GEOSCIENCE FRONTIERS 2(3) (2011) 463e474

available at www.sciencedirect.com

China University of Geosciences (Beijing)

GEOSCIENCE FRONTIERS
journal homepage: www.elsevier.com/locate/gsf

ORIGINAL ARTICLE

Finite element modeling of pore-fluid flow in the

Dachang ore district, Guangxi, China: Implications
for hydrothermal mineralization
Minghui Ju a,b, Chongbin Zhao a,b,*, Tagen Dai a,b, Jianwen Yang c

a
Computational Geosciences Research Center, Central South University, Changsha 410083, China
b
The Key Laboratory of Chinese Education Ministry for Non-ferrous Metal Metallogenic Prediction,
Central South University, Changsha 410083, China
c
Department of Earth and Environmental Sciences, University of Windsor, Canada

Received 6 January 2011; accepted 1 April 2011

Available online 14 June 2011

KEYWORDS Abstract Convective heat transfer associated with the circulation of pore-fluid in porous rocks and
Finite element modeling; fractures within the upper crust of the Earth is substantial when the temperature gradient is suffi-
Dachang ore district; ciently high. In order to understand the process of Sn-polymetallic mineralization in the Dachang
Hydrothermal ore district of Guangxi, a finite element method has been used in this study to simulate both
mineralization; pore-fluid flow and heat transfer in this district. On the basis of related geological, tectonic and
Salinity-induced buoyancy geophysical constraints, a computational model was established. It enables a computational simula-
tion and sensitivity analysis to be carried out for investigating ore-forming pore-fluid flow and other
key factors that may affect hydrothermal ore genesis in the district. The related simulation results
have indicated that: (1) permeable fault zones in the Dachang ore district can serve as preferential
pathways for pore-fluid flow on a regional-scale; and (2) the pore-fluid flow can affect the salinity

* Corresponding author. Computational Geosciences Research Center,

Central South University, Changsha 410083, China.
E-mail address: chongbin.zhao@iinet.net.au (C. Zhao).
1674-9871 ª 2011, China University of Geosciences (Beijing) and Peking
University. Production and hosting by Elsevier B.V. All rights reserved.

Peer-review under responsibility of China University of Geosciences

(Beijing).
doi:10.1016/j.gsf.2011.05.008

Production and hosting by Elsevier

464 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
distribution. This latter factor is part of the reason why Sn-polymetallic mineralization has taken
place in this district.
ª 2011, China University of Geosciences (Beijing) and Peking University. Production and hosting
by Elsevier B.V. All rights reserved.
1. Introduction
Hydrothermal circulation through porous rocks is an important
physical process for the genesis of hydrothermal ore deposits, and
numerical modeling can help understand such processes (Schmidt
Mumm et al., 2010; Zhao et al., 2010). This paper presents a case
study of the Dachang Sn-polymetallic deposits in Guangxi, China,
where the ore genesis has been debated and controlling factors of
mineralization have not been well understood. The host rocks of
the Dachang deposits are typically banded, consisting mainly of
siliceous rocks and limestones, with lesser, but considerable
amounts of alternating thin beds of sulfides and K-feldspar-rich
rocks (some of which have high BaO contents of 3.3 wt.%e
7.9 wt.%). Based on the mineral assemblages of the host rocks and
their structures and geochemistry, it has been proposed that these
banded rocks had a submarine hydrothermal-exhalative origin
(Han and Hutchinson, 1989; Han et al., 1993). On the other hand,
based on the association of the deposits with granites, it has been
suggested that the mineralization is related to hydrothermal
activities associated with granitic emplacement in the district
(Chen et al., 1985, 1993; Tanelli and Lattanzi, 1985). Despite
these controversies, it appears that the mineralization was
controlled by a combination of faults, strata (aquifers), and heat
sources. A preliminary numerical study has recently been con-
ducted to simulate tectonic deformation-driven fluid flow in the
Dachang ore district (Ju and Yang, 2010). This simulation shows
that hot basinal pore-fluids ascending along active faults may have
flowed laterally into the permeable aquifers, interacted with the
host rocks and resulted in the mineralization. In this paper, we
focus on the potential roles of heat and fluid salinities in
controlling convective pore-fluid flow.
In terms of studying the occurrence of convective pore-fluid
flow in porous media, Nield and Bejan (1992) summarized the
theoretical work that was carried out before 1992. Subsequently,
over the past 14 years, under the stimulus of understanding crustal
ore body formation and mineralization, Zhao and his coworkers
have conducted, both theoretically and computationally, system-
atic and extensive studies on convective heat transfer in geological
systems (Zhao et al., 1997, 1998a, 1998b, 1999a, 1999b, 2000,
2001, 2002, 2003, 2004, 2005, 2006a, 2006b, 2007, 2008a,
2008b, 2010). The results of these studies have been summa-
rized in a recently-published monograph (Zhao et al., 2008a) and
have led to many applications in different geological regions
(Hobbs et al., 2000; Gow et al., 2002; Ord et al., 2002; Sorjonen-
Ward et al., 2002; Schaubs and Zhao, 2002; K€uhn et al., 2006;
Yang et al., 2010; Gessner et al., 2009; Harcou€et-Menou et al.,
2009; Alt-Epping and Zhao, 2010; Fu et al., 2010; Magri et al.,
2010; Zhao et al., 2010).
Physically, convective pore-fluid flow in fluid-saturated porous
media can be caused by the following three mechanisms: thermal
diffusion, mass diffusion and a combination of both. As a result, the
resulting convective pore-fluid flow is respectively called
the thermal diffusion driven flow, the mass diffusion driven flow and
the double diffusion driven flow (Zhao et al., 2008a). Due to
a significant difference between the physical language and
geological language, these three kinds of convective pore-fluid
flows are called, respectively, in geological terms, thermal
convective flow, salinity convective flow and thermohaline
convective flow (Person and Garven, 1994).
Salinity convective flow, which is driven by salinity only, is
important in enhancing the hydrodynamic mixing of a dense
solute with less dense ambient groundwater in the field of
contaminant hydrology, where only isothermal conditions are
considered (Frind, 1982; Simmons and Narayan, 1997; Robinson
et al., 1998; Shikaze et al., 1998; Smith and Turner, 2001).
However, it is unrealistic in the upper crust of the Earth, where
geothermal gradient exists in nature. For this reason, thermohaline
convective flow, rather than salinity convective flow, should be
considered in geological systems. Because high salinity conditions
in surface and subsurface evaporative deposits are an important
factor in the development of basinal brines responsible for the
formation of base metal ore deposits (Davidson, 1998; Cooke
et al., 2000), it is necessary to investigate the relationships
between salinity distributions and ore-forming processes in
hydrothermal systems.
As one of the most important roles played by the emerging
computational geoscience in clarifying different views of a geological
process (Zhao et al., 2008a, 2009), the computational simulation
method, namely the finite element method in this particular case, is
used to test the occurrence of convective pore-fluid flow and its effect
on the ore-forming processes in the Dachang ore district. Accordingly,
a series of numerical analyses have been conducted to investigate the
effects of fault zone width, hydraulic conductivity, salinity and other
factors on the pore-fluid flow and heat transfer in this district.
2. Geological background
The Dachang ore field of South China (Fig. 1) is located in the
center of the northwest-striking NandaneHechi fold-and-thrust
belt, at the border between the Jiangnan anticline and the Tien-
gui syncline (Cai et al., 2004, 2007). The NandaneHechi district
underwent subsidence and faulting in the Devonian and Carbon-
iferous, folding in the Middle Triassic, and extension and block
faulting during the Cretaceous (Cai et al., 2004, 2007). Cretaceous
tectonic activities controlled the emplacement of contempora-
neous granitoid intrusions (Chen et al., 1985; Cai et al., 2004,
2007). In the NandaneHechi fold-and-thrust belt, the major
deposits located from northwest to southeast include the Mayang
Hg deposit, the Mangchang and Dachang Sn deposits, and the
Beixiang and Furongchang PbeZn deposits (Qin et al., 1998).
These deposits are hosted in Devonian to Triassic sedimentary
rocks that were intruded by Yanshanian (Late Mesozoic) granit-
oids (Cai et al., 2004, 2007).
In the Dachang ore district, the main strata include the Mid-
Devonian Nabiao and Luofu groups, the Upper-Devonian Liu-
jiang, Wuzishan and Tongchejiang groups, the Carboniferous
Shimen and Huanglong groups, and Permian strata (Yang et al.,
 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474 465

