Agricultural Water Management 180 (2017) 50–60

Contents lists available at ScienceDirect

Agricultural Water Management

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

Selecting the best model to estimate potential evapotranspiration

with respect to climate change and magnitudes of extreme events
Mohammad Valipour, Mohammad Ali Gholami Sefidkouhi ∗ , Mahmoud Raeini−Sarjaz
Department of Water Engineering, Sari Agricultural Sciences and Natural Resources University, Sari, Iran

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

Article history: There are a lot of investigations to select the best model to estimate potential evapotranspiration (ETo ) in
Received 19 February 2016 a certain climate or region. In this paper, the types of climate include arid, semiarid, Mediterranean, and
Received in revised form 15 August 2016 very humid. A spatial and temporal study of the ETo is the aim of this paper, according to the peak and
Accepted 23 August 2016
low events (extreme events) and climate change alarms. For this purpose, 50 years (1961–2010) monthly
Available online 4 November 2016
meteorological data of 18 regions in Iran, with various climates, were collected. For estimating the ETo , 5
temperature−based, 5 radiation−based, and 5 mass transfer−based models, were selected with respect
Keywords:
to better performance of them in different climates on the basis of past investigations. The results will
Best model
Climate variability
especially be useful in the regions where the monthly (rather than daily) meteorological data are available.
Crop water requirement The results appear that the Blaney−Criddle (BC) (root mean square error (RMSE) = 1.32 mm day−1 ) and
Iran Abtew (Ab) (RMSE = 0.83 mm day−1 ) are the best models for estimating the ETo in the arid and semiarid
Uncertainty regions, respectively. While, modified Hargreaves−Samani 2 (MHS2) represents the best performance
Weather parameters in the Mediterranean and very humid regions (RMSE = 0.30 mm day−1 & 0.68 mm day−1 , respectively).
In addition, radiation—and mass transfer−based models are proper tools to estimate the ETo in warm
and cold seasons on the basis of improving values of evaluation indices in 40% and 70% of the study
area, respectively. Increasing air temperature and decreasing minimum relative humidity for best perfor-
mance of most models alarms a climate change in most regions of Iran. As a result, the radiation−based
models were adapted with climate change better than the temperature−based and particularly mass
transfer−based models. Finally, a step by step flowchart was presented for selecting the best model to
estimate the ETo in each climate.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction (average data). In the other studies, the researchers focused on

There have been many studies into the estimation of ETo around
the world. Although there are some advancements in each study,
one or more limitation also leads to decreasing reliability of the
models those were introduced by researchers as the best for the
estimating the ETo . It can be observed using an overview on the
literature.
Ngongondo et al. (2013) claimed that the PT and HS meth-
ods underestimate the ETo by using a 37−year data in an arid
environment, in Malawi. However, there are also some inves-
tigations in which overestimation of the HS were reported
(Azhar and Perera, 2011; Ashraf et al., 2014; Patel et al., 2015).
Ahmadi and Fooladmand (2008) indicated advantages of Thorn-
thwaite’s equation, in southern Iran, by using 20−year data
∗ Corresponding author.
E-mail address: mohammad25mordad@yahoo.com (M.A. Gholami Sefidkouhi).
importance of water management which indicates the role of
accurate estimation of evapotranspiration to deal with water cri-
sis (Valipour, 2012a,b,c,d,e; Valipour, 2014; Valipour, 2015e,f,g;
Valipour, 2016a,b). Shiri et al. (2014) revealed the superiority of
the HS compared to the Mk, Turc, and PT by using a 9−year period
data.
An overview on the mentioned above indicates that anyone
of the previous investigations suffers one or more the sources of
uncertainty. One of them is lack of sufficient data which not only
leads to reduction of reliability for the results but also leads to
inapplicability for studying the role of climate variability and/or
climate change on the accuracy of the models. In addition, the
main part of the reported errors related to the peak and low
events which occurred in warm and cold seasons, respectively,
that has not been evaluated separately. Moreover, the efficiency
of the Food and Agriculture Organization of the United Nation
(FAO)−Penamn−Monteith (FPM), as the base model, has not been
characterized. There are also some other sources of uncertainty

http://dx.doi.org/10.1016/j.agwat.2016.08.025
0378-3774/© 2016 Elsevier B.V. All rights reserved.
 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60 51

such as lack of proper selection of temperature−, radiation−, and investigations (e.g. Tabari et al., 2013; Valipour, 2015b; Valipour
mass transfer−based models for accurate estimating the ETo with and Eslamian, 2014) and the results were compared with the FPM.
respect to the climate of the region. Table 2 shows the selected models with their References
In the other hand, some studies tried to reflect the peak and It should be noted that the definition and use of the term “refer-
low errors. For instance, Caporusso and Rolim (2015) compared 18 ence evapotranspiration” was developed in the 1970s (Wright and
different models to estimate the ETo in a humid region of Brazil Jensen, 1972; Doorenbos and Pruitt, 1977) to resolve ambiguities
using only 6−year meteorological data. The results indicated that involved in the definition and interpretation of “potential evapo-
Presley−Taylor (PT) and Thornthwaite (Th) estimate the ETo in ration”; the “reference” descriptor points to the use of a specific
warm and cold seasons, respectively, better than other models type of vegetation or specific definition of vegetation properties to
against the FPM. represent the evaporative index.
In the other hand, climate change impacts on the different Although there are a lot of studies in which the empirical meth-
parameters applied for various models to estimate the ETo (Table 2). ods (15 selected equations) have been compared with the FPM
Therefore, accurate analysis of climate change is very important to (e.g. Ahmadi and Fooladmand, 2008; Tabari et al., 2013; Valipour,
find the best model which has more adaptation with the variations 2015c), the ETo from different equations may has different defini-
of meteorological data in future. There are many investigations to tions which most of the mentioned investigations have neglected
deal with climate change in Iran. For example, Ashraf et al. (2014) this. The different definitions of empirical models related to ini-
showed that the most significant increasing temperature occurred tial conditions for which these models have been extracted (See
at the beginning of 21 century in all locations of Iran during a Table 2). The FPM equation refers to the potential evapotranspira-
48−year period of 1961–2008. According to precipitation anoma- tion of a grass surface, but other equations might consider the ETo
lies, all locations experienced dry and wet periods, but generally as the evaporation from an open water surface. Table 2 compares
dry periods occurred more often especially in the beginning of 21 also the different definitions of all models used in this paper.
century. To evaluate the accuracy of the models four indices were used
There are three sources of uncertainty that they are more consid- as follows:
erable in the previous studies including decreasing accuracy of the 
n
selected models in the estimation of peak and low values occurred 
 (Xi − Yi )2
in warm and cold seasons, respectively, lack of considering the 
i=1
role of climate change to select the superior model, and/or lack of RMSE = (1)
validation of the base method (i.e. the FPM in most cases). Thus, n

