Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Agricultural Systems 151 (2017) 20–32

Contents lists available at ScienceDirect

Agricultural Systems

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

Performance of AquaCrop and SIMDualKc models in evapotranspiration


partitioning on full and deficit irrigated maize for seed production under
plastic film-mulch in an arid region of China
Hui Ran a, Shaozhong Kang a,⁎, Fusheng Li b, Ling Tong a, Risheng Ding a, Taisheng Du a, Sien Li a, Xiaotao Zhang a
a
Center for Agricultural Water Research in China, China Agricultural University, Beijing 100083, China
b
College of Agriculture, Guangxi University, Nanning, Guangxi 530005, China

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

Article history: It is difficult – though important – to assess the necessary allocation of increasingly scarce water resources and
Received 6 April 2016 yield estimation, by determining the exact amount of evapotranspiration (ET) and the partitioning of ET into tran-
Received in revised form 13 October 2016 spiration (T) through the stomata of plants and evaporation (E) from the soil. An accurate and effective method
Accepted 6 November 2016
for ET partitioning and estimation is, therefore, desirable. This paper aims to evaluate the performance of two
Available online 19 November 2016
main models for ET partitioning and estimation in terms of feasibility and accuracy: the AquaCrop (Version
Keywords:
4.0) model, and the SIMDualKc model. Field experiments were carried out between 2011 and 2015 in an arid re-
Crop coefficient gion of Northwest China on two sites using full and deficit irrigation under plastic film-mulch, with the ET of
Crop transpiration maize for seed production and its partitioning components (T and E) being strictly measured by using the
Soil evaporation eddy covariance (EC) system or the sap flow system and micro-lysimeter cylinders. Subsequently, part of the
Water stress measured data was used to calibrate the two models, so that the calibrated models could then be used to assess
Deficit irrigation whether the agreement between simulation and measurement had proved successful, thus validating the
Northwest China models, or not. The results showed that the two models performed well with regard to their simulation of ET
and T under full irrigation conditions. Under deficit irrigation conditions, the ET and T values simulated by the
AquaCrop model were much closer to the actual measurement when compared with the results simulated by
the SIMDualKc model. This was particularly the case when the soil was re-watered after a period of long-term
water stress. For the simulated E, however, both models generated data that were distant from the actual mea-
surements taken under full or deficit irrigation conditions using plastic film-mulch, although when the
SIMDualKc model simulated E data, it came closer to the measurement than did the AquaCrop model. The com-
parison of two models in terms of their accuracy and feasibility based on the data analysis is discussed.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction stress is present. The two-step crop coefficient–reference ET (Kc–ET0)


procedure is a practical way of estimating the water requirements of
In arid regions, agriculture largely depends on irrigation (Kang et al., crops. If this approach is used, it is recommended that reference evapo-
2000) and, because of the shortage of water, the accurate estimation of transpiration (ET0) is calculated using the FAO Penman–Monteith meth-
crop yield and rational development of irrigation scheduling are neces- od (Allen et al., 1998), which represents primary weather-induced
sary for an appropriate management of water resources. Evapotranspi- effects on water consumption, with the crop coefficient (Kc) scaling
ration (ET) and its components constitute the main basis for the the ET0 to account for crop-specific influences on ET and their variation
development of irrigation scheduling and yield simulation, making re- during the growing season (Pereira et al., 2015a). The AquaCrop model,
search on simulating evapotranspiration and its components necessary. which adopts a “Kc–ET0” approach based on canopy cover (CC), and the
There are, in fact, several models available for partitioning ET into E SIMDualKc model, which uses the dual crop coefficient approach, are
and T, such as the Shuttleworth-Wallace, ENWATBAL, Cupid-DPEVAP, two simple crop coefficient approaches for the calculation and partition
SWEAT, TSEB and HYDRUS-1D models. These models are complicated, of ET.
however, and need many input parameters, which are generally diffi- The AquaCrop model is the newest international water-driven crop
cult to measure or estimate, especially under conditions where water model developed by FAO (Steduto et al., 2009; Raes et al., 2009). Instead
of the dual crop coefficient approach, this model adopts a “Kc-ET0” ap-
⁎ Corresponding author. proach based on green canopy cover (CC) to calculate ET and its compo-
E-mail address: kangsz@cau.edu.cn (S. Kang). nents (Pereira et al., 2015b). In the AquaCrop model, T is proportional to

http://dx.doi.org/10.1016/j.agsy.2016.11.001
0308-521X/© 2016 Elsevier Ltd. All rights reserved.
H. Ran et al. / Agricultural Systems 151 (2017) 20–32 21

CC and E is proportional to the fraction of the soil surface not covered by wind speed of 1.3 m s−1. Water resources are relatively scarce, howev-
a canopy (1 – CC). There have been many reports written on the er, with a mean annual precipitation of 164 mm, mean annual pan evap-
AquaCrop model (Hsiao et al., 2009; Heng et al., 2009; Araya et al., oration of approximately 2000 mm, and a groundwater table of below
2010a, 2010b; Andarzian et al., 2011; Paredes et al., 2014a; Tavakoli et 25 m between 1955 and 2005 (Li et al., 2015). The soil texture is that
al., 2015; Linker et al., 2016; Toumi et al., 2016), which have indicated of sandy loam, with a mean dry bulk density of 1.4 g cm−3, a mean sat-
that it is able to provide a highly accurate simulation of CC, soil moisture, urated water content of 0.41 cm3 cm−3, a mean field capacity (FC) of
biomass and yield. To prove whether a model can be used to develop ir- 0.30 cm3 cm−3, and mean permanent wilting point of 0.10 cm3 cm−3
rigation scheduling and manage water under different conditions, it is for the 0–100 cm layers. A mean saturated hydraulic conductivity of
necessary to validate daily ET, especially in conditions of water stress 500 mm d−1 was adopted as recommended for the AquaCrop model
(Stewart et al., 1977). Studies on the accurate simulation of daily ET by when used on sandy loam soil.
the AquaCrop model are relatively few and far between (Farahani et
al., 2009; Iqbal et al., 2014). Validations of daily ET partitioning by the
2.2. Experimental method
AquaCrop model are also scarce, especially under conditions when
water stress is present (Katerji et al., 2013; Pereira et al., 2015b).
The field data for two separate experiments, i.e. the full irrigation ex-
The dual crop coefficient approach proposed by FAO (Allen et al.,
periment (Site 1) and the deficit irrigation experiment (Site 2), were
1998) is an empirical method of distinguishing crop transpiration and
collected from April to September 2011–2015. The experiment was car-
soil evaporation, and is also one of the most common current methods,
ried out on Site 1 in a large area of farmland, with the eddy covariance
known for its good performance in developing irrigation scheduling
(EC) system being used to measure the ET of the canopy because of
with fewer parameters. The dual crop coefficient approach considers
the adequate fetch length. Since the experiment was carried out in
the two components of ET, namely plant transpiration and soil evapora-
plots, the EC system on Site 2 was not suitable because each plot was
tion, separately by using the basal crop coefficient (Kcb) and soil evapo-
limited. Instead, the ET was measured using the sap flow system and
ration coefficient (Ke) respectively (Kool et al., 2014). There are many
micro-lysimeter cylinders. Although there were differences between
studies, which verify this method in various climates and regions and
the two methods used, both could be used to measure the ET (Allen et
on various different crops, but mostly under well-watered conditions
al., 2011; Zhang et al., 2011; Ding et al., 2013; Kool et al., 2014). Jiang
(Ding et al., 2013; Zhang et al., 2013; Paredes et al., 2014b; Pereira et
et al. (2016) compared the two methods in the same field and noted
al., 2015a; Wu et al., 2015; Zhao et al., 2015). However, there are also
that two methods agreed with one another. Thus, on Site 1, the daily
studies that show that the dual crop coefficient approach is capable of
ET was measured using the EC system, with the daily E being obtained
substantial errors when it simulates ET, because the differences be-
by micro-lysimeters. The daily T was calculated by ET minus E. On Site
tween the crop coefficients recommended by FAO-56 and the actual
2, the daily T was measured by using the sap flow system, and the
crop can be as large as ±40%, especially for the mid-term crop coeffi-
daily E was measured using micro-lysimeters. The daily ET was obtained
cient (Katerji and Rana, 2006). In addition, there are relatively few stud-
by T plus E.
ies on the dual crop coefficient approach conducted when water stress
is present, so that it has become necessary to validate it based on ET, E,
and T measured on a scale set on a daily basis in order to determine 2.2.1. Site 1 experiment
its adaptability under conditions in which there is a shortage of water. The experiment was conducted on a farm in a large field measuring
Several studies have already compared the AquaCrop and 300 × 300 m2, which was managed by local farmers. The amount of ir-
SIMDualKc models (Paredes et al., 2015; Pereira et al., 2015b), but rigation and the time taken for each growing stage adhered to local
these studies were carried out in semi-humid monsoon and Mediterra- management and practice, with border irrigation being the method
nean climates. Neither did they simulate daily ET partitioning under used. The maize was planted so that there was one line of male plants
conditions of water stress. Paredes et al. (2014a) have considered to five lines of female plants. The surface of the soil was mulched
water stress conditions with respect to maize, but did not compare using a 1.2 m wide plastic film in the plant row with a 0.4 m-wide
their results with the measured ET. The use of plastic film-mulch to sub- area of bare soil between the two rows of film sheets. The plant spacing
stantially reduce soil evaporation is a well-developed method and is was 0.25 m and the row spacing was 0.4 m. The plant density was
widely used in the arid regions of China. Previous studies have indicated 100,000 plants ha−1. Nitrogen (N), phosphorus (P2O5) and potassium
that soil evaporation is reduced by ~ 50% when plastic film-mulch is (K2O) fertilizers were applied at 500, 240 and 50 kg ha−1 according to
used throughout the growing season, especially during early growth average long-term fertilization data for each of the five years,
when the surface of the soil is not fully covered by a canopy (Zhou et respectively.
al., 2009; Hou et al., 2010). The objective of this study is to test the per- The EC system was installed in the northwest of the field, which was
formance of the AquaCrop and SIMDualKc models to simulate ET, E, and large enough to provide an adequate fetch length for the EC measure-
T under conditions of full and deficit irrigation with five-year field data ment. All EC sensors were kept at 1.0 m above the maize canopy. The
under plastic film-mulch in an arid region, and to discover an appropri- EC system was composed of a 3-D sonic anemometer/thermometer
ate ET partitioning model for the development of irrigation scheduling (model CSAT3, Campbell Scientific Inc., Utah, USA), a krypton hygrome-
and yield forecast for maize for seed production. ter (model KH20, Campbell Scientific Inc., Utah, USA), and a tempera-
ture and humidity sensor (model HMP45C, Campbell Scientific Inc.,
2. Materials and methods Utah, USA). Models CSAT3 and KH20 measured the vertical fluctuations
of wind, temperature and water vapor density at 0.1 s intervals, and the
2.1. Experimental site and description temperature and humidity at 10 min intervals. The net radiation (Rn)
was measured by a net radiometer (model NR-LITE, Kipp & Zonen,
Field data were collected at the Shiyanghe Experimental Station at Delft, Netherlands) at a height of 1.5 m above the canopy. Two soil
China Agricultural University, located in the city of Wuwei, Gansu Prov- heat flux plates (model HFP01, Hukseflux, Netherlands) were inserted
ince, Northwest China (37°52′ N, 102°50′ E, at 1581 m elevation). The at a soil depth of 50 mm so as to measure the heat flux of the soil. All
experimental site belonged to a typical arid inland desert climate the sensors were connected to a data logger (model CR5000, Campbell
where light and heat resources are abundant, with the mean annual du- Scientific Inc., USA), and the ET was computed at 10 min intervals.
ration of sunshine being over 3000 h, the mean frost free days number- More details concerning the measurements of the ET by the EC used in
ing over 150 d and with a mean annual temperature of 8 °C, an annual this study are described by Ding et al. (2010), Li et al. (2013) and Jiang
accumulated temperature of (N0 °C) of 3550 °C and an annual average et al. (2016). The observation period of the ET was between 2011 and
22 H. Ran et al. / Agricultural Systems 151 (2017) 20–32