3. Computational methodology

3.1. Brief introduction of the finite element method

The finite element method (FEM) is perhaps the most widely used
numerical method across science and engineering fields. Since its
origin in the late 1950s and early 1960s, much FEM development
work has been specifically oriented toward structural engineering
and rock mechanics problems. This is because it was the first
numerical method with enough flexibility for dealing with mate-
rial heterogeneity, nonlinear deformation, complex boundary
conditions, in situ stresses and gravity. The method appeared in
the field of geotechnical engineering during the late 1960s and
early 1970s, when the traditional finite difference method (FDM)
with regular grids could not satisfy the essential requirements for
simulating rock mechanics problems.
In the finite element method, a problem domain is discretized by
an array of nodes and associated finite elements in a mesh pattern.
Associated with the mesh are nodal points, where the discretized
governing equations are solved to obtain unknown values. The
elements form the framework of the numerical model. In this study
the finite element method, which forms the methodological base of
an existing code (called FEFLOW; see Diersch, 2002), is employed
with a triangular mesh.
Prior to actual computation, it is important to design the finite
element mesh and to parameterize the boundary and initial condi-
tions properly, based on a conceptual model of the problem. The
distribution of the mesh needs to meet the standard error tolerance.
Calibration is necessary and important, because it helps to establish
a more reasonable geological model and produce more reliable
results. Calibration includes the adjustment of the conceptual
model, boundary/initial conditions and physical parameters.
Figure 1 Magma distributions in the Danchi metallogenic belt (after
Zeng et al., 1995).
3.2. Governing equations of the problem

1999; Pasava et al., 2003). Among them the Devonian strata are The FEFLOW code solves numerically the following governing
the main ore-bearing strata with the lithologies dominated by equations that describe the pore-fluid flow, mass transport and heat
carbonaceous shale, mudstone, reef limestone, lenticular and transfer, respectively (Diersch, 2002):
banded limestone, and siliceous rock (Yang et al., 1999). In vh vqi
S0 þ ZQr þ QEB ðC; TÞ ð1Þ
addition, the Permian strata are dominated by siliceous rocks and vt vxi
sandstone (Pasava et al., 2003).
The main structures are the NW-trending NandaneHechi (or

vC v vC vC
Danchi for short) anticline and the NandaneHechi fault, with f  Dij þ qi þ Qr CZQC ð2Þ
vt vxi vxj vxi
a series of parallel subordinate folds and faults on both sides (Cai
et al., 2007). A gravity survey indicated that the NW-trending

 vT v vT vT
faults possibly reach the lower crust and perhaps even the upper f rf cf þ ð1  fÞrs cs  lij þ rf cf qi
mantle (Cai et al., 2004, 2007). The igneous rocks are mainly vt vxi vxj vxi
Yanshanian granitoids with the lithologies dominated by biotite þ rf cf Qr ðT  T0 ÞZQT ð3Þ
granite, alaskite, monzonitic granite, pegmatite, granite-porphyry,
and dioritic porphyrite (Yang et al., 1999). The biotite granite is The Darcy equation for variable-density flow is:


