Vous êtes sur la page 1sur 13

Agricultural and Forest Meteorology 152 (2012) 3143

Contents lists available at SciVerse ScienceDirect


Agricultural and Forest Meteorology
j our nal homepage: www. el sevi er . com/ l ocat e/ agr f or met
The aerodynamics of pan evaporation
Wee Ho Lim
a
, Michael L. Roderick
a,b,
, Michael T. Hobbins
a,c
, Suan Chin Wong
a
, Peter J. Groeneveld
a
,
Fubao Sun
a
, Graham D. Farquhar
a
a
Research School of Biology, The Australian National University, Canberra, ACT 0200, Australia
b
Research School of Earth Sciences, The Australian National University, Canberra, ACT 0200, Australia
c
Colorado Basin River Forecast Center, National Weather Service, National Oceanic and Atmospheric Administration, Salt Lake City, UT 84116, USA
a r t i c l e i n f o
Article history:
Received 12 April 2011
Received in revised form10 August 2011
Accepted 19 August 2011
Keywords:
Pan evaporation
Vapour transfer
Boundary layer theory
Aerodynamics
a b s t r a c t
In response to worldwide observations reporting a decline in pan evaporation over the last 3050 years,
we developed an instrumented US Class A pan that replicated an operational pan at Canberra Airport in
Australia. The aim of the experimental facility was to investigate the physics of pan evaporation under
non-steady state conditions. By monitoring the water level at 5-min intervals we were able to calculate
the evaporation rate and thereby determine the short-term mass balance of the pan. Over the same
time intervals, we also monitored (short- and long-wave) radiation, temperature (air, water surface, bulk
water, inner and outer pan wall), atmospheric pressure as well as the air vapour pressure and the wind
speed at a standard reference height (2 m above ground level). The experimental pan was operated for
three years (20072010).
In this paper, we develop a framework for quantifying vapour transfer by coupling Ficks First Law of
Diffusion with boundary layer theory. This approach adequately represented pan evaporation measure-
ments over short time intervals (half-hourly) under non-steady state conditions provided that surface
temperature measurements, that account for the substantial cooling associated with evaporation, are
available. It involved estimating the boundary layer thickness and other properties of air above the evap-
orating surface for a pan. Our results are consistent with the envelope of theoretical curves concept for
the wind function introduced by Thom et al. (1981).
2011 Elsevier B.V. All rights reserved.
1. Introduction
Pan evaporation is the evaporation froma standard water-lled
dish and is the most widely used physical measure of the evapora-
tive demand of the atmosphere. Pan evaporation measurements
have been widely used in agricultural meteorology due to their
simplicity, low cost and proven ease of application for irrigation
scheduling (Stanhill, 2002). Due to the widespread applications,
evaporation pans in various forms have been deployed in many
regions for at least the last several decades (Brutsaert, 1982). Anal-
ysis of worldwide pan evaporation data has found changes, mostly
declines (Roderick et al., 2009a,b) despite the trend of rising global
average air temperature. This has become known as the pan evapo-
ration paradox (Peterson et al., 1995; Brutsaert and Parlange, 1998;
Roderick and Farquhar, 2002), because it has occured concurrently
with the global warming. Reduction in irradiance (Roderick and
Farquhar, 2002) and wind speed appear to have been major causes
of this phenomenon (Roderick et al., 2007).

Corresponding author at: Research School of Biology, The Australian National


University, Canberra, ACT 0200, Australia.
E-mail address: Michael.Roderick@anu.edu.au (M.L. Roderick).
For pans located above the ground, the surface area for heat
transfer is larger than for mass transfer (Kohler et al., 1955; Riley,
1966). Consequently, most studies haveemployedapancoefcient,
typically 0.7 for a US Class A pan (Stanhill, 1976), by which pan
evaporationis multipliedtogiveavaluemorerepresentativeof nat-
ural evaporation, toaccount for this largelyradiative effect (Linacre,
1994). Interms of the aerodynamics of panevaporation, Thomet al.
(1981) examined the roles of free and forced convection. Rotstayn
et al. (2006) subsequently developed the PenPan model of pan
evaporation by combining the radiative model of Linacre (1994)
with the aerodynamic model of Thom et al. (1981). The PenPan
model has been used to attribute the cause of changes in pan evap-
oration in both observations (Roderick et al., 2007; Shuttleworth
et al., 2009) andinclimate models (JohnsonandSharma, 2010). The
derivation of the PenPan model assumed steady state conditions
that require a typical integration period of around 1 week in sum-
mer and up to 1 month in winter (Roderick et al., 2009a). However,
the radiative forumulation (Linacre, 1994) that underlies the Pen-
Pan model, whilst physically based, has never been experimentally
tested. Further, the aerodynamic formulation (Thom et al., 1981)
has not, to our knowledge, been subject to an independent exper-
imental test. These experimental tests have a high priority given
the prominent role attributed to declines in windspeed (stilling)
0168-1923/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.agrformet.2011.08.006
32 W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143
and/or solar radiation in explanations of the world-wide decline in
pan evaporation (Roderick et al., 2007, 2009b).
Detailed physical investigations on pan evaporation have been
rare. Notably, Jacobs et al. (1998) have studied the sub-daily ther-
mal behaviour of a standard US Class A pan. They found that the
water in the pan was generally well-mixed. Their pan was located
on the ground and therefore in strong thermal contact with the
soil surface. We expect that the nding of a well-mixed water
body would also hold for a pan located on an elevated (150mm)
wooden platformsuch as used in standard US Class A pan installa-
tions. The question of any depression in surface temperature due
to evaporation froma thin (1mmor so) layer immediately below
the surface (Ward and Stanga, 2001; Hisatake et al., 1993, 1995)
was not explicitly addressed because Jacobs et al. (1998) used a
Penman-style formulation for the energy balance where the sur-
face temperature was eliminated from the underlying equations.
More recently, the nding of well-mixed water within the pan has
alsobeenreportedbasedonexperiments using aninstrumentedUS
Class A pan (Martinez et al., 2006). In that research, the water sur-
face temperature was assumed to be uniformover a 40mmdeep
layer. As noted above, this is unlikely to be true for an evaporat-
ing surface (Ward and Stanga, 2001; Hisatake et al., 1993, 1995),
but again, the Penman-style formulation used in that research
(Martinez et al., 2006) effectivelyeliminatedthe water surface tem-
perature from the equations. A second caveat applicable to that
study was that the external walls of the pan were insulated to min-
imise heat transfer and thereby simplify the analysis. Therefore,
that installation did not mimic standard evaporation pans.
To contribute to a better understanding of the world-wide
trends in pan evaporation, we have constructed an instrumented
experimental pan that replicates existing standard installations.
We installed specialised sensors onto a standard US Class A pan
as used by the Australian Bureau of Meteorology (BoM). The water
level, (short- andlong-wave) radiation, temperature(air, water sur-
face, bulk water, inner and outer pan wall), atmospheric pressure
as well as the air vapour pressure and the wind speed at a standard
reference height (2m above ground level) were all monitored at
5-min intervals for a three-year period (20072010).
This paper describes the rst step in the development of a new
parameterisation for a Penman-type combination equation for pan
evaporation. Here, we focus explicitly on the aerodynamic compo-
nent of pan evaporation. To do that, we develop a framework for
quantifying vapour transfer by coupling Ficks First Law of Diffu-
sionwithboundarylayer theoryassumingthat surfacetemperature
measurements are available. We investigate the underlying physics
of mass transfer and test our theory using data collected at the
experimental pan. We relate our research to the ideas put forward
by Thomet al. (1981).
2. Theory
Section 2.1 summarises existing vapour transfer equations
employed in environmental applications. Section 2.2 derives a
vapour transfer equation based on Ficks First Law of Diffusion.
Section 2.3 describes the signicance of boundary layer theory in
vapour transfer. Section 2.4 links the boundary layer theory with
Ficks First Law of Diffusion for vapour transfer and presents an
approach to quantifying vapour transfer froman evaporation pan.
2.1. Existing vapour transfer equations
Daltons Lawassumes that vapour moves fromhigh to lowpar-
tial pressure and is generally expressed as
E (e
s
(T
s
) e
a
(T
a
)) = f
v
(e
s
(T
s
) e
a
(T
a
)) (1)
where E [ms
1
] is the evaporation rate of liquid water in tradi-
tional hydrologic units of depth per unit time, e
s
(T
s
) [Pa] is the
vapour pressure at the evaporating surface, e
a
(T
a
) [Pa] is the air
vapour pressure at the same height that air temperature is mea-
sured at and f
v
[ms
1
Pa
1
] is the aerodynamic function. (Note that
e
s
(T
s
) is taken to be the saturated vapour pressure at the surface
temperature.) f
v
depends on both wind speed and the temperature
difference between the evaporating surface and the air, especially
whenwindspeeds are low(Thomet al., 1981). For simplicity (Thom
et al., 1981), f
v
has usually been taken as a function of wind speed
and Eq. (1) becomes
E = f (u)(e
s
(T
s
) e
a
(T
a
)) (2)
where f(u) is known as the wind function. In terms of the
widely usedresistance terminology (Monteith, 1965; Monteithand
Unsworth, 2008), Eq. (1) can be rewritten as
E =
M
w
M
a

