VZJ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 20 November 2007
Published in Vadose Zone J 6:959-970 (2007)
DOI: 10.2136/vzj2007.0004
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Murray, C. J.
Right arrow Articles by Wilson, J. L.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Murray, C. J.
Right arrow Articles by Wilson, J. L.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Murray, C. J.
Right arrow Articles by Wilson, J. L.
Related Collections
Right arrow Variably Saturated Fluid Flow
Right arrow Vadose Zone Processes and Chemical Transport

SPECIAL SECTION: HANFORD SITE

Influence of Clastic Dikes on Vertical Migration of Contaminants at the Hanford Site

Christopher J. Murraya,*, Andy L. Warda and John L. Wilsonb

a Natural Resources Division, Pacific Northwest National Lab., P.O. Box 999, Richland, WA 99352
b Dep. of Earth and Environmental Science, New Mexico Institute of Mining and Technology, Socorro, NM 87801
* Corresponding author (Chris.Murray{at}pnl.gov).

All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.


Received 9 January 2007.



   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Properties of Clastic Dikes
 Materials and Methods
 Discussion and Conclusions
 REFERENCES
 
Clastic dikes are subvertical sedimentary features that cut through horizontally layered sediments, and they are common at the Hanford Site. Because of their cross-cutting relationship with the surrounding matrix, they have been proposed as potential fast paths from former contaminant discharge sites at the surface to the water table. However, little was known of the detailed hydrogeologic properties of the dikes, and detailed modeling of flow and transport through the dikes had not been performed. We excavated a 2-m-wide clastic dike at the Hanford Site and characterized it using an air minipermeameter, infrared imagery, and grain size analyses. Field injection experiments were also used to characterize the system. The resulting data were used to prepare a detailed numerical model of the clastic dike and surrounding matrix for a portion of the excavation. Unsaturated flow through the system was modeled for several recharge rates. The highly heterogeneous nature of the system led to complex behavior, with the relative flux rates in the matrix and clastic dike being highly dependent on the recharge rates that were imposed on the system. The occurrence of saturation-dependent complementary flow networks suggests that the contaminant release history may be important to the choice of remedial actions. Contaminants released under high flux conditions could be inaccessible under low fluxes, and vice versa. This phenomenon may also help explain the occurrence of complex breakthrough patterns of contaminants at compliance planes.

Abbreviations: GPR, ground penetrating radar • IR, infrared • TDR, time domain reflectometry • USGAO, U.S. General Accounting Office


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Properties of Clastic Dikes
 Materials and Methods
 Discussion and Conclusions
 REFERENCES
 
The Hanford Site produced large quantities of strategic radioactive materials, especially during the 25-yr period beginning in 1944. As a result of that national defense work, about 454 million liters of liquid waste containing chemical and radioactive contaminants were discharged directly to the ground, most of that in the 200 Areas, which comprise the present waste management areas (Gephart, 2003). In addition, large quantities of high-level waste are still being stored in the waste management areas in 177 large underground tanks. Sixty-seven of these tanks are known or suspected to have leaked 3.8 million liters or more of contaminated liquid to the soil (Gephart, 2003). Original assumptions suggested that little of these tank wastes would reach the groundwater because of the combined effects of low recharge rates and the tendency for the soils to geochemically bind contaminants from the migrating fluids. However, a number of recent studies indicate that contaminants have migrated to greater depths at the Hanford Site than expected. This includes the possible migration of 137Cs in the vadose zone at the S-SX tank farm (Johnson and Chou, 1998). Other indications of more rapid vertical contaminant migration than originally expected are the presence of technetium, carbon tetrachloride, and other mobile contaminants in groundwater beneath the waste management areas (USDOE, 1998). The discrepancy between predicted and observed travel times could be the result of preferential pathways caused by geologic features, leaky water lines, improperly sealed boreholes, and surface run-on. Of the geologic features leading to preferential pathways (e.g., clastic dikes, fractures, faults, inclined bedding), clastic dikes are the principal features known to exist around tank waste leak sites and are the potential pathway investigated here.

Clastic dikes are vertical fissures filled with clay to gravel-sized sediments that cross-cut the host sediment (Fecht et al., 1998). Vertical clastic dikes occur at many locations in the waste management areas at Hanford (Fecht et al., 1998) and have been proposed as potential pathways for vertical contaminant transport (Johnson and Chou, 1998). The vertical orientation of the clastic dikes provides a possible avenue for vertical transport of contaminants that may be faster than transport through a horizontally layered sediment sequence. For example, at one field site, flow rates were found to be more than 10 times faster in a clastic dike than flow through the enclosing horizontally layered sediments (Fecht et al., 1998).

In 1998 the U.S. General Accounting Office (USGAO) conducted an audit of the vadose zone activities at the Hanford Site and found that the USDOE was not adequately performing its stated mission of protecting human health and the environment (USGAO, 1998). The USGAO forcefully stated the need for characterization of the vadose zone as an essential step toward understanding groundwater contamination and engendering credibility in the USDOE's ability to handle contamination issues at the site. According to the USGAO report, one area that was not being addressed was "narrow pathways" that might promote the movement of contaminants more rapidly than expected. Clastic dikes are such narrow pathways and have the potential to promote rapid contaminant movement.

Earlier flow and transport models for contaminant migration in the vadose zone at the 200 Areas were based on relatively simple hydrogeologic models that assumed horizontally layered sediments (e.g., USDOE, 1996), with no preferential vertical flow paths. In a recent modeling study of the SX tank farm, Ward et al. (1997) acknowledged that the probability of encountering a clastic dike (beneath SX) was substantial. However, they concluded that information on dikes and their hydrologic properties at Hanford is "more or less qualitative rather than quantitative. Modeling of the effects of a clastic dike then becomes somewhat subjective." Without data, Ward et al. (1997, p. 3.1) chose not to include clastic dikes in their model analysis. Recent work at the Hanford Site has included sensitivity cases with clastic dikes as potential preferential pathways (Wood et al., 1995, 1996; USDOE, 1999; Mann et al., 2001; Knepp, 2002; Connelly, 2005). However, these models have not incorporated the heterogeneity of physical properties that occur within the dikes.