a typical S-type granitic stock that crops out near the axis of the vh rf  rf 0
Longxianggai anticline in the central part of the Dachang ore field qi Z  Kij fm þ ej ð4Þ
vxj rf 0
and covers an area of approximately 0.5 km2 (Yang et al., 1999;
Cai et al., 2007). Alaskite, monzonitic granite and pegmatite where qi is the Darcy velocity of the pore-fluid in the xi direction,
occur as dikes derived from the Longxianggai granite stock; h is the hydraulic head, t is time, S0 is the specific storage coef-
granite-porphyry and dioritic porphyrite dikes are situated west of ficient of the porous medium, QEB(C,T ) is the term of the
the ore field (Yang et al., 1999). Field investigation has revealed extended Boussinesq approximation (Diersch, 2002), T0 is the
that the Dachang ore district is spatially associated with the reference temperature, Qx is the source/sink function for the pore-
Longxianggai granite, with the ore zones lying in the contact fluid (x Z r), the solute mass (x Z C ) or heat (x Z T ), f is
metamorphosed sedimentary rocks above the roof of the granite porosity, C is the concentration of the chemical component,
(Lattanzi et al., 1989; Cai et al., 2007). lij Zf lfij þ ð1  fÞlsij is the second-order thermal conductivity
466 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
Figure 2 Conceptual model for coupled pore-fluid flow and heat transfer in the Dachang ore district.
tensor of the porous medium, lfij and lsij are the second-order
thermal conductivity tensors of the pore-fluid and solid matrix
respectively, fm is the dimensionless dynamic viscosity of the pore-
fluid, rf and rf0 are the density and reference density of the pore-
fluid, rs is the density of the solid matrix, cf and cs are the specific
heat of the pore-fluid and solid matrix respectively, Dij is the
second-order (solute) dispersion tensor, Kij is the second-order
hydraulic conductivity tensor.
kij rf 0 g
Kij Z ð5Þ
mf 0
where kij is the second-order permeability tensor, mf0 is the
reference dynamic viscosity of the pore-fluid, g is the acceleration
due to gravity.
These governing equations involve two important parameters,
pore-fluid density rf and dynamic viscosity mf, which may influence
the pore-fluid flow. Generally, the density of pore-fluid is dependent
on temperature, pressure, and salinity. For less compressible pore-
fluid such as water, the dependence of pore-fluid density on pres-
sure can be neglected without causing considerable errors. On the
other hand, for small to moderate density variations, the Boussinesq
approximation is valid so that the dependence of pore-fluid density
on salinity can be also neglected. As a result, the temperature
dependence of the pore-fluid density can be represented using the
following polynomial interpolation functions (Molson and Frind,
1993):
 
rf ðTÞZ1000  1 þ  0:435  105 þ 0:838  108
 
 ðT  4Þ  ðT  4Þ2 ð6Þ
However, in the computer code (i.e. FEFLOW) used in this
study, Eq. (6) is replaced by the following sixth order polynomial
to improve the solution accuracy:
rf ðTÞZa þ bT þ cT 2 þ dT 3 þ eT 4 þ fT 5 þ gc T 6 ð7Þ
where the unit of the pore-fluid density is g/l, the unit of
temperature T is  C, a, b, c, d, e, f and gc are coefficients:
Figure 3 Computational model for coupled pore-fluid flow and heat transfer in the Dachang ore district.
aZ9:9957  102 ; bZ  8:45  102 ; cZ  4:10  103 ;
dZ8:0  106 ; eZ4:0  1016 ; f Z  9:0  1019 ;
gc Z  2:0  1021 ð8Þ
The dynamic viscosity dependence of temperature and solute
concentration was expressed using empirical polynomial functions
(Diersch, 2002). In FEFLOW (Diersch, 2002) the dimensionless
dynamic viscosity of the pore-fluid is assumed to be a function of
concentration C and temperature T as follows:
mf 0 1 þ 1:85 uðcZc0 Þ  4:1 u2ðcZc0 Þ þ 44:5 u3ðcZc0 Þ
fm Z Z
mf ðC; TÞ 1 þ 1:85 u  4:1 u2 þ 44:5 u3
1 þ 0:7063 2  0:04832 23
 ð9Þ
1 þ 0:7063 2ðTZT0 Þ  0:04832 23ðTZT0 Þ
C T  150
uZ ; 2Z ð10Þ
rf 100
where mf 0 Zmf ðCZC0 ; TZT0 Þ is the reference dynamic viscosity of
the pore-fluid at reference concentration C0 and reference
temperature T0, the unit of the pore-fluid density is g/l, the unit of
temperature T is  C.
The variable density of the pore-fluid in a coupled system can be
treated in different manners, depending on whether high-density
contrasts occur and density influences can become important in the
related terms of governing equations. For the most density-dependent
transport phenomena, the so-called Boussinesq approximation Qr is
often employed (Diersch, 2002). This enables a coupled nonlinear
system to be essentially simplified with the consideration of neces-
sary intrinsic coupling mechanisms for a free convection process.
Generally, the Boussinesq approximation is valid for small to
moderate density variations.
The Boussinesq approximation may become insufficient for
the large density variations of the pore-fluid where high concen-
tration brines or extremely high temperature gradients are
involved. In this case, an extra term needs to be added in the mass
 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474 467

Table 1 Major physical properties of the fault zones and host strata units.
Strata Lithology Hydraulic conductivity (m/s) Porosity Thermal Density of
Horizontal Vertical (%) conductivity rocks (kg/m3)
(W/(m  C))
Fault zone 108 108 28 2 2400
Permian Siliceous rocks, sandstone 1012 1014 18 2 2450
Carboniferous Dolomitic limestone 1011 1013 20 2 2576
Upper-Devonian Silicate, limestone 109 1011 22 2 2700
Middle Devonian Black shale 1012 1014 12 2 2844
Late Devonian Black shale, sandstone 1013 1015 10 3 2900

balance equation of the pore-fluid. In FEFLOW, this additional approximation is used in the mass conversation equation of pore-
term can be expressed as follows: fluid in FEFLOW.



