See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223142518

A computational study of salt diffusion and

heat extraction in solar pond plants

Article in Solar Energy · November 2006

DOI: 10.1016/j.solener.2005.10.015

CITATIONS READS

18 121

3 authors:

Celestino Angeli Erminia Leonardi

University of Ferrara CRS4 Centro di Ricerca, Sviluppo e Studi Sup…
121 PUBLICATIONS 2,695 CITATIONS 17 PUBLICATIONS 263 CITATIONS

SEE PROFILE SEE PROFILE

Luca Maciocco
Advanced Accelerator Applications SA
22 PUBLICATIONS 133 CITATIONS

SEE PROFILE

All content following this page was uploaded by Celestino Angeli on 06 April 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 Solar Energy 80 (2006) 1498–1508
www.elsevier.com/locate/solener

A computational study of salt diffusion and heat extraction

in solar pond plants
Celestino Angeli a, Erminia Leonardi b,*
, Luca Maciocco c

a
Dipartimento di Chimica, Università di Ferrara, Via Borsari 46, I-44100 Ferrara, Italy
b
CRS4, Center for Advanced Studies, Research and Development in Sardinia, Parco Scientifico e Tecnologico, POLARIS,
Edificio 1, 09010 Pula, CA, Italy
c
ADACAP, Advanced Acceleration Applications, Technoparc-rue Diesel 20, Fr-D1630 St. Genis Pouilly, France

Received 15 March 2005; received in revised form 26 August 2005; accepted 20 October 2005
Available online 2 February 2006

Communicated by: Associate Editor Aliakbar Akbarzardeh

Abstract

The problem of the development of salt concentration profiles in a solar pond is investigated, the thermodiffusion
contribution is also taken into account, using a one-dimensional mathematical model and a finite-difference approach.
A novel scheme of heat extraction from the solar pond is presented, along with preliminary two-dimensional
computational fluid dynamics (CFD) simulations.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Solar pond; Salt diffusion; Finite difference 1D numerical methods; Thermodiffusion

1. Introduction zone, UCZ) and the bottom (lower convective

A solar pond is an artificially constructed water
pond several meters deep, filled with stratified brine
varying in concentration from near saturation at the
bottom to approximately sea water concentration at
the surface. About 25% of the solar radiation inci-
dent at the surface penetrates to the bottom of the
pond and is absorbed there, causing the adjacent
brine to heat up. It consists of three distinct regions:
two convective regions at the top (upper convective
*
Corresponding author. Tel.: +39 070 925 0328; fax: +39 070
925 0216.
E-mail address: ermy@crs4.it (E. Leonardi).
zone, LCZ), separated by a quiescent region (non-
convective zone, NCZ) characterized by strong
temperature and salinity gradients, which prevent
the occurrence of convection currents.
The salt diffusion caused by both molecular diffu-
sion and thermodiffusion (also known as Soret
effect, Soret, 1879) tends to homogenize the system
and therefore must be carefully put under control.
In previous papers (Angeli and Leonardi, 2004,
2005), a one-dimensional model based on a finite
difference scheme was used to describe the salt diffu-
sion within a solar pond. In the present work, the
analysis of the behaviour of the solar pond has been
extended by considering other effects besides the salt

0038-092X/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.solener.2005.10.015
 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508 1499

Nomenclature

A cross-sectional area of the lower convec- UCZ upper convective zone

tive zone, m2 ~
v velocity of the injected brine, m s1
c salt concentration, kg m3 W solar power density, W m2
Cp thermal capacity, J kg1 K1 W0 solar power density at the water surface,
D salt diffusivity, m2 s1 W m2
~
J total mass flux, kg m2 s1 z solar pond height, measured from the
J~D diffusive mass flux, kg m2 s1 bottom (z = H  Z), m
J~S thermodiffusive mass flux, kg m2 s1 Z solar pond depth, measured from the
k thermal conductivity, W m1 C1 surface, m
j thermal diffusivity, m2 s1 ZL solar pond depth measured at the LCZ–
L solar pond length, m NCZ boundary, m
LCZ lower convective zone ZU solar pond depth measured at the UCZ–
w1 salt weight fraction, dimensionless NCZ boundary, m
NCZ non-convective zone
P solar power, W Greek symbol
q_ heat generated per unit time and volume, q density, kg m3
W m3
S salinity, % Subscripts
sT Soret coefficient, C1 xD diffusive x
T temperature, C xinj injected x
TL temperature in the LCZ, C xS thermodiffusive (Soret) x
TU temperature in the UCZ, C xtr traslational x

