Вы находитесь на странице: 1из 9

Engineering Geology 199 (2015) 148–156

Contents lists available at ScienceDirect

Engineering Geology

journal homepage: www.elsevier.com/locate/enggeo

Estimation of permeability function from the soil–water


characteristic curve
Qian Zhai a, Harianto Rahardjo b,⁎
a
School of Civil and Environmental Engineering, Nanyang Technological University, Block N1, B1a-01a, Nanyang Avenue, Singapore 639798, Singapore
b
School of Civil and Environmental Engineering, Nanyang Technological University, Block N1, #1B-36, 50 Nanyang Avenue, Singapore 639798, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: Because direct measurement is time-consuming and costly, the permeability function of unsaturated soil is com-
Received 9 October 2014 monly determined by estimation using the soil–water characteristic curve (SWCC). Various prediction models
Received in revised form 10 October 2015 (i.e., indirect methods) for the permeability function have been proposed by different researchers. Mualem
Accepted 1 November 2015
(1986) categorized these prediction models into three groups: empirical, macroscopic, and statistical models.
Available online 9 November 2015
Of these, the statistical model is the most rigorous and provides the most accurate results (Leong and Rahardjo
Keywords:
1997). In this paper, a new equation for the calculation of the permeability function is proposed in which the per-
Unsaturated soil meability function of unsaturated soil is correlated with the fitting parameters of SWCC. In other words, unsatu-
Soil–water characteristic curve rated hydraulic conductivity is computed using an electronic spreadsheet with input parameters (i.e., the fitting
Pore-suction distribution function parameters of SWCC). The proposed equation is shown to be the general form of the equations by Marshall
Permeability function (1958) and Kunze et al. (1968). In the equation, suction is considered as a variable and SWCC, in the form of
degree of saturation, S, is adopted as a probability function. Soil volume change is also incorporated in this equa-
tion. Lastly, the proposed equation is verified using experimental data from the literature.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction is time-consuming and costly. Various estimation/prediction models


for the calculation of the permeability function for unsaturated soil
Water flow through soil can significantly affect the stability of a have been proposed. Mualem (1986) categorized the various indirect
slope during rainfall (Rahardjo et al. 2010). Water flow through soil is measurement methods into three groups: empirical, macroscopic, and
dependent on the hydraulic conductivity of the soil as well as the hy- statistical models. Of these, the statistical model is the most rigorous
draulic gradient. If the soil volume change and chemical composition and it provides the most accurate results (Leong and Rahardjo 1997).
of the pore fluid are assumed to be insignificant, which is a reasonable Fredlund et al. (2012) recommended and illustrated the calculation of
assumption for sandy material, the hydraulic conductivity is almost a permeability function using statistical models, such as those by
constant for saturated soil; however, for unsaturated soil, the hydraulic Marshall (1958) and Kunze et al. (1968).
conductivity is a function of the volumetric water content, θw, or the Romero et al. (1999) and Romero (2013) presented experimental
degree of saturation, S. As the volumetric water content, θw, or degree results for both the macrostructure and microstructure of soil. Inter-
of saturation, S, is a function of matric suction, as defined in the soil– aggregate pores (i.e., macrovoids between soil aggregates or shielding
water characteristic curve (SWCC), the unsaturated hydraulic conduc- grains) and intra-aggregate pores (i.e., microvoids inside clay aggre-
tivity can also be considered a function of matric suction. According to gates) were presented. According to Romero et al. (2011), the intra-
Fredlund et al. (2012), the estimation procedures for obtaining unsatu- aggregate porosity displayed non-constricted porosity with no bottle-
rated soil properties are approximate but are generally satisfactory for neck effects and the water adsorption storage mechanism was not
analyzing unsaturated soil mechanics problems. In other words, the affected by porosity variation and the restricted capacity to liquid
hydraulic conductivity of unsaturated soil is commonly estimated from flow. However, the inter-aggregate porosity displayed interconnected
SWCC instead of being measured directly because direct measurement porosity and the water storage mechanism was affected by void ratio
changes. Research by Romero et al. (1999); Romero et al. (2011) and
Romero (2013) suggests that a statistical model for the prediction
of the permeability function of unsaturated soil is only applicable for
⁎ Corresponding author.
inter-aggregate governing suction.
E-mail addresses: zhaiqian@ntu.edu.sg (Q. Zhai), chrahardjo@ntu.edu.sg One of the assumptions of the statistical method is that SWCC is
(H. Rahardjo). analogous to the pore-size distribution function (PSD). Della Vecchia

http://dx.doi.org/10.1016/j.enggeo.2015.11.001
0013-7952/© 2015 Elsevier B.V. All rights reserved.
Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156 149

