xvEPA
United States
Environmental Protection
Agency
Hazardous Waste Engineering
Research Laboratory
Washington DC 20460
EPA/600/2-87/065
August 1 987
Research and Development
Construction Quality
Control and
Post-Construction
Performance
Verification for the
Gilson Road
Hazardous Waste Site
Cutoff Wall
-------
EPA/600/2-87/065
August 1987
CONSTRUCTION QUALITY CONTROL AND POST-CONSTRUCTION
PERFORMANCE VERIFICATION FOR THE
GILSON ROAD HAZARDOUS WASTE SITE CUTOFF WALL
by
Matthew J. Barvenik & John E. Ayres
Goldberg-Zoino & Associates, Inc.
320 Needham Street
Newton Upper Falls, Massachusetts 02164
Project Officer
Stephen James
Land Pollution Control Division
Hazardous Waste Engineering Research Laboratory
Cincinnati, Ohio 45268
WATER SUPPLY AND POLLUTION CONTROL COMMISSION
STATE OF NEW HAMPSHIRE
CONCORD, NEW HAMPSHIRE 03301
HAZARDOUS WASTE ENGINEERING RESEARCH LABORATORY
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
U.S. Environmental Protection Agency
Region 5, Library (PL-12J)
77 West Jackson Boulevard, 12th Floor
Chicago, IL 60604-3590
-------
NOTICE
The information in this document has been funded by the U.S.
Environmental Protection Agency and the New Hampshire Water
Supply and Pollution Control Commission under cooperative
agreement Contract No. CR811130-01 with Goldberg-Zoi no &
Associates, Inc. It has been subjected to the Agency's peer
review and administrative review, and it has been approved for
publication as a U.S. EPA document. Mention of trade names or
commercial products does not constitute an endorsement or
recommendation for use.
11
-------
FOREWARD
Today's rapidly developing and changing technologies and
industrial products and practices frequently carry with them the
increased generation of solid and hazardous wastes. These
materials, when improperly dealt with, can threaten both public
health and environment. Abandoned waste sites and accidental
releases of toxic and hazardous substances to the environment
also have important environmental and public health implications.
The Hazardous Waste Engineering Research Laboratory helps provide
an authoritative and defensible engineering basis for assessing
and solving these problems. Its products support the policies,
programs, and regulations of the Environmental Protection Agency;
the granting of permits and other responsibilities of State and
local governments; and the needs of both large and small
businesses in handling their wastes responsibly and economically.
This report describe assessment activities undertaken to
evaluate the effectiveness of a soil/bentonite backfilled cutoff
wall (slurry trench) installed for the purpose of hazardous waste
containment. The work includes development and revaluation of
field quality control tests, evaluation of electronic piezocone
instrumentation for post-construction verification of backfill
homogeneity, and evaluation of cutoff wall bulk hydraulic
conductivity via hydraulic stress testing. This information in
this report is useful to those involved in the feasibility study,
design and/or construction of cutoff walls as a hazardous waste
remediation technology.
For further information, please contact the Land Pollution
Control Division of the Hazardous Waste Engineering Research
Laboratory.
Thomas R. Hauser, Director
Hazardous Waste Engineering
Research Laboratory
11
-------
ABSTRACT
This report is intended to provide a critical assessment of
the construction and effectiveness of a soil/bentonite backfilled
cutoff wall (slurry trench) installed for the purpose of
hazardous waste containment. The wall was designed, constructed,
and monitored as part of the remediation action implemented on
the Gilson Road site in Nashua, New Hampshire (National Priority
List [NPL] #23). This work was conducted in three distinct
phases.
The first phase consisted of an evaluation of the accuracy,
precision, and applicability of two new field quality control
tests developed specifically to be used for control of wall
construction. The API fixed ring test and the methylsne blue
titration test allowed real time determination of backfill
hydraulic conductivity and percent bentonite, respectively. The
resulting field data were assessed with respect to cutoff wall
specification requirements and guidelines were developed.
The second phase of the work encompassed field testing of
electronic piezocone instrumentation to determine feasibility for
post-construction verification of backfill homogeneity. This
entailed evaluation of piezocone response characteristics with
respect to intact backfill versus non-backfill materials in a
specially constructed full scale test section. Response
signatures were then used to evaluate the homogeneity of the
Gilson Road containment wall followed by verification via direct
sampling.
The primary objective of the third phase of the project was
to hydraulically stress the containment wall to assess its bulk
hydraulic conductivity and thus its overall effectiveness with
respect to hazardous waste containment. This work was
supplemented with documentation of the spacial and temporal
variation in contaminant concentrations as they reflect the
behavior of the hydrologic system in response to containment wall
installation.
This report is intended to document actual site specific
data with respect to the effectiveness of a soil/bentonite cutoff
wall used for hazardous waste containment. The data can then be
used by those formulating and reviewing remedial action plans to
rationally assess the suitability of cutoff walls incorporated as
part of the overall site remediation proposed for other sites.
IV
-------
CONTENTS
NOTICE i i
FOREWORD iii
ABSTRACT i V
CONTENTS V
FIGURES ix
TABLES xi i
ACKNOWLEDGEMENTS xiv
1. INTRODUCTION 1-1
1.1 Overview 1-1
1. 2 Background 1-3
1.2.1 Site Description and Geology 1-3
1.2.2 Remedial Measures 1-4
1.2.3 Design and Construction 1-5
1.3 Report Organization 1-5
2. FIELD QUALITY CONTROL TESTING PROCEDURES 2-1
2.1 Evaluation of Backfill Hydraulic Conductivity
Using the Fixed Ring Apparatus 2-2
2.1.1 API Fixed Ring Test 2-4
2.1.2 Testing Program 2-6
2.1.2.1 API Fixed-Ring Test Precision 2-6
2.1.2.2 API Fixed-Ring Test Accuracy 2-8
2.1.3 Guidelines for Field use of the API Fixed-
Ring Test 2-13
2.1.3.1 Verification Testing 2-14
2.1.3.2 Production Testing 2-14
2.2 Evaluation of Backfill Bentonite Content Using
Methylene Blue Titration 2-17
2.2.1 Methylene Blue Titration 2-18
2.2.2 Testing Program 2-20
2.2.2.1 Distilled Water Matrix 2-20
2.2.2.2 Ottawa Sand/Distilled Water Matrix... 2-23
2.2.2.3 Contaminated Water Matrix 2-25
2.2.2.4 Gilson Road Backfill Matrix 2-27
2.2.3 Field Use of the Methylene Blue Test 2-31
2.3 Effect of Backfill Gradation on Hydraulic
Conductivity 2-33
2.3.1 Data Base 2-34
2.3.2 Statistical Evaluation 2-35
2.3.3 Importance of Statistical Correlation 2-37
2.4 Assessment of Gilson Road QC Data 2-38
2.4.1 Review of QC Data 2-39
2.4.1.1 QC Data from Construction Samples.... 2-39
v
-------
CONTENTS (continued)
2.4.1.2 QC Data from Post-Construction
Samples 2-43
2.4.1.3 Evaluation of the Validity of the
QC Construction Data 2-47
2.4.1.4 Implication of QC Data on Containment
Perf ormance 2-47
2.5 Quality Control Guidelines 2-50
2.5.1 General Quality Control Guidelines 2-50
2.5.1.1 Test Requirements 2-50
2.5.1.2 Quality Control Execution 2-51
2.5.2 QC Guidelines for Support Slurry 2-53
2.5.2.1 Unit Weight 2-54
2.5.2.2 Viscosity 2-55
2.5.2.3 Filtrate Loss 2-55
2.5.3 QC Guidelines for Trench Excavation 2-56
2.5.3.1 Verticality 2-56
2.5.3.2 Continuity 2-57
2.5.3.3 Key Penetration 2-57
2.5.3.4 Key Cleaning 2-58
2.5.3.5 Stability 2-59
2.5.4 QC Guidelines for Backfill 2-60
2.5.4.1 Unit Weight 2-60
2.5.4.2 Slump 2-61
2.5.4.3 API Fixed-Ring Hydraulic Conductivity
Testing 2-61
2.5.4.4 Methylene Blue Titration for
Bentonite Content 2-62
2.5.4.5 Gradation 2-62
2.6 Conclusions 2-62
3 . WINDOW DETECTION VIA PIEZOCONE SOUNDING 3-1
3.1 Piezocone Operating Principles 3-3
3.1.1 Standard Instrument Configuration 3-3
3.1.2 Instrument Modification 3-5
3.1.3 Sounding Procedure 3-7
3.1.4 Data Manipulation and Output Format 3-9
3.2 Material Identification 3-10
3.2.1 Point Resistance 3-11
3.2.2 Local Friction 3-12
3.2.3 Excess Pore Pressure 3-12
3.2.4 Relative Hydraulic Conductivity Index 3-15
3.2.5 Empirical Correlations 3-17
3.2.6 Window Identification 3-21
3.2.6.1 Intact Soil/Bentonite Backfill 3-22
3.2.6.2 Clean Granular Window 3-25
3.2.6.3 Slurry Filled Granular Window 3-27
3.2.6.4 Bentonite Support Slurry Filled
Window 3-28
vi
-------
CONTENTS (continued)
3.3 Cutoff Wall Test Section 3-29
3.3.1 Test Section Construction 3-30
3.3.1.1 Excavation 3-30
3.3.1.2 Backfilling 3-33
3.3.1.3 Window Emplacement 3-34
3.3.2 Test Section Window Verification 3-37
3.3.2.1 High Density Clean Granular Window... 3-38
3.3.2.2 Low Density Clean Granular Window.... 3-40
3.3.2.3 Slurry Filled Granular Window 3-40
3.3.2.4 Support Slurry Filled Window 3-40
3.3.3 Test Section Soundings 3-42
3.3.3.1 Intact Backfill Analysis 3-53
3.3.3.2 High Density Clean Granular Window... 3-58
3.3.3.3 Low Density Clean Granular Window.... 3-61
3.3.3.4 Slurry Filled Granular Window 3-65
3.3.3.5 Support Slurry Filled Window 3-69
3.3.4 Window Identification Guidelines 3-69
3.3.4.1 High Density Clean Granular Window... 3-70
3.3.4.2 Low Density Clean Granular Window.... 3-70
3.3.4.3 Slurry Filled Granular Window 3-71
3.3.4.4 Support Slurry Filled Window 3-71
3 . 4 Containment Wall Soundings 3-71
3.4.1 Intact Backfill; Sounding GZ-13 3-73
3.4.2 Potential Window Locations, GZ-13 3-76
3 . 5 Window Verification 3-81
3.5.1 Sample Classification 3-82
3.5.1.1 Boring GZA-13, 43 to 48 Feet 3-83
3.5.1.2 Boring TBF-lA, 39 to 46 Feet 3-84
3.5.1.3 Boring SU-7, 30.5 to 35.5 Feet,
45.5 to 48 Feet, and 49.5 to
52.0 Feet 3-84
3.5.1.4 Boring SU-4, 21 to 26 Feet 3-85
3.5.2 Sample Hydraulic Conductivity Testing 3-86
3.5.3 Containment Wall Homogeneity 3-86
3.6 Conclusions 3-87
4. EVALUATION OF CONTAINMENT LEAKAGE 4-1
4 .1 Bedrock Pumping Test 4-2
4.1.1 Pumping Well Location and Installation 4-3
4.1.2 Recharge Trenches 4-5
4.1.3 Monitoring Points 4-7
4.1.3.1 Existing Points 4-7
4.1.3.2 New Installations 4-9
4.1.4 Execution of Pump Test 4-10
4.1.4.1 General Conditions 4-10
4.1.4.2 Data Discussion 4-13
4.1.5 Analysis 4-18
4.1.6 Conclusions 4-26
vii
-------
CONTENTS (continued)
4.2 Cutoff Wall Hydraulic Efficiency 4-27
4.2.1 Test Constraints 4-28
4.2.2 Reference Data Set 4-29
4.2.3 Numerical Methodology 4-33
4.2.3.1 Model Description 4-34
4.2.3.2 Model Calibration 4-37
4.2.4 Sensitivity Analysis 4-39
4.2.5 Cutoff Wall Efficiency 4-47
4.2.6 Summary 4-49
4.3 Groundwater Quality Monitoring 4-50
4.3.1 Location and Installation of Groundwater
Monitoring Wells 4-50
4.3.2 Sampling and Analysis 4-52
4.3.2.1 Sampling of Monitoring Points 4-52
4.3.2.2 Analysis of Groundwater Samples 4-52
4.3.3 Results of Groundwater Quality Monitoring 4-53
4.3.3.1 Selection of Pertinent Groundwater
Monitoring Points 4-53
4.3.3.2 Selection of an Indicator VOC
Compound 4-55
4.3.3.3 Data Discussion 4-57
4.3.3.4 Contaminant Degradation of the
Cutoff Wall 4-61
4.4 Conclusions 4-64
5. CONCLUSIONS 5-1
REFERENCES 6-1
APPENDICES 7-1
A. API FIXED-RING PROCEDURE 7-1
B. METHYLENE BLUE TEST PROCEDURE 8-1
C. GEOPHYSICAL METHODS FOR WINDOW DETECTION 9-1
vill
-------
LIST OF FIGURES
Number Page
2-1 API Apparatus for Hydraulic Conductivity
Testing 2-5
2-2 Hydraulic Conductivity: API vs. Triaxial
(Remixed) 2-11
2-3 Triaxial Hydraulic Conductivity Results:
"Undisturbed" vs. "Remixed" Samples 2-16
2-4 Methylene Blue Test Results: Bentonite in
Distilled Water 2-22
2-5 Methylene Blue Test Results: Bentonite in
Ottawa Sand/Distilled Water 2-24
2-6 Methylene Blue Test Results: Bentonite in
Contaminated Water 2-26
2-7 Methylene Blue Test Results: Bentonite in
Borrow/Trench Spoil/Distilled Water 2-30
2-8 Hydraulic Conductivity vs. Time - Bentonite
Content vs. Time 2-40
2-9 Hydraulic Conductivity vs. Station - Bentonite
Content vs. Station 2-42
2-10 Apparent Zones of Backfill with Hydraulic
Conductivity Greater than 1 x 10-7 cm/sec 2-49
3-1 Schematic of Piezocone Instrument 3-4
3-2 Schematic of Permeameter 3-6
3-3 Permeameter Pressuri?ation/Flow Measurement
System 3-8
3-4 Excess Pore Pressure Dissipation (Wissa, 1975)... 3-16
3-5 Material Identification-Friction Ratio 3-18
3-6 Material Identification-Friction Ratio 3-19
3-7 Material Identification-Excess Pore Pressure 3-20
ix
-------
LIST OF FIGURES (continued)
3-8 Relative Piezocone Response 3-23
3-9 Location Plan 3-32
3-10 Test Section Construction Profile 3-35
3-11 Post-Construction Test Section Profile 3-39
3-12 Piezocone Test—Sounding No. GZl 3-43
3-13 Piezocone Test—Sounding No. GZ2 3-44
3-14 Piezocone Test—Sounding No. GZ3 3-45
3-15 Piezocone Test—Sounding No. GZ4 3-46
3-16 Piezocone Test—Sounding No. GZ5 3-47
3-17 Piezocone Test—Sounding No. GZ6 3-48
3-18 Piezocone Test—Sounding No. GZ7 3-49
3-19 Piezocone Test—Sounding No. GZ8 3-50
3-20 Piezocone Test—Sounding No. GZ9 3-51
3-21 Piezocone Test—Sounding No. GZlO 3-52
3-22a Piezocone Test—Sounding No. GZ13 3-74
3-22b Piezocone Test—Sounding No. GZ13 3-75
4-1 Pumping Test Monitoring Network 4-4
4-2 Pumping Well Installation 4-6
4-3 Typical Existing Multilevel Installation 4-8
4-4 Typical Pumping Test Multilevel Installation 4-11
4-5 PT-2 Time-Drawdown Curve 4-14
4-6 M-14 Time-Drawdown Curve 4-15
4-7 M-3-4/M-3-5 Time-Drawdown Curve 4-17
4-8 Measured Bedrock Drawdown Contours 4-19
-------
LIST OF FIGURES (continued)
4-9
4-10
4-11
4-12
4-13
4-14
4-15
4-16
4-17
4-18
4-19
4-20
4-21
4-22
4-23
4-24a
4-24b
Pumping Test Idealized Geometry
Modeled Bedrock Drawdown Contours
Measured Water Table Contours, June, 1983
Vertical Head Distribution at M-10R 1 M-14,
June, 1983
Cutaway View of Numerical Model
Plan View of Top Layer of Model
Jan. 82 Scenario Calibration
Oct. 83 Scenario Calibration
Sept. 85 Scenario Calibration
Modeled Vertical Head Distributions for
Different Wall Conductivities
Location of Groundwater Monitoring Points
Location of Selected Groundwater Monitoring
Points
Variations in Contaminant Distribution
Variations in Contaminant Distribution
4-20
4-21
4-24
4-25
4-30
4-32
4-35
4-36
4-40
4-41
4-42
4-43
4-45
4-51
4-56
4-58
4-59
XI
-------
LIST OF TABLES
Number Page
2-1 Results of Hydraulic Conductivity Tests in
API Fixed-Ring Apparatus: Precision
Series 2-7
2-2 Hydraulic Conductivity Testing: API Fixed-
Ring Vs. Triaxial (Remixed) 2-10
2-3 Effect of Varying Gradient in API Fixed-
Ring Test 2-13
2-4 Triaxial Hydraulic Conductivity: "Undisturbed"
vs. "Remixed" Samples 2-15
2-5 Methylene Blue Test Results On-Site (Trench
Spoil) 2-28
2-6 Methylene Blue Test Results Off-Site Borrow 2-29
2-7 Correlation Coefficients - Hydraulic Conductivity
vs. Gradation Parameters 2-36
2-8 Hydraulic Conductivity: "Undisturbed" vs.
"Remixed" Samples 2-45
3-1 Support Slurry QC Data 3-31
3-2 Soil/Bentonite Backfill QC Data 3-33
3-3 Intact Backfill 3-54
3-4 Relative Hydraulic Conductivity Index 3-55
3-5 High Density Clean Granular Window 3-59
3-6 Low Density Clean Granular Window 3-62
3-7 Slurry Filled Granular Window 3-66
4-1 Daily Precipitation During Pumping Test (9/85)... 4-13
4-2 Stress Scenarios and Simulation Fluxes 4-38
4-3 Calibration Deviation Statistics 4-39
xn
-------
LIST OF TABLES (continued)
4-4 Modeled Outward Fluxes 4-48
4-5 Sensitivity Run Fluxes 4-49
4-6 Concentrations of Tetrahydrofuran in Parts
Per Million 4-57
XHL
-------
ACKNOWLEDGEMENTS
This document was prepared by Goldberg-Zoino & Associates,
Inc. in partial fulfillment of contract No. CR811130-01, which
was funded under a cooperative agreement between the New
Hampshire Water Supply and Pollutant Control Commission and the
U.S. EPA's Hazardous Waste Engineering Research Laboratory. The
state of New Hampshire's project officer was Michael Donahue with
major input from Michael Sills. The EPA project officer was
Stephen James. Initial input was provided by Dr. Walter Grube
and Jon Herrmann, also of the U.S. EPA's Hazardous Waste
Engineering Research Laboratory.
The Principal-in-Charge of the project for Goldberg-Zoino &
Associates, Inc. was John Ayres; Matthew Barvenik was Project
Manager and principal author. Other GZA contributing authors
included William Badge, David Brown, Thomas Kern, and Brian
Dorwart.
The following individuals provided direct input to the
project via laboratory work, field observation, data analysis,
etc., during the design and construction of the Gilson Road
cutoff wall and/or during execution of the project:
Michael Powers
Charles Lindberg
Donald Schulze
Katrina Grundstrom
John Forde
Tim Cady
Paul Hertzlier
Ralph Wickson
Bob Ostrofsky
Dan Tiltgis
Joe Lewis
James Handly
GZA
GZA
GZA
GZA
GZA
GZA
State of New Hampshire
State of New Hampshire
State of New Hampshire
Case International
(now with Recosol, Tnc,
Case International
James Handly Company
xiv
-------
SECTION 1
INTRODUCTION
The work performed under this contract was funded under a
cooperative agreement between the New Hampshire Water Supply and
Pollution Control Commission and the U.S. EPA Hazardous Waste
Engineering Research Laboratory. As a cooperatively funded
project, two separate but complementary objectives were
established. The first, and primary, objective was to evaluate
the design, construction, operation, and monitoring of the Gilson
Road National Priority List site remediation. As such, this
information enhanced the overall effectiveness of the passive and
active barrier containment system. The second objective was to
conduct an evaluation which would lead to a better understanding
of containment technologies in general. This knowledge could
then be incorporated into the design and construction of
hazardous waste remediation actions on other NPL sites.
1.1 Overview
Soil/bentonite backfilled cutoff walls have successfully
been used as an adjunct to pumping for excavation dewatering and
general groundwater control for over forty years. The procedure
of adding bentonite to excavated in-situ soil and then placing
the low hydraulic conductivity mixture back in the
slurry-supported trench to form a barrier was initially used in
Europe and then in the United States. Specifications governing
this work typically fell under the dewatering section of contract
construction documents and, as such, were performance oriented.
In many cases, success was judged solely on the ability of the
barrier to limit fluid flux to quantities that could reasonably
be pumped. Due to the magnitude of the overall construction
effort typically involved, the addition of pumping to remove
leakage from the excavation with off-site disposal was often
considered insignificant.
More recently, soil/bentonite (S/B) cutoff technology has
been used as part of remedial actions at United States
Environmental Protection Agency National Priority List (NPL)
sites, as well as private industrial sites. (JRB, 1984). For
these more critical applications, leakage must be reduced to a
1-1
-------
far greater extent than typically associated with construction
dewatering.
The successful construction of a hazardous waste cutoff wall
containment is largely dependent on the quality of the
contractor's workmanship. As such, specifications must remain
generally performance oriented with the primary requirement of
attaining a long-term hydraulic conductivity less than some
specified value; typically less than 1 x 10~7 cm/sec for waste
containment. However, in recognition of the difficulties
involved in performance verification and the consequences of
detecting substandard construction only after wall completion,
contract documents must also include minimum design standards.
These center around a number of design and construction issues
such as minimum bentonite and fines content in the backfill, for
example (Ayres, 1983). This allows quality control (QC) testing
to be executed in the field during construction as a check on
attainment of the minimum standards specified. "Learning Curve"
problems can therefore be detected and addressed early in the
construction process before they become pervasive and jeopardize
the overall effectiveness of the barrier. Adequate specification
and quality control of such minimum design standards requires
intimate knowledge of the technology underlying cutoff wall
construction as tempered with practical empiricism gained via
construction experience.
As is often the case with initial technology transfer,
cutoff walls have been designed and specified by professionals
associated with the hazardous waste field, rather than those
experienced in cutoff wall design/construction. Environmental
professionals, although familiar with the hydrogeologic
constraints of the application, generally lack expertise and
experience in cutoff wall specific technology. Specifications
have therefore been strictly performance based in some cases,
thus allowing contractors to execute the work in a manner
consistent with past dewatering practices. Unfortunately, fluid
flux rates associated with standard construction practices for
dewatering purposes are unacceptable with respect to hazardous
waste containment. As a result, some of these containments have
not performed up to expectation. Cutoff walls have thus fallen
from favor with respect to hazardous waste remediation due to
poor implementation rather than limitations inherent in the
technology.
Use of procedures, such as field QC, to prevent cutoff wall
deficiencies before they occur is not only highly desirable but
necessary for hazardous waste containment applications. However,
it must also be recognized that the specifications under which
these walls are constructed are principally performance based.
The performance clauses incorporated in the specifications can
only be meaningful to the extent that equipment and methods are
1-2
-------
available to enforce them, should post-construction verification
become necessary. A demonstrated capability for performance
verification also instills added incentive to achieve the highest
quality of workmanship possible during construction. As such,
instrumentation and procedures should be developed to identify
unsatisfactory zones in completed cutoff walls as well as
hydraulically test their post-construction performance.
1.2 Background
The Gilson Road uncontrolled hazardous waste site in Nashua,
New Hampshire was the subject of the first site remediation
cooperatively funded under EPA's Superfund program. Clandestine
dumping of toxic volatile organic compounds into an abandoned
gravel pit resulted in a contaminant plume which degraded
residential air quality as the contaminants volatilized following
recharge to a local stream and threatened downstream municipal
drinking water supplies (Barvenik, 1986).
1.2.1 Site Description and Geology
The site is underlain by glacial soils consisting of
primarily granular ice-contact and glacial lake deposits. These
include fine to coarse sands and gravels, with layers and lenses
of silts and silty fine sands. Below the stratified drift is a
granular basal till. This, in turn, is underlain by a moderately
fractured schist bedrock, with a weathered, highly fractured
upper zone (Koteff, 1973).
The sand and sand/gravel deposits exhibit an average
hydraulic conductivity of 10~2 cm/sec based on pumping test
results. However, these deposits vary over a range from
10~5 cm/sec for the silts and silty fine sands to an estimated
10-1 cm/sec for open work gravel seams found in the esker
deposits. The glacial till deposits are typically less permeable
(10~5 cm/sec) than the average values for the overlying
stratified drift. However, the till is generally thin and
discontinuous in nature, thus constituting a relatively imperfect
aquitard. The hydraulic conductivity of the upper 50 feet of the
bedrock was estimated to range between 10~5 an
-------
1.2.2 Remedial Measures
Final designed remedation for the site consisted of
complementary active and passive physical barriers to control
groundwater flow, and treatment of the captured contaminant plume.
This interactive system was adopted because neither a purely
passive barrier nor a purely active barrier alone could eliminate
off-site contaminant migration. A cutoff wall to the top of
bedrock and a synthetic cap form the passive elements of the
system. The cutoff wall was 4,000 feet in length, up to 110 feet
deep, and completely encompassed the entire 20 acre site.
Extending the passive barrier through the fractured portion of
the bedrock was evaluated during initial design phases. This
could be accomplished by excavating through the bedrock or
grouting the rock fractures. Albeit feasible, neither method was
found to be a cost-efficient solution for limiting contaminant
flux through the bottom of the containment.
The active component of the total containment vessel is a
hydrodynamic isolation system which accounts for the inadequacies
of the bedrock as an aquiclude. This system redistributes the
groundwater within the containment area and eliminates head
differences across the cutoff wall to achieve a no leakage
condition (zero gradient). The groundwater extraction/recharge
associated with the hydrodynamic isolation also allows for
treatment of the recirculated water, resulting in aquifer
restoration. The restoration process requires a number of
flushing cycles for complete contaminant removal.
A purely active system of plume control via hydrodynamic
extraction/recharge was also found to be unacceptable during the
feasibility study due to the heterogeneity of the aquifer.
Groundwater extraction at a rate necessary to eliminate plume
migration would necessarily entail the pumping of significant
volumes of clean water from regions of high hydraulic
conductivity outside the plume and nearby surface waters. This
dilution of the plume at the point of recharge and the
concomitant spreading quickly results in a contaminant plume
width greater than the extraction system's radius of influence.
Even in a homogeneous aquifer, contaminants can still escape
active containment if the recharge point does not lie directly
upstream of the extraction well (i.e., along a flow line).
Finally, even with a perfect recirculation system, contaminant
dispersion across the flow lines at the edge of the containment
results in a further loss of contaminant mass from the system.
Although the cutoff wall provides enhanced hydraulic
control, no practical hydrodynamic isolation system incorporating
a finite number of extraction and recharge points, can completely
1-4
-------
eliminate all gradients across the physical barrier. Therefore/
to preclude off-site contaminant migration, the system is
operated with a conservative bias by forcing an inward gradient
everywhere along the cutoff wall. The required net recharge
deficit, relative to extraction, necessitates an external purge
stream (discharged outside of the containment). The purge stream
volume is, in part, a function of the hydraulic conductivity of
the containment perimeter (cutoff wall). To attain an extremely
high water quality in the purge stream, as mandated for off-site
disposal at this CERCLA site, -this portion of the pumpage must
undergo additional treatment steps as compared to the total
recirculation flow. Therefore, the hydraulic efficiency of the
cutoff wall directly affects purge stream treatment costs, which
are a considerable fraction of the two million dollar/year
overall operation and maintenance cost of the on-site treatment
f ac i 1 i ty.
1.2.3 Design and Construction
In recognition of the importance of the cutoff wall to the
successful operation of the hydrodynamic isolation/recirculation
system, a rigorous design phase was implemented including an
advanced soil/bentonite backfill design methodology (Schulze,
1984) and development of new equipment and procedures for quality
control field testing. This work was incorporated into the
specifications as minimum design standards keyed to specific
aspects of the actual construction process (Ayres, 1983). The
Gilson Road project therefore represents perhaps the first case
where the level of cutoff wall design, specification, and
construction quality control (Q/C) were commensurate with
application to hazardous waste containment. As such, this site
was considered ideal for use in evaluation of the suitability of
soil/bentonite cutoff wall technology as part of an overall
containment strategy for hazardous waste remediation.
1.3 Report Organization
The scope of this work was separated into three major phases.
The first phase (Section 2) encompassed evaluation of two new
testing techniques used for quality control during construction
of the Gilson Road containment. The methylene blue titration and
the API fixed ring hydraulic conductivity tests were employed to
determine the bentonite content and hydraulic conductivity of the
intact soil/bentonite cutoff wall backfill, respectively. The
project described herein demonstrated the applicability,
accuracy, precision, and practicality of these tests for quality
control.
1-5
-------
The second phase of this project (Section 3) was directed at
investigation of the feasibility of piezocone instrumentation for
verification of cutoff wall backfill homogeneity. This entailed
evaluation of piezocone response characteristics with respect to
intact backfill versus non-backfill materials in a specially
constructed full scale test section. Response signatures were
then used to evaluate the homogeneity of the Gilson road
containment wall followed by verification via direct sampling.
The primary objective of third phase of the project
(Section 4) was to hydraulically stress the containment wall to
assess its bulk hydraulic conductivity and thus its overall
effectiveness with respect to hazardous waste containment. This
work was supplemented with documentation of the spacial and
temporal variation in contaminant concentrations as they reflect
the behavior of the hydrologic system in response to containment
wall installation.
Each of the three above described major sections are
organized to be independent, self-contained documents. The
conclusions derived from the work are therefore presented after
each individual section. The report concludes with a final
section summarizing the general conclusions which can be drawn
from this project.
1-6
-------
SECTION 2
FIELD QUALITY CONTROL TESTING PROCEDURES
Quality control testing of soil/bentonite backfill during
cutoff wall construction is essential from two perspectives: 1)
it allows an on-going evaluation of construction procedures; and
2) test results form the basis of evaluating future wall
performance. QC testing provides a means of assessing whether
minimum design standards have been met for a discrete number of
representative backfill samples obtained prior to placement in
the trench. From this set of data, future performance of the
wall as a whole is extrapolated. While recognizing that wall
performance may be significantly impaired by anomalies in the
backfill not revealed by the QC data, QC testing imposes control
over a construction procedure for which there are few visual
indicators of quality. Emphasis of QC testing during
construction thus allows for the timely and efficient correction
of deficiencies. Over the long-term, QC testing during
construction also reduces the need for post-construction
verification of backfill parameters—which is costly—and the
reliance on post-construction "leakage data"--which can be
circumstantial and subject to varied interpretation. Moreover,
even if deficiencies can be detected in the completed wall, their
correction is generally neither simple nor inexpensive.
An effective QC program involves frequent testing of primary
backfill parameters, including unit weight, slump, gradation,
fines content, bentonite content, and hydraulic conductivity.
The real time data obtained through field QC testing are made
available to the contractor to allow modification of component
proportioning and mixing effort at the outset of the project to
achieve specified requirements. Standard methods for determining
unit weight, slump, fines content, and overall gradation are
easily executed in the field within a time period commensurate
with real time QC of field operations (typically within one day).
However, long periods of time are required for triaxial testing
for hydraulic conductivity (greater than one week) and standard
hydrometer testing of bentonite content (typically three to four
days). In recognition of these limitations, new S/B backfill QC
techniques, including API fixed-ring permeation, and methylene
blue hitration, were utilized on the Gilson Road project for
bentonite content and hydraulic conductivity determinations,
2-1
-------
respectively. This work demonstrated the applicability and
practicality of these new methods for field control.
The first phase of the project, presented in this chapter,
focused on the refinement of these two new QC procedures and
evaluation of their precision and accuracy using samples obtained
from the Gilson Road site (Sections 2.1 and 2.2). Guidelines for
field use were developed. To better assess the role of gradation
parameters on hydraulic conductivity, a statistical evaluation of
the QC data obtained from the Gilson Road project was also
performed (Section 2.3). The QC data obtained from the Gilson
Road project were then reviewed to assess the conformance of the
completed wall to project specifications (Section 2.4). Finally,
quality control guidelines for cutoff wall construction were
developed, encompassing trench excavation and backfill placement
as well as backfill preparation (Section 2.5).
2.1 Evaluation of Backfill Hydraulic Conductivity Using the
Fixed-Ring Apparatus
The primary hydrologic feature of a cutoff wall is its low
hydraulic conductivity and thus high impedance with respect to
groundwater flow. This characteristic results in the
effectiveness of cutoff walls as physical barriers to contaminant
migration. In recognition of the importance of the hydraulic
conductivity actually attained during cutoff construction, this
key parameter is subject to a high degree of specification in
project contract documents. For most hazardous waste
containments, an upper bound hydraulic conductivity of 1 x 10~7
cm/sec is typically required, as was the case for the Gilson Road
site.
A number of criteria must be met during construction to
achieve a soil/bentonite cutoff wall with the specified hydraulic
conductivity. The types and percentages of backfill components
have a large effect on hydraulic conductivity and are typically
determined via laboratory testing undertaken during the design
phase (Schulze, 1984). These data are generally reflected in the
project specifications as minimum design standards. (For
example, the Gilson Road specifications required a minimum of
30 percent fines and 5 percent minimum bentonite content for the
backfill). Even if these minimum criteria are met, however, the
proper hydraulic conductivity is not guaranteed due to the
importance of uniform mixing of the backfill components,
particularly with respect to dispersion of the bentonite.
Inasmuch as proper proportioning of backfill components and
sufficient mixing effort are primarily related to the quality of
workmanship on the part of the contractor, the required hydraulic
2-2
-------
conductivity for the cutoff appears as a performance
specification in the project documents.
To have any impact as a specification requirement,
performance criteria should be subject to immediate verification.
However, in the strictest sense, the actual long-term hydraulic
conductivity of the cutoff wall can only be evaluated through
post-construction monitoring. This is a costly endeavor due to
the requirement for closely-spaced multilevel piezometric head
and groundwater quality monitoring installations located around
the containment perimeter, both on the outside and inside of the
wall. In all but the simplest cases, the data thus obtained also
require interpretational techniques based on three-dimensional
computer simulation.
Beyond the high cost, the major drawback associated with
post-construction monitoring is that zones of unacceptable
hydraulic conductivity are detected only after it is too late for
prevention and/or cost-efficient repair. Therefore, while
performance criteria are still necessary for specification of
hydraulic conductivity, a more cost-effective and timely
procedure is required for prediction, and thus construction
control, of the hydraulic conductivity finally achieved by the
barrier. This function can be fulfilled via QC testing as
demonstrated on the Gilson Road project.
Though not an absolute determinant of the long term,
in-place performance of the overall cutoff, hydraulic
conductivity testing of the backfill prior to placement in the
trench provides a good indicator of the eventual intact hydraulic
conductivity of the barrier. Standard triaxial testing has
therefore been routinely employed to provide these data despite
its limitations. Due to the sophistication of the equipment
required, the tests are costly (typically greater than several
hundred dollars/test) and cannot easily be performed in the field.
As such, a statistically insignificant number of tests are
generally undertaken (normally not more than one test per
multiple thousands of cubic yards of backfill) and the data
become available only after several weeks have elapsed since the
samples were obtained. This procedure suffers from the syndrome
of "too little and too late" with respect to influence on the
eventual performance of the cutoff barrier. Hence, it can only
be viewed as a token post-construction verification effort.
An alternate procedure for obtaining hydraulic conductivity
data employs fixed-ring equipment. Although, in general, not as
accurate as triaxial testing, fixed-ring tests yield acceptable
data if the correct procedures are followed to limit boundary
leakage and the samples are conformal and initially saturated, or
nearly so (Schulze, 1984).
2-3
-------
Fixed-ring equipment was used to provide hydraulic
conductivity data during construction of the Gilson Road cutoff
wall. This inexpensive, easily-operated equipment allowed field
testing of numerous samples prior to permanent incorporation of
the backfill into the cutoff. The one day availability of
hydraulic conductivity data allowed real-time quality control of
the construction process. In light of the high degree of
usefulness and practicality of the API fixed-ring hydraulic
conductivity testing as shown on the Gilson Road project, a
testing program was undertaken to:
0 Assess the precision of the test
0 Assess the accuracy of of the test.
0 Refine and standardize the test procedure for
incorporation in general guidelines for QC testing of
S/B cutoff walls.
The test program and test results are described in the
following subsections.
2.1.1 API Fixed-Ring Test
Various equipment configurations based on fixed-ring
confinement of a test sample for evaluation of hydraulic
conductivity have been in use for many years (U.S. Army Corps
manual, 1970 and Lambe, 1951). In most cases, however, the
equipment is meant for laboratory use and not designed to
withstand the "environment" associated with field testing.
In an effort to standardize procedures around equipment
which would survive repeated field use and also be readily
obtainable, the conventional American Petroleum Institute (API)
filter press was adopted for use as a fixed-ring permeameter
(see Figure 2-1). As a filter press, this equipment is already
routinely used on cutoff wall projects to test the filter cake
forming capacity of bentonite slurries used for trench
stabilization.
To allow accurate testing of S/B backfill samples at
stresses commensurate with in-situ conditions, a number of minor
equipment and procedural modifications are required, including:
0 The pressure system should be modified to accept a
regulator for each cell along with a gage which has a
range of 1 to 15 psi for application of low pressure.
2-4
-------
FIG. 2-1 API. APPARATUS FOR HYDRAULIC CONDUCTIVITY TESTING
DIAL GAUGE
BASE CAP
STEEL MESH SCREEN
SUPPORT ROD
GRADUATED CYLINDER
GAS LINE
REGULATOR
COMPRESSED
AIR SOURCE
(APPARATUS AS SHOWN MANUFACTURED BY BAROID)
2-5
-------
0 Bentonite "paste" should be applied to the inside
diameter of the rigid wall and trimmed to a uniform
thickness of 1/32 inch using a trimming jig. This will
limit boundary flux along the backfi11-rigid wall
interface.
0 Test samples should be fabricated on a thin bed of sand
underlain by a porous screen to prevent plugging of the
bottom outlet port.
0 The test samples should be sieved through a standard 1/2
mesh prior to placement in the cell to limit the maximum
particle size to less than 1/6 of the cell diameter.
In basic terms, API hydraulic conductivity testing involves
placing a sample in the cell/ covering it with water, and forcing
the water through the sample using pneumatic pressure. Once
consolidation of the sample is complete, the steady state
hydraulic conductivity is computed based on standard Darcy's Law
relationships assuming 100 percent saturation. For typical S/B
backfill samples, approximate values of hydraulic conductivity
can be obtained in several hours and final values in less than 24
hours. As such, the procedure is ideal for real time quality
control of the construction process.
A detailed description of the standardized test procedure,
which directly reflects the results of efforts performed for this
project, is presented in Appendix A.
2.1.2 Testing Program
The overall program objective was to determine the inherent
reliability of the API fixed-ring test for measuring the
hydraulic conductivity of S/B backfill samples. This information
could then be used to evaluate the validity of the QC data
obtained during construction of the Gilson Road cutoff and, from
a more general standpoint, form a basis for standardization of QC
procedures for use on other projects. More specifically, the
project was structured to first evaluate the precision of the
test and then assess test accuracy. These two phases of the work
are described in the following subsections.
2.1.2.1 API Fixed-Ring Test Precision
The precision testing was directed towards determining if
equipment and procedures as simple as the API-fixed ring-based QC
2-6
-------
testing performed on the Gilson Road
repeatable hydraulic conductivity values.
project would provide
Five batches of the Gilson Road backfill were selected from
record samples obtained during construction. Selection
emphasized encompassing the full range of gradations and
hydraulic conductivities encountered during construction of the
actual cutoff wall. After thorough remixing in the laboratory,
three "identical" aliquots of each sample were tested at a
gradient of 40 (driving pressure of 3 psi - all pressures
reported as gage). The results are presented in Table 2-1.
TABLE 2-1
RESULTS OF HYDRAULIC CONDUCTIVITY TESTS IN API FIXED-RING
APPARATUS: PRECISION SERIES
Sample
Hydraulic Conductivity cm/sec
(3 splits of each sample)
Station 0+50
Station 7+75
Station 14+75
Station 20+25
Station 35+00
0.48 x 10-7
0.38 x 10-7
0.46 x 10-7
4.02 x 10-7
4.02 x 10-7
4.44 x 10-7
0.80 x 10-7
1.86 x 10-7
0.79 x 10-7
0.63 x 10-7
0.25 x 10-7
0.25 x 10-7
0.69 x 10-7
0.28 x 10-7
0.33 x 10-7
The data indicate a worst case precision of +57 percent
within one standard deviation of the mean value for any group of
three tests. Although this initially appears as a large degree
2-7
-------
of scatter, it is actually quite small when compared to the five
orders of magnitude (100,000 times) reduction in hydraulic
conductivity of the initial in-situ soils. Further inspection of
the data also reveals that in all five groups of tests, at least
two of the three tests in each group agreed to ^10 percent within
one standard deviation of the mean. Hence, some of the scatter
reflected in the worst case precision of ±51 percent may be
attributable to variations in aliquots of the same sample that
were not eliminated by mixing. It must be recognized that the
degree to which the precision of a test procedure can be
established is highly dependent on how identical the test
specimens are prepared. This is problematic for API fixed-ring
hydraulic conductivity testing in that a single specimen cannot
be set up more than once. Therefore, multiple specimens were
required to evaluate the degree of variability in the entire
procedure, from set-up through permeation. Unfortunately, the
coarseness of granular based backfills such as used on the Gilson
Road project can result in a somewhat heterogeneous sample from
which the individual aliquots are taken for the comparison
testing. Therefore, uniform mixing, to the extent practical in
the laboratory, was emphasized during sample preparation.
However, it is likely that some degree of heterogeneity still
existed among the comparison samples.
This hypothesis is supported by field data obtained during
subsequent QC testing for other cutoff walls utilizing primarily
clayey mixes. For these finer grained backfills, the probability
of obtaining identical samples is significantly enhanced. API
fixed-ring tests run on "identical" samples of clayey mixes
yielded closer precision than for the Gilson Road backfill.
2.1.2.2 API Fixed-Ring Test Accuracy
The accuracy of the API fixed-ring test was evaluated via
comparison with triaxial tests. Inasmuch as the value of
hydraulic conductivity attributed to a backfill specimen is
dependent on the method of measurement, determination of the
accuracy of a specific testing procedure suffers from the lack of
an absolute value of hydraulic conductivity for comparison. Even
triaxial testing, which is one of the most well developed and
flexible testing techniques for determining hydraulic
conductivity, can yield a wide range of values for the same S/B
backfill sample. This variability stems primarily from the lack
of standardization of physical testing parameters such as state
of stress, degree of saturation, gradient, etc., to which the
sample is highly sensitive (Schulze, 1984).
2-8
-------
The accuracy evaluations made herein are therefore based on
"absolute" values determined via triaxial procedures in which the
physical testing parameters were standardized to model in-situ
conditions at the top of the cutoff wall (upper 5 to 10 feet).
As such, these "absolute" values of hydraulic conductivity would
be conservative (high) relative to those expected for the same
sample tested under conditions simulating greater depths or
incomplete saturation. The philosophy of modeling the worst case
condition of "top of wall" recognizes the indeterminancy of the
final vertical location of a specific batch of backfill
represented by a QC sample taken prior to backfill placement.
During construction of the Gilson Road cutoff wall, ten
samples of the backfill were obtained specifically for triaxial
testing. Each sample was remixed to improve uniformity and
consistency and then split to permit field testing in the API
fixed-ring apparatus as part of the QC program. In addition,
five samples of backfill were obtained from the completed wall
via conventional stationary-piston, tube sampling techniques for
both API fixed-ring and triaxial hydraulic conductivity testing.
Each of these samples was remixed prior to triaxial testing and
then a split of the triaxial sample was used for API fixed-ring
testing. Therefore, to the degree that the fifteen samples taken
were homogeneous, the value of hydraulic conductivity was
determined by both triaxial and API fixed-ring procedures on
"identical" specimens. These data are presented on Figure 2-2
and are summarized in Table 2-2.
