Arab J Geosci (2016) 9:441
DOI 10.1007/s12517-016-2469-8
ORIGINAL PAPER
Evaluation of METRIC-derived ET fluxes over irrigated alfalfa
crop in desert conditions using scintillometer measurements
K. A. Al-Gaadi 1,2 & V. C. Patil 1,3 & E. Tola 1 & R. Madugundu 1 & P. H. Gowda 4
Received: 28 October 2015 / Accepted: 7 April 2016
# Saudi Society for Geosciences 2016
Abstract A field study on a 50-ha alfalfa (Medicago sativa L.)
irrigated field was conducted to investigate the performance of
the remote sensing (RS) based Mapping EvapoTranspiration at
high Resolution with Internalized Calibration (METRIC) model
in the estimation of evapotranspiration (ET) under the arid con-
ditions of Saudi Arabia. The METRIC model performance was
investigated by comparing the energy fluxes estimated by the
model to the output of a surface layer scintillometer (SLS) system
installed in the field, given the fact that the SLS is efficient in
measuring sensible heat fluxes (H) over vegetative areas.
Landsat-8 reflectance data were used as inputs for the
METRIC model. Results of the study revealed that the
HMETRIC data was strongly correlated with the HSLS data with
an R2 value of 0.74 (P > F = 0.0064) and a mean bias error
(MBE) of 6.05 W m−2 (6 %). The METRIC model showed a
good performance in estimating the hourly latent heat (LE) fluxes
compared with SLS data with an R 2 value of 0.81
(P > F = 0.0023), an MBE of 24.46 W m−2 (8 %) and a Nash–
Sutcliffe efficiency (NSE) of 0.91. Furthermore, the hourly ET
was estimated with an MBE and an NSE of 0.036 mm h−1 (8 %)
and 1.00, respectively. Compared to the SLS data, the METRIC
model was found to generally provide an efficient and an
* R. Madugundu
rmadugundu@ksu.edu.sa
1
Precision Agriculture Research Chair, King Saud University,
Riyadh, Saudi Arabia
2
Department of Agricultural Engineering, College of Food and
Agriculture Sciences, King Saud University, Riyadh, Saudi Arabia
3
Electron Science Research Institute, Edith Cowan University,
Joondalup, Australia
4
Forage and Livestock Production Research Unit, USDA-ARS
Grazing-lands Research Laboratory, El Reno, OK, USA
accurate means of energy fluxes estimation; therefore, ET esti-
mation over the studied irrigated alfalfa crop.
Keywords Centre pivot system . Evapotranspiration .
Landsat-8 data . Alfalfa field . Surface layer scintillometer
Introduction
In agriculture, evapotranspiration (ET) plays a major role in
the quantification of consumptive use of irrigation water and
precipitation. Therefore, an accurate estimation of ET is es-
sential for the determination of crop water needs and crop
water use efficiency (Gowda et al. 2007). Accurate ET mea-
surements can be achieved through weighing lysimetry, soil-
water monitoring, and micrometeorological methods (eddy
covariance, scintillometry, Bowen Ratio, etc.). Of which, the
micrometeorological methods were found to be more accu-
rate, as they worked under critical environmental conditions
without disturbing the microenvironment of the area under
investigation (Mengistu and Savage 2010).
Scintillometry is one of the popular micrometeorological
methods used by agro-meteorologists, hydrologists, and micro-
meteorologists for various applications (Thiermann and Grassl
1992; Savage et al. 2005; Odhiambo and Savage 2009). A
surface layer scintillometer (SLS) is a scientific device that
works on scintillometry techniques and is used to measure
small fluctuations of the refractive index of the air caused by
variations in temperature, humidity, and pressure (Beyrich et al.
2012). It consists of an optical or a radio wave transmitter and a
receiver at opposite ends of an atmospheric propagation path.
The receiver detects and evaluates the fluctuations in the inten-
sity of the transmitted signal, which is termed as scintillation
(Solignac et al. 2009). The variance of the logarithm of the
intensity fluctuations can be related to the structure parameter
441 Page 2 of 12
of the refractive index of the air ðC2n ). Then, it can be related to
the structure parameter of temperature ðC2T ), to derive the sen-
sible heat flux through Monin-Obukhov similarity theory
(MOST), which is required for ET estimation (Ezzahar et al.
2007; Andreas 2012; Gruber and Fochesatto 2013).
Remote sensing (RS) models are widely used for the estima-
tion of ET, which can be obtained as a residual of the surface
energy balance (EB) as described in Eq. 1 (Liu et al. 2007).
Examples of these models include the Surface Energy Balance
Algorithm for Land (SEBAL) (Bastiaanssen et al. 1998),
Mapping Evapotranspiration at high Resolution with
Internalized Calibration (METRIC) (Allen et al. 2013),
Simplified Surface Energy Balance (SSEB) (Senay et al. 2013),
Remote Sensing of Evapotranspiration (ReSET) (Elhaddad and
Garcia 2011), Modified SEBAL (Papadavid et al. 2013), and
Surface Energy Balance with Topography Algorithm (SEBTA)
(Gao et al. 2011). These models have been successfully used to
estimate ET over agricultural fields at different spatial scales
(Kalma et al. 2008; Gowda et al. 2008; Lagos et al. 2013).
LE ¼ Rn −G−H ð1Þ
Where LE is the latent heat (W m−2), Rn is the net radiation
(W m−2), G is the soil heat flux (W m−2) and H is the sensible
heat flux (W m−2).
Several studies were conducted to assess the ability
of scintillometers in evaluating the accuracy of remote
sensing estimated ET across a range of agricultural
fields (Hoedjes et al. 2007; Pauwels et al. 2008;
Ezzahar et al. 2009; Kite and Droogers 2000;
Mkhwanazi et al. 2012; Rambikur and Chavez 2012),
grasslands (Hartogensis and DeBruin 2005; Odhiambo
and Savage 2009), vineyards (DeBruin et al. 1995),
mixed vegetation-palm trees, and rice (Hemakumara
et al. 2003; Dzikiti et al. 2014). Most of these studies
were performed under humid and semi-arid conditions.
However, such studies under hyper arid or desert con-
ditions are limited, leading to a lack of knowledge in
the utility of these models in ET estimation under these
conditions. Hence, this study was conducted to evaluate
the METRIC model estimated energy fluxes against
those measured using the SLS system installed over an
irrigated alfalfa (Medicago sativa L.) field under desert
(hyper arid) conditions of Saudi Arabia.
Materials and methods
Description of the study area
The experimental work was conducted on a 50-ha field
cultivated with alfalfa crop under a centre pivot irriga-
tion system in Todhia Arable Farm (TAF) located
Arab J Geosci (2016) 9:441
between Al-Kharj and Haradh cities of Saudi Arabia
(Fig. 1), and within the latitudes of 24° 10′ 22.77 and
24° 12′ 37.25 N and the longitudes of 47° 56′ 14.60
and 48° 05′ 08.56 E. The climate of the study area was
hyper arid with hot summers (40 ± 2 °C) and cold to
moderate winters (15 ± 3 °C). The annual rainfall was
about 90 mm, most of which occurred in the months of
November and December. The major forage crops culti-
vated in the study area were alfalfa, Rhodes grass, and
corn.
Surface layer scintillometer
A dual-beam SLS (model SLS40-A) was installed on the 50-
ha alfalfa study field with a path length (beam distance be-
tween the transmitter and the receiver) of 150 m and a beam
height of 1.6 m above the soil surface (Fig. 2). At the time of
SLS installation, the path of SLS was positioned perpendicu-
lar to the prevailing N–E wind direction. The surface
above which the SLS was installed was nearly flat.
The SLS system was set to measure the structure pa-


