00 4.1000 0.41000 180.00 3.6000 0.36000 214.00 2.9000 0.29000 240.00 2.6500 0.26500 APPENDIX B Tracer Concentration Histories CICO 0.01 0.008 0.006 0.004 0.002 58 Well 2 I I I I I I T I I I I I I I I I I I I l I - C/Co (Fluorescein 1) - C/Co (tritium) - C/Co (Fluorescein 2) c o o ' IVI 1 l l l l I l - Dun 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes 0100 0.025 0.02 0.015 0.01 0.005 59 Well 3 ' C/Co (Fluorescein 1) : ClCo (tritium) . C/Co (Fluorescein 2) ‘ o : 9' v; S v : _ v ' . I- o ’w‘“ W “ «I I v‘ ""*w u ' ' v v - .. V . _ O V W v v '- .. o - . o v' I v v 'v v' v . I- v 'v v V i l' 0 0.05 0 1 0.15 0.2 0.25 0.3 0.35 0 4 Tube Volumes CICO 60 Well 4 0.2 I I I I I I I I IT I I I I l I I I I I I . f 0 * 1:1 ClCo (Fluorescein 1) . v V C/Co (tritium) . 0° ’ <> ClCo (Fluorescein 2) 0.1 5 f V . . W I O 0.1 .. l 0.05 .. 0 .MM 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes CICO 0.005 0.004 0.003 0.002 0.001 61 Well 5 IjII'IIII'IIII'IIIIlI E1 ClCo(Fluorescein 1) V CICo (tritium) : o 0 C/Co (Fluorescein 2) I 0 I " E1 % u Du 11 ' : can u 11 i 0:51:5- n time, :1 11min IIUMEI 0. 0. 25 0.3 .35 0.4 Tube Volumes CICO 62 Well 6 0.25 IIII1fIIII'IIII'IIIT' . - El C/Co (Fluorescein 1) Z 0 V C/Co (tritium) _ <> ClCo (Fluorescein 2) o 0.2 - - <> - o - o 0.15 - 0 0.1 - 0.05 - 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes CICO 63 Well 7 0.02 I I I I I 1 I I I l I I I I I I I I I I I - - El C/Co (Fluorescein 1) . ° C/Co (tritium) . C/Co (Fluorescein 2) l" . v I"; 0.015 ' ’ V . V _ r '3‘. .v ' 7 o y k I " 001 - ' ' ' . v 'v' V" n V O V a . v' 0 D ' % V 0.005 -<> v 0% g h ' o ' n l l l l 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes CICO 0.7 . 0.6 0.5 0.4 0.3 0.2 0.1 64 Well 8 El C/Co (Fluorescein 1) V C/Co (tritium) 0 C/Co (Fluorescein 2) IIII'IIrrIfrIIlIIII'IIII'IIII'IIII 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes 0100 0.05 0.04 0.03 0.02 65 Well 9 I I I I l I I I I I I I I I I I I I I l I ' v C/Co (Fluorescein 1) '_' ClCo (tritium) . ClCo (Fluorescein 2) l: v v ‘ . V I .. 65m . :' <><> , 1 F 00 v Q Q 0 I : i0 ”#58 ’w v 0 0 v : .- "'v V V '0' 7, JV T. I V 0 V v I I o o '- . 0 V I r V . 0 an . 3 D1911“ 0 0.05 0.1 0.15 0.2 0.25 Tube Volumes 0100 0.006 1 0.005 0.004 0.003 0.002 0.001 IIIII'IIII'IIII'IIII'IIII'Ifir 66 Well 10 C/Co (Fluorescein 1) C/Co (tritium) CICo (Fluorescein 2) 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes CiCo 0.2 0.15 0.1 0.05 67 Sample Port 11 I I I I r I I I I l I I I I I T I I I l I - C/Co (Fluorescein 1) . o ClCo (tritium) . 69 ClCo (Fluorescein 2) I ' 7.}: .' . .'O'.-'-_ I I [El-Ira jag—lib" % E u .- "1| ”This—"“1.“ a. a C0: . u . ' -->—1§2'5°°1°l°°m. 0 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes 0100 68 Sample Port 12 0.2 IIIII’j’IIIIIIII'IIII—II . h- 0 C/Co (Fluorescein 1) . C/Co (tritium) . C/Co (Fluorescein 2) 0.15 - _ o _ oo . 00 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 - Tube Volumes CICO 69 Sample Port 13 0.2 Ij I l I I I I l I I I I l I I I I l I . - C/Co (F luoresceln 1) . C/Co (tritium) . o C/Co (Fluorescein 2) l 0° 0.15 - ° - I o d 0.1 "" .1 0.05 - L 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes CICO 0.2 0.15 0.1 0.05 70 Sample Port 14 I I I I I I I El C/Co (Fluorescein 1) V ClCo (tritium) 0 C/Co (Fluorescein 2) I L I I l I I I I l J I I I I 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes 0100 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 71 Sample Port 15 I I I I I I1 I I I I I I I l I I I I I I 2 E1 C/Co (Fluorescein 1) V C/Co (tritium) o o 0 C/Co (Fluorescein 2) 0 V 2....L.‘ U Iona;- 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes 0100 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 71 Sample Port 15 .I I I I l I I I I I I I I I l I I I I I . <3 11 C/Co (Fluorescein 1) L V C/Co (tritium) : 00 o 0 C/Co (Fluorescein 2) L . fiflrmm v '5’ _ I!!! 242423229.