Energy and Buildings 38 (2006) 303–314

www.elsevier.com/locate/enbuild

Simultaneous heat and moisture transfer in soils combined

with building simulation
Gerson H. dos Santos, Nathan Mendes *
Pontifical Catholic University of Parana, PUCPR/CCET, Thermal Systems Laboratory—LST, Rua Imaculada Conceição,
1155 Curitiba, PR 80215-901, Brazil

Received 10 February 2005; received in revised form 25 May 2005; accepted 22 June 2005

Abstract

In order to precisely predict ground heat transfer, room air temperature and humidity, a combined model has been developed and conceived
to calculate both the coupled heat and moisture transfer in soil and floor and the psychrometrics condition of indoor air. The present
methodology for the soil is based on the theory of Philip and De Vries, using variable thermophysical properties for different materials. The
governing equations were discretized using the finite-volume method and a three-dimensional model for describing the physical phenomena
of heat and mass transfer in unsaturated moist porous soils and floor. Additionally, a lumped transient approach for a building room and a
finite-volume multi-layer model for the building envelope have been developed to integrate with the soil model. Results are presented in terms
of temperature, humidity and heat flux at the interface between room air and the floor, showing the importance of the approach presented and
the model robustness for long-term simulations with a high time step.
# 2005 Elsevier B.V. All rights reserved.

Keywords: Coupled heat and moisture transfer; Soil simulation; Building hygrothermal performance

1. Introduction
Since the 1970s, due to the worldwide energy crisis, some
countries have adopted a severe legislation aiming to
promote energy efficiency of equipment and buildings. In
Brazil, in 2001, a committee to develop regulations for
energy efficiency in buildings, such as national standards
and building codes, was created so that rational policies of
energy conservation could be applied all over the country. In
this context, to evaluate the building performance with
thermal parameters, several codes have been developed.
However, most of those codes do not take into account the
moisture presence within building envelopes. The moisture
in the furniture and envelope of buildings implies an
additional mechanism of transport absorbing or releasing
latent heat of vaporization, affecting the hygrothermal
building performance or causing mold growth.
* Corresponding author. Tel.: +55 41 3271 1322; fax: +55 41 3271 1349.
E-mail addresses: gerson.santos@pucpr.br (G.H. dos Santos),
nathan.mendes@pucpr.br (N. Mendes).
Actually, some studies have been carried out to model the
moisture storage and transport mechanisms and their effects
on heat transport through walls and roofs of buildings [1–3].
However, these works present some simplifications on their
calculation routines by not considering the three-dimen-
sional aspects of heat and moisture transfer and have not
focused these effects on highly capillary soils.
Some building physics studies involving the pure
conduction heat transfer through the ground can be found
in the literature. The first experimental study concluded that
the heat loss through the ground is proportional to the size of
its perimeter. However, Bahnfleth [4] observed that the area
and shape must be taken into account as well.
Davies et al. [5], using the finite-volume approach,
compared multidimensional models and observed that the
use of three-dimensional simulation provides better predic-
tion of building temperature and heating loads than two-
dimensional simulation, when these results are compared
with experimental data.
Computer programs for transient and steady-state pure
conduction heat transfer in two and three dimensions have

0378-7788/$ – see front matter # 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.enbuild.2005.06.011
304 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314

Nomenclature qr solar radiation (W/m2)

