Journal of Agrometeorology 21 (3) : 351-356 (September 2019)

Assessing suitability of temperature-based reference evapotranspiration methods for

semi-arid basin of Maharashtra
1* 2 3
P.S. WABLE , M.K. JHA and S.D. GORANTIWAR
1
International Crop Research Institute for the Semi-Arid Tropics, Hyderabad, India
2
Agricultural and Food Engineering Department, Indian Institute of Technology Kharagpur, India
3
Dr. A. S. College of Agricultural Engineering, Mahatma Phule Krishi Vidyapeeth, Rahuri, India
*Corresponding author email:[email protected]

ABSTRACT
FAO Penman-Monteith (FAO-PM) is deemed as a sole standard method for estimating reference
evapotranspiration (ETo). However, limited availability of meteorological data at spatial and temporal
scales restricts the application of this method. To address this issue, the FAO 56 experts suggested three
methods when only maximum and minimum temperature data are available: (i) Temperature-based
Penman-Monteith (PMT-1) method wherein Tdew ≈ Tmin (ii) PMT-2 wherein Tdew ≈ Tmin-2.5, and (iii)
Hargreaves method. These ETo methods were assessed for a semi-arid basin of Western India which
lacks adequate climatic data. The performances of the ETo methods were evaluated against the standard
FAO-PM method using salient statistical and graphical indicators, together with the sensitivity analysis.
The results of the three temperature-based methods had a tendency of over-predication of ETo in the
study area. The PMT-1 method, however, provided superior ETo estimates compared to PMT-2 and
Hargreaves methods. For estimating monthly ETo, the FAO-PM method was most sensitive to
temperature. Further, ETo of the monsoon season over the study area increased from 5 to 12% during
'drought' years compared to 'normal' years. It was concluded that PMT-1 method is the most suitable
temperature-based method for estimating ETo in semi-arid regions under limited climatic condition.

Keywords: FAO Penman-Monteith method, temperature-based Penman-Monteith method, Hargreaves method, ET

mapping, limited data, semi-arid region

Evapotranspiration (ET) is one of the important (Thornthwaite and Hargreaves, etc.), radiation-based methods
components of the water cycle and plays a central role in water (Priestley-Taylor, and FAO-24 radiation, etc.), pan methods
balance studies, hydrological modelling, planning and (Christiansen Pan and FAO-24 pan), and combination
management of irrigation and drainage systems as well as in methods (Penman-Monteith, and FAO-24 corrected Penman,
climatological studies. Crop Evapotranspiration (ETc) is etc.). Among these methods, FAO Penman-Monteith (FAO-
affected by the crop characteristics, soil factors and climatic PM) has emerged as the sole standard method for ETo
parameters, and management practices leading to a high computation (Allen et al., 1998). Also, it can be applied in
spatial and temporal variability in ETc (Allen et al., 1998). The data-scarce situations.
concept of reference evapotranspiration (ETo) evolved in the
Unfortunately, the application of FAO-PM method
last decade of the 20th century to study the evaporative demand
requires several meteorological data, some of them are
of the atmosphere independently of crop type, crop
missing or limited in spatial and temporal scales mostly in
development and management practices. ETo is a climate
developing countries. According to the modern guidelines for
parameter because it is only affected by climatic factors. The
computing crop water requirements (Allen et al., 1998), the
best estimates of ETo is done by using lysimeter but it is
following three methods can be used under data-scarce
cumbersome, time consuming, costly, and its application at a
conditions: (i) Temperature-based Penman-Monteith (PMT-1)
larger scale is restricted. To overcome this problem, numerous
method, wherein Tdew ≈ Tmin; (ii) PMT-2, wherein Tdew ≈ Tmin-2.5;
empirical methods have been developed during past six
and (iii) Hargreaves method. Among the less data-intensive
decades for estimating ETo indirectly under varying agro-
ETo methods used, temperature-based Hargreaves method has
climatic conditions (Tabari et al., 2013; Jadhav et al., 2015;
been found to be the best in arid and/or semi-arid climate of
Phad et al., 2019). These methods can be classified into four
Spain (Lopez-Urrea et al., 2006), Iran (Tabari, 2010,
groups (Jensen, 1990) of temperature-based methods
Sabziparvar and Tabari, 2010), and India (Nandagiri and
352 Temperature-based reference evapotranspiration methods for Maharashtra September 2019