the literature review shows that there are considerable sources  Np

of uncertainty in the past studies which need to a comprehensive 
4
(Xi − Yi )2 × (Xi )2
work to reduce the mentioned limitations and to improve reliability
i=1
of the introduced models for estimating the ETo in each climate. This PVC =  (2)
study aims to estimate the ETo by selecting 15 more recommended  Np

models (based on their performance in the previous investigations)  (Xi )2
with respect to values of the cold and warm seasons and climate i=1
change alarms in 18 regions of Iran, with various climates, during 
a 50−year period from 1961 to 2010.  Nl


4
(Xi − Yi )2 × (Xi )2
i=1
2. Materials and methods LVC =  (3)
 Nl

The monthly averages of meteorological data were collo-  (Xi )2
cated from Islamic Republic of Iran Meteorological Organization i=1
(IRIMO) (http://irimo.ir/eng/index.php). These data contain mean,
minimum, and maximum daily air temperature ( ◦ C), mean and X̄ − Ȳ
Student’st − test =  (4)
minimum relative humidity (%), wind speed (m s−1 ), rainfall (mm  
month−1 ), and sunshine (hr month−1 ). Table 1 shows the position
 n n
 (
2
Xi −X̄ ) + (Yi −Ȳ )
2

of all 18 synoptic stations and their climates. 

Among all stations, there is 50−year period information for 16
regions. In addition, there is 27−year and 21−year period informa-
tion for Moghan and Jiroft, respectively, that they were also added
to other 16 regions (with 50−year data).
Although the FPM model has been applied in various regions
(e.g. Rahimi et al., 2015; Valipour, 2013a,b,c; Valipour, 2015a;
Yannopoulos et al., 2015), it needs too many parameters to esti-
mate the ETo . In most regions (without synoptic stations and/or
with un−gauged stations), meteorological data are limited and
researchers cannot use the FPM model. To this end, empirical meth-
ods have been developed for the estimation of the ETo using limited
data.
Among numerous empirical methods to estimate the ETo , 5
temperature−based (HS, modified Hargreaves−Samani 1 (MHS1),
MHS2, Th, and BC), 5 radiation−based (JH, PT, Mk, Ab, and Tu), and 5
mass transfer−based models (Penman (Pe), Ivanov (Iv), Mahringer
(Ma), Trabert (Tr), and WMO), were selected with respect to better
performance of them in different climates on the basis of the past
i=1 i=1
n(n−1)
where, PVC and LVC reflect the peak and low errors occur in the
warm (June–September) and cold (November–February) seasons
(El-Shafie et al., 2009; El-Shafie et al., 2014), respectively, Xi and Yi
are the calculated ETo using the FPM and estimated ETo using 15
empirical equations, respectively; X and Y are the average of Xi and
Yi , Np is number of peak evapotranspiration greater than one−third
of the mean peak ETo observed, Nl is number of low ETo lower than
one−third of the mean low evapotranspiration observed and n is
the total numbers of data. First, the average of extreme events (ETo
in June to September) was obtained. Then, this value was dived to
3 and extreme events were compared to this value to determine
Np . Similar method (ETo in November to February) was employed
to characterize Nl . This approach has already been applied in the
previous investigations (El-Shafie et al., 2009; El-Shafie et al., 2014).
The following has been prepared in the different parts for better
understanding readers. First, the more accurate model was charac-
52 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60

Table 1
Position and climate of the stations with length of collected data.

Station Name ICAO Code North Latitude East Longitude Altitude (masl) Start Year EndYear Climate

Ahvaz 40811 31◦ 20 48◦ 40 22.5 1961 2010 Arid
Arak 40769 34◦ 6 49◦ 46 1708.0 1961 2010 Semiarid
Bushehr 40858 28◦ 58 50◦ 49 9.0 1961 2010 Arid
Esfahan 40800 32◦ 37 51◦ 40 1550.4 1961 2010 Arid
Hamedan 40768 34◦ 52 48◦ 32 1741.5 1961 2010 Semiarid
Jiroft 40866 28◦ 35 57◦ 48 601.0 1989 2009 Arid
Kerman 40841 30◦ 15 56◦ 58 1753.8 1961 2010 Arid
Mashhad 40745 36◦ 16 59◦ 38 999.2 1961 2010 Semiarid
Moghan 40700 39◦ 39 47◦ 55 31.9 1984 2010 Semiarid
Qazvin 40731 36◦ 15 50◦ 3 1279.2 1961 2010 Semiarid
Rasht 40719 37◦ 19 49◦ 37 −8.6 1961 2010 Very humid
Sanandaj 40747 35◦ 20 47◦ 0 1373.4 1961 2010 Mediteranean
Shahrekord 40798 32◦ 17 50◦ 51 2048.9 1961 2010 Semiarid
Shiraz 40848 29◦ 32 52◦ 36 1484.0 1961 2010 Semiarid
Tabriz 40706 38◦ 5 46◦ 17 1361.0 1961 2010 Semiarid
Urmia 40712 37◦ 40 45◦ 3 1328.0 1961 2010 Semiarid
Yazd 40821 31◦ 54 54◦ 17 1237.2 1961 2010 Arid
Zabol 40829 31◦ 2 61◦ 29 489.2 1961 2010 Arid