diffusion. The extraction of the heat from the system ori the direction of the thermodiffusion flux. In fact
could also cause the damage of the NCZ, and must it appears to be sensitive to details of the molecular
be performed with caution. In this paper a novel interaction potentials and cannot easily be mea-
heat extraction scheme is also presented. This sured experimentally and predicted theoretically
method allows solar energy to be used efficiently (Lin et al., 1991; Longree et al., 1980).
without interfering with the system and has been When diffusion and thermodiffusion are consid-
patented (Cabibbo et al., 2003). ered, the principle of mass transfer for the solute
The rest of this paper is organized as follows: in results in an equation of the form:
Section 2 we briefly present the mathematical model oc

used to analyze the diffusion problem; in Section 3 ¼ r ~ ~
J ¼ r ~  J~D þ J~S
ot
some preliminary simulations are presented to verify    
~ ~ c ~
the validity of a new heat extraction scheme. ¼ r  Drc  DsT c 1  rT ; ð1Þ
q
2. Salt diffusion where ~ J is the total mass flux, J~D is the diffusive
mass flux, J~S is the mass flux contribution due to
Salt diffusion within a solar pond is made up of the thermodiffusion, and q is the density. For a
two factors: molecular diffusion and thermodiffu- NaCl aqueous solution, q is given as (McCutcheon
sion, which is the separation of the components of et al., 1993):
a liquid mixture induced by temperature gradients.  
T þa 2
The thermodiffusion coefficient varies signifi- qðz; tÞ ¼ 1000  1   ðT  d Þ
cantly according to the nature of the components b  ðT þ c Þ
3
of the mixture and their concentration. Until now þ eðT Þ  S þ f ðT Þ  S 2 þ g  S 2 ; ð2Þ
there has not been a hydrodynamical explanation
and/or a microscopical model of the phenomenon, where a = 288.9414, b = 508929.2, c = 68.12963,
and therefore there is not a way to determine a pri- d = 3.9863, g = 4.8314 · 104 and
1500 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508

eðT Þ ¼ 8:24493  101  4:0899  103  T
5 2 7
þ 7:6438  10  T  8:2467  10
9 4
þ 5:3675  10 T ;
f ðT Þ ¼ 5:724  10 3
þ 1:0227  10 4
T
6
 1:6546  10 T ; 2
D is the diffusivity, obtained by a least-square fit to
the International Critical Tables for a temperature
range between 5 and 100 C (Giestas et al., 1997):
D ¼ 8:16 þ 0:255  T þ 0:00254  T 2  0:28  S

oW
þ 0:0147  S 2  1010
where S is the salinity in weight percent (S = c/
q · 100), D is in m2 s1, and sT is the Soret coeffi-
cient, which has been chosen according to the value
measured by Caldwell (Caldwell, 1973), for a 0.5
normal NaCl solution in the range of temperature
between 0 and 50 C. The curve given by Caldwell
has been linearly extrapolated T > 50 C. The over-
all curve in the range between 0 and 100 C has been
interpolated with the cubic polynomial function:
sT ¼ 0:0012597 þ 0:0001221  T  1:3495
 106  T 2 þ 6:0831  109  T 3 .
Eq. (1) can be solved numerically, using a finite
difference scheme and a second order approximate
in time Crank–Nicholson method (Numerical
Recipes, 1992), where the concentration c(z, t) is
computed on a regular grid of points in z
(fz1 ; z2 ; . . . ; znz g, zi+1  zi = Dz) and t (ft1 ; t2 ; . . . ;
3 tnt g, ti+1  ti = Dt) (Angeli and Leonardi, 2004,
T
2005). The temperature profile within the salinity-
ð3Þ gradient layer is computed by using the steady state
conduction heat equation:
 
