Water vapor enhancement factor due to temperature gradient...
Transcript of Water vapor enhancement factor due to temperature gradient...
明治大学大学院農学研究科
2016 年度
博士学位請求論文
Water vapor enhancement factor due to temperature
gradient in unsaturated soils
不飽和土壌中の温度勾配による水蒸気促進係数
学位請求者 農学専攻
サニー・ゴー・エン・ギアップ
Ph.D. Dissertation
March 2017
Water vapor enhancement factor due to temperature
gradient in unsaturated soils
Sunny Goh Eng Giap
Agriculture Program
Graduate School of Agriculture
Meiji University
i
TABLE OF CONTENTS
List of Figures ................................................................................................................................ iv
List of Tables ................................................................................................................................. vii
Chapter 1. General introduction ................................................................................................. 1
Chapter 2. Sensitivity analysis and validation for numerical simulation of water infiltration into unsaturated soil ........................................................................................ 5
2.1 Abstract ............................................................................................................................ 5
2.2 Introduction ...................................................................................................................... 6
2.3 Materials and methods ..................................................................................................... 9
2.3.1 The governing equation of water flow in unsaturated soil,
and its numerical solution ........................................................................................................ 9
2.3.2 Constitutive functions of matric pressure head ( m ) and
hydraulic conductivity ( K ) ................................................................................................... 10
2.3.3 Numerical experiment and the default setting of input
parameters of the flow problem ............................................................................................. 11
2.3.4 Statistical measures ................................................................................................. 11
2.4 Results and discussion .................................................................................................... 11
2.4.1 Simulation results and its accuracy ......................................................................... 11
2.4.2 Sensitivity analysis and simulation model validation ............................................. 12
2.5 Conclusions .................................................................................................................... 15
Chapter 3. Liquid water, water vapor, and heat flow simulation in unsaturated soils ............................................................................................................................ 24
3.1 Abstract .......................................................................................................................... 24
3.2 Introduction .................................................................................................................... 25
3.3 Materials and methods ................................................................................................... 28
3.3.1 The governing equation of water, vapor, and heat transport
in unsaturated soil, and its numerical solution ...................................................................... 28
3.3.2 Input parameters, initial and boundary conditions .................................................. 32
3.3.3 Statistical measure for parameter calibration .......................................................... 33
3.4 Results and discussion .................................................................................................... 34
3.4.1 The comparison of numerical simulation and Heitman data .................................. 34
ii
3.4.2 The uncertainty of hydraulic conductivity and vapor
enhancement factor ................................................................................................................ 36
3.4.3 Limitation and possible improvement .................................................................... 39
3.5 Conclusions .................................................................................................................... 41
Chapter 4. An improved heat flux theory and mathematical equation to estimate water vapor volume expansion-advection as an alternative to mechanistic enhancement factor ................................................................................................... 47
4.1 Abstract .......................................................................................................................... 47
4.2 Introduction .................................................................................................................... 48
4.3 Materials and methods ................................................................................................... 50
4.3.1 Cass experimental setup, parameter estimation, relation to
the present study and limitation ............................................................................................. 50
4.3.2 Theory of heat flux.................................................................................................. 52
4.3.3 The mass balance relation of liquid water and water vapor,
and its effect on water vapor volume expansion-advection .................................................. 55
4.3.4 The presence of volume expansion-advection from the
perspective of ideal gas law ................................................................................................... 56
4.3.5 The presence of air and water vapor volume expansion-
advection................................................................................................................................ 57
4.3.6 Water vapor volume expansion-advection equation ............................................... 58
4.3.7 The relation of mechanistic enhancement factor and water
vapor volume expansion-advection ....................................................................................... 59
4.4 Results and Discussion ................................................................................................... 61
4.4.1 The general trend of thermal conductivities ........................................................... 61
4.4.2 The wet porous media thermal conductivity by heat
conduction, and the sensible heat thermal conductivity by liquid water
movement .............................................................................................................................. 62
4.4.3 The thermal conductivity of water vapor diffusion and
volume expansion-advection ................................................................................................. 63
4.4.4 The water vapor thermal conductivity by latent heat and
sensible heat........................................................................................................................... 64
4.4.5 The water vapor thermal conductivity of the partial
derivative of matric pressure head effects on relative humidity with
respect to temperature ............................................................................................................ 65
iii
4.4.6 Water vapor permeability variables with respect to soil
moisture content and air pressure .......................................................................................... 65
4.4.7 Limitation of previous work, and the possibility of
improvement .......................................................................................................................... 67
4.5 Conclusions .................................................................................................................... 68
Chapter 5. The extended Darcy law and water vapor relative permeability derivation in unsaturated soils ................................................................................. 75
5.1 Abstract .......................................................................................................................... 75
5.2 Introduction .................................................................................................................... 75
5.3 Materials and methods ................................................................................................... 78
5.3.1 The derivation of extended Darcy law, and its relative
permeability and the Klinkenberg effect ............................................................................... 78
5.3.2 The integral form of relative permeability of liquid water
and water vapor ..................................................................................................................... 82
5.3.3 The thermal conductivity governed by Fick’s law and Darcy
law in heat flux equation ....................................................................................................... 87
5.4 Results and Discussion ................................................................................................... 89
5.4.1 Replacing Fick’s law with Darcy law ..................................................................... 89
5.4.2 The water vapor relative permeability, and the Klinkenberg
permeability ........................................................................................................................... 91
5.4.3 Using Darcy law in the thermal water vapor diffusivities ...................................... 92
5.4.4 Limitation, and potential area of improvement....................................................... 93
5.5 Conclusions .................................................................................................................... 94
Chapter 6. Summary and Conclusions ................................................................................... 102
References ................................................................................................................................... 107
iv
List of Figures
Figure 2.1. Comparison of simulated results with Philip's semi-analytical solution. Note: Philip(H) and Philip(K) were from Haverkamp et al. (1977) and Kabala and Milly (1990), respectively. ................................................................. 17
Figure 2.2. The rank of sensitivity coefficient. Note: θ_s and θ_r are the saturated and residual volumetric water content; ∆z, spatial spacing size; ∆t, time-step size; K_s, saturated hydraulic conductivity; θ_L(initial cond.), clay medium initial value of volumetric water content; θ_L(upper bound.), upper boundary of volumetric water content; A, B, β and α are the fitting parameters from Haverkamp, as in Eqs. 2.10 and 2.11. ................................................................................................................ 18
Figure 2.3. The cumulative effect of input parameters, on the absolute residual error at simulation time 105 s. Note: θ_L(initial cond.) (step 2), clay medium initial value of volumetric water content; θ_r (step 3), residual volumetric water content; β (step 4), α (step 5), A (step 7) and B (step 8) are the fitting parameters; K_s (step 6), saturated hydraulic conductivity; ∆t (step 9), time-step size; and ∆z (step 10), spatial spacing size. .......................................................................................................... 19
Figure 2.4. The effect of ∆z (step 10 alone), and cumulative effects of steps 2 to 9 and 2-10 in comparison with Philip(H) and Philip(K).......................................... 20
Figure 3.1. Silt loam (sil-20) simulated: (a) water content and (c) temperature distributions at temperature boundary of 15/-50, 15/-150 and 30/-150, which were 17.5-12.5, 22.5-7.5oC and 37.5-22.5oC, respectively. The (b) water content and (d) temperature distribution at reverse temperature boundary of 22.5/100, i.e. 17.5-27.5oC, after 12, 48 and 96 hours. Note: the initial water content set at 0.20 m3 m-3. The
63.8 10sK m s-1 and zeta � = �� used in the simulation. .............................. 43
Figure 3.2. Sand (s-8) simulated: (a) water content and (c) temperature distributions at temperature boundary of 15/-50, 15/-150 and 30/-150, which were 17.5-12.5, 22.5-7.5oC and 37.5-22.5oC, respectively. The (b) water content and (d) temperature distribution at reverse temperature boundary of 22.5/100, i.e. 17.5-27.5oC, after 12, 48 and 96 hours.
Note: the initial water content set at 0.08 m3 m-3. The 69.8 10sK
m s-1 and zeta � = � used in the simulation.......................................................... 44
Figure 4.1. Water vapor thermal conductivity, and the combined (wet porous media heat conduction and sensible heat by liquid water movement) thermal
conductivity (at 0r oP P P ) versus soil volumetric water content.
Note: the water vapor thermal conductivity was at a pressure ratio of 1.0, i.e. atmospheric pressure, and temperature at 32.5 oC. The water
v
vapor thermal conductivity was corrected for the pressure dependence of the thermal conductivity of air, based on the equation given by Cass. ...................................................................................................................... 71
Figure 4.2. The curve-fitting results of permeability variables as a function of the soil volumetric water content, and the pressure ratio. The permeability variables based on the adapted-improved heat flux theory for (a) the lysimeter sand, and (b) the Portneuf silt loam. Note: the graph (a) and (b) were subjected to Eq. 4.20, before curve-fitting using Eq. 4.22 to determine the permeability variables. Zero permeability variables values were added as supposed zero water vapor volume expansion-advection at fully saturated soil moisture content to improve the curve-fitting results. .............................................................................................. 72
Figure 4.3. The comparison of the experimental and curve-fitted permeability variables based on the adapted-improved heat flux theory for (a) the lysimeter sand, and (b) the Portneuf silt loam. Note: the graph (a) and (b) were subjected to Eq. 4.20, before curve-fitting using Eq. 4.22 to determine the permeability variables. Care should be placed on the fully saturated soil moisture content condition because the permeability variables estimated by Eq. 4.22 could result in negative values. ................................................................................................................... 73
Figure 4.4. (a) The relative permeability ( rk ) of lysimeter sand versus water
saturation estimated using Eq. 4.24, and (b) Enlarged graph on relative permeability value ranged 0-0.3. Note: the other relative permeability models were van Genuchten-Mualem, van Genuchten-Burdine, Corey, and Falta referring to Eqs. 4.25-4.28, respectively. .................... 74
Figure 5.1. Schematic diagram of velocity profile due to fluid flow through a small tube diameter in the porous media. ....................................................................... 96
Figure 5.2. The water vapor thermal conductivity of lysimeter sand versus water saturation at (a) 32.5oC, (b) 22.5oC, (c) Portneuf silt loam versus saturation at 32.5oC, and (d) Kanto loamy soil versus saturation at temperature ranged from 38.5 to 31.5oC. Note: the water vapor thermal conductivity estimated by Darcy law was obtained by subtracting the water vapor thermal conductivity predicted by Fick’s law from the total water vapor thermal conductivity. Equation 5.33 was used in the determination. The total water vapor thermal conductivity dataset derived from Cass et al. (1984) at different pressure ratios, temperatures, and volumetric water contents. ............................. 97
Figure 5.3. The permeability variables ( rnw P Tk k ) versus water saturation at the unity
of pressure ratio and 32.5 degree Celsius. ............................................................ 98
Figure 5.4. The (a) Portneuf silt loam, (b) lysimeter sand, and (c) Kanto loamy soil
permeability variables ( rnw P T ok k k ) data estimated by Eq. 5.34 at
different water saturations, temperatures, and pressure ratios. Note:
vi
the data were curve-fitted by the rnwk and P Tk from Eqs. 5.30 and
5.14, respectively. The equation with and without the temperature as
an independent variable in the expression of P Tk would correspond
to the model e rS T P and e rS P , respectively. From the first to
fourth wave-like curves were respectively the pressure ratios of 1, 0.53, 0.16 and 0.083, as in graph (a), and the each point of the rise and fall of the permeability variables of the wave-like curve, from left to right, correspond to the increasing water saturation. The experimental temperature was 32.5 degree Celsius. In graph (b), the first three wave-like curves correspond to temperatures of 3.5, 22.5 and 32.5 degree Celsius at a pressure ratio of unity. The following three wave-like curves correspond to the similar temperature range, but at the pressure ratio of 0.53, and it continued until the fourth (or last) three wave-like curves at a pressure ratio of 0.083. Graph (c) represented by first to fifth wave-like curves with respective pressure ratios of 1.000, 0.705, 0.495, 0.308 and 0.203. The first wave-like curve permeability variables, from left to right, corresponding to decreasing temperature (42.5 to 31.6 degree Celsius) and increasing water saturation. The rest of the wave-like curves correspond to a similar range of temperature and saturation conditions. The data on graphs (a) and (b) were derived from Cass et al. (1984), while graph (c) was from Tanimoto (2007) derived using DPHP method. .............................................................................. 100
vii
List of Tables
Table 2.1 The coefficients value from Haverkamp et al. (1977) based on the constitutive Eqs. 2.10 and 2.11. These values were used as a base case. ....................................................................................................................... 21
Table 2.2 Statistical measures (μ, μ*, σ) of elementary effect method. They were the mean of elementary effect, the mean of absolute values of the elementary effect, and the standard deviation, respectively. ................................. 22
Table 2.3 Significant digits approximation on input parameter value. ......................................... 23
Table 3.1 Input parameters used for silt loam and sand. ............................................................... 45
Table 3.2 The comparison of water content and temperature root mean square errors (RMSEs) from Heitman et al. (2008) and current simulation, at few temperature boundary conditions. ......................................................................... 46
Table 5.1 The parameter constants ( n and m ) and the Klinkenberg gas slip factor ( b ) of water vapor relative permeability. ........................................................... 101
1
Chapter 1. General introduction
The vapor enhancement factor (VEF) is a well-known but inadequately defined parameter in soil
physics. It was first considered by Philip and de Vries (1957) when they introduced Fick’s law
equation to govern water vapor diffusion in unsaturated soil. But Fick’s law underestimated the
water vapor flux. So, a coefficient, namely, the VEF multiplied the water vapor diffusion
equation to match the discrepancy. Philip and de Vries (1957) hypothesized the VEF as the
combined effects of liquid island and the ratio of the air-filled pores to the overall temperature
gradients. The VEF has been applied in various studies of liquid water, water vapor and heat
flow in unsaturated soil (Banimahd and Zand-Parsa, 2013, Heitman et al., 2008, Milly, 1982,
Milly, 1984, Milly, 1980, Moghadas et al., 2013, Nassar and Horton, 1992, Nassar and Horton,
1997, Noborio et al., 1996, Novak, 2016, Saito et al., 2006, Sakai et al., 2009, Scanlon and Milly,
1994), but its fundamental mechanism is subject to speculation. Shokri et al. (2009) claimed that
the use of capillary flow would negate the need for VEF. Parlange et al. (1998) and Cahill and
Parlange (1998) hypothesized that air volume expansion-contraction convective transport by
diurnal heating and cooling on soil surface is responsible for the VEF. Similarly, Scanlon and
Milly (1994) stated water vapor transport is influenced by thermally driven water vapor
advection, and further added that the great air fluidity in maintaining the relatively constant
atmospheric pressure and in the event of wind gust, it could lead to a net effect of air flow in and
out that transports water vapor. Nevertheless, Ho and Webb (1999) carried out a thorough
investigation on VEF and ultimately concluded its presence. The investigation by Sakaguchi et
al. (2009) and Lu et al. (2011) further confirmed the presence of VEF. However, VEF values
varied from 0.35 in a hydrophobic to 3.5 in a wettable soil, and not even pore-scale models could
2
describe these observations (Horton and Ochsner, 2011). Also, VEF varies in different soil types,
moisture saturation, and ambient temperature (Cass et al., 1984). Clearly, further investigation is
needed to reveal the mechanism controlling VEF.
In this study, the investigation focused on the VEF application and its origin. The VEF
was investigated in two case study situations. The first case study involved the application of
VEF in a simulation of liquid water, water vapor, and heat transport in unsaturated soil. The
second case study estimated to the VEF value indirectly through the use of heat flux theory,
water vapor diffusion based on Fick’s law, and thermal conductivity which was determined by
heat balance equation.
The first case required solving the isothermal liquid water flow equation (Richards, 1931)
and validating the simulated result by Philip’s semi-analytical solution (Chapter 2), before
extending the simulation of liquid water flow to include the effect of water vapor and heat flow
simulation (Chapter 3) in unsaturated soil. A similar validation procedure was employed by
Noborio et al. (1996). The approach set forth necessarily confirmed the mass (water and vapor)
and heat flow simulation developed from a verified solution of liquid water flow simulation. The
Heitman et al. (2008) study has created a controlled experimental temperature boundaries, which
was different from other studies, e.g. Saito et al. (2006) and Sakai et al. (2009). The former study
created an environmental condition successfully isolated from any possible uncertainty encounter
at the boundaries, especially those stated by Scanlon and Milly (1994). Thus, in the current
study, the experimental dataset from Heitman et al. (2008) was re-examined, and the calibration
procedure was improved from using single to using multiple temperature boundaries to
determine whether the VEF, hydraulic conductivity, or their combined effects were responsible
for the discrepancy between the simulation and the physical observation. Furthermore, the
3
outcome of the study would then be able to address the questions of whether or not the
simulation deteriorated over time, whether the VEF could adequately reproduce the water
content distribution after a reverse temperature boundary, whether the calibrated parameters
(VEF and saturated hydraulic conductivity) with the least error in simulation varied when a soil
was simulated at different initial water contents, and whether the search for the least error
simulation generated different calibrated parameters depending on either prioritization on
temperature or water content distributions.
The second case study is based on the work of Cass et al. (1984). The VEF originated as
a correction factor imposed on Fick’s law equation, which itself a constitution of the water vapor
latent heat. Water vapor latent heat combined with wet porous media heat conduction are
assumed as the only two thermal conductivity components used in the simplified heat flux
equation. Cass et al. (1984) published the thermal conductivity values and physical properties of
two soils utilized in the estimation of the VEF. Often, improvement simply identified from the
assumptions used in establishing the theory. Thus, in the current study, similar to the former first
case study, the dataset was re-examined using the heat flux theory from de Vries (1958). The
thermal conductivity contributed by liquid water flow was also considered (Shokri et al., 2009).
Also, the fact that Fick’s law needs the VEF as a correction factor implies an incomplete
phenomenon discovery and so there is an opportunity for improvement. Ho and Webb (1996) and
Ho and Webb (1999) attempted to consider water vapor flux using the Darcy law equation, but
were unsuccessful. In this study, the phase transition volume expansion from liquid water to
water vapor was examined using a mass balance equation. The ideal gas law applied on water
vapor was theoretically investigated under incremental temperature conditions, and the water
vapor advection by air volume expansion concept used in Parlange et al. (1998) was revisited
4
(Chapter 4). The limitation revealed in Chapter 4 was addressed in Chapter 5.
5
Chapter 2. Sensitivity analysis and validation for numerical simulation of
water infiltration into unsaturated soil1
2.1 Abstract
This chapter describes the validation of a FORTRAN code developed for liquid water flow in
unsaturated soil under isothermal conditions to simulate water infiltration into Yolo light clay.
Successful validation of the source code was necessary because the code could be extended to
program the equation governing water vapor and heat flux in unsaturated soil. In this study, the
liquid water flow governing equation, which is Richards’ equation, was approximated by a finite-
difference method. A normalized sensitivity coefficient was used in the sensitivity analysis of
Richards’ equation. The normalized sensitivity coefficient was calculated using one-at-a-time
method (OAT) and elementary effect method based on constitutive equations of matric suction
and hydraulic conductivity. Results from the elementary effect method provide additional insight
into model input parameters, such as input parameter linearity and oscillating sign effect.
Boundary volumetric water content (��(����� �����. )) and saturated volumetric water content
(��) were consistently found to be the most sensitive parameters corresponding to positive and
negative relations, as given by the hydraulic functions. In addition, although initial volumetric
water content (��(������� ����. ) ) and time-step size (∆� ), respectively, possessed a great
amount of sensitivity coefficient and uncertainty value, they did not exhibit significant influence
on the model output as demonstrated by spatial spacing size (∆z). The input multiplication of
parameters sensitivity coefficient and uncertainty value was found to affect the outcome of the
model simulation, in which parameter with the highest value was found to be ∆z. This finding
1 This chapter has been published in Goh and Noborio (2015).
6
would imply that the extension of the water flow equation, that is Richards’ equation, to include
water vapor and heat flux, as in Nassar and Horton (1997) and Heitman et al. (2008), would
require a small spatial spacing size to guarantee a good simulation result. Furthermore, validating
the water infiltration source code was used as a necessary step in Noborio et al. (1996) before
modeling two-dimensional model for water, heat, and solute transport in furrow-irrigated soil.
2.2 Introduction
Sensitivity analysis is used for various reasons, such as decision making or development of
recommendations, communication, increasing understanding or quantification of the system, and
model development. In model development, it can be used for model validation or accuracy,
simplification, calibration, coping with poor or missing data, and even to identify important
parameter for further studies (Pannell, 1997).
More than a dozen sensitivity analysis methods are available, ranging from one-at-a-time
(OAT) to variance-based methods (Campolongo et al., 1999, Davis et al., 1997, De Roo and
Offermans, 1995, Fox et al., 2010, Hamby, 1994, Saltelli et al., 2004, Sepúlveda et al., 2014,
Vazquez-Cruz et al., 2014, Vereecken et al., 1990). On a fundamental level, sensitivity analysis is
a tool to assess the effect of changes in input parameter value on the output value of a simulation
model. In this aspect, the sensitivity coefficient, in a normalized form, is given in the following
relationship:
��,� =���� ���⁄
��� ��⁄ (2.1)
where ��,� is referred to as the normalized sensitivity coefficient for �th input parameter at �th
observation point; ��� is the model dependent variable value at �th observation point; and �� is
the �th input parameter value. However, the method does not explore other input space factors
7
in which more than one input parameter is varied. Despite this disadvantage, Saltelli and Annoni
(2010) noted that researchers continual use of the OAT method is due to a few claimed
advantages. Although variance-based method is the best practice, Saltelli and Annoni (2010)
have suggested using the elementary effect method, which is an enhancement of the OAT
method, when computation time is expensive, for instance, in numerical simulation that is
computationally demanding.
The elementary effect method is accomplished through the use of a technical scheme to
generate trajectories. Each trajectory consists of some steps in which each step refers to an
increment or decrement of an input parameter value. The base condition for each trajectory is
different from the others, and it is selected randomly. The random version of trajectory
generation is as following (Morris, 1991, Saltelli et al., 2008, Saltelli et al., 2004):
�∗ = �����,��∗ + (∆ 2⁄ )��2� − ����,���
∗ + ����,����∗ (2.2)
where �∗ is the generated trajectory in the form of a matrix with dimension (� + 1) × �, where
� is the number of independent input parameters; ∆ is the value in [1 (� − 1)⁄ , … ,1 −
1 (� − 1)⁄ ] and � is the number of levels; ����,� is (� + 1) × � matrix of 1's; �∗ is a randomly
chosen base value; � is the low triangular matrix of 1's; � ∗ is the �-dimensional diagonal matrix
in which each element is either +1 or -1, by random generation; and �∗ is the �-by-� random
permutation matrix that each row and column with only one element equal to 1. This method has
been used by some researchers, for instance, Drouet et al. (2011) on nitrous oxide emissions at
farm level, and Chu-Agor et al. (2011) on the vulnerability of coastal habitats to sea level rise.
Campolongo et al. (2011) improved this method into a quantitative approach.
The generated trajectories can be screened to obtain a subset of trajectories with the
greatest geometric distances. The trajectories scanned for maximum geometric distance between
8
all the pairs of points between two trajectories, were determined as follows (Campolongo et al.,
2007):
��� = �∑ ∑ �∑ �����(�) (� ) − ����
(�) (�)��
����
������
������ � ≠ �
0
(2.3)
where ��� is the distance between a pair of trajectories � and �; ����(�) (�) is the �th coordinate of
the �th point of the �th trajectory; and ����(�) (� ) is the �th coordinate of the �th point of the � th
trajectory.
