You are on page 1of 13

Solar Energy 81 (2007) 463475

www.elsevier.com/locate/solener

Simulation and model validation of a horizontal shallow


basin solar concentrator
Dae Hyun Kim *, Bryan M. Jenkins, Thomas R. Rumsey, Matt W. Yore, Nam Jin Kim
Department of Biological and Agricultural Engineering, University of California, 1 Shields Avenue, Davis, CA 95616, United States
Received 13 June 2006; received in revised form 13 June 2006; accepted 15 August 2006
Available online 5 October 2006
Communicated by: Associate Editor Aliakbar Akbarzardeh

Abstract
Salt removal from drainage water is becoming increasingly important for sustainable irrigated arid land agriculture, where inadequate
drainage infrastructure exists. Solar evaporation and concentration systems are currently in development in California for this purpose.
The thermal behavior and evaporation rates of a horizontal shallow basin solar concentrator were modeled for design purposes and
investigated experimentally in order to validate the model. Three dierent evaporation rate models were evaluated and compared. Measured and predicted peak brine temperatures diered by as much as 5 C when using prescribed literature coecients without calibration.
Model prediction was improved by calibration so that peak brine temperature deviated less than 3 C when tested against independent
data sets.
Minimum root mean square error was used to calibrate the mass transfer coecient and absorptance of the collector surface for solar
radiation, which are the main factors aecting the heat transfer associated with the solar concentrator. Calibrated collector surface
absorptance for solar radiation declined while mass transfer coecients were increased from reported literature values. Under calibration, the absorptance of the collector surface was adjusted from 0.8 to 0.61, and mass transfer coecients estimated by Newell et al.
[Newell, T.A., Smith, M.K., Cowie, R.G., Upper, J.M., Cler, C.L., 1994. Characteristics of a solar pond brine reconcentration system.
Journal of Solar Energy Engineering 116 (2), 6973] from 1.36 106(1.9 + 1.065V) to 1.70 106(1.84 + 1.0V) kg m2 s1 mm Hg1,
by Manganaro and Schwartz [Manganaro, J.L., Schwartz, J.C., 1985. Simulation of an evaporative solar salt pond. Industrial & Engineering Chemistry Process Design and Development 24, 12451251] from 0.0208(1 + 0.224V) to 0.0233(1 + 0.214V) kg m2 h1 mm
Hg1, and by Alagao et al. [Alagao, F.B., Akbarzadeh, A., Johnson, P.W., 1994. The design, construction, and initial operation of a
closed-cycle, salt-gradient solar pond. Solar Energy 53 (4), 343351] from 2.8 + 3.0V to 3.0 + 3.33V W m2 C1. The calibrated models
were tested using an independent data set. Maximum deviation between measured and predicted brine temperatures diered by less than
3 C. The measured and predicted peak evaporation rates were between 1.2 and 1.4 kg m2 h1.
The calibrated Newell model was used to predict the monthly productivity and daily maximum evaporation rates at Five Points,
California for the year 2004. The productivity from April to September and from March to October was 80.7% and 94.3% of the total
annual productivity, respectively.
 2006 Elsevier Ltd. All rights reserved.
Keywords: Shallow basin solar evaporator; Evaporation rate; Drainage water; Salinity; Energy balance; Mass balance; Solar concentrator; Evaporator
productivity

1. Introduction

Corresponding author. Fax: +1 530 752 2640.


E-mail address: leekim@ucdavis.edu (D.H. Kim).

0038-092X/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.solener.2006.08.007

Irrigated arid land agriculture historically suers from


buildup of salts in soils eventually leading to reduced
productivity and if uncorrected to land retirement. Integrated farm drainage management (IFDM) systems

464

D.H. Kim et al. / Solar Energy 81 (2007) 463475

Nomenclature
evaporation surface area, m2
collector area, m2
specic heat of brine in the collector, J kg1 K1
specic heat of collector, J kg1 K1
specic heat of makeup brine, J kg1 K1
salt concentration in bulk solution, kg salt
kg solution1
di
dierence between ith estimated and ith measured values
Esal
saline evaporation, m period1
Fev
volumetric ow rate of water evaporated, m3 s1
Fmu
volumetric ow rate of makeup brine into the
concentrator, m3 s1
hb
specic enthalpy of brine, J kg1
hconv, b-a convection coecient between the air and the
brine, W m2 K1
hfg
latent heat of vaporization of water, J kg1
Gb
beam component of radiation on horizontal surface, W m2
Gd
diuse component of radiation on horizontal
surface, W m2
hconv,c-b convection coecient between the collector and
the brine, 170 W m2 K1
hmu
specic enthalpy of makeup brine, J kg1
hvapor specic enthalpy of vapor, J kg1
kG
transport coecient, kg s1 m2 mm Hg1
l
brine depth, cm
mb
mass of brine in the collector, kg
mc
mass of collector, kg
m_ mu
mass ow rate of makeup brine owing into the
collector, kg s1
evaporation rate, kg s1
m_ ev
n
number of data pairs
nair
index of refraction of air, 1.00
nwater index of refraction of water, 1.33
pa
partial pressure of water in air, mm Hg
pb
brine vapor pressure, mm Hg
psat
saturation water vapor pressure, mbar
p0w
saturation vapor pressure of pure water,
mm Hg
qsolar absorb rate of solar energy absorbed by the brine,
W m2
qconv,c-b rate of convection heat transfer from the collector to the brine, W m2
qrad,b-sky net rate of radiation heat transfer with the lower atmosphere, W m2
A
Ac
cpb
cpc
cpmu
c1

employ phytoremediation strategies to reduce salinity


impacts through sequential reuse of water on increasingly
salt tolerant crops (Cervinka et al., 1999) with eventual salt
removal. In current IFDM designs, mixed salts are produced via nal stage physical evaporation of water, usually
in a solar evaporation basin although other concepts are in