Kovoor, 2006; Meshram et al., 2013). In the recent past, a few The meteorological data of Solapur station were collected
researchers have evaluated the performance of temperature- from the India Meteorology Department, Pune. Temperature
based Penman-Monteith and Hargreaves methods and data of Kashti and Rosa meteorological stations were
reported contradictory results. In the semi-arid climate of collected from the State Data Storage Centre, Nashik, and that
Tunisia, the Hargreaves method outperformed compared to for the Rahuri meteorological station was acquired from
the PMT-1 method (Jabloun and Sahli, 2008). Sentelhas et al. Mahatma Phule Agricultural University, Rahuri.
(2010) reported that adjusted Hargreaves method can be
Estimation of ETo
preferred over the PMT-1 method in Canada. In semi-arid
climate of Mediterranean countries (Todorovic et al., 2013) FAO-PM method
both PMT-1 and PMT-2 methods performed better than the The FAO Penman-Monteith (FAO-PM) equation was
Hargreaves. Based on the review, it can be concluded that the developed to describe ET of a reference grass crop, which is
comparative evaluation of ETo estimation methods has defined as the potential rate of evapotranspiration from a
attracted the attention of researchers and also only few studies hypothetical crop with an assumed fixed height (120 mm),
(Nandagiri and Kovoor, 2006; Meshram et al., 2013) have surface resistance (70 s.m-1) and albedo (0.23), closely
been conducted in the Indian subcontinent related to the resembling the evapotranspiration from an extensive surface
performance evaluation of ETo methods under limited data of a disease-free green grass cover of uniform height, actively
conditions, particularly in arid regions. Considering this growing, completely shading the ground, and with adequate
research gap and an increasing incidence of drought water and nutrient supply (Allen et al., 1998).
(hydrological disaster) in several parts of India in general and
Maharashtra in particular, the present study was conceived The FAO-PM equation for the calculation of ETo takes the
taking Sina River basin as study area. It suffers from severe form:
water scarcity due to frequent droughts. Therefore, this study
was carried out with the objectives to: (i) explore the efficacy
of three methods for one meteorological station having all
desirable climatic data and to find out a suitable method for Where ETₒ = Reference evapotranspiration, mm day-1,
other weather stations of the basin having limited data, (ii) Rn = Net radiation at crop surface, MJm-2 day-1, G = Soil heat
quantify the sensitivity of ETₒ estimates to the meteorological flux density, MJm-2 day-1, T = Mean daily air temperature at 2
parameters, and (iii) analyse spatio-temporal variation of ETₒ m height, oC, u2 = Wind speed at 2 m height, m s-1, es =
over the basin. Saturation vapour pressure, kPa, ea = Actual vapour pressure,
kPa, es– ea = Saturation vapour pressure deficit, kPa, Δ = Slope
MATERIALS AND METHODS
of vapour pressure curve, kPa oC-1, and g= Psychometric
Study area and data collection constant, kPa oC-1.
Sina river basin (study area) is located between 17° 28′ Estimation of ETo under data-scarce condition
N and 19° 16′ N latitude, 74° 28′ E and 76° 7′ E longitude. The
As mentioned earlier, limited climatic data were
study area is of 12,707 km2 area, with an elevation ranging
available for Kashti, Rosa, and Rahuri meterological stations
from 420 to 964 m MSL. More information regarding the
in the study area. Hence, ETo for these stations were estimated
study area can be gathered from Wable and Jha (2017). The
in this study by using the following temperature-based
data used in this study were collected from various
methods mentioned below.
government organizations. Fig. 1 shows the locations of the
meteorological stations. Out of the four meteorological (a) PMT method
stations (Kashti, Rahuri, Rosa and Solapur), only Solapur
FAO Penman-Monteith method requires a number of
station records all the meteorological data (maximum and
weather parameters for ETo calculation, which were not
minimum temperatures, wind speed, relative humidity and
available for all stations. Hence, the recommendation of the
sunshine hour) on a daily basis required for the FAO-PM
procedure made by Allen et al. (1998) to estimate missing
method for the period 2001-2005. Unfortunately, only
climatic parameters for the data sets (i.e., data of Kashti,
temperature (minimum and maximum) data on a daily time
Rahuri, Solapur and Rosa stations) containing at least
scale were available at all stations for the period 1985-2009.
maximum and minimum temperatures was followed to use
Vol. 21, No. 3 WABLE et al. 353

