Computers and Electronics in Agriculture 147 (2018) 12–17

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture

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

Original papers

A refined method for rapidly determining the relationship between canopy T

NDVI and the pasture evapotranspiration coefficient
⁎
Muhammad Shahinur Alam , David W. Lamb, Muhammad Moshiur Rahman
Precision Agriculture Research Group, University of New England, NSW 2351, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: The estimation of actual crop evapotranspiration (ETc) from any given land cover or crop type is important for
Crop coefficient (Kc) irrigation water management and agricultural water consumption analysis. The main parameter used for such
Normalized Difference Vegetation Index estimations is the crop coefficient (Kc). Spectral reflectance indices, such as the normalized difference vegetation
(NDVI) index (NDVI) and the crop coefficient of a specific crop or pasture canopy are important indicators of ‘vigour’,
Evaporation chamber
namely the photosynthetic activity and rate of biomass accumulation. Measuring both parameters simulta-
neously, with a view to understanding how they interact, or for creating optical, surrogate indicators of Kc is very
difficult because Kc itself is difficult to measure. In this study a portable enclosed chamber was used to measure
ETc of a pasture and subsequently calculated Kc from reference evapotranspiration (ETo) data derived from a
nearby automatic weather station (AWS). Calibration of the chamber confirms the suitability of the device to
measure the amount of water vapour produced by local plant evapotranspiration, producing a calibration factor
(C) close to 1 (C = 1.02, R2 = 0.87). The coincident NDVI values were measured using a portable active optical
sensor. In a test involving a pasture (Festuca arundinacea var. Dovey) at two different stages of growth in two
consecutive growing seasons, the NDVI and crop coefficients were observed to be strongly correlated (R2 = 0.80
and 0.77, respectively). A polynomial regression (R2 = 0.84) was found to be the best fit for the combined,
multi-temporal Kc-NDVI relationship. The main advantages of this method include the suitability of operating at
a smaller scale (< 1 m2), in real time and repeatability.

1. Introduction theoretical approach developed by Allen et al. (1998).

Since the actual crop evapotranspiration (ETc) includes the water
Knowledge of monitoring crop evapotranspiration is important for transpired by the crops as well as the water evaporated from the soil,
planning and management of water resources especially where crop the crop coefficient (Kc) can also be alienated into basal crop coefficient
water requirements exceeds natural precipitation (Gowda et al., 2008). (Kcb) (transpiration from crop only) and soil evaporation coefficient
Current techniques of measuring the actual evapotranspiration of a crop (Ke). Measuring Kc or the components Kcb and Ke (Kc = Kcb + Ke) is
in-situ involve indirect methods such as through energy balance or soil therefore the pathway for estimating actual crop evapotranspiration.
water balance models, or directly using lysimeters, which are often Considerable research has been reported in estimating Kc for irri-
complex and expensive (Allen et al., 1998). gation water management at the regional scale using the satellite de-
Evapotranspiration can also be measured theoretically from re- rived spectral reflectance data, where Kc has been related to the
ference crop evapotranspiration. For a given crop, the reference crop Normalized Difference Vegetation Index (NDVI = (NIR − R)/
evapotranspiration (ETo) is the maximum rate of evapotranspiration (NIR + R); where NIR and R are reflectance from near infrared and red
possible for a given environmental condition and is based purely on the bands respectively) (Allen et al., 2010; Bausch and Neale, 1987;
environmental evaporative demand and a ‘standard’ canopy surface. Duchemin et al., 2006; El-Shirbeny et al., 2014; Er-Raki et al., 2007;
The crop coefficient (Kc), the ratio of actual and reference crop eva- Hunsaker et al., 2005; Johnson and Trout, 2012; Neale et al., 1990;
potranspiration is unique to a specific canopy (morphology, phenology) Neale et al., 2003; Nouri et al., 2014). Johnson and Trout (2012) de-
and by knowing the crop coefficient for a crop or pasture community, veloped Landsat-5 satellite-based time-series of Kcb for four different
the actual canopy evapotranspiration (ETc) can be calculated with the crops using mean NDVI from several experimental plots on each clear-
help of reference crop evapotranspiration data calculated from the sky satellite overpass date. The NDVI was subsequently converted to

⁎
Corresponding author.
E-mail address: [email protected] (M.S. Alam).