qconv,b-a rate of convective heat transfer from the brine


to the air, W m2
re
evaporative mass ux, kg h1 m2
rw
reection loss coecient for water for beam
radiation
rwd
reection loss coecient for water for diuse
radiation
S
salinity weight, %
Ta
ambient air temperature, C
Tb
brine temperature, C
Tc
collector temperature, C
Td
dew point temperature, C
Tmu
temperature of makeup brine owing into the
collector, C
Tsky
eective sky (lower atmosphere) temperature, C
t
time, s
V
wind speed, m s1
xb
mass fraction of salt in the brine, kg salt kg1
brine
xmu
mass fraction of salt in the makeup solution, kg
salt kg1 fresh brine
Xi
ith estimated value
Yi
ith measured value
ab
absorptance of brine for beam radiation
ad
absorptance of brine for diuse radiation
asd
diuse solar absorptance of collector surface
b
activity coecient of water dimensionless
e
emissivity of brine surface
gi
rate energy distribution
href
angle of refraction, degrees
href_di eective angle of refraction for diuse radiation,
degrees
hz
zenith angle of sun, degrees
li
extinction coecient, cm1
qev
density of water evaporated, kg m3
qmu
density of makeup brine, kg m3
r
StefanBoltzmann constant, 5.6697 108,
W m2 K4
s
transmittance
sw
transmittance of water for beam radiation
swd
transmittance of water for diuse radiation
(sa)b
beam transmittanceabsorptance product for
the collectorbrine system
(sa)d
diuse transmittanceabsorptance product for
the collectorbrine system
/
relative humidity decimal

development. Recovery of puried salts might improve the


overall economics of salt management. A critical element in
this last stage is nal concentration of the drainage water.
The behavior of a horizontal shallow basin solar concentrator which can be used for concentrating salts and disposing brine is investigated.

D.H. Kim et al. / Solar Energy 81 (2007) 463475

Physical characteristics of shallow solar ponds have previously been described empirically (Bowen, 1926; Penmam,
1948; Bonython, 1958; Ferguson, 1952). Pancharatnam
(1972) improved Fergusons energy balance equation for
a shallow solar pond by considering an additional term
of heat transfer between the water and the bottom surface.
Model validation by comparing numerical results with
experimental data has also been performed (Manganaro
and Schwartz, 1985; Losordo and Piedrahita, 1991; Alagao
et al., 1994; Newell et al., 1994; Gao and Merrick, 1996;
Jacobs et al., 1998). The study described here was conducted to develop a simulation model for design, and
obtain experimental data for model validation.

up by assuming that the makeup volumetric ow rate


equals the rate that water is evaporated

F ev ; F ev > 0
2
F mu
0;
F ev 6 0
Therefore the total mass balance becomes

F ev qmu  qev ; F ev > 0
dmb

dt
qev F ev ;
F ev 6 0

2. Mathematical models

2.1. Total mass balance


A mass balance on the brine in the concentrator can be
written as
dmb
qmu F mu  qev F ev
dt

Under normal conditions, the feed rate into the concentrator is controlled such that the level of brine is constant. If
water ows into the concentrator from precipitation or
condensation, the makeup ow is reduced or stopped
depending on the magnitude of inux. The solution in
the basin becomes diluted. We model the experimental set

Solar radiation
direct & diffuse
Convection with
ambient air

m_ ev
qev

2.2. Component mass balance on salt


The component mass balance on the salt in the basin can
be written as
dmb xb
qmu F mu xmu
dt

Expanding the left-hand side


mb

dxb
dmb
xb
qmu F mu xmu
dt
dt

Combining Eqs. (3) and (6):


For Fev > 0
mb

dxb
xb F ev qmu  qev F mu qmu xmu
dt

For Fev < 0


mb

dxb
xb qev F ev 0
dt

Simplifying gives the equation for the mass fractions of solids in the brine
( F ev
qmu xmu  xb qmu  qev ; F ev > 0
dxb
m
F evb
9
dt
qev xb ;
F ev 6 0
mb

Long wave radiation


brine/sky

Energy and mass in


with makeup solution

Energy and mass out


with evaporation

Brine solution
Convection
brine to basin

The volumetric evaporation rate is calculated from the


evaporation mass ow rate and density
F ev

To predict evaporation rate, a model of an open type


shallow basin horizontal concentrator was developed.
The general simulation system is schematically shown in
Fig. 1. The model consists of the basin structure along with
a shallow liquid layer maintained at constant depth by
makeup solution.
Two coupled components are considered in the model:
the brine in the basin and the basin structure. The two
components are coupled through the heat exchange processes between the basin and the brine and the absorption
of solar radiation by the brine before reaching the basin
surface.

465

Basin (collector)

Fig. 1. Horizontal concentrator and components considered in mass and energy balances.

466

D.H. Kim et al. / Solar Energy 81 (2007) 463475

2.3. Energy balance on brine

Substituting this into the energy balance on the brine, Eq.


