AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx

available at www.sciencedirect.com

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

Reviewing SEBAL input parameters for assessing

evapotranspiration and water productivity for the
Low-Middle São Francisco River basin, Brazil
Part A: Calibration and validation

A.H. de C. Teixeira a,*, W.G.M. Bastiaanssen b,c, M.D. Ahmad d,e, M.G. Bos f
a
Embrapa Semi-Árido, P.O. Box 23, 56302-970 Petrolina, PE, Brazil
b
WaterWatch, Generaal Foulkesweg 28, 6703 BS, Wageningen, The Netherlands
c
Delft University of Technology, Stevinweg 1, 2600 GA, Delft, The Netherlands
d
Formerly with International Water Management Institute, P.O. Box 2075, Colombo, Sri Lanka
e
Currently at CSIRO Land and Water, GPO Box 1966, Canberra act 2601, Australia
f
Formerly with International Institute for Geo-Information Science and Earth Observations, P.O. Box 6, 7500 AA,
Enschede, The Netherlands

article info abstract

Article history: There is a growing interest in quantifying regional scale actual evapotranspiration (ET) for
Received 13 May 2008 water accounting and for water productivity assessments at river basin scale. Methods that
Received in revised form provide point values fail to describe the situations at larger scales. Remote sensing mea-
3 September 2008 surements can be used at different spatial scales. This paper applies the theory of the
Accepted 24 September 2008 Surface Energy Balance Algorithm for Land (SEBAL). SEBAL was originally derived for Egypt,
Spain and Niger [Bastiaanssen, W.G.M., 1995. Regionalization of surface flux densities and
moisture indicators in composite terrain: a remote sensing approach under clear skies in
Keywords: Mediterranean climates. Ph.D. dissertation, CIP Data Koninklijke Bibliotheek, Den Haag, The
Latent heat flux Netherlands. 273 pp.] and was calibrated and validated using ground measurements from
Sensible heat flux four flux sites and from seven agro-meteorological stations in the semi-arid region of the
Soil heat flux Low-Middle São Francisco River basin, Brazil. Measured parameters included surface
Net radiation albedo, surface temperature, atmospheric and surface emissivity, soil heat flux, surface
Evaporative fraction roughness, net radiation, air temperature gradients, sensible heat flux, latent heat flux,
Actual evapotranspiration evaporative fraction, and photosynthetically active radiation. The daily ET was estimated
(RMSE of 0.38 mm d1) for mixed agricultural and natural ecosystems. The improved
coefficients for the local conditions can now be used to study the impact of expanding
irrigated agriculture on the regional water balance and to quantify the water productivity of
irrigated horticulture that is the largest water consumer in the Brazilian semi-arid region.
Both applications are described in an accompanying paper (Part B).
# 2008 Elsevier B.V. All rights reserved.

* Corresponding author. Tel.: +55 87 38621711; fax: +55 87 38621744.

E-mail address: [email protected] (A.H. de C. Teixeira).
0168-1923/$ – see front matter # 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.agrformet.2008.09.016

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

2 agricultural and forest meteorology xxx (2008) xxx–xxx

1. Introduction and Becker (1991), Kustas and Norman (1996), Bastiaansssen

et al. (1998a), Bastiaanssen et al. (1999), Kustas et al. (2004),
Irrigated agriculture in the semi-arid region of the São Courault et al., 2005 and Allen et al. (2007a) provided reviews of
Francisco River basin (North-east Brazil) constitutes an the progressive development of remote sensing algorithms for
important activity for livelihood of rural communities. The the estimation of ET during the last 20 years.
potential for fruit production in this region has been confirmed These reviews basically concluded that after these years of
by the continuous expansion of irrigated land. Vineyards and research, algorithms are sufficiently robust to be used for
mango orchards are the major crops. These land use changes water management. One of these algorithms is the Surface
affect the regional scale water balance. Energy Balance Algorithm for Land (SEBAL). SEBAL has been
Together with rainfall and runoff, the ET controls the applied to a variety of ecosystems (Bastiaansssen et al., 1998b;
availability of water at the Earth’s surface (McCabe and Wood, Bastiaanssen et al., 1999; Bastiaansssen et al., 2001; Bastiaans-
2006). The accurate determination of ET significantly reduces sen et al., 2002, 2005, 2008; Wang et al., 1998; Bastiaansssen,
uncertainties in the water balance of a (sub-)basin (Cleugh 2000; Farah and Bastiaanssen, 2001; Bastiaanssen and Ban-
et al., 2007), providing water managers with information on (i) dara, 2001; Ahmad and Bastiaanssen, 2003; Bastianssen and
water resources being consumed and thus not longer available Chandrapala, 2003; Chemin et al., 2004; Allen et al., 2005; El-
for downstream users, and (ii) water productivity, i.e. the Magd and Tanton, 2005; Akbari et al., 2007; Immerzeel et al.,
consumption of water in terms of biomass production per unit 2008).
of water (e.g. Steduto et al., 2007). SEBAL was introduced in the São Francisco River basin in
Field scale ET measurements on vineyards, mango orchard 2000 (Bastiaansssen et al., 2001). This introduction has
and natural vegetation (caatinga) were done in the Low-Middle prompted Embrapa (Brazilian Agricultural Research Corpora-
São Francisco River basin. Micro-meteorological methods tion) to use remote sensing for the up scaling of local water
were used providing point values for specific sites (Teixeira fluxes and water productivity in the basin. Although this
et al., 2007, 2008a,b). Direct extrapolations of data from algorithm was designed to calculate the energy balance at
individual flux sites to the regional scale can lead to biased regional scale using a minimum of ground data, local
estimates, because a few flux sites cannot provide a fair parameterization of any remote sensing equations can
sample of a larger area (Wylie et al., 2003). improve the accuracy of the model (Duchemin et al., 2006).
According to Nagler et al. (2005), species-level vegetation This paper combined satellite data, field measurements of
maps and species-specific algorithms would be needed for the surface radiation and energy balances and agro-meteor-
scaling tower-based ET data to larger areas. Species maps are ological stations to review the various SEBAL steps. During this
difficult to construct even with high-resolution aerial photo- process, the accuracy of individual empirical relationships, as
graphy. Pelgrum and Bastiaanssen (1996) confirmed this well as the final validation of daily regional evapotranspira-
difficult with the use of field data from 13 flux towers in a tion, was investigated. The reviewed parameters were; surface
heterogeneous landscape in central Spain with irrigated crops, albedo, surface temperature, surface and atmospheric emis-
rainfed crops and natural vegetation being insufficient, sivities, roughness length for momentum transport, net
because the spatial variations of these latent heat fluxes are radiation, soil heat flux, air temperature gradient, sensible
dictated by plant spacing, leaf area index, soil wetness, etc. heat flux, latent heat flux, and photosynthetically active
Therefore, data from flux stations are only a first estimate of radiation. The relevant equations were adapted for the semi-
ET from contrasting ecosystems (Leuning et al., 2005). arid conditions in the Low-Middle São Francisco River basin,
Allen et al. (2007b) summarized a number of examples of and then applied to each individual Landsat image.
remote sensing applications for irrigation management issues
in the Western U.S. Bastiaanssen and Harshadeep (2005)
reported a range of applications in Asia. The major advantage 2. Material and methods
of remote sensing is that the ET at a large scale can be
computed on pixel-to-pixels basis applying a consistent set of 2.1. Brief outline of SEBAL major principles
equations that utilize unique spectral radiances for each pixel.
The spatial and temporal distribution of ET can be mapped SEBAL requires spatially distributed, visible, near-infrared
from remote sensing techniques without going through and thermal infrared data together with routine weather
excessive ground truth data collection (e.g. Franks and Beven, data. The algorithm computes net radiation (Rn), sensible
1999). According to Nagler et al. (2007), remotely sensed heat flux (H) and soil heat flux (G) for every pixel and the latent
vegetation indices, obtained as a time series over a growing heat flux (lE) is acquired as a residual in energy balance
season, and micrometeorological data can be used to extra- equation. This is accomplished by first computing the
polate plot level measurements of ET and water productivity regional surface radiation balance, followed by the regional
over larger landscapes units. Another advantage of remote surface energy balance. The schematic overview to convert
sensing is that ET is sampled over many fields growing the spectral radiances into the net radiation using Landsat
same crop or over natural bush land. In this way it is possible images is presented in Fig. 1.
to determine true averages for a certain ecosystem or region The net shortwave radiation available at the earth surface
(Tasumi and Allen, 2007). depends on the incoming solar radiation (RG) and the surface
Jackson et al. (1977) pioneered in Arizona with the albedo (a0). The second parameter is calculated from satellite-
determination of ET by thermal remote sensing data, using measured spectral radiances for each narrow band, followed
infrared thermometry in wheat. Choudhury (1989), Schmugge by mathematical expressions for spectral integration and