w
P
a
(e
s
(T
s
) e
a
(T
a
))
r
a
0.622

a

w
P
a
(e
s
(T
s
) e
a
(T
a
))
r
a
(3)
where r
a
[s m
1
] is the aerodynamic resistance, M
a
[kgmol
1
] is
the molecular mass of air, M
w
[kgmol
1
] is the molecular mass of
water,
a
[kgm
3
] is the density of air,
w
[kgm
3
] is the density
of liquid water and P
a
[Pa] is the atmospheric pressure. In practice,
this is more or less equivalent to Eq. (2) since r
a
is mostly driven by
the wind (Chu et al., 2010). In subsequent sections we examine the
functional formof the relation between r
a
and wind speed.
2.2. Ficks First Law of Diffusion
Ignoring complications of inhomogeneity, a one-dimensional
form of Ficks First Law of Diffusion (see Fick (1995) for a trans-
lation) can be written as
J = D
dC
dz
(4)
where J [mol m
2
s
1
] is the ux density, D [m
2
s
1
] is the diffusion
coefcient, C [mol m
3
] is the molar concentration and z [m] is the
distance. (Note: the negative sign implies that J is positive when
diffusion is toward the lower concentration.) For vapour transfer,
Eq. (4) can be expressed in a nite formassuming an ideal gas
J =
D
v
R
(e
a
(T
a
)/T
a
) (e
s
(T
s
)/T
s
)
z
=
D
v
R
(e
s
(T
s
)/T
s
) (e
a
(T
a
)/T
a
)
z
(5)
where D
v
[m
2
s
1
] is the diffusion coefcient for water vapour in
air, R [J mol
1
K
1
] is the ideal gas constant, T
s
[K] is the water sur-
face temperature, T
a
[K] is the air temperature and z [m] is the
boundary layer thickness (see details in Sections 2.3 and 2.4). For
typical environmental conditions,
e
s
(T
s
)
T
s

e
a
(T
a
)
T
a

e
s
(T
s
) e
a
(T
a
)
T
a
,
and Eq. (5) can be simplied as
J
1
R
D
v
T
a
(e
s
(T
s
) e
a
(T
a
))
z
(6)
This is Ficks First Law of Diffusion in Daltons form for vapour
transfer (see Appendix D for relation to formulations commonly
used in plant physiology). It should be noted that D
v
is not a con-
stant, but increases with T
a
(Gilliland, 1934) and varies inversely
withP
a
(e.g., MonteithandUnsworth, 2008; Rohsenowet al., 1985).
An equation to calculate D
v
as a function of T
a
and P
a
is given in
W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143 33
Fig. 1. Aschematic diagramof the threshold model adopted here for the variation
in vapour pressure (e) with height (z) above the evaporating surface.
After Leighly (1937) and Machin (1964, 1970).
Appendix A. We can express Eq. (6) in units of depth per unit time
(i.e., E) by incorporating M
w
and
w
,
E =
M
w
R
w
D
v
T
a
(e
s
(T
s
) e
a
(T
a
))
z
(7)
This is the general formof Ficks First Lawof Diffusionfor vapour
transfer based on an ideal gas. By comparing Eqs. (1) and (7) we see
that the aerodynamic function (f
v
in Eq. (1)) is
f
v
=
M
w
R
w
D
v
T
a
1
z
(8)
and the conductance g
m,v
[mol m
2
s
1
] is given by
g
m,v
=
1
R
D
v
P
a
T
a
1
z
(9)
2.3. Boundary layer theory
Thickness of the boundary layer in air has been measured
directly at various times and by various methods. Its order of
magnitude is froma few millimeters down to some small frac-
tion of a millimeter. Within it, gradients of vapor concentration,
temperature, and velocity are linear. Leighly (1937)
Current conceptions of vapour transfer are for a thin (bound-
ary) layer of air above the evaporating surface with transport of
vapour across that layer by molecular diffusion (e.g., Giblett, 1921,
pp. 473474; Penman, 1948) and this is conrmed by observa-
tion (Doe, 1967). Further, the boundary layer thickness is known
to decrease as the wind speed increases (Machin, 1964, 1970). The
treatment consistent with those experimental results holds that
the vapour pressure at the top of the boundary layer is the same as
the vapour pressure at a reference height (e.g., 2m above ground
level) (Leighly, 1937). The simplied threshold model is depicted
in Fig. 1.
To ensure that the framework adopted here can be applied at
other evaporation pans we make the initial assumption that mea-
surements of vapour pressure, air temperature and wind speed
will be available at the reference height. Hence, the challenge is
to estimate the boundary layer thickness z using those available
measurements.
2.4. Boundary layer thickness
Vapour transfer can be conceived as due to free or forced con-
vection, or a mixture of both, often called mixed convection
(Monteith and Unsworth, 2008). It is difcult to specify precise
boundaries between these various convection regimes. Here, we
propose a convenient structure for estimating z without a pri-
ori selection of the thresholds. Following boundary layer theory
(Hisatake et al., 1993, 1995) we formulate the boundary layer thick-
ness as
z = f (Re, L) Re
q
L (10)
where Re is the Reynolds number (dimensionless) and L [m] is the
characteristic length of the evaporating surface. For a cylindrical
evaporation pan, we assume that L is the diameter (1.21m for a
US Class A pan). Here q is a dimensionless constant (range: 0 to
1.0). Conventionally, q is 0.5 for laminar owover a at surface
(Schlichting, 1960).
Traditionally, Re is calculated using the free stream velocity
of the air and adjusted using a numerical factor (e.g., Eq. 2.2
in Schlichting (1960)). Conceptually, the numerical factor is an
attempt to estimate the wind velocity immediately adjacent to
the evaporating surface. Importantly, the numerical factor of Re
will depend on the height at which wind velocity is measured. An
alternative formulation is to calculate Re using the wind velocity
immediately adjacent to the evaporating surface. To develop such
an expression for Re we note that
Re
inertial forces
viscous forces
=