et al. (2015) and Beckett and Augarde (2013) illustrated that the pore- Marshall (1958) equations, Kunze et al. (1968) presented a simple
size distribution (PSD) function can be measured using a mercury intru- Eq. (1) for the calculation of the permeability function:
sion porosimetry (MIP) test. Casini et al. (2012) investigated the evolu-
tion of the microstructure of aggregate fabric due to compaction by ks X m  
kw ðθw Þi ¼ Ad ð2j þ 1−2iÞðua −uw Þ j −2 ; i ¼ 1; 2; ⋯; m ð1Þ
analyzing changes in the pore-size distribution function under different ksc j¼i
initial conditions. Casini et al. (2012) concluded that, at low water con-
tents, the microporosity (i.e., the ratio of intra-aggregate pores to total where,
pores in soil) decreases and that the PSD function tends to be a quasi-
monomodal distribution for the driest compaction. kw(θw)i predicted coefficient of permeability for volumetric water
As an alternative to the statistical models, new technologies and content;
models have been proposed in recent years for the estimation of hy- ua air pressure (kPa);
draulic conductivity. Cosentini et al. (2012) used electrical resistivity to- uw pore-water pressure (kPa);
mography (ERT) to quantify local changes in water content within a soil (θw)i corresponds to the ith interval (m/s);
sample subject to wetting to estimate the hydraulic parameters of un- i interval number that increases as the volumetric water con-
saturated soil. Zhou et al. (2014) examined the relative hydraulic con- tent decreases;
ductivity of unsaturated granular soil using the newly-proposed SWCC j a count from “i” to “m”;
model called MFX and the Fredlund et al. (1994) equation. Hu et al. m total number of intervals between the saturated volumetric
(2013) proposed a model for the prediction of SWCC and hydraulic con- water content, θs, and the lowest volumetric water content, θL;
ductivity for deformable soils by horizontal shifting and vertical scaling ks measured saturated coefficient of permeability (m/s);
of the pore-size distribution (PSD) function. In addition, Ye et al. (2012) ksc calculated saturated coefficient of permeability (m/s);
showed that the permeability function can be affected by temperature. Ad adjusting constant;
However, temperature effects are not considered in this study.
In this paper, a new equation for the calculation of the permeability In Kunze et al.'s (1968) equation, the interval Δψ in the higher
function of unsaturated soil is derived. In the proposed model, the direct suction range is much greater than the interval Δψ in the lower suc-
division of suction is adopted, which is convenient for computational tion range. The calculation of Δψ from θw is dependent on iteration
programming. SWCC, in the form of degree of saturation, S, is consid- and cannot always be easily solved using electronic spreadsheet. In
ered analogous to the pore-size distribution function and adopted as addition, the soil volume change with respect to suction is not
the probability function of random connections. In addition, the soil vol- incorporated in the models by Childs and Collis-George (1950);
ume change is incorporated into the calculation of the permeability Marshall (1958) or Kunze et al. (1968).
function. The derivation shows that the proposed equation is a general Mualem (1976) showed that Kunze et al.'s (1968) equation can be
form of the Marshall (1958) and Kunze et al. (1968) equations. The pro- expressed in an analytical form as follows:
posed equation is verified using published experimental data from the
literature. Zθw
ðθw −ϑ Þ
As suction is always referred to as a state variable for unsaturated dϑ
soil, the pore-size distribution function is replaced with the pore- ψ2
0
kðθw Þ ¼ : ð2Þ
suction distribution function using Kelvin's capillary theory throughout Zθs
this paper. The pore-size distribution function defines the relationship ðθw −ϑ Þ

between pore size density and pore size, while the pore-suction distri- ψ2
0
bution function defines the relationship between pore size density
and matric suction. Leong and Rahardjo (1997) proposed an empirical model to
determine the permeability function for unsaturated soil from ex-
2. Literature review perimental data by modifying Fredlund and Xing's (1994) equation
as follows:
Childs and Collis-George (1950) proposed a model to predict the co-
efficient of permeability based on the theory of “cutting and random 1
kr ¼    b  c0 ð3Þ
rejoining”, which considers pores in soil as a series of capillary tubes
that are randomly connected to each other in a given cross-section. ln e þ ψa
This model makes three major assumptions: (a) pores in soil can be con-
sidered a set of randomly distributed interconnected pores character- where,
ized by a pore radius, r, and its density, f(r), and the density, f(r), is
the same for any cross-section; (b) the Poiseuille equation is applicable; kr relative hydraulic conductivity, i.e., kr = k(θw) / ks;
and (c) SWCC is analogous to the pore-size distribution function. c′ cp;
Marshall (1958) improved the model by Childs and Collis-George a, b, c fitting parameters for Fredlund and Xing's (1994) equation
(1950) by simplifying the pore-size distribution function as a uniform with correction factor C(ψ) = 1;
distribution. As statistical models use a few discrete points rather than p additional fitting parameter.
a continuous mathematical model to represent the permeability func-
tion of unsaturated soil, the accuracy of statistical models is highly de- Fredlund and Xing's (1994) equation is illustrated in Eq. (4).
pendent on the numbers and locations of these discrete points. On the 2