the FAO-PM method. conditions by the three methods (PMT-1, PMT-2 and
Hargreaves) were compared with the FAO-PM method. The
Radiation under data scared environment is calculated
performance evaluation of these ETo estimation methods was
by Eq.2. This relationship is known as Hargreaves' radiation
done using five statistical indicators, viz., Mean Absolute
formula (Allen et al., 1998), and is expressed as follows:
Error (MAE), Root Mean Squared Error (RMSE),
Nash–Sutcliffe efficiency (NSE), Index of Agreement (IOA),
Where, Ra = Extraterrestrial radiation, MJm-2day-1; Tmax and Percent Bias (PBIAS). Apart from these statistical
& Tmin = Maximum and minimum air tempertures, oC, and KRS = indicators, a visual comparison of the ETo estimated by the
Adjustment coefficient (for interior locations, KRS is 0.16). FAO-PM method with the ETo estimated with missing data
procedures was also performed using scatter plots.
Wind speed is estimated as the variation in average
wind speed over monthly periods. FAO 56 suggests its range Sensitivity analysis
from less than 1 m.s-1 (light wind areas) to greater than 5 m.s-1 In this study, the sensitivity analysis of FAO-PM ETo
(strong wind areas). The average value of wind speed at 2 m estimates was carried out on a monthly basis at Solapur station
height for over 2000 weather stations across the globe was during 2001-2005. For this weather station, four key climatic
found to be 2 m s-1 (Allen et al., 1998), which was considered parameters (mean temperature:Tmean, mean relative humidity:
for the Solapur station to evaluate temperature-based FAO- RHmean, wind speed: u2, and sunshine hours: Ssh) were changed
PM methods considered in this study. within ±20% at intervals of 5%. Thereafter, the percentage
For the missing humidity is calculated by using Eq. 3. change from the original ETₒ estimate was calculated, and the
In this case, ea is calculated as (Allen et al., 1998): sensitivity curves were plotted.
ETo mapping
Spatial variation of ETo was analyzed by generating
ETo maps using ArcGIS. Thiessen polygon method was used
The relation Tmin ≈ Tdew (dew pint temperature) holds for ETo mapping because of less number of meteorological
good for well-watered locations. In arid regions, Tmin > Tdew stations in the study area (Maidment, 1992). The analysis of
therefore, Allen et al. (1998) suggested that Tdew may be better annual spatio-temporal variability of ETo estimates over the
approximated by subtracting 2-3oC from Tmin as a correction for study area for the period 1985-2009 was carried out by
arid regions. To verify this suggestion, two cases of ETo estimating ETo at each station by using the best performing
estimation were considered in this study, namely PMT-1 temperature-based method. Further, the spatial variability of
wherein Tdew ≈ Tmin,and PMT-2 wherein Tdew ≈ Tmin ̶ 2.5°C. ETo was mapped using Thiessen polygon method. This
(a) Hargreaves method analysis was performed by considering the ETₒ estimates for
the monsoon season (June to October period), as most of the
As an alternative to the FAO-PM method, Allen et al. crops are grown in this season. As the study area was affected
(1998) suggested to use the Hargreaves method for estimating by drought in 2003 (EIS, 2006) therefore, ETo maps were
ETo under data limited conditions, which requires measured generated for monsoon seasons of 'drought' and 'normal' years.
maximum and minimum temperature data only. The The year 2009 was considered as a representative of normal
Hargreaves equation is given as (Hargreaves and Samani, years.
1985):

RESULTS AND DISCUSSION
Performance of ETo methods

In Eq. (4), the units of both ETₒ and Ra are mm day-1. It
has a tendency to under-predict ETo under high wind Performance evaluation of the three temperature-based
conditions (u2>3 m.s-1), and to over-predict ETo under the methods for estimating ETo relative to standard FAO-PM at
conditions of high relative humidity (Allen et al., 1998). Solapur station for 2001-2005 period was carried out using
five statistical indicators as presented in Table 1. It can be
Performance evaluation of ETo methods
inferred that all five statistical indicators support PMT-1 as a
The results of ETo estimated under limited data best method with lowest values of RMSE (0.70 mm day-1),
MAE (0.59 mm day-1), PBIAS (7.87%) and highest value of
354 Temperature-based reference evapotranspiration methods for Maharashtra September 2019

Table 1: Statistical summary of different ETo methods with reference to FAO-PM on monthly time step for Solapur Station
Sl. No. Method MAE, RMSE, NSE IOA PBIAS
-1 -1
mm day mm day %
1. PMT-1 0.59* 0.70* 0.66* 0.89* -7.87*
# # # #
2. PMT-2 0.80 0.93 0.40 0.84 -15.55#
3. Hargreaves 0.71 0.84 0.51 0.87 -14.07