R universal gas constant (J/(kmol K))
Ai area of the i-th surface (m2) Rlw long-wave radiation (W/m2)
Aj,f area of j-th control volumes of the floor sur- T temperature (8C)
face discretized by using the finite-volume T0 temperature calculated at the previous time
method (m2) step (8C)
c specific heat (J/(kg K)) Ti temperature of each surface of the building
cm mean specific heat (J/(kg K)) envelope (8C)
cw wall specific heat (J/(kg K)) Tint room air temperature (8C)
cair air specific heat (J/(kg K)) Text external air temperature (8C)
DTl liquid phase transport coefficient associated to Ty=H temperature of the ground or floor surfaces
a temperature gradient (m2/(s K)) (8C)
DTv vapor phase transport coefficient associated to Tx=0 temperature on the external surfaces (8C)
a temperature gradient (m2/(s K)) Tx=e temperature on the internal surfaces (8C)
Dul liquid phase transport coefficient associated to Tprev temperature calculated at the previous itera-
a moisture content gradient (m2/s) tion (8C)
Duv vapor phase transport coefficient associated to Tsur temperature of internal surfaces of surround-
a moisture content gradient (m2/s) ing walls (8C)
DT mass transport coefficient associated to a t time (s)
temperature gradient (m2/(s K)) Vair room volume (m3)
Du mass transport coefficient associated to a W0 humidity ratio calculated at the previous time
moisture content gradient (m2/s) step and (kg water/kg dry air)
Ėt energy flow that crosses the room control Wext external humidity ratio (kg water/kg dry air)
surface (W) Wint internal humidity ratio (kg water/kg dry air)
Ėg internal energy generation rate (W) Wv,f the humidity ratio of each control volume (kg
hm mass convection coefficient (m/s) water/kg dry air)
hmf floor mass convection coefficient (m/s) as soil solar absorptivity
hms soil mass convection coefficient (m/s) aw wall solar absorptivity
h convection coefficient (W/(m2 K)) es soil emissivity
hint internal convection heat transfer coefficient ef floor emissivity
between internal surfaces and room air ew wall emissivity
(W/(m2 K)) eroof roof emissivity
hext external convection heat transfer coefficient fv view factor
between external surfaces and external air f relative humidity
(W/(m2 K)) fext external relative humidity
hroof external convection heat transfer coefficient l thermal conductivity (W/(m2 K))
for the rooftop (W/(m2 K)) lf floor (concrete) thermal conductivity
hD the floor mass transfer coefficient (kg/m2 s), (W/(m K))
H depth of the ground (m) ls soil thermal conductivity (W/(m K))
jv vapor flow (kg/(m2 K)) lw wall thermal conductivity (W/(m K))
jl liquid flow (kg/(m2 K)) lroof roof thermal conductivity (W/(m K))
j total flow (kg/(m2 K)) r density (kg/m3)
L latent heat of vaporization (J/kg) r0 solid matrix density (kg/m3)
ṁinf mass flow by infiltration (kg/s) rl water density (kg/m3)
ṁvent mass flow by ventilation (kg/s) pv,1 vapor density in the surrounding air far from
ṁb water vapor flow from the breath of occupants the soil surface (kg/m3)
(kg/s) Pv,y=H vapor density at the upper surface of the soil
ṁger internal water-vapor generation rate (kg/s), domain (kg/m3)
M molecular mass (kg/kmol) rair air density (kg/m3)
m number of surfaces u volume basis moisture content (m3/m3)
n number of control volumes of the floor surface uprev volume basis moisture content calculated at
discretized by using the finite-volume method the previous iteration (m3/m3)
Ps saturated pressure (Pa) s Stefan–Boltzmann constant (W/(m2 K4))
prev previous iteration
 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314
been developed by Blomberg [6]. These codes can be used
for analyses of thermal bridge effects, heat transfer through
the corners of a window and heat loss from a house to the
ground. However, the moisture presence has been ignored.
Matsumoto et al. [7] described a program for ground
temperature data generation, using the finite-element
method. Latent heat losses by evaporation at the ground
surface and snow cover on the surface are considered in their
program.
Krarti [8] discussed the effect of spatial variation of soil
thermal properties on slab-on-ground heat transfer by using
the Interzone Temperature Profile Estimation (ITPE)
technique.
In the works mentioned above, the conductivity and the
thermal capacity are considered constant and the moisture
effect is ignored. However, the presence of moisture in the
ground implies an additional mechanism of transport: in the
pores of unsaturated soil, liquid water evaporates at the
warm side, absorbing latent heat of vaporization, while, due
to the vapor-pressure gradient, vapor condenses on the
coldest side of the pore, releasing latent heat of vaporization
[9]. This added or removed latent heat can cause great
discrepancies on the prediction of room air temperature and
relative humidity, when compared to values obtained by pure
conduction heat transfer [2].
Moreover, Brink and Hoogendoorn [10] analyzed
groundwater losses due to conduction and natural convec-
tion heat transfer modes and verified that convection heat
losses are mainly dependent on soil permeability.
Freitas and Prata [11] elaborated a numerical methodol-
ogy for thermal performance analysis of power cables under
the presence of moisture migration in the surrounding soil.
They utilized a two-dimensional finite-volume approach to
solve the governing equations and the boundary conditions
did not take any phase-change effect into account.
Ogura et al. [12] analyzed the heat and moisture behavior
in underground space in a two-dimensional model using the
quasi-linearized method. The outdoor temperature, relative
humidity, solar radiation and precipitation were investigated
as outdoor conditions.
Some simplified models can be also found in the
literature such as the model utilized by the MOHID program
[13]. This code, using the Richards’s equation, is utilized for
simulating marine processes, quality processes in reservoirs
and water flow in unsaturated media.
Combined heat and moisture transfer in soils may require
the use of very short time steps, especially at highly
permeable surfaces, which may prohibit the use of long time
step for long-term simulations. Galbraith [14] noticed the
influence of the spatial discretization on the accuracy of the
numerical simulation on the heat and mass transport in
porous media. It was observed that in one-dimensional
discretization, the two-way model, based on the technique of
control volume, can be used to avoid refined meshes.
Due to the numerical instabilities caused by the effect of
latent heat at the boundaries, Wang and Hagentoft [15]
305
presented a numerical method based on an algorithm that
combines an explicit model with relaxation schemes. In this
model, a criterion for time step determination is developed
to improve numerical stability.
For ensuring numerical stability in the present model, the
linearized set of equations was obtained by using the finite-
volume method and the MultiTriDiagonal-Matrix Algorithm
[16] to solve a three-dimensional model to describe the
physical phenomena of heat and mass transfer in sandy-silt
and sandy porous soils. In this way, the code has been
conceived to be numerically robust with a fast-simulating
procedure. The heat and moisture transfer in soils was based
on the theory of Philip and De Vries [17], which is one of the
most disseminated and accepted mathematical formulation
for studying heat and moisture transfer through porous soils,
considering both vapor diffusion and capillary migration.
Janssen et al. [18] elaborated an analysis of heat loss
through a basement and presented as false a generally
accepted postulate in building simulation: the combined heat
and mass transfer in ground can be ignored for the heat flow
calculation through the building foundation. In this context,
Onmura et al. [19] investigated the evaporative cooling
effect of roofs lawn gardens and observed a reduction of up
to 50% on the heat flux through the ceiling.
In soil simulation, some parameters such as the boundary
conditions, initial conditions, simulation time period
(including warm-up), simulation time step and grid
refinement have to be carefully chosen and combined in
order to reach accuracy without using excessive computa-
tional processing.
In this way, it is presented a mathematical model in order to
test the hygrothermal performance of buildings by consider-
ing the combined three-dimensional heat and moisture
transport through the ground for capillary unsaturated porous
soils. Heat diffusion through building envelope (walls and
roof) was calculated by using the Fourier’s law. The
importance of considering a three-dimensional approach
for the soil domain for low-rise buildings was verified by
Santos and Mendes [20], using a simply conductive model for
ground heat transfer calculation.
The room can be submitted to loads of solar radiation,
inter-surface long wave radiation, convection, infiltration and
internal gains from light, equipment and people. A lumped
approach for energy and water vapor balances is used to
calculate the room air temperature and relative humidity.
2. Mathematical model
The physical problem is divided into three domains:
ground, building envelope (walls and roof) and room air. At
the external surfaces, the heat transfer due to short wave
radiation and convection were considered as boundary
conditions and the long-wave radiation losses were taken
into account only at horizontal surfaces, i.e., ground and
roof. At the internal surfaces, besides the convection heat
306 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314

transfer, long-wave radiation exchange between the surfaces

was considered and for the ground and floor a combined heat
and mass transfer model was used.

