Assessment of aquifer heterogeneities at the Hanford Nuclear Reservation, Washington using inverse contaminant plume analysis

All rights reserved INFORMATION TO ALL USERS

The qu ality of this repro d u ctio n is d e p e n d e n t upon the q u ality of the copy subm itted. In the unlikely e v e n t that the a u th o r did not send a c o m p le te m anuscript and there are missing pages, these will be note d . Also, if m aterial had to be rem oved,

a n o te will in d ica te the deletion.

uest

ProQuest 10781060

Published by ProQuest LLC(2018). C op yrig ht of the Dissertation is held by the Author. All rights reserved.

This work is protected against unauthorized copying under Title 17, United States C o d e M icroform Edition © ProQuest LLC.

ProQuest LLC.

789 East Eisenhower Parkway P.O. Box 1346

An engineering report submitted to the Faculty and the Board of Trustees of the Colorado School of Mines in partial fulfillment of the requirements for the degree of Master of Engineering (Geological Engineering).

Golden, Colorado Date: k ) a i K j m Signed:

t tUvL

.. „

____

Samuel S. Adams, Head Department of Geology and Geological Engineering

A B STR A C T

Inverse plume analysis can be used to delineate the geology of heterogeneous aquifers. By analyzing the concentrations at pairs of points in a contaminant plume, the spatial distributions of apparent transverse and longitudinal dispersivity are determined. Since dispersivity is an intrinsic property of a porous medium, spatial distributions of the dispersivities obtained from inverse plume analyses are an indication of aquifer material transitions.

Numerical models of contaminant plumes in hypothetical aquifers aid interpretation of field data. These hypothetical aquifers consist of various configurations of material types that represent different sediments such as silt and sand. Results from the modeling studies indicate that it is possible to delineate material zones on the basis of transitions of apparent transverse and longitudinal dispersivity. The analysis of apparent transverse dispersivity is the most useful of the two procedures because it is less spatially restrictive than the procedure for determining apparent longitudinal dispersivity.

Dispersivity values obtained are ''apparent" because the assumptions that the aquifer is isotropic and homogeneous are violated. In the case of a heterogeneous aquifer, the dispersivity value obtained at a given location by inverse plume analysis is affected by the spreading o f the plume through different materials. Therefore, variations in dispersivity should be interpreted qualitatively for delineation of aquifer heterogeneity and not

quantitatively as for input to numerical transport models. Use of the procedure to identify aquifer heterogeneities is illustrated by application to a contaminant plume on the Hanford Site, Washington.

TABLE OF CONTENTS

ABSTRACT... iii

INTRODUCTION... 1

PURPOSE AND SC O PE... 1

LOCATION...1

HYDROGEOLOGY... 3

HYDROLOGY...9

CHARACTERIZATION OF THE UNCONFINED AQUIFER... 10

INVERSE PLUME ANALYSIS... 11

APPLICATION OF INVERSE PLUME ANALYSIS TO A HETEROGENEOUS AQUIFER...14

EVALUATION OF HYPOTHETICAL HETEROGENEOUS AQUIFERS...15

FIELD APPLICATION... 31

POTENTIAL USES OF INVERSE PLUME ANALYSIS TO ASSESS AQUIFER HETEROGENEITIES... 48

FUTURE W O R K ... 49

SU M M A R Y ... 49

LIST OF FIGURES

Figure 1. Location map of Hanford Reservation and geographic setting of

southcentral Washington (after Gaylord and Poeter, 1987)... . . ; . 2 Figure 2. Location of Hanford Reservation operational areas (after Lindberg and

Bond, 1979)... 4

Figure 3. Generalized cross section of the unconfined aquifer sediments extending west-east across the central portion of the Hanford

Reservation (after DOE, 1987)... 5 Figure 4. Generalized stratigraphy of the post-basalt sediments on and adjacent to

the Hanford Reservation, southcentral Washington (after Gaylord and

Poeter, 1987)... 6 Figure 5. Schematics illustrating application of inverse plume analysis to a

heterogeneous aquifer...16 Figure 6. Numerical model 1: Configuration o f materials, concentration

contours, and apparent transverse dispersivity values...19 Figure 7. Numerical model 2: Configuration of materials, concentration

contours, and apparent transverse dispersivity values... 20 Figure 8. Numerical model 3: Configuration o f materials, concentration

contours, and apparent transverse dispersivity values...21 Figure 9. Numerical model 4: Configuration of materials, concentration

contours, and apparent transverse dispersivity values... 22 Figure 10. Numerical model 5: Configuration of materials, concentration

contours, and apparent transverse dispersivity values...23 Figure 11. Numerical model 6: Configuration of materials, concentration

contours, and apparent transverse dispersivity values... 24 Figure 12. Numerical model 7: Configuration of materials, concentration

contours, and apparent transverse dispersivity values... 25 Figure 13. Selected models with data m issing... 29 Figure 14. Well network used to monitor the unconfined aquifer at the Hanford

Reservation (after DOE, 1987)... 32 Figure 15. Tritium plume in the unconfined aquifer at the Hanford Reservation as

measured in 1977 (after Freshley and Graham, 1988)... 34

Figure 16. Isocon map of 1963 tritium concentrations in the Hanford Reservation

produced in SURFER (contour interval = 200000 picocuries per liter)... 35 Figure 17. Isocon map of 1975 tritium concentrations in the Hanford Reservation

produced in SURFER (contour interval = 100000 picocuries per liter)... 36 Figure 18. Dispersivity zonation map for 1963 tritium plume... 39 Figure 19. Dispersivity zonation map for 1975 tritium plum e. ...40 Figure 20. Percent mud and silt map for the upper 60 feet of the Hanford

unconfined aquifer (1963)...41 Figure 21. Percent mud and silt map for the upper 60 feet of the Hanford

unconfined aquifer (1975)... 42 Figure 22. Overlay o f apparent dispersivity zones and percent mud and silt map

(1963)... 43 Figure 23. Overlay o f apparent dispersivity zones and percent mud and silt map

LIST O F TABLES

Table 1. Material properties used in numerical models... 18 Table 2. Results of steady-state numerical modeling...26 Table 3. Results o f transient numerical modeling... 27

ACKNOWLEDGMENTS

The research and writing of any thesis involves more than just the work of the author. There are many players within the background that aid and assist in the author's work.

I would like to thank my advisor, Dr. Eileen Poeter, for her guidance,

encouragement, and example for this work and in my life. She has touched my life in more ways than she can know. I hope that through my life I can maintain the balance and the professionalism that she has exhibited.

At Washington State University, I would like to thank the project researchers for their work and the play. Drs. Kevin Lindsey and David Gaylord helped tremendously in the geology and the interpretation parts of this work. I would also like to thank the Geology Department and Geological Engineering Section secretaries (Debbie, Dottie, Jo, and Cathy) at WSU for their kindness and letting me get away with murder. Ken Seymour and Tom Weber also deserve thanks for the good times in Albrook Lab and for helping me in their fields of expertise.

At the Colorado School of Mines, I would like to thank all my professors for the instruction you've given me. I would also like to thank my committee, Drs. Jerry Higgins and Ken Kolm, for their help and advice on my research and the writing of this thesis. I would also like to thank Marilyn and Debbie, the departmental secretaries for their help and conversation. The computing center staff also deserves a round of applause, even if I ended up not using anything they helped me with.

Graduate school is not all work. To those who have gone before me and still remain, I give a note of appreciation and encouragement We have worked hard into the night, taking classes, finishing projects and drinking beer. To Pete Zlatev and Greg Naugle, thanks for the friendship, mountain biking, and the Fourteeners (or the attempts). I will never forget the "technical mountain biking” on South Table Mountain. Thanks also to Newt, Sandy, Buz, Sarah,and Carma. The friendship, help in classes, and beers (and margiritas) made all the work somehow worth i t Hope to see you all in the professional world!