https://doi.org/10.1016/j.compag.2018.02.008
Received 28 August 2017; Received in revised form 4 January 2018; Accepted 8 February 2018
0168-1699/ © 2018 Elsevier B.V. All rights reserved.
M.S. Alam et al. Computers and Electronics in Agriculture 147 (2018) 12–17

green fractional cover (fc) using a predefined formula, and finally to Kcb et al., 2008; Macfarlane and Ogden, 2012; McJannet et al., 1996), ve-
based on lysimeter experiments conducted by Bryla et al. (2010). Allen getated rangeland (Stannard and Weltz, 2006), as well as in some
et al. (2010) explored the linear relationships between ETrF (synon- pastures and emerging crops (McLeod et al., 2004; Stannard, 1988).
ymous with the crop coefficient Kc) and top of atmosphere NDVI from The dimension of the containment chamber, instrumentation and pro-
two years of data for Southern Idaho using Mapping Evapotranspiration cedures vary depending on the purpose and circumstances but the
at high Resolution with Internalized Calibration (METRIC) energy principle of instantaneous ET measurement is similar in each study.
balance process to near-real time estimates of ETrF from Landsat or Careful calibration and device operation is important for successful use
other satellite imagery. In both cases the concurrent measurement of of this technique. When operated correctly, the performance of this
related parameters could not be attained, and reliance has been placed chamber compared with other theoretical approaches of measuring ET
on secondary data or pre-defined formula. Bausch and Neale (1987) has demonstrated the potential value of using such methods for com-
used a hand-held radiometer and electronic module to measure the pleting evapotranspiration measurements (McLeod et al., 2004). To
radiance and found that the seasonal NDVI curve was curvilinear and date these devices have only been used to understand the water demand
resembled the basal crop coefficient (Kb) curve for corn. Duchemin and evapo-transpiration dynamics of the species in question.
et al. (2006) found linear relationships between NDVI and Kcb with a With the availability of high spatial and temporal resolution mul-
good accuracy ( ± 15%) using high spatial resolution Quickbird sa- tispectral satellite systems, and the emergence of ultra-high spatial re-
tellite imagery to measure top-of-atmosphere NDVI and in-situ eddy solution aerial platforms such as drones and miniature multispectral
covariance instrumentation to calculate the actual evapotranspiration. imaging devices as their payloads, there is considerable interest in
An ET-NDVI relationship for urban vegetation was explored by Nouri being able to conveniently derive VI-Kc relationships so that these
et al. (2014) from high spatial resolution WorldView-2 imagery and systems can be used for irrigation scheduling. The specific objective of
systematic evaluation of evapotranspiration through observational- this study is to bring together the two sets of measurements; namely
based approaches. The experimental site of Nouri et al. (2014) includes spectral reflectance index measurements and Kc measurements into a
cover grass that is similar to pasture, however, because of the presence single field measurement process for determining VI- Kc relationship. In
of trees, shrubs and water bodies in the experimental area the re- this case we demonstrated the technique for a pasture species of im-
lationship derived cannot be exactly the same as pasture. All of these portance to grazing industries worldwide (Rahman et al., 2014).
existing methods of linking NDVI to Kc are based on predominantly
satellite based measurement of NDVI and evapotranspiration or Kc
2. Materials and methods
measured or collected from secondary sources.
Advances in satellite imaging systems, and data and information
2.1. Instrumentation and calibration
delivery systems is seeing a growth in interest of irrigation scheduling
