Building and Environment 46 (2011) 187e195

Contents lists available at ScienceDirect

Building and Environment

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

Swimming pools as heat sinks for air conditioners: Model design

and experimental validation for natural thermal behavior of the pool
Jonathan Woolley 1, Curtis Harrington*, Mark Modera 2
University of California Davis, Western Cooling Efficiency Center, 1450 Drew Avenue, Suite 100, Davis, CA 95618, United States

a r t i c l e i n f o a b s t r a c t

Article history: Swimming pools as thermal sinks for air conditioners could save approximately 40% on peak cooling
Received 5 April 2010 power and 30% of overall cooling energy, compared to standard residential air conditioning. Heat
Received in revised form dissipation from pools in semi-arid climates with large diurnal temperature shifts is such that pool
28 June 2010
heating and space cooling may occur concurrently; in which case heat rejected from cooling equipment
Accepted 15 July 2010
could directly displace pool heating energy, while also improving space cooling efficiency. The perfor-
mance of such a system relies on the natural temperature regulation of swimming pools governed by
Keywords:
evaporative and convective heat exchange with the air, radiative heat exchange with the sky, and
Swimming pool
Heat sink
conductive heat exchange with the ground. This paper describes and validates a model that uses
Energy efficient air conditioning meteorological data to accurately predict the hourly temperature of a swimming pool to within 1.1
C
Heat transfer model maximum error over the period of observation. A thorough review of literature guided our choice of the
Experimental validation most appropriate set of equations to describe the natural mass and energy exchange between a swim-
Free surface evaporation ming pool and the environment. Monitoring of a pool in Davis, CA, was used to confirm the resulting
simulations. Comparison of predicted and observed pool temperature for all hours over a 56 day
experimental period shows an R-squared relatedness of 0.967.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction reduce heat sink temperatures, and reduce the required temperature
differences between the refrigerant and the source and sink. For
In California, where all the large electric utilities experience their example, rejecting condenser heat to water instead of air reduces the
peak power demand in the summer, space cooling accounts for 29% of temperature difference that is needed for adequate heat transfer; air-
the total peak power demand and approximately 40% of the residential cooled condensers typically require a refrigerant temperature that is
peak demand [1]. This occurs in part because the COP for traditional air- 20
C higher than condenser inlet air, while exchange with water only
cooled vapor-compression cooling equipment diminishes significantly needs a 10
C temperature difference.
at high outdoor temperatures, such that equipment efficiency can be at The research presented in this paper provides a foundation for
its worst when cooling demand is greatest. Thermodynamics for heat the design of cooling systems that reject condenser heat to swim-
pumps dictates that the work required to transfer heat from a cooler ming pools, a strategy that has been deployed successfully in many
source to a warmer sink increases with the temperature difference installations [2,3], but that has not been widely adopted. One
between the two. In practice, for a vapor-compression system, since reason for the lack of application is the lack of research, docu-
heat exchange with the refrigerant at the condenser and evaporator is mentation and standardization. Our thesis is that a better under-
driven by the temperature differences between the refrigerant and the standing of the mechanisms that drive performance and savings
sink and source respectively, the overall temperature difference could inform the development of guidelines for appropriate design
experienced by the refrigerant is significantly larger than the temper- of these systems, and could lead to more prevalent adoption,
ature difference between the sink and source. For this reason, a large resulting in cost-effective energy and peak demand savings. The
fraction of cooling efficiency research has focused on techniques to savings should come from three mechanisms:

1. Lower sink temperature since pool water is cooler than outdoor

* Corresponding author. Tel.: þ1 530 754 7670; fax: þ1 530 754 7672. air during most cooling periods.
E-mail addresses: [email protected] (J. Woolley), csharrington@ucdavis.
edu (C. Harrington), [email protected] (M. Modera).
2. Improved heat transfer at the condenser since exchange with
1
Tel.: þ1 530 752 1101; fax: þ1 530 754 7672. water is more effective than exchange with air.
2
Tel.: þ1 530 754 7671; fax: þ1 530 754 7672. 3. Reduction of energy consumption for pool heating.

0360-1323/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.buildenv.2010.07.014
188 J. Woolley et al. / Building and Environment 46 (2011) 187e195

Nomenclature pa ambient pressure (Pa)