I would also like to thank Battelle Pacific Northwest Laboratories for the provision of the funding that enabled this work to take place. I would also like to thank, George V. Last, Pam Mitchell, Wally Walters, and others who helped, but whose names escape my feeble memory, for their technical expertise and willingness to help the project.

Thanks also go to the BBC, John Pertwee, Tom Baker, Peter Davison, Colin Baker, and Sylvester McCoy for many travels through time and relative dimensions in space.

Lastly, I would like to thank my wife and children. My family has been a source of comfort that I could not have done without. Ellen has worked on this as much as I have. She endured late nights when I was gone, my anxieties over tests, research and life (in general), and a really bad first semester at WSU. Thanks Ellen for putting up with all this and raising two wonderful people, Sara and Meagan. Sara and Meagan were always willing to give hugs and kisses, which made all this worth it. Thanks.

INTRODUCTION

The heterogeneities of the aquifer must be known in some detail in order to simulate transport of contaminants in a ground-water system. An aquifer cannot be characterized in sufficient detail with drilling because of the expense of this operation and the possibility of destroying the aquifer integrity by characterization. The lack of high resolution field data is one of the major problems that prevent the extensive and accurate use of ground-water numerical models (Gaylord and Poeter, 1987). Indirect methods (e.g., geophysical surveys) must be used to analyze and interpret the subsurface geology. Such techniques produce qualitative results, but when used in conjunction with other indirect techniques, aquifer heterogeneities can be characterized.

PURPOSE AND SCOPE

At the Hanford Nuclear Reservation in Washington, contaminants have been introduced to the unconfined ground-water system. The movement of these contaminants is, in a large part, controlled by the geology of the unconfined aquifer. The purpose of this study is to delineate aquifer heterogeneities by estimating the spatial distribution of

hydraulic and transport properties of the materials using inverse plume analysis techniques. The large amounts of geologic and hydrologic data collected at the Hanford

Reservation are advantageous for studying the use of ground-water contaminant transport properties to assess subsurface geology (Gaylord and Poeter, 1987). The character of contaminant transport properties of a small region within in the Hanford Reservation can be estimated by analyzing ground-water contaminant concentration data measured during the past four decades. These spatial distributions of transport properties can then be used to assess the distribution of sediment types in the aquifer. This sediment distribution can then be compared to inteipietations derived from existing geophysical data, drilling logs, and stratigraphic/sedimentological data to better delineate aquifer heterogeneities.

LOCATION

The Hanford Nuclear Reservation is within the Pasco Basin in southcentral

C lla n a b u t? coluwBi. 6**1# HounuiA C «bl« HANFORD r o p p t o lt li R ida* K ian tv tck hr* 1 1 * w « I U JU v « c . j til lu« 111 lomucoi*

Figure 1. Location map of Hanford Reservation and geographic setting of southcentral Washington (after Gaylord and Poeter, 1987).

03

uplifts are known as the Rattlesnake Hills and Horse Heaven Plateau on the south, the Umtanum and Yakima Ridges on the west, and the Saddle Mountains on the north (Fig. 1). To the east, the Pasco Basin is bordered by the Columbia Plateau. The Hanford Nuclear Reservation is located in the central and lowest part of the Pasco Basin to the south of the Columbia River (Fig. 1) and has an area of approximately 700 square miles. Figure 2 illustrates the operational areas of the Hanford Nuclear Reservation where nuclear processing and research activities are carried o u t

HYDROGEOLOGY

The Hanford unconfined aquifer consists of a sequence of sedimentary deposits which overlay late Miocene aged flood basalts (Myers and Price, 1979). This post-basalt sedimentary sequence comprising the unconfined aquifer consists of three formations, listed from oldest to youngest: 1) the Ringold Formation, 2) glaciofluvial sediments (Missoula flood deposits), and 3) aeolian deposits and/or surface alluvium. Figure 3 shows a generalized east-west cross section of the sediments comprising the unconfined aquifer and Figure 4 shows a generalized stratigraphic section of the sediments comprising the unconfined aquifer.

Ringold Formation

The Ringold Formation consists mainly of fluvial and lacustrine sediments deposited during the Pliocene and early Pleistocene Epoches. The Ringold Formation has been informally divided into four subdivisions: the basal Ringold Formation, the lower Ringold Formation, the middle Ringold Formation,and the upper Ringold Formation. These subdivisions do not represent a position in the stratigraphic sequence, but instead are sedimentologic distinctions; there is interfingering of the divisions within

100 0 ANO OR ICON 100 KE ANO KW 100 8 ANO C COLUMBIA RIVER VERNITA .^B R ID G E ^UMTANUM RIDGE GABLE MOUNTAIN GABLE BUTIE YAKIMA ^ « j.0G£

a

r

M et er s A bo ve M en Se a L ev el West 200 East Miles 200 West 200 East Unconfined Water Table West 183

Hanford Sediments _Columbia

River 152 East 122- ; Middle Ringold Low* Distance, Kilometers

Figure 3. Generalized cross section of the unconfined aquifer sediments extending west-east across the central portion of the Hanford Reservation (after DOE, 1987).

E o l i a n sed im ent s; 0 -3 5 f t ; s i l t an d fi ne -g ra in ed s a n d . 0) P 10 • P T3 Q) £ E (0 O 0) rH o» •o £ £ o <0 u 1 * P *0 0) £ *0 0 H 0) a o > i A P rH 0 •H P 01 0 H P 43 rH 43 •H <D 0) Q. P P mh m o m r -r -_* CN | 1 o O1 ... ro CM P P •H •H £ £ a • _a P H rH rH (0 •H (0 •H w •H > > a tj a rH c rH MH (0 MH O O •H •H U T) O cd c <0 H <0 rH o (/I o 4J p •o *H TJ rH £ H £ •H (0 01 0 0 01 0 -P >♦ T3 >1 • (0 (0 0 0 0 »H rH £ rH P 0 O •H U 0 E 0 P P O 0 U 0 0 0 rH _{a 0»} 1 a E *d O' rH rH O rH £ 43 c 43 rH a O 3 •d O' o • O T> •H £ 43 •O £ •d £ 0 £ in 0 0 O 0 (0 0 £ P 0) •d £ £ P rH ₅ 5 0 0 *d a o ■P 0 •d rH £ o in Hh P rH 43 (0 43 . 43 43 a 43 o 0 0) • O • % CN 43 f t ' d AJ P -P rH P T3 £ . f—1 MH 1 MH •* (0 •rH _{0 O} £ P 01 w rH O O 0 <u 43 O .« O 0 a a 43 rH rH rH IP -H 0 - i 0 1 rH p (0 Q*TJ O > O 43 O (0 rH £ 0 43 CO P O — 10 P ■*• O rH 0 p 01 >1 O' O • E MH 0 0 O 0 P TJ rH TJ • rH •-rH MH O 0 o £ 0 U P •* o> O £ 0 P MH p rH £ O CN -H 3h 0 P rH 0 O 1 0 0 TJ p 0 in 0 P O CN i n in cu £ 0 3 w 01 •H 1 cn O ' a 0 £ O i 0 £ T3 O i a 0) 0 rH o 03 a c - 0 rH <0 •v t j 0i * _{- O'} * •- P rH T3 rH M TJ TJ £ •d •d 0 10 0 rH O (0 rH rH O rH rH > •H P 0 O ' O 0 O U o o •H > (0 O' £ U CJ» O' O' O ' 04 a p £ -H £ * £ 0 c • £ rH 0 H 04 0 •H T3 •H rH •H TJ -H 0 MH £ 04 P 04 £ 04 43 04 £ 04 •H 0 0 0 0 43 _{i i} 0 43 •H rH iH rH 1 U 0) P O 0) rH £ o O' 0 TJ 0 a> a> U 0 0 a 0 c 0 . • a £ 5 'O ? 3 T J W r—i rH 0 f t •H H _{O £} o o O £ 0 0 O O a E mh rH (0 1—1 p rH 0 43 o *o.o’ o .q. » -•oOQ.-'fliV.oOo ,o.Q:i rsQ5?.o„«uo; oo?. .*0-. 4> X! |T O ■*“' X o *- o s « CJ “ e5 C ;->«* c5 *p T3 O 0 >* 2 § c <£ <L> 0 .S '?