in a more reliable way. However, many of the existing, spectral re-
A schematic diagram of the ET dome and instrumentation is shown
flectance index-Kc (or −Kcb) relationships have been derived where
in Fig. 1 and is based upon the designs of (McLeod et al., 2004;
spatial scale and sensor resolution are not matched, and they are only
Stannard, 1988). The diameter of the clear perspex dome is 0.68 m and
known for a limited number of agricultural-relevant plant canopies.
the height is 0.36 m including a 2 cm thick foam-rubber ring on its base
There is currently no ‘simple’ method of directly relating the spec-
to ensure a good seal with the canopy/ground surface. A humidity
tral reflectance index of choice with Kc. All previous research points to a
sensor (Vaisala Oy, HMP 35A, Helsinki) and a thermocouple (ICT In-
statistically robust relationship between the parameters but the cali-
ternational, Australia) were mounted inside the dome to monitor the
bration methods are optimal for a gross measurement of regional scale
change in the air humidity and temperature, respectively and both
evapotranspiration. Moreover, accessing satellite based NDVI data for
sensors were connected to a data logger (SM1E904, ICT International,
any calibration is constrained to the revisit time of the satellite and this
Australia) which was able to record the data at one second interval. The
can be from few days to weeks (e.g. 16 days for Landsat), further ex-
humidity sensor was specifically chosen for its fast response time. Two
acerbated by weather. While depending on satellite data for a decision
fans of 80 mm diameter were used to mix the air and accumulating
making such as irrigation scheduling is problematic in its own right, in
water vapour uniformly inside the dome. Whenever the dome was
the case of crop or pasture management, any relationship between a VI
positioned over a ‘target’ surface, the fans were running continuously
and ET may also vary with often rapidly-evolving plant growth.
while the dome was positioned to ensure rapid mixing of the vapour
Irrespective of the delivery of any decision support tools for, say, irri-
and presentation of that mixed vapour to the overhead environmental
gation management, there is a need for a consistent and time-efficient
sensors.
means of creating relationships between spectral VIs and ET, which is
A crucial first step in dome deployment is calibrating the dome to
convenient, portable and suitable for varying skill levels of the user.
ensure that water vapour accumulation, as measured by the environ-
Handheld, active, optical reflectance sensors like GreenSeeker®
mental sensors, is the same as water lost to the atmosphere contained in
(Trimble, Sunnyvale, California, USA) or the CropCircle® (Holland
Scientific Inc., Lincoln, NE, USA) can read the NDVI of the target plant
canopy instantly regardless of the environmental conditions (Rahman
et al., 2014). It is desirable to be able to measure the crop evapo-
transpiration coincident with such measurements; in situ, at the same
spatial scale and at very near, if not at the same time. One solution is
the use of a portable hemispherical evaporation chamber for directly
recording evapotranspiration in situ (Stannard, 1988). This is a con-
venient method requiring only a few minutes onsite to set up and less
than a minute of measurement time (McLeod et al., 2004). Responsive
sensors are mounted inside the chamber to record temperature and
humidity so that the vapour generated from the target area can be ac-
curately measured from the change in vapour concentration inside the
chamber when the area covered by the chamber and the air inside
mixed uniformly with the help of a fan. A number of studies have used
devices to contain the atmosphere over plant targets to quantify eva-
Fig. 1. Schematic representation of the ET dome fitted with necessary accessories.
potranspiration in conditions ranging from forest to desert (Garcia