po reference pressure (Pa)
Acond surface area of conduction to ground (m2) Pr Prandtl number (e)
As surface area used for shape factor calculations (m2) q_ ss dimensionless conduction heat rate (e)
A surface area of pool at airewater interface (m2) qcond conduction heat flux (W/m2)
CBowen Bowen coefficient3 (61.3 Pa/
C) qconv convection heat flux (W/m2)
dpoolavg average pool depth (m) qevap evaporation heat flux (W/m2)
ea vapor pressure in ambient air (Pa) qrad radiation heat flux (W/m2)
es saturation vapor pressure of air at the pool Qsolar solar heat gain (W)
temperature (Pa) RBowen Bowen ratio (e)
Esky emissivity of sky (e) RaL Rayleigh number (e)
Ew emissivity of water (e) S solar input (W/m2)
g acceleration of gravity (m/s2) Ta ambient air temperature (
C)
GrL Grashof number (e) Tdew dew point temperature (
C)
h average convection coefficient (W/m2
C) Tw swimming pool temperature (
C)
hevap wind speed function for evaporation (W/m2 Pa) Tsky effective sky temperature (
C)
HR humidity ratio (kg/kg) Tsoil soil temperature (
C)
kair thermal conductivity of air (W/m
C) V wind speed (m/s)
ksoil thermal conductivity of soil (W/m
C) a absorptivity of water (e)
L average length of pool (m) ba thermal expansion coefficient of air (1/
C)
Lc characteristic length of pool used for shape factor bw thermal expansion coefficient of water (1/
C)
calculations (m) r density of water (kg/m3)
NuL average Nusselt number (e) s Stefan-Boltzmann constant (5.67 E8 W/m2 K4)
Osky opaque sky cover (tenths) n kinematic viscosity of air (m2/s)
P perimeter of pool (m) w average width of pool (m)

The practical use of condenser heat rejection to swimming pools 2. Methodology and results
relies critically on the natural temperature regulation of pools by
conductive heat exchange with the ground, convective and evap- 2.1. Model development
orative heat exchange with the air, and radiative heat exchange
with the sky. The key is to balance heat rejection from the space An analytical model to determine the heat and mass transfer for
cooling system with heating demand for the pool, such that pool a swimming pool was developed to calculate the transient thermal
temperature is maintained in a desirable range. We expect that this behavior of a pool based on hourly weather data. The model relies on
balance will be easiest to maintain in climate regions of the western detailed information about the site and the operating characteristics
United States, or other semi-arid regions with low ambient of the pool. Based on meteorological inputs and system conditions,
humidity and relatively low nighttime temperatures. In these at each hourly time step (t), the calculations draw on empirical and
regions, heat dissipation from swimming pools is increased by high theoretical heat transfer correlations to estimate the steady state
evaporation rates in low humidity environments, and by longwave heat transfer rates for conductive, convective, radiative, and evap-
radiative cooling which increases with low ambient humidity and orative heat exchange mechanisms. Rates are integrated across the
clear skies. Anecdotal evidence suggests that heat dissipation from hour, and energy and mass storage terms are calculated to determine
pools in these climates is such that pool heating is often required to the average pool temperature at the beginning of the next hour
maintain desired water temperature, even when space cooling is (t þ 1). Meteorological inputs and system conditions at (t þ 1) are
required to maintain desired indoor temperature. In this case, heat then used to solve for system conditions at the following hour (t þ 2).
rejected from cooling equipment could directly displace energy The following sections describe the basis for calculating heat
consumed to heat a pool, while concurrently improving the COP of transfer rates for each mechanism considered in the model.
the cooling system.
The objective of this paper is to document and discuss the 2.1.1. Insolation
development of a model to simulate the energy and mass balance The heat gain (W) due to solar radiation is found by multiplying
of a swimming pool in natural interaction with its local environ- the solar insolation at the pool surface by the absorptivity and area
ment; subsequent research will validate the model for simulation of the pool.
of a swimming pool used as a heat sink for vapor-compression
air conditioning. Since there is no standardized approach to Qsolar ¼ S$a$A (1)
modeling the thermal behavior of swimming pools, this research
The concept is simple, but determining the solar insolation and
draws from the conclusions of many authors to develop a clear
absorptivity are challenging prospects. Insolation at the pool surface
and generalized method, and validates model predictions
is comprised of both direct and diffuse radiation, so when a pool is
against long-term experimental measurements from a pool in
partly shaded by nearby objects, raw meteorological data for the
Davis, CA.
global horizontal insolation is not representative of actual condi-
tions. To compensate, shading of the pool must be described for each
hour by inspection of the site and analysis of solar pathways for the
3
The Bowen Coefficient is 6.13 Pa/
C for the case when evaporation from a water latitude, season, and time period of the simulation. Diffuse insola-
surface does not significantly impact absolute humidity of the air. tion is used as the solar input for shaded periods and global
 J. Woolley et al. / Building and Environment 46 (2011) 187e195 189