2015, and the lengths of the four growth stages in each season were di- Each irrigation treatment was randomized among three replicates
vided in accordance with Allen et al. (1998) (Table 1). during a three year period. Each plot area was 86.8 (12.4 × 7) m2. Border
Three micro-lysimeter cylinders, which were made of PVC tubes irrigation was the irrigation method used. The sap flow was measured
with a height of 20 cm and a diameter of 10 cm, were placed in the using the Flow32-1 K system (Dynamax Inc., Houston, TX, USA) during
bare soil between two plastic films in order to measure soil evaporation the middle and late growth stages in the Site 2 experiment from 2013 to
(E) in 2011, 2014 and 2015. An illustration can be found in Li et al. 2015. Two or three sap flow system probes (SGB19-WS) were installed
(2013). The cylinder bottom was sealed using filter paper ensuring the for each treatment to measure crop transpiration. The lengths of the
water exchange between the soil inside and outside the cylinder. The axes for the stem cross-sections of the chosen plants ranged from 1.9
cylinders were taken out from the soil and weighed daily using an elec- to 2.3 cm; these were appropriate for the SGB19-WS. The sensors
tronic scale with the precision of 0.1 g, except when conditions were were installed at a height of N0.2 m above the ground. The sap flow
damp, i.e. when irrigation had taken place or it had rained. The micro-ly- was monitored every 60 s and recorded as 15-min averages using a
simeters were reinstalled after each irrigation and after heavy rain. CR1000 data logger (Campbell Scientific, Logan, UT, USA). Sensors
More details about the micro-lysimeter technique with plastic mulch were uninstalled every 10–15 days to avoid corrosion and protect plants
can be found in Ding et al. (2013) and Jiang et al. (2014). from continuous heating. Further details on these methods, their theo-
retical background and installation procedures can be found in Zhang
2.2.2. Site 2 experiment et al. (2011), Ding et al. (2013) and Jiang et al. (2016). In addition, the
The Site 2 experiment was carried out from 2013 to 2015 on plots sensors were disconnected before irrigation to avoid damage and,
about 200 m away from Site 1. In each irrigation treatment, water was then, reinstalled after irrigation had taken place.
applied when the water content in the soil had reached its FC criteria. Three micro-lysimeter cylinders, which were same as those used on
In 2013, two irrigation levels – i.e. 65–70% (W1) and 45–50% (W2) of Site 1, were used to measure soil evaporation for each treatment. Plant-
FC – were treated. In 2014, three irrigation levels – i.e. 65–70% (W1), ing and fertilization were the same as for Site 1.
55–60% (W2) and 45–50% FC (W3) – were treated. In 2015, three irriga-
tion treatments – i.e. full irrigation (W1), no irrigation during reproduc- 2.3. Measurements
tive growth stage (W2) and no irrigation during vegetative growth
stage (W3) – were made. W1 was irrigated five times over the whole 2.3.1. Meteorological data
growth period, and W2 and W3 were only irrigated the first twice at Meteorological data including precipitation (P), solar radiation (Rs),
the vegetative growth stage, and the last twice at the reproductive air temperature (T), wind speed (u2) and relative humidity (RH) during
growth stage, respectively. The irrigation depth was 120 mm each time. the experimental period, were continuously measured using a standard
automatic weather station (Hobo, Onset Computer Corporation, Cape
Table 1 Cod, Massachusetts, USA) at 2.0 m above ground, about 150 m away
Meteorological variables over whole growth stage of maize for seed production from 2011 from the experimental fields in Site 1 and 2. The data were collected
to 2015. at a 5 s interval and 15 min averages were calculated and recorded
Rs T RH P ET0 using a data logger. The reference crop evapotranspiration (ET0) was
Season Growth stage (W m−2) (°C) (%) (mm) (mm) calculated using the FAO Penman–Monteith method (Allen et al.,
2011 Initial (April 22–May 23) 224.0 15.0 41.0 21.0 131.8 1998). The Rs, T, RH, P and ET0 from the standard automatic weather sta-
Development (May 291.8 20.8 42.3 34.0 201.8 tion at each growing stage are shown in Table 1.
24–July 2)
Middle (July 3–August 22) 302.4 22.0 51.3 113.0 255.9 2.3.2. Plant measurements
Late (August 208.7 16.2 63.3 51.0 86.9
23–September 19)
In each treatment, nine plants were randomly selected in order to
Overall 265.6 19.1 49.0 219.0 676.4 measure plant height and the length and maximum width of the
2012 Initial (April 19–May 23) 224.6 15.6 21.5 15.0 147.1 green leaves every seven to ten days during the growing season. The
Development (May 267.5 20.6 20.3 16.0 189.7 leaf area was calculated by summing leaf length × maximum width of
24–July 2)
one side of each leaf multiplied by a factor of 0.7, which was derived
Middle (July 3–August 22) 298.0 21.1 29.5 69.0 228.0
Late (August 248.5 17.5 31.3 29.0 102.3 from the linear regression (R2 = 0.998) of the calculated and actual
23–September 20) value measured by the AM300 leaf area meter (ADC BioScientific Ltd.,
Overall 259.7 18.7 25.6 129.0 667.1 UK). The leaf area index (LAI) is the total leaf area divided by the average
2013 Initial (April 20–May 19) 232.8 16.6 33.3 6.0 119.6 ground area per plant. The canopy cover (CC) was derived from LAI
Development (May 223.0 20.0 48.0 18.2 157.4
20–June 28)
through the following empirical formula recommended by Hsiao et al.
Middle (June 29–August 205.8 21.7 61.0 32.6 181.4 (2009):
18)
Late (August 174.6 18.3 66.9 11.4 66.7 CC ¼ 1:005½1− expð−0:6LAI Þ1:2 ð1Þ
19–September 11)
Overall 210.9 19.6 52.6 68.2 525.2
2014 Initial (April 15–May 25) 223.2 13.5 44.4 55.0 148.0 The fraction of the soil covered or shaded by the crop canopy near
Development (May 237.9 20.1 52.3 22.6 155.4 solar noon (fc) was estimated from the plant height (h), as in the follow-
26–July 1) ing equation recommended by FAO-56 (Allen et al., 1998):
Middle (July 2–August 27) 228.2 20.1 66.4 151.2 214.3
Late (August 178.6 17.0 66.0 7.6 58.7  ð1þ0:5hÞ
K cb −K c min
28–September 20) fc ¼ ð2Þ
Overall 221.7 17.9 57.4 236.4 576.5 K c max −K c min
2015 Initial (April 15–May 25) 230.7 16.0 37.6 25.0 168.5
Development (May 227.7 19.4 54.7 46.2 147.4 where Kcb is the potential basal crop coefficient, Kc min is the minimum
26–July 1)
crop coefficient (Kc) for dry bare soil with no ground cover and Kc max
Middle (July 2–August 20) 243.4 20.7 58.4 53.2 202.1
Late (August 195.0 18.0 59.6 26.2 80.8 is the maximum Kc immediately after wetting.
21–September 16) The initial minimum effective rooting depth was assumed to be
Overall 227.9 18.7 52.2 150.6 598.9 0.3 m. The maximum rooting depth was found at 1.0 m below ground,
Rs is solar radiation, T is air temperature, RH is relative humidity, P is precipitation and ET0 which agrees with the reports compiled by Ding et al. (2013) and
is reference evapotranspiration. Jiang et al. (2016).
H. Ran et al. / Agricultural Systems 151 (2017) 20–32 23

