Professional Documents
Culture Documents
Agricultural Water Management
Agricultural Water Management
a r t i c l e i n f o a b s t r a c t
Article history: The FAO AquaCrop model has been conceived as a tool for simulating, on a daily scale, the canopy cover
Received 30 October 2012 (CC), biomass and the actual evapotranspiration and for simulating, on a seasonal scale, the final biomass,
Accepted 2 August 2013 the harvested yield, the cumulate actual evapotranspiration, and the crop water use efficiency. This
Available online 1 September 2013
performance was analysed after a series of tests carried on 12 crop cycles, concerning corn and tomato
grown in the Mediterranean region, and having three levels of plant water stress: absence of plant water
Keywords:
stress (control), moderately stressed and severely stressed. The results highlight the effects of three
Plant water stress
factors affecting the AquaCrop performances: the species, the level of plant water stress during the crop
Irrigation
Crop model
cycle and the output variable to simulate.
Water balance The AquaCrop adequately simulates the daily canopy cover (CC) in control treatments of tomato and
corn, and in moderate stress treatment of corn. In the severe stressed treatment of corn, the simulated
values of CC were close to the measured values only from sowing to 60 days after sowing, after that the
simulated values do not fit the measurements.
The AquaCrop model adequately simulates the daily biomass accumulation under all treatments in
tomato and under non-stressed and moderate stressed treatments in corn. However, the simulated
biomass outputs were generally overestimated during the late stages of the crop cycles and, conse-
quently, the yield also exhibited a tendency to be overestimated. Nevertheless, the yield overestimation
can be retained as acceptable because the normalised differences (D) between the simulations and mea-
sured values were less than 15% on average. An exception was the tomato yield simulated in the severely
stressed treatment, for which D was greater than 30%. In contrast, in the case of the severely stressed
treatment in corn, AquaCrop did not exhibit any aptitude for simulating the biomass or the grain yield.
In fact, the model predicts the absence of any yield production, while 5 t ha−1 of grain were actually
measured in the severely stressed treatment.
The daily actual evapotranspiration simulated by AquaCrop was consistent with the observations only
in the case of the control treatments of tomato, in all the three seasons. In contrast, for the other treatments
(all treatments in corn and all stressed treatments in tomato), the quality of the evapotranspiration
simulation was poor.
In general, AquaCrop underestimated the seasonal values of evapotranspiration. The normalised dif-
ferences between the seasonal values of the observations and simulations are acceptable in the case of
the tomato evapotranspiration (D = −7%). However, in the case of corn, the differences are related to the
level of plant water stress, and they become unacceptable (D = −36%) in the severely stressed treatments.
The overestimation of the yield and the underestimation of the seasonal evapotranspiration cause the
simulations of the water use efficiency to be overestimated. In the specific case of corn, due to the unac-
ceptable performance of the model under severely stressed treatments, the linear regression between
the observations and measurements of water use efficiency is unsatisfactory.
The potential uses of the AquaCrop model as a tool for research purposes aimed to enhance the water
efficiency and as a tool for managing irrigation have been deeply discussed. The paths that should be
followed in the future to improve the model simulations have also been suggested
© 2013 Elsevier B.V. All rights reserved.
∗ Corresponding author. Tel.: +39 080 547 5014; fax: +39 080 547 5023.
E-mail addresses: pasquale.campi@entecra.it, pasquale.campi@gmail.com (P. Campi).
0378-3774/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.agwat.2013.08.005
N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26 15
Table 1 density was of 5 plants m−2 . The crop was grown under a high
The main soil characteristics observed at the Rutigliano and Foggia sites.
input of mineral fertiliser supply (120 kg P2 O5 ha−1 before sowing
Soil characteristics Rutigliano Foggia and 100 kg ha−1 of N in two rates) and irrigated according to three
Corn Tomato
water treatments (see Section 2.3). The experimental design was
a randomised block replicated three times. Each plot was 9 m long
Depth (m) 0.6 0.6 1.2
and 9.6 m wide. The dates of the main phenological stages were
Saturated hydraulic conductivity (mm d−1 ) 100 100 100
Water content at the field saturation (vol.%) 48 48 50 observed and used to calculate the growing degree days (GDD),
Water content at the field capacity (vol.%) 34 33.5 39 for each phenological stages, following ‘method 2’ as described by
Water content at the wilting point (vol.%) 16 22.5 23 Raes et al. (2009). The values of GDD observed on control treatment
Total available soil water (mm) 108 66 192
in 1996 are reported in Table 2.
The Leaf Area Index (LAI) measurements are required for cali-
brating the canopy ground cover parameter CC (see Section 2.5.1).
The objective is to analyse the performance of the AquaCrop
The LAI was measured with an area metre (LI-3100C Area Metre, Li-
under different irrigation strategies affecting the plant water status
Cor, USA), and the dry matter was measured after drying (at 65 ◦ C
during the crop cycles and the final productivity.
for 48 h in an oven) the sampled plants. At the end of the corn cycle,
all the plants were harvested from the 10 central rows of each plot,
2. Material and methods and the yields in total biomass and in grain were determined.
Table 2
Comparison between the default values contained in the AquaCrop files (Raes et al., 2009) for corn and tomato crops and the values calibrated at the Rutigliano site.