rameter of the refractive index of the air C2n , the structure


parameter of temperature C2T , the inner scale length ( 0), the
air temperature, the atmospheric pressure, and the sensible
heat flux (HSLS) on a 2-min sampling interval. The SLS mea-
surements were taken for the period from June 3 to October 9,
2013. The SRUN (Ver. 2.0) software program was used to
analyze the SLS data. During data analysis, the rejection or
filtering procedures were adopted as described by Savage
et al. (2004). The latent heat flux of the SLS system (LESLS)
was calculated as a residual component of the energy balance
equation 2 (Samain et al. 2012).
LE SLS ¼ AE−H SLS ð2Þ
Where HSLS is the SLS measured sensible heat flux
and AE is the available energy (i.e., Rn − G). The net
radiation (Rn) and soil heat flux (G) were measured by
using the Net Radiometer (CNR-4) and soil heat flux
plates (SHFP) installed within the path of the SLS.
The details of the instruments used for measuring EB
components are provided in Table 1. Subsequently, the
obtained LESLS was transformed into ETSLS.
METRIC model
Cloud free Landsat-8 images were processed for surface
properties, such as albedo, surface emissivity, normal-
ized difference vegetation index (NDVI), leaf area index
(LAI), and surface temperature. These variables were
used as inputs for the METRIC model to estimate ET
(ET METRIC ) by solving the components of the EB
Arab J Geosci (2016) 9:441
Fig. 1 Location map of the surface layer scientillometer (SLS) system over an irrigated alfalfa field
equation (Eq. 1). In this study, the METRIC model was
executed as described by Allen et al. (2007). Initially,
the METRIC estimated LE (LEMETRIC) was computed
as a residual component of EB equation. Subsequently,
the obtained LEMETRIC was converted to ETMETRIC. The
net radiation (Rn) was computed by subtracting the out-
going from the incoming radiant fluxes (Eq. 3).
Rn ¼ Rs↓ −αRs↓ þ RL↓ −RL↑ −ð1−ε0 ÞRL↓
Where R S↓ is the incoming short-wave radiation
(W m−2); α is the surface albedo (dimension less); RL↓
is the incoming long-wave radiation (W m−2); RL↑ is the
outgoing long-wave radiation (W m−2) and ε0 is the
broad-band surface emissivity (dimension less). The
(1-ε0)RL↓ term represents the fraction of the incoming
long-wave radiation reflected from the surface.
Page 3 of 12 441
Landsat-8 images were used to compute RS↓ using the solar
constant (1367 W m−2), the cosine of the solar incidence an-
gle, the inverse squared relative earth–sun distance, and the
broad-band atmospheric transmissivity according to Allen
et al. (2007). The surface albedo (α) was computed by inte-
grating the reflectance of Landsat-8 bands (2–8). The outgo-
ing long-wave radiation (RL↑), which is emitted from the sur-
face, was derived using the Stefan–Boltzmann constant (5.67
ð3Þ
E-08 W m−2 K−4), the broadband surface emissivity (ε0), and
the surface temperature (Ts in °K). The Ts, a remotely sensed
radiometric surface temperature, was obtained after correcting
the sensor brightness temperature for the atmospheric effects
and surface emissivity (Hipps 1989; Brunsell and Gillies
2002). Subsequently, the surface emissivity was corrected
using the standard values for typical bare soil and fully vege-
tated surface emissivity as 0.93 and 0.98, respectively.
Thereafter, surface albedo for the vegetated areas was
441 Page 4 of 12 Arab J Geosci (2016) 9:441

Fig. 2 A schematic diagram of

the surface layer scintillometer
(SLS) setup

estimated using Brest and Goward (1987) model (Eq. 4), et al. 2014). In this study, Eq. 7 proposed by Bastiaanssen
based on the reflectance in the red (R) and near infrared (1998) was used to estimate G as a fraction of Rn and NDVI.
(NIR) bands. The emissivity of the air was obtained using

