4.0 Groundwater Modeling

Transcription

1 4.0 Groundwater Modeling P. D. Thorne Predicting future groundwater conditions and the movement of contaminants in groundwater is important in planning waste management and cleanup activities for the Hanford Site. Such predictions are primarily made using computer models that simulate groundwater flow and contaminant transport. Groundwater modeling activities that address problems on a Hanford site-wide scale, such as contaminant movement from the Central Plateau to the Columbia River, have been consolidated under the Hanford Groundwater Monitoring Project. The consolidation of sitewide modeling eliminates redundancy and promotes consistency of groundwater models (DOE/RL ). Other groundwater models are used for problems at a local scale (i.e., <~10 kilometers). The major application of local-scale modeling during the past several years has been to design and evaluate pump-and-treat systems for groundwater. This chapter summarizes Hanford Site groundwater modeling activities for fiscal year Section 4.1 reports progress on the continuing development of the consolidated site-wide groundwater model. Sections 4.2 through 4.4 describe specific applications of the consolidated model during the past year: completion of the System Assessment Capability initial assessment modeling to support the Hanford Site Solid Waste Environmental Impact Statement modeling of potential releases from each of the tank farm areas Sections 4.5 and 4.6 briefly describe modeling activities that have been performed over the past several years to address problems on a local scale using models other than the consolidated site-wide groundwater model. These include Computer models of groundwater help predict future groundwater conditions and the movement of contaminants in groundwater. This information is important in planning waste management and cleanup activities at the Hanford Site. During fiscal year 2002, a three-dimensional groundwater model was applied as part of the System Assessment Capability to simulate groundwater flow and transport of contaminants in the unconfined aquifer. The System Assessment Capability is an integrated system of computer models and databases used to assess the impact of waste remaining on the Hanford Site. It will help decision makers and the public evaluate cleanup options. Groundwater Modeling 4.0-1

2 modeling to support the design and operation of groundwater pump-and-treat systems and modeling of groundwater movement in the zone of interaction between the unconfined aquifer and the Columbia River. 4.1 Site-Wide Groundwater Flow and Transport Model Objectives of Hanford Site Groundwater Model A computer model of Hanford Site groundwater must be able to predict impact of Hanford activities on groundwater assess performance of waste-disposal facilities predict movement of contaminants evaluate remediation strategies A conceptual model that provides the basis for the site-wide groundwater model was developed from information on the hydrogeologic structure of the aquifer, spatial distributions of hydraulic and transport properties, aquifer boundary conditions, and distribution and movement of contaminants. Development of the basic aspects of this three-dimensional model of the unconfined aquifer system is documented in PNL-8332, PNL-8971, PNL-10195, and PNNL The model consists of nine separate hydrogeologic units, eight of which exist below the water table. Figure shows a cross section of the conceptual model layering from west to east across the central part of the Hanford Site. The groundwater flow system is bounded by the Columbia River on the north and east and by the Yakima River and basalt ridges on the south and west, respectively. Additional information on the site-wide groundwater model is presented in PNNL and PNNL Status of the Consolidated Model While computer model predictions are never 100% correct, modelers are working to quantify the degree of uncertainty in model results. Understanding the uncertainty leads to a better basis for waste-management and cleanup decisions. Quantifying uncertainty in model results is a major objective of the consolidated groundwater modeling task. Uncertainty is inherent in models because of assumptions about unknown aspects of the groundwater flow system and because the calculations used in groundwater models only approximate the processes of groundwater flow and transport. Understanding and quantifying the resulting uncertainty in model predictions will strengthen the technical defensibility of groundwater transport predictions and lead to a better basis for waste-management and cleanup decisions. The effort to incorporate uncertainty in the site-wide model began in 1999 with recommendations from an external peer review panel to establish a new modeling framework that accepts the inherent uncertainty in model conceptual representations, inputs, and outputs (PNNL-13641). This framework will produce a range of predicted results for future groundwater conditions and contaminant transport based on differences in conceptual model assumptions. As described in PNNL-13641, uncertainty in the site-wide groundwater model is being quantified through sensitivity analysis (e.g., alternative conceptual models and future scenarios) for those aspects of the analysis related to vagueness, and uncertainty analysis for those situations where the uncertainty (e.g., for parameters) can be represented by probability density functions. Current activities involve developing alternative conceptual models that encompass identified uncertainties, then applying a transient inverse calibration to each model based on historical observations of water-level changes and contaminant movement. Results of the different calibrated alternative models will then span the range of results expected based on different assumptions. See PNNL for details about the Hanford Site groundwater conceptual model and a more detailed description of the strategy for quantifying uncertainty in model results. Calibration of a base-case model and calibration of one alternative model were completed in fiscal year 2001 (PNNL and PNNL-13623). The alternative model included interaction between the unconfined aquifer system and the underlying basalt-confined aquifer. Previous site-wide modeling assumed the movement of water between these different systems was negligible. However, it is Hanford Site Groundwater Monitoring 2002