Corn Tomato
• At the Rutigliano site, the experimental design in 2006 and The probes were linked to Tektronix equipment for the corn
2007 combined two irrigation schedules: the control treatment experiments and to TDR100 (Campbell Scientific, Inc., Logan, UT,
(absence of plant stress), which restored 100% of the readily avail- USA) for the tomato experiments. The probes in turn were con-
able soil water, as calculated according the FAO-56 methodology nected to a CR1000 data logger (Campbell Scientific, Inc., Logan,
(Allen et al., 1998); and the moderately stressed treatment (STR1), UT, USA). Calibration curves were used to calculate the volumetric
which reduced the irrigation volume by 50% compared with the soil water content.
control treatment from 50 DAT until the harvest in the 2006 The drainage (Dr) was estimated as the amount of water exceed-
season, and from 30 DAT until the harvest in 2007 season. The ing the maximum water capacity in the entire soil profile.
irrigation water was supplied by a drip irrigation system 2 or This approach to measure daily ET was previously validated by
3 times per week, according to the evaporative demand of the Mastrorilli et al. (1998) in diverse soil water conditions by compar-
atmosphere and to the phenological stage of the tomato plants. ing on a daily scale the ET from the simplified soil water balance
• At the Foggia site, the experimental design in 2008 combined equation and ET measured by the Bowen ratio method (Rana and
as well two irrigation schedules: the control treatment and the Katerji, 1998). Differences within 10–15% were measured between
severe stress (STR2) treatment, which was obtained by com- the daily ET determined by the two methods (Mastrorilli et al.,
pletely withholding the irrigation supply from 50 DAT until the 1998).
harvest. This change in the experimental protocol aimed to create In the tomato trials, the soil water content data measured using
a water stress higher than that previously observed in Rutigliano. the TDR technique were only available 20 days after transplanting
because agro-technique recommends ploughing the inter-rows
At the tomato plant level, the water stress was monitored at and shaping the soil surface after transplanting and, as a conse-
both sites in all treatments during the entire vegetative cycle, by quence, the TDR probes cannot function in the disturbed soil at
measuring the pre-dawn leaf water potential ( ), following the this time. In this period, the cumulative evapotranspiration was
same experimental procedure adopted for the corn. calculated according to Eq. (1). In particular, W is the difference
between the values of the soil water content measured by the gravi-
2.4. Actual evapotranspiration measurement metric method at transplanting and at the first day when the TDR
data became available.
Daily actual evapotranspiration (ET) was indirectly measured
(Lhomme and Katerji, 1991) using a simplified soil water balance 2.5. The AquaCrop model
approach. At the Rutigliano site, runoff and capillary rise are negli-
gible because of the flat ground and the presence of a cracked rocky Similar to many other crop-growth models, AquaCrop further
layer that limits the soil depth and ascending water. At the Foggia develops sub-model components that include: the soil, the crop,
site, runoff can be neglected because the area is flat, while the cap- the atmosphere, and management. The simulations are executed
illary rise can be assumed to be zero because of the presence of a with a daily time step, using either calendar days or GDD.
calcareous layer at 1.2 m-deep that prevents the roots from expand- The crop responses to water deficits are simulated using four
ing and the water stored in the deeper soil layers from moving up modifiers that are functions of the fractional available soil water
to the soil surface. modulated by the evaporative demand, based on the differential
The simplified equation for the soil water balance can be sensitivity to water stress of four key plant processes: canopy
expressed, at a daily scale, as: expansion, stomatal control of transpiration, canopy senescence,
and the Harvest Index (HI).
ET = ±DW + P − Dr (1)
where ET, daily actual evapotranspiration (mm d−1 ); P, precipita- 2.5.1. Model input
tion or/and irrigation (mm d−1 ); ±W, the difference in volume of The application of the AquaCrop model to a given crop requires
soil water content in the whole soil profile (mm d−1 ) measured with that a series of inputs be determined. They are related to the (a)
TDR (time domain reflectometry) probes; Dr, drainage (mm d−1 ). climate, (b) crop, (c) crop management, and (d) soil properties.
The same technique for monitoring the soil water status was
used for both sites. Coaxial probes (0.3 m in length) were installed (a) Climate inputs concern the daily data of air temperature, rain-
horizontally into the soil at two layers (−15 and −45 cm from the fall and reference evapotranspiration (ETo). According to the
soil surface at Rutigliano and −20 and −60 cm at Foggia) in only FAO-56 methodology (Allen et al., 1998), the latter parame-
one block. ter is calculated based on the daily measurements of global
18 N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26
radiation, wind speed, air humidity and temperature. In this of both crops, we adopted the statistical procedure proposed by
study, all the climatic parameters have been collected from Loague and Green (1991), consisting of the best fit of the simula-
standard agro-meteorological stations within the two farms tions. This procedure is based on the Relative Root Mean Square
where the field experiments were carried out. Error (RRMSE), calculated from the following equation:
(b) The crop inputs concern two types of parameters:
• Conservative crop parameters do not change materially with
n
(P
i=1 i
− Oi )2 100
time, management practices, or geographic location. For RRMSE = · (3)
simulating productivity, evapotranspiration, and water use n Ō
efficiency of corn and tomato, they were assumed to be the
same as those listed in the AquaCrop files (Raes et al., 2009). where n is the number of observations, Pi is the value predicted
• Cultivar specific, or less conservative, parameters are affected by AquaCrop, Oi is the measured value, and Ō is the mean of the
by the climate, field management or conditions in the soil measured values.
profile. In this simulation exercise the cultivar parameters The validation is considered to be excellent when the RRMSE is
were calibrated to take into account the characteristics of the <10%, good if the RRMSE is between 10 and 20%, acceptable if the
retained cultivars. These characteristics were addressed as RRMSE is between 20 and 30%, and poor if >30% (Jamieson et al.,
follows: 1991).
i. Crop development expressed as growing degree days To assess the quality of the model to predict the biomass at the
(GDD) as reported in Table 2. end of the cycle, the yield, and the seasonal evapotranspiration in
ii. The canopy ground cover (CC) is a crucial input (Steduto each water regime treatment of the two crops, we considered the
et al., 2009) for AquaCrop, and for each species it is recom- difference between the simulations and normalised measured val-
mended to verify preliminarily the validity of the values of ues (D, in %). If the values of D did not exceed 15%, as suggested by
this parameter. In this simulation exercise CC values were Brisson et al. (2002), the simulation could be retained as acceptable
verified on the control treatments. As reported by Farahani because the differences fall within the error range that is generally
et al. (2009), the measured CC values can be derived using admitted for field crop research.
the Ritchie model (Ritchie, 1972; Belmans et al., 1983; To assess the quality of the model in predicting water use effi-
Ritchie et al., 1985) from the following equation: ciency and its components (yield and seasonal evapotranspiration)
for each crop under all water regime treatments, the linear regres-
CC = 1 − exp(−K · LAI) (2) sions between the measured and simulated values was analysed.