■8 2

the Ringold Formation (Lindberg and Bond, 1979). The discontinuous nature of the Ringold Formation is illustrated in Figures 3 and 4.

Basal Ringold Formation

The basal Ringold Formation consists mainly of a coarse to fine sand matrix supported conglomerate. The clasts are pebble- to cobble-sized basalt and extra-basin lithologies (Myers and Price, 1979). The basal Ringold Formation is believed to have been deposited as slopewash by streams flowing through the Pasco Basin (Myers and Price, 1979). Lower Ringold Formation

The lower Ringold Formation, composed of finely bedded to massive sand, silt, and clay, is usually a blue or green color (indicative of a reducing environment), but can be tan colored if oxidized. These sediments are believed to be flood plain and lacustrine deposits (Myers and Price,

1979). The lower Ringold Formation has only been detected in the subsurface by drilling; no outcrops have been discovered (Last, 1987). Middle Ringold Formation

The middle Ringold Formation is a conglomerate composed of well rounded pebbles and cobbles. The matrix is composed of coarse to fine sand and silt (Myers and Price, 1979). Lenses of sands and silts are present within the conglomerate (Myers and Price, 1979) that appear to increase in occurrence towards the top o f the unit (Lindsey, 1987). Incomplete calcareous and siliceous cementation exist within the conglomerate (Myers and Price, 1979).

Upper Ringold Formation

The upper Ringold Formation is composed of well sorted medium to fine sands, silts, and clays with minor lenses of pebbles. Caliche horizons

are common in this unit (Myers and Price, 1979). The upper Ringold Formation has a light brown to tan color and is usually capped by a caliche horizon (Myers and Price, 1979).

Myers and Price (1979) indicate that the upper Ringold Formation was deposited by low energy fluvial systems and local lacustrine systems, but the lateral extensiveness of bedding within upper Ringold Formation outcrops in the White Bluffs along the Columbia River seem to indicate a more extensive lacustrine environment with distal and prograding delta systems (Lindsey, 1987).

Glaciofluvial Deposits

During the late Pleistocene Epoch, lobes of glacial ice moved into the Clark Fork River Valley in eastern Idaho, damming the river. Impounded water filled the Clark Fork River valley and tributary valleys to the east, in Montana, creating Lake Missoula.

Eventually, the ice dam was breached by the river, resulting in the release of over 500 cubic miles of water in approximately two days (USGS, 1982).

The waters flooded across existing drainages in Eastern Washington, creating the Channelled Scablands (USGS, 1984). As the ice lobes continued to move across the Clark Fork River and subsequent ice dams were breached, many of these catastrophic flood events occurred (Bretz, 1959). These flood events were known as the Missoula Floods (Myers and Price, 1979).

The sediments deposited by these floods are known informally as the Hanford formation (Myers and Price, 1979) or Hanford sediments. Within the Reservation area, the sediments deposited by these flood events occur in two facies, the flood channel deposits (the Pasco Gravels) and the slackwater deposits (the Touchet Beds).

The Pasco Gravels are composed of uncemented pebble to boulder conglomerate and some coarse sands, silts, and silty clays (Gaylord and Poeter, 1987). The clasts are made up of basalt, granite, diorite, andporphyritic volcanics (Myers and Price, 1979). Coarser parts of the Pasco Gravels were deposited in flood current channels, while the

finer parts (sand and silts) were deposited in lower energy environments or during waning periods of flooding (Myers and Price, 1979). The Hanford Reservation contains the thickest and most extensive deposits of the Pasco Gravels in the Columbia Plateau, between 100 and 200 feet thick (Myers and Price, 1979).

The Touchet Beds are fine grained, rhythmically bedded sediments deposited by the slackwaters of the catastrophic floods. These slack waters were produced when flood waters were impounded by hydraulic dams at narrow passages in the flood drainages (USGS, 1982). The Touchet Beds consist of silt to fine sand with stringers of coarse sand and gravels (Myers and Price, 1979). They unconformably overlie and interfinger with the Pasco Gravels (Myers and Price, 1979).

Eolian Deposits and Surface Alluvium

Pleistocene- and Holocene-aged eolian deposits and surface alluvium blanket much of the Hanford Reservation land surface. These deposits consist of silt and silty fine grained sand deposits and coarse grained lag deposits (Gaylord and Poeter, 1987).

h y d r q u q i l y

The unconfined aquifer at the Hanford Reservation consists predominantly of the Ringold Formation and includes some of the Pasco Gravels of the Hanford formation. Near the Columbia River, Holocene-aged alluvial deposits are thick enough to allow flow of ground water. Most of the flux within the aquifer is through the middle Ringold Formation conglomerate and the Pasco Gravels.

Hydraulic conductivities of the middle Ringold Formation range from 17 feet per day to 420 feet per day due to the variability in cementation (Gephart and others, 1979). The uncemented Pasco Gravels have large hydraulic conductivities with values up to

144,000 feet per day (Gephart and others, 1979).

Prior to the beginning of operations at the Hanford Reservation in the 1940's, the water table was mainly within the Ringold Formation, below the base of the Pasco Gravels (Newcomb and others, 1972). With the introduction of waste water from processing

operations the water table has risen from the natural (pre-1944) levels (Gephart and others, 1979). Currently, the water table is mainly within the Pasco Gravels. In the southwestern part of the Reservation and in a mile-wide strip next to the Columbia River on the eastern side of the Reservation the water table remains within the middle Ringold Formation (Bergeron and others, 1986).

The unconfined aquifer receives natural recharge from precipitation, surface runoff, and the Columbia and Yakima Rivers during flood stages. Gephart, and others (1979) indicate that the unconfined aquifer receives little if any recharge from direct precipitation; however, even a small amount of recharge may be important. If 0.04 inches of infiltration reached the unconfined aquifer, approximately 7 million gallons per year of recharge would be contributed. Lysimeter studies in 1983 and 1984 indicate that there is a potential for recharge to the unconfined aquifer from direct precipitation (Gee and Kirkham, 1984).

Artificial recharge to the unconfined aquifer began in the 1940's in the form of waste water discharges from Hanford processing plants. In some areas on the

Reservation, the water table has risen as much as 80 feet above the pre-1944 levels

(USGS, 1987). Ground-water mounds exist beneath the principal disposal locations in the 200 Area (U, B, and Gable Mountain Ponds) (Zimmerman, and others, 1986; USGS,

1987) and the 300 Area (North and South Process Ponds and the Process Trenches) (Lindberg and Bond, 1979) (Fig. 2).

Discharge from the unconfined aquifer is primarily to the Columbia River, with a minor amount o f discharge to the Yakima River. This has caused some concern since much of the artificial recharge (from Hanford’s processing plants) contains toxic chemicals and radionuclides.