terized in each region with respect to the evaluation indices. Then,
the best performance of each model was detected with respect to
the evaluation indices. In the next step, the cumulative ETo and
its error were evaluated. The assessment of the more accurate
model and the best performance of each model have also been
done in each region with respect to the climate change. Then, the
variations of the meteorological parameters were assessed in the
best performance of each model. In addition, the accuracy of the
temperature−, radiation−, and mass transfer−based models was
analyzed in each region and climate with respect to the evaluation
indices and the error of the warm and cold seasons. In the next
step, the results were investigated for each climate and whole of
the Iran. Finally, a step by step flowchart was presented to select
the best method for estimating the ETo .
3. Results
3.1. The more accurate model in each region and best
performance of each model with respect to the evaluation indices
Fig. 1 yielded an excellent relationship (R2 > 0.9100) between
the best models and FPM for 14 regions as well as a good rela-
tionship (R2 > 0.8100) between the best models and FPM for other
regions (Ahvaz, Arak, Esfahan, and Shahrekord).
However, there are 26% and 20% error between the values cal-
culated by the FPM and estimated by the best models for Hamedan
and Qazin (both of them have a semiarid climate on the basis of
Table 1), respectively (see also Kisi et al., 2015; Shiri et al., 2014;
Tabari et al., 2013). Other regions recorded the error less than 20%
(6 regions contain 3 semiarid regions) and less than 10% (10 regions
contain 6 semiarid regions). The MHS2 is the superior model due
to better performance for 5 regions (contain all four climates and
there is underestimation for four of these regions). There is an excel-
lent relationship (R2 > 0.9100) between the best performance of 9
models (Ab, BC, HS, Iv, JH, MHS1, MHS2, PT, and Tu) and FPM, a
good relationship (R2 > 0.8100) between the best performance of
4 models (Ma, Mk, Th, and WMO) and FPM, a fairly relationship
(R2 = 0.7394) between the best performance of the Pe and FPM, and
a poor relationship (R2 = 0.6775) between the best performance of
the Tr and FPM. However, there is 42% error between the values cal-
culated by the FPM and estimated by the best performance of the
Tr. Other models recorded the error less than 20% (JH, Ma, MHS1,
Mk, Pe, Th, and WMO) and less than 10% (Ab, BC, HS, Iv, MHS2, PT,
and Tu). Therefore, the worst fitness between the values calculated
by the FPM and estimated by the best performance of the models
belongs to mass transfer−based models (4 models with the error
more than 10%) which confirm the results of Fig. 1.
3.2. The assessment of the more accurate model and best
performance of each model in each region with respect to the
climate change
As mentioned on the importance of the evaluation of the role of
climate change, it is necessary to count influence of climate change
on the performance of the different methods of estimating the ETo
to raise the reliability aspect of the best models introduced for each
region and climate. For this purpose, Fig. 2 was plotted to analyze
the spatiotemporal variations of the error of the best models.
Among 5 models which presented their best performances for
Moghan, there is no significant difference between the values of the
error of the first and second periods for the Iv model. Therefore, this
model is not able to reflect climate change in Moghan. However, the
values of the error have reduced during the second 13−year for the
other 4 models. This reduction is more for the PT with the minimum
error (38 mm) during the second half than the other models. In
addition, among 4 models which presented their best performances
for Rasht, the values of the error have increased for three of them
in the second 25−year. Therefore, the MHS2 is the best model to
estimate the ETo in Rasht with the minimum error (72 mm) with
respect to the climate change alarm (Ashraf et al., 2014; Mosaedi
et al., 2015; Roshan et al., 2015).
3.3. The assessment of the accuracy of the temperature−,
radiation−, and mass transfer−based models in each region and
climate with respect to the evaluation indices and error of the
warm and cold seasons
As shown in Fig. 3, radiation−based models estimate the ETo
in arid, semiarid, and Mediterranean climates and temperature−
and mass transfer−based models estimate the ETo in very humid
regions better than other models.
The accuracy of the empirical models, in the northern and
north−western Iran (with lowest rate of the ETo ) is higher than
other areas of Iran (see also Heydari et al., 2015). While, the mod-
els have not accurately resulted in eastern and south−eastern Iran
with arid climate (the highest rate of the ETo ).
Fig. 3 does not consider the peak and low errors. Thus, Fig. 4 was
plotted to assess the importance of The ETo recorded in the warm
and cold seasons (peak and low values, respectively).
According to Fig. 4, there is only one region (Shiraz) in which
the values of the PVC and LVC for the more accurate model with
Table 2
Selected models to estimate potential evapotranspiration including their references, formulae, and parameterization.

Model Reference(s) Formula Parameterization Definitions

900
0.408(Rn − G) +  T +273
u(es − ea )
FAO Penman-Monteith (FPM) Allen et al. (1998) ETo =
 +  (1 + 0.34u)  month2
G = 0.14(T
n
− Tmonth1 ) Potential evapotranspiration from a
Rs = as + bs Ra grass surface
N
Hargreaves-Samani (HS) Hargreaves and Samani (1985) ETo = 0.00102Ra (Tmax − Tmin )0.5 (T + 16.8) – Potential evapotranspiration from a
grass surface
Jensen-Haise (JH) Jensen and Haise (1963) ETo = 0.0102 (T + 3) Rs – Potential evapotranspiration from an
alfalfa surface
Penman (Pe) Penman (1948) ETo = (2.625 + 0.713u) (es − ea ) – Potential evaporation from an open
water surface and bare soil
 Rn − G  n
Priestley-Taylor (PT) Priestley and Taylor (1972) ETo = 1.26 Rs = as + bs Ra Potential evapotranspiration from a
+  N
pasture surface & evaporation from an
open water surface
 m 
12
T

M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60

1.514
Thornthwaite (Th) Thornthwaite (1948) ETo = 16 10 I= (0.2Ti ) Potential evapotranspiration
I
i=1
m = 6.75 × 10−7 I 3 − 7.71 × 10−5 I 2

