Computers and Geotechnics 108 (2019) 107–116

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Research Paper

An analytical model for evaporation from unsaturated soil T

a a a,⁎ b a,c
Jidong Teng , Xun Zhang , Sheng Zhang , Chenjun Zhao , Daichao Sheng
a
National Engineering Laboratory for High-Speed-Railway Construction, Central South University, China
b
School of Naval Architecture, Ocean & Civil Engineering, Shanghai Jiaotong University, China
c
ARC Centre of Excellence in Geotechnical Science and Engineering, The University of Newcastle, Australia

ARTICLE INFO ABSTRACT

Keywords: Evaporation from unsaturated soil is characterized by vapor transfer in the upper part and liquid water transfer
Unsaturated soil in the lower part of the soil. This study develops an analytical model for identifying the vaporization plane where
Vaporization plane the evaporation occurs, and the model consists of three partial differential equations that respectively govern the
Evaporation rate vapor flow, liquid water flow and heat transfer. These equations are solved simultaneously for the transient
Analytical model
water content profile, evaporation rate, transient temperature profile and location of the vaporization plane. A
series of experiments are used to validate the proposed analytical model, which indicates that this model can
reasonably well predict the temporal water content profile and evaporation rate during the evaporation process.
The result shows that the evaporation rate during falling rate stage is proportional to the inverse of the square
root of elapsed time, and the proportionality is affected by the vapor diffusion coefficient, heat diffusion coef-
ficient, and critical water content. The depth of vaporization plane is found to be independent of soil hydraulic
properties, but only dependent on the heat diffusion coefficient of the soil. It is also revealed that heat diffusion
coefficient has a pronounced influence on the evaporation process, which has not been observed in previous
studies. A larger thermal diffusion coefficient leads to a faster advancing and a deeper vaporization plane, as well
as a faster decreasing evaporation rate. The analytical model provides a useful tool for investigating the me-
chanism of the evaporation process.

1. Introduction stage of evaporation and lasts a short time. During this period, the
actual evaporation rate (Ea) is primarily controlled by the atmospheric
Evaporation is an important process involved in many applications condition, and is approximately equal to the potential evaporation rate
of geotechnical and geo-environmental engineering and soil physics, (Ep, Fig. 1b). Stage 2 begins when the liquid water in the soil is not
such as soil cover systems in mines and landfills [45,4], soil salinization continuous, i.e., the vaporization plane begins to form and gradually
in arid and semi-arid regions [28,48,13,40], and long-term resistance to recedes to a deeper position in the soil [49,11,29,37]. During Stage 2,
desiccation cracks in pavements, embankments, dams and buildings water transfer occurs in form of the flow of liquid water (Zone 2,
[9,36,40,34]. Evaporative water loss significantly affects the hydro- Fig. 1c), as well as the diffusion of water vapor in the top dry layer
thermal properties of soil at a shallow depth. Therefore, understanding (Zone 1). As the top dry layer expands, the evaporation rate gradually
the mechanism of evaporation is important for geotechnical design and decreases and finally reaches an extremely low value [26,39].
damage prevention. To understand the mechanism of evaporation process, a number of
The evaporation process can be regarded as the process of liquid theoretical studies have been carried out, and a series of mathematical
water changing into vapor and flowing from the soil pores into the models have been proposed. These models can be generally divided into
atmosphere. It involves simultaneous heat and mass transfer, phase two categories: numerical models and analytical models. Many re-
change, and coupling of the soil-atmosphere boundary [17,2]. Many searchers have developed numerical codes for solving the coupled
studies that describe the drying behavior of soil have been published movement of heat and water in unsaturated soil based on the finite
[46,53,30,23,38,56]. In general, the evaporation process can generally element or finite difference methods. These methods are capable of
be divided into two stages: a relatively high and constant rate stage coupling with the atmospheric boundary to describe the evaporation
(Stage 1) followed by a lower and gradually decreasing evaporation (or infiltration) process [10,45,31,9,12,32]. When dealing with the
rate (Stage 2) as shown in Fig. 1a. Stage 1 usually occurs at the initial evaporation process, these numerical models either set a constant value

⁎
Corresponding author at: Railway Campus of Central South University, Shaoshan South Road No. 68, Changsha, Hunan Province 410075, China.
E-mail address: zhang-sheng@csu.edu.cn (S. Zhang).

https://doi.org/10.1016/j.compgeo.2018.12.005
Received 18 August 2018; Received in revised form 3 December 2018; Accepted 7 December 2018
0266-352X/ © 2018 Published by Elsevier Ltd.
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

process.
Many researchers have proposed various analytical or semi-analy-
tical mathematical models based on the Richards equation
[22,50,19,55,37,21,39]. These analytical solutions provide an effective
way to simulate the evaporation process, but they do little to in terms of
revealing the mechanism of evaporation process. Analytical models
usually treat the evaporation flux as a fixed value or equivalent to a
simple function. Few models can capture the transition from Stage 1 to
Stage 2 or identify parameters that affect this transition [35,50,15]. In
addition, most analytical models are used to solve the Richards equa-
tion, and the solutions are therefore only valid for the conditions in
which the liquid water in the soil is continuous. However, when the
vaporization plane forms and penetrates into soil, two distinct zones
emerge, as shown in Fig. 1. In this case, the analytical models based on
the Richards equation cannot be applied to the entire soil profile
[11,30]. Finally, few of the existing analytical models have been vali-
dated against laboratory or field tests [21,42]. Therefore, an analytical
model that can overcome the above deficiencies, i.e., having an explicit
physical meaning and a concise mathematical expression, is the key to
revealing the mechanism of evaporation.
The primary objective of this study is to develop a physics-based
analytical model that can reveal the evaporation mechanism. By
quantifying the movement of the vaporization plane, the physical
process in the soil profile can be formulated in three partial differential
equations that describe the liquid water flow, vapor diffusion and heat
transfer. The solution of the model presents the expressions of several
important variables, including the transient liquid water content dis-
tribution, evaporation rate at the soil surface, and the time-dependent
depth of the vaporization plan. The proposed model is then evaluated
by experimental data from the literature. Finally, some key input
parameters are analyzed, in attempt to better understand the me-
chanism of the evaporation process.