a
u
s
L

a
(11)
where u
s
[ms
1
] is a three-dimensional wind velocity immedi-
ately adjacent to the evaporating surface and
a
[kgm
1
s
1
] is the
dynamic viscosity of air. (Note that
a
is a function of air tem-
perature, see Appendix B.) Here, u
s
f(u
V
, u
H
) where u
V
[ms
1
]
and u
H
[ms
1
] are the vertical (analogous to free convection) and
horizontal (analogous to forced convection) components respec-
tively. When u
V
u
H
, u
s
is dominated by the vertical component,
i.e., free convection dominates. Alternatively, when u
H
u
V
, u
s
is
dominated by the horizontal component, i.e., forced convection
dominates. When u
V
and u
H
are of similar magnitude, mixed con-
vection occurs.
In our experiment, the wind velocity components above the
evaporating surface (i.e., u
V
andu
H
) are not measureddirectly. Here
we assume that u
H
is some fraction of the horizontal wind speed at
the reference height, u
ref
as follows
u
H
= nu
ref
(12)
wherenis a dimensionless constant (range: 0to1.0) andu
ref
[ms
1
]
is the horizontal wind speed measured at the reference height, e.g.,
2mabove groundlevel. Usingthe standardtheorybasedonthe ver-
tical gradient in air density (see Appendix C for details), we derive
that u
V
can be calculated as
u
V
= ku
V,C
(13)
where k is another dimensionless constant (range: 0) and u
V,C
[ms
1
] is the characteristic speed of air movement in the vertical
direction.
One way of combining u
H
and u
V
to estimate u
s
is
u
s
=

u
V

+u
H

1/
(14)
where is a dimensionless constant (range: 1 to ). For exam-
ple, is equivalent to the assumption that u
s
is the maximum
34 W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143
of the free or forced convection component (McAdams, 1954; Ball
et al., 1988). Alternatively, setting = 1 implies that u
s
is the sum
of free and forced convection components (Adams et al., 1990).
Intermediate values between these extremes change the relative
contributions of u
V
and u
H
to u
s
.
Combining Eqs. (10)(14) the boundary layer thickness is given
by
z =

a
[(ku
V,C
)

+(nu
ref
)

]
1/
L

q
L (15)
Substituting Eq. (15) into Eq. (7), the nal formof the evapora-
tion equation is
E =
M
w
R
w
D
v
T
a
(e
s
(T
s
) e
a
(T
a
))
[(
a
[(ku
V,C
)

+(nu
ref
)

]
1/
L)/
a
]
q
L
(16)
3. Materials and methods
3.1. Field installation
The experimental US Class A Pan (with bird guard) was located
at the BoM eld station at Canberra Airport (Australia, Latitude:
35.3

S, Longitude: 149.2

E, elevation 578m) and was directly


adjacent (5m) to a BoM operational US Class A pan. Once oper-
ational, our experimental pan was replenished with water daily
(at 9am local time) by the BoM duty ofcer following standard
operating procedure. We commenced installation in September
2006 and the pan was operational fromearly 2007 until 20 January
2010.
The experimental pan facility is depicted in Fig. 2. The pan
was equipped with a water-level sensor (see details below) that
enabledus to calculate the mass balance at 5-minintervals. Inaddi-
tion, Pt100 temperature sensors (Type GW2105, dimension 2mm
2.3mm, Degussa, Hanau, Germany) were located on the interior
and exterior pan wall at the four cardinal compass points, and at
three different levels (25, 100 and 175mm from the bottom), to
characterise the thermal dynamics of the pan. Temperature sen-
sors were also placed at the same three levels at the centre of
the pan to record the bulk water temperature. The temperature
of the water surface was measured using an infrared thermome-
ter (Model: M50-1C-06-L, Mikron Instrument Co. Inc., Oakland, NJ,
USA) (Fig. 2b).
The evaporation rate was calculated from water height mea-
surements made using a magnetostrictive linear displacement
transducer (MLDT) (MagneRule Plus, MRU-4001-015, Schawitz
Sensors, Hampton, VA, USA) with a spherical oat. The oat was
installed in a stilling well connected to one side of the pan. After
initial experimentation, and contrary to the manufacturers spec-
ications, we found that the output of the MLDT was sensitive to
variations inambient temperature. Toovercome that limitation, we
attached a proportional-integral-derivative (PID) controlled heater
to the casing of the sensor head to maintain a constant tempera-
ture of 40

