532
Proceedings of the 18
th
International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013
fluctuating glacial advance and retreat produced a complex
distribution of over-consolidated glacial till layers, separated by
interstadial and interglacial stratified deposits of glacio-
lacustrine plastic silt/clays and non-plastic silt/sands.
The subsurface overburden encountered during the site
investigation were initially classified into 17 different soil types
(Types 1 through 17). The soil classification system followed
the modified version of Unified Soil Classification System.
Identification of soil origin as “till” was based on their
heterogeneous structure, the relatively broad grain size
distributions and the documented local geology. Many of the
different soil types demonstrate relatively comparable
engineering characteristics and may possibly have similar
geological origin. Consequently, the various soil types were
consolidated into six engineering classes (Classes A through F).
The six soil classes are as follows:
Class A: Fill and Topsoil
Class B: Interstadial Sand to Gravel
Class C: Interstadial Silt to Sand
Class D : Non-Plastic Till
Class E : Plastic Glacio-lacustrine
Class F : Plastic Till
Class B was divided into two subclasses based on the
percentage of silt and clay particles (<75 μm). Sandy soils with
less than 20% silt and clay particles were grouped under Class
B2,3,4 and the rest (> 20% silty and clay) under Class B5,6.
3 ESTABLISHING HYDRAULIC CONDUCTIVITY
Glacial deposits by nature comprise of variable soils types in
relatively short distances. Due to the inherent variable nature of
the glacial deposits at project area, conventional filed pumping
tests may not provide fully reliable results for a proper
dewatering calculation as the zone of influence of a pump test
may only extend a few tens of meters. On the other hand, the
actual dewatering volume of a structure is affected by the
characteristics of surrounding soil within a few hundreds of
meters. Furthermore, the pumping tests were not necessarily at
the exact location of some structures.
It became necessary to complement the hydraulic
conductivity values obtained through field testing in order to
expand the test results to a larger domain or be able to focus on
any specific area. It was decided to use the available semi-
empirical methods/formulae in literature to complement
hydraulic conductivity values obtained through filed testing
with predicted values based on index properties such as grain
size distributions, pore size distributions and/or specific surface.
The following sections will outline the procedure followed to
predict hydraulic conductivities and provide design parameters.
3.1
Kozeny-Carman formula
Since Kozeny (1927) introduced his theory for a series of
capillary tubes and Carman (1938 and 1956) followed this work
and provided formulations that takes into the account the
tortuosity of the flow path of a fluid in a porous medium. The
following formula presented by Carman was then referred to as
the Kozeny-Carman (K-C) formula (Carrier, 2003).
Details of the formula can be found in the subject references.
In summary, the hydraulic conductivity of the soil can be
estimated as follows:
)]
1/( )[
/1( )}
/( { /%100 10 99.1
3 2
2 5.0
5.0
4
e
e SF
D D f
K
si
li
i
(1)
Where,
e
is the void ratio; SF is a shape factor;
f
i
is the
fraction of particles between two sieves (%), denoting the larger
sieve with (l) and the smaller one as (s) in, and
D
ave-i
=
(D
li
×D
si
)
0.5
is the average particle size, in cm, between two sieve
sizes
.
.
The Kozeny-Carman formula takes into account specific
surface area of full range of particle sizes and soil void ratio
which leads to better accuracy than the famous Hazen formula
(Lambe and Whitman 1969) in predicting the hydraulic
conductivity for a wide range of soils. Notwithstanding the
above, the application of K-C formula is constrained by almost
the same limitations as Hazen (Carrier 2003). Such constrains,
as discussed below, arise when dealing with soils at the
extremes of any spectrum such as the grain size, particle size
distribution, particle shape, and particles orientation
(anisotropy).
The formula does not account for the electrochemical forces
between particles and particles and water which disqualify the
formula from being applied to clayey soils. In addition, the
formula assumes laminar flow, which may not be satisfied in
gravels and gravelly sands. The formula does not produce a
close estimate to the specific surface area of particles with
extreme shapes such as platy or flakey particles. Therefore, the
K-C formula may not be applicable in these cases or can be
applied after replacing the calculated specific surface area by
the measured value. Also, K-C formula does not account for soil
anisotropy which is more pronounced in natural deposits than
for laboratory constructed samples.
Locat et al (1984) measured the specific surface area (S) for
several clays and found that clays with low plasticity (8 < PI <
15) have S between 23 and 30 m
2
/kg and is independent of the
percentage of soil finer than 2
m. Chapuis and Aubertin (2003)
picked a constant number between 23 and 30 m
2
/kg as an
estimate for S of the soil fraction finer than 2
m and calculated
S for the fraction coarser than 2
m as per original K-C
formula. Consequently, the results of these hybrid methods in
using K-C formula were in good agreement with measured
hydraulic conductivities in laboratory for clayey soils with
PI<15. In this study, the approach proposed by Chapuis and
Aubertin (2003) was followed for plastic glacial tills with PI
less than 15. However, the effect of weathering and factures in
the upper portion of the clayey till deposits must be considered
in any assessment (McKay, 1993; Hendry, 1982).
3.2
Site specific correlation factor for K-C formula
This section outlines the work completed in the field to obtain
in-situ hydraulic conductive (K) for the different soil classes
and explains the approach followed to establish site specific
correlation factor for using K-C formula.
Hydraulic conductivities for each soil class were measured in
the field by a combination of pumping tests and/or falling or
rising head slug tests. The results of 8 pumping tests with
associated observation wells and 88 slug tests conducted along
the tunnel alignment, distributed among six soil classes are used
in this study. The number of the field tests performed on the
aquifers’ materials was greater than those performed on the
other soil types. However, a significant number of the tests were
performed on both plastic and non-plastic tills.
One grain size distribution analysis was conducted, as
minimum, on the soil samples recovered from within the screen
interval of the 88 slug tests and pumping tests with associated
observation wells. These grain size distributions were
determined by undertaking sieve analysis, in accordance to
ASTM C136-06, and the hydrometer test, in accordance to
ASTM D422-63. These grain size distributions analyses were
used to calculate K based on the K-C formula. After excluding
the tests for samples with PI >15 and/or field test conducted in
the clayey till deposits with obvious signs of weathering and
fracture, K-C formula was applied to about 80 grain size
analyses that were screened as suitable (not within the
limitations of the formula) and correspond with K obtained
from field tests. As a result, for every in-situ measured K in the
field (K
field
) there is a corresponding predicted K from applying
KC formula to the grain size analysis associated with the screen
interval (K
KC
), as shown in Figure 1.