2-9
-------
TABLE 2-2
HYDRAULIC CONDUCTIVITY TESTING:
API FIXED-RING VS TRIAXIAL (REMIXED)
Station API x 10~7 cm/sec Triaxial x 10~7 cm/sec
0+00
1+00
1+00
3+45
3+45
4+75
9+00
11+00
12+50
16+50
20+50
24+50
28+50
32+50
36+50
0.93
0.50
0.80
1.0
0.80
0.38
2.1
0.70
0.50
0.40
0.57
0.76
1.3
0.16
0.30
1.1
1.0
0.80
4.0
1.5
0.34
3.0
0.30
0.34
0.42
0.56
0.75
0.90
0.48
0.50
The data indicate that the API fixed-ring values match the
triaxial values to within +66 percent in fourteen out of the
fifteen cases, with one case of the API fixed ring value 133
percent higher than the triaxial. This error band not only
encompasses the accuracy of the API fixed-ring test but also
incorporates errors associated with heterogeneity between
"identical" samples and precisional errors inherent in both the
2-10
-------
TT-Z
API HYDRAULIC CONDUCTIVITY x I0-7cm./sec.
ro
i
ro
c
I-
o
o
o
o
c
o
H
>
-o
2
>
X
m
2
X
m
o
-------
API fixed-ring and triaxial procedures. The error associated
with heterogenity between "identical" samples could not be
quantified. Precisional error for API fixed-ring testing was
found to be +57 percent. Precisional error for triaxial testing
was not quantified, but experience suggests an error of
+50 percent. As such, the data indicate a very good level of
agreement between the simple field API fixed-ring test and the
more sophisticated laboratory triaxial test.
It should be noted that the applied pressure in thirteen of
the fifteen API fixed-ring tests described above was 10 psi. The
other two tests were run at 2 psi. Ten psi is considered an
upper limit for applied pressure. A higher applied pressure
would result in consolidation of the sample to levels not
commensurate with "top of wall" conditions. As such,
artificially low (unconservative) values of hydraulic
conductivity could be obtained. A higher applied pressure could
also lead to piping and thus unrealistically high values of
hydraulic conductivity.
It was further found that API fixed-ring hydraulic
conductivity did not decrease significantly as applied pressure
was increased from 2 psi to 10 psi. This trend was observed in
three of the samples cited above which were initially subjected
to 2 psi and subsequently increased to 3 psi, 5 psi, 7 psi, and
10 psi as well as in four other samples in which gradient was
varied (refer to Table 2-3 for summary of results). The increase
in applied pressure over this range appeared to have an
insignificant effect on consolidation of the granular Gilson Road
backfill (this conclusion might not be true for a more clayey
backfill in which sample consolidation would be a larger factor).
Test results indicated that increasing the pressure by
400 percent (from 2 to 10 psi) caused a decrease in hydraulic
conductivity of less than 50 percent.
2-12
-------
TABLE 2-3
EFFECT OF VARYING GRADIENT IN API FIXED-RING TEST
Station
Depth
Hydraulic Conductivity
x 10-7 cm/Sec
1+00
1+00
3+45
3+45
3 + 45
3 + 45
11+00
20'-22'
41'-43'
3.5'-5.5'
ll'-13'
22'-24'
32--341
9--10.51
28
0.45
2.1
1.5
0.88
0.55
1.5
3.9
41
0.58
1.4
1.4
2.0
0.78
1.0
3.1
Gradient
69
0.64
0.82
1.2
1.1
0.73
0.9
2.8
97
0.56
0.67
1.1
0.99
0.27
0.8
2.3
138
0.55
0.52
0.8
0.97
—
0.8
2.3
2.1.3 Guidelines for Field Use of the API Fixed-Ring Test
Based on the data presented herein, and the field data from
actual cutoff wall projects undertaken to date, the API
fixed-ring test appears to yield reliable and accurate values of
hydraulic conductivity. The reliability and accuracy combined
with the relative simplicity, short testing time, and low cost of
the test make it ideal for field quality control. The test can
therefore be used as the primary field measure (production test)
of backfill hydraulic conductivity provided accuracy is verified
for each new project.
Verification testing and production testing of backfill are
described in the following subsections. It will be seen that the
key to verification testing is obtaining an API fixed-ring sample
as representative of the corresponding triaxial sample as
possible, while the key to production testing is obtaining a
sample as representative of the in-place wall backfill as
possible. This difference reflects the different objectives of
2-13
-------
the two series of tests: verification testing validates test
accuracy; production testing validates backfill hydraulic
conductivity.
2.1.3.1 Verification Testing
Verification testing should begin during the design phase of
a project and continue during construction. During the design
phase, a limited number of triaxial tests should be run to verify
the site specific accuracy of the API fixed ring procedure. This
check between API fixed-ring and triaxial data should be made
after backfill component selection and proportioning are
established. As part of the verification program, the driving
pressure used for API fixed-ring testing should be varied to
determine which value yields the best API/triaxial match.
Samples selected for testing should represent the contract design
mix (or mixes). Backfill samples from which aliquots will be
taken for API/triaxial comparisons should be throughly mixed to
be as homogeneous as possible to yield aliquots as identical as
possible. Splits of a poorly mixed sample may reveal significant
variations in gradation which will lead to different hydraulic
conductivities. The variation in hydraulic conductivity would
then be misinterpreted as testing inaccuracy.
This design testing should then be followed by triaxial
testing of actual backfill samples obtained during construction.
Samples selected for testing should cover the complete range of
hydraulic conductivities and backfill types encountered on the
project as well as represent the entire wall perimeter. To
enhance the usefulness of triaxial data obtained during
construction, samples should also be tested as soon as they are
taken. Hence, the site specific validity of the API fixed ring
data can be established well before the end of construction. The
field samples from which aliquots are taken for the API/triaxial
comparisons must also be as homogeneous as possible. Therefore,
additional mixing far beyond that actually attained in the field
is mandatory prior to testing.
2.1.3.2 Production Testing
Although a high degree of additional backfill mixing is
required for the API/triaxial comparison samples, mixing of field
samples taken for QC production testing in excess of that
achieved during construction may yield unconse rvat i vely low
values of hydraulic conductivity. Hydraulic conductivity is a
function not only of the proportions in which backfill components
2-14
-------
are mixed, but also of the degree to which they are mixed.
Therefore, additional mixing of a sample obtained for QC testing
may improve the degree of blending to the point where the sample
is no longer representative of the field mix and yields a
falsely low value of hydraulic conductivity
Triaxial data from five undisturbed tube samples obtained
from the completed wall indicate that up to an order of magnitude
reduction in hydraulic conductivity can occur after the
undisturbed sample is mixed thoroughly and retested. Results are
illustrated on Figure 2-3 and summarized in Table 2-4. In each
of the five cases, the remixed sample yielded a lower hydraulic
conductivity than the undisturbed samples. Therefore, in some
cases, the backfill may have been mixed uniformly in the field to
a scale similar to that of sample size but not to the grain size
scale which controls hydraulic conductivity.
TABLE 2-4
TRIAXIAL HYDRAULIC CONDUCTIVITY:
"UNDISTURBED" VS "REMIXED" SAMPLES
Station
Depth
Hydraulic Conductivity
x 10-7 cm/sec
Undisturbed Remixed
1+00
1+00
3+45
3 + 45
11+00
20-22
80-82
11-13
22-24
22-24
2.
10.
12.
3.
5.
1
0
0
6
0
1.0
0.8
4.0
1.5
0.3
To obtain the most representative values of hydraulic
conductivity, samples for QC production testing should be taken
and placed in the test chamber with as little additional mixing
as possible while still removing air voids. Failure to observe
this procedural "detail" can result in artificially low
(unconservative) values of hydraulic conductivity. It also
2-15
-------
N)
I
FIG. 2-3 TRIAXIAL HYDRAULIC CONDUCTIVITY RESULTS
"UNDISTURBED" vs. "REMIXED" SAMPLES
i.o 10.0
TRIAXIAL HYDRAULIC CONDUCTIVITY x I0"7cm./sec.
(REMIXED SAMPLE)
-------
follows directly from this discussion that both the API
fixed-ring and triaxial data sets obtained for verification
testing do not qualify as QC production data. As indicated, the
sole purpose of these tests is to demonstrate the site specific
accuracy of the API fixed-ring procedure.
The driving pressure used during API fixed-ring QC testing
should be that value which, based on the verification testing,
yields a hydraulic conductivity closest to that derived from
triaxial testing. It is emphasized that, for the reasons
discussed previously, significantly higher driving pressures
should not be used in an effort to increase the flow per unit
time so as to make effluent volume measurements easier. Rather,
smaller graduated cylinders should be employed to increase the
reading accuracy for the small effluent volumes resulting from
particularly low hydraulic conductivity backfills.
2.2 Evaluation of Backfill Bentonite Content Using Methylene
Blue Titration
The cutoff wall at the Gilson Road site was constructed
using a backfill consisting of on-site soils, off-site borrow,
and bentonite. The on-site soils were highly permeable sands and
gravelly sands excavated from the trench (trench spoil).
Off-site borrow consisted of a quartz based fine sandy silt and
was mixed with the trench spoil to reduce its hydraulic
conductivity. Additional reduction of the mix hydraulic
conductivity to design values «1 x 10~7 cm/sec) was accomplished
via bentonite addition, first as a dry powder and then as a
bentdnite/water slurry.
For QC of backfill proportioning and mixing in the field,
testing of percent bentonite is second in importance only to
measurement of the actual hydraulic conductivity. For
predominantly granular backfills in particular, bentonite is the
singularly most significant component responsible for reduction
in hydraulic conductivity of the mix. Up to a critical level
(previously determined by laboratory testing) the higher the
bentonite content in the backfill, the lower and thus more
desirable the overall hydraulic conductivity. For this reason,
project specifications for S/B cutoff wall construction typically
include a minimum requirement for percent bentonite (5% for the
Gilson Road wall). In recognition of the above, the magnitude of
percent bentonite in a backfill sample is the first indicator to
be evaluated in the event that the API fixed-ring hydraulic
conductivity testing yields unacceptable values.
2-17
-------
In the past, field measurements of bulk quantities of
backfill soils and bentonite have been used to determine
bentonite percent by applying basic weight/volume relationships.
This method is approximate due to the limited accuracy of the
bulk weight/volume estimates, and optimistically assumes the
bentonite is evenly mixed throughout the backfill. Thus, bulk
measurements can be used as a gross check on the percentage of
bentonite in the backfill, but are inadequate for quantitative
field control.
As an alternative, the use of conventional hydrometer test
procedures have been attempted for bentonite determination. This
procedure is routinely used to determine the amount of silt and
clay size particles in samples of natural soils. Inherent
limitations in the test, however, restrict its use for backfill
QC testing. Lambe (1951) reports that the hydrometer test can
measure colloidal factors to an accuracy of only +3 percent, an
unacceptable error band when compared to the specified bentonite
content of 5 percent. Moreover, the test cannot differentiate
bentonite from other naturally occurring clay-size particles,
necessitating additional hydrometer tests on the other components
of the mix. Equally restrictive is the three to four days
required to complete each group of tests. This time period is
excessive relative to the rate at which backfill is mixed and
placed. Hence, the test is considered unsuitable for real time
quality control of the construction process.
A laboratory testing program was therefore carried out to
investigate the applicability of the quicker and potentially more
accurate methylene blue test for determination of percent
bentonite in S/B backfill. A primary objective of this work was
to verify the accuracy and precision of the methylene blue test
procedures previously adopted for evaluation of bentonite content
on the Gilson Road project. As such, primarily granular mixes in
which the bentonite additive is the predominant clay component
were used during the testing to simulate the backfill used at
Gilson Road. In addition, evaluations of the test's
applicability for general use in field control of other sites
employing S/B cutoff walls were made. The test and testing
program are described in the following sections.
2.2.1 Methylene Blue Titration
Methylene blue titration is a laboratory technique commonly
used to determine bentonite content in bentonite/water slurries
(Alther, undated). Primary applications in the past have been in
the oil drilling industry for testing drilling mud. The
procedure is based on the proportionality between clay content
2-18
-------
and ion exchange capacity. In a suitable aqueous environment,
such as an electrolytic clay- water suspension, clay materials
are associated with naturally occurring adsorbed cations of a
fixed total charge (typically sodium, calcium, magnesium, or
potassium). These cations are either retained by the clay or
exchanged for other cations subsequently introduced into solution.
The degree to which this exchange may occur, referred to as
cation exchange capacity, increases with the proportion and
activity of the clay in suspension.
The determination of bentonite content via the methylene
blue test therefore relies on the proportionality between the
amount of methylene blue cations exchanged and the amount of
bentonite in the backfill sample tested. The introduction of a
methylene blue solution into a bentonite-water suspension will
cause such an exchange. Bentonite content can then be determined
by quantifying the amount of cations exchanged. In a similar
sense, bentonite content in a suspension of S/B backfill and
water can also be determined using the proportionality with
exchanged cations.
The quantity of methylene blue cations exchanged is
determined by titration. The titration procedure involves the
incremental addition of methylene blue solution to a suspension
consisting of the backfill sample and a predetermined volume of
distilled water and dispersent until the cation exchange capacity
of the suspension is exceeded (end point). Methylene blue
consumption at the end point is then used to determine the dry
weight of bentonite in the sample via a pre-established
correlation.
The correlation is based on a series of methylene blue tests
performed on laboratory prepared samples with known
concentrations of the individual backfill components (bentonite,
trench spoil, and borrow). The test results indicate the cation
exchange capacity of each component over the range of
concentrations anticipated in the backfill. The correlation
curve for the backfill is then established via the superposition
of these results. As bentonite has a high cation exchange
capacity relative to the granular trench spoil and borrow, cation
exchange capacity of the backfill is highly dependent on
bentonite content and only marginally affected by changes in
spoil and borrow content.
The methylene blue test is well suited for quality control
of S/B backfills due to the relatively short testing time
(approximately 30 minutes per test), low cost, and minimal
laboratory requirements. A detailed description of the
standardized test procedure, which directly reflects the results
of efforts performed for this project, is described in Appendix B.
2-19
-------
The procedure is based on general bentonite titration methods
(Alther, undated and Baroid, undated), but includes modifications
necessary for specific application to quality control of S/B
cutoff wall construction. These modifications affected primarily
sample preparation and computational techniques.
2.2.2 Testing Program
Evaluation of the accuracy and precision of the methylene
blue test for quality control of S/B backfill was separated into
four distinct phases. Phasing was implemented to progressively
assess the impact attributable to increasingly more complicated
soil/water matrices accompanying the bentonite. These matrices
encompassed the simplest case of distilled water without soil to
more realistic combinations approximating actual backfill
samples, including natural soils with non-bentonite related
cation exchange capacity and interstitial water contaminated with
hazardous wastes. The applicability of linear proportionality,
matrixing effects and superposition were thus investigated.
Testing accuracy was assessed through comparisons of measured
bentonite content with that actually added to the laboratory
prepared samples.
The four phases were:
0 Bentonite in a distilled water matrix;
0 Bentonite in an Ottawa sand and distilled water matrix;
0 Bentonite in a contaminated water matrix;
0 Bentonite in an off-site borrow, trench spoil and
distilled water matrix.
The rationale leading to the selection of the above four
matrices, as well as the results of the work, are presented in
the following subsections.
2.2.2.1 Distilled Water Matrix
Bentonite in a matrix of distilled water and dispersant was
chosen for the first phase of testing to assess the
proportionality and inherent accuracy/precision of the methylene
blue titration for bentonite content. This series of tests
allowed evaluation of the basic test procedure without the
2-20
-------
complicating effects that additional backfill components might
present.
Three to four aliquots each of ten slurries containing from
0.0 to 0.9 grams bentonite were prepared. These bentonite
amounts represented 0 to 9 percent bentonite by dry weight in
10-gram backfill samples. Standardization of a 10-gram sample
size reflects sampling, testing, and typical backfill gradation
considerations. Designed S/B backfill mixes for hazardous waste
containment generally require 2 to 5 percent bentonits and as
such fall well within the range of tests performed. The same
bentonite was used throughout the testing and conformed to the
American Petroleum Institute's API grade, minimum 91 barrel
yield.
For each of the 36 individual tests, the volume of methylene
blue solution added to reach the end point of titration
(completion of cation exchange) was plotted versus the known
percent bentonite referenced to a 10-gram sample (see Figure 2-4).
(Although the titration volume actually corresponds to the dry
weight of bentonite, the data are presented herein as an
equivalent bentonite percent in the sample. This format is used
in recognition of the ultimate objective of standardizing this
procedure for QC of S/B backfill mixes).
The trend portrayed by the data not only demonstrates
proportionality but also indicates a linear relationship between
methylene blue titration volume and percent bentonite. Linear
regression analysis performed on the data yields a line referred
to as the "bentonite calibration curve" which fits the data set
to a high degree as reflected by a coefficient of determination
equal to 0.999.
Assumption of linearity, as supported by the physical
processes involved in the titration as well as the data, allows
evaluation of the test's precision and accuracy. Inasmuch as the
calibration curve itself is determined by the data, accuracy and
precision of the calibration are synonomous and defined as the
degree to which an individual data point would be expected to
correspond to the relationship defined by the entire data set
(the regression line). Standard statistical methods were used to
establish 95 percent confidence limits about the mean response of
the data set (also shown on Figure 2-4). The resulting 0.2
percent offset of upper and lower bound error bands from the
calibration curve is therefore the primary measure of both
accuracy and precision. Hence, given a slurry sample containing
an unknown amount of bentonite, it would be expected that the
titration volume in conjunction with the previously established
calibration curve would yield the correct percent bentonite to
within +0.2 percent with a 95 percent degree of confidence.
2-21
-------
FIG. 2-4 METHYLENE BLUE TEST RESULTS: BENTONITE IN DISTILLED WATER
NJ
I
IV)
NJ
IW
QO
_j ao
*»
w 7O-
-3 ru
. i fin.
Ill OW
z
UJ
v 50
X
h-
!= AO-
^ *rW
3O-
20-
Ifi"
1-11 1
95% C
(3)-— i
\S^
(3) — —j*
^1
o • ,
DNFIDENCE
(4) •» «
•^
1
: INTERVA
• BENTC
CURVI
•a ^
.2).^. —~"
K
)NITE CALI
(3)—^I
K
BRATION
L ^ e
H
•^
! 1
.0.2ml.
S~
rf
1
r t
(2) 1
10.2%
» n
% BENTONITE IN DISTILLED WATER
-------
The bentonite/distilled water samples used to establish the
basic relationship between titration volume and percent bentonite
excluded all other possible backfill components. This "best
case" condition is therefore expected to yield the highest degree
of accuracy. The curve defined by this data is the baseline
calibration against which the results of subsequent tests on more
realistic backfill samples are compared and testing accuracy
assessed.
2.2.2.2 Ottawa Sand/Distilled Water Matrix
For the second phase of testing, known amounts of bentonite
were added to a matrix composed of Ottawa sand, distilled water,
and dispersant. These samples were one degree more complex than
those used for Phase I testing and also one step closer to a
typical backfill sample. Ottawa sand was selected in that it is
almost entirely quartz with no clay minerals present.
Furthermore, quartz is the most prevalent mineral in the Gilson
Road backfill due to the silica base of the on-site soils and
off-site borrow used as the predominant mix components
(determined via x-ray diffraction). As would be expected based
on cation exchange principles, quartz, and thus the Ottawa sand,
has no exchange capacity. The specific objective of this testing
phase was therefore to determine whether the presence of backfill
components having no cation exchange capacity would interfere
with the bentonite/methylene blue cation exchange. This
interference, referred to as "matrixing effect," would include
any physical, chemical, or electrical phenomena which alter the
surface activity of the bentonite clay mineral.
Three identical samples were individually prepared at each
of five bentonite contents: 0, 2, 4, 6, and 8 percent (by dry
weight) . Each sample contained appropriate dry weights of Ottawa
sand and bentonite to yield 10-gram "backfill" samples with the
specified bentonite percentage. These samples were added to the
standardized volume of distilled water/dispersant and slurried
for testing. For each of the 15 tests, the methylene blue
titration volume at end point was plotted versus the known
bentonite percentage (Figure 2-5). The previously determined
calibration curve and 95 percent confidence interval (from
Figure 2-4) are also presented with the data for comparison.
The data indicate that, within the +0.2 percent error band
established via the 95 percent confidence interval, the Ottawa
sand has no effect on the methylene blue/bentonite cation
exchange process. Therefore, matrixing problems are not probable
due to the presence of quartz-based backfill components with no
cation exchange capacity.
2-23
-------
FIG. 2-5 METHYLENE BLUE TEST RESULTS! BENTONITE IN OTTAWA SAND/DISTILLED WATER
ro
I
K)
BENTONITE CALIBRATION CURVE
95% CONFIDENCE INTERVAL
12345678
% BENTONITE IN OTTAWA SAND/DISTILLED WATER
-------
2.2.2.3 Contaminated Water Matrix
During methylene blue testing, backfill samples are, in all
cases, mixed with distilled water to create a suspension which is
then titrated. Hence, the water used during the testing is
always contaminant free. However, contaminants may be added to
the suspension via the backfill sample itself. Where the cutoff
wall is excavated through contaminated portions of the aquifer,
trench spoil incorporated into the backfill will contribute
contaminants carried in its interstitial fluid. Furthermore,
trench slurry is typically added to the backfill during mixing to
increase the slump. Water added in this way typically accounts
for over 50 percent of the backfill water content. While the
slurry may also pick up contaminants from the aquifer's
interstitial fluid via osmosis and/or entrainment, of greater
importance is the chemical make-up of the water used to produce
the slurry.
In light of the above, addition of contaminants to the test
suspension may be more the norm rather than the exception. This
was in fact the case for the Gilson Road site where the primary
aquifer contaminants consisted of volatile organic industrial
solvents, and the slurry makeup water exhibited significant
hardness. To assess the possible matrixing effects of these
contaminants on the accuracy of the methylene blue test, a worst
case scenario was utilized for sample preparation. In addition
to incorporation of contaminants via the backfill itself,
contaminated water was used to form the test suspension instead
of distilled water.
Contaminated water from three different sources on the
Gilson Road site were obtained for this phase of the work. The
first two, designated well water and stream water, were sources
actually used during construction for slurry makeup water. The
third and most contaminated source was leachate as taken from
monitoring installations on the site. Backfill samples were
prepared at 0, 2, 5, and 8 percent bentonite in a manner similar
to Phase I testing. Two aliquots were tested for each bentonite
content using the stream and well water (total of 16 samples).
Only one series was performed using leachate at the four
bentonite contents (4 samples).
Test results, presented in Figure 2-6, indicate that the
contaminants tested had no significant effect on the test results.
All but one of the individual data points fall within +0.2
percent of the original bentonite calibration curve. As such,
the initial calibration accuracy of +0.2 percent with a 95
percent confidence level was maintained in spite of the Gilson
Road contaminants introduced into the matrix tested.
2-25
-------
FIG. 2-6 METHYLENE BLUE TEST RESULTS! BENTONITE IN CONTAMINATED WATER
I
ro
100
90
80
U 70
GO
UJ
z
UJ
Ul
60
30
20
10
(2)
• WELL WATER
* STREAM WATER
• LEACHATE
-(2)
-(2
(2)
(2)-
BENTONITE CALIBRATION-
CURVE
95% CONFIDENCE INTERVAL
234567
% BENTONITE IN CONTAMINATED WATER
8
-------
For volatile organic contaminants, noninterference would be
expected inasmuch as they are non-ionic and therefore should not
exhibit a high cation exchange capacity. In addition, the
backfill sample is dried in an oven at 105°C and the
backfill/water/dispersant matrix is boiled prior to titration.
The temperature involved in the above procedural steps would be
expected to remove the volatiles.
For other sites, particularly those containing highly ionic
contaminants, the cation exchange potential of the species
present should be compared to that of methylene blue with respect
to affinity for the bentonite clay mineral. In cases where high
non-volatile solvent concentrations exist, reduced solvent
(water) polarity effects, as they relate to methylene blue
solubility, should also be investigated. This can be done on a
theoretical basis or by utilizing the simple testing procedure
described herein. It should be recognized however that this
testing procedure can yield "false positives" in that the
quantity of contaminants introduced into the matrix far exceeds
that which is likely to be added via an actual backfill sample.
2.2,2.4 Gilson Road Backfill Matrix
The fourth phase of testing employed prepared samples
duplicating the backfill used for the Gilson Road project. As
previously discussed, the Gilson Road backfill consisted of
approximately equal proportions of trench spoil (sand and gravel)
and off-site borrow (sandy silt) with 3 to 5 percent bentonite
added. The objectives of this series of tests were to 1)
evaluate the impact of a cation exchange capacity attributable to
non-bentonite backfill components on the determination of percent
bentonite in the backfill; 2) determine if the bentonite
calibration could be corrected to account for non-bentonite
cation exchange capacity utilizing superposition; and 3) based on
the above, evaluate the accuracy of the methylene blue test
results previously obtained for the Gilson Road project.
The on-site soils (trench spoil component of the backfill)
used for testing were obtained during a post-construction boring
program. Samples were selected from various locations, both with
respect to wall perimeter and aquifer depth. They exhibited
gradations ranging from fine sand and silt to fine to coarse sand
and gravel and represent the full variation of soils encountered
during cutoff construction. The ten samples were mixed with
distilled water and dispersant in a manner similar to Phase II
testing except that no bentonite was added. The on-site soils
exhibited average cation exchange capacities equivalent to 0.2
2-27
-------
percent bentonite with a range of 0.1 to 0.4 percent. Results
are summarized in Table 2-5.
TABLE 2-5
METHYLENE BLUE TEST RESULTS
ON-SITE (TRENCH SPOIL)
Station
2+50
9+30
9+30
13+00
13+00
34+70
34+70
25+30
25+30
38 + 50
Offset
20'S
50'N
50 "N
0
0
60'NW
60'NW
20 'NE
20 'NE
40'E
Depth
Feet
0-2
33-37
11-13
24-26
44-46
54-55.5
79-80.5
29-31
69-71
24-26
Description
fine sand and silt
fine sand
fine to coarse sand
fine to medium sand
fine to coarse sand
glacial till
glacial till
fine to coarse sand
and gravel
fine to medium sand
fine to coarse sand
sand gravel
Methylene
Blue, Ml
4
3
1
2
4
1
1
2
2
3
Nine samples of the off-site borrow were also selected from
record samples taken during construction. Again the objective of
sample selection was to encompass the full variability of the
materials. Eight of these samples yielded bentonite equivalents
of 0.6 to 0.7 percent with one sample yielding 0.4 percent. Test
results are summarized in Table 2-6.
2-28
-------
TABLE 2-6
METHYLENE BLUE TEST RESULTS
OFF-SITE BORROW
Station From Which
Borrow Was Obtained
Millileters of Methylene
Blue Required to Complete
Titration
4+45
9+00
12 + 50
18+00
25+50
28+00
32+00
35+00
38 + 50
7
7
6
7
4
6
7
6
6
This preliminary testing for the exchange capacity of
individual non-bentonite backfill components provided a basis for
prediction of the bias expected relative to the bentonite
calibration previously developed. An anticipated ' offset
equivalent to 0.4 percent bentonite was derived using a weighted
average of the individual exchange capacities based on the
relative amounts of borrow and trench spoil in the backfill.
Fifteen splits of backfill were then prepared containing
equal proportions of the on-site soil and the off-site borrow
(ratio used during construction of the Gilson Road cutoff).
Bentonite was added to the resulting mixture (a silty sand and
gravel) to obtain three samples at each of 5 bentonite contents:
1, 3, 5, 7, and 9 percent by dry weight. The samples were
titrated with methylene blue solution until endpoint; the data
are plotted on Figure 2-7.
2-29
-------
FIG. 2-7 METHYLENE BLUE TEST RESULTS!
BENTONITE IN BORROW/TRENCH SPOIL/DISTILLED WATER
i
U)
o
ID
_l
03
UJ
Z
X
h-
UJ
BENTONITE CALIBRATION CURVE
CORRECTED FOR BORROW
a SPOIL
BENTONITE
CALIBRATION
CURVE
= BENTONITE EQUIVALENT
IN BORROW a SPOIL
(0.4%)
I 2 3 45 6 7 8
% BENTONITE IN BORROW/TRENCH SPOIL/DISTILLED WATER
-------
Test results exhibited a linear trend, but, as expected,
also demonstrated bias reflecting greater exchange capacity than
attributable to the bentonite alone. Superposition of the
previously predicted 0.4 percent non-bentonite derived off-set on
the original calibration curve yielded a "corrected calibration
curve" presented as a dashed line on Figure 2-7. The original
bentonite calibration curve developed during the first phase of
testing is also presented for comparison.
Although this corrected -calibration does not exactly match
the best fit line through the actual data, it does agree to +0.3
percent bentonite. These data therefore support the validity of
superposition in that the small discrepancy obtained corresponds
to inaccuracies incurred by averaging the data upon which the
corrected calibration curve was based. These results indicate
that the procedure of superimposing average exchange capacities
of non-bentonite backfill components on the bentonite calibration
curve, as was used on the Gilson Road project, yields a new
corrected calibration curve with an expected accuracy of +0.3
percent bentonite. This error band represents less than
+6 percent of the specified bentonite content (5 percent of
backfill, by dry weight) and verifies that the data amassed
during the cutoff wall construction at Gilson Road were in fact
sufficiently accurate for field QC use.
It should be emphasized at this point that unless the
bentonite calibration curve is corrected for the cation exchange
capacity attributable to non-bentonite backfill components, the
bentonite contents measured during QC testing will be
unconservatively high. In addition, the accuracy of this
superposition-based correction is directly dependent on the
uniformity of the cation exchange capacities of the individual
components and the proportioning of these components in the
backfill mix. Therefore, if either of these parameters are
varied during construction of the cutoff, then separate
calibration curve corrections representing individual sections of
the wall would be required for QC testing.
2.2.3 Field Use of the Methylene Blue Test
The laboratory program described herein was used to assess
the accuracy of the methylene blue test and was therefore
structured to measure the unknown quantity of methylene blue
required to titrate backfill samples with known bentonite
contents. The data thus developed demonstrated the accuracy of
the test for field QC in cases similar to the Gilson Road project.
During actual use for QC in the field, however, the methylene
blue consumption would be measured and used to determine the
2-31
-------
unknown bentonite content in the backfill via pre-established
calibrations. Thus, the testing methodology used in field
applications initially would follow a logic similar to the
laboratory program for establishment of a calibration curve, but
thereafter utilizes methylene blue consumption as the independent
variable and bentonite content as the dependent variable.
Implementation of the methylene blue test for quality
control therefore consists of the following procedure:
0 Establish a bentonite calibration curve representing the
relationship between methylene blue consumption and
bentonite content for ben tonite/distilied water
suspensions. The same bentonite to be used during
construction must be used during this step. The
procedure would be similar to the Phase I testing
described herein except that the methylene blue
consumption would be plotted on the x-axis as the
independent variable.
0 Determine whether contaminants are likely to be present
in the backfill samples and evaluate the probable effect
on the calibration.
0 Determine average cation exchange capacity (methylene
blue consumption) of the non-bentonite backfill soils
combined in the same proportions they will be mixed in
the field. Use the bentonite calibration curve to
determine the bentonite equivalent for this volume.
0 Offset the initially determined bentonite calibration
curve to account for the bentonite equivalent of the
non-bentonite soils and/or contaminants to establish a
field calibration curve. Note that multiple field
calibration curves may be required if significant
variation in the offset values are anticipated due to
variations in the site geology or construction
procedures.
0 Perform the methylene blue test on backfill samples and
enter the field calibration curve with the volume of
methylene blue consumed at the end point to
determine bentonite content.
It should be pointed out that although the test is quick,
inexpensive, and accurate, an initial level of practice is
required to accurately assess the end point of titration.
Therefore, only technicians having experience with the titration
procedure should be allowed to perform the tests under the rigors
imposed by field quality control of fast track construction.
2-32
-------
Finally, the procedures presented herein are not directly
applicable to predominantly clay-based backfills. For this class
of cutoff, the high proportion of non-bentonite clay components
and the variability of their total percentage in the backfill can
make correction of the bentonite calibration curve "sample
specific." Therefore, at the very least, independent knowledge
of the amount of non-bentonite clay in the sample would be
required. A procedure has been developed for this case allowing
determination of bentonite content through solution of
simultaneous equations. This work will be presented in a future
report once the precision and accuracy of the procedure have been
established.
2.3 Effect of Backfill Gradation on Hydraulic Conductivity
As discussed in the previous sections, the addition of
relatively small percentages of bentonite to a granular soil will
reduce its hydraulic conductivity several orders of magnitude.
This effect is widely recognized and has been clearly
demonstrated in laboratory tests (Schulze, 1984) and field
applications. It is also recognized, though less well
understood, that the overall gradation of a backfill sample also
effects its hydraulic conductivity. Inherently, the more
"well-graded" the sample, the denser the particle packing and
hence, the lower the hydraulic conductivity. It is not clear,
however, what effect various gradation parameters have on
hydraulic conductivity.
As part of this project, therefore, backfill data collected
both during and after construction were statistically analyzed to
determine to what extent gradation characteristics of a backfill
sample could be used to predict hydraulic conductivity. The
ultimate objective of this effort was to identify which aspects
of backfill gradation play the greatest role in determining
backfill hydraulic conductivity. This information would then
serve to enhance quality control procedures by providing
additional insight into a backfill mix which API fixed-ring
testing indicates has an unacceptably high hydraulic conductivity.
With such a correlation, gradation analyses of backfill could
become a more integral part of the quality control program. At
the present, gradation parameters are typically used only to
control the percentage of fines within a minimum and maximum
value. No further criteria to control, much less optimize,
backfill gradation are included in specifications because the
influence of varying gradation characteristics is not
sufficiently understood.
2-33
-------
The method used to evaluate the correlation consisted of the
following:
0 Establish a data base of gradation parameters for a
statistically meaningful number of backfill samples.
For each of these samples, determine hydraulic
conductivity.
0 Perform linear regression analyses on the individual
parameters to establish the degree of dependency of
hydraulic conductivity on each.
0 Evaluate the interaction of the individual parameters
using multiple polynomial regression analyses.
0 Establish confidence envelopes for the relation between
hydraulic conductivity and the soil parameters.
This procedure and the results of the analyses are described
in the following sections.
2.3.1 Data Base
A body of gradation parameters and hydraulic conductivity
data was collected for eighty-six backfill samples obtained from
the Gilson Road site both during and after construction. The
primary backfill parameters for use in the correlation were
percent bentonite, percent clay sizes, and percent silt sizes
since hydraulic conductivity is considered most dependent upon
percentage and nature of fines (material passing no. 200 sieve).
Additional gradation parameters were included to describe the
overall distribution of particle sizes and the grain size curve.
Several density parameters were also included since hydraulic
conductivity is seen as a secondary function of backfill density.
The backfill gradation and density parameters used in the
analyses are listed below:
0 Percentage of material passing the No. 4 sieve (by dry
weight).
0 Percentage of material passing the No. 200 sieve,
i.e., silt and clay sizes (by dry weight).
0 Percentage of material finer than 2 microns, i.e., clay
sizes (by dry weight).
2-34
-------
0 Percentage of bentonite (by dry weight; determined using
methylene blue titration).
0 Grain size at which 60 percent of the sample, by dry
weight, is finer (D60).
0 Uniformity coefficient (D60/D10).
0 Coefficient of curvature (D30)2/(D/10) (D60).
0 Unit Weight.
0 Water content (wt. water/wt. dry soil).
0 Void ratio (volume of voids/volume of soil).
Values of hydraulic conductivity used in the analysis were
determined by API fixed-ring testing.
2.3.2 Statistical Evaluation
As the second step in the development of a correlation, the
dependence of hydraulic conductivity on each backfill parameter
was evaluated. Using standard statistical methods, regression
lines defining the mean response of each gradation/density
parameter and hydraulic conductivity were determined. For each
set of data, the correlation coefficient describing the scatter
of the data about the regression line was also calculated as an
index of the degree of dependency. With a correlation
coefficient of 1.0 (plus or minus) indicative of perfect
correlation (i.e., all data points located on regression line), a
correlation coefficient of 0.75 was initially chosen to
distinguish those backfill parameters having "primary" influence
from those having "secondary" influence, for later use in
performing polynomial regression analyses.
Correlation coefficients were first calculated based on an
arithmetic scale plotting of hydraulic conductivity versus each
soil parameter. Results are tabulated in Table 2-7. None of the
soil parameters produced a correlation coefficient greater than
0.75; the highest coefficient was 0.42. The correlation
coefficient between percent bentonite and hydraulic conductivity
was 0.36; the correlation coefficient for percent clay sizes was
also 0.36. For total percent fines, the coefficient was 0.27.
Thus, the parameters thought to be most important to hydraulic
conductivity did not emerge as such in the statistical
evaluation.
2-35
-------
TABLE 2-7
CORRELATION COEFFICIENTS
HYDRAULIC CONDUCTIVITY VS. GRADATION PARAMETERS
Y = CON-API
X vs. Y X vs. LOG(Y) LOG(X) vs. Y LOG(X) vs. LOG(Y)
X = %-No.4
X = %-Nb.200
X = %-2u
X = DIG
X = D30
X = D60
X = Cu
X = Cz
X = GAMMA T
X = W
X = e
X = % Bent
X = CON-API
0.048
-0.274
-0.355
0.187
0.422
0.141
-0.314
0.047
0.311
-0.347
-0.363
-0.359
1.000
0.095
-0.240
-0.420
0.231
0.388
0.089
-0.393
-0.004
0.352
-0.380
-0.399
-0.440
0.894
0.049
-0.309
-0.362
0.376
0.387
0.130
-0.300
0.094
0.310
-0.351
-0.368
-0.400
0.894
0.097
-0.269
-0.418
0.438
0.382
0.091
-0.362
0.074
0.353
-0.376
-0.395
-0.471
1.000
NOTES:
CON-API = API fixed-ring hydraulic conductivity.
%-No.4 = Percentage of material passing the No. 4 sieve (by dry
weight); similar for "%-No.200" and "%-2u".
D10 = Grain size at which 10 percent of the sample (by dry weight) is
finer; similar for "D30" and "D60".
Cu = D6o/Di0
C2 =
GAMMA T = Unit weight.
W = Water Content.
e = Void ratio.
% Bent = Percent bentonite determined using methylene blue test.
Correlation coefficients were then calculated based on
combined arithmetic-logarithmic ("semi-log") scale plottings and
on a logarithmic ("log-log") scale plotting to seek a
correlation. None of these relationships exhibited a
coefficient in excess of 0.5. Results are also
Table 2-7.
better
correlation
summarized in
As none of the individual parameters was statistically found
to be a primary determinant of hydraulic conductivity, additional
2-36
-------
statistical evaluation to determine interaction of individual
parameters and then confidence envelopes was not pursued. The
lack of a predominant dependency of hydraulic conductivity on
percent bentonite, percent clay sizes or percent fines and lesser
dependency on other parameters indicated that the overall
correlation between hydraulic conductivity and soil gradation
parameters may be extremely complex and might not be fully
reflected by the available data set. This may be due to the
variable degree of blending of backfill components in the samples
tested. Increased blending of field samples was found to
decrease hydraulic conductivity (see Section 2.1.3.2), but it is
not known to what degree the field samples were homogeneous on a
grain-size scale. In addition, the correlation is affected by
the previously observed variations in hydraulic conductivity
measurements of identical backfill samples. The worst case
precision of the test was found to be +57 percent within one
standard deviation of the mean value; this degree of variation
would partially mask the existence of a close correlation.
The poor statistical correlation does not, however, alter or
invalidate the premise that hydraulic conductivity is largely
determined by percent bentonite or percent clay sizes.
Experience has demonstrated the key roles played by these factors.
Therefore, it is simply concluded that the underlying correlation
between the key soil parameters and hydraulic conductivity may
have been obscured by the inherent precision of the API
fixed-ring test.
2.3.3 Importance of Statistical Correlation
The establishment of a statistical correlation between
gradation parameters and hydraulic conductivity was investigated
primarily to gain further insight into the relative importance of
the various parameters on backfill hydraulic conductivity.
Better correlations would have rendered gradation analyses a more
integral component of field quality control testing than at the
present. As a supplement to methylene blue testing, gradation
analyses might have provided several key indicators of potential
deficiencies in a backfill sample having an unacceptably high
hydraulic conductivity.
In the absence of such a correlation, however, quality
control capabilities are not seriously impaired. API fixed-ring
and methylene blue testing remain the primary and secondary
quality control tests. Both tests are more direct indicators of
backfill quality than gradation analyses and both are quicker to
run.
2-37
-------
The greater value in establishing a statistical correlation
between quality control and hydraulic conductivity is for
backfill design. If the effect of various gradation parameters
on hydraulic conductivity could b^e quantified, preliminary
backfill design could consist of assessing gradation
characteristics of on-site soils and determining via the
correlation the various changes in gradation required to achieve
desired hydraulic conductivity.
2.4 Assessment of Gilson Road QC Data
The use of the API fixed-ring and methylene blue tests
during construction of the Gilson Road cutoff wall reflected the
current state of the art in field quality control procedures.
Although the precision and accuracy of these tests were not
assessed at the time of construction, the tests were used because
of their respective advantages over alternative test procedures.
Moreover, the data generated during construction ultimately
played a vital role in the construction control program. Tests
were run on a daily basis to assess the success of the contractor
in attaining design standards; deficiencies thus identified were
brought to the contractor's attention and immediate corrective
measures were typically taken to remedy the problem. It should
be emphasized, however, that evaluation of the overall quality of
the final product at the end of construction was based to a large
degree on test procedures for which reliability had not yet been
determined. As a result, in spite of the state of the art field
control program, the degree to which the quality control data
represented the quality of wall construction, and therefore, the
degree to which the completed wall satisfied project
specifications, was not known.
The present effort intended to achieve two major objectives:
1) assess the reliability of QC procedures used at Gilson Road;
and 2) evaluate the data from wall construction in the light of
this new information. In previous sections, both the API
fixed-ring test and methylene blue test were found to be
sufficiently accurate and repeatable to validate their use at
Gilson Road. In this section, the QC data generated during and
after construction are reviewed to assess the degree to which
project specifications were satisfied. Where specifications were
not satisfied, the impact on cutoff wall performance has been
evaluated. Further, the rationale for accepting backfill which
does not meet specifications is discussed.
2-38
-------
2.4.1 Review of QC Data
Effectively two bodies of QC data were collected at Gilson
Road. The first set represented tests performed on "grab"
samples obtained during construction. The second set consisted
of tests performed on "undisturbed" tube samples obtained from
the completed wall one to three years after construction. The
first set of data indicated that the hydraulic conductivity
specification was generally satisfied once initial "Contractor
Learning Curve" problems with backfill proportioning and mixing
were overcome. The second set of data, however, indicated that
the in-situ hydraulic conductivity was potentially an order of
magnitude higher than that determined on the "grab" samples. The
results of the QC testing are described below and evaluated with
respect to specification requirements.
2.4.1.1 QC Data From Construction Samples
Seventy-five backfill samples were collected during wall
construction. The samples were collected immediately after
completion of the field mixing effort, just prior to placement of
the backfill in the trench. Typically, each sample consisted of
approximately 25 pounds of backfill to provide splits for
hydraulic conductivity testing, methylene blue testing, gradation
testing, and record samples. Prior to splitting, each sample
received additional mixing by hand to improve the homogeneity of
the mix and thus the similarity of splits.
API fixed-ring hydraulic conductivity tests were run on each
of the seventy-five samples. In addition, triaxial tests were
run on ten of the samples.
The results of the eighty-five tests ranged from
9.9 x 10~9 cm/sec to 4.3 x 10~7 cm/sec. The mean value was
5 x 10-8 qm/sec. Sixty-six tests (78 percent) satisfied project
specifications for hydraulic conductivity, yielding values less
than or equal to 1.0 x 10~7 cm/sec. Of the nineteen samples that
yielded hydraulic conductivities greater than the specified
maximum value, eleven were between 1.0 x 10-7 and
2.0 x 10~7 cm/sec, eight were between 2.0 x 10~7 cm/sec and
4.3 x 10-7 cm/sec.
A chronological plot of hydraulic conductivity versus
construction date (Figure 2.8a) reveals that six of the eight
highest values were recorded during the initial days of
construction (9/11/82 - 9/17/82). Throughout the remainder of
2-39
-------
FIG. 2-8 a. HYDRAULIC CONDUCTIVITY vs. TIME
lu.o
1.0
O.I
i
(J
•
® •
* •
4
(
»
f> ®
TRIAXIAL/
(TYR)
I
^MAXIMUM t
/ PER SPECI
/ 1.0 x 10 C
• <
i
•l.& "
'
I
ALLOWABLE
FICATION
m!?sec.
:
* 0
FIXED RING/
(TYR)
1
.