3
other hand, the locations of these discrete points are dependent on ψ
6 ln 1 þ 7
θs 6 Cr 7
the manner in which the entire range of suction is discretized into divi-
θ ¼ C ðψÞ n h  n iom ¼ 61− ! 7 n h θs  io ð4Þ
sions. Kunze et al. (1968) proposed equally dividing the volumetric 6 7 n m
ln e þ ψa 4 106 5 ln e þ ψ
ln 1 þ a
water content, θw, into intervals, Δθw, and calculating the interval of Cr
matric suction, Δψ, accordingly. Equal division of the volumetric water
content, θw, makes the density, f(r), unique for all pore radii, meaning where,
that the pore-size distribution function follows a uniform distribution.
With improvements to the Childs and Collis-George (1950) and a, n, m fitting parameters;
150 Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156

Cr an input value related to the residual suction, which can be distribution function defines the degree of saturation, as illustrated in
roughly estimated as Cr = 1500 kPa for most cases Fig. 2. The overall area under the pore-suction distribution function
(Fredlund and Xing (1994) and Zhai and Rahardjo (2012b)). 106
θs saturated volumetric water content. should be equal to 1 (i.e., ∫ f ðψÞdψ ¼ 1). Therefore, the SWCC in the
0:01
form of degree of saturation can be considered analogous to the pore-
Leong and Rahardjo (1997) suggested that the accuracy of the suction distribution function.
results of the permeability function could be improved by using The pore-suction distribution function is a function of suction. In
Eq. (3) as the best fit equation for relative permeability data from other words, the pore-suction distribution function changes with
experiments. changes in suction because a soil volume change occurs during the dry-
By expressing the volumetric water content, θw, in the form of ing or wetting process (Fredlund and Pham 2006; Cuisinier et al. 2014).
degree of saturation, S, Fredlund and Xing's (1994) equation can be At suction ψ1, the pore-suction distribution function is a curve, as shown
expressed as follows: on the left-hand side of Fig. 3. If the suction increases to ψ2, the pore-
suction distribution function may change into the curve on the right-
1 hand side of Fig. 3, similar to results presented by Hu et al. (2013). Al-
S ¼ C ðψÞ n h  n1 iom1 ð5Þ
ln e þ ψ though the pore-suction distribution function may change with a
a1
change in suction, the definition of the degree of saturation using the
pore-suction distribution function is still applicable, as illustrated in
where, a1, n1 and m1 are the fitting parameters that differ from a, n and Fig. 3.
m in Eq. (4) if soil volume change is considered (Zhai and Rahardjo The entire range of pores in the soil can be divided into a series of
2014). Eq. (5) is adopted in this paper for the prediction of the perme- groups. Each group can be treated as a capillary tube with a certain di-
ability function as this equation defines SWCC in the form of degree of ameter, as illustrated in Fig. 4. The diameter of the capillary tube repre-
saturation. sents the pore radius of an individual group of pores. The minimum
value, the maximum value or the mean value of the pore radius can
3. Theory be adopted as the diameter of the capillary tube. Normally, the mean
value gives the most accurate result. However, if the interval is very
The equation for the calculation of unsaturated hydraulic conductiv- small, the minimum or the maximum value may give satisfactory re-
ity is derived in this section. There are four assumptions adopted in this sults. The normalized volume (i.e., the ratio of the volume of pores
derivation: (a) pores in soil can be represented by a limited number of from an individual group to the volume of all pores in the soil) of an in-
pores with a pore radius of r and a pore size density of f(r), and f(r) is dividual capillary tube can be calculated from the area under the pore-
the same for any cross-section; (b) Poiseuille's law for stream-line suction distribution function, which is equal to ΔS as illustrated in Fig.
flow is also applicable for water flow in unsaturated soil; (c) SWCC, in 4. Therefore, the probability of a capillary tube with a larger diameter
the form of degree of saturation, is analogous to the pore-suction distri- connecting to a capillary tube with a smaller diameter is ΔS1 ∗ ΔS2 as il-
bution function and can be referred to as the probability function of ran- lustrated in Fig. 4.
dom connections; and (d) there is no entrapped air in the pores.