G
Eq. 5 (Brutsaert 1975). ¼ ðT s −273:15Þð0:0038 þ 0:0074αÞ 1−0:98 NDVI 4 ð7Þ
Rn
α ¼ 0:512 R þ 0:418 NIR ð4Þ
Where Ts is the surface temperature (°K) and α is the sur-
face albedo.
The sensible heat flux (H) was calculated by using Eq. 8.
εa ¼ 0:85 ð−lnτsw Þ0:09 ð5Þ
dT
Where εa is the effective atmospheric emissivity (dimen- H ¼ ρair C P ð8Þ
rah
sion less); τsw is the broad-band atmospheric transmissivity for
short-wave radiation, which was calculated as described by Where ρ is the air density (kg m−3), Cp is the specific heat
ASCE-EWRI (2005) (Eq. 6). capacity of the air at constant pressure (J kg−1 K−1), dT is the
"  0:4 # near surface temperature difference (°K), and rah is the aero-
−0:00146P W dynamic resistance (s m−1) to near surface heights Z1 and Z2
τ sw ¼ 0:35 þ 0:627exp −0:075 ð6Þ
K t cosθhor cosθhor (generally 0.1 and 2.0 m) and was computed as a function of
the estimated aerodynamic roughness of the particular pixel.
To compute rah, the Monin–Obukhov theory was applied in an
Where P is the atmospheric pressure (kPa); W is the amount
iterative process considering buoyancy effects until rah was
of water in the atmosphere (mm), θhot is the solar zenith angle
stabilized (Eq. 9).
over a horizontal surface and Kt (dimension less) is the turbid-
 
ity coefficient 0 < Kt ≤ 1.0. Z2
The soil heat flux (G) is the rate of heat storage into the soil ln
Z1
and vegetation due to conduction (Gowda et al. 2011). rah ¼ ð9Þ
u* k
Different empirical equations have been developed for the
estimation of G in various experimental fields (Bastiaanssen Where Z1 and Z2 are the heights in meters above the zero
et al. 1998; Chavez et al. 2005; Singh et al. 2008; Danelichen plane displacement (d) of the vegetation, u* is the friction

Table 1 Instruments used for

measuring energy balance Measured parameter Sensor description No. of sensors
components
Wind direction, solar radiation, air Weather Station (Davis – Vantage Pro2) 1
temperature, RH, perspiration,
VPD, etc.
Sensible heat flux and atmospheric Surface layer scintillometer 1
pressure, temperature (SLS-40A), Scientic AG.
Net radiation CNR-4 Net Radiometer (Kipp & Zonen) 1
Soil heat flux Soil heat flux plates (HFP01) – 80 mm (dia.) 5
Soil moisture Soil Moisture Probe (Theta Probe ML2x) 5
Arab J Geosci (2016) 9:441
velocity (m s−1) which quantifies the turbulent velocity fluctua-
tions in the air, and k is the Von Karmon’s constant (0.41). The
friction velocity (u*) was computed using the logarithmic wind
law for neutral atmospheric conditions (Andreas 1992) (Eq. 10).
ku200
u* ¼  .  ð10Þ
ln 200 Z 0m
−1
Where u200 is the wind speed (m s ) at a blending height
assumed to be 200 m, and Z0m is the momentum roughness
length (m) for each pixel and was calculated as a function of
LAI (Eq. 11).
Z 0m ¼ 0:018 LAI ð11Þ
Cold and hot pixel selection
To compute the sensible heat flux (H) from Eq. 8, the near
surface temperature (dT) for each pixel needs to be defined.
Hence, the dT was assumed to have a linear relationship with
surface temperature Ts (Eq. 12) and was determined by
selecting two anchor pixels (cold and hot pixels), as a linear
relation of dT and the radiometric surface temperature, and the
relationship was used to explain the use of a and b coefficients
(Eqs. 13 and 14).
dT ¼ aT s þ b ð12Þ
dT hot −dT cold
a¼ ð13Þ
T shot −T scold
b ¼ dT hot −a  T shot ð14Þ
In this study, the temperature of the cold pixel (Hcold) was
computed by selecting the high NDVI pixel (>0.68), which
represented the full vegetation cover (Eq. 15).
H cold ¼ ðRn −GÞcold −LE cold ð15Þ
Uniform (pure) thermal pixel was selected for the compu-
tation of Hcold. To avoid the thermal contamination from sur-
rounding areas, LE of the cold pixel was assumed to be
1.05*ETr*λ (where ETr is the reference ET, λ is the latent heat
of vaporization) as proposed by Allen et al. (2007). The ETr
was calculated on an hourly basis employing the standardized
ASCE Penman–Monteith method (ASCE-EWRI 2005) on
0.5 m tall alfalfa crop. The data required for the computations
was collected from the automated weather station installed
near the experimental field.
Pure high temperature pixel was selected for the computa-
tion of the hot pixel temperature (dThot), corresponding to the
low NDVI (<0.08) pixels (i.e., the borders of the study field).
Page 5 of 12 441
It was ensured that the selected hot pixels were not in touch
with irrigation water and were purely representing bare or
nearly bare soil conditions, as recommended by Allen et al.
(2005, 2007). The selected hot pixel was assumed to have zero
ET. Once the hot pixel was identified, the value of Hhot was
obtained using Rn and G of that hot pixel (Eq. 16).
H hot ¼ ðRn −GÞhot −LE hot ð16Þ
Calculation of evapotranspiration
ET at the instant of the satellite image (ETinst) was calculated,
for each pixel, from the relation between LE and the latent heat
of vaporization as given in Eq. 17.
LE
ET inst ¼ 3600 ð17Þ
λρw
Where ETinst is the instantaneous ET (mm h−1), ρw is the
density of water (~1000 kg m−3) and λ is the latent heat of
vaporization (J kg−1) representing the heat absorbed when a
kilogram of water evaporates, and was computed using Eq. 18.
λ ¼ ½2:501−0:00236ðT s −273:15Þ106 ð18Þ
The reference ET fraction (ETrF) was calculated as the ratio of
the computed instantaneous ET (ETinst) from each pixel to the
reference ET (ETr) computed from the weather data using Eq. 19.
ET inst
ET r F ¼ ð19Þ
ET r
Where ETinst (mm h−1) is from Eq. 18 and ETr is for the
standardized 0.5 m tall alfalfa reference at the time of the image.
Performance evaluation of the METRIC model
against the SLS system
The METRIC model estimated energy balance (EB) compo-
nents were assessed against those measured by the SLS system.
The assessment was achieved by using different performance
indicators, such as the root mean square error (RMSE), the mean
bias error (MBE), and the Nash–Sutcliffe efficiency (NSE), as
suggested by Krause et al. (2005) and Gowda et al. (2008).
Results and discussion
SLS measured parameters
The SLS measured structural parameters were analyzed, and
the summary results are provided in Table 2. The results indi-
cated that the SLS measured stability parameter varied from
−0.28 to 0.07 (negative means unstable; positive means
441 Page 6 of 12 Arab J Geosci (2016) 9:441