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 3

Fig. 1 – Schematic flowchart for the radiation balance.

atmospheric corrections. Regional RG in the present research where the effects from the land surface roughness could be
was computed using interpolated measurements from pyr- eliminated. A height of 200 m was considered (u200). The first
anometers of seven agro-meteorological stations. estimate of u* is used—conjunctively with surface roughness
The incoming longwave radiation (RL#) is calculated using estimates—to infer the first values of aerodynamic resistance
the Stefan–Boltzmann equation with an empirically deter- (ra). Corrections for atmosphere stability are obtained itera-
mined atmospheric emissivity (ea) and the field measure- tively for each pixel (u*_corr), requiring a series of iterations to
ments of air temperature (T). Outgoing longwave radiation determine new values of u*_corr and ra_corr before obtaining
(RL") is obtained using the Stefan–Boltzmann equation with numerical stability.
an empirically determined surface emissivity (e0) and SEBAL computes the near-surface vertical difference in air
surface temperature (T0) acquired by the satellite after temperature DT that governs the sensible heat flux by
atmospheric correction. The final term in radiation balance assuming the linear relationship between this gradient and
equation, (1  e0)RL#, represents the fraction of RL# that is lost surface temperature (T0) together with the coefficients of this
from the surface due to reflection. The local calibration of relationship acquired following an internal calibration proce-
radiation balance terms is explained in detail in the next dure (see also Allen et al., 2007a). This calibration is image and
sections. occasion dependent.
The second step of the SEBAL is to compute the soil (G) and The algorithm uses two ‘‘anchor’’ pixels at which a value
the sensible (H) heat fluxes. The term G is calculated by its for H can be estimated on the basis of a priori knowledge
relation with the Rn by the ratio G/Rn. Most literature describes (Fig. 2) of the fluxes over dryland (dry pixel) and wet terrain
this last ratio to be a function leaf area index—LAI or (cold pixel). For this study, the cold pixel and the hot pixel
Normalized Difference Vegetation Index—NDVI (e.g. Clothier were considered on the plots of micro sprinkler irrigated
et al., 1986; Choudhury, 1989). In contrast with these mango orchard and natural vegetation, respectively, where
commonly accepted parameterizations, the role of T0 in the the eddy covariance systems were installed. The sensible
physical description of heat diffusion is recognized in the heat fluxes in these sites (Hcold and Hhot) were calculated
present research by incorporating it into the parametric applying the energy balance equation, and the annual
expression G/Rn. H is obtained by near-surface temperature generalized curves of the ratios ETa/ET0 together with the
gradients (Fig. 2). ET0 for the anchor pixels at satellite overpass time. The
According to Fig. 2, the first value for friction velocity (u*) is reference evapotranspiration (ET0) is the evapotranspiration
computed for neutral stability, using data from agro-meteor- rate from a reference surface, not short of water, which can
ological stations. The near-surface wind speed is converted to be a hypothetical grass surface with specific characteristics
a value at the blending height, i.e. a height above the stations (Allen et al., 1998).

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

4 agricultural and forest meteorology xxx (2008) xxx–xxx

Fig. 2 – Schematic flowchart for the energy balance of the SEBAL algorithm.

The instantaneous estimation of the latent heat flux (lE) TM band 6 provides data for longwave (thermal) radiation.
can be scaled to longer periods by using the evaporative The spectral radiance in band 6 (L6) of Landsat can be
fraction [lE/(Rn  G)] following Bastiaanssen et al. (2005) or converted into a radiation temperature applicable at the top
applying the ratio ETa/ET0 following Allen et al. (2007a). The of the atmosphere, the brightness temperature (Tsat), by
essential SEBAL steps are elucidated in Allen et al. (2007a) and inversion of Plank’s law in the 10.4–12.5 mm band width:
in Koloskov et al. (2007).
K2
Tsat ¼ (3)
ln ðK1 =ðL6 þ 1ÞÞ
2.2. Landsat measurements
where L6 is the uncorrected thermal radiance from the land
According to Fig. 1 the planetary albedo (ap_b) for each Landsat surface; and K1 (607.76 and 666.09 for Landsat 5 and 7, respec-
satellite band is: tively) and K2 (1260.56 and 1282.71 for Landsat 5 and 7, respec-
tively) are conversion coefficients (Markham and Barker, 1986;
Lb pd2 Landsat 7 Science Data User Data Handbook Chapter 11, 2002).
ap b ¼ (1) One Landsat 7 and nine Landsat 5 images were used in this
Ra b cos ’
study (Table 1).
where Lb is the spectral radiance for wavelengths of the band
b, d is the relative earth–sun distance; Ra b is the mean solar
exo-atmospheric irradiance (or atmospheric irradiance) for Table 1 – Satellite data over the period July 2001 to
January 2007.
each band (W m2 mm1) and w the solar zenith angle. The
broadband planetary albedo (ap) is calculated as the total sum Date DOY Satellite
of the different narrow-band ap b values according to weights 10 September 2001 253 Landsat 7
for each band (wb ). 04 October 2001 277 Landsat 5
X 06 July 2003 187 Landsat 5
ap ¼ wb ap b (2) 24 September 2003 267 Landsat 5
12 October 2004 286 Landsat 5
14 November 2004 328 Landsat 5
The weights for the different bands are computed as the
15 October 2005 288 Landsat 5
ratio of the amount of incoming shortwave radiation from the 16 November 2005 320 Landsat 5
sun in a particular band and the sum of incoming shortwave 30 July 2006 211 Landsat 5
radiation for all the bands. The bands 1–5 and 7 provide data 22 January 2007 022 Landsat 5
for the visible and near infrared bands used for surface albedo
DOY—Day of the year.
calculations.

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 5

Fig. 3 – The São Francisco River Basin and the semi-arid region around the town of Petrolina investigated. The agro-
meteorological stations (black) and the flux towers (white) on irrigated crops (wine grape, table grape, mango orchard) and
natural vegetation (caatinga) are indicated.