a vC vT a vC vT
QEB ðC; TÞZ  f b  qi b ð11Þ
Cmax vt vt Cmax vxi vxi 3.3. Computational model of the problem
where Cmax is the maximum concentration of the system, a is the
ratio of the pore-fluid density difference resulting from the A conceptual model in Fig. 2 is established for the cross section
maximum concentration (i.e. Cmax) of the system to the reference along line AB in Fig. 1, with a set of assumptions that reflect the
pore-fluid density as follows (Diersch, 2002): subsurface geological conditions of the Dachang ore district as
well as other features that are relevant to pore-fluid flow in this
rf ðCmax Þ  rf 0 0:7Cmax
aZ z ð12Þ district. The conceptual model consists of five strata and three
rf 0 rf 0 major faults. Among the five strata, the Upper-Devonian stratum is
considered to be the most permeable aquifer. Based on this
b is the thermal volumetric expansion coefficient of the conceptual model, a computational model with the finite element
pore-fluid. mesh (Fig. 3) is constructed to investigate pore-fluid flow in the
1 h
ore district.
bðTÞZ  b þ 2cT0 þ 3dT02 þ 4eT03 þ 5fT04 þ 6gc T05 The upper boundary of the model represents a paleo-seafloor,
rf 0
 which is permeable, has a temperature of 15  C, and a seawater
þ c þ 3dT0 þ 6eT02 þ 10fT03 þ 15gc T04 ðT  T0 Þ salinity of 9 wt.% NaCl (Fu et al., 1993; Pasava et al., 2003). The
 lower boundary is assumed to be impermeable to pore-fluid flow
þ d þ 4eT0 þ 10fT02 þ 20gc T03 ðT  T0 Þ2
 and to have a constant heat flux of 70 mW/m2. Salinity along the
þ e þ 5fT0 þ 15gc T02 ðT  T0 Þ3 þðf þ 6gc T0 ÞðT  T0 Þ4 lower boundary is 20 wt.% NaCl as determined by fluid inclusion
i
þ gc ðT  T0 Þ5 ð13Þ data (Fu et al., 1993; Pasava et al., 2003). The two lateral
boundaries are assumed to be impermeable and adiabatic, so as to
where b, c, d, e, f and gc are the same coefficients as defined in avoid heat and mass exchange to the outside of the model from
Eq. (8). them. The initial temperature varies linearly with depth as a result
Eq. (11) is called the extended Boussinesq approximation and of applying the paleo-geothermal gradient of 30  C/km (Bachu
represents an additional term to incorporate mass-dependent and and Cao, 1992) to the computational model. The initial salinity
temperature-dependent compression effects. Note that the first is assigned to increase linearly with depth from the top boundary
term of this equation becomes unimportant if the temporal value of 9 wt.% NaCl to the bottom boundary value of 20 wt.%
changes in concentration and temperature vanish when the NaCl.
system approaches a steady state, while the second term of this The following parameters are commonly used in this investi-
equation only vanishes if the density gradient is essentially gation: reference concentration C0 is 9 wt.% NaCl, reference
orthogonal to the pore-fluid flow. The extended Boussinesq temperature T0 is 15  C, the specific heat of the pore-fluid is

Table 2 Major physical properties of the fault zones and host strata units.
Strata Lithology Hydraulic conductivity (m/s) Porosity Thermal Density of
Horizontal Vertical (%) conductivity rocks (kg/m3)
(W/(m  C))
Fault zone 107 107 28 2 2400
Permian Siliceous rocks, sandstone 1012 1014 18 2 2450
Carboniferous Dolomitic limestone 1011 1013 20 2 2576
Upper-Devonian Silicate, limestone 108 1010 22 2 2700
Middle Devonian Black shale 1012 1014 12 2 2844
Late Devonian Black shale, sandstone 1013 1015 10 3 2900
468 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
Figure 4 Temperature contours at 100,000 years. vmax Z 5.89  103 m/d in the case of less permeable aquifer and faults.
4174 J/(kg  C), the specific heat of the solid matrix is 920 J/
(kg  C), the thermal conductivity of the pore-fluid is 0.6 W/
(m  C), the specific storage coefficient of the porous medium is
104/m. Since the second-order (solute) dispersion tensor is
defined as a function of molecular diffusion and mechanical
diffusion in FEFLOW (Diersch, 2002), the molecular diffusion is
assumed to be 6.2  108 m2/s, whereas longitudinal and trans-
verse dispersivities are assumed to be 20 m and 2 m, respectively.
Other major physical properties of the fault zones and host strata
units are given in Table 1 or 2, depending on the specific cases
simulated in the computational model.
4. Simulation results and discussions
Two scenarios are considered to investigate the potential pore-
fluid flow patterns in the Dachang ore district. In the first
scenario, only the geothermal gradient is applied to the compu-
tational model. Since the salinity is neglected in this scenario,
geothermal convective pore-fluid flow can only occur if the
geothermal gradient applied to the model is high enough. In the
second scenario, both geothermal and salinity gradients are
considered in the computational model. This scenario allows
thermohaline convective pore-fluid flow to take place if the
salinity at the top is higher than that at the bottom of the
computational model.
4.1. Scenario 1: consideration of geothermal gradient only
Three cases are considered in this scenario. In the first case, it is
assumed that the Upper-Devonian aquifer is the most permeable
strata and that the fault zones are also more permeable than
surrounding rocks. In the second case, to examine the permeability
effects of the aquifer and fault zones on convective pore-fluid flow,
the permeabilities of both the aquifer and fault zones are increased
by one order of magnitude. For this reason, the first case is called the
less permeable aquifer and faults case, whereas the second case is
called the more permeable aquifer and faults case. In the third case,
different widths of the fault zones are considered in order to examine
their effects on pore-fluid flows in the computational model.
Figure 5 Temperature contours at 100,000 years. vmax Z 1.12  102 m/d in the case of more permeable aquifer and faults.
4.1.1. Less permeable aquifer and faults case
Table 1 shows the main physical parameters of the fault zones and
host strata units simulated in the computational model, while Fig. 4
shows the temperature distribution of the model at 100,000 years.
It can be seen from Fig. 4 that the temperature contours bend only
a little along the fault zone due to the relative low values of the
hydraulic conductivity used for the Devonian aquifer and fault zones.
In this situation, the maximum velocity of the hydrothermal pore-
fluid in the fault zone is 5.89  103 m/d at 100,000 years of the
simulation time.
4.1.2. More permeable aquifer and faults case
In order to explore pore-fluid flow patterns at 100,000 years when
the Devonian aquifer and fault zones are more permeable, higher
values of hydraulic conductivity are assigned as shown in Table 2.
Fig. 5 shows the temperature distribution of the corresponding
computational model at 100,000 years. In this case, the maximum
value of the pore-fluid velocity can reach 1.12  102 m/d in the
fault zones. Since the maximum value of the pore-fluid velocity in
this case is much higher than that in the case of less permeable
aquifer and fault zones, the resulting temperature contours in the
current case bend with a higher curvature around the fault zone
than in the previous case. In the current case where the aquifer and
fault zones are more permeable, the temperature in the aquifer
around the fault zone can reach ca. 180  C, which is much higher
than the first case resulting aquifer temperature (ca. 80  C). These
results indicate that the value and distribution of hydraulic
conductivity can significantly control both the flux of ascending
hot pore-fluids and the temperature of the aquifer.
4.1.3. Effects of fault width on pore-fluid flow
To investigate how the fault width influences pore-fluid flow, the
width of the fault zone in the middle part of the computational
model is increased from 300 m in sub-Section 4.1.1 to 500 m. All
other parameters remain the same as those used in sub-Section
4.1.1.
Fig. 6 displays the temperature contours due to two different values
of fault zone width at 100,000 years. As expected, the increase of the
fault zone width causes an increase in the pore-fluid velocity in the
 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474 469

Figure 6 Temperature contours corresponding to different fault zone width: (a) 300 m and (b) 500 m.

fault zone, which in turn results in an increase in the temperature of the compared to that within the host formation. Our results indicate that
aquifer adjacent to the fault zone. Consequently, the basal hot pore- the fault provides an important pathway for the ascending hydro-
fluids are brought up into the aquifer with a maximum temperature thermal pore-fluids, which is driven by the buoyancy force caused by
of 142  C, and thus enhance a temperature increase within the fault, the temperature gradient applied to the computational model.