   Properties of Clastic Dikes
TOP
 ABSTRACT
 INTRODUCTION
 Properties of Clastic Dikes
Materials and Methods
 Discussion and Conclusions
 REFERENCES
 
Clastic dikes are vertical to subvertical sedimentary structures that cross-cut normal sedimentary layering (Fig. 1 ). They are found in a wide range of geologic environments and ages and occur as both isolated intrusions and swarms (Maltman, 1994). Clastic dikes are a common geologic feature of the Pleistocene flood deposits of the Hanford formation, although they also have been found in the underlying Ringold Formation and the Columbia River Basalt Group and intercalated sedimentary interbeds. Individual clastic dikes commonly cut more than one stratigraphic unit.


Figure 1
View larger version (111K):
[in this window]
[in a new window]

  		FIG. 1. Photograph of clastic dike cutting horizontal layers of Hanford formation.

 
Clastic dikes are common features at the Hanford Site (Fecht et al., 1998). Price and Fecht (1976) observed clastic dikes at most of the single-shell tank farms and throughout the waste management areas and were able to map them at the 241-SY tank farm. Fecht et al. (1998) documented the presence of clastic dikes at the Fast Flux Test Facility, the U.S. Ecology low-level waste burial trenches, the Environmental Restoration Disposal Facility, the tank waste vitrification privatization site, and in several boreholes. In addition, several clastic dikes were mapped in the floors and walls of the Integrated Disposal Facility trench (Reidel and Fecht, 2005). Despite their numerous occurrences and potentially important hydrologic implications, most clastic dike studies have been largely descriptive, and little consensus has been reached on their origin.

Clastic dikes in the Columbia Basin typically occur in swarms and form four types of networks: (i) regular-shaped polygonal patterns, (ii) irregular-shaped, polygonal patterns, (iii) preexisting fissure fillings, and (iv) random occurrences (Fecht et al., 1998). Regular polygonal networks resemble 4- to 8-sided polygons and typically range from 3 cm to 1 m in width, from 2 to >20 m in depth and from 1.5 to 100 m along strike. Smaller dikelets, sills (horizontal instrusions), and small-scale faults and shears are commonly associated with master dikes that form the polygons. The regular-shaped polygonal clastic dike network is the dominant type near the waste management areas at Hanford.

Several features of clastic dike networks suggest that they could influence the migration of vadose zone fluids. These include the geometry and size of the dikes, the nature of the materials that fill them, small structures associated with the dikes, and the relationships of clastic dikes with clastic sills and host sediments. Polygon size seems to be related to the host sediment texture and bedding characteristics. The more variable the texture and bedding in the host sediment, the more variable the polygon size. Also, in many regular-shaped polygon networks, the widths of the major dikes seem to be related to the widths of the polygons such that polygon dimensions increase as dike width increases (Fecht et al., 1998). In any given polygonal-patterned dike network, the lengths of most polygons are within 20% of the median polygon length. Taken as a whole, clastic dikes at the Hanford Site typically show a wide range of widths, depths, and lengths. The vertical extent of clastic dikes has been observed to range from 30 cm to >55 m. Clastic dike widths ranges from about 1 mm to >2 m, and their length varies from as little as 0.3 to >100 m.

In general, a clastic dike is composed of an outer skin of clay with coarser infilling material. Clay linings are commonly 0.03 to 1.0 mm in thickness, but linings up to about 10 mm are known. The clay skins may have a great influence on transport and sorption of waste both within and adjacent to the clastic dikes. The width of individual infilling layers ranges from as little as 0.01 mm to more than 30 cm, and their length can vary from about 0.2 to >20 m. Infilling sediments are typically poor to well-sorted sand but may contain clay, silt, and gravel. Further quantitative information on the composition as well as the spatial distribution and shapes of clastic dikes can be found in Murray et al. (2003) and Fecht et al. (1998).

Hydrologic Properties of Clastic Dikes
Little detailed quantitative work has been done on the hydrologic properties of clastic dikes at Hanford. Fecht et al. (1994, 1998) reported the results of several small-scale field tests designed to measure the hydraulic conductivity of clastic dikes and their host sediments. They found effective hydraulic conductivities (Kfs) in the center of dikes to be about 10–3 cm s–1, horizontal conductivity (Kx) across clay linings to vary from 10–4 to 10–7 cm s–1, and Kfs of host sediments to be about 10–3 m s–1. Three methods were used in these tests: mapping the wetting front, a Guelph permeameter, and an ultracentrifuge. Mapping the wetting front might indicate the general hydraulic nature of the sediment, but it does not provide quantitative measurements of hydraulic properties. The Guelph permeameter method assumes a certain infiltration pattern away from the infiltration hole so that a specific analytical solution can be used to estimate the saturated conductivity. That assumption is not valid for clastic dikes if they behave differently from the surrounding matrix. The third technique, the ultracentrifuge, is useful, but caution should be exercised with those results because sampling tends to disturb the dike structure and the ultracentrifuge samples are very small (1–2.5 cm).

A major concern at Hanford is the possibility that clastic dikes could provide a fast vertical conduit to groundwater by short-circuiting the anisotropy of the layered sediments. The anisotropy ratio, the ratio of horizontal conductivity (Kx) to vertical conductivity (Kz), within the horizontally bedded Hanford sediments is unknown but was estimated by Connelly et al. (1992) at approximately 10:1. Within clastic dikes, layering is vertical, and anisotropy ratios can be expected to be much smaller, and even less than 1:1. This is due to the presence of clay skins along the walls of the dikes, which would act to reduce Kx and prevent lateral spreading of water. Thus, these features could represent a very significant increase in vertical transport over normal sedimentary conditions. Field observations suggest that dikes could act as preferential pathways for the movement of liquids; however, reliable hydrologic data do not exist to quantify this effect.