where LAI was measured while K (extinction coefficient)
was assumed to be 0.60 for corn (Hsiao et al., 2009; Heng
et al., 2009) and 0.75 for tomato (Rinaldi et al., 2011). 3. Results
iii. The minimum effective rooting depths (Table 2) were
measured directly at the earliest development stages by 3.1. Experimental measurements of the plant water status for all
removing the corn and the tomato plants from the soil. treatments of corn and tomato
Whereas the maximum effective rooting depths were
taken as equal to the depth of the soil for the crops cul- Fig. 1(a) and (b) shows the temporal variation in the predawn
tivated at the Rutigliano site (0.6 m) and at the Foggia site leaf water potential ( ) for all treatments in corn during the two
(1.2 m). crop seasons. Differentiation between the treatments began 49 and
iv. The Harvest Index (HI) values were determined 40 days after sowing in 1996 and 1997, respectively. In the STR1
(yield/total biomass) from samples collected at the treatments, it was possible to identify in each year three cycles of
end of the crop cycles. drought when the measured became significantly lower than
The three inputs concerning the cultivar characteristics (i, that measured in the control treatment. In the STR2 treatments,
ii, and iv) were calibrated using data measured in the control there were only two drought cycles, but in each drought cycle the
treatments, they were collected during the 1996 season for corn drought intensity was higher than that measured in the STR1 treat-
and during the 2007 season for tomato. ments because reached very low levels.
(c) For the management inputs, the actual irrigation amount and During the two crop seasons, the rainfall differed in intensity
timing of each treatment were provided as input. The effect and in time distribution. The situation differed at the end of the
of soil fertility on yield was not addressed because sufficient crop seasons, when the extended rainless period in 1997 prolonged
amounts of mineral fertiliser were supplied to all irrigation the water stress period in corn plants until the harvest.
treatments to ensure the realisation of the full genetic potential. The time variations in the soil water availability determined for
(d) Soil inputs concern the parameters previously reported in each treatment during 1996 and 1997 seasons are shown Fig. 1(c)
Table 1. Moreover, the soil water content at the beginning of the and (d). The changes in the soil water availability were synchro-
crop cycle was also supplied as an input. They were measured nised with those in for all water regime treatments during both
using the gravimetric method. years.
Fig. 2(a)–(c) shows the temporal variation in for all tomato
In Table 2, calibrated values of the two crops are reported, for treatments during the three seasons. Differentiation between the
Rutigliano site, in comparison with the default values contained in STR1 and control treatments began 50 and 30 days after transplan-
the AquaCrop files. Generally, minor discrepancies can be observed ting in 2006 and 2007, respectively. Throughout the crop cycle,
between the calibrated and default values. the values of were on average 0.16 MPa in 2006 and 0.11 MPa
in 2007 lower than those measured in the control treatments.
2.5.2. Model validation Differentiation between the high stress treatment STR2 and the
The validation concerns crop parameters (ET, biomass, yield, and control treatments began 51 days after transplanting in 2008. The
WUE) which were not utilised in the calibration process. differences between values measured in the STR2 and con-
To assess the quality of the model to predict the daily values trol treatments increased progressively until the highest value
of canopy cover, biomass, and actual evapotranspiration (ET) mea- 0.37 MPa was reached at the end of the crop cycle, which is 2.7 times
sured during the crop seasons for each water regime treatment higher than the average value measured in the STR1 treatments.
N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26 19
1996 1997
a) b)
Predawn leaf water potential (MPa)
DAS DAS
0 20 40 60 80 100 120 0 20 40 60 80 100 120
0
-0.2
-0.4
-0.6
-0.8 Control
IRR
STR1
-1
STR2
-1.2
c) d)
Soil water avaiability (mm)
250
200
150
100
Control
50 STR1
STR2
0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
DAS DAS
Fig. 1. Values measured in corn during two cropping seasons (1996 and 1997) for three water treatments (control – not stressed, moderately – STR1, and severely stressed
– STR2) of: predawn leaf water potential (a and b); soil water availability (c and d). In c and in d, the values of the wilting point (- - - - - -) and of the field capacity () are also
shown. DAS is the acronym of ‘Days After Sowing’.
-0.1
-0.3
-0.5
Control
IRR
Control
IRR
-0.7 Control
IRR
STR2
STR1 STR1
-0.9
d) e) f)
Soil water avaiability (mm)
Fig. 2. Values measured in tomato during three cropping seasons (2006, 2007 and 2008) for three water treatments (control, STR1 and STR2) of: predawn leaf water potential
(a–c); soil water availability (d–f). In panel d, in e, and in f, the values of the wilting point (- - - - - -) and of the field capacity () are also shown. DAT is the acronym of ‘Days
After Transplanting’. The definition of the three treatments (control, STRI and STR2) is provided in the caption to Fig. 1.
20 N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26
STR2-meas
60 Season Treatment CC ET Biomass
Corn
40 1996 4 37 20
Control
1997 7 41 32
1996 29 50 14
20 STR1
1997 15 42 18
1996 96 70 65
STR2
0 1997 81 68 55
0 20 40 60 80 100 120
Tomato
DAS
2006 23
b) Corn – 1997 2007 11 29 19
Control
100 Control-sim
2008 29 14
STR1-sim
2006 STR1 38
STR2-sim
Control-meas 2007 STR1 41 24
80
STR1-meas 2008 STR2 44 25
Canopy cover (%)
STR2-meas
60
clear trends were lacking. It was only possible to determine that the
60 days when the simulated ET was underestimated alternated with
days when the daily ET was overestimated.
40 The AquaCrop aptitude to forecast the biomass in control treat-
Control-sim
Control-meas ments is good overall, with an exception for the corn grown in 1997.
20 However, the aptitude to predict the daily ET in these treatments
is poor (Table 3).