The resulting broadband planetary albedo and radiometric experiments as well in agro-meteorological stations at the
temperature from Eqs. (2) and (3) needs to be corrected vicinities of each site. Table 4 shows the averaged daily
atmospherically for acquiring the land surface counterparts. meteorological ground conditions during the overpasses days
of the Landsat satellite, as well the vegetated surface types
2.3. Field measurements considered for each image. These days cover periods with a
high (6.2 mm d1) and low (3.8 mm d1) ET0. The average value
The region with the field energy balance experiments and the of this last parameter for all days was 5.1 mm d1.
network of agro-meteorological stations used in this study is
shown in Fig. 3. The green areas in the study region represent
irrigated crops while the darker ones consist of natural 3. Calibration of the empirical relationships
vegetation. The thin blue lines are drainage lines.
The coordinates and details of each plot being equipped 3.1. Satellite-measured variables
with flux towers are shown in Table 2. The experimental
setups are described in Teixeira et al. (2007, 2008a,b). The 3.1.1. Surface albedo, a0
routine weather data from the seven automatic agro-meteor- The atmosphere disturbs the signal reaching the satellite
ological stations were used for SEBAL calculations and to sensor. The satellite registered radiances are, therefore,
compute the reference evapotranspiration (ET0) by the Pen- affected by the atmospheric interaction in the radiative
man-Monteith method (Allen et al., 1998). The stations are transfer path. Part of RG is scattered back to the satellite
equipped with pyranometers, anemometers, net radiometers, before it reaches the earth surface. A correction is deemed
sensors for air temperature and relative humidity and soil heat necessary to estimate surface albedo (a0). Following Menenti
flux plates. Table 3 resumes the data collection in all field et al. (1989), a simplified linear relationship of the following

Table 2 – Field energy balance experiments used to calibrate and validate SEBAL in Low-Middle São Francisco River basin.
Vegetation Location (latitude, longitude) Area (ha) Irrigation system Method of measurements
0 00 0 00
Wine grape (WG) 9816 01.73 S; 40811 32.62 W 4.13 Drip Bowen ratio
Table grape (TG) 98180 40.8400 S; 408220 29.4700 W 5.13 Micro sprinkler Bowen ratio
Mango (MG) 98220 32.2000 S; 408330 54.2300 W 11.92 Micro sprinkler Eddy covariance
Caatinga (CT) 98030 30.7100 S; 408190 45.2100 W N/A Not irrigated Eddy covariance

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

6 agricultural and forest meteorology xxx (2008) xxx–xxx

Table 3 – The field data over the period July 2001 to January 2007 in the four energy balance field experiments and agro-
meteorological stations.
Site Instrumentation

Wine grape  Tower–Bowen ratio system: gradients of air temperature and humidity.
 Radiation: short wave radiation (downward and upward) and net of all
wave radiation above the canopy.
 Microclimatic conditions: air temperature and relative humidity at canopy level and
gradient of wind speed above the canopy.
 Soil heat flux at a depth of 2 cm.
 Soil moisture at the depths of: 20, 40 and 60 cm

Table grape  Tower–Bowen ratio system: gradients of air temperature and humidity.
 Radiation: short wave radiation (downward) and net of all wave radiation above and
under the canopy.
 Microclimatic conditions: wind speed above the canopy (one level), air temperature and
relative humidity at canopy level.
 Soil heat flux at a depth of 2 cm.
 Soil moisture at the depths of: 20, 40 and 60 cm

Mango orchard  Tower–Bowen ratio and eddy correlation systems: gradients of air temperature and humidity;
fluctuations of wind speed, air temperature and absolute humidity; and frictional velocity.
 Radiation: short wave radiation (downward and upward) and net of all wave radiation.
 Microclimatic conditions: wind speed above the canopy (one level), air temperature and
humidity at canopy level.
 Soil heat flux at a depth of 2 cm.
 Soil moisture at the depths of: 20, 40, 60, 80, 100, 120 cm.

Caatinga  Tower–Eddy correlation system: fluctuations of wind speed, air temperature and absolute
humidity; and frictional velocity.
 Radiation: short wave radiation (downward and upward), long wave radiation (downward
and upward) and net of all wave radiation.
 Microclimatic conditions: wind speed above the vegetation (one level), air temperature and
humidity at leaves level.
 Soil heat flux at a depth of 2 cm.
 Air pressure
 Rainfall above the vegetation

Agro-meteorological stations Wind speed and direction, air temperature and relative humidity, incident short wave radiation,
net of all wave radiation and soil heat flux at a depth of 2 cm bellow grass.

type has been sought for correcting all atmospheric dis- Combination of field measurements of a0 with Landsat
turbances: calculations of ap resulted in physical meaningful results:
a = 0.61 and b = 0.08. The coefficient a is the two-way
a0 ¼ a ap þ b (4)
transmittance through the atmosphere, leading to an average

Table 4 – Daily values of weather variables on Landsat overpass days. The averaged data of seven agro-meteorological
stations and land surface types used for SEBAL calibration and validation are presented: extra-terrestrial solar radiation
(Ra); global solar radiation (RG), air temperature (T), relative humidity (RH), vapour pressure deficit (D), wind speed at 3 m
above the ground (u) and reference evapotranspiration (ET0). All values represent 24 h periods.
Day/year Surface type Ra (MJ m2 d1) RG (MJ m2 d 1) T (8C) RH (%) D (kPa) u (m s1) ET0 (mm d1)

253/2001 WG 35.7 25.0 27.4 52 1.8 3.0 4.4

2772001 WG 37.1 24.5 29.2 46 2.2 3.2 4.5
187/2003 TG, CT 28.7 21.5 23.5 54 1.3 1.9 3.8
267/2003 TG, CT 38.3 22.2 25.8 48 1.7 2.4 5.0
286/2004 MG, CT 37.1 26.7 28.1 46 2.1 3.4 6.0
318/2004 MG, CT 39.4 28.0 31.4 39 2.8 2.5 6.2
288/2005 MG, CT 37.0 27.4 28.1 45 2.1 3.0 5.8
320/2005 MG, CT 39.4 28.0 28.2 48 2.0 2.6 6.0
211/2006 MG, CT 31.1 21.8 27.0 53 1.7 2.0 4.0
022/2007 MG, CT 40.7 27.7 29.6 46 2.3 1.4 5.2

Mean – 36.5 25.3 27.4 52 1.8 3.0 5.1

Irrigated crops: Wine grape (WG), table grape (TG) and mango orchard (MG).
Natural vegetation: Caatinga (CT).

