You are on page 1of 11

agricultural water management 96 (2009) 711–721

available at www.sciencedirect.com

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

A discussion on and alternative to the Penman–Monteith


equation

Peter Widmoser
Hydrology and Water Resources Department, Ecology Centre, University of Kiel, D-24098 Kiel, Germany

article info abstract

Article history: The Penman–Monteith (PMe) equation that estimates evaporation from leaf/canopy sur-
Received 29 May 2008 faces is based on a few approximations. Several authors discussed ensuing errors and
Accepted 19 October 2008 suggested improvements. This paper reminds those discussions which ended in the early
Published on line 5 December 2008 nineties. It compares linearized PMe with non-linear iterative solutions and illustrates
resulting deviations. It differentiates between deviations for daily and hourly evaporation
Keywords: rate estimates. The latter are found to be higher. It also demonstrates deviations obtained at
Evaporation two different altitudes above sea level. Considering present tendencies to refine evaporation
Energy balance estimates for practical purposes and making use of easily available methods for solving non-
Penman–Monteith equation linear equations this paper offers a new method to estimate evaporation.
Surface temperature In a first step, a simple algebraic term, the surface temperature control sum, is intro-
duced to find approximate differences between air and evaporating surface (leaf, canopy)
temperatures. It suggests to concentrate research on the rs/ra ratio. A new formula is derived
for estimating leaf/canopy surface temperatures for non-water stressed plants.
In a second step, the estimates of temperature differences are used to calculate evapora-
tion estimates. This two-step approximation leads to appreciably smaller errors as com-
pared to the PMe-solution over the full range of input parameters of agro-meteorological
relevance. It is, however, less accurate than some of methods proposed in literature. The
method is meant for practical application in agricultural water management.
# 2008 Elsevier B.V. All rights reserved.

basis, too (Alexandria and Kerkides, 2003; Lecina et al., 2003;


1. Introduction Allen et al., 2006; Perez et al., 2006), as well as a general
framework to interpret leaf or canopy energy and water
One method of estimating vapor fluxes from plant leaves or budgets even on an instantaneous basis. It is generally agreed
canopies is the Penman–Monteith equation (PMe). It offers the that the PMe gives acceptable evaporation estimates for
simultaneous solution of surface energy balance and of heat practical purposes (e.g., Batchelor, 1984; Allen, 1986; Kustas
and mass (i.e., water vapor) transport, introducing some et al., 1996; Klaasen, 2001; Rana, 2001; Sumner and Jacobs,
approximations and thereby resulting in an explicit equation 2005; Xu and Chen, 2005; Perez et al., 2006).
that avoids the non-linearity of the equation system. The most Errors introduced by the approximations were studied by
relevant approximation replaces the temperature of the comparing PMe-solutions with field measurements (e.g.,
evaporating surface – neither easy to define nor to measure Grant, 1975; Gavin and Agnew, 2004) or with mathematically
(Alves and Pereira, 2000) – by air temperature (Penman, 1948). more exact solutions, e.g., by developing the non-linear part of
It is usually applied on a time scale of days as recommended by the equation into a higher order Taylor series (e.g., Thom, 1972;
Penman. There are, however, tendencies to use it on an hourly Paw and Gao, 1988; Milly, 1991; Paw, 1992) or by iterative

E-mail address: widmoser@hydrology.uni-kiel.de.


0378-3774/$ – see front matter # 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.agwat.2008.10.003
712 agricultural water management 96 (2009) 711–721

methods (e.g., Brown and Rosenberg, 1973; Tracy et al., 1984).


Nomenclature
Those solutions point to possible errors of up to 30% (Paw and
Gao, 1988). Comparisons with field data usually suffer from
A (W m2) external energy available for evaporation
uncertainties of input variables, especially the stomatal or
as difference between net radiation flux
canopy resistance.
density (Rn) and soil and/or vegetation
Nowadays, it is trivial to solve non-linear equations to any
heat flux density (G); A = Rn–G
realistically desired accuracy by automated iterative approx-
C (W m2) convective (sensible) heat flux density
imation techniques, also widely available and even installed in
cp (J kg1 K1) mass heat capacity of dry air under
some programmable pocket calculators. However, in practice,
constant atmospheric pressure, ca
this obvious solution does not appear to be appreciated. There
1010 J kg1 K1
3 1 may be two arguments against its use: (1) some of the
cv (J m K ) volumetric heat capacity of (moist) air at
evaporation controlling input variables are not known to an
constant atmospheric pressure, depend-
accuracy higher than the PMe-deviations referred to, and (2)
ing on Ta, Ts, ea and pa (ca 1200 J m3 K1)
numerical energy balance solutions give no straightforward
dT (8C) difference between air and (leaf/canopy)
insight into the evaporation process, which in fact is one of the
surface temperature
merits of the PMe.
dE (mm d1) absolute deviation as difference PMe
Developments over the last decade in measurement
minus IT-solution
techniques (e.g., eddy covariance, telemetric temperature
dP relative percentage deviation as 100 dE/IT-
observations) and demands on evaporation estimates (Allen
solution
et al., 1998; e.g., hourly estimates of evaporation; irrigation
E* (kg m2 s1) evaporation mass flux density
scheduling) make it appear appropriate to revive the topic.
E (mm d1) 86:4  106 rw 1 E
0 1 This paper recalls in its first part the theoretical back-
E (mm d ) evaporation estimated
ground, compares PMe-approximations with solutions
e (Pa) air vapor pressure
derived from a mathematical iteration technique and specifies
IT solution for surface temperature by itera-
to some extent the conditions leading to deviations, found to
tion as used in this paper
go up to around 40% and +9%. It also shows that errors
p (Pa) atmospheric pressure
increase with reduced, e.g., hourly vs. daily, evaporation
r (s m1) resistance
periods. In its second part, the paper offers an alternative
Rn (W m2) net radiation flux density at evaporating
approximate two-step concept that avoids extreme deviations
surface
and can be handled with pocket calculators featured with
rh relative air humidity
solver routines.
rho correlation coefficient
The term evaporation will be used in this paper synonym to
r1 1 + rs/ra
evapotranspiration or transpiration, considering a comment
SD (mm d1) standard deviation
by Monteith (1985).
T (8C) temperature
VPD (Pa) vapor pressure deficit