Table 2 SLS measured structural parameters and sensible heat fluxes at the time of satellite overpass (07:17 ± 5 min of UTC)

Date (2013) Plant height (m) 0 Stability parameter (ζ) Condition BR H (W m−2)
C 2n (mm) C 2T
(m(−2/3)) (K2 m(−2/3))

June 03 0.39 7.48E-13 5.649 1.18E + 00 0.02 Stable −0.17 −64.30

June 19 0.48 4.58E-13 4.766 7.55E-01 −0.15 Unstable 0.70 134.47
July 05 0.49 5.66E-13 3.493 9.26E-01 0.07 Neutral −0.24 290.59
July 21 0.66 9.04E-13 6.309 1.38E + 00 −0.24 Unstable 0.63 141.92
Aug. 22 0.24 8.76E-13 5.514 1.38E + 00 −0.28 Unstable −0.42 14.54
Sept. 07 0.52 4.56E-13 5.099 7.53E-01 0.04 Stable −0.42 213.88
Sept. 23 0.29 4.57E-13 4.950 6.55E-01 −0.03 Unstable 0.14 26.61
Oct. 09 0.52 6.05E-13 5.183 9.77E-01 0.04 Stable −0.13 100.56

refractive index structure parameter, 0 inner scale of turbulence, temperature structure parameter, BR bowen ratio, H sensible heat flux
C 2n C 2T

stable). The stability condition ( 0) was more than 4.0 mm, and
the Bowen Ratio (β) varied between −0.13 and 0.70. Due to
the hot climatic conditions during the summer period (June to
August), there was an inconsistency in β values, which ranged
from −0.17 to 0.70. It was also noted that as the wet conditions
progressed (September to October), the β value was increased
from −0.42 to 0.14.
The nature of the SLS measured stability parameter (the
stable and unstable conditions) mainly relied on the tempera-
ture and the vegetation height at the time of measurement. In
general, the stability parameter became Bunstable^ with the
decrease in temperature and vegetation height; however, it
turned to be Bstable^ with the increase in temperature and
vegetation height. Hence, based on the results of this study,
most of the recorded SLS measurements at the time of satellite
overpass were under unstable condition (Table 2). It was also
noticed that in mid-summer (June 3 to July 21, 2013), the
increase in the plant height was associated with a decrease in
surface temperature, which was attributed to the increase in
wetness condition of alfalfa crop caused by the increased LAI.
Similar trend occurred in the period from August 22 to
October 9, 2013. However, the magnitude of β varied drasti-
cally, which can be due to fluctuations in the measured 0.
Energy balance components
The results of the comparative analysis of the surface energy
fluxes (Rn, G, H, and LE) along with the hourly ET estimated
by the METRIC model and the SLS approach are provided in
Table 3. The evaluation of the METRIC-derived EB compo-
nents was performed against the SLS measurements as de-
scribed in the following sections.
Net radiation (Rn)
In this study, the hourly CNR-4 measured Rn, for the eight
observation dates, was extracted and analyzed as provided in
Table 3. The results of the measured Rn showed a mean var-
iation of about 10 %. The magnitude and trends in the
METRIC-estimated Rn were also analyzed for discrepancies.
It was found that the METRIC-estimated Rn was achieved
with a mean variability of about 5 % across the studied
Landsat-8 images (Fig. 3). A linear regression analysis was
performed between the METRIC predicted and the CNR-4
measured Rn and the results showed a good correlation be-
tween them, with an R2 value of 0.54 (P > F = 0.0381). On the
other hand, the values of the RMSE, the MBE, and the NSE
performance indicators were 18.32 W m−2 (4 %), 8.66 W m−2
(2 %), and 1.0, respectively (Table 4). Although the results
showed low values of RMSE and MBE, the trend was not a
perfect fit. This was due to the differences in scales, where the
Rn measured by CNR-4 was a point measurement and the
view footprint of the downward pyranometer from its height
of 3.5 m was approximately 26 m, which deviated from the
aggregated Rn estimates over a large area of the METRIC
model. Hence, the ground-based versus the satellite-based es-
timates of Rn indicated that the METRIC model overestimated
Rn values by 4 %. This is in agreement with the results of
previous studies (Singh et al. 2008; Mkhwanazi and Chavez
2012) with RMSE of 22.1 W m −2 (4 %) and MBE of
17.8 W m−2 (3 %).
Soil heat flux (G)
Results indicated that the measured soil heat flux (G) ranged
between 17 and 155 W m−2 with mean measurement errors of
9–16 %. The relationship between the METRIC estimated
(G METRIC) and the measured (GMeasured) soil heat flux,
depicted in Fig. 4, was relatively good with an R2 value of
0.67 (P > F = 0.0131) and an NSE value of 0.59, while the
RMSE and MBE values were 28.46 W m−2 (45 %) and
12.42 W m−2 (19 %), respectively. However, significant error
was observed at the stages just after harvesting (RMSE = 32 to
46 %), when the soil was dry and exposed. These errors were
Arab J Geosci (2016) 9:441 Page 7 of 12 441