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 7

Fig. 4 – Relationship between surface albedo from the field energy balance experiments and their counterparts from satellite
overpass measurements: (a) instantaneous values of a0 and ap; (b) daily (subscript 24) and instantaneous (subscript inst)
values of a0.

single pathway atmospheric transmissivity of a0.5 = 0.78. The found excellent agreement for dense fully closed wheat.
latter is a value for clear sky radiation (e.g. Allen et al., 1998). However, when this crop is sparse, large differences were
The instantaneous regression analysis is shown in Fig. 4a. observed; especially when LAI was lower than 2. For sparse or
Eq. (4) was applied to all Landsat images for determining composite vegetation, the radiometric surface temperature
the spatial variation of a0. The Landsat overpass measure- was found to be much higher than T0 from flux inversion
ments yielded instantaneous values of surface albedo that (Kalma and Jupp, 1990).
were systematically lower than those for 24 h. Hence, a second In addition to the differences between radiometric and
regression equation was necessary to retrieve the daily from aerodynamic temperatures, part of the spectral thermal
instantaneous values (Fig. 4b). The relationship depicted in infrared radiation from the land surface is absorbed by the
Fig. 4b has been considered in the computations of the 24-h atmosphere and emitted upwards (path radiance). The
values of net radiation required for the determination of daily thermal radiation measured by satellite sensors thus needs
energy partition and ET. to be corrected for both atmosphere emission and the
difference between radiometric and aerodynamic surface
3.1.2. Surface temperature, T0 temperature. To correct these conjugated effects the field
The standard equations for heat transfer require aerodynamic results from Eq. (5) were used to fit a linear relationship
surface temperature. The determination of the components of between T0 (T0 = T + DT) and Tsat:
the surface radiation and energy balances according to the
T0 ¼ aTsat þ b (6)
parameterizations outlined required the vertical temperature
difference (DT) between two heights, which was derived from The coefficients for multiple Landsat days appear to be
the radiation temperature measured by the field data sets, H, a = 1.07 and b = 20.17 (Fig. 5). Eq. (6) was applied to all Landsat
air temperature above the canopies, and the aerodynamic images to obtain the spatial variation of T0 to be used in solving
resistance—ra (Smith et al., 1989). H.
It is realized that a physically based atmospheric correction
Hra
DT ¼ (5) would include transmittance and reflectance of spectral
ra c p
radiances (e.g. Schmugge et al., 1998), but for practical reasons
where ra and cp are the air density and air specific heat at Eqs. (4) and (6) are doable.
constant pressure, respectively. Field values of ra were deter-
mined by flux profile in vineyards (Teixeira et al., 2007) and by 3.2. Land measured variables
eddy covariance measurements in mango orchard and caa-
tinga (Teixeira et al., 2008a,b). 3.2.1. Broadband apparent emissivity of the atmosphere, ea
International experiments have demonstrated that, other Measured RL# over caatinga—in combination with T—allowed
than for a thin surface such as bare soil or a short canopy, a us to inspect the instantaneous apparent emissivity of the
difference of several degrees can be observed between atmosphere (ea). The SEBAL-based equation for the satellite
radiometric and aerodynamic surface temperature (Troufleau overpass time reads as:
et al., 1997). The latter implies that the radiometric tempera-
ea ¼ aðln t sw Þb (7)
ture from Landsat is an ill-provision of aerodynamic surface
temperature required for solving H.
Cleugh and Dunin (1995) compared aerodynamic surface where the regression coefficients become a = 0.942 and
temperatures (T0) with canopy radiometric temperatures and b = 0.103 as shown in Fig. 6, and tsw, the atmospheric trans-

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

8 agricultural and forest meteorology xxx (2008) xxx–xxx

Fig. 6 – Satellite overpasses relationship between the

broadband apparent atmospheric emissivity and
Fig. 5 – Relationship between surface temperature from the shortwave atmospheric transmissivity.
field energy balance experiments and their counterparts
from satellite overpass measurements.

NDVI values as mango orchard, but at a substantially lower

leaf area index (LAI).
missivity, is the ratio of instantaneous solar radiation The simplified approach below was tested for the mixture
measured at the ground surface to that for the top of the of agricultural and natural ecosystems:
atmosphere (RG/Ra). The subscript sw means shortwave   
NDVI
radiation. z0m ¼ exp a Þ þb (9)
a0
Note that the daily values of Eq. (7) are with a = 0.936 and
b = 0.114 slightly different (Teixeira et al., 2008b), and also The regression coefficients were a = 0.26 and b = 2.21
differ from those obtained for Idaho (a = 0.85; b = 0.09; source: (Fig. 8). The coefficient of determination is rather encouraging
Allen et al., 2000) and Egypt (a = 1.08; b = 0.265; source: (R2 = 0.92) for describing a difficult land surface parameter by
Bastiaanssen, 1995). In fact, the instantaneous values for some simplified remote sensing variables. Eq. (9) was scaled up
the Low-Middle São Francisco River basin are between those and used in calculating regional friction velocity (u*) and then
for Idaho and Egypt. regional H.
Eq. (7) was thereafter applied, together interpolated T, to all
Landsat images to obtain the regional long wave emitted from 3.2.4. Soil heat flux, G
the atmosphere (RL#) in the radiation balance (Fig. 1). Field measurements of G and Rn were used together with
satellite measurements of surface temperature (T0) in 8C,
3.2.2. Broadband surface emissivity, e0 surface albedo a0, and NDVI to test the following multiple
Field values of emitted long wave radiation–jointly with regression equation:
estimates of T0–gave us the opportunity to isolate and thus
G
to quantify the thermal infrared emissivity (e0). Following Van ¼ T0 ðaa0 þ bÞð10:98 NDVI4 Þ (10)
Rn
der Griend and Owe (1993), the results of e0 from radiation
balance experiments were correlated with NDVI from Landsat
images (Fig. 7):

e0 ¼ a ln NDVI þ b (8)

The values found were 0.059 and 1.004 for the coefficients a
and b, respectively. The original coefficients are a = 0.047 and
b = 1.009, hence slightly different (Bastiaanssen, 1995). The
values of e0 were used for the estimation of RL" in the radiation
balance (Fig. 1).

3.2.3. Surface roughness length for momentum transfer, z0m

Field values of roughness length for momentum transfer (z0m)
in natural vegetation and irrigated crops are described by
Teixeira et al. (2008b). To estimate regional values of z0m Allen
et al. (2007a) suggested a simplified expression based on a0 and
NDVI. The inclusion of these both indices helps to distinguish Fig. 7 – Relationship between broadband surface emissivity
between vegetation having different architecture but similar of vegetated surfaces and the Normalized Difference
values of NDVI. For example table grape can have the same Vegetation Index (NDVI).

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 9

Fig. 9 – Relationship between the ratio of soil heat flux (G)

Fig. 8 – Relationship between the roughness length for over net radiation (Rn) with surface temperature (T0),
momentum transport (z0m) and the ratio of the Normalized surface albedo (a0) and the Normalized Difference
Difference Vegetation Index (NDVI) to surface albedo (a0). Vegetation Index (NDVI).

The results reveal a = 0.11, b = 0.02 to be the best set of surface temperature difference (DT) for each pixel is given as
locally calibrated coefficients. These coefficients were (T1 – T2), the temperature difference between two heights (z1
obtained from the multiple regression of G/Rn with a0, T0 and z2). The regionalization of DT occurs by assuming a linear
and NDVI (Fig. 9). The original coefficients are different with relationship with T0:
a = 0.0074 and b = 0.0038 for the condition where they were
DT ¼ aT0 þ b (11)
found (Bastiaanssen, 1995). The differences can be explained
by the conditions of daily irrigation and bigger LAI of irrigated where a and b are the fitting coefficients.
fruit crops in the Low-Middle São Francisco River basin, The generalized annual curves of ET/ET0 for the time of
Brazil. G is a difficult term to evaluate and care should be used satellite overpass (9:30) of these two contrasting vegetation
in its computation by Eq. (10). The values should be checked types have been used to estimate H (and thus DT) at these
against actual measurements on the ground for specific specific locations (Fig. 10). There were two complete years of
conditions. flux data in mango orchard and caatinga, however, data of
2005 were used because during the year of 2004 unusual
3.2.5. Sensible heat flux, H storms occurred. These ET/ET0 curves were used for other
The computation of H by SEBAL algorithm requires two years together the values of ET0 for the satellite overpass days.
‘‘anchor’’ pixels, also referred to as the hot and the cold pixels. Fig. 10 shows that the lE of natural vegetation, caatinga—
In this study, the hot and cold pixels were represented by the for instance—is not negligible, but has an ET/ET0 value around
natural vegetation (caatinga) and irrigated crop (mango 0.20 during the dry season and even more than 0.90 during the
orchard), respectively. To compute H from Eq. (5), the near wet season.