0 In the moderately stressed treatments STR1 (Fig. 4(c) and (d)),
0 20 40 60 80 100 120
DAT the measured and simulated biomass values were very similar,
mainly during the entire season in 1996. During the successive
Fig. 3. Canopy cover measured (meas) and simulated (sim) for three water treat- cropping season, significant differences became apparent between
ments (control, STR1 and STR2) during two cropping seasons of corn (1996 and 1997, the measured and simulated biomass values. The trend and the time
a and b) and during one cropping season of tomato for the control treatment (2007,
when these differences appeared were similar to those measured
c). DAS and DAT are the acronyms of ‘Days After Sowing’ and ‘Days After Transplan-
ting’. The definition of the three treatments (control, STRI and STR2) is provided in previously in the control treatments.
the caption to Fig. 1. In STR1 (Fig. 5(c) and (d)), AquaCrop underestimated the daily
values of ET during the drought cycles (for example, see the ET
values simulated and measured between 50 and 60 DAS in 1996
In addition, during the tomato crop cycle, the changes in soil and 1997 and between 100 and 120 DAS in 1997). Furthermore,
water availability (Fig. 2(d)–(f)) were noticeably synchronised to this means that AquaCrop overestimated the drought effect on the
for all water treatments during both years. daily ET. In addition, the model underestimates the maximal values
of the daily ET measured soon after the irrigation supplies.
3.2. Simulations and measurements of CC, biomass, and ET The AquaCrop aptitude (Table 3) for predicting the biomass in
during the crop cycles STR1 was more effective than in the case of the control treatments.
However, the model exhibited a poor aptitude to predict the daily
3.2.1. The case of corn grown under contrasting watering ET (Table 3) under the same conditions.
conditions In the severely stressed treatments STR2, the model correctly
AquaCrop was able to simulate accurately the canopy cover (CC) simulated the biomass only for the first 50 DAS (Fig. 4(e) and
development in the control treatments (Fig. 3(a) and (b); Table 3). (f)). After the first drought cycle, AquaCrop predicted that the
In the moderately stressed treatments STR1, simulation could be biomass would not exhibit further growth, consistently with the
retained to a certain extent acceptable, since the model systemati- simulations predicting the absence of CC (Fig. 3(a) and (b)). As a
cally underestimates the maximal CC values. In STR2, the simulated consequence, the differences between the measured and simulated
and measured values of CC were close only from sowing to 60 DAS values increased until reaching the maximal value at the end of the
(days after sowing), after that the simulated values do not fit the crop cycle.
N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26 21
1996 1997
a) b)
25
Control Control
20
15
t ha-1
10 measured
simulated
5
c) d)
25
STR1 STR1
20
15
t ha-1
Biomass
10
0
e) f)
25
STR2 STR2
20
15
ha-1
10
t
0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
DAS DAS
Fig. 4. Biomass measured and simulated by AquaCrop model during two corn cropping seasons (1996 and 1997) for three water treatments: control (a and b), STR1(c and
d) and STR2 (e and f). DAS is the acronym of ‘Days After Sowing’. The definition of the three treatments (control, STRI and STR2) is provided in the caption to Fig. 1.
The simulation of the daily ET in the STR2 treatments (Fig. 5(e) higher under water stress conditions (Fig. 7(d)–(f)). AquaCrop apti-
and (f)) confirms and amplifies the observations previously tude to predict the biomass under water stress conditions was
reported about the STR1 treatments (underestimation of the min- satisfactory. Moreover, the model aptitude to predict the daily ET
imal and maximal values of daily ET). The model shows a poor became poor (Table 3) in all water stress treatments.
aptitude (Table 3) to predict the biomass and ET on a daily scale
in the STR2 conditions. 3.2.3. Simulations and measurements under well watered
treatments: final biomass, yield, seasonal ET and WUE
3.2.2. The case of tomato growing under contrasting watering In terms of grain yield and seasonal ET, the measured corn values
conditions of STR1 decreased on average 22% and 14% with respect to the con-
In the control treatments, the simulated CC (Fig. 3(c)) and trol treatments (Table 4). These reductions were more pronounced
biomass (Fig. 6(a) and (b)) values followed the same trend (on average 43% and 30%, respectively for yield and ET) in the STR2
previously measured in the control treatments of corn. The dif- treatments.
ferences between the simulation and measurements of daily ET In STR1, the yield in fruits and the seasonal ET of tomato
(Fig. 7(a)–(c)) were generally less, particularly in 2006 season, than decreased by an average of 37% and 24% with respect to the control
those previously described for the control treatments of corn. treatments. These reductions were noticeably more pronounced in
As for the three control treatments, the model exhibited a good STR2 (69% for yield and 40% for ET). In both species, the yield was
aptitude (Table 3) to predict CC (in 2007 season) and biomass (2007 reduced more than the seasonal ET.
and 2008 seasons). The daily ET simulations were also acceptable The differences (D) in the final biomass (15–20%) and yield (13%)
(Table 3). measured in the control treatments of corn between the observa-
AquaCrop overestimated the tomato biomass in the stressed tion and simulation during both cropping seasons indicate that the
treatments (Fig. 6(c) and (d)), mainly after 40 DAT and 60 DAT for model overestimates these two parameters. However, the simu-
STR2 and STR1, respectively. In addition, differences between mea- lated values are acceptable, particularly for the yield. Less difference
sured and simulated values were noticeably more pronounced than was observed in the STR1 conditions, ±10% and ±4% for biomass
those measured in the irrigated treatments of the same crop. This and yield, respectively. In contrast, in the case of the STR2 treat-
observation can also be extended to the daily ET values because ments, AquaCrop did not exhibit any aptitude to predict production
the deviations between the simulated and measurements became as the biomass or especially as grain. In the case of corn biomass, the
22 N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26
1996 1997
a) b)
10 simulated
Control observedd
measure Control
8
mm
4
0
c) d)
10
Daily Actual Evapotranspiration
STR1 STR1
8
6
mm
4
1
0
e) f)
10
STR2 STR2
8
6
mm
0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
DAS DAS
Fig. 5. Actual daily evapotranspiration measured and simulated by AquaCrop model during two cropping seasons of corn (1996 and 1997) for three treatments: control (a
and b), STR1(c and d) and STR2 (e and f). DAS is the acronym of ‘Days After Sowing’. The definition of the three treatments (control, STRI and STR2) is provided in the caption
to Fig. 1.