ð4Þ o oT
q_ þ k ¼ 0; ð7Þ
oZ oZ
where k is the thermal conductivity, q_ is the solar
energy absorbed from the water per unit time and
per unit volume:
ð5Þ q_ ¼  ; ð8Þ
oZ
W(Z) is the solar power density at the depth Z
measured from the water surface.
The diffusion process within the salinity-gradient
layer has been simulated taking into account the
thermodiffusion effect as well. The solar pond has
been dimensioned with the following thicknesses:
1 m for LCZ, 2 m for NCZ and 0.5 m for UCZ.
Impermeable walls have been chosen as boundary
conditions, and therefore the salt mass must be con-
served in time within the solar pond. Moreover, at
each time step of the simulation a homogenization
ð6Þ
of the UCZ and LCZ has been carried out to take
into account the effect of the convective motion on
the salt distribution. An initial salinity profile vary-
ing linearly from 200 g kg1 at the bottom to
35 g kg1 at the surface has been chosen, while the
temperature at the bottom and at the surface have

200
Initial salinity
180
Diffusion
160

140
Salinity (g kg-1)

Thermodiffusion
120 +
Diffusion
100

80

60

40

0 0.5 1 1.5 2 2.5 3 3.5

Solar pond height (m)
Fig. 1. Salinity profile after 2 years, with and without thermodiffusion.
 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508 1501

been fixed at 85 and 25 C, respectively. An average second method, but which avoids temperature
solar power density of 300 W m2 complete the set discontinuities.
of data for the simulation. The temperatures and As shown in Fig. 2, the solar pond is U shaped,
solar power density are assumed constant in time. with a heat exchanger located between the inlet
As shown in Fig. 1, where the development of the and outlet sides of the solar pond, thus reducing
salinity profile after 2 years is given, the diffusion the length of pipelines and the energy cost incurred
process tends to destroy the salinity-gradient, and by the need to pump the brine along the close cir-
the presence of the thermodiffusion makes the situ- cuit. In order to apply this scheme, the translational
ation even worse. Therefore, in order to ensure the motion of the brine in the LCZ must not affect the
effectiveness of a solar pond for long periods of stability of the salinity-gradient layer; therefore the
time, it is necessary to inject concentrated brine translational velocity of the brine must be lower
form the bottom and flush the surface with fresh than the brine velocity due to natural convection.
or low salinity water.
3.1. Computational approach
3. Heat extraction
The new extraction scheme has been studied by
A crucial aspect of solar pond technology is the means of computational fluid dynamics (CFD) sim-
heat extraction from the LCZ (Sherman, 1989; Lu ulations, using the commercial code STAR-CD
et al., 2001; Kumar and Kishore, 1999). Presently (Star-CD, 2001).
it is carried out in one of the following two ways: Table 1 resumes the characteristic parameters
by using a submerged heat exchanger located on used for the simulations of the solar pond, while
the bottom of the pond, or by pumping brine from
the lower convective zone to an external heat
exchanger and then returning it to the pond. The Table 1
Simulation parameters
first method implies high costs for the maintenance
of the heat exchanger, as it is fully submerged in the UCZ thickness 0.5 m
NCZ thickness 2.0 m
hot brine, which is a highly corrosive medium. The
LCZ thickness 1m
second method generates local temperature differ- Solar pond length 1000 m
ences, whose final effect is to destabilize the salin- Solar pond width 25 m
ity-gradient layer, because this brine extraction/ Inlet velocity (mean storage-layer 1.5 · 102 m/s
injection process is performed at a specific point of translation velocity) (1.5 · 103 m/s)
Inlet temperature 85 C
the solar pond, generally close to the centre of a
Ambient temperature 25 C
circular solar pond (this geometry maximizes the Solar radiation 300 W/m2
volume to perimeter ratio, and then minimizes the Ground temperature, thermal resistance 10 C, 1 m2/W
heat losses). Surface velocity 5 · 102 m/s
Below we discuss a heat extraction scheme, Initial salinity (storage, gradient 200, linear, 35 g/kg
and surface layers)
which can be considered a modification of the

Fig. 2. Heat extraction scheme (arbitrary scale).