2.1. Soil and floor domain

The governing equations, based on the theory of Philip

and De Vries [17], to model heat and mass transfer through
porous media, are given by Eqs. (1) and (2). The energy
conservation equation is written in the form:
@T Fig. 1. Physical domain for ground and floor.
r0 cm ðT; uÞ ¼ r
ðlðT; uÞrTÞ  LðTÞðr
jv Þ (1)
@t
and the mass conservation equation as:

 Surface 2 (in contact with external air):
@u j
¼ r
; (2)

@t rl @T
ls ðT; uÞ þ ðLðTÞ jv Þy¼H
@y y¼H
where r0 is the solid matrix density (m3/kg); cm, the mean
specific heat (J/(kg K)); T, temperature (8C); l, thermal ¼ hs ðText  Ty¼H Þ þ as qr
conductivity (W/(m K)); L, latent heat of vaporization (J/ þ LðTÞhms ðrv;ext  rv;y¼H Þ  es Rlw ; (5)
kg); u, volumetric moisture content (m3/m3); jv, vapor flow
(kg/(m2 K)); j, total flow (kg/(m2 K)); rl is the water density where h(T1  Ty=H) represents the heat exchanged by con-
(kg/m3). vection with the external air, described by the surface
The total three-dimensional vapor flow (j) – given by conductance h, asqr is the absorbed short-wave radiation
summing the vapor flow (jv) and the liquid flow (jl) – can be and LðTÞhm ðrv;1  rv;y¼H Þ, the phase-change energy term.
described as: The long-wave radiation loss is defined as Rlw (W/m2) and e

 is the surface emissivity. The solar absorptivity is repre-
j @T @u sented by a and the mass convection coefficient by hm,
¼  DT ðT; uÞ þ Du ðT; uÞ i
rl @x @x which is related to h by the Lewis’ relation.


@T @u Similarly, the mass balance at the upper surface is written
 DT ðT; uÞ þ Du ðT; uÞ j as:
@y @y



@T @u @Kg @u @T hmf
 DT ðT; uÞ þ Du ðT; uÞ þ k; (3) Duf ðT; uÞ þ DTf ðT; uÞ ¼ ðr  ry¼H Þ;
@z @z @z @y @y y¼H rl v;int
with DT = DTl + DTv and Du = Dul + Duv, where DTl is the (6)
liquid phase transport coefficient associated to a temperature
gradient (m2/(s K)), DTv, vapor phase transport coefficient for surface 1 (in contact with internal air), and as:
associated to a temperature gradient (m2/(s K)), Dul, liquid

@u @T hms
phase transport coefficient associated to a moisture content Dus ðT; uÞ þ DTs ðT; uÞ ¼ ðr  ry¼H Þ;
@y @y y¼H rl v;ext
gradient (m2/s), Duv, vapor phase transport coefficient asso-
ciated to a moisture content gradient (m2/s), DT, mass (7)
transport coefficient associated to a temperature gradient
for surface 2 (in contact with external air).
(m2/s K) and Du is the mass transport coefficient associated
The other soil domain surfaces were all considered
to a moisture content gradient (m2/s).
adiabatic and impermeable.
Fig. 1 shows the physical domain. According to Fig. 1,
Eqs. (6) and (7) show a vapor concentration difference,
the boundary conditions, for the most generic case (three-
Drv, on their right-hand side. This difference is between the
dimensional), can be mathematically expressed as:
porous surface and air and is normally determined by using
Surface 1 (in contact with internal air):
the values of previous iterations for temperature and


@T moisture content, generating additional numerical instabil-
lf ðT; uÞ þ ðLðTÞ jv Þy¼H ity. Due to the instability created by this source term, the
@y y¼H
solution of the linear set of discretized equations normally
X
m
requires the use of very small time steps, which can be
4 4
¼ hf ðTint  Ty¼H Þ þ fv ef sðTsur  Ty¼H Þ
i¼1 exceedingly time consuming especially in long-term soil
simulations; in some research cases, a time period of several
þ LðTÞhmf ðrv;int  rv;y¼H Þ (4) decades has to be simulated, taking into account the three-
 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314 307

dimensional heat and moisture transfer through a very convection heat transfer. In this way, the external boundary
refined grid. condition (x = 0) can be mathematically expressed as:
In order to raise the simulation time step, Mendes et al.

[16] presented a procedure to calculate the vapor flow, @T
 lw ¼ hext ðText  Tx¼0 Þ þ aw qr : (10)
independently of previous values of temperature and @x x¼0
moisture content. In this way, the term (Drv) was linearized
as a linear combination of temperature and moisture content, On the internal side (x = L), the inter-surface long-wave
viz., radiation was included as:

 X
m
ðrv;1  rv ðsÞÞ ¼ M1 ðT1  TðsÞÞ þ M2 ðu1  uðsÞÞ þ M3 ; @T 4 4
lw ¼ hint ðTint  Tx¼e Þ þ fv ew sðTsur  Tx¼e Þ:
(8) @x x¼e i1

where (11)
M On the other hand, for the roof, long-wave radiation
M1 ¼ A f;
R losses were considered (Rlw) so that Eq. (10) has assumed