2. Materials and methods

2.1. Basic assumptions

According to our current understanding, the vaporization plane is

initially located at the soil surface and gradually moves downward
during evaporation. Relative to the position of vaporization plane, the
soil profile can be divided into the vapor diffusion zone (Zone 1) and
the liquid water migration zone (Zone 2), as shown in Fig. 1c. To
Fig. 1. Schematic diagram for the two processes of evaporation, (a) typical simplify the quantitative description of the evaporation process, the
evaporation curve moisture status at soil surface (after Teng et al. [39]), (b) following assumptions are made:
water content profile during Stage 1, and (c) water content profile during Stage
2. Ea and Ep are the actual and potential evaporation rate, respectively, and s(t) (1) The soil profile is a homogeneous and semi-infinite space.
is the depth of the vaporization plane. (2) In Zone 1, the pore water in the soil is not continuous. Compared to
the vapor diffusion, the liquid water flow in this zone is so minor
of evaporation flux, define a critical matric potential (or a characteristic that it can be neglected.
variable) at the onset of Stage 2, or impose a computational formula of (3) In Zone 2, the liquid water flow is assumed to be due to capillary
evaporation rate that is related to the soil status and atmospheric action. The domain of Zone 2 is assumed to be isothermal be-
conditions. These approaches can generate a reasonable result for the cause the liquid water flow driven by temperature gradient in
analysis of the evaporation problem at the field scale, for which these this zone is negligible compared to that driven by matric suction
cases have a lower requirement of computation accuracy. However, gradient [27,57]. Moreover, some measured results of tempera-
they can produce a considerable deviation from the evaporation results ture profile in literature show that Zone 2 can reach an iso-
at the laboratory scale or even pore scale [54,23]. The applicability and thermal state quickly [33]. This assumption indicates that the
complexity of these numerical models also vary. Some of them require temperature at the vaporization plane is consistent with the Zone
sophisticated algorithms to overcome convergence and mass con- 2 temperature.
servation problems, while others may suffer accuracy and applicability (4) During Stage 1 of evaporation, because the soil is sufficiently wet,
limits due to the adopted assumptions [3,39]. Therefore, we shift our the vaporization plane is located at the soil surface and does not
efforts to finding an analytical model to describe the evaporation move.

108
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

2.2. Governing equations T2 (s, t ) T1 (s , t ) ds (t )

k2 k1 = *L v
x x dt (5)
The water vapor concentration of Zone 1 can be defined as C(x, t) where Lv is the latent heat of vaporization (J/kg); k1 and k2 are the
(kg/m3), which is a function of position x and time t. The temperature volume average thermal conductivity (W/K * m) of Zone 1 and Zone 2,
of Zone 1 also changes with time and position; thus, it can be expressed respectively; and θ* is the critical volumetric water content at the va-
as T1(x, t) (K). The volumetric water content and temperature in Zone 2 porization plane (m3/m3).
are expressed as θ (m3/m3) and T2(x, t) (K), respectively. During the The boundary conditions of the evaporation process can be for-
evaporation process, the evaporation plane moves downward, and its mulated as follows:
position is defined as s(t) (m). Therefore, the range of Zone 1 is
0 < x < s(t), while the range of Zone 2 is x > s(t). C (0, t ) = C0 (6)
In Zone 1, the process of the vapor diffusion in the soil is described
C (s, t ) = Cs (7)
by Fick’s law. It has been reported that heat flux due to thermal con-
ductivity is always two orders of magnitude larger than those of the (s , t ) = * (8)
latent heat flux due to the vapor phase change in unsaturated soils
[27,18]. Therefore, the heat transfer is assumed to occur through heat ( , t) = (9)
conduction in Zone 1. The mass and heat movement are expressed as
T1 (0, t ) = T0 (10)
follows:
C (x , t ) 2 C (x , t) T1 (s, t ) = Ts = T2 (s, t ) (11)
= a1 0 < x < s (t )
t x2 (1)
T2 (x , 0) = T2 (x , t ) = T2 ( , t ) = T (12)
T1 (x , t ) 2T (x , t)
= a2 1
0 < x < s (t ) Eq. (6) uses a constant water vapor concentration C0 at the soil
t x2 (2) surface (kg/m3), which is balanced with the vapor concentration in the
In the equations above, a1 is the diffusion coefficient of the vapor in atmosphere. Eqs. (7) and (8) define the vapor concentration and critical
the soil (m2/s) and a2 is the average heat diffusion coefficient (m2/s) of water content at the vaporization plane as Cs (kg/m3) and θ* (m3/m3),
the volume [43]. respectively. Eq. (9) indicates that the water content at an infinite depth
In Zone 2, the Richards equation governs the liquid water flow in is θ∞ (m3/m3). Eqs. (10) and (11) show that the temperatures at the soil
the unsaturated soil. Finding an analytical solution to the Richards surface and at the vaporization plane are T0 (K) and Ts (K), respectively.
equation is challenging due to the different forms of the integrals or Eq. (12) shows that the initial temperature of the soil is T∞ (K), which is
series involved. One popular way to simplify the Richards equation is to equal to the temperature at Zone 2.
adopt an exponential formula for the soil-water characteristic curve and Eqs. (1)–(12) constitute the mathematical model that describes the
an exponential model of the unsaturated permeability [39,52]. The evaporation process of an unsaturated soil. The analytical solution to
relations are expressed as follows: this model can be obtained by simultaneously solving the four partial
differential equations with the boundary or initial conditions.
k ( ) = ks e (3a)