2.3.3. Soil water content 2.4. Model description


The soil water content (SWC) was measured using the gravimetric
method at a depth of 10, 20, 40, 60, 80 and 100 cm with two replicates 2.4.1. AquaCrop model
in each plot before sowing, and the weighted average of the SWC values The AquaCrop model is a crop model with the growth engine being
observed for the various soil layers was calculated with respect to the water-driven, with transpiration being calculated first and then trans-
entire root zone. In the Site 1 experiment, the SWC data were collected lated into biomass through water productivity. The crop yield is calcu-
before sowing took place, while for the Site 2 experiment, the SWC data lated by using the above-ground biomass and harvest index. The
were collected every seven days using a gravimetrically calibrated time AquaCrop model uses a “Kc-ET0” approach to estimate potential crop
domain reflectometer (TRIME-PICO-IPH, IMKO, Ettlingen, Germany) in transpiration (T) and soil evaporation (E) depending on the canopy
2013 and 2014, and collected one to three days before and after each ir- cover (CC) (Steduto et al., 2009; Raes et al., 2009). T is calculated as:
rigation in 2015, using the gravimetric method in addition to the mea-
surements taken before sowing. The irrigation schedule for each T ¼ KsCC  KcTr;x ET 0 ð6Þ
treatment is shown in Table 2.
where Ks is the soil water stress coefficient including water logging
2.3.4. Crop transpiration (Ksaer), stomatal closure (Kssto) and soil salinity stress (Kssto,salt), CC⁎ is
As, in this study, T was simulated on a daily basis so as to calculate the actual canopy cover adjusted for micro-advective effects, KcTr,x is
the diurnal sap flow, the 15-min rate sap flow course (within the day) the coefficient for maximum crop transpiration (well watered soil and
was integrated with the daily cumulative sap flow. The mean daily T complete canopy, canopy cover = 100%), and ET0 is the reference
in each treatment was derived from scaling up the sap flow measure- evapotranspiration calculated using the FAO Penman-Monteith method
ments using the leaf area index as suggested by Zhang et al. (2011), (mm d−1) (Allen et al., 1998).
Ding et al. (2013) and Jiang et al. (2016): CC⁎ is calculated as:

  CC  ¼ 1:72CC−CC 2 þ 0:30CC 3 ð7Þ


1 Xn Qi
T¼  LAI ð3Þ
n i¼1 Ai The development and senescence of the green canopy under optimal
conditions is described as:
where T is the mean transpiration for each treatment (mm d−1), n is the
CC ¼ CC 0 et CGC
for CC ≤CC x =2 ð8Þ
sampling number, Qi is the sap flow of the ith individual (l d−1), Ai is the
leaf area through which the water transfer in the plant occurs (m2), and
LAI is the leaf area index (m2 m−2). ðCC x Þ2 −t CGC
CC ¼ CC x −0:25 e for CCNCC x =2 ð9Þ
CC 0
2.3.5. Soil evaporation h  CDC i
The soil evaporation was calculated from the daily weight changes CC ¼ CC x 1−0:05 eCCx t −1 for CC decline ð10Þ
recorded by the micro-lysimeters as:
where CC0 is the initial canopy cover at 90% emergence, CCx is the max-
ΔM imum canopy cover, CGC is the canopy growth coefficient (d−1), CDC is
Es ¼ 10 ð4Þ the canopy decline coefficient (d−1), and t is time (d).
Ae
CGC and CDC are two of the most important parameters for building
up the CC, and is modified by soil water stress as:
where Es is the mean of the bare soil evaporation depth (mm d−1), ΔMi
is the mean daily weight changes of the micro-lysimeter (g d−1), and Ae CGC adj ¼ Ks exp;w CGC ð11Þ
is the cross sectional area of the micro-lysimeters (cm2).

The actual soil evaporation in conditions when plastic film-mulch CDC adj ¼ 1−Ks8 sen CDC ð12Þ
was used was calculated as:
where CGCadj is CGC adjusted for water stress (d−1), Ksexp,w is the water
stress coefficient for canopy expansion (0–1), CDCadj is CDC adjusted for
E ¼ ð1− f r mulch ÞEs ð5Þ
water stress (d−1), and Kssen is the water stress coefficient for early can-
opy senescence (0–1).
where E is the actual soil evaporation (mm d−1), fr mulch is a fraction of The soil evaporation under plastic film-mulch is calculated as:
the plastic film-mulch, and fr mulch is measured at about 0.7 for the pur-
poses of this study. E ¼ Krð1− f m f r mulch Þð1−CC  ÞKex ET 0 ð13Þ

Table 2
Treatments and the irrigation amount during the growing season of maize for seed production from 2011 to 2015.

2011 2012 2013 2014 2015

S1 S1 S1 W1 W2 S1 W1 W2 W3 S1 W1 W2 W3

256a 273a 259a 290a 290a 259a 290a 286a 282a 259a 242a 247a 270a

DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm DAP mm
47 120 50 120 47 120 53 63 99 48 56 120 60 81 89 81 108 115 56 120 53 90 53 90 114 120
64 120 70 120 69 120 86 99 106 84 78 120 89 81 123 98 77 120 76 120 76 120 133 120
84 120 90 120 84 120 106 62 97 120 108 81 97 120 96 120
103 120 110 120 103 120 123 115 131 120 129 31 117 120 114 120
130 120 130 120 122 120 144 120 133 120

S1 is full irrigation experiment in Site 1, W is irrigation treatment in Site 2 and DAP is days after planting.
a
Initial soil water content (SWC, in mm) in the profile defined by maximum rooting depth (1 m) for each treatment.
24 H. Ran et al. / Agricultural Systems 151 (2017) 20–32

where Kr is the evaporation reduction coefficient, fm is the adjustment is the depth of water that can evaporate without water availability re-
factor for the effect of the mulch on soil evaporation, which varies be- strictions (mm).
tween 0.5 for a mulch made of plant material and is close to 1.0 for plas- few is calculated as:
tic mulch, fr mulch is a fraction of the mulch (fr mulch = 0.7), and Kex is the
maximum soil evaporation coefficient for a fully wet and unshaded soil f ew ¼ minð1− f c ; f w Þ ð20Þ
surface, which varies from 0.95 to 1.30 with a default value of 1.10 in the
AquaCrop model. where 1-fc is the average exposed soil fraction not covered (or shaded)
In order to compute Kr, the model performs a daily soil water bal- by vegetation, and fw is the average fraction of the soil surface wetted by
ance of the surface layer. Kr is calculated as: irrigation or precipitation. In conditions where plastic mulch is used, fc is
estimated as the maximum value between fc and fr mulch, where fc is the
e f k W rel −1 fraction of soil covered by the crop and fr mulch is the fraction of soil cov-
Kr ¼ ≤1 ð14Þ
e f k −1 ered by the plastic sheet (Rosa et al., 2012).
ET is the sum of T and E. More details concerning the dual crop coef-
where fk is a decline factor which varies from 1 to 8 with a default value
ficient approach and the SIMDualKc model can be found in Allen et al.
of 4 in the AquaCrop model, and Wrel is the relative water content of the
(1998) and Rosa et al. (2012).
soil layer through which water moves to the evaporating soil surface
layer.
2.5. Model calibration and validation
The ET is calculated by T plus E. For more information on the
AquaCrop model, please refer to Steduto et al. (2009) and Raes et al.
So as to be completely representative, all the data from the experi-
(2009).
ments carried out on Site 1 and Site 2 were used both to calibrate and
to validate the model. The first and last years in which the data were
2.4.2. SIMDualKc model
tested were selected for calibration, and the corresponding remaining
The SIMDualKc model is an irrigation scheduling simulation model
data were then used for the validation. The details of these arrange-
that uses a daily time step to compute the soil water balance in the
ments are as follows:
root zone in accordance with research carried out by Allen et al.
The ET data collected from the Site 1 experiment in the first year of
(1998). The SIMDualKc model adopts the FAO dual crop coefficient ap-
measurement (i.e. 2011) were chosen to calibrate the AquaCrop and
proach for a separate computation of daily T and E (Rosa et al., 2012). T is
SIMDualKc models, while the remaining data collected from same ex-
computed as:
periment in the remaining years (i.e. from 2012 to 2015) were used to
T ¼ K s K cb ET 0 ð15Þ test the results predicted by the two models against the measurements
taken under fully irrigated conditions.
where Ks is the water stress coefficient (0–1). The ET data collected from the Site 2 experiment – when there was
Ks is expressed as a linear function of depletion in the effective root no irrigation during the reproductive growth state (i.e. W2 treatment)
zone: – in the last year of measurement (i.e. 2015) were chosen to calibrate
the AquaCrop and SIMDualKc models, while the remaining data collect-
TAW−Dr TAW−Dr ed from the Site 2 experiment were used to test the results predicted by
Ks ¼ ¼ ð16Þ
TAW−RAW ð1−pÞTAW the two models against the measurements taken under conditions of
water stress.
where TAW is the total available soil water relative to the root depth The output of AquaCrop and SIMDualKc against observed field mea-
(mm), Dr. is the water depletion in the effective root zone (mm), surements were assessed using a regression coefficient through the or-
RAW is the readily available soil water relative to the root depth igin (b0), the coefficient of determination (R2), the root mean square
(mm), and p is the average fraction of TAW that can be depleted error (RMSE), the normalized root mean square error (NRMSE), and
from the root zone before moisture stress (reduction in ET) occurs Willmott's index of agreement (d) (Hsiao et al., 2009; Pereira et al.,
(0–1). 2015b), which are calculated as:
E is computed as:
Xn
Mi Si
E ¼ K e ET 0 ð17Þ b0 ¼ Xi¼1
n ð21Þ
i¼1
Mi 2
where Ke is the evaporation coefficient.
Ke is calculated as: 8 92
>
> 
Xn   >
>
>
< M i −M Si −S >
=
K e ¼ min½K r ðK c max −K cb Þ; f ew K c max  ð18Þ 2 i¼1
R ¼       ð22Þ
>
> Xn  2
0:5 X 2 0:5 >
>
>
: M i −M
n
Si −S >
;
where Kr is a dimensionless evaporation reduction coefficient depen- i¼1 i¼1
dent on the depletion of soil water (cumulative depth of evaporation)
from the topsoil layer, Kc max is the maximum value of Kc (i.e.,
Kcb + Ke) following rain or irrigation, and few is the fraction of the soil ex- "Xn #0:5
posed to radiation and wetted by rain or irrigation. ðMi −Si Þ2
RMSE ¼ i¼1
ð23Þ
The model performs a daily soil water balance of the surface layer in n
order to compute Kr. Kr is calculated as:
"Xn #0:5
TEW−De;i−1 100 ðMi −Si Þ2
Kr ¼ ð19Þ NRMSE ¼ i¼1
ð24Þ
TEW−REW M n