C. The resolution of the MLDT for our installation was


10m.
In addition, standard meteorological measurements included
radiation, wind speed, air temperature, air vapour pressure and
atmospheric pressure. All components of the radiation balance
(incomingandoutgoingshort- andlong-waveradiation) weremea-
sured using a Kipp & Zonen CNR 1 Net Radiometer attached to
a swinging-arm (Fig. 2). Most of the time, the swinging arm was
parked to the southeast of the pan with the downward sensors fac-
ing the ground. At 5-min intervals, a motor swung the armover the
centre of the pan where the downward sensors sampled radiation
fromthe water surface for a 20-s period. Forty radiometer readings
are takenineachdirectional swing. (Apaper focusing onthe energy
balanceof panevaporationis inpreparation.) Windspeedwas mea-
sured using a cup anemometer at 2m above ground level (u
2
). In
addition, a mast was installed 5maway fromthe experimental pan
to enable installation of temperature, vapour pressure and atmo-
spheric pressure sensors and a 2D ultrasonic anemometer (Wind
Observer II, Gill Instrument Ltd.). The 2D ultrasonic anemome-
ter was located at the same height as the cup anemometer to
enable us to check the performance of the cup anemometer. Atmo-
spheric pressure was measuredwitha Vaisala Pressure Transmitter
(Model: PTB101B). Air temperature and air vapour pressure were
measured with a Vaisala Humitter (Type 50Y, Vaisala, Helsinki,
Finland). Air temperature, vapour pressure and the water level sen-
sor were all calibrated in the laboratory and periodically checked
on-site after eld installation. All analog sensors signals were con-
vertedviaa16-bit analog-to-digital converter, averagedover 5-min
intervals and stored in a single-board computer.
3.2. Data sampling
Short-termoscillations in the water level were apparent in the
5-min data. To avoid the high-frequency noise, all water level data
were aggregated, and the resulting evaporation rate calculated, at
half-hourlyintervals. The change inwater level due todailyrelling
(at 9amlocal time) was takenintoaccount. All other meteorological
data were resampled to half-hourly intervals.
From the database, we identied 160 days of elite data that
had no missing half-hourly totals and included 40 (rainless) days
each in spring, summer, autumn, and winter, respectively. The nal
elite database contains (= 160 days 48 samples per day) a total
of 7680 half-hourly measurements.
3.3. Parameter estimation and validation
We split the data fromthe 160 days into two databases. The rst
subset (60 days, including 15 days in each of the four seasons) was
used to estimate the model parameters. The second subset (100
days, including 25 days in each of the four seasons) was used to
validate the model.
While the optimumvalue of (Eq. (16)) was >2, the effect onthe
t to the data was not strong, and we chose =2 based on a vecto-
rial combination assumption (Adams et al., 1990). The parameters
(k, n, q) were estimated using 2880 half-hourly measurements (=
60 days 48 samples per day) using a least squares optimisation
approach. To do that we initially assumed values of the parame-
ters and then computed pan evaporation E
pan
for that parameter
combination. Many possible combinations were tested using an
automated computer algorithm and the parameter combination
with the lowest root mean square error (RMSE) was selected.
4. Results
4.1. Meteorological data, calibration and validation
The wind speed measured by the cup anemometer (u
2
, also
known as the (horizontal) wind speed hereafter) was compared
to measurements by the 2D ultrasonic anemometer. We used half-
hourly data fromall 160 days (7680 samples) (Fig. 3). At lowwind
speeds (<2ms
1
), the cup anenometer estimates were slightly
lower. However, at higher wind speeds, when most of the evap-
oration occurs, the measurements were consistent.
Half-hourlymeteorological data for eight typical days areshown
in Fig. 4. The characteristic diurnal cycle (T
s
, T
a
) is clearly evident
with E
pan
usually a maximumin the mid-afternoon. The horizontal
wind speed (u
2
) also shows a pronounced diurnal cycle in all 160
days in the database. On a daily basis, E
pan
varies from a high of
1015mmd
1
in summer to a lowof 2mmd
1
in winter.
W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143 35
Fig. 2. Experimental pan installation at Canberra Airport BoMstation. (a) The US Class A pan is 1.21min diameter (4ft) and 0.254min height (10in.), equipped with instru-
ments measuring (short- and long-wave) radiation, wind, temperature (water surface, bulk water, inner and outer pan wall) and water level. (b) The infrared thermometer
installation (inside the white PVC pipe facing the water surface in the pan) measuring long-wave radiation emitted by the water surface.
Fig. 3. Comparison of wind speed measurements from the cup anemometer and the 2D ultrasonic anemometer over half-hourly intervals at the experimental pan (7680
data points, y =1.07x 0.44, R
2
=0.97, RMSE=0.42ms
1
).
The least squares estimates of the parameters are k =0.20,
n=0.10, and q=0.64 (half-hourly samples: 2880, regression
of estimated versus observed E
pan
: slope =0.86, inter-
cept=6.710
6
mms
1
, R
2
=0.82, RMSE=3.110
5
mms
1
).
The estimate for q is close to the previously noted value of 0.5
for laminar ow. The estimate of n is also sensible in that it makes
the horizontal component of wind velocity immediately above the
surface only 10% of the wind speed at 2mabove ground level. With
36 W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143
Fig. 4. Pan evaporation Epan (observed versus estimated), conductance gm,v (observed versus estimated per Eq. (17)), vapour pressures (es(Ts), ea(Ta)), temperatures (Ts, Ta),
u
2
(wind speed at 2mabove ground level) and atmospheric pressure Pa for the experimental pan for diurnal cycles in: spring (a and b), summer (c and d), autumn (e and f),
and winter (g and h).
W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143 37
Fig. 4. (Continued).
38 W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143
Fig. 5. Estimated versus observed Epan by integrating half-hourly results to a daily basis for 100 days: (a) Ts available (y =1.04x +0.39, R
2
=0.98, RMSE=0.51mmd
1
); (b)
assume Ts =Ta (y =1.56x +1.43, R
2
=0.91, RMSE=3.06mmd
1
).
Fig. 6. Temperature depression of water surface at half-hourly intervals for the experimental pan (7680 data points). (a) Frequency distribution of Ts Ta (binsize =0.1K).
(b) Frequency distribution of Ts Tw (binsize =0.1K). (c) Observed Epan versus Ts Ta.
those results, the pan evaporation E
pan
(semi-empirical equation
fromEq. (16)) is
E
pan
=
M
w
R
w
D
v
T
a
(e
s
(T
s
) e
a
(T
a
))