3 known that communication takes place, particularly along faults and in the area between Gable Mountain and Gable Butte, where erosion has removed the upper layers of confining basalt. Transient inverse calibration of these alternative models was completed with respect to historical observations of water-level changes. Calibration against historical patterns of contaminant movement is planned as a future activity. The calibration process utilized inverse-modeling procedures in UCODE, a universal inverse-modeling code developed jointly by the U.S. Geological Survey and the International Groundwater Modeling Center of the Colorado School of Mines (Poeter and Hill 1998). During fiscal year 2002, development of the site-wide model focused on both improvements to the base-case, site-wide model and development of additional alternative conceptual models. The alternative models address uncertainty in the extent and distribution of major mud units within the unconfined aquifer and uncertainty in the distribution of hydraulic conductivity zones within the Hanford formation. The distribution of mud units within the aquifer affects vertical migration of contaminants and also affects lateral movement, particularly where mud units exist at the water table. The Hanford formation is critically important in the transport of groundwater contaminants to the Columbia River because of its relatively high hydraulic conductivity. Therefore, the distribution of Hanford formation zones that have unique values of hydraulic conductivity may have a large impact on contaminant movement from the Central Plateau to the Columbia River and the accessible environment. Advanced geostatistical techniques were applied to develop the new conceptual models, partially through a cooperative project between Pacific Northwest National Laboratory and the Russian Academy of Sciences Institute for Nuclear Safety (IBRAE). This allowed us to use the expertise of IBRAE staff in mathematical geology, and particularly, the application of advanced techniques for analyzing patterns in spatial data. Based on the available data, they applied several different spatial analysis techniques. Some methods used a geostatistical approach, in which the phenomenon (e.g., existence of a mud unit) is considered as a random function and the data samples as a realization of that random function. This accounts for variability that is reflected by the structure of the spatial distribution of the phenomenon. Spatial correlation between the samples was analyzed and modeled using variograms (i.e., second-order moments of the distributions). The resulting model of spatial variability was then used in various stochastic simulation algorithms to create many different realizations. In addition to the geostatistical methods, three artificial intelligence techniques were applied including: probabilistic neural network method artificial neural network multilayer perceptron method support vector machine method During fiscal year 2002, scientists from Pacific Northwest National Laboratory and the Russian Academy of Sciences Institute for Nuclear Safety cooperated on a project to apply advanced geostatistical techniques to a groundwater model. These techniques are also known as machine learning algorithms in which a mathematical construct is trained using known data points. The algorithm then creates estimates of unsampled values based on its training, which are informed by the spatial variability of the original data. Artificial intelligence methods are model-free estimators. They are not based on a theoretical or statistical model. The output results depend only on the input data, network architecture, and type of learning used. A principal advantage of the artificial intelligence techniques is the ability to discover patterns that are imperceptible to standard statistical methods. Additional details of the geostatistical results and the development of alternative conceptual models are provided in Sections and Results from this Groundwater Modeling 4.0-3