M Ps ðsÞ prev @f prev the following form:
M2 ¼ ;
R TðsÞ @uðsÞ


  @T
M Ps ðsÞ prev  lroof ¼ hroof ðText  Tx¼0 Þ þ aroof qr  eroof Rlw ;
M3 ¼ Rðuprev ðsÞÞþf1 ðRðT1 ÞRðT prev ðsÞÞÞ : @x x¼0
R TðsÞ
(12)
In the equations above, the index(s) represents the surface
on contact with air and (1) where the term eroof represents the roof emissivity at the
 the air far from that surface, R is surface.
a residual function of PTs ; Ps, saturated pressure (Pa); R,
universal gas constant (J/(kmol K)); M, molecular mass (kg/ It has been assumed that surrounding surfaces that face
kmol); f, relative humidity, prev, previous iteration; A is the the building envelope and the building envelope surfaces are
nearly at the same temperature. In this way, the long-wave
straight-line
Ps
 coefficient from the approximation
radiation term was only considered in Eq. (12).
T ¼ AT þ B.
The solar radiation (direct, reflected and diffuse) came
2.2. Building envelope domain from models presented by ASHRAE [21] and are
conveniently projected to each surface considered in both
As the scope of this work is to analyze the coupled three- envelope and soil domains. In this way, the numerical value
dimensional heat and moisture transfer through the ground, a of ‘‘qr’’, shown in Eqs. (1)–(12), is modified according to the
simple one-dimensional conductive heat transfer model was projection of the solar beam at each simulation time step.
considered for the building envelope including walls and
roof (Fig. 2). In this way, the internal surface temperature is 2.3. Internal air domain
calculated by an energy balance equation, in an elemental
control volume, using the Fourier’s law: The present work uses a dynamic model for analysis of
hygrothermal behavior of a room without HVAC system.
@T @2 T Thus, a lumped formulation for both temperature and water
rc ¼l 2: (9)
@t @x vapor is adopted. Eq. (13) describes the energy conservation
equation applied to a control volume that involves the room
On the external side of the room, the walls, roof, doors
air, which is submitted to loads of conduction, convection,
and windows are exposed to solar radiation and to
short-wave solar radiation, inter-surface long-wave radia-
tion and infiltration:

Fig. 2. Building envelope schematic representation.
dTint
Ėt þ Ėg ¼ rair cair Vair ; (13)
dt
where Ėt is the energy flow that crosses the room (W), Ėg the
internal energy generation rate (W), rair the air density (kg/
m3), cair the specific heat of air (J/(kg K)), Vair the room
volume (m3) and Tint is the room air temperature (8C).
The term Ėt , on the energy conservation equation,
includes loads associated to the building envelope (sensible
heat) and latent conduction from floor, fenestration
(conduction and solar radiation), and openings (ventilation
and infiltration).
308 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314

The sensible heat released by the building envelope and extremely time consuming. A third method used was the one
floor is calculated as: obtained by the Matlab program, which provides analytical
X
m expressions to be solved in a real simultaneous way.
Qs ðtÞ ¼ hint Ai ½Tx¼e ðtÞ  Tint ðtÞ (14) However, these expressions are time consuming due to their
i¼1 great size, even though they require less iterations due to the
numerical robustness.
for the sensible conduction load and as:
In order to avoid limitations such as the requirement of
X
n
small time steps and high computer run time, it was shown
Qfloor ðtÞ ¼ LðTy¼H ðtÞÞhmf A j;f ½rv;int ðtÞ  rv;f ðtÞ (15) that the use of a semi-analytical method could be a good
j¼1
strategy to solve the differential governing equations for the
for the latent load from floor. room air, as it combines robustness and rapidness, which are
In Eq. (14) Ai represents the area of the i-th surface (m2), important criteria in whole-building simulation programs.
hint the convection heat transfer coefficients (W/(m2 K)), This last method solves analytically each equation (mass and
Tx=e(t) the temperature at the i-th internal surface of the energy balances), but with numerical coupling between each
building (8C) and Tint(t) the room air temperature (8C). In other. In this way, for the proposed problem, Eqs. (13) and
Eq. (15), n is the number of control volumes of the floor (16) can be written as:
surface discretized by using the finite-volume method; L, the dTint
vaporization latent heat (J/kg); hmf, the floor mass A  BTint ¼ rcV (17)
dt
convection coefficient (m/s); Aj,f, the area of j-th control
volumes of the floor surface discretized by using the finite- and
volume method (m2); rv,int, internal water vapor density (kg/ dWint
m3); rv,f the water vapor density of each control volume (kg/ C þ DWint ¼ rV ; (18)
dt
m3). The temperature and vapor density are calculated by the
combined heat and moisture transfer model described in which gives
Section 2.1 by using the values of temperature, moisture B
ercV Dt ðA þ BT0 Þ  A
content and sorption isotherm. Tint ¼ (19)
B
In terms of water vapor mass balance, different
contributions were considered: ventilation, infiltration, and
internal generation, people breath and floor surface. In this D

way, the lumped formulation becomes: erV Dt ðC þ DW0 Þ  C

Wint ¼ ; (20)
D
ðṁinf þ ṁvent ÞðWext  Wint Þ þ ṁb þ ṁger
  where T0 and W0 are the temperature and the humidity ratio
Xn
dWint calculated at the previous time step and
þ hD A j;f Wv;f  Wint ¼ rair Vair ; (16)
j¼1
dt X
m
A¼ hAi Ti þ Ėgs þ Ėgl ðWprev Þ;
where ṁint is the mass flow by infiltration (kg/s); ṁvent , the i¼1
mass flow by ventilation (kg/s); Wext the external humidity X
m

ratio (kg water/kg dry air); Wint, the internal humidity ratio B¼ hAi ;
i¼1
(kg water/kg dry air); ṁb , the water vapor flow from the X
n
breath of occupants (kg/s); ṁger , the internal water-vapor C ¼ ṁinf Wext þ ṁresp ðTprev Þ þ ṁger þ hD A j;f Wv;f and
generation rate (kg/s); hD, the mass transfer coefficient (kg/ ! j¼1
m2 s); Aj,f represents the area of j-th control volumes of the X m