(0.20u
V,C
)
2
+(0.10u
2
)
2
L

/
a

0.64
L
(17)
Note that the numerical value of L is 1.21m, i.e., the diameter of
a US Class A pan.
We subsequently used Eq. (17) to estimate E
pan
for the remain-
ing (4800) half-hourly totals. (The resulting half-hourly totals are
compared with measurements over eight typical days in Fig. 4.)
The half-hourly totals were then summed into 100 daily totals and
compared with measurements (Fig. 5a). The model explained 98%
of the observed variance in the daily E
pan
with an overall RMSE of
0.51mmd
1
giving us condence in the parameterisation.
4.2. Water surface temperature and evaporative cooling
The model parameters and results to date were derived using
our (infrared) measurement of the water surface temperature T
s
.
W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143 39
Fig. 7. Half-hourlyboundarylayer thickness z (per Eq. (15); =2, k =0.20, n=0.10,
q =0.64) versus u
2
(wind speed at 2m above ground level) for the experimental
pan (7680 data points).
Fig. 8. Half-hourly aerodynamic function fv (per Eq. (17)) versus u
2
(wind speed at
2 mabove ground level) for the experimental pan (7680 data points).
In practical applications, and especially for evaluating the histori-
cal records, T
s
is unknown. Further, inspection of Fig. 4 shows that
temperature of the water surface T
s
can be quite different fromthat
of the air T
a
, as would be expected for a freely evaporating surface.
The following questionarises: canE
pan
be estimated accurately with-
out measurements of T
s
in the absence of radiation measurements? To
answer this, we assumed that the water surface was at the air tem-
perature (i.e., T
s
=T
a
) and accordingly recalculated the saturated
vapour pressure at the surface. The results, using totals from the
100 elite-days, show that this is a bad assumption for a purely
aerodynamic formulation of evaporation (Fig. 5b). In general, the
estimateof dailyE
pan
basedontheassumptionthat T
s
=T
a
was much
larger than the observations at high rates of E
pan
. That result means
that the water surface is cooler than the air when E
pan
is high and
implies evaporative cooling. The same phenomenon is visible in
Fig. 4, where T
s
is substantially lower than T
a
in the mid-afternoon
when E
pan
tends to be highest.
To investigate further, we examined the relationships between
T
s
, T
a
and the bulk water temperature T
w
(taken as the average of
measurements at 25, 100 and 175mm below the water surface).
The bulk water was found to be well mixed (results not shown) but
the evaporating surface was often up to 3K warmer or 5K cooler
than the bulk water, with an overall average of around 2K cooler
than the bulk water (Fig. 6b). We also found that the evaporating
surface could be up to 5K warmer or 11K cooler than the air, and
was on average, 1K cooler than the air over the 160 days (Fig. 6a).
In general, the temperature of the water surface was much lower
than the air when E
pan
was high (Fig. 6c), conrming our earlier
deductions about evaporative cooling. In summary, the magnitude
of the surface cooling due to evaporation was substantial.
4.3. Estimates of boundary layer thickness and aerodynamic
function
Estimates of boundary layer thickness z using the estimated
parameter values are plotted as a function of wind speed at 2m
above ground level (u
2
) in Fig. 7. For u
2
>1ms
1
, the results are
dominated by forced convection with z in the range 14mm.
For u
2
<1ms
1
, z is highly variable within an envelope constraint
imposed by the free convection regime.
The resulting aerodynamic function has been computed using
all available half-hourly data (Fig. 8). The overall features of the
aerodynamic function are consistent with the ideas put forward by
Thomet al. (1981, Figs. 2 and 5). In particular, at low wind speeds
(u
2
<1ms
1
) we see the variation in the aerodynamic function due
to a mixture between free and forced convection. At higher wind
speeds when forced convection dominates, the relation collapses
to be u
2
0.64
, which is nearly a straight line.
4.4. Estimates of pan evaporation without a bird guard
Our experimental pan used the same bird guard as used by the
Australian BoM (Fig. 2). The effect is to reduce the mass transfer
by 7% in comparison to a similar pan without a bird guard (van
Dijk, 1985). For comparative purposes, it is useful to estimate the
impact of the bird guard. On the boundary layer formulation used
here (Fig. 1), the bird guard would affect the evaporation rate by
reducing the wind speed near the evaporating surface and thereby
increasingtheboundarylayer thickness z. Hence, wecanincorpo-
rate that effect by adjusting the numerical value of the n parameter.
Assuming all else is held constant, we nd a decrease in n of close
to 10% will reduce E
pan
by around 7%. In summary, in the absence
of a bird guard, the parameter estimates are k =0.20, n=0.11,
q=0.64.
To compare our adjusted formulation for a pan without a bird
guard with previous research, we plotted the aerodynamic func-
tion f
v
as a function of wind speed (2m above ground level) u
2
for various differences between the water surface temperature T
s
and the air temperature T
a
under typical conditions (P
a
=101.3kPa,
e
a
(T
a
) =1kPa, T
a
=293.15K (20

C)). The resulting envelope of the-


oretical curves (for different T
s
T
a
) (Fig. 9a) is consistent with the
concepts proposed by Thomet al. (1981, Figs. 2 and 5). This enve-
lope covers the majority(but not all) of aerodynamic functions used
in previous studies when free or mixed convection dominates (i.e.,
u
2
<1ms
1
); and has a linear or near-linear form of the classical
wind functions under high wind speeds when forced convection
dominates (Fig. 9b).
5. Discussion and summary
The theoretical formulation derived here is based on the idea
of vapour transport, predominantly by diffusion, fromthe liquid to
vapour phases across a well-dened boundary layer (Fig. 1). This
40 W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143
Fig. 9. Aerodynamic function fv versus u
2
(wind speed at 2m above ground level) for pan evaporation under typical conditions (Pa =101.3kPa, ea(Ta) =1kPa, Ta =293.15K
(20