Given the uncertainty with regard to the properties of clastic dikes, and their potential impact on transport of contaminants to groundwater, a study was performed to address the geometry and internal properties of clastic dikes and their effect on vertical transport. This included excavating a clastic dike, taking detailed measurements, performing an infiltration test, and flow modeling. Although the properties of clastic dikes are variable, detailed data and modeling on flow through a clastic dike at Hanford should provide information that bear on the scientific hypothesis that clastic dikes at the Hanford Site waste management areas provide vertical preferential pathways for the movement of water and dissolved contaminants through the vadose zone. This paper presents some of the results of that study, including detailed measurements of air permeability and model results for relative rates of fluid movement in clastic dikes and surrounding matrix at varying recharge rates.


   Materials and Methods
TOP
 ABSTRACT
 INTRODUCTION
 Properties of Clastic Dikes
 Materials and Methods
Discussion and Conclusions
 REFERENCES
 
Experimental Field Site
In 2001 we performed a study of the small-scale hydrogeologic properties of clastic dikes at a site near Army Loop Road (Fig. 2 ). A ground penetrating radar (GPR) survey was used to supplement air photo and field mapping to select a site to trench across a clastic dike (Murray et al., 2003). The dikes in the Army Loop Road area appear to be representative of regular-shaped polygonal clastic dikes, the most common type of clastic dike at the Hanford Site (Fecht et al., 1998). In June 2001 a clastic dike at the Army Loop Road site was trenched with a backhoe to a depth of ~3.5 m. The exposed clastic dike cut vertically through sediment layers in the sand-dominated facies of the Hanford formation. The dike excavated at the Army Loop Road site was approximately 2 m wide.


Figure 2
View larger version (38K):
[in this window]
[in a new window]

  		FIG. 2. Map of clastic dike networks exposed at the Hanford Site. Clastic dikes appear to be common across the entire Hanford Site but are usually covered by surface sediment deposits or cultural features. Field experiments described here were conducted at the Army Loop Road site.

 
The dike was excavated in three different levels, each approximately a meter high, labeled Tier 1, Tier 2, and Tier 3 from top to bottom of the excavation. After the excavation of each tier, the slopes surrounding the excavation were pushed back, and the excavation was taken down another level. In this way, we were able to record images of the excavation faces and measure the properties of cross-sections of the dike and matrix on the three different tiers of the excavation, which were approximately one on top of the other, without creating a safety hazard in the unstable sediment. The face exposed at each level was mapped, and sediment samples were taken for laboratory analysis. We captured images of each face using digital thermal infrared (IR) camera and 35-mm cameras, measured air permeability and hydraulic conductivity, and collected sediment samples for measurement of grain size, mineralogy, and hydraulic properties.

Measurement of Geological and Hydrogeological Properties
A number of methods were used to measure or estimate geological and hydrogeological properties of the clastic dikes and enclosing matrix sediments. Techniques included intrinsic air permeability by minipermeameter; hydraulic conductivity by a minidisk infiltrometer; grain size distributions; moisture content using GPR, neutron probe, and time domain reflectometry (TDR); and matric potential using tensiometers. Mineralogy was characterized using X-ray diffraction, and correlations were calculated between IR imagery and the various measured physical properties.

We used an air minipermeameter system (Tidwell and Wilson, 1997) to measure the air permeability of clastic dike and matrix sediments exposed in the excavations at the Army Loop Road site. Members of the project team from the New Mexico Institute of Mining and Technology made substantial improvements to the air minipermeameter system used for the measurements. The standard LSAMP II air minipermeameter developed by the New Mexico Institute of Mining and Technology had a practical range of ~6.2 x 10–13 to 2.4 x 10–10 m2, corresponding to fine to medium sand. The system was modified for this study, extending the range by almost an order of magnitude so it could be used to make measurements in some of the finer-grained sediments present in the clastic dikes.

A total of about 450 measurements were made on the three tiers of the excavation, one-third in the dike and two-thirds in the host matrix. The results indicate the median air permeability of the dike is about an order of magnitude lower than the permeability in the matrix. The variability of the data from the dike is much higher than that of the matrix, with a coefficient of variation (i.e., ratio of standard deviation to the mean) of 1.2 in the dike compared with 0.6 in the matrix. The overall variability of air permeability in the dike–matrix system is about four orders of magnitude (Fig. 3 ). This is an important observation because some methods used for upscaling permeability data assume the variability in the system is low, about an order of magnitude, which means that applying those methods to the clastic dike and its surrounding sediments would be questionable.


Figure 3
View larger version (15K):
[in this window]
[in a new window]

  		FIG. 3. Air permeability measurements from Army Loop Road excavation.

 
In addition to the air permeability measurements, Kfs was measured in situ on Tier 2 using the Guelph permeameter (Reynolds and Elrick, 1985). Guelph permeameter measurements showed a mean Kfs for the host matrix of 3.24 x 10–2 cm s–1 and 2.58 x 10–3 cm s–1 for the dike. The mean Kfs in an adjacent sill, a dense layer of horizontally laminated fine-textured sediments, was 5.9 x 10–4 cm s–1. Several measurements of Kfs were also made every 0.3 m along a 7.5-m transect on Tier 2 using a minidisk infiltrometer using deionized water (Zhang, 1997). Results show an order of magnitude difference between the sandy host matrix (2.29 x 10–2 cm s–1) and the composite dike (fine plus coarse vertical layers) region (2.04 x 10–3 cm s–1). The mean of measurements made on isolated fine-textured skins within the dike wide enough to fit the minidisk were almost five times smaller (4.68 x 10–3 cm s–1) than the host sand. The mean Kfs in the dike (fine plus coarse) material was 1.48 x 10–6 cm s–1, and the coefficient of variation was much higher (2.1) than in the matrix (0.9).