4 research were presented at the International Containment and Remediation Technology Conference in Orlando, Florida (Thorne et al. 2001) and at the annual meeting of the International Association for Mathematical Geology in Berlin (Savelieva et al. 2002) Improvements to the Base-Case Model In fiscal year 2002, the site-wide conceptual model was modified to more accurately represent the groundwater flow system. The changes resulted in more realistic estimates of hydraulic parameters and helped the computer model fit historical water-level data more closely. Transient inverse calibration of the base-case, site-wide groundwater model (PNNL-13447) significantly improved the ability of the model to simulate historically observed water-level changes over the entire Hanford Site, particularly near major discharge facilities in the 200 West Area. However, some hydraulic parameters took on unrealistic values indicating that changes in parameter zonation or other conceptual model improvements were needed. Most parameter estimates derived from the transient inverse calibration were consistent with prior knowledge. However, estimates for the specific yield of the Hanford formation (from 0.06 to 0.07) and the Ringold Formation (between 0.20 and 0.21) were not consistent with current understanding of these sediments. Calibration of the alternative model, including interaction with the basalt-confined aquifer system (PNNL-13623), resulted in a slight measurable improvement in overall model fit over the entire prediction period when compared to the baseline model. Estimates of specific yield, however, continued to take on unrealistic values for both the Hanford and Ringold Formations, indicating that additional conceptual model improvements were needed. Based on these results, some adjustments were made during fiscal year 2002 to the base-case, site-wide model with the objective of producing realistic estimates of hydraulic parameters and better fits to historical water level data. The changes are summarized below: Account for delay of discharged wastewater caused by transit-time through the vadose zone. Create a separate hydrogeologic unit composed of gravels deposited on top of the Ringold Formation but prior to the Missoula floods. These gravels are found in the east-central portion of the site and were previously lumped with Hanford formation (Missoula flood) sediments to form model unit 1. Adjust hydrogeologic unit contacts southeast of 200 East Area, consistent with well data, to create a channel where saturated Hanford formation sediments exist below the water table. Adjust the extent of hydraulic parameter zones for the Hanford formation. Preliminary calibration of the refined base-case model resulted in significant improvements to the parameter estimates. However, refinement of the base-case model is continuing, particularly with regard to the distribution of parameter zones within the Hanford formation (model unit 1) Alternative Model Mud Unit Extent and Distribution One of the sets of alternative models under development during fiscal year 2002 addresses uncertainty in the extent and distribution of major mud units within the unconfined aquifer. Because of their low permeability, mud units within the aquifer system impede vertical movement of groundwater and migration of contaminants. Mud units also affect lateral movement, particularly where mud units exist at the water table; the three-dimensional distribution of mud units may cause contaminants to move deeper into the aquifer in some areas Hanford Site Groundwater Monitoring 2002

5 Three mud units are defined below the water table for the site-wide groundwater model (see Figure 4.1-1). Model unit 4 (upper Ringold mud) is composed of Lindsey s (BHI-00184) silt-dominated facies association III portion of the member of Taylor Flat. It does not include the sand-dominated facies association II. In the model, sand-dominated upper-ringold sediments generally occur beneath the silt-dominated sediments and are grouped with Lindsey s (BHI-00184) underlying gravel and sand-dominated units E and C to form model unit 5. The sanddominated sediments are expected to have similar hydraulic properties. Model unit 6 corresponds to fine-grained over bank and paleosol deposits described by Lindsey (BHI-00184) that separate geologic unit B from overlying unit C in the east part of the Hanford Site. Model unit 8 is equivalent to Lindsey s (BHI-00184) lower mud unit and forms an aquitard across much of the Hanford Site. This unit is a combination of fine-grained paleosols and lacustrian deposits. The mud in this unit is often described as blue or green, sticky clay, and frequently includes a white ash. Uncertainty in the distribution of mud units arises from both the possibility of misinterpretation and from actual spatial variability of the physical system. Uncertainty from misinterpretation can arise from the incorrect identification of a unit at a borehole or from a unit being missed in the borehole interpretation. Samples are often logged only every 1.5 to 3 meters, which may not be enough to intercept a relatively thin unit. Incomplete descriptions or errors in descriptions and field interpretations can also lead to misinterpretation of units at a borehole location. However, even if the interpretation of units at boreholes were 100% accurate, unit continuity between boreholes would be uncertain because of the variability of the aquifer system (i.e., the problem of interpolation and extrapolation under spatial variability). For example, an erosional hole may exist in a mud unit but is not represented in the model because no boreholes were in the area of the hole. On the other hand, a unit may be present in an area where no boreholes exist and not be represented in the model. The extent of mud units in the basecase model is based on data that are dense in a few areas and sparse over most of the site. The presence/absence of a particular mud unit is more uncertain in the areas of sparse data. Several geophysical methods discussed in Section were applied by researchers from IBRAE to explore the spatial variability of the presence/absence of each mud unit. The spatial analyses utilized data on the presence/absence of mud units at 405 wells, as well as the thickness of a unit if it was present. The first stage of the analysis focused on mapping the presence/absence of the mud, while the second stage focused on the variability of mud unit thickness. Maps showing probability of the mud unit presence/absence and thickness were then generated using several spatial methods. For example, Figures through are maps showing the probability of presence generated for each of the mud units using the probabilistic neural network method. A stochastic simulation method was applied to create a set of 100 realizations that span the range of likely extent/distribution for each mud unit. These initial simulations were based of the probabilities of presence. The realizations were merged to create a final set of 10,000 different possible combinations of the realizations. Modelers then ranked the merged realizations according to the total area of mud present and the tortuosity (i.e., a measure of connectedness) for each mud unit. Ranking the merged realizations provides a range of cases that can be selected from for use in model calibration. Figure shows cross sections of alternative conceptual models generated from the extreme (least mud and greatest mud) and median cases. These and other cases will be used in the inverse model calibration to create calibrated alternative conceptual models. The calibrated Because of their low permeability, mud units within the aquifer system impede vertical movement of groundwater and migration of contaminants. An alternative model in fiscal year 2002 investigated uncertainty in the extent of major mud units. Groundwater Modeling 4.0-5