NSE (0.66) and IOA (0.89). Table 2: Trends of annual ETo at different stations during
1985-2009
Moreover, comparison of the ETo estimated by FAO-
PM and other temperature-based methods was performed Stations Trend equation
using scatter plot at Solapur station (Fig. 1). It is apparent from
Kashti Y = -11.336 X +1925.3
this figure that PMT-1 provides better results compared to
other two temperature based methods, which confirms finding Rahuri Y = 2.465 X +1784.3
of statistical indicators. Among the methods examined, the Rosa Y = 1.745 X +1692.6
performance of PMT-2 method was the worst (Table 1).
Furthermore, negative PBIAS values (Table 1) revealed that Solapur Y = 2.147 X +1808.1
all three less data-intensive methods had tendency to over-
weather parameters. The mean temperature of the station thus
predict monthly ETo estimates with lowest PBIAS value
was the most sensitive parameter to monthly ETo estimation.
(7.87%) for PMT-1 method. It could be inferred that the PMT-
1 method was most suitable for estimating ETo for the study Spatio-temporal variability of ETo
area when only temperature data are available. This finding is
The spatio-temporal variability of ETo during 1985-
consistent with the results obtained by Todorovic et al. (2013)
2009 period was studied using best performed PMT-1 method.
for the semi-arid regions of Mediterranean countries. It should
The annual trends of ETo at four stations are given in Table 2.
be noted that the aridity correction Tdew ≈ Tmin-2.5 suggested by
The trend analysis indicated that ETo was decreasing at rate of
Allen et al. (1998) for arid regions does not hold good for the
11 mm year-1 in Kashti station and it was found to increase at
river basin under study. Also, performance of Hargreaves
Rahuri, Rosa and Solapur stations at the rates of 2.47, 1.75,
method in this study was contrary to that reported in Allen et
2.15 mm year-1, respectively.
al. (1998).
The spatial variation of ETo during monsoon season
Sensitivity analysis
over the study area for normal and drought years using
Monthly ETo for Solapur station was estimated using Thiessen polygon method is shown in Figs. 5(a,b). The
FAO-PM method. At this station, the extent of change in ETo average ETₒ during monsoon season (June-October) varied
with respect to 5% to 20% (increase and decrease) in each from 3.8 mm day-1 to 4.7 mm day-1 in normal years [Fig. 3(a)],
climatic variables (Tmean, RHmean, u2, and Ssh) on monthly time whereas it varied from 4.0 mm day-1 to 5.0 mm day-1 in drought
scale during 2001-2005 were performed and is presented in years [Fig. 3(b)]. In northern, southern and the north-western
Fig. 2. The slope of the sensitivity curves of ETo climatic part of the study area, there was an increase in ETo estimate
variables were positive for Tmean,u2and Ssh, whereas negative for from 4 to 6% during drought years, encompassing about 50%
RHmean. Also, the change in temperature had the highest effect of the study area (6132 km2). On the other hand, the maximum
on ETo ranging from (-) 15% to 16% for (±) 20% variation in increase in ETo for the remaining 50% of the study area
the temperature, followed by relative humidity and sunshine covering the central portion was found to be 12% in drought
hours, both within the range of (±) 6%. Thus, sensitivity years as compared to normal years. This mainly covered
analysis indicated that a change in wind speed had the least Bhoom, Jamkhed, Barshi, Madha, Karmala and Parenda
impact on the ETₒ estimation (±) 5% as compared to other blocks of the study area [Figs. 3(a,b)].
Vol. 21, No. 3 WABLE et al. 355

Fig. 1: Scatter Plots showing the over and under prediction of different ETₒ estimation methods with the FAO-PM for Solapur station

Fig. 2: Percentage change in ETo due to changes in weather

parameters at Solapur weather station on monthly time scale
CONCLUSIONS
This study focuses on the performance evaluation of
three temperature-based ETo methods (PMT-1, PMT-2 and
Hargreaves) with reference to the standard FAO-PM method
under data-scarce conditions in a semi-arid river basin of
western India. Among the three temperature-based methods,
the PMT-1 method yielded better ETo estimates with RMSE of
0.59 mm day-1 and IOA of 0.89. The spatial variation of ETo in
the study area during the monsoon season ranged from 3.8-4.7
mm day-1 for normal years, and 4.0-5.0 mm day-1 for drought
years. The maximum increase in ETo during drought years was
12% as compared to normal years in half of the study area. This
finding emphasizes the urgent need for efficient management Fig. 3 : Spatial variation ETo over the study area during
of water resources to ensure sustainable water supply during (a) Normal year and (b) Drought year
drought years for domestic and agriculture purposes. The
developed ETo maps can provide a scientific basis for effective data conditions. The methodology presented in this study can
planning and management of water resources at a micro-scale provide useful guidelines for semi-arid regions of Indian
(block or district level) in the study area under scanty climatic subcontinent as well as for other regions of the world.
356 Temperature-based reference evapotranspiration methods for Maharashtra September 2019