The sensitivity coefficient of an input parameter in the elementary effect method as ��,
which is the mean of the elementary effect (����). ��
∗ is the mean of absolute values of the
elementary effect, which is to avoid cancellation of difference signs in the mean value. The
sensitivity measures (��, ��∗, �) and ���
� are given by (Saltelli et al., 2008):
����=
��(���∆�)���(��)
∆� (2.4)
�� =�
�∑ ���
����� (2.5)
��∗ =
�
�∑ ����
���
��� (2.6)
��� =
�
���∑ ����
�− ���
����� (2.7)
where ��(��) and ��(�� + ∆�) are the simulation result before and after increment or decrement
of ∆ value, i.e. ∆� which can either positive or negative value; � is referring to the total number of
trajectories; ����
is the elementary effect of � input parameter at � trajectory; and �� is the
standard deviation of � input parameter.
The aim of the current work was to carry out sensitivity analyses on water infiltration into
unsaturated soil as governed by Richards’ equation and to use it as an assessment method to
9
validate the simulation source code with the analytical solution. Thus, the objective of this study
was to use the sensitivity coefficient, under a hypothetical assumption to validate the model
simulation with Philip's semi-analytical solution using literature data. The water infiltration
results from Haverkamp et al. (1977) and Kabala and Milly (1990) were used for the verification.
2.3 Materials and methods
2.3.1 The governing equation of water flow in unsaturated soil, and its numerical solution
The governing equation for transient liquid water flow in soil may be described as (Celia et al.,
1990, Richards, 1931):
���
��=
�
�����
���
�������
��− ���⃑ � (2.8)
where �� is the volumetric water content (m3 m-3); � is the simulation time (s); � indicates the
vertical distance in the simulation (m); � is the hydraulic conductivity of the medium (m s-1); ��
is the matric pressure head (m); ��⃑ is the vector unit with a value of positive one when it is
vertically downwards.
Equation 2.8 was approximated numerically, and its algebra was implemented in
FORTRAN 2008 using Simply Fortran Integrated Development Environment. The spatial
discretization method used was termed as the cell-centered finite difference (Zheng and Bennett,
2002). To avoid unnecessary redundancy, we only provide the algebra for Eq. 2.8, that was used
for sensitivity analysis in the current study, as follows:
��(�)������(�)
�
∆�=
���
��
�������
���
��
∆��(�.�∆������.�∆��)���(���)
��� − ��(�)����
−
���
��
�������
���
��
∆��(�.�∆����.�∆����)���(�)
��� − ��(���)���� −
��,�,��
��
��⃑ ���,�,��
��
��⃑
∆�� (2.9)
10
where � indicates the cell-centered number in z-direction in Cartesian coordinate system; ∆� (s)
is the time-step size; ��(�)� (m3 m-3) and ��(�)
��� (m3 m-3) indicate volumetric water contents at
initial time (n) and new time (n+1), respectively; ���� �⁄ (m s-1) is the hydraulic conductivity at
the interface between cell � and � + 1 ; ���� �⁄ (m s-1) is the hydraulic conductivity at the
interface between cell � − 1 and �; (��� ���⁄ )��� �⁄ is the partial derivative of �� with respect
to �� at the interface between the cell � and � + 1; (��� ���⁄ )��� �⁄ is the partial derivative of
�� with respect to �� at the interface between the cell � − 1 and �; ∆���� (m), ∆�� (m) and
∆���� (m) are corresponding to the spatial sizes of spacing of cell � + 1 , � and � − 1 .
��(���)��� (m3 m-3), ��(�)
��� (m3 m-3) and ��(���)��� (m3 m-3) are the volumetric water
contents at new time level of cell � + 1, � and � − 1, respectively. An iterative method was used
to solve the mathematical algebra of Eq. 2.9 (Tu et al., 2007). A convergence factor criterion was
used to indicate the condition for the iteration termination process, that is the absolute maximum
difference ���(�)��� − ��(�)
�� for every single cell.
2.3.2 Constitutive functions of matric pressure head ( m ) and hydraulic conductivity ( K )
The constitutive functions implemented were from Haverkamp et al. (1977):
�� = −10����� ��(�����)
�����− ��
�
� (2.10)
� = ���
��(������)� (2.11)
where � , � , � and � are the fitted parameters; �� (m3 m-3) is the residual volumetric water
content; �� (m3 m-3) is the saturated volumetric water content; and �� (m s-1) is the saturated
hydraulic conductivity.
11
2.3.3 Numerical experiment and the default setting of input parameters of the flow
problem
Water infiltration into Yolo light clay was used in the numerical experiment. The coefficients
values for the medium (Table 2.1) were based on the constitutive function equations (Eqs. 2.10
and 2.11). The initial condition for the volumetric water content was 0.2376 m3 m-3. Lower
boundary was set to permeable inflow and outflow of water, and the upper boundary was set at
0.495 m3 m-3. After considering the mass balance ratio (Celia et al., 1990) and iteration number,
the time-step size, spatial spacing size, and convergent value were set at corresponding values of
500 s, 1 cm, and 10-12 m3 m-3, respectively.
2.3.4 Statistical measures
A statistical equation was used to determine the goodness of fit between the data and the
simulated results. The equation was known as absolute residual errors (MA) as follows (Zheng
and Bennett, 2002):
�� =�
�∑ |���� − ����|���� (2.12)
where ���� is the simulated data at cell �, and ���� is the analytical solution at cell �.
2.4 Results and discussion
2.4.1 Simulation results and its accuracy
Based on the conditions as stated in the previous section, water infiltration into Yolo light clay
was simulated for up to 105 s. Philip's semi-analytical solution data were collected from
Haverkamp et al. (1977) and Kabala and Milly (1990), hereafter referred to as Philip(H) and
Philip(K), respectively. Simulation results were compared with the data to verify the simulation
12
(Fig. 2.1). Before referring to any statistical measure, it was evident that the simulation results
slightly under-estimated the water flow infiltration front.
Also, Fig. 2.1 shows that there was a small discrepancy between Philip(K) and Philip(H),
but the former was relatively closer to the simulation results than the latter. Thus, a sensitivity
analysis was carried out to determine the sensitivity coefficient for all input parameters. The
sensitivity analysis results were used to assess the model simulation based on the assumption that
possibly the cumulative effect of input parameters, regarding significant digits approximation,
could be contributing the under-estimation of the volumetric water content of the simulation. The
sensitivity analysis was one of the most important steps in evaluating the effect of input
parameter on simulation results. Moreover, other researchers used sensitivity analysis for model
validation (Gosling and Arnell, 2011, Min et al., 2006, Nathan et al., 2001, Stange et al., 2000).
2.4.2 Sensitivity analysis and simulation model validation
Negligible sensitivity response could be the result of too small a perturbation in size, and
inaccuracy in sensitivity response could be due to too large a perturbation in size (Poeter and
Hill, 1998). The input parameters values were subjected to a perturbation in size of between -5%
and 5%, as suggested by Zheng and Bennett (2002). Seven input parameters, as listed in Table
2.1, and four additional input parameters were tested, i.e. initial volumetric water content
(��(������� ����. )), boundary volumetric water content (��(����� �����. )), time-step size
(∆�), and spatial spacing size (∆�), were tested. A depth of 15.5 cm from the ground surface was
used for observation.
The normalized sensitivity coefficients are shown in Fig. 2.2. There were two groups of
sensitivity coefficients, i.e. positive and negative relationships. In the positive relationship group,
13
the boundary volumetric water content had the highest sensitivity coefficient. This was followed
by initial volumetric water content and saturated hydraulic conductivity. The smallest sensitivity
coefficient in the group was the residual volumetric water content. In negative relationship
group, saturated volumetric water content had the highest sensitivity coefficient, and in this
group spatial spacing size and time-step size had the smallest sensitivity coefficients.
For comparison purposes, the elementary effect method was also used to calculate
normalized sensitivity coefficient. Only random generation in �-dimensional diagonal matrix
(� ∗), and Eq. 2.2 was used to generate 50 trajectories. Equation 2.3 was used to screen for four
selected trajectories. The mean of elementary effect was used to calculate the normalized
sensitivity coefficient (Table 2.2). Similar values of � and �∗ indicate linear effects on a few
input parameters in the positive (�, ��, ��(������� ����. )) and negative (��, �) relationships.
Other input parameters have shown the effect of oscillating sign which result in different values
of � and �∗. Also, both sensitivity analysis methods generated comparable trends of sensitivity
coefficients.
We assumed that a minor deviation in each input parameter, regarding its significant
digits approximation, could contribute some effect to the simulation outcome that could explain
the gap between the simulated results and Philip's semi-analytical solution (Fig. 2.1). In other
words, the parameter values used in the computer simulation by Haverkamp et al. (1977) could
have been different from the exact data that they published. Thus, the sensitivity analysis positive
and negative relationships could be used to determine the hypothetical approximation values in
Table 2.3 for investigation. The cumulative effect begins from step 1 for the base case to step 10
for spatial spacing size. The ��(������� ����. ) value (0.2376499 m3 m-3) was used as a second
simulation (in step 2) after the base case simulation. A third simulation (in step 3) was applied
14
using �� value as 0.124499 m3 m-3 by keeping the ��(������� ����. ) value from the second
simulation. In each simulation, the absolute residual error (MA) (Eq. 2.12) was calculated for the
discrepancy between the simulation and Philip's semi-analytical solution. Of the eleven
parameters in Table 2.3, only ∆� and ∆� were parameters without any limit of variation. Thus,
they were reduced by 98% and 90% from 500 s and 1 cm to 10 s and 0.1 cm, respectively. The ��
and ��(����� �����. ) were not included in the study because decreasing and increasing the
former and latter value would result in simulation failure.
A consistent reduction in MA value from ��(������� ����. ) to � input parameter was
observed; and a steep slide was observed on the ∆z input parameter simulation (Fig. 2.3). The
reducing ∆� value should lead to a reduction in MA value, but an incremental simulated result
was observed. This may be explained from an elementary effect perspective in that ∆� has
different values of � and �∗, which indicate the capability of sign oscillation (Table 2.3).
Figure 2.4 showed that the simulation of the cumulative effect of steps 2-9, which
combined the effect from ��(������� ����. ) (step 2) to ∆� (step 9), did not contribute a
significant effect to the advancement of the water infiltration front; it only resulted in a reduction
of 7.8% in MA value, from 0.0254 to 0.02343 m3 m-3. In addition, these eight input parameters
would have to vary in the way as tabulated in Table 2.3 in order to cause the stated percentage
reduction. Therefore, the significant digits approximations might not be the main cause of the
problem in considering the greater effect of ∆� on the advance of the water infiltration front as
shown in Fig. 2.4. A further step to include ∆� in the simulation, that is, the cumulative effect of
steps 2-10, which combined the effect from ��(������� ����. ) (step 2) to ∆� (step 10), found a
54.7% reduction in MA value of step 9, from 0.02343 to 0.0106 m3 m-3. This indicates that
spatial spacing size was the main cause of the advance of the water infiltration front. Therefore,
15
the simulation was repeated for a last time to assess the effect of spatial spacing size, Step 10
alone, (Fig. 2.4), showing a good agreement between the simulation results and the Philip(K).
This observation could be explained using Eq. 2.1, after rearranging it into the following form,
which we termed it as percentage variation in simulation results:
∆���
���=
∆��
��× ��,� (2.13)
where ∆�� ��⁄ is the normalized input parameter value (%); ∆��� ���⁄ is the normalized output
parameter value (%); and ��,� is the normalized sensitivity coefficient (%/%). Equation 2.13 was
simply a multiplication of the percentage change in input parameter value from the base case and
the normalized sensitivity coefficient.
From Eq. 2.13, the percentage variation in simulation results from the ∆� and ∆� input
parameters caused an increment of 4.95% and 0.06%, respectively, despite ∆� having the highest
reduction in percentage (-98%) from the base case. This observation can be summarized as
follows: (1) the input parameter with the highest sensitivity coefficient does not imply the
greatest effect on the simulation result, for example, ��(������� ����. ) ; and (2) the input
parameter with the highest percentage of change also does not imply the greatest effect on the
simulation result, for example, ∆� . Therefore, only the parameter that produces the highest
multiplication of sensitivity coefficient, and greatest percentage change on the input parameter
(or the uncertainty) would cause the most substantial effect on the simulation result.
2.5 Conclusions
The equation governing transient water flow in unsaturated, under isothermal condition, was
approximated numerically by finite-difference solution. It was implemented in FORTRAN
programming language, simulated and verified by Philip's semi-analytical solution on water
16
infiltration into Yolo light clay, using data from the literature.
One-at-a-time (OAT) and elementary effect methods (EE) were used in the sensitivity
analysis. A common trend of sensitivity was observed across the methods in both positive and
negative relationships. The latter method allowed exploration of additional characteristics of
input parameters at different input spaces, such as linearity and sign oscillation effect. The sign
oscillation effect observed on input parameters explained the possibility of its deviation from
those observed in the OAT method at different input spaces, for example, time-step size.
A hypothetical case that was established to study the cumulative effect of input
parameters on the discrepancy between the simulated result and Philip's semi-analytical solution,
regarding significant digits approximation (from base case), was found to be unlikely.
Surprisingly, neither the high normalized sensitivity coefficient of initial volumetric water
content nor the largest percentage changes in time-step size, contributed any substantial impact
on simulation results when compared to spatial spacing size, alone. This observation leads to the
conclusion that the uncertainty of the input parameters and normalized sensitivity coefficients of
the input parameters both control the outcome of the simulation.
Furthermore, the finding of this study indicates the simulation code of Richards’ equation
was indeed working properly, and it could be extended to include water vapor and heat flux
governing equations. The final developed simulation code would have to use a small spatial
spacing size to generate accurate simulation output.
17
Figure 2.1. Comparison of simulated results with Philip's semi-analytical solution. Note:
Philip(H) and Philip(K) were from Haverkamp et al. (1977) and Kabala and Milly (1990),
respectively.
0.0
5.0
10.0
15.0
20.0
25.0
30.0
0.20 0.25 0.30 0.35 0.40 0.45 0.50
De
pth
, z (
cm)
Volumetric water content (m3 m-3)
Simulated (10^5 s) Philip(H) @ 10^5s Philip(K) @ 10^5s
18
Figure 2.2. The rank of sensitivity coefficient. Note: θ_s and θ_r are the saturated and residual
volumetric water content; ∆z, spatial spacing size; ∆t, time-step size; K_s, saturated hydraulic
conductivity; θ_L(initial cond.), clay medium initial value of volumetric water content;
θ_L(upper bound.), upper boundary of volumetric water content; A, B, β and α are the fitting
parameters from Haverkamp, as in Eqs. 2.10 and 2.11.
-4.9
7E+
00
-1.0
7E+
00
-5.8
9E-
01
-5.5
0E-
02
-6.1
9E-
04
3.8
5E-
02
8.2
5E-
02
2.0
8E-
01
5.0
8E-
01
6.4
4E-
01
4.7
1E+
00
-11.00
-9.00
-7.00
-5.00
-3.00
-1.00
1.00
3.00
5.00
7.00
9.00
11.00
θ_s B β ∆
z
∆t
θ_r α A
K_s
θ_L
(in
itia
l co
nd
.)
θ_L
(u
pp
er b
ou
nd
.)
No
rmal
ized
sen
siti
vity
co
effc
ien
t (%
/%)
Input parameter
19
Figure 2.3. The cumulative effect of input parameters, on the absolute residual error at
simulation time 105 s. Note: θ_L(initial cond.) (step 2), clay medium initial value of volumetric
water content; θ_r (step 3), residual volumetric water content; β (step 4), α (step 5), A (step 7)
and B (step 8) are the fitting parameters; K_s (step 6), saturated hydraulic conductivity; ∆t (step
9), time-step size; and ∆z (step 10), spatial spacing size.
0.0
25
4
0.0
25
3
0.0
25
3
0.0
24
5
0.0
24
5
0.0
24
5
0.0
24
5
0.0
23
38
0.0
23
43
0.0
10
6
0.010
0.015
0.020
0.025
0.030
0.035
Bas
e C
ase
(Ste
p 1
)
θ_L
(in
itia
l co
nd
.) (
Step
2
) θ_r
(St
ep 3
)
β (
Step
4)
α (
Step
5)
K_s
(St
ep 6
)
A (
Step
7)
B (
Ste
p 8
)
∆t
(Ste
p 9
)
∆z
(Ste
p 1
0)
Ab
solu
te r
esid
ual
err
or
(MA
), (
m3
m-3
)
Input parameter
20
Figure 2.4. The effect of ∆z (step 10 alone), and cumulative effects of steps 2 to 9 and 2-10 in
comparison with Philip(H) and Philip(K).
16
17
18
19
20
21
22
23
0.27 0.29 0.31 0.33 0.35 0.37 0.39
De
pth
, z (
cm)
Volumetric water content (m3 m-3)
Philip(H) Base Case Steps 2-9
Steps 2-10 Philip(K) Only Step 10
21
Table 2.1 The coefficient values from Haverkamp et al. (1977) based on the constitutive Eqs.
2.10 and 2.11. These values were used as a base case.
Parameter Value
� 739
�� 0.124 m3 m-3
�� 0.495 m3 m-3
� 4
� 124.6
� 1.77
�� 1.23x10-7 m s-1
Note that �� is the residual volumetric water content, �� is the saturated volumetric water
content, �� is the saturated hydraulic conductivity, and �, �, � and � are the fitting coefficients.
22
Table 2.2 Statistical measures (μ, μ*, σ) of elementary effect method. They were the mean of
elementary effect, the mean of absolute values of the elementary effect, and the standard
deviation, respectively.
μ (%/%) μ* (%/%) σ
�� -6.03E+00 6.03E+00 9.48E-01
� -1.85E+00 1.85E+00 9.38E-01
� -2.07E-01 3.20E-01 3.70E-01
� -4.14E-02 1.25E-01 1.47E-01
∆� -3.35E-02 3.95E-02 4.56E-02
∆� -1.66E-04 5.25E-04 6.21E-04
�� 4.44E-03 3.37E-02 4.02E-02
� 3.13E-01 3.13E-01 7.41E-02
�� 5.24E-01 5.24E-01 6.73E-02
��(������� ����. ) 8.84E-01 8.84E-01 3.05E-01
Note: �� is the residual volumetric water content, �� is the saturated volumetric water content, ��
is the saturated hydraulic conductivity, ∆� is the spatial spacing size, ∆� is the time-step size,
��(������� ����) is the initial value of volumetric water content, and � , � , � and � are the
fitting coefficients.
23
Table 2.3 Significant digits approximation on input parameter value.
Parameter Value
� 739.499 (≈739)
�� 0.124499 (≈0.124) m3 m-3
�� 0.495 m3 m-3
� 3.95 (≈4)
� 124.64 (≈124.6)
� 1.765 (≈1.77)
�� 4.4284×10-2 (≈4.428x10-2) cm hr-1
��(������� ����. ) 0.2376499 (≈0.2376) m3 m-3
��(����� �����. ) 0.495 m3 m-3
∆� 10 s, the base case was 500 s
∆� 0.1 cm, the base case was 1 cm
Note: �� is the residual volumetric water content, �� is the saturated volumetric water content, ��
is the saturated hydraulic conductivity, ∆� is the spatial spacing size, ∆� is the time-step size,
��(������� ����) is the initial value of volumetric water content, ��(����� ����) is the upper
boundary of volumetric water content, and �, �, � and � are the fitting coefficients.
24
Chapter 3. Liquid water, water vapor, and heat flow simulation in
unsaturated soils
3.1 Abstract
Liquid water, water vapor, and heat movement in soil are important for agriculture and
engineering applications. Successfully in simulating the mass and heat transport of soil water and
temperature distributions could be used for land management strategy, and also to reduce the cost
of continuous instrumental measurement. It inspired researcher to program the equation to
simulate the observations. Heitman et al. (2008) attempted to describe the observation after
parameter (��, �) calibration using a single temperature boundary condition out of nine available
boundary conditions, in which each condition has a duration of 96 hrs. Thus, the calibrated
parameters were not optimized. In the current study, the water, vapor, and heat flux equation was
solved by finite difference method and coded to re-examine the experimental data from Heitman
et al. (2008). The calibrated parameters were obtained after taking into account all the imposed
temperature boundary conditions. The results showed that saturated hydraulic conductivity (��)
at its original value produced low error, i.e. root mean square error (RMSE). The zeta (�) value, a
variable that represents vapor enhancement factor, has an uncertain value between 2 and 6, which
corresponds to accuracy prioritization on either temperature or water content, respectively. The
model has potential for further improvement in the inaccuracy noted in the low- and high-
temperature edges of the soil column by implementing the equation from Fayer and Simmons
(1995). The vapor enhancement factor requires further theoretical examination and to
experimentally measure the value, otherwise, it continues as an adjustable parameter without real
physical representation. Furthermore, the reverse temperature boundary simulation caused a
25
significant error revealing the necessity of a new guiding force, for example, vapor buoyancy.
3.2 Introduction
The movement of water, vapor, and heat has important applications in agriculture and
engineering (Saito et al., 2006). In the agricultural sector, the variation of temperature and water
content could influence biological activities, seed germination and plant growth (Banimahd and
Zand-Parsa, 2013). Seed germination can occur with only water supply in the form of water
vapor (Wuest et al., 1999). Also, heat, water, and vapor fluxes have an influence on soil
salinization (Gran et al., 2011). In engineering applications, the coupling of heat, water, and
vapor simulation is necessary for landfill management, nuclear waste repositories, and the
improved detection of buried land mines (Moghadas et al., 2013). Kroener et al. (2014) used the
coupling simulation to study heat dissipation of underground electric power cables, which could
lead to a better design of cable installation. The coupling simulation could also be used for water
balance prediction to estimate potential leachate generation as part of landfill management
(Khire et al., 1997). Furthermore, the coupling simulation has been used to study airflow
pathway development that could be applied to predict vapor intrusion into building basement
(Sakaki et al., 2013).
In soil physics, the governing equation for water, vapor, and heat transport was initially
derived by Philip and de Vries (1957) and de Vries (1958), hereafter known as PdV. The PdV
equation transformed into a matric suction-based equation by Sophocleous (1979). Based on the
work of previous authors and also by Milly (1980) and Milly (1982), the PdV equation was
extended by Nassar and Horton (1997) to include a solute transport mechanism. The extended
PdV form was also used to study furrow-irrigated soil (Noborio et al., 1996). Also, the PdV
26
equation was adapted to simulate coupling of water, vapor, and heat in unsaturated freezing soils
(Hansson et al., 2004, Zhang et al., 2016). Sakai et al. (2009) used the equation to study the
effect of water vapor condensation and evaporation on temperature and water content
distribution in a sandy column, while Kelleners et al. (2016) included a snow water flow
equation in addition to the water, vapor, and heat transport. The PdV equation has also been used
to couple with global climate model prediction to study dry area (Pfletschinger et al., 2014). The
use of PdV equation in diverse applications such as wood drying, concrete drying and fracturing,
sand-water-steam systems, soil during wildfires, and wet porous thermal barriers, was reported in
Massman (2015). Also, Moghadas et al. (2013) integrated the PdV model with ground-
penetrating radar model to improve estimation of soil hydraulic properties. Dong et al. (2016)
attempted to estimate the soil water content, temperature, thermal and hydraulic properties of the
PdV equation by assimilating soil temperatures.
Although the PdV equation has been subjected to continuous improvement over decades,
there is a fundamental problem regarding the model prediction. The problem arises from the
uncertainty of the model parameters. Rose (1968) described the possibility air flux as a possible
phenomenon to explain the vapor enhancement factor due to a temperature gradient, as noted by
Bittelli et al. (2008). The air flux possibility was then investigated by Parlange et al. (1998) and
Cahill and Parlange (1998), and the air volume expansion and contraction by diurnal temperature
variation were claimed to be responsible for additional vapor flux. Or and Wraith (2000)
questioned the findings’ validity because diurnal temperature variation could affect the measured
water content due to time domain reflectometry. Later, Novak (2016) found the PdV equation
without the effect of airflow was sufficient to describe field measurements. Besides, Scanlon and
Milly (1994) noticed other types of uncertainty, such as that the saturated hydraulic conductivity
27
value should be lower in value, air movement resulting from atmospheric pressure fluctuation
due to a wind gust, and the neglected hysteresis in the water retention function.