Fig. 10 – Curves of (ET/ET0) for the cold (mango orchard) and hot (caatinga) pixels. Data for 2005 were taken.

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

10 agricultural and forest meteorology xxx (2008) xxx–xxx

The advantage of using known land use together with ET/ et al. (1971) for solving the integrated stability functions of
ET0 values is that the assumptions of zero sensible or latent temperature (Ch) and momentum (Cm). The following expres-
heat fluxes are not strictly required as originally (Bastiaans- sions were used for unstable situations:
sen, 1995), which makes it possible to calculate regional
ET also during the rainy period.      1=4
zd 1 þ x2 zd
In the cold pixel (irrigated mango orchard) the value of Hcold Yh ¼ 2 ln ; with x ¼ 1  16 (15)
L 2 L
could be approximated as:

       
ET zd 1þx 1 þ x2 p
Hcold ¼ Rn cold  Gcold  ET0 (12) Cm ¼ 2 ln þ ln  2 arctan ðxÞ þ (16)
ET0 cold L 2 2 2

where (ET/ET0)cold is the ratio valid for the mango orchard where x is an empirical parameter, z for Ch was considered 5 m
according to the time of the year (Fig. 10), Rn_cold is the net and for Cm it was considered 200 m. L was calculated as:
radiation, Gcold is the soil heat flux and ET0 is the reference
evapotranspiration at satellite overpass for the place where ra c p u3 T
the flux tower was in the irrigated mango orchard. The energy L¼ (17)
kgH
balance terms were taken from satellite results, while for ET0
the weather values at the agro-meteorological stations were where ra is the air density (kg m3), cp is the air specific heat
interpolated by using the seven agro-meteorological stations. (1004 J kg1 K1), T is the air temperature (K), g is the gravitational
At the hot pixel, the values of Hhot were calculated in the constant (9.81 m s2), and H is the sensible heat flux (W m2).
same way using values of (ETa/ET0)hot valid for caatinga The friction velocity (u* expressed in m s1) by:
according to the time of the year (Fig. 10); and net radiation
(Rn_hot), soil heat flux (Ghot) and reference evapotranspiration uk
u ¼ (18)
(ET0) for the place where the flux tower was in the natural ln ðz  d=z0m Þ  C m
vegetation:
where u is the wind speed at 200 m (m s1), k is von Karman’s
 
ET constant (0.41), z0m is the roughness length for each pixel (m), k
Hhot ¼ Rn hot  Ghot  ET0 (13)
ET0 hot is von Karman’s constant (0.41), and Cm is the stability correc-
tion for momentum transport at the height z = 200 m.
The values of Hcold and Hhot were used to infer DT applying The regional values for roughness length for moment
Eq. (5) in an iterative process starting with neutral conditions transfer was obtained from Eq. (9) and corrected value for ra
till stabilization of DT at the cold and the hot pixels after was computed during each iteration using Eq. (14) for z = 5 m.
successive atmospheric stability corrections for ra. The Next, the final corrected value for H was obtained at each pixel
calculations of the regression coefficients of DT versus T0 (see Fig. 2), which was used in the computation of the regional
(Eq. (11)) were done in the spreadsheet and were variable for instantaneous latent heat flux.
each Landsat image acquisition day. The a and b values based
on degrees Kelvin varied from 0.54 to 1.16 and from 160.8 to 3.2.6. Evaporative fraction, EF
348.2, respectively, found for the images of January (lowest a According to our field data for mango orchard and natural
and highest b) and November (highest a and lowest b), vegetation, the instantaneous evaporative fraction at daylight
representing the wettest and the warmest months of the time was lower than for the 24 h periods, while vineyards
year, respectively (Fig. 11). Eq. (11) and the image date specific results showed no big differences between these time scales
regression coefficients were applied to all Landsat images for (Teixeira et al., 2008b). A straightforward regression equation
acquirement of H. The values of T were interpolated from the 7 was then applied to scale the instantaneous ET to daily values
agro-meteorological stations. according to:
The values of H for each pixel were calculated using Eq. (5)
ET24 ¼ aEFisnt Rn 24 (19)
first assuming neutral conditions. The drag force between land
and atmosphere is accounted for ra, and this parameter where ET24 and Rn_24 are the actual evapotranspiration and the
required the surface roughness conditions z0h and friction net radiation for 24 h (mm d1) and a is a correction coefficient,
velocities u* to be specified: which was found to be 1.18 for the Brazilian semi-arid condi-
tions. Note that the EFinst represents the instantaneous value
ln ðz  dÞ=z0h  ch ððz  dÞ=LÞ
ra ¼ (14) at satellite overpass.
ku
The net radiation for 24 h was calculated with albedo
where z0h (m) is the roughness length governing transfer of values (a0_24), global solar radiation (RG_24) and atmospheric
heat and vapour away from the land surface into the atmo- transmissivity (tSW_24) for this time scale, applying the Slob
sphere, k is the von Karman’s constant (0.41), z (m) is the equation (Bruin de, 1987; Bruin de and Stricker, 2000):
reference height taken as 5 m, d (m) is the displacement height
taken as 4.67z0m, ch is the stability correction due to buoyancy Rn 24 ¼ ð1  a0 24 ÞRG 24  at SW 24 (20)
and L is Obukhov length.
The atmospheric surface-layer similarity theory was The coefficient a averaged 143 for our semi-arid conditions
applied, using universal functions suggested by Businger with R2 = 0.80. This value is higher than the originally

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 11

suggested value of 110 found for the Netherlands. The 3.3. Photosynthetically active radiation, PAR
difference can be ascribed to different thermal conditions
between the macroclimates. As it varied as a function of T a The photosynthetically active radiation (PAR) is of paramount
regression equation between a and T was applied to scale Rn_24 importance for the computation of biomass production. PAR
(Teixeira et al., 2008b). (0.4–0.7 mm) is a fraction of RG (0.3–3 mm) that can be

Fig. 11 – Relationship between the vertical temperature difference and the surface temperature in the cold and hot pixels for
Day/year 253/2001, 277/2001, 187/2003, 267/2003, 286/2004, 318/2004, 288/2005, 320/2005, 211/2006 and 022/2007.