* / *
•
>
E
o
O
O
O
o
DC
Q
10 15 20
SEPTEMBER
25 30 5
(1982)
10 15 20
OCTOBER
30
b. BENTONITE CONTENT vs. TIME
PERCENT BENTONITE
O f° * 0» OB C
1
A
» * • <
•
•
•
t
» • 4
«
•
• ^
' i *
i
i <
.. .[ -v
i
.
: i
r
L«.
» •
i
10 15 20 25 30 5 10 15 20 25 3
SEPTEMBER (1982) OCTOBER
2-40
-------
the project, values in excess of the specification were
encountered on five construction days spaced over the first three
weeks in October. A plot of hydraulic conductivity versus
station along the wall alignment (Figure 2-9a) indicates that the
higher values were found between stations 0+00 and 3+00, between
stations 6+00 and 12+00, and between stations 25+00 and 29+00.
The primary cause of the higher hydraulic conductivities
measured during construction appears to be insufficient bentonite
content. On a time plot illustrating percent bentonite versus
construction date (Figure 2-8b), the lowest percent bentonite
(1.3 percent) was recorded on 9/11/82 (start of construction).
The time plot further illustrates that percent bentonite was
gradually increased over the first few weeks of construction to
values as high as 7 and 8 percent bentonite before leveling off
at a range of 2 to 5 percent. The gradual increase in percent
bentonite at the beginning of construction reflects the
contractor's response to the initially low values. These low
values resulted from "Learning Curve Problems"! associated
with the inaccuracies of bentonite determination via bulk
measurements. The increase in percent bentonite from
approximately 9/11/85 to 9/24/85 is directly reflected in a
steady decrease in hydraulic conductivity over the same period
(refer to Figure 2-8a). Beyond 9/24/85, the occurrence of the
lowest values of percent bentonite occurred between 10/2 and 10/4
and between 10/12 and 10/18, coinciding with the occurrence (in
both time and location) of hydraulic conductivities in excess of
specifications.
A secondary cause of the high values of hydraulic
conductivity, particularly during early stages of construction,
may have been inadequate mixing effort. While bentonite contents
of 2 percent or lower were measured during three periods
throughout construction, high values of backfill hydraulic
conductivity were most frequently measured during the initial
days of construction. This situation is also reflective of a
learning period that construction personnel typically experience
during initial days of backfill preparation as they adjust to
site specific conditions.
cutoff wall construction is a specialty contracting job,
commonly executed by local equipment operators working under the
supervision of two or three experienced cutoff wall personnel, a
certain period of time is required for "operator education" at
the initial stage of construction. The "learning curve"
phenomenon can result in erratic quality for the first several
hundred feet of a wall. Therefore, during initial project
start-up, additional conservatism with respect to bentonite
content and mixing effort should be executed.
2-41
-------
PERCENT BENTONITE
HYDRAULIC CONDUCTIVITY x IO"7 cm./sec.
ro
I
ro
CO
o
•f
o
0
o
4.
o
o
to
H
HO
Oo
g
o
o
•
*
4
'
•• 4
• *
1 0
m a
t
•
•
^ • <
.;•*•
•
• •
*
•
\.
i
•
O
ro
i
to
c
n
o
o
o
o
d
5
o
-------
The results of the hydraulic conductivity and methylene blue
tests run on grab samples thus indicated that, in general, the
hydraulic conductivity specification was satisfied while the
bentonite specification was not. Although hydraulic conductivity
data indicated that approximately 22 percent of the samples
tested had a hydraulic conductivity in excess of the maximum
specified value of 1.0 x 10~7 cm/sec, the higher values were
judged to have little impact on wall performance. The impact of
lower than specified bentonite values was also limited as the
specified value of 5.0 percent bentonite incorporated a safety
factor of 1.67 over the value required based on design testing
(3.0). This factor of safety was adopted to account for the less
than "laboratory perfect" proportioning and mixing expected
during actual construction. Therefore, the mean value of 3.5
value reflected a technically acceptable level.
2.4.1.2 QC Data From Post-Construction Samples
Eleven backfill samples were obtained in 1983 and 1985 from
the completed wall. The samples were obtained from various
depths at three stations along the wall alignment, station 1+00,
station 3+45, and station 11+00. These locations were chosen on
the basis of the relatively high hydraulic conductivities
measured on backfill samples obtained from these areas during
construction (refer to Figure 2-9a). It was intended that the
post-construction samples would represent "worst case" backfill.
The samples were obtained using a stationary-piston tube sampler,
commonly used to obtain "undisturbed" samples of fine-grained
soils. Sample depths ranged from 2 to 82 feet. The tubes were
x-rayed prior to sample extrusion to determine the relative
gradation of the backfill samples (predominantly coarse versus
fine) as well as the location of gravel pieces within the tube.
The primary objective was to test a variety of backfill samples
while at the same time avoiding specimens with large pieces of
gravel. A large piece of gravel would result in an
unrepresentatively low hydraulic conductivity as it would block
off a large percentage of the cross-sectional area of the sample
and the hydraulic conductivity of the gravel itself is nearly
zero. The total area is used in the hydraulic conductivity
computation.
Each of the eleven samples was tested "undisturbed" in the
triaxial apparatus immediately upon extrusion from the sampling
tube. The tests were first run at low confining pressures
commensurate with top of wall conditions although eight of the
samples were taken from depths below 10 feet. "Top of wall"
stresses were selected to correlate with the stress range used
for API fixed-ring testing executed during construction.
2-43
-------
Quality control testing during construction was conducted using
"top of wall" stresses inasmuch as the final vertical location of
backfill in the wall cannot be predicted when the sample is
obtained prior to backfill placement (see Section 2.1.2.2).
It was recognized that the undisturbed samples taken from
the completed wall would have undergone consolidation
commensurate with depth in the wall. Therefore, the soil fabric
of these samples would correspond to the API fixed-ring samples
only to the extent that the sample swelled in response to stress
relief. The samples would thus be tested in an overconsolidated
state. The significance of overconsolidation is discussed below.
Seven of the eleven samples were subsequently tested at a
second, higher stress level, commensurate with the approximate
vertical effective stress at the depth from which the sample was
taken. This higher stress level yields hydraulic conductivities
representative of the actual performance of the wall. The lower
stress levels would be expected to yield higher and thus
conservative hydraulic conductivity values.
The eleven tests at low stresses yielded hydraulic
conductivities ranging from 6.6 x 10~8 to 1.2 x 10~6 cm/sec
(refer to Table 2-8). The mean value was 5.6 x 10~7 cm/sec.
Only two of the eleven tests satisfied project specifications for
hydraulic conductivity. However, seven of the tests yielded
hydraulic conductivities less than 5.5 x 10~7 cm/sec. Thus, as
hydraulic conductivities of the construction samples taken from
these areas ranged from 9.9 x 10~9 to 4.3 x 10~7 cm/sec, these
"undisturbed," triaxial results appear consistent with
construction results. The remaining four test results are
notably higher, exceeding the previous maximum by half an order
of magnitude and the specified limit by one order of magnitude.
2-44
-------
TABLE 2-8
HYDRAULIC CONDUCTIVITY: "UNDISTURBED" VS "REMIXED" SAMPLES
Station
Depth
(ft.)
Triaxial Hydraulic Conductivity,
K(x 10~7 citv/sec)
Undisturbed Undisturbed Remixed
low stress high stress low stress
API Fixed-Ring
Hydraulic Conductivity
(x 10~7 cm/sec)
1+00
1+00
1+00
1+00
1+00
3+45
3+45
3+45
3+45
11+00
11+00
9.5-9.9
20.6-21.1
31.0-31.4
42.5-42.9
80.6-81.0
4.5-5.0
12.5-12.9
22.4-22.8
33.0-33.5
9.2-9.7
22.4-22.9
0.66
2.1
1.5
0.85
10.0
12.0
12.0
3.6
5.5
10.0
5.0
5
5
5
5
3
5
5
5
5
5
3
0.47
1.3
3.8
7.8
2.5
1.7
2.5
10
20
1.0
37 0.8
10
20
40
4.0
1.5
5
5
13 0.26
0.28
0.78
1.0
0.82
0.8
1.2
1.1
0.78
0.90
2.8
0.7
d = effective confining pressure (psi).
When seven of the tests were further consolidated to
approximate in-situ stress levels, the result in each case was a
reduction in hydraulic conductivity. Within this subset of seven
tests, hydraulic conductivity ranged from 4.7 x 1 0 ~ ^ to
7.8 'x 10~' cm/sec at the higher stress levels, with a
of 2.9 x 1017 cm/sec. The range for the lower stress
cm/sec, with a mean
Results are summarized in Table 2-8.
6.6 x 10~8 to 1.2 x 10~6
5.6 xlO~7 cm/sec.
mean value
level was
value of
This data indicates that the Gilson Road QC results are
likely to be conservative relative to actual in-situ conditions
due to the lower than in-situ stresses used for QC testing. The
data indicates a reduction in hydraulic conductivity by a factor
of two. The actual reduction is probably even greater than this
since the low stress data was obtained from samples which had
previously been consolidated to higher in-situ stresses.
Although the stresses were lowered to "top of wall" levels prior
to testing, the soil fabric should still have reflected the
previously higher stresses (i.e., the sample was
overconsolidated ) . It would therefore yield hydraulic
conductivities lower than those associated with identical samples
which had not previously experienced the higher stress level.
2-45
-------
Thus, the change in hydraulic conductivity with stress is likely
to be greater than the above data indicates.
Five of the eleven "undisturbed" triaxial samples were then
remixed and retested in the triaxial apparatus to assess whether
the higher hydraulic conductivities measured in the "undisturbed"
samples were related to the lesser degree of mixing when compared
to the construction samples. Each sample was mixed thoroughly
and retested in a "disturbed" state, modeling the triaxial tests
performed during construction. The tests were run at low
confining stresses. Test results, summarized in Table 2-8,
indicated that the additional mixing lead to lower hydraulic
conductivities in all five cases and by approximately one order
of magnitude in three of the cases. Furthermore, the results on
the remixed samples were consistent with the results from the
samples obtained during construction, ranging from 2.6 x 10~8
cm/sec to 4.0 x 10~1 cm/sec.
API fixed-ring tests were subsequently performed on each of
the eleven samples after remixing. These tests also yielded
values consistent with those obtained during construction.
Results ranged from 2.8 x 10~8 cm/sec to 2.8 x 10"' cm/sec with a
mean value of 1.0 x 10~7 cm/sec (refer to Table 2-8). Once
again, remixing resulted in significantly lower hydraulic
conductivities as compared to the undisturbed triaxial tests.
Furthermore, remixed API fixed-ring tests yielded results very
compatible to remixed, low stress triaxial tests (within an
average factor of two). The additional mixing utilized to
prepare the API fixed-ring samples from the remixed triaxial
samples resulted in a further reduction in hydraulic
conductivity.
Therefore, the results from the triaxial tests and API
fixed-ring tests on the remixed post-construction samples
indicate that remixing beyond the level achieved in the field to
obtain duplicate sample splits may result in a lower hydraulic
conductivity than that which exists in the field. This reduction
in hydraulic conductivity for the Gilson Road site was nearly
one-half an order of magnitude as based on the five backfill
samples which were tested in the triaxial apparatus, first
"undisturbed" and then remixed. (The average hydraulic
conductivity of the "undisturbed" samples was 6.5 x 10~? cm/sec
and the remixed samples was 1.5 x 10~7 cm/sec.)
2-46
-------
2.4.1.3 Evaluation of the Validity of the QC
Construction Data
The primary purpose of the data obtained from
post-construction samples is to validate the use of QC
construction data for evaluating wall performance. Although the
API fixed-ring test has been established as a reliable and
accurate quality control procedure (Section 2.1), the validity of
the QC construction data at the Gilson Road site also depends
upon a viable comparison of test procedures and conditions with
actual field conditions (i.e., in-situ stresses, no remixing),
which are more closely reflected in the post-construction testing.
Based on the post-construction data, two quality control testing
procedural details have been identified which may result in the
hydraulic conductivity of the Gilson Road QC construction samples
being unrepresentative of that actually attained in the field.
The first such detail is the additional mixing of the
backfill after sampling to increase homogeneity prior to
splitting the sample for duplicate testing. The
post-construction data indicate that this may have resulted in
nearly a one-half order of magnitude, unconservative error in the
Gilson Road QC construction data. This error can be easily
avoided in future projects by eliminating the remixing step and
accepting some variability in the duplicate sample aliquots. The
second procedural step involves testing the samples at "top of
wall" stresses in recognition of the indeterminancy of final
backfill vertical location during sampling. The above data
indicate that this procedure resulted in greater than a
one-quarter order of magnitude conservative error in the Gilson
Road* QC data. This conservative bias is an unavoidable but
desirable facet of the QC testing program.
The combination of the above two factors indicates that the
Gilson Road construction QC data is likely unconservative by a
factor of approximately two. Again, it is emphasized the
elimination of the remixing step on future projects should result
in a somewhat conservative bias to the QC data.
2.4.1.4 Implication of QC Data on Containment
Performance
The quality control data obtained during construction
indicated at the time that the hydraulic conductivity
specification was generally satisfied while the percent bentonite
specification was not. Although the hydraulic conductivity data
indicated that approximately 22 percent of the samples tested had
2-47
-------
a hydraulic conductivity in excess of the maximum specified value
of 1.0 x 10~7 cm/sec, the higher values were judged to have
little impact on wall performance. The most important
considerations in assessing the impact of the higher than
specified hydraulic conductivities were the magnitude of the
variation from the specified limit and the location along the
wall alignment of the higher hydraulic conductivity zones. The
average hydraulic conductivity of the nineteen tests above
1.0 x 10~7cm/Sec was 2.1 x 10~7 cm/sec. This margin is small
relative to the four to five order of magnitude difference
between backfill hydraulic conductivity and the hydraulic
conductivity of the natural soil deposits (2 x 10"? cm/sec vs.
1 x 10~3 to 1 x 10~2 cm/sec). It should also be recognized that
the samples were tested at conservatively low confining stresses,
corresponding to "top of wall" conditions. It is likely that
some of these samples reflected backfill which ultimately
advanced to a deeper location in the wall and was subjected to
higher stresses in-situ than those in the test. Since an
increase in confining stress would cause a reduction in hydraulic
conductivity, the in-situ hydraulic conductivity is likely lower
than the tested values.
The major impact of the higher permeability zones is that
groundwater flow into the containment area, induced by pumping,
will be marginally greater in these areas. Overall, however, as
the average wall hydraulic conductivity was 5 x 10~8 cm/sec,
groundwater inflow through the wall would be less than
anticipated for a wall with an average hydraulic conductivity of
1 x 10~7 cm/sec. The presence of the bentonite filter cake along
trench walls, not previously mentioned should also help
compensate for marginally high hydraulic conductivity zones in
the backfill.
In addition to the relatively small degree by which the
backfill exceeded the specification limit, the higher hydraulic
conductivity zones appear to be located in areas of relatively
low hydrologic impact. As indicated, these areas include from
station 0+00 to 3+00, 6+00 to 12+00, and 25+00 to 29+00 (see
Figure 2-10). The direction of groundwater flow is predominantly
parallel to these sections of the wall, resulting in
comparatively low gradients across the wall. (Areas of largest
gradients would be those sections perpendicular or nearly so to
flow, approximately from station 13+00 to 19 + 00 and from station
32+00 to 0+00). The low gradients would result in comparatively
low flow quantities across the wall.
The subsequent quality control testing program on samples
obtained from the completed wall indicated that the construction
2-48
-------
FIGURE 2-10 APPARENT ZONES OF BACKFILL WITH
HYDRAULIC CONDUCTIVITY GREATER
NO
I
-V
THAN
"7
X
00
LYLE REED
BROOK
0' 50' 100'
200'
-------
QC data may be somewhat unconservative. Average hydraulic
conductivity of "undisturbed" backfill samples tested at in-situ
stresses was approximately two times greater than that of the
same sample remixed, suggesting that the hydraulic conductivities
of samples obtained during construction may be half the actual
hydraulic conductivity of in-situ backfill. Thus, the average
hydraulic conductivity of the wall backfill may be approximately
1 x 10~7 cm/sec, the specified maximum, rather than 5 x 10~^
cm/sec as indicated by the construction data. In addition, the
percentage of data above 1 x 10~7 cm/sec is likely greater than
the 22 percent indicated during construction. Nevertheless, the
somewhat higher average backfill hydraulic conductivity is judged
to have little impact on wall performance.
2.5 Quality Control Guidelines
The previous sections have focused on the use of API
fixed-ring testing and methylene blue titration for quality
control of soi1/bentonite backfill. These tests, however,
represent only part of the quality control effort required for
cutoff wall construction. In addition to further testing of the
backfill, the quality control program must also address the
support slurry, trench excavation, and backfill placement,
because all of these aspects will affect the quality of the
finished product. Specific guidelines for each of these areas
are presented in the sections following a discussion of broad
guidelines to quality control testing in general.
2.5.1 General Quality Control Guidelines
2.5.1.1 Test Requirements
The quality control program should be based on measurable
indices of. wall construction for which minimum design standards
have been established in the project specifications. The
contractor's workmanship can then be evaluated on the basis of
conformance to these standards. Field tests used to measure
these indices must meet four criteria to be acceptable for field
quality control:
1. Timeliness; QC testing must be "real time" in nature to
allow the timely modification of contractor1s procedures
early in the construction process, before inadequacies
become pervasive and jeopardize the effectiveness of the
wall. Tests which require several days to yield a
2-50
-------
result do not allow early detection of a potential
problem.
2. Ac curacy/Freei si on; The tests specified must be
sufficiently accurate and precise (repeatable) to yield
a meaningful representation of the measured parameter or
property.
3. Simplici ty : QC testing must be based on simple
equipment and procedures to allow execution in the
field.
4. Low Cost; Testing must be inexpensive enough to allow a
statistically significant number of tests to be
performed.
As discussed in previous sections, the API fixed-ring
hydraulic conductivity test and methylene blue tests both meet
these criteria. Other quality control tests discussed in the
following sections meet these criteria as well.
2.5.1.2 Quality Control Execution
In the past, field control has often been the sole
responsibility of the contractor. This was a reflection, in
part, of performance-based specifications which indiscriminately
assign all responsibility for satisfactory cutoff construction to
the contractor. However, because the contractor's primary
interest is in getting the job done, QC testing is often
relegated to a secondary importance. With limited construction
staffing when a construction schedule slips, the priority of the
contractor shifts to increasing production rather than quality.
QC testing performed by the contractor tends to serve as a
check on work completed rather than a control over work to be
done, analogous to the testing of concrete cylinders during
building construction. Therefore, tests are often performed
after a substantial portion of work is done, too late to have a
significant impact on construction. Because the objective of
quality control testing is not so much to document construction
as to identify deficiencies in construction procedures as early
as possible, "check" testing is an insufficient means of quality
control. For quality control testing to effectively serve as a
control over construction procedures, the tests must be performed
and reviewed on a timely basis, ideally by a qualified
representative of the owner (the design engineer) rather than the
2-51
-------
contractor. The engineer as quality control agent provides a
number of important advantages as compared to the contractor:
A. Understanding of Relationship between Design and QC
Data:
The design engineer has the highest degree of
understanding of the assumptions and constraints under
which the project was designed. As such, he is the most
capable of assessing the meaning of the QC testing data.
In addition, unanticipated variability in subsurface
conditions, etc., which can only be determined via
construction observation, may in themselves appear
inconsequential (to the contractor) but may have a major
impact on overall performance if not accounted for in
design. Again, the design engineer is best equipped to
evaluate this information as it affects the specific
construction at hand and/or subsequent construction
phases. The contractor may not have full knowledge of
all the design considerations and may not as readily
appreciate the significance of seemingly aberrant test
results.
B. Modifications:
Modification of QC procedures is relatively easily
accomplished by the engineer in response to
unanticipated contingencies. On the other hand, a
contractor following QC procedures outlined in the
specifications may not be in a position to judge what
modifications are necessary if unexpected contingencies
arise.
C. Responsiveness:
Inasmuch as a QC team of engineers is not burdened with
responsibility for the actual construction, their
efforts can be concentrated on taking samples,
performing tests, and providing real time data. If
difficulties develop during construction, the contractor
may try to reallocate QC testing manpower to activities
directly related to "getting the job done." Hence,
during precisely the period when QC testing may be most
important, its emphasis may be reduced.
D. Objectivity:
If the contractor is responsible for QC testing,
objectivity during sample selection and test execution
may not exist to a sufficient degree.
2-52
-------
Quality:
If the contractor is responsible for QC testing,
obtaining quality testing equipment and adequately
experienced technicians are not guaranteed.
The contractor should do limited testing of his own, but it
should not be incorporated in the engineer's quality control
program. The
use should
data.
contractor's data should be gathered for his own
a dispute arise with the engineer's quality control
2.5.2 QC Guidelines for Support Slurry
During cutoff wall construction, the stability of the trench
excavation depends to a large part on maintaining a positive head
of a bentonite slurry within the trench.
Design parameters
Unit Weight -
for the slurry include:
the slurry density is an indirect
indicator of bentonite content; the
density should be slightly higher than
that of water, while not being so heavy
as to impede the movement of the
advancing backfill face.
Viscosity
Filtrate
Loss
the viscosity of the slurry is an
indicator of the rate at which the filter
cake will develop along trench walls; the
higher the viscosity the faster the
filter cake will form due to higher
bentonite concentration in the slurry.
filtrate loss is also an indicator of the
capacity of the slurry to form a filter
cake of bentonite particles along the
trench walls.
Because the support slurry is ultimately displaced and not a
permanent element of the completed cutoff, its impact on wall
performance is significantly less than that of, say, backfill
preparation or placement. The filter cake which remains does
contribute to overall wall impermeability, typically having a
hydraulic conductivity of 1 x 10"^ cm/sec or less. However, the
contribution of the filter cake is not accounted for when
2-53
-------
evaluating wall performance and is thus regarded as a safety
factor. The primary impact of an inadequate slurry is trench
collapse, which poses construction problems (i.e., delays), but
does not necessarily have a long-term effect if the collapsed
section is reexcavated properly. Thus, the contractor will be
more concerned with a trench collapse, or the potential for one,
than the engineer.
Accordingly, quality control of the support slurry is often
the contractor's highest priority during construction. The
frequency of his slurry testing typically far exceeds that of
other quality control tests because the consequence of inadequate
slurry is costly in terms of time lost for reexcavation. At the
same time, because of the contractor's concern with construction
delays and the comparatively low impact of slurry quality on the
completed wall, the engineer need not place great emphasis on
slurry testing when planning his quality control program.
The engineer should include within his quality control
program a series of tests on slurry, but the required testing
frequency can usually be less than that adopted by the contractor.
Slurry quality control tests include: unit weight, viscosity,
and filtrate loss. Their use is described in the following
paragraphs.
2.5.2.1 Unit Weight
Unit weight of the slurry may be determined by numerous
means of weight/volume measurements. However, unit weight is
commonly determined using the mud balance test, which consists of
a specially designed scale calibrated to indicate density when
loaded with a prescribed volume of slurry. As indicated, the
slurry should always have a density greater than that of water
while not being so dense as to impede backfill placement.
Execution of the test is straightforward, and the
verification of a minimum unit weight relatively simple. A
slurry sample obtained from the top of the trench is sufficient
to establish whether the slurry density is higher than 62.4 pcf.
Slurry unit weight generally increases with depth, however, due
to the slow sedimentation of entrained sand and, in certain
cases, the settling of flocculated bentonite particles.
Determination of maximum slurry unit weight must then be based on
a sample obtained from the bottom of the trench. Collecting such
a sample can be difficult.
The use of a backhoe bucket to obtain a sample from the
bottom of the trench is generally inadequate. By the time the
2-54
-------
bucket emerges from the trench bottom, the slurry in the bucket
is most representative of slurry from the top of trench,
particularly if the bucket is designed with holes to drain water,
as is common. Therefore, alternative means of sample collection
must be used. A sampling device which traps slurry from the
trench bottom is necessary. Groundwater samplers with ballcheck
valves are commercially available, but are typically inadequate
for this use. A slurry sample which contains entrained sand
particles may interfere with proper seating of the valve. In
addition, if the slurry is heavy and highly viscous, it will not
pass the sampler orifice and the sampler will displace the heavy
slurry.
2.5.2.2 Viscosity
Viscosity is determined using the Marsh Funnel test which
measures the time for a fixed volume of slurry to empty from a
standard Marsh Funnel. Typically, a minimum time of forty
seconds is specified to achieve the proper viscosity. A
contractor will try to maintain a Marsh Funnel time as close to
forty seconds as possible without dropping below forty seconds.
Because viscosity is proportional to the bentonite content and
barrel yield^ of the slurry, a viscosity significantly in
excess of the specified value is indicative of excess bentonite.
Because the cost of bentonite generally represents the highest
materials cost, the contractor will seek to use the minimum
bentonite required.
2.5.2.3 Filtrate Loss
Filtrate loss tests are run in the API filter press cell.
The standard criteria for support slurry are less than 30 cc of
filtrate loss when subjected to a 100 psi pressure for thirty
minutes. The tests are typically run on fresh mixed slurry prior
to the introduction into the trench. A test on a sample of the
same slurry taken from the trench would yield a lower volume of
filtrate due to filtrate loss occurring in the trench. It should
also be added, however, that the filtrate loss test is a good
indicator of adverse chemical effects on the slurry. Groundwater
contaminants resulting in bentonite flocculation would lead to a
slurry with a relatively high value for filtrate loss.
^Barrel yield is defined as the number of 55-gallon drums of
bentonite slurry produced by one ton of dry bentonite.
2-55
-------
2.5.3 QC Guidelines for Trench Excavation
Control aspects of slurry trench excavation include:
0 trench verticality
0 trench continuity
0 depth of penetration into key (aquiclude)
0 bottom cleaning
0 trench stability
Quality control of trench excavation generally receives
little emphasis, particularly when compared with quality control
testing of slurry and backfill. To some degree, this is because
the excavation process is regarded as a relatively common
construction operation. This attitude also reflects the fact
that slurry trench excavation is performed "blindly:" there is
no way to observe the sides or bottom of the trench. Another
reason is that quality control of the trench excavation is
largely subjective. Aside from trench width and depth, there are
no distinct, measurable parameters.
As compared with quality control of slurry or backfill,
quality control of excavation consists more of a series of
measurements than tests. Methods of evaluating trench
verticality, continuity, depth of penetration into key, the
condition of the bottom, and trench stability are discussed in
the following paragraphs.
2.5.3.1 Verticality
Verticality of the trench is assessed by measuring the
verticality of the excavation equipment. The measurement is best
made using a 4-foot level on the backhoe boom or on the backhoe
tracks. (The backhoe typically sits on a compacted clay working
pad constructed along the trench alignment. The pad is primarily
intended to provide stable walls at the top of the trench).
Verticality is not generally regarded as a critical aspect
of wall construction, particularly to the contractor, since the
wall does not have to be perfectly vertical to function as
designed. However, wall verticality is essential to
post-construction investigation, particularly if backfill samples
are required from the completed wall and the wall is deep. A
small deviation of several degrees from the vertical at ground
surface can result in deviation of several feet at the bottom of
a 50-foot trench. The time and money spent in "finding" the wall
2-56
-------
at depth via borings can be a major obstacle to post-construction
verification.
Therefore, a horizontal working pad should be maintained
during construction. If the excavating equipment deviates by
more than 1 degree from vertical, the pad should be regroomed to
restore levelness. If the pad is properly established, however,
there should be little need to regroom because there should be
little construction activity along the wall alignment prior to
excavation.
2.5.3.2 Continuity
Continuity of the trench is largely verified by the sweep of
the backhoe stick. Especially along relatively straight sections
of wall, the action of the backhoe should indicate the continuity
of the excavation. Problems are most likely to arise at sharp
bends in the wall alignment. The continuity of the trench may
be affected by a discontinuous, or interrupted,excavation
procedure. The contractor may have to reposition his backhoe
several times to negotiate the bend. Each repositioning creates
a potential for a discontinuity.
To reduce the risk of discontinuity, the contractor can
overexcavate one leg of the trench and then start a new trench
angled anywhere from 45 to 90 degrees from the old. The new
trench should be started sufficiently behind the old trench to
intersect at full depth. The main disadvantage associated with
this type of overexcavation is that backfill placement is more
difficult to accomplish properly.
The quality control engineer can monitor trench continuity
by dragging a rigid, vertical probe along the trench bottom. A
break in trench continuity will block the advance of the probe.
2.5.3.3 Key Penetration
Adequate penetration of the trench into a naturally
occurring aquiclude is fundamental to the performance of a cutoff
wall. Thus, it is critical to know at what depth the aquiclude
is first encountered in order to determine the minimum acceptable
depth of trench to satisfy required penetration. The contractor
typically relies on his backhoe operator to determine when the
aquiclude is encountered and when sufficient penetration is
achieved. The depth of the trench is then commonly checked by
2-57
-------
the contractor using a weighted rope calibrated in foot
increments.
The quality control engineer must employ a more precise
method of measurement. The measurement tool should be a rigid
probe calibrated in foot increments. When the backhoe operator
indicates he has encountered the aquiclude, a depth measurement
should be taken. Verification that the aquiclude has been
reached is a simple matter of checking the trench spoil. The
aquiclude, when probed, should also "feel" different than the
generally coarser soils which overly it, particularly if the
aquiclude is a clay or silt. This first measurement should
become the record depth to top of aquiclude even if the backhoe
may have penetrated 6 to 12 inches into it. Key penetration
should be referenced to this depth. As the contractor continues
to excavate into the aquiclude, subsequent measurements should be
taken until the minimum specified depth of penetration is
achieved. The depth measurements should be referenced to
pre-established stations along the wall alignment to maintain a
record of excavation depth and key penetration depth.
2.5.3.4 Key Cleaning
The removal of sediment from the key zone prior to backfill
placement is important for two reasons. First, the key
penetration is effectively reduced if a portion of the key zone
contains coarse sands which have sedimented out of the slurry.
The reduced penetration may jeopardize the integrity of the
cutoff. Second, sediment at the trench bottom can be pushed as a
wave by the toe of the advancing backfill slope to the extent
that the sediment mounds and becomes incorporated in the backfill.
The advancing backfill pushes this pile of sediment until it
becomes too heavy to move. The backfill then rides over the
mound of sediment, leaving a "window" within the backfill. This
is most likely to occur at a bend in the wall alignment.
The removal of sediment from the trench bottom is
most-commonly accomplished by the backhoe bucket. The quality
control engineer, however, must have the ability to "feel" the
trench bottom himself to assess the effectiveness of the cleaning.
This evaluation can generally be made using a rigid probe. Also,
the variation in the vibration of the probe as it passes through
the slurry and sediment can be detected. It is unlikely that
this determination can be made by the contractor using the
backhoe bucket because the heavy equipment is not sensitive
enough to detect the transition from slurry to sediment.
2-58
-------
The probe technique can also be used to detect the build-up
of sediment at the toe of the advancing backfill slope. Frequent
probing for this problem should be incorporated into quality
control procedures. As required, the contractor should excavate
the mound of sediment before it becomes trapped within the
backfill.
2.5.3.5 Stability
Stability of the uppermost section of the trench is
relatively easy to monitor. Typically, a failure of the
excavation will be signaled by the development of tension cracks,
roughly parallel to the trench alignment. The failure itself
will consist of a conchoidal wedge of soil which slides into the
trench. The remedy in such a case is simply to excavate the
fallen soil.
Stability of the trench at greater depths, when there may be
no signs of failure at ground surface, is more difficult to
assess. The sloughing of a trench wall can occur below the
slurry level and may involve no visible movement of soil at the
surface. If the sloughing occurs in a section of the trench that
has not been completely excavated or cleaned, the sloughed
material will most likely be excavated. If the sloughing occurs
in a section of trench through which the backfill is advancing,
however, the sloughed soil will get trapped within the backfill,
leaving a high hydraulic conductivity window.
Though there are no direct means of assessing trench
stability below the slurry level, the quality control engineer
can detect sloughing failures using an indirect approach. This
consists of frequently monitoring the profile of the backfill
slope. The backfill should normally slump forward in the trench
on a relatively consistent slope. This can be verified by
probing at intervals along the trench alignment. The frequency
of probing and spacing will be determined on a case-by-case basis.
A sloughing failure would deposit soil on the backfill slope and
alter its profile. The mound of soil created by the failure
should be detectable by probing, showing up as an irregularity in
the backfill profile. Profiling at the beginning and end of each
working day will allow the timely detection of a sloughing
failure.
2-59
-------
2.5.4 QC Guidelines for Backfill
The focal point of the engineer's quality control program
should be the backfill. Further, QC of backfill should involve a
series of tests which address aspects of both backfill placement
and long-term performance.
QC tests related to backfill placement include:
0 Unit Weight - the backfill must be sufficiently dense
to displace the support slurry.
0 Slump - the backfill must have a high enough
slump to "flow" into the trench without
being too fluid and thus mixing with the
support slurry.
QC tests related to long-term backfill performance include:
0 Hydraulic - API fixed-ring testing is the most direct
Conductivity quality control measure to assess wall
performance.
0 Methylene - bentonite content determination should be
Blue used to supplement API fixed-ring results
Titration particularly if hydraulic conductivity is
too high.
0 Gradation - gradation testing during construction is
perhaps the least critical quality
control test since backfill components
and proportioning should have already
been determined during design testing;
gradation testing does, however, provide
an ongoing check on fines content.
2.5.4.1 Unit Weight
Determining the unit weight of the backfill involves a
straightforward weight-volume calculation. Most critical is
obtaining a large enough sample to yield a representative unit
weight. Thus, the size of the sample must be large relative to
the largest particle size in the backfill. A standard concrete
cylinder should generally suffice.
Unit weight testing rarely has an impact on construction
procedures. Most backfill mixes, whether granular or clay-based,
2-60
-------
will easily satisfy the minimum unit weight requirement with no
special effort.
2.5.4.2 Slump
Slump testing of backfill is performed using the same
procedure as slump testing of concrete (ASTM C-143-78). The
minimum and maximum slump values should be selected based on
backfill placement considerations. A low slump mix will not
"flow" easily into the trench and will result in relatively steep
backfill slopes within the trench. The steeper the slope, the
greater the probability of a slope failure which could
potentially trap a pocket of slurry. A high slump mix will move
easily into the trench and result in a relatively flat backfill
slope. The greatest risk of a high slump mix is that the slurry
will not be positively displaced and there may be some mixing of
the slurry and backfill. A secondary risk is that a relatively
flat backfill slope will result and thus long sections of trench
will be open for an inordinately long time, increasing the
potential for a sloughing failure. The deeper the trench, the
more likely the chance of a failure.
Assuming that the contractor has prepared the backfill using
the proper proportions of backfill components, results of the
slump test should be used as the main determinant of whether the
backfill is ready for placement in the trench. It is not
practical to wait (even several hours) for the results of
hydraulic conductivity results on the backfill. Therefore, once
the backfill has achieved an acceptable slump and visual
homogeneity, placement is allowed.
2.5.4.3 API Fixed-Ring Hydraulic Conductivity Testing
Guidelines for use of the API fixed-ring test for quality
control were outlined in Section 2.1.3. To summarize:
0 At the design stage and during the early stages of
construction the site specific accuracy of the API
fixed-ring test should be assessed by comparing APA
fixed-ring test results to triaxial hydraulic
conductivity results on identical backfill samples.
0 During construction, samples obtained for API fixed-ring
testing should be transported and placed in the test
apparatus with as little disturbance as possible.
Specifically, no additional mixing should take place.
2-61
-------
2.5.4.4 Methylene Blue Titration for Bentonite Content
Guidelines for use of the methylene blue test were outlined
in Section 2.2.3. Besides its obvious value of determining
conformance with specifications, the methylene blue test should
be used to complement API fixed-ring testing. If backfill
hydraulic conductivity is too high, it is necessary to identify
the reason. Because insufficient bentonite content is the most
probable cause of a high hydraulic conductivity, methylene blue
testing can be used to verify bentonite content.
2.5.4.5 Gradation
Gradation testing of backfill samples involves conventional
soil testing procedures. Because gradation specifications for
backfill are of necessity broad, the primary value of gradation
testing is in the determination of percent fines. As with
methylene blue testing, gradation testing should be considered a
complement to API fixed-ring hydraulic conductivity testing,
particularly where API fixed-ring tests yield unacceptable
results.
2.6 Conclusions
The first phase of this project focused on the evaluation of
two quality control procedures for soil/bentonite cutoff walls:
API fixed-ring hydraulic conductivity testing and methylene blue
titration for determination of bentonite content. These
procedures were adapted for use at the Gilson Road site in 1982.
As part of the project, testing procedures were refined and
standardized for future use. Following assessment of these
procedures, the quality control data obtained at Gilson Road was
reviewed to assess conformance to specifications. Based on
experiences gained at the Gilson Road site and other subsequent
cutoff wall projects, guidelines for additional aspects of
quality control were developed.
Major conclusions of this phase of the research effort are:
0 Both the API fixed-ring hydraulic conductivity test and
methylene blue test are suitable for field quality
control of cutoff wall construction; they are
sufficiently accurate and precise to allow a meaningful
2-62
-------
evaluation of future wall performance, they require
relatively short testing times, they are relatively
simple to perform and they are relatively inexpensive on
a per-test basis.
A quantitative correlation between hydraulic conductivity
and backfill gradation characteristics did not emerge
from a statistical evaluation of the data obtained at the
Gilson Road site; the importance of such a correlation,
however, diminished -over the course of the research
effort as the suitability of API fixed-ring testing and
methylene blue testing was confirmed.
Hydraulic conductivity data obtained at the Gilson Road
site indicate that project specifications for the
completed wall were generally satisfied; approximately
20 percent of the data were in excess of the maximum
allowable limit, but by a margin deemed to have minor
impact on overall wall performance; the bentonite
contents for these samples were generally below the
minimum specification limit, but were generally above the
required minimum based on a design testing.
2-63
-------
SECTION 3
WINDOW DETECTION VIA PIEZOCONE SOUNDING
Appropriate use of quality control field testing allows a
high degree of certainty to be developed with respect to the
adequacy of soil/bentonite backfill proportioning and mixing
during cutoff wall construction (Section 2). This, coupled with
data obtained during the backfill design phase prior to
construction (Schulze, 1984), provides confidence that the intact
backfill will meet the long term hydraulic conductivity
specifications once it is placed in the trench. However,
placement of the backfill in the trench results in a set of
inherent conditions which can allow entrapment of non-backfill
material ("windows") in the cutoff wall. While the intact
backfill may meet specification, the windows are typically of a
higher hydraulic conductivity and thus degrade the overall
performance of the containment.
The formation of windows may result from various scenarios
such as:
0 Sloughing of in-situ soils from the sides of the trench
onto the advancing backfill face.
0 Sedimentation of relatively course material, temporarily
suspended in the support slurry during excavation, onto
the backfill face.
0 Inclusion of pockets of bentonite support slurry during
slurry displacement by the backfill and
0 Failure of the excavation tools to remove all the in-situ
material and/or trench sediment from the key profile,
particularly when keying to the top of bedrock.
These occurrences can be due to either geologic conditions,
such as highly pervious open worked in-situ soils causing local
instabilities and thus sloughing, or construction inadequacies
such as insufficient or excessive backfill fluidity.
Based on the probable mechanisms leading to window
formation, they are expected to be relatively small in size;
3-1
-------
hypothesized order of magnitude less than 10 to 100 square feet
in profile dimension. Despite their small size, windows can have
a significant impact on the overall efficiency of the cutoff wall
barrier. This is a direct result of the groundwater mounding and
thus increased gradient in the vicinity of the barrierl
(Powers 1981).
The ability to cost-effectively detect such defects is
instrumental to the verification of successful adherence to the
performance specification for backfill homogeneity should a
post-construction verification program become necessary.
Furthermore, the demonstration of a capability for
post-construction window detection may provide added incentive to
the contractor to strive for the highest quality of workmanship
during construction.
The most direct method of window detection is to obtain
undisturbed samples of the completed cutoff wall. A quantitative
assessment of the hydraulic conductivity of the window material
can then be obtained via laboratory testing. However, due to the
anticipated small size of a typical window as compared to the
cross sectional area of the cutoff wall,2 a large number of
samples are required to detect a statistically significant
percentage of the windows. The high unit cost of sample
retrieval and laboratory testing therefore render this solution
impractical3.
As an alternative to direct sampling, the feasibility of
remote sensing geophysical techniques was considered. Down-hole
and cross-hole techniques have gained wide acceptance by the oil
related industries. More recently, surface techniques have been
developed for hazardous waste investigative work. However, much
of the current technology was judged to be unacceptable for
detection of anomalies in soil/bentonite cutoff walls as based on
•••The mounding and increased gradient caused by the intact portions
of the barrier elevate the flux through the windows above that
which would be calculated based on the cross-section of the
window and the gradient which existed before barrier
construction.
^For the Gilson Road project, a 100 square foot window would
account for less than five one-hundredths of one percent of the
cross sectional area of the 4,000-foot long wall.
^Undisturbed shelby tube samples taken at 10-foot depth intervals
at 50-foot centers along the Gilson Road cutoff wall would cost
approximately $400,000 including triaxial testing. This figure
approaches one-half the cost of wall construction.
3-2
-------
theoretical considerations. Limitations of these techniques are
discussed in Appendix C.
The considerable experience existing in the geotechnical
exploration and geophysical testing fields indicated that
piezocone equipment might be ideally suited for window detection.
Commercially available instruments have been routinely used for
over fifteen years for foundation design and off-shore work in
soft soils. Originally developed for stratigraphic
identification, these instruments provide continuous soil ..index
data as they are advanced from the ground surface. Their rapid
penetration rate also offer the potential for cost-efficient
continuous soundings at many locations along a cutoff wall. The
primary objective of Phase Two was therefore, to evaluate the
suitability of electronic piezocone instrumentation for window
detection.
The first step in this phase was to evaluate the need for
modification of existing equipment to make it more suitable for
detection of windows in soil/bentonite cutoff walls (Sections 3.1
and 3.2). The second step was to construct a cutoff wall test
section incorporating various types of windows at known locations
(Section 3.3). The ability of the piezocone to detect windows in
the test section was then evaluated as the third step in the
research (Section 3.3). Steps four and five encompassed use of
the piezocone in the actual Gilson Road containment wall (Section
3.4) and obtaining undisturbed samples of materials identified as
potential windows (Section 3.5).
3.1 Piezocone Operating Principles
3.1.1 Standard Instrument Configuration
Current electronic piezocone instrumentation is an outgrowth
of the mechanical Dutch Cone equipment used in Europe for many
years (Sanglerat 1972) and the piezoprobe developed in the early
1970's (Wissa 1975). The piezocone used for this project was
developed by Geotechniques International, Inc. and is depicted in
Figure 3-1. The instrument has a 60 degree apex angle conical
point, 3.56 cm in diameter, with a projected area of 10 cm2 as
based on the Fugro design. Overall probe length is approximately
50 cm.
The conical point of the instrument is connected to a load
cell which allows measurement of the point resistance developed
during sounding. The end of the point houses a porous section
3-3
-------
FIG. 3-1 SCHEMATIC OF PIEZOCONE INSTRUMENT
U)
I
I. CONICAL POINT (10 cm2)
2. POROUS TIP
3. STRAIN GAGES; a = POINT, b = SLEEVE
4. FRICTION SLEEVE (225cm2)
5. ADJUSTMENT RING
6. WATERPROOF BUSHING
7. CABLE
8. CONNECTION WITH PERMEAMETER
a PRESSURE TRANSDUCER
-------
which is hydraulically connected to a pressure transducer for
measuring excess pore pressures generated during the sounding.
The frictional and/or adhesion force on the side of the
instrument is measured by an independent load cell. The load
cell is connected to a 225 cm2 area friction sleeve which is 20
cm long and is positioned immediately behind the point. The data
generated by these transducers are transmitted to the data
acquisition system at the surface via an electrical cable.
Additional description of the standard piezocone, its
development, and application can be found in Sanglerat 1972.
3.1.2 Instrument Modification
As part of this effort, the piezocone was modified to
include an additional data channel for measurement of a "relative
hydraulic conductivity index." The rationale for this
modification was based on recognition that the backfill hydraulic
conductivity was the primary parameter of interest. Furthermore,
the standard piezocone data channels for point resistance, excess
pore pressure, and side friction primarily respond to material
density, consistency, and shear characteristics, rather than
hydraulic conductivity. Therefore, a more direct indication of
hydraulic conductivity was sought to custom tailor the piezocone
to soil/bentonite backfill sounding. Inasmuch as a direct
measure of in-situ backfill hydraulic conductivity was not
possible, a relative quantitative index was considered
appropriate.