( )= r +( s r) e (3b) 2.3. Theoretical derivation

where ks is the saturated hydraulic conductivity coefficient (m/s); ψ is Eq. (2) is solved first. A general solution of Eq. (2) that satisfies the
the matric suction of the soil (kPa); α the desaturation coefficient in the boundary condition of Eq. (10) can be obtained, as Eq. (A8) in the
exponential model of the soil-water characteristic curve (1/m); and θs appendix. Substituting Eq. (11) into Eq. (A8) gives the following re-
and θr are the saturated water content and residual water content, re- lationship:
spectively (dimensionless).
The Richards equation can then be changed into a form of the wave Ts = T1 (s, t ) = T0 + merf
s
equation that is easy to solve, as follows: 2 a2 t (13)
2
(x , t )
= a3
(x , t )
x > s (t ) where m is a constant and erf(x) is the error function. Since the vari-
t x2 (4) ables Ts, T0 and m are constant, it is easy to describe the value of the erf
where a3 is the equivalent liquid water diffusion coefficient (m2/s); (x) function as a constant, i.e.,
which can be represented as a3 = ks/(α(θs−θr)). It is noted that the s
=
gravitational term in Richards equation is neglected in Eq. (4) for two 2 a2 t (14)
reasons. First, the gradient of the gravitational head is found to be
negligible compared to the strong matric suction gradient during eva- where λ is a constant. Substituting Eq. (14) into Eq. (13) and elim-
poration process [51,8]. Second, it is a challenge to obtain an analytical inating the parameter m, we can obtain the following:
solution of the Richards equation with gravitational term because many Ts T0 x
difficult forms of integrals or series will be involved. Even if the solution T1 = T0 + erf
erf ( ) 2 a2 t (15)
can possess a clear formulation, it would somehow diverge from the
mechanism established by the boundary conditions [39]. By substituting Eqs. (14) and (15) into the energy balance equation
There are four unknowns, namely C, θ, T1 and s, in the three gov- at the vaporization plane, i.e., Eq. (5), the expression of the critical
erning equations (Eqs. (1), (2) and (4)). A closed solution of these water content θ* can be obtained as follows:
equations requires additional information that describes the relation T0 Ts 1 2
among the variables. Here, the energy conservation must be satisfied at * = k1 e
erf ( ) a2 L v (16)
the vaporization plane; that is,

109
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

As for Zone 2, the solution of Eq. (4) satisfying the boundary con- when the water content at the soil surface decreases to θ* and the
ditions of Eqs. (8) and (9) can be obtained by referring to Eq. (A9) in the evaporation rate starts to gradually decrease.
appendix, as follows: In this model, Eqs. (15), (17) and (23) describe the temperature
profile of Zone 1, the water content profile of Zone 2 and the water
* x
= erfc vapor concentration profile of Zone 1, respectively. Eqs. (14) and (24)
erfc ( a2/ a3 ) 2 a3 t (17) present the evolution of the position of the vaporization plane and
When homogenizing Eq. (4), the exponential function of the soil- evaporation rate at the soil surface, respectively. The proposed model
water characteristic curve of the unsaturated soil is adopted. This has a simple form and is easy to compute, which provides a concise
function can be applied to the vaporization plane. Substituting the approach to gaining a better understanding of the mechanism of the
critical water content θ* into this exponential function gives the fol- evaporation process.
lowing relationship: As for the calculation process of the proposed analytical model. The
critical water content θ* is determined first. Then, Eq. (16) is applied to
*( ) = r +( s r) e (18) compute the intermediate variable λ. The water vapor concentration Cs
Applying the local thermal equilibrium hypothesis to the water- at the vaporization plane and the concentration profile C of the water
vapor interface [25], the relationship between the matric suction and vapor in Zone 1 are obtained by substituting θ* and λ into Eqs. (22) and
the relative humidity in the soil is established by using the Kelvin (23), respectively. The location of the vaporization plane s(t) and the
equation: temperature distribution T1 of Zone 1 are determined by substituting λ
into Eqs. (14) and (15), respectively. The water content profile of Zone
h v = exp
M 2 is obtained by substituting θ* and λ into Eq. (17).
RTs (19) It is noted that the definition of the critical water content θ* in this
study refers to that in soil physics. The vaporization plane forms if the
where hv is the relative humidity (dimensionless), M is the molar mass
water content of the surface soil reduces to the field water retention
of the water (0.018 kg mol−1), g is the acceleration of gravity (m s−2),
capacity, which is the natural water content in equilibrium with free
and R is the general gas constant (J mol−1 K−1).
drain condition. At that moment, the evaporation rate starts to decrease
The concentration of water vapor at the vaporization plane can be
gradually. Therefore, θ* is defined as the field water retention capacity
expressed as follows:
of the soil, or specifically the volumetric water content when the matric
Cs = h v Cs,sat (20) suction of the soil is 33 kPa (10 kPa for sand) or the hydraulic con-
where Cs,sat is the saturated vapor concentration at the temperature Ts. ductivity of the soil is 0.1 mm/day [16,20]. Here, a discussion on the
The relation between the saturated vapor concentration and the tem- definition of θ* should be given since no consensus has been reached in
perature can be obtained by integrating the Clausius-Clapeyron equa- the literature. For example, Novak [22] suggests that θ* is equal to the
tion. However, this approach is rather complicated. An empirical rela- residual moisture content θr. Kondo et al. [14] give an empirical ex-
tion proposed by Tetens [41] is widely used as a reasonable pression of θ* for different soil textures based on evaporation tests.
replacement and is adopted in this study, as follows: Wilson et al. [47] believed that θ* is the water content corresponding to
the soil matric suction of 3000 kPa. Teng et al. [38] show that the value
M T 273.15 of θ* changes with the variation in the wind speed through a series of
Cs,sat = 0.611 exp 17.27 s
RTs Ts 36 (21) laboratory tests. Tran et al. [42] conclude that the value of θ* lies be-
tween the water content corresponding to the air entry value and the
By substituting Eq. (18) into Eq. (19) to eliminate the parameter ψ, residual water content. In addition, van de Griend and Owe [44],
and then substituting Eqs. (19) and (21) into Eq. (20), we can establish Aluwihare and Watanabe [1], and Bittelli et al. [5] present different
the relation between Cs and θ*. empirical formulas of θ*. In summary, defining θ* as the field capacity
M M * r Ts 273.15 in this paper seems more reasonable because it has been validated by a
Cs = 0.611 exp ln + 17.27 number of tests for different soil textures and meteorological condi-
RTs RTs s r Ts 36 (22)
tions.
Finally, the vapor diffusion equation, Eq. (1), can be solved ac-
cording to Eq. (A8) given in the appendix, and the boundary conditions
Eqs. (6) and (7) are satisfied. The solution can be expressed as follows: 3. Results and discussion
Cs C0 x
C = C0 + erf 3.1. Model validation
erf ( a2 / a1 ) 2 a1 t (23)
The evaporation rate at Stage 1 is equal to the potential evaporation A large body of literature describes the results of laboratory tests
rate Ep, which can be computed by a widely accepted method, Penman performed on evaporation process in unsaturated soils. Among this
equation [24]. When the evaporation continues into Stage 2, Ea can be
determined by the derivative of the vapor concentration C at x = 0 Table 1
according to Fick’s law. Therefore, the expression of the evaporation Test conditions (after Song et al. [33,34]).
rate Ea(t) at the soil surface is as follows: Case number Wind speed (m/s) at Air temperature (K) Test duration
5 cm above soil surface (day)
Ep t tc
Ea (t ) = a1 Cs C0 1 0.5 323.15 11.5
t > tc
t erf ( a2 / a1 ) (24) 2 0.46 473.15 11.5
3 0.34 323.15 17.5
where tc is the transition time from Stage 1 to Stage 2 and is determined 4 0.34 473.15 30
by setting Eq. (17) equal to θ*. The physical meaning of tc is the time