Table 4
Seasonal evapotranspiration (ET), final biomass and yield measured (meas) and simulated (sim) by AquaCrop model for corn and tomato crops grown under three water
treatments: control, STR1 and STR2. The normalised differences (D in %) between simulated and measured values are also reported. The definition of the three treatments
(control, STRI, STR2) is provided in the caption to Fig. 1.
Control 509 483 −5.1 19.9 23.0 15.3 9.2 10.4 13.3
1996 STR1 425 367 −13.6 16.1 14.5 −10.2 6.8 6.5 −4.3
STR2 372 267 −28.2 13.1 5.4 −59.2 5.0 0.0 –
Corn
Control 498 465 −6.6 18.7 22.3 19.5 8.7 9.8 13.0
1997 STR1 442 382 −13.6 15.7 17.3 10.5 7.2 7.5 4.2
STR2 338 214 −36.7 11.6 5.3 −54.3 5.2 0.0 –
2007 2008
a) b)
12
Control Control
10
8 measured
t ha-1
6 simulated
4
2
biomass
0
c) d)
12
STR1 STR2
10
8
ha-1
6
t
4
2
0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
DAT DAT
Fig. 6. Biomass measured and simulated by AquaCrop model during two cropping seasons of tomato (2007 and 2008) for three water treatments: control (a and b), STR1(c)
and STR2(d). DAT is the acronym of ‘Days After Transplanting’. The definition of the three treatments (control, STRI and STR2) is provided in the caption to Fig. 1.
difference between the simulated and measured values was above In general, the simulated values in tomato exhibited the same
54–59%. Moreover, the model predicted the absence of any yield behaviour previously described for corn, consisting in the overes-
production in the STR2 treatments, while approximately 5 t ha−1 timation of final biomass and fruit, and in the underestimation
of grain were actually measured. of the seasonal ET (Table 4). However, the differences in the
The differences in the seasonal ET between the simulation and biomass and seasonal ET observed in all treatments are consid-
measurements observed in all of the corn treatments highlight that ered acceptable (if D < 15%). In the specific case of the fruit yield,
AquaCrop systematically underestimates the seasonal ET. In addi- the differences are acceptable only on the control and STR1 treat-
tion, the D values increased as a function of the level of plant water ments. However, this difference doubles (D = 30%) in the STR2
stress. These differences, however, can be considered unacceptable treatment. The yield simulated in this treatment condition is unac-
(D > 15%) in the STR2 treatments. ceptable.
6
mm
4
Daily Actual Evapotranspiration
0
d) e) f)
10
STR1 STR1 STR2
8
6
mm
0
0 20 40 60 80 100 120 0 20 40 60 80 100 120 0 20 40 60 80 100 120
DAT DAT DAT
Fig. 7. Actual daily evapotranspiration measured and simulated by AquaCrop model during three cropping seasons of tomato (2006, 2007 and 2008)) for three treatments:
control (a–c), STR1 (d and e) and STR2 (f). DAS is the acronym of ‘Days After Sowing’. The definition of the three treatments (control, STRI and STR2) is provided in the caption
to Fig. 1.
24 N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26
a) 600 4. Discussion
Tomato The analysis reported in the previous sections indicates that the
Corn performance of the AquaCrop depends on three factors: the species
Seasonal ETsimulated (mm)
500 (corn or tomato), the water stress level experienced by the plants
during the crop cycle (control, moderately, or severely stressed),
and the output variable (CC, biomass, ET, yield, or WUE).
TOMATO: The validation tests on the control treatments of tomato crop
400 y = 0.95x + 9.5 underline the model aptitude to simulate CC, biomass, ET and yield.
R² = 0.98 This observation is consistent with previous results obtained by
Rinaldi et al. (2011) and Palumbo et al. (2012) on the same species.
CORN: The validation test under water stress conditions reported here
300 y = 1.57x - 313 can be considered as original results, because at the moment they
R² = 0.99 cannot be compared to other observations taken from the litera-
ture. However, these results should be confirmed by further studies
carried out under water stress conditions which consider specific
200
observations on the canopy cover. After that an exhaustive analysis
200 300 400 500 600
can be drawn on the aptitude of AquaCrop to simulate tomato crop
Seasonal ET measured (mm)
productivity and water requirement under soil water stress condi-
b) 12 tion. Nevertheless, the results here obtained in corn and in tomato
Corn
crops under severe water stress (STR2), are coherent and allow con-
Tomato
cluding that AquaCrop showed a poor aptitude in simulating the
Yield simulated (t ha-1)
The inhibition of the canopy expansion, calculated by the model studies under water stress conditions. On the seasonal scale, the ET
under the first drought cycle on the STR2 treatments, appears to validation does not allow for demonstrating this aptitude of the
be highly overestimated. In fact, it is possible to observe that, at model. For example, the seasonal ET values simulated in this study
the end of drought cycle, CC suddenly disappears and it stands for were close to those measured in tomato crops under all water treat-
the condition of LAI dropping to zero. However, the measured CC ments. This low deviation could result from the model aptitude in
values indicate a decrease by 10% during the same period (Fig. 3(a) simulating daily ET close to those measured during the whole crop
and (b)). The absence of LAI acts firstly on simulated transpiration season, as was observed in the case of the control treatment of
and it explains the difference between the model output and the tomato in 2006. In addition, this low deviation could indicate that
measured ET. In the second half of the corn cycle, in comparison the periods when the simulated daily ET is overestimated are bal-
with the measurements, AquaCrop simulates low values of daily ET. anced by periods when the simulated daily ET is underestimated,
This is due to the fact that simulated ET corresponds uniquely to the as was observed in the stressed treatments of tomato crops.