The "permeameter" consisted of a 50 micron pore size ultra
high molecular weight polyethylene porous stone. The stone was
23 cm2 in exposed area and was located approximately 60 cm behind
the point (see Figure 3-2). It was hydraulically isolated from
the pore pressure transducer at the point of the piezocone and
also from the annular space within the drill rods. The stone was
maintained at a constant driving pressure4 via a 1/4-inch
diameter nylon tube which connected the stone to a pressurized
water reservoir at the surface. The reservoir was pressurized
via a precision regulator and a tank of compressed gas.
4The total driving pressure in the stone was increased as the
depth of the instrument increased during a sounding. The total
piezometric head in the stone was thus maintained at a constant
preset level above the total piezometric head in the in-situ
soils surrounding the cutoff wall as based on steady state
hydrostatic conditions.
3-5
-------
FIG. 3-2 SCHEMATIC OF PERMEAMETER
OJ
I
KEY:
I. PIEZOCONE
2. BOND BREAKER
3. PERMEAMETER
4. POROUS STONE
5. FITTING FOR WATER ENTRY TO STONE
6. 0.25 INCH NVLON TUBE
7. N W COUPLING TO DRILL STRING
8. PIEZOCONE CABLE
NOTE: DASHED LINE INDICATES INITIAL PROFILE.
-------
During a sounding, the volume of water which exited the
stone into the soil/bentonite backfill was monitored versus depth
of probe penetration. Flow volume monitoring for the low flow
rates associated with the relatively impermeable backfill was
accomplished using a commercially available unit which is
commonly employed for triaxial testing (manufactured by
Geotechniques International, Inc., see Figure 3-3). This unit
records the volume of water which is expelled from the surface
reservoir via a sliding piston. The piston travel, which is
directly proportional to volume, was monitored with a linear
direct current displacement transducer (DCDT). The output of the
DCDT was fed into the data acquisition system. The rate of flow
into the backfill at various depths provided an index indicating
relative backfill hydraulic conductivity. An auxiliary pressure
application/flow measuring system was also incorporated into the
equipment to monitor higher flow rates but at a reduced accuracy.
This portion of the equipment was not used due to limitations in
the rate of flow which could be transmitted through the porous
stone.
3.1.3 Sounding Procedure
To execute a sounding, the piezocone was pushed into the
formation via a drill rig at a constant rate of penetration (1 to
2 cm/sec). The sounding was stopped at 5-foot intervals to add
drill rod. The depth of the piezocone was monitored by an
external transducer which is connected to the drill string. Data
output from this transducer was fed to the data acquisition
system. The drilling procedure utilized during this project was
in accordance with ASTM 3441.
Inasmuch as the permeameter portion of the instrument was
not standard, procedures for its operation are not covered by
ASTM. The procedure used is therefore described as follows.
Initially, a determination of the magnitude of driving pressure
to be applied to the stone was made. Selection of this pressure
represents a tradeoff between two mutually exclusive objectives;
1) a high pressure provides high volume flow into the backfill
which increases measurement accuracy and 2) a low pressure
reduces the probability of hydraulically fracturing the formation
or creating boundary flow along the drill string, particularly at
shallow depths. The pressure utilized during the project
described her,ein was generally 2.5 psi for the test section
soundings and and 15 psi for the containment wall soundings.
This driving pressure was applied to the stone prior to each
sounding to ensure that the porous stone, pressurization and
volume measuring systems were operating correctly. The rate of
3-7
-------
FIG. 3-3 PERMEAMETER PRESSURIZ AT I ON/FLOW MEASUREMENT SYSTEM
TO
PERMEAMETER
U)
I
00
TO COMPUTER
TO PRESSURIZED GAS SUPPLY
5 (TYPICAL)
KEY'
TO
WATER RECHARGE
SUPPLY
I. 0-100 PSI REGULATOR
2. DCDT TRANSDUCER ON PISTON
3. 2X INTENSIFIER (LOW FLOW)
4. VISUAL FLOW METER (HIGH FLOW)
5. VALVES
6. HIGH VOLUME WATER RESERVOIR
-------
flow was recorded and compared to previous data during each
sounding. The driving pressure was maintained at a constant
value throughout the sounding. During each drill rod addition
(5-foot intervals), flow to the stone was terminated and the
surface water supply reservoir was refilled.
Data from the three internal piezocone transducers (point
resistance, dynamic pore pressure and side friction) and the two
external transducers (permeameter flow and instrument depth) were
transmitted through electrical cables as independent 4 to
20 milliamp analogue signals. The signals were processed to
digital format by an Analog Devices A to D converter. The
digitized data was stored directly in memory and subsequently on
floppy disk using an Analog Devices Mac Sym 150 Computer.
3.1.4 Data Manipulation and Output Format
The telemetry data for point resistance, excess pore
pressure, local friction and permeameter volume flow was
initially in chronological format. A data reduction program
then transformed the data into a depth correlated format5-
The program also subtracted the value of the steady state pore
pressure at each point in the formation surrounding the wall
(computed assuming hydrostatic conditions) from the total pore
pressure measured at the point of the piezocone. Excess pore
pressure, as generated by quasi-static penetration6f was thus
computed. In addition, the program manipulated the independent
transducer outputs to develop composite parameters. These
included friction ratio (side friction divided by point
resistance) and A factor (excess pore pressure divided by point
resistance). The input data used for these computations was
depth correlated and, as such, represented transducer response
for the material at a given depth.
5The individual piezocone sensors, located along the 60 cm length
of the instrument, transmit data immediately as it is taken.
Therefore, at any instant of time, each sensor is responding to
backfill characteristics at a different depth. To be useful, the
data from the individual sensors must be shifted in time such
that the individual parameter values correspond to the same
backfill depth.
6As the cone penetrates the material of interest, it generates an
excess pore pressure (a pore pressure different from that which
existed before the sounding) due to cavity expansion displacement
and shearing processes. This excess pore pressure varies in
magnitude and can be either positive or negative depending on the
material characteristics.
3-9
-------
The data acquisition system performed the above described
data reduction on a real time basis. As such, the output was
displayed directly on a CRT screen to allow the operator to
monitor piezocone operation during the sounding. The following
data was presented on the CRT as plotted against a depth axis:
Point resistance in kg/cm2
Local friction in kg/cm2
Excess pore pressure in kg/cm2
Friction ratio in percent
A-factor in percent
0 Permeameter volume displacement in cm3/ft
At the end of the sounding, a paper copy of these plots was
generated.
Each time the sounding was stopped to add another length of
drill rod (5-foot intervals), the excess pore pressure at the
piezocone point was monitored with time?. The excess pore
generated provided
characteristics. These
on the CRT and then as
these data collected at
when desired by the
pressure equilibration behavior thus
additional data with respect to material
data were initially presented in real time
hard copy. The sounding was stopped and
points other than drill rod addition
operator.
3.2 Material Identification
Electrical piezocone equipment is well suited for
differentiating between various soil types and thus determining
the geologic stratigraphy encountered during a sounding.
Material identification is accomplished through the use of
empirical correlations which relate transducer output to material
properties. The following sections first discuss the expected
response for each individual transducer when sounding various
soil types. It is recognized that each such index, by itself,
does not provide definitive soil identification. However, the
discussion serves to demonstrate the characteristics and range of
responses to be expected. The relative output of multiple
transducers, producing indices such as friction ratio and A
factor is then presented as related to empirical correlations
which have been developed for soil type identification. Finally,
'The excess pore pressure is generated by piezocone penetration of
the material. Therefore, once penetration is stopped, the excess
pore pressure decays to the steady state value which existed
prior to the sounding.
3-10
-------
the expected response of the piezocone to intact soil/bentonite
backfill and various potential window types is discussed.
3.2.1 Point Resistance
During the development of the electrical piezocone, the
first index measured was point resistance. This is defined as
the load required to penetrate the point into the material of
interest** divided by the projected area of the point
(typically 10 cm2). in addition to the standardization of point
area, the point apex angle has also been standardized at 60°.
Penetration is controlled to a constant, relatively slow rate
(1 to 2 cm/sec) and is therefore considered quasi-static. As
such, dynamic inertia effects are assumed to be zero.
The penetration resistance encountered by the point is
primarily related to the density and shear strength of the
material penetrated. The greater the density and/or shear
strength, the greater the magnitude of point resistance. This is
true for both cohesive (clayey) soils and cohesionless (granular,
non-plastic) soils. As such, point resistance by itself cannot
be used to distinguish soil type. A loose/dense sand can yield
the same point resistance as a soft/stiff clay depending on the
relative degree of the material density/consistency.
In soils with a significant percentage of gravel size
constituents, erratic and artificially high readings may be
obtained locally. This typically occurs when the point engages a
piece of gravel, particularly one with a projected area greater
than 10 cm2 f an<3 pushes it through the soil matrix. This problem
is alleviated to some extent by averaging algorithms contained in
the data reduction program9.
°The load generated by point penetration is measured by a
transducer connected directly to, and positioned immediately
behind, the point. In this way, forces generated by side
friction/adhesion on the aides of the instrument and the drill
string is not measured.
the system used during this project, the analog transducer
output is digitized at 60 hertz. The individual digital values
are then averaged over a one-quarter second time span
(penetration depth of 1/4 to 1/2 cm) for data presentation.
3-11
-------
3.2.2 Local Friction
The second parameter to be incorporated into cone
penetrometer instruments was measurement of local friction.
Local friction is defined as the vertical shear load engaged by
the friction sleeve during penetration divided by the friction
sleeve area (typically 150 cm2f but 225 cm2 for the instrument
used during this project in recognition of the probable low shear
values). The friction sleeve is approximately 20 cm long and is
the same diameter as the base of the conical point. It is
located immediately behind the point as shown in Figure 3-1. In
practice, adhesion forces are measured as well as purely
frictional forces. Viscous drag forces are limited to the extent
possible via the quasi-static rate of penetration.
For granular, cohesionless soils, the magnitude of local
friction encountered during a sounding is related primarily to
the lateral effective stresslO and the friction angle of the
material. The greater the lateral effective stress and/or
friction angle, the larger the value of local friction. In
addition, the elasto-plastic behavior of the soil affects local
friction measurements to the extent that the lateral effective
stress is increased via cavity expansion upon penetration.
For fine-grained clayey (cohesive) soils, the measured local
friction is a function of the undrained shear strength as well as
frictional forces to the varying extent that the material behaves
undrained (shear strength) versus drained (frictional). For a
given point resistance, it is typically expected that a cohesive
soil will yield higher local friction values as compared to a
cohesionless soil. As for the case of point resistance, the
value of local friction, in and of itself, cannot be used for
definitive soil type identification.
3.2.3 Excess Pore Pressure
Measurement of the excess pore pressure generated as the
piezocone penetrates the formation is a relatively recent
development. Initially, the pore pressure transducer and porous
stone system was placed in an 18° apex angle probe and used for
measurement of steady state pore pressures only. This
instrument, the piezoprobe (Wissa 1975), incorporated no
luThe lateral effective stress is that portion of the total lateral
stress that is due to forces transmitted via inter-granular
contact. Pore water fluid pressures are therefore not included
in effective stress (Lambe, 1969, Pg. 241).
3-12
-------
provisions for measurement of point resistance or local friction.
In fact, no effort was made to measure the instantaneous values
of excess pore pressure generated as the probe was inserted to
the depth of interest. The probe was therefore a reusable, very
rapidly responding piezometer. The high stiffness value for the
system allowed it to equilibrate with the formation very quickly
and also minimized the change in formation pore pressure in the
vicinity of the piezometer due to the act of measuring the pore
pressure11. This characteristic made it ideal for measurement
of rapidly varying pore pressures such as exist in clay
formations subject to embankment loading.
After initial problems with transducer overload from too
fast an insertion rate in soft clays, pore pressure was monitored
during insertion to protect the transducer. It was then noticed
that the pore pressure generated during insertion was highly
sensitive to material type, particularly sand senses in a clay
formation or the converse. Piezoprobe capability was therefore
expanded for use as an aid in stratification identification and
then incorporated in the standard cone penetrometer (point and
side resistance only) to form the piezocone.
The excess pore pressure generated due to constant rate
penetration depends on three general soil characteristics. The
first is the soils elasto-plastic response to cavity expansion,
the second is the degree of dilatancy12 exhibited by the soil
in response to shear strain and the third is the soil's hydraulic
conductivity. While the soil's behavior is actually a complex
composite of the interrelationship between these effects, each
^Stiffness is a measure of the volume of water which must flow
into/out of the measurement system for it to respond to a given
change in formation pore pressure. The larger the volume
required, the longer the equilibration time required. This is
particularly critical in clayey formations where the low
inherent hydraulic conductivity places severe limits on the rate
of flow to the piezometer. This constraint not only applies to
equilibration time following installation, but also to the time
for the piezometer to respond to a change in formation pore
pressure. If the volume required for system response is high
(low stiffness) as compared to formation hydraulic conductivity,
the response time may be greater than the duration of the change
in formation pore pressure. As such, the act of measurement
changes the parameter being measured and a true reading is never
obtained.
12An increase in soil volume upon shearing due to the geometric
interlocking of individual soil grains in dense soils (Lambe,
1969 Pg. 129).
3-13
-------
will be treated as if independent in the following paragraphs to
demonstrate general trends (Lexadoux 1980).
As the piezocone instrument penetrates a soil mass, it
compresses it due to volume displacement. The total stress
therefore increases in the local vicinity of the sounding. The
magnitude of the increase in total stress depends on the
elasto-plastic response of the soil to the volume displacement
(cavity expansion). An increase in the total stress tends to
increase the pore pressure above the steady state value which
existed prior to the sounding. The trend is therefore towards
positive excess pore pressure generation for all soil types, but
the magnitude is soil type dependent.
During penetration, the soil experiences severe shear strain
as well as compression. Shear strain also tends to change the
pore pressure. For this case however, not only is the magnitude
of the change soil type dependent, but so is the direction of the
change (positive vs. negative). The shearing process in loose
sands and soft normally consolidated*3 clays tends to drive the
excess pore pressure in a positive direction and thus complements
compressional effects. As such, penetration of these materials
generates a positive excess pore pressure which is superimposed
on the previously existing steady state conditions. For dense
sands and stiff overconsolidatedll clays, dilatancy effects
drive the shear related pore pressure in a negative direction.
Hence, the shear component of the excess pore pressure opposes
the compressional component. The total excess pore pressure can
therefore be either positive or negative depending primarily on
the degree of dilatancy exhibited by the soil (density of sands,
degree of overconsolidation of clays).
As excess pore pressure is generated due to cavity expansion
and shearing processes, it immediately begins to dissipate to the
soil mass surrounding the sounding. The rate of drainage, as
compared to the rate of penetration, directly affects the
magnitude of excess pore pressure attained. Therefore, given
that other parameters remain constant, a lower hydraulic
conductivity soil (clays), will typically generate a higher
excess pore pressure than a high hydraulic conductivity soil
(sands). This is true for both positive and negative pore
pressure generation.
13A normally consolidated soil is one which currently exists in
equilibrium at a state of stress equivalent to the maximum stress
state it has ever been subjected to. An overconsolidated soil
currently exists at a state of stress less than that which has
previously existed (Lambe 1969, pg. 74).
3-14
-------
The rate of excess pore pressure dissipation is indicative
of soil hydraulic conductivity. Low hydraulic conductivity soils
drain slower than high hydraulic conductivity soils and thus the
rate of excess pore pressure decay is slower. A relative
indication of soil hydraulic conductivity can therefore be
obtained by monitoring excess pore pressure decay when
penetration is stopped. This is typically done when drill rod is
added (at 5-foot increments). Figure 3-4 presents typical curves
showing excess pore pressure and rate of decay for various soil
types (Baligh, 1980).
3.2.4 Relative Hydraulic Conductivity Index
As indicated in Section 3.1.2, the standard piezocone was
modified to include a permeameter for evaluating the relative
hydraulic conductivity of the cutoff wall backfill. This
technique had not been used prior to this project and, as such,
expected response could not be determined based on past
experience. The following discussion is therefore based on
theoretical considerations, and is presented in the context of
window identification.
If the permeameter penetrates a material of constant
hydraulic conductivity with depth, the water take14 would be
constant. This is because the driving pressure to the stone is
maintained at a constant preset value above the formation pore
pressure, independent of depth. For homogenous backfill placed
as a vertical cutoff wall, it would be expected that the in-place
backfill hydraulic conductivity would decrease somewhat with
depth. The decrease in hydraulic conductivity results from
increased consolidation stresses and thus density due to the
weight of the overlying backfill. Therefore, water take would be
expected to decrease with depth in the absence of windows. The
rate of decrease with depth would be backfill type dependent.
The response obtained when penetrating a window depends
primarily on three variables; 1) the hydraulic conductivity of
the window, 2) the degree of communication between the window and
the in-situ formation surrounding the cutoff wall and 3) the size
of the window. As the hydraulic conductivity of the window
increases, the water take would be expected to increase. If the
window is in direct communication with a highly pervious
•^Defined as the volume of flow into the formation per unit depth
penetrated. Inasmuch as time is not incorporated in this
correlation, the rate of penetration must be held constant.
3-15
-------
EXCESS PORE PRESSURE-METERS OF WATER
co
O
o
-o
r
o
o
o
o
CO
m
3
CO
10
I
X
o
X
5
TJ
(-
>
CO
o
o
o
m
CO
s
CO
~ rn
m
z
CO
m
co
CO
O
Ut
_
en o
CD m
> co
~_ co
Jo "o
->1 O
01 10
— ' m
m
co
co
c
7)
m
o
co
co
TJ
O
z
DENSE/STIFF
LOOSE/SOFT
-------
formation surrounding the cutoff wall, then the water taken would
increase in proportion to the window hydraulic conductivity
(within the constraints that the hydraulic conductivity of the
window remains low as compared to the formation and the flow
required does not cause significant head loss in the permeameter
system). However, if the window is totally contained within the
backfill, then water take will be limited by the rate at which
water can flow out of the window into the surrounding backfill.
In this case, water taken will increase in proportion to the
surface area of the window and thus window size (within the
constraints that the backfill hydraulic conductivity is low in
comparison to that of the window).
3.2.5 Empirical Correlations
Empirical correlations have been developed between piezocone
response and soil type. These correlations are typically based
on the relationship between a "secondary" index (local friction
or excess pore pressure) and the primary index of point
resistance. These relationships are generally expressed in a
normalized fashion.
The first correlation to be developed was between local
friction and point resistance. In this case, the value of local
friction was normalized with respect to point resistance
resulting in the friction ratio index (computed as the local
friction divided by point resistance and reported in percent).
Figures 3-5 and 3-6 present two such correlations which
demonstrate that: 1) for a given point resistance, increasing
friction ratio indicates a finer grained, more cohesive soil and
2) an increasing point resistance at a given friction ratio,
indicates a coarser grained more cohesionless soil.
The second major correlation of interest is between the
excess pore pressure generated during a sounding and point
resistance. One such example is presented in Figure 3-7. In
this case, excess pore pressure has been normalized with respect
to steady state hydrostatic pore pressure and the point
resistance has been normalized with respect to the total vertical
stress. This correlation demonstrates that 1) for a given point
resistance, increasing excess pore pressure generation indicates
a finer grained, more cohesive soil and 2) an increasing point
resistance, for a given excess pore pressure, indicates a coarser
grained, more cohesionless soil.
The above correlations can be used to define soil type on a
general basis, only. The standard correlations are typically
3-17
-------
FIG. 3-5 MATERIAL IDENTIFICATION-FRICTION RATIO
(DOUGLAS, 1981)
10,000
SENSITIVE MIXEQ
SOILS
10 •*
2 4
FRICTION RATIO{%)
NONCOHESIVE
COARSE GRAINED
NONCOHESIVE COARSE
AND FINE GRAINED
COHESIVE-NONCOHESIVE
FINE GRAINED
COHESIVE
FINE GRAINED
3-18
-------
FIG. 3-6MATERIAL IDENTIFICATION-FRICTION RATIO
(SCHMERTMANN.I977)
CM
6
u
UJ
o
h-
(O
(O
UJ
o:
o
a.
100 "V
6 i
345
FRICTION RATIO (%)
VERY SHELLY SANDS,
LIMEROCKS
SAND
SILT-SAND MIXTURES
CLAYEY SANDS 8 SILTS
SANDY a SILTY CLAYS
INSENSITIVE, NON-FISSURED
INORGANIC CLAY
ORGANIC CLAYS a MIXED
SOILS
NOTE: BASED ON CORRELATIONS DEVELOPED FOR THE NORTH CENTRAL
FLORIDA GEOLOGY.
3-19
-------
FIG. 3-7 MATERIAL IDENTIFICATION-EXCESS PORE PRESSURE
(JONES,VANZYL, 1981)
6 nrr
o
ZD
I
rj
Ut = TOTAL PORE PRESSURE
Uo = HYDROSTATIC PORE PRESSURE
-------
modified based on empirical data generated for a local's specific
geology. Hence the soil identification curves can vary
significantly as demonstrated by Figures 3-5 and 3-6,
particularly at low point resistance. Therefore, caution must be
exercised when applying the general correlations to deposits for
which no prior data exists.
3.2.6 Window Identification
The occurrence of windows in completed cutoff walls and
their characterization is poorly documented. As such, not only
is there a lack of empirical correlations for window
identification, but the composition of actual windows has not
been investigated. Therefore, the following discussion of
expected piezocone response is based on the probable composition
and behavior of windows as based on hypothesized mechanisms
leading to their formation. An understanding of the response to
be expected, as based on theoretical considerations, is critical
to the interpretation of field data obtained during sounding of
the test section and the Gilson Road containment wall as
presented in subsequent sections.
The three most likely types of windows of concern are:
' 1. Clean granular window - Composed of either 1)
unexcavated in-situ formation soils, or 2) on-site
formation soils which sloughed off the trench wall,
remained intact, and became entrapped in the backfill.
In the first instance, it is expected that the material
would remain unaltered with respect to the parent
formation and thus retain it's density and strength. In
the second case, it is anticipated that the material
would show a decrease in density and strength. In
either case, the window would contain original
groundwater as the pore fluid.
2. Slurry-filled granular window - Composed of on-site
formation soils entrained in the slurry during
excavation which sediment out and contain bentonite
support slurry as the pore fluid.
3. Bentonite slurry window - Composed of bentonite support
slurry entrapped during slurry displacement by the
backfill.
For each of the above windows, the expected response for the
four piezocone transducers will be discussed individually in the
following sections. This is preceded by a discussion of expected
3-21
-------
response for intact soil/bentonite backfill. Figure 3-8 presents
a schematic representation of the expected relative piezocone
response to backfill and windows.
3.2.6.1 Intact Soil/Bentonite Backfill
The soil/bentonite backfill specified for the Gilson Road
site consisted of a primarily sand and gravel mixture augmented
to contain not less than 30 percent fines and not less than 5
percent bentonite. The material was blended and wetted with
trench support slurry until it reached a slump of 4 to 6
inches^5. This plastic material was placed in the trench with
a consistency of toothpaste and it consolidated with time to a
soft to medium consistencyl6.
A. Point Resistance
The soil/bentonite backfill used at the Gilson Road site
was far less dense than the existing in-situ granular
formations. Point resistance should therefore be
relatively low. It was also expected that the material
would behave as a plastic, rather than granular, matrix
further reducing expected point resistance.
With depth, the backfill is subjected to greater
stresses and point resistance was therefore expected to
increase. However, it should still remain well below
values associated with the in-situ formation.
Localized, higher point resistance "spikes",
superimposed on this general trend, would be anticipated
due to cobbles and gravel in the plastic soil matrix.
B. Local Friction
The absolute value of local friction attributable to the
soil/bentonite backfill is again expected to be much
lower than that of the existing in-situ formation.
Although the backfill is plastic, thus indicating
relatively high local friction due to adhesion
characteristics, the much greater density of the in-situ
granular formation results in greater frictional forces.
-'•-'Slump was measured with a standard concrete slump cone in
accordance with ASTM Standard C 143-78.
consistency of plastic soil deposits are standardly rated
from very soft to hard in accordance with data presented in Lambe
1969, pg. 72.
3-22
-------
FIG. 3-8 RELATIVE PIEZOCONE RESPONSE
POINT
0 PROFILE 0 RESISTANCE
LOCAL
FRICTION
FRICTION
RATIO
U)
I
CO
EXCESS PORE
PRESSURE
WATER
FLOW
m
«?/«
-.iX>*>.
Hi
& si
S.F. 6. -
S.S.F. -
SLURRY FILLED GRANULAR WINDOW
SUPPORT SLURRY FILLED WINDOW
- LOW DENSITY CLEAN GRANULAR WINtfOW
HIGH DENSITY CLEAN GRANULAR WINDOW
-------
With depth, local friction should increase due to the
associated higher state of stress and thus higher
undrained shear strength. Although the magnitude of
local shear is expected to be less for the backfill than
for the in-situ granular formation, the value of
friction ratio should be more consistent with the
plastic nature of the backfill. Hence, the friction
ratio would be anticipated to be greater for the
backfill than the granular native formation.
C. Excess Pore Pressure
The soil/bentonite backfill is both almost completely
granular and relatively plastic at the same time. As
such, it is atypical of most natural soil deposits.
This behavior is due to the exceptionally high
activity!? of the bentonite component. However, by
separating the two characteristics, it can be shown that
the excess pore pressure generated while penetrating the
backfill should generally be positive. If a completely
cohesionless soil existed at the same density as the
backfill, it would be considered loose. As such it
would not exhibit dilatant behavior and thus would
generate positive excess pore pressures. A totally
cohesive material mixed and placed under the same
conditions as the backfill, would exist under normally
consolidated conditions. Again, this would indicate
that positive excess pore pressures should be expected.
While the prediction of positive excess pore pressure
generation is straight forward as based on theoretical
considerations, the magnitude of the expected pressures
cannot be quantitatively defined. However, the excess
pore pressures generated should be greater for the
backfill as compared to the formations existing in-situ
at the Gilson Road site.
If the sounding is stopped, such as during drill string
addition, it is expected that the excess pore pressure
will dissipate at a slow rate. The rate of dissipation
should be consistent with the low hydraulic conductivity
of the backfill (1 x 1Q-7 cm/sec).
17Activity is a standard soil mechanics parameter which indicates
the degree of plasticity a given soil component (typically clay
sizes) will impart to a soil mixture, as related to the
percentage of that component in the soil; Lambe, 1969, pg. 34.
3-24
-------
D. Relative Hydraulic Conductivity
By definition, the water take in the backfill is
expected to be consistent with the backfill hydraulic
conductivity (1 x 10~7 cm/sec). Given the backfill
hydraulic conductivity, the applied driving pressure,
and the appropriate geometries, it is possible to
determine, quantitatively, the expected flow rate.
However, this analysis was not performed in that the
scope of the project was directed toward investigation
of the suitability of this index for window detection in
the cutoff wall. As such, the effort was limited to
establishing relative values of water take.
3.2.6.2 Clean Granular Window
Material composing a clean granular window would retain the
original formation gradation. The hydraulic conductivity and
strength of the window would be dependent on the mechanism
leading to its formation. However, the magnitude of each of
these parameters is expected to remain significantly greater than
exhibited by the intact backfill.
A. Point Resistance
Point resistance is proportional to the strength of the
material penetrated. Therefore, it is expected to show
an increase when penetrating a clean granular window.
The magnitude of the increase is dependent on whether
the window is of low density, such as that formed by
material sloughing off the trench side wall and falling
onto the backfill, or high density, as would be expected
for unexcavated material.
B. Local Friction
The decreased adhesion and coarse grained nature of the
window, as compared to the backfill, would tend to
decrease the value of local friction. However, the
increased strength of the window material would increase
the value of local friction. Therefore, the direction
of the change in local friction cannot be predicted.
However, the friction ratio (local friction divided by
point resistance) would be expected to decrease when
encountering the window.
3-25
-------
C. Excess Pore Pressure
To the extent that the hydraulic conductivity of the
in-situ materials are preserved in the window, it would
be expected that excess pore pressure would decrease
when entering the window. The magnitude of the decrease
should be dependent on the dilatant nature of the
materials. Therefore, the decrease in excess pore
pressure should be greater for high density windows than
for low density windows.
D. Relative Hydraulic Conductivity
Water take would typically be expected to show a marked
increase when the sounding penetrates the window. This
is directly due to the large increase in hydraulic
conductivity for the clean granular materials as
compared to the backfill (10-7 cm/sec). If a
significant portion of the window is in direct
communication with the in-situ formation (no separation
by intact backfill or filter cake), then the water take
will be limited only by head loss within the permeameter
system. This is a direct result of the high hydraulic
conductivity associated with the in-situ deposits at the
site. However, if the window is totally encapsulated
within intact backfill, then the actual magnitude of
water take will be a function of the size of the window.
In this case, the water take is limited by the flux at
the window/backfill interface. Therefore, the window
becomes analogous to having a porous permeameter stone
with an area equal to that of the window. This is
attributable to the disparity between window and
backfill hydraulic conductivities. Even in the case of
a totally encapsulated window, water take is expected to
increase significantly as compared to the intact
backfilllS.
Based on the above discussion, it is anticipated that a
clean, granular window would most easily be discriminated using
the parameters of point resistance and relative hydraulic
conductivity. Both indexes are expected to exhibit significant
increases in magnitude when the sounding encounters a window of
this type.
18The surface area of even a small window 1 Ft2 in dimension
(1,728 in2) is over four orders of magnitude greater than the
permeameter stone area (3.6 in2).
3-26
-------
3.2.6.3 Slurry Filled Granular Window
The sedimentation mechanism leading to the formation of this
type of window should result in uniformly graded (poorly sorted),
low density material with support slurry filling the void spaces.
The actual grain size of the material comprising the window would
be dependent on the gradation of the in-situ materials being
excavated, the viscosity of the support slurry, and the time
period between excavation and backfilling.
A. Point Resistance
Due to the low density of the window, it is anticipated
that point resistance will be low. However, inasmuch as
the resistance to penetration in the backfill is also
inherently low, the point resistance may show an
increase relative to intact backfill. The magnitude of
point resistance in such a window is expected to be
highly grain size dependent. (Slurry deposited fine
sand would be expected to exhibit a very loose
structure, where as a gravel filled window would be
relatively difficult to penetrate.) Therefore, the
point resistance for the window may actually be much
greater than that for the backfill in specific
instances.
B. Local Friction
The slurry filled granular window is expected to be an
essentially granular deposit. However, the slurry
filling the voids may impart significant adhesion,
depending on its consistency and thixotropy!^. As
such, local friction could not be predicted. As
indicated above, point resistance, as well as local
friction, cannot be predicted. Therefore, friction
ratio can not be predicted.
C. Excess Pore Pressure
The low density of the material should preclude dilatant
behavior. Therefore, it is anticipated that a positive
excess pore pressure would be generated when sounding
through a slurry filled granular window. The relative
magnitude of the excess pressure compared to that for
the backfill, as well as the absolute magnitude, is
-^Strength increase with time at constant composition and stress
(Lambe, 1969, Pg. 458).
3-27
-------
indeterminate with the existing theoretical
understanding.
D. Relative Hydraulic Conductivity
It is expected that the water take for this type of
window would be similar to that for a support slurry
filled window (see Section 3.2.6.4). However, for a
given viscosity slurry in the window, it would be
expected that the water take for the slurry filled
granular window would be somewhat less than the support
slurry filled window. This results from the added
resistance to flow commensurate with the increase
tortuosity associated with the small intergranular pore
size. In any case, the water take is still expected to
be greater in the window than in the intact backfill.
Based on the above criteria, it is expected that a slurry
filled granular window will be the most difficult to detect of
the three types of windows discussed. However, this type of
window is also expected to have a relatively low hydraulic
conductivity, high stability, and thus the least detrimental
affect on the overall efficiency of the cutoff wall; except in
special cases. (If the window is filled with gravel size
material and is adjacent to a very coarse in-situ deposit, a high
gradient can remove the slurry via piping20> This will leave a
very high hydraulic conductivity window which totally breaches
the barrier.)
3.2.6.4 Bentonite Support Slurry Filled Window
The window in this case is composed of the bentonite support
slurry which has become entrapped by the advancing backfill. It
is hypothesized that this condition could result if the slurry is
allowed to become too heavy and is overridden rather than
displaced by the backfill.
A. Point Resistance
Inasmuch as the window is a fluid, point resistance is
expected to be zero. However, a small point resistance
may be measured due to the thixotropic behavior of the
bentonite slurry.
^uFor a definition and mechanisms leading to piping, the reader is
referred to Middlebrooks, (1953), and Cedergren, (1967).
3-28
-------
B. Local Friction
Local friction values within a fluid are also zero. A
small value may be measured due to viscous drag forces
if the slurry is very thick. Adhesion forces may also
develop due to thixotropic effects.
C. Excess Pore Pressure
Assuming the backfill, has totally consolidated and the
slurry window remains intact, the fluid pressure in the
window will be consistent with hydrostatic conditions.
Inasmuch as the slurry is a fluid, shear generated
excess pore pressure would be zero. Therefore, excess
pore pressure generation would be consistent with
volumetric displacement of the slurry as the probe
penetrates the window and therefore should be positive.
The magnitude of the excess pore pressure would be
dependent on the rate at which the "compressed" slurry
could permeate out of the window. Factors influencing
this behavior include surface area of the window and
thus its size, viscosity of the slurry, and hydraulic
conductivity of the sides of the window (either
soi1/bentonite backfill, slurry trench filter cake
and/or natural formation).
D. Relative Hydraulic Conductivity
The water take would typically be expected to increase
when penetrating a slurry filled window from an intact
backfill. The water introduced into the window would
pressurize the slurry and cause the slurry to permeate
into the backfill. While the viscosity of the slurry as
a permeant is greater than that of water, it is likely
that the increased surface area (of the window as
compared to permeameter stone in direct contact with
backfill) would more than offset this factor.
Based on the above discussion, it is expected that the most
definitive indicators of a slurry filled window would be severely
reduced point resistance and local shear as compared to intact
backfill.
3.3 Cutoff Wall Test Section
While the standard piezocone has been widely used for
stratigraphic investigation over the past 10 years, its technical
and economic feasibility has not been demonstrated for cutoff
3-29
-------
wall anomaly detection. To evaluate technical feasibility, a
full scale cutoff wall test section was constructed. The
objective of test section construction was to incorporate windows
of known type and location such that the response of the
instrument, when probing a known condition, could be evaluated.
3.3.1 Test Section Construction
The test section was constructed in the northwest corner of
the Gilson Road containment (see Figure 3-9) and was
approximately 3.5 feet wide by 24 feet long by 21 feet deep.
Construction followed the same methodology as used during
construction of the containment cutoff wall with the exception
that windows were purposely incorporated into the backfill.
3.3.1.1 Excavation
Prior to excavation, the bentonite support slurry was mixed
using Baroid National Premium sodium bentonite, a 90 barrel yield
Premium Grade^l bentonite (as specified for construction of the
Gilson Road containment wall). The mixing proportions were
7 percent bentonite in water, by weight. The water used was
drinking water quality spring water. The support slurry was
placed in a holding pond after mixing to allow complete hydration.
Marsh funnel viscosity and unit weight quality control data are
presented in Table 3-1.
21In accordance with testing standards developed by the American
Petroleum Industry, Method 13A and B (API, 1981 and API, 1982).
3-30
-------
TABLE 3-1
SUPPORT SLURRY QC DATA
Marsh Funnel Unit Weight
Sample #
1
2
3
4
Sample Location
SE Corner
NE Corner
NW Corner
SW Corner
Viscosity (sec.)
38
42
37
45
(PCF)
65.5
—
—
66.5
Notes
1
1
1
1
1. Sample taken from hydration pond just prior to excavation,
2. — indicates not tested.
3-31
-------
\
FI6. 3-9 LOCATION PLAN
I
3'
* — -
/\
«-*/ \
*sA
XV
\
\
\
is^oo
1
1
1
/
\. T* BTEST
X. Sj, | SECTION
^^X
*""V^ \ «
^^\ \
LYLE REED \ \ >• _
-------
Trench excavation was completed under slurry stabilization.
The trench remained open for only one hour between excavation and
backfilling. Therefore, full filter cake development was
unlikely.
3.3.1.2 Backfilling
The backfill was composed-of a mixture of the excavated sand
and gravel (trench spoil), imported sandy silt borrow, dry
powdered bentonite and support slurry in the same proportions
used for the Gilson Road containment. The mixture was blended
via bulldozer until it reached a uniform consistency, as
determined visually. Quality control testing data is presented
in Table 3-2.
TABLE 3-2
SOIL/BENTONITE BACKFILL QC DATA
Hydraulic4
Sample
No.
1
2
3
Slumpl
( inches)
5.5
5.5
5.0
Unit
Weight
(PCF)
123.2
124.0
125.0
Fines2
Content
(percent)
24
22
25
Bentonite^
Content
(percent)
2.8
2.6
2.4
Conduct-
ivity
(cm/sec)
1.7 x
2.0 x
2.0 x
10-7
10-7
10-7
Notes
5,6
5,7
5,8
1. ASTM Method C-143-78
2. Percent passing number 200 sieve
3. Determined using methylene blue titration
4. Determined using API Fixed Ring Test
5. Samples taken prior to backfill placement
6. Placed in bottom 3 foot of trench
7. Placed between 10 and 15 foot depth
8. Placed at top of trench
The backfill was placed in the trench one bucket at a time.
The backhoe lowered the filled bucket into the slurry to the top
of the backfill surface. The bucket was then slowly rotated to
deposit the backfill. This method was similar to backfill
placement at the beginning of actual containment construction
3-33
-------
(initial backfill placement until it rose above slurry surface).
However, it does not simulate the lateral movement characteristic
of production backfill placement during most of the Gilson Road
containment wall construction. The associated backfill placement
techniques could not be simulated on the small scale of the test
section wall.
3.3.1.3 Window Emplacement
Four different types of windows were emplaced in the test
section. These were: 1) a high density clean granular window,
2) a low density clean granular window, 3) a slurry filled
granular window, and 4) support slurry filled windows. The
design location and dimensions of these windows is depicted in
Figure 3-10. These dimensions were verified by measurements
along the trench axis with a weighted tape after placement of
each layer.
It is emphasized that the mechanisms leading to window
formation and their composition has not been previously
documented. The objective of this effort was therefore to
simulate windows which were judged likely to occur during actual
cutoff wall construction as based on previous field observations
and theoretical considerations. These windows were constructed
using the following techniques.
A. High density clean granular window (HDCG)
This window was meant to simulate the characteristics of
clean granular unexcavated material (Section 3.2.6).
The high density was required to be consistent with the
character of the in-situ formations encountered on-site.
This window was the easiest to create, inasmuch as the
unexcavated material below the bottom of the trench
matches this hypothesized type of window perfectly. The
materials comprising the window were in direct hydraulic
communication with the in-situ granular materials
on-site. The profile of the trench bottom and thus the
top of this window is shown on Figure 3-10.
B. Low Density Clean Granular Window (LDCG)
A low density counterpart to the window described above
was also emplaced in the test section. The objective in
this case was to simulate a window formed due to
sloughing of intact portions of the trench side wall
3-34
-------
FIG. 3-10 TEST SECTION CONSTRUCTION PROFILE
WEST
EAST
INTACT SOIL/BENTONITE
BACKFILL
HIGH DENSITY CLEAN
GRANULAR WINDOW
LOW DENSITY CLEAN
GRANULAR WINDOW
SLURRY FILLED
GRANULAR WINDOW
SUPPORT SLURRY
FILLED WINDOW
HORIZONTAL SCALE
SSa55HSiZ!—3
2' 4' 6' 6' 10'
NOTE: HIGH DENSITY CLEAN GRANULAR WINDOW IS MATERIAL IN WHICH
TRENCH WAS EXCAVATED.
3-35
-------
which retained the existing pore fluid but not the
existing density. This window would be hydraulically
isolated from the in-situ parent formation by the filter
cake.
The window was constructed after the trench was
partially filled with soil/bentonite backfill by 1)
placing concrete sand in the backhoe bucket, 2)
saturating the sand with water, 3) lowering it to the
surface of the in place backfill, and 4) rotating the
bucket to deposit the sand. This process was repeated
until a layer of material, as shown in Figure 3-10, was
created. Subsequent to sand placement, additional
soil/bentonite backfill was placed over this window
using the same placement technique.
C. Slurry Filled Granular Window (SFG)
The slurry filled granular window was meant to simulate
a window formed via the sedimentation of coarse material
out of the slurry (see Section 3.2.6 for further
description). In this case, concrete sand was dropped
into the top of the slurry filled trench and allowed to
settle to the top of the previously placed
soil/bentonite backfill. After an appropriate volume of
material was placed and allowed to settle, additional
soil/bentonite backfill was placed as described above.
It is also likely that a similar slurry filled window
incorporating the in-situ granular formation material
was, to varying extents, created at the bottom of the
trench. In all cases, a filter cake of bentonite slurry
would form at the trench bottom. However, this filter
cake would only be about 1/4-inch thick. A much thicker
slurry filled granular window would be created where the
backhoe teeth scarified the trench bottom during
excavation and thus mixed support slurry with those
deposits. Where this occurred, the window may be 6 to
12 inches thick in vertical dimension. The exact
locations of these areas along the bottom of the trench
were not known.
D. Support Slurry Filled Window (SSF)
The slurry filled window was installed after the test
section was completely backfilled. Although this format
does not chronologically correspond to that anticipated
during actual containment construction, it was the only
way window location could be controlled.
3-36
-------
Two slurry filled windows were constructed, one at each
end of the test section (see Figure 3-10). The
installation procedure consisted of driving a casing to
the depth of the window and then washing out the casing.
A side ported washing bit was then inserted into the
backfill below the casing. Slurry was pumped out of the
bit under high pressure to erode the intact backfill.
The slurry, carrying the cuttings, was brought to the
surface through the annular space between the jetting
tool and the casing. The slurry was recirculated
through a sedimentation tub and the cuttings removed.
The cuttings consisted primarily of coarse to fine sand
and silt with about 10 percent fine gravel. The
gravelly backfill fractions could not be suspended in
the return slurry and were thus left at the bottom of
the cavity.
The vertical thickness of the windows ranged between 12
and 18 inches as determined by measurement below the
casing. The cavities were expanded until no further
cuttings were removed by the slurry. The lateral extent
of the cavities, as shown on Figure 3-10, was estimated
by evaluating the erosional capabilities of the slurry
jet in the backfill at the surface of the test section.
3.3.2 Test Section Window Verification
As previously stated, the objective of the test section was
to provide a known condition with which to evaluate piezocone
response. As such, numerous measurements were taken before and
after each lift of material was placed. These data were used to
generate Figure 3-10. However, once a lift was placed, there was
no method available to ensure that it remained in position as the
subsequent lifts were placed on top. Displacement of previously
placed lifts was possible due to:
0 Inability of the backhoe operator to "feel" the material
as he rotated the bucket to add new material.
0 Low unit weight, low stability materials (slurry filled
granular window) placed below higher unit weight
backfill.
0 Unknown long-term stability of the window (support
slurry filled window).
Ideally, an independent verification method would be used to
document actual in-place conditions in the test section after it
3-37
-------
had stabilized. Such verification methods would include
continuous undisturbed sampling at close spacings along the
trench alignment or complete excavation. While these methods are
feasible if performed after piezocone sounding, the associated
costs placed them outside the scope of this project.
To the extent that the piezocone response to intact backfill
and windows is unknown, the sounding data cannot be used to
verify window location. In fact, it is the piezocone response
which was being investigated. However, while piezocones have not
previously been employed to investigate soil/bentonite backfilled
cutoff walls, theoretical considerations and empirical data from
other applications provided insight into expected response
(Section 3.2.6). The piezocone data was therefore used to verify
window location as based on expected trends.
The piezocone sounding data was analyzed, correlated, and
then used to construct a post-construction, as-built test section
profile. (The individual soundings and window types are
discussed in detail in Section 3.3.3.) This profile, presented
in Figure 3-11, reveals some significant discrepancies when
compared to the profile developed from construction measurements
(Figure 3-10). These discrepancies are discussed individually in
the following sections along with potential explanations for the
deviations.
3.3.2.1 High Density Clean Granular Window
The high density clean granular, H.D.C.G., window was
simulated by the unexcavated materials below the trench bottom.
As such, the backfill/window interface was the trench bottom.
The only discrepancy found in this case, existed at the extreme
west end of the trench. It appears that the initially vertical
west end of the trench may have sloughed off or been dragged down
by the backhoe bucket. The material thus raised the elevation of
the trench bottom. This hypothesis is supported by point
resistance data from soundings GZ-9 and GZ-10. In both cases,
the magnitude of point resistance nearly doubles at depths
corresponding to the predicted trench bottom. Another plausible
explanation is that the trench was not excavated to the full
depth. However, this is not supported by the measurements taken
during construction.