C H A R A C T E R IZ A T IO N O F T H E UN CONFINED A Q U IFE R

As indicated by the previous section, the geology of the unconfined aquifer is heterogeneous. In order to study contaminant transport on the Hanford Reservation, these heterogeneities should be characterized. Most geohydrologic field techniques measure bulk properties of the porous medium. Since it is uneconomical to characterize the aquifer with drilling, indirect methods must be used. Because indirect methods yield qualitative results,

a number of such methods must be integrated with each other and borehole data to define heterogeneities.

One such technique is developed herein, inverse plume analysis as applied to heterogeneous aquifers. The following section discusses the application of inverse plume analysis to heterogeneous aquifers in a generic sense to illustrate its validity in definition of aquifer heterogeneities. In subsequent sections, the method is applied to the Hanford Reservation.

Domenico and Robbins (1985) and Domenico (1987) developed methods to determine aquifer and contaminant source characteristics from the spatial distribution of contaminant concentration in a contaminant plume in homogeneous aquifers. These

methods are termed inverse plume analysis. Such analysis is analogous to aquifer pumping tests, but instead of using drawdown relationships to determine transmissivity and

storativity, contaminant concentration data is used to determine the three orthogonal dispersivities, the center of mass of the contaminant plume, and the contaminant source strength and dimensions. Inverse plume analysis has been performed on chloride plumes at the Hanford Reservation (Lavenue and Domenico, 1986) and on the Idaho National Engineering Laboratory (Fryer and Domenico, in press). The method has also been verified using a three-dimensional finite element code (Domenico, 1987).

Inverse plume analysis assumes the aquifer is isotropic and homogeneous (Domenico and Robbins, 1985) and utilizes an extended pulse model of the form: C(x,y,z,t) = (Co/8){erfc [(x - vt)/2(Pxt)0S]}

{erf [(y + Y/2)/2(Dyx/v)0-5] - erf [(y - Y/2)/2(Dyx/v)0'5] }

{erf [(z + Z/2)/2(Dzx/v)0'5] - erf [(z - Z/2)/2(D2x/v)0'5] } (1) where C(x,y,z,t) = contaminant concentration

Y, Z = contaminant source dimensions of a parallelpiped source, where Y is the transverse dimension of the source and Z is the vertical dimension of the source

Dx, Dy, D z = principal values of the dispersion tensor (dispersion coefficient)

x, y, z = spatial coordinate of concentration

t = time since introduction of contaminant to the system v = average linear flow velocity

erf = the error function

erfc = the complimentary error function.

A similar equation which accounts for the decay of chemical species is given by Domenico (1987). The dispersion coefficient is a function of dispersivity, molecular diffusion, and the flow field velocity. Dispersivity is dependent upon characteristics of the porous medium (Ogata, 1970; Freeze and Cherry, 1979; Wang and Anderson, 1982; Anderson, 1984). The dispersion coefficient is defined as (Freeze and Cherry, 1979):

D = a v + D* (2)

where a = dispersivity

v - average flow velocity

D* = coefficient o f molecular diffusion.

The second term (diffusion) in equation (2) is usually insignificant relative to the first term (dispersivity-velocity). The definition and treatment of dispersion (and

dispersivity) and the validity o f the advection-dispersion equation are under much scrutiny at this time due to the inability of current approaches to accurately predict contaminant plume development, particularly over more than one scale. Alternative approaches to

mathematically describing contaminant plume migration are being considered (Schwartz, 1977; Gelhar and others, 1979; Matheron and De Marsily, 1980; Gelhar and Axness, 1983; Guven and others, 1984; Neuman and others, 1987). Inverse plume analysis utilizes the classic definitions of dispersivity and the advection-dispersion equation, but will still be applicable if a better representation of dispersion is developed. In essence, searching for aquifer heterogeneities with inverse plume analysis utilizes the advection-dispersion equation as a relative measure of the degree of spreading of a contaminant.

The basic procedure for applying inverse plume analysis to homogeneous materials in a one-dimensional, constant velocity field as described by Domenico and Robbins (1985) is summarized as follows:

1) Select two concentrations at the same x-coordinate and different y- coordinates. The ratio of these concentrations is a function of the spatial coordinates, transverse dispersivity, and the Y source dimension (i.e., the ratio of equation (1) at the two points).

2) Solve the concentration ratio equation for the two unknowns (transverse dispersivity and Y source dimension). This results in an infinite number of solutions. Results are plotted as transverse dispersivity (cq) versus Y source dimension.

3) Use another set of concentration ratios at a different x-coordinate and plot at versus the Y source dimension on the same graph.

4) The intersection of the curves produced by steps 2) and 3) is then recorded as the unique value for transverse dispersivity and transverse source dimension.

5) Use concentrations selected from constant x- and y-coordinates with different z-coordinates and the procedure described above can be used to determine vertical dispersivity and the vertical source dimension (if three- dimensional data is available).

6) Determine source concentration by selecting a concentration within the steady state part of the plume along the centerline and solving (Domenico and Robbins, 1985):

C(x,0,0) = Co erf[Y /4(ayx)0'5] erf[Z/2(azx)0-5] (3) where C(x,0,0) is the steady state concentration along the plume's

centerline. Having conducted steps 1 through 5, the source concentration is the only unknown in the equation.

7) Calculate the velocity and longitudinal dispersivity, using data from a transient part of the contaminant plume and procedures similar to those described above.

Although many of the parameter values determined in the inverse plume analysis process could be used to identify zones of different character in an aquifer, the transport properties utilized to assess the subsurface geology are the longitudinal and transverse dispersivity. Since the development of the inverse plume analysis technique assumes a homogeneous and isotropic aquifer, spatial variance in the calculated dispersivities indicate variation of the subsurface geology.

H E T E R O G E N E O U S A Q U IFE R

If a contaminant plume develops in a one-dimensional, constant velocity flow field within a homogeneous aquifer, then the successive application o f steps (1) through (3) of the previous section will yield the same dispersivity and source dimension for all

combinations o f concentrations from various locations within the contaminant plume. If concentration data from a wide distribution of points in a contaminant plume within a heterogeneous aquifer were analyzed, a unique intersection point for the entire aquifer would not occur. To analyze concentration data in heterogeneous aquifers, steps (1) through (3) from the previous section should be performed on data from four nearby points. The computed dispersivity should be noted as representative of the bulk apparent

dispersivity for the zone delineated by those four points . Steps (1) through (3) are then repeated for another set of four proximal points and the dispersivity noted.

After a number of dispersivity values have been computed and mapped over the contaminant plume area, zones of similar dispersivity will become apparent and

hydrogeologic transitions can be hypothesized between these zones. Figure 5 illustrates the application of inverse plume analysis to a heterogeneous aquifer.

Once dispersivity zones are delineated, they should be evaluated in the light of existing geologic and geophysical data. Such evaluation may lend credibility to the dispersivity zonations and may aid in further delineation or characterization of the

subsurface. Knowledge of the dispersivity zones can be used to guide future drilling and further field studies. Or, if further field studies are not conducted but the area is

numerically modeled, then the zones can be considered during model calibration and evaluations.

EV A LU A TIO N O F H Y PO T H E T IC AT, H E T ER O G EN EO U S AQ UIFERS Before examining a field situation, hypothetical situations were investigated.

Numerical models of hypothetical heterogeneous aquifers were constructed and a source of contamination was introduced to the hypothetical ground-water flow field. Inverse plume analysis was used to examine the resulting contaminant plumes and the spatial variance of the dispersivities were compared to the original zonation of the different geologic materials.

