United States
Environmental Protection
Agency
Hazardous Waste Engineering
Research Laboratory
Cincinnati OH 45268
'/.^
Research and Development
EPA/600/S2-86/097 Jan. 1987
Project Summary
Precision and Reliability of
Laboratory Permeability
Measurements
John Bryant and Andrew Bodocsi
In this study, a limited set of labora-
tory test data on clay liner permeabili-
ties was gathered to create a data bank
suitable for a preliminary statistical
analysis. The methods used to collect,
organize and analyze the data are
brifely described.
In the main part of the study, the find-
ings from the data analysis are pre-
sented. In the analysis, consideration
was given to the degree of variability
found in replicated permeability tests,
the question of sample equilibration
with water and chemicals, and the ef-
fect of the magnitude of gradient on
permeability. Permeability test results
using flexible-wall and fixed-wall per-
meameters were also compared. Con-
sideration was also given to the effects
of parameters of sample preparation,
such as the molding water content,
which may be strongly related to per-
meability, and the amount of variability
of permeabilities that may be found in
location-to-location sampling from
within a source of liner material. In ad-
dition, the statistical aspects of the de-
sign and analysis of permeability exper-
iments were discussed and alternate
test protocols were suggested.
This Project Summary was devel-
oped by EPA's Hazardous Waste Engi-
neering Research Laboratory, Cincin-
nati, OH, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering infor-
mation at back.)
Introduction
Clay liners and clay caps are impor-
tant components of hazardous waste
disposal facilities. In engineering prac-
tice the permeability of proposed clay
liner material is customarily determined
in geotechnical engineering laborato-
ries, usually using water as the perme-
ant. A few years ago, (1981) however, it
became evident that certain chemical
leachates have pronounced effects on
clay liner permeabilities. Since then, the
Land Pollution Control Division, Con-
tainment Branch, of the Hazardous
Waste Engineering Research Labora-
tory of the U.S. EPA has sponsored ex-
tensive research on the permeabilities
of clay liner materials with hazardous
chemicals and leachates.
The investigations revealed that cer-
tain chemicals may indeed have dra-
matic effects on compacted clays. How-
ever, considerable variability was
observed in the equilibrium permeabili-
ties from the various laboratories and
individual researchers. There may be
several reasons for this. Some investi-
gators used rigid-wall permeameters,
while others used flexible-wall perme-
ameters. Further sources of variability
may have been the varying percentages
of molding water contents, the individ-
ual testing techniques used, the meth-
ods of deairing the samples and the de-
gree of water-saturation achieved
before the introduction of chemicals,
the type and concentration of chemicals
used, and the various driving gradients
applied to the test specimens.
Having access to a considerable
quantity of accumulated data on clay
liner permeabilities, the Hazardous
Waste Engineering Research Labora-
tory was interested in establishing the
overall reliability of the laboratory
measurements of clay liner permeabili-
ties and their variability. For this pur-
-------
pose, the present study was undertaken
with the following objectives:
a. To collect available historic data on
permeability measurements from
both public and private sources and
to identify the test methods and
parameters used.
b. To analyze the collected data in
order to determine the degree of
variability likely to be found among
the results of permeability tests and
to identify sources of errors in the
testing procedures and estimate
their significance.
c. To make recommendations to en-
hance the applicability of statistical
methods to the analysis of future
clay liner permeability tests.
Methods and Procedures
Data Collection and
Organization
Permeability data were collected from
laboratories that were active either in
research or design of clay liners. They
were selected based on established
contacts, past and on-going research,
data availability and publications. These
included three EPA contract sources,
one government laboratory, and two
private consulting firms.
The typical permeability data re-
ceived contained information on the
clays, testing methods, types of perme-
ameters, types of chemicals, etc. used,
and a tabulation of permeability versus
flow data for each test. The data were
organized by filling out a "Permeability
Test Specifications Sheet" and a "Clay
Type Sheet" for each test and clay, re-
spectively, and attaching them to the
corresponding permeability test data
sheets.
Data Base Limitations
It was known at the initiation of the
project that the data base resulting from
the data collection effort would be lim-
ited in certain respects:
a. Data sources were identified based
on the knowledge that they pos-
sessed relevant and available data
on permeability testing. The se-
lected sources were in no sense
chosen randomly from any pool of
available sources.
b. For many of the tests obtained from
private sources, concomitant infor-
mation concerning the preparation
of specimens and the physical char-
acteristics of the clay soils was not
as complete as would be desired.