3-38
-------
FIG. 3-11 POST-CONSTRUCTION TEST SECTION PROFILE
INTACT SOIL/BENTONITE
BACKFILL
HIGH DENSITY CLEAN
GRANULAR WINDOW
LOW DENSITY CLEAN
GRANULAR WINDOW
JPgpp$ POSSIBLE LOW DENSITY
MJS&&I CLEAN GRANULAR WINDC
WINDOW
SLURRY FILLED GRANULAR
WINDOW
0'
HORIZONTAL SCALE
2' 4' 6' 8'
10'
NOTE I. HIGH DENSITY CLEAN GRANULAR WINDOW IS MATERIAL IN WHICH TRENCH
WAS EXCAVATED.
2. * INDICATES NO DATA
3-39
-------
3.3.2.2 Low Density Clean Granular Window
The low density clean granular, L.D.C.G., window appears to
have been successfully constructed in general accordance with
Figure 3-10. The top of the window is located at the correct
elevation. However, it is somewhat thicker than anticipated over
the center and eastern portions of the trench and nonexistent at
the extreme west end of the trench. It is likely that the
material slid down the apparently steep slope at the west end of
the trench (Section 3.3.2.1) and increased the window thickness
over the rest of the trench.
3.3.2.3 Slurry Filled Granular Window
The piezocone soundings did not detect a consistent layer
corresponding to the slurry filled granular, S.F.G. window
anywhere within the trench profile. The data indicated that the
window material probably exists as discrete pockets located in a
random manner.
It is hypothesized that the layer of window material was
displaced when the backfill was placed on top of it. The S.F.G.
window would be more prone to instability than the L.D.C.G.
window because 1) it had a lower unit weight than the L.D.C.G.
window (probably also lower than the backfill) and 2) the
material would behave in a more cohesive manner due to the
bentonite slurry in the voids. The second property may have
allowed the material to behave as a viscous liquid during
placement of subsequent backfill lifts.
3.3.2.4 Support Slurry Filled Window
Behavior of undisturbed soil/bentonite samples from the
completed Gilson Road containment wall indicated that the
material exhibited significant strength. This was further
documented during post-construction excavation into the top 5
feet of the containment wall. In this case, the backfill would
easily support a person's weight. Based on these observations,
it was judged that a support slurry filled, S.S.F., window could
exist in the containment wall and thus could probably be
constructed in the test section. As described in Section
3.3.1.3, two such windows were constructed in the test section.
Measurements made during window installation verified that the
slurry filled cavities were, in fact, created.
3-40
-------
Upon completion of the soundings, however, it was evident
that the two S.S.F. windows were not detected in the trench
profile at all. These windows would have been easily
differentiated from other materials in that fluids provide no
point resistance or local friction. The complete absence of
these windows elicited additional laboratory work to arrive at a
plausible explanation.
Laboratory testing indicated that the soil/bentonite
backfill does in fact exhibit significant strength in the
undrained condition22. However, as drained conditions are
approached over time, the backfill loses most of its cohesive
strength. The drained cohesive strength of the backfill does not
appear high enough to support a cavity in the wall via arching.
It is therefore probable that after the cavity was created, it
remained stable for a short time as supported by the relatively
high undrained strength. With time, the backfill approached
drained conditions with a resulting loss in strength. The top of
the cavity then progressively failed until the original cavity
became filled. The cavity thus propagated towards the surface of
the trench. Evidence from sounding GZ-5 indicates that the
failure zone may never have reached the surface. An anomalous
zone was detected extending from 5 to 8 feet deep. In this zone,
the friction ratio was very low for intact backfill. In fact, it
corresponds to that indicative of a L.D.C.G window. However, no
significant deviation in point resistance was obtained as would
be the case for a L.D.C.G. window (See Section 3.3.3.3). This
response may indicate the presence of a very coarse and very low
density material as would be anticipated based on the revealing
scenario.
In summary, laboratory testing of the strength properties of
the granular soil/bentoinite backfill and field measurements
taken during construction indicate that the support slurry filled
windows were created during test section construction. However,
they would be expected to be a transient phenomenon. It is
therefore likely that they did not exist intact at the time the
soundings were executed.
22Cohesive soils behave differently depending on the rate of load
application as compared to the rate at which the pore pressure
within the soil can change in response to the loading. The two
extremes in this behavior are referred to as undrained (rapid
loading) and drained (slow loading). Additional discussion can
be found in Lambe 1969, pg. 423.
3-41
-------
3.3.3 Test Section Soundings
Prior to sounding, the test section was allowed to stabilize
for a period of three months after its construction. In this
way, full consolidation of the backfill materials under their own
weight was ensured. In addition, it was anticipated that creep
effects were sufficiently accounted for via this time period23.
As such, the test section should adequately model the conditions
existing in the two year old Gilson Road containment wall.
Ten soundings were conducted in the test section spaced at
approximately 2 foot intervals along the trench alignment (see
Figure 3-11). The soundings were executed in two phases. The
objective of the first phase, consisting of soundings GZ-1 to
GZ-6 , was to establish baseline data for the three standard
piezocone indices; point resistance, local friction and excess
pore pressure. The second phase objective was to investigate the
behavior of the relative hydraulic conductivity index during the
remaining four soundings; GZ-1 to GZ-10 inclusive. The data
obtained for each sounding is included herein as Figures 3-12
through 3-21.
Before the index values could be established for the intact
backfill and windows, the location of these materials within the
test section needed to be established. Depth measurements alone
were expected to provide sufficient information for this task as
based on the construction profile (Figure 3-10). However, the
discrepancies encountered between the actual as-built profile
(sounding data) and the anticipated construction profile
complicated this process. As discussed in Section 3.3.2, the
test section did not provide the expected known condition
pursuant to window location. Therefore, window locations were
determined based on the sounding data and the expected
qualitative piezocone response for the individual window
materials. The piezocone response values were then quantified
once window locations were established.
The following subsections are organized to first present
data for piezocone response in the intact backfill followed by
evaluation of window induced deviations from the backfill indices.
The data from each sounding was correlated into a profile view of
the test section which summarizes detected window locations
(Figure 3-11). The variation between the window locations
detected by the piezocone and the initial placement of the
windows during test section construction was previously discussed
in Section 3.3.2.
effects following consolidation typically follow an
exponential decay function. See Lambe 1969, pg. 419.
3-42
-------
SOUNDING DEPTH, FEET
* • u
ro
-o
ui
.' a o
g
R
9
Li "
-------
FIG.3-13 PIEZOCONE TEST—SOUNDING NO. GZ2
POINT 5TRESS-OC .K6/CH1 LOCflL FRiaiON-F ,KG/CH" EXCESS PORE PRESS. KG/CM1 FRICTION RRTIO-FR ,1
H-FBCTOR ,1
MTER FLOV OC
CO
I
-------
Sfr-E
SOUNDING DEPTH, FEET
•?•••-I-•••?
'a
3
1
a n
-------
SOUNDING DEPTH, FEET
•*. • -iT- • • '?
m
,r
-------
FI6.3-I6 PIEZOCONE TEST—SOUNDING NO. GZ5
POINT STRESS-K ,«C6/CMJ LOCH. FRiaiON-F ,K£/QP ', EXCESS PORE PRESS. KG/CM1 FRICTION RDTIO-FR .X
» • i -.1 I .1 I U
WTER FLOW CM1
U)
I
X
I—
Q_
a'
in
-------
FK5.3-I7 PIEZOCONE TEST—SOUNDING NO. GZ6
U)
I
CO
POINT STRESS-QC .KB/CM1 LOCAL FRICTION-F .KG/CM1 EXCESS PORE PRESS. K6/CM3 FRICTION RRTIO-FR ,1
B-FRCTOR .X
WTER FLOW CM1
IDS-
,
I
-------
LJ
I
FIG.3-18 PIEZOCONE TEST—SOUNDING NO. GZ7
POINT STRESS-OC ,KS/Ot' LOCAL FRICTIOH-F .K/Of EXCESS PORE PRESS. Kf/Of FRiaiON RATIO-FR .X
It I 3 -.< I .1
R-FRCTOR «Z
VOTER FLOW CM*
-------
os-e
SOUNDING DEPTH, FEET
a*
.5s
-------
FIG.3-20 PIEZOCONE TEST—SOUNDING NO. GZ9
POINT STRESS-OC .KG/CM1 LOOL FRiaiON-F .KE/CM1 EXCESS PORE PRESS. KG/CM9 FRICTION RRTIO-FR .X
fl-FBCTOR
VOTER FLOW CM'
I
l/l
-------
ss-e
SOUNDING DEPTH, FEET
3
5
r "D
3 R
as
o
z
in
z
o
3
3
-------
3.3.3.1 Intact Backfill Analysis
The first step in analysis of the sounding data encompassed
identification of zones of intact backfill. Zones of intact
backfill were identified via visual inspection of the sounding
data as based on the following guidelines:
0 Depth of the sounding as compared to the anticipated
placement of materials in the test section (Figure
3-10).
0 Consistency of the data with trends expected based on
theoretical considerations (Section 3.2.6.I)/ and
0 General uniformity of the data over a given zone.
The locations of intact backfill zones are summarized in
Figure 3-11. Piezocone response in the backfill was then
evaluated and average response values for the individual indices
parameters of point resistance, local friction, excess pore
pressure, and relative hydraulic conductivity were established.
Average response values were determined via linear
interpolation performed directly on the graphs. These data are
summarized in Tables 3-3 and 3-4. Deviations from these values
formed the basis for subsequent window identification.
3-53
-------
Notes: 1.
2.
3.
TABLE 3-3
INTACT BACKFILL
Sounding
Number
GZ-1
GZ-2
GZ-3
GZ-4
GZ-5
GZ-6
GZ-7
GZ-8
GZ-9
GZ-10
Average
Average
Point
Resistance
(kg/cm2)
1.3
1.2
2.0
1.3
1.3
1.2
1.0
1.3
2.8
2.1
1.5
Average
Frictional
Ratio
(%)
2.6
2.8
3.1
2.9
1.8
1.5
2.8
2.0
1.1
0.9
2.2
Excess
Pore
Pressure
(kg/cm2)
+0.2
0
—
0
+0.2
+0.1
—
—
0
0
+0.1
Notes
1
Abnormally low friction ratio (0.3) encountered above
zone of attempted support slurry window. This value
omitted in average for GZ-5.
Highest and lowest values omitted from average.
— indicates data not taken.
3-54
-------
TABLE 3-4
RELATIVE HYDRAULIC CONDUCTIVITY INDEX
Material
Type
Intact
Intact
Intact
Intact
Intact
Intact
Intact
I ntact
Intact
Intact
Intact
Intact
I nt'act
Weighted
H.D.C.G.
H.D.C.G.
H.D.C.G.
H.D.C.G.
Weighted
L.D.C.G.
L.D.C.G.
L.D.C.G.
L.D.C.G.
Weighted
S.F.G.
Weiahted
Sounding
Number
GZ-7
GZ-7
GZ-7
GZ-7
GZ-8
GZ-8
GZ-8
GZ-8
GZ-9
GZ-9
GZ-9
GZ-9
GZ-9
Average
GZ-9
GZ-9
GZ-9
GZ-9
Average
GZ-7
GZ-8
GZ-8
GZ-8
Average
GZ-8
Average
Depth
(ft.)
0.
4.
8.
12.
1.
7.
9.
11.
0.
3.
6.
8.
12.
13.
14.
16.
18.
15.
15.
15.
17.
2.
5- 2
0- 7
4-11
0-13
3- 2
1- 9
8-11
2-14
0- 3
6- 6
2- 7
0-12
8-13
8-14
9-15
4-17
4-19
4-16
8-15
9-16
0-17
7- 3
.3
.4
.4
.0
.8
.7
.0
.0
.0
.2
.9
.0
.8
.7
.8
.8
.1
.4
.9
.8
.6
.3
Average
Water
Takel
cm3/ft Notes
4.
10.
15.
29.
2.
23.
20.
18.
8.
9.
2.
8.
11.
12.
111.
71.
62.
52.
74.
90.
166.
125.
166.
123.
71.
71.
9
1
9
4
9
3
4
9
2
0
6
4
8
2
1
4
5
6
0
9
7
0
7
1
4
4
3-55
-------
Notes: 1. Defined as the volume of flow into the formation per
unit depth penetrated.
2. Weighted average computed based on material
thickness.
3. 2.5 psi driving pressure (above hydrostatic) used for
testing.
4. H.D.C.G. - High Density Clean Granular Window
5. L.D.C.G. - Low Density Clean Granular Window
6. S.F.G. - Slurry Filled Granular Window
A. Point Resistance
In general, point resistance in the intact backfill was
nearly constant with depth for a given sounding. The
only exception was sounding GZ-3 (Figure 3-14) where
point resistance increased somewhat with depth.
The magnitude of average point resistance from sounding
to sounding ranged from 1.0 kg/cm2 to 2.8 kg/cm2. When
the individual averages for each sounding were again
averaged, the value obtained for point resistance in the
backfill was 1.5 kg/cm2. This value fell below the
range of point resistance given on empirical
correlations (see Figures 3-5 and 3-6). As such, the
very low point resistance indicated that the backfill
was very soft.
B. Local Friction
The values of local friction measured in the intact
backfill were very low, typically less than 0.05 kg/cm2.
As can be seen on the figures, the standard software
scaling factors could not provide adequate resolution
for measurements in the intact backfill while still
remaining within scale for the entire sounding.
However, the graph of friction ratio did provide
adequate resolution.
The magnitude of average friction ratio for each of the
ten soundings ranged from 0.9 percent to 3.1 percent.
The average friction ratio for the ten soundings
combined was 2.2 percent. The very low value of point
resistance (1.5 kg/cm2) combined with the friction ratio
indicated that, although the backfill was primarily
granular (by dry weight), it behaved as a plastic
material due to the bentonite. This is demonstrated on
Figure 3-6 which indicates that the backfill is a very
soft clayey sand/silt.
3-56
-------
C. Excess- Pore Pressure
The excess pore pressures generated while sounding the
backfill were small in comparison to the operational
range of the transducer incorporated in the piezocone (7
kg/cm^). In particular, the data magnitude was small
when compared to possible errors in establishing the
zero point for the transducer. As such, the data was
interpreted in a qualitative sense only.
The soundings yielded excess pore pressures that
averaged near zero or slightly positive (up to +0.2
kg/cm2)e Spikes in the data ranged from a +0.6 to a
-0.2 kg/cm^. Based on theoretical considerations, it is
unlikely that the intact backfill should yield negative
excess pore pressures (see Section 3.2.6.1). It is
therefore probable that either the value used for
hydrostatic pore pressure24 was too high or the
transducer zero stability was not adequate due to its
high range. However, if an average excess pore pressure
of +0.1 to +0.2 kg/cm2 is used for the backfill, then
the data indicate that the material is a clayey sand
(see Figure 3-7).
D. Relative Hydraulic Conductivity
The permeameter equipment was employed to obtain water
take values in the backfill as an index of relative
permeability. This equipment was used for soundings
GZ-7 through GZ-10, only. The graph of water take vs.
depth for each sounding was divided into regions where
the intact backfill exhibited a uniform rate of flow.
In all cases, data from the upper 12 inches of the test
section soundings was disregarded. This procedure was
followed in recognition of the preferential boundary
flow conditions which may have existed between the
piezocone drill string and the backfill25> The water
^4Hydrostatic pore pressure is subtracted from the total pore
pressure measured by the transducer to compute excess pore
pressure.
25gpon insertion, it is likely that a preferential flow channel
would form along the drill string due to boundary effects.
Higher water takes would therefore be expected due to the
communication of this flow channel with the surface. This effect
can be seen in soundings GZ-7 and GZ-8. Based on this data, a
12 inch zone was established where the values obtained were
considered inaccurate.
3-57
-------
take value for each zone was then established by visual
linearization of the curve. This data is summarized on
Table 3-4. The water take values ranged from 2.6 cm3/ft
to 29.4 cm^/ft. A weighted average (based on zone
thickness) of 12.2 cm3/ft was computed when all the
intact backfill was considered.
3.3.3.2 High Density Clean Granular Window
The high density granular window (H.D.C.G.) was simulated by
the unexcavated sand and gravel formation remaining below the
trench bottom. The data (summarized in Tables 3-4 and 3-5)
indicatedthat point resistance and relative hydraulic
conductivity were the most useful parameters for detecting this
type of window. The high value of local friction and the low
friction ratio also allowed discrimination between the window and
the backfill. The response was less abrupt, however. This was
probably due to the 20 cm length of the friction sleeve. Excess
pore pressure, although qualitatively useful as corroboratory
data, provided little quantitative identification. It is
expected that a more sensitive transducer (lower operational
range) would provide better performance. The locations of
H.D.C.G. windows within the test section are summarized in Figure
3-11. The following subsections treat these parameters in more
detail on an individual basis.
3-58
-------
TABLE 3-5
HIGH DENSITY CLEAN GRANULAR WINDOW
Sounding
Number
GZ-1
GZ-2
GZ-3
GZ-4
GZ-5
GZ-6
GZ-7
GZ-8
GZ-9
GZ-10
Average^
Depth
(Ft.)
19.0-23.2
19.0-20.5
18.8-20.0
20.0-21.0
19.0-21.0
19.6-20.1
—
18.8-19.4
13.8-21.2
11.3-20.0
Point
Resist-
ance^
kg /cm ^
100
45
65
65
90
100
—
40
90
100
69
Excess
Frictional Pore
Ratio
(%)
1.0
0.6
1.0
1.0
0.6
—
—
—
0.7
1.0
0.9
Pressure Water
(kg/cm^) Take
+0.1
-0.3
_ —
-0.2
-0.1
-0.3
_ —
—
-0 . 3 Yes
-0.2
-0.1 to
-0.2
Notes
1,2,3
1,2
Notes: 1. Friction sleeve did not penetrate window.
2. Permeameter did not penetrate window.
3. Sounding stopped before window.
4. Average computed with omission of highest and lowest
values.
5. Point resistance greater than 10 kg/cm2 is off-scale
on figures. Peak data taken from digital output.
6. — indicates data not taken.
A. Point Resistance
Point resistance provided a clear indicator when this
window was encountered. The values of point resistance
ranged from 40 kg/cm2 to over 100 kg/cm2. The lower end
of this range is considered artificially reduced in that
penetration was typically terminated before full point
resistance was reached so as to protect the transducer
from an overload condition. The average value of point
resistance for the H.D.C.G. window was 69 kg/cm^ as
compared to 1.5 kg/cm2 for the intact backfill.
3-59
-------
B. Local Friction
Local friction values for the H.D.C.G. window can only
be established in cases where the entire friction sleeve
penetrated the window. This required a point
penetration distance of nearly 1 foot. Therefore,
soundings GZ-6, GZ-7, and GZ-8 did not provide local
friction data. The average value of local friction
obtained from the remaining soundings (0.44 kg/cm2) was
significantly above that yielded by the intact backfill
«0.05 kg/cm*).
Friction ratio values were also computed for the
H.D.C.G. window. They ranged from 0.6 percent to 1.0
percent with an average 0.9 percent. This value is less
than one-half that exhibited by the intact backfill
(average = 2.2 percent) and, along with point
resistance, is indicative of a dense, coarse grained
noncohesive material (Figure 3-6). This
classification,based on piezocone data, agrees well with
the visual identification of the material.
C. Excess Pore Pressure
The magnitude of excess pore pressure generated when an
H.D.C.G. window was encountered was small as compared to
the transducer range. In general/ however, they were
negative as would be expected for a dense noncohesive
material. The sounding data indicated that a reasonable
average value was -0.1 to -0.2 kg/cm2. This was best
exemplified by soundings GZ-9 and GZ-10 where over 7
feet of the window were penetrated. This value was
further corroborated by the other soundings with lesser
penetration distances. However, GZ-1, which penetrated
over 4 feet into the window, yielded positive excess
pore pressures. This anomalous behavior could not be
explained based on the available data. A value of -0.1
to -0.2 kg/cm2 for excess pore pressure and a point
resistance of 69 kg/cm2, indicated that the material is
noncohesive as based on Figure 3-7. Again this matches
the visual classification.
From the standpoint of window detection, it is also
useful to examine at the change in excess pore pressure
as compared to that for intact backfill. While the
average value for the intact backfill (+0.1 kg/cm2) was
not greatly different from the average window generated
value (-0.1 to -0.2 kg/cm2)* the differences for each
sounding were more significant. The typical response
was a significant decrease in excess pore pressure at
3-60
-------
the interface between the backfill and the window. On
average, this change amounted to -0.4 to -0.5 kg/cm^ in
the cases where the window was located immediately below
the backfill. Other cases were less definitive due to
intervening windows of different types.
D. Relative Hydraulic Conductivity
The permeameter stone penetrated the H.D.C.G. window in
only one sounding. Penetration into the window required
over 2 feet of point penetration due to permeameter
location on the instrument. In general, significant
point penetration of the window was not possible due to
limited point resistance transducer capacity.
As shown on sounding GZ-9, the hydraulic conductivity
index increased significantly upon encountering the
window. Four separate determinations of water take were
made. As summarized on Table 3-4, the average value was
74.0 cm3/ft. This is significantly above the average
value for intact backfill of 12.2 cm3/ft.
3.3.3.3 Low Density Clean Granular Window
As for the H.D.C.G. window, point resistance and relative
hydraulic conductivity were the most dramatic parameters with
respect to change when encountering a low density clean granular
window. However, point resistance was not as definitive as for
the H.D.C.G. window due to the lower density and lack of gravel
size constituents. While change in friction ratio provided good
correlation with point resistance determined window location,
local friction values were too low for discrimination on the
graphs. Excess pore pressure was useful only in a qualitative
sense, as was the case for the H.D.C.G. window.
The data for the above parameters are summarized in Tables
3-4 and 3-6. The locations of the L.D.C.G. windows are shown in
Figure 3-11. The following subsections discuss each parameter in
more detail.
3-61
-------
TABLE 3-6
LOW DENSITY CLEAN GRANULAR WINDOW
Sounding Depth
Number (Ft.)
Excess
Point Friction Pore
Resistance Ratio Pressure Water
(kg/cm2) (%> (kg/cm2) Take Notes
GZ-1
GZ-1
GZ-2
GZ-3
GZ-4
GZ-4
GZ-4
GZ-5
GZ-5
GZ-6
GZ-7
GZ-7
GZ-8
GZ-9
GZ-9
GZ-9
GZ-9
GZ-10
Average
Notes :
14.7-16.2 16
16.2-17.8 4
15.3-16.2 8
14.5-16.8 17
4.4- 5.0 4
13.6-14.3 7
15.2-16.5 11
10.7-11.2 4
15.5-16.1 4
14.3-18.5 16
1.0- 1.5 5
15.4-16.8 16
15.8-17.7 19
2.3-2.8 5
4.0- 4.5 4
5.2-5.8 4
12.9-13.4 4
3.0- 4.0 6
10.3
0.4 -.1
1.0 -.1 — 1
0.2 -.2
0.7
0.9 -.3 ~ 1
0.4 -.4
0.3 - . 2
0.4 -.1 — 1
<0.2 +.4 — 1
<0 . 2 - . 2
0.9 — No 1,3
0.4 — Yes
0.5 — Yes
0.4 — No . 1,3
1.2 ~ No 1,2,3
1.2 ~ No 1/2,3
1.1 -- — 1,2,3
0.6 — ~ 1,2
0.4 0. to
-0.2
1. Potential window - does not meet point resistance
criteria, but point resistance significantly above
intact backfill.
2. Window ruled out based
3. Window ruled out based
on friction ratio criteria.
on water take criteria.
4. Average computed with omission of highest and lowest
values. Windows ruled out based on notes 2) and 3)
above also omitted.
5. — indicates data not
taken.
A. Point Resistance
Due to the lower density and lack of gravel sizes in
this window, point resistance was not expected to be as
high as for the H.D.C.G. window. Guidelines were
3-62
-------
therefore established to discriminate between a
potential window and the intact backfill as based on
point resistance alone. Final determination of window
existence was then based on confirmatory data from the
other parameters.
Point resistance magnitudes greater than 7 kg/cm2 were
considered indicative of L.D.C.G. window material. This
boundary was established, via inspection of the sounding
data, at five times the magnitude previously established
for intact backfill. Values of point resistance which
were less than 7 kg/cm2, but were still significantly
greater than the response in the backfill, were also
considered subject to confirmatory data from the other
piezocone parameters.
As shown in Table 3-6, the average point resistance was
10.3 kg/cm2. This value incorporated data from windows
with point resistances less than the cutoff point of 7
kg/cm2, but which could not be ruled out based on other
piezocone parameters. This average is thus considered a
lower bound average but was used in recognition that the
7 kg/cm2 cutoff point was somewhat arbitrary. The lower
bound average of 10.3 kg/cm2 is still significantly
above that generated by the intact backfill. The abrupt
development of these relatively high point resistances,
as compared to intact backfill, provided significant
confidence that the L.D.C.G. windows locations could be
identified.
B. Local Friction
The values of local friction measured in the L.D.C.G.
windows were quite low, typically less than 0.1 kg/cm2.
At this magnitude of response, the difference .between
the backfill and the window could not be resolved on the
graphs. However, the friction ratio provided a better
index of window presence. Average friction ratio for
the L.D.C.G. window was 0.4 percent as compared to 2.2
percent for the intact backfill. A friction ratio upper
limit of 1.1 percent (one-half the average intact value)
was established as a guideline for L.D.C.G.
windowidentification.
The average value of friction ratio, for the L.D.C.G.
window combined with an average point resistance of 10.3
kg/cm2, indicated that the material was a loose
noncohesive soil type(Figures 3-5 and 3-6). This
description, from the empirical correlations, agrees
3-63
-------
well with the actual material type placed as the
windows.
C. Excess Pore Pressure
The excess pore pressures generated while penetrating
the L.D.C.G. window materials were typically slightly
negative; average magnitudes ranged between -0.1 to -0.2
kg/cm2 as was the case for the H.D.C.G. window. It
would not be expected that sands existing at low
relative densities would produce negative excess pore
pressures. In fact, very loose sands, such as would be
probable for the L.D.C.G. window (deposited in a
submerged condition), would be expected to generate
slightly positive or zero excess pore pressures. The
low magnitude of the data, as compared to the transducer
operating range, may have introduced a zero shift error.
This hypothesis would result in zero excess pore
pressure which would be more consistent with theoretical
considerations and past empirical data.
Independent of the actual magnitude of excess pore
pressure generated, the change in this parameter at the
transition from the backfill to a L.D.C.G. window
provided useful window identification data. In all
cases, the trend was for an abrupt decrease in excess
pore pressure when entering the window as would be
expected. The average decrease was 0.3 kg/cm2.
The actual values of negative excess pore pressure
generated during the soundings indicated that the
material fell outside the range of the empirical
material identification chart (Figure 3-7). However, if
the excess pore pressures are adjusted slightly in the
positive direction, to account for the hypothesized zero
shift, then the empirical correlation correctly
identifies the sandy composition of the window. Even
without adjustment, the correlation indicates that the
material is likely to be noncohesive.
D. Relative Hydraulic Conductivity
The permeameter was operational during penetration of
two windows satisfying the point resistance and friction
ratio guidelines previously established to identify
L.D.C.G. material. In both cases, the relative
hydraulic conductivity index increased markedly. Four
separate determinations of water take were made in those
two windows as summarized on Table 3-4. The water take
3-64
-------
values ranged from 90.9 to 166.7 cm3/ft. with a weighted
average of 123.1 cm3/ft.
The average value of relative hydraulic conductivity
obtained for the L.D.C.G. windows were an order of
magnitude greater than that representing the backfill
(12.2 cm^/ft.). As such, identification of these more
permeable zones was straight forward. The relative
hydraulic conductivity index was therefore also used to
rule out marginal spikes in point resistance from
further consideration as potential windows.
3.3.3.4 Slurry Filled Granular Window
The slurry filled granular windows were most easily
identified based on the combination of high point resistance and
high local friction, as compared to the intact backfill. The
relative hydraulic conductivity index increased markedly in the
S.F.G. window, but only one determination of its magnitude was
made. Excess pore pressures typically exhibited negative trends
but the magnitudes were low compared to the operating range of
the transducer. Excess pore pressure was thus used as a
qualitative index only.
The individual sounding data for the S.F.G. windows is
summarized on Table 3-7. The locations of the windows are
presented on Figure 3-11. The response of the individual
piezocone indices is described in further detail in the following
sections.
3-65
-------
TABLE 3-7
SLURRY FILLED GRANULAR WINDOW
Sounding Depth
Number (Ft.)
Excess
Point Friction Pore
Resistance Ratio Pressure Water
(kg/cm2) (%) (kg/cm2) Take Notes
GZ-1
GZ-1
GZ-2
GZ-2
GZ-2
GZ-4
GZ-5
GZ-6
GZ-6
GZ-8
GZ-8
GZ-8
GZ-9
GZ-10
Average*
10.8-12.4
18.2-19.0
4.3- 5.3
5.7- 6.0
7.5- 8.7
3.4- 4.0
3.4- 3.9
3.7- 4.8
18.5-19.5
2.7- 3.6
4.7- 5.2
17.9-18.6
11.3-12.5
10.2-11.1
25
12
5
4
6
10
5
17
18
7
5
30
10
10
11.4
>10
>10
>10
1.5
7
>10
>10
>10
6
8
7
>10
5
9
7.9
-.3
0
-.2
-.1
-.1
-.1
-.5
-.2
-.2
—
—
—
-.4
-.3
-0.2
3
1
1,2
1
—
1
—
3
Yes
1
3
—
3
Notes: 1. Potential window - does not meet point resistance
criteria.
2. Window ruled out based on friction ratio criteria.
3. Filter cake and/or formation soil mixed with slurry
by bucket teeth at trench bottom.
4. Average computed with omission of highest and lowest
values. Windows ruled out based on Note 2) above
also omitted.
5. — indicates data not taken.
A. Point Resistance
The point resistance encountered when penetrating the
S.F.G. window was similar to that for the L.D.C.G.
window. The average value computed was 11.4 kg/cm2 as
compared to 10.3 kg/cm2 for the L.D.C.G. window. The
range of point resistance obtained from the individual
windows was also comparable. This result was expected
based on the similar composition and depositional
3-66
-------
methods used to construct the windows. The same point
resistance guidelines established to identify L.D.C.G.
windows were therefore also adopted for this window
type.
B. Local Friction
The magnitude of local friction generated when
penetrating the S.F.G. window was often greater than 0.1
kg/cm2 with an average value of 0.2 kg/cm2. This value
was high as compared to that for intact backfill and the
L.D.C.G. window. The high average local friction for
the S.F.G. windows did not produce friction ratios
significantly greater than the intact backfill at the
point of maximum point resistance. In fact, the average
point resistance and average local friction values
combine to yield an average friction ratio of 1.8
percent. This is identical to that for the intact
backfill (2.2 percent) to within the accuracy of the
data. However, high friction ratios were typically
detected immediately above and below the peak in point
resistance. This behavior was not evident for any of
the other test section materials. It is hypothesized
that the point resistance, when initially entering a
S.F.G. window, was low due to lack of confinement by the
soft, plastic backfill. This would also be true when
the point approaches the bottom of the window26. The
local friction, however, would be high over the entire
window thickness due to the adhesion attributed to the
bentonite filling the soil voids. Therefore, friction
ratios would be high entering and exiting the S.F.G.
windows.
A friction ratio of greater than 5 percent was required
for high point resistance to be considered a S.F.G.
window. The friction ratio reported for this window was
the highest value recorded when the window was entered
and exited. The thickness of the window summarized on
Table 3-7 also reflected the occurrence of high friction
ratio in the context of the above hypothesis.
The above data yielded an average friction ratio of
7.9 percent for the top and bottom of the window and an
average value of 1.8 percent at the location of maximum
point resistance. An average friction ratio of 7.9
percent, in conjunction with an average point resistance
of 11.4 kg/cm2, indicated that the window material is a
hypothesis is supported by the behavior of end bearing piles
in multilayered systems (Poulos, 1980).
3-67
-------
mixed soil type as per Figure 3-6. The lower average
friction ratio (1.8 kg/cm^) yields a loose sand material
identification. Both classifications are supported by
the material type used to construct the window. The
important point, however, is that a S.F.G. window yields
a consistent piezocone signature and thus can be
identified.
C. Excess Pore Pressure
Excess pore pressures generated when sounding a S.F.G.
were similar to those representing the L.D.C.G. window,
but somewhat more negative. The average value was -0.2
kg/cm^. The change in excess pore pressure at the
transition from the backfill to the window was also
typically abrupt and in the negative direction. The
average change was -0.5 kg/cm^. Again, the magnitude of
the change was greater for the S.F.G. window than the
L.D.C.G. window.
General discussion of the excess pore pressure
generation behavior in loose sand materials applies to
the S.F.G. window as well as the L.D.C.G. window. This
discussion is found in Section 3.3.3.3. An explanation
for the greater negative response in the S.F.G. is not
possible given the available data.
D. Relative Hydraulic Conductivity
The permeameter penetrated only one slurry filled
granular window. Upon entering the window, from the
intact backfill above, the water take abruptly increased
from 2.9 cm3/ft. to 74 cm3/ft. As shown in the GZ-8
sounding data (Figure 3-19), the water take maintained a
constant rate until the permeameter pressurization
reservoir emptied and required refilling.
The water take obtained in the S.F.G. window was high as
compared to the intact backfill as would be expected.
However, it was also approximately 60 percent of the
value obtained for the L.D.C.G. window. Given that both
windows were constructed of the same soil material
(concrete sand), the difference in water take can be
attributed to two differing conditions. The first
concerns pore fluid. The S.F.G. window contains
bentonite slurry in the void spaces which is about 40
3-68
-------
percent more viscous than clean water27^ qi^e second
condition relates to window size. As previously
indicated, the water take for a window totally entrapped
in intact backfill will be related to the flux through
the window/backfill interface (Section 3.2.6.3). The
available data, as summarized on Figure 3-11, indicated
that the L.D.C.G. window was probably significantly
larger than the S.F.G. window. Both of the above
conditions should result in a decreased relative
hydraulic conductivity index for the S.F.G. window as
compared to the L.D.C.G. window.
3.3.3.5 Support Slurry Filled Window
As discussed in Section 3.3.2.4, the support slurry filled
window was not detected. It is hypothesized that the slurry
filled cavity collapsed prior to the piezocone soundings.
3.3.4 Window Identification Guidelines
Based on the data obtained from the test section soundings,
a set of general guidelines were developed to quantitatively
model the signature corresponding to each type of window
investigated. These guidelines were formulated as deviations in
the piezocone indices obtained when sounding a window as compared
to average values obtained in the intact backfill. These
guidelines, summarized below by individual window type, were used
to analyze the containment wall soundings for potential window
locations.
It is emphasized that zones labeled as "potential windows"
were identified based solely on deviations in piezocone indices
as compared to values associated with intact backfill. Without
confirmatory data, the existence of an actual window cannot be
determined. Furthermore, the impact of the potential windows on
overall containment wall effectiveness cannot be evaluated as
based on piezocone data. Such evaluations can only be made by
hydraulically stressing the cutoff wall and analyzing its effect
on the hydrologic response of the aquifer/aquiclude system. This
work is discussed in Section 4.
As determined via marsh funnel testing. The support slurry used
exhibited a marsh funnel of approximately 41 seconds, where as
clean water has a marsh funnel viscosity of 29 seconds.
3-69
-------
It should also be recognized that the test section upon
which the following guidelines were based was constructed to
model the type of backfill used at the Gilson Road site. The
test section backfill was therefore a granular soil/bentonite
type mix^S. other types of soil/bentonite mixes, or different
backfill types all together (cement/bentonite for example), would
not be expected to follow the same guidelines.
3.3.4.1 High Density Clean Granular Window
0 Point resistance greater than twenty times that for the
intact backfill.
0 Point resistance consistent with the known material
type, strength, and density of the in-situ native
formation (values > 30 kg/cm^ for the Gilson Road site).
0 Friction ratio less than one-half the value obtained in
the intact backfill.
0 Reduction in excess pore pressure from values associated
with the intact backfill (to near zero or negative).
0 Relative hydraulic conductivity index greater than five
times that for the intact backfill.
3.3.4.2 Low Density Clean Granular Window
0 Point resistance greater than five times that for the
intact backfill.
0 Friction ratio less than one-half that corresponding to
the intact backfill.
0 Reduction in excess pore pressure to a near zero
magnitude.
0 Relative hydraulic conductivity index greater than five
times that for the intact backfill.
^8Soil/bentonite backfills can be separated into granular or clayey
type mixes. Granular mixes are those composed primarily of
noncohesive materials (by dry weight) which rely heavily on small
percentages of bentonite to achieve low hydraulic conductivity.
Clayey mixes contain high percentages of plastic materials which
reduce the backfill hydraulic conductivity to low values prior to
bentonite augmentation.
3-70
-------
3.3.4.3 Slurry Filled Granular Window
0 Point resistance greater than five times that for the
intact backfill.
0 Relatively high local friction values, greater than
twice the intact backfill magnitude.
0 Friction ratio greater than twice the intact backfill
value upon entering and exiting the window.
0 Reduction in excess pore pressure to a near zero
magnitude.
0 Relative hydraulic conductivity index greater than five
times that for the intact backfill.
3.3.4.4 Support Slurry Filled Window
This type of window is not expected to remain stable in a
granular type soil/bentonite backfill such as used at the Gilson
Road site. However, this window type, if existing in a cutoff
wall, would be easily identified through a lack of significant
point resistance or local friction.
3.4 Containment Wall Soundings
Thirteen soundings were executed in the Gilson Road
containment wall after completion of the test section trial
soundings. Five of the soundings (GZ-11, GZ-12, GZ-13, S-2E, and
S-10) were concentrated in two areas where windows were most
likely to exist. The first area (Station 1+00) was the location
3-71
-------
of the deepest part of the containment wall. The second area
(Station 3+45) coincided with the occurrence of a localized
trench instability encountered during construction^. The
remaining eight soundings were distributed along the downgradient
portion of the containment wall. The individual locations of
these soundings are provided in Figure 3-9.
The containment wall soundings were executed in two phases.
The first phase (GZ-11, GZ-12, and GZ-13) followed immediately
after the test section soundings. The second phase was delayed
for approximately one month to allow modification of the
permeameter equipment^.
In general, the containment wall soundings exhibited trends
similar to those corresponding to the intact backfill zones in
the test section. Deviations in piezocone response from the
intact backfill values were also noted in some cases. At depths
where deviations were noted in one piezocone index, the other
indices were examined to identify signatures resembling window
data from the test section. The windows were analyzed based on
the guidelines established during the test section soundings
(Section 3.3.4).
An example sounding is analyzed in the following subsections.
This sounding corresponded to the deepest portion of the cutoff
wall (Section 1+00). This area was selected for investigation
due to the 100 foot depth of the wall and thus the higher
potential for windows. Of the three soundings performed in this
excavation, a 2 inch wide, 50 foot long, conchoidal shaped
crack developed at the surface 15 to 20 feet behind the trench.
It is hypothesized that this resulted from slurry loss into an
esker deposit with the associated local instability at depth.
Slurry was pumped to the trench from the two slurry ponds
constructed in recognition of the esker deposits. The excess
slurry head and overall stability was thus maintained.
Thirty-six inch diameter pipes were then inserted into the trench
to provide added support. The trench bottom was recleaned and
the trench was backfilled.
^Initially, the permeameter porous stone was the same diameter as
the drill string. The potential for preferencing boundary flow
along the drill string was recognized during test section work.
A step taper was therefore added to the design (Figure 3-2). In
addition, it was found that a large permeameter
pressurization/volume measurement reservoir was required. This
equipment was also modified prior to the final phase of
soundings.
3-72
-------
area, GZ-13 exhibited the most complete record31 ana also
provided the greatest variety of response. The record for
sounding GZ-13 is provided as Figures 3-22a and 3-22b.
3.4.1 Intact Backfill; Sounding GZ-13
The point resistance encountered in the soil/bentonite
backfill exhibited a nearly constant trend at 2 kg/cm2 to a depth
of 50 feet. From 50 feet to the bottom of the sounding at 95
feet, a linearly increasing point resistance to 4 kg/cm2 was
evident. This data compared favorably with that developed during
soundings in the test section (test section point resistance =
1.5 kg/cm2). In fact, a trend of increasing point resistance
with depth better follows the theoretical trend for a uniform
material.
Local friction typically averaged 0.1 kg/cm2 with a slightly
increasing trend towards the bottom of the sounding. This value
is somewhat higher than that established during the test section
soundings «0.05 kg/cm2). The proportionally higher local
friction, as compared to point resistance, also yielded a
somewhat higher friction ratio of 3.0 to 3.5 percent (test
section = 2.2 kg/cm2). The point resistance and friction ratio
values indicate that the material is a sandy/silty clay as based
on the empirical identification charts (Figure 3-6).
Excess pore pressures generated during sounding GZ-13 were
significantly higher than observed in the test section (0.2
kg/cm2). The average magnitude ranged from 0.6 kg/cm2 to 1.0
kg/cm2, with the higher values encountered over the bottom
section of the wall. These larger excess pore pressures are in
better agreement with theoretical considerations than the data
from the test section. This lends a higher degree of credibility
to the hypothesis that the low, sometimes negative, average
excess pore pressures may have been due to an inaccurate zero
value. Figure 3-7 indicates that the material behaves as a
clayey sand or a silty clay, as based on the excess pore pressure
and point resistance indices.
Relative hydraulic conductivity values for the intact
backfill averaged 7.6 cm3/ft., within a range of 4.5 cm3/ft. to
11.5 cm3/ft. With the exception of a few higher hydraulic
conductivity ?ones, (potential windows) the water take values
31The excess pore pressure transducer failed during sounding S2E
and the point resistance transducer failed in sounding S-10.
3-73
-------
FIG. 3-22o PIEZOCONE TEST—SOUNDING NO. GZ13
POINT STRESS-OC .KG/CH1 POINT STRESS-OC -KG/CM3 LOCAL FRICTION-F .KG/CM* PORE PRESSURE-U .KB/CM1 FRICTION RBTIO-FR ,Z
. y
fl-fRCTOR
UJS
UJ
U_
CO ».
WRTER FLOW CM*
\
-------
FK3.3-22b PIEZOCONE TEST—SOUNDING NO. GZ13
POINT STRESS-OC .KG/CM1 POINT STRESS-OC .KB/CM* LOCH. FRICTIOM-F .KG/CM' PORE PRESSURE-U .KG/CM* FRICTION RRTIO-FR ,Z
A-FACTOR .1
VRTER FLOV Or
m z
Si
-------
were relatively consistent. It should be noted that the head
loss across the porous stone increased during the test section
and containment wall soundings. The increase was due to smearing
of the high density polyethylene stone and internal surface
loading due to suspended solids (iron precipitate) in the water.
Driving pressures were therefore increased after the test section
soundings to maintain adequate flow rates. As such, the water
take values are comparable within a given sounding but not
generally from sounding to sounding (the same driving pressure
was used for all test section soundings).
3.4.2 Potential Window Locations, GZ-13
Sounding GZ-13 exhibited signatures which corresponded to
all three types of potential windows detected in the test section.
These are the high density clean granular (H.D.C.G.) window, the
low density clean granular (L.D.C.G.) window and the slurry
filled granular (S.F.G.) window. Each window type will be
discussed individually below.
A. High Density Clean Granular Window
The occurrence of this window type is most likely to be
the result of incomplete excavation of native formation
materials. This is particularly true of sites where the
cutoff wall has been keyed to the top of bedrock as was
the case at Gilson Road. Incomplete excavation to the
top of bedrock typically results when the scale of
bedrock surface convolutions is small relative to the
size of the excavation
As based on the previously established guidelines, it
appears that sounding GZ-13 encountered native formation
J^At the Gilson Road site a great deal of quality control was
exercised to ensure excavation to the top of bedrock. Quality
control procedures included excavation until no additional
material could be removed, verification as based on borings at
100 to 250 foot spacing, and re-excavation just ahead of the
advancing backfill face to clean the key (Ayres, 1983). During
excavation, bedrock, as identified via fracture plane and
lithology, was typically removed from the bottom of the trench
for visual inspection. However, it is likely that occasional
small pockets of glacial till remained in the smaller bedrock
convolutions after backfilling. The specification required air
lifting to clean these pockets but it was not enforced due to the
enhanced potential for side wall instability which air lifting
would induce.
3-76
-------
material at a depth of 88.5 feet. In this location the
point resistance increased to over 50 kg/cm2 and
friction ratio decreased to about 1 percent (intact
backfill values of 2 to 4 kg/cm2 and 3.0 to 3.5 percent,
respectively). Excess pore pressure dropped from an
average value of +1.0 kg/cm2^ in the backfill to an
instantaneous low of < -0.6 kg/cm2 upon encountering
this zone. The relative hydraulic conductivity index
peaked in this zone at a maximum value of 40 cm3/ft.