insolation is used for un-shaded periods; periods of fractional 1 As


shading receive a corresponding fraction of global and diffuse qcond ¼ q_ ss ksoil Tsoil  Tpool (2)
2Lc Acond
insolation.
Absorptivity is even more complicated due to the phenomenon where
of multiplicative reflection and absorption in transparent materials.
As ¼ 2D2 þ 4Dd (3)
Only a portion of the solar radiation available at the pool surface is
accumulated as heat in the water volume; a fraction is reflected by d ¼ 2dpoolavg (4)
the water surface, and of the portion that passes unabsorbed
through to the pool, a fraction is absorbed by the pool bottom and  0:5
As
a fraction is reflected back into the pool. Water clarity affects the Lc ¼ (5)
4p
absorption rate per unit depth, and the pool construction, espe-
0:5
cially color, affects absorption at the pool bottom. Moreover, the D ¼ Acond þ d2 d (6)
fraction of radiation reflected from the water surface changes with
solar incident angle, so the net absorptivity of the pool varies by
time and season. However, for the model presented here, an annual 2.1.3. Radiation
average absorptivity was calculated using a method presented by Exchange of longwave radiation with the sky is one of the most
Wu [4], which uses latitude, longitude, refractive index of water and significant pool cooling effects; it occurs continuously, separate
air, pool bottom absorptivity, and average depth. The approach from solar radiation. The magnitude of this heat flux is calculated
divides solar insolation into separate spectral bands to account for using equation (7), the standard radiative heat transfer equation.
the fact that energy content and extinction rates vary as a function The approach relies on the effective sky temperature e a value that
of wavelength; it considers multiplicative reflection and absorp- reduces the complex phenomenon of radiant exchange between
tion, and the impact of incident angle. According to validation by the pool, the semi-opaque atmosphere, and space beyond, to
Wu, the method predicts absorptivity to within 3.67% of experi- a simple radiative exchange between the pool and a much larger
mental observations over the course of a day. surface of representative temperature. Walton [10] developed two
methods to determine the effective sky temperature; one is
2.1.2. Conduction a function of infrared radiation from the sky, the other relies on the
Conduction between the swimming pool and ground is simpler to dew point temperature, sky emissivity and opaque sky cover. The
model, and in most circumstances accounts for less than 1% of the total latter approach was used here, as the infrared sky radiation is not
energy loss from the pool [5e7]. Since the time constant of thermal generally measured by standard meteorological stations.
response for the earth is very large, pool temperature only affects h
4 i
ground temperature very near the pool walls, and transient pool qrad ¼ sEw Tsky þ 273 ðTw þ 273Þ4 (7)
temperature has very little effect on the daily temperature distribution
in the ground. Therefore, most authors assume that soil temperature

0:25
remains constant, and since the temperature difference is small and Tsky ¼ ðTa þ 273Þ$ Esky  273 (8)
conduction is minimal compared to other heat transfer mechanisms,
many authors ignore this heat transfer component altogether. A review   
T þ 273
of the literature identified several different models to estimate Esky ¼ 0:787 þ 0:764$log dew
273
conduction effects. All authors rely on a standard one-dimensional h i
conduction equation; some use the thermal conductivity of the wall 1 þ 0:224$Osky  0:0035$Osky þ 0:00028$O2sky
2
ð9Þ
and a constant ground temperature, while others approximate a total
thermal resistance for the pool wall plus soil in a temperature transi- This radiative heat transfer decreases with increases in dew point
tion zone. Govaer [6] accounts for seasonal changes in ground temperature, opaque sky cover, and ambient air temperature. It can
temperature, but few account for the effect of a vertical temperature account for up to 60% of the total thermal losses at night in arid
profile in the ground. Hull developed a semi-empirical method which climates with low humidity, minimal cloud cover, and low night-
uses the distance from the bottom of the pool to a constant tempera- time temperatures. Compared with the other two parameters, dew
ture sink, the ground conductivity and pool dimensions [8]. The model point temperature, an indicator of ambient humidity, has a rela-
presented here includes an analysis of conduction, but assumes tively small impact on the overall longwave radiative heat transfer.
a constant ground temperature, even across seasons, and a soil However, there is an obvious correlation between low ambient
conductivity of 0.52 W/m-K [9]. humidity and a low degree of cloud cover. Fig. 1 and Fig. 2 illustrate
While a three dimensional conduction model could be used, the magnitude of longwave radiation loss (W/m2) as a function of
since the ground temperature is assumed to remain uniform and opaque sky cover and ambient temperature respectively. Note that
constant, this model simplifies the conduction problem to a one- heat flux into the pool is the positive convention, so negative
dimensional function by using a shape factor to account for exchange of longwave radiation represents cooling of the pool. It is
geometry effects. Shape factors have been developed analytically critical for the model’s accuracy to obtain cloud cover data for each
for many different geometric cases; for this model the swimming hour of the day because, as Fig. 1 shows, the radiative heat transfer
pool is approximated as a cuboid embedded in an infinite medium, can differ significantly between clear sky and cloudy conditions.
and a shape factor given by Incropera [9] is adapted to account for
heat transfer through five faces of the cuboid with ground interface 2.1.4. Evaporation
area similar to that of the pool. Evaporation is an especially complicated phenomenon to model
for swimming pools since:

d/D q_ ss 1. There is no commonly accepted theoretical approach for esti-

0.1 0.943 mating evaporation rates from free water surfaces [11]
1.0 0.956 2. Empirical equations may only be appropriate under the
2.0 0.961
conditions for which they were developed
190 J. Woolley et al. / Building and Environment 46 (2011) 187e195

0 These equations were developed experimentally by correlating

water temperature in several lakes to meteorological measure-
-10 ments of air temperature, relative humidity, and wind speed. A
negative value for qevap indicates that water is evaporated, sensible
Longwave Radiation (W/m2)