Long et al. (1996) successfully applied imaging techniques using IR wavelengths to map detailed permeability distributions in porous media at outcrop scales, with resolution in the millimeter range. They related slight temperature differences in exposed sediment faces to differences in evaporative cooling, which appear to reflect available surface moisture. In sandy material, moisture retention is largely a function of unsaturated permeability, with more permeable sediments exhibiting higher temperatures and lower permeability sediments exhibiting lower temperatures (Long et al., 1996). Infrared and digital photographic images were collected under ambient light conditions in the excavation, and the IR temperature response was correlated to independent observations of permeability collected using the air minipermeameter.

Bulk samples were collected from the trench faces and analyzed for mineralogy, water content, water retention, particle size distribution, and particle density. Particle size analyses were conducted on the <2-mm size fraction using a combination of sieving and hydrometer techniques (Gee and Bauder, 1986). Dry sieving was performed using sieves with openings of 2, 1, 0.5, 0.25, 0.106, 0.075, and 0.053 mm, with final collection in the pan. Silt- and clay-size fractions were determined by hydrometer. Particle density was also performed on the <2-mm size fraction using the pycnometer method (Blake and Hartge, 1986). Bulk density ({rho}b) measurements were conducted on undisturbed samples using the clod method. Porosity ({phi}) was estimated from {rho}b and the mean particle density ({rho}a) as {phi} = 1 – {rho}b/{rho}a.

Field Infiltration Experiments
A large-scale infiltration experiment was conducted at the Army Loop Road site in 2001 (Ward et al., 2006). A drip irrigation system was used to apply the specified fluxes. The application area was centered on the dike and aligned with the longer axis perpendicular to the dike (Fig. 4 ). Three fluxes of water were applied to the clastic dike and surrounding matrix, and the progress of the infiltrating water was monitored for each flux rate. Water content, matric potential, and electrical conductivity were measured throughout the tests using a neutron probe, cross-borehole radar, tensiometers, and TDR probes. The TDR probes and tensiometers were installed vertically to a depth of 0.5 m on a plot oriented perpendicular to the dike. Five rows of multipurpose TDR probes (TDR + tensiometers) were installed with a spacing of 30 cm along the length of the plot (perpendicular to the dike) and 50 cm along the width (parallel to the dike).


Figure 4
View larger version (33K):
[in this window]
[in a new window]

  		FIG. 4. Layout of infiltration area and monitoring instruments. TDR, time domain reflectometry.

 
The three fluxes applied were 0.001, 0.01, and 0.1 Ks of the host matrix, in that order. For typical soils, this translates roughly into flux values of 10–5, 10–4, and 10–3 cm s–1. Similar fluxes have been used in previous field tests of surface soils at the Hanford Site (e.g., Khaleel, 1999). These fluxes are a compromise between the length of time it would take to run an experiment at input fluxes representative of Hanford recharge conditions (0.1–100 mm yr–1 or 3.2 x 10–10 to 3.2 x 10–7 cm s–1) and project time constraints. Relating each flux to the resulting equilibrium water content, {theta}, provides a measure of the unsaturated conductivity function, K({theta}) (Youngs, 1964). The water content and matric potential ({psi}) data provide a direct measure of in situ water retention.

Once steady state was achieved with the third (and highest) flux rate, the irrigation supply tank was switched to a solution of KBr and Brilliant Blue FCF dye. The effect of KBr on the electrical impedance of the soil allowed tracking of solute movement and the construction of solute breakthrough curves using TDR measurements. Irrigation continued until the KBr moved below the TDR sensing zone (about 0.5 m). Further movement of the water was monitored with neutron probe and cross-borehole radar measurements.

The excavation began after the application of the tracer in the infiltration area. The main excavation face was approximately 8 to 10 m from the edge of the infiltration zone, so that the moisture would not affect the air permeability measurements or IR thermal imaging. However, after construction of the main excavation area was complete, an additional face was cut at the edge of the infiltration area so the distribution of the tracers could be examined. The upper portion of Fig. 5 shows a composite color photographic image of what we term the "dye" face, with the lower portion of the figure being a map of the moisture distribution in the face. The photographic image shows the very heterogeneous distribution of the blue dye. The dike is in the center-right area of the image, from 3 to 5 m, and shows the highest moisture contents but appears to transmit less dye. In fact, the deepest penetrations of the dye occur in restricted bands within the dike that coincide with regions of coarser texture (Fig. 5).


Figure 5
View larger version (48K):
[in this window]
[in a new window]

  		FIG. 5. Top: Composite image of dye penetration. Bottom: Contour map of clastic dike. Water content measurements for surface are shown in upper image.

 
The map of the moisture distribution in the lower portion of Fig. 5 was made using TDR probe measurements on a 20-cm horizontal and 10-cm vertical grid across the entire face. The moisture map captures the main features of moisture distributions seen in the photographic image with the fine-textured regions being wetter and the sandy regions being drier. However, these results highlight some difficulties in characterizing flow in heterogeneous unsaturated porous media over the multiple scales of heterogeneity that appear to control flow. Owing to the spacing of TDR probes and that water content is averaged over the length of the vertical probe, many of the small-scale features are averaged out. Nevertheless, the difference between the distributions of dye and moisture reflects the nonlinearity in hydraulic properties and the dependence on texture. In heterogeneous systems, there can exist complementary transport domains that depend on saturation. In this case, we hypothesize that there are two complementary networks: (i) a fine-textured high-permeability network that is active at relatively low fluxes owing to a K({theta}) higher than the coarser host matrix; and (ii) a coarse-textured, high-permeability network that is active at high fluxes when the K({theta}) of the coarse host matrix exceeds that of the fine-textured dike. These complementary flow networks can mask the underlying heterogeneity as water is redirected around regions of lower permeability at a given flux, depending on saturation.