110
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

Table 2 Song et al. [33] and the parameters θs, θr, α. Parameters a1, a2, and k1
Inputs of the analytical model for simulating of Song et al. [33] experiments. are determined by referring to the database in Hydrus-1D code, where
Case 1 Case 2 Case 3 Case 4 the basic hydro-thermal properties of some kinds of soils are summar-
ized [32]. The latent heat of vaporization (Lv) is a constant, and con-
T0 (K) 292.15 295.15 292.15 295.15 stant λ is computed by Eq. (16).
Ts (K) 289.15 291.65 290.15 293.65
It is noted that the inputs in this model can be divided into two
θ* (m3/ m3) 0.107
k1 (W/K * m) 0.0183
groups: (1) the parameters accounting for boundary conditions (T0, Ts,
C0 (kg/m3) 1.0 × 10−3 C0, θ∞), and (2) the parameters accounting for soil hydro-thermal
λ 2.65 2.62 2.58 2.57 properties (a1, a2, a3, θs, θr, θ*, α, and k1). The outputs of this model are
a1 (m2/s) 1.0 × 10−8 transient water content, evaporation rate, the depth of vaporization
a2 (m2/s) 3.9 × 10−11
plane, and so on.
a3 (m2/s) 6.7 × 10−9
α (kPa−1) 0.353 The measured and computed water content profiles for the four test
θr (m3/m3) 0.01 conditions are presented in Fig. 3. A good agreement between the
θs, θ∞ (m3/m3) 0.36 computed results and the measured data can be observed, indicating
Lv (J/kg) 2.46 × 106
that the proposed model reasonably well predicts the water content
distribution during the evaporation process. It can be observed that the
soil near the surface undergoes a clear decrease in moisture under the
four evaporative cases. The water content varies almost linearly with
the soil depth in the near-surface zone. In addition, the computed re-
sults also reveal that all the Cases advance to the Stage 2 of the eva-
poration, i.e., the surface water contents are lower than the critical
water content after a certain time. The computed time of the shift from
Stage 1 to Stage 2 for Cases 1, 2, 3 and 4 are 228.6 h, 141.9 h, 187.2 h
and 151.2 h, respectively.
The computed and measured evaporation rates and depth of va-
porization plane during Stage 2 are presented in Fig. 4. It is noted that
Case 1 is not discussed here since its period of Stage 2 is only 47.4 h. It
is so short that the variation trends of evaporation rate and vaporization
plane depth are not significant. As shown in Fig. 4a, the computed re-
sults of evaporation rate generally match the measured data well.
Fig. 4b presents the measured depth of vaporization plane gradually
increase with the elapsing of time during Stage 2, which is matched by
the computed result quite well. The proposed model is competent to
simulate the Stage 2 evaporation. A small mismatch can be observed
between the measured and computed results in Fig. 4a at the onset of
Fig. 2. The measured and simulated soil-water characteristic curve of the tested Stage 2 for the three cases. This mismatch may be caused by using an
soil. exponential function for the soil-water characteristic curve (Eq. (3a)) to
homogenize the Richards equation, and this approach always generates
work, Song et al. [33] reported a series of large-scale evaporation ex- a linear relationship between the evaporation rate and t−0.5. It is noted
periments based on an environmental chamber. They selected Fontai- that Brutsaert and Chen [6] and Shokri and Or [30] found an empirical
nebleau sand, a fine sand, as the test material. The atmospheric para- formula to describe the linear relation between the inverse of the square
meters (air flow rate, relative humidity, and temperature) and response of evaporate rate and the elapsed time. The proposed model (Eq. (24))
of the soil (volumetric water content, temperature, and soil suction) gives a theoretical expression for this linear relationship. That is, the
were monitored simultaneously during the test. Four different eva- outcome of this model is consistent with the experimental observation.
poration conditions were conducted by controlling the circulating wind
speed and air temperature, as shown in Table 1. The published results 3.2. Parametric study
provide a useful database for investigating the mechanism of eva-
poration process. To improve our understanding of the mechanism of evaporation
The validity of the proposed analytical model is checked against the from unsaturated soil, a parameter study is performed based on the
data of Song et al. [33,34]. The assignments of the parameters in the analytical model. Some key inputs are analyzed, including the critical
model are given in Table 2. The temperatures T0 is an average of the water content θ*, the vapor diffusion coefficient a1 of Zone 1, the vo-
measured temperatures at 0.75 cm above the soil surface, and Ts is an lume average heat diffusion coefficient a2 of Zone 1, and the equivalent
average of the measured temperatures at 0.25 cm below than the soil liquid water diffusion coefficient a3 of Zone 2. The control value and
surface. The parameters θ∞ is set according to the test boundary con- the comparative values are shown in Table 3. The outcomes of the
ditions. The parameter C0 is assigned to be an extremely low value, i.e. model consist of the water content distribution, the evaporation rate
0.001, to account for the controlled dry air in the test of Song et al. and the migration of the vaporization plane. It is noted that the other
[33]. The critical water content (θ*) is determined based on the mea- input parameters are consistent with those given in Case 2 of Table 2.
sured soil-water characteristic curve, and parameters θs, θr and α are Since the critical water content θ* is defined as the field water re-
obtained by fitting the soil-water characteristic curve, as shown in tention capacity, its value is largely controlled by the fine content of the
Fig. 2. The parameter a3 is computed from parameter ks that given in soil. A higher fine content leads to a larger value of θ*. In the analytical
model, θ* also represents the upper boundary of the liquid water flow