soil evaporation, while measured values of ET take into account also The model aptitude to simulate the daily ET appears to be
the transpiration losses (Fig. 5(e) and (f)). In turn, the absence of leaf acceptable only in the case of the three control treatments in the
surface, as described by Raes et al. (2009), provokes a stop in the tomato crops. Nevertheless, two out of three treatments generated
biomass accumulation (Fig. 4(e) and (f)), and, finally the inhibition RRMSE values equal to 29%, close to the threshold value of 30%,
of the process of yield formation (Table 4), and, as a consequence, which indicates poor quality of the simulation. Moreover, in control
the inappropriate estimation of WUE (Fig. 8). treatments of corn and all stressed treatment of corn and tomato,
In the STR2 treatments the canopy expansion does not react to this aptitude was poor. The AquaCrop model does not appear to be
the plant rehydration, after the irrigation supply (Fig. 1). In the case an innovative tool for managing irrigation water, as was previously
of the STR1 treatments, at the end of a soil drought cycle, AquaCrop ascertained in cotton (Farahani et al., 2009).
simulates to a certain extent the CC raise (Fig. 3(a) and (b)). It should be noted here that the AquaCrop model adopts the
These remarks are in agreement with those previously reported Allen et al. (1998) approach to calculating ET, which is based on the
by Heng et al. (2009), since they underline the need of revising, two-step approach: reference evapotranspiration (ETo) and crop
under severe water treatments, the equations used in AquaCrop coefficient (Kc ). This double approach has generated much criti-
for calculating the water stress. In particular, according the obser- cism when it was applied, mainly in the Mediterranean region (Testi
vations here reported, it is required to revise the estimate of the et al., 2004; Katerji and Rana, 2006; Orgaz et al., 2007; Lovelli et al.,
canopy expansion coefficient. 2007; Irmak and Mutiibwa, 2009). Alternative solutions, based on
In addition, such revision is supported by further considerations. one steep approach, have been proposed to improve the ET calcu-
The same experimental data-set on corn was used in previous stud- lation (Rana et al., 2012). To improve the simulation of the grain
ies to validate the simulation provided by the CERES-Maize (Ben yield in corn, Heng et al. (2009) attempted to increase by 15% the
Nouna et al., 2000) and the STICS (Katerji et al., 2010) models. In ETo calculated according to Allen et al. (1998) but it was unsuccess-
the case of the STR2 treatments, the normalised differences (D) ful. This result suggests avoiding alternative solutions in search of
between simulated and measured values of final yield ranged from a better determination of the daily ET.
15% (in the case of STICS) to 22% (in the case of CERES-Maize).
Moreover, Doorenbos and Kassam (1979) proposed a linear 5. Conclusions
model relating the seasonal evapotranspiration deficit and the yield
deficit for corn. The slope of this linear relation is equal to 1.25. If Based on this study, a series of conclusions are derived which
this simple linear model is used for both STR2 treatments, after could be taken into account for advancing the AquaCrop model.
inputting the simulated values of the evapotranspiration deficit, Because the model performance has been demonstrated to vary
the resulting yield deficits of STR2 are equal to 44% and 37% of the according the level of plant water stress, the model validation tests
control treatments in 1996 and in 1997, respectively. in future should be carried out under very contrasting plant water
The simulations of yield, obtained on STR2 treatments by CERES- stress. The level of water stress, measured at the plant scale, should
Maize, STICS and Doorenbos and Kassam models, are more reliable be clearly identified in any further validation test of the model.
than the absence of a yield on this treatment, as simulated by Additional validation tests under severe water deficit conditions
AquaCrop (Table 4). are required for some species. These tests will allow the proper
Several authors have adopted the AquaCrop as a suitable adjustment of the water stress coefficients which are now available
research tool for studying the optimisation of irrigation water sup- for the model users.
ply and for recommending the appropriate water management To improve the ET simulated by AquaCrop, simple corrections of
decisions (Andarzian et al., 2011; Rinaldi et al., 2011; Garcia- the ETo or Kc values are not appropriate. For calculating ET, alterna-
Vila and Fereres, 2012). In the study from Garcia-Vila and Fereres tive method, based on the one step approach, should be considered.
(2012), performed using four species, including corn, the authors These concluding remarks, if correctly considered, can ame-
consider a wide range of irrigation water allocations to the crops liorate and make more reliable the AquaCrop simulations and
and derive economic analyses, which were obtained by introducing any further analyses concerning the optimisation of the irrigation
an economic sub-routine into AquaCrop. strategies and the economic returns derived from the water supply.
The last studies hypothesise that the validation of the AquaCrop,
performed exclusively on the basis of the yield measured in irri-
gated or moderately stressed crops, can be retained also as the References
validation of this model for simulating other output (yield, ET,
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration: guide-
and WUE), under any condition of plant water stress. The results lines for computing crop water requirements. Irrigation and Drainage paper No.
obtained in this study do not support this hypothesis, particularly in 56. FAO, Rome, pp. 300.
Andarzian, B., Bannayan, M., Steduto, P., Mazraeh, H., Barati, M.E., Barati, M.A., Rah-
the case of corn. However, this hypothesis could be considered real-
nama, A., 2011. Validation and testing of the AquaCrop model under full and
istic for the tomato crop if some corrections are taken into account deficit irrigated wheat production in Iran. Agric. Water Manage. 100, 1–8.
for the model outputs, such as those reported in Fig. 8 of this study. Belmans, C., Wesseling, J.G., Feddes, R.A., 1983. Simulation of the water balance of a
The aptitude of the model to be used as a tool for irrigation cropped soil: SWATRE. J. Hydrol. 63, 271–286.
Ben Nouna, B., Katerji, N., Mastrorilli, M., 2000. Using the CERES-Maize model in a
scheduling should be proven on the basis of its performance in sim- semi-arid Mediterranean environment. Evaluation of model performance. Eur.
ulating the daily ET. This parameter was never analysed in previous J. Agron. 13, 309–322.