(average value in intact backfill of 7.6 cm3/ft.). With
further penetration, the point resistance fell off
dramatically (8 kg/cm2) the relative hydraulic
conductivity index decreased (7.4 cm^/foot) and the
excess pore pressure increased to positive values (+1
kg/cm2). After three additional feet of penetration,
this trend was again followed by high point resistance
(40 kg/cm2), markedly increased relative hydraulic
conductivity (100 cm^/ft.) and decreased excess pore
pressure (+0.2 kg/cm2). The only inconsistent parameter
was friction ratio which continued to decrease in a
linear fashion.
The piezocone data, in and of themselves, indicate that
the cutoff wall was not constructed to the top of
bedrock. As based on a penetration of this material of
over 6 feet, the data further indicated that bedrock is
at least-33 that distance below the wall. However,
another explanation for the data exists which is more
consistent with the quality control data obtained during
construction as well as the piezocone data. It is
hypothesized that the sounding penetrated the side of
the cutoff wall and then proceeded into the native
formation. It is further hypothesized that this was due
resistance did not indicate bedrock was encountered. The
sounding was terminated when enough down pressure was applied to
lift the drill rig off the ground. This is not reflected in the
point loads because a significant portion of the penetration
resistance was likely to have been mobilized at the enlarged
shoulder corresponding to the drill string (see Figure 3-2).
3-77
-------
to a non-vertical cutoff wall rather than an inclined
sounding-^. The following data support this
conclusion:
After excavation, the slurry trench was sounded to
determine its depth and a measurement of 102 feet was
obtained at station 1+00. This depth corresponded to
the top of bedrock as verified via the above mentioned
quality control measures. These data indicate that the
trench was in fact excavated to the top of bedrock.
Post-construction verification borings also penetrated
the sidewall of the trench prior to reaching the bottom
of the wall. It is highly likely that these boreholes
are very nearly vertical (see Section 3.5).
A general trend of the linearly increasing point
resistance and linearly decreasing friction ratio are,
as would be expected, based on exit of the side wall at
a high penetration angle (slow increase in confinement
with distance from the cutoff wall).
The two high peaks in the point resistance, relative
hydraulic conductivity and excess pore pressure, along
with the intervening more moderate zone, are postulated
to indicate that the piezocone was moving in and out of
the backfill/filter cake/native formation due to
irregularities in the side wall.
Approximately 3 feet prior to encountering a point
resistance indicative of a H.D.C.G. window, the relative
hydraulic conductivity index increased somewhat from
values associated with the intact backfill. It is
postulated that this reflects the proximity or partial
contact of the porous stone with the native formation.
Most of the other containment wall soundings also
encountered native formation materials prior to reaching
bedrock depth.
special procedures were followed to obtain vertical
soundings, wall verticality was only specified to within 1
percent. This could result in a translation of about 1 foot at
the bottom of the wall. However, the actual verticality of the
wall was governed by how level the backhoe tracks were positioned
on the working surface. A 6 inch error in the level of the
working mat, over the track width, would result in a translation
of 2.5 feet at the bottom of the wall.
3-78
-------
The above data does not support the conclusion that the
material encountered below 88.5 feet represents
unexcavated native material. However, it is recognized
that another possible explanation encompasses a window
composed of native material which sloughed off the side
wall "in mass" and filled the bottom of the trench prior
to backfill placement. However, the available data also
does not support this scenario to an extent which would
render it plausible.
Independent of the exact nature and cause of the
"potential window"/sidewall penetration identified below
88.5 feet in sounding GZ-13, the individual piezocone
indices obtained from this location exhibit a
significant degree of consistency. In addition, the
magnitude and direction (positive vs. negative) of the
deviations from intact backfill indices agree with
theoretical trends that would be expected for a
noncohesive, highly permeable material. This data
demonstrates the potential usefulness of the piezocone
for window detection.
B. Low Density Clean Granular Window
The occurrence of this type of window is most likely to
be the result of side wall material which had sloughed
onto the advancing backfill face. Two such zones were
initially identified in sounding GZ-13. The first zone,
located from 17.5 to 18.5 feet deep in the wall
exhibited the point resistance, friction ratio and
excess pore pressure signature characteristics
indicative of the L.D.C.G. windows encountered in the
test section. However the relative hydraulic
conductivity index did not support the classification of
this zone as a potential window. It was therefore
dropped from further consideration due to the
definitiveness of this index.
The second potential L.D.C.G. window was encountered at
a depth of 43 to 48 feet. In this case, the point
resistance abruptly increased to about 20 kg/cm2 and
remained relatively constant at an average value of 10
to 15 kg/cm2. Average point resistances immediately
proceeding and following this zone were commensurate
with intact backfill (2 kg/cm2). The friction ratio
decreased from about 3 percent to values approaching 1
percent in this zone. Excess pore pressure decreased
abruptly from an average of +1.0 kg/cm2 to an average of
0.0 kg/cm2 and then restabilized at 1.0 kg/cm2 upon
exiting this zone. Finally, the relative hydraulic
3-79
-------
conductivity index increased to about 30 cm3/ft. in this
zone. This index yielded values of 8.7 cm^/ft. and
11.5 cm^/ft. prior to and after encountering this zone,
respectively.
The above data is internally consistent and follows
trends exhibited in the test section. As such, it
indicates the presence of a potential L.D.C.G. window.
Interpretation in this instance is not complicated by
other significantly plausible explanations for the data.
In fact, the identification of intact backfill both
above and below the potential window add credibility to
this interpretation.
C. Slurry Filled Granular Window
Three zones were encountered in sounding GZ-13 which
initially indicated the potential for the presence of a
S.F.G. window. The first case occurred at a depth of
28.0 to 29.5 feet. This zone yielded a characteristic
signature with respect to point resistance, local
friction, friction ratio and excess pore pressure.
However, no deviation in relative hydraulic conductivity
from intact backfill values was detected. This zone was
therefore omitted from further investigation.
The second zone occurred at a depth of 88.5 feet, just
above the potential H.D.C.G. window previously discussed.
It is likely that the response is indicative of
penetration of the filter cake at the side of the cutoff
wall as indicated by the very high friction ratio spike.
This zone is also not considered in more detail herein,
due to the complications addressed above for the
H.D.C.G. window.
The third zone of interest was located at a depth of
58.5 to 59.5 feet. This zone yielded piezocone indices
which were nearly identical to the signature observed
during test section soundings. The local friction value
increased to about 0.4 kg/cm2. The point resistance
peaked at about 15 kg/cm2 in the middle of this zone.
This yielded the double peak in friction ratio (6 to 10
percent) encountered in the test section soundings.
Excess pore pressure dropped to a near zero magnitude
and relative hydraulic conductivity peaked at 40 cm^/ft.
These indices yielded values commensurate with the
intact backfill both before and after penetration of
this zone. As such, this zone has a high probability of
being a slurry filled granular window.
3-80
-------
3.5 Window Verification
The primary piezocone indices of point resistance, local
friction and excess pore pressure indicated that a significant
number of potential windows existed in the containment wall.
However, a relative hydraulic conductivity index commensurate
with intact backfill allowed most of these potential windows to
be dropped from further consideration. The next step in the
window verification procedure encompassed obtaining samples of
the remaining potential windows for visual classification and
laboratory testing. These samples were obtained using standard
split spoon and Osterberg undisturbed Shelby tubes in boreholes
supported with heavy weight drilling fluid^S.
A. total of twelve borings were executed in the containment
wall. These borings were located in areas exhibiting the highest
probability of containing windows as based on the QC data
obtained during construction and the post-construction piezocone
soundings. The locations of these borings are presented on
Figure 3-9. As shown on the figure, some locations contain
multiple borings in a relatively concentrated area. This
reflects two basic problems that were encountered when attempting
to obtain samples from the wall. The first concerns locating the
boring at the centerline of the wall. Although the as-built
survey of wall location was sufficiently accurate for subsequent
construction at the site, it was not accurate enough for wall
sampling. Initially, ground penetrating radar (GPR) was employed
to find wall centerline. The location of the top of the wall was
easily identified. However, the top of the wall was often 8 to
10 feet wide for the first 5 to 10 feet of depth and then
narrowed to the standard 3 to 4 foot
The GPR could not penetrate through this upper zone and
therefore could not accurately identify wall centerline. Hence,
the borings sometimes missed the cutoff wall and required
relocation. The second difficulty concerns wall verticality. As
discussed in Section 3.4.2, it is hypothesized that the cutoff
wall is not completely vertical. This has no impact on the
effectiveness of the wall. However, due to the verticality of
^Additional information on sampling and drilling techniques can be
found in Winterkorn, 1975, Pgs. 15 and 44.
3*>During excavation, the top 5 to 10 feet of the trench was
typically sloped back on a 30 to 45 degree angle from vertical.
This was done to prevent surface materials at the edge of the
trench from raveling in and potentially resulting in windows.
3-81
-------
the borings37, many of them penetrated the side of the trench
and proceeded into the native formation before reaching the
bottom of the containment wall. This also resulted in the
requirement for additional borings.
During the sampling program, a total of fifty-four split
spoon samples and fifty six tube samples were obtained. The
split spoon samples were used for visual classification and as an
aid in determination of tube sampling intervals. The tube
samples were used to provide undisturbed material for triaxial
hydraulic conductivity testing as well as visual classification.
Prior to sample designation for triaxial testing, the unopened
tubes were radiographed to ensure that representative samples
were selected-^.
3.5.1 Sample Classification
Six of the samples retrieved from potential window locations
yielded material which, if continuous through the wall, would
result in actual windows. These locations are:
0 Boring GZU-13, cutoff wall station 1+00 at a depth of 43
to 48 feet.
0 Boring TBF-1A, cutoff wall station 3 + 45 at a depth of 39
to 46 feet.
borings were advanced using drilling fluid supported,
open-hole techniques. Large diameter drill rods (NX) were used
and the rigs were precision leveled prior to and during the
drilling. The combination of the leveling, the weight of the
rods, and the soft consistency of the cutoff wall backfill foster
a high degree of verticality. These techniques, as used on other
projects where the boreholes were surveyed with inclinometers,
typically result in a maximum error of 3 to 6 inches at the
bottom of 100 foot deep bore holes.
^Radiography allows general classification of the soil in the
Shelby tubes prior to opening them. The method is based on x-ray
transmission characteristics of the tube contents. The procedure
proved valuable as a tool to identify representative portions of
the samples for triaxial testing, particularly with respect to
intact backfill. Samples containing large pieces of gravel could
thus be rejected from testing even when the gravel was not
visible at the surface of the sample. Samples containing large
pieces of gravel (greater than one-sixth the sample diameter) can
yield artificially low values of hydraulic conductivity.
3-82
-------
0 Boring SU-7, cutoff wall Station 33+30 at depths of 30.5
to 35.5 feet, 45.5 to 48 feet, and 49.5 to 52 feet.
0 Boring SU-4, cutoff wall Station 37+50 at a depth of 21
to 26 feet.
Each of the above potential windows are discussed
individually in the following subsections. In addition to the
above enumerated potential window samples, two cases were
encountered where a sample could not be retrieved from a
potential window location. These were boring GZ-13U, 58.5 feet
to 59.5 feet, boring TBF-lA, 9.5 feet to 10.5 feet. As such, a
final determination of material classification could not be made.
In cases where the borehole encountered dense native formation
materials above bedrock, the available data indicate that this
was the result of trench sid e w all penetration due to
non-vert icali ty of the cutoff wall rather than unexcavated
material at the bottom of the wall. Samples of this material are
therefore not treated further herein.
3.5.1.1 Boring GZU-13, 43 to 48 Feet
This potential window was identified at station 1 + 00 in
sounding GZ-13. The boring, located 5 feet down-station from the
sounding, encountered the potential window at a depth of 48 feet.
Two samples of the potential window were obtained. The first
sample was a split spoon from 48 to 50 feet and the second was a
tube from 50 to 52 feet. The samples were classified as a loose,
light brown, coarse to fine sand, trace fine gravel, trace silt,
in accordance with the Burmister system. This material is
consistent with the piezocone identification as a low density
clean granular window.
After the above two samples were obtained, the boring was
advanced and remained in granular deposits until refusal was
encountered at a depth of 90 feet. The deeper granular deposits
appear to be the result of penetration of the trench sidewall.
Inasmuch as additional backfill was not encountered below the
potential window, the identification of this material as a window
is inconclusive. In fact, this material may also correspond to
sidewall penetration rather than the zone identified in sounding
GZ-13.
3-83
-------
3.5.1.2 Boring TBF-lA, 39 to 46 Feet
This potential window was identified at station 3+45. The
window consisted of two separate materials. The upper 4.5 feet
contained a loose, light brown, coarse sand, trace medium sand,
trace silt. The material exhibited a uniform grain size and the
voids were filled with bentonite slurry. As such, this material
corresponded to a slurry filled granular window which was caused
by sedimentation of material suspended in the support slurry.
The lower 2.6 feet of this window contained a medium dense,
gray, fine to coarse sand, trace gravel, trace silt. It appears
that this material was glacial till which was located in a
bedrock surface convolution and therefore could not be excavated.
Boring TBF-1, located 5 feet down-station of TFB-lA also
encountered this window. In this case, the window was 6.5 feet
thick. Sounding GZ-12, located an additional 5 feet down-station
of TBF-lA, did not encounter the window. In the up-station
direction, sounding GZ-11, located 5 feet from TBF-lA,
encountered the window over a thickness of 3.5 feet. This data
indicates that the window is greater than 10 feet long in the
direction of the cutoff wall alignment.
3.5.1.3 Boring SU-7, 30.5 to 35.5 Feet, 45.5 to 48
Feet and 49.5 to 52.0 Feet
These three potential windows were initially identified at
station 33+30 in sounding S-7. The piezocone signatures for each
of the three windows were very similar. They each exhibited
significant point resistance similar to a L.D.C.G. or S.F.G.
window (15 to 20 kg/cm^ as compared to 5 kg/cm2 for the intact
backfill). Local friction values were also relatively high as
would be expected for a S.F.G. window (0.4 to 0.6 kg/cm2). These
indices yielded friction ratios which averaged from 3 to 5
percent with local peaks as high as 7 percent. Although the
magnitude of the indices are similar to those expected for a
S.F.G. window as based on the test section soundings, the
friction ratio signature did not exhibit the characteristic shape.
Instead of the well defined friction ratio spikes upon entering
and exiting the window, friction ratio in this case yielded
consistently high values. This behavior would be expected if the
piezocone encountered the filter cake at trench sidewall
irregularities. This hypothesis can not be further supported by
the piezocone data in that excess pore pressure and relative
hydraulic conductivity data was not available.
3-84
-------
Boring SU-7 was located 5 feet down-station of the sounding.
Sampling confirmed two zones corresponding to the above potential
windows. The depths of these zones were 29.0 to 35 feet and 48
to 50 feet. As such, they are somewhat displaced from that
indicated by the sounding. The samples were classified as loose,
brown, coarse to fine sand, little fine gravel, trace silt.
Bentonite slurry was not noted in the soil pore space. These
samples were obtained via split spoon. Tube samples were
attempted but they yielded no recovery.
The samples support classification as a L.D.C.G. window.
However, this window type is not corroborated by the piezocone
data. As with the piezocone sounding, it is probable that the
boring penetrated into native formation material at locations of
irregularities in the trench sidewall. This hypothesis is
supported by:
0 A material type which corresponds to native formation
deposits.
0 Penetration of the sidewall below a depth of 52 feet.
The boring thereafter remained in native formation
material until refusal at 60 feet.
0 Piezocone response uncharacteristic of window types
anticipated in the cutoff wall but consistent with
partial sidewall penetration.
0 An identical piezocone response in sounding S-9 from 23
to 28 feet. Boring SU-9 encountered the same zone at 22
feet but remained in the native formation until refusal
at 30 feet.
The data is inconclusive as to the cause of the non-backfill
material encountered. While the existence of some type of window
not modeled in the test section is certainly possible, it is more
likely that the sounding and boring both penetrated
irregularities in the sidewalls of the trench. Such
irregularities are often observed in concrete backfilled slurry
trenches when the soil is removed from the walls. Irregularity
magnitudes of 1 to 3 feet are not uncommon, particularly in
granular formations.
3.5.1.4 Boring SU-4, 21 to 26 Feet
This potential window was identified at Station 37+50 in
Sounding S-4. The piezocone signature in this case was identical
to the three zones described in Section 3.5.1.3 (Sounding S-7).
3-85
-------
As such, it is hypothesized that Sounding S-4 also penetrated an
irregularity in the trench sidewall or initially penetrated the
sidewall and was deflected back into the soft backfill by the
denser native deposits. Boring SU-4, located 5 feet
down-station, did not encounter these potential window materials.
Continuous samples were taken starting 3 feet above the top of
the potential window and proceeded 4 feet below it's bottom.
Only intact backfill was recovered therefore indicating that the
boring did not encounter the trench sidewall.
3.5.2 Sample Hydraulic Conductivity Testing
In only two cases could undisturbed samples of potential
windows be obtained. These were from boring GZU-13 and TBF-1A.
While triaxial hydraulic conductivity tests could have been
performed on those samples, the effort was judged unnecessary.
The grain size distribution of the material in the samples
indicated a probable hydraulic conductivity of 10~^ to 10~^
cm/sec. Further refinement of this estimate provides no
significant information in that this value is high as compared to
the design specification for the cutoff wall of 1 x 10"' cm/sec.
Therefore, if these materials exist in a continuous state
transecting the cutoff wall, then a window would certainly exist.
Investigation with respect to the hydraulic conductivity of
undisturbed intact backfill samples was considered a more
valuable endeavor. As such, representative tube samples of intact
backfill were subjected to triaxial testing. The results of this
work and the conclusions drawn from the data are presented in
Section 2.
3.5.3 Containment Wall Homogeneity
One of the objectives of the containment wall soundings was
to provide an estimate of the probable frequency of window
occurrence in the cutoff wall. Nearly 1,500 feet of sounding and
borehole installation yielded only two cases where the
probability of the existence of a window was high. The first
case was boring GZU-13 as described in Section 3.5.1.1. In this
case, window identification although probable, was inconclusive
due to the lack of an intact backfill sample below the potential
window. However, it is believed that the second case (boring
TBF-lA, section 3.5.1.2) was conclusively identified as a
non-backfill deposit located between the bottom of the cutoff
wall and the bedrock. In this case, the same undisturbed tube
sample contained intact backfill directly above S.F.G. window
3-86
-------
material. It is unlikely that this data fits a plausible
non-window explanation.
The above data indicated that few windows are likely to
exist in the Gilson Road containment wall. However, it is
recognized that much of the sounding data was inconclusive due to
complications arising from trench side wall penetration. While
this hypothesis is probable, it can not be proven with the
available data. The inability to adequately sound and sample the
containment wall was a direct result of the unknown location of
exact wall center line and the probable non-verticality of the
wall. While these issues do not result in deficiencies with
respect to containment performance, they do render
post-construction verification a very difficult, expensive and
inclusive undertaking. As such, greater quality control should
be exercised in these areas.
Finally, it is emphasized that even in the case where the
existence of a window was highly probable, its impact on the
efficiency of the cutoff wall can not be determined via sample
retrieval. Sampling provides no indication of the degree to
which the window material is continuous through the wall. In
fact, it is probable that under the worst case conditions, a
filter cake exists between the window material and the native
formation. A filter cake, such as would be anticipated, would
render the hydraulic conductivity of the overall wall
cross-section orders of magnitude below that of the native
formation. The effectiveness of the overall cutoff wall can only
be evaluated by hydraulically stressing the wall and observing
the piezometric response of the system. This task is treated in
Section 4.
3.6 Conclusions
The following conclusions can be drawn from the data
presented herein:
0 The test section did not provide the known condition
expected for piezocone performance evaluation. The lack
of support slurry filled window long-term stability
provided valuable insight into the improbability of the
existence of such a window in the containment wall.
However, the inability to construct other window types to
a known geometery impaired verification of piezocone
suitability for window detection in soi1/bentonite
backfilled cutoff walls.
3-87
-------
The response of the piezocone when encountering windows
in the test section was in general accordance with that
anticipated based on prior empirical data and theoretical
considerations. As such, window locations could be
identified in the test section. The windows produced
repeatable and identifiable signatures in the piezocone
indices. Based on the test section soundings, it appears
that the piezocone is capable of window detection in
soil/bentonite backfilled cutoff walls.
As an adjunct to the standard piezocone indices, the
relative hydraulic conductivity index provided valuable
additional insight with respect to potential window
identification. To enhance the utility of this system,
redesign should be undertaken to allow high flow volumes
at low driving pressures. This would better allow
evaluation of the hydraulic significance of a potential
window while minimizing preferential boundary flow and
hydraulic fracturing.
Potential window identification must utilize a
multi-index analysis. Many cases were found in which one
or two indices indicate window existence but the other
indices did not. In particular, the relative hydraulic
conductivity index often demonstrated that a high
hydraulic conductivity zone was not present in locations
initially identified as potential windows based on the
other indices.
While anomalies in the cutoff wall backfill were
relatively easy to identify as based on piezocone
response, significant judgment was required to identify
characteristic responses which were likely to be
associated with potential windows. This judgment must be
based on a thorough understanding of piezocone
operational principles as well as the stress, strain,
hydraulic conductivity characteristics of soil deposits.
Based on the data acquired during this portion of the
project, it appears that the piezocone could provide a
useful, routine post-construction verification capability.
However, additional work must be conducted to build up a
data base upon which empirical correlations can be
constructed. This magnitude of data generation would
likely require use of the instrument on a number of
different cutoff wall projects following additional
controlled test section sounding.
For piezocone sounding to be viable as a
post-construction verification tool, construction quality
3-88
-------
control of the cutoff wall must include wall alignment
and verticality provisions. The accuracy of as-built
wall alignment documentation and actual wall verticality
must be increased significantly above that required for
hydraulic performance only.
The piezocone soundings and undisturbed sampling of the
Gilson Road cutoff wall encountered windows in only a few
instances. These investigations were concentrated in
areas with the highest probability of window occurrence
as based on construction quality control data. This data
therefore reinforces the conclusion that the cutoff wall
behaves essentially as an intact barrier. However, it is
recognized that the data is somewhat inconclusive due to
complicating factors associated with the trench side wall
penetration issue. Sidewall penetration was likely due
to the probable non-verticality of the cutoff wall.
3-89
-------
SECTION 4
EVALUATION OF CONTAINMENT LEAKAGE
Passive physical containment (in-situ impervious barriers)
forms an integral part of the overall containment strategy
adopted for the remediation of the Gilson Road site. The three
major components of the passive containment system are:
1. The synthetic surface cap which acts as a barrier to
the vertical infiltration of precipitation and thus
forms the top of the containment vessel,
2. The soi1/bentonite cutoff wall which impedes the
horizontal flow of groundwater into and out of the site
and acts as the sides of the containment, and
3. The natural glacial till/bedrock aquiclude which forms
the bottom of the containment.
Imperfections in any one of these three components may
result in leakage and thus off-site migration of contaminants.
From the standpoint of environmental damage and human health
risk, if there is a breach in the containment, it may not be
particularly important which element is leaking. However, from
the standpoint of evaluating the inherent viability of a specific
containment technology or determining an appropriate hydrodynamic
isolation system (active physical barrier) to augment the passive
containment, as was the case for the Gilson Road site, it is
critical to assess the integrity of each component individually.
The objective of this, the third phase of the project, was
to evaluate the hydraulic effectiveness of the soil/bentonite
cutoff wall. The construction quality control testing and the
post-construction wall sampling/testing (Phase One, as reported
in Section 2.0) demonstrated that the intact hydraulic
conductivity of the cutoff wall generally met the contract
specification of 1 x 10~7 cm/sec (3 x 10~4 ft/day). However, the
effectiveness of the barrier is proportional to the cutoff wall
bulk hydraulic conductivity, of which intact hydraulic
conductivity is only a part. Ideally, the evaluation of bulk
cutoff wall hydraulic conductivity would be independent of the
influence of the other barriers. Inasmuch as the cap is located
4-1
-------
in the vadose zone, it only indirectly influences groundwater
flow. Its effect on the "analysis of bulk cutoff wall
conductivity can therefore be easily addressed. Unfortunately,
both the cutoff wall and aquiclude directly impact groundwater
movement, and component separation in this case is more
difficult.
Two approaches were used to separate the cutoff wall and
aquiclude, such that the effectiveness of the cutoff wall alone
could be evaluated. The first approach was based on the use of a
3-dimensional numerical model of the site which was previously
used to design the hydrodynamic isolation/recirculation system
for the project. The first step in this approach was to better
evaluate the aquiclude hydraulic conductivity via a bedrock
pumping test (Section 4.1). The second step was to employ a
numerical sensitivity analysis based on vertical piezometric head
distributions to evaluate bulk cutoff wall hydraulic conductivity
(Section 4.2) The second approach consisted of evaluation of
contaminant concentration versus time and depth (Section 4.3).
4.1 Bedrock Pumping Test
The primary objective of the bedrock pumping test was to
evaluate the hydraulic conductivity of the fractured rock which
forms the bottom of the containment below the cutoff wall.
Direct determination of bedrock hydraulic conductivity was an
essential step in the evaluation of the bulk hydraulic
conductivity of the cutoff wall. Particular emphasis was placed
on determining the bedrock hydraulic conductivity in the
direction perpendicular to the cutoff wall. This value would
control the leakage under the wall should the bedrock fracture
pattern exhibit anisotropic characteristics.
Documentation of the hydraulic conductivity of the rock has
also increased confidence in the calibration of the numerical
model which was used to design the hydrodynamic isolation system.
During model calibration, it was necessary to use conductivities
for the bedrock which were significantly greater than those
measured during packer testing. The "unjustifiably" high values
were required to match vertical heads simulated by the model to
the field observations. The pumping test has demonstrated that
these higher conductivities were essentially correct.
The secondary objective of the pumping test was to assess
the degree of hydraulic interconnection between the fractured
rock and the high hydraulic conductivity sand and gravel deposits
4-2
-------
of the upper aquifer. This was achieved via evaluation of an
effective homogeneous leakancel for the glacial till aquitard.
4.1.1 Pumping Well Location and Installation
The pumping well, PT-1, was located at the western end of
the site as shown in Figure 4-1. There were several
considerations involved in this selection. First, this portion
of the site experiences the greatest head loss across the cutoff
wall. Therefore, the bedrock hydraulic conductivity at this
location more strongly affects leakage out of the containment,
particularly because this is the downgradient end. Second, the
glacial till thickness increases in this area, providing greater
isolation between the bedrock and overburden aquifers.
Therefore, evaluation of bedrock hydraulic conductivity is
simplified and confidence in the evaluation is increased. Third,
the downgradient portion of the cutoff wall would be the most
highly stressed during the subsequent overburden test, thereby
allowing better comparison between the two tests. Finally, a
greater abundance of monitoring instrumentation was already in
place in the vicinity of the downgradient portion of the wall.
The pumping well was placed inside the cutoff wall so as not
to draw contaminants outside of the containment during execution
of the test. At the downgradient portion of the site, the
bedrock dips sharply in the direction perpendicular to the wall.
Therefore the well was placed close to the cutoff wall so as to
stress those materials which most directly affect leakage rates,
i.e., those immediately below the bottom of the wall. In
addition, the horizontal component of flow induced in the
overburden would be roughly radial, and therefore parallel to the
low hydraulic conductivity barrier. This minimized the
inaccuracies associated with approximating the overburden.aquifer
as homogeneous in the mathematical analysis.
The objective of the well installation was to allow pumping
from the bedrock only, and therefore the overburden and glacial
till were sealed off. The entire column of rock was to be
stressed, down to at least 50 feet below the glacial till, in
order to include the major fracture zones. Earlier drilling at
well R-l, including continuous coring and classification by RQD
^Hydraulic resistance to vertical flow between aquifers separated
by an aquitard.
4-3
-------
FIG. 4-1 PUMPING TEST MONITORING NETWORK
4
LYLE REED
BROOK
o
o
PUMPING WELL
MULTI-LEVEL
MONITORING INSTALLATION
SURFACE WATER POINT
RECIRCULATION WELL
" i RECIRCULATION TRENCH
6
o' so' too1
200'
-------
(rock quality designation2)f had shown that, in addition to a
highly fractured upper zone, a 5 foot thick highly fractured zone
(RQD = 20%) existed at a depth of 50 feet. Therefore the
possibility existed that leakage was occurring at significant
depths, and not merely in the highly fractured upper portion of
the bedrock. The depth of the borehole was therefore established
to penetrate 100 feet into the bedrock.
A 6-inch casing was installed through the overburden,
including the glacial till, and drilling of the bedrock began
using the air-rotary method. However, the top 10 feet of the
highly fractured rock was unstable and would not remain open.
Therefore the casing was driven through this layer as well. The
hole was then advanced uncased through 89 feet of mostly
competent rock. Finally, the lower 10 feet of casing was
perforated to reestablish hydraulic connection with the upper
portion of the bedrock.
The hole was tested several times during the drilling
process for its water yield. The water yield, or water make, was
estimated by forcing air into the bottom of the borehole, and
measuring the flow rate of water which was forced out of the top
of the casing at steady state. Before the casing was perforated,
the yield was approximately 10 gpm (gallons per minute), but
after perforation the yield increased to about 150 gpm. This
demonstrated the dominance of the upper zone regarding the
transmi ssi vity of the formation. In addition, the yield of 150
gpm indicated that the upper fractured zone was far more
permeable than had been inferred from the previous packer test
data.
After drilling and development operations were completed,
the pump was placed in the casing at a depth of 75 feet. A
piezometer was also included to provide water level measurements
in the well. Details of the installation are shown in
Figure 4-2.
4.1.2 Recharge Trenches
Water pumped during the test had to remain on the site due
to its contamination. Therefore, the water was recharged to the
overburden via three recharge trenches already in place. These
trenches, numbers 4, 5, and 6 as shown in Figure 4-1, were
index used to quantify rock mass integrity with respect to
degree of fracturing.
4-5
-------
FIG. 4-2 PUMPING WELL INSTALLATION
fl-
IC •
20-
30-
40-
50-
«- 60-
o>
*"„ 70-
X
»-
LU
O
90-
100-
*•
4.
150-
160-
170-
180-
•
CASING LEFT
IN PLACE
^_^
t
EL. 55
EL 7' —
PERFORATED ,
CASING ^
PUMP^
EL 83
SLOTTED PVC^-^
PIEZOMETER
'f t
OPEN HOLE — +-
EL, 172
J
,
/
/
\
*^^— .
SAND 8 GRAVEL
_
GLACIAL
TILL
HIGHLY FRACTURED
BEDROCK
MODERATELY
FRACTURED
2^ BEDROCK
J BOTTOM OF
BOREHOLE
4-6
-------
already operating as part of the hydrodynamic isolation system.
The pumpage from the bedrock pumping test was added to the
existing flow.
4.1.3 Monitoring Points
Drawdown resulting from pumping the bedrock well was
measured in the pumping well and 72 pre-existing and recently
installed piezometers and observation wells. Observation points
were divided among 24 borings, most of them multi-level
installations (see Figure 4-1). Specific emphasis was placed on
measurements in the upper rock layer and in the overburden just
above the glacial till. The upper, highly fractured portion of
the bedrock was of primary interest in that it controlled leakage
under the cutoff wall. Measurements just above the glacial till
were also emphasized inasmuch as the difference in drawdowns
across the glacial till controls the amount of induced leakage
through the aquitard and thus recharge to the bedrock.
The locations of monitoring points ranged horizontally from
as close as 10 feet to as far as 500 feet away from the pumping
well, and vertically from the water table to 100 feet below the
top of bedrock. In addition, water levels were monitored at five
points in nearby surface waters.
4.1.3.1 Existing Points
Prior to the pumping test, numerous observation wells and
gas-drive sampler/piezometers (Barvenik, 1983) existed at the
site as the result of previous engineering studies (e.g., design
and operation of hydrodynamic isolation systems for emergency
response and final remediation). In particular, there were
numerous multi-level installations both inside and outside the
cutoff wall for the purpose of estimating leakage through the
site. These monitoring installations typically included a 1-1/2
inch PVC observation well at the water table, one or two
sampler/piezometers in the stratified drift, and one in the top
10 to 15 feet of the bedrock. A typical installation is
presented as Figure 4-3.
The method used to construct these installations was
typically as follows. The casing was driven and washed out with
a roller bit down to the top of the bedrock. The hole was then
advanced by coring 10 to 15 feet into the bedrock. A
sampler/piezometer was placed in the bedrock inside a sand filter.
The casing was then extracted incrementally as a compacted
4-7
-------
FIG. 4-3 TYPICAL EXISTING MULTILEVEL INSTALLATION
o
10
9f\
30
•^
I
00 *•
o> 40'
•2?
JE 50-
A
*^
UJ
Q
60'
70-
80-
9O-
STRATIFIED
DRIFT
T-
>
EL. 53
/*** A /* I A 1
GLACIAL
TILL
1 1 LL
EL-72
BEDROCK
EL. 82
•v:
'•%?
•"';
'* V-
~5
fijj
K?
.:-:;
<&
C'*-'-
iS
^•ii
1
/
•'."•".
(^ ^
'• ".'.
'/
/
1
;•.•!
=
~
s
r.t:
'.
$'
*.*t
i
/
/
• .-
^ ^
-.•
^
/
1
\ r
*.«;
•v>
5^
r •*
^
:'."•
*Vv
^
•"^
«k^j
»•£
>*;
•\?:
'•*•*
^
/
/
/
,/
^V-
COLLAPSED
FORMATION
l'/2" PVC OBSERVATION
WELL WITH SAND PACK
p/\i i ADCtrn
w WLL. Ar O C U
FORMATION
-BENTONITE SEAL
GAS-DRIVE
MONITORING INTERVAL
DFRITtf^MITC OP A 1
OC.N IUNI 1 t, OCAL
-COLLAPSED FORMATION
WITH BENTONITE SEALS
BENTONITE SEAL
GAS-DRIVE
MONITORING INTERVAL
— CONCENTRIC
SAMPLING TUBES
-BENTONITE
SEAL
SANO PACK
POROUS STONE
SAMPLER
-OPEN HOLE
-BENTONITE
SEAL
-------
bentonite seal was placed to seal the borehole at the level of
the glacial till aquitard. This process was repeated for each
sampler/piezometer with intermediate seals placed such that the
vertical hydraulic conductivity of the installation was less than
that of the parent formation. An observation well was then
installed with the screen positioned so as to straddle existing
and projected water table elevations.
To allow groundwater samples to be extracted from the
formation, the gas-drive sampler/piezometers include one-way
check-valves, which allow unrestricted flow from the formation
into the access tube but which restrict, without completely
preventing, flow in the opposite direction during sampling. They
are therefore not capable of responding adequately to rapid
drawdowns associated with pumping tests in confined aquifers.
If, however, drawdowns are occurring more slowly, the instrument
response to decreasing piezometric head is consistent with that
of the aquifer. In particular, the check-valve does not
interfere at all with steady state water level measurements in
that transient effects are, by definition, absent. This will be
documented in a subsequent section. In addition, the
sampler/piezometers are considered satisfactory for recovery
responses, where heads are rising and the check-valve is open and
thus unrestrictive . In fact, response time for the
sampler/piezometers is better than for standard observation wells
when heads are increasing, due to the lower storage volume
associated with the small diameter riser tubes (Barvenik 1984).
4.1.3.2 New Installations
Three new monitoring installations (PT-2, PT-3, and PT-4)
were constructed specifically for this study. The primary reason
for the new installations was to provide rapid response .drawdown
data at small distances from the pumping well. In addition, they
were drilled to about the same depths as PT-1 in order to obtain
an understanding of the importance of the deeper zones of flow in
the bedrock. Two of the wells, PT-3 and PT-4, were lined up with
PT-1 perpendicular to the cutoff wall, because that directional
hydraulic conductivity controls the specific discharge from the
containment. PT-4 was located just outside of the cutoff wall in
order to better isolate the section of bedrock directly
underneath the wall. PT-2 was included to estimate the degree of
horizontal anisotropy.
The instrumentation differed from that in the pre-existing
installations in that each sampler/piezometer was supplemented
with a separate small-diameter PVC piezometer. The access tubes
for those piezometers were designed specifically to accommodate
4-9
-------
pressure transducers to measure the small-time transient response
of the aquifer. Transducers were required because/ due to the
semi-confined nature of the bedrock aquifer, drawdowns were
expected to occur too quickly for manual water level measurement
techniques. In addition, problems associated with the time lags
due to the check-valves in the sampler/piezometers were thus
bypassed.
There were several objectives to be met in locating the
piezometers within the new installations. The piezometers in the
bedrock were to be located in the most densely fractured zones.
They also required a minimum vertical separation in order to
avoid redundancy and allow adequate vertical distance to
establish an effective seal. Finally, measurement points were
desired directly above and below the glacial till in order to
assess the amount of induced leakage through the aquitard. In
order to meet these objectives, one point was placed as deep as
possible into the bedrock such that it coincided with a zone
exhibiting a high fracture density. The second monitoring point
was located in the highly fractured upper 10 feet of the bedrock.
A bentonite/cement grout seal at least 20 feet long was situated
in the more competent rock between these two locations. A third
monitoring location was placed in the sand/gravel aquifer just
above the glacial till, with a seal placed over the entire
glacial till thickness. Finally a 1-1/2-inch PVC observation
well was screened in the upper portions of the stratified drift
phreatic aquifer. A typical installation is presented in
Figure 4-4. Additional details regarding bedrock monitoring
installations of this type can be found in Barvenik (1983).
4.1.4 Execution of Pump Test
4.1.4.1 General Conditions
Due to the high water yield observed in the pumping well
during drilling, it was anticipated that the transmissivity of
the bedrock might be higher than had been estimated from previous
packer test results. Therefore, prior to execution of the
full-scale test, a preliminary test was executed to determine the
response time and degree of drawdown which might be expected.
The response time was quite fast, indicating that transducers
should be used. However, the amount of drawdown observed (less
than 3 feet) was much lower than expected. The accuracy of the
transducers was determined to be insufficient for monitoring the
full scale test. Therefore, water level readings had to be taken
manually. Electric water level readers, typically dedicated to
4-10
-------
FIG. 4-4 TYPICAL PUMPING TEST MULTILEVEL INSTALLATION
20-
30'
40' •
50
60-
70-•
80-
90- •
STRATI-
FIED
DRIFT
-38—
GLACIAL
TILL
•71-
BEDROCK
OBSERVATION
WELL WITH PEASTONE
PACK
COLLAPSED
FORMATION
GAS-DRIVE
MONITORING
INTERVAL
GROUT
SEAL
GAS-DRIVE
MONITORING
INTERVAL
GROUT SEAL
IOOT
IIO--
BEDROCK
120"
130- •
140"
150- •
160-
I70--
I80-1-
GROUT
SEAL
172 I
GAS-DRIVE
MONITORING
INTERVAL
CONCENTRIC
SAMPLING TUBES
Jfi-GROUT
SEAL
SAND
PLUG
t.1 "
3/4 PVC
PIEZOMETER
FILTER BAG
POROUS STONE
SAMPLER
SAND PACK
PEASTONE
BACKFILL
GROUT SEAL
-------
each monitoring point, allowed readings as often as every five
seconds. Readings were taken to within + 0.01 feet, but the
final accuracy was probably + 0.05 feet when all potential
reading errors are accounted for.
The pumping test in the bedrock was run from September 6
through September 16, 1985, for a total of ten days. Water was
extracted from the pumping well in the bedrock and returned to
the overburden via recharge trenches 4, 5, and 6 located
approximately 700 feet upgradient. These trenches were already
operating as part of the recirculation system, and the pumpage
was added to the existing flow. Please note, however, that the
system had been operating for several months prior to the test,
and had achieved approximate steady state. Therefore, changes in
piezometric head were attributed to the increment in recharge.
The pumping rate was constant throughout the test at 88 gpm.
This rate was measured by adding the flows through two parallel
flow meters, and was verified by Doppler ultrasonic measurements
of fluid velocity in the pipe between the meters and the trenches.
The pump was run on a direct three-phase 220-volt line, thereby
eliminating the attendant problems of using a portable generator.
At the beginning of the test, emphasis was placed on reading
water levels at those points closest to the pumping well. For
these points, readings were initially taken about every ten
seconds. As will be discussed subsequently, even this rapid
reading schedule proved to be too slow to capture the full
transient drawdown curve due to the rapid response of the bedrock
aquifer. This shortcoming in the data was primarily limited to
the points in the highly fractured upper zone of the bedrock
immediately adjacent to the pumping well. As time progressed and
the monitoring wells closest to the pumping well approached
steady state, emphasis shifted to the points further away, both
horizontally and vertically. Some of the furthest points did not
show significant drawdowns for more than one hour after pumping
began. Thus, numerous readings at the early times were not
justified for these locations.
The occurrence of heavy rains during execution of the
pumping test affected the drawdowns measured in the monitoring
wells. Table 4-1 shows the daily total precipitation at a nearby
NOAA weather data station (Pennichuck Water Works). During the
test, fluctuations of up to 0.4 feet in the water level of Lyle
Reed Brook rendered transient response data for the system
unsuitable for analysis, as will be discussed in the following
sections.
4-12
-------
TABLE 4-1
DAILY PRECIPITATION DURING PUMPING TEST (9/85)
Date
Precipitation (in.)
9/06/85
9/07/85
9/08/85
9/09/85
9/10/85
9/11/85
9/12/85
9/13/185
9/14/85
9/15/85
9/16/85
1.30
1.06
0.00
trace
0.45
0.01
0.00
0.00
0.00
0.00
0.00
4.1.4.2 Data Discussion
Once pumping began, steady-state conditions in the rock
layers were achieved in about two days. In the overburden,
achievement of steady-state conditions took longer, typically on
the order of seven days. At some points however, the drawdown
was still increasing slowly after ten days. Figures 4-5 and 4-6
show the time-drawdown history at two multi-level installations.
The first, PT-2, is representative of the majority of the data.
The second, M-14, demonstrates anomalous behavior due to a
previous error in location of the seal between monitoring
intervals.
At PT-2, drawdowns were higher in the bedrock than in the
stratified drift. The highest drawdown occurred in the
upper bedrock (PT-2-3). At this point, the majority of the
drawdown took place before measurements could be made (less than
5 seconds), indicating a low storage coefficient. The deep rock
piezometer (PT-2-4) ultimately developed a drawdown near that of
PT-2-3. However, the response time was two to three orders of
magnitude larger. This indicates that the connection between
points deep in the bedrock is quite tortuous, with a low
hydraulic conductivity.
4-13
-------
FIG. 4-5 PT-2 TIME-DRAWDOWN CURVE
2.4
1.8
I
M
*>.
4)
V
O
O
<
oc
Q
1.2
• PT-2-1
A PT-2-2
• PT-2-3
+ PT-2-4
WATER TABLE
TOP OF TILL
TOP OF ROCK
DEEP ROCK
-/•
0.6
0.0
TIME, minutes
-------
FIG. 4-6 M-14 TIME-DRAWDOWN CURVE
I
H
un
Q)
0>
o
o
tr.
o
TIME, min utes
-------
At PT-2-2 (above the glacial till), the drawdown was
generally 2 feet less than that at PT-2-3. The small-time
response for this point was not observed. The water table
(PT-2-1) was the slowest to respond, since its storage
coefficient is equal to the specific yield of the sands and
gravels, and thus much larger than that of the lower points. The
water table also experienced the least drawdown of any point in
PT-2.
The induced components of flow in the aquifer near PT-2 can
be deduced from the differences in drawdowns between various
measurement points. The induced flow occurs from regions of low
drawdown to those of high drawdown. For example, in the vicinity
of PT-2, flow is occurring upwards in the bedrock and downwards
in the overburden. Note that the difference in final drawdown
between PT-2-1 and PT-2-2 (both in the stratified drift) is 0.3
feet, while the difference between PT-2-2 and PT-2-3 (spanning
the glacial till) is 1.6 feet. This illustrates the fact that
the till is much less permeable than the sands above it.
The drawdown history for M-14 illustrates the importance of
seal location and integrity. M-14-4 was placed in the fractured
rock zone, while M-14-3 was placed in the sand and gravel aquifer
just above the glacial till. However, both points show nearly
identical drawdowns. This is because the lower seal separating
the two monitoring points was placed in the top of the bedrock,
as is standard practice, and not in the less permeable glacial
till. Another seal was placed below M-14-3 in the top of the
glacial till. However, a significant section of the installation
through the glacial till was not sealed. Figure 4-6 therefore
illustrates the consequence of breaching the aquitard and thus
establishing a connection between the two points. This small
artificial leak through the aquitard probably has little effect
on the overall response of the two aquifer system. However, its
local effect in the vicinity of the monitoring point is
significant enough to render this data misleading. This isolated
case is highlighted because future sections will address the
vertical head profiles at this installation.