13
M.S. Alam et al. Computers and Electronics in Agriculture 147 (2018) 12–17

the dome from a given source (for example the plant canopy). The basic
principle of calibration was to produce water vapour inside the dome at
a known rate and compare this value with the calculated rate of ac-
cumulated water vapour inside the dome using the environmental
sensors (McLeod et al., 2004; Stannard, 1988). This was achieved by
placing a small, open-top container of water on the digital balance
(PA413C, Ohaus, USA, Precision balance) inside the dome. The rate of
evaporation from the container was varied (0.0007–0.02 g/s) by
changing the water temperature and the diameter of the open portion of
the water container. The exact water temperature and container di-
mensions required to achieve this was determined by trial and error
using an unenclosed container in the laboratory. The balance was
connected to a laptop computer to record the weight of the water at a
1 s interval. On placing the dome over the water container, the hu-
midity and temperature data were used to calculate the vapour density
(g/m3) inside the dome using Eqs. (1) and (2), which was then multi-
plied by the dome volume to yield the quantity of vapour (g) inside the
Fig. 2. Collection of field data with the evaporation dome.
dome at any moment.

0.622RH(%)es ⎞
Vapour Density, ρv (g/m3) = ⎛ ⎜ ⎟× 10 experiment the data was collected in a 30 min window close to noon
⎝ d ( T+ 273.15) ⎠
R (1) under clear sky conditions with the help of portable evaporation dome
and NDVI sensor. Fig. 2 shows the dome in operation in experimental
17.50T ⎞
Saturated vapour pressure, es (Pa) = 6.11f(p)exp ⎛ site.
⎝ 241.2+T ⎠ (2)
For each ET measurement, the dome was lowered onto the ground
Here RH (%) is the relative humidity, T is the temperature in °C, Rd with the fans running at the selected speed. The target was covered by
is the gas constant (287.04 J/kg K), 0.622 is a factor derived from the the dome for approximately 60–90 s and the temperature and humidity
ratio of molecular weight of water and of dry air (g/mol), es is the of the enclosed atmosphere recorded as per the initial laboratory cali-
saturated vapour pressure (Pa) and f(p) is related to atmospheric bration. Again the calculated accumulation of vapour which was
pressure which is assumed to be a constant value of 100.47. plotted against time and the slope (M) of the plotted curve (g/s) used to
The increase in the amount of vapour inside the dome as inferred calculate the instantaneous crop evapotranspiration. The crop evapo-
from the environmental sensors was plotted against time and the slope transpiration (ETc) in mm/hr was than calculated using the equation of
of the fitted line was used to determine the rate of accumulation of (Stannard, 1988):
moisture in the dome (g/s). The rate of water lost from the water
MVC
container (g/s) was plotted against the rate of vapour accumulation in Crop Evapotranspiration (ETc) = 3.6 mm/hr
A (3)
the dome and the slope of the fitted line, passing through the origin was
taken as the dome calibration factor (C). Mixing of airflow is an im- where M is the slope of the curve, C is the calibration factor derived
portant consideration in the veracity of measurements to estimate va- from the laboratory calibration, A is the area of the target crop covered
pour accumulation in the dome. Three different fan speeds, producing by the dome (m2) and 3.6 is the unit conversion factor.
an average air velocity in the vicinity of the fan face of 1.88 m/s (Low), The reference crop evapotranspiration (ETo) value, necessary to
3.4 m/s (Medium) and 5.36 m/s (High), respectively were tested during calculate the associated Kc value for each set of measurements was
the calibration process and the selection of final fan speed for sub- collected from the nearby SMART Farm automatic weather station
sequent field use was determined from the accuracy of the environ- (AWS). The AWS was configured to measure ETo every 10 min using an
mental sensor data to explain the accumulation of the water into the array of environmental sensors and the Penman-Monteith equation. The
atmosphere contained inside the dome. maximum time interval between any field measurements of ETc and the
ETo values used to calculate Kc was 5 min.
2.2. Field data collection and analysis Immediately following a dome measurement of ETc, the NDVI of the
same target sample was recorded using handheld GreenSeeker®
Field campaigns were conducted to collect ET and NDVI data from (Trimble, Sunnyvale, California, USA) following the recommended in-
pasture on the University of New England SMART Farm (NSW, dustry protocol.
Australia; Latitude −30.481088° S, Longitude 151.644757° E) during
the peak summer pasture growing seasons of November 2016 – 3. Results
February 2017 and November 2017 to December 2017. The pasture
species of interest was Tall Fescue (Festuca arundinacea var. Dovey). A 3.1. ET instrumentation calibration and data processing
biomass measuring site was prepared, measuring 50 m long and 20 m
across. This site contained ten 10 m × 10 m plots comprising of 3 re- An example plot of the calculated vapour accumulation over an
plicates of nitrogen applied at ‘low’, ‘medium’ and ‘high’ rates; suffi- open evaporating dish at a fan speed of 5.36 m/s is given in Fig. 3.
cient to achieve discernible variations in biomass accumulation, and a As can be seen in Fig. 3, when the dome is lowered over the target,
single plot of exposed soil. In preparation for the field measurements of there is a short delay (∼5 s) while the contained atmosphere stabilizes
both growing seasons, the site was first mown to a height of approxi- and the moist air from adjacent to the plant canopy is circulated to the
mately 5 cm high, and then regularly irrigated to promote vigorous overhead atmospheric sensors (denoted as point ‘A’. The point identi-
pasture growth. A first set of field measurements was undertaken two fied as ‘B’ in Fig. 3 can be considered the point at which it can be
weeks after mowing when the pasture was in its initial growing stage assumed that the sensors are responding to accumulating moisture.
and the soil was only partially covered (biomass ranging from 589 to However with prolonged containment, the accumulating moisture in-
2364 kg DM/ha). A second set of data was collected at the third week side the dome reduces the vapour pressure deficit which also reduces
after mowing when the pasture had achieved complete ground cov- evapotranspiration (for example Eamus and Shanahan (2002) and
erage (biomass ranging from 1581 to 3200 kg DM/ha). During each Ocheltree et al. (2014)) and the rate of accumulation decreases beyond

14
M.S. Alam et al. Computers and Electronics in Agriculture 147 (2018) 12–17

0.6
Calculated vapour inside the chamber (g)
3
y = 0.04x + 0.99 Kc = (0.56 ± 0.05) x NDVI + (0.10 ± 0.03)
2.5 R² = 0.99 R² = 0.80
0.5
2
C