where TEW is the maximum depth of the water that can evaporate from Xn
the evaporation layer when the soil has been completely wetted (mm), ðSi −M i Þ2
d ¼ 1−X 
i¼1
2 ð25Þ
De,i-1 is the cumulative depth of evaporation (depletion) from the sur- n
Si −S þ Mi −M
face layer of the soil at the end of dayi-1 (the previous day), and REW i¼1
H. Ran et al. / Agricultural Systems 151 (2017) 20–32 25

where Si is the simulated values, Mi is the measured values, S is the mean After the calibration of CC, the ET collected in 2011 during the experi-
of the simulated values and M is the mean of the measured values. It in- ment carried out on Site 1 was used to calibrate the REW and KcTr,x
dicates that the simulated values are statistically close to the measured under conditions of full irrigation. The calibrated REW and KcTr,x are pre-
ones, that is, if b0, d and R2 values are close to 1. It was reported that the sented in Table 3. The ET of W2 measured in 2015 during the Site 2 ex-
results from the simulation were considered acceptable in watershed periment was used to calibrate the stomatal conductance threshold
simulations if the value of R2 is N0.5 (Santhi et al., 2001; Van Liew et (psto) and the stomata stress coefficient curve shape for conditions in
al., 2003; Moriasi et al., 2007). Regarding the value of RMSE, the agree- which water stress was present; these are presented in Table 3. The
ment between the simulation and measurement becomes good when base temperature, upper temperature and initial canopy cover (cco)
the value of RMSE is close to zero. The value of NRMSE may indicate were adopted as default values in this study (Table 3).
the extent of agreement as a whole and has been applied in several The initial and calibrated values of the parameters used in the
studies (Jamieson et al., 1991; Bannayan and Hoogenboom, 2009). The SIMDualKc model are presented in Table 4. The ET collected in 2011 dur-
simulation was labelled (a) excellent; (b) good; (c) fair and (d) poor if ing the experiment on Site 1 was used to calibrate the Kcb of all the
the value of NRMSE was smaller than 10%, between 10% and 20%, be- growth stages and the soil evaporation parameters Ze and REW by min-
tween 20% and 30% and N30%, respectively (Jamieson et al., 1991; imizing the differences between the simulated and measured ET under
Bannayan and Hoogenboom, 2009). However, there has not been a sin- conditions of full irrigation. The TEW was calculated according to the
gle parameter that can be generally accepted as an assessment criterion water holding capacity of the soil and textural characteristics using
by the majority. We also recognize the complexity of partitioning the equation recommended by FAO-56 (Allen et al., 1998). The ET of
evapotranspiration into evaporation and transpiration, so we consid- W2, which was measured in 2015 in the Site 2 experiment, was used
ered all the metrics including b0, d, R2, RMSE, and NRMSE to indicate to calibrate the p value for all growth stages. When the SIMDualKc
the efficiency and accuracy of the models (AquaCrop and SIMDualKc) model was used, Kcb was adjusted to suit climatic conditions in accor-
used in this study. dance with the method recommended by FAO-56 (Allen et al., 1998),
but not adjusted for the plant density because the experiments involved
the same density in this study. Furthermore, the deficit irrigation treat-
3. Results and discussion ments in the Site 2 experiment had no effect on the seedling emergence
rate for the high initial soil water content (Table 2). Runoff was not
3.1. Calibration of model parameters taken into consideration, and the deep percolation was estimated by
the SIMDualKc model by using the simplified procedure described in
The default and calibrated parameters relative to AquaCrop model Doorenbos and Pruitt (1977) in which they noted that all deep percola-
are presented in Table 3. The CC was calibrated first of all owing to the tion occurs during the day on which excess water is applied, this is au-
AquaCrop's ability to produce sound predictions for ET, E and T depend- tomatically computed by the model without any input data
ing on the CC curve. Referring to Pereira et al. (2015b), the distinct pa- requirements. Both the AquaCrop and SIMDualKc models used the
rameterization of the CC curves for five crop seasons was performed same initial soil water content, fraction of plastic mulch, soil, irrigation
using data from the Site 1 experiment. The calibrated CCx, CGC and schedules and climate data.
CDC for each year under conditions of full irrigation are presented in
Table 3. The five CC fitting curves showed no tendency for over- or 3.2. Comparison in ET partitioning on Site 1
under-estimation, with b0 = 0.995 and R2 = 0.980 for the linear regres-
sion through the origin (Fig. 1a-e). Based on the calibration of Site 1, the In the Site 1 experiment, the results showed that the AquaCrop and
measured CC on Site 2 was used to calibrate the CC for water stress con- SIMDualKc models performed well with regard to simulating ET and T
ditions. The calibrated leaf growth upper threshold (pexp,upper), leaf under full irrigation conditions, but the results pertaining to the simula-
growth lower threshold (pexp,lower), leaf growth stress coefficient curve tion of E in the maize for seed production with plastic film-mulch under
shape, senescence stress coefficient (psen) and senescence stress coeffi- full irrigation conditions were found to be unacceptable. As only a small
cient curve shape are presented in Table 3. There was generally a good proportion of the total ET is E when field tests are carried out under con-
match between the simulated CC and those measured (b0 = 0.968 ditions in which plastic film-mulch is used, in this study the influence of
and R2 = 0.937) for the irrigation treatment on Site 2 (Fig. 1f–m). the results concerning the prediction of E on the simulation results of

Table 3
Default and calibrated parameters of maize for seed production assigned in AquaCrop model.

Description Units or meaning Value

Conservative parameters Adopted


Base temperature °C 8.0
Upper temperature °C 30.0
cco cm2, Canopy cover per seedling at 90% emergence 6.5
Calibrated parameters Default Calibrated
Leaf growth upper threshold (pexp,upper) Fraction of TAWa, above this leaf growth is inhibited 0.14 0.14
Leaf growth lower threshold (pexp,lower) Leaf growth stops completely at this p 0.72 0.72
Leaf growth stress coefficient curve shape Moderately convex curve 2.9 2.9
Stomatal conductance threshold (psto) Above this stomata begin to close 0.69 0.5
Stomata stress coefficient curve shape Highly convex curve 6.0 6.0
Senescence stress coefficient (psen) Above this early canopy senescence begins 0.69 0.5
Senescence stress coefficient curve shape Moderately convex curve 2.7 2.7
Crop coefficient for transpiration at CC = 100% Full canopy transpiration relative to ETo 1.03 1.20
Readily evaporable water mm 10 6
Parameters of the canopy cover curve 2011 2012 2013 2014 2015
Maximum canopy cover, CCx % 65–99 94 85 85 90 95
Canopy growth coefficient (CGC) %, increase in CC relative to existing CC per GDDb 1.2–1.3 1.03 1.09 1.00 1.31 1.07
Canopy decline coefficient (CDC) %, decrease in CC relative to CCx per GDD 1.0 0.71 0.43 0.34 0.50 0.68
a
Total available soil water of the root zone.
b
Growing degree day(s).
26 H. Ran et al. / Agricultural Systems 151 (2017) 20–32

100 100 100 100


(a) (b) (d)
80 80 80 (c) 80
2011S1 2012S1 2014S1
2013S1
60 60 60 60
CC (%)

CC (%)
CC (%)

CC (%)
40 40 40 40
Measured
20 Simulated 20 20 20

0 0 0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
DAP (d) DAP (d) DAP (d) DAP (d)
100 100 100 100
(e) (f) (g) (h)
80 2015S1 80 2013W1 80 2013W2 80
2014W1

60 60 60 60
CC (%)

CC (%)

CC (%)

CC (%)
40 40 40 40

20 20 20 20

0 0 0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
DAP (d) DAP (d) DAP (d) DAP (d)
100 100 100 100
(i) (j) (k) (l)
80 2014W2 80 80 80
2014W3 2015W1 2015W2

60 60 60 60
CC (%)

CC (%)

CC (%)

CC (%)
40 40 40 40

20 20 20 20

0 0 0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
DAP (d) DAP (d) DAP (d) DAP (d)
100