Also, there was virtually no repli-
cation of tests from within the
same samples, and all private
source tests used only water as a
permeant.
c. The data received from EPA con-
tract sources was highly project
specific in the sense that re-
searchers in general used different
types of equipment and test meth-
ods, investigated different clay-
chemical combinations, and in
some cases, no replications of
tests on specific clay-permeant
combinations were conducted.
d. Criteria for test termination varied
considerably, sometimes making
the assignment of a terminal per-
meability rather subjective. Often,
tests were terminated prior to the
point in time where equilibrium
was clearly established. This led to
extreme complications in attempt-
ing to reasonably quantify the de-
gree of variation between tests.
e. As a general rule, precise informa-
tion on the level at which tests
were replicated was not available.
In most cases it was not known if
replicate specimens were pre-
pared from a single moisturized
batch of clay soil, or from sepa-
rately moisturized batches. In no
case were tests systematically
replicated at both the within and
between-batch levels, as would be
required for a rigorous analysis of
the components of variance in test
results to be attributed to the oper-
ations of moisturization, com-
paction and testing.
f. In general, there appeared to be
little conscious effort to randomize
experimentation with regard to
either the assignment of experi-
mental material to treatments or
time sequence of testing.
Statistical Analysis Methods
All computations performed in the
analyses described in the report were
performed at the University of Cincin-
nati, Cincinnati, Ohio on either an
Amdahl 470 V/7A or an IBM 3081D com-
puter. Data transformations and sum-
marizations were accomplished by use
of SAS arithmetic capabilities, built-in
functions or data summarization proce-
dures. All statistical analyses performed
or discussed in the report are available
through SAS procedures or in the
BMDP statistical software computer
programs, with the exception of those
proposed for the confirmation of equi-
librium of permeability tests.
Data Analysis Results and
Conclusions
In the main part of the study, analyses
of the variability among the results of
laboratory permeability tests were
made and the factors which appeared to
contribute to that variability were inves-
tigated. The highlights of the results and
the most important conclusions
reached are:
a. Given that the permeability test
samples have achieved equi-
librium, the variability of perme-
ability test results using standard
aqueous leachate, standardized
compaction, and samples pre-
pared from a uniformly moistur-
ized batch of homogeneous clay
soil appears to be reasonably
small relative to other sources of
error. The variability of tests using
chemicals as permeants differed
substantially, with some clay-
chemical combinations yielding
variability characteristics com-
parable to water permeation,
while others resulted in extreme
inconsistency.
b. Certain series of tests in the col-
lected data indicated that, ever
with water as the permeant liquid,
reaching liquid flow equilibrium
may be somewhat more difficult tc
establish than is commonly be
lieved. Furthermore, the error in £
permeability determination whicr
may be caused by premature ter
mination may not be small in com
parison to the variability of the ini
tial permeabilities of replicatec
tests. This implies that the lengtf
of testing time commonly found ir
private sector tests in the collecte(
data may be insufficient. Because
the measured permeability of <
test often undergoes a prolonget
period of transitory behavior be
fore leveling off at an equilibriun
value, deciding when a test has ex
hibited stable permeability lorn
enough to conclude that equi
librium has been achieved is no
straightforward.
c. The permeability of a soil sampl
is extremely sensitive to th
parameters of its preparatior
such as the molding water center
at which it is compacted. This ap
pears to be true not only for wate
as permeant but for chemical pei
-------
meants as well. Minor variations in
these parameters can cause
changes in permeability which
may be large relative to the accu-
racy of the test itself.
d. Several series of tests on which
data have been provided appear to
indicate that the gradient at which
tests are run may have an effect on
the resulting permeabilities in
both rigid-wall and flexible-wall
tests, the magnitude of which in
some cases appears to be practi-
cally significant. Such a phe-
nomenon could be caused by fac-
tors whose effects are opposite in
nature. That is, increased gradient
could cause sample consolidation
in flexible-wall permeameters
leading to decreased permeability,
which appears to have occurred in
several series of tests discussed in
this study. In tests of chemical per-
meants in rigid-wall permeame-
ters, on the other hand, there are
data that suggest that higher gra-
dients may promote channel for-
mation and increased permeabil-
ity.