Figure 7 Temperature and salinity distributions at two different times: (a) at 10,000 years and (b) at 100,000 years.
470 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
4.2. Scenario 2: consideration of both geothermal and
salinity gradients
As demonstrated in a previous study (Zhao et al., 2000), the
consideration of a (downward) positive concentration gradient of
salinity (i.e. salinity decreases toward the top) in the upper crust of
the Earth can stabilize thermal convective pore-fluid flow; in other
words, it makes the thermal convective pore-fluid flow less likely to
occur in the upper crust of the Earth. To the contrary, the consid-
eration of a (downward) negative concentration gradient of salinity
(i.e. salinity increases toward the top) in the upper crust of the Earth
can destabilize the thermal convective pore-fluid flow, so that it
makes thermal convective pore-fluid flow more likely to occur in
the upper crust. Based on this conclusion, two cases are investigated
on the effects of variable geothermal and salinity distributions on
convective pore-fluid flow in the Dachang ore district. In the first
case, a (downward) positive concentration gradient of salinity is
applied to the whole domain of the computational model, implying
that the concentration of salinity at the bottom is higher than that at
the top of the computational model. However, in the second case,
a (downward) negative concentration gradient of salinity is applied
to the whole domain of the computational model, implying that the
concentration of salinity at the bottom is lower than that at the top of
the computational model.
Figure 8 Temperature and salinity distributions at two different times: (a) at 10,000 years and (b) at 100,000 years.
4.2.1. Effects of (downward) positive concentration gradient
of salinity
In this case, two situations are considered. In the first situation, the
salinity concentrations of 9 wt.% NaCl and 20 wt.% NaCl are
applied to the top and the bottom of the computational model. It is
assumed that within the computational model, the initial salinity
varies linearly from the top of 9 wt.% NaCl to the bottom of
20 wt.% NaCl. All other parameters remain the same as those used
in sub-section 4.1.2.
Fig. 7 shows the temperature and salinity distributions at two time
instants. As can be seen from this figure, the thermally-induced
buoyancy force overwhelms the saline gravity force and thus
drives the saline water at depth to move upwards along the fault zones
into the shallow aquifer. The solute and heat are brought into the
aquifer and participate in the convective circulation therein. At
100,000 years, the maximum values of salinity and temperature
within the aquifer around the faults are equal to 20 wt.% NaCl and
131  C respectively. Obviously, permeable faults are important
pathways for the transport of the ore-forming pore-fluids. These
results support the case that permeable faults are crucial for ore-
forming processes, because most Sn-polymetallic mineralization
favors high salinity environments (Rankin et al., 1992; Sherman
et al., 2000). However, according to most studies, the temperature
simulated here is lower than the temperature range (200e400  C)
 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
Figure 9 Temperature and salinity distributions at two different times: (a) at 10,000 years and (b) at 100,000 years.
that is favorable for Sn mineralization (Sherman et al., 2000). This
may be because it might take longer time for the heat transfer caused
by convective pore-fluid to raise the temperature to the required value
favorable for the mineralization, or because the bottom temperature
should be higher than we have used because of the existence of
granitic intrusions in the Dachang ore district. Therefore, in future
research further investigations are needed to find geological evidence
to evaluate these two possibilities.
According to the study of paleohydrology in the Dachang
region, there was a high salinity distribution at depth (Fu et al.,
1993), so it is desirable to test how an increase in the salinity at
the bottom of the computational model can affect the required
temperature for Sn mineralization in the ore district. For this
purpose, in the second situation a salinity of 32 wt.% NaCl is
applied at the bottom of the computational model, instead of the
20 wt.% NaCl used previously. It is also assumed that the salinity
at the top is equal to 9 wt.% NaCl and that within the computa-
tional model, the initial salinity varies linearly from the top 9 wt.%
NaCl to the bottom 32 wt.% NaCl.
Fig. 8 illustrates temperature and salinity distributions when the
salinity at the bottom of the computational model is raised to the
value of 32 wt.% NaCl. Although the thermally-induced buoyancy
force overwhelms the saline gravity, the higher salinity at depth
impedes the upwelling hydrothermal pore-fluids. As a result, the
471
maximum value of temperature within the aquifer around the faults is
even lower than that obtained from the previous model where the
salinity of 20 wt.% NaCl is applied at the bottom. Generally, the
modeling results in this situation are very similar to those obtained
from the first situation where the salinity of 20 wt.% NaCl is applied
at the bottom of the computational model. The difference between
these two situations is that in the second situation, the distribution of
high salinity area is enlarged due to the application of a higher salinity
value at the bottom of the computational model. Taking the ore-
forming processes into account, our modeling also indicates that
when the salinity at the bottom of the computational model is raised
to the value of 32 wt.% NaCl, a more favorable high salinity condition
can be provided for Sn mineralization in the Dachang district.
4.2.2. Effects of (downward) negative concentration gradient
of salinity
Modeling results stated above indicate that a high bottom salinity
distribution impedes ascending pore-fluid flow and, therefore,
hydrothermal Sn-polymetalization. Consequently a reflux model
with a high salinity distribution of 32 wt.% NaCl fixed at the upper