Apart from all these limitations, Sakai et al. (2009) found a good match between PdV
simulation and experimental observation. The study unveiled an inseparable inverse relationship
between the pore-connectivity coefficient from unsaturated hydraulic conductivity and the zeta
coefficient from vapor enhancement factor. The relationship highlighted the necessity of a low
value of pore-connectivity coefficient at high zeta coefficient, and vice versa. Such calibration
was not an isolated case, as Heitman et al. (2008) found that calibration of the zeta value and the
saturated hydraulic conductivity was required to match the simulation with observation.
The research work by Heitman et al. (2008) originated back in the early work by Nassar
and Horton (1989), when using the destructive (gravimetric) method to determine water content
and the failure of a one-dimensionally designed soil column resulted in two-dimensional
temperature distribution. Subsequently, Prunty and Horton (1994) managed to achieve a one-
dimensional temperature, i.e. free from the ambient temperature interference (ATI), in a 5 cm
column length, but it was too short in length to allow a measuring probe, e.g. TDR, placement in
between. Then, Prunty (2003) designed a soil column length of 10 cm, but was unable to remove
ATI. Later, Zhou et al. (2006) improved the design up to a 10 cm length soil column without
noticeable ATI. Using the improved soil column design, together with the in situ measurement
solution from Ren et al. (2005) for water content and soil thermal property, Heitman et al. (2008)
modeled water, vapor, and heat transport in the closed soil column.
This study re-examines the simulation results from Heitman et al. (2008). The
experimental works by Heitman and his co-workers were relatively thorough, but the model
calibration effort was limited only to a single case of boundary condition, even though multiple
28
sequentially applied boundary conditions, i.e. nine temperature gradients, were available.
Finding optimized parameters would require multiple boundary conditions to be simultaneously
accounted in the model calibration. Thus, the objective of the current study is to calibrate the
model with multiple boundary conditions simultaneously so that either the vapor enhancement
factor, hydraulic conductivity, or the combined effect could be single out as the main culprit for
the uncertainty in simulation output. This work built on the previous chapter with the simulation
code of Richards equation for liquid flow and gravity being extended to include vapor and heat
transport in unsaturated soil. The simulation would implement fine spatial discretization size to
generate reliable simulation results (Goh and Noborio, 2015), as found in the previous chapter.
3.3 Materials and methods
3.3.1 The governing equation of water, vapor, and heat transport in unsaturated soil, and
its numerical solution
The study used the water content based equation, as in the original PdV equation, to govern
water and vapor fluxes, which can be described as:
�1 +��
��
���
���
���
���−
��
������
��+
��
������
��+
���
���
���
�����
��
=�
���(��� + ���)
��
��+ ���� + �
���
�������
��− ���⃑ � (3.1)
��� = �����
������
��+
���
���
���
��� (3.2)
��� =����
������
���
���
���� (3.3)
��� = ����
�� (3.4)
where Eqs. 3.2, 3.3 and 3.4 are constitutive functions of Eq. 3.1. � is the water vapor
enhancement factor (dimensionless), � is the vapor diffusion coefficient (m2 s-1), Ω is the
29
tortuosity factor (dimensionless), �� is matric suction (m), �� is the volumetric air content (m3
m-3), �� is the liquid water density (kg m-3), �� is the water vapor density (kg m-3), �� is the
volumetric water content (m3 m-3), � is the time (s), � is the temperature (K), � is the vertical
distance (m), ��� is the thermal vapor diffusivity (m2 s-1 K-1), ��� is the thermal liquid
diffusivity (m2 s-1 K-1), ��� is the isothermal vapor diffusivity (m s-1), � is the hydraulic
conductivity (m s-1), and ��⃑ is the vector unit with a value of positive one when it is vertically
downwards. Equation 3.1 was adapted from Heitman et al. (2008), which was originally in
matric suction based form. Although some researchers such as Milly (1982) have expressed the
necessity of using the matric suction based form equation, the water content based equation used
in the current study to examine a soil water content that was far below saturation level, so it
would not be an issue of concern (Heitman et al., 2008).
The heat flux equation was also adapted from Heitman et al. (2008), which originated
from de Vries (1958). The equation can be described as:
{��}��
��+ {��}
���
��=
�
���{�+ ��}
��
��+ {��}
���
��− ��(� − ��)�����⃑ � (3.5)
�� = �� + [(� − ��)�� + ��]�� ����
��+
���
���
���
��� (3.6)
�� = ��(���� + ����) + ������ + ������ (3.7)
�� = (� − ��)(���� − ����) − ���� − ��� + [(� − ��)�� + ��]�����
���
���
��� (3.8)
�� = [�� − (�� − ��)(� − ��)]����� + ��(� − ��)�����
+ ��(� − ��)������
��� (3.9)
� = �∗ + ������ = �� + ���� + �����.� (3.10)
�� = ��(� − ��)����� + ��(� − ��)����� (3.11)
where Eqs. 3.6, 3.8, 3.9, 3.10 and 3.11 are constitutive functions of Eq. 3.5, while Eq. 3.7 is part
30
of Eq. 3.6. �� is the bulk density (kg m-3), �� = 1900 and �� = 730 are the respective organic
and mineral specific heats (J kg-1 K-1), �� and �� are the respective organic and mineral mass
fractions (dimensionless), �� is the reference temperature (273.15 K), �� is the liquid water
specific heat (J kg-1 K-1), �� is the water vapor specific heat at constant pressure (J kg-1 K-1), �� is
the latent heat of vaporization (J kg-1) at the reference temperature, � is the differential heat of
wetting (J kg-1), �∗ is the thermal conductivity of the porous media in the presence of liquid
water (W m-1 K-1), � is the latent heat of vaporization (J kg-1) at the temperature �, and ��, ��
and �� are the fitting parameters.
The water vapor density was derived from the equations of saturated vapor density
(Kimball et al., 1976), matric head effect on relative humidity (Philip and de Vries, 1957), and
water retention curve (Campbell, 1974):
�� = 10����� �19.819−����.�
�+
����
�����
�����
���[−�(� − ��)]� (3.12)
where � is the gravitational constant (9.81 m s-2), � is the water molecular weight (0.01801528
kg mol-1), � is the ideal gas constant (8.3145 J mol-1 K-1), �� is the saturated water content (m3
m-3), �� and � are the fitting parameters of water retention curve, and � is a constant (2.09 x 10-3
K-1).
Taking the derivative of Eq. 3.12 on matric suction gives:
���
���= 10����� �19.819−
����.�
����
����� �
��
���� �
��
�����
���[−�(� − ��)]� (3.13)
The water retention curve relation from Campbell (1974) can be taken as the derivative of
the matric suction on water content, together with the matric head effect on relative humidity
from Philip and de Vries (1957):
���
���= −
���
�����
�������
���[−�(� − ��)] (3.14)
31
Again, Eq. 3.12 can be taken as the derivative of the water vapor density on temperature
to give:
���
��= �
����.�
��−
����
����10����� �
����
��� ��� �19.819−
����.�
�� (3.15)
Based on the similar derivative procedure as in Eq. 3.14, but this time the derivative of
matric suction with respect to temperature:
���
��= −��� �
��
�����
���[−�(� − ��)] (3.16)
The vapor diffusion coefficient (Kimball et al., 1976), tortuosity factor (Milly, 1984),
unsaturated hydraulic conductivity (Bear et al., 1991, Campbell, 1974), and vapor enhancement
factor (Cass et al., 1984) respectively, are expressed as shown below:
� = 2.29× 10�� ��
���.����.��
(3.17)
Ω = (� − ��)�
� (3.18)
� = �� ���
������� �(��)
�(�) (3.19)
� = � + ��� − (� − 1)��� �−��1 + � √�⁄ ������ (3.20)
where � is the soil porosity (m3 m-3), �� is the saturated hydraulic conductivity (m s-1), �(��) is
the dynamic viscosity at reference temperature (kg m-1 s-1), �(�) is the dynamic viscosity at
temperature �, zeta (�), � and � are respectively 8, 6 and 4, � is the soil particle density (2.6 Mg
m-3), and � is the clay fraction of soil.
Numerical approximation was applied to Eq. 3.1, and the algebra implemented in
FORTRAN 2008. The cell-centered finite difference was used as the spatial discretization
method. The algebra for water and vapor transport (Eq. 3.1) is:
�1 +��(�)
��(�)����
���������
�����−
��(�)
��(�)��
∆����(�)
��� − ��(�)��
32
+ ��(�)
��(�)��
���
����+ �
���
���������
�������
�������
∆�
=��
����������
����������
����������
���
�.�∆������.�∆������
����������
����������
����������
���
�.�∆����.�∆�����
∆��
+
������
�����
����
�������
���
��
∆��(�.�∆������.�∆��)���(���)
��� − ��(�)����
−
������
�����
����
�������
���
��
∆��(�.�∆����.�∆����)���(�)
��� − ��(���)���� −
���
��
��⃑ ����
��
��⃑
∆�� (3.21)
where ∆� is the time-step size (s), ∆� is the spatial discretization size (m), � is the middle cell,
� − 1 is the cell located spatially before cell �, � + 1 is the cell after cell �, � −�
� is the interface
between cells � − 1 and �, � +�
� is the interface between cells � and � + 1, � is the initial time
level, and � + 1 is the new time level.
Similar discretization procedure implemented on the heat flux equation of Eq. 3.5:
��(�)
�����
��� − ���� + ��(�)
��(�)������(�)
�
��
=���
��
������
���
���(�.��������.����)�����
��� − ������ −
���
��
������
���
���(�.������.������)���
��� − ��������
+�����
�
��
��(���)������(�)
���
���(�.��������.����)− �
�����
��
��(�)������(���)
���
���(�.������.������)
−�����
�����
����
������������������
��
��⃑
���+
�����
�����
����
������������������
��
��⃑
��� (3.22)
3.3.2 Input parameters, initial and boundary conditions
All data in this section were retrieved from Heitman et al. (2008). In this study, sand and silt
loam water content (��) and temperature (�) distribution profiles in the vertically closed soil
column were simulated. There were four separate sets of simulation, they were sil-10 (silt loam
at 0.10 m3 m-3), sil-20 (silt loam at 0.20 m3 m-3), s-8 (sand at 0.08 m3 m-3), and s-18 (sand at 0.18
m3 m-3). Ten transient temperature boundary conditions, were imposed on the vertical closed soil
column, they were 17.5-12.5, 20-10, 22.5-7.5, 25-20, 27.5-17.5, 30-15, 32.5-27.5, 35-25, 37.5-
33
22.5, 17.5-27.5oC, in each case corresponding to the top (left) and bottom (right) temperature
boundaries in the 10 cm soil column. The respective boundary conditions in the form of mean
temperature (oC)/temperature gradient (oC m-1) are 15.0/-50, 15.0/-100, 15.0/-150, 22.5/-50,
22.5/-100, 22.5/-150, 30.0/-50, 30.0/-100, 30.0/-150, and 22.5/100. The boundary conditions of
32.5-27.5oC were simulated, but not included in the calibration as in Heitman et al. (2008). Each
transient temperature boundary simulation lasted for 96 hours. The last boundary condition
(17.5-27.5oC) was also held for 96 hours but temperature gradient was reversed, and both �� and
� were recorded at 12, 24, 48 and 96 hours. Hence, there were eight transient boundary
conditions with eight recorded pairs of �� and �, together with another four recorded pairs in the
last boundary condition, to give a total of twelve pairs for each simulation set, e.g. sil-10. The
input parameters are listed in Table 3.1. The study assumed uniform water content and
temperature as in the initial condition. The 10 cm vertical column was spatially discretized into
200 cells, and 100s was used as the time-step size. Nine different temperature boundaries were
equally divided over 864 hours, together with the reverse temperature gradient for 96 hours, to
give a total simulation of the exactly 40 days.
3.3.3 Statistical measure for parameter calibration
Heitman and his co-workers conducted a single boundary condition model calibration for each
soil measured, and then used the single calibrated model to simulate other boundary conditions.
It is noted that the root mean square error (RMSE) used in their study to estimate the error of
simulation from the experiment, was a RMSE at a single boundary condition of the soil, and then
they re-calculated the RMSE again at a few additional boundary conditions. For instance, on silt
loam at 0.20 m3 m-3 water content, the RMSE was used to calibrate parameters at 22.5/-100 and
34
then the calibrated parameters were applied to forecast and re-calculate the RMSE at the
additional boundary conditions of 15/-50 and 22.5/-150. A similar procedure used in calibrating
the silt loam (0.10 m3 m-3) and the sand (0.08 and 0.18 m3 m-3). It should be noted that there were
nine boundary conditions, i.e. twelve pairs of �� and � data, available for each soil type at
specific water content, thus, the calibration approach used by Heitman and his co-workers could
not adequately optimize the parameters. As stated in section 3.2, the simultaneous calibration of
multiple boundary conditions could single out the influence of either the vapor enhancement
factor or hydraulic conductivity on the discrepancy observed between experimental and
simulated results. To allow comparison with the Heitman et al. (2008) simulation, the root mean
square error (RMSE) equation was used as the statistical measure for parameter calibration:
���� = �∑ (���������)��
���
� (3.23)
where ���� is the simulated data at cell � , ���� is the experimental data at cell � , � is the
number of data.
3.4 Results and discussion
3.4.1 Comparison of numerical simulation and Heitman data
At the beginning, the soil water content was homogeneous throughout the vertical soil column.
Then, as soon as the simulation started, the first imposed temperature boundary (17.5-12.5oC)
resulted in the movement of water, vapor, and heat fluxes. The temperature gradient led to mass
and heat fluxes as described by Kroener et al. (2014). The high-temperature region produced
high water vapor concentrations and a flux towards the low-temperature region for water vapor
condensation. The water vapor flux carried a significant amount of heat flux that was necessary
to reduce the temperature in the high-temperature region. As the supply of liquid water from the
35
neighboring low-temperature region declined due to the reduced unsaturated hydraulic
conductivity by declining matric suction (low water content), the soil temperature in the warmer
region started to increase in response to low water content available for vaporization to remove
heat in the form of vapor. The counter balance behavior between temperature gradient and water
content gradient was similarly noted in other studies, for example, Sakai et al. (2009). This
explained the observation of low water content in the high-temperature region and high water
content in the low-temperature region, as in Figs. 3.1 and 3.2.
As noted earlier in section 3.2, the required calibrations were the hydraulic conductivity
and vapor enhancement factor. The saturated hydraulic conductivity (��), Eq. 3.19, and the zeta
(�) coefficient of vapor enhancement factor, Eq. 3.20, were calibrated, as in Heitman et al.
(2008). From now on, it is referred as Heitman. The simulation calibration has to compromise
between the accuracy of water content and temperature, as was also noted in Massman (2015).
The calibrations in Figs. 3.1 and 3.2 were based on the best combination of �� and � to achieve
the lowest root mean square error (RMSE) of water content, at the expense of temperature
accuracy if necessary. For instance, to achieve the lowest temperature RMSE for sil-20, the �
value would have to reduce to 4. A reasonably good match between simulation and Heitman data
for sil-20 was observed for the first eight temperature boundary conditions (Figs. 3.1(a) and
3.1(c) as forward boundary), but a greater discrepancy was found in Figs. 3.1(b) and 3.1(d) after
the reverse boundary condition was initiated. The forward boundary simulation of s-8 water
content (Fig. 3.2(a)) was not as good as for sil-20 (Fig. 3.1(a)), while the reverse boundary
simulation had similar characteristics to that of sil-20. Also, the sil-10 and s-18 could not
adequately reproduce low and high water contents at the close edge regions of high- and low-
temperature, respectively.
36
In the search for the lowest water content and temperature RMSE, a large number of
possible combinations of �� and � were investigated. The outcome of the simulations could be
used to suggest the most probable value for �� and � that would satisfy each soil type at different
water contents was discussed in the following section.
3.4.2 The uncertainty of hydraulic conductivity and vapor enhancement factor
Silt loam at an initial water content of 0.20 m3 m-3 was first simulated for parameter calibration
by Heitman at a single temperature boundary, i.e. 22.5/-100. The calibrated parameters (�� =
0.38 μm s-1, � = 4) were claimed to produce a low RMSE of water content and temperature, and
a reasonably good match with the experimental data. Then, the parameters were used to simulate
at other temperature boundaries, e.g. sil-20 at 15.0/-50 and 22.5/-150, and sil-10 at 22.5/-50 and
22.5/-150. The calculated RMSE for each boundary is shown in Table 3.2. Similarly, parameter
calibration for sand at 0.08 m3 m-3 was performed, initially, by Heitman at the 22.5/-100
temperature boundary. The parameters (�� = 9.8 μm s-1, � = 4) determined from the calibration
were used for simulations at s-8 (15.5/-50, 22.5/-150) and s-18 (22.5/-50, 22.5/-150) boundaries.
In the current study, following the same procedure, simulations for three of out the five
boundaries of silt loam were found to generate lower �� and � RMSE values, i.e. better accuracy,
than those estimated by Heitman (Table 3.2). Similarly, in the current study, simulations for three
out of the five boundary conditions for sand also generated lower RMSE than that found by
Heitman. Only the �� and � RMSE values for the five boundaries were reported by Heitman, but
our findings, at least, indicate the simulation code was performing properly. Note that for each
soil type and water content, e.g. sil-20, there were twelve pairs of �� and � data available for
calibration, as stated in section 3.3.2. The sil-20 calibration was based on three temperature
37
boundaries, while sil-10, s-8 and s-18 were calibrated with two, three and two temperature
boundaries, respectively. Clearly, some other temperature boundaries were not adequately
considered, for instance, the reverse temperature boundary.
After considering all the temperature boundaries, the best fitted �� simulation produced
different calibrated parameters (��, �) for sil-10 (3.8 μm s-1, 16), sil-20 (3.8 μm s-1, 12), s-8 (9.8
μm s-1, 6) and s-18 (5.0 μm s-1, 16). The calibrated parameters could be different when the
priority of best fitting was prioritized towards accurate � fitting. For instance, the sil-20 lowest �
RMSE was found at � = 4 , while maintaining the �� . Moreover, more uncertainty of the
parameters could be found when a selected individual temperature boundary was used as fitting
criterion. For example, sil-20 (15.0/-50) at �� = 3.8 μm s-1 and � = 16 produced 0.013 m3 m-3 of
�� RMSE that was lower than the one shown in Table 3.2. Therefore, it may not be the best
approach to focus on a specific temperature boundary to obtain the best fitting for either �� or �,
and then to further assume it as a good representation at other temperature boundaries.
In this study, different �� and � values were tested for each soil. For sil-10 and sil-20, all
possible combinations of �� values (0.1, 0.38, 3.8 μm s-1) and � values (2, 4, 6, 8, 10, 12, 14, 16),
were simulated. s-8 and s-18 soils were also simulated at all possible combination of �� and �
values, with �� values of 5.0, 6.0, 7.0 and 9.8 μm s-1 and � values that were the same as for the
silt-loam soils. The � values were limited to 2-16 for investigation, after considering the work of
Cass et al. (1984) and Heitman et al. (2008), while the �� values were set to equal to or lower
than the values, from Scanlon and Milly (1994) and Luckner et al. (1989). By taking an average
of the �� RMSE values at different �� for both sil-10 and sil-20, � = 6 was found to give the
most accurate simulation when compared to other � values. This could be written as
���� �� ����(sil-10,20) = 6. � = 2 was the best simulation for the average taken over � RMSE
38
values at different ��, i.e. ���� � ����(sil-10,20) = 2. �� = 3.8 μm s-1 was the best simulation for
the average of the �� RMSE values at different � in both sil-10 and sil-20, and �� = 3.8 μm s-1
was also valid for the average over the � RMSE values at different �. They can be respectively
simplified as ����� �� ����(sil-10,20) = ����� � ����
(sil-10,20) = 3.8 μm s-1. Performing the
same procedure on s-8 and s-18, it was found ���� �� ����(s-8,18) = 6 and ���� � ����(s-8,18) =
2, while ����� �� ����(sil-10,20) = ����� � ����
(sil-10,20) = 9.8 μm s-1. If the procedure was
repeated to consider both the soils and the water contents, then ���� �� ����(sil-10,20,s-8,18) = 6
and ���� � ���� (sil-10,20,s-8,18) = 2. Overall, it could be concluded that � could be varied
between 2 and 6, and the exact �� value from the experiment should be used.
The simulation accuracy deteriorated with time and increasing mean temperature. This
trend was observed for the �� RMSE values from sil-10, sil-20 and s-8, and the � RMSE values
from sil-10 and sil-8. The reverse temperature boundary resulted in a greater deviation from the
exact data (Figs. 3.1(b), 3.1(d) and 3.2(b), 3.2(d)). The average �� RMSE (average of �� RMSE
values at different times) of the reverse boundary simulation was also found greater than the
forward boundaries for sil-10, sil-20 and s-18. When the average � RMSE was taken, a high
RMSE value for reverse simulation was found in sil-20 and s-18. Furthermore, taking the
average of both the soils and the water contents (sil-10, sil-20, s-8, s-18), the average �� RMSE
for forward simulation (0.021 m3 m-3) was less than for reverse simulation (0.0248 m3 m-3).
Similarly, the average � RMSE for forward simulation (0.4800 oC) was less than reverse
simulation (0.5385 oC). If the same procedure is applied to calibrate the forward and reverse
simulation parameters separately, but averaged over soils and water contents, the forward
���� �� ����(sil-10,20,s-8,18) = 6 and the reverse ���� �� ����(sil-10,20,s-8,18) = 16, while the
39
forward ���� � ����(sil-10,20,s-8,18) = 4 and the reverse ���� � ����(sil-10,20,s-8,18) = 14. The
apparently large error in the reverse boundary simulation that requires a high value of � to
compensate, it suggests that some other water vapor phenomenon may need to be accounted for.
3.4.3 Limitation and possible improvement
The previous section 3.4.2 has shown the difficulty encountered when trying to narrow down the
influence of either hydraulic conductivity or vapor enhancement factors as the cause of the
simulation discrepancy from the real physical data. The examination of the �� value indicates the
exact experimental value should be used rather than reducing its value as conducted in Heitman.
In fact, sil-10 and s-8 showed improvement in the �� RMSE value when the �� values were
increased to 15.8 and 18.8 μm s-1, respectively. These values were excluded from the
investigation because such high values would not be a real physical representation of the
experiment. Another aspect for consideration was the � value used in the vapor enhancement
factor. Irrespective of the soil type over different water contents, the � value changed from 2 to 6
when the priority of accuracy was shifted from � to �� RMSE. When a distinction was made
between forward and reverse boundary, the � value could be varied between 4 or 6 and 14 or 16,
respectively. Such an uncertainty was partly due to the result of the inability of researchers to
properly define the vapor enhancement factor. Hence, it was open to speculation (Cahill and
Parlange, 1998, Parlange et al., 1998, Shokri et al., 2009) and ultimately be one of the parameters
to be calibrated in simulation (Heitman et al., 2008, Sakai et al., 2009). Further effort to better
define the mechanism guiding the vapor enhancement factor so that it was able to be measured
through experimentation would definitely remove the parameter as one of the uncertain
parameters for calibration. The greater � value was required in the reverse than the forward
40
simulation could be partly due to the absence of buoyancy-induced flux. The buoyancy
phenomenon was also neglected in the pore scale mechanisms study (Shahraeeni and Or, 2012).
Moreover, the assumptions establishing the derivations of Eqs. 3.1 and 3.5 could be re-
examined due to the weakness of the mass and heat equations that could possibly be partly
responsible for the discrepancy observed. The equations have similar drawbacks as those stated
in Nassar and Horton (1997) that soil was assumed to be rigid and inert, and they the neglected
hysteresis of the water retention curve. It assumed a local thermodynamic equilibrium that the
water, vapor, solid and air have the same temperature. The equation was limited to one-
dimensional flow in the column test. The force driving mass and heat fluxes is governed by
gravity, water content gradient, temperature gradient, heat conduction, sensible heat flux and the
latent heat flux. Also, the current model assumed soil water vapor in equilibrium with liquid
water so that the phase transition from liquid to vapor was instantaneous (Massman, 2015).