1502 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508
Fig. 3. Diagram of a three-layer solar pond.
Fig. 3 shows the characteristic variables involved in
the definition of the time scale of the problem.
A time step of the order of 1 s is adequate to
appropriately describe natural convection. The
characteristic time of the translational velocity is
about 106 s (12 days), which comes from the follow-
ing equation for the translational velocity, Vtr:
P
V tr ¼
AqC p DT
50; 000 ½m2   300 ½W=m2   0:25
¼
25 ½m2   1116 ½kg=m3   3515 ½J=kg K  20 ½K
 0:002 ½m=s; ð9Þ
where P is the total solar power reaching the surface
of the SP, A is the cross-sectional LCZ area, q is the
average density, Cp is the average thermal capacity,
and DT is the inlet–outlet temperature difference.
Therefore, indicating with Lt the total length of
the pond, ttr = Lt/Vtr  106 s (12 days). The
thermal inertia characteristic time (tth) of the three
layers can be estimated as tth = h2/j, h being the
height of the layer and j = k/qCp the thermal diffu-
sivity. For the gradient layer, k  0.6 W/m/K and
qCp  4 · 106, yielding tth,NCZ  107 s.
Finally, a reasonable characteristic time of the
variation of environmental conditions can be con-
sidered of the order of 103 s.
From these estimations it results that different
approaches must be used depending on the phenom-
enon to be studied. In particular, a detailed CFD
model is used to simulate the convection mecha-
nism, its effect on the heat transfer, the salinity gra-
dient erosion and the stability of the system, but
limited to a small stretch of length and to specified
overall thermal conditions. A simplified CFD
model, where the convection mechanism is modelled
on the basis of the results from the detailed model,
thus eliminating the shorter time scales, is used for
the simulation of the thermal dynamics of the solar
pond. These two approaches can be used in parallel
for the analysis and the optimization of the system.
3.2. Detailed 2D model
For a detailed 2D simulation, the solar pond is
modelled as an area 20 m long and 3.5 m deep, with
about 25,000 computational cells. The computa-
tional grid also includes an inlet divergent duct: in
this way the inlet conditions are not influenced by
the convection vortices developed in the LCZ. The
bottom of the solar pond is simulated as a layer of
solid material, where the temperature and flow fields
are calculated considering coupled equations. A dia-
gram of the computational grid is shown in Fig. 4.
All fluid properties (density, viscosity, thermal
conductivity, specific heat and salt diffusivity) are
calculated as local functions of temperature and salt
concentration. The boundary conditions of the
problem are: assigned brine velocity and tempera-
ture at the inlet and fixed mass flow rate at the out-
let. Moreover, the free surface is considered as a
wall with a locally assigned heat flux: the heat
removed by convection of the ambient air and by
evaporation is considered a function of the surface
temperature (calculated) and of the assigned ambi-
ent conditions (temperature and humidity), through
empirical formulas (Gill, 1982). The drag effect of
the wind on the surface is taken into account by
assigning a sliding velocity to the wall, together with
a noslip condition for the velocity (if this effect is not
considered, a slip wall is used). The bottom wall is
the interface between the fluid and the solid bottom,
and it is considered as a conducting wall. The
 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508
Fig. 4. Computational model for the detailed 2D simulation.
ground is considered as a wall with a fixed temper-
ature (representing the temperature of the ground at
a given depth) and with a thermal resistance calcu-
lated on the basis of the ground thermal conductiv-
ity. The absorption of the solar radiation within the
brine is simulated as a source term, whose distribu-
tion is given by
W ¼ W 0 ½a  b lnðxÞ; ð10Þ
where x is the distance crossed by the sun rays in the
water, a = 0.442, b = 0.0929, and W0 is the power
density at x = 0. It is assumed that the surviving radi-
ation passing through the brine is completely ab-
sorbed by the bottom surface, and it is assigned as
a source term to the first layer of cells of the solid bot-
tom. The turbulence in the two convective layers is
simulated with a two-equation Chen k– model.
The solar pond is always non-stationary, even in
the case of stationary boundary conditions. In fact,
even if stationary convection is possible in the case
of a layer bounded by two walls at different temper-
atures (Rayleigh–Bernard flow), the presence of the
source term and of the overall translation flow do
not allow the formation of stationary vortices.
Therefore, a transient simulation is always necessary.
The time step of 1 s has been used for this simulation.
After a simulation time of 6 h the system reaches a
pseudo steady state condition, with natural convec-
tion confined to the UCZ and LCZ, with a maximum
value in the LCZ of the order of 100 mm/s, that is,
one order of magnitude larger than the mean transla-
tional velocity of the brine. Therefore, the proposed
heat extraction method does not affect the stability of
the NCZ. The salinity field, expressed in salt g/solu-
tions kg, is shown in Fig. 5.
1503
The temperature field is shown in Figs. 6 and 7.
In particular, as shown in Fig. 7, the temperature
variations in the LCZ are less than 1 C.
Fig. 8 shows a comparison between the initial salt
concentration profile and the profile after the simu-
lation of 6 and 13 h of solar pond operation: the
erosion of the NCZ due to the natural convection
in the LCZ is reproduced by the calculation. This
phenomenon, although very slow, makes the salin-
ity-gradient smoother with time.
In general, in order to overcome this problem
and to make the solar pond operative for long peri-
ods of time (years), it is necessary to inject concen-
trated brine at the boundary between the LCZ and
the NCZ, and to flush the UCZ with fresh (or low
salinity) water. Such an effect has been evaluated
by solving the modified version of Eq. (1), where
new flux term, ~ vc, has been added to the right side:
   