boundary is considered in this sub-section. This alternative model
represents an evaporative condition of surface seawater. The initial
salinity throughout the model domain is assumed to have been
12 wt.% NaCl.
472 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
Fig. 9 shows the temperature and salinity distributions at two
different times. It is noted that the cold, saline surface pore-fluids
sink down across the top Permian layer into the Upper-Devonian
aquifer. In contrast, thermally-induced buoyancy forces tend to
drive the hot, deep but less saline pore-fluids upwards along the
faults into the aquifer. As time progresses, the thermally-induced
buoyancy forces overcomes the top high saline gravity and drive
the pore-fluids upwards. Consequently, a series of convection cells
have developed within the aquifer. At 100,000 years, both salinity
and temperature in the aquifer around the faults have been raised.
At the early stage, the sinking pore-fluids brings sufficient
solute into the aquifer but leads to low temperatures, whereas at
the later stage the aquifer around the faults reaches high temper-
atures but has low salinity. Considering the common temperature
and salinity conditions for Sn deposition in most mineral deposits,
the simulation results shown in Fig. 9 indicate that the top part of
this area is not supportive for hydrothermal Sn-polymetallic
mineralization. Even if the salinity is high enough in the top
part of the model, the environment is still not favorable due to its
significantly lower temperature. However, the bottom still can be
a reasonable location for Sn deposition due to its moderate
temperature and salinity. Thus, this reflux model may not be
a good explanation for the genesis of the tin mineral deposits in
the Dachang ore district.
5. Discussions and conclusions
Subsurface geological systems are very complicated as a result of
geological events. In such systems many factors, e.g. distributions
of temperature, pressure, pH, salinity and other chemical
components, can influence the process of hydrothermal Sn-
polymetallic mineralization. This generally requires the consid-
eration of a fully coupled system involving medium deformation,
pore-fluid flow, heat transfer, mass transport and multiple-
components chemical reactions in both time and three-
dimensional space. Due to the complexity of the problem and
the limited modeling capability of FEFLOW, only a simplified
two-dimensional model is used to produce some preliminary
results for investigating the possible convective pore-fluid flow
patterns in the Dachang ore district. In this simplified model, rock
deformation and chemical reactions were not considered. For this
reason, the scenarios described here represent only some possible
paleo-geological settings between tectonic deformation episodes,
which certainly simplify the geological system in reality, but to
some extent may reflect the pore-fluid flow system and diagenetic
environment for Sn mineralization in this ore district.
Numerical simulation results (and geologic studies over the
years) indicate that permeable faults can play a critical role in
controlling hydrothermal pore-fluid flow. Upwelling pore-fluids
along permeable fault zones bring salinity and heat from under-
lying rock units at depth into the shallow aquifer, whereas the
downward flowing pore-fluids within the permeable faults cool the
basement. The width of the permeable fault zones obviously
influences pore-fluid flow patterns, salinity, and temperature
distributions, because the wider fault zones allow more pore-fluids
to be channeled through them.
Our computational simulations demonstrate that the salinity
distribution strongly controls the pore-fluid flow patterns. A higher
salinity distribution (32 wt.% NaCl) at depth impedes the
upwelling hydrothermal pore-fluid flow and, therefore, enhances
the cold seawater to move downwards, resulting in a relatively low
temperature in the Upper-Devonian aquifer. However, low saline
and less dense pore-fluids at depth (12 wt.% NaCl) could be
brought upwards into the Upper-Devonian aquifer by a thermally-
induced buoyancy force, resulting in a relatively high temperature
in the Upper-Devonian aquifer, which would be compatible with
hydrothermal Sn-polymetallic mineralization. Surface evaporative
conditions enable high saline pore-fluids to move downwards into
the aquifer, but prevent hot pore-fluids flowing into it from depth.
According to the prerequisites for hydrothermal Sn-
polymetallic mineralization, both high salinity and high temper-
ature are required. However, our simulation results indicate that
the Upper-Devonian aquifer cannot reach both high salinity and
high temperature simultaneously if only a buoyancy force is
involved as the driving mechanism. Thus, most of the numerical
results presented in this study do not favor the formation of
hydrothermal Sn-polymetallic deposits in the Dachang ore district.
As a result, more realistic models are needed to further investigate
the ore-forming mechanisms in this ore district. Such models
should consider the full coupling between rock deformation, pore-
fluid flow, heat transfer and reactive mass transport in realistic
geochemical environments. For instance, if rock deformation is
incorporated, higher and more localized pore-fluid flux might be
expected during deformation episodes as a result of a significant
change in flow channels due to strain localization within and
around the fault zones in the Dachang ore district.
Acknowledgments
This work is financially supported by the Natural Science Foun-
dation of China (Grant No: 10872219). We would like to thank
Prof. Guoxiang Chi and Dr. Yanhua Zhang for critical and
constructive reviews of the manuscript.
References
Alt-Epping, P., Zhao, C., 2010. Reactive mass transport modeling of
a three-dimensional vertical fault zone with a finger-like convective
flow regime. Journal of Geochemical Exploration 106, 8e23.
Bachu, S., Cao, S., 1992. Present and past geothermal regimes and source
rock maturation, Peace river arch area, Canada. AAPG Bulletin-