Although not all of those limitations could be significantly affecting the simulation, the
discrepancy observed between the simulation and the physical measure data would remain until
the key factors are further considered.
The obvious key factors that should be addressed first are the unsaturated hydraulic
conductivity and the vapor enhancement factor. The poor fitting of �� at the edges of dry and wet
regions, estimated by the Campbell (1974) equation, could be improved with the Fayer and
Simmons (1995) equation that was modified from van Genuchten (1980) to improve the estimate
at low water content (Tuller and Or, 2001). In order to adequately describe the flow of water in
dry soil, the measurement methods suggested by Sakai et al. (2009) could be applied, such as the
water flow under accelerated gravitational conditions, multistep outflow method, and the
evaporation method. Also, the theoretical re-examination of the vapor enhancement factor could
41
be established from the work of Cass et al. (1984) that built on the theory of heat flux (de Vries,
1958). The Darcy law extended by Klinkenberg (1941) to account for the gas slippage effect as a
possible alternative to Fick’s law should also be examined, as attempted in Ho and Webb (1999).
3.5 Conclusions
The transport equations of water, vapor, and heat were successfully extended from the Richards’
equation. The lesson learned from Chapter 2 was implemented in this study by applying a small
spatial discretization size to improve the model accuracy. The new code could predict
temperature and water content distributions that the overall error, i.e. RMSE, was less than those
reported in Heitman.
The simulation results revealed some previously unknown factors. A single boundary
condition could not provide an adequate calibration when multiple boundaries are available for
the calibration because the multiple boundaries calibration was able to identify the parameters
that provide the best compromise for all the boundaries. The best-calibrated parameters (��, �)
for sil-10, sil-20, s-8 and s-18 were different, and further uncertainty arise when accuracy has to
prioritize either the temperature or the water content distribution. The simulation accuracy
deteriorated over time and with the increasing average temperature, with the exception of the
reverse temperature boundary simulation. The high � value, as part of the vapor enhancement
factor, required for reverse boundary simulation implied that possibly some transport
mechanisms were present such as buoyancy force and water retention hysteresis. Rather than a
single value for the �, as reported in Heitman, the current study revealed that the value was
varied between 2 and 6. The second calibrated parameter, �� was found to be best when
maintained at its original value, whereas Heitman found the value had to be reduced to improve
42
fitting, e.g. 90% reduction of the silt loam �� value. Although the reason for such discrepancy is
not clear, the findings of the current study should prevail because the �� value was justified
based on the lowest error after averaging the twelve soil boundary RMSE values for two
different water contents. Furthermore, the inability of the current model to simulate water content
distribution at dry and wet ends could potentially be improved using the Fayer and Simmons
(1995) equation.
Of all the uncertainties found, the vapor enhancement factor was one of the least known
parameters, primarily because the mechanism involved was unclear and it is continuously open
to speculation (Heitman et al., 2008, Ho and Webb, 1999, Parlange et al., 1998). Also, the
complexity of the theory and experimentation further complicating the investigation. The
published experimental data from Cass et al. (1984) would be an important step to re-examine
the theory, and potentially be used to derive and validate a new mathematical expression to
describe the vapor enhancement factor.
43
(a) (b)
(c) (d)
Figure 3.1. Silt loam (sil-20) simulated: (a) water content and (c) temperature distributions at
temperature boundary of 15/-50, 15/-150 and 30/-150, which were 17.5-12.5, 22.5-7.5oC and
37.5-22.5oC, respectively. The (b) water content and (d) temperature distribution at reverse
temperature boundary of 22.5/100, i.e. 17.5-27.5oC, after 12, 48 and 96 hours. Note: the initial
water content set at 0.20 m3 m-3. The 63.8 10sK m s-1 and zeta (�) = 12 used in the
simulation.
0.05
0.10
0.15
0.20
0.25
0.30
0 0.05 0.1
Wat
er c
on
ten
t (m
3/m
3)
Distance from warm end (m) 15/-50 15/-150
30/-150 Sim (15/-50)
Sim (15/-150) Sim (30/-150)
0.05
0.10
0.15
0.20
0.25
0.30
0 0.05 0.1
Wat
er c
on
ten
t (m
3/m
3)
Distance from warm end (m) R12, 22.5/100 R48, 22.5/100
R96, 22.5/100 Sim (R12, 22.5/100)
Sim (R48, 22.5/100) Sim (R96, 22.5/100)
280
290
300
310
0 0.05 0.1
Tem
per
atu
re (
K)
Distance from warm end (m) 15/-50 15/-150
30/-150 Sim (15/-50)
Sim (15/-150) Sim (30/-150)
290
295
300
305
0 0.05 0.1
Tem
per
atu
re (
K)
Distance from warm end (m) R12, 22.5/100 R48, 22.5/100
R96, 22.5/100 Sim (R12, 22.5/100)
Sim (R48, 22.5/100) Sim (R96, 22.5/100)
44
(a) (b)
(c) (d)
Figure 3.2. Sand (s-8) simulated: (a) water content and (c) temperature distributions at
temperature boundary of 15/-50, 15/-150 and 30/-150, which were 17.5-12.5, 22.5-7.5oC and
37.5-22.5oC, respectively. The (b) water content and (d) temperature distribution at reverse
temperature boundary of 22.5/100, i.e. 17.5-27.5oC, after 12, 48 and 96 hours. Note: the initial
water content set at 0.08 m3 m-3. The 69.8 10sK m s-1 and zeta (�) = 6 used in the
simulation.
0.00
0.05
0.10
0.15
0.00 0.05 0.10Wat
er c
on
ten
t (m
3/m
3)
Distance from warm end (m)
15/-50 15/-150
30/-150 Sim (15/-50)
Sim (15/-150) Sim (30/-150)
0.00
0.10
0.20
0 0.05 0.1Wat
er c
on
ten
t (m
3/m
3)
Distance from warm end (m) R12, 22.5/100 R48, 22.5/100
R96, 22.5/100 Sim (R12, 22.5/100)
Sim (R48, 22.5/100) Sim (R96, 22.5/100)
280
290
300
310
320
0.00 0.05 0.10
Tem
per
atu
re (
K)
Distance from warm end (m)
15/-50 15/-150
30/-150 Sim (15/-50)
Sim (15/-150) Sim (30/-150)
290
295
300
305
0.00 0.05 0.10
Tem
per
atu
re (
K)
Distance from warm end (m)
R12, 22.5/100 R48, 22.5/100
R96, 22.5/100 Sim (R12, 22.5/100)
Sim (R48, 22.5/100) Sim (R96, 22.5/100)
45
Table 3.1 Input parameters used for silt loam and sand.
Parameter Silt loam Sand
Wetting differential heat (�, m2 s-2) 4620 200
Bulk density (��, kg m-3) 1200 1600
Organic mass fractions (��) 0.044 0.006
Mineral mass fractions (��) 0.956 0.994
Soil porosity (�, m3 m-3) 0.547 0.396
Soil clay fraction (�) 0.249 0.011
Saturated hydraulic conductivity (��, m s-1) 3.8 x 10-6 9.8 x 10-6
Soil thermal conductivity, Eq. 3.10 fitting parameters
��
��
��
-0.952
-4.31
6.00
-0.394
-5.11
7.23
Water retention curve-fitted parameters
�� (m)
�
-0.13
6.53
-0.03
3.38
Note: parameters compiled from Heitman et al. (2008).
46
Table 3.2 The comparison of water content and temperature root mean square errors (RMSEs)
from Heitman et al. (2008) and current simulation, at few temperature boundary conditions.
��(m�m��), �(K) RMSE
Heitman et al. (2008) Current Simulation
Calibrated at �� = 0.38 μm s-1, � = 4
Silt loam at 0.20 m3 m-3
22.5oC/-100oC m-1 0.013, 0.58 0.010, 0.38
15.0oC/-50oC m-1 0.019, 0.18 0.017, 0.15
22.5oC/-150oC m-1 0.011, 0.91 0.010, 0.45
Silt loam at 0.10 m3 m-3
22.5oC/-50oC m-1 0.014, �� 0.018, 0.10
22.5oC/-150oC m-1 0.017, �� 0.030, 0.78
Calibrated at �� = 9.8 μm s-1, � = 4
Sand at 0.08 m3 m-3
22.5oC/-100oC m-1 0.049, 0.60 0.023, 0.48
15.5oC/-50oC m-1 0.012, 0.14 0.012, 0.23
22.5oC/-150oC m-1 0.023, 1.28 0.021, 0.81
Sand at 0.18 m3 m-3
22.5oC/-50oC m-1 0.009, 0.07 0.010, 0.29
22.5oC/-150oC m-1 0.016, 0.43 0.014, 0.38
Note: �� referring to as not available.
47
Chapter 4. An improved heat flux theory and mathematical equation to
estimate water vapor volume expansion-advection as an alternative to
mechanistic enhancement factor2
4.1 Abstract
The current study extended the theory of heat flux which has been normally used to assess the
mechanistic enhancement factor. The present study improved heat flux theory to include three
new phenomena that were excluded from the simplified heat flux equation. Those three
phenomena were sensible heat by liquid water movement, sensible heat by water vapor
movement, and the effects of partial derivative of matric pressure head on relative humidity with
respect to temperature. The first phenomenon was found to be an important factor in relatively
wet porous media, whereas the third phenomenon was found important in near dry porous media
condition. Moreover, mathematical descriptions were derived to allow direct conversion of the
mechanistic enhancement factor to a water vapor volume expansion-advection term. The study
used the basic mass balance equation, ideal gas law, and water vapor advection by air volume
expansion to describe the water vapor volume expansion-advection mechanism. The water vapor
volume expansion-advection equation was found to be able to describe the influence of soil
moisture content and air pressure. Also, a bell-like curve of water vapor relative permeability
was needed to comply with the Cass et al. (1984) experimental data. A mathematical function
was proposed as a temporary solution. Further investigation of the water vapor relative
permeability mathematical relationship may be necessary to improve the estimation at saturated
soil moisture content, and also, to reveal its theoretical foundation.
2 This chapter has been published in Goh and Noborio (2016).
48
4.2 Introduction
Philip and de Vries (1957) was the first to suggest the presence of a vapor enhancement factor in
soil. A mechanistic enhancement factor ( ) was used to describe the discrepancy between
observed and calculated vapor diffusion by Fick’s law, in the presence of a temperature gradient.
The liquid island effect between solid particles, and the local temperature gradients in the air-
filled pores that were higher than the average temperature gradients of the bulk soil were claimed
to be responsible for the factor. Alternatively, Cary and Taylor (Cary, 1963, Cary, 1964, Cary,
1965, Cary and Taylor, 1962) derived a phenomenological enhancement factor ( ), based on
irreversible thermodynamics, to study similar phenomena. The air porosity and the tortuosity
effect (Penman, 1940) of porous media could be used to relate the mechanistic enhancement
factor to the phenomenological enhancement factor. Therefore, a solution to the mechanistic
enhancement factor alone is sufficient, and it can be translated into the phenomenological
enhancement factor when necessary.
Cass et al. (1984) proposed a simplified theory of the heat flux equation by assuming
only the presence of heat conduction and latent heat transfer, without the presence of liquid water
movement. They applied the assumption to study the effect of the mechanistic enhancement
factor on lysimeter sand and Portneuf silt loam. Their findings demonstrated that their proposed
mathematical function was able to model the variation of the mechanistic enhancement factor
with soil moisture content. Hiraiwa and Kasubuchi (2000) utilized the simplified heat flux theory
to study the effect of temperature on thermal conductivity. Sakaguchi et al. (2009) also used the
theory to estimate vapor flux of the heat pipe phenomenon in Ando soil. Similarly, Lu et al.
(2011) implemented the simplified heat flux theory and the mathematical function to the
estimation of the mechanistic enhancement factor of the silty clay soil and correlated it with the
49
soil moisture content. However, Shokri et al. (2009) found that liquid water movement alone is
sufficient to account for the mechanistic enhancement factor.
Ho and Webb (1996) defended the presence of vapor enhancement diffusion through a
simple mass balance estimation on pore-scale analysis. The analysis estimated that vapor
enhancement diffusion could be ten times the amount of diffusion by Fick’s law. The subsequent
work of Ho and Webb (1999) further established the presence of vapor enhancement diffusion
through experiments with pore scale, fracture aperture, and porous media. The research findings
on pore-scale modeling generated data showing the relationship between the mechanistic
enhancement factor and modeled water saturation. Ho and Webb (1999) proposed that the
mathematical relationship given in McInnes (1981) and Cass et al. (1984) was a preferable
function. The research work of Cass remained valid, and it established a simplified heat flux
theory and a mathematical formulation to relate the mechanistic enhancement factor with soil
moisture content in the study of vapor enhancement diffusion. However, the assumptions used to
establish the simplified heat flux theory are a possible drawback. Clearly, further research is
required to test the validity of the assumptions.
Ho and Webb (1996) have highlighted drawbacks in the vapor diffusion studies. One of
the suggestions for improvement was to use the numerical code TOUGH2 (Pruess, 1991) that
could be applied in geothermal, environmental remediation, and nuclear waste management. The
code consisted of a pressure gradient term used to govern vapor advection that was able to
account for part of the observed vapor enhancement diffusion. The subsequent work of Ho and
Webb (1999) suggested the use of an advection-dispersion model to govern vapor enhancement
phenomenon, but they claimed the dusty-gas model to be more superior in terms of accuracy.
Webb and Pruess (2003) found the diffusive flux of Fick’s law comparable to the dusty-gas
50
model after including the Knudsen effect. A complete description of the dusty-gas model
comprising the Knudsen effect, viscous flow, and Fick’s law is available in Cunningham and
Williams (1980) and Mason and Malinauskas (1983). Furthermore, the research findings of
Cahill and Parlange (1998) and Parlange et al. (1998) found discrepancies between field
observations and the prediction of the Philip and de Vries (1957) model. The discrepancy
resulted from the thermally-driven advection transport mechanism that was governed by
alternating expansion and contraction of the soil air. Clearly, additional investigations of Fick’s
law including the Knudsen effect and advection term are necessary.
In this study, we investigated the assumptions used by Cass et al. (1984) in formulating
the simplified heat flux theory and formulae. From now on it is referred as Cass. The
assumptions have neglected three phenomena. These are sensible heat by liquid water transfer,
sensible heat effect by water vapor transfer, and the partial derivative of matric pressure head
effects on relative humidity with respect to temperature. This study employed a basic mass
balance approach, ideal gas law relation, and air volume expansion-advection mechanism to
justify the necessity of a new mechanism to account for the water vapor enhancement effect. A
mathematical formula was established to relate the mechanistic enhancement factor to the water
vapor volume expansion-advection mechanism.
4.3 Materials and methods
4.3.1 Cass experimental setup, parameter estimation, relation to the present study and
limitation
Cass set up a research method to allow estimation of thermal conductivities at various soil
pressures (0, 103, 517, and 1034 kPa gauge), moisture contents (0.002, 0.009, 0.021, 0.024,
51
0.054, 0.060, 0.061, 0.108, 0.186, and 0.190 m3 m-3), and temperatures (average temperature of
3.5, 22.5, and 32.5oC). Stainless steel tubes (1.5 mm wall thickness, 47 mm inner diameter, and
230 mm length) were used to contain the sieved and repacked soil samples (lysimeter sand, and
Portneuf silt loam) and they were inflated to the preset gauge pressure. The pressured-steel tube
first was equilibrated in one bath to a preset temperature and then was immersed in another bath.
A temperature difference of 5oC was imposed on each experimental condition. Copper
(Evenohm) constantan thermocouples were used to measure temperatures within and outside the
steel tube. A data logger used to store the data every 12 s until the maximum temperature
difference of 5oC fell to near 0oC. The soil thermal conductivity was estimated by the Riha et al.
(1980) method.
Thermal conductivity values estimated for particular soil moisture content at a few
discreet pressures allowed approximation of the thermal conductivity under different pressure
environments. Extrapolation of a linear line to a vertical intercept on a thermal conductivity
versus pressure ratio ( rP ) graph facilitated the derivation of thermal conductivity at an extreme
pressure condition. The thermal conductivity value estimated at the extreme pressure condition
had significant importance because it represented the heat flux of the media without the presence
of water vapor. Thermal conductivity values at different pressures, except at the extreme
pressure, included the phenomena of three phases. They were the liquid water movement
(sensible heat), wet porous media (liquid water plus dry porous media) heat conduction, and
water vapor heat fluxes (latent heat and sensible heat). The Cass method allowed the isolation of
thermal conductivity of water vapor from the other heat fluxes.
In general, water vapor transfer was assumed to be governed by the diffusion equation.
The assumption required a mechanistic enhancement factor to be applied to the diffusion
52
equation to compensate for the excessively large value contributed by the water vapor flux. In
this study, we re-examined the heat flux theory to reconstruct the temperature gradient heat flux
equation, and also to identify the physical phenomena lacking in the simplified heat flux
provided from Cass. Also, the mathematical equation was derived and validated using the results
of Cass experiments at different pressures and soil volumetric water contents. The present study
used the published data on lysimeter sand and Portneuf silt loam from Cass.
4.3.2 Theory of heat flux
Cass adopted the method of Parikh et al. (1979) to conduct measurements of the thermal
conductivity as a function of temperature, soil moisture content, and pressure. The heat flux
theory used in the calculation of the mechanistic enhancement factor was a simplified version of
the heat flux density flow given in de Vries (1958). From now on it is referred as deVries. The
complete expression of heat flux density given by deVries is as follows, except for the last term
on the right side of the equation:
*h o v v o v L o L L o r
dTq L q c T T q c T T q c T T q
dx (4.1)
where hq is the total heat flux density (J m-2 s-1), * is the thermal conductivity in the porous
media in the presence of liquid water (J m-1 s-1 K-1), oL is the latent heat (J kg-1) of water at
temperature oT (K), vq is the water vapor flux (kg m-2 s-1), vc is the specific heat of water vapor
at constant pressure (J kg-1 K-1), Lc is the specific heat of liquid water (J kg-1 K-1), Lq is the liquid
water flux (kg m-2 s-1), and rq is the liquid water flux by root (kg m-2 s-1).
On the right side of Eq. 4.1, the first term is the heat flux from pure heat conduction, and
the second term is the latent heat flux by water vapor movement. The third term is the sensible
53
heat flux by water vapor movement while the fourth term is the sensible heat flux by liquid water
movement. The fifth term is the sensible heat flux by root water movement. This fifth term was
not important in the vapor enhancement studies because the soils used in Cass experiments were
mainly clay, silt, and sand, excluding the influence of roots, e.g. Caldwell et al. (1998). In Cass’s
research, pure heat conduction and transfer of latent heat by vapor movement were the only flux
terms assumed to govern the heat transfer used in the mechanistic enhancement studies. Cass
excluded the third and fourth terms in his study. In the present study, the first four terms on the
right-hand side of Eq. 4.1 were included in the calculation of the mechanistic enhancement factor.
Water vapor diffusion is governed by Fick’s law (Fick, 1995, Jackson et al., 1963) and it
is an isothermal equation. Under the influence of spatial temperature gradient, Fick’s law can be
rewritten as follows (Cass et al., 1984, Philip and de Vries, 1957):
vv a
d dTq D
dT dx
(4.2)
where is the mechanistic enhancement factor that is used to account for increased vapor
diffusion over the ordinary Fick’s law equation, D is the molecular diffusivity of water vapor in
air (m2 s-1), is the tortuosity factor, dimensionless, a is the volumetric air content of the
media (m3 m-3), and v is the water vapor density (kg m-3).
According to Eq. 4.3, the water vapor density is affected by saturated water vapor density
( s , kg m-3) and matric pressure head effects on relative humidity ( mh , dimensionless). In the
chain rule, the partial derivative of water vapor density with respect to temperature resulted in
Eq. 4.4. Philip and de Vries (1957) introduced the second term on the right side of Eq. 4.4. Later,
Nakano and Miyazaki (1979) proposed the first term of the right side of Eq. 4.4.
v s mh (4.3)
54
v m ss m
hh
T T T
(4.4)
The liquid water flux was governed by the Buckingham-Darcy flux law (Jury and Horton,
2004). Similar to Eq. 4.2, under the influence of temperature gradient the Buckingham-Darcy
flux law became:
mL L
d dTq K
dT dx
(4.5)
where L is the soil liquid water density (kg m-3), K is the hydraulic conductivity (m s-1), and
m is the matric pressure head (m).
By placing Eqs. 4.2-4.5 into Eq. 4.1, and performing a simple mathematical
rearrangement, an improved heat flux equation could be rewritten as follows:
*m m s
h L L a o v s m
d h dTq c T K D L c T h
dT T T dx
(4.6)
Cass measured thermal conductivity at a temperature gradient, and at different soil
moisture contents, and pressures. The method used was based on Riha et al. (1980). At each soil
moisture content and temperature, thermal conductivity values were plotted against the ratio of
the atmospheric pressure to the experimentally imposed pressure values. From the curve-fitted
straight line on the graph, the thermal conductivity at an extreme pressure was determined by
extrapolating to the vertical intercept. The thermal conductivity at extreme pressure should be
given by the first two terms of Eq. 4.6, instead of only the first term as used by Cass. This is
because those two terms are governed by solid and liquid of heat movement.
The simplified heat flux equation proposed by Cass was as follows:
*s
h a o m
dTq D L h
T dx
(4.7)
55
Equation 4.7 excludes the influence of sensible heat in the water vapor movement and
liquid water movement, and the partial derivative of matric pressure head effects on relative
humidity with respect to temperature. In this study, the mechanistic enhancement factor
estimated by Eqs. 4.6 and 4.7 were compared.
4.3.3 The mass balance relationship between liquid water and water vapor, and its effect
on water vapor volume expansion-advection
If a unit mass of liquid water vaporizes fully into the air, the logical assumption is that the
vaporized mass of liquid water is equivalent to the mass of vapor in the air. By knowing that
mass is a product of density and volume, equating both masses of liquid water and water vapor,
the following simple mass balance relation was produced:
vaporwater
vapor water
V
V
(4.8)
It is a known fact that the density of liquid water is greater than the density of water
vapor. For instance, at 20oC, saturated water vapor density is 0.0229 kg m-3, calculated from the
Kimball et al. (1976) equation while the liquid water density is 996.55 kg m-3. The ratio given by
Eq. 4.8, at the temperature, indicates that water vapor volume is 43,578 times greater than the
volume of liquid water. Hence, the volume occupied by water vapor is significantly greater than
the volume of liquid water. The simple mass balance equation indicates that a mass of liquid
water transformed into water vapor would take up a larger volume of pore space than the same
mass in the form of liquid water. Under the constant air pore space in porous media, the
expanding water vapor volume, resulted from vaporizing liquid water, was directly translated to
the volume expansion-advection of water vapor apart from vapor diffusion by Fick’s law.
56
Therefore, the temperature related variation in density and volume, between liquid and vapor
phases, was the reason for water vapor volume expansion-advection. The volume expansion-
advection term could be governed by Darcy law as described in Ho and Webb (1996) and Ho and
Webb (1999).
4.3.4 The presence of volume expansion-advection from the perspective of ideal gas law
The ideal gas law is as follows:
PV nRT (4.9)
where P is the pressure of the gas (Pa), V is the volume of the gas (m3), n is the amount of gas
in moles, and R is the ideal gas constant (J K-1 mol-1).
By taking the partial derivative of Eq. 4.9 with respect to temperature, it resulted in the
following form:
V P nP V nR RT
T T T
(4.10)
where Eq. 4.10 is based on the rationale that the increase in temperature leads to a rise of the
water vapor moles, due to vaporization. The increase in water vapor moles would either be
accompanied by an increase in the volume of gas phase or gas partial pressure or a combination
of both in equilibrium.
Consider a column of soil, under the influence of a temperature gradient, the side of the
column with higher temperature has greater vaporization of liquid water than the other end with
lower temperature. In a soil column under constant pore-space, the evaporation of liquid water to
water vapor would result in high partial pressure of water vapor and water vapor density as
governed by Eq. 4.10. Hence, it implies that partial pressure gradients of water vapor and water
vapor density gradient are present in the soil column. The pressure gradient governing the
57
volume expansion-advection flow would be an adequate approach to represent the vapor
enhancement factor. Clearly, it indicates that the temperature gradient introduced to the system
has resulted in volume expansion-advection of water vapor.