C)). (a) Our model without a bird guard (per Eq. (16); =2, k =0.20, n=0.11, q=0.64) for various differences between water surface and air temperature; (b) fv from
previous studies.
is well established in laboratory studies (Doe, 1967; Machin, 1964,
1970) and the resulting formulation (Eq. (7) has a direct physical
interpretation. Hence, the utility of this approach rests on whether
the conceptual framework (Fig. 1) is useful.
The assumptionthat =2is basedonthe idea of vector average
of the uxes associated with the two physical processes (Adams
et al., 1990). In contrast, the idea of taking a maximum value of
free or forced convection (McAdams, 1954; Ball et al., 1988) means
having , which might be useful over very short time inter-
vals (e.g., minutes); yet could be less appropriate over longer time
intervals (e.g., half hours) since a mixture of both free and forced
convection is most likely to be the case under outdoor conditions.
Typical wind function approaches of the formf(u) =a +bu implic-
itly assume the free convection component (a) to be a constant
(Adams et al., 1990). This is similar to setting =1 and q=1 in
Eq. (16).
Once was set =2, the remaining parameters were estimated
using the measurement database consisting of half-hourly data for
60days. Theresultingparameters (k =0.20, n=0.10, q=0.64) were
closetobroadexpectations. Theestimatedku
V,C
range(00.6ms
1
)
is within our expected range (<1ms
1
), suggesting that our result
(k =0.20) is reasonable. Based on the logarithmic wind prole
assumption (von Karman constant of 0.41; zero plane displace-
ment of 0; roughness length for open water surface ranging from
0.0020.006m, per MonteithandUnsworth(2008, Table 16.1)), the
wind speed at 5cm above the evaporating surface (and at same
level as the rim of the pan) would range from 0.36u
2
to 0.47u
2
(i.e., n0.4). Our result (n=0.10) implies that the horizontal com-
ponent of wind velocity u
H
(see Section 2.4) probably locates at
311mm above the evaporating surface. For laminar ow over a
at surface we expect q to be 0.5 (Schlichting, 1960) and up to
0.8 for turbulent ow (Schlichting, 1960; Grace, 1977; Brenner
and Jarvis, 1995). Our result (q=0.64) implies that E
pan
u
s
0.64
and is close to E
pan
u
2
0.54
obtained by Penman (1948) using a
sunken pan, and also within the general range u
ref
0.30.8
found
for heat transfer of leaves (Brenner and Jarvis, 1995; Stokes et al.,
2006). When free convection dominates, Eq. (17) (associated with
Eq. (C.3) in Appendix C) becomes E
pan
(/)
0.32
and is con-
sistent with the ndings fromfree convection experiments, which
suggest E
pan
(/)
1/3
(Bower and Saylor, 2009, Eq. 30).
With the parameter estimates (k =0.20, n=0.10, q=0.64) we
were able to calculate the boundary layer thickness z. When the
wind speed is high (u
2
>1ms
1
), forced convection dominates, and
the air above the pan is quickly replaced by the surrounding air.
Under those conditions, z is predominantly a function of wind
speed. However, under low wind speeds (u
2
<1ms
1
), free con-
vection dominates, and other factors also play important roles in
determiningz. Inparticular, thespatial gradient of temperaturein
the vertical direction (a surrogate for density difference) becomes
important at lowwind speeds.
Our theory (Eq. (17)) and subsequent results (Fig. 8) show that
a unique wind function does not exist. It also suggests a depen-
dence on atmospheric pressure. Thus it would be interesting to test
whether the formulation presented here (Fig. 9a) could be applica-
ble under a wider set of conditions, such as at high altitude sites
(Blaney, 1960; Giambelluca and Nullet, 1992) where the atmo-
spheric pressure is substantially reduced. Our formulationassumes
that the water surface is always close to the top of the pan and
thereby avoids the shelter effect (Chu et al., 2010). This is most
easily achieved in operational settings by relling the pan each
day.
Measuring the water surface temperature T
s
(in absence of
radiation measurements) proved to be important for accurately
estimating E
pan
using the aerodynamic approach when the evapo-
ration rate was high and the associated evaporative cooling of the
surface was at a maximum (Figs. 5 and 6). On average, the evap-
orating water surface was cooler than both the air and the bulk
water (Fig. 6a andb), as foundpreviously inlaboratory experiments
(Hisatake et al., 1995, Fig. 5). This implies that the very thin layer of
surface water, is, on average, a net absorber of sensible heat from
both the air above it and from the bulk water below it. We were
surprised by the magnitude of this cooling effect.
Acknowledgements
We thank the BoMstaff; Tony McCarthy, Ross Heareld, Kirsty
Rhind, Nigel Smedley, David Pottage, Neil McArthur and Kenn Batt
for their help in maintaining our experimental pan at the Canberra
Airport and Liang Li for his contribution in setting up the exper-
imental pan database. We acknowledge the Australian Research
Council (ARC) for the nancial support of this study through the
grant DP0879763. We are grateful to two anonymous reviewers
for helpful comments.
W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143 41
Fig. A.1. Diffusion coefcient Dv versus air temperature Ta at different values of
atmospheric pressure Pa.
Appendix A. Diffusion coefcient of water vapour
The diffusion coefcient of water vapour D
v
is calculated based
on Pruppacher and Klett (1997):
D
v
= 2.11

T
a
273.15

1.94

P
o
P
a

10
5
[m
2
s
1
] (A.1)
where T
a
[K] is the air temperature, P
a
[Pa] is the atmospheric pres-
sure, P
o
[Pa] is the atmospheric pressure at the mean sea level
(101.325kPa). Fig. A.1 shows the change of D
v
with T
a
and P
a
using
Eq. (A.1).
Appendix B. Dynamic viscosity of air
The dynamic viscosity of air
a
(assumed dry air for simplicity)
is calculated based on Jacobson (2005):

a
= 1.8325

416.16
T
a
+120

T
a
296.16

1.5
10
5
[kgm
1
s
1
] (B.1)
Although
a
is based on dry air instead of moist air, the differ-
ence is small (Maxwell, 1866; Kestin and Whitelaw, 1964). Fig. B.1
illustrates the change of
a
with T
a
using Eq. (B.1).
Appendix C. Derivation of the speed of air in the vertical
direction above evaporating surface
The vertical circulation of air above the evaporating sur-
face is determined by the air density difference (Schlichting,
1960; Incropera and DeWitt, 1990; Holman, 2002; Monteith
and Unsworth, 2008), which results from temperature gradients,
vapour concentrationgradients, or acombinationof both(Monteith
and Unsworth, 2008).
In principle, the speed of air in the vertical direction can be
derived from Reynolds and Grashof numbers using dimensional
analysis. An equivalent Reynolds number in the vertical direction
(Re
V
) is the ratio of inertial forces (in the vertical direction) to vis-
cous forces, i.e.,
Re
V
=

a
u
V
L

a
(C.1)
where
a
[kgm
3
] is the density of air, u
V
[ms
1
] is the speed of
air in the vertical direction, L [m] is the characteristic length of the
Fig. B.1. Dynamic viscosity of air a versus air temperature Ta.
evaporating surface and
a
[kgm
1
s
1
] is the dynamic viscosity of
air (calculation of
a
is given in Appendix B).
The Grashof number (Gr) is equal to buoyancy forces times iner-
tia forces divided by the square of viscous forces. Since the origin
of Gr is in heat transfer studies (Karwe and Deo, 2003), it is com-
monly calculated based on a spatial temperature difference. Here,
we calculate Gr by using the spatial density difference. The purpose
is to take into account both temperature and concentration (water
vapour and dry air) differences between the evaporating surface
and the reference height, i.e.,
Gr =

2
a
g(
a
/
a
)L
3

2
a
(C.2)
where g [ms
2
] is the gravitational acceleration,
a
[kgm
3
] is the
average air density between the evaporating surface and the refer-
ence height and
a
[kgm
3
] is the spatial difference inthe density
of air between the evaporating surface and the reference height.
Assuming constant atmospheric pressure between the evapo-
rating surface and the reference height, u
V
[ms
1
] can be derived
fromEqs. (C.1) and (C.2) as follows,
Re
2
V
Gr, Re
V