Kc
1.5 0.4
B
1
A
0.3
0.5

0.2
0
0 10 20 30 40 50 0.25 0.35 0.45 0.55 0.65 0.75
Time (s) NDVI
Fig. 3. Example of progressive water vapour accumulation over an open evaporating dish (a)
at fan speed = 5.36 m/s. 0.9

0.8 Kc = (1.36 ± 0.13) x NDVI - (0.32 ± 0.09)

‘C’ (Fig. 3). It could be argued that identifying the exact time and range
R² = 0.77
over which to fit the regression line to the time series data is somewhat
0.7
subjective. Certainly a measure of instantaneous evaporation rate, at
the exact time the dome is lowered onto the target canopy is the desired 0.6

Kc
goal but the simple act of lowering the dome onto the target disrupts
airflow over the sensors, and time is required for the moisture-laden air 0.5
to circulate to the sensors. Experience through trial and error, and a
visual examination prior to fitting the rate line to the data remains the 0.4
only viable method.
0.3
A graph of the rate of vapour production from the water container,
versus the calculated rate of vapour accumulation from the combina- 0.2
tion temperature and humidity sensor data is given in Fig. 4. The slope 0.4 0.5 0.6 0.7 0.8 0.9
of the fitted, line is 1.02 (R2 = 0.87) and this is the calibration factor NDVI
(C). This plot was generated using the High fan speed. The Low fan
speed (Figure not shown) caused poor mixing and presentation of the
(b)
vapour to the overhead temperature and humidity sensors; resulting in Fig. 5. Graphs of crop coefficient (Kc) versus measured NDVI for Tall Fescue pasture at (a)
a calibration factor of 1.58 (R2 = 0.82). The medium fan speed (also 2 and (b) 3 weeks post mowing under irrigated conditions. The uncertainties in each
not shown), while improving the mixing and presentation, yielding an regression parameter (slope and intercept) are included. Fan speed = 5.36 m/s.
improved calibration factor C = 1.07 (R2 = 0.99) similar to Macfarlane
and Ogden (2012) (C = 1.08, R2 = 0.97) in their improved evaporation water container with a wire mesh reduced noise in the balance readings
dome. and produced a calibration factor of 1.02 (R2 = 0.87) for the High fan
The High fan speed tested was observed to create turbulence in the speed. This noise was not evident in the subsequent field trials invol-
enclosed atmosphere and vibrated the container/balance introducing ving pasture canopies and so this fan speed (5.36 m/s) was used from
noise into the readings. It was observed that enclosing the balance and hereon in.

0.025
Rate of vapour production from water

3.2. Field measurements of ET (Kc) versus NDVI

y = 1.02x
R² = 0.87 Graphs of derived Kc values versus NDVI for Tall Fescue for the two
0.02
sampling periods (2 and 3 weeks post mowing at two consecutive
growing seasons) are given in Fig. 5(a) and (b).
container (g/s)

0.015 In this study the Kc for Tall Fescue ranges from 0.25 with low
ground coverage to a maximum value of 0.90 when the pasture was
well developed and fully covered the ground surface.
0.01 Both plots in Fig. 5(a) and (b) exhibit strong linear relationship
(R2 = 0.80 & 0.77 respectively).
0.005 Data from 2016 and 2017 seasons were used to generate those plots.
However, the slopes and intercepts of Fig. 5(a) and (b) are significantly
different for the different growing stages (2 weeks and 3 weeks after
0 mowing). Interestingly when the data are combined into a single plot
0 0.005 0.01 0.015 0.02 0.025 (Fig. 6), the Kc – NDVI relationship appears continuous and non-linear,
Rate of vapour accumulation calculated from sensor data and the variance in the Kc – NDVI relationship can be explained using a
(g/s) single polynomial regression curve (R2 = 0.84).
Fig. 4. A graph of the rate of vapour production from the water container, as measured by
the progressive loss of weight of the water container using the balance, versus the cal- 4. Discussions
culated rate of vapour accumulation in the enclosed atmosphere from the combination
temperature and humidity sensor data. Fan speed = 5.36 m/s. The calibration factor (C) The complete sequence of lowering dome and acquisition of suffi-
is the slope of the fitted line passing through the origin.
cient data is less than 20 s. This performance is consistent with the