Table 3 Comparison of METRIC-derived hourly surface energy fluxes against measurements

EB Component Method June July Aug Sept Oct

03 19 05 21 22 07 23 09

Rn (W m−2) METRIC 488.66 491.46 491.07 490.64 502.95 489.37 491.8 477.75
Measured 468.20 500.60 506.22 463.27 513.17 476.86 469.6 456.49
G (W m−2) METRIC 28.62 92.54 143.73 80.65 97.92 48.16 19.98 98.31
Measured 21.64 44.11 155.18 70.33 37.54 54.98 16.50 110.30
H (W m−2) METRIC −57.19 184.15 286.10 160.45 01.77 95.75 39.98 195.62
SLS −64.30 134.47 290.59 141.92 14.54 213.88 26.61 100.56
LE (W m−2) METRIC 574.47 214.76 61.25 249.55 500.20 345.46 551.7 183.82
SLS 510.86 322.02 60.45 251.02 461.09 208.00 426.5 245.63
ET (mm h−1) METRIC 0.85 0.32 0.09 0.37 0.74 0.51 0.81 0.27
Calculated 0.75 0.48 0.09 0.37 0.68 0.31 0.63 0.36
G soil heat flux, Rn net radiation, H sensible heat flux, LE latent heat flux, ET evapotranspiration

found to be relatively higher than values reported in similar
previous studies. For example, Mkhwanazi and Chavez
(2012) reported an RMSE value of 14.2 W m−2 (28 %) and
an MBE value of −3.0 W m−2 (−6 %) in estimating G using
the METRIC model for irrigated alfalfa.
Sensible heat flux (H)
The results of the METRIC estimated sensible heat flux
(HMETRIC) compared to that measured using the SLS system
(HSLS) are provided in Tables 1 and 2 and illustrated in Fig. 5.
It was observed that HMETRIC was significantly correlated to
HSLS with an R2 value of 0.74 and P > F of 0.0064. Further
assessment of the performance of the HMETRIC was achieved
through other performance indicators (RMSE, MBE, and
NSE) as shown in Table 4. The results of both MBE
(6.05 W m−2 B6 %^) and NSE (0.75) confirmed the significant
correlation between the HMETRIC and the HSLS. Similar results
were reported by Carrasco-Benavides et al. (2013), where the
METRIC model overestimated H value by a mean absolute
error (MAE) of 10 % compared to that measured using eddy
covariance system. In contrast, the high value of RMSE
(57.24 W m−2 B53 %^) indicated inaccuracy in estimating H
using the METRIC model compared to the SLS data. The
observed discrepancy between the measured (HSLS) and the
estimated H (i.e. HMETRIC) might be attributed to the dry and
heterogeneous surface conditions (Mkhwananzi et al. 2012).
Also, the temporal and spatial fluctuations in the air tempera-
ture are expected to introduce uncertainties in H estimation
(Lagouarde et al. 2002; Brohan et al. 2006). This was evident

Fig. 3 Relationship between the

METRIC estimated net radiation
(Rn) and that measured using
CNR-4 net radiometer
441 Page 8 of 12 Arab J Geosci (2016) 9:441

Table 4 METRIC performance statistics were due to the high wind speed and low relative humidity
EB R 2
P > F RMSE MBE NSE conditions, which are common in the desert arid environ-
component ments. These variations were clearly observed in the values
Amount % Amount % of NDVI obtained for the hot pixels (0.04 to 0.08) and the cold
pixels (0.68 to 0.81).
Rn (W m−2) 0.54 0.0381 18.32 3.80 8.66 1.80 1.000
G (W m−2) 0.67 0.0131 28.46 44.59 12.42 19.45 0.590
H (W m−2) 0.74 0.0064 57.24 53.36 6.05 5.63 0.754
Latent heat flux and ET
LE (W m−2) 0.81 0.0023 83.28 26.8 24.46 7.87 0.914
ET (mm hr−1) 0.92 0.0001 0.123 26.80 0.036 7.87 1.000 The comparison of the METRIC estimated latent heat flux
RMSE root mean square error, MBE mean bias error, NSE Nash–Sutcliffe efficiency (LE) (LEMETRIC) against the SLS-derived observations is pre-
sented in Fig. 6. It was observed that the LEMETRIC values
were slightly higher than LESLS values, with mean values of
335.154 and 310.695 W m−2, respectively. The statistical anal-
in the retrieval of H on August 22, 2013 (mid-summer), where ysis indicated that the LEMETRIC accounted for 27 % (RMSE)
the remote sensing estimated H was 1.77 W m−2 compared to variability compared to the LESLS hourly fluxes. Although the
the SLS measured H of 14.54 W m−2. In addition, the wind RMSE (83.28 W m−2) was relatively high, the results of both
speed and the aerodynamic surface characteristics also posed a MBE (24.46 W m−2 B8 %^) and NSE (0.91) confirmed the
major challenge, where the sensor field of view changed from high significant correlation between LEMETRIC and LESLS. A
one angle to another. The variations might occur in the re- similar study by Carrasco-Benavides et al. (2013) reported an
ceived signal based on the density of vegetation over the field MAE of 14 % in estimating LE using the METRIC model
of view of the SLS sensor. In addition, neglecting the effect of compared to the ground-based eddy covariance
water vapor on C2T could form a source of error (Singh et al. measurements.
2008; Gowda et al. 2008). The MBE of the LEMETRIC for the studied eight images
The sensitivity of HSLS was influenced by u* and more was 24 %. When the image acquired on August 22, 2013
significant by the stable conditions (Hartogensis et al. 2003). was omitted, the average absolute error was reduced to
For example, on August 22, 2013, the HMETRIC was calculated 18 %. This was due to the fact that the alfalfa crop was har-
at 1.77 W m −2 compared to the measured value of vested only 2 days before the satellite overpass, and eliminat-
14.54 W m−2; where the plant height was 0.24 cm (minimal ing the associated image of August 22, 2013 would definitely
vegetation) and the stability parameter was −0.28 (unstable increase the accuracy of data. A non-uniformity in wetness
condition), along with the drier condition provided the low associated with August 22, 2013 image was due to differential
LEcold value, which has significantly affected the determined drying as one side of the field was with fully grown alfalfa and
HMETRIC. The results also revealed that the highest values of the other side was harvested. Hence, the performance of the
HMETRIC (286 W m−2) occurred in mid-summer (July 5th) METRIC model in the estimation of dT (the H of anchor