4.3.5 The presence of air and water vapor volume expansion-advection
The mass balance of air equation and the ideal gas equation from the research work of Parlange
et al. (1998) were implemented to support the claim that volume expansion-advection of water
vapor does exist. The use of the mass balance equation of air resulted in the following relation:
air airair air air airu
t t x
(4.11)
where air is the density of air (kg m-3), air is the porosity of air space (m3 m-3), and airu is the
velocity of air (m s-1).
By taking the partial derivative of equation air air airP M RT , converted from Eq. 4.9,
with respect to time, and assuming that pressure does not change with time, it would give the
following relationship:
air air T
t T t
(4.12)
Puting Eq. 4.12 into Eq. 4.11 would generate the following relationship that is the
equation used by Parlange et al. (1998) to estimate air volume expansion:
2
1
1 xair air
air air airxair
Tu x
T t t
(4.13)
They assumed that the air velocity given in Eq. 4.13 governs water vapor flow. Clearly,
air volume expansion caused water vapor advection is present as a result of temperature variation
in the soil.
58
4.3.6 Water vapor volume expansion-advection equation
The equation governing water vapor volume expansion-advection could be described by Darcy
law (Muskat and Wyckoff, 1946):
p vv v
L v v
k gq Pd
dx g
(4.14)
where vq is the water vapor flux (kg m-2 s-1), pk is the gas permeability (m2), g is the gravity
acceleration, 9.81 m s-2, v is the dynamic viscosity of water vapor (kg m-1 s-1), and vP is the
partial pressure of water vapor (kg m-1 s-2).
Equation 4.13 was derived by assuming the pressure of air as a constant value. Hence,
the change of air temperature directly affects the air density without any impact on the air
pressure. The general relationship occurs by allowing the effect of temperature on either water
vapor pressure or water vapor density, or a combination of both. The general form of the ideal
gas law before applying the partial derivative with respect to time is as follows:
v
v v
P RT
M (4.15)
where vM is the molecular weight of water vapor, 0.01801528 kg mol-1.
By applying Eq. 4.15 to Eq. 4.14, it allows any variation of partial pressure and density
of water vapor to have a direct relationship with the temperature of water vapor. Hence, it brings
about the following relationship:
p r L v
v
v v
k k R dTq
M dx
(4.16)
where rk is the relative permeability, dimensionless, and it was applied as in Ho and Webb
(1999).
59
4.3.7 The relation of mechanistic enhancement factor and water vapor volume expansion-
advection
From the simple mass balance equation, the ideal gas law, and air volume expansion, it is evident
that water vapor is not only diffused through porous media, but also that the water vapor volume
expansion-advection mechanism is present. The nature of the experimental setup by Cass
enabled estimation of the mechanistic enhancement factor due to the temperature gradient effect.
Water vapor volume expansion-advection contributed to the enhancement factor. Therefore, the
mechanistic enhancement factor could be detached from the water vapor diffusion term, in Eq.
4.6, to form a new term, i.e. water vapor volume expansion-advection term, as follows:
m sa o v s m
hD L c T h
T T
m so v a o v s m
hf L c T D L c T h
T T
(4.17)
where f is the water vapor volume expansion-advection factor. In simplified form:
1 m sa s m
hf D h
T T
(4.18)
From Eq. 4.7, similarly it is written as:
1 sa mf D h
T
(4.19)
Equation 4.18 is only suitable for newly constructed experimental research while Eq.
4.19 can be used for both new experimental data analysis, and also the existing published data.
To utilize Cass published data, Eq. 4.20 is necessary. The first term in Eq. 4.20 represents the
water vapor diffusion of the simplified heat flux, and the second term represents the water vapor
diffusion of the improved heat flux. The equation used in the current study is as follows:
60
a o m s m s
a s m
o v
D L h hf D h
L c T T T T
(4.20)
In the present study, Eqs. 4.18-4.20 are named as improved, simplified, and adapted-
improved water vapor volume expansion-advection factors, respectively.
The water vapor volume expansion-advection factor ( f ) from the relationship in Eq.
4.16:
p r L v
v v
k k Rf
M
(4.21)
After rearranging Eq. 4.21 it becomes:
p rv v
o L v o
k kf M
k R k
(4.22)
According to Ho and Webb (1999), the permeability is used for the slip or Knudsen
effect, and it has the Klinkenberg factor (b ). The permeability used is as follows:
1 ep o rk k bP (4.23)
where ok is the intrinsic permeability (m2), r oP P P is the pressure ratio where oP is the
atmospheric pressure at 101325 Pa and P is the imposed air pressure when measuring the
thermal conductivity, and e is a fitting parameter.
The relative permeability is a function of soil moisture content. Hence, the following
empirical relationship was proposed to govern the relationship between relative permeability and
soil moisture content:
242
1 2 3
Lr L
L L
k aa a a
(4.24)
The left side of Eq. 4.22 is the dependent variable, while the pressure ratio and soil
61
moisture content are independent variables. The relationship was curve-fitted to determine
variable constants for 1a , 2a , 3a , 4a , b , and e . It is important to note that at zero soil volumetric
water content, the relative permeability, given by Eq. 4.24, would be zero, which would
terminate the water vapor volume expansion-advection. At fully saturated soil moisture content,
the relative permeability would fail to generate a zero value.
van Genuchten-Mualem (Mualem, 1976, van Genuchten, 1980), van Genuchten-Burdine
(Burdine, 1953, van Genuchten, 1980), Corey (Corey, 1954), and Falta (Falta et al., 1989),
advocated other relative permeability equations, shown below, respectively:
20.5 1
1 1 1 1
mm
L Lr
s s
k
(4.25)
2 1
1 1 1 1
mm
L Lr
s s
k
(4.26)
2 2
1 1L Lr
s s
k
(4.27)
3
1 Lr
s
k
(4.28)
4.4 Results and Discussion
4.4.1 The general trend of thermal conductivities
The thermal conductivity of water vapor was lower than the combined thermal conductivities of
wet porous media and liquid water sensible heat. The wet porous media consisted of heat
conduction from liquid water and dry porous media. For instance, in lysimeter sand, the water
62
vapor thermal conductivity had about 0.03-59.6% of the thermal conductivities of wet porous
media and liquid water sensible heat combined, at pressure ratio ( oP P P ) of 1.0. The
variation in water vapor thermal conductivity narrows to 0.03-31.6, 0.03-9.56, 0.03-4.98% at the
respective pressure ratios of 0.53, 0.16, and 0.083. The pressure ratio was given by atmospheric
over air pressure. For example, 1.0 pressure ratio is at atmospheric pressure, whereas the 0.083
pressure ratio is a much higher air pressure than the atmospheric pressure. The high percentage
of water vapor thermal conductivity at low pressure indicated high water vapor transfer. A
similar observation was demonstrated in the Portneuf silt loam. Also, water vapor thermal
conductivity values initially rose at low volumetric water content, before curving downward at
higher soil moisture content observed under various pressure conditions (Fig. 4.1). Campbell
(1985) reported a similar trend for sand, silt loam, and clay soils. However, at extremely high
pressure, both the wet porous media and liquid water sensible heat thermal conductivities
increased with the increasing soil volumetric water content, without falling trend (Fig. 4.1). The
downward curve behavior is consistent with the physical expectation of porous media because as
the soil moisture content increases less air space is available for vapor flux. Hence, at saturated
soil moisture content, the water vapor thermal conductivity should approximate to zero.
4.4.2 The wet porous media thermal conductivity by heat conduction, and the sensible
heat thermal conductivity by liquid water movement
The thermal conductivity at extreme pressure consists of two components, there are thermal
conductivities of wet porous media (liquid water plus dry porous media) and sensible heat by
liquid water movement. They are respectively the first and second terms on the right side of Eq.
4.6. Both thermal conductivities increased with increasing soil moisture content in the porous
63
media. The thermal conductivity by liquid water movement is much less than that of the wet
porous media thermal conductivity, but the increment rate of the former is far greater than the
latter. For instance in a near dry porous region, i.e. 0.014 m3 m-3, the liquid water movement had
125.33 10 % of the wet porous media thermal conductivity value. In the wet region, i.e. 0.295
m3 m-3, it increased to 0.039%. The highest soil moisture content studied was about 65.5% of
fully saturated soil volumetric water content. Hence, at soil moisture content greater than the
studied value, more than 0.04% of liquid water movement over wet porous media thermal
conductivity would be expected. The Portneuf silt loam also demonstrated a similar trend. The
observation indicated that thermal conductivity by liquid water movement is a more important
factor in the wet than in the near dry porous media regions. However, the liquid water movement
thermal conductivity is one of the three factors that were excluded from the simplified heat flux
theory by Cass.
4.4.3 The thermal conductivity of water vapor diffusion and volume expansion-advection
The thermal conductivity of water vapor is composed of two parts - water vapor diffusion, and
water vapor volume expansion-advection. The water vapor diffusion was solved by Fick’s law,
and the water vapor volume expansion-advection by Darcy law was used as the remaining vapor
flux. The observed thermal conductivities of water vapor diffusion and volume expansion-
advection decreased with increasing pressure. The finding is consistent with the earlier
observation that the high pressure condition reduced the availability of water vapor. Therefore,
less water vapor is available for diffusion and volume expansion-advection. In a near dry porous
media region, i.e. 0.014 m3 m-3, the diffused water vapor thermal conductivity from the improved
heat flux equation increased by 7-12% from the values of water vapor diffusion derived from
64
simplified heat flux. At other wet regions, the thermal conductivity of water vapor diffusion was
only increased by 0.3-0.5%. On the other hand, the volume expansion advected water vapor
thermal conductivity at the near dry porous media region was found to decrease by 10-90% from
those conductivities using the simplified heat flux equation. In the wet region, the thermal
conductivity shrank by 0.03-0.3%. When volume expansion-advection and diffusion were
compared, the volume expansion-advection of water vapor thermal conductivity was about 1 to
67% of the values of water vapor diffusion, at the near dry region 0.014 m3 m-3. In the wet
region, it could reach as high as 1600%. In short, the percentage increased with increasing soil
moisture content. Therefore, the diffusion of water vapor thermal conductivity is a major
contributor to water vapor flux only in the near dry porous media region. The importance of
water vapor diffusion was enhanced through the utilization of the improved heat flux theory. The
new terms that contributed to the enhancement of diffusion in the near dry region were excluded
from the simplified heat flux by Cass. The new terms were the sensible heat flux by water vapor
movement, and the partial derivative of matric pressure head effects on relative humidity with
respect to temperature, which were shown in the improved heat flux theory in Eq. 4.6. In the wet
porous regions, the volume expansion-advection of water vapor thermal conductivity was
significantly higher, i.e. it could be up to 16 times higher than those values given by the diffusion
process. Both the lysimeter sand and Portneuf silt loam have manifested similar characteristics as
discussed.
4.4.4 The water vapor thermal conductivity by latent heat and sensible heat
Water vapor movement is influenced by two phenomena - latent heat flux and sensible heat flux.
The water vapor thermal conductivity due to the sensible heat flux has approximately 0.3-0.4%
65
of the latent heat flux of water vapor thermal conductivity. The percentage does not vary with
pressure or soil moisture content, only with temperature because both the latent of vaporization
and the specific heat of water vapor are a function of temperature.
4.4.5 The water vapor thermal conductivity of the partial derivative of matric pressure
head effects on relative humidity, with respect to temperature
Water vapor density is a function of the saturated water vapor density and the matric pressure
head effects on relative humidity (refer to Eq. 4.3). Eq. 4.4 resulted from subjecting Eq. 4.3 to
the partial derivative with respect to temperature. The second term on the right side of Eq. 4.4 is
the original function term used in the simplified heat flux, by excluding the first term of the
equation (refer to Eq. 4.7). Similar to section 4.4.3, in drier porous media regions the first term
has 7-12% of the value of the right side second term of Eq. 4.4, i.e. 0.014 m3 m-3. In wet regions,
it is 65.0 10 - 0.146%. Therefore, the partial derivative of matric pressure head effects on
relative humidity with respect to the temperature term appeared to be more essential in near dry
porous media region than in wet regions. Moreover, in relative terms, the partial derivative term
was the main reason for increased water vapor diffusion in the drier porous media. In other
words, the partial derivative term has a more significant influence than the sensible heat flux
term in the drier region.
4.4.6 Water vapor permeability variables with respect to soil moisture content and air
pressure
In variably saturated porous media, water vapor flux is subjected to the influence of soil moisture
content and air pressure. In the water vapor diffusion process, the variation in soil moisture
66
content directly changes the matric suction head that in turn would affect the relative humidity.
The relative humidity has a direct influence on water vapor density, which is one of the
parameters to affect the water vapor diffusion process. On the other hand, the pressure imposed
on the porous media would have a direct influence on pressure ratio. The pressure ratio would
then change the diffusion of water vapor in the porous media. In terms of water vapor volume
expansion-advection, the soil moisture content would influence the relative permeability (Eq.
4.24) which would then affect the water vapor volume expansion-advection. The pressure would
affect the permeability (Eq. 4.23) that is responsible for the Knudsen effect. In Eq. 4.22 which is
used for curve-fitting the left side of the equation is the dependent variable. The right side of the
equation consists of permeability and relative permeability, that are respectively influenced by
the soil volumetric water content and the pressure ratio as the independent variables. The curve-
fitting results are shown in Fig. 4.2. The permeability variables ( p r ok k k ) predicted from the
simplified and improved heat flux theories possessed concave downward curve characteristics
that were influenced by soil volumetric water content. The characteristic was carried over from
the relationship between water vapor thermal conductivity and soil volumetric water content as
in Fig. 4.1. The concave characteristic was most profound in low pressure regions, while
appeared to be less significant at high pressure. Moreover, the permeability variables decreased
with increasing pressure. The observation is consistent with the fact that the water vapor flux
decreased in high pressure regions because less water can vaporize in a high pressure
environment. The influence of soil volumetric water content and pressure on permeability
variables show a promising relationship. Its trend is similar to the relationship of heat flux, soil
volumetric water content, and pressure as reported by Sakaguchi et al. (2009). Also, the
comparison of experimental and predicted data has demonstrated a reasonably good compliance,
67
with R-squared values of 0.93 as indicated in Figs. 4.3(a)-4.3(b). A consistent relationship was
observed from both lysimeter sand and Portneuf silt loam.
The derived relative permeability values appeared to predominate at lower values when
compared to other models (refer to Fig. 4.4(a)). The other models begin with a unity relative
permeability value at an extremely dry condition. The value reduces with increasing soil
moisture content, and it approximates to zero relative permeability in fully water saturated
porous media. On the other hand, the estimated relative permeability value increases from zero
relative permeability to reach a peak, before it reduces to approximately zero in fully water
saturated porous media (Fig. 4.4(b)). A similar relationship was shown between water vapor
conductivity and matric suction head for sand, silt loam, and clay soils (Campbell, 1985). The
relative permeability given by other models is only applicable to conditions when an external
pressure force is applied to a porous media. In the case of temperature gradient generated relative
permeability, in the absence of external pressure force, the water vapor volume expansion-
advection by temperature-driven vaporization would have to depend on the soil moisture content.
4.4.7 Limitation of previous work, and the possibility of improvement
There were some limitations of the Cass research work that could be considered for further
improvement. For instance, a single measurement of thermal conductivity, at each point of
pressure and water content, and more than two measurements were necessary to establish
standard deviation or error. A temperature gradient of 5oC was imposed on the soil tube, and the
possible effect of different gradients was not investigated. Also, water content at near saturation
was not investigated. The thermal conductivity was measured at only four pressure values to
estimate thermal conductivity at extreme pressure, and so measurement at additional pressure
68
values could further improve the estimation.
Moreover, some factors, related to soil heat capacity and thermal conductivity, are also
worth assessing for future improvement. The findings of the following researchers corroborate
this statement. Bilskie et al. (1998) found variation in α-alumina colloidal suspensions in water
caused changes in soil heat capacity and thermal conductivity measured by DPHP probe. Also,
dissolved compounds like salt decreased thermal conductivity as the concentration increased
(Abu-Hamdeh and Reeder, 2000, Noborio and McInnes, 1993). Apart from the volumetric
proportions of various soil components, both structure, e.g. orientation of soil particles, and
packing, e.g. nature of contacts between soil particles, could affect soil thermal conductivity
(Winterkorn and Fang, 1991). Also, soil porosity, the chemical composition of soil particles and
water structure served as important factors for soil thermal conductivity (Farouki, 1981). The
presence of a vapor enhancement factor was undeniably significant and the proposed governing
equation for water vapor phenomenon, in this study, would be applicable. Since Fick’s law
poorly predicted the vapor flux and required a significant contribution from the water vapor
volume expansion-advection by Darcy law, the possibility of the latter equation to completely
replace the former equation should be examined. The mathematical solution to the water vapor
relative permeability presented in the study was temporary, so a theoretically derived equation
with the ability to handle its value at saturated soil moisture content would require further
examination.
4.5 Conclusions
The resulting thermal conductivity from the liquid water movement sensible heat term possesses
a rapid rate of increment when compared to wet porous media heat conduction that led to thermal
69
conductivity with increasing soil moisture content. This fact suggests that the sensible heat by
liquid water movement is an important factor to consider in highly wet porous media, and in
variably saturation porous media. In relatively dry porous media conditions, the water vapor
diffusion flux was significantly enhanced by the partial derivative of matric pressure head effects
on relative humidity with respect to temperature. Hence, this phenomenon appeared to be
important at the opposite spectrum of the important soil moisture content region of the sensible
heat by liquid water movement. The current proposed improved heat flux theory was established
from the work of de Vries (1958), Cass et al. (1984), and Ho and Webb (1999). Moreover, the
improved heat flux equation would not be limited by the constraint set forth by Cass to only
apply to experimentation work between the relatively dry to moderately wet porous media
conditions. This is because the sensible heat by liquid water movement could account for such
limitation, which is one of three phenomena proposed for the theory and equation improvement.
Also, Shokri et al. (2009) have shown the significance of liquid water movement in evaporation
experiments to account for the excessive vapor flux observed.
Although the vapor enhancement factor had been used by numerous researchers, e.g.
Milly (1982), Shurbaji and Phillips (1995), Noborio et al. (1996), Nassar and Horton (1997),
Saito et al. (2006) and Heitman et al. (2008), until now, an equation to represent its physical
phenomenon has not yet been derived in its fundamental form. The segregation of mechanism is
an important process to allow full understanding of the basic factor governing the phenomena.
As a result, the mathematical description of water vapor volume expansion-advection is
proposed because it represents a more conceivable phenomenon of vapor flux than the
mechanistic enhancement factor imposed-correction on the diffusion equation of water vapor.
The water vapor volume expansion-advection mathematical relation is a function of soil
70
volumetric water content, and air pressure. Further investigation would be needed to derive a
theoretical equation for the water vapor relative permeability that would be able to predict at
saturated soil moisture content. Also, the effect of different average temperatures on water vapor
flux should be included into the equation.
71
Figure 4.1. Water vapor thermal conductivity, and the combined (wet porous media heat
conduction and sensible heat by liquid water movement) thermal conductivity (at 0r oP P P )
versus soil volumetric water content. Note: the water vapor thermal conductivity was at a
pressure ratio of 1.0, i.e. atmospheric pressure, and temperature at 32.5 oC. The water vapor
thermal conductivity was corrected for the pressure dependence of the thermal conductivity of
air, based on the equation given by Cass.
0.2
0.6
1
1.4
0.00
0.10
0.20
0.30
0.40
0.00 0.10 0.20 0.30
Co
mb
ined
th
erm
al c
on
du
ctvi
ty a
t P
r=0
(J
m-1
s-1K
-1)
Wat
er v
apo
r th
erm
al c
on
du
ctiv
ity
(J m
-1s-1
K-1
)
Volumetric water content (m3 m-3)
Pr=1.0, T=32.5 Celsius
Without water vapor (Pr=0)
72
(a)
(b)
Figure 4.2. The curve-fitting results of permeability variables as a function of the soil volumetric
water content, and the pressure ratio. The permeability variables based on the adapted-improved
heat flux theory for (a) the lysimeter sand, and (b) the Portneuf silt loam. Note: the graph (a) and
(b) were subjected to Eq. 4.20, before curve-fitting using Eq. 4.22 to determine the permeability
variables. Zero permeability variables values were added as supposed zero water vapor volume
expansion-advection at fully saturated soil moisture content to improve the curve-fitting results.
73
(a)
(b)
Figure 4.3. The comparison of the experimental and curve-fitted permeability variables based on
the adapted-improved heat flux theory for (a) the lysimeter sand, and (b) the Portneuf silt loam.
Note: the graph (a) and (b) were subjected to Eq. 4.20, before curve-fitting using Eq. 4.22 to
determine the permeability variables. Care should be placed on the fully saturated soil moisture
content condition because the permeability variables estimated by Eq. 4.22 could result in
negative values.
0.0E+00
2.5E-06
5.0E-06
7.5E-06
0.0E+00 2.5E-06 5.0E-06 7.5E-06
Pre
dic
ted
kpk r
/ko
Actual data kpkr/ko
R2 = 0.927No. data = 68
0.0E+00
5.0E-05
1.0E-04
1.5E-04
0.0E+00 5.0E-05 1.0E-04 1.5E-04
Pre
dic
ted
kpk r
/ko
Actual data kpkr/ko
R2 = 0.978No. data = 28
74
(a)
(b)
Figure 4.4. (a) The relative permeability ( rk ) of lysimeter sand versus water saturation estimated
using Eq. 4.24, and (b) Enlarged graph on relative permeability value ranged 0-0.3. Note: the
other relative permeability models were van Genuchten-Mualem, van Genuchten-Burdine,
Corey, and Falta referring to Eqs. 4.25-4.28, respectively.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0.0 0.2 0.4 0.6 0.8 1.0
Rel
ativ
e p
erm
eab
ility
, kr
(dim
en
sio
nle
ss)
Water saturation, θL/θs (m3/m3)
van Genuchten-Mualemvan Genuchten-BurdineCorey
Falta
Estimated relativepermeability
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.0 0.2 0.4 0.6 0.8 1.0
Rel
ativ
e p
erm
eab
ility
, kr
(dim
en
sio
nle
ss)
Water saturation, θL/θs (m3/m3)
75
Chapter 5. The extended Darcy law and water vapor relative permeability
derivation in unsaturated soils
5.1 Abstract
This study re-examined the relationship between water vapor relative permeability and soil
moisture saturation. The permeability of water vapor was found to increase with increasing soil
moisture saturation at lower water saturation levels. However, from near mild levels of soil
moisture saturation, the permeability was found to decrease with the increasing soil moisture
saturation. The water vapor relative permeability equation was derived to govern this
observation. Two types of equations used in the derivation were – the integral equations, and the
empirical equation from van Genuchten. The derived equation could adequately describe the
shape of the curvature and its peak location with reasonable accuracy, with an R-squared value
greater than 0.93 for both Portneuf silt loam and lysimeter sand, and greater than 0.8 for Kanto
loamy soil. The vapor enhancement factor imposed on Fick’s law was limited to the effects of
soil moisture saturation, whereas Darcy law could be extended to account for and validate the
ambient pressure and temperature; where very complex environmental conditions could be
considered. In this study, Darcy law was proposed as a replacement for the vapor enhancement
factor imposed-Fick’s law which resulted from the temperature gradient.