REFERENCES N . V. ( 2 0 1 3 ) . C o m p u t a t i o n o f r e f e r e n c e c r o p
evapotranspiration of Nasik station of Maharashtra,
Allen, R.G., Perevia L.S, Raes, D. and Smith, M. (1998). Crop
India. Mausam, 64(2): 357–362.
Evapotranspiration: Guidelines for Computing Crop
Water Requirements. FAO Irrigation and Drainage Nandagiri, L. and Kovoor, G.M. (2006). Performance
Paper No. 56, Food and Agriculture Organization, evaluation of reference evapotranspiration equations
Rome, Italy, pp: 300. across a range of Indian climates. J. Irrig. Drain. Eng.,
132(3): 238–249.
EIS (2006). Disaster Management in Maharashtra.
Environmental Information System (EIS). Phad, S.V., Dakhore, K.K. and Sayyad, R.S. (2019).
Environmental Department, Environmental Comparison of different methods for estimation of
Information System (ENVIS) Newsletter, Government reference evapotranspiration at Parbhani,
of Maharashtra, India, 1, pp: 1-10. Maharashtra. J. Agrometeorol., 21(2): 236-238.
Hargreaves, G.H. and Samani, Z.A. (1985). Reference crop Sabziparvar, A.A. and Tabari, H. (2010). Regional estimation
evapotranspiration from temperature. Appl. Eng. of reference evapotranspiration in arid and semiarid
Agric., 1(2): 96–99. regions. J. Irrig. Drain. Eng., ASCE, 136(10):
724–731.
Jabloun, M.D. and Sahli, A. (2008). Evaluation of FAO-56
methodology for estimating reference Sentelhas, P.C., Gillespie, T.J. and Santos, E.A. (2010).
evapotranspiration using limited climatic data: Evaluation of FAO Penman–Monteith and alternative
Application to Tunisia. Agric. Water Manage., 95(6): methods for estimating reference evapotranspiration
707–715. with missing data in Southern Ontario, Canada. Agric.
Water Manage., 97(5): 635–644.
Jadhav, P.B., Kadam, S.A. and Gorantiwar, S.D. (2015).
Comparison of methods for estimating reference Tabari, H. (2010). Evaluation of reference crop
evapotranspiration (ET o ) for Rahuri region. J. evapotranspiration equations in various climates.
Agrometeorol., 17(2): 204-207. Water Resour. Manage., 24(10): 2311–2337.
Jensen, M.E., Burman, R.D. and Allen, R.G. (1990). Tabari, H., Grismer, M.E. and Trajkovic, S. (2013).
E v a p o t r a n s p i r a t i o n a n d I r r i g a t i o n Wa t e r Comparative analysis of 31 reference evapo
Requirements. ASCE Manual and Report on transpiration methods under humid conditions. Irrig.
Engineering Practices No. 70, ASCE, New York, pp: Sci., 31(2): 107–117.
528.
Todorovic, M., Karic, B. and Pereira, L.S. (2013). Reference
Lopez-Urrea, R., de Santa Olalla, F.M., Fabeiro, C. and evapotranspiration estimate with limited weather data
Moratalla, A. (2006). Testing evapotranspiration across a range of Mediterranean climates. J. Hydrol.,
equations using lysimeter observations in a semiarid 481: 166–176.
climate. Agric. Water Manage., 85(1): 15–26.
Wable, P.S. and Jha, M.K. (2017). Application of Archimedean
Maidment, D.R. (1992). Handbook of Hydrology. McGraw- copulas to the impact assessment of hydro-climatic
Hill, Inc., New York, USA, pp: 1424. variables in semi-arid aquifers of western India.
Hydrogeol J., 26(1): 89-108.
Meshram, D.T., Gorantiwar, S.D., Mittal, H.K. and Singh,

Received : October 2018 : Accepted : August 2019