Figure 4-7 shows the time-drawdown plot for the two
monitoring points M-3-4 and M-3-5. The former is a
sampler/piezometer and the latter is a piezometer only. Both
were installed in the same borehole at the same depth. Note that
due to its check valve, the sampler/piezometer responds more
slowly at early times, but that the two points respond
identically (within reading accuracy) at large times. This plot
illustrates the accuracy of the gas-drive sampler/piezometers in
providing steady-state hydraulic information.
4-16
-------
FIG. 4-7 M-3-4/M-3-5 TIME-DRAWDOWN CURVE
0.8
0.6
I
I-1
•^1
o>
o
o
<
tr
Q
0.2
o.o
•
• M-3-4 GAS-DRIVE SAMPLER
A M-3-5 PVC PIEZOMETER
t
A
1 1 1 1 1 1 II
^W
^A- — ~H_»*
A A, A ^^ ^— — *"•
' * y
•^
1 1 1 1 1 1 II
^
1 1 1 1 1 1 II
^:^
i 1 1 1 1 1 ii
.A
/v
<:?
1 1 1 1 1 1 II
0
K
1 1 1 1 1 1 II
10"
IOU
10'
JO*
I03
TIME, minutes
-------
Figure 4-8 shows horizontal contours of the measured
steady-state drawdowns in the fractured bedrock layer. Three
features of this plot are significant. First, note that the
maximum drawdown, occurring at the pumping well (PT-1), is only
2.98 feet. This is unexpectedly low, indicating a very high
hydraulic conductivity. Second, the contours are nearly
circular, indicating that the hydraulic conductivity is nearly
isotropic in the horizontal plane. Both features strongly
support the representation of this uppermost fractured bedrock
zone as an equivalent porous medium (Brown and Gelhar, 1986).
Third the existence of the cutoff wall appears not to have
affected drawdowns in the bedrock.
Steady-state drawdowns in the overburden were vertically
averaged and contoured in Figure 4-9. Drawdowns were less than 1
foot and near the recharge trenches were negative, due to the
incremental mounding. The effect of the cutoff wall is seen in
the 0.4 foot contour, which is skewed towards the northwest. The
skewing likely occurs for two reasons. First, flow from the
northwest experiences greater resistance, thereby developing
larger drawdowns. Second, the recharge from the trenches is more
likely to be felt inside than outside of the wall. Nevertheless,
the former effect may have been partly ameliorated by placement
of the pumping well close to the cutoff wall so that the
(ideally) radial flow paths would be parallel to the wall.
4.1.5 Analysis
Hantush (1967) presented closed-form analytical solutions
for the drawdowns in two aquifers, separated by a semi-pervious
layer, due to pumping from one of the aquifers. This solution is
inherently unsteady, because all of the pumpage must ultimately
come from storage in the aquifers. Therefore, for large time,
drawdowns will diverge. However, during the pumping test, all of
the pumped water was recharged to the overburden via trenches 4,
5, and 6. This recharge provides the source of water allowing
the system to reach steady-state, both in the field and, by
superposition, in the analysis.
Figure 4-10 shows the idealized geometry for the pumping
test analysis. The aquifers are assumed to be infinite, with
homogeneous and isotropic (but distinct) transmissivities T^ and
T2, and uniform storativities Si and 82- (Here, as elsewhere,
subscripts 1 and 2 will refer to the overburden and the bedrock,
respectively). The aquitard is also infinite, with uniform
thickness, b', and homogeneous vertical hydraulic conductivity,
K'. Flow is assumed to be horizontal in the aquifers and
4-18
-------
-]!—"-'
i!
II
„.,
-------
I
ro
o
°o _
FIG. 4-9 MEASURED OVERBURDEN DRAWDOWN CONTOURS
LYLE REED
BROOK
r
J-L.
\
\
p
1
1
1
1
1
1
•
\
1
15 +00
1 '
1 1
1 '
1
/
0' 50' 100' 200'
-------
OD
m
o
%
o
o
inimii
nnTTrmmnr
en
k
rn
t
iiiiniiiniiiiiiiniiiiiiiiiinn
*> Nl!
o 5;'
™
3>
z
m
98
Is
O H
m o
H£
m
z
o
X
m
V)
p
i
O
TJ
C
m
c/>
H
6
m
N
m
o
o
m
o
z
m
-------
vertical in the aquitard. The amount of leakage through the
aquitard is proportional to the difference in heads between the
two aquifers. In addition, storage in the aquitard is neglected.
The actual elevations of the stratigraphic units vary
considerably throughout the site. For example, the top of
bedrock under the recharge trenches is higher than the top of
till near the pumping well. However, approximating these units
as horizontal is reasonable due to the semi-confining nature of
the till.
The recharge trenches have been replaced mathematically by a
single fully-penetrating (through the stratified drift) recharge
well, as shown in Figure 4-10. While the recharge trenches
actually add water only at the water table, approximating the
trenches with a fully-penetrating well results in insignificant
errors when calculating drawdowns near the pumping well.
Let s^j be the drawdown in aquifer i due to pumping at a
rate Qj in aquifer j. Then, Hantush's solution for large time
due to the pumping in the bedrock may be expressed as:
92 =
= 92 { In(ar2) + K0(0r2) }
= 92 { In(ar2) - (Ti/r2) K0(/3r,) }
Qi
r,)
1/2
l\LlO
b'
where *2 is the horizontal distance from the bedrock pumping
well, t is the time since pumping began, and K(X) is the Modified
Bessel Function of the second kind and order zero, of argument x.
By symmetry,
su = 91
s12 = 9! (In(ari) + K0((3ri) }
Qi
91 ~27r(7\
4-22
-------
where r-^ is the horizontal distance from the overburden well
The total drawdowns s^ and 82 are then given by
= q (In(r2/n) + K0(0r2) +- (T./T,)
= 9 { In(r2/n) - (Ti/rj) K0(0r2) -
These solutions were evaluated numerically and contoured for
varying values of the parameters T^, T2» and > commonly known as
the leakance of the aquitard. This procedure was iterative. For
example, if the calculated drawdowns in the bedrock were too
large, then either the transmissivity of the bedrock was
increased, or the leakance of the till was increased, or both.
The values of these parameters which produced analytical contours
providing the best fit to the measured drawdown contours (Figures
4-8 and 4-9) are:
0 overburden transmissivity = 3,500
0 bedrock transmissivity = 6,500
0 glacial till leakance =0.33 d~l
The analytical drawdown contours calculated using these
parameters are shown in Figures 4-11 and 4-12 for the fractured
rock layer and the overburden, respectively. There appears to be
good correspondence between the measured and computed drawdowns
in the bedrock. The overburden match is less accurate. However,
consideration must be accorded the heterogeneous nature of the
geologic setting (especially the existence of the cutoff wall),
the surface water boundaries nearby, and the highly idealized
geometrical assumptions which were made.
The parameters thus calculated in this analysis are
effective homogeneous values. As such, the analytical model can
only represent a kind of mean behavior of the system. In
particular, the irregularities in the overburden must be
neglected.
4-23
-------
I
(0
FIG. 4-11 MODELED BEDROCK DRAWDOWN CONTOURS
.V-
°0
•X
LYLE REED
BROOK
5
o
o
0' 50' 100' 200'
-------
FIG. 4-12 MODELED OVERBURDEN DRAWDOWN CONTOURS
I
t>o
4
.J.J
LYLE REED
BROOK
8
-H
m
O1 50' 100' 200'
-------
In the bedrock, the calculated transmissivity is higher than
had previously been recognized. Assuming the transmissivity is
concentrated in the upper 10 feet of the bedrock, as is
consistent with the logs of the pumping well, PT-1, and
observation wells PT-2, -3, and -4, then this zone has a
hydraulic conductivity of 650 ft/d. The high bedrock hydraulic
conductivity and its observed isotropy are mutually consistent
and agree with the observed high level of fracturing in that
stratum.
The effective homogeneous transmissivity (3,500 ft2/<=i)
calculated for the overburden, however, has probably been
strongly affected by the presence of the cutoff wall, and
therefore, is considered to be somewhat lower than the
transmissivity of the native material alone. Therefore, the
calculated transmissivity is considered to be useful only for
lumping the behavior of the overburden as it affects the other
stratigraphic units in the analysis.
Near the pumping well, the glacial till is approximately 15
feet thick. Given this thickness and the computed leakance, the
glacial till would have an effective homogeneous vertical
hydraulic conductivity of 5.0 ft/d. This is reasonable
considering the discontinuous nature of this stratum, and the
relatively small percentage of fines found in the glacial till
underlying the site.
4.1.6 Conclusions
Based on a comparison between the measured drawdowns in the
bedrock and stratified drift during a pumping test and analytical
drawdowns similarly calculated using a mathematical model,
effective homogeneous transmissivities have been evaluated for
both aquifers. An effective homogeneous leakance for the glacial
till aquitard was also derived. These calculated parameters
compared favorably with descriptive field data.
The transmissivity of the bedrock was found to be 6,500
ft^/d. Based on data obtained during drilling, it is apparent
that the upper ten feet of the bedrock is highly fractured and is
probably responsible for the bulk of the transmissivity. The
hydraulic conductivity of this layer is therefore calculated to
be 650 ft/d.
The calculated transmissivity of the overburden was 3,500
ft2/d. However, the existence of the cutoff wall has undoubtedly
affected this calculation. Therefore, this transmissivity is
4-26
-------
useful only for lumping the behavior of the stratified drift as
it affects the other stratigraphic units.
The leakance of the glacial till was estimated to be
0.33 d~l, a value which has probably been affected by the
discontinuous nature of this stratum. Nevertheless, if the till
is considered to be fifteen feet thick (on average), then the
vertical hydraulic conductivity of the till can be estimated to
be 5 ft/d.
4.2 Cutoff Wall Hydraulic Efficiency
The objective of this portion of the work was to evaluate
the hydraulic efficiency of the cutoff wall. Cutoff efficiency
is dependent upon the wall's bulk, as opposed to intact,
hydraulic conductivity. The intact hydraulic conductivity
represents the in-place backfill material alone, whereas the bulk
hydraulic conductivity incorporates the combined effects of the
intact material and any possible cracks and/or windows in the
backfill and inadequacies of the key to bedrock. It is the bulk
hydraulic conductivity which determines the large-scale hydraulic
behavior of the wall.
The cutoff wall bulk hydraulic conductivity can only be
evaluated indirectly, by hydraulically stressing the upper
aquifer and cutoff wall while observing the hydrologic response.
Ideally, the hydrologic response would be directly related only
to cutoff wall hydraulic conductivity. At the Gilson Road site,
however,both the cutoff wall in the upper aquifer and the
underlying bedrock aquifer affect the flow of groundwater. In
addition, the glacial till aquitard separating the two aquifers
is relatively permeable, rendering it impossible to stress one
aquifer without stressing the other. Therefore, it was ne.cessary
to separate multi-aquifer effects upon the groundwater system
before the performance of the cutoff wall could be assessed.
Data yielded by the bedrock pumping test (Section 4.1) provided a
good estimate of the bulk, large-scale transmissivi ty of the
bedrock aquifer and the leakance of the glacial aquitard. Thus,
the effects of these units could be accounted for in the overall
response of the upper aquifer.
A three-dimensional finite-difference flow model of the site
had previously been calibrated for the design of the hydrodynamic
isolation system. In addition, numerous multilevel monitoring
installations existed at the site, both inside and outside of the
cutoff wall. Therefore, cutoff wall bulk hydraulic conductivity
could be evaluated by artificially increasing the cutoff
hydraulic conductivity in the numerical model and performing a
4-27
-------
sensitivity analysis comparing simulated vertical piezometric
head distributions to the actual field heads. The actual bulk
hydraulic conductivity of the cutoff wall was taken to fall
within the range of conductivities for which the computed and
measured heads matched.
Based upon the above defined range of bulk conductivities,
it was possible to assess the efficiency of the cutoff wall.
Cutoff efficiency was defined as the ratio of flux through the
overburden which had been cut off to that which would have
occurred without the wall. These fluxes were also determined
using the model.
4.2.1 Test Constraints
The investigation of post-construction cutoff wall
efficiency should encompass the entire cutoff wall. The impact
of design/construction issues such as varying wall depth,
overburden type, magnitude and direction of bedrock slope
relative to backfilling direction, etc. could then be addressed.
However, the large effort associated with this work required a
limited scope. As such, the effort concentrated on a 1,000 foot
section of the downgradient wall. This portion of the wall was
judged likely to provide the most salient data with respect to
containment performance for a number of reasons. First, the
large depth of the wall in this location (up to 105 feet)
increased the potential for imperfections and thus for bulk
conductivities exceeding specification limits. Second, under
non-pumping conditions, this area exhibited the greatest head
loss across the wall and thus the highest gradients to drive
leakage through windows, if they existed. Third, the highest
contaminant concentrations were found in the downgradient portion
of the site. As such, a given degree of leakage in this area
would result in the most significant environmental impact.
Fourth, the greatest degree of groundwater manipulation was to be
implemented in this area via the hydrodynamic isolation system.
Additional advantages associated with investigation of the
downgradient wall from a programmatic standpoint were previously
presented in Section 4.1.1.
To evaluate cutoff wall bulk hydraulic conductivity via
sensitivity analysis, the cutoff wall must be hydraul ically
stressed such that the variation in piezometric head in the
vicinity of the cutoff is large in comparison to the calibration
accuracy of the numerical model. To create this condition, the
head difference across the cutoff wall must exceed two feet.
Initially, it was envisioned that the 5 pumping wells associated
with the existing hydrodynamic isolation system (see Figure 4-1)
4-28
-------
could be over-pumped to stress the downgradient wall. However,
the large bedrock transmissivity, as documented by the bedrock
pumping test (Section 4.1), precluded the development of adequate
drawdown in the vicinity of the wall.
The criterion that testing should not cause off-site
contaminant migration eliminated the possibility of increasing
the internal heads as compared to the heads outside the wall.
This could have been accomplished by shutting down the
hydrodynamic isolation system and allowing the groundwater to
mound against the inside face of the downgradient wall. If
necessary, the head difference could have been further augmented
by dewatering Lyle Reed Brook.
The above constraints eliminated potential mechanisms which
could significantly stress the downgradient cutoff wall.
Fortunately, previous water level monitoring data included rounds
conducted after cutoff wall construction but prior to
implementation of the hydrodynamic isolation system. One such
data set, taken in June 1983, showed head differences across the
wall of over 3 feet. This data set was therefore adopted as the
reference data set for the sensitivity analysis.
4.2.2 Reference Data Set
The June 1983 reference data set includes piezometric heads
obtained at 117 points in 56 multi-level and single level
monitoring installations. The observation wells and piezometers
used were installed according to the methods outlined in Section
4.1.3.1 In addition, water levels were taken at 16 surface
stations along Lyle Reed Brook. This data set predates the
bedrock pumping test, and as such, the monitoring installations
installed for that test were not in place. However, the.re were
enough multi-level installations to adequately define the
vertical piezometric head distribution inside and outside of the
downgradient cutoff wall.
Figure 4-13 shows contours of water table elevations during
the June, 1983 reference period. The piezometric heads were
measured over a three-day period, from June 3 to June 5, 1983.
This data was taken after a very wet spring season, and
represents the highest internal/external head differences
measured at the site since cutoff wall construction. At the time
the data was taken, the cutoff wall had just been installed and
the recirculation system was not yet in operation. Near the
eastern (upgradient) end of the containment, the heads outside
the wall built up in response to the low hydraulic conductivity
4-29
-------
'983
83
T
83
o'
-------
of the barrier. The total drop across the upgradient wall was
approximately 2.5 feet. Along the north and south sides, flow
was inward, although the head differences were typically less
than one foot. At the downgradient end of the containment, the
levels inside the wall were over 3.0 feet higher than outside the
containment in response to mounding associated with flux through
the site. Here the water levels outside the wall were largely
determined by Lyle Reed Brook, which has an average slope of
0.002 in this section.
The vertical head distributions near the downgradient
portion of the cutoff wall are illustrated by wells M-14 and
M-10R. Installation M-14 is located inside the containment and
M-10R is located outside of the containment (see Figure 4-1 for
location). The head distributions are shown in Figure 4-14.
Note that due to an error in seal placement, as discussed in
Section 4.1.4.2, the data points for M-14-3 and M-10R-3 were not
used. However, the shapes of the head distributions presented
reflect a compilation of data from other multilevel installations
adjacent to this location.
A number of trends in this data should be noted. First, in
the upper aquifer, the data indicates a head drop of over 3 feet
across the wall. The mounding on the inside of the wall reflects
the impedance generated by the wall to flux through the
containment. Second, on the inside of the wall, piezometric
heads decrease with depth indicating downward flow. The converse
is true outside of the wall. Note that if both the wall and the
bottom aquiclude were totally impervious (no flow), the head
distribution would be vertical on both sides of the wall (no head
loss with depth). Third, the vertical gradient is greater in the
glacial till than in the sand and gravel overburden as would be
expected based on the disparity in hydraulic conductivity of the
deposits. Finally, the head in the bedrock inside the
containment is nearly identical to that outside the containment.
This is consistent with a bedrock of high hydraulic conductivity
(low head loss with flow) confined by a lower hydraulic
conductivity glacial till, as defined during the bedrock pumping
test (Section 4.1). The above data indicates that flux from
upgradient in the containment mounds when it encounters the low
hydraulic conductivity cutoff wall, the mounding drives flow
through the glacial till aquitard, and the flow out of the
containment is then concentrated in the fractured rock below the
wall.
4-31
-------
FIG. 4-14 VERTICAL HEAD DISTRIBUTIONS AT M-IOR 8 M-14, JUNE , 1983
>£>
I
00
NJ
90--
70-
SO-
ia 30 +
«^-
•z.
o
Ul
_l
U)
.0 +
-10--
-30-1-
PIEZOMETRIC HEAD (ft.)
76 77 78 79
OUTSIDE CONTAINMENT
M-IOR
PIEZOMETRIC HEAD (ft.)
78 79 80 81
M-14
INSIDE CONTAINMENT
-------
4.2.3 Numerical Methodology
The complexity of the hydrologic system at the Gilson Road
site renders closed-form solutions unsuitable for analysis of
cutoff bulk hydraulic conductivity. This is particularly true
because the distribution of piezomenric heads reflects the
geologic heterogeneity in both the vertical and horizontal planes.
In this case, over 1,000 lineal feet of cutoff wall was to be
investigated. Therefore, a three-dimensional numerical model was
required to correlate the piezometric head data generated when
stress testing the wall. Such a model had previously been
calibrated to the site for design of the hydrodynamic isolation
system and thus, this capability was readily available.
As previously discussed, the Gilson Road site represents a
coupled system where response to hydraulic stress is a function
of not only the cutoff wall bulk hydraulic conductivity, but also
the characteristics of the glacial till aquitard and the
underlying bedrock aquifer. The first step in isolating the
behavior of the cutoff wall from the rest of the hydrologic
system was to better define the transmissivity and leakance of
the bedrock aquifer and glacial till aquitard, respectively.
This was accomplished via a bedrock pumping test as outlined in
Section 4.1. The hydrologic parameters for the upper sand and
gravel aquifer were known from pumping tests executed as part of
the Remedial Investigation/Feasibility Study. Therefore, the
only remaining hydrologic unit for which the parameters were
unknown was the cutoff wall itself.
Under hydrologic stress conditions, the horizontal vs.
vertical flux through the upper aquifer is a function of the
unknown bulk hydraulic conductivity of the cutoff wall as
compared to the known conductivities of the underlying
aquiclude/aquifer system. The distribution of vertical
piezometric heads reflects the flux vector. For high wall
conductivities, relative to conductivities below the wall, the
flux would primarily be horizontal through the upper aquifer,
yielding little variation in head with depth. For low relative
wall hydraulic conductivities, flux would be more vertical
commensurate with flow under the wall. This condition would be
reflected in varying piezometric head with depth. Naturally, the
degree of variation in the vertical heads is also dependent on
the relative conductivities of the units adjacent to and below
the wall. The multi-level monitoring network established at the
site allowed measurement of the vertical head distribution for
various stress states. This data, in conjunction with the
numerical model, provided the capability for cutoff wall bulk
hydraulic conductivity evaluation via numerical sensitivity
analysis.
4-33
-------
A range of bulk conductivities were input into the
calibrated model and the vertical heads generated were compared
to the actual field head distribution for a given stress state,
the June 1983 scenario. Bulk conductivities which produced heads
matching the field data, to within the model calibration
accuracy, were taken to be possible representations of the actual
state of the cutoff wall. Conversely, simulated bulk
conductivities which did not generate matching heads were assumed
to be invalid.
4.2.3.1 Model Description
The sensitivity analysis was performed using a
three-dimensional finite-difference program (described in
Trescott (1975) and Torak (1982)). Features of this code include
aquifer heterogeneity and anisotropy, steady or transient
simulations, variable grid spacing and a phreatic surface in the
uppermost layer. The solution is found using the strongly
implicit procedure.3
A cutaway view of the model is shown in Figure 4-15. A plan
view of the top layer is shown in Figure 4-16, with the nodes
representing the cutoff wall, Lyle Reed Brook, and Trout Brook
indicated. The model has five layers vertically, and a 48 x 50
grid horizontally. The horizontal grid spacing varies from 180
feet near the boundaries to 40 feet in the vicinity of the cutoff
wall. The bottom layer of the model is 50 feet thick, with its
center at elevation -28 feet (with respect to the City of Nashua
datum). The next layers are 30, 20, 20, and 50 feet thick,
respectively. The top layer is artificially thick to allow the
water table to vary without restraint. This extra thickness is
not shown in Figure 4-15.
Most of the borders of the model, including those
representing constant heads, were no-flow boundaries. The one
exception is at the downgradient end of the model, where a fixed
constant flux was simulated by a series of virtual pumping wells.
Precipitation-generated recharge was applied uniformly as a
source to the nodes in the top layer, except in areas covered by
the impermeable surface cap. The recharge which would have been
applied in these areas was added to nodes around the cap, in
accordance with the known cap contours.
3A numerical technique used to solve the piezometric head matrix
4-34
-------
ELEVATION,feet
Ul
o
1
LAYER
o
oj en -si to
O 0 O 0
i i i i
r r r r
$55$
3333
r-O
.o
*^o
o
X
o
5
M
O
CO
o
m
(O
ro
_o
^
OJ
_o
o
o
o
o
o
2)
p
i
oi
O
c
i
5
<
m
o
-n
z
c
S
m
o
o
O
m
-------
FIG. 4-16 PLAN VIEW OF TOP LAYER OF MODEL
I
CO
MARSHES
o1
-------
During the calibration procedure, the cutoff wall hydraulic
conductivity simulated was the design hydraulic conductivity of 1
x 10-7 cm/sec, as based on quality control test data (Section 2).
The actual conductivities of the model nodes representing the
wall were, however, ten times greater in order to account for the
40 foot node width as compared to the 3 to 4 foot wall thickness.
In order to reduce complexity, the model was used to
simulate steady-state conditions only. No transient simulations
were performed for this effort. Each simulation treated all of
its boundary conditions as constant, although the values of the
boundary parameters were changed from simulation to simulation.
In cases where a condition was actually time-varying, it was
replaced by an equivalent steady-state condition. The most
obvious example of this is the areal recharge due to
precipitation. For example, in the calibration for the June,
1983, scenario, the applied recharge was 42 inches/year. This
value exceeds the probable yearly average recharge to the area by
a factor of two. However, it is reasonable as an effective
steady-state recharge for the three-month period preceding that
date, as based on meteorological data.
The other transient conditions simulated by the model were
judged to be changing slowly enough that their values at any
point in time could be (and were) used as steady-state input.
For example, constant head nodes were used to represent Lyle Reed
Brook and Trout Brook, as well as the swamps and marshes
upgradient of the site. The specified heads varied spatially and
between simulations, but were constant at any one node during any
one simulation.
4.2.3.2 Model Calibration
The model had originally been calibrated for two water level
data sets, representing the system's state in January, 1982, and
June, 1983. The current effort included further calibration to
reproduce two additional water level data sets, (October 1983 and
September 1985), for a total of four different stress conditions.
(The four data sets will be abbreviated as JAN82, JUN83, OCT83,
and SEP85). The stress conditions which these scenarios
represent are shown in Table 4-2. The additional calibration
effort incorporated improved understandings of system behavior as
well as new, better and/or more extensive data. In particular,
the bedrock transmissivity which was determined from the bedrock
pumping test was explicitly incorporated. Also, further data
regarding seasonal variations in surface water levels proved
helpful in improving the ability of the model to represent the
behavior of the aquifer/cutoff wall system in response to
different recharge rates.
4-37
-------
Parameters varied during model calibration included the
hydraulic conductivities and anisotropies of the soils, the
amount and distribution of recharge, and the amount of subsurface
drainage (constant flux) out of the model at the downgradient end.
The values of recharge and constant flux used in each of the
calibrated simulations are shown in Table 4-2. In all cases,
parameter variation remained within ranges previously established
based on the subsurface data amassed for the site.
TABLE 4-2
STRESS SCENARIOS AND SIMULATION FLUXES
Date
Wall &
Cap
H.I.
On
1
Sys.
Constant2
Flux (cfs)
Recharge
(in/yr) Descriptive
JANS 2
JUN83
OCT83
SEP85
no
yes
yes
yes
no
no
no
yes
0.240
0.363
0.128
0.240
20
42
6
8
medium
high
low
low
1 Hydrodynamic isolation system
2 Flux rate for downgradient boundary condition
Three measures of the degree of calibration were used.
First, statistics were calculated for the deviations between
measured piezometric heads and those calculated at the homologous
nodes in the model. This provided a quantitative estimator of
the degree of calibration as a whole. Second, visual comparison
of water table elevation contours ensured that important
qualitative features were preserved. Finally, care was exercised
to see that the vertical distributions of modeled heads matched
well with the field data.
The statistical measures used were the maximum and minimum
deviations between measured and modeled heads, the average
deviation, the average of the absolute deviation, and the RMS
deviation. The latter measure is equal to the standard deviation
when the average deviation is zero. These statistics are
presented for each calibration run in Table 4.3 and indicate that
all of the runs are calibrated to a high degree.
4-38
-------
TABLE 4-3
CALIBRATION DEVIATION STATISTICSl
Date Maximum Minimum Average Ave. ABS RMS
JANS 2
JUN83
OCT83
SEP 8 5
0.84
1.11
0.75
1.33
-0.77
-0.70
-0.87
-0.64
-0.08
0.06
0.04
0.17
0.27
0.33
0.30
0.29
0.35
0.41
0.36
0.43
1A11 values in feet. Data analysis included the piezometric
head data from all five layers.
Figures 4-17, 4-18, 4-19, and 4-20 show the measured water
table contours for the four calibrated scenarios and their
modeled counterparts. In general, the calibration was quite good
as demonstrated by the figures, although there were some minor
discrepancies which could not be resolved. In JAN82, most of the
contours match, although the modeled heads were somewhat lower
than the measured heads in the southern portion of the site. The
OCT83 attempt however, suffers with respect to the contours from
78 to 80 feet. In SEP85, the field data shows a steep gradient
from trenches 4, 5, and 6 towards pumping wells A through E, but
then a very flat gradient from the wells to the wall. This
abrupt change in slope was difficult to capture in the model,
resulting in a high maximum deviation for the run. It is
hypothesized that the mismatch was due to the effects of
simulating the wells and trenches with the relatively coarse grid.
Nevertheless, the discrepancy should not strongly affect leakage
estimates because it is located away from the highly sensitive
upgradient and downgradient ends of the site. Note that the
modeled head differences across the downgradient portion of the
wall are similar to those shown by the field data.
4.2.4 Sensitivity Analysis
The sensitivity analysis was performed by varying the cutoff
wall bulk hydraulic conductivity (the parameter under
investigation) in the model and evaluating the degree of
correspondence between the numerically generated head
4-39
-------
FIG. 4-17 JAN. 82 SCENARIO CALIBRAT ION
-------
Oo,
,0
-------
FIG. 4-19 OCT. 83 SCENARIO CALIBRATION
76
\
I
^
CO
83
78
79
MEASURED
MODELED
0' 50' 100' 200'
-------
FIG.4-20SEPT. 85 SCENARIO CALIBRATION
\
i
*»
u>
8] 81
82 82 83
76
83
77
'78
MEASURED
MODELED
79
o1 so1 100' 200'
-------
distributions and the actual field measured data. The bulk
hydraulic conductivity was varied over four orders of magnitude
in five equal steps. The initial value input for bulk hydraulic
conductivity was 1 x 10~7 cm/sec. This value, equal to the
specification criterion, was used during model calibration^.
The next four sensitivity runs used precisely the same input
data, except that the cutoff wall bulk hydraulic conductivity in
each run was increased by a factor of ten over the previous run.
The greatest bulk hydraulic conductivity used during the
sensitivity analysis was 1 x 10"3 cm/sec. This value is within
one order of magnitude of the average hydraulic conductivity of
the upper sand and gravel aquifer (10~2 cm/sec).
The results of the sensitivity analysis are shown in
Figure 4-21. For each run, the computed heads at the nodes
corresponding to multilevel wells M-14 and M-10R were plotted and
compared to the field data. The first run (1 x 10~7 cm/sec)
generated a head distribution which matched the field data to
within about one quarter of a foot. On the inside of the wall
the simulated heads were somewhat higher than, but closely
followed the trend of, the measured heads. Outside the
containment, the simulated heads again followed the trend
exhibited by the measured heads, but were not consistently
greater or less than the measured heads. This data set
represents the calibration accuracy of the numerical model in the
vicinity of the downgradient wall. As previously indicated, it
was generated as the final run used to calibrate the numerical
model for the JUN83 scenario. The second run (1 x 10-6 cm/sec)
generated heads that were nearly identical to the first run. A
single piezometric head distribution was therefore presented on
Figure 4-21 for both runs.
For the third run, the bulk hydraulic conductivity of the
cutoff wall was increased to 1 x 10~5 cm/sec. The generated
heads in this case diverged noticeably from the first two runs.
However, they still correlated with the field measurements to
within the error band defined by the accuracy of the initial
calibration. Therefore, 1 x 10~5 cm/sec, as well as lOx 1~6 and
1 x 10~7 cm/sec, represents a possible cutoff wall bulk
hydraulic conductivity to the extent that it meets the
constraints imposed by the sensitivity analysis. Further
increases in bulk hydraulic conductivity (runs 4 and 5) resulted
in head distributions which no longer matched the field data. As
such, these values (1 x 10~4 and 1 x 1Q-3 cm/sec) do not
4The value of 1 x 10~7 cm/sec used for model calibration was
selected as based on quality control field data obtained during
cutoff wall construction; see Section 2.
4-44
-------
FIG. 4-21 MODELED VERTICAL HEAD DISTRIBUTIONS FOR DIFFERENT WALL CONDUCTIVITIES
PIEZOMETRIC HEAD (ft.)
76 77 78 79
PIEZOMETRIC HEAD (ft.)
78 79 80 81
90 -r
70--
50--
I
£>
Ul
OJ
30--
UJ I0-|-
-10--
-30-1-
OUTSIDE CONTAINMENT
INSIDE CONTAINMENT
SIMULATED BULK
CONDUCTIVITY
I07, I06 CM/SEC
10s CM/SEC
VERTICAL HEAD
DISTRIBUTION
10 CM/SEC —-.<
SIMULATED BULK
CONDUCTIVITY
103 CM/SEC
MEASURED, 6/83
VERTICAL HEAD
DISTRIBUTION
-------
represent valid possibilities for the bulk hydraulic conductivity
of the cutoff wall.
The above analysis indicates that bulk conductivities lower
than 1 x 10~5 cm/sec are all, to the extent definable by the
sensitivity analysis, plausible representations of the cutoff
wall's impedance to flux. The analysis also indicates that bulk
conductivities in excess of 1 x 10~5 Cm/sec begin to yield head
distributions which no longer match the field condition and are
thus considered invalid. It is emphasized that the analysis only
demonstrates that the value of 1 x 10~5 cm/sec is an upper bound
on the cutoff wall hydraulic conductivity. This should not be
misinterpreted to mean that the actual bulk hydraulic
conductivity of the cutoff wall is 1 x 10~5 cm/sec. In fact,
significant data indicates that the actual value is much lower,
in the range of 1 x 10"7 cm/sec (see Sections 2 and 3).
The inability of the sensitivity analysis to provide a
definitive evaluation of the cutoff wall bulk hydraulic
conductivity, with respect to the specification criterion of 1 x
10~7 cm/sec, is a direct result of the much higher conductivities
associated with the containment bottom (glacial till and bedrock).
This is due to the coupled hydraulic behavior of the containment
system which is analogous to an electrical circuit containing
resistors in parallel. As such, the sensitivity analysis can
only resolve the hydraulic conductivity of a specific individual
hydraulic element to a degree proportional to the highest
hydraulic conductivity exhibited by any element in the system.
This is demonstrated by the data in Figure 4-21. As the cutoff
wall hydraulic conductivity is increased from 10-7 cm/sec, the
bedrock flux continues to dominate the heads, and thus the amount
of fluid flux through the wall increases roughly linearly with
it's hydraulic conductivity. The increase in flux through the
wall is small enough, however, that it has no discernable effect
on the heads. As the wall flux eventually becomes significant
with respect to the bedrock flux, a concomitant decrease in the
overall gradient occurs as reflected in the vertical head
distribution. The high bedrock transmissivity, therefore
effectively masks the cutoff wall hydraulic conductivity below
10~5 cm/sec. If the bedrock had a lower transmissivity, then
this critical masking level would occur at a lower wall
hydraulic conductivity. An upper bound could then be established
which would be more consistent with the value specified for the
cutoff wall bulk hydraulic conductivity (1 x 10-7 cm/sec). This
limitation is inherent in any hydraulic stress analysis.
Furthermore, in as much as hydraulic analysis is the only
mechanism available to evaluate bulk hydraulic conductivity, the
limitation applies to the viability of post-construction
verification in general.
4-46
-------
4.2.5 Cutoff Wall Efficiency
The effectiveness of the cutoff wall as a barrier technology
is a function of its ability to arrest fluid flux. It is
therefore necessary to determine the amount of flux before and
after the installation of the wall. The effectiveness of the
wall may then be quantified as the ratio of flux through the
overburden which has been eliminated by the cutoff wall, to that
which would have existed without the wall. This ratio is defined
herein as the "wall efficiency".
An alternative definition of efficiency is the ratio of
total flux through the entire site (overburden and rock) which
has been cut off to total flux through the entire site before the
installation of the wall. However, this ratio is suited only to
an evaluation of the effectiveness of the entire passive
containment system (cap, cutoff wall and bottom aquiclude) and
not the effectiveness of the wall as a technology per se. This
ratio is termed the "containment efficiency".
The flux values required to compute the above ratios were
determined using a postprocessor adjunct to the flow model. In
this program, the transmissivity and head matrixes are used to
calculate the fluid flux out of the site through each node in or
under the simulated cutoff wall. The resulting fluxes are
reported by layer, by horizontal node, and by element type (wall
or rock).
In addition to the three calibration runs5 JAN82, JUN83,
and OCT83, six more runs were performed on the model. These six
runs represented the system under the combined conditions of low,
medium, and high recharge, with and without the cutoff wall. A
cutoff wall hydraulic conductivity of 1 x 10~7 cm/sec was used
for the wall. The postprocessor was used to calculate the fluid
fluxes through the rock and overburden/cutoff wall nodes in each
of these runs. The results are shown in Table 4-4.
5The SEP85 run is not included herein because it models a scenario
with the hydrodynamic isolation system in operation. As such,
flux through the site/wall is a function of pumping rates as well
as barrier conductivities.
4-47
-------
TABLE 4-4
MODELED OUTWARD FLUXESl
No Containment Wall and Cap
Recharge Wall2 Rock Wall3 Rock
low
medium
high
74,386
89,248
121,982
29,635
32,833
39,116
73
79
93
65,001
66,578
73,668
values in gallons per day
2native overburden permeabilities
3wall hydraulic conductivity = 1 x 10~7 cm/sec
Table 4-4 shows that, before installation of the cutoff
wall, the flux through the rock is about one-fourth of the total
flux through the site. After cutoff installation, the rock flux
is roughly doubled, due to the increased gradients created by the
wall. However, the flux through the overburden is effectively
eliminated. Therefore, the containment efficiency of the passive
physical barriers at the Gilson Road site is approximately
55 percent under worst case precipitation induced stress
conditions. Please note, however, that this does not include the
effects of the hydrodynamic isolation system, (the active
component of the total containment) which is designed to
completely eliminate off-site migration of contaminated
groundwater.
The sensitivity runs used to establish an estimate for
cutoff wall bulk hydraulic conductivity were also passed through
the postprocessor. The calculated fluxes out of the site for
these runs are shown in Table 4-5. Based on these fluxes and the
entry in Table 4-4 for high recharge^ no containment, it is
possible to calculate the wall efficiency for each run. Based on
the sensitivity analysis, the cutoff wall bulk hydraulic
conductivity was determined to be less then 1 x 10~5 cm/sec.
Therefore, Table 4-5 shows that the wall efficiency must be
greater than 93 percent.
6The high recharge data for the no containment case was selected
to match the JUN83 scenario upon which the sensitivity analysis,
and thus the data in Table 4-5, was based.
4-48
-------
TABLE 4-5
SENSITIVITY RUN FLUXES
Cutoff Wall
Bulk K
(cm/ sec)
10-7
10-6
10-5
10-4
10-3
Flux through the Site
(
Total
73,761
74,117
77,417
95,800
126,165
gallons per
• Wall
93
830
8,361
45,036
90,814
day)
Rock
73,668
73,287
69,056
50,764
35,351
Wall
Efficiency
99.92%
99.32%
93.15%
63.08%
25.55%
4.2.6 Summary
The bulk hydraulic conductivity of the downgradient section
of the Gilson Road cutoff wall was evaluated using a numerical
sensitivity analysis. Three-dimensional finite-difference flow
modeling provided comparison of simulated vertical piezometric
head distributions to actual field measures head distributions
for varying values of cutoff wall bulk hydraulic conductivity.
An upper bound of 1 x 10~5 cm/sec was thus established, for the
cutoff wall. Further definition of the actual bulk hydraulic
conductivity below this value was not possible using stress
testing due to the large bedrock tramsmissivity and the practical
limits on the calibration of the model. However, the
considerable quality control data obtained during construction
(Section 2) indicates that the hydraulic conductivity of the wall
is approximately 1 x 10"^ cm/sec. This value is consistent with
the stress testing results in that stress testing yielded an
upper bound only.
Inasmuch as only an upper bound value for the cutoff wall
bulk hydraulic conductivity could be determined via stress
testing, evaluation of the wall efficiency (ratio of the
overburden flux eliminated to the overburden flux without the
containment) was limited to a lower bound solution. Based on the
4-49
-------
results of the sensitivity analysis, the wall efficiency for the
cutoff wall was found to be greater than 93 percent. This is
based solely upon the upper bound value for bulk hydraulic
conductivity established by the sensitivity analysis, and as such
is a worst case condition. A more probable value for wall
efficiency is over 99% as obtained based on the quality control
data which indicates a wall hydraulic conductivity of
approximately 1 x 10~7 cm/sec.
4.3 Groundwater Quality Monitoring
As part of the evaluation of the effectiveness of the cutoff
containment wall installed at the Gilson Road site, groundwater
samples were obtained and analyzed. The data generated by the
groundwater sampling and analysis program provided direct
evidence of remedial action performance by establishing
contaminant location in temporal and spatial dimensions. The
groundwater quality data, when used in concert with the hydraulic
evaluations (Sections 4.1 and 4.2), aided in the evaluation of
the long-term environmental impact and overall effectiveness of
the cutoff wall containment.
Useful information, such as the determination of containment
flow paths, was gained by comparing groundwater contamination
data from monitoring points located within the same geologic
strata but on opposite sides of the containment wall.
Additionally, groundwater quality data collected over a
relatively long period of time allowed evaluation of the
effectiveness of milestone events which have occurred at the
Gilson Road site.
4.3.1 Location and Installation of Groundwater Monitoring
Wells
The installation of groundwater monitoring wells on the
Gilson Road site dates back to 1979. Since that time both single
and multi-level monitoring wells have been placed at strategic
locations to provide data on the horizontal and vertical
distribution of contaminants (see Figure 4-22). Monitoring wells
installed by the state of New Hampshire in 1984 were placed
primarily on the downgradient, or western, end of the site. The
location of the monitoring points were selected so as to provide
water quality data in similar geologic formations but on opposite
sides of the cutoff wall. The geologic formations consist of a
relatively thick sand and gravel deposit overlying a glacial till
4-50
-------
20'
FIG. 4-22 LOCATION OF GROUNDWATER MONITORING POINTS
M-
+ _^r
&
o
o
8
_f_
-+-
-+-
Ae
+ +*
-f
-f
-4-
LYLE REED
BROOK
A EXTRACTION WELL
r™_T_rr RECHARGE TRENCH
MULTI-LEVEL MONITORING POINT
it'OO'
-f- SINGLE-LEVEL MONITORING POINT
STREAM SAMPLING POINT
200'
-------
which, in turn, overlies fractured bedrock. Monitoring points
were placed within the sand and gravel deposits and at the
interface of the sand and gravel and till deposits and at the
till-bedrock interface.
Installation of the monitoring points was consistent with
the methodology discussed in Section 4.1.3. A typical
installation is depicted in Figure 4-3.
4.3.2 Sampling and Analysis
4.3.2.1 Sampling of Monitoring Points
The sampling of the monitoring points to obtain groundwater
samples has occurred since 1979. Samples were collected over a
series of milestone events which have occurred on the site.
These events include the initial site assessment period, the
installation of an emergency recirculation system, the
emplacement of a cutoff wall and impermeable cap, and the
operation of a permanent groundwater hydrodynamic
isolation/recirculation system. Over 330 chemical analyses were
completed for the site during this period. Sampling rounds just
prior to and subsequent to cutoff wall installation have included
70 analyses.
Observation wells were sampled using 1-inch stainless steel
bailers. Three times the initial volume of the well was
evacuated prior to collection of groundwater quality samples.
Gas-driven sampling points were sampled using ultra-high purity
nitrogen gas in accordance with the methodology outlined in
Barvenik, 1984. Groundwater samples collected from the site were
placed in ice-filled coolers and delivered to EPA certified
analytical laboratories for chemical analyses.
4.3.2.2 Analysis of Groundwater Samples
Groundwater samples collected from The Gilson Road site were
initially analyzed for the full range of priority pollutants. In
addition, a number of non-priority pollutants were also
identified. This data indicated the primary contaminants on-site
were volatile organic compounds (VOCs). VOCs were therefore used
to characterize and monitor contaminant levels both spacially and
temporally. VOC analyses were executed using EPA Method 624 for
priority list pollutants and equivalent methodologies for VOCs
4-52
-------
not found on the priority pollutant listing but known to exist
on-site.
4.3.3 Results of Groundwater Quality Monitoring
4.3.3.1 Selection of Pertinent Groundwater Monitoring
Points
As previously mentioned, a large number of groundwater
monitoring points have been installed on the Gilson Road site.
These points were sampled for groundwater quality information
since 1979. As a result, a large volume of data exists regarding
the distribution of contaminants on the site. In order to
evaluate changes in groundwater quality in response to the
installation of the cutoff wall, a subset of all of the
groundwater monitoring points was chosen for detailed analysis.
The monitoring points selected for this analysis satisfied the
following criteria:
0 The monitoring points were located at the downgradient
portion of the site where head loss across the cutoff
wall, and therefore resulting contaminant migration out
of the site, is greatest;
0 The monitoring points were composed of multi-level
installations located within the three major geologic
strata present on the site, whenever possible, in order
to provide vertical contamination distribution
information, and
0 The monitoring points were sampled over an extended
period of time covering the major milestone events
which occurred during remediation of the site.
Based upon the above requirements, five groundwater
monitoring points were chosen for detailed analysis. These five
monitoring points show contaminant concentrations which are
representative of the general trend in contaminant distribution
of the downgradient portion of the site. In addition, a surface
water sampling point located downgradient of the cutoff wall and
along the monitoring point cross-section line was included in the
4-53
-------
analysis?, Tt should be noted that while the monitoring well
concentrations are representative of the site in general, data
gaps do exist throughout the sampling history. These data gaps
were typically the result of changing conditions on the site
during the remediation effort. Monitoring well placement and
sample selection were primarily in response to specific data
needs associated with remediation and as such were not always
compatible with the cutoff wall evaluation effort.
The location of monitoring wells, as they ultimately relate
to the cutoff wall evaluation, were influenced by several
factors:
0 As monitoring wells were destroyed and replaced over the
relatively long history of the site, new locations for
wells were selected so as to reflect crucial data
requirements at the time of installation. As such,
destroyed monitoring locations were not always replaced.