As pointed out by Roth (1995), this phenomenon is essential to the differences observed between unsaturated flow in heterogeneous systems and flow in homogenous or in saturated systems. The discrepancy between the water and dye patterns in Fig. 5 is a reflection of such a mechanism. Dye was transported into the coarse-textured regions under the higher flux conditions, bypassing the dike, because of the higher K({theta}). Following cessation of irrigation, these coarse-textured regions also drained faster than the dike region, leaving the dye in place. The dike remained wetter because of the lower permeability and was generally free of dye except for the coarse-textured inclusions. Thus, it appears that clastic dikes could very well enhance vertical flow, but only under high flux conditions in which water is able to move in the coarser zones as the fine-textured regions of the dikes diverts water to the higher K({theta}) zones. To be a major path to groundwater, these coarse-grained zones would have to be continuous over a significant depth interval. Under typical Hanford conditions, K({theta}) of the fine-textured regions would be higher than the coarse regions, but the actual values (≤10–6 cm s–1) would be too low to cause contaminants to move significant distances.

Modeling of Field Infiltration Experiments
Geologic Interpretation and Numerical Model Development
A combination of particle size analysis and digital photogrammetry was used to build a two-dimensional geologic model of the excavation in a five-step procedure. In the first step, high-resolution digital visible and infrared images were compiled into a mosaic with a resolution of 0.111 cm per pixel. Intrinsic air and water permeabilities measured in the field were assigned to their spatial locations on the mosaic. We found that the IR temperature data and the air permeability data collected from the dike and surrounding matrix at the Army Loop Road were positively correlated with one another, with a linear correlation coefficient of 0.73 (Fig. 6 ). This analysis was performed on a normal score transform of the data. The normal score transform is a graphical algorithm that transforms a variable so that its histogram is Gaussian with a mean of 0 and a variance of 1 (Deutsch and Journel, 1998). Variogram analysis of the normal scores of the two datasets indicated that the spatial continuity of the air permeability and IR temperature data were very similar. The relationship between IR temperature and air permeability (Fig. 6) was used to predict the air permeability across the entire domain by kriging using the variogram model derived from the IR temperature distribution of the mosaic (Fig. 7 ). Discontinuities between panels in the IR mosaic were first removed by filtering using digital photogrammetric techniques (Seedahmed and Ward, 2005).


Figure 6
View larger version (24K):
[in this window]
[in a new window]

  		FIG. 6. Cross-plot of air permeability data and infrared (IR) data.

 

Figure 7
View larger version (61K):
[in this window]
[in a new window]

  		FIG. 7. Spatial variability for Tier 2 of the excavation. (a) Spatial variation in soil textures from a digital camera; (b) measured infrared temperatures; (c) vertical saturated hydraulic conductivity, inferred from the relationship between geometric mean grain size and intrinsic permeability.

 
Particle size distributions derived from the outcrop samples were also assigned to corresponding locations on the mosaic. We constructed scatterplots of IR temperatures, air permeability, and intrinsic permeability versus geometric mean grain diameter and determined the correlations between the variables so as to generate a distribution of grain size statistics (mean grain diameter, dg, and geometric standard deviation, {sigma}g). The cooler temperatures in Fig. 7b represent areas of higher moisture content, which are mostly regions occupied by silt and silt loam. These soils represent the fine-textured high-permeability networks under the ambient, low-recharge conditions present at the Hanford Site. The warmer temperatures represent the drier, coarse-textured sediments that would form the high-permeability networks under high fluxes. Sediment types were differentiated by texture based on the content of sand, silt, and clay according to the USDA textural triangle.

Figure 8 shows a log-probability plot of particle diameter, d, as a function of cumulative mass percent less than d for several samples. The wide range of textures exemplifies the variability at the field site. Even though these samples were all from material that is from the sand-dominated upper Hanford formation, or appears to be derived from that material, there were many samples with size distributions reflecting a finer texture than the typical Hanford sand. Soil textures, derived from the ratio of sand, silt, and clay, ranged from coarse sands (AL-HT2-1) with >95% sand for a sample taken from the matrix sediment surrounding the dike, to medium and fine sands (AL-CDT2-2) in coarse segments of the clastic dike, to loam and silt loam (AL-CD-T2-7) with as much as 19% clay in the fine-grained portions of the clastic dike. This wide variation in particle size distribution within clastic dike material can be expected to influence water retention characteristics and permeability and, ultimately, the flow and transport properties under unsaturated conditions.


Figure 8
View larger version (40K):
[in this window]
[in a new window]

  		FIG. 8. Comparison of particle size distribution from <2-mm fractions from the Army Loop Road field site. Essentially, five soil types were identified from the analyses: silt loam (e.g., IN-A1-5, •), loam (e.g., AL-CDT3-1, {blacksquare}); sandy loam (e.g., AL-CDT2-8, {diamondsuit}); loamy sand (e.g., AL-CDT2-5, {circ}); and sand (e.g., Al-HT1-1, {square}).

 
Simulation of Water Flow
The water operational mode of the STOMP simulator was used to simulate water flow. The STOMP simulator is a three-dimensional, fully implicit, integrated finite difference code that has been used to simulate a variety of multiphase (White and Oostrom, 2000; Oostrom and Lenhard, 1998) and variably saturated flow systems (Ward et al., 1997; Rockhold et al., 1999; Zhang et al., 2003, 2004). This simulation mode assumes the porous medium to be incompressible, and fluid properties are assumed to be time invariant. The governing partial differential equation was discretized following the integrated-volume finite difference method by integrating over a control volume. Using backward Euler time differencing to yield a fully implicit scheme, a series of nonlinear algebraic expressions was derived. The algebraic forms of the nonlinear governing equations were solved with a multivariable, residual-based Newton–Raphson iterative technique. In this series of simulations, internodal conductance was computed as the product of the harmonic mean of the saturated permeability in adjacent grid blocks and upgradient averaged relative permeability. Harmonic averaging was used for all other flux components. The maximum number of Newton–Raphson iterations was set to eight, while the tolerance was set at 10–6.