Gr,

a
u
V
L

2
a
g(
a
/
a
)L
3

2
a
,
u
V

a
L, u
V
= k

a
L, u
V
= ku
V,C
(C.3)
where k is a dimensionless constant (range: 0) and u
V,C
[ms
1
]

g(
a
/
a
)L

is the characteristic speed of air in the verti-


cal direction. We calculate
a
and
a
as

a
=
1
R

(P
a
e
a
(T
a
))M
a
+e
a
(T
a
)M
w
T
a

(P
a
e
s
(T
s
))M
a
+e
s
(T
s
)M
w
T
s

(C.4)
42 W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143
and

a
=
1
2R

(P
a
e
a
(T
a
))M
a
+e
a
(T
a
)M
w
T
a
+
(P
a
e
s
(T
s
))M
a
+e
s
(T
s
)M
w
T
s

(C.5)
respectively, where R [J mol
1
K
1
] is the ideal gas constant, T
s
[K]
is the water surface temperature, T
a
[K] is the air temperature, P
a
[Pa] is the atmospheric pressure, e
a
(T
a
) [Pa] is the air vapour pres-
sure, e
s
(T
s
) [Pa] is the vapour pressure at the evaporating surface,
M
a
[kgmol
1
] is the molecular mass of air and M
w
[kgmol
1
] is the
molecular mass of water. We set
a
=0 for conditions when the
air density at the evaporating surface is greater than that at the ref-
erence height, i.e., minimal buoyancy forces; subsequently u
V,C
=0
and thus u
V
=0 in Eq. (C.3).
Appendix D. Equivalent scheme in plant physiology
Here, we showthe relation between our theoretical framework
and the scheme widely used in plant physiology.
For vapour transfer from inside a leaf to the atmosphere, z is
the combinationof the equivalent thickness of the parallel stomatal
pores z
l
[m] and the aerodynamic boundary layer thickness z
a
[m]. Therefore, Eq. (6) becomes
J =
1
R
D
v
T
a
(e
s
(T
s
) e
a
(T
a
))
z
l
+z
a
(D.1)
Based on the ideal gas law,
RT
a
= P
a
V
m
(D.2)
where V
m
[m
3
mol
1
] is the molar volume of moist air. Substituting
Eq. (D.2) into Eq. ((D.1), we have
J =
(e
s
(T
s
)/P
a
) (e
a
(T
a
)/P
a
)
(V
m
z
l
/D
v
) +(V
m
z
a
/D
v
)
(D.3)
By replacing e
s
(T
s
)/P
a
, e
a
(T
a
)/P
a
, V
m
z
l
/D
v
and V
m
z
a
/D
v
with
w
s
, w
a
, r
m,l
and r
m,a
respectively, we obtain
J =
w
s
w
a
r
m,l
+r
m,a
(D.4)
where w
s
is the mole fraction of water vapour inside the leaf, w
a
is
themolefractionof water vapour intheair (at thereferenceheight),
r
m,l
[m
2
s mol
1
] is the leaf resistance, and r
m,a
[m
2
s mol
1
] is the
aerodynamic resistance. Eq. (D.4) is identical to Cowan (1977, Eq.
95).
References
Adams, E.E., Cosler, D.J., Helfrich, K.R., 1990. Evaporation from heated water bod-
ies: predicting combined forced plus free convection. Water Resour. Res. 26,
425435, doi:10.1029/WR026i003p00425.
Ball, M.C., Cowan, I.R., Farquhar, G.D., 1988. Maintenance of leaf temperature and
the optimisation of carbon gain in relation to water loss in a tropical mangrove
forest. Aust. J. Plant Physiol. 15, 263276, doi:10.1071/PP9880263.
Blaney, H.F., 1960. Evaporation from water surfaces in mountain areas of western
United States. Hydrol. Sci. J. 5, 2737, doi:10.1080/02626666009493161.
Brenner, A.J., Jarvis, P.G., 1995. A heated leaf replica technique for determination of
leaf boundary-layer conductance in the eld. Agric. For. Meteorol. 72, 261275,
doi:10.1016/0168-1923(94)02160-L.
Bower, S.M., Saylor, J.R., 2009. A study of the SherwoodRayleigh relation for water
undergoing natural convection-driven evaporation. Int. J. Heat Mass Trans. 52,
30553063, doi:10.1016/j.ijheatmasstransfer.2009.01.034.
Brutsaert, W., 1982. Evaporation into the Atmosphere: Theory, History, and Appli-
cations. D. Reidel Publishing Company, Boston, MA.
Brutsaert, W., Parlange, M.B., 1998. Hydrologic cycle explains the evaporation para-
dox. Nature 396, 30, doi:10.1038/23845.
Chu, C.R., Li, M.H., Chen, Y.Y., Kuo, Y.H., 2010. A wind tunnel experiment on
the evaporation rate of Class A evaporation pan. J. Hydrol. 381, 221224,
doi:10.1016/j.jhydrol.2009.11.044.
Cowan, I.R., 1977. Stomatal behaviour and environment. Adv. Bot. Res. 4, 117228,
doi:10.1016/S0065-2296(08)60370-5.
Doe, P.E., 1967. Measurement of a mass transfer boundary layer. Nature 216,
11011103, doi:10.1038/2161101a0.
Fick, A., 1995. On liquid diffusion (Reprinted fromFick, A., 1855. On liquid diffusion.
The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Sci-
ence 10, 3039). J. Membr. Sci. 100, 3338, doi:10.1016/0376-7388(94)00230-V.
Giambelluca, T.W., Nullet, D., 1992. Evaporation at high elevations in Hawaii. J.
Hydrol. 136, 219235, doi:10.1016/0022-1694(92)90012-K.
Giblett, M.A., 1921. Some problems connectedwithevaporationfromlarge expanses
of water. Proc. R. Soc. Lond. Ser. A 99, 472490.
Gilliland, E.R., 1934. Diffusion coefcients in gaseous systems. Ind. Eng. Chem. 26,
681685, doi:10.1021/ie50294a020.
Grace, J., 1977. Plant Response to Wind. Academic Press, London.
Hisatake, K., Fukuda, M., Kimura, J., Maeda, M., Fukuda, Y., 1995. Experimental and
theoretical studyof evaporationof water inavessel. J. Appl. Phys. 77, 66646674,
doi:10.1063/1.359079.
Hisatake, K., Tanaka, S., Aizawa, Y., 1993. Evaporation rate of water in a vessel. J.
Appl. Phys. 73, 73957401, doi:10.1063/1.354031.
Holman, J.P., 2002. Heat Transfer, 9th ed. McGraw-Hill, NewYork.
Incropera, F.P., DeWitt, D.P., 1990. Fundamentals of Heat and Mass Transfer, 3rd ed.
John Wiley, NewYork.
Jacobs, A.F.G., Heusinkveld, B.G., Lucassen, D.C., 1998. Temperature variation
in a class A evaporation pan. J. Hydrol. 206, 7583, doi:10.1016/S0022-