0 Prior to the installation of the cutoff wall, there was
no reason to place monitoring wells in close proximity to
each other. Such an alignment only became productive
once the wall was in place to allow comparison of
contaminant levels inside and outside of the containment.
Given the large number of sampling points located on the
Gilson Road site, not all monitoring points could be sampled
during each round of data collection. The selection of sampling
points was also influenced by several factors:
0 The selection of sample locations was in response to the
historic location of the contaminant plume on the site.
The dynamics of the plume meant that different sampling
points would be monitored so as to track plume movement;
0 The selection of sample locations under the current
contract was performed under a limited budget and was to
meet multiple objectives. The groundwater quality data
was collected for the purpose of cutoff wall evaluation
and also to supply data to the State of New Hampshire for
performance evaluation of the groundwater extraction and
treatment system.
'In order to bridge the milestone events, data from two surface
water sampling points were combined into one point. The two
surface water monitoring points are located in close proximity to
each other and show similar concentration levels and trends on
dates where analyses were performed on both sample points.
4-54
-------
The result of the competing factors for monitoring well
placement and sampling location yields a history of groundwater
quality on the site which contains some data gaps. However, the
monitoring points chosen for detailed analysis give a general
trend in contaminant distribution which can be used to evaluate
the effectiveness of the cutoff wall. The monitoring point
cross-section line is shown in Figure 4-23.
4.3.3.2 Selection of an Indicator VOC Compound
The selection of a single compound as an indicator for total
VOCs allows for accurate comparison of data collected over an
extended period of time. This allows the data to be less
sensitive to changes in the priority and non-priority pollutant
VOC list for which the samples were analyzed. Since a large
percentage of the total VOCs detected on the site were
non-priority pollutant VOCs, the groundwater samples were not
always analyzed for the same suite of compounds. Therefore, the
comparison of data collected over a long period of time is more
accurate in th'is case through the use of an indicator compound
than through comparison of total VOC concentrations.
The results of the chemical analyses for VOCs on groundwater
samples from the Gilson Road site indicated the presence of one
VOC at very high levels in almost all samples. Tetrahydrofuran
(THF) comprised more than 90 percent of the VOCs present in most
of the groundwater samples collected from the site. This
compound was selected an an indicator for total VOC distribution
on the site for the following reasons:
0 Tetrahydrofuran, in most samples, comprises a large
percentage of the total VOCs present in the sample. In
most cases it is present in excess of 90 percent of the
total VOCs;
o
THF is relatively soluble in water and is not readily
absorbed by soil particles; and
0 THF is not significantly biodegradable in the subsurface
environment.
As a result, the compound makes a good indicator for
contaminant distribution and allows for comparison of groundwater
quality data over relatively long periods of time. This data is
summarized in Table 4-6.
4-55
-------
FIG. 4-23 LOCATION OF SELECTED GROUNDWTER MONITORING POINTS
i
171
A EXTRACTION WELL
MULT I-LEVEL MON ITORING POINT
SINGLE-LEVEL MONITORING POINT
•+• STREAM SAMPLING POINT
LYLE REED
BROOK-
0' SO' 100' 200'
-------
TABLE 4-6
CONCENTRATIONS OF TETRAHYDROFURAN IN PARTS PER MILLION
Well Sampling Late
Location 7/81 1/82 3/82 5/82 7/82 11/82 4/83 1/85* 10/85
A-10
M-23-1
M-23-2
M-23-4
M-3-1
M-3-2
M-3-3
M-3-4
NHDB-04
M-4-1
M-4-2
M-4-3
M-4-4
M-7-2
M-7-4
850
16
26
68
170
12
54
34
— ' 4.4
— —
— —
— —
572 46
3.2 5.0
25.1 17.8
34.2 —
— —
1.2 1.3
— —
— —
— —
— —
10
5.2 —
178 55
0.89 —
1.8 ~
2.2 12
68
0.4 0.9
3.7
33
1.3
46
8.8
— —
59
150
83
84
17
1.4
1.9
—
110
—
110
76
10
0.04
0.69
5.0
0.3
—
—
ND
35
—
—
0.045
10.0
80
ND
—
—
—
—
0.06
80
85
80
— Indicates data not taken
*QA/QC data indicates a possible laboratory dilution error.
4.3.3.3 Data Discussion
The distribution of THF for the selected groundwater
monitoring points is displayed both spatially and temporally in
Figures 4-24 a and b. An examination of the data contained in
the figures reveals several trends in the distribution of the
contaminants in response to the milestone events which occurred
on the site.
A. Contaminant Distribution During Initial Site Assessment
Contaminant distribution on the site, as indicated by
the initial site assessment data, shows the movement of
contaminants from the center of the site towards Lyle
4-57
-------
T5
it*
I
Ln
CO
45
CO
Z J,
o
t-
-15
FIG. 4-24b VARIATIONS IN CONTAMINANT DISTRIBUTION
M-4-
LEGEND:
SAND ft GRAVEL
GLACIAL TILL
BEDROCK
20' 40*
HORIZONTAL SCALE
•0'
SAMPLE
DESIGNATION
1
t
3
4
s
,
T
•
SAMPLE
DATE
T/SI •
l/St f
S/U /
j
ft/tt y
r/M /
^
^
4/«S
I/W
IO/S9
MILESTONE
EVENT
INITIAL IITE
ASSEMIIENT
CMIt«ENCT
• ECI*CULATION
SYSTEM
CUTOrP WALL
INSTALLATION
PERMANENT
RECIKCULATION
srsTE*
CONCENTRATIONS OF TETRAHYDROFURAN (THF)INPARTS
PER MILLION (•«/!). ND INDICATES NO THF DETECTED.
-------
85
43-
W 35
I
U1
5
0-
-s-
-IS'
-29-
-35..
FIG. 4-24b VARIATIONS IN CONTAMINANT DISTRIBUTION
CUTOFF *
WALL — M-23-
A'
M-3-
20' 40'
HORIZONTAL SCALE
80'
A-10-
-------
Reed Brook, on the northwest boundary of the site. This
movement is reflected in increasing concentrations of
contaminants in water samples collected both from the
brook and in the vicinity of the future cutoff wall.
Concentrations in Lyle Reed Brook peaked at 12 parts per
million during the July 1981 sampling round.
Concentrations of THF at M-3 also showed peak
concentrations during this time frame ranging from 16 to
170 parts per million at the various sample levels
(concentrations at M-3 increased with depth).
B. Contaminant Distribution in Response to Emergency
Recirculation System
Concentrations at the downgradient portion of the site
began to decrease after the start-up of the Emergency
Recirculation System. Levels decreased during operation
of the system (February, 1982 through November, 1982) in
the stream sampling point, NHDB-04, and in the
downgradient wells, M-3 and A-10. This response in
concentrations to recirculation system operation shows
the system retarded the advance of the contaminant plume
by recirculating water back to the central portion of
the site.
C. Contaminant Distribution in Response to Installation of
Cutoff Wall
The response of contaminant distribution to the
installation of the cutoff wall were twofold. Just
prior to installation of the cutoff wall at the
downgradient portion of the site, the emergency
recirculation system was shut down in order to allow
grading, pumping well extension, and impermeable cap
placement. As a result, the advancement of the plume,
which had been halted and reversed during the
functioning of the system, once again began. This plume
movement was reflected in contaminant levels at the
downgradient monitoring wells.
The cutoff wall was completed in November, 1982.
Monitoring points located downgradient of the wall
continued to show an increase in concentration. This
increase was in response to movement of contaminants
beyond the cutoff wall location while the emergency
recirculation system was shut down. This pulse of
contaminants continued to be observed over time as it
moved downgradient. In general, at the time of the
4-60
-------
completion of the cutoff wall, the levels of
contaminants were highest within the bedrock.
With the completion of the wall, contaminant levels in
the area of the stream again decreased. In addition,
concentrations inside the cutoff wall remained higher
than levels outside the cutoff wall. Contaminant levels
in the vicinity of the cutoff wall were highest in the
bedrock. These levels indicate flow in the vicinity of
the cutoff wall was occurring within the fractured
bedrock and not through the wall itself.
D. Contaminant Distribution in Response to the Permanent
Recirculation System
Currently, contaminant levels within the cutoff wall
have decreased as compared to levels measured subsequent
to cutoff wall completion. This decrease is primarily
due to the functioning of the permanent recirculation
system. Pumping of groundwater within the cutoff wall
has had the effect of mixing or homogenizing the
groundwater. As a result, levels within the cutoff wall
have dropped as highly contaminated zones were mixed
with zones exhibiting lower concentrations. However,
even with the homogenizing effect, levels within the
cutoff wall are higher than those outside of the wall.
4.3.3.4 Contaminant Degradation of the Cutoff Wall
-The cutoff wall installed at Gilson Road has performed as a
multi-functional containment structure. Initially, the wall
served as a temporary measure to impede the off-site migration of
contaminants. With the start up of the permanent hydrodynamic
isolation system, the cutoff wall now functions as a clean water
exclusion barrier to inhibit the flux of clean water back onto
the site in response to groundwater pumping. While performing
the initial containment function, the wall was exposed to
contaminants. As a result, the potential for chemical
degradation of the wall was investigated.
Two major factors control the time required for cutoff wall
degradation:
0 The number of pore volume displacements required to
affect a chemically mediated change in the wall hydraulic
conductivity;
0 The rate at which leachate flows through the wall.
4-61
-------
Long-term hydraulic conductivity testing of the Gilson Road
backfill with worst case leachate from the site indicated that
displacement of two to three pore volumes are required to effect
changes in the cutoff wall hydraulic conductivity (Schulze, 1984).
The testing, which simulated in-situ conditions, indicated a
maximum increase to twice the initial hydraulic conductivity (1.2
x 10~8 to 2.5 x 10~8 cm/sec). These results agree with
literature documentation on chemical degradation of clay based
barriers.
The flow of contaminated groundwater through the cutoff wall
intact backfill is dependent upon the hydraulic gradient (j.)
across the wall and the intact hydraulic conductivity of the wall
(fc). The hydraulic gradient is determined by the ratio of the
total head difference across the wall as compared to the width of
the wall (L). The intact hydraulic conductivity of the wall was
determined by hydraulic conductivity testing of cutoff wall
samples (Section 2).
Flow through a porous medium is governed by Darcy's Law (Q =
kiA) where Q is the volume of water flowing through a
cross-sectional area of the medium (A). The average velocity of
the flow through the medium is described by the equation Vs =
ki/n , where n is the porosity of the material. The time
required to displace one pore volume can be determined by the
equation t = W/Vs, where W is the width of the medium through
which flow is occurring.
In order to determine the time to displace one pore volume
of water through the wall, certain conservative, or worst-case,
assumptions were made in the selection of parameter values.
The greater the hydraulic gradient (^) across the wall, the
faster the wall will degrade. The maximum head difference
observed on the site was three feet. This condition only
occurred during one period of particularly high precipitation
without hydrodynamic controls. A minimum width of the wall would
be equal to the width of the bucket used to excavate the
cutoff wall trench. A 3 foot bucket was used. This results in a
maximum hydraulic gradient of j_ = 1.
The greater the hydraulic conductivity of the wall, the
faster degradation should occur. The hydraulic conductivity of
the wall is inversely proportional to the level of stress on the
wall. The stress on the wall increases with depth to a maximum
of 50 pounds per square inch (psi) at the bottom of the wall (105
feet). In order to simulate worst case conditions, hydraulic
conductivity testing of the cutoff wall was conducted using
stress conditions corresponding to the top of the cutoff wall, or
4-62
-------
3 to 5 psi. Although design phase laboratory testing and
construction QC testing in the field yielded an average cutoff
wall hydraulic conductivity of 5 x 10~8 cm/s, a more conservative
value of 1 x 10"7 cm/s was used for determining pore volume
displacement rates as based on work completed under phase one
(Section 2) of this contract.
The porosity (n) of a material is that portion of the
material not occupied by solid matter relative to its total
volume. If all other factors remain constant, a lower porosity
will result in a higher flow velocity, which will cause a faster
degradation of the cutoff wall. Based on QC data, a conservative
value of n = 0.4 was chosen.
A computation of the time to displace a single pore volume
is determined using the equations stated previously:
t = W; where
Vs
Vs = ki
n
Using the conservative, or worst case, values stated above,
the form of the equations becomes:
Vs = 1 x 10-7 cm/s (i) - 2.5 x 10~7 cm/s.
0.4
t = 91.44 cm = 3.7 x 108 seconds;
2.5 x 10-/ cm/s
t = 11.6 years
The time to displace one volume of pore water using worst
case parameters is calculated to be approximately eleven years.
However, under the average conditions which have actually
occurred on the site, the time required to displace a single pore
volume is approximately fifty years. As the testing of the
degradation of the cutoff wall material has shown, chemical
degradation of the material was complete only after two pore
volume displacements, thereby doubling the likely time estimates
quoted above.
It is realized that chemical degradation of the wall would
begin immediately at the surface in contact with the leachate.
As a result, the hydraulic conductivity of the backfill at the
surface, over a thickness dx, would increase. Assuming as a
worst case that the increase was infinite, then the gradient
across the wall would increase commensurate with the decrease in
4-63
-------
effective wall width. Integration of the appropriate equations
demonstrates that the rate of chemical degradation would increase
by a factor of two as compared to that computed assuming constant
gradient. The assumption of infinite increase in hydraulic
conductivity due to chemical degradation, however, is
conservative in that actual testing indicates only a two fold
increase. As such, the increase in rate of degradation due to a
changing gradient would be less than two.
Given the data and computations performed above, it is
unlikely that the Gilson Road cutoff wall has undergone
significant chemical degradation in the 2.5 years since its
construction began. The calculation of potential cutoff wall
degradation assumed a hydraulic gradient of 1. However, the
operation of the hydrodynamic isolation system, in place since
April 1985, has acted to balance the hydraulic heads inside and
outside the wall, thereby reducing the hydraulic gradient to
essentially zero. In addition, the operation of the groundwater
treatment plant under construction at the site includes a
treatment purge stream located outside the cutoff wall. As a
result, pumping of groundwater within the wall will cause a water
deficit within the cutoff wall and result in flux into the site.
This flux into the site will cause relatively clean groundwater
to flow through the wall, thereby reversing the degradation
caused by the flow of leachate through the wall if any has
actually occurred.
4.4 Conclusions
The hydraulic stress testing and contaminant migration
analyses both indicate that the cutoff wall appears to be
functioning as an essentially intact barrier. However, it is
apparent that the fractured bedrock which forms the bottom of the
containment is highly pervious and would result in a major
leakage path without the hydrodynamic isolation systems
incorporated in the overall containment design. The following
more specific conclusions can be drawn from the Phase Three work
as summarized herein.
0 The three-dimensional numerical modeling of the site
predicted that the bedrock located at the downgradient
portion of the containment was more pervious than
initially indicated via packer testing data. The bedrock
pumping test verified this prediction, yielding bedrock
hydraulic conductivities in the range of 10~1 cm/sec.
This value is large as compared to that specified for the
cutoff wall (1 x 10~7 cm/sec). The pumping test also
demonstrated that the glacial till existing just above
4-64
-------
the bedrock was also quite pervious and probably
discontinuous. This data supports conclusions reached
during the RI/FS.
The cutoff wall and bottom aquitard/aquifer units form a
hydraulically coupled system. Analysis of cutoff wall
bulk hydraulic conductivity therefore must rely on
numerical modeling to correlate the stress test data and
separate the behavior of the wall from that of the
bedrock. The accuracy of the bulk hydraulic conductivity
computed via the sensitivity analysis is therefore
inherently limited to the calibration accuracy of the
numerical model. As such, post-construction verification
efforts are based on somewhat circumstantial data in the
form of piezometric head distributions and thus could be
opened to varying interpretations.
Analyses of cutoff wall bulk hydraulic conductivity must
be based on hydraulic stress testing of the containment.
The value obtained will therefore inevitably be in the
form of an upper bound solution. The proximity of the
upper bound value obtained via analysis to that specified
for the cutoff wall will be limited by not only the
success of the cutoff wall construction effort, but also
by the hydraulic conductivity of the containment bottom.
The very cases which are likely to require
post-construction verification studies (containment
leakage) are therefore those for which hydraulic stress
analysis may be the least conclusive. In these
instances, specifications based solely on performance
criteria may prove difficult to enforce.
The bulk hydraulic conductivity of the Gilson Road cutoff
wall was found to be less than 10~5 cm/sec. The degree
to which the actual value falls below this upper bound
cannot be determined due to the high hydraulic
conductivity of the containment bottom. However, a worst
case value of 10~5 cm/sec yields a cutoff wall efficiency
of greater than 93%. The actual value of wall hydraulic
conductivity is probably approximately 1 x 10"^ cm/sec as
based on quality control testing (Section 2). This value
yields a cutoff wall efficiency in excess of 99%.
The overall passive containment efficiency, including
bedrock leakage, is also important. The overall
efficiency of the passive containment elements (cutoff
wall, cap and fractured bedrock) was found to be
approximately 55% (under worst case precipitation induced
stress conditions) with the major loss flowing out
through the bedrock aquifer. It was recognized during
4-65
-------
the RI/FS phases of the project that the glacial
till/bedrock aquitard/aquifer containment bottom would
leak. However, additional passive barriers such as
grouting to stop such leakage were not found to be
economically feasible. Hydrodynamic isolation systems
were therefore incorporated in the overall containment
design. Hydrodynamic isolation not only proved more cost
efficient than additional physical barriers, such as
grouting the bedrock, but also provided for
recirculation/treatment capabilities. Following this
approach, a purge stream pumpage was instituted in 1986
when the groundwater treatment system went on line. At
this time, over 72,000 GPD are being discharged outside
the cutoff wall in order to implement the hydrodynamic
system elements.
0 The contaminant migration analysis supported the overall
conclusions derived form the hydraulic stress testing.
The data indicated that contaminant flux, in the absence
of hydrodynamic isolation, would be through the bedrock
below the wall. However, the contaminant data was
inconclusive when analyzed independently. This stems
from the lack of sufficient data due to cost and
conflicting objectives governing sample selection as well
as delays in the start-up of the hydrodynamic isolation
system due to construction contract difficulties.
0 The data obtained and the computations executed during
this study in combination with long-term
leachate/backfill compatibility testing undertaken as
part of the cutoff wall design process indicate that
significant chemical degradation of the cutoff wall is
unlikely over the 2.5 years since its construction. The
two pore volume exchanges required for chemical
degradation are computed to take over 20 years under
worst case conditions. Leachate/backf ill compatibility
testing indicates that the worst case leachate found at
the Gilson Road site only increases the hydraulic
conductivity of the intact backfill by a factor of two.
4-66
-------
SECTION 5
CONCLUSIONS
The following general conclusions can be drawn from the work
executed under this contract. More specific and detailed
conclusions with respect to the individual work phases can be
found at the end of their respective sections.
0 Work on the Gilson Road project demonstrated that a high
degree of emphasis on field quality control procedures
was required to achieve and document satisfactory cutoff
wall installation. To be meaningful, the groundwork for
the field QC testing should be initiated during backfill
mix design and leachate compatibility testing. The
results of this design phase work should be reflected in
the project construction specifications as minimum
design standards. QC testing should then be executed in
the field during construction with emphasis on immediate
data availability such that it can be used to guide the
construction on a day to day basis.
0 The API fixed-ring hydraulic conductivity and methylene
blue titration procedures, developed for quality control
of the Gilson Road cutoff wall, proved to be highly
applicable and practical testing methods. The hydraulic
conductivity and percent bentonite content data that
these tests provided were sufficiently precise and
accurate for evaluation of conformance to the project
specifications. As such, these tests should be
routinely incorporated into the quality control program
for soil/bentonite cutoff walls used to contain
hazardous wastes.
0 Modified piezocone instruments appear to be capable of
detecting windows in soil/bentonite backfilled cutoff
walls. However, the present state of the art requires
significant judgment with respect to data interpretation.
This judgment must be based on a thorough understanding
of the equipment used and general soil mechanics theory.
Prior to routine use as a post-construction verification
tool, a significant amount of additional work must be
completed to build up a data base upon which empirical
correlations can be constructed.
5-1
-------
For piezocone sounding or direct sampling to be viable
as post-construction' verification tools, construction
quality control of the cutoff wall must include wall
alignment and verticality provisions. The accuracy of
as-built wall alignment documentation and actual wall
verticality must be increased significantly above that
required for hydraulic performance only.
Post-construction verification of cutoff wall bulk
hydraulic conductivity is an inherently costly endeavor.
Under most actual field conditions of interest/
sensitivity analyses, based on calibrated 3-dimensional
numerical models, will be required to evaluate hydraulic
stress testing data. The value of bulk hydraulic
conductivity obtained for the cutoff wall will therefore
typically be derived from somewhat circumstantial
evidence and will be inconclusive in cases where the
bottom aquiclude is pervious as compared to the value
specified for the cutoff wall. Therefore, use of
quality control testing during construction is typically
preferable to reliance on post-construction verification
for enforcement of project specifications.
Based on the work performed under this contract, the
Gilson Road cutoff wall appears to behave as an
essentially intact barrier to flow. As such, it is
performing as intended with a lower bound efficiency of
greater than 93 percent (an actual efficiency of over 99
percent is probable as based on hydraulic conductivity
values obtained from quality control testing). However,
the fractured bedrock, which forms the bottom of the
containment, appears to be more pervious than initially
anticipated. As such, pumping rates for the
hydrodynamic iso1 ation/recircu1ation system,
incorporated as part of the overall containment system
to account for the fractured rock, will likely have to
be increased relative to initially predicted values.
Work on the Gilson Road project has demonstrated that
cutoff walls can be successfully constructed for
hazardous waste containment. These physical barriers
can prove highly cost efficient as an integral component
of the overall permanent containment system. This is
particularly true for sites where the objective is to
eliminate, rather than just reduce, off-site contaminant
migration. In these cases, the cutoff wall should be
viewed as a clean water exclusion barrier used to
increase hydrodynamic isolation/recircu1 ati o n
effectiveness and reduce treatment costs.
5-2
-------
REFERENCES
Alther, George. Design, Testing & Construction of Bentonite
Liners. International Minerals & Chemical Corporation,
(undated).
API. American Petroleum Institute. API Recommended Practice:
Standard Procedure for Testing Drilling Fluids. API RP13B.
American Petroleum Institute, Dallas, Texas, 1982.
API. American Petroleum Institute. API Specification for Oil
Well Drilling Fluid Materials. API Spec. 13A American
Petroleum Institute, Dallas, Texas, 1981.
Ayres, J. E., and M. J. Barvenik, and D. C. Lager. The First EPA
Superfund Cutoff Wall; Design and Specifications.
Proceedings of the 3rd National Symposium and Exposition on
Aquifer Restoration and Groundwater Monitoring, sponsored
by NWWA, Columbus, OH, May 1983.
Baligh, M. M., and J. N. Levadoux. Pore Pressure Dissipation
After Cone Penetration. Publication No. R80-11, Order No.
662, Department of Civil Engineering, MIT, Cambridge, MA,
April 1980.
Bafoid/HL Industries. Field Testing of Drilling Muds, NL
Baroid/NL Industries, Inc., Houston, Texas (undated).
Barvenik, M. J. Dedicated Gas-Drive Groundwater Sampling
Instruments, Lecture and Notes presented as part of the
Design, Installation, and Sampling of Groundwater
Monitoring Wells Short Course sponsored by the National
Water Well Association, multiple courses in 1984.
Barvenik, M. J. Methods for Passive Physical Containment of
Groundwater Contamination and Field Applications of Active
and Passive Physical Containment of Contaminated
Groundwater. Lectures and notes presented as part of the
Short Course entitled: Corrective Action for Containing
and Controlling Groundwater Contamination, sponsored by the
National Water Well Association; multiple courses in 1985
and 1986.
Barvenik, M. J. and R. M. Cadwgan. Multi-level Gas-Drive
Sampling of Deep Fractured Rock Aquifers in Virginia,
Groundwater Monitoring Review, Vol. 3, No. 4, Fall 1983.
Brown, D. M., and L. W. Gelhar. The Hydrology of Fractured
Rocks: A Literature Review. MIT Parsons Laboratory
Technical Report #304, 1986.
6-1
-------
Cedergren, H. R. Seepage, Drainage, and Flow Nets. John Wiley &
Sons, Inc., NY, NY, pp. 6, 1967.
Douglas, B. J. and R. S. Olsen. Soil Classification Using
Electric Cone Penetrometer. Sym. on Cone Penetration
Testing and Experience. Geot. Eng. Div., ASCE, St. Louis,
Missouri, Oct. 1981.
Hantush, M. S. Flow to Wells in Aquifers Separated by a
Semi-pervious Layer, Journal of Geophysical Research, 72
(6), 1709-1720, 1967.
JRB. Slurry Trench Construction for Pollution Migration Control.
U.S. EPA Document EPA-540/2-84-001, Cincinnati, OH, 1984.
Jones, G. A., and D. Van Zyle, and E. Rust. Mine Tailings
Characterization by Piezometer Cone. Sym. on Cone
Penetration Testing and Experience. Geot. Eng. Div., ASCE,
Oct. 1981.
Koteff, C. and R. P. Volchman. Surficial Geologic Map of the
Pepperell Quadrangle. USGS, GP-1118, 1973.
Lambe, T. W., and R. V. Whitman. Soil Mechanics. John Wiley &
Sons, Inc., NY, NY, 1969.
Levadous, J. N. and M. M. Baligh. Pore Pressures During Cone
Penetration in Clays, Research Report. Department of Civil
Engineering, MIT, Cambridge, MA, 1980 (in preparation).
Middlebrooks, T. A. Earth Dam Practice in the United States.
Trans. ASCE (centennial volume), pp. 697, 1953.
Poulos, H. G. and E. H. Davis. Pile Foundation Analysis and
Design. John Wiley & Sons, Inc., NY, NY, 1980.
Powers, M. A., and P. P. Virgadamo, and W. E. Kelly. Groundwater
Control at Hazardous Waste Sites. ASCE National Convention
on Hazardous Waste, St. Louis, Missouri, 1981.
Sanglerat, G. The Penetrometer and Soil Exploration. Elsevier
Publishing Company, Amsterdam-London-New York, 1972.
Schmertmann, J. H. Guidelines for CPT Performance and Design.
U.S. Dept. of Transportation, FHWA-T5-78-209, Feb. 1977.
6-2
-------
Schulze, D., and M. J. Barvenik, and J. E. Ayres. Design of the
Soil/Bentonite Backfill Mix for the First EPA Superfund
Cutoff Wall. Presented at the Fourth National Symposium on
Aquifer Restoration and Groundwater Monitoring, sponsored
by NWWA, Columbus, OH, 1984.
Torak, L. J. Modifications and Corrections to the
Finite-Difference Model for Simulation of Three-Dimensional
Groundwater Flow. USGS Water Resources Investigation,
82-4025, 1982.
Trescott, P. C. Documentation of Finite-Difference Model for
Simulation of Three-Dimensional Groundwater Flow. USGS
Open File Report 75-438, 1975.
U.S. Army. Laboratory Soils Testing. Department of the Army,
Manual EM 1110-2-1906, November 1970.
Winterkorn, H. F., and H. Y. Fang. Foundation Engineering
Handbook. Van Nostrand Reinhold Company, NY, NY, 1975.
Wissa, A. E. Z., and R. T. Martin, and J. E. Garlanger. The
Piezo Probe, Proceedings. ASCE Specialty Conference on
In-Situ Measurement of Soil Properties, Raleigh, NC, Vol.
I, pp. 536-545, 1975.
6-3
-------
APPENDIX A
API FIXED-RING PROCEDURE
Purpose
The purpose of this procedure is to determine the hydraulic
conductivity of a sample of backfill using the standard API
filter press apparatus.
Materials
1. Standard filter press and mud cell assembly (as manufactured
by Baroid, refer to Figure 2-1 of text)
2. Pressure source and feed lines
3. Pressure gage
4. Pressure regulator
5. Graduated cylinder (10 millileter capacity marked at 0.1 ml
intervals)
6. Dry bentonite
7. Saturated Ottawa sand
8. Paste trimmer
9. Spatula
10. Analytical scale (accuracy: 0.01 grams)
11.' Drying oven
12. Approximately 600 grams of backfill sample
Test Procedure
1. With a spatula, smear a paste consisting of bentonite and
water on the inside of the API cell. The paste should be
dry enough to adhere to the sides of the cell, but wet
enough to saturate the dry bentonite. The purpose of the
paste is to create a seal between the edge of the backfill
sample and the inner surface of the cell. Be careful to
fully cover the wall over the 2-inch high area to be
occupied by the backfill.
2. Place the cell in the paste trimmer. Adjust the trimming
edge so a bentonite seal of approximately 1/32-inch will be
obtained. Evenly rotate the cell in the trimmer to remove
the excess bentonite paste from the cell walls. Remove the
cell from the trimmer and check if a 2-inch high zone
starting 1/4-inch up from the bottom of the cell is entirely
covered by a smooth, 1/32-inch bentonite seal. The paste
will line the inside of the cell over the section occupied
by the backfill sample.
7-1
-------
3. With a spatula and paper towel, remove any excess bentonite
on the cell wall, outside'of the area to be in contact with
the backfill.
4. Assemble the base cap with filtrate tube, rubber gaskets,
steel mesh screen, and cell.
5. Carefully place saturated Ottawa sand in the cell to a depth
of 1/4-inch. The sand should just be in contact with the
bottom edge of the bentonite seal. This sand will act as a
lower filter. Sand is most evenly spread by blocking the
drainage channel with a finger, submerging the sand (in
water), and tapping the outside of the cell to smooth the
surface. Outlet is then opened to let excess water drain
off. If backfill mix is such that migration of fines and
possible subsequent "blow-through" is a concern, carefully
place a 2-inch diameter filter paper on top of sand. (Fines
on the filter paper are also a good indicator of migration
after test is completed).
6. Prepare the backfill sample for testing. Pass the wet
backfill sample through the 1/2-inch sieve. On an
analytical scale, weigh the test sample which should consist
only of material passing the 1/2-inch sieve. Using the
spatula, place the sample material into the cell. The
sample should be placed so as to leave no air voids and
without scraping the bentonite paste seal. After 2 inches
of material has been placed, smooth the sample surface using
the spatula.
7. Weigh the portion of backfill test material not placed in
the cell and record the number of grams of backfill placed
in the cell. With the remaining material, determine the
moisture content of the sample at preparation.
8. Place the cell with the prepared soil sample into the filter
press frame. Slowly, fill the cell above the soil sample
with water. Do not disturb the fine materials at the sample
surface while adding the water. This can be facilitated by
placing a piece of filter paper on top of the backfill mix,
then slowly adding water with a lab squeeze bottle on top of
the filter paper. This will reduce the chance of sample
degradation by scouring. Be sure to remove filter paper
when cell is full (it should float to the top). The water
in the cell should remain clear.
9. Place the top cap and its rubber gasket on the cell. Clamp
the unit in place with the T-screw.
10. Set an empty graduated cylinder under the filtrate tube.
7-2
-------
11. Apply a "seating" pressure to the sample corresponding to
the lowest setting of the regulator (typically 1-1.5 psi).
12. Increase the pressure to 2 psi. Let sit for thirty
minutes, then increase to 3 psi.
13. Read the water level in the graduated cylinder every two
hours for six to eight hours. Record the time, the water
level, the time increment since the last reading (in
minutes), and the change in water level since the last
reading (in cubic centimeters). Calculate the permeability
for each time increment (using change in water volume and
elapsed time since last reading). Also, calculate flow and
volume of water in head space to determine when sample must
be refilled (i.e., particularly important for overnight
testing).
14. If the permeability has stabilized after the initial two
hours water level readings, take readings every eight to
fourteen hours. If the permeability is fluctuating,
continue to take water level readings frequently (every two
hours) until stabilization.
15. Run the test for a minimum of twenty-four hours.
16. If test data at a higher pressure desired, increase the cell
pressure and repeat steps 13-15.
17. Keep a record of the total volume of water permeating the
soil sample. Refill the cell above the soil sample with
additional water when needed.
7-3
-------
APPENDIX B
METHYLENE BLUE TEST PROCEDURE
I. PURPOSE OF TEST
The methylene blue test is a technique used to measure the
cation exchange capacity of a soi1/bentonite mixture. The
purpose of the test is to me-asure bentonite content in soil/
bentonite backfill utilizing the proportionality which exists
between bentonite content and cation exchange capacity.
II. GENERAL PROCEDURE
A sample of soil and bentonite weighing 10 grams is mixed
with a solution of the dispersant tetrasodium pyrophosphate
(TSPP). The TSPP disperses the bentonite and replaces
exchangeable cations of bentonite with sodium which facilitates
the exchange process. A solution of methylene blue in distilled
water is added to the slurry of soil/bentonite and dispersant in
measured volumes until the exchange capacity of the sample is
exceeded. At each increment in the titration, a drop of the
slurry/methylene blue mixture is placed on a piece of filter
paper. Prior to reaching the exchange capacity of the bentonite
(endpoint), the drop appears as a uniformly dark blue circle.
The end point is indicated by the appearance of a light blue halo
around the dark blue dot. At this point, the exchange process is
complete and free methylene blue cations are in solution (causing
the halo).
III. PREPARATION OF METHYLENE BLUE SOLUTION
A. Materials and Apparati
1. Methylene blue crystals: U.S.P. grade, formula
weight of 373.9, zinc free (available from EM
Science)
2. Distilled water
3. Analytical balance (accuracy: 0.01 grams)
4. 1000 - milliliter volumetric flask
5. Funnel
6. Magnetic stirrer and stirring bar
8-1
-------
7. Dark brown glass bottle with air-tight screw-on
cap
B. Procedure
1. Weigh out 3.74 grams of methylene blue (MB)
crystals,
2. Transfer the crystals to a 1000 - milliliter
volumetric flask using a funnel,
3. Add approximately 500 milliliters of distilled
water to the flask,
4. Stir the MB solution for one-half hour using the
magnetic stirrer and stirring bar.
5. Remove the stirring bar from the solution and
carefully fill the flask to the 1000 - milliliter
calibration mark with distilled water,
6. Stir the MB solution for another half hour,
7. Transfer the solution to a dark brown glass
storage container,
8. Repeat steps 1-7 for additional methylene blue
solution (if desired),
9. Continuously stir the MB solution for twelve hours
using the magnetic stirrer,
10. Label the storage container, including the date of
preparation; store the MB solution at room
temperature and out of direct light; the storage
life is approximately three months.
IV. PREPARATION OF DISPERSANT SOLUTION OF TETRASODIUM
PYROPHOSPHATE
A. Materials
1. Tetrasodium pyrophosphate (TSPP) crystals: spe-
cifically Na4 P2 07 . 10H20
2. Distilled water
3. One liter flask
4. Analytical balance (accuracy: 0.01 grams)
8-2
-------
B. Procedure
1. Weigh out 33.30 grams of the TSPP crystals,
2. Transfer the crystals to a flask containing one
liter of distilled water,
3. Shake flask for five to ten minutes, until
crystals are completely dissolved (determined
visually).
V. PREPARATION OF SOIL SAMPLES
A. Materials
1. Soil sample and tare
2. Drying oven
3. Mortar and rubber pestle
4. No. 10 sieve
5. Analytical flask
6. 250 millimeter erlenmeyer flask
7. 2 percent solution of TSPP (prepared per
Section IV)
8. Magnetic stirrer
9. Glass stirring rod
10. Hotplate
B. Procedure
1. Weigh a representative backfill sample,
2. Oven dry the sample,
3. Weigh the dry sample,
4. Using a mortar and rubber pestle, break up
hardened clumps of dry soil,
5. Pass the sample through a No. 10 sieve,
8-3
-------
6. Weigh the portion of the dry sample passing the
No. 10 sieve.
7. Weight out 10.00 grams of the sample which passes
the No. 10 sieve in a 250-ml erlenmeyer flask,
8. Add one hundred milliliters of 2 percent hydrated
tetrasodium pyrophosphate (TSPP) solution to the
10.00 gram sample,
9. Mix the resultant slurry with a magnetic stirrer.
Use a glass stirring rod to force any clay
sticking to the flask walls into the solution,
10. On a hotplate, heat the sample to a slow rolling
boil and maintain for five minutes. Swirl the
flask and again use a glass stirring rod to remove
any clay clinging to the flask walls.
11. Bring the slurry to room temperature.
VI. TITRATION WITH METHYLENE BLUE
A. Materials
1. Soil sample dispersed with TSPP solution
2. Magnetic stirrer and stirring bar
3. Methylene blue solution
4. Whatman's filter paper (hardened No. 50)
5. 1 - 50-ml beaker
6. 1 - 50-ml glass syringe with needle
7. 1 - 3-ml glass syringe with needle
B. Procedure
1. Stir the MB solution in its storage container for
thirty minutes before titration (this should be
done once at the start of each day of testing),
2. Transfer a portion of the MB solution to a wide
mouthed glass jar,
3. Mix the slurry of soil and dispersant for thirty
seconds to one minute,
8-4
-------
4. Place a piece of filter paper atop of a beaker,
such that the paper's edge is overhanging the
beaker (label the paper by test number, date).
5. Titrate the MB solution by syringe. Using the
50-ml glass syringe, transfer enough MB solution
to the slurry so the color of the mix turns from a
greenish-blue to a dark blue,
6. Mix the slurry with the magnetic stirrer for two
minutes,
7. Dip a glass stirring rod into the slurry and place
a drop of the dark blue mixture onto the edge of
the filter paper. Watch for a light blue halo
around the circumference of the resulting dark
blue dot. If a halo does not appear, add an
additional 1 milliliter of MB solution to the
sample using the 3-ml syringe. Mix for thirty
seconds and repeat Step 7,
8. When a blue-green halo appears, do not add more MB
solution. Stir the sample for another two to
three minutes. Check for the halo again by
placing another drop on the filter paper. If the
halo does not appear, repeat Step 7, but after
each 1 milliliter addition of MB solution, mix the
slurry for two minutes. When the halo appears on
both the initial drop and a check, the end point
is reached. Record the test result as milliliters
of MB solution required to reach the end point.
9. The test result (ml MB) is then entered on t-he
bentonite calibration cure to yield a percent
bentonite. This is the percent bentonite in the
material passing the No. 10 sieve. The percent
bentonite in the total backfill sample would then
be
(percent bentonite in minus No. 10 material)
(percent of total sample passing No. 10 sieve).
8-5
-------
APPENDIX C
GEOPHYSICAL METHODS FOR WINDOW DETECTION
A number of geophysical methods currently exist which
potentially could be used to detect windows in soil/bentonite
backfilled cutoff walls. A number of these methods were briefly
investigated on a theoretical basis.
For a remote sensing investigatory method to be feasible for
window detection, its signal must respond differently to a window
as compared to the intact backfill. Furthermore, the signal must
first reach the window and then, in its altered state, propagate
to the instrument for detection. This requires that the
backfill, or formation surrounding the backfill, be relatively
transparent to the investigatory signal. In addition, the window
of interest must exhibit parameters that are significantly
different from the backfill with respect to signal propagation.
C.I Investigation Methodology
Two general investigative approaches exist for window
detection. The first, defined as "PROBE" methods, continuously
detect physical parameters of the backfill in the immediate
vicinity of a probe as it is inserted through the backfill.
These methods have resolution similar to direct sampling of the
wall, but are much faster and less expensive than actual sampling.
These methods, however, will not detect windows which fall in
between probe hole location. Hence, the frequency of windows
would be based on statistical analyses.
The second type or "SCAN" methods, pick up changes in
physical parameters as electromagnetic waves pass between
tranmission/relieving stations and thus through the wall. These
methods "sample" the entire wall but do not have the resolution
of the probe methods. They also tend to require more
sophisticated equipment and data reduction. Based on work done
to date, potential candidates of each type will be treated
individually below.
C.2 Probe Methods
The three primary classes of probe methods are soil borings
with undisturbed sampling, down-hole geophysical methods and
piezocone sounding.
9-1
-------
C.2.1 Borings/Sampling
This method provides the most definitive evaluation of
window characteristics once they have been located. This results
directly from the sample retrieval capability. However, the
process is very slow and costly. Past experience indicates that
the wall sampling and sample testing effort needed to
statistically evaluate window occurrence can easily cost up to
50 percent of the construction cost of the wall.
C.2.2 Down-Hole Geophysics
A number of individual sensing mechanisms can be
incorporated in a down-hole probe such as acoustical fracture
mapping, nuclear density, conductivity, etc. While some of these
mechanisms may respond well to differences in windows vs. intact
backfill, they all require borehole installation prior to use.
As such, these methods were ruled as impractical due to the costs
associated with the required boring program as above. However,
these methods may become practical for window detection if
packaged in a direct insertion probe.
C.2.3 Piezocone Soundings
The standard configuration of probe simultaneously detects
point resistance, side friction, and excess pore pressure
generation during insertion into the wall. Correlation of the
three parameters indicates the type of material penetrated. The
probe is inserted by a drill rig at 1 to 2 cm/sec. About 250
linear feet of probing can be done a day, and as such, the method
is reasonable from a cost standpoint.
The low density and strength of the backfill in conjunction
with rigid push rods result in vertical penetration. The small
diameter of the probe and rods (less than 2.5 inches) renders the
size of these vertical holes unimportant to the integrity of the
wall. In addition, it is expected that upon extraction of the
probe, the hole will close due to the plastic nature of the
backfill.
9-2
-------
C.3 Scan Methods
A number of "scan type" methods exist. The following
discussion will be limited to surface methods in that cross-hole
geophysics was judged too costly due to the requirement for
borehole installation. Surface geophysical methods include:
ground penetrating radar, various seismic techniques, acoustic
monitoring, terrain conductivity, and resistivity. The following
sections provide a brief outline of these methods and their
probable limitations.
C.3.1 Ground Penetrating Radar (GPR)
Initial work indicated that radar would probably not be
applicable in soil/bentonite cutoff wall because the backfill has
a very high conductivity, thus resulting in high attenuation
rates. In addition, the verticality of the wall relative to the
horizontal' antenna array yields poor backscattering
characteristics. These observations were confirmed when GPR was
used to locate the centerline of the cutoff wall at the Gilson
Road site. While the wall was easily distinguishable from the
native sand and gravel formation, penetration was limited to
approximately 10 to 15 feet.
C.3.2 Seismic
Seismic methods were easily ruled out in that the wave
velocity through intact backfill and windows is expected to be
nearly identical. This is a direct result of the high influence
of water on wave velocity and the fully saturated nature of all
backfill materials including windows.
C.3.3 Acoustic Emission
Standard acoustic emission monitoring techniques are not
appropriate for window detection because differences in
intergranular. shear strain are not expected in the backfill due
to the presence of windows. In addition, even though seepage
velocities through windows in the wall would be much higher than
those through intact backfill, it is not expected that the
stresses induced would generate significant emissions.
Furthermore, the natural soil on either side of the wall would
9-3
-------
typically preclude piping and the associated highly turbulent
flow (large emission potential).
More recent developments center around monitoring the
acoustic emissions generated by probe induced material
displacement during a sounding. Equipment packaging is similar
to the piezoconeinstrument. This probe technique may offer
potential for anomaly detection in soil/bentonite cutoff walls,
but was not yet available during execution of this research.
C.3.4 Terrain Conductivity
Terrain conductivity has been shown effective in locating
changes in the conductivity caused by contaminant plumes.
However, a number of constraints render this method inappropriate
for investigation of the occurrence of windows in the Gilson Road
cutoff wall.
0 This method will not respond to the physical differences
between windows and the intact backfill due to .the same
limitations discussed under "GPR" methods and standard
resistivity sounding methods.
0 Cutoff walls are constructed to contain typically the
most highly contaminated portion of a plume and not the
entire plume. As such, contaminated water exists on both
sides of the wall thus, it is difficult to locate windows
via detection of contaminants escaping through wall.
0 It is not likely that the detection range associated with
terrain conductivity methods would allow detection of
contaminated water escaping from windows located at
depth, even if the water outside the wall was clean.
This is particularly true for contaminant plumes which
contain only organic chemicals in the parts per million
concentration range.
0 Windows located near the top of the wall could not be
detected via differences in the conductivity of the
groundwater in areas (1) where clean water exists on both
sides of the wall (upgradient) and (2) where there is no
gradient across the wall.
C.3.5 Resistivity
Resistivity methods using a standard electrode
configurations for sounding with depth such as Wenner &
9-4
-------
Schlumberger arrays, appear to encompass too large a mass outside
the wall at trench depths approaching 100 feet. This is a direct
result of the wall geometry; disproportionate depth to width
distances. These geometrical limitations tend to mask the
potential perturbation caused by small windows. Therefore, it is
anticipated that probes would have to be inserted inside the wall
and current flow between electrodes measured directly.
While this method appeared feasible, it would require
significant development work and as such, was considered outside
the scope of this effort.
------- |