(10):

Energy is received by the concentrator as direct radiation from the sun and indirectly from the atmosphere. Part
of this energy is reected, part absorbed, and the rest is
transmitted to the collector bottom surface of the concentrator. Some of the energy absorbed by the collector is
transferred to the brine via convective heat transfer, some
is lost to the ambient environment by convection and radiation. The brine in the basin also receives energy from the
make-up solution whose temperature is assumed to be the
same as the ambient air temperature although this is not
necessarily the case in practice especially where brine is
stored in-ground such as in a drainage sump. The brine
loses energy by radiation exchange with the environment
(lower atmosphere), by convection with the air, and by
evaporation of water. The energy balance on the brine is
shown schematically in Fig. 2.
The brine is assumed to be well-mixed with respect to
temperature and composition and brine depth is assumed
to remain constant by make-up solution over the time period. The specic heat of the brine is assumed to be independent of brine concentration, surface dust has no eect, and
the heat transfer is one-dimensional. The overall energy
balance on the brine is
dmb hb
qsolar absorb Ac qconv;c-b Ac  qrad;b-sky Ac
dt
 qconv;b-a Ac  m_ ev hvapor m_ mu hmu

10

The energy balance equation is simplied as follows:


dmb hb
dmb
dhb
hb
mb
dt
dt
dt
and
dmb
qmu F mu  qev F ev m_ mu  m_ ev
dt
The specic enthalpies of the brine and makeup solutions are
hb cpb T b and hmu cpmu T mu
dhb dhb dT b
dT b

cpb
dt
dT b dt
dt
dmb hb
dT b
cpb T b m_ mu  m_ ev mb cpb
dt
dt

qsolarabsorb

qrad,b-sky

.
mevhvapor

qconv,b-a

Brine solution

qconv,c-b

Basin (collector)

Fig. 2. Energy balance on brine.

m mu hmakeup

mb cpb

dT b
qsolar absorb Ac qconv;c-b Ac  qrad;b-sky Ac
dt
 qconv;b-a Ac  m_ ev hvapor  cpb T b
m_ mu cpmu T mu  cpb T b

11

Models are presented for each of the energy ux terms in


the energy balance.
2.3.1. Solar radiation absorbed by the brine
Net solar radiation absorbed by the brine per unit area
is written as
qsolar absorb ab Gb ad Gd

12

The absorptance of the brine for beam radiation was


determined using the net radiation method (Tsilingiris,
1998)
ab

1  rw
1  rwd swd 1  asd swd
 1  swd sw 1  asd  asd sw 

ad

1  rwd
1  rwd swd 1  asd swd
 1  swd swd 1  asd  asd swd 

13

14

The reectance at the water surface, rw, is given by


Fresnels equation with the angle of refraction calculated
from Snells law and the angle of incidence in this case is
the zenith angle of the sun
"
#
1 sin2 href  hz tan2 href  hz
rw

;
2 sin2 href hz tan2 href hz


nair
1
sin hz
15
href sin
nwater
The reectance for the diuse radiation is a constant and
is calculated by assuming the incidence angle is 60
(Tsilingiris, 1997, 1998)
"
#
1 sin2 href diff  60 tan2 href diff  60

rwd
;
2 sin2 href diff 60 tan2 href diff 60


nair
1
sin 60
16
href diff sin
nwater
The zenith angle, Gb, and Gd are calculated from Due
and Beckman (1991).
The transmittance of the brine for beam radiation is calculated using the model developed by Tsilingiris (1988,
1998) for pure water

 X
5
l
sw s
gi eli l= cos href
17

cos href
i1
The transmittance for diuse radiation reected from
the bottom absorbing surface is treated as beam radiation
with an equivalent incidence angle of 60 (Tsilingiris, 1998)

D.H. Kim et al. / Solar Energy 81 (2007) 463475


Table 1
Parameter values for wavelength intervals of Eq. (18) (source: Tsilingiris,
1998)
Wavelength
band (i)

Wavelength
(lm)

1
2
3
4
5

0.20.60
0.600.75
0.750.90
0.901.20
1.203.00


swd s

l
cos 60

5
X

li (cm1)

gi
0.237
0.193
0.167
0.179
0.224

0.00032
0.0045
0.03
0.35
18

gi eli l= cos 60

18

i1

Parameter values for Eq. (18) are listed in Table 1. The


experimental data for solar radiation is in the form of total
radiation (beam plus diuse) on a horizontal surface. An
empirical correlation is used to separate the beam and diffuse components (Reindl et al., 1990).
2.3.2. Convection from collector to brine
Convection between the collector surface and the brine
is modeled by
W m2

qconv;c-b hconv;c-b T c  T b

19

The heat transfer coecient is based on turbulent convection arising from wind and from the temperature dierence
between the collector surface and the brine (Pancharatnam,
1972). For turbulent convection, the convection coecient
is 170 (W m2 C1) or higher (McAdams, 1954).
2.3.3. Convection from brine to ambient air
The rate of convective heat transfer from the brine to the
air depends on the temperature dierence between the brine
and the air as well as the wind speed. The eect of wind
speed is modeled as a linear dependence over the range
of wind speeds of interest (Due and Beckman, 1991)
qconv;b-a hconv;b-a T b  T a W m2

20

hconv;b-a 2:8 3:0V W m2  C2

21

2.3.4. Longwave radiation exchange between the brine


surface and the sky
The energy transferred by long-wave radiation exchange
with the lower atmosphere can be expressed as
4

qrad;b-sky erT b 273:15  T sky 273:15  W m2


22
The eective sky temperature, Tsky, is (Sartori, 1996)

1=4
T d 200
T sky T a 273:15
 273:15
23
250
2.3.5. Evaporation
Only a few evaporation models consider a change of
salinity in the pond. Manganaro and Schwartz (1985) for-

467

mulated equations for an evaporative solar salt pond system. In their study, evaporation rate was expressed as
follows:
re 0:0416Cpb  pa 1 0:224V
pb

p0w 1

 0:7c1



1:8T  545:4
0
pw 31:82 exp 17:42
1:8T 0:6

24
25
26

Newell et al. (1994) investigated a brine reconcentration


system by spraying and exposing brine to ambient air.
Evaporation occurred as the brine owed down the water
droplet area. They used Pancharatnams (1972) formulation for the heat lost by evaporation. The overall energy
balance for this system was the same as that used by Pancharatnam (1972) except for the addition of ground conduction. The evaporation rate is expressed as
m_ ev k G Apb  pa
pb
1:0  0:01158S
p0w