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

12 agricultural and forest meteorology xxx (2008) xxx–xxx

Table 5 – Summary of the instantaneous regression

analysis for satellite and land surface variables: surface
albedo—a0; planetary albedo—ap; surface temperature—
T0; brightness temperature—Tsat; atmospheric emissiv-
ity—ea, shortwave atmospheric emissivity—tsw; surface
emissivity—e0, roughness length for momentum trans-
fer—z0m, soil heat flux—G; net radiation—Rn; photo-
synthetically active radiation—PAR.
Variable Equation a b R2

a0 a0 = aap + b 0.61 0.08 0.90

T0 T0 = aTsat + b 1.02 20.2 0.93
ea ea = a(ln tsw)b 0.94 0.10 0.75
e0 e0 = a ln NDVI
h + b  i 0.05 1.01 0.83
z0m z0m ¼ exp a NDVI a0 þb 0.26 2.21 0.92
G G 4
Rn Rn ¼ T0 ðaa0 þ bÞð1  0:98NDVI Þ 0.11 0.02 0.87
PAR PAR = aRG 0.44 – 0.99
Fig. 12 – 24 h Relationship between photosythetically active
radiation (PAR) and global solar radiation (RG).
and ozone amount, among others (e.g. Froin and Pinker, 1995;
Moran et al., 1995).

PAR ¼ aRG (21)

potentially absorbed by chlorophyll for photosynthesis in the
plants. Measurements of RG and PAR (Kipp & Zonen, Delft, the A value of a = 0.44 is then recommended in the semi-arid
Netherlands) above natural vegetation allowed the parame- region of Sao Francisco River basin.
terizations between these two solar variables (Fig. 12). The A summary for all instantaneous regression analysis is
coefficient a was found to be 0.44, being slightly lower than given in Table 5. All relationships presented high values of
normally reported values (0.45 to 0.50). The variability of the coefficient of determination, being the lowest for apparent
PAR/RG fraction is usually ascribed to visibility, optical depth atmospheric emissivity.

Fig. 13 – Relationship between instantaneous energy balance components obtained from SEBAL (sat) and from field energy
balance measurements (field): (a) Rn—net radiation; (b) lE—latent heat flux; (c) H—sensible heat flux; and (d) G—soil heat
flux. The black and white symbols represent natural vegetation and irrigated crops, respectively.

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 13

Fig. 14 – Relationships of daily actual evapotranspiration obtained from SEBAL and field measurements (ETsat and ETfield): (a)
considering EF similar for instantaneous and daily scales and (b) applying the correction factor for EF. The black and white
symbols are for natural vegetation and irrigated crops, respectively.

4. Validation of the energy balance the field measurements have their own sources of errors, the
components and actual evapotranspiration accuracy of the actual study is considered satisfactory.
The results of regional ET and PAR were then scaled up and
The SEBAL procedure has been applied to all 10 different used for regional actual evapotranspiration, biomass produc-
Landsat images using the calibrated empirical relationships tion and crop water productivity analyses in semi-arid region
based on field measurements. The satellite energy balance of São Francisco River basin and discussed in the accompany-
results—and notably the fluxes—were validated with field ing paper (Part B).
measurements. Fig. 13 shows scatter plots for instantaneous
satellite and field measurements for Rn; for lE; for H; and for G
involving all vegetation types studied. 5. Conclusions
The results showed excellent agreements on Rn and lE. The
coefficient of determination for Rn was R2 = 0.94, with a root Field energy balances in natural vegetation and irrigated crops
mean square error (RMSE) of 17.5 W m2 and an average have been used to calibrate and validate the SEBAL algorithm.
deviation of 1%. The results for lE were R2 = 0.93, average The satellite images covered the semi-arid region of the Low-
deviation of 10% and a RMSE of 33.8 W m2. The differences Middle São Francisco River basin, North-eastern Brazil, from
occur typically for the higher values of lE. The latter is a result 2001 to 2007.
of the bias that is observed at the lower end H-values. Good The ground data consisted of radiation and energy
agreements are achieved for H and G. For H, the R2 was of 0.83. balances from four field experiments and seven agro-
Part of the deviations could be attributed to the estimations of meteorological stations. It is one of the first datasets that
z0m. The RMSE was of 41.8 W m2, with an average deviation could be used to validate individual steps of SEBAL since the
3%. Deviations are higher in the lower end of H values–thus development of the model.
centred on the irrigated pixels. For G, the local calibration Most of the expressions required a slight correction for
yielded a R2 of 0.81, with the RMSE of 13.3 W m2 and an acquiring agreement with the field data. This does not
average deviation 6% occurring typically at the lower end of G necessarily imply that the use of default equations will lead
values. to inappropriate results of actual evapotranspiration, because
As instantaneous values of actual evapotranspiration are SEBAL self-calibrates the value of sensible heat flux.
not interesting for water managers, the daily 24 h values were The coefficients of determination were for: surface albedo
computed as well (Fig. 14a). The coefficient of 1.18 of the (R2 = 0.90), daytime integration surface albedo (R2 = 0.95),
regression equation was applied in Eq. (19) (Fig. 14b). After the surface temperature (R2 = 0.93), apparent emissivity of the
application of this regression coefficient, the relation between atmosphere (R2 = 0.75), surface emissivity (R2 = 0.83), instan-
the SEBAL and field measurements of daily ET presented then taneous net radiation (R2 = 0.94), integration of net radiation
a R2 of 0.91 and a RMSE of 0.38 mm d1 with a very small (R2 = 0.80), surface roughness for momentum transport
deviation of 1% (ETsat = 1.01ETfield). (R2 = 0.92), soil heat flux/net radiation fraction (R2 = 0.87),
According to Bastiaanssen et al. (2008), the SEBAL model instantaneous soil heat flux (R2 = 0.80), instantaneous sensible
has been validated in grapes, peaches and almonds from heat flux (R2 = 0.83), integration of evaporative fraction
Spain, Turkey and California. These validations revealed that (R2 = 0.91), instantaneous latent heat flux (R2 = 0.93), integra-
accumulated values of ET predicted by the model are within tion of actual evapotranspiration (R2 = 0.91) and photosynthe-
several percent from the measured values. Considering that tically active radiation (R2 = 0.99). The apparent emissivity of