111
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

0.00 0.00
(a) (b)
0.05 0.05 t = 96 h
t = 96 h
t = 144 h
Depth, m 0.10 0.10

Depth, m
0.15 t = 192 h 0.15 t = 96 h
t = 96 h t = 144 h t = 276 h
0.20 t = 192 h t = 276 h 0.20 t = 192 h t = 240 h
t = 276 h t = 240 h t = 192 h
0.25 Computed result 0.25 t = 276 h
Computed result
0.30 0.30
5 10 15 20 25 30 35 5 10 15 20 25 30 35
Volumetric water cotent, % Volumetric water content, %
0.00 0.00
(c) (d)
0.05 0.05
t = 96 h t = 96 h
0.10 0.10
Depth, m

Depth, m
t = 192 h
0.15 0.15 t = 192 h
t = 96 h t = 96 h
t = 192 h t = 288 h t = 192 h
0.20 0.20 t = 288 h
t = 288 h t = 288 h
t = 420 h t = 720 h
0.25 0.25
t = 420 h Computed result t = 720 h
Computed result

0.30 0.30
5 10 15 20 25 30 35 5 10 15 20 25 30 35
Volumetric water content, % Volumetric water content, %
Fig. 3. Simulated and measured water content profiles for the Fontainebleau sand under four test conditions: (a) Case 1; (b) Case 2; (c) Case 3; (d) Case 4. The
symbols are the measured data while the solid lines indicate the simulated result.

zone (Eq. (2)). Its influence on the water content distribution is re- potential evaporation rate. This result can be explained as follows: a1
flected in Eq. (17). The water content profiles at t = 100 h and 200 h for controls the diffusion velocity of the water vapor in Zone 1, and the
different θ* value are depicted in Fig. 5. It is shown that the value of θ* evaporation rate is the upper boundary of the vapor diffusion zone, i.e.,
controls the top boundary water content and the slope of water content Zone 1. According to Fick’s law, the slope of the linear relationship is
profile. the gradient of the vapor concentration at z = 0.
The effects of θ* on the evaporation rate and the depth of vapor- Eqs. (14), (17) and (24) indicate that the heat diffusion coefficient
ization plane are shown in Fig. 6. It can be observed that a smaller value a2 has an influence on the depth of vaporization plane, water content
of θ* results in an earlier transition from Stage 1 into Stage 2, a smaller distribution of Zone 2 and evaporation rate. In addition, Eq. (16) shows
evaporation rate, and a deeper vaporization plane. In addition, θ* seems that a2 is related to the critical water content, and thus it can affect the
to have only a minor influence on the curve patterns. The evaporation transition time from Stage 1 into Stage 2. It is noted that the majority of
rate gradually decreases with time and eventually converges to a steady literature on evaporation process deals with the unsaturated liquid
value, irrespective of the θ* values. The depth of the vaporization plane water or vapor flow. Few of these studies paid attention to the influence
decreases with increasing θ* value, but the decreasing rate of the va- of the heat transfer on the evaporation [30,23,39]. Fig. 8 shows the
porization plane seems to be independent of the critical water content. influence of a2 on the water content distribution, while the embedded
In general case, a finer soil corresponds to a higher θ* value, and hence chart shows the relation between the transition time and a2. It can be
should lead to a lower evaporation rate and a deeper vaporization plane observed that transition time exponentially decreases with increasing
under same evaporation condition [11,38], which seems to be the op- a2. When a2 is assigned to the values of 1 × 10−10, 3.9 × 10−11 and
posite of the results in Fig. 6. The reason is that soil type does not only 1 × 10−11, the corresponding transition times are 58 h, 141.9 h and
affect the critical water content, but also controls hydraulic con- 550 h, respectively. If a2 is less than 6 × 10−12, the evaporation con-
ductivity and thermal conductivity. The result implies that the soil tinues in Stage 1 and does not have a transition time. The result in-
hydraulic or thermal properties have a more pronounced influence on dicates that the heat transfer process in the soil significantly affects the
evaporation process than critical water content. evaporation process, and a smaller thermal diffusion coefficient leads to
It is interesting to note that the vapor diffusion coefficient (a1) does a longer duration of Stage 1 in the evaporation process. The water
not affect the water content distribution or the depth of vaporization content profiles for different a2 values at t = 200 h are shown in Fig. 8
plane (Eqs. (14)–(17)), but it is closely related to the evaporation rate as well. The shape of the three curves are generally consistent. A closer
through Eq. (24). The evaporation rate versus a1 at different times is position to the vaporization plane corresponds to a larger difference
shown in Fig. 7, which shows that the evaporation rate increases line- between the curves. In addition, a smaller value of a2 leads to a larger
arly with increasing a1, while the rate of increase decreases with time. It critical water content.
should be noted that the actual evaporation rate is limited by the The heat diffusion coefficient a2 has an influence on the evaporation