27
28

Alagao et al. (1994) modeled a closed-cycle salt-gradient


pond. In their study, evaporation rate was related to the
dissolved salt concentration. Salts dissolved in water
decrease water evaporation by lowering the free energy of
solution. In addition, the saturation vapor pressure at the
brine surface is decreased as salinity increases due to the
interfering presence of salts. The energy lost by evaporation was then formulated as
Esal ff V gfbSpsat T  upsat T a g

29

psat T 10 expf5:8002206  10 T  5:5162560


 4:8640239  102 T 4:1764768  105 T 2
 1:4452093  108 T 3 6:5459673 lnT g
30
bS 1  0:00309S  0:000039S

31

The coecient f(V) is usually empirical and is a linear


function of wind speed. Meyer (1915), Rohwer (1931),
Penmam (1948), Pancharatnam (1972) and Due and
Beckman (1991) proposed similar forms of the mass transport coecient for calculating evaporation rate and for this
study, the Due and Beckman coecients were used as
follows:
f V 2:8 3:0V

32

The curve for activity coecient, (b(S)), is within 0.02%


accuracy by comparing the calculated results with measured activity coecient at dierent salinities (Alagao
et al., 1994).
2.4. Energy balance on collector
The energy balance on the collector shown in Fig. 3 is
made assuming the collector can be modeled using the
lumped capacity method (Cengel, 1998)

468

D.H. Kim et al. / Solar Energy 81 (2007) 463475

()bGb

()dGd

Brine solution
qconv,c-b

Basin (collector)

Fig. 3. Energy balance on collector.

mc cpc

dT c
sab Gb sad Gd  hconv;c-b T c  T b Ac
dt
33

The transmittanceabsorptance product for the beam


radiation was found using the net radiation method (Due
and Beckman, 1991)
sab

Fig. 4. Density versus concentration for various salt solutions at 20 C.

asd sw 1  rw
1  rwd swd 1  asd sw

34

and the product for diuse radiation is


sad

asd swd 1  rwd


1  rwd swd 1  asd swd

35

2.5. Final model equations for horizontal concentrator


The energy equations are solved along with the mass
balances to yield the brine and collector temperatures
and the evaporation rate. The three evaporation methods
described above were incorporated into the energy and
mass balances to compare predictions. For example, in
the case of the Newell model, the nal energy equations are
Brine:
mb cpb

dT b
ab Gb ad Gd Ac hconv;c-b Ac
dt
h

3. Experimental methods and procedure

i
 er T b 273:15  T sky 273:15 Ac
4

 hconv;b-a T b  T a Ac  k G Ac pb  pa
 hvapor  cpb T b m_ mu cpmu T mu  cpb T b
36
Collector:
mc cpc

(Campbell Scientic, Logan, Utah). Horizontal solar radiation was measured with a LiCor LI-200SA pyranometer
(LI-COR, Lincoln, Nebraska). All sensors were located
adjacent to the experimental collector system on the
white-surfaced roof of a three-story oce building in
Davis, California. Data were collected and recorded on
an electronic datalogger (Model CR21X, Campbell Scientic, Logan, Utah). The experiment was performed from
06/22 to 6/28/2003.
Density data for brine as a function of both temperature
and salt concentration were obtained from the literature.
Several data sets are compared in Fig. 4. For the results
reported here, density data (Table 2) for sodium sulfate
solutions were used (Sohnel and Novotny, 1985).

dT c
sab Gb sad Gd  hconv;c-b T c  T b Ac
dt
37

2.6. Solution procedure and model assumptions


The model equations were coded into Matlab v.6.5.0.
(Mathworks Inc., 2002). Ambient meteorological conditions were measured at one minute intervals and averaged
and recorded over 30 minutes. Wind speed was measured
by a MetOne 014A three cup anemometer (MetOne,
Grants Pass, Oregon). Air temperature and relative humidity were measured by a Campbell Scientic CS500 sensor

A small-scale prototype horizontal concentrator was


built and used to test the predictions of the model. Brine
was maintained in the basin as a quiescent liquid. A constant head brine supply system maintained a constant
depth of brine in the basin.
The concentrator included the basin, water supply system, and instrumentation (Fig. 5), all located on the open
roof of a large three-story building. The basin was constructed of plastic-coated 19 mm thick plywood lined on
the inside with black polyethylene plastic lm. The basin
was 1.5 m2 with a maximum allowable depth of 35 mm.
The basin was supported level at a height of 0.9 m above
the roof surface. The brine supply system was composed
of an inverted brine tank suspended from a load cell to
measure the weight and situated so as to maintain a constant depth of brine.
Salt concentration of agricultural drainage water from a
model farm in the Central Valley, California is approximately 1%. Batches of brine with the same concentration
as the drainage water were prepared by mixing pre-weighed

D.H. Kim et al. / Solar Energy 81 (2007) 463475

469

Table 2
Density of sodium sulfate solutions (Sohnel and Novotny, 1985)

100 C
95 C
90 C
85 C
80 C
75 C
70 C
65 C
60 C
55 C
50 C
45 C
40 C
35 C
30 C
25 C
20 C
15 C
10 C
5 C
0 C

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

22%

24%

26%

28%

958
962
965
969
972
975
978
981
983
986
988
990
992
994
996
997
998
999
1000
1000
1000

976
979
983
986
989
992
995
998
1001
1003
1006
1008
1010
1012
1014
1015
1017
1018
1019
1019
1019

994
997
1000
1003
1007
1010
1013
1015
1018
1021
1023
1025
1028
1030
1032
1033
1035
1036
1037
1038
1038