5.2 Introduction
Permeability is an important factor for gas, oil and water flow in porous media such as rocks,
industrial products, and soils. Conventionally, gas permeability, under fully saturated gas
concentration, decreases with increasing wetting phase saturation, e.g. liquid water and water
76
vapor respectively as wetting and non-wetting phases. In Chapter 4, the non-wetting phase
relative permeability, e.g. water vapor, appeared to have behaved differently from conventional
gas permeability (Goh and Noborio, 2016). The water vapor relative permeability was derived
from the thermal conductivity data published by Cass et al. (1984). The thermal conductivity
data obtained from the experiment, under which the closed steel tubes subjected to external heat
energy and pressure, was measured at different concentrations of water saturation. The condition
successfully allowed the removal of any external source of the influx of water vapor. Therefore,
the heat flux theory could estimate the thermal conductivity value that can quantify the water
vapor flux. The heat flux theory employed in Cass et al. (1984) was modified from de Vries
(1958). The experimental procedure used was adapted from Parikh et al. (1979), and the
mathematical methods applied to approximate the thermal conductivity data were adopted from
Riha et al. (1980).
In Chapter 4, an improved heat flux theory was proposed. The theory, accompanied by an
empirical relative permeability equation, was put forward to describe the observed relative
permeability data. With regards to the benefits and limitations, discussed in Chapter 4, the
equation successfully captured the curved downward characteristic of relative permeability with
increasing soil moisture saturation under various pressure conditions. However, the equation
failed at fully saturated soil moisture contents. Clearly, more research effort is necessary to
derive a theoretical equation that can fully describe the governing phenomenon. Successful
achievement of this would be a significant step in completing the theory of water vapor flux and
also help to solve the mysteries of the water vapor enhancement factor that have been puzzling
the soil physics community for more than 50 years, since the work of Philip and de Vries (1957).
Also, Horton and Ochsner (2011) acknowledged that the existing theory could not explain the
77
phenomenon that led to large different thermal vapor enhancement factors in soils.
The relative permeability phenomenon, in general, could be governed by capillary,
statistical, empirical and network models (Dullien, 1992, Honarpour et al., 1986). Fatt and
Dykstra (1951) derived an integral equation based on the capillary approach utilizing a bundle of
capillary tubes of various diameters with a fluid path length greater than the sample length. The
equation by Fatt and Dykstra (1951) was similar to Purcell (1949), and was described in Saraf
and McCaffery (1985). Childs and Collis-George (1950) employed a statistical method based on
the probability of two pore sizes encountering each other and subjected the relationship to
integration to generate an effective permeability function in the form of summation or
integration. Marshall (1958), Millington and Quirk (1961) and Mualem (1976) also used a
similar approach. Burdine (1953) derived the integral equation through the hydraulic radius
theory, and Bear (1972) showed an alternative derivation method based on Burdine’s integral
equation. In empirical models, the relative permeability is derived from the combination of either
the capillary or statistical model with an empirical equation to obtain the best approximation of
relative permeability. For instance, van Genuchten (1980) used the statistical model from
Mualem (1976) and proposed a modified equation from Haverkamp et al. (1977) to derive a new
closed-form empirical equation. Corey (1954), Brooks and Corey (1966), Campbell (1974) and
Li and Horne (2006) also utilized a comparable approach. In the present study, we limit the
investigation to the integral equations from the capillary and statistical models, and the empirical
equation used by van Genuchten (1980) which is widely claimed to generate a good compliance
with experimental data. Also, we explored the possibility of using Darcy law to govern water
vapor flux, rather than the Fick’s law alone or the combination of Fick’s law and Darcy law.
78
5.3 Materials and methods
The study was based on the experimental data from Cass et al. (1984). The experimental setup,
parameter estimation, and limitations were thoroughly discussed in Chapter 4. Some other
details, e.g. thermal conductivity data of lysimeter sand and Portneuf silt loam, and also matric
suction data of lysimeter sand, can be retrieved from Cass et al. (1981) and Cass et al. (1984).
The Portneuf silt loam permeability was derived from Robbins (1977). For comparison, this
study included a Kanto loamy soil dataset from Tanimoto (2007), and the thermal conductivity
was measured with the dual probe heat pulse (DPHP) method. The derivation of the improved
heat flux theory, including those from Cass et al. (1984), and the proposed temporary empirical
equation for water vapor relative permeability was discussed in detail in Chapter 4.
5.3.1 The derivation of extended Darcy law, and its relative permeability and the
Klinkenberg effect
As described by Brooks and Corey (1964), the Navier-Stokes equation could be reduced to a
relationship between pressure force and viscous drag force by assuming that the fluid flow in a
small diameter tube (Fig. 5.1) is incompressible, and has negligible inertia and body force:
' ' '
1 1P u u ur r
x r r r r x x
(5.1)
With further assumptions, such as the symmetrical tube and fluid flow in steady state, i.e.
no variation of flow velocity in 'x -direction (constant velocity profiles 1 and 2), the equation
would be described as:
'
dP d dur
dx r dr dr
(5.2)
where P is the pressure imposed on the fluid system (kg m-1 s-2), is the dynamic viscosity of
79
the fluid (kg m-1 s-1), u is the fluid velocity (m s-1), and r is the radius of the tube where the
fluid flow (m). To solve Eq. 5.2, two times integration was used. The first integration was
derived from the center of the tube where the tube radius was 0r to an unspecified location of
r r . Then Eq. 5.2 would be expressed as:
' 2
dP r du
dx dr
(5.3)
In the second integration, Eq. 5.3 was integrated with respect to the tube radius from an
unspecified location r r to the edge of the tube where or r . Then, Eq. 5.3 is described as:
2 2
' 4o
o
r rdPu u
dx
(5.4)
where ou is the velocity of the fluid at the wall of the tube (m s-1), i.e. slip effect. In the presence
of the fluid slip effect at the edge of the tube, the variable ou became significant. The following
relation was used in Klinkenberg (1941):
o
o
r r
duu c
dr
(5.5)
where c is the proportionality constant, is the mean free path (m), and the derivative is the
velocity gradient (s-1) at the wall of the tube. Putting Eq. 5.5 into Eq. 5.4, and using Eq. 5.3 to
find the velocity gradient at or r , the fluid flow velocity at a location ( r) between 0 and or
could be estimated by:
2 2
'
12
4o o
dPu c r r r
dx
(5.6)
Equation 5.6 is similar to the equation derived by Klinkenberg (1941). The volumetric
flow rate ( q , m3 s-1) of fluid could be estimated by multiplying Eq. 5.6 with 2rdA rdr and
80
then integrating the function with respect to the radius from 0 to or . The volumetric flow rate
would be:
4
'
4 11
8o
o
r c dPq
r dx
(5.7)
Equation 5.7 without the gas slip effect was known as Hagen-Poiseuille’s equation
(Sutera and Skalak, 1993). Moreover, based on the mathematical and theoretical description of
Atkins and De Paula (2006), the mean free path ( ) could be derived through the consideration
of molecules’ mean speed, molecules’ relative mean speed, collision frequency, number of
molecules, volume under consideration, collision cross section ( , m2), and the ideal gas law
equation. The resulting mathematical description is described as:
2
kT
P
(5.8)
where k is the Boltzmann constant (kg m2 s-2 K-1), T and P are the temperature (K) and
pressure (kg m-1 s-2) of the fluid in the tube, respectively. Atkins and De Paula (2006) claimed
that in a sample of constant volume; the pressure was proportional to the temperature, i.e. the
ratio of /T P would remain constant when the temperature increases. Hence, the would be
free from the temperature of a sample of fluid. In the current study, we claimed that the ratio of
temperature to pressure was not entirely constant. The observation from the results of the present
study would be used to support the claim. Hence, a constant b would be used to represent the
variables:
4
2 o o
ckb
r P (5.9)
where b is the Klinkenberg gas slip factor (K-1) that was slightly different from the one
employed in Klinkenberg (1941) and Tanikawa and Shimamoto (2006). Li and Horne (2004)
81
used a different constant derived from Rose (1948). In this study, Eq. 5.9 was sufficient, and
hence Eq. 5.7 could be rewritten as:
4
'
11
8o
r
r dPq bTP
dx
(5.10)
where r oP P P with oP is 101325 kg m-1 s-2 and P is the absolute pressurized air (kg m-1 s-2).
From a single pore space (see the right-hand side of Fig. 5.1), to multiple pores (see the left-hand
side of Fig. 5.1), the total flux (Q , m3 s-1) of fluid could be assumed as the sum of N tubes of
small diameter, hence, Q Nq . Knowing that ' cosx x , and then assuming the tortuosity as
the ratio of the actual travel distance of fluid to the sample path, 'x x , and also taking
porosity as 2 ' 2o oN r x Ax N r A (Dvorkin, 2009), then the total flux of fluid would be
written as:
2
21
8o
r
r A dPQ bTP
dx
(5.11)
Since the mean velocity fluid flowing in the porous media would be given by u Q A ,
and the relative permeability of fluid was provided by the ratio of effective permeability to
saturated permeability as r ok k k , where 2 28ok r , Eq. 5.11 was written as:
1r or
k k dPu bTP
dx (5.12)
Equation 5.12 could be used to govern the water vapor flux:
1rnw o v vv r
v v
k k Pdu bTP
dx
(5.13)
where the subscript v is referring to water vapor. Using the ideal gas law relationship to
governing the water vapor as v v vP RT M , and the improved gas Klinkenberg permeability
82
that includes the effect of temperature 1P T o rk k bTP , Eq. 5.13 was written as:
rnw P T vv
v v
k k R dTu
M dx
(5.14)
where vM is the molecular weight of water vapor (kg mol-1), R is the ideal gas constant (kg m2
s-2 K-1 mol-1), vP is the partial pressure of water vapor (kg m-1 s-2), rnwk is the relative
permeability of water vapor (dimensionless), P Tk is the improved water vapor Klinkenberg
permeability (m2) and v is the dynamic viscosity of water vapor (kg m-1 s-1). Then, Eq. 5.14
extended from the mean velocity of water vapor ( vu , m s-1) to vq (kg m-2 s-1) in porous media,
where v v Lu q , would be given by:
rnw P T v Lv
v v
k k R dTq
M dx
(5.15)
where L is the density of liquid water (kg m-3). The necessity of converting Eq. 5.14 to 5.15
will be explained in a later section 5.3.3.
Equation 5.15 was a derivation of the extended Darcy law, except for the pressure ratio to
the power e , used in Chapter 4 to govern water vapor flux after the Fick’s law. Moreover, the
derivation validated the mathematics employed in Chapter 4. Hereafter, the derived rk variable
from Eqs. 5.11 to 5.12 could be examined further to obtain the integral form for practical
application in estimating the relative permeability value.
5.3.2 The integral form of relative permeability of liquid water and water vapor
In unsaturated soil, the volumetric water content ( L , m3 m-3) was smaller than the soil porosity
given in Eq. 5.11. At the residual water content ( r , m3 m-3), liquid water became immobile. The
83
portion of soil porosity that contributed to liquid water flux would be given by L r . The
equation can be related as ( ) ( )L r s r eS , where s (m3 m-3) is the saturated water
content and eS (dimensionless) is the water saturation. Hence, the effective permeability from
Eq. 5.12 would be given by:
2
2
( )
8s r
o ek r S
(5.16)
If the derivative of the effective permeability ( k , m2) was taken with respect to the water
saturation, and it was followed by integrating differential dk from zero to k , which corresponds
to integrating differential edS from zero to a value of water saturation ( eS ) occupied by the
wetting phase, Eq. 5.16 would be defined as:
2
20
( )
8
eSs r
o ek r dS
(5.17)
Based on the capillary law the matric suction ( m , m) is inversely proportional to pore
radius, i.e. o mr a as given in Mualem (1976), then Eq. 5.17 could be converted into:
2
220
( )
8
eSs r e
m e
a dSk
S
(5.18)
A different form of the capillary equation was provided by Purcell (1949), and ultimately
it only caused variation in the constant a value which would not affect the estimation of the
relative permeability. The relative permeability was given by the ratio of the effective
permeability to the saturated permeability. The saturated permeability could be derived by
following the same steps as those used in deriving Eq. 5.18, except this time the differential dk
would be integrated from zero to sk which corresponds to the differential edS that would be
integrated from zero to unity. Therefore, the relative permeability resulted in the following form:
84
220
1
220
1
1
eSe
m e
rwe
m e
dS
Sk
dS
S
(5.19)
where rwk is known as the wetting phase relative permeability in which the wetting phase refers
to liquid water in porous media. If the variable could be assumed not to vary with water
saturation ( eS ), the following Eq. 5.20 could be derived, which is similar to those from Purcell
(1949) and Gates and Lietz (1950):
20
1
20
eSe
m e
rwe
m e
dS
Sk
dS
S
(5.20)
If the ratio of unsaturated to saturated tortuosity was assumed as a function of water
saturation to the power of two, Eq. 5.19 resulted in the following relationship:
20
2
1
20
eSe
m e
rw ee
m e
dS
Sk S
dS
S
(5.21)
Equation 5.21 is widely known as Burdine’s integral equation (Burdine, 1953). If the
tortuosity squared was assumed as inversely proportional to the pore radius, Eq. 5.19 could be
written as:
30
1
30
eSe
m e
rwe
m e
dS
Sk
dS
S
(5.22)
Equation 5.22 was derived by Fatt and Dykstra (1951), which they found it to have good
85
agreement with experimental data.
Equation 5.16, which was derived from Eq. 5.11, was due to the consideration of a
parameter, i.e. porosity. Another parameter that could be considered in the study is specific
surface area per unit volume ( vS , m-1). The equation could be written as
'(2 ) 2v o oS N r x Ax r . By placing the equation into the effective permeability from Eq.
5.11, and followed by a simple rearrangement, the new relationship could be rewritten as
2 22o vk N r S A . One can use the relation
2 ' 'o o or r r where the square of or is
assumed to be comprised of two different pore spaces. Then, by subjecting the equation to a
similar mathematical operation from Eqs. 5.16 to 5.19, but taking two times differentiation
followed by two times of integration of the effective permeability with respect to water
saturation, the new relative permeability would be:
'
2 '0 0
'1 1
2 '0 0
1
1
e eS Se e
v m e m e
rw
e e
v m e m e
dS dS
S S Sk
dS dS
S S S
(5.23)
If the tortuosity and the specific surface area could be assumed as a function of water
saturation to the power of n , based on Mualem (1976), Eq. 5.23 could be reduced to the
following form:
2
0
1
0
eSe
m enrw e
e
m e
dS
Sk S
dS
S
(5.24)
Hence, it showed that the Mualem (1976) integral equation could be derived by a non-
statistical method. If the power of two was changed to 2.5, and the matric suction power
86
increases to two, and 1n , then the result would be the Timmerman (1982) equation.
In Chapter 4, it was found the water vapor relative permeability curved downward at
higher soil moisture saturation. The empirical equation proposed to govern the relative
permeability provided a good approximation of the experimental data from Cass et al. (1984).
However, the equation failed under fully water saturated conditions, especially when there was
insufficient data at the saturation conditions for estimation. Therefore, it is necessary to find a
better equation. The water vapor relative permeability curvature could be separated into two
parts. The first part would be at lower water saturation where relative permeability would be
limited by the availability of the water vapor itself. The second part would be at higher water
saturation, where the water vapor behaved like a typical non-wetting phase with relative
permeability decreasing as the water saturation increased. To address the curvature described, the
integral equations from Purcell (1949), Burdine (1953), Fatt and Dykstra (1951), Mualem (1976)
and Timmerman (1982) were respectively modified as:
1
2 20
2
1
20
e
e
Se e
Sm e m e
rnw
e
m e
dS dS
S Sk
dS
S
(5.25)
1
2 20
22
2
1
20
1
e
e
Se e
Sm e m e
rnw e e
e
m e
dS dS
S Sk S S
dS
S
(5.26)
1
3 30
2
1
30
e
e
Se e
Sm e m e
rnw
e
m e
dS dS
S Sk
dS
S
(5.27)
87
1
0
21
0
1
e
e
Se e
Sn m e m enrnw e e
e
m e
dS dS
S Sk S S
dS
S
(5.28)
1
2 20
2
1
20
1
e
e
Se e
Sm e m e
rnw e e
e
m e
dS dS
S Sk S S
dS
S
(5.29)
where rnwk is the non-wetting relative permeability which in this study was the water vapor in
porous media. Then, Eqs. 5.25 to 5.29 were solved by the van Genuchten (1980) water
characteristic curve equation, i.e. the relation of e mS , to derive a closed-form equation. The
derived equations from Eqs. 5.25 to 5.29 can be summarized as:
1 1
1 1 1 1
m m
nn m mrnw e e e ek S S S S
(5.30)
where n and m are parameter constants. The power n is the only indicator distinguished
between Eqs. 5.25 to 5.29. The power values as 0, 0.5, 1, 2 or n itself correspond to Purcell
(1949), Mualem (1976), Timmerman (1982), Burdine (1953) and Mualem (1976), respectively.
Equation 5.30 was used in Eqs. 5.32-5.34 to determine the parameter constants of n , m , and the
Klinkenberg slip factor ( b ) by curve-fitting.
5.3.3 Thermal conductivity governed by Fick’s law and Darcy law in heat flux equation
The original Cass heat flux equation was employed with only Fick’s law (Eq. 5.31). The
improved-Cass heat flux equation included both the Fick’s law and Darcy law (Eq. 5.32). In
addition to these two laws, in Chapter 4 a heat flux equation was proposed that included three
88
other phenomena (Eq. 5.33). In Chapter 5, the heat flux equation was examined for improvement
to include only Darcy law (Eq. 5.34). The four equations can be expressed respectively as:
*s
h a o m
d dTq D L h
dT dx
(5.31)
*rnw P T v L s
h o a o m
v v
k k R d dTq L D L h
M dT dx
(5.32)
*m rnw P T v L
L L o v
v v
h
m sa o v s m
d k k Rc T K L c T
dT M dTq
dxhD L c T h
T T
(5.33)
*m rnw P T v L
h L L o v
v v
d k k R dTq c T K L c T
dT M dx
(5.34)
where hq is the total heat flux density (J m-2 s-1), is the vapor enhancement factor, * is the
thermal conductivity of the porous media in the presence of liquid water (J m-1 s-1 K-1), oL is the
latent heat (J kg-1) of water at temperature oT (K), vc is the specific heat of water vapor at
constant pressure (J kg-1 K-1), Lc is the specific heat of liquid water (J kg-1 K-1), D is the
molecular diffusivity of water vapor in air (m2 s-1), is the tortuosity factor (dimensionless), a
is the volumetric air content of the media (m3 m-3), s is the saturated water vapor density (kg m-
3) and mh is the matric pressure head effects on relative humidity (dimensionless), K is the
hydraulic conductivity (m s-1), m is the matric pressure head (m), and T is the temperature
difference between the outside and the middle of the closed steel tube. The first term of Eqs.
5.31-5.34 is the thermal conductivity of the porous media in the presence of liquid water, and the
second term of Eqs. 5.33-5.34 is the thermal conductivity by liquid water movement. The third
89
term of Eqs. 5.33-5.34 and the second term of Eq. 5.32 is the thermal conductivity by water
vapor movement governed by Darcy law. The second term of Eq. 5.31, the third term of Eq. 5.32
and the fourth term of Eq. 5.33 is the thermal conductivity by water vapor movement governed
by Fick’s law.
5.4 Results and Discussion
5.4.1 Replacing Fick’s law with Darcy law
The water vapor thermal conductivity was found to decrease with increasing ambient pressure
( P ), which led to a decreasing pressure ratio as given by r oP P P (Fig. 5.2). The water vapor
thermal conductivity displayed a curved concave downward pattern with increasing water
saturation ( eS ). The water vapor thermal conductivity was also found to increase with rising
temperature (Figs. 5.2(b) to 5.2(a) corresponding to 22.5 and 32.5 oC, respectively). In Chapter 4,
the total water vapor thermal conductivity was separated into two terms – the thermal
conductivities by Fick’s law and Darcy law, as in Eqs. 5.32 and 5.33. In the Chapter 5 study,
Darcy law was introduced to replace the vapor enhancement factor while retaining Fick’s law to
govern water vapor diffusion. Figures 5.2(a) to 5.2(c) clearly show that the water vapor thermal
conductivity estimated by Darcy law equation was in close approximation to the experimentally
derived total water vapor thermal conductivity, while a fairly good approximation was also
demonstrated in Fig 5.2(d). The discrepancy between the experimentally derived total water
vapor thermal conductivity and that estimated by Darcy law reduced as the ambient pressure
increased. The decreased temperature also reduced the discrepancy as shown in the change
between Figs. 5.2(a) and 5.2(b). The relatively small discrepancy value suggests that the total
water vapor thermal conductivity could be described by Darcy law alone, as in Eq. 5.34.
90
Therefore, Darcy law could be used to replace the combined Fick’s law and Darcy law as applied
in Chapter 4 (i.e. Eqs. 5.32 and 5.33) and Fick’s law only used in Cass et al. (1984) (i.e. Eq.
5.31). Also, the vapor enhancement factor was found to be less than unity at volumetric water
content less than 0.059 and 0.15 m3/m3 for the Portneuf silt loam and Kanto loamy soil,
respectively. This below unity value from Eqs. 5.32 and 5.33 would generate inaccurate
estimates which would support the proposal to use Darcy law as a replacement for the heat flux
equation. Furthermore, replacing the Fick’s law temperature gradient term with Darcy law would
not affect the water vapor flux by the matric suction gradient or the moisture content gradient
term of the equation used to govern water and vapor flow in soil. Nassar and Horton (1997) and
Heitman et al. (2008) can be referred to for the partial differential equations of water and vapor
flux that differentiate the terms use for matric suction and temperature gradients. In fact, Fick’s
law was found largely inadequate to describe the temperature gradient resulting in water vapor
flux. Ho and Webb (1996) had estimated that, under a specific environmental condition, a value
of ten times the Fick’s law value was needed to account for the discrepancy from the total water
vapor flux due to the temperature gradient. Also, Cass et al. (1984) had found that the total water
vapor flux in lysimeter sand could be as large as sixteen times the value given by Fick’s law, and
twelve times in Portneuf silt loam. Clearly, the significant discrepancy observed for water vapor
flux due to temperature gradient governed by the Fick’s law equation indicates an insufficiency
and also suggests the need to propose an alternative governing equation. Consequently, the
Darcy law equation was proposed as an alternative replacement to describe the temperature
gradient-resulting water vapor flux. Hence, the heat flux equation would be given as in Eq. 5.34.
Also, the R-squared value obtained from including the water vapor relative permeability (Eq.
5.30) in the heat flux equations (Eqs. 5.32-5.34) was reasonably high. The R-squared value of the
91
proposed heat flux equation (Eq. 5.34, R2=0.933) for lysimeter sand was found to be comparable
to those from the improved-Cass (R2=0.932), and the Chapter 4 heat flux equations (R2=0.933).
The Portneuf silt loam R-squared value was around 0.971 for the three heat flux equations,
whereas the Kanto loamy soil was reasonably good at 0.849 from Eq. 5.34. The R-squared
results imply that Darcy law could practically replace the Fick’s law equation for the
phenomenon caused by the temperature gradient in the soil, without sacrificing much of the
accuracy of estimating the total water vapor flux.
5.4.2 The water vapor relative permeability, and the Klinkenberg permeability
The derived water vapor relative permeability ( rnwk ), i.e. Eq. 5.30, consisted of two parameter
constants – n and m . The constants m and n combined to affect the level of water saturation
that corresponds to the peak in the curve of permeability variables ( rnw P Tk k ). The results indicate
the water saturation peak value for lysimeter sand was less than that for Kanto loamy soil or
Portneuf silt loam, as shown in Fig. 5.3. This observation is consistent with their respective
values of intrinsic permeability of 111.09 10 , 122.91 10 and 134.52 10 m2. Thus, a soil with
high intrinsic permeability possibly indicates an early peak of the permeability variable than for a
soil with lower permeability. Most importantly, Fig. 5.3 illustrates the derived relative
permeability is able to capture zero permeability values at initial and fully saturated soil moisture
conditions. Furthermore, the Klinkenberg gas slip factor ( b ) from the Kanto loamy soil was
found to be greater than that for the Portneuf silt loam and lysimeter sand (Table 5.1). This factor
indicates that increasing the temperature and/or reducing the ambient pressure in the Kanto
loamy soil would have a greater impact on increasing the water vapor slip effect than in the
Portneuf silt loam or lysimeter sand. It was evident from Eq. 5.9 that soil with a greater
92
Klinkenberg gas slip factor ( b ) due to the greater proportional constant ( c ), enhanced the
velocity gradient ( or r
du dr
) of water vapor at the soil particle walls. The phenomena then lead
to increased slip velocity ( ou ), as given by Eq. 5.5. Hence, in theory the Kanto loamy soil could
be more susceptible to the water vapor slip effect as pressure and temperature vary than the
Portneuf silt loam or lysimeter sand.