Water flow was simulated on a two-dimensional vertical outcrop of the dike (Tier 2 of the Army Loop Road excavation; see Fig. 7), 4.7 m long and 2 m deep, encompassing the dike and surrounding host matrix (Fig. 9 ). Simulations were performed with specified flux conditions at the upper boundary, no flow conditions on the east and west, and a unit gradient condition at the lower boundary. To represent the local-scale flow and transport heterogeneities, grid resolution must be consistent with the local scale processes. Thus, node spacing was uniform and ranged from 0.111 cm in a microscale model of only the dike region to as much as 10 cm in the coarsest upscaled model of the entire outcrop. The total number of nodes in each grid ranged from more than 3.5 million nodes in the microscale model of the entire transect to 8600 in the coarsest upscaled model. Seven cases were simulated with fluxes ranging from 0.1 to 105 mm yr–1 (Table 1).


Figure 9
View larger version (30K):
[in this window]
[in a new window]

  		FIG. 9. Flow domain showing boundary conditions. The surface boundary condition is a specified flux, Jw0, whereas the bottom and side boundary conditions are specified as no flow conditions by setting the gradient in matric potential, {psi}, in the x and z directions equal to zero.

 

View this table:
[in this window]
[in a new window]

  		TABLE 1. Summary of cases simulated and boundary type.

 
Model Parameters
Secondary variables—that is, those parameters not directly computed from the solution of the governing equations—were computed from the primary variables through the constitutive relations. In this study, the Brooks–Corey model (Brooks and Corey, 1964) was used to represent the relationship between hydraulic conductivity, water content, and matric potential (K–{theta}–{psi}). Hydraulic conductivity was computed as a function of water content using a modified form of Burdine's (1953) relative permeability model, integrated using the Brooks and Corey (1964) relationship.

The values for the key parameters for the moisture retention curve ({theta}–{psi}) and the unsaturated hydraulic conductivity function K({theta}) were determined for each model grid block based on the assigned particle size distribution. The moisture characteristic of a given soil has been shown to be dependent on the pore size distribution, which is in turn related to the proportions of each particle size class and the bulk density (Campbell, 1974). Particle size distributions were assumed to be log normally distributed with a geometric mean grain diameter, dg, and a geometric standard deviation, {sigma}g. These parameters were computed from particle distributions by calculating the contribution of each of the textural fractions sand, silt, and clay. The power of the {theta}({psi}) curve, {lambda}, and the matric potential at air entry or bubbling pressure, {psi}b, were also estimated from the particle size data according to Campbell (1974). The saturated water content, {theta}s, was estimated as the product of {phi} and the effective porosity. Typical values of effective porosity range from 0.80 to 0.95 (Arya et al., 1999). Measured grain size distributions were analyzed to determine geometric parameters defined above and to predict hydraulic properties at unsampled regions of the experimental site. Relationships between the intrinsic water and air permeabilities were used with dg to interpolate the intrinsic water permeability. For the microscale model, the hydraulic conductivity field was assumed isotropic at the local scale. There was a disparity between the scale of measurement and the scale at which the flow domain could be discretized for simulation in a reasonable time. The dimensions of the full domain for Tier 2 was 3.59 million nodes. The combination of infrared and visible images coupled with permeability measurements provided estimates of permeability on a grid on 0.111 cm. Thus, constitutive properties were upscaled from the IR "measurement" scale to the grid block using methods described in Rockhold et al. (1999, 2002) and Meyer et al. (2002). Briefly, in this upscaling approach, the principal components of the effective saturation, Se, and the conductivity tensor Ke,jf are approximated for the model grid block (coarse scale) from the fine scale as a function of the variable {gamma}e,jf({psi}):

Formula 1[1]
In Eq. [1], the superscripts c and f represent the coarse- and fine-scale values, respectively. When {gamma}e,jf({psi}) represents the effective saturation, Se,jf, at the fine scale, the upscaled (coarse-scale) effective saturation is given by fec({psi}). Similarly, when {gamma}e,jf({psi}) represents the fine-scale hydraulic conductivity, Ke,jf, the upscaled hydraulic conductivity is given by fec({psi}). The upscaled conductivity of a model grid block is rendered anisotropic by virtue of length and area weighted averaging. Length-weighted harmonic means of K at the scale of "measurement" on a fine grid were first calculated for each grid cell for an orientation parallel to the flow direction. Area-weighted arithmetic means of these values were then computed for the horizontal direction according to Rockhold et al. (1999).

Figure 10a shows typical water retention curves for the range of textures observed at the outcrop. The characteristics derived from the granulometric data capture the essential features of heterogeneous unsaturated soils. These curves show a clear dependence of saturated water content, {theta}s, on texture and a dependence of water content, {theta}, on the matric potential, {psi}. In general, {theta} is higher in the fine-textured silt loam and silts than in coarse-textured sands across the range of matric potentials.


Figure 10
View larger version (20K):
[in this window]
[in a new window]

  		FIG. 10. Hydraulic functions for five typical soils from the Army Loop Road Dike Site: (a) water retention, {theta}({psi}), and (b) hydraulic conductivity, K({psi}). Textures included silt loam (IN-A1-5), loam (AL-CDT3-1), sandy loam (AL-CDT2-8), loamy sand (AL-0CDT2-5), and sand (AL-HT1-1). Matric potential, {psi}, is in meters, and hydraulic conductivity is in centimeters per second.

 
For large negative values of {psi} [i.e., large log (–{psi})], {theta} in the finer sediments can be three times higher than in the coarser sediments. In addition, the range of air entry pressures, {psi}b, from about 5 cm for coarse sands to >60 cm for silt loam is consistent with laboratory and field measurements. These characteristics will impact the relative permeability functions as shown below.

Figure 10b shows typical hydraulic conductivity functions for the range of textures observed at the outcrop. For small negative values of {psi} [i.e., small log (–{psi})], at and near saturation, K of the coarse sediments is higher than that of the finer sediments. The air entry pressure is also lower in the coarse sediments. This relation is reversed as {psi} decreases [i.e., as log (–{psi}) gets larger]. As {psi} decreases, large pores empty earlier and the sediment becomes less permeable.