oc ~ ~ c ~
¼ r  Drc  DsT c 1  rT þ~ vc ;
ot q
ð11Þ
where ~ v is the rising velocity of the NCZ, positive
upward and constant with respect to time and space.
In particular, at the boundary between the LCZ and
the NCZ, we have considered the injection of
concentrated brine (cinj = 300 kg m3) with v =
2.5 · 104 m s1. Thermodiffusion, as already
shown, is a minor effect and, therefore, it has been
neglected. A comparison of the results between the
case with and without concentrated brine injection
is shown in Fig. 9, where salinity is plotted at
t = 0 s, and after 4 years.
Although important to maintain the salinity
gradient, the term ~ vc has not been introduced in this
1504 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508

Fig. 5. Distribution of salt concentration.

Fig. 6. Temperature field in the pond.

 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508 1505

Fig. 7. Temperature field in the storage layer.

Fig. 8. Erosion of the salinity gradient comparison between the concentration profiles at t = 0 s, t = 6 h, and t = 13 h.
1506 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508

preliminary CFD simulation, it will be the argument
of future study.
3.3. Simplified 2D model
The thermal evolution of the solar pond needs a
time scale much larger than that used to describe the
natural convection, and, therefore, this last phe-
nomenon cannot be described at the same time.
Therefore, a simplified model is used where the stor-
age layer is simulated as a 1D flow, moving with the
layer translational velocity, Vtr. This approximation
is justified by the small temperature difference in the
LCZ with respect to the temperature variation in the
NCZ, as confirmed in the detailed 2D simulation.
The computational grid is shown in Fig. 10.
The natural convection heat exchange is mod-
elled by assigning to the fluid a thermal conduction
coefficient equivalent to the overall heat exchange
coefficient, estimated on the basis of the results of
the detailed simulation (it could also be considered
a function of the pond length or of the local thermal
conditions). In the hypothesis of a stable salinity
gradient (also confirmed by the detailed simulation),
the gradient layer can be reasonably simulated as a
conducting solid, with a thermal conduction

0.2
With injection of
concentrated brine

0.15
Salinity (g kg-1)

Without injection of
concentrated brine

0.1

Initial condition
0.05

0 0.5 1 1.5 2 2.5 3 3.5

Solar pond height (m)
Fig. 9. Initial salinity and salinity profiles after 4 years, with and without injection of concentrated brine (cinj = 300 kg m3,
v = 2.5 · 104 m s1).

Fig. 10. Simplified solar pond model.