The effect of including the temperature as an independent variable into the Klinkenberg
permeability ( P Tk ) resulted in an improvement in model prediction. The R-squared value for
lysimeter sand was found to improve from 0.915 to 0.933 (Fig. 5.4(b)), estimated by Eq. 5.34,
and also a similar improvement was observed in the estimations by Eqs. 5.32-5.33. The R-
squared value for the Portneuf silt loam was maintained at 0.971 (Fig. 5.4(a)) because the
thermal conductivity data at 32.5 degree Celsius was the only temperature measured by Cass et
al. (1984). A similar improvement to the lysimeter sand was also noted for the Kanto loamy soil.
Figures 5.4 shows that the proposed heat flux equation of Eq. 5.34, the derived relative
permeability of Eq. 5.30 and the Klinkenberg permeability from Eq. 5.15 were able to describe
the three soils behavior reasonably well at different soil temperatures, pressures, and water
saturations.
5.4.3 Using Darcy law in the thermal water vapor diffusivities
As stated earlier, the Darcy law equation could account for the observed water vapor flux and
could also replace Fick’s law equation. Darcy law could be used to govern the effect of volume
expansion-generating water vapor pressure gradient due to the imposed temperature gradient on
the soil porous media, as described in Chapter 4. Subsequently, it would require modification of
the parameter in one of the terms used in the governing equation of heat and mass balance of
93
liquid water and water vapor in soil media (Heitman et al., 2008, Nassar and Horton, 1997). The
parameter was originally based on Eq. 5.31, as given by:
a m sTv s m
L
D hD h
T T
(5.35)
where TvD is known as the thermal water vapor diffusivities (m2 s-1 K-1). Based on the improved-
Cass (Eq. 5.32) and the Chapter 4 (Eq. 5.33) heat flux equations, the thermal vapor diffusivities
can be summarized as:
rnw P T v a m sTv s m
v v L
k k R D hD h
M T T
(5.36)
In the current study, based on Eq. 5.34, the following expression was proposed:
rnw P T vTv
v v
k k RD
M
(5.37)
Equation 5.36 is an improvement on Eq. 5.35 because it could explain the mystery of the
vapor enhancement factor ( ). The drawback is that it had more mathematical and
computational demand than Eq. 5.35, and also a possible inaccurate estimation of vapor flux at
low water saturation, as pointed out earlier. On the other hand, Eq. 5.37 preserved the quality of
Eq. 5.36 so that it could describe the effects of ambient pressure, temperature, and water
saturation, and maintain its accuracy without the need for the complexity found in Eq. 5.36.
Hence, Eq. 5.37 appeared to be a better solution than Eqs. 5.35 and 5.36. In other words, the first
term of Eq. 5.36 could replace the second term.
5.4.4 Limitation, and potential area of improvement
The limitations of the Cass research work were already discussed in Chapter 4. Moreover, other
factors related to the present study could be considered for future improvement and investigation.
94
The findings of the following researchers corroborate this statement. The network model
considers the interconnectedness of the pore structure and the microscopic flow mechanisms as
described in Dullien (1992). Honarpour et al. (1986) compiled a list of factors that were found to
affect relative permeability – rock properties, wettability, saturation history, overburden pressure,
porosity and permeability, interfacial tension and density, viscosity, and the initial wetting phase
saturation. Also, in addition to that list, Saraf and McCaffery (1985) reported the gravitational
force relating to the Bond Number as a factor that could affect the relative permeability. While
all these factors are relevant, the water vapor flux as a function of water saturation, ambient
pressure and temperature, under the influence of temperature gradient, were successfully
addressed in the present study. Finally, it is important to stress that the mathematical solution
provided in this work was found to model the observed phenomenon well. The curve-fitted
parameters’ ( n , m and b ) exact values and their relative values from one soil to another may
not be their final values; simply because the alternative combination of parameter values, from
curve-fitting solution, could generate similar calculated output. Any experimental effort, in any
possible way, should be directed to estimate the parameter values. Nonetheless, the current
mathematical solution would be a significant step forward for future investigation.
5.5 Conclusions
The water vapor relative permeability relationship showing a concave downward curve pattern
with the increasing soil moisture saturation was successfully derived. The integral equation was
extended from the work of Purcell (1949), Burdine (1953), Fatt and Dykstra (1951), Mualem
(1976) and Timmerman (1982). The empirical equation and derivation method used was based
on the work of van Genuchten (1980). The Klinkenberg (1941) permeability relationship was
95
extended to include the effect of temperature so that both the ambient pressure and temperature
could be incorporated to account for the water vapor slip effect in the soils. Darcy law was found
to be able to replace Fick’s law in governing the water vapor flux in soils, based both on the R-
squared values and graphical examination. One of the advantages is that it could describe the
effect of temperature and pressure on the water vapor slip effect, whereas Fick’s law confined
this to a single unknown vapor enhancement factor which could not account for both the effects.
Also, mathematically and computationally Darcy law did not increase the complexity more than
that from Fick’s law, and it was simpler than the addition of Darcy law to Fick’s law as proposed
in Chapter 4. Moreover, the mathematical relationship derived in the present study should be
able to address challenging ventures such as that stated by Sakaguchi et al. (2009) about the need
to understand ambient pressure variation to reflect the regolith on planets such as Mars.
96
Figure 5.1. Schematic diagram of velocity profile due to fluid flow through a small tube
diameter in the porous media.
97
(a) (b)
(c) (d)
Figure 5.2. The water vapor thermal conductivity of lysimeter sand versus water saturation at (a)
32.5oC, (b) 22.5oC, (c) Portneuf silt loam versus saturation at 32.5oC, and (d) Kanto loamy soil
versus saturation at temperature ranged from 38.5 to 31.5oC. Note: the water vapor thermal
conductivity estimated by Darcy law was obtained by subtracting the water vapor thermal
conductivity predicted by Fick’s law from the total water vapor thermal conductivity. Equation
5.33 was used in the determination. The total water vapor thermal conductivity dataset derived
from Cass et al. (1984) at different pressure ratios, temperatures, and volumetric water contents.
0.00
0.10
0.20
0.30
0.40
0.00 0.20 0.40 0.60 0.80
Wat
er
vap
or
ther
mal
co
nd
uct
ivit
y
(J m
-1s-1
K-1
)
Water Saturation, Se (dimensionless)
Total, Pr=1.000 Darcy, Pr=1.000Total, Pr=0.530 Darcy, Pr=0.530Total, Pr=0.160 Darcy, Pr=0.160Total, Pr=0.083 Darcy, Pr=0.083
0.00
0.10
0.20
0.30
0.40
0.00 0.20 0.40 0.60 0.80
Wat
er v
apo
r th
erm
al c
on
du
ctiv
ity
(J m
-1s-1
K-1
)
Water Saturation, Se (dimensionless)
Total, Pr=1.000 Darcy, Pr=1.000Total, Pr=0.530 Darcy, Pr=0.530Total, Pr=0.160 Darcy, Pr=0.160Total, Pr=0.083 Darcy, Pr=0.083
0.000
0.100
0.200
0.300
0.400
0.00 0.20 0.40 0.60 0.80 1.00
Wat
er v
apo
r th
erm
al c
on
du
ctiv
ity
(J·m
-1·s
-1·K
-1)
Water Saturation, Se (dimensionless)
Total, Pr=1.000 Darcy, Pr=1.000Total, Pr=0.530 Darcy, Pr=0.530Total, Pr=0.160 Darcy, Pr=0.160Total, Pr=0.083 Darcy, Pr=0.083
0.000
0.100
0.200
0.300
0.400
0.20 0.40 0.60
Wat
er v
apo
r th
erm
al c
on
du
ctiv
ity
(J·m
-1·s
-1·K
-1)
Water Saturation, Se (dimensionless)
Total, Pr=1.000 Darcy, Pr=1.000
Total, Pr=0.495 Darcy, Pr=0.495
Total, Pr=0.203 Darcy, Pr=0.203
98
Figure 5.3. The permeability variables ( rnw P Tk k ) versus water saturation at the unity of pressure
ratio and 32.5 degree Celsius.
0.0E+00
1.0E-17
2.0E-17
3.0E-17
4.0E-17
5.0E-17
6.0E-17
7.0E-17
8.0E-17
0 0.2 0.4 0.6 0.8 1
k rn
wk p
-T(m
2)
Water saturation (Se)
Portneuf silt loam Lysimeter sand
Kanto loamy soil
99
(a)
(b)
(c)
0.0E+00
2.0E-05
4.0E-05
6.0E-05
8.0E-05
1.0E-04
1.2E-04
1.4E-04
1.6E-04k r
nw
k P-T
/ko
Data point
Data
Model Se-Pr-T
Model Se-Pr
0.0E+00
1.0E-06
2.0E-06
3.0E-06
4.0E-06
5.0E-06
6.0E-06
7.0E-06
8.0E-06
k rn
wk P
-T/k
o
Data point
Data
Model Se-Pr-T
Model Se-Pr
0.0E+00
2.0E-06
4.0E-06
6.0E-06
8.0E-06
1.0E-05
1.2E-05
k rn
wk P
-T/k
o
Data point
Data
Model Se-Pr-T
Model Se-Pr
100
Figure 5.4. The (a) Portneuf silt loam, (b) lysimeter sand, and (c) Kanto loamy soil permeability
variables ( rnw P T ok k k ) data estimated by Eq. 5.34 at different water saturations, temperatures,
and pressure ratios. Note: the data were curve-fitted by the rnwk and P Tk from Eqs. 5.30 and
5.14, respectively. The equation with and without the temperature as an independent variable in
the expression of P Tk would correspond to the model e rS T P and e rS P , respectively. From
the first to fourth wave-like curves were respectively the pressure ratios of 1, 0.53, 0.16 and
0.083, as in graph (a), and the each point of the rise and fall of the permeability variables of the
wave-like curve, from left to right, correspond to the increasing water saturation. The
experimental temperature was 32.5 degree Celsius. In graph (b), the first three wave-like curves
correspond to temperatures of 3.5, 22.5 and 32.5 degree Celsius at a pressure ratio of unity. The
following three wave-like curves correspond to the similar temperature range, but at the pressure
ratio of 0.53, and it continued until the fourth (or last) three wave-like curves at a pressure ratio
of 0.083. Graph (c) represented by first to fifth wave-like curves with respective pressure ratios
of 1.000, 0.705, 0.495, 0.308 and 0.203. The first wave-like curve permeability variables, from
left to right, corresponding to decreasing temperature (42.5 to 31.6 degree Celsius) and
increasing water saturation. The rest of the wave-like curves correspond to a similar range of
temperature and saturation conditions. The data on graphs (a) and (b) were derived from Cass et
al. (1984), while graph (c) was from Tanimoto (2007) derived using DPHP method.
101
Table 5.1 The parameter constants ( n and m ) and the Klinkenberg gas slip factor ( b ) of water
vapor relative permeability.
Soil type n m b (K-1) R-squared
Lysimeter sand 0.535 1.550 1.883 0.933
Portneuf silt loam 1.280 1.160 116.892 0.971
Kanto loamy soil 3.867 2.542 1310.853 0.849
Note: the permeability variables ( /rnw P T ok k k , dimensionless), calculated from the parameters
n , m and b above, should be multiplied by 10-7 to obtain its original value. Equation 5.34 was
used to determine the parameters.
102
Chapter 6. Summary and Conclusions
The aim of the study was to describe the mechanism governing the vapor enhancement factor, by
investigating in two case studies. The first case study involved the use of the vapor enhancement
factor as a calibration parameter to simulate liquid water, water vapor, and heat flow in
unsaturated soil. The second case study required a re-examination of the theory and equation
used to derive the governing equation of water vapor flux and its constitutive equation.
Therefore, this study was divided into two main stages to address the problem. The first stage of
investigation involved the extension of source code from the simulation of water infiltration in
unsaturated soil (Chapter 2) to simulate water vapor, and heat flow movement, in addition to
liquid water movement (Chapter 3) in unsaturated soil. Key information learned in Chapter 2
was employed in Chapter 3. Also, the outcome of the simulation in Chapter 3 highlighted some
potential areas to improve the uncertainty of the vapor enhancement factor, mainly due to its
undefined mechanism, which lead to the research in the second stage of the overall study. The
second stage proposed an improved heat flux equation, an alternative equation for the vapor
enhancement factor (Chapter 4) and identified a necessary correction for water vapor relative
permeability (Chapter 5). The alternative equation can replace the enhanced-vapor diffusion
equation with a newly derived water vapor relative permeability equation.
In Chapter 2, the source code to simulate water infiltration into an unsaturated soil was
validated using Philip’s semi-analytical dataset. The elementary effect method was used in
addition to the one-at-a-time method in the validation procedure, where the sign oscillation effect
could be identified, for example, the time-step size. A hypothetical case study set up to
investigate the discrepancy observed between the dataset and the simulation revealed that
103
parameters with a high sensitivity coefficient and a high uncertainty range have the greatest
influence on the simulation output. For example, the initial moisture content and time-step size
had the highest sensitivity coefficient and highest uncertainty range, respectively, but neither of
them had the greatest effect on the water infiltration. The factor with the most influence on
model output was spatial spacing size, which implies that the extension of the current source
code to simulate the heat and mass flow of liquid water and water vapor would require a fine
spatial size to generate a reliable simulation result.
The liquid water, water vapor and heat flow source code was developed by extending the
water infiltration source code used in Chapter 2. By using a fine spatial size a reliable simulation
with a low error was obtained, that was able to re-simulate the temperature and water content
distributions based on the experimental conditions used in Heitman et al. (2008). Calibrations
with multiple temperature boundaries was superior to a single temperature boundary because the
optimum determined pair of hydraulic conductivity and vapor enhancement factor values was
more widely applicable and with generally lower error than any other pairs. However, the
observed optimum pair varied with soil type and initial water contents used in the study. For
example, the silt loam at 0.10 and 0.20 m3 m-3 had different optimum pairs, and similarly
different optimum pairs were observed for different soil types at the same water content. The
uncertainty was refined by using different optimum pairs to achieve the required accuracy for
either temperature or water content distributions. The model accuracy deterioration, or increased
error, was enhanced with the simulation time, and with an increasing average temperature
gradient. A higher error in the reverse compared with the forward temperature boundaries
suggested the presence of an unaccounted phenomenon such as buoyancy force and/or water
retention hysteresis. The zeta coefficient used in the vapor enhancement factor ranged between 2
104
for the least error in temperature distribution and 6 for the water content distribution. The values
were obtained irrespective of whether it was based on a single soil type averaged over different
water contents, or averaged over different soils and water contents. The least error value for
saturated hydraulic conductivity was obtained at its original value, without any decrement
necessary. The weak compliance of the simulation at low and high water contents suggested the
need for future improvement with the Fayer and Simmons (1995) equation. Based on a
comparison of both the calibrated parameters, the vapor enhancement factor was the least
understood parameter. This may be due to the theoretical and experimental complexity, which
has deterred adequate investigation by researchers. Hence, the vapor enhancement factor was
investigated in Chapter 4, in particular its theoretical aspects.
In Chapter 4, an alternative equation, based on Darcy law, was proposed to govern the
water vapor enhancement factor and it was validated using published data from Cass et al.
(1984). As part of the solution, an improved heat flux equation was proposed with additional
mechanisms included. A rapid increment rate of thermal conductivity resulting from liquid water
movement sensible heat with increasing water saturation was observed when compared to the
wet porous media heat conduction. The other two added mechanisms were water vapor sensible
heat and the partial derivative of matric pressure head effects on the relative humidity with
respect to temperature. The improved heat flux equation enabled the derivation of an equation for
experimental conditions ranging from relatively dry to saturated moisture conditions, because the
equation accounted for the liquid water flow in wet porous media condition. The Darcy law
equation was proposed to govern the water vapor enhancement factor after considering the
simple mass balance equation, ideal gas law, and the water vapor advection by air volume
expansion. The simple mass balance equation based on the ratio of liquid water to water vapor
105
densities resulted in a clear expansion in estimated water vapor volume. Again, the ideal gas law
described the presence of water vapor volume expansion under the effect of temperature
increment. Hence, not only was water vapor advection caused by air volume expansion, but the
water vapor itself experienced volume expansion as it vaporized. The water vapor density used in
Fick’s law density gradient equation to govern the water vapor flux was right after the water
vapor volume expansion, and thus, an enhanced-Fick’s law equation was required to account for
the under prediction. Water vapor volume expansion-advection, which is regarded as the physical
phenomenon to govern the vapor enhancement factor, would be a more reasonable approach than
the enhanced-water vapor diffusion by Fick’s law. The equation described the effect of water
saturation and air pressure on water vapor flux reasonably well. However, the effect of average
temperature variation on the water vapor flux was not accounted for, and so an equation for
water vapor relative permeability was proposed as a temporary solution. Both the temperature
effect and relative permeability were addressed in Chapter 5.
In Chapter 5, the extended Darcy law equation was derived from Navier-Stokes equation
to obtain the Klinkenberg permeability and relative permeability. The derivation of Klinkenberg
permeability allowed the identification of a theoretical basis for including the average
temperature variable into the equation. Similarly, the derivation of relative permeability was
extended to derive water vapor relative permeability, which was able to describe the concave
downward curve with increasing soil moisture saturation. The integral and empirical equations
were based on the work of previous researchers, and the derivation method was solely based on
the work of van Genuchten (1980). The Fick’s law diffusion equation was found largely
inadequate for describing the observed water vapor flux. Thus, the Darcy law, water vapor
volume expansion-advection equation was proposed to replace the enhanced-vapor diffusion
106
equation. The newly derived Klinkenberg and water vapor relative permeability equations were
able to describe the lysimeter sand and Portneuf silt loam from Cass et al. (1984), and the Kanto
loamy soil from Tanimoto (2007).
107
References
Abu-Hamdeh, N.H. and R.C. Reeder. 2000. Soil Thermal Conductivity Effects of Density,
Moisture, Salt Concentration, and Organic Matter. Soil Science Society of America Journal
64: 1285-1290. doi:10.2136/sssaj2000.6441285x.
Atkins, P. and J. De Paula. 2006. Atkins’ physical chemistry. 8 ed.,WH Freman, New York.
Banimahd, S.A. and S. Zand-Parsa. 2013. Simulation of evaporation, coupled liquid water, water
vapor and heat transport through the soil medium. Agricultural Water Management 130:
168-177. doi:10.1016/j.agwat.2013.08.022.
Bear, J. 1972. Dynamics of fluids in porous media. Dover Publications, Inc., New York.
Bear, J., J. Bensabat and A. Nir. 1991. Heat and mass transfer in unsaturated porous media at a
hot boundary: I. One-dimensional analytical model. Transport in Porous Media 6: 281-298.
doi:10.1007/bf00208954.
Bilskie, J.R., R. Horton and K.L. Bristow. 1998. Test of a dual-probe heat-pulse method for
determining thermal properties of porous materials. Soil Science 163: 346-355.
Bittelli, M., F. Ventura, G.S. Campbell, R.L. Snyder, F. Gallegati and P.R. Pisa. 2008. Coupling
of heat, water vapor, and liquid water fluxes to compute evaporation in bare soils. Journal
of Hydrology 362: 191-205. doi:10.1016/j.jhydrol.2008.08.014.
Brooks, R.H. and A. Corey. 1966. Properties of porous media affecting fluid flow. Journal of the
Irrigation and Drainage Division 92: 61-90.
Brooks, R.H. and A.T. Corey. 1964. Hydraulic Properties of Porous Media. Colorado State
University, Fort Collins, Colorado.
Burdine, N. 1953. Relative permeability calculations from pore size distribution data. Journal of
108
Petroleum Technology 5: 71-78.
Burdine, N.T. 1953. Relative Permeability Calculations From Pore Size Distribution Data. Trans.
Am. Inst. Min. Metall. Pet. Eng. 198: 71-78. doi:10.2118/225-G.
Cahill, A.T. and M.B. Parlange. 1998. On water vapor transport in field soils. Water Resources
Research 34: 731-739. doi:10.1029/97WR03756.
Caldwell, M.M., T.E. Dawson and J.H. Richards. 1998. Hydraulic lift: consequences of water
efflux from the roots of plants. Oecologia 113: 151-161. doi:10.1007/s004420050363.
Campbell, G.S. 1974. A Simple Method for Determining Unsaturated Conductivity From
Moisture Retention Data. Soil Science 117: 311-314.
Campbell, G.S. 1985. Soil physics with BASIC: transport models for soil-plant systems.
Elsevier, Amsterdam; New York.
Campolongo, F., J. Cariboni and A. Saltelli. 2007. An effective screening design for sensitivity
analysis of large models. Environmental Modelling & Software 22: 1509-1518.
doi:10.1016/j.envsoft.2006.10.004.
Campolongo, F., A. Saltelli and J. Cariboni. 2011. From screening to quantitative sensitivity
analysis. A unified approach. Computer Physics Communications 182: 978-988.
doi:10.1016/j.cpc.2010.12.039.
Campolongo, F., S. Tarantola and A. Saltelli. 1999. Tackling quantitatively large dimensionality
problems. Computer Physics Communications 117: 75-85. doi:10.1016/S0010-
4655(98)00165-9.
Cary, J.W. 1963. Onsager's relation and the non-isothermal diffusion of water vapor. The Journal
of Physical Chemistry 67: 126-129. doi:10.1021/j100795a030.
Cary, J.W. 1964. An evaporation experiment and its irreversible thermodynamics. International
109
Journal of Heat and Mass Transfer 7: 531-538. doi:10.1016/0017-9310(64)90050-X.
Cary, J.W. 1965. Water flux in moist soil: thermal versus suction gradients. Soil Science 100:
168-175.
Cary, J.W. and S.A. Taylor. 1962. The interaction of the simultaneous diffusions of heat and
water vapor. 413-420, Soil Science Society of America Proceedings.
Cass, A., G.S. Campbell and T.L. Jones. 1981. Hydraulic and thermal properties of soil samples
from the Buried Waste Test Facility. Pacific Northwest Lab., Richland, WA (USA).
Cass, A., G.S. Campbell and T.L. Jones. 1984. Enhancement of Thermal Water Vapor Diffusion
in Soil. Soil Science Society of America Journal 48: 25-32.
doi:10.2136/sssaj1984.03615995004800010005x.
Celia, M.A., E.T. Bouloutas and R.L. Zarba. 1990. A general mass-conservative numerical
solution for the unsaturated flow equation. Water Resources Research 26: 1483-1496.
doi:10.1029/WR026i007p01483.
Childs, E.C. and N. Collis-George. 1950. The Permeability of Porous Materials. Proceedings of
the Royal Society of London A: Mathematical, Physical and Engineering Sciences 201:
392-405. doi:10.1098/rspa.1950.0068.
Chu-Agor, M.L., R. Muñoz-Carpena, G. Kiker, A. Emanuelsson and I. Linkov. 2011. Exploring
vulnerability of coastal habitats to sea level rise through global sensitivity and uncertainty
analyses. Environmental Modelling & Software 26: 593-604.
doi:10.1016/j.envsoft.2010.12.003.
Corey, A.T. 1954. The interrelation between gas and oil relative permeabilities. Producers
monthly 19: 38-41.
Cunningham, R.E. and R.J.J. Williams. 1980. Diffusion in gases and porous media. Plenum
110
Press, New York.
Davis, A., S. Kamp, G. Fennemore, R. Schmidt, M. Keating, K. Hoenke, et al. 1997.
Environmental Policy Analysis, Peer Reviewed: A Risk-Based Approach to Soil
Remediation Modeling. Environmental Science & Technology 31: 520A-525A.
doi:10.1021/es9725662.