Hypothetical Aquifer Configurations

The Golder Groundwater Computer Package (Poeter, 1983) was used in the aquifer simulations. Seven two-dimensional numerical models were constructed to simulate

simplified versions of geometries observed in heterogeneous aquifers by assigning

representative material properties to zones in the finite element grid. The models consisted of a regular grid 1900 feet square with 100 feet between nodes for models 1,2, and 3. Fifty feet spacing between nodes was used for models 4 through 7 to obtain numerically stable results in the more complex models. A conservative contaminant was introduced at a

POSSIBLE MATERIAL TRANSITIONS

CALCULATED APPARENT TRANSVERSE DISPERSIVITY

POINT WHERE CONTAMINANT CONCENTRATION IS ESTIMATED FROM ANALYSIS OF A WATER SAMPLE OR FROM INTERPOLATION OF AN ISOCON MAP OF THE CONTAMINANT PLUME AND USED TO CALCULATE APPARENT TRANSVERSE DISPERSIVITY.

CURVES FROM ALL POINTS

60-CURVES FROM ADJACENT POINTS

q .30-1 ' 1 I- ’ T ' " T s o u r c e d i m e n s i o n « 5 0 4 0 © 3 0 .52 2 0 "O 10 i— i— i— i- - - -- r 1 s o u r c e d i m e n s i o n Figure 5: Schematics illustrating application of inverse plume analysis to a

continuous source concentration of 1000 parts per million with background concentrations at 0 parts per million.

Table 1 presents the material properties used in the numerical models. Dispersivity values for the sand were lower (specified in parentheses in Table 1) when a finer finite element mesh was used to avoid numerical instabilities in models 3 through 7. Figures 6 through 12 illustrate the various distribution of aquifer materials in the models.

A gradient of 5.3 x 10‘3 was applied to each model by fixing head on both sides of the model and a steady state flow field was established. R ow in the illustrations is from right to left, with a line source of contaminant introduced on the right boundary of the illustration. Both steady state and transient contaminant plumes were examined. Transient contaminant plumes were examined to evaluate the usefulness of determining longitudinal dispersivity to delineate aquifer heterogeneities. Once the results from the numerical simulation were obtained, various points in the contaminant plume were selected for analysis in the previous chapter. These points were selected within the material zones in order to determine the ability of inverse plume analysis to assess aquifer heterogeneities. However, it is not possible to know the location of all sedimentologic transitions in a field situation, so a systematic evaluation of grid points is recommended.

Results of Analyzing Hypothetical Contaminant Plumes

The results of the inverse plume analyses performed on the numerical model output indicate that it is possible to use inverse plume analysis to assess aquifer heterogeneities. As data points are analyzed in each of the material property zones, dispersivities obtained reflect different sediment types. Tables 2 and 3 present the dispersivities calculated by applying inverse plume analysis to the contaminant plumes produced by the numerical modeling. Table 2 presents the results of the steady state numerical runs analyzed for transverse dispersivity and Table 3 presents the longitudinal dispersivities calculated by analysis of transient contaminant plumes.

TABLE 1

Material Properties Used In Numerical Models

SAND SILTY SAND SILT

Hydraulic Conductivity (ft/day) 13.4 0.134 0.00134 Effective Porosity 0.30 0.25 0.15 Longitudinal Dispersivity (ft) 100 (50)* 10.0 1.0 Transverse Dispersivity 10 (5)* 1.0 0.1 (ft)

200 10.6 9.0 600 1000 10.5 9.0 600 200 KEY SILT SILTY SAND | [SAND g s ? °

V A DATA POINTS AND CALCULATED DISPERSIVITY VALUES ~20©-s CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 ft.

Figure 6. Numerical model 1: Configuration of materials, concentration contours, and apparent transverse dispersivity values.

10.6

9.0

a* = 10.0

V J DATA POINTS AND CALCULATED DISPERSIVITY VALUES CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 f t.

Figure 7. Numerical model 2: Configuration of materials, concentration

contours, and apparent transverse dispersivity values.

9.3

7.0

10.0

6.0

a* = 10.0

a* = 10.0

V j DATA POINTS AND CALCULATED DISPERSIVITY VALUES "2 0 0 ^ CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 ft.

Figure 8. Numerical model 3: Configuration of materials, concentration

contours, and apparent transverse dispersivity values.

7777Z _22222222 KEY D l SILT SILTY SAND □ sand gv?-°

V \ DATA POINTS AND CALCULATED DISPERSIVITY VALUES "20©N CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 f t .

Figure 9. Numerical model 4: Configuration of materials, concentration

contours, and apparent transverse dispersivity values.

r //7 ///jr // % y/A Z ? /////////* '///a s/ /j

r

v

a t= 5.0

V j DATA POINTS AND CALCULATED DISPERSIVITY VALUES “ 20frs CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 f t .

Figure 10. Numerical model 5: Configuration of materials, concentration

contours, and apparent transverse dispersivity values.

a* = 5.0

'//s*y/s'A vaMwyMza

CL

= 5.0

KEY

h h s.lt

E%j SILTY SAND ["""I SAND e * °

V J DATA POINTS AND CALCULATED DISPERSIVITY VALUES * 2 0 ^ CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 f t.

Figure 11.: Numerical model 6: Configuration of materials, concentration

contours, and apparent transverse dispersivity values.

KEY D IJsiL T

SILTY SAND | [SAND

V J DATA POINTS AND CALCULATED DISPERSIVITY VALUES

-2 0 9 s CONTOUR INTERVAL = 200 PPM

| CONTAMINANT SOURCE = 1000 PPM

1 0 0 f t.

Figure 12. Numerical model 7: Configuration of materials, concentration

contours, and apparent transverse dispersivity values.

TABLE2

Results O f Steady-State Numerical Modeling

MODEL# F IG .# MODEL REGION TRANS VERSE DISPERSIVITY (FT) input calculated

1 6 SAND 10.0 9.0 - 10.6

2 7 SAND 10.0 9 .0 - 11.0

SILT 0.1 0.1

SAND & SILT NOT APPLICABLE 1.2

3 8 LEFT SILTY SAND 1.0 2.5

RIGHT SILTY SAND 1.0 0.5

SAND LENS 10.0 3.0

4 9 TOP SILTY SAND 1.0 1.5- 3.5

BOTTOM SILTY SAND 1.0 1.6- 1.9

SAND 5.0 5 .0 - 6.0

5 10 TOP SAND 5.0 2 .0 - 3.7

BOTTOM SAND 5.0 1.8- 2.0

SILTY SAND CHANNEL 1.0 1.0- 1.2

6 11 LEFT SAND 5.0 6 .0 - 7.0 RIGHT SAND 5.0 9 .3 -1 0 .0 SILT 0.1 0 .6 - 0.7 7 12 SILT 0.1 2 .0 - 2.2 SILTY SAND 1.0 3.0 SAND 5.0 3 .8 - 5.0

TABLE3

Results of transient numerical modeling

MODEL # MODEL REGION LONGITUDINAL DISPERSIVITY (FT) input calculated

1 SAND 50.0 47.0

2 SAND 50.0 45.0 - 50

SILT 1.0 22.0 - 23.0

3 LEFT SILTY SAND 10.0 17.0-18.0

RIGHT SILTY SAND 10.0 0.5

SAND LENS 50.0 65.0 - 68.0

4 TRANSIENT RUNS NOT CONDUCTED

5 TRANSIENT RUNS NOT CONDUCTED

6 7 LEFT SAND 50.0 26.0 - 30.0 RIGHT SAND 50.0 28.0 SILT 1.0 16.0 7 8 SILT 1.0 18.0 SILTY SAND 10.0 10.0 SAND 50.0 29.0

D is c u s s io n

Dispersivities calculated by the inverse plume analysis differentiate the hypothetical aquifer materials, but are not equal to the initial values of dispersivities input. This is believed to be caused by two mechanisms. The first, and probably most significant mechanism is the violation of the assumption that the aquifer is homogeneous. In a

heterogeneous aquifer the divergence and convergence of flow lines caused by variations of hydraulic conductivity produces variation in the degree of spreading of contaminants. This impact of the flow field combines with the variable intrinsic dispersivity of the materials to yield the apparent dispersivity computed by the inverse plume analysis. For example, in Figure 9, flow lines converging on the right side of the sand lens produce an apparent dispersivity lower than the value input for the dispersivity in the silt. On the left side of the lens flow lines diverge, yielding an apparent dispersivity higher than the dispersivity input for the silt. Limited dispersion of the contaminant in the silt to the right of the sand lens and perhaps lateral confinement of the sand lens results in a lower apparent dispersivity in the sand than the dispersivity input for the sand. The lack of lateral confinement is is exhibited in Figure 11. The second cause of differences between the input dispersivities and the results obtained from the analysis is due to numerical truncations from the finite element calculations. A finer mesh would reduce errors resulting from the finite element solution.