American Association of Petroleum Geologists 76, 1533e1549.
Cai, M.H., Liang, T., Wu, D.C., Huang, H.L., 2004. Structural feature and
its control of mineralization of the NandaneHechi metallogenic belt in
Guangxi Province. Geology and Exploration 6, 5e10 (in Chinese with
English abstract).
Cai, M.H., Mao, J.W., Liang, T., Franco, P., Huang, H.L., 2007. The origin
of the TongkengeChangpo tin deposit, Dachang metal district,
Guangxi, China: clues from fluid flow inclusions and He isotope
systematics. Mineralium Deposita 42, 613e626.
Chen, Y., Huang, M., Xu, Y., Ai, Y., Li, X., Tang, S., Meng, L., 1985.
Geological features and metallogenetic series of the dachang cassiterite-
sulfide-polymetallic belt. Acta Geological Sinica 3, 228e240.
Chen, Y.C., Huang, M.Z., Xu, J., 1993. Tin Geology in Dachang. Geological
Publishing House, Beijing (in Chinese with English abstract).
Cooke, D.R., Bull, S.W., Large, R.R., McGoldrick, P.J., 2000. The
importance of oxidized brines for the formation of Australian Prote-
rozoic stratiform sediment-hosted PbeZn (Sedex) deposits. Economic
Geology 95, 1e18.
Davidson, C.F., 1998. Alkali alteration styles and mechanicsms, and their
implications as brine factory sources of base metals in the rift-related
McArthur Group, Australia. Australian Journal of Earth Sciences 45,
33e50.
 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
Diersch, H.J.G., 2002. FEFLOW Reference Manual. Wasy GmbH, Berlin.
Frind, E.O., 1982. Simulation of long-term transient density-dependent
transport in groundwater. Advances in Water Resoures 5, 73e88.
Fu, F.Q., McInnes, B.I.A., Evans, N.J., Davies, P.J., 2010. Numerical
modeling of magmaticehydrothermal systems constrained by U-Th-
Pb-He timeetemperature histories. Journal of Geochemical Explora-
tion 106, 90e109.
Fu, M., Kwak, T.A.P., Mernagh, T.P., 1993. Fluid inclusion studies of
zoning in the Dachang Tin-polymetallic ore field, People’s Republic of
China. Economic Geology 88, 283e300.
Gessner, K., K€ uhn, M., Rath, V., Kosack, C., Blumenthal, M.,
Clauserss, C., 2009. Coupled process models as a tool for analysing
hydrothermal systems. Surveys in Geophysics 30, 133e162.
Gow, P., Upton, P., Zhao, C., Hill, K., 2002. CoppereGold mineralization
in the New Guinea: numerical modeling of collision, fluid flow and
intrusion-related hydrothermal systems. Australian Journal of Earth
Sciences 49, 753e771.
Han, F., Hutchinson, R.W., 1989. Evidence for exhalative origin for rocks
and ores of the Dachang Tin polymetallic field: the ore-bearing
formation and hydrothermal exhalative sedimentary rocks. Mineral
Deposits 89, 25e40 (in Chinese with English abstract).
Han, F., Shen, J., Hutchinson, R.W., 1993. Adularia-an important indicator
mineral of syngenetic origin for stratiform mineralization at the
Dachang tin-polymetallic deposit. Mineral Deposits 12, 330e337 (in
Chinese with English abstract).
Harcou€et-Menou, V., Guillou-Frottier, L., Bonneville, A., Adler, P.M.,
Mourzenko, V., 2009. Hydrothermal convection in and around miner-
alized fault zones: insights from two- and three-dimensional numerical
modeling applied to the Ashanti belt, Ghana. Geofluids 9, 116e137.
Hobbs, B.E., Zhang, Y., Ord, A., Zhao, C., 2000. Application of coupled
deformation, fluid flow, thermal and chemical modelling to predictive
mineral exploration. Journal of Geochemical Exploration 69,
505e509.
Ju, M., Yang, J., 2010. Preliminary numerical simulation of tectonic
deformaiton-driven fluid flow: implications for ore genesis in the
Dachang district, South China. Journal of Geochemical Exploration
106, 133e136.
K€uhn, M., Dobertb, F., Gessner, K., 2006. Numerical investigation of the
effect of heterogeneous permeability distributions on free convection in
the hydrothermal system at Mount Isa, Australia. Earth and Planetary
Science Letters 244, 655e671.
Lattanzi, P., Corazza, M., Corsini, F., Tanelli, G., 1989. Sulfide mineralogy
in the polmetallic cassiterite deposits of Dachang P.R. China. Min-
eralium Deposita 24, 141e147.
Magri, F., Akar, T., Gemici, U., Pekdeger, A., 2010. Deep geothermal
groundwater flow in the SeferihisareBalçova area, Turkey: results
from transient numerical simulations of coupled fluid flow and heat
transport process. Geofluids 10, 388e405.
Molson, J.W., Frind, E.Q., 1993. Heatflow: Density-dependent Flow and
Thermal Energy Transport Model in Three Dimensions, User Guide.
Waterloo Center for Groundwater Research, University of Waterloo,
Canada.
Nield, D.A., Bejan, A., 1992. Convection in Porous Media. Springer-
Verlag, New York.
Ord, A., Hobbs, B.E., Zhang, Y., Broadbent, G.C., Brown, M., Willetts, G.,
Sorjonen-Ward, P., Walshe, J., Zhao, C., 2002. Geodynamic modelling
of the Century deposit, Mt Isa Province, Queensland. Australian
Journal of Earth Sciences 49, 1011e1039.
 ak, K., Fan, D., Zhang, T.,
Pasava, J., Krıbek, B., Dobes, P., Vavrın, I., Z
Boiron, M.C., 2003. Tin-polymetallic sulfide deposits in the eastern
part of the Dachang tin-field (South China) and the role of black shales
in their origin. Mineralium Deposita 38, 39e66.
Person, M., Garven, G., 1994. A sensitivity study of driving forces on fluid
flow during continental-rift basin evolution. Geological Society of
America Bulletin 106, 461e475.
Qin, D., Chen, J., Tian, Y., 1998. Geology and origin of the Changpo tin
ore deposit, Dachang, Guangxi. Geological Exploration for Non-
Ferrous Metals 7, 146e151.
473
Rankin, A.H., Ramsey, M.H., Coles, B., Thomas, C.R., 1992. The
composition of hypersaline, iron-rich granitic fluids based on laser-ICP
and Synchrotron-XRF microprobe analysis of individual fluid inclusion
in topaz, Mole granite, eastern Australia. Geochimica et Cosmochi-
mica Acta 56, 67e79.
Robinson, N.I., Sharp, J.M., Kreisel, I., 1998. Contaminant transport in
sets of parallel finite fractures with fracture skins. Journal of
Contaminant Hydrology 31, 83e109.
Schaubs, P., Zhao, C., 2002. Numerical modeling of gold-deposit forma-
tion in the BendigoeBallarat zone, Victoria. Australian Journal of
Earth Sciences 49, 1077e1096.
Schmidt Mumm, A., Brugger, J., Zhao, C., Schacht, U., 2010. Fluids in