Fig. 4 The METRIC estimated

soil heat flux (G) against that
measured using the SHFP
Arab J Geosci (2016) 9:441
Fig. 5 Comparison between the
METRIC estimated sensible heat
flux (H) and the SLS data
pixels, dTcold, and dThot) was exaggerated, where the SLS
beam was passing through both the harvested and the non-
harvested areas of the field. This can be attributed to the error
associated with the estimation of the individual components of
the available energy (Rn − G), which may have resulted in
estimation errors for both short- and long-wave components.
Moreover, the differentiation between the direct and diffuse
radiation has not been made in the METRIC model and only
the total Rn flux was considered. In comparison, a commonly
quoted standard error for ET prediction equations based on the
weather data, such as Penman or Penman–Monteith equa-
tions, was about 10 % for daily estimates (Wright and
Jensen 1978). Hence, a significant difference in LE was highly
Fig. 6 Comparison between the
METRIC estimated LE fluxes
and the SLS data
Page 9 of 12 441
expected since the differences in water use efficiency was
neglected in this study.
The ET dynamics of alfalfa crop across the study period is
provided in Fig. 7. Comparisons of the METRIC-estimated
hourly and daily ET against SLS-derived data are illustrated in
Fig. 8. The results revealed a highly significant linear relation-
ship between the hourly ETMETRIC and ETSLS (R2 = 0.92,
P > F = 0.0001). The results of the performance indicators
confirmed this significant correlation with the values of
0.123 mm h−1 (27 %), 0.036 mm h−1 (8 %), and 1.0, for
RMSE, MBE, and NSE, respectively (Table 4). These results
are in agreement with the findings of Mkhwanazi and Chavez
(2012) who assessed the performance of the METRIC model
441 Page 10 of 12 Arab J Geosci (2016) 9:441