112
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

2.4
Measured date (Case 2) (a)
2.0 Computed result (Case 2)
Computed result (Case 3)
Evaporation rate, mm/d

Computed result (Case 4)

1.6

1.2
Measured date (Case 3)

0.8

Measured date (Case 4)

0.4
0 100 200 300 400 500
Time, h
Fig. 5. Effect of critical water content θ* on water content distribution.
0.25
Computed result (Case 4) (b)
0.20
Depth of vaporization plane, m

0.15

Computed result
0.10 (Case 2) Computed result
(Case 3)

0.05 Measured data (Case 2)

Measured data (Case 3)
Measured data (Case 4)
0.00
0 100 200 300 400 500
Time, h
Fig. 4. (a) Comparison between simulated and measured evaporation rates Fig. 6. Effects of critical water content θ* on evaporation rate of Stage 2 and
during Stage 2 for Cases 2, 3 and 4. Continuous dash polylines are the measured depth of vaporization plane.
results, solid lines denote the results computed by the proposed model. (b)
Comparison between simulated and measured depth of vaporization plane
during Stage 2 for Cases 2, 3 and 4. Symbols are the measured data, and solid
lines denote the results computed by the proposed model.

Table 3
The values of the selected parameters.
Control value Comparative value

θ* (m3/m3) 0.107 0.05, 0.15

a1 (m2/s) 1.0 × 10−8 1.0 × 10−9 ∼ 1.2 × 10−7
a2 (m2/s) 3.9 × 10−11 1 × 10−10, 1 × 10−11
a3 (m2/s) 6.7 × 10−9 1 × 10−9, 2 × 10−8

rate and the depth of the vaporization plane. Their relationships are
shown in Fig. 9. The divergence of the curves along the abscissa is at-
tributed to the different transition times. A comparison of the three
evaporation rate curves indicates that a larger a2 leads to a faster de-
creasing evaporation rate during Stage 2 and a smaller residual value,
when the evaporation becomes a slow and gentle process. For the re-
lationship between a2 and the depth of the vaporization plane, a larger Fig. 7. Effect of vapor diffusivity of soil on the evaporation rate at different
a2 corresponds to a faster advancing rate and a greater depth of the times.
vaporization plane. The reason for this relationship is as follows: Eqs.
(10) and (11) indicate that the temperatures at the vaporization plane is greater. In addition, Fig. 9 depicts that a greater value of the depth of
and soil surface are constant, while a greater a2 in Eq. (2) means that the vaporization plane corresponds to a lower value of the evaporation
the temperature gradient from the soil at a certain depth to at the rate, which is consistent with the actual situation.
boundary is much smaller, such that the depth of the vaporization plane The equivalent water diffusion coefficient a3 is positively correlated

113
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

that a3 affects the water content distribution and the transition time,
but has no influence on the evaporation rate and the movement of
vaporization plane. Fig. 10 shows the relationship between a3 and the
water content distribution at the times of t = 100 h and 200 h. A greater
value of a3 corresponds to a much gentler profile of water content,
meaning that the water is evaporated over a larger depth. With in-
creasing time, the water content profile becomes gentler because a
greater diffusion coefficient is beneficial to rapidly transferring liquid
water to the vaporization plane and to supplying the evaporation de-
mand. As such, the water content of the soil is changed over a larger
depth. In the case of a3 equal to 2 × 10−8, the water content profile at
t = 200 h becomes almost linear, which is attributed to the exponential
model of the soil-water characteristic curve (Eq. (3a)). When the pro-
cess reaches a steady state, the water content is linearly distributed with
depth, as the solution of Eq. (4).

Fig. 8. Effect of thermal diffusion coefficient a2 on water content profile at

t = 200 h. The embedded chart shows the relation between transition time and 4. Conclusion
a2 .
Based on the movement of the vaporization plane, a model for
evaporation from unsaturated soil is established, which includes three
partial differential equations respectively governing the vapor diffu-
sion, liquid water flow and heat transfer. The outcomes of this model
include the evaporation rate, the depth of vaporization plane, and the
transient profiles of vapor concentration and water content, which all
have explicit expressions. A series of evaporation experiments from the
literature are used to evaluate the performance of the proposed model.
The simulated results satisfactorily reproduce the laboratory experi-
ments as the analytical solutions are very close to the measured values
of the water content profiles and evaporation rates.
The result shows that the evaporation rate during Stage 2 is pro-
portional to the inverse of the square root of time, with the pro-
portionality being affected by the vapor diffusion coefficient, heat dif-
fusion coefficient, and critical water content. The depth of vaporization
plane is however independent of soil hydraulic properties such as vapor
diffusion coefficient, hydraulic conductivity, water retention capacity,
Fig. 9. Effects of thermal diffusion coefficient a2 on the evaporation rate and but dependent on heat diffusion coefficient.
depth of the vaporization plane. A parametric study is performed in attempt to gain some insights of
evaporation process. The results show that a smaller value of the critical
water content leads to an earlier transition from Stage 1 into Stage 2, a
smaller evaporation rate, and a greater depth of the vaporization plane.
However, compared to the thermal and hydraulic properties of the soil,
the influence of the critical water content on evaporation process seems
relatively minor.
The heat diffusion coefficient has a significant influence on the
evaporation process that has not been previously addressed. The tran-
sition time from Stage 1 to Stage 2 decreases exponentially with in-
creasing thermal diffusion coefficient. A greater value of the thermal
diffusion coefficient results in a faster decreasing evaporation rate, a
faster advancing rate and a greater depth of the vaporization plane.
The analytical model developed in this paper is relatively simple
and practically useful. It can provide an alternative approach for
studying the evaporation mechanisms in unsaturated soils.

Acknowledgements
Fig. 10. Effect of water diffusion coefficient a3 on water content profile, at
t = 100 h and 200 h. This research was supported by National Natural Science
Foundation of China (No. 51878665), National Basic Research Program
with the permeability of the soil, reflecting the penetration capacity of China (No. 2014CB047001), and National Natural Science
into the soil by fluid. It can be found from Eqs. (14), (16), (17) and (24) Foundation of China (No. 51722812).