1011
1015
1018
1021
1024
1027
1030
1033
1036
1038
1041
1043
1046
1048
1050
1052
1054
1055
1056
1057

1030
1033
1036
1039
1042
1045
1048
1051
1054
1056
1059
1062
1064
1066
1068
1071
1072
1074
1075

1048
1051
1054
1057
1060
1063
1066
1069
1072
1075
1078
1080
1083
1085
1087
1089
1091
1093

1067
1070
1073
1076
1079
1082
1085
1088
1091
1094
1096
1099
1102
1104
1107
1109
1111

1086
1089
1092
1095
1098
1101
1104
1107
1110
1113
1116
1118
1121
1124
1126
1128
1131

1105
1108
1111
1114
1117
1120
1123
1126
1129
1132
1135
1138
1141
1143
1146
1148
1151

1125
1128
1131
1134
1137
1140
1143
1146
1149
1152
1155
1158
1161
1164
1166
1169

1145
1148
1151
1154
1157
1160
1163
1166
1170
1173
1176
1179
1181
1184
1187
1190

1165
1168
1171
1175
1178
1181
1184
1187
1190
1193
1196
1199
1202
1205
1208

1186
1189
1192
1195
1199
1202
1205
1208
1211
1215
1218
1221
1224
1227
1229

1207
1210
1213
1217
1220
1223
1227
1230
1233
1236
1239
1242
1246
1249
1251

1228
1232
1235
1238
1242
1245
1249
1252
1255
1258
1262
1265
1268
1271
1274

30%

32%

1261
1264
1268
1271
1274
1278
1281
1284
1287
1290
1294

1310
1314
1317

Brine tank
LC

Concentrated
brine

Makeup brine

Air vent when new


makeup brine added

New makeup brine inlet

Horizontal
concentrator

Data logger

LC: Load cell


V: Valve

Fig. 5. Schematic of horizontal concentrator.

amounts of salt obtained from a solar pond at Rainbow


Ranch (now Andrews Agriculture), California, with deionized and distilled water.
Makeup brine was added daily into the reservoir. Tank
weight and hence evaporation rate was measured using a
calibrated load cell and recorded electronically. In addition
to the mass balance, salt concentration was also monitored
by electrical conductivity (Model 30, YSI, Inc, Yellow
Springs, OH) twice daily, at the beginning and end of the
peak evaporation period (i.e., 9 a.m. and 6 p.m.). Brine
temperature in the basin and the collector surface (bottom
surface of the basin) temperature were measured by type T
thermocouples shielded from direct solar radiation.
4. Results and discussion
Experimental results were compared with predicted values under the same meteorological conditions and basin
characteristics. For calculating evaporation rates, the models of Manganaro and Schwartz (1985), Newell et al.
(1994), and Alagao et al. (1994) were used.

Fig. 6. Comparison of measured and predicted surface temperatures from


the uncalibrated models.

Predicted temperatures and evaporation rates from the


uncalibrated model utilizing the original published coecients generally follow the experimental results but deviate
by more than 5 C and 0.4 kg m2 h1 during the peak
evaporation periods of the day (Figs. 68). Although the
coecients from Newell et al. (1994) consistently predicted
higher collector and brine temperatures than measured, the
predicted evaporation rates were lower than measured during the peak because the Newell model gives the lowest sensitivity of the transport coecients on wind speed among
three models so that heat transfer from the brine body to
the ambient air by evaporation is relatively low, therefore,
higher brine temperatures were predicted.

470

D.H. Kim et al. / Solar Energy 81 (2007) 463475

Fig. 7. Comparison of measured and predicted brine temperatures from


the uncalibrated models.

Fig. 8. Comparison of measured and predicted evaporation rates from the


uncalibrated models.

4.1. Uncertainty of measurements and predictions


The uncertainty associated with the measurements was
evaluated according to Coleman and Glenn Steele (1989).
The uncertainty is given by
"
#1=2
j
X
2
wr
hi wxi
38
i1

where the wxi are the uncertainties in the variables xi and


or
hi ox
where the result, r = r(x1, x2, . . . , xj) is the experii
mental result dependent on the xi.
The uncertainty in the measurement of the evaporation
rate depends on the individual uncertainties associated with
the calibration of the load cell and the time at which
makeup brine is discharged from the reservoir into the
basin. The makeup discharge is not continuous, but instead
occurs intermittently as the liquid level in the basin declines
suciently for the head in the reservoir to overcome the
atmospheric pressure force. At such time, an unmetered
amount of water is discharged from the reservoir. In some
instances this discharge occurs in a data sampling interval
following the primary period of evaporation; hence the
recorded evaporation rate in an interval may either be
under- or over-estimated. The relative error associated with
this intermittency decreases with the frequency of discharge
and so is lower during the peak evaporation period. This
uncertainty in the evaporation rate measurement is greater
than that associated with the calibration of the load cell
used to record reservoir weight.
The experimental uncertainty in the predicted values is
associated with the measurements of the meteorological
and concentration parameters used in the model. The
parameters needed in Eq. (38) for both the experimental
and predicted evaporation rates, m_ ev , are listed in Table 3.
The uncertainties in the predicted rates range from 0.12
to 0.22 kg m2 h1 with a mean of 0.16 kg m2 h1 for the
sampling period (Table 3). By comparison, the uncertainties in the experimental evaporation rates are
0.27 kg m2 h1 over the same period. The absolute uncertainties in the measurement are 0.27 kg m2 h1, mainly by
the frequency of discharge from the reservoir involved.
Relative uncertainties are high (>50%) when experimental
evaporation rate is near zero (i.e., at night) but decrease
to as low as 14% during periods of peak evaporation. Without calibration, predicted peak evaporation ranged from
1.1 to 1.3 kg m2 h1 compared with measurements yielding 1.3 to 1.6 kg m2 h1 (Table 4). The model was calibrated against the experimental data in order to reduce
the deviation between predicted and measured evaporation
rate.
4.2. Calibration of model
A number of statistical indicators have been proposed
and used to evaluate models. Stone (1993) discussed several
indicators including the root mean square error

Table 3
Uncertainty (kg m2 h1) in maximum daily evaporation rate for 7 days
Day
a

Meteorological parameters (Ta, Tb, RH , wind speed)


Load cell
Discharge
a

RH relative humidity.