Fig. 7 The METRIC estimated hourly ET map for the experimental field

Fig. 8 The METRIC estimated

hourly ET and that calculated
from the SLS data
Arab J Geosci (2016) 9:441
in estimating the hourly ET against the values measured
using a large monolith weighing lysimeter located in the
middle of an alfalfa field. They reported an R2 value of
0.83, RMSE of 014 mm h − 1 (18 %), MBE of
−0.08 mm h−1 (−10 %), and NSE of 0.71 for the rela-
tionship between their METRIC and lysimeter data.
These results are in partial agreement with the results
reported by Pocas et al. (2014), which revealed that the
METRIC model slightly overestimated ET (R2 = 0.85
and MBE = 13 %) compared to the ground-based eddy
covariance measurements for the first year (2011) of
their study. For the second year (2012), their results
indicated that the METRIC model exhibited a higher
error (MBE = 31 %) in estimating ET compared to
the measured values. In general, these results can be
recognized within most of the remote sensing methods
used for the estimation of evapotranspiration, which are
associated with an average RMSE of 15–30 % (Kalma
et al. 2008).
Conclusions
A field study was conducted to evaluate the performance of
the METRIC model in estimating surface energy fluxes and
ET by comparing its data to the output of a surface layer
scintillometer (SLS) system installed over an irrigated alfalfa
field under arid conditions. The specific conclusions drawn
from the study are as follows:
& The METRIC model was observed to be accurate and
efficient in estimating energy components (H and LE).
The HMETRIC was strongly correlated with the HSLS
(R2 = 0.74; P > F = 0.0064), and the hourly latent heat
(LE) fluxes were significantly correlated as well
(R2 = 0.81; P > F = 0.0023).
& A high significant linear relationship was observed be-
tween the METRIC model estimated hourly ET and that
calculated from the SLS data (R2 = 0.92, P > F = 0.0001).
These statistics exhibit the good accuracy of the model in
providing ET data.
& The METRIC model provided an efficient RS-based
means to determine the important factor of ET. The model
was proved to provide a good estimation of ET under the
arid desert conditions of Saudi Arabia.
Acknowledgments This project was financially supported by King
Saud University, Vice Deanship of Research Chairs. The assistance pro-
vided by the graduate students M.E. Abass, A.M. Zeyada, and A.G.
Kayad in the accomplishment of the field research work was quite valu-
able. The unstinted cooperation and support extended by Mr. Jack King,
Mr. Alan King, and their team in carrying out the research are gratefully
acknowledged.
Page 11 of 12 441
References
Allen GR, Tasumi M, Morse A, Kramber WJ, Bastiaanssen WGM (2005)
Computing and mapping of evapotranspiration p 73–90 In:
Aswathanarayana U (Eds) Advances in water science methodolo-
gies. AA Balkema Publishers Leiden The Netherlands. ISBN0–
203–08684-8
Allen RG, Burnett B, Kramber W, Huntington J, Kjaersgaard J, Kilic A,
Kelly C, Trezza R (2013) Automated calibration of the METRIC-
Landsat evapotranspiration process. J Am Water Resour Assoc
49(3):563–576
Allen RG, Tasumi M, Morse A, Trezza R, Wright J, Bastiaanssen WGM,
Kramber W, Lorite I, Robison C (2007) Satellite-based energy bal-
ance for mapping evapotranspiration with internalized calibration
(METRIC)—applications. J Irrig Drain Eng 133(4):395–406
Andreas E (1992) Uncertainty in a path-averaged measurement of the
friction velocity u*. J Appl Meteorol Climatol 31:1312–1321
Andreas E (2012) Two experiments on using a scintillometer to infer the
surface fluxes of momentum and sensible heat. J Appl Meteorol
Climatol 51:1685–1701 DOI:101175/JAMC-D-11-02481
ASCE-EWRI (2005) The ASCE standardized reference evapotranspira-
tion equation. In: Allen RG, Walter IA, Elliot RL, Howell TA,
Itenfisu D, Jensen ME, Snyder RL (Eds) ASCE standardization of
Reference Evapotranspiration Task Committee Final Rep Reston
Va: pp 70
Bastiaanssen WGM, Menenti M, Feddes RA, Holtslag AAM (1998)
Remote sensing surface energy balance algorithm for land
(SEBAL):1 formulation. J Hydrol 212-213(1–4):198–212
Beyrich F, Bange J, Hartogensis OK, Raasch S, Braam M, Dinther Dvan,
Graf D, Kesteren Bvan, Kroonenberg AC van den, Maronga B,
Martin S, Moene AF (2012) Towards a validation of scintillometer
measurements:the LITFASS-2009 experiment. Boundary-Layer
Meteorol 144:83–112
Brest CL, Goward SN (1987) Driving surface albedo measurements from
narrow band satellite data. Int J Remote Sens 8:351–367
Brohan P, Kennedy JJ, Harris I, Tett SFB, Jones PD (2006) Uncertainty
estimates in regional and global observed temperature changes: a
new dataset from 1850. J Geophys Res 111:D12106. doi:101029/
2005JD006548
Brunsell NA, Gillies R (2002) Incorporating surface emissivity into a
thermal atmospheric correction. Photogramm Eng Remote Sens
68(12):1263–1269
Brutsaert W (1975) On a derivable formula for long-wave radiation from
clear skies. Water Resourc Res 11:742–744
Carrasco-Benavides M, Ortega-Farías S, Lagos LO, Kleissl J (2013)
Assessment of the METRIC model in the estimation of instanta-
neous values of sensible and latent heat fluxes over a drip-irrigated
Merlot vineyard using Landsat images. Anais XVI Simposio
Brasileiro de Sensoriamento Remoto - SBSR Foz do Iguaçu PR
Brasil 13 a 18 de abril de 2013 INPE (Accessed on August 27
2014 from:http://wwwdsrinpebr/sbsr2013/files/p0656pdf)
Chavez JL, Neale CMU, Hipps LE, Prueger JH, Kustas WP (2005)
Comparing aircraft-based remotely sensed energy balance fluxes
with eddy covariance tower data using heat flux source area func-
tions. J Hydrometeorol 6:923–940
Danelichen VHD, Biudes MS, Souza MC, Machado NG, Da-Silva BB,
Nogueira JD (2014) Estimation of soil heat flux in a neotropical
wetland and region using remote sensing techniques. Revista
Brasileira de Meteorologia 29(4):469–482
DeBruin HAR, Hurk BJJM van den, Kohsiek W (1995) The scintillation
method tested over a dry vineyard area. Boundary-Layer Meteorol
76:25–40
Dzikiti S, Jovanovic NZ, Bugan R, Israel S, Maitre DC Le (2014)
Measurement and modelling of evapotranspiration in three fynbos
vegetation types. Water SA 40(2):189–198
441 Page 12 of 12
Elhaddad A, Garcia LA (2011) ReSET-raster: surface energy balance
model for calculating evapotranspiration using a raster approach. J
Irrig Drain Eng 137(4):203–210
Ezzahar J, Chehbouni A, Er-Raki S, Hanich L (2009) Combining a large
aperture scintillometer and estimates of available energy to derive
evapotranspiration over several agricultural fields in a semi-arid re-