-20 heat is lost from the water mass, and latent heat is gained in the air
mass. The empirical wind speed function for evaporation, hevap,
accounts for the latent energy content of water vapor and the rate
-30
at which water vapor diffusion occurs under different wind
conditions, while the equation for qevap accounts for the evapora-
-40
tive potential due to the difference between vapor pressure in the
ambient air and saturation conditions at the water surface
-50 temperature. It’s worth noting that the evaporation rate is driven by
the relationship between pool temperature and the absolute
-60 humidity of the ambient air, but it is not directly correlated to the
ambient dry bulb temperature.
-70
0 2 4 6 8 10
2.1.5. Convection
Cloud Cover (Tenths) Evaporative and convective heat transfer phenomena are
Fig. 1. Exchange of longwave radiation as a function of cloud cover (Ta ¼ 30
C,
related; they operate by very similar mechanisms and can be
HR ¼ 0.010, Tw ¼ 25
C). reasonably conceptualized as a single process of coupled heat and
mass transfer. Mass transfer and the associated transformation of
3. Evaporation is sensitive to local environmental conditions, which sensible heat to latent heat occur by evaporation, while sensible
differ from available meteorological data due to the proximity of heat transfer occurs by convection. A difference in absolute
obstructions such as buildings and trees, and microclimatic humidity, or vapor pressure, is the driving potential for evapora-
patterns related to neighborhood scale phenomena tion, and a temperature difference is the driving potential for
convection. Wind affects both by increasing the total effective
The model developed herein couples empirical formulae for interface for heat and mass transfer, and notwithstanding the role
mass transfer from free water surface evaporation with empirical of wind, evaporation and convection are rate limited by mass
and theoretical heat transfer correlations to develop a more diffusivity and thermal diffusivity respectively. The evaporation
complete model of the evaporative heat and mass transfer mech- equations indicate that the water mass provides all sensible heat for
anisms at play in a swimming pool. phase change to latent heat through evaporation. However, if you
Evaporation is driven by wind speed, and by the difference consider evaporation and convection together, it is clear that as the
between the saturated vapor pressure of air at the pool surface water cools sensibly due to evaporation, convective heat transfer
temperature and the vapor pressure of ambient air. Thus, evapo- rates will shift, and given that the air is warmer than the water
ration is greater in arid climates, and is typically the most signifi- surface some sensible heat for evaporation will effectively be drawn
cant heat transfer mechanism for the overall energy balance of from the air by convection. In contrast, if evaporation occurs under
a pool. Equations (10) and (11) were developed by McMillan [12], conditions where the air is cooler than the water, all sensible heat
and confirmed by Sweers [13] and Sartori [11]. They describe the for evaporation must necessarily be derived from the water mass.
relationship between evaporative heat transfer and relevant envi- Thus, as the two phenomena are closely related, the convective
ronmental conditions. heat transfer rate can be derived theoretically as a function of the
evaporative heat transfer rate. Bowen [14] expresses the relation-
qevap ¼ hevap $ðes  ea Þ (10) ship as a ratio:
hevap ¼ 0:0360 þ 0:0250$V (11) qconv
¼ Rbowen (12)
where the wind velocity V is corrected to a height of 3 m. qevap
which can be calculated by:
0

pa ðTw  Ta Þ
-20 Rbowen ¼ CBowen $ $ (13)
po ðes  ea Þ
Longwave Radiation (W/m2)

-40 Bowen developed this formula from first principles, based on

a control volume analysis of sensible and latent heat densities in
-60 a differentially small element of air, subject to molecular and
thermal diffusivity, and forced air movement. Note that the formula
-80 accounts for the impact of ambient pressure on the ratio of
convective and evaporative heat transfer. Using the Bowen ratio,
-100
the convective heat transfer rate is determined simply by multi-
plying by the evaporative heat transfer rate:
-120
qconv ¼ Rbowen $qevap (14)

-140 The net heat transfer by the coupled process of convection and
15 20 25 30 35 40
evaporation is simply the sum of qconv and qevap. Fig. 3 plots Rbowen
Ta, Ambient Air Temperature (C)
alongside the net heat transfer by convection plus evaporation, for
Fig. 2. Exchange of longwave radiation as a function of ambient air temperature (clear several different ambient dry bulb temperature conditions, all as
sky, HR ¼ 0.010, Tw ¼ 25
C). a function of humidity ratio. Note that the humidity ratio at low
 J. Woolley et al. / Building and Environment 46 (2011) 187e195 191

4 200

Net Convective & Evaporative Heat Transfer (W/m2)

3 150
B, Ta=15
B, Ta=20
2 100
B, Ta=25
B, Ta=30
Bowen Ratio 1 50
B, Ta=35
0 0 B, Ta=40
qconv+qevap, Ta=15
-1 -50
qconv+qevap, Ta=20
qconv+qevap, Ta=25
-2 -100
qconv+qevap, Ta=30
-3 -150 qconv+qevap, Ta=35
qconv+qevap, Ta=40
-4 -200
0.005 0.01 0.015 0.02 0.025
Humidity Ratio

Fig. 3. Plot of Bowen ratio and net heat transfer by convection and evaporation, for several different ambient dry bulb temperatures, as a function of humidity ratio. Wind
speed ¼ 1 m/s, water temperature ¼ 25
C.