Appendix

Eqs. (1), (2) and (4) belong to the same type of partial differential equation. In this appendix, this type of partial differential equation is solved for
the general solution of the different boundary conditions. Assume that the function v(x,t) satisfies the above equation, i.e.,

114
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

v 2v
=k 2
t x (A1)
where k is a constant. According to the solution of the partial differential equations in Carslaw and Jaeger [7], we let
1 x2
v (x , t ) = t 2 e 4kt (A2)
Substituting Eq. (A2) into Eq. (A1), we can obtain the following relationships:
2v 1 x2 x2 x2
= e 4kt + e 4kt
x2 3
2kt 2
5
2k 2t 2 (A3)
2 2
v 1 x x2 x
= e 4kt + e 4kt
t 2t 2
3
2kt 2
5
(A4)
According to the definition of the error function, transforming the function of v leads to
x
x 1 x2 1 2
t 2 e 4kt dx = 2k 2 2 kt e d (A5)
0 0

The error function erf(x) and complementary error function erfc(x) are defined as follows:
2 x 2
erf (x ) = e d
0 (A6)
2 2
erfc (x ) = e d
x (A7)
Considering the expressions of Eqs. (5), (6) and (7), we can obtain a solution for the partial differential equation when satisfying the condition of
x → 0 and v → v0, as follows:

x
v = v0 + Aerf
2 kt (A8)
where A is a constant. The other solution of the partial differential equation can be solved when it meets the condition of x → ∞ and v → v∞, as
follows:

x
v=v Berfc
2 kt (A9)
where B is a constant.

References [16] Lee TJ, Pielke RA. Estimating the soil surface specific humidity. J Appl Meteorol
1992;31:480–4.
[17] Lehmann P, Assouline S, Or D. Characteristic lengths affecting evaporative drying of
[1] Aluwihare S, Watanabe K. Measurement of evaporation on bare soil and estimating porous media. Phys Rev E 2008;77:056309. https://doi.org/10.1103/PhysRevE.77.
surface resistance. J Environ Eng 2003;129(12):1157–68. 056309.
[2] Assouline S, Narkis K, Gherabli R, Lefort P, Prat M. Analysis of the impact of surface [18] Li Q, Sun S, Xue Y. Analyses and development of a hierarchy of frozen soil models
layer properties on evaporation from porous systems using column experiments and for cold region study. J Geophys Res Atmos 2010;115(D3):315–7. https://doi.org/
modified definition of characteristic length. Water Resour Res 2014;50:3933–55. 10.1029/2009JD012530, 2010.
[3] Aydin M, Yang SL, Kurt N, Yano T. Test of a simple model for estimating eva- [19] Menziani M, Pugnaghi S, Vincenzi S. Analytical solutions of the linearized Richards
poration from bare soils in different environments. Ecol Model equation for discrete arbitrary initial and boundary conditions. J Hydrol
2005;182(1):91–105. 2007;332:214–25.
[4] Blight GE. Interaction between the atmosphere and the earth. Géotechnique [20] Milly PCD. Potential evaporation and soil moisture in general circulation models. J
1997;47(4):715–67. Clim 1992;5(3):209–26.
[5] Bittelli M, Ventura F, Campbell GS, Snyder RL, Gallegati F, Pisa PR. Coupling of [21] Nasseri M, Daneshbod Y, Pirouz MD, Rakhshandehroo GR, Shirzad A. New analy-
heat, water vapor and liquid water fluxes to compute evaporation in bare soils. J tical solution to water content simulation in porous media. J Irrig Drain Eng
Hydrol 2008;362:191–205. 2012;138(4):328–35.
[6] Brutsaert W, Chen D. Desorption and the two stages of drying of natural tall grass [22] Novak MD. Quasi-analytical solutions of the soil water flow equation for problems
prairie. Water Resour Res 1995;31(5):1305–13. of evaporation. Soil Sci Soc Am J 1988;52(4):916–24.
[7] Carslaw HS, Jaeger JC. Conduction of heat in solids. 2nd edition England: [23] Or D, Lehmann P, Shahraeeni E, Shokri N. Advances in soil evaporation physics–a
Clarendon. Oxford; 1959. p. 99–114. review. Vadose Zone J 2013;12(4):1–45. https://doi.org/10.2136/vzj2012.0163.
[8] Chen JM, Tan YC, Chen CH, Parlange JY. Analytical solutions for linearized [24] Penman HL. Natural evaporation from open water, bare soil and grass. Proceedings
Richards’ equation with arbitrary time-dependent surface fluxes. Water Resour Res of the Royal Society of London. 1948. p. 120–45.
2001;37(4):1091–3. [25] Philip JR, de Vries DA. Moisture movement in porous materials under temperature
[9] Cui YJ, Ta AN, Hemmati S, Tang AM, Gatmiri B. Experimental and numerical in- gradients. Trans Am Geophys Union 1957;38(2):222–32.
vestigation of soil-atmosphere interaction. Eng Geol 2013;165:20–8. [26] Rose DA, Konukcu F, Gowing JW. Effect of water table depth on evaporation and
[10] Fayer MJ, Jones TL. UNSAT-H version 2.0: Unsaturated soil water and heat flow salt accumulate on from saline groundwater. Soil Res 2005;43(5):565–73.
model. Richland, Washington: Pacific Northwest Laboratory; 1990. PNL-6779. [27] Sakai M, Toride N, Šimunek J. Water and vapor movement with condensation and
[11] Gowing JW, Konukcu F, Rose DA. Evaporative flux from a shallow water table: the evaporation in a sandy column. Soil Sci Soc Am J 2009;73(3):707–17.
influence of a vapour-liquid phase transition. J Hydrol 2006;321:77–89. [28] Shimojimaa E, Yoshioka R, Tamagawa I. Salinization owing to evaporation from
[12] Hitthe EM, Graf T. Non-iterative adaptive time-stepping scheme with temporal bare soil surfaces and its influence on the evaporation. J Hydrol 1996;178:109–36.
truncation error control for simulating variable-density flow. Adv Water Resour [29] Shokri N, Lehmann P, Vontobel P, Or D. Drying front and water content dynamics
2012;49:46–55. during evaporation from sand delineated by neutron radiography. Water Resour Res
[13] Jalila J, Moncef B, Hatema E. Salt removal from soil using the argil porous ceramic. 2008;44(6):663–71.
Desalination 2016;379:53–67. [30] Shokri N, Or D. What determines drying rates at the onset of diffusion controlled
[14] Kondo J, Saigusa N, Sato T. A parameterization of evaporation from bare soil sur- stage-2 evaporation from porous media? Water Resour Res 2011;47(9):1900–4.
face. J Appl Meteorol 1990;29(5):385–9. [31] Šimunek J, van Genuchten MTh, Šejna M. Development and applications of the
[15] Konukcu F, Istanbulluoglu A, Kocaman I. Determination of water content in drying HYDRUS and STANMOD software packages, and related codes. Vadose Zone J
soils: incorporating transition from liquid phase to vapour phase. Aust J Soil Res 2008;7(2):587–600.
2004;42(1):1–8. [32] Šimunek J, van Genuchten MTh, Šejna M. The HYDRUS-1D software package for