(m)
80
2015W3

60
CC (%)

40

20

0
0 20 40 60 80 100 120 140 160
DAP (d)

Fig. 1. Comparisons of simulated and measured canopy cover (CC) of (a-e) Site 1 (2011–2015), and (f-m) Site 2 (2013–2015).

the daily ET and T is small. It should also be noted that the AquaCrop models generally matched the measured data, although there were var-
model required the additional calibration of the annual CC parameter. iations for a few days after planting had taken place.
For the AquaCrop model, the values of b0, R2, RMSE, NRMSE and d
3.2.1. Evapotranspiration (ET) during all five-years were 0.956, 0.849, 0.814 mm d− 1, 25.6% and
Generally speaking, both the AquaCrop and SIMDualKc models sim- 0.956, respectively (Table 5). The results were comparable to those
ulated daily ET with acceptable accuracy under full irrigation conditions. shown by other studies. For example, the study carried out by Katerji
As shown in Fig. 2, the data simulated by the AquaCrop and SIMDualKc et al. (2013) on the application of the AquaCrop model with regard to
maize grown in a Mediterranean climate demonstrated that
underestimated daily ET alternates with overestimated daily ET under
Table 4
Default and calibrated parameters of maize for seed production assigned in SIMDualKc
well-watered conditions, and that the NRMSE values between simulat-
model. ed and measured ET over a two-year period were 37% and 41%, respec-
tively. Toumi et al. (2016) used eddy covariance data to validate the
Default (maize) Calibrated (maize for seed production)
AquaCrop model on winter wheat and showed that the precision rate
Crop parameters for daily ET simulation was reliable overall, but that ET was
Kcb ini 0.15 0.10
underestimated during the decline stage. Our research showed that
Kcb mid 1.15 1.10
Kcb end 0.5 0.3 daily ET could be simulated using the AquaCrop model under well-
pini, pmid, and pend 0.55 0.55 watered conditions, producing a NRMSE value ranging from 17.0 to
30.0% with the overall NRMSE being 25.6% over five years, which was
Soil evaporation parameters
REW (mm) 5–12 6 11.4% to 15.4% lower when compared with the results obtained by
TEW (mm) 20–37.5 37.5 Katerji et al. (2013), indicating a better prediction rate for the AquaCrop
Ze (m) 0.1–0.15 0.15 model in this study. In addition, in this study, ET was underestimated
REW is readily evaporable water, TEW is total evaporable water and Ze is depth of the sur- during the senescence period, which was similar to the results from
face soil layer for evaporation. Toumi et al. (2016).
H. Ran et al. / Agricultural Systems 151 (2017) 20–32 27

With regard to the SIMDualKc model, the values of b0, R2, RMSE, with regard to the maize for seed production using plastic film-mulch
NRMSE and d during the whole five-year period were 0.973, 0.852, under full irrigation conditions. However, since the proportion of E is
0.806 mm d− 1, 25.4% and 0.957, respectively (Table 5). The results much smaller than T in the total ET under plastic film-mulch, the unsuc-
were also comparable to those obtained in other studies. For example, cessful performance of the two models in predicting E had an insignifi-
Liu and Luo (2010) and Zhang et al. (2013) also showed that the dual cant impact on the simulation results of the daily ET and T. For example,
crop coefficient approach simulated daily ET effectively under well- Jiang et al. (2016) reported that E only accounted for 13–16% of the total
watered conditions. ET under plastic mulch.
For the AquaCrop model, the results of the simulation showed that
3.2.2. Soil evaporation (E) the values of b0, R2, RMSE, NRMSE and d during a period of three
Unfortunately, both the AquaCrop and SIMDualKc models produced years overall were 1.062, 0.235, 0.345 mm d−1, 138.4% and 0.577, re-
unacceptable E values when compared with the measurements made spectively (Table 5). The R2 value of 0.235 was much smaller than 0.5

10 10
2011 2011
8 Measured 8 AquaCrop

Simulated ET (mm)
AquaCrop SIMDualKc
SIMDualKc 1:1
ET (mm)

6 6

4 4

2 2

0 0
0 20 40 60 80 100 120 140 160 0 2 4 6 8 10
DAP (d) Measured ET (mm)
10 10
2012 2012
8 8

Simulated ET (mm)
ET (mm)

6 6

4 4

2 2

0 0
0 20 40 60 80 100 120 140 160 0 2 4 6 8 10
DAP (d) Measured ET (mm)
10 10
2013 2013
8 8
Simulated ET (mm)
ET (mm)

6 6

4 4

2 2

0 0
0 20 40 60 80 100 120 140 160 0 2 4 6 8 10
DAP (d) Measured ET (mm)
10 10
2014 2014
8 8
Simulated ET (mm)
ET (mm)

6 6

4 4

2 2

0 0
0 20 40 60 80 100 120 140 160 0 2 4 6 8 10
DAP (d) Measured ET (mm)
10 10
2015 2015
8 8
Simulated ET (mm)
ET (mm)

6 6

4 4

2 2

0 0
0 20 40 60 80 100 120 140 160 0 2 4 6 8 10
DAP (d) Measured ET (mm)

Fig. 2. Comparisons of daily measured versus simulated evapotranspiration (ET) by AquaCrop and SIMDualKc model under full irrigation condition (Site 1) from 2011 to 2015.
28 H. Ran et al. / Agricultural Systems 151 (2017) 20–32

Table 5 model assigned greater weight to the impact of canopy coverage on E.


Statistical analysis of measured and simulated daily evapotranspiration, transpiration and Studies carried out by Pereira et al. (2015b) on barley grown in a Med-
soil evaporation with AquaCrop and SIMDualKc models on maize for seed production un-
der full irrigation conditions (Site 1) from 2011 to 2015.
iterranean climate and Paredes et al. (2015) on soybeans grown in the
North China Plain also showed higher simulation results with regard
n b0 R2 RMSE (mm d−1) NRMSE (%) d to E when calculated by the SIMDualKc model than the AquaCrop
ETAquaCrop model, producing similar findings to ours.
2011 149 0.951 0.881 0.914 25.2 0.968
2012 93 0.907 0.889 0.741 17.0 0.958
2013 145 0.871 0.860 0.769 29.1 0.951 3.2.3. Crop transpiration (T)
2014 146 1.001 0.784 0.843 30.0 0.938 Both the AquaCrop and SIMDualKc models simulated daily T with an
2015 133 1.086 0.863 0.756 26.8 0.948 acceptable degree of accuracy under full irrigation conditions. For the
overall 666 0.956 0.849 0.814 25.6 0.956
AquaCrop model, the values of b0, R2, RMSE, NRMSE and d during a pe-
ETSIMDualKc riod of three years overall were 1.007, 0.853, 0.912 mm d−1, 30.5% and
2011 149 0.945 0.869 0.935 25.8 0.964 0.955, respectively (Table 5). The R2 value of 0.853 was much N0.5 and
2012 93 0.944 0.832 0.975 22.3 0.940
the NRMSE value of 30.5% was close to 30%. For the SIMDualKc model,
2013 145 0.962 0.870 0.699 26.5 0.962
2014 146 0.980 0.838 0.723 25.7 0.956 the values of b0, R2, RMSE, NRMSE and d during a period of three
2015 133 1.080 0.865 0.707 25.1 0.952 years overall were 0.961, 0.898, 0.703 mm d− 1, 23.5% and 0.971
overall 666 0.973 0.852 0.806 25.4 0.957 (Table 5), respectively. The R2 value of 0.898 was much N0.5 and the
TAquaCrop NRMSE value of 23.5% was smaller than 30%.
2011 76 0.922 0.915 0.932 24.5 0.971
2014 69 1.061 0.770 1.003 45.4 0.922
2015 79 1.142 0.898 0.802 27.8 0.948
3.3. Comparison in ET partitioning on Site 2
overall 224 1.007 0.853 0.912 30.5 0.955
For the Site 2 experiment, the results showed that both AquaCrop
TSIMDualKc
2011 76 0.888 0.948 0.792 20.8 0.977
and SIMDualKc models exhibited different degrees of decreasing simu-
2014 69 0.977 0.842 0.706 32.0 0.955 lation accuracy for ET and T when deficit irrigation treatments were
2015 79 1.094 0.925 0.602 20.9 0.968 used, and unacceptable results were found when simulating E in
overall 224 0.961 0.898 0.703 23.5 0.971 maize for seed production using plastic film-mulch under different irri-
EAquaCrop gation treatments. It was also found that under deficit irrigation condi-
2011 76 0.772 0.108 0.309 124.0 0.508 tions, the ET and T values simulated by the AquaCrop model were much
2014 69 1.150 0.202 0.399 136.6 0.555 closer to the actual measurements when compared with the results
2015 93 1.220 0.354 0.329 151.5 0.634
simulated by the SIMDualKc model. This strongly demonstrated the ad-
overall 238 1.062 0.235 0.345 138.4 0.577
vantage of the “Kc-ET0” approach based on CC utilized in the AquaCrop
ESIMDualKc model.
2011 76 0.859 0.515 0.166 66.7 0.813
2014 69 0.693 0.582 0.218 74.7 0.806
2015 93 1.169 0.364 0.342 157.8 0.631 3.3.1. Evapotranspiration (ET)
overall 238 0.894 0.389 0.261 105.1 0.718 Consistent with the results from the Site 1 experiment, both the
Measured data of evapotranspiration, evaporation and transpiration in 2011 were used for AquaCrop and SIMDualKc models simulated daily ET with acceptable
the calibration of AquaCrop and SIMDualKc model, and the rest of the data were for vali- accuracy during well-watered treatments. It was also found that the ac-
dation under full irrigation conditions. Evaporation and transpiration in 2012 and 2013
curacy of simulation achieved by the AquaCrop and SIMDualKc models
were not measured. b0 is regression coefficient through the origin; R2 is the coefficient
of determination; RMSE is root mean square error; NRMSE is normalized root mean in the Site 2 experiment for water stress treatments was slightly less
square error; and d is Willmott's index of agreement. when compared with the well-watered treatments.
Table 6 shows that the simulated results obtained by the AquaCrop
model were far closer to the measured results when compared with
and the NRMSE value of 138.4% was much N 30%. For the SIMDualKc the SIMDualKc model. For the AquaCrop model, the values of b0, R2,
model, the results of the simulation showed that the values of b0, R2, RMSE, NRMSE, and d in all irrigation treatments were 0.955, 0.812,
RMSE, NRMSE and d during a period of three years overall were 0.894, 0.659 mm d−1, 20.9% and 0.933, respectively. For the SIMDualKc
0.389, 0.261 mm d−1, 105.1% and 0.718, respectively (Table 5). The R2 model, the values of b0, R2, RMSE, NRMSE, and d were 1.060, 0.641,
value of 0.389 was smaller than 0.5 and the NRMSE value of 105.1% 1.017 mm d−1, 32.3% and 0.850, respectively. Katerji et al. (2013) inves-
was much N 30%. tigated the applicability of the AquaCrop model in simulating daily ET
In this study, the unsuccessful results for simulating E by the with different degrees of water stress in Mediterranean climates. Their
AquaCrop and SIMDualKc models may have been due to the fact that results showed that the AquaCrop model tended to underestimate the
(a) an additional complexity in terms of the E simulation was caused measured daily ET, particularly when the soil used for testing was re-
by using plastic mulch. For example, film cover can increase soil temper- watered, which indicated that the model slightly overestimates the lin-
ature, thus influencing soil evaporation. However, this influence was gering effect of water stress on daily ET. This differed from the results of
not taken into consideration in the equations used to calculate E equa- our study, the reason potentially being related to the different durations
tions in the AquaCrop and SIMDualKc models; (b) the amount of and degrees of water stress between the two studies. Furthermore,
water extracted from the roots of the crops was not considered by the Pereira et al. (2015b) and Paredes et al. (2015) considered that the
micro-lysimeters, resulting in higher soil water content in the micro-ly- AquaCrop model was inapplicable to the irrigation system because it
simeters than in the surrounding area. The micro-lysimeters might, used the empirical method based on crop factor (CC) instead of the
therefore, have overestimated soil evaporation, as shown in the study dual crop coefficient approach, resulting in a relatively low accuracy in
carried out by Klocke et al. (1990). terms of the simulation of ET partitioning. It should be noted, however,
Based on the results of “Goodness-of-fit” indexes, the value of E sim- that we applied deficit irrigation treatments in our study, which was dif-
ulated by the SIMDualKc model was slightly closer to the measurement, ferent from their studies.
when compared with the results simulated by the AquaCrop model. As The AquaCrop model considered the crop factor (CC) and caused the
shown in Eqs.(13) and (17), the method for calculating E was, in fact, results of ET to be slightly closer to the measurement when compared
different in the AquaCrop and SIMDualKc models. The AquaCrop with the SIMDualKc model applied under soil water stress conditions.
H. Ran et al. / Agricultural Systems 151 (2017) 20–32 29