The higher air entry pressure of the fine-textured sediments cause them to retain water at lower values of {psi}; consequently, K of the fine-textured sediments exceeds that of the coarser sediments. There is a value of {psi} at which K for the different textures are equal; this value of {psi} would correspond to the crossover pressure defined by Rucker et al. (2005). Different soils will therefore conduct water at the same rate during constant flux infiltration only if the boundary flux, Jw0, corresponds to the associated value of K({theta}). This phenomenon is the fundamental difference between unsaturated flow in heterogeneous systems and flow in homogenous and saturated systems. Under typical field conditions in which flow is transient, this phenomenon will cause flow networks and transport behavior to change with the input flux and even boundary type. Depending on the degree of heterogeneity, flow networks and transport rates will change with saturation, can be equal in different soil textures, and can even alternate between soil textures, depending on the surface flux and boundary type. Because a limited amount of data are typically available for modeling the effects of clastic intrusions across multiple scales, modelers have often resorted to using average or effective properties. Such an approach is incapable of reproducing the complex flow patterns observed due to the presence of clastic dikes and will lead to erroneous conclusions about their effect on field-scale transport.

Subsurface Moisture Patterns and Their Relation to Surface Flux
Figure 11 shows typical simulation results for the subsurface distribution of water content in response to steady input fluxes ranging from 10–10 to 10–5 cm s–1. As shown in Fig. 7, the flow domain exhibited heterogeneities over a wide range of spatial scales. This heterogeneity is reflected in the distribution of water content, {theta}. Values of {theta} vary over the entire range of possible values, even at the lowest flux (Fig. 11a). The gradient {nabla}{psi}, together with gravity, is the driving force of the water flux, and this flux will tend to minimize the potential gradients. In contrast, {theta} plays no role in the driving force for unsaturated flow and simply adjusts to bring about the required potential. Consequently, {psi} is a smooth function even in heterogeneous systems while {theta} varies in space as a reflection of the underlying heterogeneity. As the surface flux is increased, the effect of the fine-scale heterogeneities on {theta}, for example, within the dike, is less apparent. However, differences between the coarse and fine textures persist.


Figure 11
View larger version (89K):
[in this window]
[in a new window]

  		FIG. 11. Simulated distributions of volumetric water content ({theta}) during steady infiltration under a constant surface flux of (a) 1 mm yr–1, (b) 10 mm yr–1, (c) 102 mm yr–1, and (d) 103 mm yr–1. Domain properties were upscaled to 2-cm grid.

 
Simulated water contents of 0.15 to 0.20 m3 m–3 in the coarser sediments and 0.30 to 0.40 m3 m–3 in the fine sediments for Jw0=103 mm yr–1 (Fig. 11d) are consistent with observations at the experimental site. Higher water contents, however, do not necessarily mean higher transport velocities. Owing to the dependence of K on {theta}, it is often assumed that higher {theta} corresponds to higher transport velocities. Although this may be the case in homogeneous systems or those described by a single set of effective properties, it is not true in complex environments.

Subsurface Flow Networks and Their Relation to Surface Flux
The nonlinear dependence of K on {theta} results in a much larger range of K than is observed in the distributions of {theta} (Fig. 11). The effects of variability in the {theta}({psi}) function, coupled with the differences in air entry pressure, can cause internal water fluxes to range over several orders of magnitude. These fluxes can even exceed the input flux and may lead to different flow networks being active depending on the input flux. This phenomenon is illustrated in Fig. 12 , which plots the normalized vertical flux, that is, the node centered flux, |Jw| as a ratio to the input flux, Jw0. It is clear that the underlying heterogeneity leads to the development of a complicated flow network that does not reflect the distribution of {theta}. Furthermore, the flow network may even mask the underlying structure of the heterogeneity. At low input fluxes, water content is highest in the fine-textured regions of the domain, and K({theta}) is consequently highest in those regions (Fig. 12a). Flow channels, therefore, form at or near the upper boundary and propagate downward following the paths of highest conductivity. In the early stages of infiltration, the origin and distribution of these paths are clearly dependent on the underlying structure of the heterogeneity with water being redirected or funneled from regions of low conductivity to regions of high conductivity. Once the channel has formed, it transmits enough water to overcome the water entry pressure of low-permeability regions resulting in wetting of small isolated regions. Larger regions of low permeability, however, would require larger amounts of water to overcome the water entry than might be available from these channels; consequently, flow is diverted around these regions at low-input fluxes. This is illustrated in Fig. 12a, 12b, and 12c in the coarse textured region near x = 3.6 m. Water is diverted around this region until the input flux reaches 101 cm s–1 while smaller regions within the dike, such as at x = 2.5 m, show increased velocities at much lower fluxes.


Figure 12
View larger version (94K):
[in this window]
[in a new window]

  		FIG. 12. Simulated distributions of dimensionless water flux, |Jw|/Jw0, during steady infiltration under a constant flux surface flux, Jw0, of (a) 1 mm yr–1, (b) 10 mm yr–1, (c) 102 mm yr–1, and (d) 103 mm yr–1. Domain properties were upscaled to 2-cm grid.

 
As the input flux increases, the differences in K({theta}) between the different textures become less as {psi} approaches the value at which K({theta}) is equal. Consequently, the differences in the flux ratio become less apparent. As shown in Fig. 12c, both the coarse and fine regions show similar flux ratios when the input flux is about 102 mm yr–1. As the input flux is increased, log(–{psi}) continues to decrease, and K({theta}) of the coarse-textured regions surpasses that in the fine-textured regions (Fig. 12c, 12d). A reversal in the distribution of the flux ratio becomes apparent, and regions that were being bypassed earlier because of lower K({theta}) become the dominant regions of transport (Fig. 12d). This phenomenon is conceptually identical to that of the complementary flow networks reported by Roth (1995) based on simulations in numerically generated heterogeneities; in this case the heterogeneities are based on field observations.