Errors of the finite element solution will not be present in a field plume. Therefore, the various dispersivities calculated for a field plume will result from aquifer heterogeneities and errors and uncertainties associated with the field data. Given either reasonably good field data or a valid definition of the uncertainties associated with the data, one can determine zones where the nature of the aquifer materials change.

As illustrated by models 3 and 4 (Fig. 13a and 13b, respectively), the application of inverse plume analysis can lead to inference of material zones which are not penetrated by boreholes. For example, consider model 4 (Fig. 13b) and envision that there are no boreholes in the silt (but enough boreholes to the right and left of the silt to delineate the contaminant plume). Geologic logs of boreholes in the sand would not indicate the presence of silt. However, the differences in apparent dispersivities in the right and left

X SSSSSSSSSSSSSSSSSSSSSSJ

N o t e h o w a p p a r e n t t r a n s v e r s e d i s p e r s i v i t y v a l u e s r e f l e c t t h e p r e s e n c e o f a n u n k n o w n z o n e i n b o t h m o d e l s .

Figure 13. Selected models with data missing.

sand zones would suggest that a zone of lower dispersivity exists between the boreholes right and left of the silt. These observations suggest that future boreholes be placed in the zone where differing geologic materials are suspected to occur.

When soft information (such as an inferred silt zone from an inverse plume

analysis, the results of geophysical surveys, the nature of sedimentary environments, and the geologic history of an area) is coupled with other soft information, as well as hard data (water levels and hydraulic test data), a better definition of aquifer heterogeneities will emerge. Such soft information, when provided to numerical models, puts constraints on the number of material zones and the range of the relevant parameters in each material zone. Thus, the constraints enhance the uniqueness of the solution.

A sensitivity analysis was performed in the sand region of model 2 in order to determine if the location of data points used in the analysis (relative to the geometry of the contaminant plume) would affect the results of the analyses. Points were selected that were directly across the plume axis, along the plume axis, on the same side o f the plume axis, and diagonal across the plume axis. As can be seen in Figure 7, there is no appreciable difference for the various point configurations selected from the contaminant plume. This is advantageous during field application of this method because concentrations from non- uniformly spaced boreholes may be used in the analysis. When using data from boreholes, the overall contaminant plume geometry must be known in order to determine the borehole coordinates relative to the centerline of the contaminant plume and the contaminant source location. Because of this, one would anticipate the need for a considerable amount of field data is needed to establish the origin and the centerline of the contaminant plume.

However, Lavenue and Domenico (1986) indicate that the analysis of large scale

contaminant plumes may be relatively insensitive to errors associated with the x-dimension of the contaminant plume geometry. Therefore, borehole networks that aid in the location of the contaminant plume's centerline would be the most cost effective.

Transient contaminant plumes were simulated in some o f the models and apparent longitudinal dispersivities were calculated (Table 3). Comments similar to those

concerning the calculation of apparent transverse dispersivity can be made on the nature of the resulting distribution of apparent longitudinal dispersivities. The analyses required

more effort, and are limited to the centerline of the contaminant plume. Yet, little more insight is provided to aquifer heterogeneities than the calculation of apparent transverse dispersivity. Calculation of longitudinal dispersivity is the last of seven aquifer parameters calculated with the inverse plume analysis and requires that the previous six parameters be known. Any error in the calculation of these parameters compounds the error involved in the calculation of longitudinal dispersivity. Hence, we conclude that the determination of transverse dispersivity is more useful in identifying aquifer heterogeneities than the determination of longitudinal dispersivity.

In general, the inverse plume method is useful for the qualitative evaluation of aquifer heterogeneities. Inverse plume analysis yields an apparent dispersivity that reflects not only the dispersivity of the zone being analyzed, but also the flow patterns and the dispersivity of other zones which the contaminant plume has migrated through. Calculation of apparent transverse dispersivity is more useful than apparent longitudinal dispersivity for delineating zones of aquifer heterogeneities.

F IE L D A PPL IC A T IO N

Once the theoretical aspects of using inverse plume analyses to assess aquifer heterogeneity were established, the method was applied to a study area within the Hanford Reservation. The study area is located in the Hanford Reservation’s 600 Area, near the 200 West and 200 East Areas (the Separations Areas), where plutonium is separated from uranium (Fig. 2).

Within the 600 Area, there are contaminant plumes of various chemicals and radionuclides (Lindberg and Bond, 1979; Eddy and Wilbur, 1980). Contaminant

concentrations have been measured in wells within the 600 Area since 1948 (Lindberg and Bond, 1979). The 600 Area is an excellent choice for analysis due to the large base of contaminant concentration data available (USGS, 1987) and the relatively stable flow field. Figure 14 shows the well network used to monitor the unconfined aquifer at the Hanford Reservation.

Tritium contaminant plumes were used to assess the subsurface geology because tritium moves readily with the ground water and is not adsorbed by geologic materials

.41

wi«.« •

S P R IN G WELL LOCATION

WELL COORDINATE DESIGNATION (AREA COOE. i.«. 1 9 9 . 2 9 9 . 3 9 9 .

4 9 9 . 6 9 9 ARE OELETEO FOR SPACE REASONS)

ESTIMATED BASALT OUTCROP ABOVE WATER TABLE

YAKIMA RIVER

KILOMETERS

M A R C H 1 9 9 B

Figure 14: Well network used to monitor the unconfined aquifer at the Hanford Reservation (after DOE, 1987).

(Eddy and Wilbur, 1980). Figure 15 shows the tritium plume at Hanford (Freshley and Graham, 1988). The configurations of the tritium plume in 1963 and 1975 were selected to analyze for the heterogeneities in the unconfined aquifer because in 1963 the discharges are fairly constant and frequent and because the 1975 configuration provides an area of overlap with the 1963 plume to check the configuration of the identified apparent dispersivity zones.

Isocon Maps

The 1963 and 1975 tritium plumes were delineated by constructing isocon maps of the region in the 600 Area near the 200 East Area (Fig. 2) to determine the contaminant source locations and the direction of plume movement The contouring of this map was

approached using two different semi-variogam models. Initially, the concentration data was kriged and contoured using SURFER, from Golden Software, Golden, CO.

SURFER fits a linear semi-variogram to the data is linear. The nature of the contours were also examined using the U.S. Geological Survey’s Statistics Package (STATPAC). In this case, a spherical semi-variogram model was fitted to the data. The spherical model was chosen because it represents the "ideal” shape for a semi-variogram (Clark, 1979). Kriging of the data and contouring resulted in an isocon map very similar to the one obtained using SURFER. Figure 16 and 17 show the isocon maps obtained from SURFER. Because of the similarity of the isocon maps obtained from the different programs, maps from

SURFER were used because o f the ease of manipulation (enlargement of certain areas, changing of contour intervals, etc.). The assumption of a linear semi-variogram is not considered invalid or unreasonable since the initial curves of the other semi-variogram models can be approximated by a linear function (Davis, 1986).