floor surface (m2); Wv,f, the humidity ratio of each control
volume (kg water/kg dry air); rair, the air density (kg dry air/
m3); Vair is the room volume (m3).
The water-vapor mass flow from the people breath is
calculated as shown in ASHRAE [21], which takes into
account the room air temperature, humidity ratio and
physical activity as well.
Santos and Mendes [22] presented and discussed
different numerical methods used to integrate the differential
governing equations in the air domain (Eqs. (13) and (16)),
showing the results in terms of accuracy and computer run
time. In their analysis, it was shown that the use of explicit
methods such as Euler and Modified Euler requires
imperatively very small time steps, making simulations
D ¼  ṁinf þ ṁvent þ hD A j;f ;
i¼1
with Ėgs as the sensible energy (infiltration + generation) (W),
Ėgl , the latent energy (infiltration + generation) (W), Ti, the
temperature of each surface of the building envelope (8C).
3. Simulation procedure
A C-code was elaborated for the prediction of the
building hygrothermal performance. For the simulation, a
25-m2 single-zone building with two windows (single glass
layer) and one door, distributed as shown in Fig. 3 was
considered. For the conduction load calculation, 0.19-m
 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314 309

Fig. 4. Dimensions of the soil and floor domain used in the simulation.

Fig. 3. Dimensions of the single-zone building studied (units are in meters).

In this work, the properties of soil (sandy silt soil and
sand) strongly affected by temperature and moisture content
thick walls composed of three layers were used: mortar were taken from Oliveira et al. [25]. The basic dry-basis
(2 cm), brick (15 cm) and mortar (2 cm). In those typical material properties are shown in Table 2.
Brazilian walls, the contact resistance between two different In this work, internal generation of both energy and mass
layers was not considered. was not considered and an infiltration rate of 1 L/s was
The differential equations of energy conservation for adopted. The Sun effect (short-wave radiation) on the
each node of the building envelope (walls and roof) were ground was considered on East side until noon, and on West
discretized by using the finite-volume method [23], with a side, from noon until 6:00 p.m. On North side, solar
central difference scheme, a uniform grid and a fully implicit radiation was considered during all day and at no moment on
approach. The solution of the set of algebraic equations was the South side as the building is located in the city of
obtained by using the TriDiagonal-Matrix Algorithm Curitiba (south of Brazil at a latitude of 25.48).
(TDMA). The external climate was represented by Eqs. (21)–(23)
For the building envelope, a one-dimensional model was for temperature, relative humidity and solar radiation,
considered since the temperature gradients are much higher respectively. It was considered a yearly average temperature
on the normal direction. The thermophysical properties of of 20 8C with a daily variation of 5 8C. A yearly variation of
the building envelope materials were gathered from 5 8C was considered for the peak values. For the external
Incropera [24] and considered constant as shown in Table 1. relative humidity, a daily variation between 50 and 70% was
For the presented single-zone building, a 0.35-m concrete considered. A yearly variation on the peak values is not used
floor was considered within the soil domain as it can be seen in this case.
in Fig. 4. The governing partial differential equations were The value for total solar radiation (direct + diffuse) is
discretized by using the control-volume formulation method valid between 6:00 a.m. and 6:00 p.m., with a peak value at
[23]. The spatial interpolation method used is the control- noon, and, elsewhere, it is equal to zero.
difference scheme (CDS) and the time derivatives are


integrated using a fully implicit approach. In the three- pt pt
Text ¼ 20 þ 5 sin p þ þ 5 sin p þ
dimensional model for the ground and floor, an amount of 31536000 43200
9261 nodes were used. (21)
The MultiTriDiagonal-Matrix Algorithm (MTDMA;
[16]) was used to solve a three-dimensional model to

pt
describe the physical phenomena of the strongly coupled fext ¼ 0:60  0:10 sin p þ (22)
43200
heat and mass transfer in porous soils. In this algorithm, the
dependent variables are obtained simultaneously, avoiding


pt 3p pt
numerical divergence caused by the evaluation of coupled qrad ¼ 300 þ 100 sin p þ sin þ
terms from previous iteration values. 31536000 2 43200
(23)
Table 1
Thermophysical properties [24]
Material l (W/(m K)) r (kg/m3) cp (J/(kg K)) Table 2
Dry-basis properties of soils
Mortar 0.72 1860 780
Brick 0.72 1920 835 Soil r0 (m3/kg) cm (J/(kg K)) Porosity
Wood 0.16 720 1255 Sandy silt 1280 880 0.517
Glass 1.4 2500 750 Sand 1650 800 0.380
310 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314
Fig. 5. Temperature profiles of the sandy silt soil for different time steps.
4. Results
The boundary conditions, the pre-simulation time period
(warm-up), the size of the physical domain, the simulation
time step, the grid refinement, the convergence errors and the
required computer run time are important simulation
parameters, which have to be chosen very carefully in
order to accurately predict temperature and moisture content
profiles in soils under different sort of weather data.
Fig. 6. Moisture content profiles of the sandy silt soil for different time
steps.
Fig. 7. Temperature profiles of the sandy silt soil for different node
numbers.
Fig. 8. Moisture content profiles of the sandy silt soil for different node
numbers.
In the sensitivity analysis, the results presented in Figs. 5–
10 were obtained by carrying out simulations for the sandy
silt soil only. In Section 4.4, the comparison results shown
in Figs. 11–17 were obtained for both sandy silt and sand
soils.
In all cases, sinusoidal functions (Eqs. (21)–(23)) have
been considered to represent the weather. In the soil domain
of the building located in the city of Curitiba-Brazil (south
latitude of 25.48), a constant convection heat transfer
coefficient of 10 W/(m2 K), an absorptivity of 0.5 and a
constant long-wave radiation loss of 30 W/m2 were
considered as a boundary condition at the upper surface.
Fig. 9. Temperature profiles in a soil physical domain with a 5-m depth.
Fig. 10. Moisture content profiles in a soil physical domain with a 5-m
depth.
 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314
Fig. 11. Variation of the internal temperature after two years of pre-
simulation.
Fig. 12. Internal humidity ratio after two years of pre-simulation period.
The other surfaces were considered adiabatic and imperme-
able.
In the analysis of moisture effects on indoor air (Section
4.4), three different boundary conditions were considered
for the external ground upper surface: (i) solar radiation; (ii)
rain with no solar radiation effect; (iii) rain followed by solar
radiation. For the purely conductive model, rain was not
considered.
Fig. 13. Heat transfer rate from floor (25 m2) in the simulation period of one
month.
311
Fig. 14. Internal temperature after two years of pre-simulation period with
an infiltration rate of 10 L/s.
Fig. 15. Internal temperature considering soil and floor as sand.
4.1. Time step sensitivity
For the time step sensitivity analysis, the temperature and
moisture content profiles presented in Figs. 5 and 6 cor-
respond to the one obtained on December 31st at 12:00 p.m. of
the fifth year of simulation period. In this comparison, time
step of 1, 10, 30 and 60 min were considered.
A maximum temperature difference of approximately
0.3 8C on the ground surface was noticed in Fig. 5, by
comparing the different time steps. However, the moisture
content did not present any plausible difference, as it can be
Fig. 16. Internal humidity ratio considering soil and floor as sand.
312 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314
Fig. 17. Heat transfer rate from floor considering soil and floor as sand in
the simulation period of one month.
seen in Fig. 6. The low sensitivity to the time step was due to
the use of a robust numerical method [26].
4.2. Grid size refinement
In a similar way, in the analysis of sensitivity to the grid
size refinement, the values of temperature and moisture
content correspond to the one attained at 12:00 p.m. on
December 31st of the fifth year of simulation period. The
number of nodes (300, 500 and 1000) was varied on the
vertical direction only, since this is the direction where the
temperature and moisture content gradients are predomi-
nant, when no solar shading effect is considered. A time step
of 10 min was used in this case.
Increasing the number of nodes on the gravity direction
axis from 300 to 1000, a maximum temperature difference of
1.8 8C (Fig. 7) is noticed. On the other hand, according to the
results presented in Fig. 8, the moisture content profile
showed a highest absolute difference of nearly 0.1 (vol.%) at
the soil top surface.
The simulation performed showed a higher sensitivity to
the grid refinement than to the time step, which has to be
carefully done as the required computer run time is much
more sensitive to the grid size than to the time step.
4.3. Depth sensitivity
Santos and Mendes [27] have presented results compar-
ing soil domains with different depth, showing that a 5-m
depth soil domain could provide reasonable results,
especially near the upper surface. In this way, for the
weather functions presented in Eqs. (21)–(23), the tem-
perature and moisture content distributions vary with time
according to the results presented in Figs. 9 and 10, for a 1-h
time step and a grid of 4961 nodes.
It is possible to notice in Figs. 9 and 10 that the
temperature and moisture content values remain constant, in
essence, at the upper surface. A very slow evolution
especially in the moisture content distribution was
remarked. The time evolution differences in the temperature
and moisture content profiles are mainly attributed to the
high thermal and hygric soil capacities.
4.4. Moisture effects on indoor air
A warm-up period of two years was employed for the
analysis of the moisture effects on the heat and mass transfer
calculation through the floor and ground. In Figs. 11–14, for
the soil a sandy-silt type was considered, which properties
were gathered from Oliveira et al. [25] and for the floor,
properties from mortar [28] were used. Those strongly
moisture-content-dependent properties include specific heat,
density, thermal conductivity and liquid and vapor transport
coefficients associated to temperature and moisture content
gradients.
Three different boundary conditions were considered for
the external ground upper soil surface: (i) solar radiation; (ii)
rain with no solar radiation effect; (iii) rain followed by solar
radiation. For the purely conductive model, rain was not
considered.
Convection heat transfer has been taken into account in
all cases. For all external surfaces, a constant convection
heat transfer coefficient of 12 W/(m2 K) was adopted and,
for all internal surfaces, 3 W/(m2 K) was considered. A
constant long-wave radiation loss of 100 W/m2 was
attributed for the ground and roof as well as an absorptivity
of 0.5 for the ground and 0.3 for other surfaces. In the soil
domain, the laterals and lower surface were considered
adiabatic and impermeable.
In Figs. 11–17, a temperature of 20 8C and a volumetric
moisture content of 4% were considered as initial conditions.
Fig. 11 shows the variation of the room air temperature.
During the simulation period, it was not verified a significant
variation between the purely conductive model and the
model that takes moisture transport through the ground and
floor into account. The decrease of the internal temperature
in the rain case is mainly due to the absence of solar radiation
during the 1-month period for the first case and 3 days, for
the second case. Therefore, it can be clearly noticeable that
solar radiation, in this case, is the uppermost thermal load
source considered in the room air energy balance.
A higher difference of approximately 15% on the internal
humidity ratio was verified in Fig. 12, when the purely
conductive model for the ground and floor is considered. It
was also noticed a low daily variation of humidity ratio when
moisture is not neglected, i.e., a lower buffer effect of
humidity in air occurs when moisture buffer capacity
materials are employed.
In the case moisture is disregarded, water vapor is only
exchanged by ex- or infiltration. Nevertheless, when
moisture is considered, the vapor flow exchanged with the
floor can be a significant counterpoise on the room air
moisture balance, providing different results in terms of
room air relative humidity.
In this context, some research [29,30] has also been
conducted to aware about the importance of indoor relative
 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314
humidity, which significantly affects: (i) thermal and
respiratory comfort; (ii) perception of indoor air quality;
(iii) occupant health; (iv) durability of building materials (v)
energy consumption. For example, there will be twice as
many occupants dissatisfied with the indoor comfort
conditions at 24 8C and 70% relative humidity than at
24 8C and 40% [30]. At the same time, the occupants will
perceive the indoor air quality (IAQ) to be better at lower
humidity (in fact enthalpy) and recent research results show
that ventilation rates could be decreased notably by
maintaining a moderate enthalpy in spaces. Humidity also
influences the growth of dust mites and fungi and the
occurrence of respiratory infections, with values between 30
and 55% relative humidity recommended [30].
In addition, Salonvaara and Ojanen [31] showed also that
materials with hygroscopic capacity have the ability to
improve the performance of building envelope structures
even to such level that condensation and mold growth
conditions are eliminated.
Furthermore, there is a need to improve the indoor air
relative humidity prediction by incorporating accurate
moisture models in building thermal simulation programs
[32]. The realization of this need led to formation of an
International Research Project, in the framework of the
International Energy agency, which started in late 2003 and
comprising as many as 19 countries [30].
In Fig. 13, the heat exchanged with the floor can be
observed for the same three scenarios presented for the
relative humidity results (Fig. 12). The absence of solar
radiation during the rain period, causes the decrease of the
internal temperature of the building, increasing the heat flux
from the floor. In this case, difference of up to 100% was
observed in Fig. 13, for the heat flux between the model that
does not consider moisture and the one that considers
moisture with boundary condition of rain.
Fig. 14 shows the evolution of room air temperature when
an infiltration rate of 10 L/s is applied. In this case, a
maximum difference of 0.5 8C was verified between the
peak values, attributed to the increase of vapor concentration
difference between internal air and floor surface, caused by
an increase of the infiltration flow.
In Figs. 15–17, a sand floor was considered in order to
observe a higher evaporation rate potential. In this case, the
effect of solar radiation was also ignored and an infiltration
rate of 20 L/s was adopted.
Fig. 15 shows a small difference between the temperature
peaks for the two models.
Fig. 16 illustrates a difference of 10% on the internal
humidity ratio for the two first days of simulation. Although
the values of heat transfer rate does not present a significant
variation on the internal temperature, it is observed, in
Fig. 17, a great difference of up to 110% between the results.
In this way, the heat transfer rate at the floor surface can
be very important on the calculation of sensible heat factor
for the building, depending on the direction of the latent and
sensible heat flux exchanged with the floor.
313
5. Conclusion
Building simulation codes present some simplifications
on their calculation routines of heat transfer through the
ground. Regarding the ground heat transfer some aspects
should be clarified: the multidimensional phenomenon; the
transient behavior; the great number of involved parameters,
mainly when moisture is considered.
A review in the literature showed divergences about
moisture effects on the ground heat transfer. In addition,
most of programs do not consider the three-dimensional
asymmetrical effect in the soil.
In this way, some comparison exercises were carried out
to show how important moisture in soils can be in different
scenarios, taking into account a coupled three-dimensional
heat and moisture transfer model integrated to a single-zone
building model.
Simulations carried out for sensitivity analyses of time step
and grid size shown the importance of refinement at the upper
surface, which is in contact with air. On the other hand, the
results were not so sensitive to time step, allowing the
adoption of high values of time step with no loss of accuracy
thanks to the robust algorithm and to the linearization of vapor
concentration difference at soil top surface.
We also noticed the importance of a pre-simulation
period of several years for the correct estimation of
temperature and moisture content profiles. As computer
run time plays an important role in long-term soil
simulations, a depth of 5 m was suggested whereas the
main temperature and moisture content gradients are close to
the upper surface, affecting somewhat the thermal behavior
of low-rise buildings.
Concerning the moisture effects on indoor air analysis, it
is showed a very slight difference in terms of room air
temperature between the purely conductive model for the
ground and the moisture model, due to the importance of
solar gains compared to ground heat transfer in the room
energy balance and due to the different time scales between
room air and soil. However, a significant difference of 15%
was noticed on the room air humidity ratio, even for the high
infiltration rate case. Therefore, higher energy consumption
could be expected when an air conditioning system is used
due to the augmentation of building latent loads.
Besides the energy related aspects, moisture models shall
be incorporated to building simulation codes due to other
indoor relative humidity effects such as perception of indoor
air quality, occupant’s health and durability of building
materials.
Regarding the heat flux through building floor, it was
noted a small thermal load contribution in both models.
However, the absence of solar radiation during the rain
period caused a room temperature reduction, occasioning an
increase of ground heat transfer of approximately 100%.