26 N. Katerji et al. / Agricultural Water Management 130 (2013) 14–26
Ben Nouna, B., Katerji, N., Mastrorilli, M., 2003. Using the CERES-maize model in a multi-model ensemble seasonal hindcasts in a northern Italy location by means
semi-arid Mediterranean environment. New modelling of leaf area and water of a model of wheat growth and soil water balance. Tellus A 57 (3), 488–497.
stress functions. Eur. J. Agron. 19 (2), 115–123. Marletto, V., Ventura, V., Fontana, G., Tomei, F., 2007. Wheat growth simulation and
Brisson, N., Mary, B., Ripoche, D., Jeuffroy, M.H., Ruget, F., Nicoullaud, B., Gate Ph yield prediction with seasonal forecasts and a numerical model. Agric. Forest
Devienne-Barret, F., Antonioletti, R., Durr, C., Richard, G., Beaudoin, N., Recous, Meteorol. 147 (1-2), 71–79.
S., Tayot, X., Plenet, D., Cellier, P., Machet, J.M., Delecolle, R., 1998. STICS: a Mastrorilli, M., Katerji, N., Rana, G., Ben Nouna, B., 1998. Daily actual evapotranspi-
generic model for the simulation of crops and their water and nitrogen bal- ration measured with TDR technique in Mediterranean conditions. Agric. Forest
ances. I. Theory and parametrization applied to wheat and corn. Agronomie 18, Meteorol. 90, 81–89.
311–346. Mastrorilli, M., Katerji, N., Ben Nouna, B., 2003. Using the CERES-maize model in
Brisson, N., Ruget, F., Gate Ph Lorgeou, J., Nicoullaud, B., Tayot, X., Plenet, D., Jeuffroy, a semi-arid Mediterranean environment. Validation of three revised versions.
M.H., Bouthier, A., Ripoche, D., Mary, B., Justes, E., 2002. STICS a generic model Eur. J. Agron. 19 (2), 125–134.
for simulating crops and their water and nitrogen balances. II. Model validation Molden, D., Oweis, T., Coordinating leading authors, 2007. Pathways for increas-
for wheat and maize. Agronomie 22, 69–92. ing water productivity. In: Molden, D. (Ed.), Water for Food, Water for Life: A
Brisson, N., Gary, C., Justes, E., Roche, R., Mary, B., Ripoche, D., Zimmer, D., Sierra, Comprehensive Assessment of Water Management in Agriculture. International
J., Bertuzzui, P., Burger, P., Bussiere, F., Cabidoche, Y.M., Cellier, P., Debaeke, P., Water Management Institute, London: Earthscan and Colombo, pp. 279–310.
Gaudillere, J.P., Maraux, F., Seguin, B., Sinoquet, H., 2003. An overview of the crop Orgaz, F., Villalobos, F.J., Testi, L., Fereres, E., 2007. A model of daily mean canopy
model STICS. Eur. J. Agron. 18, 309–332. conductance for calculating transpiration of olive canopies. Funct. Plant Bi 34,
Campi, P., Palumbo, A.D., Mastrorilli, M., 2009. Effects of tree windbreak on micro- 178–188.
climate and wheat productivity in a Mediterranean environment. Eur. J. Agron. Palumbo, A.D., Vitale, D., Campi, P., Mastrorilli, M., 2012. Time trend in reference
30, 220–227. evapotranspiration: analysis of a long series of agrometeorological measure-
Castrignanò, A., Katerji, N., Karam, F., Mastrorilli, M., Hamdy, A., 1998. A modified ments in Southern Italy. Irrig. Drain. Syst. 25, 395–411.
version of CERES-Maize model for predicting crop response to salinity stress. Raes, D., Steduto, P., Hsiao, T.C., Fereres, E., 2009. AquaCrop—the FAO crop model to
Ecol. Model. 111, 107–120. simulate yield response to water: II. Main algorithms and software description.
De Benedetto, D., Castrignano, A., Sollitto, D., Modugno, F., Buttafuoco, G., lo Papa, G., Agron. J. 101, 438–447.
2012. Integrating geophysical and geostatistical techniques to map the spatial Rana, G., Katerji, N., 1998. A measurement based sensitivity analysis of
variation of clay. Geoderma 171–172, 53–63. Penman–Monteith actual evapotranspiration model for crops of different height
Deng, X.P., Shan, L., Zhang, H., Turner, N.C., 2006. Improving agricultural water use and in contrasting water status. Theor. Appl. Climatol. 60, 141–149.
efficiency in arid and semiarid areas of China. Agric. Water Manage. 80 (1–3), Rana, G., Katerji, N., Lazzara, P., Ferrara, R.M., 2012. Operational determination of
23–40. daily actual evapotranspiration of irrigated tomato crops under Mediterranean
Doorenbos, J., Kassam, A.H., 1979. Yield Response to Water. FAO Irrigation and conditions by one-step and two-step models: multiannual and local evaluations.
Drainage Paper No 33. FAO, Rome. Agric. Water Manage. 115, 285–296.
Farahani, H.J., Izzi, G., Oweis, T.Y., 2009. Parameterization and evaluation of the Rinaldi, M., Garofalo, P., Rubino, P., Steduto, P., 2011. Processing tomatoes under
AquaCrop model for full and deficit irrigated cotton. Agron. J. 101, 469–476. different irrigation regimes in Southern Italy: agronomic and economic assess-
Garcia-Vila, M., Fereres, E., 2012. Combining the simulation crop model AquaCrop ments in a simulation case study. Ital. J. Agrometeorol. 3, 39–56.
with an economic model for the optimization of irrigation management at farm Ritchie, J.T., 1972. Model for predicting evaporation from a row crop with incomplete
level. Eur. J. Agron. 36, 21–31. cover. Water Resour. Res. 8 (5), 1204–1213.
Heng, L.K., Hsiao, T.C., Evett, S.R., Howell, T.A., Steduto, p., 2009. Testing of Ritchie, J.T., Basso, B., 2008. Water use efficiency is not constant when crop water
FAO AquaCrop model for rainfed and irrigated maize. Agron. J. 101 (3), supply is adequate or fixed: the role of agronomic management. Eur. J. Agron.