3.1. Probability function of random connections 3.2. Interval of suction, Δψ

The SWCC and the pore-suction distribution function are illustrated The radius, r, and its pore size density, f(r), are dependent on the
in Fig. 1. Degree of saturation, S, defines the ratio of the volume of water interval of suction, Δψ. The entire suction range can be divided into
in the soil as the total volume of voids (or pores) in the soil. Fredlund intervals so that f(r) is unique for all the radii, as illustrated in Fig.
and Rahardjo (1993) showed that suction, ψ, is related to the radius of 5(a). Alternatively, the suction range can be divided equally so that
the meniscus using Kelvin's capillary law: a higher suction results in a Δψ is unique for all suction ranges, as illustrated in Fig. 5(b).
smaller radius. If the radius of the meniscus is smaller than the radius Kunze et al. (1968) adopted the first approach, but used SWCC in the
of the pore, air will break through the meniscus and the water in the form of volumetric water content (instead of SWCC in the form of de-
pore will drain out. In other words, the area under the pore-suction gree of saturation) as the probability function for random connections.

Fig. 1. Illustration of the SWCC and pore-suction distribution function.


Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156 151

Fig. 2. Illustration of the relationship between the degree of saturation and pore-suction distribution function.

In Kunze et al.'s (1968) equation, the interval of suction Δψ must be cal- rectangles are used to represent the soil particles and pores, respective-
culated from θw. In addition, in Kunze et al.'s (1968) equation, the inter- ly. There are six possible scenarios of connections, which are all illustrat-
val Δψ in either very high or very low suction ranges is much larger than ed in Fig. 8. Scenario (a) illustrates a soil particle connected to a soil
in a medium suction range, which makes the calculated results less ac- particle; (b) illustrates a dry pore connected to a soil particle;
curate, as illustrated in Fig. 6. No results are obtained between suctions (c) illustrates a saturated pore connected to a soil particle;
of 0.01 kPa and 0.3 kPa, and only two data points are obtained beyond a (d) illustrates a dry pore connected to a dry pore; (e) illustrates a dry
suction of 5 kPa. The calculated results mostly gather between suctions pore connected to a saturated pore; and (f) illustrates a saturated pore
of 0.3 kPa and 5 kPa. It seems that the results calculated using Kunze connected to a saturated pore.
et al.'s (1968) equation are less accurate for low and high suction ranges, It is assumed that scenarios (a) to (e) are impervious as the water
as the results mostly cluster in the medium suction range. In this paper, phase is discontinuous; only scenario (f) allows water to flow through
the equal interval Δψ is adopted and the calculated results are evenly both capillary tubes. The permeability of a section is dependent on the
distributed across the entire suction range. smaller diameter of the capillary tube, as illustrated in scenario (f).
Therefore, the permeability of a section is dependent on the probability
3.3. Equivalent effective radius for the permeability of saturated soil of scenario (f) occurring and the diameter of the smaller capillary tube.
If the soil is fully saturated, there are no dry pores and scenarios (b),
Consider a soil element cut into two pieces. The two sections along (d) and (e) cannot occur. The equivalent effective radius for the perme-
the cutting are defined as section A and section B. The groups of pores ability of saturated soil is dependent on the probability of connections of
in the two sections are considered capillary tubes with different diame- pores and can be expressed as follows:
ters (i.e., r1, r2, —, ri, —, rn, where r1 N r2 N — N ri N — N rn). Pores with dif-
ferent radii are distributed following the pore-suction distribution
function in both sections, as illustrated in Fig. 7. !
N h
X i
The soil particles and pores with different radii in section A are ran- r 2t ¼ n2 ½1−Sðψi Þ2 −½1−Sðψi−1 Þ2 r 2i ; where Sðψ0 Þ ¼ 1: ð6Þ
domly connected to soil particles and pores in section B. Circles and i¼1

Fig. 3. Illustration of degree of saturation using pore-suction distribution functions at different suctions.
152 Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156

Fig. 4. Illustration of equivalent capillary tubes associated with the pore-suction distribution function.