+1.7912 × 10−2 I + 0.49239

Trabert (Tr) Trabert (1896) ETo = 3.075u0.5 (es − ea ) – Potential evaporation from an open
water surface
Ivanov (Iv) Romanenko (1961) ETo = 0.00006(25 + T )2 (100 − RH) – Potential evaporation from an open
water surface
0.5
Mahringer (Ma.) Mahringer (1970) ETo = 2.86u (es − ea ) – Potential evaporation from an open
water surface
WMO WMO (1966) ETo = (1.298 + 0.934u) (es − ea ) – Potential evaporation from an open
water surface
Rs Tmax
Abtew (Ab) Abtew (1996) ETo = 0.01786 – Potential evapotranspiration from a

grass surface
 Rs
Makkink (Mk) Makkink (1957) ETo = 0.61 − 0.12 – Potential evapotranspiration from a
+ 

grass surface
at = 1 RH ≥ 50%
Tat
Turc (Tu) Turc (1961); Xu et al. (2008) ETo = (0.3107Rs + 0.65) Potential evapotranspiration from a
T + 15 50 − RH
at = 1 + RH < 50% grass surface
70
n
Blaney-Criddle (BC) Blaney and Criddle (1950) ETo = a + bP (0.46T + 8.13) (1 + 0.0001H) a = 0.0043RHmin − − 1.41 Potential evapotranspiration from a
N
n grass surface
b = 0.82 − 00.41RHmin + 1.07
N
n
+0.66u − .006RHmin
N

Modified Hargreaves-Samani 1 (MHS1)
Modified Hargreaves-Samani 2 (MHS2)
Droogers and Allen (2002)
Trajkovic (2007)
−0.0006RHmin × u
ETo = 0.0005304Ra (Tmax − Tmin − 0.0123R)0.76 (T + 17) – Potential evapotranspiration from a
grass surface
ETo = 0.0009384Ra (Tmax − Tmin )0.424 (T + 17.8) – Potential evapotranspiration from a
grass surface

ETo is the potential evapotranspiration (mm day−1 ), Rn is the net radiation (MJ/m2 /day), G is the soil heat flux (MJ/m2 /day), ␥ is the psychrometric constant (kPa/ ◦ C), es is the saturation vapor pressure (kPa), ea is the actual vapor
pressure (kPa),  is the slope of the saturation vapor pressure–temperature curve (kPa/ ◦ C), T is the average daily air temperature ( ◦ C), u is the mean daily wind speed at 2 m (m/s), H is the elevation (m), ␸ is the latitude (rad),
Tmin is the minimum air temperature ( ◦ C), Tmax is the maximum air temperature ( ◦ C), RH is the average relative humidity (%), RHmin is the minimum relative humidity (%), P is the mean annual percentage of daytime hours that

can be obtained from Doorenbos and Pruitt (1977), n is the actual duration of sunshine (hr), Rs is the solar radiation (MJ/m2 /day), Ra is the extraterrestrial radiation (MJ/m2 /day), ␭ is the latent heat of vaporization 
(MJ/kg), Tmonth2
is the temperature at the end of the period ( ◦ C), Tmonth1 is the temperature at the beginning of the period ( ◦ C) and the solar radiation gaps were filled using the Angstrom equation (Allen et al., 1998): Rs = as + bs Nn Ra where Ra
is the extraterrestrial radiation (MJm−2 day−1 ), n is the actual duration of sunshine (h), N is the maximum possible duration of sunshine or daylight hours (h), as is the regression constant, expressing the fraction of extraterrestrial
radiation reaching the earth on overcast days (n = 0) and as + bs is the fraction of extraterrestrial radiation reaching the earth on clear days (n = N).

53
54 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60

Fig. 1. The methods that presented the best performance for each station compared to FPM model, HS indicates Hargreaves−Samani, JH indicates Jensen−Haise, Ab
indicates Abtew, Mk indicates Makkink, Tu indicates Turc, BC indicates Blaney−Criddle, MHS1 indicates Modified Hargreaves−Samani 1, and MHS2 indicates Modified
Hargreaves−Samani 2.
 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60 55

Fig. 2. Temporal analysis of error of the methods that presented the best performance for each station, HS indicates Hargreaves−Samani, JH indicates Jensen−Haise, Ab
indicates Abtew, Mk indicates Makkink, Tu indicates Turc, BC indicates Blaney−Criddle, MHS1 indicates Modified Hargreaves−Samani 1, and MHS2 indicates Modified
Hargreaves−Samani 2.
56 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60

Fig. 3. The error (millimetre per day) of radiation− (R), temperature− (T), and mass transfer− (MT) based models in each station, red colour indicates arid regions, yellow
colour indicates semiarid region, blue colour indicates Mediterranean region, and green colour indicates very humid region. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.).

respect to the RMSE is similar to the more accurate model with
respect to the error of the warm (June−September) and cold
(November−February) seasons. Therefore, separation of the ETo
estimating of cold and warm seasons from other seasons leads to
increasing accuracy of the models (the PVC and LVC less than 20%
with the exception of Mediterranean climate).
ETo in arid, semiarid, (see also these papers for arid and semi-
arid region: Sabziparvar and Tabari, 2010; Shiri et al., 2014; Tabari
et al., 2013) and Mediterranean regions and whole of Iran (average
RMSE < 2.00 mm day−1 , average RMSE = 1.18 mm day−1 , average
RMSE = 1.37 mm day−1 , and average RMSE = 1.52 mm day−1 respec-
tively). The temperature−based models are more accurate only in
very humid region (average RMSE = 0.55 mm day−1 ).

3.4. The assessment of the results for each climate and whole of
the Iran
The results (Table 3) appears that the BC
(RMSE = 1.32 mm day−1 ) and Ab (RMSE = 0.83 mm day−1 ) are
the best model for estimating the ETo in arid and semiarid regions,
respectively. While, the MHS2 represents the best performance
in very humid and Mediterranean regions (RMSE = 0.30 mm day−1
and 0.68 mm day−1 , respectively).
However, the PVC and LVC were calculated in the warm
(June−September) and cold (November−February) seasons were
approximately more than 30% and 40%, respectively, for these
best models. Therefore, the best method to estimate the ETo in
warm and cold seasons should be determined for each region
(not climate) separately (Fig. 4). As a result, the radiation−based
models are more accurate than the other models to estimate the
3.5. flowchart step by step to select the best method for
estimating the ETo
Finally, Fig. 5 presents a flowchart step by step to select the best
method for estimating the ETo with respect to values of warm and
cold seasons and climate change.
In the first step, user should determine the climate of area. A list
of 10 more suitable models (with their priority) has been suggested
for each climate with respect to the results of the evaluation indices.
In the next step, the values of error in both warm and cold seasons
are analyzed. The next step in this condition (an also if there is no
considerable peak/low error) is evaluation of the climate change
alarms for the region. If there are no climate change alarms or the
selected model well adapted with the variations, the selected model
 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60 57