ambient dry bulb temperatures is limited by saturation, and that where:

the Bowen ratio diverges asymptotically at the saturation humidity
for air at the water surface temperature.
For conditions where humidity of the ambient air is less than NuL $kair
h ¼ (16)
the saturation humidity at the water surface temperature (the left Lc
side of Fig. 3), a positive Bowen ratio indicates that heat is lost A
Lc h (17)
from the water mass by both evaporation and convection. As P
illustrated by the net convective and evaporative heat transfer

1=3
lines in Fig. 3, these conditions result in the maximum water NuL ¼ RaL 107 (RaL (1011 (18)
cooling effect. As ambient dry bulb temperature approaches the
water surface temperature the Bowen ratio approaches zero. At and for conduction with non-circulating air:
this point no heat is exchanged by convection, though heat may
still be lost from the water by evaporation. A negative Bowen ratio ka ðTw  Ta Þ
qconv ¼ 0:932$ (19)
indicates that evaporation and convection have opposing effects. Lc
A ratio between 0 and -1 means that the cooling effect of evapo- where:
ration is dominant, a ratio of 1 indicates a net-zero balance of
rffiffiffiffiffiffiffi
convection and evaporation, and a ratio beyond 1 means that As
convective heat gains to the pool dominate evaporative losses, Lc ¼ (20)
4p
resulting in a net heat gain to the water. For conditions where
ambient humidity is greater than the saturation humidity at the
As ¼ 2wL (21)
water surface temperature (the right side of Fig. 3), condensation
and convection will both contribute heat to the water mass. Note The result of this validation shows that at zero wind speed, deriving
that for such conditions the Bowen ratio must be positive, since the average convective heat transfer rate from the evaporative heat
such absolute humidity conditions cannot occur for air tempera- transfer rate and the Bowen Ratio, as described by equation (10),
tures below the water surface temperature. agrees with other well developed models for buoyancy driven
At zero wind speed, the solution of equation (14) should equate convection or conduction in a non-circulating air mass to within 5%.
to other well developed theoretical models for heat transfer. When
the water surface temperature is greater than the air temperature 2.1.6. Other mechanisms
the solution should agree with models for buoyancy driven free Other mechanisms that affect the pool temperature include
convection. If water surface temperature is less than the air swimmers, rain, makeup water, pool covers, and the thermal effect of
temperature, air movement should stagnate above the water mass pumps. Accounting for swimmers is particularly difficult because of the
and equation (14) should yield similar results to models for myriad variables involved, though some authors have considered it. For
conduction with a semi-infinite non-circulating mass. Equations example, Molineaux [15] assumes a heat addition of approximately
(15) through (18), presented by Incropera [9], were used to validate 400 calories per swimmer per hour, which would have a measurable
equation 14 at zero wind speed for cases of buoyancy driven free effect on water temperature in a pool with heavy use. The impacts of
convection; equations (19) through (21) were used similarly for rain and makeup water may be significant in certain instances and can
conduction with a stagnant air mass. be calculated by accounting for the mass and temperature of the added
For buoyancy driven convection: water, though these values may be difficult to estimate. Pool covers can
significantly impact the thermodynamics of a pool, mostly by elimi-
qconv ¼ h$ðTw  Ta Þ (15) nating evaporation and reducing longwave radiative losses. Each pool
192 J. Woolley et al. / Building and Environment 46 (2011) 187e195