   Discussion and Conclusions
TOP
 ABSTRACT
 INTRODUCTION
 Properties of Clastic Dikes
 Materials and Methods
 Discussion and Conclusions
REFERENCES
 
It is clear that formation of the preferential channels shown in the simulation results presented in Fig. 12 is controlled completely by the structure of the small-scale heterogeneity seen in Fig. 7. However, propagation and persistence is somewhat more complex and appears to be dominated by heterogeneities at the larger scale, as these tend to redirect flow at low fluxes. Representation of such phenomena requires, in the least, a multidimensional mechanistic model that reflects the dominant components of the underlying heterogeneity structure. These phenomena cannot be simulated with one-dimensional models, particularly the screening models used to date for retrieval performance evaluations performed at the Hanford Site. These results are consistent with the two complementary flow networks hypothesized from the field experiments, as well as the results of theoretical analyses reported in the literature (Roth, 1995; Birkholzer and Tsang, 1997).

The modeling and experimental results provide insight into the existing distribution of contaminants in the subsurface and possibly the breakthrough behavior of contaminants at the water table. The larger the input flux, the larger the local potential gradients, and the larger the number of flow paths that are active in transporting water and dissolved solutes. The nature of the flow paths depend strongly on how they are formed near the surface; therefore, under certain conditions, multipeaked breakthrough curves may be observed. The shape of breakthrough curves is expected to change as the input flux varies. As the input flux increases, breakthrough curves could be expected to change from multipeaked at low fluxes to a single peak at the critical flux when the flow networks are similar and back to multipeaked as the network switches from high-permeability, fine-textured regions to high-permeability, coarse-textured regions. The travel time characteristics of the breakthrough would also be expected to change. An indication of this behavior was seen in the field data. Figure 13 , for example, compares the breakthrough of bromide observed at the 0.8-m monitoring depth in the host matrix and the dike. Despite the complexity of the internal structure of the clastic dike that was studied, the bimodal breakthrough in the dike illustrated in Fig. 13 can be attributed simply to the existence of the two flow networks within the dike itself.


Figure 13
View larger version (14K):
[in this window]
[in a new window]

  		FIG. 13. Example of solute breakthrough curves observed in the field at the Army Loop Road dike site. Breakthrough in the dike is multimodal, reflecting at least two flow domains. A dike continuous to a compliance plane could result in multipeaked breakthrough curves at that plane.

 
These results also suggest that contaminants deposited in the vadose zone under one type of boundary condition may subsequently become inaccessible unless similar conditions can be re-created to allow the originally active pore space to be assessed. For example, contaminants deposited in heterogeneous systems under rapid, large-volume leaks or recharge conditions would be bypassed under normal recharge (i.e., low flux) conditions, and vice versa. These findings may provide insight into existing contaminant distributions and episodic contaminant breakthrough at the water table.

While the results of this study do not provide sufficient data to test the hypothesis that clastic dikes provide vertical fast paths for transport at the Hanford Site, the experimental and modeling results for the clastic dike studied at the Army Loop Road site suggest that clastic dikes may indeed serve as fast paths for transport of water in the vadose zone. However, the simulation results discussed above suggest that enhanced vertical transport relative to the matrix would only occur for certain flux conditions. Even if clastic dikes do allow vertical transport of water at certain conditions of saturation, this may only apply to conservative contaminants like tritium and technetium-99. The fine-grained nature of clastic dikes, with up to 19% clay-sized material in the finer-grained portion of the Army Loop Road clastic dike, suggests that more reactive contaminants would be unlikely to travel far in clastic dikes.

Further work should be performed to examine a broader range of clastic dikes, including studies in which the host matrix is coarser than in the current study and studies that focus on reactive transport of contaminants in clastic dikes, which would be useful to further examine the potential for clastic dikes to serve as fast flow paths for contaminants at the Hanford Site.


   ACKNOWLEDGMENTS
 
Funding for this study was provided by the U.S. Department of Energy's Environmental Management Science Program, Project No. 70193, and by the Remediation and Closure Science Project at Pacific Northwest National Laboratory, funded through USDOE Richland Operations. The authors thank P.E. Long and R.L. Kirkham for collection and processing of the infrared imagery; Y. Xie for geostatistical analysis; G. Seedahmed for digital photogrammetry and textural segmentation of the digital images; W. Clement for ground penetrating radar surveys; B.J. Lechler, Z. Brown, M.L. Fayer, B.N. Bjornstad, and D.G. Horton for field support; and K. Fecht for information and insights based on his extensive studies of clastic dikes at the Hanford Site. Comments and suggestions received from three anonymous reviewers are greatly appreciated.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Properties of Clastic Dikes
 Materials and Methods
 Discussion and Conclusions
 REFERENCES
 




This article has been cited by other articles:


Home page
 Soil Sci.Home page
M. A. Mayes, G. Tang, P. M. Jardine, L. D. McKay, X. L. Yin, M. N. Pace, J. C. Parker, F. Zhang, T. L. Mehlhorn, and R. Dansby-Sparks
Influence of Sedimentary Bedding on Reactive Transport Parameters under Unsaturated Conditions
Soil Sci. Soc. Am. J., October 21, 2009; 73(6): 1938 - 1946.
[Abstract] [Full Text] [PDF]

Home page
 Vadose Zone JHome page
A. L. Ward and Z. F. Zhang
Effective Hydraulic Properties Determined from Transient Unsaturated Flow in Anisotropic Soils
Vadose Zone J., November 20, 2007; 6(4): 913 - 924.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Murray, C. J.
Right arrow Articles by Wilson, J. L.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Murray, C. J.
Right arrow Articles by Wilson, J. L.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Murray, C. J.
Right arrow Articles by Wilson, J. L.
Related Collections
Right arrow Variably Saturated Fluid Flow
Right arrow Vadose Zone Processes and Chemical Transport

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome