Sei sulla pagina 1di 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233623794

Methods of interpretation of borehole falling-head tests performed in


compacted clay liners

Article  in  Canadian Geotechnical Journal · February 2005


DOI: 10.1139/t04-068

CITATIONS READS
22 3,959

1 author:

Paul Chiasson
Université de Moncton
13 PUBLICATIONS   97 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

interpretation of field falling head tests for measuring hydraulic conductivity View project

Stochastic modelling of soil spatial variability View project

All content following this page was uploaded by Paul Chiasson on 02 September 2015.

The user has requested enhancement of the downloaded file.


79

Methods of interpretation of borehole falling-head


tests performed in compacted clay liners
Paul Chiasson

Abstract: The interpretation of falling-head tests in cased boreholes is discussed. These tests are commonly used to
measure hydraulic conductivity of compacted clay liners and are often part of the construction quality assurance pro-
gram. Three methods of interpretation are reviewed with data sets collected from real tests. Two of these methods have
been the subject of past research by other authors: the Hvorslev, or time-lag, method and the velocity method. After
the limitations of these two approaches have been underlined, a third method is proposed. It uses a best linear unbiased
estimator to fit the theoretical head difference function in a plot of falling water column elevation as a function of time
(Z–t method). The Hvorslev method is found unreliable and is not recommended. The velocity method is theoretically
sound, but statistical uncertainty can become high when this method is used in testing materials with low hydraulic
conductivity, such as clay liners. Materials with low hydraulic conductivity tend to produce scattered velocity plots,
creating considerable uncertainty for the estimated k value. The proposed Z–t method is less sensitive to inaccuracies in
measurements, yielding a more reproducible result. An interpretation method for stages I and II of two-stage borehole
tests is also proposed. This method yields the anisotropy of the liner and the vertical hydraulic conductivity. As a result
of inaccuracies in measurements and limited difference between the geometries of stages I and II, the computed aniso-
tropy exhibits significant uncertainty.
Key words: clay liners, clay covers, hydraulic conductivity, permeability, in situ test, anisotropy, interpretation.

Résumé : L’interprétation d’essais à charge variable dans des forages tubés est discutée. Ces essais sont communément
utilisés pour évaluer la conductivité hydraulique des barrières d’argile compactée. Ils font généralement partie du pro-
gramme d’assurance qualité pendant la construction. Trois méthodes d’interprétation sont abordées. Deux de ces métho-
des sont d’emploi courant et ont été le sujet de plusieurs travaux de recherche : celle de Hvorslev et celle des vitesses.
Après une discussion mettant en contraste les particularités et les limites de ces deux méthodes, une troisième méthode
est proposée. Celle-ci utilise la technique du meilleur estimateur linéaire non-biaisé pour lisser la courbe théorique de
la différence d’élévation de la colonne d’eau en fonction du temps. La comparaison conclut que, pour les argiles com-
pactées, la méthode de Hvorslev conduit à des valeurs non fiables parce qu’on y suppose a priori la position d’un ni-
veau piézométrique inconnu. Celle des vitesses permet en théorie de trouver ce niveau piézométrique mais l’incertitude
statistique peut être élevée dans le cas des faibles conductivités hydrauliques. La méthode Z–t proposée est moins sen-
sible aux incertitudes de mesure. Une méthode d’interprétation pour les étapes 1 et 2 de l’essai à deux étapes en forage
(« two-stage borehole ») est proposée. Elle permet de calculer l’anisotropie de la barrière et la conductivité hydraulique
verticale. Cette anisotropie ne peut pas être connue avec précision étant données les incertitudes de mesure et la faible
différence entre les géométries des étapes I et II de l’essai.
Mots clés : barrière hydraulique d’argile, couverture en argile, conductivité hydraulique, perméabilité, essai in situ, ani-
sotropie, interprétation.

Chiasson 90

Introduction restrict outflow of contaminated liquids (mainly contami-


nated water). Cell designs typically include a hydraulic bar-
In the early 1980s, a number of regulatory bodies adopted rier layer overlaid by a leachate drainage system. During
programs to manage all domestic solid wastes produced operations, wastes are gradually mounded in the cell. Once
within a geographically defined region. Such programs have the design capacity is reached, the cell is capped with a sec-
permitted the construction of controlled regional sanitary ond hydraulic barrier (cover), thus encapsulating and isolat-
landfills that have gradually replaced local unsupervised ing the waste.
open-pit dumps. Wastes are now stored in isolation cells that
Modern regulations for base-layer hydraulic barriers in
landfills have the goal of protecting groundwater (Chen and
Received 16 December 2002. Accepted 24 June 2004. Liew 2003). Typically, the regulation requires a minimum
Published on the NRC Research Press Web site at breakthrough criterion, such as 25 years under a unit hydrau-
http://cgj.nrc.ca on 15 February 2005. lic gradient flow. In the case of the U.S. Environmental Pro-
P. Chiasson. Secteur de génie civil, Faculté d’ingénierie, tection Agency (U.S. EPA 1991), the regulations provide
Université de Moncton, Moncton, NB E1A 3E9, Canada two design options for owners of municipal solid-waste
(e-mail: chssp@umoncton.ca). landfills. The first option is a composite liner system, whose

Can. Geotech. J. 42: 79–90 (2005) doi: 10.1139/T04-068 © 2005 NRC Canada
80 Can. Geotech. J. Vol. 42, 2005

minimum requirements include a 0.60 m thick compacted many orders of magnitude more permeable than samples
clay liner (CCL) with a hydraulic conductivity no greater compacted on the wet side (Bjerrum and Huder 1957;
than 1 × 10–9 m/s. Quality control for liner construction has Lambe 1958a, 1958b; Mitchell et al. 1965), CQA programs
led to the development and in some instances standardiza- did not use acceptance criteria based on this widely demon-
tion of a number of field hydraulic conductivity testing strated experimental result until quite recently. Work by
methods. Sealed double-ring infiltrometer (SDRI) tests, Benson and Boutwell (1992), Leroueil et al. (1992), Daniel
American Society for Testing and Materials (ASTM) test and Koerner (1995), Daniel (1998), Benson et al. (1999),
method D5093 (ASTM 2004a), and lysimeter tests are ex- and many others gradually permitted the evolution of a CQA
amples. These permit one to test CCL compliance by mea- based on compaction control points falling within a zone of
suring flow through both large- and small-scale defects, such acceptance. This zone of acceptance is defined by a plot of
as bad lift-to-lift bonding and insufficient destruction of unit dry mass versus water content (the same as for a com-
clods and interclod macropores. Unfortunately, they require paction curve). On the basis of optimum conditions for hy-
long testing times, which can exceed 3 weeks (Trautwein draulic conductivity, the zone is delimited on its left side
and Boutwell 1994). This renders these tests impractical for (lowest water contents) by the line of optimums and on its
construction quality assurance (CQA). They are used, rather, right side (highest water contents) by the saturation curve.
on test pads to demonstrate compliance during the design Other criteria, such as shear strength and the potential to
phase. In this design philosophy, field tests serve to establish generate shrinkage cracks, may be added to circumscribe an
construction procedures that best minimize defects in the overall acceptance zone (Daniel and Wu 1993).
liner. These procedures are then used to construct the actual Even when compaction control points fall on the wet side
liner. During construction, the procedures adopted must be within the overall zone of acceptance, there is still potential
followed exactly and controlled through a well-conceived for noncompliance. Other defects can yield a CCL that fails
CQA program. hydraulic conductivity requirements because of bad bonding
Time being an important factor in regions where the con- between lifts; insufficient water content blending in the liner
struction season is rather short, a need exists for faster hy- material during compaction; or environmental factors such
draulic conductivity tests. One of these tests, which aims to as desiccation and freezing. Thus, field hydraulic conductiv-
decrease testing times, is the two-stage borehole (TSB) test ity tests are still warranted and should be part of a good
(test method D6391; ASTM 2004b). This test combines two CQA program.
well-established borehole test methods: end of casing and A safer practice would be to include such controls on the
injection through an extended borehole chamber (CAN– CCL during and after final construction. For practical
BNQ 1988a, 1998b; AFNOR 1992). Trautwein and Bout- reasons, controls must be fast and reliable. Otherwise, con-
well (1994) indicated that this test is more than three times struction may be severely delayed, creating excessive expen-
faster than SDRI. ditures. TSB tests fall into this category. They are fast, with
Trautwein and Boutwell (1994) proposed special shape results obtained in 24 h. They also have minimal destructive
factors for stage I (end-of-casing test) and stage II (extended impact on the CCL, thus minimizing subsequent repair
borehole). These (as advanced by the authors) could be com- costs.
bined and solved for horizontal and vertical components of A final performance check of the liner can be done
hydraulic conductivity. Chapuis (1999) demonstrated these through a full-scale test for total leakage versus water load
shape factors to be incorrect. Furthermore, Chapuis doubted within the cell (Chapuis 2002). Such a test has the merit of
whether the anisotropy ratio, kh/kv, could be extracted from detecting defects that have been missed by field tests or the
such a procedure. Chapuis gave two arguments to support CQA program. Performing the test for a long duration
his affirmation: the two stages have nearly spherical (not el- (greater than 3 weeks) can also help evaluate whether the
liptical) equipotentials; and smearing effects introduce high liner is prone to suffosion (internal erosion). Although such
uncertainty into this ratio. Also worthwhile pointing out is a test may appear as overkill, detection of defects before the
an assumption implicitly made by Trautwein and Boutwell liner enters into service makes it less costly to repair than it
(1994): the liner is homogeneous in k. Liners are built from would be once the site is in operation.
a succession of lifts that may vary in hydraulic conductivity This paper reviews data interpretation of cased borehole,
for reasons such as variations in dry density and water con- specifically TSB, tests. Two known methods of interpreta-
tent, soil gradation, and clay mineral content. Cassan (2000) tion are reviewed: the Hvorslev method and the Chapuis ve-
pointed out that the hypothesis of homogeneity in k is legiti- locity method. A third method is also introduced. All three
mate at a testing scale where water flows through all repre- methods are compared for their precision in the statistical
sentative soil layers. This does not appear to be the case in a sense when field hydraulic conductivity, k, is estimated. Rec-
TSB test. ommendations are proposed for interpreting such measure-
ments.
Even with a good CQA program, the importance of hy-
draulic conductivity field tests must be emphasized. It is
through such tests that bad or faulty construction practice Site description and testing program
was identified in the past. Early liner construction often
showed large discrepancies between field and laboratory A number of cased borehole falling-head hydraulic con-
hydraulic conductivity values (Daniel 1984; Elsbury et al. ductivity tests were performed at the Red Pine Regional
1990). Although it was well known that hydraulic conductiv- Sanitary Landfill. The landfill is located in northeastern New
ity measurements on samples compacted on the dry side are Brunswick (Canada), 20 km south of Bathurst. It is part of a

© 2005 NRC Canada


Chiasson 81

government program adopted in 1985 and designed to man- of casing was inspected and manually cleaned according
age solid waste in a safe and controlled manner. The site to ASTM D6391 recommended practice (ASTM 2004b).
currently collects and stores 86 000 t/year of waste. Finally, permeability tests in compacted clay were preceded
The cell design at Red Pine includes a base liner of com- by a wetting period of 24–48 h.
pacted clay (1.4 m) overlaid by a high-density polyethylene
geomembrane. To meet the New Brunswick Department Interpretation of falling-head sealed
of the Environment breakthrough criteria, this clay layer borehole tests
must have a hydraulic conductivity of no more than 1.4 ×
10–9 m/s. Chapuis (1998) indicated that when the deformations of
The construction material, clayey red till and gray transi- the soil can be neglected, falling-head tests are governed by
tion till that is readily available on the site, has the potential the Laplace equation. Its solutions, the harmonic functions,
to meet requirements. Chiasson et al. (1998) reported labora- have several properties. One of them relates the flux in the
tory hydraulic conductivity values on specimens compacted soil (Qsoil) to the flow into the pipe (Qinj) through a mass-
at optimum water content under modified energy. Both ma- balance equation,
terials and their mixtures were found to have values no
[1] Qinj = Qsoil = ckH
greater than 0.29 × 10–9 m/s. Mineralogical studies showed
high clay mineral content for both native soils, thus explain- where c is a shape factor that depends on the geometry of
ing their potential to attain low hydraulic conductivity. the injection zone and on the hydraulic boundaries of the
Means and standard deviation (SD) for plastic limits, liquid problem; H is the applied hydraulic head difference; and k is
limits, and plasticity indexes of the two materials are, re- the hydraulic conductivity. This equation is the starting point
spectively, 18.9% (1.1%), 30.0% (2.5%), and 11.1% (2.2%). of both the Hvorslev and the Chapuis methods. Another
During the course of landfill operations, a number of test equation is the starting point of another method for cases
pads were built and field tested. Data presented in this paper where soil deformation is assumed to be elastic and not
are for one test pad built in 1995, where an extensive testing negligible (Cooper et al. 1967). However, the mathematical
program was performed. Construction crews and practices model of this method does not correspond to the physical
for the pad were the same as for the real storage cell being model. This is according to mathematical, physical, and nu-
built at that time. The 1995 test pad was 1 m thick and built merical proofs by Chapuis (1998) and experimental proofs
of clayey till only. Very little variation in water content was by Chapuis and Chenaf (2002). According to the equations
reported during construction. Compaction controls yielded of Chapuis (1998), the effect of soil deformation can be ne-
an average maximum dry density of 2.015 Mg/m3, with av- glected when the soil is an aquifer or an overconsolidated
erage water content of 10%, corresponding to 80% satura- aquitard. Soil deformation is no longer negligible for com-
tion. Following construction of the test pad, an SDRI test pressible aquitards when they are tested by using either a
was performed (Jacques Whitford Ltd. 1995). Total duration falling-head test with a very small injection pipe or a pulse
of infiltration was 25 days. Shelby tube samples, extracted test between packers. Chapuis and Cazaux (2002) suggested
before SDRI testing, showed an average water content of methods for handling the instantaneous (elastic) and delayed
10%, with little variation throughout the sampled depth of deformations in such cases. Because this paper deals only
0–300 mm. Tensiometers were also installed within the outer with compacted clays, which are overconsolidated aquitards,
ring of the SDRI at depths of 50, 100, and 150 mm. Briefly the use of eq. [1] is thus justified. In a falling-head test, Qinj
after installation and before SDRI testing, suction readings is the flow through the inflow pipe (often a standpipe con-
rapidly stabilized at 80 kPa. Following initiation of SDRI in- nected to the borehole casing) of internal cross section Sinj.
filtration, suction readings gradually declined to stabilize at
dH
5 kPa after 8, 15, and 21 days at the three respective depths. [2] Qinj = − Sinj
This indicates that compacted clay had attained satiation at dt
these depths at those times of testing. Following removal of
where t is time. Equations [1] and [2] yield
the SDRI apparatus, water content of 14% was measured up
to a depth of 220 mm. Cased borehole tests reported in this dH ckH
[3] =−
paper were performed on this pad the next year, in July dt Sinj
1996. Although water content was not measured when these
tests were performed, the 1 year resting period must have fa- Rearranging gives
voured dissipation of suction values throughout the thickness
of the liner, yielding satiated compacted clay. dH ck
[4] =− dt
Cased borehole tests performed on the 1995 test pad were H Sinj
of the end-of-casing type. Acrylonitrile butadiene styrene
pipes had an internal diameter of 102 mm (4 in.). The testing Integrating leads to Hvorslev’s solution (1951):
campaign used medium sodium bentonite chips (9.5 mm)
⎛H ⎞ c
and sodium bentonite grout to seal the annular space be- [5] ln⎜ 1 ⎟ = −k (t1 − t 2) = −kC (t1 − t 2)
tween borehole sidewall and pipe. As reported by ASTM ⎝ H2 ⎠ Sinj
(2004b), Chapuis (1998), and Trautwein and Boutwell
(1994), wall smearing of the injection zone may influence where H1 and H2 are, respectively, the applied differences in
results. To prevent this problem, the soil surface at the end total head at times t1 and t2; and C = c/Sinj. If the falling-

© 2005 NRC Canada


82 Can. Geotech. J. Vol. 42, 2005

head test is made with a standpipe of internal diameter d, Fig. 1. Typical setup of a cased borehole test in a compacted
then clay liner.
πd 2
[6] Sinj =
4

Shape factors
The following shape factors are best suited for the prob-
lem at hand (Chapuis 1999). For the end-of-casing test, the
shape factor C is
11D
[7] C =
πd 2
For the Lefranc test, which is with a cylindrical injection Fig. 2. Semilog plot of falling-head test. Datum and assumed
zone, the sphere formula is recommended when 1 ≤ L/D ≤ 8: piezometric level at top of compacted clay liner (see data in Ta-
ble 1).
L 1
[8] c = 2π D +
D 4
which gives
c 8D L 1
[9] C = = 2 +
Sinj d D 4

when 1 ≤ L /D ≤ 8. These latter two equations are, according


to Chapuis (1999), accurate to a relative error of 10%. More
accurate shape factors can be obtained from modified equa-
tions that take into account the upper and lower boundaries
of the barrier (Chapuis 1989).

Hvorslev method of interpretation


In the Hvorslev method, data are plotted on a semilog In normal practice, the operator is too often unaware of
graph with ln [H(t = 0)/H(tj)] on the ordinate and time t on the unknown height Ho. The operator thus sets the PL at an
the abscissa. According to eq. [5], this should theoretically assumed level and sets this level as the datum. By doing this,
give a straight line. Let m be the slope of this line, such that the operator is not plotting ln[H(t = 0)/H(tj)] but rather
m = [ln (H1/H2)]/(t1 – t2). Then, from eq. [5], ln[Z(t = t0)/Z(tj)] versus tj, which from eq. [5] gives
[10] m = –kC ⎡ Z(t ) ⎤
[13] ln ⎢ 0 ⎥ = −kC (t 0 − t j ) + ln Z(t 0)
and hence ⎢⎣ Z(t j ) ⎥⎦
[11] k = –m/C − ln[Z(t 0) − Ho(1 − e− kC ( t0 − tj ) )]
In the case of TSB tests in a CCL, the true difference in
where t0 and Z(t0) are initial readings of time and water
hydraulic head at time t is unknown. What is measurable at
time t is elevation Z of the water column from a set refer- column elevation; and tj and Z(tj) are the jth readings.
ence level. Typically, this is chosen as the surface of the This plot gives a straight line only when the datum is set
barrier. The unknown component is the height of the piezo- at the same level as the correct PL of the soil (in other
metric level (PL) of the soil around the injection zone from words, when the hypothesis Ho = 0 is correct). When Ho ≠ 0,
the set reference (Fig. 1). The correct difference in head at the plot of ln [Z(t0)/Z(tj)] versus tj yields a more or less pro-
time t is nounced upward or downward curve (Fig. 2). Furthermore,
[12] H(t) = Z(t) – Ho inspection of eq. [13] shows that when Ho ≠ 0, eq. [13] can-
not be simplified to eq. [5], and therefore the relationship
where Ho is the unknown height of the soil PL from the set between hydraulic conductivity and the slope of the plot as
reference level (positive, if over; negative, if under). This no- expressed by eq. [11] is no longer valid. Anyone who is not
tation is in accordance with that of a number of authors aware of this will naturally assume that eq. [11] still applies.
(Schneebeli 1954; Chapuis et al. 1981; CAN–BNQ 1988a, This would lead to an erroneous interpretation of the plot
1988b) who use Ho to symbolize this unknown height. This and erroneous hydraulic conductivity values. Because soils
should not be confused with the definition employed in the composing a CCL are in an unsaturated state, a PL below
Hvorslev method (1951), where Ho is the initial head differ- the top of the liner should be expected. In the author’s expe-
ence. rience, the correct PL for a fairly humid climate is often

© 2005 NRC Canada


Chiasson 83

Table 1. Falling-head test in a cased borehole on the 1995 test pad.


Reading, j Time, tj (min) Z(tj) (cm) ln [Z(to)/Z(tj)] m (min–1) k (m/s) v (cm/min) Zm (cm)
0 0 8 0
1 4.53 7 0.13 0.029 2.42×10–9 0.2209 7.5
2 9.10 6 0.29 0.034 2.76×10–9 0.2186 6.5
3 13.72 5 0.47 0.039 3.24×10–9 0.2166 5.5
4 18.49 4 0.69 0.047 3.83×10–9 0.2094 4.5
5 23.36 3 0.98 0.059 4.85×10–9 0.2055 3.5
6 28.35 2 1.39 0.081 6.66×10–9 0.2004 2.5
7 33.41 1 2.08 0.137 1.12×10–8 0.1974 1.5
8 35.99 0.5 2.77 0.269 2.21×10–8 0.1940 0.75
Note: Datum and assumed piezometric level of soil at top of compacted clay liner (Lefranc type test, shape factor
C = 2032 cm–1).

found somewhere between the top and the bottom of the Fig. 3. Hydraulic conductivity computed for assumed piezometric
liner (see also Chapuis 1999). In an illustrative example (Ta- level at top, bottom, and correct level (44.55 cm below top) with
ble 1, Fig. 2), the datum is set at the top of a 1 m CCL, and Hvorslev’s method (see data in Table 1).
the soil PL is assumed to be at the same level. This leads to
the erroneous conclusion that hydraulic conductivity in-
creases with time (Table 1, Fig. 3: PL assumed to be at the
top of the CCL). If the datum and assumed PL are set at the
bottom of the liner (add 100 cm to Z in Table 1), the hydrau-
lic conductivity decreases slightly with time, thus suggesting
a relatively constant value of k = 1.4 × 10–10 m/s (Fig. 3: PL
assumed to be at the bottom). Finally, choosing the correct
PL, at 44.55 cm below the top of the liner, yields a constant
k = 3.51 × 10–10 m/s throughout the duration of the test.
Three different results are found for three different assumed
PL levels. Test interpretation through the Hvorslev method
yields a hydraulic conductivity, k, that is a function of the
assumed PL. Use of this method does not appear to be rec-
ommendable.

The velocity (Chapuis) method Fig. 4. Velocity method plot of a cased borehole test (see data in
Table 1).
According to eq. [3], plotting of falling-head velocity as a
function of applied head difference should yield a straight
line. Because the applied head difference is usually un-
known, combining eqs. [3] and [12] yields
dZ
[14] = −Ck(Z − Ho)
dt
and rearranging gives
dZ 1
[15] Z =− + Ho
dt Ck
Note that in a falling-head test, the velocity, dZ/dt, is neg-
ative, as a result of the downward direction of flow. Accord-
ing to this last equation, the plot should yield a straight line
for whatever set datum. The intercept Ho is the error, or the
In practice, the mean elevation, Zm, during time increment
height separating the datum and the correct PL of the soil
∆t is plotted as a function of average velocity, v = ∆Z/∆t. A
(Fig. 1). The slope of the plot is mv = –1/Ck. Therefore,
plot of data from Table 1 gives a straight line (Fig. 4) with a
[16] k = –1/Cmv slope mv = 233.7 min and an intercept Ho = –44.55 cm.
Computing k gives 3.51 × 10–10 m/s. The Hvorslev method
The slope, mv , is not a function of the unknown PL of the gives this same value if the datum is set 44.55 cm below the
soil. Thus, a hydraulic conductivity computed from a plot of top of the liner.
height Z as a function of velocity dZ/dt is independent of the To resume, the velocity method does not depend on the
PL of the soil. set datum (or assumed PL); furthermore, it yields the correct

© 2005 NRC Canada


84 Can. Geotech. J. Vol. 42, 2005

Table 2. Falling-head test in a piezometer installed Fig. 5. Velocity method plot for a falling-head test in a piezo-
in Champlain clay below the weathered surface meter installed in Champlain clay (Lefranc test; see data in Ta-
layer and phreatic line. ble 2).
Time (min) z (cm) v (cm/min) Zm (cm)
0 122.90
2 122.00 0.450 122.45
4 120.80 0.600 121.40
6 119.80 0.500 120.30
8 118.80 0.500 119.30
10 118.00 0.400 118.40
12 117.00 0.500 117.50
14 116.10 0.450 116.55
16 115.00 0.550 115.55
18 113.90 0.550 114.45
20 113.10 0.400 113.50
22 112.30 0.400 112.70
24 111.50 0.400 111.90
26 110.50 0.500 111.00 Fig. 6. Semilog plot for a falling-head test in a piezometer in-
28 109.60 0.450 110.05 stalled in Champlain clay (Lefranc test; see data in Table 2).
30 108.80 0.400 109.20
32 108.10 0.350 108.45
34 107.10 0.500 107.60
Note: Lefranc type test, shape factor C = 2353 cm–1.

PL of the soil. Unfortunately, this method has its pitfalls. In-


accuracies in measurements can generate larger inaccuracies
in computed velocities. This is due to the problem of double
inaccuracy in derivatives, with the potential to produce con-
siderable scatter in the velocity plot (see data in Table 2 and
Fig. 5). Least squares yields an intercept Ho = 100.97 cm
and a slope mv = 29.59 min, which corresponds to k = 2.4 ×
10–9 m/s. Applying the Hvorslev method to the falling eleva-
tion data of Table 2, with a new datum set at 100.97 cm
above the original one, yields a curved plot in the semilog and combining with eq. [12] yields
graph (Fig. 6). This curvature underlines an inappropriate [19] Z(t) = [Z(0) – Ho]e–at + Ho
correction of the PL, as demonstrated earlier by eq. [13]. Af-
ter inspection of the velocity plot and the semilog plot (see Because the expression in brackets corresponds to H(0),
Figs. 5 and 6), one would not put much confidence in the the following is preferred:
velocity method value of k or in the test itself. The velocity
plot has a mediocre fit (R2 = 0.1832), whereas the semilog [20] Z(t) = H(0)e–at + Ho
plot is clearly curved downward. The inaccuracy amplified where Z(t) is water column elevation as a function of time t;
in the velocity plot of Fig. 5 may lead one to classify this and H(0) is the (true and unknown) hydraulic head differ-
test as poor or defective, and the temptation may be to reject ence at t = 0. A best unbiased estimator is used to evaluate
it. Results such as this motivated the development of the the unknown parameters: H(0), Ho, and a. The residual be-
method described in the following section. tween the estimator of the function Z * (t) and the measure-
ment Z(tj) is
Proposed interpretation method: Z–t [21] ε j = Z j − Z * (t j )
The objective is to develop a method less sensitive to The solution for Z * (t) is obtained by minimizing the sum
measurement inaccuracy. Because relative inaccuracies on of squared residuals,
water column elevation Z and time t are generally small
(<1%), a plot of these data should display little scatter ⎛ n ⎞
[22] Min ⎜ ∑ ε 2j ⎟
around the trend. Rearranging eq. [5] results in H(t) becom- ⎜ ⎟
ing ⎝ j =1 ⎠

[17] H(t) = H(0)e–kCt subject to the unbiased condition,


n
Setting
[23] ∑εj =0
[18] a = kC j =1

© 2005 NRC Canada


Chiasson 85

where n is the number of measurements Z(tj). The complex- Fig. 7. Elevation of water column as a function of time for
ity of eq. [20] does not permit an analytical solution to this falling-head test data in a piezometer installed in Champlain clay
problem. Numerical optimization techniques are therefore (Lefranc test, see data in Table 2).
required. These are readily available in most modern spread-
sheets. The solution for hydraulic conductivity, k, will be
[24] k = a/C
Deriving eq. [20] as a function of time t yields the slope
of the velocity graph:
[25] mv = dZ/dt = –1/a = –1/kC
Plotting the data of Table 2 and applying the method yields
H(0) = 71.79 cm, Ho = 51.15 cm, and a = 0.00730 min–1
(Fig. 7). Equation [24] gives k = 5.2 × 10–10 m/s. As expected,
the statistical scatter around the trend is low. The hydraulic
conductivity from this method is smaller by a factor of 4.6
than that obtained from the velocity method. The goodness of
fit illustrated in Fig. 7 gives more confidence in this hydrau-
Fig. 8. Semilog plot of corrected falling-head test data in a
lic conductivity value and suggests that the test is of good
piezometer installed in Champlain clay. Data corrected by Z–t
quality. Furthermore, a regression line through a semilog
method (Lefranc test; see data in Table 2).
plot of corrected falling elevation data from Table 2, with a
new datum set 51.15 cm above the original reference, yields
a correlation coefficient, R2, of 0.9996 (Fig. 8). This implies
that the estimated Ho obtained by the proposed method is a
better estimate of the true correction to be applied to the
data.

Comparing methods
For comparison of statistical robustness, the three studied
methods—the velocity method, the Hvorslev method cor-
rected by Ho as estimated from the velocity method, and the
Hvorslev method corrected by Ho as estimated by the Z–t
method—and the proposed Z–t method were applied on 19
falling-head test trials. These tests were performed on a CCL
test pad built in 1995 on the Red Pine site. The clay was Fig. 9. Cumulative frequency of hydraulic conductivity for 19
compacted at 94.5% of the modified Proctor dry density. All tests as computed with the velocity method; the Hvorslev method
19 test trials were performed in borehole No. 5 and used the corrected by Ho as estimated by the velocity method; the pro-
same end-of-casing geometry. The velocity and the Z–t me- posed Z–t method; and the Hvorslev method corrected by Ho as
thods of interpretation yield values of the same order of estimated by the Z–t method. All tests were performed in bore-
magnitude, although the average hydraulic conductivity of 1.9 hole No. 5, using the same end-of-casing geometry, on a com-
× 10–9 m/s for the proposed method is lower than the 2.4 × pacted clay liner built in 1995.
10–9 m/s obtained from the velocity approach. The velocity
method also yields higher statistical scatter than the Z–t
method (Fig. 9). With the exception of one test, the proposed
method systematically yields lower hydraulic conductivity
values (Figs. 9 and 10). It also gives the lowest SD for k,
with 0.6 × 10–9 m/s versus 0.9 × 10–9 m/s. Thus, the preci-
sion of computed k values is 33% better with the proposed
Z–t method.
In the case of the velocity method, the standard error on
the slope permits one to compute the 95% confidence inter-
val for k (Neter et al. 1989). As a basis of comparison for
both methods, the coefficient of determination (R2) com-
puted for each trial run is used. Both quantities are used as
indexes to characterize statistical dispersion of each test
trial. Test trials with low data scatter have narrower error
bars than tests with higher scatter (Fig. 10).
In tests with low scatter, similar hydraulic conductivity
values are computed whatever the method used. This is not
true for tests with high data scatter. In such tests, the veloc-

© 2005 NRC Canada


86 Can. Geotech. J. Vol. 42, 2005

Fig. 10. Hydraulic conductivity computed with the proposed Z–t Fig. 12. Hydraulic conductivity computed with the Z–t method
method versus that computed with the velocity method with er- as a function of the coefficient of determination for 19 test trials
ror bars (end-of-casing tests performed in borehole No. 5, 1995 (end-of-casing tests performed in borehole No. 5, 1995 test pad).
test pad).

posed method is statistically more robust. This is evident


Fig. 11. Hydraulic conductivity computed with the velocity when comparing the goodness of fit of the velocity plots
method as a function of the coefficient of determination for 19 (Fig. 5) with that of the Z–t plots (Fig. 7). With the velocity
test trials (end-of-casing tests performed in borehole No. 5, 1995 method, small inaccuracies in measurements generate rather
test pad). large relative errors in the velocity values. Scatter is ampli-
fied in the velocity plot, from which follows the high uncer-
tainty of k.
The Hvorslev method does not prove to be very useful.
When the correct head difference of eq. [12] uses Ho com-
puted with the velocity method, the Hvorslev method yields
the same k value as obtained from the velocity approach
(Fig. 9). If the same is done using Ho obtained from the pro-
posed Z–t method, the Hvorslev method then gives the same
result as the proposed method (Fig. 9). Thus, the Hvorslev
method will agree with the method with which the Ho cor-
rection is obtained, even though the velocity method and the
proposed method yield different hydraulic conductivity val-
ues.

Evaluating the anisotropy of hydraulic conductivity


Soils commonly display anisotropy in hydraulic conduc-
ity and Z–t methods of interpretation yield different hydrau- tivity, α = kh/kv ≠ 1. In a stratified soil, this ratio will always
lic conductivity values. For example, test trial No. 19 yields be greater than 1. According to eq. [1] and Cassan (2000),
3.0 × 10–9 m/s with the velocity method, whereas the pro- the flow from an injection zone into a transformed medium
posed Z–t method gives 1.1 × 10–9 m/s. The latter value is is
2.8 times smaller. As shown in Fig. 11, hydraulic conductiv- Qinj
ity computed with the velocity method tends to increase [26] ′ =
Qinj = Qsoil
′ = c′ k′ H ′
with decreasing quality of fit (decreasing R2). In the case of α
the proposed method, hydraulic conductivity is not sensitive where c′ is the shape factor of the injection zone for the
to random inaccuracies in measurements, because no signifi- transformed medium; k′ = kv; and H′ = H.
cant correlation is found between computed hydraulic con-
In the shape factor c′, D′ = D/(α1/ 2). The shape factor c′ is
ductivity and quality of fit (R2 = 0.064 in Fig. 12). Note also
a function of the injection zone aspect ratio, λ′, in the trans-
that the proposed method yields very high coefficients of de-
formed medium, where
termination. This indicates that the curve fits the data well
(such as illustrated in Fig. 7) and explains well the interrela- L
[27] λ′ = λ α = α
tionship between time t and falling water column elevation D
Z. The same data, when interpreted with the velocity
method, yield variable coefficients of determination, with Depending on the ratio of anisotropy, the transformed ra-
many tests appearing to be of poor quality. Meanwhile, very tio can be significantly greater than the true physical aspect
high coefficients of determination are systematically ob- ratio. Therefore, the shape factor function for the trans-
tained for the proposed method (Table 3). Thus, the pro- formed medium can be different from that for the physical

© 2005 NRC Canada


Chiasson 87

Table 3. Hydraulic conductivity (k), coefficient of determination (R2), and relative error on
k (ε k/k) as computed with velocity and Z–t methods.
Velocity method Z–t method
Trial No. k (m/s) R2 ε k/k (%) k (m/s) R2
–9 –9
1 2.4×10 0.936 705 16 2.2×10 0.999 976
2 2.4×10–9 0.955 260 16 2.1×10–9 0.999 966
3 2.9×10–9 0.935 620 20 2.7×10–9 0.999 968
4 3.1×10–9 0.963 077 14 2.9×10–9 0.999 967
5 1.8×10–9 0.927 941 20 1.7×10–9 0.999 978
6 1.9×10–9 0.967 652 13 1.8×10–9 0.999 988
7 2.3×10–9 0.759 273 38 1.6×10–9 0.999 928
8 1.7×10–9 0.954 769 14 1.6×10–9 0.999 959
9 1.8×10–9 0.886 522 24 1.4×10–9 0.999 940
10 1.2×10–9 0.952 773 15 1.1×10–9 0.999 981
11 1.4×10–9 0.926 404 18 2.8×10–9 0.999 974
12 1.7×10–9 0.982 498 9 1.8×10–9 0.999 956
13 4.1×10–9 0.723 131 47 2.8×10–9 0.999 138
14 4.7×10–9 0.535 798 70 2.8×10–9 0.999 450
15 2.2×10–9 0.867 935 32 2.0×10–9 0.999 898
16 2.6×10–9 0.933 218 22 1.8×10–9 0.999 515
17 2.2×10–9 0.630 636 62 9.7×10–10 0.999 903
18 2.8×10–9 0.607 390 61 1.5×10–9 0.999 481
19 3.0×10–9 0.446 227 91 1.1×10–9 0.999 817
Note: All 19 test trials were performed in the same borehole on the 1995 test pad.

medium. Following development of the velocity method, from velocity plots (or Z–t plots and eq. [25]) for tests per-
eq. [15] can be rewritten as formed in stage I and stage II. Computing the quotient of
dZ 1 eqs. [31] and [32] gives
[28] Z =− + He
dt α C ′ k′ C 2′ m
[33] = 1
C1′ m 2
where C ′ = c′/ Sinj .
The velocity plot of average elevation, Zm, during a time Both C1′ and C 2′ are functions of the ratio of anisotropy,
increment as a function of dZ/dt has the slope α = kh/kv. During stage I of the TSB test described by
1 Trautwein and Boutwell (1994), shape factor C1′ is, accord-
[29] mv = − ing to eq. [7],
α kv C ′
11D′ 1 11D
The vertical hydraulic conductivity of the medium is then [34] C1′ = =
πd 2 α πd 2
1
[30] kv = − Because of the limited thickness of the CCL, injection
αC ′ m v
zones for stage II of TSB tests rarely extend beyond one di-
The shape factor C′ of the transformed injection chamber ameter in length. Generally, the shape factor for stage II will
is a function of the unknown anisotropy, α = kh/kv. To solve correspond to eq. [9], giving
unknowns α and kv, Trautwein and Boutwell (1994) sug-
1 8D L α 1
gested performing borehole tests of varying geometry. The [35] C 2′ = +
test is first performed with the casing flush with the hole α d2 D 4
bottom (end-of-casing test). This is stage I of the test. Stage
II infiltration is measured with an extended borehole. During when 1 < (Lα1/2)/D < 8. Note that (Lα1/2)/D is the trans-
stage I, formed aspect ratio λ 2′ , and the shape factor C is as defined
in eq. [9]. Equation [33], combined with shape factors for
1
[31] kv = − stage I and II, then gives
α C1′ m1 2
m12 ⎛ 11 ⎞ 1
and during stage II, [36] λ 2′ = 2 ⎜ ⎟ −
m 2 ⎝ 8π ⎠ 4
1
[32] kv = − when 1 < λ 2′ < 8 and by eq. [27],
α C 2′ m 2
2
where C1′ and C 2′ are, respectively, shape factors for stage I ⎛λ′ D ⎞
[37] α=⎜ 2 ⎟
and stage II; and slopes m1 and m2 are obtained, respectively, ⎝ L ⎠

© 2005 NRC Canada


88 Can. Geotech. J. Vol. 42, 2005

Fig. 13. Hydraulic conductivity ratio of anisotropy as a function low hydraulic conductivity. Furthermore, computations will
of slopes m1 and m2 obtained from velocity plots of stages I and yield higher k values as scatter in the velocity plot increases.
II. Relation valid when, for stage II, 1 < λ′ < 8. Dashed lines In a velocity plot, average water column elevation (Zm)
are confidence intervals obtained when relative errors on slopes and falling-head velocity, v = dZ/dt, are both dependent vari-
m1 and m2 are both 10%. ables. Least-squares estimation is theoretically based on one
dependent variable being a function of another that is inde-
pendent. With low scatter, the choice of one or the other as
the dependent variable has little consequence. With high
scatter, k values depend on which variable is set as the inde-
pendent variable and which is set as the dependent variable.
The data in Table 2 yield k = 2.4 × 10–9 m/s when y = v and
x = Zm. The same data produce k = 4.4 × 10–10 m/s if y = Zm
and x = v. The k value being a function of which variable is
set as x and which is set as y is attributable to a hypothesis
behind the least-squares method: x is an independent vari-
able with no inaccuracies, and y is a dependent variable with
unknown random inaccuracies in measurements. Unfortu-
nately, both velocity, v, and average elevation, Zm, are depen-
dent variables, with both having some random inaccuracies.
Hence, they do not respect, in the strict sense, the hypothesis
behind the least-squares method. Where inaccuracies are
small, they will respect the hypothesis in a relaxed sense. In
such cases, they will yield k values comparable with those of
Theoretically, eqs. [36] and [37] give the ratio of aniso- the Z–t method and will not be sensitive to which variable is
tropy from measurements for stages I and II of a TSB test, set as x and which is set as y (data in Table 1, for example).
and eq. [31] or eq. [32] permits one to compute the vertical These observations lead us to conclude that the velocity
hydraulic conductivity, kv. The ratio of anisotropy as a func- method is not statistically robust. It should be used with cau-
tion of the slope ratio for a typical TSB test, where λ′ is tion when scatter in the velocity plot is observed.
within the range 1–8, is illustrated in Fig. 13. From a practi- The proposed Z–t method is found to be statistically more
cal point of view, inaccuracies in measurement cause high robust. It is independent of random inaccuracies in measure-
uncertainty for anisotropy. An excellent velocity plot, such ments (at least in the orders of magnitude found in routine
as illustrated in Fig. 4, gives 15% relative error on the slope tests). With this method, the intrinsic hypothesis mentioned
(based on the 95% confidence interval on slope value; Neter earlier for least-squares estimation is respected. Water col-
et al. 1989). Figure 13 shows a confidence interval where umn elevation (Z) and time (t) are, respectively, true depend-
relative errors on slopes m1 and m2 are both 10%. This confi- ent and independent variables. Plots also typically display
dence interval for the ratio of anisotropy spans a wide range. little data scatter, which clearly shows the theoretical interre-
For example, a TSB test in which the ratio for stages I and II lationship of t and falling water column elevation (Z) of
(m1/m2) is 4.22 gives a ratio of anisotropy (α) of 10, with a eq. [20].
confidence interval extending from 3.7 to 22. This is in Theoretically, the Hvorslev method should yield a correct
agreement with observations made by Chapuis (1999) and as k when the correct PL of the soil is used for computing head
earlier described in the introduction. Therefore, these equa- difference. In practice, inaccuracies in measurements yield
tions should be used with caution. high uncertainty in the estimation of the correct PL. This is
due to an important trend extrapolation to obtain Ho in both
the proposed and the velocity methods. The Hvorslev
Conclusion method will agree with the method with which the PL is es-
timated, even though the velocity and proposed methods
Three methods of interpretation for borehole falling-head
yield different hydraulic conductivity values. In light of this,
tests were reviewed. Hydraulic conductivity values com-
the direct use of the Hvorslev method without questioning
puted with the Hvorslev method were found to strongly de-
the PL value is not recommended.
pend on the correct PL of the soil. Because the PL is usually
An interpretation method for stages I and II of TSB tests
unknown, a level must be assumed. Depending on the eleva-
was also proposed. It yields the anisotropy and vertical hy-
tion at which PL is set, the method can yield an increasing k,
draulic conductivity of the liner. As a result of inaccuracies
a decreasing k, or a k that appears constant during the test.
in measurements and limited difference in geometry between
With the velocity method, the unknown PL of the soil is
stages I and II, computed anisotropy has high uncertainty.
not needed for computing hydraulic conductivity. Therefore,
an error in the assumed PL has no consequence for computa-
tions. When scatter in the velocity plot is low, the method Acknowledgements
yields good hydraulic conductivity values (these are found to
agree with the proposed method). Unfortunately, even small This research on the interpretation of falling-head bore-
inaccuracies in measurements create large inaccuracies in hole tests was supported by the Nepisiguit-Chaleur Solid
computed velocities, particularly when testing materials with Waste Commission and in part by the Natural Sciences and