e. Other variations in testing meth-
ods may be expected to affect test
results as well. In one series of
tests on a bentonite clay, using
xylene as the permeant and rigid-
wall permeameters, the initial per-
meabilities in nonsaturated tests
were four orders of magnitude
greater than those found in corre-
sponding tests which were presat-
urated by permeation of .01N
CaS04 prior to addition of the
xylene. Even after two or more
pore volumes of flow, the perme-
abilities of the presaturated tests
had not reached the level of the
nonsaturated tests.
f. There were almost no within-lab
comparisons of flexible-wall and
rigid-wall permeability tests with
concentrated chemical permeants
in the data sets. The data that were
available, together with between-
lab comparisons under admittedly
dissimilar conditions, suggest that
often quantitatively different re-
sults may be obtained by the two
methods. They also suggest that
flexible-wall test behavior may be
somewhat less erratic in testing
certain chemicals.
g. It is important to assess the degree
of uncertainty of permeability de-
terminations due to the nature of
the test itself. It is equally impor-
tant to estimate the variability of
permeabilities that might occur in
location-to-location sampling
from within a proposed source of
liner material in order to judge the
amount of sampling required to
ascertain the acceptability of the
soil within the source. Limited data
from the private sector suggest
that permeabilities of samples
taken from various positions
within a site exhibit far more vari-
ability than do the replicate tests
on uniform clays observed in the
research setting, even when the
samples are relatively homoge-
neous with respect to Unified Soil
Classification, Atterberg limits and
grain size information. Although it
seems reasonable to ascribe this
variability in large part to location-
to-location heterogeneity of the
soil, for a number of reasons dis-
cussed in the study this factor can-
not be completely separated in the
data from variability which has
been caused by differences in
sample preparation and testing.
Experimental Plans for
Permeability Studies and
Statistical Methods for the
Analysis of Data
A number of experimental plans were
discussed in the study which may be of
use in proposed laboratory research
studies designed to compare the effect
of various factors on permeability tests.
It is not suggested that a single protocol
is appropriate for all studies which
might be proposed, since the best ap-
proach in a particular situation is
specific to the research objectives of in-
terest, and also is a function of the de-
gree to which the experimenter is capa-
ble of controlling such extraneous
factors as soil heterogeneity, molding
water content and compactive effort.
The study describes the basic features
of the potentially useful plans to com-
pare their relative strengths and weak-
nesses, and to illustrate the concepts of
randomization, replication and block-
ing. The features of completely random-
ized design, randomized complete block
design, factorial experiments, random-
ized incomplete block design and split-
plot designs were discussed. In addi-
tion, the construction of analysis of
variance tables for the various experi-
mental plans was discussed. Also,
methods for the determination of ade-
quate replication of experiments were
given, and diagnostic checks of analysis
assumptions were outlined.
Generally, the information presented
is directed toward the non-statistician
who is involved in the development of
experimental plans. However, it was
suggested that an experienced statisti-
cian should be consulted as required,
both during the experimental planning
phase of the project and during the
statistical analysis of the data.
Permeability Tests of Proposed
Borrow Material
In this section of the study, general
considerations concerning the design of
a testing program for the evaluation of
borrow material, which is under consid-
eration for use in the construction of a
liner, were discussed. In such a case the
major goal of the investigation is the
estimation of the permeability of the
soil within the proposed site in order to
determine whether it will meet specified
performance criteria.
One consideration is that the soil con-
tained within the proposed site may ex-
hibit a good deal of heterogeneity in its
chemical and physical properties. This
may lead to considerable variation in
permeability so that an estimate of the
average permeability throughout the
site may not be very relevant in judging
its overall acceptability. That is, a source
whose average permeability might be
considered to be acceptably small for
use as liner material might nevertheless
contain a significant amount of material
having substandard permeability char-
acteristics. For this reason, it is sug-
gested that the site be partitioned into
smaller sub-sites, which will be referred
to as "cells", in such a way that the char-
acteristics of the soil found within any
one cell are reasonably uniform. This
could be done based on a preliminary
determination of characteristics such as
Atterberg limits, grain size distribution
and water content-density relation-
ships. Such cells should not be made up
of noncontiguous parts, should include
a single soil layer, and at a minimum
should contain soils having the same
Unified Soil Classification.