115
J. Teng et al. Computers and Geotechnics 108 (2019) 107–116

simulating the. One-dimensional movement of water, heat, and multiple solutes in diffusion under semiarid conditions. Water Resour Res 1994;30(2):181–8.
variably-saturated media. Version 4.17. Riverside (California, USA): Department of [45] Wilson GW. The evaluation of evaporative fluxes from unsaturated soil surfaces PhD
Environmental Sciences, University of California Riverside; 2013. thesis: University of Saskatchewan; 1990.
[33] Song WK, Cui YJ, Tang AM, Ding WQ, Tran TD. Experimental study on water [46] Wilson GW, Fredlund DG, Barbour SL. Coupled soil-atmosphere modeling for soil
evaporation from sand using environmental chamber. Can Geotech J evaporation. Can Geotech J 1994;31(2):151–61.
2014;51(2):115–28. [47] Wilson GW, Fredlund DG, Barbour SL. The effect of soil suction on evaporative
[34] Song WK, Cui YJ, Ye WM. Modelling of water evaporation from bare sand. Eng Geol fluxes from soil surface. Can Geotech J 1997;34:145–55.
2018;233:281–9. [48] Xue Z, Akae T. Maximum surface temperature model to evaluate evaporation from a
[35] Stewart JM, Broadbridge P. Calculation of humidity during evaporation from soil. saline soil in arid area. Paddy Water Environ 2012;10(2):153–9. https://doi.org/10.
Adv Water Resour 1999;22(5):495–505. 1007/s10333-011-0286-y.
[36] Tang C, Shi B, Liu C, Gao L, Inyang HI. Experimental investigation of the desiccation [49] Yamanaka T, Takeda A, Sugita F. A modified surface-resistance approach for re-
cracking behavior of soil layers during drying. J Mater Civ Eng 2011;23(6):873–8. presenting bare-soil evaporation: wind tunnel experiments under various atmo-
[37] Teng J, Yasufuku N, Liu Q, Liu S. Analytical solution for soil water redistribution spheric conditions. Water Resour Res 1997;33(9):2117–28.
during evaporation process. Water Sci Technol 2013;68(12):2545–51. [50] Yamanaka T, Yonetani T. Dynamics of the evaporation zone in dry sandy soils. J
[38] Teng J, Yasufuku N, Liu Q, Liu S. Experimental evaluation and parameterization of Hydrol 1999;217:135–48.
evaporation from soil surface. Nat Hazards 2014;73(3):1405–18. [51] Yanful EK, Choo LP. Measurement of evaporative fluxes from candidate cover soils.
[39] Teng J, Yasufuku N, Zhang S, He Y. Modelling water content redistribution during Can Geotech J 1997;34(3):447–59.
evaporation from sandy soil in the presence of water table. Comput Geotech [52] Yanful EK, Mousavi SM, Yang M. Modeling and measurement of evaporation in
2016;75:210–24. moisture-retaining soil covers. Adv Environ Res 2003;7(4):783–801.
[40] Teng J, Shan F, He Z, Zhang S, Sheng D. Experimental study of ice accumulation in [53] Yiotis AG, Tsimpanogiannis IN, Stubos AK, Yortsos YC. Pore-network study of the
unsaturated clean sand. Géotechnique 2018. https://doi.org/10.1680/jgeot.17.P. characteristic periods in the drying of porous materials. J Colloid Interface Sci
208. 2006;297(2):738–48.
[41] Tetens O. Uber einige meteorologische Begriffe. Zeitschrift fur Geophysik [54] Yuan F, Lu Z. Analytical solutions for vertical flow in unsaturated, rooted soils with
1930;6:297. variable surface fluxes. Vadose Zone J 2005;4(4):1210–8.
[42] Tran DTQ, Fredlund DG, Chan DH. Improvements to the calculation of actual [55] Zarei G, Homaee M, Laight AM, Hoorfar AH. A model for soil surface evaporation
evaporation from bare soil surfaces. Can Geotech J 2016;53(1):118–33. based on Campbell’s retention curve. J Hydrol 2010;380:356–61.
[43] Tu X, Dai F. Analytical solution for one-dimensional heat transfer equation of soil [56] Zhang C, Li L, Lockington D. A physically based surface resistance model for eva-
and evaluation for thermal diffusivity. Chin J Geotech Eng 2008;30(5):652–7. (in poration from bare soils. Water Resour Res 2015;51(2):1084–111.
Chinese)). [57] Zhang S, Teng J, He Z, Liu Y, Liang S, Yao Y, et al. Canopy effect caused by vapour
[44] Van de Griend AA, Owe M. Bare soil surface resistance to evaporation by vapor transfer in covered freezing soils. Géotechnique 2016;66(11):927–40.

116