15
M.S. Alam et al. Computers and Electronics in Agriculture 147 (2018) 12–17

1
evaporation term (Ke) and a leaf transpiration term (Kcb), was princi-
Kc = (1.84 ± 0.41) x NDVI2 - (1.03 ± 0.48) x NDVI + (0.42 ± 0.14)
0.9
R² = 0.84
pally governed by evaporation from the underlying soil (Ke), since the
lower-end value of approximately 0.2 observed in this work (assumed
0.8
to be Ke) is consistent with those simulated values of Zeleke and Wade
0.7 (2012) for crops in south eastern Australia.
0.6 Er-Raki et al. (2007) reported that the Kcb may begin to saturate
Kc

before the maximum NDVI of a plant canopy is reached. In other words

0.5
the increased photosynthetic activity manifest by the increasing NDVI
0.4 Kc Full can continue beyond the point at which the maximum net evapo-
0.3 transpiration of the canopy has been reached. While no reason was
Kc Initial
ascribed to this behavior in Er-Raki et al. (2007), it is likely that at this
0.2
saturation point the microclimate deeper in the canopy may be limiting
0.1 the contribution of the lower leaves to net canopy ET, even while ad-
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
ditional photosynthetic activity is taking place with increasing LAI. In
NDVI
the results of this present work however, there appears to be no such
Fig. 6. Graph of crop coefficient (Kc) versus measured NDVI for Tall Fescue pasture at all indications of any potential saturation in Kc at higher NDVI. If it is
development stages under irrigated conditions. The uncertainty in each regression assumed that Kcb is dominating the Kc value at the higher NDVI in this
parameter (slope and intercept) is included. Fan speed = 5.36 m/s. present work, the plateau at which ET no longer increases with in-
creasing canopy LAI has not been reached. This is due to availability of
observations of others using similar apparatus (McLeod et al., 2004; water in the underlying soil or simply the canopy has not reached to its
Stannard, 1988). maximum LAI to saturate ET.
The values of Kc derived from this experiment are comparable with
the tabulated time-averaged values recorded by Allen et al. (1998); Kc 5. Conclusion
ini = 0.40 and Kc mid = 0.85–1.05 for ‘rotational grazing pastures’ and
Kc ini = 0.30 and Kc mid = 0.75 for ‘extensive grazing pastures’. A more In this study an enclosed portable chamber was successfully applied
detailed comparison is difficult due to the absence of details in Allen to quickly and conveniently measure evapotranspiration from Tall
et al. (1998) concerning pasture type and, for example fraction of se- Fescue pasture variety with calibration factor C = 1.02 (R2 = 0.87) at
nesced (‘standing dead’) material. Allen et al. (1998) does state that “for high fan speed (5.36 m/s). The device was found to be reliable in ac-
non-active or only moderately active ground cover (active indicates curately measuring the actual evapotranspiration of the target canopy
green and growing ground cover with LAI > about 2–3), Kc should be and therefore facilitated the comparison with active optical sensor de-
weighted between Kc for no ground cover and Kc for active ground rived NDVI collected in a very small time difference (Less than 1 min).
cover, with the weighting based on the “greenness” and approximate The strong positive correlation between active optical sensor derived
leaf area of the ground cover.” NDVI and Kc at two pasture growth stages in non-limiting soil moisture
Kondoh and Higuchi (2001) observed a linear relationship between condition shows a good prospects to estimate ETc and consequently the
Landsat 5 TM –derived NDVI (30 m spatial resolution) and daily eva- daily or seasonal crop water requirement using simple measurement of
potranspiration from a 160 m diameter site covered with ‘mixed pas- NDVI from a large field. The NDVI-Kc relationships were found to be