488–498. 28 (3), 273–281.
Hsiao, T., Steduto, P., Fereres, E., 2007. A systematic and quantitative approach to Ritchie, J.T., Godwin, D.C., Otter-Nacke, S., 1985. CERES-Wheat: A Simulation Model
improve water use efficiency in agriculture. Irrig. Sci. 25, 209–231. of Wheat Growth and Development. Texas A&M University Press, College
Hsiao, T.C., Heng, L.K., Steduto, P., Raes, D., Fereres, E., 2009. AquaCrop—model Station, TX.
parameterization and testing for maize. Agron. J. 101 (3), 448–459. Scholander, P.F., Hammel, H.T., Bradstreet, E.D., Hemmingsen, E.A., 1965. Sap pres-
Irmak, S., Mutiibwa, D., 2009. On the dynamics of evaporative losses from sure in vascular plants. Science 148, 339–346.
Penman–Monteith with fixed and variable canopy resistance during partial and Steduto, P., Raes, D., Hsiao, T.C., Fcreres, E., Heng, L.K., Howell, T.A., Evett, S.R., Rojas-
complete canopy. Trans. ASABE 52 (4), 1139–1153. Lara, B.A., Farahani, H.J., Izzi, G., Oweis, T.Y., Wani, S.P., Hoogeveen, J., Geerts, S.,
Jamieson, P.D., Porter, J.R., Wilson, D.R., 1991. A test of the computer simulation 2009. Concepts and applications of AquaCrop: the FAO crop water productivity
model ARCWHEAT1 on wheat crops grown in New Zealand. Fields Crop Res. 27, model. In: Cao, W., White, J.W., Wang, E. (Eds.), Crop Modeling and Decision
337–350. Support. Tsinghua University Press, pp. 175–191.
Jones, C.A., Kiniry, J.R., 1986. CERES-Maize. A Simulation Model of Maize Growth and Steduto, P., Hsiao, T.C., Fereres, E., Raes, D., 2012. Crop Yield Response to Water –
Development. Texas A & M University Press, College Station, pp. 194. FAO Irrigation and Drainage Paper 66, Rome (Italy), 500.
Jones, C.A., Dyke, P.T., Williams, J.R., Kiniry, J.R., Benson, V.W., Griggs, R.H., 1991. Stewart, J.L., Danielson, R.E., Hanks, R.J., Jackson, E.B., Hagon, R.M., Pruit, W.O.,
EPIC: an operational model for evaluation of agricultural sustainability. Agric. Franklin, W.T., Riley, J.P., 1977. Optimizing Crop Production Through Control
Syst. 37 (4), 341–350. of Water and Salinity Levels in the Soil. Utah Water Res. Lab., PR 151-1, Logan,
Katerji, N., Bethenod, O., 1997. Comparaison du comportement hydrique et de la Utah, pp. 191.
capacité photosynthétique du mais et du tournesol en condition de contrainte Stöckle, C.O., Nelson, R.L., 2000. CropSyst User’s Manual (Version 3.0), Biological
hydrique. Conclusions sur l’efficience de l’eau. Agronomie 17, 17–24. Systems Engineering Dept. Washington State University, Pullman, WA.
Katerji, N., Rana, G., 2006. Modelling evapotranspiration of six irrigated crops under Stöckle, C.O., Donatelli, M., Nelson, R., 2003. CropSyst, a cropping systems simulation
Mediterranean climate conditions. Agric. Forest Meteorol. 138, 142–155. model Europ. J. Agron. 18, 289–307.
Katerji, N., Mastrorilli, M., Rana, G., 2008. Water use efficiency of crops cultivated in Stricevic, R., Cosic, M., Djurovic, N., Pejic, B., Maksimovic, L., 2011. Assessment
the Mediterranean region: Review and analysis. Eur. J. Agron. 28, 493–507. of the FAO AquaCrop model in the simulation of rainfed and supplemen-
Katerji, N., Mastrorilli, M., Cherni, H.E., 2010. Effects of corn deficit irrigation and tally irrigated maize, sugar beet and sunflower. Agric. Water Manage. 98 (10),
soil properties on water use efficiency. A 25-year analysis of a Mediterranean 1615–1621.
environment using the STICS model. Eur. J. Agron. 32, 177–185. Testi, L., Villalobos, F.J., Orgaz, F., 2004. Evapotranspiration of a young irrigated olive
Loague, K.M., Green, R.E., 1991. Statistical and graphical methods for evaluating orchard in southern Spain. Agric. Forest Meteorol. 21 (1–2), 1–18.
solute transport models: overview and application. J. Contam. Hydrol. 7, 51–73. Todorovic, M., Albrizio, R., Zivotic, L., Abi Saab, M.T., Stöckle, C., Steduto, P., 2009.
Lovelli, S., Perniola, M., Ferrara, A., Tommaso, T.D., 2007. Yield response factor to Assessment of AquaCrop, CropSyst, and WOFOST models in the simulation of
water (ky) and water use efficiency of Carthamus tinctorius L. and Solanum mel- sunflower growth under different water regimes. Agron. J. 3, 509–521.
ongena L. Agric. Water Manage. 92, 73–80. Van Diepen, C.A., Wolf, J., Van Keulen, H., Rappoldt, C., 1989. WOFOST. A simulation
Lhomme, J.P., Katerji, N., 1991. A simple modelling of crop water balance for agrom- model of crop production. Soil Use Manage. 5, 16–24.
eteorological application. Ecol. Model. 57, 11–25. Zwart, S.J., Bastiaanssen, W.G.M., 2004. Review of measured crop water productivity
Marletto, V., Zinoni, F., Criscuolo, L., Fontana, G., Marchesi, S., Morgillo, A., Van Soe- values for irrigated wheat, rice, cotton, and maize. Agric. Water Manage. 69,
tendael, M., Ceotto, E., Andersen, U., 2005. Evaluation of downscaled DEMETER 115–133.