1694(98)00087-0.
Jacobson, M.Z., 2005. Fundamentals of Atmospheric Modeling, 2nd ed. Cambridge
University Press, NewYork.
Johnson, F., Sharma, A., 2010. A comparison of Australian open water body
evaporation trends for current and future climates estimated fromclass A evap-
oration pans and general circulation models. J. Hydrometeorol. 11, 105121,
doi:10.1175/2009jhm1158.1.
Karwe, M.V., Deo, I., 2003. Grashof number. In: Heldman, D.R. (Ed.), Encyclope-
dia of Agricultural, Food, and Biological Engineering. Taylor & Francis, pp.
454456.
Kestin, J., Whitelaw, J.H., 1964. The viscosity of dry and humid air. Int. J. Heat Mass
Trans. 7, 12451255, doi:10.1016/0017-9310(64)90066-3.
Kohler, M.A., Noredenson, T.J., Fox, W.E., 1955. Evaporation fromPans and Lakes. US
Weather Bureau Research Paper 38. US Weather Bureau, Washington, D.C.
Leighly, J., 1937. A note on evaporation. Ecology 18, 180198.
Linacre, E.T., 1994. Estimating U.S. Class A pan evaporation from few climate data.
Water Int. 19, 514, doi:10.1080/02508069408686189.
Machin, J., 1964. Evaporation of water from Helix Aspersa. II. Measurement of air
owand diffusion of water vapour. J. Exp. Biol. 41, 771781.
Machin, J., 1970. The study of evaporation from small surfaces by the
direct measurement of water vapour pressure gradients. J. Exp. Biol. 53,
753762.
Martinez, J.M.M., Alvarez, V.M., Gonzalez-Real, M.M., Baille, A., 2006. A simulation
model for predictinghourlypanevaporationfrommeteorological data. J. Hydrol.
318, 250261, doi:10.1016/j.jhydrol.2005.06.016.
Maxwell, J.C., 1866. The Bakerian Lecture: on the viscosity or internal friction of air
and other gases. Philos. T. R. Soc. Lond. 156, 249268.
McAdams, W.H., 1954. Heat Transmission. McGraw-Hill, NewYork.
Monteith, J.L., 1965. Evaporation and environment. Symp. Soc. Exp. Biol. 19,
205234.
Monteith, J.L., Unsworth, M.H., 2008. Principles of Environmental Physics, 3rd ed.
Elsevier Academic Press, San Diego.
Penman, H.L., 1948. Natural evaporation fromopen water, bare soil and grass. Proc.
R. Soc. Lond. Ser. A 193, 120145.
Peterson, T.C., Golubev, V.S., Groisman, P.Y., 1995. Evaporation losing its strength.
Nature 377, 687688, doi:10.1038/377687b0.
Pruppacher, H.R., Klett, J.D., 1997. Microphysics of Clouds and Precipitation, second
rev & enl. ed. Kluwer Academic Publishers, Dordrecht, Netherlands.
Rayner, D.P., 2007. Windrunchanges: the dominant factor affectingpanevaporation
trends in Australia. J. Clim. 20, 33793394.
Riley, J.J., 1966. Heat balance of Class A evaporation pan. Water Resour. Res. 2,
223226, doi:10.1029/WR002i002p00223.
Roderick, M.L., Farquhar, G.D., 2002. The cause of decreased pan evaporation over
the past 50 years. Science 298, 14101411, doi:10.1126/science.1075390-a.
Roderick, M.L., Hobbins, M.T., Farquhar, G.D., 2009a. Pan evaporation trends and
the terrestrial water balance. I. Principles and observations. Geogr. Compass 3,
746760, doi:10.1111/j.1749-8198.2008.00213.x.
Roderick, M.L., Hobbins, M.T., Farquhar, G.D., 2009b. Pan evaporation trends and the
terrestrial water balance. II. Energy balance and interpretation. Geogr. Compass
3, 761780, doi:10.1111/j.1749-8198.2008.00214.x.
Roderick, M.L., Rotstayn, L.D., Farquhar, G.D., Hobbins, M.T., 2007. On the
attribution of changing pan evaporation. Geophys. Res. Lett. 34, L17403,
doi:10.1029/2007GL031166.
Rohsenow, W.M., Hartnett, J.P., Ganic, E.N., 1985. Handbook of Heat Transfer Fun-
damentals, revised ed. McGraw-Hill, NewYork.
Rotstayn, L.D., Roderick, M.L., Farquhar, G.D., 2006. Asimple pan-evaporation model
for analysis of climate simulations: evaluationover Australia. Geophys. Res. Lett.
33, L17715, doi:10.1029/2006GL027114.
Schlichting, H., 1960. Boundary Layer Theory, 4th ed. McGraw-Hill, NewYork.
Shuttleworth, W.J., Serrat-Capdevila, A., Roderick, M.L., Scott, R.L., 2009. On the the-
ory relating changes in area-average and pan evaporation, Quart. J. R. Meteorol.
Soc. 135, 12301247, doi:10.1002/qj.434.
Stanhill, G., 1976. The CIMOInternational Evaporimeter Comparisons. WMO Report
No. 449. WMO, Geneva, Switzerland.
W.H. Limet al. / Agricultural and Forest Meteorology 152 (2012) 3143 43
Stanhill, G., 2002. Is the Class A evaporation pan still the most practical and
accurate meteorological method for determining irrigation water require-
ments? Agric. For. Meteorol. 112, 233236, doi:10.1016/S0168-1923(02)
00132-6.
Stokes, V.J., Morecroft, M.D., Morison, J.I.L., 2006. Boundary layer conductance for
contrasting leaf shapes in a deciduous broadleaved forest canopy. Agric. For.
Meteorol. 139, 4054, doi:10.1016/j.agrformet.2006.05.011.
Thom, A.S., Thony, J.L., Vauclin, M., 1981. On the proper employment of evaporation
pans andatmometers inestimating potential transpiration. Quart. J. R. Meteorol.
Soc. 107, 711736, doi:10.1002/qj.49710745316.
van Dijk, M.H., 1985. Reduction in evaporation due to the bird screen used in the
Australian class A pan evaporation network. Aust. Meteorol. Mag. 33, 181183.
Ward, C.A., Stanga, D., 2001. Interfacial conditions during evaporation or condensa-
tion of water. Phys. Rev. E 64, 051509, doi:10.1103/PhysRevE.64.051509.

Vous aimerez peut-être aussi