ture’ (including a Poa species ‘Kentucky bluegrass’ – Poapratesis l. and linear at both initial and full growing stages of the pasture. However the
‘African lovegrass’ – Eragrostis curvula Nees). Like Tall Fescue these are relationship gradients differ from each other in two growing stages and
leafy, predominantly erectophile pastures which also occur in the re- combining the data over the full range of NDVI yielded a polynomial
gion of this present work. Tasumi et al. (2006) observed a linear re- regression. This method and the relationship derived can substitute the
lationship between an alfalfa-reference Kc (labelled Kcr) and at-satellite dependency on satellite based remote sensing methods where simulta-
NDVI (labelled NDVIas) for a range of crops including alfalfa, and edible neous measurements of NDVI and ET is quite impossible rather relied
beans, field and sweet corn, small grains, peas, potatoes and sugarbeet; on secondary data. The protocol can be implemented for other similar
here typical field sizes were 400 m × 400 m up to 800 m × 800 m and crops for more convenient estimation of Kc at large scale. Moreover, this
again 30 m spatial resolution Landsat TM imagery was used. Rafn et al. study warrants further research regarding better understanding and
(2008) used the METRIC model to estimate Kc together with MODIS partitioning evapotranspiration in a variable soil moisture condition.
(Moderate Resolution Imaging Spectroradiometer)-derived NDVI data
over two different study areas. While they too observed a linear re- Acknowledgement
lationship between NDVI and predicted Kc they also observed that the
relationships for each site were statistically different, cautioning again The authors wish to acknowledge the technical assistance of Derek
the application of region specific relationships to other regions. Schneider in preparing the field sites, and Dr Malem McLeod (NSW
A collation of data from numerous authors on the transpiration to Department of Primary Industries, Australia) for loan of the ET dome.
potential evaporation ratio (E/EP) of crops as a function of leaf area One of us (MSA) wish to acknowledge receipt of a Strategic
index (LAI) is given in Choudhury (1994). There is a clear asymptotic International Postgraduate Research Scholarship from the University of
trend at higher LAI similar to the asymptote observed widely in NDVI- New England, Australia.
LAI relationships. Furthermore, Choudhury (1994) demonstrates the
theoretical validity of approximating a linear relationship between E/EP References
and vegetation index related to photosynthetic activity.
The departure from linearity in combined Kc-NDVI relationship, or Allen, R., Robison, C.W., Garcia, M., Trezza, R., Tasumi, M., Kjaersgaard, J., 2010. ETrF vs
alternatively the fact that the linear relationship between Kc and NDVI NDVI relationships for Southern Idaho for rapid estimation of evapotranspiration.
Report to Idaho Department of Water Resources, 1–17. doi: https://www.idwr.idaho.
has a smaller gradient at lower NDVI, can be explained by the lower gov/Browse/WaterInfo/ESPAM/ET_Subcommittee/References/Allen_UI_Report_to_
ground cover and/or reduced shading of the ground at lower NDVI. The IDWR_on_ETrF_NDVI_May_3_2010.pdf.
absolute lower limit of achievable NDVI was attributed to the under- Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration-Guidelines
for computing crop water requirements-FAO Irrigation and drainage paper 56. FAO,
lying bare soil and any non-photosynthetically active vegetation (for Rome 300 (9), D05109.
example a layer of dead leaves). The value of Kc which includes both an Bausch, W.C., Neale, C.M., 1987. Crop coefficients derived from reflected canopy radia-
tion: a concept. Trans. ASAE 30 (3), 703–709.