3.4. Equivalent effective radius for the permeability of unsaturated soil 3.5. Coefficient of permeability for saturated and unsaturated soil

If the soil is unsaturated, dry pores exist and scenarios (b), (d) and Substituting Eq. (6) into Poiseuille's equation, the permeability for
(e) can occur. The number of dry pores is dependent on the suction saturated soil can be obtained as follows:
state; higher suction results in more dry pores in the soil. Any dry ( )
N h i
pore or soil particle in a section make the section impervious. The equiv- 1 2 X
alent effective radius for the permeability of unsaturated soil is depen- ks ¼ n ½1−Sðψi Þ2 −½1−Sðψi−1 Þ2 r 2i ; ð8Þ
8 i¼1
dent on the probability of connections of saturated pores. If the
suction state in the soil is ψm, all pores with a radius greater than rm.
where, S(ψ0) = 1.
(i.e. rm. = 2T / ψm) are dry pores, while pores with a radius less than
Substituting Eq. (7) into Poiseuille's equation results in a relative
rm. are saturated pores. The radii of dry pores are named r1, r2 to rm.,
permeability function kr for unsaturated soil as follows:
while the radii of saturated pores are rm + 1, rm + 2 to rn. The probability
of connections of saturated pores is illustrated in Fig. 9. N h i
2 X
Therefore, the effective radius for the permeability of unsaturated Sðψm Þ−S ψmþ1 ψ−2m þ ðSðψm Þ−Sðψi ÞÞ2 −ðSðψm Þ−Sðψi−1 ÞÞ2 ψ−2
i
soil at a suction state of ψm can be expressed as follows: n2m i¼mþ1
kr ¼
n2s N h
X i
½1−Sðψi Þ2 −½1−Sðψi−1 Þ2 ψ−2
i
! i¼1
  2 N h
X i
r 2t ¼n 2
Sðψm Þ−S ψmþ1 r2m þ ½Sðψm Þ−Sðψi Þ2 −½Sðψm Þ−Sðψi−1 Þ2 r 2i : ð9Þ
i¼mþ1

ð7Þ where, ks = saturated coefficient of permeability;

Fig. 5. Two approaches to divide the entire suction range into different groups.
Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156 153

Fig. 6. Calculated relative hydraulic conductivity kr (i.e. kr = k(ψ) / ks) using Kunze et al.'s (1968) equation.

ψm = suction state in the soil; i

ns = porosity corresponding to saturated state; ∑ f ðψi Þ ¼ Ni into Eq. (8), the hydraulic conductivity can be obtained as
j¼1
nm = porosity corresponding to suction ψm. given in Eq. (11), which is the same as Marshall's (1958) equation:
If the coefficient of permeability k(ψm), with respect to a suction of
ψm, is known, then the coefficient of permeability k(ψm + i), with re- N h i
spect to a suction of ψm + i, can be calculated using Eq. (10): n2 X n2  
k¼ 2
ðiÞ2 −ði−1Þ2 r2i ¼ 2 r 21 þ 3r 22 þ 5r23 þ …ð2n−1Þr 2n =8: ð11Þ
N i¼1 N
nmþi 2
kðψmþi Þ ¼ kðψm Þ
nm 2
8 9 By substituting S(ψm) − S(ψi) = (i − m) / N into Eq. (9) and ignor-
< 2 XN  2  2  =
Sðψmþi Þ−Sðψmþiþ1 Þ r 2mþi þ Sðψmþi Þ−Sðψ j Þ − Sðψmþi Þ−Sðψ j−1 Þ r 2j ing the soil volume change, the relative hydraulic conductivity of unsat-
: ;
j¼mþiþ1 urated soil can be obtained as follows:
( ) :
 2 N 
X 2  2 
Sðψm Þ−Sðψmþ1 Þ r2m þ Sðψm Þ−Sðψi Þ − Sðψm Þ−Sðψi−1 Þ r2i "

2
#
i¼mþ1
X
N
m i m i−1 2 −2
1− − 1− − 1− −1− ψi
ð10Þ kðψm Þ i¼m N N N N
kr ¼ ¼ "

2

#
ks XN
i i−1 2 −2
1− 1− − 1− 1− ψi
Eq. (10) adopts SWCC in the form of degree of saturation and poros- N N
i¼1
ity ratio (i.e., n2m + i / n2m), where nm is the porosity of soil correspond- ð12Þ
XN h i
ing to suction ψm and nm + i is the porosity of soil corresponding to ði−mÞ2 −ði−m−1Þ2 ψ−2 i
suction ψm + i. Both SWCCs in the form of degree of saturation and po- i¼m
¼ N h
:
rosity ratio help incorporate the effect of soil volume change into the X i
prediction of the permeability function from SWCC. As Eq. (10) corre- ðiÞ2 −ði−1Þ2 ψ−2
i
lates the relative permeability function directly to the SWCC, the uncer- i¼1

tainty of relative permeability function can be estimated using Zhai and


Rahardjo (2013b) equations. Eq. (12) is the same as Kunze et al.'s (1968) equation given in Eq. (1).
Therefore, the proposed equation is proven to be a general form of
3.6. Relation between the proposed equation and equations by Marshall Marshall's (1958) and Kunze et al.'s (1968) equations. As an equal divi-
(1958) and Kunze et al. (1968) sion of matric suction is adopted in the proposed equation, while an
equal division of volumetric water content was adopted in Kunze
If the pore-suction distribution function is considered to be uniformly et al.'s (1968) equation, the proposed equation will perform better in
distributed, the pore size density f(ψi) will equal 1 / N. Substituting the prediction of a permeability function from a bimodal SWCC.