Table 6 unacceptable compared with the measurements taken on the maize


Statistical analysis of measured and simulated daily evapotranspiration, transpiration and for seed production using plastic film-mulch under different irrigation
soil evaporation with AquaCrop and SIMDualKc models on maize for seed production un-
der different water stress conditions (Site 2) from 2013 to 2015.
treatments (shown in Table 6). It should also be noted that the propor-
tion of E is much smaller than T in total ET under plastic film-mulch, so
n b0 R2 RMSE (mm d−1) NRMSE (%) d that the unsuccessful performance in predicting E had an insignificant
ETAquaCrop impact on the simulated results of daily ET and T.
2013 W1 19 0.789 0.719 1.053 28.8 0.845 For the AquaCrop model, the simulated results showed that b0, R2,
2013 W2 19 1.190 0.743 0.646 33.1 0.867
RMSE, NRMSE, and d were 1.176, 0.410, 0.288 mm d− 1, 117.9% and
2014 W1 19 0.850 0.809 0.715 19.1 0.877
2014 W2 19 1.171 0.806 0.602 23.0 0.859 0.690, respectively. The R2 value of 0.410 was slightly smaller than 0.5
2014 W3 19 1.216 0.784 0.618 28.3 0.823 and the NRMSE value of 117.9% was much N30%.
2015 W1 35 0.944 0.837 0.643 14.7 0.949 For the SIMDualKc model, the simulated results showed that b0, R2,
2015 W2 35 0.996 0.953 0.435 13.1 0.987 RMSE, NRMSE, and d were 0.986, 0.536, 0.241 mm d− 1, 98.5% and
2015 W3 35 0.923 0.705 0.607 23.0 0.904
overall 200 0.955 0.812 0.659 20.9 0.933
0.776, respectively. The R2 value of 0.536 was close to 0.5 but the
NRMSE value of 98.5% was much N 30%.
ETSIMDualKc
2013 W1 19 0.946 0.840 0.592 16.2 0.952
3.3.3. Crop transpiration (T)
2013 W2 19 1.445 0.586 1.318 67.6 0.679
2014 W1 19 0.904 0.803 0.585 15.6 0.918 Both the AquaCrop and SIMDualKc models simulated daily T with
2014 W2 19 1.307 0.782 0.971 37.1 0.747 acceptable accuracy under the well-watered treatments in the Site 2 ex-
2014 W3 19 1.541 0.821 1.304 59.6 0.589 periment, although the performance in terms of a successful agreement
2015 W1 35 0.973 0.805 0.742 17.0 0.915
between simulation and the measurement decreased in different de-
2015 W2 35 0.972 0.955 0.406 12.2 0.988
2015 W3 35 1.318 0.310 1.563 59.2 0.574 grees for water stress treatments. In addition, it was found that the T
overall 200 1.060 0.641 1.017 32.3 0.850 value simulated by the AquaCrop model was far closer to the actual
measurement when compared with the results simulated by the
TAquaCrop
2013 W1 30 0.819 0.763 0.894 29.4 0.890 SIMDualKc model, as shown in Table 6 and Fig. 3.
2013 W2 30 1.289 0.683 0.798 51.2 0.815 With respect to the W1 and W2 treatments, for the AquaCrop model
2014 W1 38 0.887 0.694 0.765 24.5 0.877 in 2013, the b0, R2, RMSE, NRMSE and d values were (0.819 and 1.289),
2014 W2 38 1.236 0.722 0.833 38.3 0.815 (0.763 and 0.683), (0.894 and 0.798 mm d− 1), (29.4% and 51.2%),
2014 W3 38 1.305 0.682 0.806 45.9 0.770
(0.890 and 0.815), respectively. For the year 2014, with regard to the
2015 W1 55 0.989 0.915 0.580 17.5 0.978
2015 W2 54 1.029 0.956 0.447 17.0 0.988 W1, W2 and W3 treatments, the b0, R2, RMSE, NRMSE and d values
2015 W3 55 0.929 0.846 0.513 27.0 0.957 were (0.887, 1.236 and 1.305), (0.694, 0.722 and 0.682), (0.765, 0.833
overall 338 0.998 0.823 0.688 27.9 0.943 and 0.806 mm d− 1), (24.5%, 38.3% and 45.9%), (0.877, 0.815 and
TSIMDualKc 0.770), respectively. The results showed that when water stress was in-
2013 W1 30 0.953 0.817 0.678 22.3 0.944 creased, the accuracy of T when simulated by the AquaCrop model de-
2013 W2 30 1.595 0.632 1.367 87.7 0.675 creased, and the model overestimated T under conditions of soil water
2014 W1 38 0.943 0.735 0.683 21.9 0.909
stress.
2014 W2 38 1.355 0.735 1.064 48.9 0.762
2014 W3 38 1.678 0.735 1.450 82.5 0.601
For the year 2015, with regard to the W1, W2 and W3 treatments,
2015 W1 55 1.001 0.899 0.673 20.3 0.966 the b0, R2, RMSE, NRMSE and d values were (0.989, 1.029 and 0.929),
2015 W2 54 1.006 0.958 0.415 15.8 0.989 (0.915, 0.956 and 0.846), (0.580, 0.447 and 0.513 mm d−1), (17.5%,
2015 W3 55 1.360 0.670 1.349 71.1 0.786 17.0% and 27.0%), (0.978, 0.988 and 0.957), respectively. The negligible
overall 338 1.100 0.728 1.010 40.9 0.886
difference in terms of error between the simulated results under W1
EAquaCrop and W2 treatments would indicate that the AquaCrop model simulated
2013 W1 92 0.640 0.221 0.156 70.8 0.635 T well when water stress occurred continuously without re-watering. It
2013 W2 88 0.972 0.375 0.118 73.0 0.767
was also found that the accuracy of simulation for the W3 treatment de-
2014 W1 88 1.262 0.345 0.354 119.8 0.630
2014 W2 85 1.486 0.558 0.326 125.0 0.722 creased slightly when compared with the W1 treatment, which indicat-
2014 W3 92 1.330 0.554 0.307 108.0 0.768 ed that the AquaCrop model showed a slight discrepancy in cases of
2015 W1 88 0.998 0.289 0.337 123.1 0.612 long-term water stress followed by re-watering. For all irrigation treat-
2015 W2 88 1.014 0.321 0.332 139.6 0.670 ments, the b0, R2, RMSE, NRMSE, and d values for the T simulation were
2015 W3 87 1.398 0.521 0.292 128.5 0.703
overall 713 1.176 0.410 0.288 117.9 0.690
0.998, 0.823, 0.688 mm d−1, 27.9% and 0.943, respectively, indicating an
acceptable T result simulated by the AquaCrop model under different ir-
ESIMDualKc rigation treatments.
2013 W1 92 0.902 0.685 0.139 63.2 0.856
2013 W2 88 0.713 0.402 0.146 90.2 0.726
For the SIMDualKc model, the results were similar to the AquaCrop
2014 W1 88 1.040 0.593 0.275 93.1 0.766 model in 2013 and 2014 (Table 6, Fig. 3), but its simulation error
2014 W2 85 0.946 0.632 0.217 83.0 0.834 under water stress conditions was greater than that obtained by the
2014 W3 92 0.851 0.569 0.242 85.1 0.824 AquaCrop model. For the year of 2015, with respect to the W1, W2
2015 W1 88 1.134 0.475 0.305 111.5 0.709
and W3 treatments, the b0, R2, RMSE, NRMSE and d values were
2015 W2 88 1.127 0.582 0.279 117.5 0.790
2015 W3 87 1.035 0.395 0.276 121.6 0.690 (1.001, 1.006 and 1.360), (0.899, 0.958 and 0.670), (0.673, 0.415 and
overall 713 0.986 0.536 0.241 98.5 0.776 1.349 mm d−1), (20.3%, 15.8% and 71.1%), (0.966, 0.989 and 0.786), re-
Measured evapotranspiration, transpiration and evaporation of W2 in 2015 were used for
spectively. The negligible error difference between the simulated results
the calibration of AquaCrop and SIMDualKc model, and the rest of the data were for vali- under the W1 and W2 treatments would indicate that the SIMDualKc
dation under different water stress conditions. b0 is regression coefficient through the or- model performed well in simulating T when water stress occurred con-
igin; R2 is the coefficient of determination; RMSE is root mean square error; NRMSE is tinuously without re-watering. It was also found that the accuracy of
normalized root mean square error; and d is Willmott's index of agreement.
simulation for the W3 treatment decreased significantly when com-
pared with the W1 treatment, which indicated that the SIMDualKc
3.3.2. Soil evaporation (E) model had failed to simulate T under long-term water stress followed
Consistent with the results from the Site 1 experiment, both the by re-watering. As the T value simulated by the SIMDualKc model im-
AquaCrop and SIMDualKc models produced E values that were mediately recovered to that of the W1 level when the soil was re-
30 H. Ran et al. / Agricultural Systems 151 (2017) 20–32