6 models will then produce a range of results for hydraulic head and contaminant transport that can then be compared to historical measurements Alternative Model Parameter Zonation of Unit 1 Alternative models addressing uncertainty in the distribution of different Hanford formation sediments also were developed during fiscal year Alternative models addressing uncertainty in the distribution of different Hanford formation sediments (with significantly different hydraulic properties) also were developed during fiscal year The Hanford formation (model unit 1) is especially important in transport of groundwater contaminants to the Columbia River because of its high hydraulic conductivity compared to Ringold sediments. Therefore, the distribution of Hanford formation zones that are given unique values of hydraulic parameters in the model may have a large impact on simulations of contaminant movement from the Central Plateau to the Columbia River. The uncertainty was evaluated using a two-dimensional approach to make the problem more manageable and because the most conductive sediment type will likely be dominant at any particular location. However, this approach does not account for water-table changes over time that can cause different sediments to be dominant at different times. Uncertainty in the distribution of model unit 1 sediments comes from both the possibility of misinterpretation and from actual spatial variability. Uncertainty from misinterpretation can arise from the incorrect identification of a sediment facies type at a data location (well). Incomplete descriptions or errors in descriptions and field interpretations can also lead to misinterpretation. Uncertainty due to spatial variability arises because there are a limited number of data points and the sediment type in other locations is not known with certainty. Five different sediment types were identified for unit 1 that were expected to have significantly different hydraulic properties. These were silt, sand, and three different gravel types (G1, G2, and G3). Data was available at 229 wells. Researchers from IBRAE used mathematical methods to study the spatial distribution of sediment types for unit 1. They developed methods that used not only the spatial distribution of sediment type data, but also soft information including a confidence level for each data point and sediment-type boundaries expected on the basis of geologic depositional environment. Both geostatistical and machine learning methods were used to generate interpolated maps of the unit 1 zonation. A stochastic simulation method was used to create 100 different realizations that span the range of likely sediment-type distribution for model unit 1. The realizations were ranked according to overall hydraulic conductivity and tortuosity (a measure of connectedness). A hydraulic conductivity score was calculated for each realization by multiplying the area of each sediment type by the log of its mean hydraulic conductivity, as determined from historical aquifer tests. The conductivity score was given equal weight with the tortuosity in the ranking. Only the areas where unit 1 is within the model domain and below the water table were included in determining the rankings. Figures through show the distribution of different sediment types for the extreme (lowest and highest) and median conductivity realizations. These and other realizations will be calibrated to create alternative models. The calibrated models will then produce a range of results for hydraulic head and contaminant transport that can then be compared to historical measurements Hanford Site Groundwater Monitoring System Assessment Capability An initial assessment performed using the System Assessment Capability was completed in fiscal year Results including those for the groundwater module are presented in PNNL