geological processes: the present state and future outlook. Journal of
Geochemical Exploration 106, 1e7.
Sherman, D.M., Ragnarsdottir, K.V., Oelkers, E.H., Collins, C.R., 2000.
Speciatin of tin (Sn2þ and Sn4þ) in aqueous Cl solutions from 25  C
to 350  C: an in situ EXAFS study. Chemical Geology 167,
169e176.
Shikaze, S.G., Sudicky, E.A., Schwartz, F.W., 1998. Density-dependent
solute transport in discretely-fractured geological meida: is
prediction possible? Journal of Contaminant Hydrolology 34,
273e291.
Simmons, C.T., Narayan, K.A., 1997. Mixed convection processes below
a saline disposal basin. Journal of Hydrology 197, 263e285.
Smith, A.J., Turner, J.V., 2001. Density-dependent surface watere
groundwater interaction and nutrient discharge in the Swan-canning
Estuary. Hydrological Processes 15, 2595e2616.
Sorjonen-Ward, P., Zhang, Y., Zhao, C., 2002. Numerical modelling of
orogenic processes and mineralization in the south eastern part of the
Yilgarn Craton, Western Australia. Australian Journal of Earth
Sciences 49, 935e964.
Tanelli, G., Lattanzi, P., 1985. The cassiterite-polymetallic sulfide deposits
of Dachang (Guangxi, People’s Republic of China). Mineralium
Deposita 20, 102e106.
Yang, B., Luo, D., He, H., 1999. Sourrce of ore-forming materials and
origin of the Dachang ore field. Geological Exploration for Non-
Ferrous Metals 8, 466e469.
Yang, J.W., Feng, Z., Luo, X., Chen, Y., 2010. Three-dimensional
numerical modeling of salinity variations in driving basin-scale ore-
forming fluid flow: example from Mount Isa Basin, northern Australia.
Journal of Geochemical Exploration 106, 236e243.
Zeng, Y., Liu, W., Chen, H., Zheng, R., Zhang, J., 1995. Evolution of
sedimentation and tectonics of the Youjiang composite basin, South
China. Acta Geologica Sinica 69, 113e124 (in Chinese with English
abstract).
Zhao, C., Hobbs, B.E., Hornby, P., Ord, A., Peng, S., 2006a. Numerical
modelling of fluids mixing, heat transfer and non-equilibrium redox
chemical reactions in fluid-saturated porous rocks. International Jour-
nal for Numerical Methods in Engineering 66, 1061e1078.
Zhao, C., Hobbs, B.E., M€uhlhaus, H.B., 1998a. Finite element modelling
of temperature gradient driven rock alteration and mineralization in
porous rock masses. Computer Methods in Applied Mechanics and
Engineering 165, 175e187.
Zhao, C., Hobbs, B.E., M€uhlhaus, H.B., 1999a. Theoretical and numerical
analyses of convective instability in porous media with upward
throughflow. International Journal for Numerical and Analytical
Methods in Geomechanics 23, 629e646.
Zhao, C., Hobbs, B.E., M€uhlhaus, H.B., 1999b. Effects of medium thermo-
elasticity on high Rayleigh number steady-state heat transfer and miner-
alization in deformable fluid-saturated porous media heated from below.
Computer Methods in Applied Mechanics and Engineering 173, 41e54.
Zhao, C., Hobbs, B.E., M€uhlhaus, H.B., Ord, A., Lin, G., 2000. Numerical
modelling of double diffusion driven reactive flow transport in
deformable fluid-saturated porous media with particular consideration
of temperature-dependent chemical reaction rates. Engineering
Computations 17, 367e385.
Zhao, C., Hobbs, B.E., Ord, A., 2008a. Convective and Advective Heat
Transfer in Geological Systems. Springer, Berlin.
474 M. Ju et al. / Geoscience Frontiers 2(3) (2011) 463e474
Zhao, C., Hobbs, B.E., Ord, A., 2008b. Investigating dynamic mechanisms
of geological phenomena using methodology of computational geo-
sciences: an example of equal-distant mineralization in a fault. Science
in China Series D: Earth Sciences 51, 947e954.
Zhao, C., Hobbs, B.E., Ord, A., 2009. Fundamentals of Computational
Geoscience: Numerical Methods and Algorithms. Springer, Berlin.
Zhao, C., Hobbs, B.E., Ord, A., 2010. Theoretical and numerical investi-
gation into roles of geofluid flow in ore forming systems: integrated
mass conservation and generic model approach. Journal of Geochem-
ical Exploration 106, 251e260.
Zhao, C., Hobbs, B.E., Ord, A., Hornby, P., Peng, S., Liu, L., 2007. Mineral
precipitation associated with vertical fault zones: the interaction of
solute advection, diffusion and chemical kinetics. Geofluids 7, 3e18.
Zhao, C., Hobbs, B.E., Ord, A., K€uhn, M., M€uhlhaus, H.B., Peng, S.,
2006b. Numerical simulation of double-diffusion driven convective
flow and rock alteration in three-dimensional fluid-saturated geological
fault zones. Computer Methods in Applied Mechanics and Engineering
195, 2816e2840.
Zhao, C., Hobbs, B.E., Ord, A., M€uhlhaus, H.B., Lin, G., 2003. Effect of
material anisotropy on the onset of convective flow in three-dimensional
fluid-saturated faults. Mathematical Geology 35, 141e154.
Zhao, C., Hobbs, B.E., Ord, A., Peng, S., M€uhlhaus, H.B., Liu, L.,
2004. Theoretical investigation of convective instability in inclined
and fluid-saturated three-dimensional fault zones. Tectonophysics
387, 47e64.
Zhao, C., Hobbs, B.E., Ord, A., Peng, S., M€uhlhaus, H.B., Liu, L., 2005.
Double diffusion-driven convective instability of three-dimensional
fluid-saturated geological fault zones heated from below. Mathemat-
ical Geology 37, 373e391.
Zhao, C., Hobbs, B.E., Walshe, J.L., M€uhlhaus, H.B., Ord, A., 2001. Finite
element modelling of fluid-rock interaction problems in pore-fluid
saturated hydrothermal/sedimentary basins. Computer Methods in
Applied Mechanics and Engineering 190, 2277e2293.
Zhao, C., Lin, G., Hobbs, B.E., Wang, Y., M€uhlhaus, H.B., Ord, A., 2002.
Finite element modeling of reactive fluids mixing and mineralization in
pore-fluid saturated hydrothermal/sedimentary basins. Engineering
Computations 19, 364e387.
Zhao, C., M€uhlhaus, H.B., Hobbs, B.E., 1997. Finite element analysis of
steady-state natural convection problems in fluid-saturated porous
media heated from below. International Journal for Numerical and
Analytical Methods in Geomechanics 21, 863e881.
Zhao, C., M€uhlhaus, H.B., Hobbs, B.E., 1998b. Effects of geological
inhomogeneity on high Rayleigh number steady-state heat and mass
transfer in fluid-saturated porous media heated from below. Internal
Journal of Computation and Methodology: Numerical Heat Transfer
33, 415e431.