0.18
5 104
0.27

0.14
5 104
0.27

0.12
5 104
0.27

0.14
5 104
0.27

0.18
6 104
0.27

0.22
5 104
0.27

0.14
7 104
0.27

D.H. Kim et al. / Solar Energy 81 (2007) 463475

471

Table 4
Mean and range of measured and predicted peak evaporation rate over a 7 day sampling period
Day
2

Maximum evaporation (kg m

1

h )

Predicted

Experiment

RMSE1

n
1X
d2
n i1 i

Mean
Min.
Max.
Mean
Min.
Max.

!1=2
39

The smaller the RMSE1, the better the model t. A drawback of this statistic is that a few large errors in the sum can
produce a signicant increase in RMSE1. According to
Stone (1993), another measure using the relative root mean
square normalized deviation can be applied
!1=2
n  2
1X
di
40
RMSE2
n i1 Y i

1.11
0.93
1.29
1.31
1.04
1.58

1.17
1.03
1.31
1.45
1.18
1.72

1.17
1.05
1.29
1.38
1.11
1.64

1.19
1.05
1.33
1.41
1.14
1.68

1.27
1.09
1.45
1.52
1.25
1.79

1.27
1.05
1.49
1.40
1.13
1.65

1.31
1.17
1.45
1.64
1.37
1.90

the temperature of the makeup brine (Table 5). Daily maximum heat convection between the air and the brine varies
between 80 and 120 W m2 or about 15% of total heat loss.
Maximum long-wave radiation is between 120 and
180 W m2 and accounts for about 20% of total heat transferred from the solar concentrator which is approximately
1000 W m2 at maximum evaporation (Fig. 9). Evaporation accounts for more than 60% of the total heat loss during this time. The absorptance of the collector is the most
important factor for this analysis relative to the energy
input to the system by solar radiation. The evaporative mass transfer coecient and the absorptance of the

RMSE2 for m_ ev is calculated only for times during which


the evaporation rate is non-zero to avoid normalization
errors.
There is another parameter to evaluate models:
R-squared (r2)
2
32
 Pn
 Pn
  Pn

6
7
n
i1 X i Y i 
i1 X i
i1 Y i
7
r2 6
i
h
i
4rh
2 Pn 2 Pn
2 5
Pn 2 Pn
n i1 X i 
n i1 Y i 
i1 X i
i1 Y i
41
r2 becomes 1 in the absence of error and zero if there is no
relationship between measured and estimated values.
As noted earlier, uncalibrated model predictions underestimate actual evaporation rates during periods of high
evaporation (Fig. 8). Model evaporation is dependent on
assumed values for the convective heat transfer coecients
between ambient air and brine and the brine and collector,
the emissivity for long-wave radiation, the evaporative
mass transfer coecient, the absorptance of the collector
surface for solar radiation, specic heat of the brine, and

Fig. 9. Heat transfer rates by evaporation, long-wave radiation, and


convection between brine and air as estimated from the uncalibrated
model.

Table 5
Initial model assumptions
asd
k
e
hconv,b-a
hconv,c-b
cpb
Tmu
a
b
c
d

Diuse solar absorptance of collector surface


Mass transfer coecient (kg m2 s1 mm Hg1)
Emissivity of brine surface
Convection coecient between the brine and the air (W m2 C1)
Convection coecient between the collector and the brine (W m2 C1)
Specic heat of brine in the collector (J kg1 C1)
Temperature of makeup brine (C)

Due and Beckman (1991).


Newell et al. (1994).
McAdams (1954).
Bromley et al. (1967).

0.8a
1.35623 106 (1.9 + 1.065V)b
0.95a
2.8 + 3.0Va
170c
Same as sea waterd
Same as air temperature

472

D.H. Kim et al. / Solar Energy 81 (2007) 463475

The numbers in parentheses in Table 7 are the data


points used for calculation. The Newell model is the most
improved through calibration, showing the greatest reduction in RMSE values, from 2.53 to 1.43 in RMSE1 for the
brine temperature (Table 7). However, RMSE1 and
RMSE2values for the brine temperature from the Alagao
model have increased from 2.16 to 2.56, and 0.08 to 0.09
respectively after calibration. Because of the uncertainty
in terms of frequency of discharge, r2 for m_ ev was calculated
using the daytime data which had relatively lower uncertainty. Although the calibrated mass transport coecient
for the Alagao model increased RMSE values and
decreased r2 for brine temperature, the calibrated model
decreased RMSE values and increased r2 for the evaporation rate. RMSE values for the evaporation rate from all
three models were reduced.

4
collector temperature
brine temperature

RMSE 01

0
0

0.2

0.4

0.6

0.8

Absorptance

Fig. 10. Sensitivity of RMSE1 to collector surface absorptance.

4.3. Validation of calibrated model

collector are the most inuential factors in determining


evaporation rate.
Model calibration was performed by minimizing RMSE
(Fig. 10). For both absorptance and the evaporative mass
transfer coecient, initial calibration values were derived
as the average of the RMSE1 and RMSE2 values individually minimized. Final calibration values were then obtained
by minimizing RMSE1 for simultaneous changes in absorptance and mass transfer coecient (Table 6).