cover has different absorptive, emissive, and insulative properties and from the California Department of Water Resources’ California
must be modeled accordingly. Pumps contribute heat to the pool, in Irrigation Management Information System (CIMIS) [16] which
part through frictional interactions with piping and dissipation of reports measurements for a meteorological station in Davis, CA, as
kinetic energy, and in part by way of heat generated within the pump. well as hundreds of other sites throughout the state. The meteo-
The design, operation, and site-specific meteorological characteristics rological station used to obtain hourly insolation values was close
of each pool will impact the relative importance of each of these heat by and assumed to be representative of the test site. The direct and
additions, such that in certain cases they should be included in the diffuse portions of this measurement were calculated using a quasi-
model. physical model for converting hourly global horizontal to direct
normal insolation developed by Maxwell and published by the
2.2. Model validation NREL Solar Energy Research Institute [17].
Since the meteorological measurements of insolation are fully
To validate the calculations discussed herein, an experiment was exposed, whereas pools are often surrounded by obstructions, an
conducted at a residential swimming pool in Davis, CA, a relatively hourly shading factor for the pool was developed by inspection of
hot and dry region in California Climate Zone 12. The thermal the site and an analysis of solar pathways for the latitude, season,
behavior of the pool was monitored over a 56 day period in spring and time period of the experiment. The model uses diffuse inso-
2009, and observations were compared to results from the model lation as the solar input for shaded periods and global insolation for
using hourly meteorological data for the same time period. The un-shaded periods; periods of fractional shading receive a corre-
pool was left uncovered, no swimmers were permitted, no makeup sponding fraction of global and diffuse insolation.
water was added, and the filter pump was set to run continuously at For the experimental validation presented here, ambient
a constant flow rate (Fig. 4). temperature, relative humidity, and wind speed were measured on
site. Some error is incurred due to slight variations in meteoro-
2.2.1. Methodology for experimental validation logical conditions at different points on site; however, the location
The model requires definition of an initial pool temperature, as of each measurement was selected to avoid significant misrepre-
well as several meteorological variables on an hourly basis, sentation of conditions at the pool surface. For example, the
including: global horizontal insolation, pool shading, cloud cover, anemometer was placed to avoid eddies and vorticies that could
ambient dry bulb temperature, ambient humidity, wind speed, and occur very near a building. The wind speed measurement was
barometric pressure. Future iterations of the model will accept the corrected to a 3 m height using standard atmospheric boundary
definition of an hourly thermal input from vapor-compression layer methods [18] since the McMillan wind speed coefficient
cooling equipment, though this study focuses on validation of the presented in equation (11) is derived for wind speed at that height.
thermodynamic balance between a passive swimming pool and the These measurements could all be taken from regional meteoro-
environment. logical data, or from typical meteorological year resources, though
Note that an accurate initial pool temperature is not absolutely microclimatic variations between a meteorological station and an
necessary for the model to appropriately predict the long-term hourly actual site introduces errors that are not associated with the
temperature behavior of the pool, though it may take up to a few weeks mathematical model itself. Wind speed is an especially sensitive
for the model to track the actual pool temperature if the initial condi- input variable. Since meteorological stations tend to be located in
tions are off by 5
C. The results presented here use measured pool open, unobstructed areas, and sites of interest are often surrounded
temperature as an initial condition for the simulation. by nearby obstructions such as trees, fences, and buildings,
Although global horizontal solar insolation could be measured measurements near to the ground do to not scale well using stan-
on site, in certain instances it is impossible to install a pyranometer dard atmospheric boundary layer methods to correct for terrain
in a completely un-shaded location; thus the model is designed to differences. For example, over the test period presented here, CIMIS
allow input from offsite meteorological measurements. The global wind speed observations at 2 m in open terrain and corrected to
horizontal insolation for each hour of the experiment was obtained 3 m in a highly obstructed urban area were consistently high as

Fig. 4. Photo of pool in Davis California used for experimental validation.

 J. Woolley et al. / Building and Environment 46 (2011) 187e195 193

compared to actual measurements, with an RMS error of 1.6 m/s. 24.5

This overestimation of wind speed would result in a consistent 24.3
under estimation of pool temperature. Relative humidity and
ambient temperature measurements vary as well, such that the 24.1

pool temperature predictions from a simulation using CIMIS data 23.9

Temperature (°C)
and a simulation using site data differ with an RMS error of 3.0
C.
23.7
Calculation of longwave radiative exchange requires informa-
tion about the fractional portion of the sky dome that is obstructed 23.5
by clouds, though this data is not regularly collected by all mete-
23.3
orological stations. For this simulation, data was obtained from the
National Oceanic and Atmospheric Administration’s (NOAA) 23.1
Quality Controlled Local Climatological Database (QCLCD) [19] for
22.9
a station in Sacramento, CA, which is about 20 miles from the
experimental site. 22.7
Conduction between the pool and earth was calculated using Predicted Measured
22.5
a constant soil temperature of 15
C. Transient effects due to diurnal 5/22/09 0:00 5/22/09 6:00 5/22/09 12:00 5/22/09 18:00 5/23/09 0:00
and seasonal heat transfer from the pool were ignored, and the pool Date
geometry was approximated as a cuboid as described previously by
Fig. 6. Predicted and measured temperatures for a case of very good accuracy.
equation (6).
The temperature of the pool was measured at 10, 40 , and 70 from the
pool bottom to develop an average pool temperature and to describe horizontal insolation, since those values were measured at offsite
the extent of thermal stratification. For this experiment, since the filter meteorological stations they could differ somewhat from local
pump ran continuously at a constant flow rate, the pool was relatively conditions. The error during this period can be reduced almost
well mixed and no thermal stratification was observed. completely if values for cloud cover are inflated, but there are no
theoretical grounds to include such adjustments in the model. The
2.2.2. Validation results periods of error in the middle of the test are not related to any
The model described herein used the initial inputs and hourly obvious meteorological event, and are not easily explained. In all
meteorological conditions to determine heat and mass exchange instances the discrepancy rarely reaches 1.0
C. It is noteworthy
between the pool and environment, and to predict hourly average that the model recovered from poor prediction periods automati-
pool temperature. The predicted values were then compared to the cally, without any input other than the new meteorological data,
observed temperature history and analyzed for accuracy, Fig. 5 which suggests that it is a robust model.
illustrates the results. Figs. 6 and 7 illustrate typical diurnal cycles for the pool during
The results suggest that, given input of appropriate meteorological the observed period. Fig. 6 is an example of one day for which the
conditions, an accurate prediction of the pool temperature can be model gives a very accurate prediction, while Fig. 7 is for a relatively
made. The RMS error of the pool temperature prediction compared to poor prediction. In both instances the simulation is in phase with the
measured values is 0.4
C and the largest discrepancy is only 1.1
C. The measurements; though the measured pool temperature transitions
temperature sensors used in the experiment had an absolute error of gradually between heating and cooling, while the predicted pool
0.2
C, so accuracy of the model is very near sensor accuracy. temperature responds more abruptly. This is likely due to the hourly
The most significant periods of error between measured and time step implemented in the simulation. In a physical system
predicted pool temperature occur for approximately one week near meteorological conditions change continuously while the model
the beginning of the test period, and for several days near the relies on constant values for each hour. If weather data were resolved
middle of the test period. The first instance is likely due to a storm more continuously the model would respond more gradually.
that brought cloud cover and measurable precipitation. Although Fig. 8 plots all hourly predicted pool temperatures against all
the model responds to data for both opaque cloud cover and global hourly measured pool temperatures. A perfectly accurate model