Fig. 7. Illustration of the distribution of pores in both sections.


154 Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156

Fig. 8. Illustration of possible scenarios for connections between the two sections.

4. Applications and discussions insignificant because the R2 of log(k(ψ)) only improved from 0.48 to
0.51. The SWCC data for Touchet silt loam did not cover a sufficient suc-
Due to the limited data with SWCC in the form of degree of satura- tion range up to the residual suction and, as a result, unreasonable
tion and hydraulic conductivity available in the literature, only four fitting parameters and residual suction may have been obtained from
sets of data were used for verification of the proposed equation. These the best fit procedure. Consequently, the accuracy of the predicted re-
four data sets are for volcanic sand, glass beads, fine sand and touch sults for the permeability function decreased.
silt loam (Brooks and Corey 1964). The unsaturated hydraulic conduc- Experimental data for Yolo light clay from Moore (1939), for which
tivity for sandy material decreases much faster than the conductivity there is only a SWCC in the form of gravimetric water content, were
for clayey soil with an increase in matric suction. The unsaturated hy- used to examine Kunze et al.'s (1968) equation. The SWCC for Yolo
draulic conductivity within the low suction range for sandy soil is light clay is illustrated in Fig. 13. As the SWCC in the form of gravimetric
more important than the conductivity in the high suction range. Most water content is not really analogous to the pore-suction distribution
of the soils used for verification in this paper were sandy soils. There- function, the calculated results using Kunze et al.'s (1968) equation
fore, the results obtained using the proposed model could only be com- were in weak agreement with the experimental data, as illustrated in
pared to experimental data within a suction range of 0.1 kPa to 30 kPa. Fig. 14. Leong and Rahardjo (1997) presented the similar comparison
The SWCCs (in degree of saturation) for the four types of soil are il- results as Fig. 14.
lustrated in Fig. 10. The measured and calculated unsaturated hydraulic Fredlund and Houston (2013) presented and discussed the effect of
conductivity at a maximum suction of 106 kPa for the four soils are illus- soil volume change on SWCC measures for oil sands tailings and Regina
trated in Fig. 11. Fig. 11 indicates that the results calculated using the clay, showing that the void ratio decreased with an increase in matric
proposed equation agree with the measured data from Brooks and suction. In other words, the difference between the degree of saturation,
Corey (1964). However, the calculated results do not fit well with the S1, by considering soil volume change and S2, by ignoring the soil vol-
experimental data. The proposed equation was derived from Poiseuille's ume change, increased with an increase in matric suction. As no data
equation, which is applicable for stream-line flow. If the suction in- for soil volume change were recorded by Moore (1939), the SWCC illus-
creases beyond the residual suction, the water phase in the soil becomes trated in Fig. 13 was re-calculated by assuming a volume change which
discontinuous and Poiseuille's equation may no longer be applicable. results in a slower rate of decrease in saturation as shown in Fig. 15. The
Therefore, the residual suction ψr is suggested as the maximum suction difference between S1 and S2 increased gradually with an increase in
adopted in the proposed equation for calculation of the permeability matric suction. The calculated results and the experimental data are il-
function of unsaturated soil. lustrated in Fig. 16. Fig. 16 indicates that the results computed using
The calculated relative hydraulic conductivity using a maximum the proposed equation with an assumed soil volume change agreed
suction of ψr is illustrated in Fig. 12. The fitting parameters and SWCC with the experimental data. However, the accuracy of the results is de-
variables, such as air-entry value (AEV) and residual suction ψr, of the pendent on the assumption of the soil volume change used in the calcula-
four types of soil were determined using Zhai and Rahardjo (2012a); tion. Figs. 14 and 16 suggest that soil volume change should be monitored
Zhai and Rahardjo (2013a) equations, as given in Table 1. and recorded during SWCC measurement.
Fig. 12 indicates that the calculated results for volcanic sand, glass
beads and fine sand significantly improved because the calculated coef- 5. Conclusions
ficient of determination (R2) of log(k(ψ)) for volcanic sand, glass beads
and fine sand improved from 0.54, 0.56 and 0.71 to 0.98, 0.72 and 0.86, A new equation for the prediction of permeability function from
respectively, when ψr was adopted as the maximum suction in the pro- SWCC in the form of degree of saturation using capillary model was pro-
posed equation. However, improvement for Touchet silt loam was posed in this paper. Water flow in soil was derived from the theory of

Fig. 9. Illustration of the probability of connections of saturated pores at a suction of ψm.


Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156 155

Table 1
Fitting parameters and SWCC variables for four types of soil.

Fitting parameters SWCC variables


Types of soil
a (kPa) n m AEV (kPa) ψr (kPa)

Volcanic sand 1.88 9.44 0.67 1.66 3.40


Glass beads 3.03 28.97 0.82 0.86 5.39
Fine sand 4.23 8.34 0.84 3.65 7.40
Touchet silt loam 8.30 9.64 0.45 7.40 16.52

Fig. 10. SWCCs of four types of soil (from Brooks and Corey 1964).

Fig. 13. SWCC in the form of gravimetric water content for Yolo light clay (Moore 1939).

an electronic spreadsheet with input parameters (e.g., fitting parame-


ters a, n and m for Fredlund and Xing's (1994) equation). The soil vol-
ume change was also incorporated into the proposed equation. The
derivation showed that the proposed equation was a general form of
Marshall's (1958) and Kunze et al.'s (1968) equations. As a direct divi-
sion of matric suction was adopted in the proposed equation, more
data could be captured between the first residual suction and the sec-
Fig. 11. Comparison of calculated and measured permeability functions for four types of
soil (from Brooks and Corey 1964) using a maximum suction of 106 kPa. ond air-entry value for the bimodal SWCC. Thus, the proposed equation
will have a better performance for the prediction of the permeability
function from a bimodal SWCC.
random connections and Poiseuille's law. The unsaturated hydraulic
conductivity was correlated with the fitting parameters of the SWCC.
The unsaturated hydraulic conductivity could then be computed using

Fig. 12. Calculated results using the proposed equation and a maximum suction of ψr for Fig. 14. Comparison of calculated results using Kunze et al.'s (1968) equation and experimen-
four types of soil. tal data for Yolo light clay (Moore 1939).
156 Q. Zhai, H. Rahardjo / Engineering Geology 199 (2015) 148–156

Fig. 15. Illustration of SWCC for Yolo light clay assuming a certain soil volume change.
Fig. 16. Comparison of calculated results using the proposed equation and experimental
data for Yolo light clay.