2. Steady-state evaporation
Greek letters
2.1. Theoretical background
1
D (Pa K ) slope of the saturation vapor pressure curve;
depends on surface temperature Ts The following explanations mainly follow Monteith (1965),
g (Pa K1) psychrometric coefficient, see Eq. (7); depends and Monteith and Unsworth (1990). They refer to an
on surface temperature evaporating leaf or canopy surface. Energy conservation
l (J kg1) latent heat of evaporation (2454 MJ kg1 at requires that across the specified surface
20 8C), depends on surface temperature Ts A  lE þ C (1)
r (kg m3) density
S (8C) surface temperature control sum, Eq. (9b) with A = (Rn  G) available external energy; Rn = net radiation
flux density; G = heat flux density of soil and in the case of
estimating evaporation at canopy level also from plant cano-
Indices pies; E* = evaporation rate = mass flux density of vapor trans-
ferred; l = latent heat of vaporization, varying with Ts (see
a air Eq. (8)); C = convective (sensible) heat flux density. For a com-
H heat plete list of abbreviations and units see end of paper.
s surface The surface will lose mass by evaporation proportional to
sat saturated latent heat transfer
v volumetric
V vapor cv ½esat ðTs Þ  ea 
lE ¼ (2)
w liquid water grV
dash0 approximated with Ts = temperature of evaporating surface; esat = saturation
vapor pressure of the air; ea = actual vapor pressure; cv
agricultural water management 96 (2009) 711–721 713

= rcp = volumetric heat capacity of (moist) air at constant


atmospheric pressure, depending on air temperature Ta, air
vapor pressure ea and atmospheric pressure p (values for cv
vary between approximately 1.14–1.28 kJ m3 K1); cp = mass
heat capacity of dry air under constant atmospheric pressure
(ca. 1.01 kJ kg1 K1); r = density of (moist) air depending on air
temperature, vapor pressure of the air and total air pressure p
(see Eq. (6)); g = psychrometric coefficient which varies with Ts,
cp and atmospheric pressure p (see Eq. (7)); rV = mass transfer
resistance for water vapor.
Heat flux is given by

cv ðTs  Ta Þ
C¼ (3)
rH

with rH = resistance to sensible heat transfer.


Inserting (2) and (3) into (1), replacing rH by ra (aerodynamic
Fig. 1 – Illustration of Eqs. (4): curve 2–3–4 = esat(T);
resistance) as common in literature and rV by ra + rs, following
Line 1–2 = slope with Sr1g; point 1 represents Ta and ea of
the suggestion of Monteith and Unsworth (1990) (rs = stomatal
the atmosphere. For more details see text.
or canopy resistance) gives the energy balance

cv esat ðTs Þ  ea cv
¼ A  ðTs  Ta Þ (4) (3) Knowing Ts and inserting it into Eq. (2) or Eqs. (1) and (3)
rs þ ra g ra
gives the evaporation lE.
Eq. (4) is fundamental for the following.
Iteration was performed using a Mathcad1 routine.
2.2. Considerations on the iteration solution The energy balance as formulated in Eq. (4) can be interpreted
geometrically extending a suggestion by Monteith and Uns-
Ta, ea and A in Eq. (4) can be measured or estimated. The worth (1990) (see there Fig. 11.12). Fig. 1 shows how to find Ts:
resistances ra and rs usually have to be estimated (for more Line 1–2 with slope r1g passes through the starting point 1 with
information see, e.g., Jones, 1992), cv usually is considered Ta and ea given. This line intercepts the saturation curve at 2,
constant (e.g., 1.2 kJ m3 K1). On the other side, Ts as well as which represents the adiabatic solution with Tw (=wet bulb
g, a function of Ts, are unknowns and esat(Ts) can be estimated temperature; Tw  Ta ). The shift from 2 to 3 (drawing a parallel to
by line 1–2 at distance þAra =cv ) is caused by external energy
available for evaporation (=A) and results in the solution Ts.
bTs
esat ðTs Þ ¼ a exp (5)
c þ Ts

with the following parameters (Murray, 1967): a = 0.6108 kPa; 3. Comparison of methods
b = 17.27; c = 237.3 8C.
From the above it follows that Eq. (4) is highly non-linear The IT-solutions obtained as described above will be compared
with respect to the unknown Ts. One way to solve it is by with linearized PMe-solutions. The comparison made consists
iteration techniques (abbreviated as ‘‘IT’’ in the following). In in calculating absolute differences dE as PMe minus IT-
this paper, the IT-method considers the following relation- solutions as well as relative differences dP as dE/IT-solutions.
ships in Eq. (4): No description of the PMe will be given here. It is presented
in a number of papers (e.g., Penman, 1948, 1963; Monteith,
(1) Air density r as a function of Ta, ea and p is calculated (Stahl, 1965; Lhomme, 2001) and text books (e.g., Monteith and
1980) with Unsworth, 1990; Jones, 1992). Its approximation technique,
however, is illustrated in Fig. 2. Replacing the saturation curve
  2–3–4 by the tangent 20 –30 –4 (with slope D) shifts point 3 to
273:15 0:378ea
r ¼ 1:293 1 (6) point 30 leading to a difference between the exact value of
273:15 þ Ta p
esat(Ts) and the approximation eś as well as to a difference
(2) The psychrometric coefficient g is given by between the corresponding temperatures Ts and Ts0 . Since g(Ts)
is replaced by g(Ta), the slope-line 1–2 is also an approximation
pc p (Monteith and Unsworth, 1990; Jones, 1992). Furthermore, one
gðTs Þ ¼ (7) may note that the approximations also depend on Ta as well as
0:622lðTs Þ
on r1g(Ta) where r1 = 1 + rs/ra.
and latent heat of vaporization as
3.1. Considerations on the comparison
lðTs Þ ¼ 2501:02  2:369Ts (8)

which is a linear approximation derived from values given in The comparison requires some basic considerations on the
Table A3 in Monteith and Unsworth (1990), valid for Ts between data sets to be compared. This will be set out in the following.
0 and 45 8C. The five input parameters (Ta, ea or rh, A, ra and rs) were chosen
714 agricultural water management 96 (2009) 711–721

included finally some 2500 admitted test runs for daily and
about 2900 for hourly evaporation values.
Furthermore, a test was performed to check the influence
of different atmospheric pressures (101.3 kPa corresponding to
sea level and 70.2 kPa, corresponding to 3000 m above sea
level) on deviations dE (Table 2a). Atmospheric pressure enters
the calculations through the psychrometric coefficient g
(Eq. (7)). The intention was to evaluate errors of the linearized
PMe varying with altitude.

3.2. Some results of the test

The results were used to illustrate deviations for daily


(Table 2a) and hourly (Table 2b) evaporation estimates as
well as for various selected combinations of input parameters
(Figs. 3–7). In addition, Table 2a gives deviations depending on
two selected altitudes.
Referring to daily deviations dE (Table 2a) the average is
0.13 mm d1 with a standard deviation of around
Fig. 2 – Illustration of Penman’s approximation: curve 2–3–
0.23 mm d1 and a correlation coefficient (rho) between PMe
4 = esat(T); Line 1–2 = slope with negative r1g(Ta); point 1
and IT-solutions around 0.995. Thus, the tests prove an
represents Ta and ea. Approximation: curve 2–3–4 is
excellent overall performance of the PMe. This observation
replaced by dotted line 20 –30 –4 (=slope D); point 20 results
does not hold, however, for some maxima deviations. They
from intersection of D-line with g-line with its
0 can reach some +9% and 37% for daily and almost 50% for
(approximated) coordinates Tw and e0w , i.e., the adiabatic
hourly evaporation rates (bold data in Tables 2a and 2b).
temperature and vapor pressure. The shift from the
Deviations change for different time scales due to different
adiabatic point 20 –30 follows from the additional external
inputs (see Table 1).
heat supply A (diabatic case); VPD = esat(T) S ea.
Some examples are also presented in Figs. 3–7.
No detailed comparisons were made between solutions of
this paper with previous literature. At first sight they appear in
within the following range considered to cover the situations line, e.g., with the results of Paw and Gao (1988), who quote
under which agricultural water management is practiced relative accuracy improvement of 25–30% with quadratic and
(5 < Ta < 45 8C; 5 < rh < 98%; 0 < A < 600 W m2; quartic methods respectively over PMe-solutions when com-
1 1
5 < ra < 200 s m ; 70 < rs < 200 s m ; as detailed in Table 1). pared with lysimeter data. They contradict Milly (1991), who
Combinations within those ranges lead to a certain evapora- quotes only negative errors up to 18%.
tion rate. With a combination of only three levels for each one Checking the test results more into detail reveals some
of the five parameters (say: low, medium, high), we will end up tendencies under which high relative errors dP occur. Those
with 243 results, which may include solutions outside the conditions are: (1) negative errors (PMe underestimates) for: (a)
scope of this paper, like negative evaporation (=condensation) low air temperatures and (b) for high air temperatures and low
or extremely high evaporations (say beyond 10 mm d1 or for air humidity if r1 and A are low; (2) positive errors (PMe
shorter time intervals, say more than 0.83 mm h1) usually not overestimates) (a) at any r1 with low A-values for high air
encountered in agro-forestry practice. temperatures and high air humidity and (b) for high r1 and A
Thus, the following upper limits for maximum evaporation with high air temperatures (Table 3). For the terms low and
rates were set: high compare Table 1. No sharp lines can be given.
Emax = 10 mm d1 for daily values along with the input
variables of Table 1, upper line, and for hourly values
Emax = 0.83 mm h1 (=20 mm d1) along with the input vari- 4. The two-step concept for estimating
ables of Table 1, bottom line. All evaporation rates resulting in evaporation
values greater than Emax were excluded from the evaluation
presented in Tables 2 and 4. Results with Ts-values below 0 8C Limitations of the PMe, as shown above and also by previous
were also excluded. Under those restrictions the comparison authors (e.g., Paw and Gao, 1988), increase for hourly

Table 1 – Input parameters selected for test runs with two different criteria for Emax.
Ta (8C) rh (%) A (W m2) ra (s m1) rs (s m1) r1
1
Daily Emax < 10 mm d 5, 10, 15, 20, 5, 18.3, 31.6, 44.9, 0, 60, 120, 180, 5, 10, 50, 50, 70, 100, 200 1.25 up to 41
25, 30, 35, 40 58.1, 71.4, 84.7, 98 240, 300 100, 200

Hourly Emax > 20 mm d1 As above As above 0, 120, 240, 360, As above As above As above
480, 600
agricultural water management 96 (2009) 711–721 715

Table 2a – Daily mean evaporation deviations (med = PM minus IT-solutions) sorted out for E < 10 mm dS1. Statistics for
eight ra/rs-combinations, each one with 384 different Ta-, ea- and A-combinations (see Table 1 upper row) and for two
atmospheric pressures p. n = admitted test runs. Maxima in bold.
p (kPa) ra rs med SD max min max min rho n r1
(s m1) (s m1) (mm d1) (mm d1) (mm d1) (mm d1) (%) (%)

101.3 = 0 m a.s.l. 5 70 0.04 0.19 +0.58 0.39 +8.2 5.6 0.9976 216 15
10 70 0.05 0.17 +0.49 0.37 +7.6 5.8 0.9980 233 8
50 70 0.13 0.19 +0.25 1.13 +6.2 15.5 0.9971 320 2.4
50 100 0.11 0.17 +0.30 0.90 +6.4 17.4 0.9976 337 3
100 100 0.18 0.25 +0.21 1.27 +6.0 24.1 0.9946 364 2
10 200 +0.04 0.16 +0.64 0.16 +8.5 6.2 0.9991 341 21
100 200 0.18 0.27 +0.25 1.05 +6.4 29.8 0.9933 378 3
200 200 0.33 0.45 +0.17 S1.94 +6.0 S36.9 0.9838 380 2
means 0.13 0.23 +0.36 0.90 +6.91 17.66 0.9951 321

70.2 = 3000 m a.s.l 5 70 0.05 0.22 +0.71 0.56 +8.5 5.7 0.9972 166 15
10 70 0.06 0.19 +0.51 0.54 +7.5 6.2 0.9976 188 8
50 70 0.17 0.23 +0.19 1.34 +5.7 13.8 0.9958 284 2.4
50 100 0.14 0.20 +0.27 1.06 +6.0 14.2 0.9968 306 2
100 100 0.20 0.29 +0.19 S1.94 +5.4 23.7 0.9935 349 3
10 200 0.02 0.17 +0.65 0.22 +9.0 5.8 0.9985 305 21
100 200 0.17 0.25 +0.23 1.21 +6.0 24.6 0.9945 366 3
200 200 0.26 0.37 +0.16 1.75 +5.4 S29.5 0.9899 374 2
means 0.13 0.24 +0.36 1.08 +6.69 15.44 0.9954 292

Table 2b – Hourly mean evaporation deviations (med = PM minus IT-solutions) sorted out for E < 20 mm dS1. Statistics for
eight ra/rs-combinations, each one with 384 different Ta-, ea- and A-combinations (see Table 1 lower row). n = admitted test
runs. Maxima in bold.
p (kPa) ra rs med SD max min max min rho n r1
(s m1) (s m1) (mm d1) (mm d1) (mm d1) (mm d1) (%) (%)

101.3 = 0 m a.s.l. 5 70 0.03 0.31 +1.20 0.42 +8.2 6.0 0.9987 307 15
10 70 0.01 0.27 +0.96 0.47 +7.6 10.4 0.9990 324 8
50 70 0.35 0.51 +0.45 2.14 +6.2 26.7 0.9942 371 2.4
50 100 0.35 0.54 +0.50 2.11 +6.4 29.8 0.9933 378 3
100 100 0.65 0.89 +0.35 3.87 +6.0 36.9 0.9938 380 2
10 200 +0.11 0.29 +0.97 0.16 +8.5 6.2 0.9996 384 21
100 200 0.82 1.04 +0.32 4.01 +6.4 44.9 0.9750 384 3
200 200 1.31 1.55 +0.17 S6.33 +6.0 48.9 0.9631 384 2
means 0.45 0.68 +0.62 2.44 +6.91 26.23 0.9895 364

Fig. 4 – Evaporation deviations dE in mm dS1 as a function


Fig. 3 – Evaporation estimated by PMe-(dotted lines) and by of Ta and rh. A = 120 W mS2; ra = 5 s mS1; rs = 200 s mS1;
IT-solutions (full lines); ra = 100 s mS1; rs = 70 s mS1; p = 101.3 kPa. Negative deviations = PMe underestimates.
rh = relative air humidity in %; p = 101.3 kPa. Within shaded area E > 10 mm dS1.
716 agricultural water management 96 (2009) 711–721

Fig. 7 – Percentage evaporation deviations dP as a function


of Ta, and rh. A = 240 W mS2; ra = 100 s mS1; rs = 200 s mS1;
p = 101.3 kPa. Negative deviations = PMe underestimates.
Within light shaded area E > 10 mm dS1, dark shaded area
E > 20 mm dS1.
Fig. 5 – Percentage evaporation deviations dP as a function
of Ta, A and rh; ra = 100 s mS1; rs = 70 s mS1; p = 101.3 kPa.
Negative deviations = PMe underestimates.
4.1. Basic considerations

The concept involves two steps:

- estimation of the surface temperature Ts different from the


PMe linearization;
- inserting Ts in Eq. (2) or Eqs. (1) and (3) to obtain estimates for
evaporation.

Starting point is the energy balance Eq. (4), which after


some transformation can also be written as

esat ðTs Þ Ara ea


þ r1 Ts ¼ r1 Ta þ r1 þ (9)
gðTs Þ cv gðTs Þ

The right side terms will be lumped, so that Eq. (9) reads

esat ðTs Þ
þ r1 Ts ¼ S (9a)
gðTs Þ

where S is given by
Fig. 6 – Percentage evaporation deviations dP as a function ra ea
S ¼ r1 ðTa þ AÞ þ (9b)
of Ta and rh. A = 0 W mS2; ra = 5 s mS1; rs = 70 s mS1; cv gðTs Þ
p = 101.3 kPa. Negative deviations = PMe underestimates.
The term S has the unit 8C and controls (determines) together
with r1 the Ts-value (see Fig. 10). It will be called the surface
temperature control sum.
A graphical solution can be derived from Eq. (9)
estimates, the use of which appears to be promoted in after slight modification. Fig. 8 gives details: S/r1 is
recent literature (Alexandria and Kerkides, 2003; Lecina plotted along the abscissa up to point 5. The line 3–5 has
et al., 2003; Allen et al., 2006; Lopez-Urrea et al., 2006; Perez the slope r1g(Ts). The intersection of line 3–5 with the vapor
et al., 2006). This in mind we offer an alternative pressure saturation curve 2–3–4, i.e., point 3, has the
approximation method, which results in far less deviations abscissa Ts which is the solution. The line 1–2 is
from the linearized PMe even in extreme climatic cases. almost parallel to 3–5, since g(Ts) changes only slightly with
Thereby Eq. (4) is not linearized but approximated in a way Ts (see Section 4.2). It corresponds to the adiabatic case with
that allows for easy use of common mathematical tools wet bulb temperature Tw ¼ Ts as solution (not indicated in
equipped with a solver routine. It does not, however, reach Fig. 8).
the accuracy of IT or other non-linear solutions quoted In the following a mathematical solution of Eqs. (9) will be
above. presented making use of an approximation.
agricultural water management 96 (2009) 711–721 717

Table 3 – Tendencies of negative/positive relative errors dP in relation to r1, A, air temperature (cold/hot) and air humidity
(dry/wet). Gray fields with pronounced errors.

 
4.2. Estimation of the surface temperature (step 1) a bTs
exp þ r1 Ts  S0 (10a)
g ðTs Þ c þ Ts
The proposed approximation consists in eliminating the
unknown Ts-terms on the right side of Eq. (9), i.e., in with
Eq. (9b): The function cp(Ts) is replaced by the constant
Ara ea
cp = 1.01 kJ kg1 K1, and g(Ts) is substituted by g(Ta). On the S0 ¼ r1 ðTa þ Þþ (10b)
1:01ra gðTa Þ
left side of Eq. (9), no replacement of Ts takes place.
These right side approximations – also used by the PMe – abbreviations and units being the same as in Eqs. (1)–(4). The
are justified by the following two considerations: test runs for this paper resulted in S0 -values ranging from 5 to
approximately 800 8C.
(1) Ts-values are assumed to range between 0 and 40 8C for The non-linear, but simplified, Eqs. (10) can be solved for Ts
practical application in agriculture. This means a change of by means of any mathematical device featuring a solver
cp-values from about 1.006 to 1.013 kJ kg1 C1, i.e., 0.7% routine.1 One may also apply Newton’s method of successive
only. approximations (see e.g., Kreyszig, 1999, p 841) which reads for
(2) The maximum difference between air (Ta) and surface (Ts) the actual problem
temperature for the tests in this paper lies around 25 8C.   
esat ðti Þ Dðti Þ 1
This comes to no more than 1.4% difference when inserting tiþ1 ¼ ti  þ r1 ti  ST0 r1  ði
gðti Þ gðti Þ
Ta instead of Ts-values in the g(T)-function.
¼ 0; 1 . . . nÞ (11)
Thus cp and g are considered here to be of weak influence on
the approximate solutions. The purpose of re-arranging Eq. (4) where D(ti) is the 1st derivative of Eq. (5) with respect to ti, that
according to Eqs. (9) is to have Ts-values of strong influence is D(ti) = esat(ti)cb(c + ti)2 and esat(ti) is given by Eq. (5).
[i.e., esat(Ts) increases exponentially with Ts] on the left side and After n steps, tn will approximate Ts? (tn ffi Ts0 ). Starting point
to have on the right side of Eq. (9) values of weak influence [i.e., is t0 that should be close to Ts0 . Trials show, however, that for
g and cp] together with all five input parameters (Ta, ea, A, ra, rs), most cases of practical relevance, Ta is an efficient starting
which are assumed to be known. point to reach the solution after about 4–10 approximation
In this way, S turns into an independent although steps. Fig. 10 can give additional support to select t0: with r1
approximate input variable S0 and Eq. (9a), also using and S0 given one can estimate starting-t0 as Ts0 from Fig. 10. For
Eq. (5), can thus be rewritten as cases with extreme meteorological conditions (say e.g., rh near
to 0% or 100%), Eq. (11) may become instable and a check for
plausibility of the result is recommended even if the ti-series
converges.
Approximate solutions for Ts0 are presented in Figs. 9 and
10.
The leftmost curve in Fig. 9a represents the situation with
rs = 0, i.e., r1 = 1, and can be excellently fitted (R2 = 0.999,
SD  0.1 8C) by Eq. (12)

Ts00 ¼ 28:35 lnðS0 þ 34:7Þ  107:4 (12)


1
with Ts00
= surface temperature in 8C for rs = 0 s m ; S0 is
inserted dimensionless.
This equation gives surface temperature estimates for non-
water stressed plants. Fig. 9b compares Eq. (12) (full line) with
IT-solutions (symbols) where ra = 30 s m1 and all other input
variables are in the range of Table 1, bottom line. Fig. 9b shows
the corresponding comparison with approximate solutions as
suggested by Jackson et al. (1981) and Jackson (1982).
Fig. 8 – Illustration of Eq. (9): curve 2–3–4 = esat(T); line 1–2
(dotted): adiabatic line with slope Sr1g(Ts); line 3–5: 1
Programs for Mathcad1 and for the HP1-48GX pocket calcu-
diabatic line parallel to line 1–2 in distance R ¼ Ara =cv . lator are available from the author.
718 agricultural water management 96 (2009) 711–721

Fig. 10 – (a) Ts0 as a function of S0 (0–300 8C) and r1-values,


two-step approximations. (b) Ts0 as a function of S0 (0–
1000 8C) and r1-values, two-step approximations.

  
rc p esat T0s  ea =ðra þ rs Þ
E0a ¼ c f  (13a)
rw 2501:2  2:369T0s gðT0s Þ
Fig. 9 – (a) Surface temperature Ts as a function of S0 -values
for different r1-values. Symbols correspond to IT-solutions
whereas the right side solution yields
(*: A = 0 W mS2; ~: A = 200 W mS2; &: A = 400 W mS2; ^:

A = 600 W mS2). (b) Comparison of Eq. (11) (full line) with A  rc p T0s  Ta =ra
E0b ¼ c f  (13b)
IT-solutions (with symbols *: A = 0 W mS2; ~: rw 2501:2  2:369T0s
A = 300 W mS2; &: A = 600 W mS2) for ra = 30 s mS1, r1 = 1;
with E0a and E0b = approximate evaporation; cf = conversion
all other input variables are within the range of Table 1,
factor (e.g., 8.64  107 s d1 mm m1 to change from m s1 to
bottom line. (c) Comparison of Eq. (11) (full line) with
mm d1); rw = water density changing less than 1% within
Jackson-approximations; Jackson et al. (1981); (with
the air temperature range under consideration (here taken
symbols *: A = 0 W mS2; ~: A = 300 W mS2; &:
as 998 kg m3); rcp = the volumetric heat capacity of air,
A = 600 W mS2) for ra = 30 s mS1, r1 = 1; all other input
where r is calculated using Eq. (6) and cp is taken as
variables are within the range of Table 1, bottom line.
1.01 kJ kg1 K1.
For IT-solutions both equations yield identical results.
Inserting approximated Ts0 -values, however, leads to differ-
ences. One may expect bigger deviations using Eq. (13a), where
errors of Ts0 increase exponentially (Eq. (5)). Test runs prove
4.3. Evaporation evaluated from surface temperatures that Eq. (13b) gives better results. In the following, we refer
(step 2) only to E0b -solutions, i.e., Eq. (13b).
These solutions will be called two-step solutions. To
Using surface temperature Ts0 as determined in step 1, one compare results of the different methods see Tables 4 (for
obtains evaporation by inserting Ts0 into the left or into the PMe-solutions compare bottom line in Table 4).
right side of Eq. (4). The left side ‘‘latent heat’’ solution thus Fig. 11 compares two-step solutions with IT-solution for
yields evaporation.
agricultural water management 96 (2009) 711–721 719

Table 4a – Daily mean evaporation deviations (med = two-step method minus IT-solutions) sorted out for E < 10 mm dS1.
Statistics for eight ra/rs-combinations, each one with 384 different Ta-, ea- and A-combinations (see Table 1a upper row).
n = admitted test runs. Extreme deviations in bold. Last line: means of PMe deviations from Table 2a for comparison.
pa (kPa) ra rs med SD max min max min rho n r1
(s m1) (s m1) (mm d1) (mm d1) (mm d1) (mm d1) (%) (%)

101.3 = 0 m a.s.l. 5 70 0.00 0.01 +0.04 0.02 +1.0 0.9 1.0000 216 15
10 70 0.00 0.01 +0.05 S0.04 +0.9 0.9 1.0000 233 8
50 70 +0.00 0.01 +0.04 0.02 +0.6 1.2 1.0000 320 2.4
50 100 0.00 0.01 +0.05 0.02 +0.7 1.1 1.0000 337 3
100 100 0.00 0.01 +0.04 0.02 +0.5 S1.3 1.0000 364 2
10 200 0.00 0.02 +0.04 0.01 +1.4 1.0 1.0000 341 21
100 200 0.00 0.01 +0.06 0.01 +0.7 1.1 1.0000 378 3
200 200 0.00 0.01 +0.05 0.01 +0.6 S1.3 1.0000 380 2
means two-step 0.00 0.01 +0.04 0.02 +0.80 1.1 1.0000 321
means PMe 0.13 0.23 +0.36 0.90 +6.91 17.66 0.9951 321

Table 4b – Hourly mean evaporation deviations (med = two-step method minus IT-solutions) sorted out for E < 20 mm dS1.
Statistics for eight ra/rs-combinations, each one with 384 different Ta-, ea- and A-combinations (see Table 1a lower row).
n = admitted test runs. Extreme deviations in bold. Last line: means of PMe deviations from Table 2b for comparison.
pa (kPa) ra rs med SD max min max min rho n r1
(s m1) (s m1) (mm d1) (mm d1) (mm d1) (mm d1) (%) (%)

101.3 = 0 m a.s.l. 5 70 0.00 0.01 +0.08 0.03 +1.2 1.0 1.0000 307 15
10 70 0.00 0.01 +0.11 S0.04 +1.1 0.9 1.0000 324 8
50 70 +0.00 0.02 +0.10 0.03 +0.6 1.2 1.0000 371 2.4
50 100 0.00 0.02 +0.11 0.03 +0.7 1.1 1.0000 378 3
100 100 0.00 0.02 +0.10 0.02 +0.6 S1.3 1.0000 380 2
10 200 0.00 0.01 +0.08 0.01 +1.6 1.0 1.0000 383 21
100 200 0.00 0.02 +0.13 0.03 +0.8 1.1 1.0000 384 3
200 200 0.00 0.02 +0.10 0.02 +0.6 S1.3 1.0000 384 2
means two-step 0.00 0.02 +0.10 0.03 +0.90 1.11 1.0000 363
means PMe 0.45 0.68 +0.62 2.44 +6.91 26.23 0.9895 364

5. Summary and conclusions

The PMe offers the advantage of encapsulating the general


equations describing heat and vapor exchange over wet
surfaces in a compact formula with elementary algebra which
can easily be handled but still reflects the essential physical
processes of evaporation. This is achieved by a linearization
leading to acceptable results in a wide range of applications as
shown above. However, under certain conditions (compare
Tables 2 and 3) errors may go up beyond tolerable limits.
PMe-solutions for daily (hourly) evaporation are shown to
be more than 30% (45%) smaller (negative errors) than the
corresponding IT-solutions for extreme cases. This compares
with errors of 30% as quoted Paw and Gao (1988).
This paper gives evidence of also positive errors which may
go up to about +9% (Tables 2a and 2b). This contrasts with Milly
(1991), who argues from a theoretical point of view that errors
are always non-positive, reaching extreme values of around
18%. On the other hand, field investigations by Gavin and
Agnew (2004), on wet grassland, by Grant (1975), on barley and
Fig. 11 – Evaporation as a function of temperature de Medeiros et al. (2006), on beans, confirm possible over-
difference dT0 = (Ts0 S Ta); Symbols represent IT-solutions estimation by the PMe (positive errors). Generally, comparison
(*: A = 60 W mS2, ~: A = 120 W mS2 and &: with field data is hampered by the difficulty to find good
A = 300 W mS2), lines represent two-step solutions.; estimates for rs-values (see, e.g., Kustas et al., 1996).
ra = 10 s mS1 (full symbols); ra = 140 s mS1 (empty In this paper surface temperature is approximated by a
symbols); rs = 70 s mS1; ea = 1.9 kPa; p = 101.3 kPa. simplified non-linear equation containing the surface tem-
720 agricultural water management 96 (2009) 711–721

perature control sum S0 , in which g(Ts) is replaced by g(Ta) and Medeiros de, G.A., Arruda, F.B., Sekai, E., 2006. Crop coefficient
cp is kept constant. Thus all five input parameters (Ta, ea, A, ra, for irrigated beans derived using three reference
evaporation methods. Agric. For. Meteorol. 135, 135–143.
rs), assumed to be known, are packed into an algebraic
Gavin, H., Agnew, C.A., 2004. Modelling actual, reference and
equation that allows its simplified non-linear solution
equilibrium evaporation from a temperate wet grassland.
(Eq. (10)). It is worth to point out that under this manipulation, Hydrol. Process. 18, 229–246.
one and the same S0 -value – whatever are the (realistic) values Grant, D.R., 1975. Comparison of evaporation from barley with
of its three additive terms (see Eq. (9b)) – leads for any Penman estimates. Agric. Meteorol. 15, 49–60.
particular r1-value to one and the same Ts0 -solution. This Jackson, R.D., Idso, S.B., Reginato, R.J., Pinter, P.J., 1981. Canopy
suggests to concentrate on the rs/ra ratio (instead of rs-values temperature as a crop stress indicator. Water Resour. Res.
17, 1133–1138.
only) in research on plant water use.
Jackson, R.D., 1982. Canopy temperature and crop water stress.
Surface temperatures of non-water stressed plants Adv. Irr. 1, 43–85.
(rs = 0 s m1) can be estimated very closely to IT-solutions Jones, H.G., 1992. Plants and Microclimate, second edn. Univ.
(SD  0.1 8C) by applying Eq. (11), which can further be used to Press, Cambridge.
estimate potential evaporation (Lhomme, 2001). Klaasen, W., 2001. Evaporation from rain-wetted forest in
On the basis of approximated Ts0 -values obtained, evapora- relation to canopy wetness, canopy cover and net radiation.
Water Resour. Res. 37, 3227–3236.
tion rates are estimated using Eq. (13b). The two-step concept
Kreyszig, E., 1999. Advanced Engineering Mathematics, 8th
results in much smaller errors than the PMe-solution in
Edition. John Wiley & Sons, Inc..
special cases. Mean daily (hourly) relative errors dP are about 8 Kustas, W.P., Stannard, D.I., Allwine, K.J., 1996. Variability in
(6) times smaller for positive and 17 (23) times smaller for surface energy flux partitioning during Washiba ‘92:
negative deviations (Table 4; compare with bottom line). These Resulting effects on Penman–Monteith and Priestly-Taylor
solutions, however, cannot compete with accuracies obtained parameters. Agric. For. Meteorol. 82, 171–183.
by IT or quadratic–quaternic solutions as suggested by Paw Lecina, S., Martinez-Cob, A., Perez, P.J., Villabos, F.J., Baselga, J.J.,
2003. Fixed versus variable bulk canopy resistance for
and Gao (1988).
reference evapotranspiration estimation using the
The use of the two-step approach is meant for practical
Penman–Monteith equation under semiarid conditions.
purposes in agricultural water management. Agric. Water Manage. 60, 181–198.
Lhomme, J.P., 2001. Stomatal control of transpiration:
Examination of Jarvis-type representation of canopy
Acknowledgments resistance in relation to humidity. Water Resour. Res. 37,
689–699.
Lopez-Urrea, R., Martin de Santa, O., Fabeiro, C., Moratalla, A.,
This paper would not have been written without the
2006. An evaluation of two hourly reference
challenging and encouraging criticism of Prof. I. Ferreira, evapotranspiration equations for semiarid conditions.
Lisbon. I also thank Dr. R. Vogt, Basel, and Prof. Th. Dracos, Agric. Water Manage. 86, 277–282.
Zurich for valuable suggestions. Milly, P.D.C., 1991. A refinement of the combination equations
for evaporation. Surveys Geophys. 12, 145–154.
Monteith, J.L., 1965. The state and movement of water in living
references organisms. In: Proceedings of the XIX Symposium of
Society for Experimental Biology, Swansea. Cambridge
University Press, Cambridge, pp. 205–234.
Monteith, J.L., 1985. Evaporation from land surfaces. In: Progress
Alexandria, S., Kerkides, P., 2003. New empirical formula for in analysis and prediction since 1948; Proceedings of the
hourly estimations of reference evapotranspiration. Agric. National Conference on Advances in Evapotranspiration,
Water Manage. 60, 157–180. December 16–17, pp. 4–12.
Allen, R.G., 1986. A Penman for all seasons. J. Irrig. Drain. Eng. Monteith, J.L., Unsworth, M.H., 1990. In: Arnold, Edward (Ed.),
112, 348–368. Principles of Environmental Physics. second edn. A division
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotran- of Hodder & Stoughton, London, N.Y., Melbourne, Auckland.
spiration: Guidelines for Computing Crop Water Require- Murray, F.W., 1967. On the computation of saturation vapor
ments. United Nations Food and Agricultural Organization. pressure. J. Appl. Meteorol. 6, 203–204.
Irrigation and Drainage Paper 56. Rome, Italy, 300 pp. Paw U, K.T., Gao, W., 1988. Applications of solutions to non-
Allen, R.G., Pruitt, W.O., Wright, J.L., Howell, T.A., Ventura, F., linear energy budget equations. Agric. For. Meteorol. 43,
Snyder, R., Itenfisu, D., Stedudo, P., Berengena, J., Yrisarry, 121–145.
J.B., Smith, M., Pereira, L.S., Raes, D., Perrier, A., Alves, I., Paw U, K.T., 1992. A discussion of the Penman form equations
Walter, I., Elliot, R., 2006. A recommendation on and comparisons of some equations to estimate latent
standardized surface resistance for hourly calculation of energy flux density. Agric. For. Meteorol. 57, 297–304.
reference ETo by the FAO 56 Penman–Monteith method. Penman, H.L., 1948. Natural evaporation from open water, bare
Agric. Water Manage. 81, 1–22. soil and grass. Proc. R. Soc., Lond., A 193, 120–146.
Alves, I., Pereira, L.S., 2000. Non-water-stressed baselines for Penman, H.L., 1963. Vegetation and hydrology. tech. Comm. No.
irrigation scheduling with infrared thermometers: a new 53, Commonwealth bureau of soils, Harpenden, UK, 125 pp.
approach. Irrig. Sci. 19, 101–106. Perez, P.J., Lecina, S., Castelloi, F., Martiny-Cob, A., Villalobos,
Batchelor, C.H., 1984. The accuracy of evaporation estimated F.J., 2006. A simple parameterization of bulk canopy
with the FAO modified Penman equation. Irrig. Sci. 5, 223– resistance from climatic variables for estimating hourly
233. evapotranspiration. Hydrol. Processes 20, 515–532.
Brown, K.W., Rosenberg, N.J., 1973. A resistance model to Rana, G., 2001. Evaporation of sweet sorghum: a general model
predict evapotranspiration and its applications to a sugar and multilocal validity in semiarid environmental
beet field. Agron. J. 65, 341–347. conditions. Water Res. Res. 37, 3237–3246.
agricultural water management 96 (2009) 711–721 721

Stahl, P.H., 1980. Feuchtigkeit und Trocknen in der Tracy, C.R., van Berkum, F.H., Tsuji, J.S., Stevenson, R.D.,
pharmazeutischen Technologie. UTB, Stuttgart, Germany. Nelson, J.A., Barnes, B.M., Huey, R.B., 1984. Errors resulting
Sumner, D.M., Jacobs, J.M., 2005. Utility of Penman–Monteith, from linear approximations in energy budget equations. J.
Priestley-Taylor, reference evapotranspiration, and pan Therm. Biol. 9, 261–264.
evaporation methods to estimate pasture Xu, C.-Y., Chen, D., 2005. Comparison of seven models for
evapotranspiration. J. Hydrol. 308, 81–104. estimation of evapotranspiration and groundwater
Thom, A.S., 1972. Momentum, mass and heat exchange of recharge using lysimeter measurements data in Germany.
vegetation. Quart. J. R. Met. Soc. 98, 124–134. Hydrol. Processes 19, 3717–3734.

You might also like