7 The System Assessment Capability is an integration of several linked computer models designed to simulate the movement of contaminants from waste sites through the vadose zone, groundwater, and Columbia River to receptors and to then assess the risk to human health, other living systems, the local economy, and cultures. The System Assessment Capability starts with waste inventory and simulates contaminant release from the various waste forms. It also incorporates linked modules to simulate transport through the vadose zone, groundwater, and Columbia River. Additional modules calculate the risks. The assessment uses a stochastic analysis, which means that selected parameters are represented by probability distributions from which values are selected. The initial System Assessment Capability met its primary objectives. The original scope of the effort was to develop and successfully test a site-wide assessment capability addressing composite risks from a suite of representative Hanford contaminants for subsurface and surface water pathways over a 1,000-year period. For the initial assessment, the transport of 10 different radionuclide and chemical contaminants released from 890 wastes sites from 1944 through 3050 was simulated. The stochastic capability involves a systems approach and Monte Carlo analysis of significant operational, physical, chemical, biological, and socioeconomic features of the Hanford Site and its environs. Completion of the initial assessment demonstrates that a site-wide analysis can be completed. The groundwater module of the System Assessment Capability receives contaminant flux from the vadose zone module. It simulates contaminant movement through the uppermost aquifer system to the Columbia River and other potential exposure locations such as wells or seeps. The concentration of contaminants in groundwater is then used in the risk module calculations and contaminant flux is passed on to the Columbia River module. The groundwater module of the initial assessment completed in fiscal year 2002 used the base-case model described in Section A two-dimensional variablethickness version of the site-wide groundwater model had been used in earlier simulations done with the System Assessment Capability. Improvements in computer hardware made it possible to apply the three-dimensional model and still generate results in a reasonable amount of time. Use of the three-dimensional model led to significant improvements in the groundwater transport simulations. The System Assessment Capability is an integrated system of computer models and databases used to assess the impact of waste remaining on the Hanford Site. 4.3 Modeling to Support the Receptor Risk Model for Tank Farms The site-wide groundwater model was applied to determine the flow path and travel time for potential contaminant releases at each of the tank farms located in the 200 East or 200 West Areas to the Columbia River. Eighteen tank farms were assessed in this evaluation. Because of the model grid spacing (~375 meters) in the Central Plateau, releases from some tank farms were combined at single location, resulting in consolidation of the eighteen tank farms being considered to eight locations, five in the 200 East Area and three in the 200 West Area. The groundwater simulations were performed using the base-case model that had been calibrated to water level changes from 1944 to 1996 (PNNL-13641) (see Section 4.1.1). Because of the long-term nature of the simulations being made, the flow system was assumed to reflect natural steady state conditions after the effect of Hanford operational discharges have ceased. The applied Columbia River boundary conditions reflect long-term annual average flow and stage conditions. The simulations were based on a unit release at each tank farm for five discrete sorption coefficient (K d ) classes at each location. The sorption coefficient classes applied were 0, 0.2, 0.5, 0.8, 1, and 3. The 0 K d class represents a contaminant Computer models provide a sitewide context for cleanup decisions that must be made at individual waste sites. Groundwater Modeling 4.0-7

8 that moves with the groundwater flow. A K d class of 3 represents an upper limit at which the contaminant is strongly absorbed to the sediment matrix in the flow path. All simulations assumed that the contaminant is conserved (i.e., there is no decay). Simulated results showed that for all sites, the majority of the contaminant plumes move to the north, through the gap between Gable Mountain and Gable Butte toward the Columbia River. A lesser component of the plumes moves to the east toward the Columbia River south of Gable Mountain. Earlier flow modeling results by Cole et al. (PNNL-11801) suggested that as the water table drops in the central part of the Hanford Site and the saturated thickness of the unconfined aquifer decreases, groundwater flow northward from the 200 Areas may be cut-off by relatively impermeable basalt in the area just north of the 200 East Area. The water table is within a few meters of the currently interpreted basalt surface and in the natural recharge and boundary fluxes that control predictions of future water-table elevation. Therefore, the potential for movement of contaminants northward through the gap is also uncertain. These issues are currently being investigated as part of the site-wide groundwater modeling task. 4.4 Modeling for the Solid Waste Environmental Impact Statement A version of the site-wide model described in PNNL was applied to predict transport from low-level burial grounds located in the 200 West and 200 East Areas. This version of the model utilizes a distribution of hydraulic conductivity based on a steady-state calibration of the model. The contaminant source term for the modeling included both low-level waste that has been previously placed in the burial grounds and waste that is forecast to be placed in the burial grounds before The long-term assessment required estimating the cumulative dose using a suite of models that estimated source-term release, vadose zone flow and transport, and groundwater flow and transport. Results are presented in DOE/EIS-0286D. 4.5 Local-Scale Modeling of Pump-and-Treat Systems During fiscal year 2002, computer models that can evaluate the hydraulic effects of remedial action sites were updated to reflect the changing water-table elevation and changes in pumping rates at the pump-and-treat locations. The Hanford environmental restoration contractor has performed local-scale modeling during the past several years to design and evaluate pump-and-treat systems for groundwater. The Micro-FEM code was used to model capture and injection zones of extraction and injection wells, respectively, and to estimate the area affected by the pump-and-treat systems over time. The model was used to evaluate the hydraulic effects of the remedial action sites in several different operational areas. The operational areas and the contaminants of concern being treated at each are listed below: 100-KR-4 Operable Unit (100-K Area) hexavalent chromium 100-NR-2 Operable Unit (100-N Area) strontium HR-3 Operable Unit (includes both 100-D and 100-H Areas) hexavalent chromium 200-UP-1 Operable Unit (200 West Area) technetium-99 and uranium 200-ZP-1 Operable Unit (200 West Area) carbon tetrachloride Hanford Site Groundwater Monitoring 2002