gion. Plant Biosyst 143(1):209–221
Ezzahar J, Chehbouni A, Hoedjes JCB, Chehbouni A (2007) On the
application of scintillometry over heterogeneous grids. J Hydrol
334:493–501
Gao ZQ, Liu CS, Gao W, Chang NB (2011) A coupled remote sensing
and the surface energy balance with topography algorithm (SEBTA)
to estimate actual evapotranspiration over heterogeneous terrain.
Hydrol Earth Syst Sci 15:119–139
Gowda PH, Chavez JL, Colaizzi PD, Evett SR, Howell TA, Tolk JA
(2007) Remote sensing based energy balance algorithms for map-
ping ET: current status and future challenges. Trans ASABE 50(5):
1639–1644
Gowda PH, Chavez JL, Colaizzi PD, Evett SR, Howell TA, Tolk JA
(2008) ET mapping for agricultural water management: present sta-
tus and challenges. Irrig Sci 26:223–237
Gowda PH, Howell TA, Paul G, Colaizzi PD, Marek TH (2011) SEBAL
for estimating hourly ET fluxes over irrigated and dryland cotton
during BEAREX08. Proc 2011 World Environ Water Resour Congr
2787–2795
Gruber M, Fochesatto GJ (2013) A new sensitivity analysis and solution
method for scintillometer measurements of area-averaged turbulent
fluxes. Boundary-Layer Meteorol 149:65–83
Hartogensis O, DeBruin HAR (2005) Monin-Obukhov similarity func-
tions of the structure parameter of temperature and turbulent kinetic
energy dissipation rate in the stable boundary layer. Boundary-Layer
Meteorol 116(2):253–276
Hartogensis O, Watts C, Rodriquez JC, DeBruin HAR (2003) Derivation
of an effective height for scintillometers: La Poza experiment in
Northwest Mexico. J Hydrometeorol 4:915–928
Hemakumara HM, Chandrapala L, Moene AF (2003) Evapotranspiration
fluxes over mixed vegetation areas measured from large aperture
scintillometer. Agric Water Manag 58:109–122
Hipps LE (1989) The infrared emissivities of soil and artemisia tridentate
and subsequent temperature corrections in a shrub-steppe ecosys-
tem. Remote Sens Environ 27:337–342
Hoedjes JCB, Chehbouni A, Ezzah J, Escadafal R, DeBruin HAR (2007)
Comparison of large aperture scintillometer and eddy covariance
measurements: can thermal infrared data be used to capture
footprint-induced differences? J Hydrometeorol 8:144–159
Kalma JD, McVicar TR, McCabe MF (2008) Estimating land surface
evaporation: a review of methods using remotely sensed surface
temperature data. Surv Geophys 29:421–469
Kite G, Droogers P (2000) Comparing evapotranspiration estimates from
satellites hydrological models and field data. J Hydrol 229(1–2):3–18
Krause P, Boyle DP, Base1 F (2005) Comparison of different efficiency
criteria for hydrological model assessment. Adv Geosci 5:89–97
Lagos LO, Lillo-Saavedra M, Fonseca D, Gonzalo C (2013)
Evapotranspiration of partially vegetated surfaces from remote sens-
ing. In: Lasaponara R, Masini N, Biscione M (Eds) Proc 33rd
EARSeL Sympo p613–624
Lagouarde JP, Jacob F, Gu XF, Olioso A, Bonnefond JM, Kerr YH,
McAneney KJ, Irvine M (2002) Spatialization of sensible heat flux
over a heterogeneous landscape. Agronomie 22:627–633
Liu S, Hu G, Lu L, Mao D (2007) Estimation of regional evapotranspi-
ration by TM/ETM data over heterogeneous surfaces. Photogramm
Eng Remote Sens 73(10):1169–1178
Arab J Geosci (2016) 9:441
Mengistu MG, Savage MJ (2010) Surface renewal method for estimating
sensible heat flux. Water SA 36(1):9–18
Mkhwanazi M, Chavez JL (2012) Using METRIC to estimate surface
energy fluxes over an alfalfa field in Eastern Colorado. Hydrology
Days, Colorado State University USA (Accessed on August 26th
2014 from: http://hydrologydays.colostate.edu/Papers_2012/
Mcebisi_paper.pdf)
Mkhwanazi M, Chavez JL, Rambikur EH (2012) Comparison of large
aperture scintillometer and satellite-based energy balance models in
sensible heat flux and crop evapotranspiration determination. Int J
Remote Sens Appl 2(1):24–30
Odhiambo GO, Savage MJ (2009) Surface layer scintillometry for esti-
mating the sensible heat flux component of the surface energy bal-
ance. S Afr J Sci 105:2008–2016
Papadavid G, Hadjimitsis DG, Toulious L, Michaelides S (2013) A mod-
ified SEBAL modeling approach for estimating crop evapotranspi-
ration in semi-arid conditions. Water Resour Manag 27:3493–3506
Pauwels VRN, Timmermans WJ, Loew A (2008) Comparison of the
estimated water and energy budgets of a large winter wheat field
during AgriSAR 2006 by multiple sensors and models. J Hydrol
349:425–440
Pocas I, Paco TA, Cunha J, Andrade A, Silvestre J, Sousa A, Santos FL,
Pereira LS, Allen RG (2014) Satellite-based evapotranspiration of a
super-intensive olive orchard: application of METRIC algorithms.
Biosyst Eng 128:69–81
Rambikur EH, Chavez JL (2012) Scintillometry for evapotranspiration
estimation over irrigated alfalfa and dry grassland p 110–118 In:
Proc Hydrol Days 2012 March 21–23 2012 North Ball Room-
Lory Student Center Colorado State University USA
Samain B, Simons GWH, Voogt MP, Defloor W, Bink NJ, Pauwels VRN
(2012) Consistency between hydrological model large aperture scin-
tillometer and remote sensing based evapotranspiration estimates for
a heterogeneous catchment. Hydrol Earth Sys Sci 16(7):2095–2107
Savage MJ, Everson CS, Odhiambo GO, Mengistu MG, Jarmain C
(2004) Theory and practice of evapotranspiration measurement with
special focus on SLS as an operational tool for the estimation of
spatially-averaged evaporation. WRC Report No 1335/1/04 Water
Res Comm Pretoria S Afr 204 pp (http://wwwreadperiodicalscom/
201001/1965015051html#ixzz3Iqa2Ockm)
Savage MJ, Odhiambo GO, Mengistu MG, Everson CS, Jarmain C
(2005) Theory and practice of evaporation measurement.
The 12th SANCIAHS National Hydrol Symp MidRand SA
(September 2005)
Senay GB, Bohms S, Singh RK, Gowda PH, Velpuri NM, Alemu H,
Verdin JP (2013) Operational evapotranspiration mapping using re-
mote sensing and weather datasets: a new parameterization for the
SSEB approach. J Am Water Resour Assoc 49(3):577–591
Singh RK, Irmak A, Irmak S, Martin DL (2008) Application of SEBAL
model for mapping evapotranspiration and estimating surface
energy fluxes in south-Central Nebraska. J Irrig Drain Eng
134(3):273–285
Solignac PA, Brut A, Selves JL, Beteille JP, Gastellu-Etchegorry JP,
Keravec P, Beziat P, Ceschia E (2009) Uncertainty analysis of com-
putational methods for deriving sensible heat flux values from scin-
tillometer measurements. Atmos Meas Tech 2:741–753
Thiermann V, Grassl H (1992) The measurement of turbulent surface-
layer fluxes by use of bichromatic scintillation. Boundary-Layer
Meteorol 58:367–389
Wright JL, Jensen ME (1978) Development and evaluation of evapo-
transpiration models for irrigation scheduling. Trans ASAE 21(1):
88–96