- 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Tube Volumes LIST OF REFERENCES Athavale, R. N., S. C. Murti, and R. Chand (1980). “Estimation of Recharge to the Phreatic Aquifers of the Lower Maner Basin, India, By using the Tritium Injection Method.” Journal of Hydrology 45: 185-202. Behrens, H., and N. Demuth (1992). “Measurement of Light Input into Surface Waters by Photolysis of Fluorescent Dye Tracers.” Tracer Hydrology: 49-56. Corwin, D. L., and R. D. LeMert (1994). “Construction and Evaluation of an Inexpensive Weighing Lysimeter for Studying Contaminant Transport.” Journal of Contaminant I_-I_ydrology15(1-2): 107-123. Garabedian, S. P., D. R. LeBlanc, L. W. Gelhar and M. A. Celia (1991). “Large-Scale Natural Gradient Tracer Test in Sand and Gravel, Cape Cod, Massachusetts 2. Analysis of Spatial Moments for a Nonreactive Tracer.” Water Resources Research 27(5): 911- 924. Garnier, J. M., N. Crampon, C. Preaux, G. Porel, and M. Vreulx (1985). “Tracage Par 13C, 2H, I-, Et Uranine Dans La Mappe De La Craie Senonienne En Ecoulement Radial Convergent.” Journal of Hydrology 78: 379-392. 72 73 Grisak, G. E., J. F. Pickens, and J. A. Cherry (1980). “Solute Transport Through Fractured Media 2. Column Study of Fractured Till.” Water Resources Research 16(4): 731- 739. Hvorslev, M. J. (1951). Time Lag and Soil Permeability in Groundwater Observations. Vicksburg Mississippi, U. S. Army Corps of Engrs. Waterways Exp. Sta. Bull. Knopman, D. S., C. I. V055, and S. P. Garabedian (1991). “Sampling Design for Groundwater Solute Transport: Tests of Metods and Analysis of Cape Cod Tracer Test Data.” Water Resources Research 27(5): 925-949. Lanzarone, N. A., and P. L. McCarty (1990). “Column Studies on Methanotrophic Degradation of Trichloroethene and 1-2-Dichloroethane.” Ground Water 28(6): 910- 919. Lever, D. A., M. H. Bradbury, and S. J. Hemingway (1985). “The Effect of Dead-End Porosity on Rock Matrix Diffusion.” Journal of Hydrology 80: 45-76. Maloszewski, P., and A. Zuber (1985). “On the Theory of Tracer Experiments in Fissured Rocks with Porous Matrix.” Journal of Hydrology 79: 333-3 58. McDonald, M. G., and A. W. Harbaugh (1988). “A modular three-dimensional finite- difference ground-water flow model”, United States Geological Survey, I WI 6-A1. 74 Moreno, L., I. Neretnieks, and T. Eriksen (1985). “Analysis of Some Laboratory Tracer Runs in Natural Fissures.” Water Resources Research 21(7): 951-958. Myrand, D., R. W. Gillham, E. A. Sudicky , S. F. OHannesin, and R. L. Johnson (1992). “Diffusion of Volatile Organic Compounds in Natural Clay Deposits; Laboratory Tests.” Journal of Contaminant Hydrology 10(2): 159-177. Neretnieks, I., Y. Eriksen, and P. Tahtinen (1982). “Tracer movement in a Single Fissure on Granite Rock; Some Experimental Results and Their Interpretation.” Water Resources Research 18(4): 849-858. Ptak, T. a. G. T. (1994). “Forced and Natural Gradient Tracer Tests in a highly Heterogeneous Porous Aquifer: Instrumentation and Measurements.” lournal of Hydrology 159: 79-104. Seigrist, H., and P. L. McCarty (1987). “Column Methodologies for Determining Sorption and Biotransforrnation Potential for Chlorinated Aliphatic Compounds in Aquifers.” Journal of Contaminant Hydrology 2(31-50). Silliman, 5., C. Cady, and C. Cunningham (1992). “Particle Transport in Heterogeneous Media; Initial Laboratory Investigation of Hydrodynamic Mechanisms.” fig 73 (43): 167. 75 Silliman, S. E., and E. 8. Simpson (1987). “Laboratory Evidence of the Scale Effect in Dispersion of Solutes in Porous Media.” Water Resources Research 23(8): 1667- 1673. Sillirnan, S. E., L. F. Kinokow, and C. I. Voss (1987). “Laboratory Investigation of Longitudinal Dispersion in Anisotropic Porous Media.” Water Resources Research 23(11): 2145-2151. Skibitzke, H. E., and G. M. Robertson (1963). “Dispersion in Groundwater Flowing Through Heterogeneous Materials.” U. S. Geological SurveLProfessional Paper 386- B. Smart, P. L. a. I. M. S. L. (1977). “An Evaluation of Some Fluorescent Dyes for Water Tracing.” Water Resources Research 13(1): 15-33. Vanclooster, M., D. Mallants, J. Vanderborght, J. Deils, J. VanOrshoven, and J. Feyen (1995). “Monitoring Solute Transport in a Multi-Layered Sandy Lysimeter using Time Domain Reflectometry.” Soil Science Society of Americaburnal 59(337-344). Wagner, B. J., and S. M. Gorelick (1987). “Optimal groundwater quality management under parameter uncertainty”, Water Resources Research 23(7): 1162-1174. 76 Wood, B. D., C. D. Dawson, J E. Szecosody, and G. P. Streil (1994). “Modeling Contaminant Transport and BIodegradation in a Layered Porous Media System.” Water Resources Research 30(6): 1833-1845. Zahn, M T., and H. Behrens (1992). “Quantitative Investigations on the Transport of Fluorescein in Fluvio-glacial Gravels.” Tracer Hydrology: 135-138. Zheng, C. (1992). “MT 3D: A modular three-dimensional transport model”, 8.8. Papadopulos 86 Associates, Inc.