The calibrated model was tested using an independent


experimental data set collected from 07/06/03 to 07/11/03
but using the same horizontal shallow basin solar concentrator and sensors.
Most RMSE values for brine temperatures and evaporation rate found using the original literature models were
higher than those using the calibrated models, and the values for the validation data set were generally lower than the
nal values after calibration for the calibration data set

Table 6
Calibrated model coecients
Original value

Calibrated value

Diuse solar absorptance of collector surface

0.8

0.61
6

1.36 10

Mass transfer coecient

2 1

(1.9 + 1.065V) (kg m

1 b

mm Hg )

0.0208(1 + 0.224V) (kg m2 h1 mm Hg1)c


2.8 + 3.0V (W m2 C1)a
a
b
c

1.70 106
(1.84 + 1.0V)
0.0233(1 + 0.214V)
3.0 + 3.33V

Due and Beckman (1991).


Pancharatnam (1972).
Manganaro and Schwartz (1985).

Table 7
Relative errors for Tb and m_ ev before and after model calibration (number of data points shown in parentheses)
Tb

m_ ev

Original

Calibrated

Original

Calibrated

Newell

RMSE1
RMSE2
R-square

2.53 (336)
0.08 (336)
0.9834 (336)

1.43 (336)
0.06 (336)
0.9837 (336)

0.20 (336)
0.67 (148)
0.7598 (148)

0.19 (168)
0.66 (73)
0.7735 (148)

Manganaro

RMSE1
RMSE2
R-square

2.29 (336)
0.10 (336)
0.9844 (336)

2.18 (336)
0.08 (336)
0.9847 (336)

0.19 (336)
0.73 (148)
0.7804 (148)

0.18 (168)
0.69 (73)
0.7883 (148)

Alagao

RMSE1
RMSE2
R-square

2.16 (336)
0.08 (336)
0.9709 (336)

2.56 (336)
0.09 (336)
0.9688 (336)

0.21 (336)
0.68 (148)
0.7395 (148)

0.19 (168)
0.66 (73)
0.7572 (148)

D.H. Kim et al. / Solar Energy 81 (2007) 463475

473

Table 8
Relative errors for Tb and m_ ev with the calibrated model run on the
validation data (number of data points shown in parentheses)
Tb

m_ ev

Newell
RMSE1
RMSE2
R-square

1.48 (288)
0.04 (288)
0.9846 (288)

0.17 (288)
0.89 (129)
0.7690 (129)

Manganaro
RMSE1
RMSE2
R-square

1.34 (288)
0.04 (288)
0.9904 (288)

0.16 (288)
0.85 (129)
0.8049 (129)

Alagao
RMSE1
RMSE2
R-square

2.25 (288)
0.06 (288)
0.9722 (288)

0.17 (288)
0.82 (129)
0.7444 (129)

(compare Tables 7 and 8). Maximum deviation between


measured and predicted brine temperatures using the calibrated models dier by no more than 3 C (Fig. 12), and
there is no apparent bias between the predicted and measured brine temperature for the uncalibrated models (compare Figs. 11 and 13). The measured and predicted peak
evaporation rate is between 1.2 and 1.4 kg m2 h1
(Fig. 14). The larger deviation in evaporation rate in
Fig. 15 is due to the rapid transients in the morning and
late afternoons indicating the eect of the frequency of discharge on the experimental evaporation rates. The makeup
discharge from the brine tank occurs intermittently by
accumulating water evaporated in the basin. At such time,
no experimental evaporation is recorded especially from
evening to next day morning although predictions of evaporation are continuous. The large deviation in Fig. 15,
hence, is a result of the limited resolution of the experimental detection of evaporation.

Fig. 12. Comparison of measured and predicted brine temperature from


the calibrated models.

Fig. 13. Measured and predicted brine temperature from the calibrated
models. Lines are 1:1.

5. Model application

Fig. 11. Measured and predicted brine temperature from the uncalibrated
models. Lines are 1:1.

The calibrated Newell model was applied to predict


evaporation rates and accumulated evaporation rates (productivity) at Five Points, California using hourly CIMIS
data from January through December 2004 (California
Irrigation Management Information System, Department
of Water Resources, http://wwwcimis.water.ca.gov/cimis)
assuming 1% salt concentration of the agricultural drainage water. Table 9 gives the hourly productivity expressed
as mass of water evaporated per unit area and per month
predicted along with the predicted average daily maximum
evaporation rate for each month and compares it with a
standard pan measuring evapotranspiration (ETo) which
is the sum of evaporation and plant transpiration obtained
from CIMIS data.

Parameter calibration of the shallow basin brine concentration model improved the predictions of peak temperatures and evaporation rates by as much as 10%

21
40
93
164
195
214
213
219
179
52
22

122

12
0.12
33
0.26
110
0.54
220
0.87
250
0.93
280
0.99
282
1.01
285
1.02
227
0.91
150
0.67
48
0.34
16
0.16

May
April
March
February
January

1

Productivity (kg m mo )
Daily peak evaporation rate
(kg m2 h1)
ETo (mm mo1)

December
November
October

6. Conclusions

2

Seasonal productivity is an important consideration


for concentrator operation. As Table 9 shows, there is little evaporation at this site during the winter months of
November to February. About 81% of the total evaporation occurs between April and September, with 94% of
evaporation occurring between March and October
(Table 10). Comparing the total productivity with the
annual ETo in Table 9, the horizontal concentrator has
a productivity about 20% higher than ETo due to the
solar energy gain by the collector.

Month

Fig. 15. Measured and predicted evaporation rate from the calibrated
models. Lines are 1:1.

Table 9
Predicted monthly productivity and daily maximum evaporation rate at Five Points, CA

June

July

August

September

Fig. 14. Comparison of measured and predicted evaporation rate from


the calibrated models.