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

14 agricultural and forest meteorology xxx (2008) xxx–xxx

the atmospheric is one of the most problematic parameters, Monograph 108, American Geophysical Union (special issue
and this has been known for some time. after joint AGU Chapman/SSSA Outreach Conference
Riverside): 179–196.
The direct comparison between field measured latent heat
Bastiaansssen, W.G.M., 2000. SEBAL-based sensible and latent
flux and the application of calibrated equations in the SEBAL
heat fluxes in the irrigated Gediz Basin, Turkey. J. Hydrol.
algorithm to estimate actual evapotranspiration revealed that 229, 87–100.
it is essential to distinguish instantaneous and daily time Bastiaanssen, W.G.M., Bandara, K.M.P.S., 2001. Evaporative
scales. For instantaneous values it was required to apply the depletion assessments for irrigated watersheds in Sri
hot and cold pixel calibration for every individual image. For Lanka. Irrig. Sci. 21 (1), 1–15.
daily scale, the values of the instantaneous evaporative Bastiaansssen, W.G.M., Brito, R.A.L., Bos, M.G., Souza, R.A.,
Cavalcanti, E.B., Bakker, M.M., 2001. Low cost satellite data
fraction needed to be adjusted.
for monthly irrigation performance monitoring:
It was useful to use curves of the ratio of actual to reference benchmarks from Nilo Coelho, Brazil. Irrig. Drain. Syst. 15,
evapotranspiration instead the average values of the pixel 53–79.
with the lowest and highest surface temperatures in the Bastiaanssen, W.G.M., Ahmed, M.-ud-D., Chemin, Y., 2002.
calibration processes as this ratio changes during the year. Sattellite surveillance of evaporative depletion across the
Indus Basin. Water Resour. Res. 38, 1273–1282.
Bastianssen, G.M.W., Chandrapala, L., 2003. Water balance
references variability across Sri Lanka for assessing agricultural and
environmental water use. Agric. Water Manage. 58 (2),
171–192.
Bastiaanssen, W.G.M., Harshadeep, N., 2005. Managing scarce
Ahmad, M.D., Bastiaanssen, W.G.M., 2003. Retrieving soil water resources in Asia: the nature of the problem and can
moisture storage in the unsaturated zone from satellite remote sensing help? Irrig. Drain. Syst. 19 (3–4), 269–284.
imagery and bi-annual phreatic surface fluctuations. Irrig. Bastiaanssen, W.G.M., Noordman, E.J.M., Pelgrum, H., Davids,
Drain. Syst. 17, 141–161. G., Allen, R.G., 2005. SEBAL for spatially distributed ET under
Akbari, M., Toomanian, N., Droogers, P., Bastiaanssen, W.G.M., actual management and growing conditions. J. Irrig. Drain.
Gieske, A., 2007. Monitoring irrigation performance in Eng. 131, 85–93.
Esfahan, Iran, using NOAA satellite imagery. Agric. Water Bastiaanssen, W.G.M., Pelgrum, H., Soppe, R.W.O., Allen, R.G.,
Manage. 88, 99–109. Thoreson, B.P., Teixeira, A.H. de C., 2008. Thermal infrared
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop technology for local and regional scale irrigation analysis in
evapotranspiration. Guidelines for computing crop water horticultural systems. ISHS Acta Hort. 792.In: Proc. Vth IS on
requirements. FAO Irrigation and Drainage Paper 56, Rome, Irrigation of Horticultural Crops.
Italy, 300 pp. Bruin de, H.A.R., 1987. From Penman to Makkink. In: Hooghart,
Allen, R.G., Hartogensis, O., de Bruin, H.A.R., 2000. Long-Wave J.C. (Ed.), Proceedings and Information: TNO Commitee on
Radiation over Alfafa During the RAPID Field Campaign in Hydrological, vol. 39, Gravennhage, The Netherlands, pp.
Southern Idaho. Research Report, Kimberly, Univ. of 5–31.
Idaho, ID. Bruin de, H.A.R., Stricker, J.N.M., 2000. Evaporation of grass
Allen, R.G., Tasumi, M., Morse, A., Trezza, R., 2005. A Landsat- under non-restricted soil moisture conditions. Hydrol. Sci. J.
based energy balance and evapotranspiration model in 45 (3), 391–406.
Western US rights regulation and planning. Irrig. Drain. Businger, J.A., Wyngaard, J.C., Izumi, Y., Bradley, E.F., 1971. Flux-
Syst. 19, 251–268. profile relationships in the atmospheric surface layer. J.
Allen, R.G., Tasumi, M., Trezza, R., 2007a. Satellite-based energy Atm. Sci. 28, 189–191.
balance for mapping evapotranspiration with internalized Chemin, Y., Platonov, A., UI-Hassan, Abdullaev, I., 2004. Using
calibration (METRIC)–Model. J. Irrig. Drain. Eng. ASCE 133 (4), remote sensing data for water depletion assessment at
380–394. administrative and irrigation-system levels: case study of
Allen, R.G., Tasumi, M., Morse, A., Trezza, R., Wright, J., the Ferghana Province of Uzbekistan. Agric. Water Manage.
Bastiaanssen, W.G.M., Kramber, W., Lorite, I., Robson, C.W., 64, 183–196.
2007b. Satellite-based energy balance for mapping Choudhury, B.J., 1989. Monitoring global land surface using
evapotranspiration with internalized calibration (METRIC)– Nimbus-7 37 GHz data. Theory and examples. Int. J. Rem.
applications. J. Irrig. Drain. Eng. ASCE 133 (4), 395–405. Sens. 10, 1579–1605.
Bastiaanssen, W.G.M., 1995. Regionalization of surface flux Cleugh, H.A., Dunin, F.X., 1995. Modeling sensible heat fluxes
densities and moisture indicators in composite terrain: a from a wheat canopy: evaluation of resistance energy
remote sensing approach under clear skies in balance model. J. Hydrol. 164, 127–152.
Mediterranean climates. Ph.D. dissertation, CIP Data Cleugh, H.A., Leuning, R., Mu, Q., Running, S.W., 2007. Regional
Koninklijke Bibliotheek, Den Haag, The Netherlands. 273 pp. evaporation estimates from flux tower and MODIS satellite
Bastiaansssen, W.G.M., Menenti, M., Feddes, R.A., Holtslag, data. Rem. Sens. Envir. 106, 285–304.
A.A.M., 1998a. A remote sensing surface energy balance Clothier, B.E., Clawson, K.L., Pinter, P.J., Moran, M.S., Reginato,
algorithm for land (SEBAL) 1. Formulation. J. Hydrol. 212– R.J., Jackson, R.D., 1986. Estimating of soil heat flux from net
213, 198–212. radiation during the growth of alfafa. Agric. For. Meteorol.
Bastiaansssen, W.G.M., Pelgrum, H., Wang, J., Ma, Y., Moreno, 37, 319–379.
J.F., Roerink, G.J., Wal van der, T., 1998b. A remote sensing Courault, D., Seguin, B., Olioso, A., 2005. Review on estimation of
surface energy balance algorithm for land (SEBAL). 2. evapotranspiration from remote sensing data: from
Validation. J. Hydrol. 213–229. empirical to numerical modeling approaches. Irrig. Drain.
Bastiaanssen, W.G.M., Sakthivadivel, R., Van Dellen Van, A., Syst. 19, 223–249.
1999. Spatially delineating actual and relative Duchemin, J., Hadria, R., Er-Raki, S., Boulet, G., Maisongrande,
evapotranspiration from remote sensing for modelling non- P., Chehbouni, A., Escadafal, R., Ezzahar, J., Hoedjes, J.,
point pollutants, in (eds) Corwin, D. et al., Assessment of Karroui, H., Khabba, S., Mougenot, B., Olioso, A., Rodriguez,
non-source pollution in the vadose zone, Geophysical J.-C., Simonneaux, V., 2006. Monitoring wheat phenology