watered after a long period of water stress, this did not match the mea- 1.010 mm d−1, 40.9% and 0.886, respectively, indicating that the T re-
sured T (Figs. 3, 2015 W3). For all irrigation treatments, the b0, R2, RMSE, sults simulated by the SIMDualKc model under different irrigation treat-
NRMSE, and d values for the T simulation were 1.100, 0.728, ments were unacceptable.

10 8
2013W1 2013W1
8 Measured
AquaCrop
6

Simulated T (mm)
AquaCrop SIMDualKc
6
T (mm)

1:1
4
4

2 2

0
0
80 90 100 110 120 130 140 150 160
0 2 4 6 8
DAP (d) Measured T (mm)
10 8

2013W2 2013W2
8 6

Simulated T (mm)
6
T (mm)

4
4

2
2

0 0
80 90 100 110 120 130 140 150 160 0 2 4 6 8
DAP (d) Measured T (mm)
10 8
2014W1 2014W1
8
6
Simulated T (mm)

6
T (mm)

4
4

2 2

0 0
80 90 100 110 120 130 140 150 160 0 2 4 6 8
DAP (d) Measured T (mm)

10 8
2014W2
8 2014W2
6
Simulated T (mm)

6
T (mm)

4
4

2 2

0 0
80 90 100 110 120 130 140 150 160 0 2 4 6 8
DAP (d) Measured T (mm)
10 8
2014W3
2014W3
8
6
Simulated T (mm)

6
T (mm)

4
4

2
2

0 0
80 90 100 110 120 130 140 150 160 0 2 4 6 8
DAP (d) Measured T (mm)

Fig. 3. Comparisons of daily Measured versus simulated transpiration (T) by AquaCrop and SIMDualKc model under different water stress conditions (Site 2) from 2013 to 2015.
H. Ran et al. / Agricultural Systems 151 (2017) 20–32 31

8
10
2015W1
2015W1
8 6

Simulated T (mm)
6
T (mm) 4
4
2
2

0 0
80 90 100 110 120 130 140 150 160 0 2 4 6 8
DAP (d) Measured T (mm)
8
10
2015W2
2015W2
8 6

Simulated T (mm)
6
T (mm)

4
4
2
2

0 0
80 90 100 110 120 130 140 150 160 0 2 4 6 8
Measured T (mm)
DAP (d)
8
10
2015W3
2015W3
8 6

6 Simulated T (mm)
T (mm)

4
2
2

0
0 0 2 4 6 8
80 90 100 110 120 130 140 150 160 Measured T (mm)
DAP (d)

Fig. 3 (continued).

The AquaCrop and SIMDualKc models implement different methods factors in response to water stress, and an improved dual crop coef-
to calculate T (as shown in Eqs.(6) and (15) for the AquaCrop model and ficient approach is needed for further work.
the SIMDualKc model, respectively). As a result, there were large differ-
ences in the accuracy of T when simulated under deficit irrigation
treatments. 4. Conclusions
Both models showed a decreasing precision in simulating T when
water stress treatments were carried out, with the AquaCrop model Based on the data collected from the Site 1 and Site 2 experiments
performing much better than the SIMDualKc model. The reason carried out between 2011 and 2015, the main conclusions arising
might be that the response of the SIMDualKc model to water stress from this study are as follows:
only considers the impact of soil water content, which is high after
re-watering, so the simulated T immediately recovers to its full irri- • Both the AquaCrop and SIMDualKc models performed well in simulat-
gation level as soon as the soil water stress disappears. But this ig- ing ET and T under full irrigation conditions. For the AquaCrop model,
nores the physiological damage to plants caused by long-term additional calibration of the annual CC parameter was required during
water stress. In this study, the T of the maize could not recover to the process of ET partitioning. Under deficit irrigation conditions, the
its full irrigation level after experiencing a long period of water ET and T values simulated by the AquaCrop model were much closer
stress. Instead, the AquaCrop model not only considered the effect to the actual measurement when compared with the results simulat-
of the soil water content, but also took into account the plant itself ed by the SIMDualKc model, particularly when the soil was re-
during its response to water stress, namely, the change in CC. watered after experiencing an extended period of water stress.
Long-term water stress led to a decline in the CC, and the CC could • Both the AquaCrop and SIMDualKc models produced unacceptable re-
not recover to its full irrigation level after re-watering. This would sults in simulating E under full or deficit irrigation conditions using
indicate that the T calculated by the AquaCrop model based on CC plastic film-mulch, although when the SIMDualKc model simulated
is closer to the measured value. Thus, in addition to soil moisture, E data, they were relatively closer to the actual measurement than
the SIMDualKc model needs to involve more of a crop's physiological with the AquaCrop model.
32 H. Ran et al. / Agricultural Systems 151 (2017) 20–32