 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508
coefficient expressed as a function of the local
temperature and concentration. The surface layer
is modelled in the same way as the storage layer
(it could even be considered a solid, the mean trans-
lation velocity being null).
The free-surface and bottom wall boundary con-
ditions of the simplified model are the same as in the
detailed model. Two more conductive walls separate
the storage and surface layers from the gradient
layer. Fixed pressure and temperature are imposed
at the inlet, while only pressure is imposed at the
outlet. The inlet–outlet pressure difference is set in
order to have a brine translational velocity of about
1.5 · 103 m s1. For the bottom wall a thermal
conductivity of 0.5 W m1 K1 and a thickness of
0.1 m have been assumed, while the ground has a
thermal resistivity of 2 m2 K W1, and a tempera-
ture of 10 C. Time-varying boundary conditions
can be applied to simulate the variation of ambient
and operational conditions. A test case of a 2 km
long pond has been simulated, in the same condi-
tions as those reported in Table 1, with an inlet tem-
perature of 70 C. Two-meter-long cells were used
in the horizontal direction, allowing a time step of
1000 s. Figs. 11 and 12 show the heat balances along
the pond’s length and the temperature distribution
in the storage and surface layers, respectively after
5000 time steps (2 months). The outlet tempera-
ture is about 93 C, corresponding to a thermal effi-
ciency of the system of about 23%.
It should be noted that these results must be con-
sidered preliminary, with more accurate calculations
Fig. 11. Heat fluxes along the solar pond length after 2 months.
1507
Fig. 12. Temperature distribution along the solar pond length
after 2 months.
with more realistic boundary conditions still to be
performed.
4. Conclusion
The salt diffusion within the solar pond has been
accurately studied, evaluating at the same time the
magnitude of thermodiffusion. Results obtained by
solving the diffusion equation, with the finite differ-
ence method, show that thermodiffusion, although
acting, as molecular diffusion, in the direction of
destabilizing the salinity-gradient, is a relatively
small contribution, but it cannot be neglected. The
stabilizing effect of the salinity-gradient by injection
of concentrated brine at the boundary between the
LCZ and NCZ has been evaluated. A preliminary
CFD simulation of a simplified system has been also
performed, in order to verify the possibility of heat
extraction by means of a slow movement of the
LCZ brine, through a close circuit, with a heat
exchanger located between the inlet and outlet of
the solar pond.
It was shown that both the short (natural convec-
tion) and long (thermal dynamics) timescale phe-
nomena of the system can be approached using
suitable CFD models. Preliminary results are prom-
ising, showing that it is possible to apply the pro-
posed heat extraction scheme. However, more
reliable quantitative results have still to be obtained,
using more realistic boundary conditions (variable
atmospheric conditions, ground characteristics, bot-
tom insulation, etc.).
Future work will be addressed to the model-
ling of the solar pond’s response to changes in
1508 C. Angeli et al. / Solar Energy 80 (2006) 1498–1508
meteorological and operating conditions as well as
the stability of the salinity-gradient, the free surface
wind effects, and the optimal geometry of the solar
pond.
Acknowledgments
The authors thank Bruno D’Aguanno for useful
discussions and suggestions and Fabio Bettio for the
help in the realization of the figures. This work has
been carried out with the financial support of the
‘‘Regione Autonoma della Sardegna’’ and of the
University of Ferrara.
References
Angeli, C., Leonardi, E., 2004. A one-dimensional numerical
study of the salt diffusion in a salinity gradient solar pond.
Int. J. Heat Mass Transfer 47, 1–10.
Angeli, C., Leonardi, E., 2005. The effect of thermodiffusion on
the stability of a salinity gradient solar pond. Int. J. Heat
Mass Transfer 48, 4633–4639.
Cabibbo, N., Leonardi, E., Rosa-Clot, M., Maciocco, L., CRS4
patent 2003. No. FI2003A000008: Apparato e metodo per la
produzione di calore mediante stagno solare.
Caldwell, D., 1973. Deep Sea Research 20.
View publication stats
Giestas, M., Joyce, A., Pina, H., 1997. The influence of non-
constant diffusivities on solar ponds stability. Int. J. Heat
Mass Transfer 18, 4379–4391.
Gill, A.E., 1982. Atmosphere–Ocean Dynamics. Int. Geographics
Series 30.
Kumar, A., Kishore, V.V.N., 1999. Construction and operational
experience of a 6000 m2 solar pond at Kutch, India. Solar
Energy 65, 237–249.
Lin, J.L., Taylor, W.L., Rutherford, W.M., Millat, J., 1991. In:
Wakeman, W.A., Nagashima, A., Sengers, J.V. (Eds.),
Measurement of the Transport Properties of Fluids. Blackwell
Scientific, Oxford, p. 321.
Longree, D., Legros, J.C., Thomaes, G., 1980. J. Phys. Chem. 84,
3480–3483.
Lu, H., Walton, J.C., Swift, A.H.P., 2001. Desalination coupled
with salinity-gradient solar ponds. Desalination 136, 13–
23.
McCutcheon, S.C., Martin, J.L., Barnwell Jr., T.O., 1993. Water
Quality in Maidment in Handbook of Hydrology. McGraw-
Hill, New York, NY, p. 11.3.
Numerical Recipes in FORTRAN 77: the art of scientific
computing (ISBN 0-521-43064-X), Cambridge University
Press, 1992, pp. 838–842.
Sherman, B.S., 1989. Modelling and control of a solar pond. PhD
thesis, The University of Western Australia.
Soret, C., 1879. Arch. Geneve 3, 48.
Star-CD, 2001. Version 3.15 manual, Computational Dynamics,
London.