Once the site has been stratified into
cells, replicated permeability tests can
be made on samples drawn from differ-
ent locations within each cell. This will
allow an estimation of the permeability
of the soil within each cell, which may
be used to determine which, if any, of
the cells are not suitable for use as liner
material.
-------
Methods to Confirm
Equilibrium of Research
Permeability Tests
Because the measured permeability
of a test often undergoes a prolonged
period of transitory behavior before lev-
eling off at an equilibrium value, decid-
ing when the test has exhibited stable
permeability long enough to conclude
that equilibrium has been achieved is
not straightforward. Development of
objective criteria, which may be useful
in confirming the experimenter's judg-
ment in this decision, is a topic which
deserves considerable attention in fu-
ture permeability research. In this sec-
tion of the study, two approaches were
described which are currently under in-
vestigation. The first is an adaptation of
a test for trend originally proposed by
Mann, which has been heuristically
modified for use as a test termination
rule. The second approach is based on a
Bayesian analysis in which the posterior
probability that equilibrium has been
achieved is updated after each new ob-
servation, until this posterior probabil-
ity first falls below a specified bound.
In the authors' opinion, the proposed
methods represent reasonable ap-
proaches to the test equilibrium prob-
lem which are deserving of additional
consideration and empirical testing.
The methods described will probably be
of practical value only in situations
where permeability tests approach their
equilibrium levels in a relatively contin-
uous manner. They are not intended to
be able to predict the behavior of tests
which are subject to sudden break-
throughs leading to episodic increases
in permeability.
Recommendations
Based on the analysis results and con-
clusions reached, the following recom-
mendations are made:
a. In comparative tests on chemical
permeants, the use of identically
prepared and concurrently run
water control tests is recom-
mended in order to reduce ambi-
guity in distinguishing between
the effects of equilibration and
chemical.
b. In some situations in which chemi-
cal permeants have strong effects
on the clay samples and lead to
crack formation and channeling,
which in turn lead to massive in-
creases in conductivity, it may be
that the concept of specimen equi-
libration is less useful than meas-
ures based on the notion of time to
failure. If this is the case, then
there is a need for the develop-
ment of practically meaningful fail-
ure criteria for use in research
studies.
c. The permeability of some soil
samples is very sensitive to the
conditions of their preparation,
such as molding water content
and compaction effort. Therefore,
methods of statistical experimen-
tal design, such as randomization
and blocking, should be applied in
comparative permeability re-
search experiments to control the
effects of the above factors.
d. The sensitivity of permeability to
factors such as molding water con-
tent and compaction effort also in-
dicates a need for research which
would estimate the bounds within
which these factors can be main-
tained during the actual construc-
tion of a liner. If these factors vary
substantially during liner construc-
tion, the results of tests performed
only at their target values may give
little indication of the future per-
formance of the liner.
e. It was observed in the limited data
from the private sector that perme-
abilities of samples from various
locations within a site exhibit far
more variability than do replicate
tests on uniform clays in the re-
search laboratory. Although it is
reasonable to attribute most of
this variability to the location-to-
location heterogeneity of the soil
deposit, this cannot be completely
separated from the variability
caused by sample preparation and
testing. Additional research in this
area is indicated.
f. Since deciding when a permeabil-
ity test has achieved equilibrium is
not straightforward, the develop-
ment of objective criteria, which
may be useful in confirming the
experimenter's judgment in this
decision, is a topic which deserves
considerable attention in future
permeability research. Two
approaches under consideration
as decribed in the study, need fur-
ther investigation.
-------
John Bryant and Andrew Bodocsi are with the University of Cincinnati,
Cincinnati, OH 45221.
Jonathan G. Herrmann is the EPA Project Officer (see below).
The complete report, entitled "Precision and Reliability of Laboratory
Permeability Measurements," (Order No. PB 87-113 791/AS; Cost: $18.95,
subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Hazardous Waste Engineering Research Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
US-OFRCIALMAt
Official Business
Penalty for Private Use $300
EPA/600/S2-86/097
i i U1 U S i
0000329 PS
U S fNVIR PROTECTION AGENCT
REGION 5 LIBRARY
230 S DEAR&ORN STREET
CHICAGO IL 60604
------- |