1913
0.66 (annual
avg)
1534

D.H. Kim et al. / Solar Energy 81 (2007) 463475

Total

474

D.H. Kim et al. / Solar Energy 81 (2007) 463475


Table 10
Seasonal productivity
2

Productivity (kg m )
Percent (%)

Total

April to September

March to October

1913
100

1544
80.7

1804
94.3

compared to original published values of heat and mass


transfer coecients. Deviations between predicted and
measured peak brine temperatures and evaporation rates
were reduced from 5 C and 0.4 kg m2 h1 to less than
3 C and 0.2 kg m2 h1, respectively.
Calibration was accomplished by minimizing root mean
square error (RMSE) between predicted and measured
temperatures and evaporation rate. RMSE on collector
surface absorptance was reduced by roughly 0.2 and mass
transfer coecients were increased by minimizing RMSE1
through simultaneous calibration of concentrator surface
absorptance and heat and mass transfer coecients. Validation of the calibrated models was performed using an
independent experimental data set.
Predicted seasonal productivity (Table 10) suggests that
concentrator designs oversized by only 20% compared with
full year operation would be sucient to handle summertime irrigated agricultural drainage ows within the primary six month growing season. Although horizontal
basins of the type modeled here may not be the preferred
design choice due to environmental considerations, these
conclusions are likely to extend to other solar concentrator
designs handling agricultural drainage brines that are the
subject of related papers.
References
Alagao, F.B., Akbarzadeh, A., Johnson, P.W., 1994. The design,
construction, and initial operation of a closed-cycle, salt-gradient
solar pond. Solar Energy 53 (4), 343351.
Bonython, C.W., 1958. The inuence of salinity upon the rate of natural
evaporation. Climatology and microclimatology. In: Proceedings of
Canberra Symposium, pp. 6570.
Bowen, I.S., 1926. The ratio of heat losses by conduction and by
evaporation from any water surface. Physical Review 27, 779787.
Bromley, L.A., Desaussure, V.A., Clipp, J.C., Wright, J.S., 1967. Heat
capacities of sea water solutions at salinities of 1 to 12% and
temperatures of 2 to 80 C. Journal of Chemical and Engineering Data
12 (2), 202206.
Cengel, Y.A., 1998. Heat Transfer, Practical Approach. WCB McGrawHill, Boston.

475

Cervinka, V., Diener, J., Erickson, J., Finch, C., Martin, M., Menezes, F.,
Peters, D., Shelton, J., 1999. Integrated system for agricultural
drainage management on irrigated farmland. Bureau of Reclamation,
US Department of the Intrior, Final Research Report, 4-FG-20-11920.
Coleman, H.W., Glenn Steele Jr., W., 1989. Experimentation and
Uncertainty Analysis for Engineers. John Wiley & Sons, Inc.
Due, J.A., Beckman, W.A., 1991. Solar Engineering of Thermal
Processes, second ed. John Wiley & Sons, Inc., New York.
Ferguson, J., 1952. The rate of natural evaporation from shallow ponds.
Australian Journal of Scientic Research 5, 315330.
Gao, J., Merrick, N.P., 1996. Simulation of temperature and salinity in a
fully mixed pond. Environmental Software 11 (13), 173178.
Jacobs, A.F.G., Heusinkveld, B.G., Lucassen, D.C., 1998. Temperature
variation in a class A evaporation pan. Journal of Hydrology 206, 75
83.
Losordo, T.M., Piedrahita, R.H., 1991. Modeling temperature variation
and thermal stratication in shallow aquaculture ponds. Ecological
Modeling 54, 189226.
Manganaro, J.L., Schwartz, J.C., 1985. Simulation of an evaporative solar
salt pond. Industrial & Engineering Chemistry Process Design and
Development 24, 12451251.
McAdams, W.H., 1954. Heat Transmission, third ed. McGraw-Hill, New
York, NY.
Meyer, A.F., 1915. Computing run-o from rainfall and other physical
data. Transactions of ASCE 79, 10561224.
Newell, T.A., Smith, M.K., Cowie, R.G., Upper, J.M., Cler, C.L., 1994.
Characteristics of a solar pond brine reconcentration system. Journal
of Solar Energy Engineering 116 (2), 6973.
Pancharatnam, S., 1972. Transient behavior of a solar pond and
prediction of evaporation rates. Industrial & Engineering Chemistry
Process Design and Development 11 (2), 287292.
Penmam, H.L., 1948. Natural evaporation from open water, bare soil and
grass. Proceedings of the Royal Society of London. Series A,
Mathematical and Physical Sciences 193 (1032), 120145.
Reindl, D.T., Beckman, W.A., Due, J.A., 1990. Diuse fraction
correlations. Solar Energy 45 (1), 17.
Rohwer, C., 1931. Evaporation from free water surfaces. US Department
Agricultural Technical Bulletin 271, 96.
Sartori, E., 1996. Solar still versus solar evaporator: a comparative study
between their thermal behaviors. Solar Energy 56 (2), 199206.
Sohnel, O., Novotny, P., 1985. Densities of aqueous solutions of inorganic
substancesPhysical Sciences Data, vol. 22. Elsevier, Amsterdam.
Stone, R.J., 1993. Improved statistical procedure for the evaluation of
solar radiation estimation models. Solar Energy 51 (4), 289291.
Tsilingiris, P.T., 1988. An accurate upper estimate for the transmission of
solar radiation in salt gradient ponds. Solar Energy 40 (1), 4148.
Tsilingiris, P.T., 1997. Design, analysis and performance of low-cost
plastic lm large solar water heating systems. Solar Energy 60 (5), 245
256.
Tsilingiris, P.T., 1998. On optical performance and directional characteristics of plastic lm liquid layer solar water heaters. Solar energy 63 (5),
293302.

You might also like