Acknowledgements Kool, D., Agam, N., Lazarovitch, N., Heitman, J.L., Sauer, T.J., Ben-Gal, A., 2014. A review of
approaches for evapotranspiration partitioning. Agric. For. Meteorol. 184, 56–70.
Li, S., Kang, S., Zhang, L., Ortega-Farias, S., Li, F., Du, T., Tong, L., Wang, S., Ingman, M., Guo,
We greatly appreciate the careful reviews and valuable comments W., 2013. Measuring and modeling maize evapotranspiration under plastic film-
made by the anonymous reviewers and the editors, who improved the mulching condition. J. Hydrol. 503 (1), 153–168.
Li, S., Kang, S., Zhang, L., Du, T., Tong, L., Ding, R., Guo, W., Zhao, P., Chen, X., Xiao, H., 2015.
manuscript. This research received financial support from the National Ecosystem water use efficiency for a sparse vineyard in arid Northwest China. Agric.
Natural Science Foundation of China (grants 91425302 and Water Manag. 148 (148), 24–33.
51321001), the Government Public Research Funds for Projects of the Linker, R., Ioslovich, I., Sylaios, G., Plauborg, F., Battilani, A., 2016. Optimal model-based
deficit irrigation scheduling using AquaCrop: A simulation study with cotton, potato
Ministry of Agriculture (201503125), and the Discipline Innovative and tomato. Agric. Water Manag. 163, 236–243.
Engineering Plan (111 Program, B14002). Liu, Y., Luo, Y., 2010. A consolidated evaluation of the FAO-56 dual crop coefficient approach
using the lysimeter data in the North China plain. Agric. Water Manag. 97 (1), 31–40.
Moriasi, D.N., Arnold, J.G., Liew, M.W.V., Bingner, R.L., Harmel, R.D., Veith, T.L., 2007. Model
References
evaluation guidelines for systematic quantification of accuracy in watershed simula-
tions. Trans. ASABE 50, 885–900.
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration. Guidelines for
Paredes, P., de Melo-Abreu, J.P., Alves, I., Pereira, L.S., 2014a. Assessing the performance of the
Computing Crop Water RequirementsFAO Irrigation and Drainage Paper 56. FAO,
FAO AquaCrop model to estimate maize yields and water use under full and deficit irri-
Rome, Italy (300 pp.).
gation with focus on model parameterization. Agric. Water Manag. 144, 81–97.
Allen, R.G., Pereira, L.S., Howell, T.A., Jensen, M.E., 2011. Evapotranspiration information
Paredes, P., Rodrigues, G.C., Alves, I., Pereira, L.S., 2014b. Partitioning evapotranspiration,
reporting: I. Factors governing measurement accuracy. Agric. Water Manag. 98 (6),
yield prediction and economic returns of maize under various irrigation management
899–920.
strategies. Agric. Water Manag. 135, 27–39.
Andarzian, B., Bannayan, M., Steduto, P., Mazraeh, H., Barati, M.E., Barati, M.A., Rahnama,
Paredes, P., Wei, Z., Liu, Y., Xu, D., Xin, Y., Zhang, B., Pereira, L.S., 2015. Performance assess-
A., 2011. Validation and testing of the AquaCrop model under full and deficit irrigated
ment of the FAO AquaCrop model for soil water, soil evaporation, biomass and yield
wheat production in Iran. Agric. Water Manag. 100 (1), 1–8.
of soybeans in North China plain. Agric. Water Manag. 152, 57–71.
Araya, A., Habtu, S., Hadgu, K.M., Kebede, A., Dejene, T., 2010a. Test of AquaCrop model in
Pereira, L.S., Allen, R.G., Smith, M., Raes, D., 2015a. Crop evapotranspiration estimation
simulating biomass and yield of water deficient and irrigated barley (Hordeum
with FAO56: Past and future. Agric. Water Manag. 147, 4–20.
vulgare). Agric. Water Manag. 97 (11), 1838–1846.
Pereira, L.S., Paredes, P., Rodrigues, G.C., Neves, M., 2015b. Modeling malt barley water use
Araya, A., Keesstra, S.D., Stroosnijder, L., 2010b. Simulating yield response to water of Teff
and evapotranspiration partitioning in two contrasting rainfall years. Assessing
(Eragrostis tef) with FAO's AquaCrop model. Field Crop Res. 116 (1–2), 196–204.
AquaCrop and SIMDualKc models. Agric. Water Manag. 159, 239–254.
Bannayan, M., Hoogenboom, G., 2009. Using pattern recognition for estimating cultivar
Raes, D., Steduto, P., Hsiao, T.C., Fereres, E., 2009. AquaCrop-the FAO crop model to simu-
coefficients of a crop simulation model. Field Crop Res. 111, 290–302.
late yield response to water: II. Main algorithms and software description. Agron. J.
Ding, R., Kang, S., Li, F., Zhang, Y., Tong, L., Sun, Q., 2010. Evaluating eddy covariance meth-
101 (3), 438–447.
od by large-scale weighing lysimeter in a maize field of Northwest China. Agric.
Rosa, R.D., Paredes, P., Rodrigues, G.C., Fernando, R.M., Alves, I., Pereira, L.S., Allen, R.G.,
Water Manag. 98 (1), 87–95.
2012. Implementing the dual crop coefficient approach in interactive software: 2.
Ding, R., Kang, S., Zhang, Y., Hao, X., Tong, L., Du, T., 2013. Partitioning evapotranspiration
Model testing. Agric. Water Manag. 103, 62–77.
into soil evaporation and transpiration using a modified dual crop coefficient model
Santhi, C., Arnold, J.G., Williams, J.R., Dugas, W.A., Srinivasan, R., Hauck, L.M., 2001. Valida-
in irrigated maize field with ground-mulching. Agric. Water Manag. 127, 85–96.
tion of the SWAT model on a large river basin with point and nonpoint sources. J. Am.
Doorenbos, J., Pruitt, W.O., 1977. Crop Water Requirements. Irrigation and Drainage Paper
Water Resour. Assoc. 37 (5), 1169–1188.
24. Italy, FAO, Rome (197 pp.).
Steduto, P., Hsiao, T.C., Raes, D., Fereres, E., 2009. AquaCrop—the FAO crop model to sim-
Farahani, H.J., Izzi, G., Oweis, T.Y., 2009. Parameterization and evaluation of the AquaCrop
ulate yield response to water: I. Concepts and underlying principles. Agron. J. 101 (3),
model for full and deficit irrigated cotton. Agron. J. 101 (3), 469–476.
426–437.
Heng, L.K., Hsiao, T., Evett, S., Howell, T., Steduto, P., 2009. Validating the FAO AquaCrop
Stewart, J.I., Hagan, R.M., Pruitt, W.O., Danielson, R.E., Franklin, W.T., Hanks, R.J., Riley, J.P.,
model for irrigated and water deficient field maize. Agron. J. 101 (3), 488–498.
Jackson, E.B., 1977. Optimizing crop production through control of water and salinity
Hou, X., Wang, F., Han, J., Kang, S., Feng, S., 2010. Duration of plastic mulch for potato
levels in the soil. Reports. Paper 67.
growth under drip irrigation in an arid region of Northwest China. Agric. For.
Tavakoli, A.R., Moghadam, M.M., Sepaskhah, A.R., 2015. Evaluation of the AquaCrop model
Meteorol. 150 (1), 115–121.
for barley production under deficit irrigation and rainfed condition in Iran. Agric.
Hsiao, T.C., Heng, L., Steduto, P., Rojas-Lara, B., Raes, D., Fereres, E., 2009. AquaCrop-the
Water Manag. 161, 136–146.
FAO crop model to simulate yield response to water: III. Parameterization and testing
Toumi, J., Er-Raki, S., Ezzahar, J., Khabba, S., Jarlan, L., Chehbouni, A., 2016. Performance as-
for maize. Agron. J. 101 (3), 448–459.
sessment of AquaCrop model for estimating evapotranspiration, soil water content
Iqbal, M.A., Shen, Y., Stricevic, R., Pei, H., Sun, H., Amiri, E., Penas, A., Rio, S., 2014. Evalua-
and grain yield of winter wheat in Tensift Al Haouz (Morocco): Application to irriga-
tion of the FAO AquaCrop model for winter wheat on the North China plain under
tion management. Agric. Water Manag. 163, 219–235.
deficit irrigation from field experiment to regional yield simulation. Agric. Water
Van Liew, M.W., Arnold, J.G., Garbrecht, J.D., 2003. Hydrologic simulation on agricultural
Manag. 135, 61–72.
watersheds: Choosing between two models. Trans. ASAE 46 (6), 1539–1551.
Jamieson, P.D., Porter, J.R., Wilson, D.R., 1991. A test of the computer simulation model
Wu, Y., Liu, T., Paredes, P., Duan, L., Pereira, L.S., 2015. Water use by a groundwater depen-
ARC-WHEAT1 on wheat crops grown in New Zealand. Field Crop Res. 27, 337–350.
dent maize in a semi-arid region of Inner Mongolia: Evapotranspiration partitioning
Jiang, X., Kang, S., Tong, L., Li, F., Li, D., Ding, R., Qiu, R., 2014. Crop coefficient and evapo-
and capillary rise. Agric. Water Manag. 152, 222–232.
transpiration of grain maize modifiedby planting density in an arid region of North-
Zhang, Y., Kang, S., Ward, E.J., Ding, R., Zhang, X., Zheng, R., 2011. Evapotranspiration compo-
west China. Agric. Water Manag. 142, 135–143.
nents determined by sap flow and microlysimetry techniques of a vineyard in North-
Jiang, X., Kang, S., Li, F., Du, T., Tong, L., Comas, L., 2016. Evapotranspiration partitioning
west China: dynamics and influential factors. Agric. Water Manag. 98 (8), 1207–1214.
and variation of sap flow in female and male parents of maize for hybrid seed produc-
Zhang, B., Liu, Y., Xu, D., Zhao, N., Lei, B., Rosa, R.D., Paredes, P., Paço, T.A., Pereira, L.S.,
tion in arid region. Agric. Water Manag. 176, 132–141.
2013. The dual crop coefficient approach to estimate and partitioning evapotranspi-
Kang, S., Liang, Z., Pan, Y., Shi, P., Zhang, J., 2000. Alternate furrow irrigation for maize pro-
ration of the winter wheat–summer maize crop sequence in North China plain.
duction in an arid area. Agric. Water Manag. 45 (3), 267–274.
Irrig. Sci. 31 (6), 1303–1316.
Katerji, N., Rana, G., 2006. Modelling evapotranspiration of six irrigated crops under Med-
Zhao, P., Li, S., Li, F., Du, T., Tong, L., Kang, S., 2015. Comparison of dual crop coefficient
iterranean climate conditions. Agric. For. Meteorol. 138 (1), 142–155.
method and Shuttleworth–Wallace model in evapotranspiration partitioning in a
Katerji, N., Campi, P., Mastrorilli, M., 2013. Productivity, evapotranspiration, and water use
vineyard of Northwest China. Agric. Water Manag. 160, 41–56.
efficiency of corn and tomato crops simulated by AquaCrop under contrasting water
Zhou, L., Li, F., Jin, S., Song, Y., 2009. How two ridges and the furrow mulched with plastic
stress conditions in the Mediterranean region. Agric. Water Manag. 130, 14–26.
film affect soil water, soil temperature and yield of maize on the semiarid loess pla-
Klocke, N.L., Martin, D.L., Todd, R.W., DeHaan, D.L., Polymenopoulos, A.D., 1990. Evapora-
teau of China. Field Crop Res. 113 (1), 41–47.
tion measurements and predictions from soils under crop canopies. Trans. ASAE 33,
1590–1596.

You might also like