De Roo, A.P.J. and R.J.E. Offermans. 1995. LISEM: a physically-based hydrological and soil
erosion model for basin-scale water and sediment management. Modelling and
Management of Sustainable Basin-scale Water Resource Systems (Proceedings of a
Boulder Symposium), IAHS Publication.
de Vries, D.A. 1958. Simultaneous transfer of heat and moisture in porous media. Eos,
Transactions American Geophysical Union 39: 909-916. doi:10.1029/TR039i005p00909.
Dong, J., S.C. Steele-Dunne, T.E. Ochsner and N.v.d. Giesen. 2016. Estimating soil moisture and
soil thermal and hydraulic properties by assimilating soil temperatures using a particle
batch smoother. Advances in Water Resources 91: 104-116.
doi:10.1016/j.advwatres.2016.03.008.
Drouet, J.L., N. Capian, J.L. Fiorelli, V. Blanfort, M. Capitaine, S. Duretz, et al. 2011. Sensitivity
analysis for models of greenhouse gas emissions at farm level. Case study of N2O
emissions simulated by the CERES-EGC model. Environmental Pollution 159: 3156-3161.
doi:10.1016/j.envpol.2011.01.019.
Dullien, F.A.L. 1992. 5 - Multiphase Flow of Immiscible Fluids in Porous Media. In: F. A. L.
Dullien, editor Porous Media (Second Edition). Academic Press, San Diego. p. 333-485.
Dvorkin, J. 2009. Kozeny-Carman Equation Revisited Stanford University. p. 1-16.
Falta, R.W., I. Javandel, K. Pruess and P.A. Witherspoon. 1989. Density‐driven flow of gas in
111
the unsaturated zone due to the evaporation of volatile organic compounds. Water
Resources Research 25: 2159-2169.
Farouki, O.T. 1981. Thermal properties of soils. CRREL Technical Information Analysis Center
(TIAC).
Fatt, I. and H. Dykstra. 1951. Relative Permeability Studies. Trans. Am. Inst. Min. Metall. Pet.
Eng. 192: 249-256. doi:10.2118/951249-G.
Fayer, M.J. and C.S. Simmons. 1995. Modified Soil Water Retention Functions for All Matric
Suctions. Water Resources Research 31: 1233-1238. doi:10.1029/95WR00173.
Fick, A. 1995. On liquid diffusion. Journal of Membrane Science 100: 33-38. doi:10.1016/0376-
7388(94)00230-V.
Fox, G.A., R. Muñoz-Carpena and G.J. Sabbagh. 2010. Influence of flow concentration on
parameter importance and prediction uncertainty of pesticide trapping by vegetative filter
strips. Journal of Hydrology 384: 164-173. doi:10.1016/j.jhydrol.2010.01.020.
Gates, J.I. and W.T. Lietz. 1950. Relative permeabilities of California cores by the capillary-
pressure method. Drilling and Production Practice, American Petroleum Institute.
Goh, E.G. and K. Noborio. 2015. Sensitivity Analysis and Validation for Numerical Simulation
of Water Infiltration into Unsaturated Soil. International Scholarly Research Notices 2015:
7. doi:10.1155/2015/824721.
Goh, E.G. and K. Noborio. 2016. An Improved Heat Flux Theory and Mathematical Equation to
Estimate Water Vapor Advection as an Alternative to Mechanistic Enhancement Factor.
Transport in Porous Media 111: 331-346. doi:10.1007/s11242-015-0596-4.
Gosling, S.N. and N.W. Arnell. 2011. Simulating current global river runoff with a global
hydrological model: model revisions, validation, and sensitivity analysis. Hydrological
112
Processes 25: 1129-1145. doi:10.1002/hyp.7727.
Gran, M., J. Carrera, J. Massana, M.W. Saaltink, S. Olivella, C. Ayora, et al. 2011. Dynamics of
water vapor flux and water separation processes during evaporation from a salty dry soil.
Journal of Hydrology 396: 215-220. doi:10.1016/j.jhydrol.2010.11.011.
Hamby, D.M. 1994. A review of techniques for parameter sensitivity analysis of environmental
models. Environmental Monitoring and Assessment 32: 135-154.
doi:10.1007/BF00547132.
Hansson, K., J. Šimůnek, M. Mizoguchi, L.-C. Lundin and M.T. van Genuchten. 2004. Water
Flow and Heat Transport in Frozen Soil. Numerical Solution and Freeze–Thaw
Applications 3: 693-704. doi:10.2113/3.2.693.
Haverkamp, R., M. Vauclin, J. Touma, P.J. Wierenga and G. Vachaud. 1977. A Comparison of
Numerical Simulation Models For One-Dimensional Infiltration. Soil Science Society of
America Journal 41: 285-294. doi:10.2136/sssaj1977.03615995004100020024x.
Heitman, J.L., R. Horton, T. Ren, I.N. Nassar and D.D. Davis. 2008. A Test of Coupled Soil Heat
and Water Transfer Prediction under Transient Boundary Temperatures. Soil Science
Society of America Journal 72: 1197-1207. doi:10.2136/sssaj2007.0234.
Hiraiwa, Y. and T. Kasubuchi. 2000. Temperature dependence of thermal conductivity of soil
over a wide range of temperature (5–75°C). European Journal of Soil Science 51: 211-218.
doi:10.1046/j.1365-2389.2000.00301.x.
Ho, C.K. and S.W. Webb. 1996. A review of porous media enhanced vapor-phase diffusion
mechanisms, models, and data: Does enhanced vapor-phase diffusion exist? Sandia
National Laboratories, New Mexico, US.
Ho, C.K. and S.W. Webb. 1999. Enhanced Vapor-Phase Diffusion in Porous Media - LDRD
113
Final Report. Sandia National Laboratories, New Mexico, US.
Honarpour, M., F. Koederitz and A. Herbert. 1986. Relative permeability of petroleum reservoirs.
CRC Press, Boca Raton, Florida.
Horton, R. and T. Ochsner. 2011. Soil Thermal Regime. Handbook of Soil Sciences: Properties
and Processes. CRC Press. p. 1-24.
Jackson, R.D., D.R. Nielsen and F.S. Nakayama. 1963. On diffusion laws applied to porous
materials USDA-ARS, Washington, D.C.
Jury, W.A. and R. Horton. 2004. Soil physics. 6th ed.,John Wiley & Sons.
Kabala, Z.J. and P.C.D. Milly. 1990. Sensitivity analysis of flow in unsaturated heterogeneous
porous media: Theory, numerical model, and its verification. Water Resources Research 26:
593-610. doi:10.1029/WR026i004p00593.
Kelleners, T.J., J. Koonce, R. Shillito, J. Dijkema, M. Berli, M.H. Young, et al. 2016. Numerical
Modeling of Coupled Water Flow and Heat Transport in Soil and Snow. Soil Science
Society of America Journal 80: 247-263. doi:10.2136/sssaj2015.07.0279.
Khire, M.V., C.H. Benson and P.J. Bosscher. 1997. Water Balance Modeling of Earthen Final
Covers. Journal of Geotechnical and Geoenvironmental Engineering 123: 744-754.
doi:10.1061/(ASCE)1090-0241(1997)123:8(744).
Kimball, B.A., R.D. Jackson, R.J. Reginato, F.S. Nakayama and S.B. Idso. 1976. Comparison of
Field-measured and Calculated Soil-heat Fluxes. Soil Science Society of America Journal
40: 18-25. doi:10.2136/sssaj1976.03615995004000010010x.
Klinkenberg, L. 1941. The permeability of porous media to liquids and gases. Drilling and
production practice, American Petroleum Institute.
Kroener, E., A. Vallati and M. Bittelli. 2014. Numerical simulation of coupled heat, liquid water
114
and water vapor in soils for heat dissipation of underground electrical power cables.
Applied Thermal Engineering 70: 510-523. doi:10.1016/j.applthermaleng.2014.05.033.
Li, K. and R.N. Horne. 2004. Experimental Study of Gas Slippage in Two-Phase Flow. SPE
Reservoir Evaluation & Engineering 7: 409 - 415. doi:10.2118/89038-PA.
Li, K. and R.N. Horne. 2006. Comparison of methods to calculate relative permeability from
capillary pressure in consolidated water-wet porous media. Water Resources Research 42:
1-9. doi:10.1029/2005WR004482.
Lu, S., T. Ren, Z. Yu and R. Horton. 2011. A method to estimate the water vapour enhancement
factor in soil. European Journal of Soil Science 62: 498-504. doi:10.1111/j.1365-
2389.2011.01359.x.
Luckner, L., M.T. Van Genuchten and D.R. Nielsen. 1989. A consistent set of parametric models
for the two-phase flow of immiscible fluids in the subsurface. Water Resources Research
25: 2187-2193. doi:10.1029/WR025i010p02187.
Marshall, T.J. 1958. A Relation between Permeability and size Distribution of Pores. Journal of
Soil Science 9: 1-8. doi:10.1111/j.1365-2389.1958.tb01892.x.
Mason, E.A. and A. Malinauskas. 1983. Gas transport in porous media: the dusty-gas model.
Elsevier Science Ltd.
Massman, W.J. 2015. A non-equilibrium model for soil heating and moisture transport during
extreme surface heating: the soil (heat–moisture–vapor) HMV-Model Version 1. Geosci.
Model Dev. 8: 3659-3680. doi:10.5194/gmd-8-3659-2015.
McInnes, K.J. 1981. Thermal conductivities of soils from dryland wheat regions of Eastern
Washington. Washington State University, Washington.
Millington, R.J. and J.P. Quirk. 1961. Permeability of porous solids. Transactions of the Faraday
115
Society 57: 1200-1207. doi:10.1039/TF9615701200.
Milly, P.C.D. 1982. Moisture and heat transport in hysteretic, inhomogeneous porous media: A
matric head-based formulation and a numerical model. Water Resources Research 18: 489-
498. doi:10.1029/WR018i003p00489.
Milly, P.C.D. 1984. A Simulation Analysis of Thermal Effects on Evaporation From Soil. Water
Resources Research 20: 1087-1098. doi:10.1029/WR020i008p01087.
Milly, P.C.D. 1980. The coupled transport of water and heat in a vertical soil column under
atmospheric excitation. Massachusetts Institute of Technology, Massachusetts Institute of
Technology.
Min, C.H., Y.L. He, X.L. Liu, B.H. Yin, W. Jiang and W.Q. Tao. 2006. Parameter sensitivity
examination and discussion of PEM fuel cell simulation model validation: Part II: Results
of sensitivity analysis and validation of the model. Journal of Power Sources 160: 374-385.
doi:10.1016/j.jpowsour.2006.01.080.
Moghadas, D., K.Z. Jadoon, J. Vanderborght, S. Lambot and H. Vereecken. 2013. Effects of Near
Surface Soil Moisture Profiles During Evaporation on Far-Field Ground-Penetrating Radar
Data: A Numerical Study. Vadose Zone Journal 12. doi:10.2136/vzj2012.0138.
Morris, M.D. 1991. Factorial sampling plans for preliminary computational experiments.
Technometrics 33: 161-174. doi:10.2307/1269043.
Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous
media. Water Resources Research 12: 513-522.
Muskat, M. and R.D. Wyckoff. 1946. The flow of homogeneous fluids through porous media. J.
W. Edwards, In., Ann Arbor, Michigan, US.
Nakano, M. and T. Miyazaki. 1979. The diffusion and nonequilibrium thermodynamic equations
116
of water vapor in soils under temperature gradients. Soil Science 128: 184-188.
Nassar, I.N. and R. Horton. 1989. Water Transport in Unsaturated Nonisothermal Salty Soil: I.
Experimental Results. Soil Science Society of America Journal 53: 1323-1329.
doi:10.2136/sssaj1989.03615995005300050004x.
Nassar, I.N. and R. Horton. 1992. Simultaneous Transfer of Heat, Water, and Solute in Porous
Media: I. Theoretical Development. Soil Science Society of America Journal 56: 1350-
1356. doi:10.2136/sssaj1992.03615995005600050004x.
Nassar, I.N. and R. Horton. 1997. Heat, Water, and Solution Transfer in Unsaturated Porous
Media: I -- Theory Development and Transport Coefficient Evaluation. Transport in Porous
Media 27: 17-38. doi:10.1023/a:1006583918576.
Nathan, R., U.N. Safriel and I. Noy-Meir. 2001. Field Validation and Sensitivity Analysis of a
Mechanistic Model for Tree Seed Dispersal by Wind. Ecology 82: 374-388.
doi:10.1890/0012-9658(2001)082[0374:fvasao]2.0.co;2.
Noborio, K. and K.J. McInnes. 1993. Thermal Conductivity of Salt-Affected Soils. Soil Science
Society of America Journal 57: 329-334. doi:10.2136/sssaj1993.03615995005700020007x.
Noborio, K., K.J. McInnes and J.L. Heilman. 1996. Two-Dimensional Model for Water, Heat,
and Solute Transport in Furrow-Irrigated Soil: II. Field Evaluation. Soil Science Society of
America Journal 60: 1010-1021. doi:10.2136/sssaj1996.03615995006000040008x.
Novak, M.D. 2016. Importance of soil heating, liquid water loss, and vapor flow enhancement
for evaporation. Water Resources Research. doi:10.1002/2016WR018874.
Or, D. and J.M. Wraith. 2000. Comment on “On water vapor transport in field soils” by Anthony
T. Cahill and Marc B. Parlange. Water Resources Research 36: 3103-3105.
doi:10.1029/2000WR900124.
117
Pannell, D.J. 1997. Sensitivity analysis of normative economic models: theoretical framework
and practical strategies. Agricultural Economics 16: 139-152. doi:10.1016/S0169-
5150(96)01217-0.
Parikh, R.J., J.A. Havens and H.D. Scott. 1979. Thermal Diffusivity and Conductivity of Moist
Porous Media. Soil Science Society of America Journal 43: 1050-1052.
doi:10.2136/sssaj1979.03615995004300050047x.
Parlange, M.B., A.T. Cahill, D.R. Nielsen, J.W. Hopmans and O. Wendroth. 1998. Review of
heat and water movement in field soils. Soil and Tillage Research 47: 5-10.
doi:10.1016/S0167-1987(98)00066-X.
Penman, H.L. 1940. Gas and vapour movements in the soil: I. The diffusion of vapours through
porous solids. The Journal of Agricultural Science 30: 437-462.
doi:10.1017/S0021859600048164.
Pfletschinger, H., K. Prömmel, C. Schüth, M. Herbst and I. Engelhardt. 2014. Sensitivity of
Vadose Zone Water Fluxes to Climate Shifts in Arid Settings. Vadose Zone Journal 13.
doi:10.2136/vzj2013.02.0043.
Philip, J.R. and D.A. de Vries. 1957. Moisture movement in porous materials under temperature
gradients. Eos, Transactions American Geophysical Union 38: 222-232.
doi:10.1029/TR038i002p00222.
Poeter, E.P. and M.C. Hill. 1998. Documentation of UCODE: A computer code for universal
inverse modeling. DIANE Publishing.
Pruess, K. 1991. TOUGH2: A general-purpose numerical simulator for multiphase fluid and heat
flow.
Prunty, L. 2003. Soil Water Retention and Conductivity When Vapor Flow Is Important. Journal
118
of Irrigation and Drainage Engineering 129: 201-207. doi:10.1061/(ASCE)0733-
9437(2003)129:3(201).
Prunty, L. and R. Horton. 1994. Steady-State Temperature Distribution in Nonisothermal,
Unsaturated Closed Soil Cells. Soil Science Society of America Journal 58: 1358-1363.
doi:10.2136/sssaj1994.03615995005800050011x.
Purcell, W.R. 1949. Capillary Pressures - Their Measurement Using Mercury and the Calculation
of Permeability Therefrom. Trans. Am. Inst. Min. Metall. Pet. Eng. 1: 39-48.
doi:10.2118/949039-G.
Ren, T., Z. Ju, Y. Gong and R. Horton. 2005. Comparing Heat-Pulse and Time Domain
Reflectometry Soil Water Contents from Thermo-Time Domain Reflectometry Probes.
Vadose Zone Journal 4: 1080-1086. doi:10.2136/vzj2004.0139.
Richards, L.A. 1931. Capillary Conduction of Liquids through Porous Mediums. Journal of
Applied Physics 1: 318-333. doi:10.1063/1.1745010.
Riha, S.J., K.J. McInnes, S.W. Childs and G.S. Campbell. 1980. A Finite Element Calculation for
Determining Thermal Conductivity. Soil Science Society of America Journal 44: 1323-
1325. doi:10.2136/sssaj1980.03615995004400060039x.
Robbins, C.W. 1977. Hydraulic conductivity and moisture retention characteristics of southern
Idaho's silt loam soils. University of Idaho College of Agriculture.
Rose, C. 1968. Water transport in soil with a daily temperature wave. II. Analysis. Soil Research
6: 45-57. doi:10.1071/SR9680045.
Rose, W.D. 1948. Permeability and Gas-Slippage Phenomena. Production Practice and
Technology: 209-217.
Saito, H., J. Šimůnek and B.P. Mohanty. 2006. Numerical Analysis of Coupled Water, Vapor, and
119
Heat Transport in the Vadose Zone. Vadose Zone Journal 5: 784-800.
doi:10.2136/vzj2006.0007.
Sakaguchi, I., T. Momose, H. Mochizuki and T. Kasubuchi. 2009. Heat pipe phenomenon in soil
under reduced air pressure. European Journal of Soil Science 60: 110-115.
doi:10.1111/j.1365-2389.2008.01095.x.
Sakai, M., N. Toride and J. Šimůnek. 2009. Water and Vapor Movement with Condensation and
Evaporation in a Sandy Column. Soil Science Society of America Journal 73: 707-717.
doi:10.2136/sssaj2008.0094.
Sakaki, T., P.E. Schulte, A. Cihan, J.A. Christ and T.H. Illangasekare. 2013. Airflow Pathway
Development as Affected by Soil Moisture Variability in Heterogeneous Soils. Vadose
Zone Journal 12. doi:10.2136/vzj2011.0118.
Saltelli, A. and P. Annoni. 2010. How to avoid a perfunctory sensitivity analysis. Environmental
Modelling & Software 25: 1508-1517. doi:10.1016/j.envsoft.2010.04.012.
Saltelli, A., M. Ratto, T. Andres, F. Campolongo, J. Cariboni, D. Gatelli, et al. 2008. Elementary
Effects Method. Global Sensitivity Analysis. The Primer. John Wiley & Sons, Ltd. p. 109-
154.
Saltelli, A., S. Tarantola, F. Campolongo and M. Ratto. 2004. Global Sensitivity Analysis for
Importance Assessment. Sensitivity Analysis in Practice. John Wiley & Sons, Ltd. p. 31-
61.
Saltelli, A., S. Tarantola, F. Campolongo and M. Ratto. 2004. The Screening Exercise.
Sensitivity Analysis in Practice. John Wiley & Sons, Ltd. p. 91-108.
Saraf, D.N. and F.G. McCaffery. 1985. Chapter 4 Relative Permeabilities. In: G. V. C. Erle C.
Donaldson and Y. Teh Fu, editors, Developments in Petroleum Science. Elsevier. p. 75-118.
120
Scanlon, B.R. and P.C.D. Milly. 1994. Water and heat fluxes in desert soils: 2. Numerical
simulations. Water Resources Research 30: 721-733. doi:10.1029/93wr03252.
Sepúlveda, F.D., L.A. Cisternas and E.D. Gálvez. 2014. The use of global sensitivity analysis for
improving processes: Applications to mineral processing. Computers & Chemical
Engineering 66: 221-232. doi:10.1016/j.compchemeng.2014.01.008.
Shahraeeni, E. and D. Or. 2012. Pore scale mechanisms for enhanced vapor transport through
partially saturated porous media. Water Resources Research 48: n/a-n/a.
doi:10.1029/2011WR011036.
Shokri, N., P. Lehmann and D. Or. 2009. Critical evaluation of enhancement factors for vapor
transport through unsaturated porous media. Water Resources Research 45: 1-9.
doi:10.1029/2009wr007769.
Shurbaji, A.-R.M. and F.M. Phillips. 1995. A numerical model for the movement of H2O,
H218O, and 2HHO in the unsaturated zone. Journal of Hydrology 171: 125-142.
doi:10.1016/0022-1694(94)02604-A.
Sophocleous, M. 1979. Analysis of water and heat flow in unsaturated-saturated porous media.
Water Resources Research 15: 1195-1206. doi:10.1029/WR015i005p01195.
Stange, F., K. Butterbach-Bahl, H. Papen, S. Zechmeister-Boltenstern, C. Li and J. Aber. 2000. A
process-oriented model of N2O and NO emissions from forest soils: 2. Sensitivity analysis
and validation. Journal of Geophysical Research: Atmospheres 105: 4385-4398.
doi:10.1029/1999jd900948.
Sutera, S.P. and R. Skalak. 1993. The History of Poiseuille Law. Annual Review of Fluid
Mechanics 25: 1-19.
Tanikawa, W. and T. Shimamoto. 2006. Klinkenberg effect for gas permeability and its
121
comparison to water permeability for porous sedimentary rocks. Hydrology and Earth
System Sciences Discussions 3: 1315-1338. doi:10.5194/hessd-3-1315-2006.
Tanimoto, S. 2007. Development of a method for easy separation of latent heat transport from
the apparent heat conductivity of the soil. Meiji University, Japan.
Timmerman, E.H.H. 1982. Practical Reservoir Engineering. PennWell Corporation.
Tu, J., G.H. Yeoh and C. Liu. 2007. Computational fluid dynamics: a practical approach.
Butterworth-Heinemann.
Tuller, M. and D. Or. 2001. Hydraulic conductivity of variably saturated porous media: Film and
corner flow in angular pore space. Water Resources Research 37: 1257-1276.
doi:10.1029/2000WR900328.
van Genuchten, M.T. 1980. A Closed-form Equation for Predicting the Hydraulic Conductivity
of Unsaturated Soils. Soil Science Society of America Journal 44: 892-898.
doi:10.2136/sssaj1980.03615995004400050002x.
Vazquez-Cruz, M.A., R. Guzman-Cruz, I.L. Lopez-Cruz, O. Cornejo-Perez, I. Torres-Pacheco
and R.G. Guevara-Gonzalez. 2014. Global sensitivity analysis by means of EFAST and
Sobol’ methods and calibration of reduced state-variable TOMGRO model using genetic
algorithms. Computers and Electronics in Agriculture 100: 1-12.
doi:10.1016/j.compag.2013.10.006.
Vereecken, H., J. Maes and J. Feyen. 1990. Estimating unsaturated hdraulic conductivity from
easily measured soil properties. Soil Science 149: 1-12.
Webb, S. and K. Pruess. 2003. The Use of Fick's Law for Modeling Trace Gas Diffusion in
Porous Media. Transport in Porous Media 51: 327-341. doi:10.1023/A:1022379016613.
Winterkorn, H. and H.-Y. Fang. 1991. Soil Technology and Engineering Properties of Soils. In:
122
H.-Y. Fang, editor Foundation Engineering Handbook. Springer US. p. 88-143.
Wuest, S.B., S.L. Albrecht and K.W. Skirvin. 1999. Vapor Transport vs. Seed–Soil Contact in
Wheat Germination. Agronomy Journal 91: 783-787. doi:10.2134/agronj1999.915783x.
Zhang, S., J. Teng, Z. He and D. Sheng. 2016. Importance of vapor flow in unsaturated freezing
soil: a numerical study. Cold Regions Science and Technology 126: 1-9.
doi:10.1016/j.coldregions.2016.02.011.
Zheng, C. and G.D. Bennett. 2002. Applied Contaminant Transport Modeling. 2 ed.,John Wiley
& Sons, New York.
Zhou, J., J.L. Heitman, R. Horton, T. Ren, T.E. Ochsner, L. Prunty, et al. 2006. Method for
Maintaining One-Dimensional Temperature Gradients in Unsaturated, Closed Soil Cells.
Soil Science Society of America Journal 70: 1303-1309. doi:10.2136/sssaj2005.0336N.
123