Fig. 4. Comparison of the models that presented the best performance for each station with respect to RMSE (column bars 2 and 4) with the models that presented the best
performance for each station with respect to peak (PVC) and low (LVC) errors (column bars 1 and 3), column bars 1 and 2 show PVC and column bars 3 and 4 show LVC.
HS indicates Hargreaves−Samani, JH indicates Jensen−Haise, PT indicates Priestley−Taylor, Tr indicates Trabert, Ma indicates Mahringer, Ab indicates Abtew, Mk indicates
Makkink, Tu indicates Turc, BC indicates Blaney−Criddle, MHS1 indicates Modified Hargreaves−Samani 1, and MHS2 indicates Modified Hargreaves−Samani 2.

is the best model for estimating the ETo in this region with respect
to values of cold and warm seasons and climate change.
Table 2 represents that the HS, Ab, Mk, Tu, BC, MHS1, and MHS2
have the same definition used for the FPM (potential evapotranspi-
ration from a grass surface). All of these 7 models have the highest
accuracy to estimate the ETo in various regions (see the Figures and
Table 3). In addition, the JH and PT models have similar conditions
(use of alfalfa for the JH and pasture for the PT). These two meth-
ods have fairly accuracy (more than mass transfer-based models
and less than the 7 mentioned models). However, mass transfer-
based models have different definition compared with the FPM.
The Tr, WMO, Ma, and Iv reflect evaporation from an open water
surface and the Pe reflects evaporation from an open water sur-
face and bare soil. This also justify clearly poor performance of
mass transfer-based models compared with the other methods in
this study. Meanwhile, the Th reflects potential evapotranspira-
tion which have the worst performance among temperature-based
models (Table 3). Therefore, the priority that supports both the
definitions of the equations and results is:
Priority 1: BC, MHS2, MHS1, and HS (temperature-based mod-
els) as well as Ab, Mk, and Tu (radiation-based models)
Priority 2: JH and PT (radiation-based models)
Priority 3: WMO, Iv, Ma, Tr, and Pe (mass transfer-based models)
as well as Th
4. Discussion
4.1. The variations of the ETo with climate change
As both the Pe and Tr are mass transfer−based models and wind
speed and saturated vapor pressure deficit are the main parameters
of them, the abnormal behaviour of these models can alarms cli-
mate change for Rasht (Ashraf et al., 2014; Molavi-Arabshahi et al.,
2015; Mosaedi et al., 2015).
Although a climate change may also be anticipated for 12 regions
(all of the arid regions and most of the semiarid regions) due to a
considerable difference between the values of the error calculated
for the first and second halves of the study periods, these differences
are increasing only for Bushehr, Jiroft, Shiraz, and Zabol. The climate
of three of these four regions is arid and the best models for esti-
mating the ETo in this regions are three temperature−based models
and one radiation−based models. Thus, increase of temperature
may be an affective factor which leads to climate change in these
regions (Banimahd et al., 2014; Esmaeili, 2009; Kousari et al., 2013).
 58
Table 3
Evaluation indices obtained for each model according to different climates and their averages (for whole of Iran), HS indicates Hargreaves−Samani, JH indicates Jensen−Haise, Pe indicates Penman, PT indicates Priestley−Taylor, Th
indicates Thornthwaite, Tr indicates Trabert, Iv indicates Ivanov, Ma indicates Mahringer, Ab indicates Abtew, Mk indicates Makkink, Tu indicates Turc, BC indicates Blaney−Criddle, MHS1 indicates Modified Hargreaves−Samani
1, and MHS2 indicates Modified Hargreaves−Samani 2.

Climate Arid Semiarid Very humid Mediterranean Iran

Model RMSE PVC LVC t-test RMSE PVC LVC t-test RMSE PVC LVC t-test RMSE PVC LVC t-test RMSE PVC LVC t-test
Ab 1.53 35.93 47.97 −0.48 0.83 34.85 42.35 −0.39 1.75 59.99 63.36 −1.80 1.02 35.60 69.88 −0.72 1.28 41.59 55.89 −0.85
BC 1.32 38.49 41.96 −0.65 0.95 38.13 47.43 −0.51 0.40 25.97 46.74 −0.14 1.36 42.88 75.87 −0.78 1.01 36.37 53.00 −0.52
HS 2.14 45.12 51.69 0.19 1.85 51.21 60.70 −0.61 0.61 11.55 68.36 −0.78 3.56 72.13 97.68 −1.54 2.04 45.00 69.61 −0.69
Iv 4.14 62.41 89.15 −2.49a 2.36 54.92 86.70 −1.43 0.75 40.86 36.26 0.76 3.37 66.92 90.02 −1.74 2.65 56.28 75.53 −1.23
JH 2.37 54.03 46.77 −0.80 1.83 54.81 48.35 −0.80 1.09 50.03 47.69 −0.76 2.44 59.98 58.60 −1.10 1.93 54.71 50.35 −0.87
Ma 5.96 85.79 47.76 −2.07 2.34 60.65 40.22 −0.99 1.09 48.67 49.19 1.57 2.96 66.64 48.54 −1.23 3.09 65.44 46.43 −0.68
MHS1 1.61 38.57 48.55 0.08 1.10 41.19 52.07 −0.55 0.42 18.15 28.80 −0.34 1.41 42.62 83.68 −0.98 1.13 35.13 53.28 −0.45
MHS2 2.16 46.03 45.50 1.57 0.96 35.39 35.41 0.76 0.30 27.67 33.24 −0.09 0.68 29.77 49.08 0.52 1.03 34.71 40.81 0.69
Mk 1.82 41.35 47.64 0.84 0.90 29.95 59.25 −0.09 1.34 37.18 99.52 −1.88 0.74 30.12 70.53 −0.06 1.20 34.65 69.24 −0.30
−3.06a −1.90 −2.02 −1.63

M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60

Pe 9.52 96.84 80.82 4.74 86.92 71.28 0.55 28.79 30.98 0.47 5.32 88.80 84.22 5.53 75.34 66.82
PT 2.41 48.17 51.06 1.91 1.14 37.92 52.14 1.04 0.50 33.87 18.93 −0.46 1.19 39.63 46.62 1.13 1.31 39.90 42.19 0.90
Th 4.02 57.21 74.40 0.52 1.59 38.76 70.47 1.61 0.68 32.65 62.40 −0.21 1.51 35.31 64.67 1.54 1.95 40.98 67.98 0.87
Tr 6.67 89.55 51.76 −2.26a 3.21 69.72 48.63 −1.29 0.96 44.95 45.48 1.32 3.47 72.13 55.54 −1.43 3.58 69.09 50.36 −0.91
Tu 1.83 39.39 55.84 −0.99 1.20 38.66 54.85 −0.88 1.96 53.34 92.13 −2.55a 1.45 44.06 69.59 −1.02 1.61 43.86 70.60 −1.36
WMO 4.71 72.67 32.40 −1.38 1.46 41.84 31.75 −0.27 1.49 58.65 58.88 2.34a 1.61 48.55 26.09 −0.50 2.32 55.43 37.28 0.05

The bolds indicate the minimum values for each index.

a
Significant at the confidence level 95%.
MHS2 indicates Modified Hargreaves−Samani 2.
BC indicates Blaney−Criddle, MHS1 indicates Modified Hargreaves−Samani 1, and
indicates Ivanov, Ab indicates Abtew, Mk indicates Makkink, Tu indicates Turc,
PT indicates Priestley−Taylor, Th indicates Thornthwaite, Tr indicates Trabert, Iv
HS indicates Hargreaves−Samani, JH indicates Jensen−Haise, Pe indicates Penman,
to cold and warm seasons values, climate change, and lysimeter measurements,
Fig. 5. A flowchart to select the best model for estimation of ETo with respect
 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60
However, the other best models well adapted with climate change
conditions occurred in the regions. There is also some climate vari-
ability. For instance, an underestimation is observable for the Ab
in Yazd in the first 12−year, then an overestimation has occurred
during the next 25−year and finally an underestimation is observ-
able again in the last 13−year. In addition, an underestimation is
observable for the Ab in Kerman in the first 15−year while adjusted
in the next years. As a result, the radiation−based models were
adapted with climate change better than the temperature−based
and particularly mass transfer−based models. This underlines the
role of sunshine and other meteorological variables to select the
best model in each region.
The PT is the best model for estimating the ETo in Moghan with
respect to climate change alarm. However, the HS is the more accu-
rate model to estimate the ETo in Moghan with respect to the
values of the evaluation indices. However, a climate change may be
observed in recent decades that if this trend be continued in future,
the selection of the best model to estimate the ETo , only based on
the results of the evaluation indices, may leads to increasing the
error in the future and is not reasonable.
The Mk and WMO are more suitable for the regions in which the
climate change leads to increasing wind speed and consequently
decreasing both temperature and minimum relative humidity. In
addition, increasing temperature (mean, maximum, and minimum)
and decreasing minimum relative humidity for the best perfor-
mance of the most models also alarms a climate change in the most
regions of Iran.
4.2. Sources of error
In this study, the average monthly meteorological data were
used only in the calculation (see Section 2. Materials and Methods)
it is hard to intercompare the results with other results obtained
in similar conditions. In addition, the performance of the spe-
cific equation depends commonly on calculation time steps, for
instance, the BC method was recommended for periods of one
month or longer, while the FPM methods was considered to offer
the best results with minimum possible error for hourly or daily
calculations as the FPM equation is a close, simple representation
of the physical and physiological factors governing the evapotran-
spiration process, reflecting the impact of feedback mechanisms
between the vegetation and overlying boundary layer. This is why
the FPM equation is maintained as a standard method for the
computation of potential evapotranspiration from meteorological
data. So, in this study, the authors calculated potential evapotran-
spiration using daily FPM equation (the formula that the authors
presented in this manuscript is the daily FAO−PM form), and adopt-
ing the average monthly meteorological data as input. This estimate
would result in unneglectable error.
Although there are a lot of studies in which only average
monthly data have been applied for estimating the ETo in Iran
(Sabziparvar and Tabari, 2010; Tabari et al., 2013; Valipour, 2015d),
however it is better to compare daily and monthly scales to address
this question that; is there any difference in the model performance
after using daily or monthly meteorological data? Because if there
is a considerable difference between the ETo estimated by the daily
meteorological data and those obtained for the monthly meteoro-
logical data, the average monthly data will not give reliable results
to predict the ETo in future.
5. Conclusion
The spatiotemporal analysis of the ETo appears that the selecting
the most accurate model to estimate the ETo needs to high sensitiv-
ity. Because the various factors control the accuracy of the empirical
59
models developed for the estimation of the ETo . If the user aims to
select a superior model with the lowest error and/or the highest
degree of reliability to predict possible climate change, a compre-
hensive assessment of the variations of the evaluation indices and
meteorological parameters is necessary for each area in both the
warm and cold seasons and during different years/decades. The best
fitness between the values calculated by the FPM and estimated by
the best empirical models belongs to arid and very humid regions
compared to Mediterranean and particularly semiarid regions. The
selection of the best model to estimate the ETo , only based on
the results of the evaluation indices, may leads to increasing the
error in the future and is not reasonable. Increasing temperature
(mean, maximum, and minimum) and decreasing minimum rela-
tive humidity for the best performance of the most models alarms
a climate change in the most regions of Iran. The accuracy of the
empirical models, in the northern and north−western Iran (with
lowest rate of the ETo ) is higher than other areas of Iran. While, the
models have not accurately resulted in eastern and south−eastern
Iran with arid climate (the highest rate of the ETo ).
References
Abtew, W., 1996. Evapotranspiration measurements and methoding for three
wetland systems in South Florida. J. Am. Water Resour. Assoc. 32 (3), 465–473.
Ahmadi, S.H., Fooladmand, H.R., 2008. Spatially distributed monthly reference
evapotranspiration derived from the calibration of Thornthwaite equation: a
case study, South of Iran. Irrig. Sci. 26 (4), 303–312.
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop
Evapotranspiration-Guidelines for Computing Crop Water Requirements-FAO
Irrigation and Drainage Paper 56, 300. FAO, Rome, pp. 6541.
Azhar, A.H., Perera, B.J.C., 2011. Evaluation of reference evapotranspiration
estimation methods under Southeast Australian conditions. J. Irrig. Drain. Eng.
137 (5), 268–279.
Ashraf, B., Yazdani, R., Mousavi-Baygi, M., Bannayan, M., 2014. Investigation of
temporal and spatial climate variability and aridity of Iran. Theor. Appl.
Climatol. 118 (1–2), 35–46.
Banimahd, S.A., Khalili, D., Kamgar-Haghighi, A.A., Zand-Parsa, S., 2014. In-depth
investigation of precipitation-based climate change and cyclic variation in
different climatic zones. Theor. Appl. Climatol. 116 (3-4), 565–583.
Blaney, H.F., Criddle, W.D., 1950. Determining Water Requirements in Irrigated
Areas from Climatological and Irrigation Data. Soil Conservation Service
Technical Paper 96, Soil Conservation Service. US Department of Agriculture,
Washington, USA.
Caporusso, N.B., Rolim, G.D.S., 2015. Reference evapotranspiration models using
different time scales in the Jaboticabal region of São Paulo, Brazil. Acta Sci.
Agron. 37 (1), 1–9.
Doorenbos, J., Pruitt, W.O., 1977. Crop water requeriments. Rome: FAO, 1977.
179p. Irrigation and Drainage Paper, 24.
Droogers, P., Allen, R.G., 2002. Estimating reference evapotranspiration under
inprecise data conditions. Irrig. Drain. Syst. 16, 33–45.
El-Shafie, A., Abdin, A.E., Noureldin, A., Taha, M.R., 2009. Enhancing inflow
forecasting model at Aswan high dam utilizing radial basis neural network and
upstream monitoring stations measurements. Water Resour. Manage. 23 (11),
2289–2315.
El-Shafie, A., Najah, A., Alsulami, H.M., Jahanbani, H., 2014. Optimized neural
network prediction model for potential evapotranspiration utilizing ensemble
procedure. Water Resour. Manage. 28 (4), 947–967.
Esmaeili, A., 2009. Environmental change and fishery management in the northern
Persian Gulf. J. Environ. Plann. Manage. 52 (8), 1071–1081.
Hargreaves, G.L., Samani, Z.A., 1985. Reference crop evapotranspiration from