16
M.S. Alam et al. Computers and Electronics in Agriculture 147 (2018) 12–17

Bryla, D.R., Trout, T.J., Ayars, J.E., 2010. Weighing lysimeters for developing crop and litter evaporation beneath re-growth and old-growth mountain ash forest.
coefficients and efficient irrigation practices for vegetable crops. HortScience 45 (11), Cooperative Research Centre for Catchment Hydrology Report (96/1), 60.
1597–1604. McLeod, M.K., Daniel, H., Faulkner, R., Murison, R., 2004. Evaluation of an enclosed
Choudhury, B.J., 1994. Synergism of multispectral satellite observations for estimating portable chamber to measure crop and pasture actual evapotranspiration at small
regional land surface evaporation. Remote Sens. Environ. 49 (3), 264–274. scale. Agric. Water Manage. 67 (1), 15–34.
Duchemin, B., Hadria, R., Erraki, S., Boulet, G., Maisongrande, P., Chehbouni, A., Neale, C.M., Bausch, W.C., Heermann, D.F., 1990. Development of reflectance-based crop
Escadafal, R., Ezzahar, J., Hoedjes, J., Kharrou, M., 2006. Monitoring wheat phe- coefficients for corn. Trans. ASAE 32 (6), 1891–1900.
nology and irrigation in Central Morocco: On the use of relationships between eva- Neale, C.M., Jayanthi, H., Wright, J.L., 2003. Crop and irrigation water management
potranspiration, crops coefficients, leaf area index and remotely-sensed vegetation using high resolution airborne remote sensing. In: Proceedings of 54th IEC Meeting of
indices. Agric. Water Manage. 79 (1), 1–27. the International Commission on Irrigation and Drainage (ICID) Workshop Remote
Eamus, D., Shanahan, S., 2002. A rate equation model of stomatal responses to vapour Sensing of ET for Large Regions, 17.
pressure deficit and drought. BMC Ecol. 2 (1), 8. Nouri, H., Beecham, S., Anderson, S., Nagler, P., 2014. High spatial resolution
El-Shirbeny, M.A., Ali, A., Badr, M.A., Bauomy, E.M., 2014. Assessment of wheat crop WorldView-2 imagery for mapping NDVI and its relationship to temporal urban
coefficient using remote sensing techniques. World Res. J. Agric. Sci. 1 (2), 12–17. landscape evapotranspiration factors. Remote Sens. 6 (1), 580–602.
Er-Raki, S., Chehbouni, A., Guemouria, N., Duchemin, B.ı., Ezzahar, J., Hadria, R., 2007. Ocheltree, T., Nippert, J., Prasad, P., 2014. Stomatal responses to changes in vapor
Combining FAO-56 model and ground-based remote sensing to estimate water con- pressure deficit reflect tissue-specific differences in hydraulic conductance. Plant, Cell
sumptions of wheat crops in a semi-arid region. Agric. Water Manage. 87 (1), 41–54. Environ. 37 (1), 132–139.
Garcia, C.A., Johnson, M.J., Andraski, B.J., Halford, K.J., Mayers, C.J., 2008. Portable Rafn, E.B., Contor, B., Ames, D.P., 2008. Evaluation of a method for estimating irrigated
chamber measurements of evapotranspiration at the amargosa desert research site crop-evapotranspiration coefficients from remotely sensed data in Idaho. J. Irrig.
near Beatty, Nye County, Nevada, 2003–06. US Geol. Survey Sci. Investigations Drain Eng. 134 (6), 722–729.
Report 5135 (10). Rahman, M., Stanley, J., Lamb, D., Trotter, M., 2014. Methodology for measuring fAPAR
Gowda, P.H., Chavez, J.L., Colaizzi, P.D., Evett, S.R., Howell, T.A., Tolk, J.A., 2008. ET in crops using a combination of active optical and linear irradiance sensors: a case
mapping for agricultural water management: present status and challenges. Irrig Sci. study in Triticale (X Triticosecale Wittmack). Precis Agric. 15 (5), 532–542.
26 (3), 223–237. Stannard, D.I., 1988. Use of a hemispherical chamber for measurement of evapo-
Hunsaker, D., Barnes, E., Clarke, T., Fitzgerald, G., Pinter Jr, P.J., 2005. Cotton irrigation transpiration: Dept. of the Interior. US Geological Survey: Books and Open-File
scheduling using remotely sensed and FAO-56 basal crop coefficients. Trans. ASAE 48 Reports [distributor].
(4), 1395–1407. Stannard, D.I., Weltz, M.A., 2006. Partitioning evapotranspiration in sparsely vegetated
Johnson, L.F., Trout, T.J., 2012. Satellite NDVI assisted monitoring of vegetable crop rangeland using a portable chamber. Water Resour. Res. 42 (2).
evapotranspiration in California’s San Joaquin Valley. Remote Sens. 4 (2), 439–455. Tasumi, M., Allen, R.G., Trezza, R., 2006. Calibrating satellite-based vegetation indices to
Kondoh, A., Higuchi, A., 2001. Relationship between satellite-derived spectral brightness estimate evapotranspiration and crop coefficients. In: Proceedings of the 2006 USCID
and evapotranspiration from a grassland. Hydrol Process. 15 (10), 1761–1770. Water Management Conference, Ground Water and Surface Water Under Stress:
http://dx.doi.org/10.1002/hyp.238. Competition, Interaction, Solutions, pp. 103–112.
Macfarlane, C., Ogden, G.N., 2012. An improved evaporation dome for forest environ- Zeleke, K.T., Wade, L.J., 2012. Evapotranspiration estimation using soil water balance,
ments. Comput. Electron. Agric. 89, 126–129. http://dx.doi.org/10.1016/j.compag. weather and crop data. Evapotranspiration-Remote Sensing and Modeling: InTech
2012.09.004. publisher.
McJannet, D., Vertessy, R., Tapper, N., O’Sullivan, S., Beringer, J., Cleugh, H., 1996. Soil

17