27.0 24.5

24.3
25.0
24.1
Temperature (°C)

23.9
Temperature (°C)

23.0
23.7
21.0
23.5

23.3
19.0
23.1

17.0 22.9
Predicted Measured
22.7
15.0 Predicted Measured
4/27/2009 5/7/2009 5/17/2009 5/27/2009 6/6/2009 6/16/2009 6/26/2009 22.5
Date 6/10/09 0:00 6/10/09 6:00 6/10/09 12:00 6/10/09 18:00 6/11/09 0:00
Date
Fig. 5. Predicted and measured pool temperatures for a pool in Davis California
observed 4/29/2009e6/22/2009. Fig. 7. Predicted and measured temperatures for a case of poor accuracy (error of 0.6
).
194 J. Woolley et al. / Building and Environment 46 (2011) 187e195

26.0

25.0
Solar Insolation
24.0 51%
Longwave
23.0 Radiation 26%
Predicted (°C)

22.0

21.0 Evaporation 20%

R2 = 0.967
20.0

19.0

18.0

17.0 Convection 2% Conduction 1%

17.0 18.0 19.0 20.0 21.0 22.0 23.0 24.0 25.0 26.0
Measured (°C) Fig. 9. Relative magnitude of each heat transfer mechanism integrated across the
duration of the experiment.
Fig. 8. Comparison of predicted and measured pool temperature for all hourly
observations.

3. Discussion
would have a one-to-one relationship between the values. For the
period of validation this simulation has very good fit with an R2 3.1. Experimental considerations
value of 0.967.
The relative impact of each heat transfer mechanism over the Although average pool temperature has been presented as the
duration of the experiment is illustrated in Fig. 9. Note that solar metric by which to validate simulations, mass evaporation could be
insolation is the only heat gain, the sum of all heat losses used as well. The predicted mass evaporation rate can be calculated
balances with the solar gains, and that evaporation and emis- directly in the model by relating heat transfer by evaporation to the
sion of longwave radiation dominate over conduction and latent heat of vaporization of water; and the actual cumulative
convection. evaporation can be measured directly by monitoring the water
Fig. 10 plots the magnitude of each heat transfer mechanism and level. However, since the mass rate of evaporation is small
the total heat accumulated across two typical days from the compared to pool volume, it is very difficult to accurately measure
experiment. Convection and longwave radiative exchange with the changes in depth on an hourly basis. The barometrically corrected
sky are affected directly by diurnal airetemperature cycles while water depth sensors used for our experimentation are accurate to
evaporation is not. Note that heat flux into the pool is the positive within 0.0035 m, so for a pool with 50 m2 surface area and 5 kg/h
convention, so negative values represent cooling of the pool. evaporation the hourly change in depth of 0.0001 m cannot be
Longwave radiative exchange with the sky is consistently negative reliably observed, especially considering the noise associated with
since the effective sky temperature never exceeds the pool naturally occurring disturbances to the water surface. The issue was
temperature during the plotted period. further complicated in this experimental validation because the

40 2000

35
Cumulative Energy Stored (MJ)

30 1500

25
Heat Flow (kW)

20 1000

15

10 500

0 0

-5

-10 -500
6/1/09 0:00 6/1/09 12:00 6/2/09 0:00 6/2/09 12:00 6/3/09 0:00
Date
Evaporation Longwave Radiation Insolation
Convection Cumulative Energy Stored

Fig. 10. The magnitude of each heat transfer mechanism and heat accumulation for two typical days of the experiment (conduction heat flow was omitted from this figure as it was
effectively zero).
 J. Woolley et al. / Building and Environment 46 (2011) 187e195 195