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016
AGMET-3973; No of Pages 15

agricultural and forest meteorology xxx (2008) xxx–xxx 15

and irrigation in Central Morocco: on the use of relationship MODIS and data from eddy covariance and Bowen ratio flux
between evapotranspiration, crops coefficients, leaf area towers. Rem. Sens. Envir. 97, 337–351.
index and remotely-sensed vegetation indices. Agric. Water Nagler, P.L., Glenn, E.P., Kim, H., Emmerich, W., Scott, R.L.,
Manage. 79, 1–27. Huxman, T.E., Huete, A.H., 2007. Relationship between
El-Magd, I.A., Tanton, T.W., 2005. Remote sensing and GIS for evapotranspiration and precipitation pulses in a semiarid
estimation of irrigation crop water demand. Int. J. Rem. rangeland estimated by moisture flux towers and MODIS
Sens. 26 (10), 2359–2370. vegetation indices. J. Arid Env. 70 (3), 443–462.
Farah, H.O., Bastiaanssen, W.G.M., 2001. Impact of spatial Pelgrum, H., Bastiaanssen, W.G.M., 1996. An inter comparison of
variations of land surface parameters on regional techniques to determine the area-averaged latent heat flux
evaporation: a case study with remote sensing data. Hydrol. from individual in situ observations: a remote sensing
Process. 15 (9), 1585–1607. approach using the European Field Experiment in a
Franks, S.W., Beven, K., 1999. Conditioning a multiple patch Desertification-Threatened Area delta. Water Resour. Res. 9,
SVAT model using uncertain time-space estimates of latent 2775–2786.
heat fluxes as inferred from remotely sensed data. Water Schmugge, T., Becker, F., 1991. In: Andre, J.C., Schmugge, T.
Resour. Res. 35, 2751–2761. (Eds.), Remote Sensing Observations of the Monitoring of
Froin, R., Pinker, R.T., 1995. Estimating photosynthetically Land Surface Fluxes and Water Budget in Determination of
radiation (PAR) at the earth’s surface from satellite Land Surface Fluxes during HAPEX. Springer Verlag, pp.
observations. Rem. Sens. Envir. 51, 98–107. 337–347.
Immerzeel, W.W., Gaur, A., Zwart, S.J., 2008. Integrating remote Schmugge, T., Hook, S.l., CoII, C., 1998. Recovering surface
sensing and a process-based hydrological model to evaluate temperature and emissivity from thermal infrared
water use and productivity in a South Indian catchment. multispectral scanner data. Rem. Sens. Envir. 65, 121–131.
Agric. Water Manage. 95 (1), 11–24. Smith, R., G.C., Barrs, H.D., Meyer, W.S., 1989. Evaporation from
Jackson, R.D., Reginato, R.J., Idso, S.B., 1977. Wheat canopy irrigated wheat estimated using radiative surface
temperatures; a practical tool for evaluating water temperature: an operational approach. Agric. For. Meteorol.
requirements. Water Resour. Res. 13, 651–656. 48, 331–344.
Kalma, J.D., Jupp, D.L.B., 1990. Estimating evaporation from Steduto, P., Kijne, J.W., Hanjra, M.A., Binibraban, P.S., 2007.
pasture using infrared thermometry: evaluation of Pathways for increasing agricultural water productivity. In:
one layer resistance model. Agric. For. Meteorol. 51, Molden, D. (Ed.), Water for Food, Water for Life: A
223–246. Comprehensive Assessment of Water Management in
Koloskov, G., Mukhamejanov, Kh., Tanton, T.W., 2007. Monin– Agriculture. International Water Management Institute,
Obukhov length as a cornerstone of the SEBAL calculations Earthscan, and Colombo, London, pp. 279–310.
of evapotranspiration. J. Hydrol. 335 (1–2), 170–179. Tasumi, M., Allen, R.G., 2007. Satellite-based ET mapping to
Kustas, W.P., Norman, J.M., 1996. Use of remote sensing for assess variation in ET with timing of crop development.
evapotranspiration monitoring over land surface. IAHS Agric. Water Manage. 88, 54–62.
Hydrol. Sci. J. 4, 495–516. Teixeira, A.H. de C., Bastiaanssen, W.G.M., Bassoi, L.H., 2007.
Kustas, W.P., Li, F., Jackson, T.J., Prueger, J.H., MacPherson, J.I., Crop water parameters of irrigated wine and table grapes to
Wolde, M., 2004. Effects of remote sensing pixel resolution support water productivity analysis in Sao Francisco River
on modeled energy flux variability of croplands in Iowa. basin, Brazil. Agric. Water Manage. 94, 31–42.
Rem. Sens. Envir. 92 (4), 535–547. Teixeira, A.H. de C., Bastiaanssen, W.G.M., Moura, M.S.B.,
Landsat 7 Science Data User Handbook, Chapter 11, 2002–Data Soares, J.M., Ahmad, M–ud–D, Bos, M.G., 2008a. Energy and
Products http://ltpwww.gsfc.nasa.gov/IAS/handbook/ water balance measurements for water productivity
handbook_htmls/chapter11/chapter11.html. analysis in irrigated mango trees, northeast Brazil. Agric.
Leuning, R., Cleugh, H.A., Zegelin, S.J., Hughes, D., 2005. Carbon For. Meteorol. 148, 1524–1537.
and water fluxes over a temperate Eucalyptus forest and a Teixeira, A.H. de C., Bastiaanssen, W.G.M., Ahmad, M.D., Moura,
tropical wet/dry savanna in Australia: measurements and M.S.B., Bos, M.G., 2008b. Analysis of energy fluxes and
comparison with MODIS remote sensing estimates. Agric. vegetation-atmosphere parameters in irrigated and natural
For. Meteorol. 129, 151–173. ecosystems of semi-arid Brazil. J. Hydrol, doi:10.1016/
Markham, B.L., Barker, J.L., 1986. Landsat MSS and TM Post- j.jhydrol.2008.08.011.
calibration Dynamic Ranges, Exoatmospheric Reflectances Troufleau, D., Lhomme, J.-P., Monteny, B., Vidal, A., 1997.
and At-Satellite Temperatures: EOSAT Landsat Technical Sensible heat flux and radiometric surface temperature
Notes, vol. 1, pp. 3–8. over sparse sahelian vegetation: is the kB1 a relevant
McCabe, M., Wood, E.F., 2006. Scale influences on the remote parameter. J. Hydrol. 189, 815–838.
estimation of evapotranspiration using multiple satellite Van der Griend, A.A., Owe, M., 1993. Determination of
sensors. Rem. Sens. Envir. 105, 271–285. microwave vegetation optical depth and single scattering
Menenti, M., Bastiaanssen, W.G.M., Van Eick, D., 1989. albedo from large scale moisture and Nimbus/SMMR
Determination of hemispheric reflectance with thematic satellite observations. Int. J. Rem. Sens. 14 (10), 1975–1996.
mapper data. Rem. Sens. Envir. 28, 327–337. Wang, J., Bastiaanssen, W.G.M., Ma, Y., Pelgrum, H., 1998.
Moran, M.S., Maas, S.J., Pinter, P.J., 1995. Combining remote Aggregation of land surface parameters in the oasis-desert
sensing and modeling for estimating surface evaporation systems of Northwest China. Hydrol. Process. 12, 2133–2147.
and biomass production. Rem. Sens. Rev. 12, 335–353. Wylie, B., Johnson, D., Laca, E., Saliendra, N., Gilmanov, T., Reed,
Nagler, P.L., Scott, R.L., Westenburg, C., Cleverly, J.R., Glenn, E.P., B., Tieszen, L., Worstell, B., 2003. Calibration of remotely
Huete, A.R., 2005. Evapotranspiration on western U.S. rivers sensed, coarse resolution NDVI to CO2 fluxes in a sage-
estimated using the Enhanced Vegetation Index from brush-steppe ecosystem. Rem. Sens. Envir. 85s, 243–255.

Please cite this article in press as: Teixeira, A.H.C., et al., Reviewing SEBAL input parameters for assessing evapotranspiration and water
productivity for the Low-Middle São Francisco River basin, Brazil. Agric. Forest Meteorol. (2008),
doi:10.1016/j.agrformet.2008.09.016