In order to increase the evaporative flux, the floor surface
was considered composed uniquely of sand. In this case, no
significant variations on the room temperature values were
314 G.H. dos Santos, N. Mendes / Energy and Buildings 38 (2006) 303–314
verified between the two models. However, these results
could be more expressive for the calculation of sensible heat
factor for building zones, depending on the direction of the
latent and sensible heat fluxes at the floor surface.
Although the moisture effect has caused no significant
difference on the room air temperature, a building with a
greater area of contact with the ground or in underground
zones where the solar radiation effect is not predominant, the
moisture flux through the floor could contribute more
effectively for the room air energy balance.
To conclude, some recommendations are addressed for
further work: (i) Simulation of underground zones and
include the moisture effects on the building envelope as
well; (ii) determination empirical correlations for ground
heat transfer; (iii) improvement of the rain model.
Acknowledgments
The authors thank the Brazilian Research Council
(CNPq) of the Secretary for Science and Technology of
Brazil for support of this work.
References
[1] H.M. Künzel, Simultaneous heat and moisture transport in building
components: one- and two-dimensional calculation using simple
parameters, Fraunhofer-Informationszentrum Raum and Bau, IRB
Verl, Stuttgard, 1995.
[2] N. Mendes, Modelos para previsão da transferência de calor e de
umidade em elementos porosos de edificações, Tese de Doutorado,
Universidade Federal de Santa Catarina (UFSC), 1997.
[3] A. Holm, H.M. Kuenzel, K. Sedlbauer, The hygrothermal behavior of
rooms: combining thermal building simulation and hygrothermal
envelope calculation, in: Eighth International IBPSA Conference,
vol. 1, International Building Performance Simulation Association,
Eindhoven, Netherlands, 2003, pp. 499–505.
[4] W.P. Bahnfleth, Three-dimensional modelling of slab-on-grade heat
transfer, in: Building Simulation Conference—IBPSA 89, 1989, 133–
138.
[5] M. Davies, A. Tindale, J. Litter, Importance of multi-dimensional
conductive heat flow in and around buildings, Building Service
Engineering Research Technology 16 (2) (1995) 83–90.
[6] T. Blomberg, Heat Conduction in Two and Three Dimensions, Report
TVBH-1008, Lund University, Department of Building Technology,
Building Physics, 1996.
[7] S. Matsumoto, H. Yoshino, H. Akasaka, A program for ground
temperature data generation based on the expanded amedas weather
data cd-roms, in: Sixth International IBPSA (BS’99), Kyoto, Japan,
1999, pp. 869–876.
[8] M. Krarti, Effect of spacial variation of soil thermal properties on slab-
on-ground heat transfer, Building and Environment 31 (1) (1996) 51–57.
[9] M.P. Deru, A.T. Kirkpatrick, Ground-coupled heat and moisture
transfer from building. Part 2. Aplication, Journal of Solar Energy
Engineering 124 (2002) 17–21.
[10] G.J. Brink, C.J. Hoogendoorn, Ground water flow heat losses for
seasonal heat storage in the soil, Solar Energy 30 (4) (1983) 367–371.
[11] D.S. Freitas, A.T. Prata, Thermal performance of underground power
cables with constant and cyclic currents in presence of misture
migration in the surrounding soil, IEEE Transactions on Power
Delivery 11 (3) (1996) 1159–1170.
[12] D. Ogura, T. Matsushita, M. Matsumoto, Analysis of heat and moisture
behavior in underground space by quasilinearized method, in: Sixth
International IBPSA Conference (BS’99), Kyoto, Japan. 2, 1999, pp.
755–762.
[13] R. Neves, P.C. Leitão, Modelação numérica da circulação da água no
solo, O modelo MOHID, Instituto Superior Técnico, Universidade
Técnica de Lisboa, Pedologia, Oeira, vol. 28, 2000, pp. 46–55.
[14] G.H. Galbraith, C. Ji, R. Mclean, P. Baker, The influence of space
discretization on the accurancy of numerical simulation of heat and
moisture transport in porous building materials, Journal of Thermal
Envelope and Building Science 25 (2) (2001) 143–160.
[15] J. Wang, C.E. Hagentoft, A numerical method for calculating com-
bined heat, air and moisture transport in building envelope compo-
nents, Nordic Journal of Building Physics 2 (2001) 1–17.
[16] N. Mendes, P.C. Philippi, R. Lamberts, A new mathematical method to
solve highly coupled equations of heat and mass transfer in porous
media, International Journal of Heat and Mass Transfer 45 (2002)
509–518.
[17] J.R. Philip, D.A. de Vries, Moisture moviment in porous media under
temperature gradients, Transactions of the American Geophysical
Union 38 (1957) 222–232.
[18] H. Janssen, J. Carmeliet, H. Hens, The influence of soil moisture in the
unsatured zone on the heat loss from building via the ground, Journal
of Thermal Envelope and Building Science 25 (4) (2002) 275–298.
[19] S. Onmura, M. Matsumoto, S. Hokoi, Study on evaporative cooling
effect of roof lawn gardens, Energy and Building 33 (2001) 653–666.
[20] G.H. Santos, N. Mendes, Multidimensional effects of ground heat
transfer on the dynamics of building thermal performance, ASHRAE
Transactions 110 (2004) 345–354.
[21] American Society of Heating Refrigeration and Air-Conditioning
Engineering (ASHRAE), Handbook-Fundamentals, ASHRAE,
Atlanta, 1997.
[22] G.H. Santos, N. Mendes, Analysis of numerical methods and simula-
tion time step effects on the prediction of building thermal perfor-
mance, Applied Thermal Engineering 24 (2004) 1129–1142.
[23] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere
Publishing Corporation, 1980.
[24] F.P. Incropera, Fundamentos de Transferência de Calor e de Massa,
fourth ed., LTC Editora, 1998.
[25] A.A.M. Oliveira Jr., D.S. Freitas, A.T. Prata, Influência das Proprie-
dades do Meio nas Difusividades do Modelo de Phillip e de Vries em
Solos Insaturados, Relatório de Pesquisa, UFSC, 1993.
[26] N. Mendes, P.C. Philippi, Multitridiagonal-matrix algorithm for
coupled heat transfer in porous media: stability analysis and computa-
tional performance, Journal of Porous Media 7 (3) (2004).
[27] G.H. Santos, N. Mendes, Unsteady combined heat and moisture
transfer in unsaturated porous soils, Journal of Porous Media 8 (6)
(2005).
[28] B. Perrin, Etude des Transferts Couplés de Chaleur et de Masse dans
des Matériaux Poreux Consolidés non Saturés Utilisés en Génie Civil,
Thèse Docteur d’Etat, Université Paul Sabatier de Toulouse, Toulouse,
France, 1985.
[29] H.J. Moon, G. Augenbroe, Evaluation of hygrothermal models for
mold growth avoidance prediction, in: Eighth International IBPSA
Conference, vol. 2, International Building Performance Simulation
Association, Eindhoven, Netherlands, 2003, pp. 895–901.
[30] H. Hens, Proposal for a New Annex. Whole Building Heat, Air and
Moisture Response (MOIST-ENG), Katholieke Universiteit Leuven,
Belgium, 2003.
[31] M. Salonvaara, T. Ojanen, A. Karagiozis, Indoor air humidity varia-
tions and its effects on the moisture performance of building envelope,
in: Eighth International IBPSA Conference, vol. 3, International
Building Performance Simulation Association, Eindhoven, Nether-
lands, 2003, pp. 1163–1169.
[32] C. Rode, N. Mendes, K. Grau, Evaluation of Moisture Buffer Effects
by Performing Whole-Building Simulation, ASHRAE Transactions,
USA 110 (2) (2004) 783–794.