The water phase in the soil may become discontinuous when the Fredlund, D.G., Xing, A., Huang, S., 1994. Predicting the permeability function for unsatu-
rated soils using the soil–water characteristic curve. Can. Geotech. J. 31, 533–546.
suction increases beyond the residual suction, and so Poiseuille's law
Hu, R., Chen, Y.F., Liu, H.H., Zhou, C.B., 2013. A water retention curve and unsaturated
for stream-line flow might not be applicable. As the statistical model hydraulic conductivity model for deformable soils: consideration of the change in
was derived based on Poiseuille's law, it might not work for suctions pore-size distribution. Geotechnique 63, 1389–1405.
Kunze, R.J., Uehara, G., Graham, K., 1968. Factors important in the calculation of hydraulic
exceeding the residual suction. Thus, it is suggested that the residual
conductivity. Proc. Soil Sci. Soc. Am. 32, 760–765.
suction be adopted in the statistical model for calculation of the perme- Leong, E.C., Rahardjo, H., 1997. Permeability function for unsaturated soils. J. Geotech.
ability function of unsaturated soil. As SWCC in the form of degree of Geoenviron. Eng. 123 (12), 1118–1126.
saturation (instead of other forms, e.g., volumetric water content or Marshall, T.J., 1958. A relation between permeability and size distribution of pores. J. Soil
Sci. 9, 1–8.
gravimetric water content) is analogous to the pore-suction distribution Moore, R.E., 1939. Water conduction from shallow water tables. Hilgardia 12, 383–426.
function, soil volume change should be monitored during conventional Mualem, Y., 1976. A new model for predicting the hydraulic conductivity of unsaturated
SWCC measurement. porous media. Water Resour. Res. 12, 513–522.
Mualem, Y., 1986. Hydraulic conductivity of unsaturated soils: prediction and formulas.
In: Klute, A.A. (Ed.), Methods of Soil Analysis. Part 1, Physical and Mineralogical
Methods, 2nd ed. Am Soc of Agronomy, Inc., and Soil Sci. Soc. of Am., Inc., Madison,
References Wis., pp. 799–823.
Rahardjo, H., Satyanaga Nio, A., Leong, E.C., Ng, Y.S., 2010. Effects of groundwater table po-
Beckett, C., Augarde, C.E., 2013. Prediction of soil water retention properties using pore-
sition and soil properties on stability of slope during rainfall. ASCE J. Geotech.
size distribution and porosity. Can. Geotech. J. 50, 435–450.
Geoenviron. 136 (11), 1555–1564 (November).
Brooks, R.H., Corey, A.T., 1964. Hydraulic properties of porous medium. Hydrology Paper
Romero, E., Gens, A., Lloret, A., 1999. Water permeability, water retention and microstruc-
no.3. Civ Engrg. Dept., Colorado State Univ., Fort Collins. Colo.
ture of unsaturated Boom clay. Eng. Geol. 54, 117–127.
Childs, E.C., Collis-George, N., 1950. The permeability of porous materials. Proc. R. Soc.
Romero, E., Della Vecchia, G., Jommi, C., 2011. An insight into the water retention proper-
A201, 392–405.
ties of compacted clayey soils. Geotechnique 61 (4), 313–328.
Casini, F., Vaunat, J., Romero, E., Desideri, A., 2012. Consequences on water retention prop-
Romero, E., 2013. A microstructural insight into compacted clayey soils and their hydrau-
erties of double-porosity features in a compacted silt. Acta Geotech. 7, 139–150.
lic properties. Eng. Geol. 165, 3–19.
Cosentini, R.M., Della Vecchia, G., Foti, S., Musso, G., 2012. Estimation of the hydraulic pa-
Ye, W.M., Wan, M., Chen, B., Chen, Y.G., Cui, Y.J., Wang, J., 2012. Temperature effects on the
rameters of unsaturated samples by electrical resistivity tomography. Geotechnique
unsaturated permeability of the densely compacted GMZ01 bentonite under con-
62 (7), 583–594.
fined conditions. Eng. Geol. 126, 1–7.
Cuisinier, O., Masrouri, F., Stoltz, G., Russo, G., 2014. Multi-scale analysis of the swelling
Zhai, Q., Rahardjo, H., 2012a. Determination of soil–water characteristic curve variables.
and shrinkage of a lime-treated expansive clayey soil. Unsaturated Soils: Research
Comput. Geotech. 42, 37–43.
& Application Proceedings of the Sixth International Conference on Unsaturated
Zhai, Q., Rahardjo, H., 2012b. Reply to the discussion by Bellia et al. On “determination of
Soils UNSAT 2014 Sydney, Australia 2–4 July, pp. 441–447.
soil–water characteristic curve variables” by Zhai Q, Rahardjo H. ComputGeotech
Della Vecchia, G., Dieudonne, A.C., Jommi, C., Charlier, R., 2015. Accounting for evolving
2012;42:37–43. Comput. Geotech. 45, 151–152.
pore size distribution in water retention models for compacted clays. Int. J. Numer.
Zhai, Q., Rahardjo, H., 2013a. Soil–water characteristic curve variables. Conference Paper,
Anal. Methods Geomech. 39 (7), 702–723.
Proceeding of the International Symposium on Unsaturated Soil Mechanics and Deep
Fredlund, D.G., Houston, S.L., 2013. Interpretation of soil-water characteristic curves when
Geological Waste Disposal, Shanghai, China, 07–10 July 2013.
volume change occurs as soil suction is changed. In: Caicedo, et al. (Eds.), Advances in
Zhai, Q., Rahardjo, H., 2013b. Quantification of uncertainty in soil-water characteristic
Unsaturated Soils. Taylor & Francis group, London (© 2013). (ISBN 978–0-415-
curve associated with fitting parameters. Eng. Geol. 163, 144–152.
62095-6).
Zhai, Q., Rahardjo, H., 2014. Variability in soil–water characteristic curve associated with
Fredlund, D.G., Rahardjo, H., 1993. Soil Mechanics for Unsaturated Soils. Wiley, New York.
fitting parameters. Conference paper, Unsaturated Soils: Research & Application
Fredlund, D.G., Rahardjo, H., Fredlund, M.D., 2012. Unsaturated Soil Mechanics in Engi-
Proceedings of the sixth international conference on unsaturated soils UNSAT 2014
neering Practice. Wiley, New York.
Sydney, Australia 2–4 July, pp. 441–447.
Fredlund, D.G., Pham, H.Q., 2006. A volume-mass constitutive model for unsaturated soil
Zhou, W.H., Yuen, K.V., Tan, F., 2014. Estimation of soil–water characteristic curve and
in terms of two independent stress state variables. Keynote Address, 4th Internation-
relative permeability for granular soils with different initial dry densities. Eng. Geol.
al Conference on Unsaturated Soils. ASCE, Carefree, Arizona, pp. 105–134 (April 2–6).
179, 1–9.
Fredlund, D.G., Xing, A., 1994. Equations for the soil–water characteristic curve. Can.
Geotech. J. 31 (3), 521–532.

Вам также может понравиться