9 During fiscal year 2002, these models were only updated to reflect the changing water-table elevation in the aquifer and changes in pumping rates. Additional information on these models is provided in DOE/RL and DOE/RL Local-Scale Modeling of Groundwater River Interaction Local-scale modeling of water movement in the zone of interaction between the unconfined aquifer and the Columbia River also has been conducted over the past few years to support the Groundwater/Vadose Zone Integration Project. The results of this modeling effort were published in PNNL Groundwater path lines were calculated to illustrate the direction and rate-of-flow within the zone of interaction, using Pacific Northwest National Laboratory s Subsurface Transport Over Multiple Phases (STOMP) code (PNNL-11218). Graphics software was then used to animate water movement and show how the flow field responds to the fluctuating river stage over one complete seasonal cycle of the river. The model was developed for the zone of interaction at the 100-H Area. This twodimensional simulation of water movement in the near-river unconfined aquifer clearly demonstrated the strong influence exerted by fluctuations in the Columbia River stage. The rise and fall of the river cause the direction and rate of groundwater flow to constantly change, with complete reversals in direction of flow and porewater velocities varying from no motion up to 10 meters per day. The dynamic nature of this flow field has implications for monitoring strategies, environmental restoration remedial actions, and river impact assessments (PNNL-11218). Computer models show how groundwater flow responds to fluctuating river stage. Groundwater Modeling 4.0-9

10 Elevation (m) West Area 200 East Area Water Table Unit 1 Hanford Gravel/Sand Unit 2 CO3-Rich Paleosol Unit 3 Cold Creek Gravel/Sand/Silt Unit 4, 6 and 8 Ringold Mud Unit 5, 7 and 9 Ringold Gravel/Sand/Silt Basalt Figure Cross Section of the Conceptual Model from West to East Across the Hanford Site Hanford Site Groundwater Monitoring 2002

11 Figure Map Showing Probability of Unit 4 Mud Presence Generated Using the Probabilistic Neural Network Method Figure Map Showing Probability of Unit 6 Mud Presence Generated Using the Probabilistic Neural Network Method Groundwater Modeling

12 Figure Map Showing Probability of Unit 8 Mud Presence Generated Using the Probabilistic Neural Network Method Hanford Site Groundwater Monitoring 2002

13 elevation (m) Least Total Mud gravel/sand mud basalt horizonal scale (m) Median Total Mud elevation (m) horizonal scale (m) Most Total Mud elevation (m) horizonal scale (m) Figure Cross Sections of Alternative Conceptual Models with the Least Total Mud Extent, Median Total Mud Extent, and Greatest Total Mud Extent Groundwater Modeling

14 S ilt Sand Gravel Type 3 Gravel Type 2 Gravel Type 1 Figure Alternative Distribution of Hanford Formation Sediment Types with the Lowest Overall Hydraulic Conductivity Silt Sand Gravel Type 3 Gravel Type 2 Gravel Type 1 Figure Alternative Distribution of Hanford Formation Sediment Types with the Median Overall Hydraulic Conductivity Hanford Site Groundwater Monitoring 2002

15 Silt Sand Gravel Type 3 Gravel Type 2 Gravel Type 1 Figure Alternative Distribution of Hanford Formation Sediment Types with the Highest Overall Hydraulic Conductivity Groundwater Modeling