• MONITORING WELL

Figure 15: Tritium plume in the unconfined aquifer at the Hairford Reservation as measured in 1977 (after Freshley and Graham, 1988).

COLUMBIA RIVER

119

3 0

46

20

YAKIMA RIVER

20000

SCALE

Figure 16: Isocon map of 1963 tritium concentrations in the Hanford Reservation produced in SURFER (contour interval = 200000 picocuries/liter).

2 0 0 E A S T AREA

C O L U M B I A RI VER

119*30'

YAKI MA RI VER

4 6* 2 0 '

20000

SCALE

Figure 17: Isocon map of 1975 tritium concentrations in the Hanford Reservation produced in SURFER (contour interval = 100000 picocuries per liter).

In order to apply inverse plume analysis to the Hanford tritium plume, the following assumptions were made:

1) The plume is assumed to be two-dimensional in the horizontal plane. Due to the lack of three-dimensional data (analytical samples collected at several depths in the same well), it was difficult to obtain a spatial configuration of the plume. Most of the water quality samples were collected from the upper 40 to 60 feet of the unconfined aquifer (Eddy and others, 1978). Vertical dispersivity was calculated using the available data from nested piezometers (Eddy and others, 1978) and a value of approximately 0.2 feet was obtained. This is small with respect to the reported longitudinal and transverse dispersivity for the unconfined aquifer (100 and 60 feet, respectively) (Arnett and others,1977). Hence, it is expected that the tritium plume is essentially limited to the upper part of the unconfined aquifer.

2) The source is considered to be a single, continuous source of constant strength. In reality, the tritium is introduced through multiple, variable strength, noncontinuous sources (Zimmerman and others, 1986). However, the sources are close together relative to the distance traveled by the plume, so the problem of multiple sources is not considered severe. The maximum concentration has drifted to the east of the source areas and the concentration distribution has the appearance of a contaminant plume produced by one instantaneous source or by the injection of a contaminant (see Figs. 16 and 17).

3) A uniform flow field is assumed. The ground-water flow has varied over the life of the plume in response to variations in the location and rate of waste water discharge and mounding from waste-water discharge, causing a somewhat radial flow field (Eddy and others, 1978). However, because the inverse plume analysis is utilized to identify aquifer material zones and not the actual values of the transport parameters, the violation of this assumption o f uniform flow field is not of substantial significance.

Although the history and condition of this contaminant plume violates many o f the assumptions required for standard inverse plume analysis, the qualitative results obtained

from analysis of the tritium plume are useful in describing aquifer heterogeneities. To analyze the contaminant plume, the area of maximum concentration in the central part of the contaminant plumes was taken as the origin.

Comparison to Geologic Models

Maps showing the zones of apparent dispersivity delineated from inverse plume analysis for the 1963 (Fig. 18) and 1975 (Fig. 19) tritium plumes were compared to geologic cross sections and a percent mud and silt maps (percent fines map)(Figs. 20 and 21) of the upper 60 feet of the saturated zone existing in 1963 and 1975. Figures 22 and 23 illustrate overlays of the 1963 and 1975 apparent dispersivity zones on the 1963 and 1975 percent fines maps, for the readers' convenience.

In general, trends from low apparent dispersivities to higher apparent dispersivities on the apparent dispersivity zone maps (Fig. 18 and 19) corresponded to trends from higher percent fine grained material to higher percent coarse grained material on the percent mud and silt maps (Figs. 20 and 21). This correspondence of lower dispersivities with finer grained sediments (and vice versa) is consistent with the fact that finer grained materials typically exhibit lower dispersivities (Perkins and Johnston, 1963; Klotz and others, 1980).

A possible explanation for the correspondence of low dispersivities with fine­ grained sediments and high dispersivities with coarse-grained sediments involves the tortuous nature of flow paths through porous media. The coarser a sediment is, the more tortuous the ground-water flow paths become (Blatt and others, 1980). As the ground­ water flow paths become more tortuous, a contaminant can be dispersed more, thus increasing the dispersivity of the material. Similar statements can be made for the

relationship between low dispersivities and fine-grained sediments. This correspondence of dispersivities with sediment grain sizes also reflects the fact that the coarser-grained materials of the Hanford sediments are more heterogeneous and exhibit greater lateral continuity than the finer-grained materials at the macro-scale (10’s of feet). Percent maps are more useful for comparison of geologic data and apparent dispersivity zones than the geologic cross sections because sedimentologic trends can be seen with greater ease.

200 EAST AREA

ZONE

OOLLM3IA RIVER

80-90

46 20

20000/

SCALE

200 EAST AREA

20-40

80-90

119*30

46 20

SCALE

2 0 0 E A S T A R E A

lv

C O L U M B I A R I V E R

119*30*

YAKI MA R I V E R

46*20*

Cl = 10 %

20900

SCALE

Figure 20: Percent mud and silt map for the upper 60 feet of the Hanford unconfined aquifer (1963).

2 0 0 E A S T A R E A

C O L U M B I A R I V E R

119° 3 0 '

YAKI MA R I V E R

46° 20'

Cl = 10 %

SCALE

2 0 9 0 0

Figure 21: Percent mud and silt map for the upper 60 feet o f the Hanford unconfined aquifer (1975).

O' 2 0 0 0 0 ' S C A L E

Figure 22: Overlay of apparent dispersivity zones and percent mud and silt map (1963).

\ / A _____ t ✓ 2 0 0 0 0 O' ■ ____________t S C A L E

Figure 23: Overlay of apparent dispersivity zones and percent mud and silt map (1975).

Analysis of 1963 Plume

For the 1963 plume, several apparent dispersivity zones correspond to features seen in the percent map. In the northern part of the apparent dispersivity zone map (Fig. 18), zones A and B (low apparent

dispersivities) correspond to an area of high percentages of mud and silt. Zones C and D in the northeastern part of the apparent dispersivity zone map (Fig. 18) correspond to a region displaying a southeasterly trend from finer grained material to coarser grained materials (Fig. 20). In the southern part of the map (Fig. 18), the E zone corresponds to low percentages of muds and silts.

There are two areas on the apparent dispersivity zone map (Fig. 18) that do not correspond with identifiable features of the percent fines map (Fig. 201). In the southern part of the map (Fig. 18) there are two zones of low apparent dispersivity (A and B) that do not correlate with a high

percentage of fine-grained sediments. Instead the percent fines map

indicates low percentages of fine-grained sediments in an area which should yield zones of high apparent dispersivities. However, data are relatively sparse, considering the scale of the southern part of the plume and to the south, and if boring data were available, a zone of fine-grained materials might be observed. The percent map (Fig. 20) also does not provide an explanation for the zone of high apparent dispersivity in the northern part of the apparent dispersivity map (zone E) (Fig. 18). It is possible that zone E is an area of coarser grained sediments that is not indicated on available geologic cross sections, geologic drill logs, or sediment percentage maps. Because characteristics other than grain-size may be important in controlling dispersivity, these zones should not be considered as anomalous or

erroneous.

The zones may reflect the presence or absence of the Ringold Formation in the upper 60 feet o f the unconfined aquifer. If the upper part of the unconfined aquifer is largely contained within either the Ringold

Formation sediments or the glaciofluvial sediments, this will affect the dispersion of contaminants. The glaciofluvial materials will exhibit higher dispersivities because they are generally coarser-grained and more

heterogeneous at the macro-scale (10's of feet) than the sediments of the Ringold Formation.