temperature. Appl. Eng. Agric. 1 (2), 96–99.
Heydari, M.M., Tajamoli, A., Ghoreishi, S.H., Darbe-Esfahani, M.K., Gilasi, H., 2015.
Evaluation and calibration of Blaney–Criddle equation for estimating reference
evapotranspiration in semiarid and arid regions. Environ. Earth Sci. 74 (5),
4053–4063.
Jensen, M.E., Haise, H.R., 1963. Estimation of evapotranspiration from solar
radiation. J. Irrig. Drain. Div. 89, 15–41.
Kisi, O., Sanikhani, H., Zounemat-Kermani, M., Niazi, F., 2015. Long-term monthly
evapotranspiration modeling by several data-driven methods without climatic
data. Comput. Electron. Agric. 115, 66–77.
Kousari, M.R., Ahani, H., Hendi-Zadeh, R., 2013. Temporal and spatial trend
detection of maximum air temperature in Iran during 1960–2005. Global
Planet. Change 111, 97–110.
Mahringer, W., 1970. Verdunstungsstudien am Neusiedler See. Arch. Meteorol.
Geophys. Bioclimatol. Ser. B 18, 1–20.
Makkink, G.F., 1957. Testing the Penman formula by means of lysimeters. J. Inst.
Water Eng. 11, 277–288.
Molavi-Arabshahi, M., Arpe, K., Leroy, S.A.G., 2015. Precipitation and temperature
of the southwest Caspian Sea region during the last 55 years: their trends and
60 M. Valipour et al. / Agricultural Water Management 180 (2017) 50–60
teleconnections with large-scale atmospheric phenomena. Int. J. Climatol,
http://dx.doi.org/10.1002/joc.4483, in press (Assessed date: 14.09.15).
Mosaedi, A., Abyaneh, H.Z., Sough, M.G., Samadi, S.Z., 2015. Quantifying changes in
reconnaissance drought index using equiprobability transformation function.
Water Resour. Manage. 29 (8), 2451–2469.
Ngongondo, C., Xu, C.Y., Tallaksen, L.M., Alemaw, B., 2013. Evaluation of the FAO
Penman-Montheith, Priestly-Taylor and Hargreaves models for estimating
reference evapotranspiration in southern Malawi. Hydrol. Res. 44, 706–722.
Patel, J., Patel, H., Bhatt, C., 2015. Modified hargreaves equation for accurate
estimation of evapotranspiration of diverse climate locations in India. Proc.
Nat. Acad. Sci. India Sect. B: Biol. Sci. 85 (1), 161–166.
Penman, H.C., 1948. Natural evaporation from open water, bare soil and grass.
Proc. R. Soc. London Ser. A 193, 120–145.
Priestley, C.H.B., Taylor, R.J., 1972. On the assessment of surface heat flux and
evaporation using large-scale parameters. Mon. Weather Rev. 100 (2), 81–92.
Rahimi, S., Gholami Sefidkouhi, M.A., Raeini-Sarjaz, M., Valipour, M., 2015.
Estimation of actual evapotranspiration by using MODIS images (a case study:
Tajan catchment). Arch. Agron. Soil Sci. 61 (5), 695–709.
Romanenko, V.A., 1961. Computation of the autumn soil moisture using a
universal relationship for a large area. In: Proceedings, Ukrainian
Hydrometeorological Research Institute, No. 3. Kiev.
Roshan, G., Yousefi, R., Fitchett, J.M., 2015. Long-term trends in tourism climate
index scores for 40 stations across Iran: the role of climate change and
influence on tourism sustainability. Int. J. Biometeorol, http://dx.doi.org/10.
1007/s00484-015-1003-0, In press (Assessed date: 05.15).
Sabziparvar, A.A., Tabari, H., 2010. Regional estimation of reference
evapotranspiration in arid and semiarid regions. J. Irrig. Drain. Eng. 136 (10),
724–731.
Shiri, J., Nazemi, A.H., Sadraddini, A.A., Landeras, G., Kisi, O., Fard, A.F., Marti, P.,
2014. Comparison of heuristic and empirical approaches for estimating
reference evapotranspiration from limited inputs in Iran. Comput. Electron.
Agric. 108, 230–241.
Tabari, H., Grismer, M.E., Trajkovic, S., 2013. Comparative analysis of 31 reference
evapotranspiration methods under humid conditions. Irrig. Sci. 31 (2),
107–117.
Thornthwaite, C.W., 1948. An approach toward a rational classification of climate.
Geog. Rev. 38, 55–94.
Trabert, W., 1896. Neue Beobachtungen uber Verdampfungsgeschwindigkeiten.
Meteorol. Z. 13, 261–263.
Trajkovic, S., 2007. Hargreaves versus Penman—Monteith under humid condition.
J. Irrig. Drain. Eng. 133, 38–42.
Turc, L., 1961. Estimation of irrigation water requirements, potential
evapotranspiration: a simple climatic formula evolved up to date. Ann. Agron.
12, 13–49.
Valipour, M., 2012a. Comparison of surface irrigation simulation models: Full
hydrodynamic, zero inertia, kinematic wave. J. Agric. Sci. 4 (12), 68–74.
Valipour, M., 2012b. Critical Areas of Iran for Agriculture Water Management
According to the Annual Rainfall. Eur. J. Sci. Res. 84 (4), 600–608.
Valipour, M., 2012c. A Comparison between horizontal and vertical drainage
systems (include pipe drainage, open ditch drainage, and pumped wells) in
anisotropic soils. IOSR J. Mechanic. Civil Eng (IOSR-JMCE) 4 (1), 7–12, http://dx.
doi.org/10.9790/1684-0410712.
Valipour, M., 2012d. Ability of Box-Jenkins models to estimate of reference
potential evapotranspiration (a case study: Mehrabad synoptic station, Tehran,
Iran). IOSR J. Agric. Veterin. Sci. (IOSR-JAVS) 1 (5), 1–11, http://dx.doi.org/10.
9790/2380-0150111.
Valipour, M., 2012e. Hydro-module determination for Vanaei village in Eslam Abad
Gharb, Iran. J. Arpn J. Agric. Biol. Sci. 7 (12), 968–976.
Valipour, M., 2013a. Increasing irrigation efficiency by management strategies:
cutback and surge irrigation. Arpn J. Agric. Biol. Sci. 8 (1), 35–43.
Valipour, M., 2013b. Necessity of irrigated and rainfed agriculture in the world.
Irrig. Drain. Syst. Eng., e001, http://omicsgroup.org/journals/necessity-of-
irrigated-and-rainfed-agriculture-in-the-world-2168-9768.S9-e001.
php?aid=12800.
Valipour, M., 2013c. Use of surface water supply index to assessing of water
resources management in colorado and Oregon, US. Adv. Agric. Sci. Eng. Res. 3
(2), 631–640, http://vali-pour.webs.com/13.pdf.
Valipour, M., Eslamian, S., 2014. Analysis of potential evapotranspiration using 11
modified temperature-based models. Int. J. Hydrol. Sci Technol. 4 (3), 192–207.
Valipour, M., 2014. Use of average data of 181 synoptic stations for estimation of
reference crop evapotranspiration by temperature-based methods. Water
Resour. Manage. 28 (12), 4237–4255.
Valipour, M., 2015a. Temperature analysis of reference evapotranspiration models.
Meteorol. Appl. 22 (3), 385–394.
Valipour, M., 2015b. Investigation of Valiantzas’ evapotranspiration equation in
Iran. Theor. Appl. Climatol. 121 (1–2), 267–278.
Valipour, M., 2015c. Calibration of mass transfer-based models to predict reference
crop evapotraspiration. Appl. Water Sci, http://dx.doi.org/10.1007/s13201-
015-0274-2, In press (Assessed date: 05.13.15).
Valipour, M., 2015d. Comparative evaluation of radiation-based methods for
estimation of potential evapotranspiration. J. Hydrol. Eng. 20 (5), 04014068.
Valipour, M., 2015e. Analysis of potential evapotranspiration using limited weather
data. Appl. Water Sci., http://dx.doi.org/10.1007/s13201-014-0234-2, In Press.
Valipour, M., 2015f. Evaluation of radiation methods to study potential
evapotranspiration of 31 provinces. Meteorol. Atmos. Physic. 127 (3), 289–303.
Valipour, M., 2015g. Long-term runoff study using SARIMA and ARIMA models in
the United States. Meteorol. Appl. 22 (3), 592–598.
Valipour, M., 2016a. How much meteorological information is necessary to achieve
reliable accuracy for rainfall estimations? Agriculture 6 (4), 53.
Valipour, M., 2016b. In: Peter, K.V. (Ed.), Hydro-climatology for Arid Horticulture
Crops: A Phonological Approach. Innovations in Horticultural Sciences. New
India Publishing Agency, New Delhi, India, pp. 497–529.
WMO, 1966. Measurement and estimation of evaporation and evapotranspiration.
Tech. Pap. (CIMO−Rep) 83. Genf.
Wright, J.L., Jensen, M.E., 1972. Peak water requirements of crops in southern
Idaho. J. Irrig. Drain. Div. 98, 193–201.
Xu, C.Y., Singh, V.P., Chen, Y.D., Chen, D., 2008. Evaporation and evapotranspiration.
In: Singh VP (ed) Hydrology and hydraulics, 1 st edn. Water Resources Pubns,
USA, pp 229–276.
Yannopoulos, S.I., Lyberatos, G., Theodossiou, N., Li, W., Valipour, M., Tamburrino,
A., Angelakis, A.N., 2015. Evolution of water lifting devices (Pumps) over the
centuries worldwide. Water 7 (9), 5031–5060.