depth sensor was mounted to a pole at the end of a diving board, under multiple environmental conditions including clear and
which seemed to expand and contracted slightly with diurnal cloudy scenarios, as well as cool and very hot conditions. However,
temperature cycles, causing sensor movement and misrepresen- the model was not validated for extended cold periods, heavy rain
tation of fluctuations in water depth. In subsequent field tests the conditions, mechanical thermal loading, or extreme climates.
depth sensor will be placed at a fixed point, and the water column
above the sensor will be isolated from small disturbances to the
5. Legal notice
pool surface.
This report was prepared as a result of work sponsored by the
3.2. Future work
California Energy Commission (Energy Commission) and the
University of California (UC). It does not necessarily represent
The next phase of model development and validation involves
the views of the Energy Commission, UC, their employees, or the
the addition of heat from vapor-compression space cooling equip-
State of California. The Energy Commission, the State of California,
ment, and the development of design guidelines for such heat
its employees, and UC make no warranty, express or implied, and
pump systems in various climate zones in the western United
assume no legal liability for the information in this report; nor does
States. A preliminary simulation was conducted for the pool studied
any party represent that the use of this information will not infringe
here with the addition of heat rejected from a condenser. The
upon privately owned rights. This report has not been approved or
condenser heat for each hour was calculated for a 3.5 ton heat
disapproved by the Energy Commission or UC, nor has the Energy
pump assuming a constant COP, and was based on cooling loads
Commission or UC, nor has the Energy Commission or UC passed
generated in MICROPAS [20] for a 1764 square foot, single story
upon the accuracy or adequacy of the information in this report.
home in California Climate Zone 12. Under this scenario the pool
temperature never exceeded 28.5
C. Another experiment will be
conducted to compare this model with observations from References
a geothermal heat pump system that is coupled to a swimming pool
with a gas-fired pool heater, solar thermal pool heaters, night [1] Brown EG. In: Justice DO, editor. Framework document for central air condi-
tioners and heat pumps; 2008. Oakland.
radiative coolers, and fountains for evaporative water cooling. The [2] Turpin JR. Homeowners dive into savings. In: Air conditioning, heating &
intent is to account for the impact of all system components in the refrigeration news; 2009.
model in order to simulate performance under various configura- [3] Randazzo R. Shasta pools develops unique cooling system. Pheonix: The Ari-
zona Republic; 2009.
tions in multiple climate zones and develop guidelines to reduce [4] Wu H. A mathematical procedure to estimate solar absorptance of shallow
energy consumption for space cooling while preventing over- water ponds. Energy Conversion and Management 2009;50:1828e33.
heating. Additionally, future work will explore the potential to [5] Hahne E. Monitoring and simulation of the thermal performance of solar
heated outdoor swimming pools. Solar Energy 1994;53(1):9e19.
offset pool heating costs during swing seasons when pool [6] Govaer D. Analytical evaluation of direct solar heating of swimming pools.
temperatures are low yet space cooling is required. Research is Solar Energy 1981;27:529e33.
needed to clarify when this occurs and how much energy could be [7] Rakopoulos CD. A model of the energy fluxes in a solar heated swimming pool
and its experimental validation. Energy Conversion Management 1987;27
saved in various climate zones, and with different degrees of pool
(2):189e95.
shading. [8] Hull JR. Dependence of ground heat loss upon solar pond size and Perrimeter
insulation e calculated and experimental results. Solar Energy 1984;33
4. Conclusions (1):25e33.
[9] Incropera. Introduction to heat transfer. 5th ed. Wiley; 2007.
[10] Walton G. Thermal analysis research program reference manual. National
Predictions from the mathematical model developed match well Bureau of Standards, NBSIR; 1983. p. 83e2655.
with measured pool temperature results, suggesting that it could [11] Sartori E. A critical review on equations employed for the calculation of the
evaporation rate from free water surfaces. Solar Energy 1999;68:77e89.
be used to accurately analyze the temperature response of a pool [12] McMillan W. Heat dispersal e lake Trawsfynydd cooling studies. Symposium
used as a thermal sink for a heat pump during the cooling season, or on Freshwater Biology and Electrical Power Generation 1971;1:41e80.
as a thermal source for a heat pump in the heating season. The [13] Sweers HE. A nomogram to estimate the heat-exchange coefficient at the
airewater interface as a function of wind speed and temperature; a critical
accuracy of the model is impressive, and is due mostly to the survey of some literature. Journal of Hydrology 1976;30:375e401.
extensive theoretical and empirical research by other authors to [14] Bowen IS. The ratio of heat losses by conduction and by evaporation from any
explain each heat transfer mechanisms at play in this scenario. It water surface. Physical Review 1926;27:779e87.
[15] Molineaux B. Thermal analysis of five outdoor swimming pools heated by
should be noted that our methodology to describe shading of the
unglazed solar collectors. Solar Energy 1994;53(1):21e6.
pool each hour is the only variable that was not derived from other [16] California Irrigation Management Information System. California Department
published work or directly measured with instrumentation, and of Water Resources, Office of Water Use Efficiency; 2009.
[17] Maxwell EL. A quasi-physical model for converting hourly global horizontal to
that no “correction factors” have been used to calibrate the model
direct normal insolation. Midwest Research Institute; 1987.
against the measurements. Although the model is very accurate, if [18] ASHRAE handbook, fundamentals. 2009 ed. American Society of Heating,
used as a design tool it should be noted that meteorological Refrigeration, and Air-Conditioning Engineers, Inc; 2009.
conditions at a site may differ significantly from available data, and [19] Quality Controlled Climatological Database, US Department of Commerce,
National Oceanic and Atmospheric Administration, National Environmental
that predictions may not be as accurate as the validation results Satellite, Data, and Information Service, National Climatic Data Center.
presented here. The test period allowed for validation of the model [20] MICROPAS; 2010.