Examination of Figure 18 reveals that the southern low apparent dispersivity zones A and B are situated south of a high apparent dispersivity zone (E). Calculations utilizing the Ringold Formation/glaciofluvial

sediments contact (obtained from geologic drilling logs) indicate that the upper part of the unconfined aquifer in this location consists of increasingly more Ringold Formation (rather than glaciofluvial sediments) from north to south. In the vicinity o f zone B, up to 75% of the upper 60 feet of the unconfined aquifer. The opposite trend occurs with the northern high apparent dispersivity zone (E). Calculations using the location of the Ringold Formation/glaciofluvial sediments contact indicate that the unconfined aquifer consists entirely of glaciofluvial sediments in this location, while to the east and to the west, the unconfined aquifer consists entirely of Ringold Formation sediments.

Analysis of 1975 Plume

By comparing the apparent dispersivity zone map for the 1975 analysis (Fig. 19) with the percent fines map (Fig. 21), similar trends are again seen. As before, areas o f coarse-grained sediments tend to fall within high apparent dispersivity zones, while finer-gained sediments fall within low apparent dispersivity zones. In the eastern section of the 1975 apparent dispersivity zone map, the A and B zones correspond to an increase in the percentages of fine-grained sediments in the same area. These low apparent dispersivity zones could also be a reflection of a westward trending lobe of fine-grained sediments nearby. The C and small B zones in the northern part of the analyzed area roughly parallel the 20% fines contour, giving an intermediate apparent dispersivity zone in an area of intermediate

percentages of fine-grained sediments. In the same area, an east-west trending E zone parallels a low (0-10%) percentage pattern, giving a high apparent dispersivity in coarser-grained sediments. In the central part of the apparent dispersivity zone map, there is a trend from a large apparent

dispersivity (E) in the east to a smaller apparent dispersivity (C) to the west. This corresponds to an increasing percentage of finer-grained sediments from east to west in the same area.

As in the 1963 tritium plume analysis, the 1975 analysis contains some apparent dispersivity zones that do not correspond to trends

observable on the percent fines map. In the southern part of the analysis area, there are low apparent dispersivity zones (A and B) where there are dominantly coarse-grained sediments indicated on the percent map. In the northern part of the analysis area, there are large apparent dispersivity zones (D and E) in an area where fine-gained sediments dominate. However, since characteristics other than grain-size affect dispersivity, these zones should not be considered as anomalous or erroneous. Discussion of Field Applications

It is apparent from the above results that there is a strong correspondence of the apparent dispersivity zones and dominant grain size in the upper part of the unconfined aquifer. Where there are low percentages of fine-grained sediments, there is generally higher apparent dispersivities, and vice versa.

The zones obtained from the inverse plume analysis on the 1963 and 1975 plumes that do not correspond to trends seen in the percent fines maps could be explained by trends in the location of the Ringold Fonnation/glaciofluival deposits contacts and the position of the water table. These trends suggest that the amount of Ringold Formation within the upper 60 feet o f the unconfined aquifer will control, to a large extent, the apparent dispersivities obtained from inverse plume analysis. The apparent dispersivity zones obtained from the analyses of the tritium plume are explained as reflecting the amount of Ringold Formation within the upper part of the unconfined aquifer. The greater the

percentage of Ringold Formation within the upper 60 feet of the unconfined aquifer, the lower the apparent dispersivity will be. The lower the amount of the percentage of Ringold Formation (i.e., the more glaciofluvial deposits) within the upper 60 feet of the unconfined aquifer, the higher the apparent dispersivity will be.

• Sediment characteristics other than grain size influence resulting dispersivity. Examples of these other characteristics include the amount and consistency of cementation. Therefore, a lack of direct correspondence with the percent fines maps does not indicate that the method produces erroneous results.

Unfortunately, the area of overlap between the 1963 and 1975 tritium plume analyses falls mainly within the large low dispersivity zone (A) in the 1975 analysis which was considered anomalous. It was hoped that the area of overlap between the two analyses would aid in the separation o f the effects of sediment character and the effects of

contaminant sources and the ground-water flow system on the resulting apparent dispersivity zones.

Problems

It should be noted that some of the zones determined by the previously described analyses could be effects of the multiple contaminant sources and the changing ground­ water flow field. As can be seen by comparing Figures 18 and 19, there are consistent apparent dispersivity zones appearing on both maps. It is hypothesized that the consistent predictions are reflecting variations in the character of the sediments while the inconsistent predictions are the result of contaminant source and ground-water flow dynamics.

POTENTIAL USES OF INVERSE PLUME ANALYSIS TO ASSESS AQUIEER HETEROGENEITIES

Using inverse plume analysis to assess subsurface geology has many potential uses. These include:

(1) Inferring the presence of geologic materials in locations where no borehole data exists;

(3) Calibrating numerical models;

(4) Generating a series of two-dimensional slices to illustrate geologic trends in three dimensions. Three-dimensional data can be collected from a series of multilevel sampling devices as described by Pickens and others (1978);

(5) Analysis of existing contaminant plumes to assess subsurface geology as an aid to aquifer restoration and/or contaminant plume containment;

(6) Introduction of ground-water tracers to analyze the resulting plume to assess subsurface geology;

(7) Analyzing natural chemistry variations in ground water to assess subsurface geology;

Applying inverse plume analyses as a means to assess aquifer heterogeneities can provide conceptual models of aquifers that can guide further data collection and input to ground­ water numerical models.

EUXUBEJffiQRK

As with any research project, this one is not complete. Much work needs to be done. Because isotropic and homogeneous aquifers are rarely found in nature, the effects of anisotropy and on the resulting apparent dispersivities need to be examined. More numerical modeling studies could examine these effects and the effects of spatial and temporal variations of the contaminant source on the resulting apparent dispersivities. Since the method is somewhat work intensive at this time a way to automate should be investigated. In order to do this, a statistical assessment of the delineation of the apparent dispersivity zones should be undertaken. It is rather subjective at this time.

SUMMARY

A method for assessing the subsurface geology of an area has been developed by using an inverse plume technique to determine aquifer dispersivities and examining the spatial variation and distribution of these dispersivities. Since dispersivity depends on the aquifer geology, the variation of dispersivity will be a reflection of aquifer geology.

It appears that the inverse plume method is useful for the qualitative delineation of aquifer heterogeneities since the method does not return the absolute value of dispersivity but rather, an apparent value of the dispersivity. Despite this, inverse plume analysis can still be used to determine aquifer heterogeneities by examining spatial variations in

dispersivity. From the numerical model analysis results, it also appears that transverse dispersivity is more useful than longitudinal dispersivity because there is less affect from the violation of the assumptions of an isotropic and homogeneous aquifer.

Maps of apparent dispersivity zones that reflect aquifer material changes in the subsurface where no boreholes penetrate can be constructed using the inverse plume analysis technique. When geologic cross sections and sediment trend maps are compared to the dispersivity zones delineated from analysis on the 1963 and 1975 Hanford tritium plume similar trends are observed. Where the dispersivity zone map indicates a trend from high to low dispersivity, the geologic data indicates a trend from low percentages of fine­ grained sediments to high percentages of fine-grained sediments. This positive correlation leads to optimism for using the hard information obtained from inverse plume analysis (apparent dispersivities) to assess and quantify soft information such as the nature, size, and distribution of aquifer heterogeneities which are significant at the scale of study. This method, when used in conjunction with other available data, may prove to be invaluable in establishing the heterogeneity of an aquifer. By being able to assess the aquifer geology in greater detail, more accurate and realistic ground-water numerical models can be developed.