© 2005 NRC Canada


Chiasson 89

Engineering Research Council of Canada and by the Faculté Chapuis, R.P. 2002. The 2000 R.M. Hardy lecture: full-scale hy-
des études supérieures et de la recherche de l’Université de draulic performance of soil–bentonite and compacted clay lin-
Moncton. The author is grateful for the assistance offered by ers. Canadian Geotechnical Journal, 39(2): 417–439.
Mr. Raymond Bryar, manager of the Commission, and by Chapuis, R.P., and Cazaux, D. 2002. Pressure-pulse test for field
Mr. Marc Antoine Caissie, engineer with the Roy Consul- hydraulic conductivity of soils: Is the usual interpretation
tants Group. Field testing performed by Benjamin Chiasson, method adequate? In Evaluation and remediation of low perme-
Rémi Godin, and Wamytan Rezza is also acknowledged. ability and dual porosity environments. ASTM STP 1415.
The author wishes to thank Mr. Camille Vautour, technician, Edited by N.N. Sara and L.G. Everett. ASTM International,
and Mrs. Jolaine Landry, laboratory engineer, for their tech- West Conshohocken, Penn., pp. 66–82.
Chapuis, R.P., and Chenaf, D. 2002. Slug tests in a confined aqui-
nical assistance throughout this research program. Special
fer: experimental results in a large soil tank and numerical mod-
thanks are due to Dr. Robert Chapuis of the École poly-
eling. Canadian Geotechnical Journal, 39(1): 14–21.
technique de Montréal for his constructive comments and
Chapuis, R.P., Paré, J.J., and Lavallée, J.G. 1981. In situ variable
helpful advice during the preparation of this paper. head permeability tests. In Proceedings of the 10th International
Conference on Soil Mechanics and Foundation Engineering,
References Stockholm, Sweden, 15–19 June. Edited by N. Foldin, J.
Lindgren, H. Fagerström, and B. Steen. A.A. Balkema, Rotter-
AFNOR. 1992. Norme NF P94-132. Essai d’eau Lefranc. Associa-
dam. Vol. 1., pp. 401–406.
tion française de normalisation, Paris.
Chen, W.F., and Liew, J.Y.R. 2003. The civil engineering hand-
ASTM. 2004a. Standard test method for field measurement of
book. 2nd ed. CRC Press, New York.
infiltration rate using a double-ring infiltrometer with a sealed-
inner ring (D5093). In Annual book of ASTM standards. Chiasson, P., Massiéra, M., Comeau, S., and Caissie, M.A. 1998.
Vol. 04.08. American Society for Testing and Materials, West Characterization of till materials used for liners at the Red Pine
Conshohocken, Penn. Regional Sanitary Landfill, Canada. In Proceedings of the 8th
International Congress of the International Association for Engi-
ASTM. 2004b. Standard test method for field measurement of hy-
neering Geology and the Environment, Vancouver, B.C., 21–
draulic conductivity limits of porous materials using two stages
25 September 1998. Edited by D.P. Moore and O. Hungr. Vol. 4.
of infiltration from a borehole (D6391). In Annual book of
A.A. Balkema, Brookfield, Vt., pp. 2491–2498.
ASTM standards. Vol. 04.08. American Society for Testing and
Materials, West Conshohocken, Penn. Cooper, H.H., Jr., Bredehoeft, J.D., and Papadopulos. 1967. Re-
Benson, C.H., and Boutwell, G. 1992. Compaction control and sponse of a finite-diameter well to an instantaneous charge of
scale-dependent hydraulic conductivity of clay liners. In Pro- water. Water Resources Research, 3(1): 263–269.
ceedings, 15th Annual Madison Waste Conference, University Daniel, D.E. 1984. Predicting hydraulic conductivity of clay lin-
of Wisconsin, Madison, Wis., 23–24 September 1992. Edited by ers. Journal of Geotechnical Engineering, ASCE, 110(2): 285–
Department of Engineering Development, University of Wiscon- 300.
sin at Madison. pp. 62–83. Daniel, D.E. 1998. Landfills for solid and liquid wastes. In Pro-
Benson, C.H., Daniel, D.E., and Boutwell, G.P. 1999. Field perfor- ceedings of the 3rd International Congress on Environmental
mance of compacted clay liners. Journal of Geotechnical and Geotechnics, Lisbon, Portugal, 7–11 September 1998. Edited by
Geoenvironmental Engineering, 125(5): 390–403. P.S.S. e Pinto. Vol. 4. A.A. Balkema, Brookfield, Vt., pp. 1231–
Bjerrum, L., and Huder, J. 1957. Measurement of the permeability 1246.
of compacted clays. In Proceedings of the 4th International Con- Daniel, D.E., and Koerner, R.M. 1995. Waste containment sys-
ference on Soil Mechanics and Foundation Engineering, Lon- tems: guidance for construction, quality assurance, and quality
don, U.K., 12–24 August 1957. Edited by L.F. Cooling, A.W. control of liner and cover systems. ASCE Press, New York.
Bishop, H.Q. Golder, and D.J. Maclean. Butterworths Scientific Daniel, D.E., and Wu, Y.K. 1993. Compacted clay liners and cov-
Publications, London. pp. 6–10. ers for arid sites. Journal of Geotechnical Engineering, ASCE,
CAN–BNQ. 1988a. Soils—determination of permeability at the 119(2): 223–237.
end of a casing. Canadian Standards Association and Bureau de Elsbury, B.R., Daniel, D.E., Sraders, G.A., and Anderson, D.C.
normalisation du Québec, Rexdale, Ont. CAN/BNQ 2501-130- 1990. Lessons learned from compacted clay liners. Journal of
M-88. Geotechnical Engineering, ASCE, 116(11): 1641–1660.
CAN–BNQ. 1988b. Soils—determination of permeability by the Hvorslev, M.J. 1951. Time lag and soil permeability in ground
Lefranc method. Canadian Standards Association and Bureau de water observations. U.S. Army Corps of Engineers, Engineer
normalisation du Québec, Rexdale, Ont. CAN/BNQ 2501-135- Research and Development Center, Waterways Experimental
M-88. Station, Vicksburg, Miss. Bulletin 36.
Cassan, M. 2000. Application des essais Lefranc à l’évaluation du Jacques Whitford Ltd. 1995. Permeability testing of the Red Pine
coefficient d’anisotropie hydraulique des sols aquifères. Revue landfill liner. Jacques Whitford and Associates Ltd., Bathurst,
française de géotechnique, 90: 25–43. N.B.
Chapuis, R.P. 1989. Shape factors for permeability tests in bore- Lambe, T.W. 1958a. The structure of compacted clay. Journal of
holes and piezometers. Ground Water, 27(5): 647–654. the Soil Mechanics and Foundations Division, ASCE, 84(SM2):
Chapuis, R.P. 1998. Overdamped slug test in monitoring wells: 1654-1 – 1654-34.
review of interpretation methods with mathematical, physical, Lambe, T.W. 1958b. The engineering behaviour of compacted clay.
and numerical analysis of storativity influence. Canadian Geo- Journal of the Soil Mechanics and Foundations Division, ASCE,
technical Journal, 35(5): 697–719. 84(SM2): 1655-1 – 1655-35.
Chapuis, R.P. 1999. Borehole variable-head permeability tests in Leroueil, S., Le Bihan, J.P., and Bouchard, R. 1992. Remarks on
compacted clay liners and covers. Canadian Geotechnical Jour- the design of clay liners used in lagoons as hydraulic barriers.
nal, 36(1): 39–51. Canadian Geotechnical Journal, 29: 512–525.

© 2005 NRC Canada


90 Can. Geotech. J. Vol. 42, 2005

Mitchell, J.K., Hooper, D.R., and Campanella, R.G. 1965. Perme- tivity tests for compacted soil liners and caps. In Hydraulic con-
ability of compacted clay. Journal of the Soil Mechanics and ductivity and waste contaminant transport in soils. ASTM STP
Foundations Division, ASCE, 91(SM4): 41–65. 1142. Edited by D.E. Daniel and S.J. Trautwein. American Soci-
Neter, J., Wasserman, W., and Kutner, M.H. 1989. Applied linear ety for Testing and Materials, Philadelphia, pp. 184–226.
regression models. 2nd ed. Irwin, Boston, Mass. U.S. EPA. 1991. Code of Federal Regulations. Title 40, Protection
Schneebeli, G. 1954. La mesure in situ de la perméabilité d’un ter- of Environment. Part 258, Criteria for municipal solid waste
rain. Comptes-rendus des 3ièmes journées d’hydraulique, Alger, landfills. U.S. Environmental Protection Agency, Washington,
pp. 270–279. D.C.
Trautwein, S.J., and Boutwell, G.P. 1994. In situ hydraulic conduc-

© 2005 NRC Canada

View publication stats

Potrebbero piacerti anche