Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Received: 29 May 2023 Revised: 21 November 2023 Accepted: 11 December 2023

DOI: 10.1002/ird.2917

RESEARCH ARTICLE

Comparison of the AquaCrop and CERES-Maize models for


simulating maize phenology, grain yield,
evapotranspiration and water productivity under different
irrigation and nitrogen levels

Ebrahim Amiri 1 | Suat Irmak 2 | Davood Barari Tari 3

1
Department of Water Engineering,
Lahijan Branch, Islamic Azad University, Abstract
Lahijan, Iran The CERES-Maize and AquaCrop models were evaluated for their perfor-
2
Department of Agricultural and mance in simulating maize (Zea mays L.) phenology, evapotranspiration
Biological Engineering, The Penn State
(ETc), grain yield and crop water productivity (CWP). Model input data for
University, State College, Pennsylvania,
USA calibration, validation and simulations were obtained from field experiments
3
Department of Agrotechnology, conducted from 2011 to 2014 that imposed three different irrigation levels
Ayatollah Amoli Branch, Islamic Azad (full irrigation treatment [FIT], limited irrigation treatment [75% FIT] and
University, Amol, Iran
rainfed control) and five nitrogen (N) management treatments. Both models
Correspondence performed well in simulating the maize phenological stages, with simulated
Suat Irmak, Department of Agricultural
values being within 1–4 days of measured values for both models. Both
and Biological Engineering, The Penn
State University, 105A Agricultural models simulated grain yield well during the calibration with a normalized
Engineering Building, University Park, root mean squared error (RMSEn) of 5% (0.6 t ha⁻1) for CERES-Maize and
State College, PA 16802, USA.
8% (0.80 t ha⁻1) for AquaCrop. The model error for ETc varied from 2 to
Email: sfi5068@psu.edu
+14% for CERES-Maize and ranged from 3 to +11% for AquaCrop. High
N levels resulted in higher accuracy in predicting ETc. The AquaCrop model
performed better than CERES-Maize in simulating irrigated crop production
under different N levels under rainfed conditions. The performance of both
models for rainfed systems was poorer than their performance in irrigated
systems, indicating that both models need further improvements in simulat-
ing rainfed maize production systems.

KEYWORDS
crop production, model error, modelling, Nebraska, water requirement

Article title in French: Comparaison des modèles AquaCrop et CERES-maïs pour simuler la phénologie du maïs, le rendement en grains,
l'évapotranspiration et la productivité de l'eau sous différents niveaux d'irrigation et d'azote.
Disclaimer: Mentions of commercial products are solely for the information of the readers and do not constitute an endorsement or recommendation
for use by the authors or their institutions.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any
medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2024 The Authors. Irrigation and Drainage published by John Wiley & Sons Ltd on behalf of International Commission for Irrigation and Drainage.

Irrig. and Drain. 2024;1–17. wileyonlinelibrary.com/journal/ird 1


15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 AMIRI ET AL.

Résumé
Les modèles CERES-Maïs et AquaCrop ont été évalués compte tenu de leurs per-
formances dans la simulation de la phénologie du maïs (Zea mays L.), de l'évapo-
transpiration (ETc), du rendement en grains et de la productivité de l'eau des
cultures (CWP). Les données d'entrée du modèle pour l'étalonnage, la validation
et les simulations ont été obtenues à partir d'expériences sur le terrain menées de
2011 à 2014 qui imposaient trois niveaux d'irrigation différents [traitement d'irri-
gation complet (FIT), traitement d'irrigation limité (FIT à 75%) et contrôle plu-
vial] et cinq traitements de gestion de niveaux d'azote (N). Les deux modèles ont
bien fonctionné dans la simulation des stades phénologiques du maïs, les valeurs
simulées se situant dans un délai de 1 à 4 jours par rapport aux valeurs mesurées
pour les deux modèles. Les deux modèles ont bien simulé le rendement en grains
lors de l'étalonnage avec une erreur moyenne quadratique normalisée (RMSEn)
de 5% (0,6 t ha⁻1) pour CERES-Maïs et de 8% (0,80 t ha⁻1) pour AquaCrop.
L'erreur du modèle pour l'ETc variait de 2 à 14% pour le CERES-Maïs et de -3 à
11% pour l'AquaCrop. Des concentrations élevées d'azote ont entraîné une plus
grande précision dans la prévision de l'ETc. Le modèle AquaCrop a donné de
meilleurs résultats que le CERES-Maïs dans la simulation de la production de
cultures irriguées à différents niveaux d'azote dans des conditions pluviales. La
performance des deux modèles pour les systèmes pluviaux était inférieure à celle
des systèmes irrigués, ce qui indique que les deux modèles doivent encore être
améliorés dans la simulation des systèmes de production de maïs pluvial.

MOTS CLÉS
production végétale, erreur de modèle, modélisation, Nebraska, besoin en eau

1 | INTRODUCTION obtaining high crop yield. However, maize has a greater


potential for productivity per unit of water and nitrogen
Maize is one of the most widely used cereal crops world- than other cereal crops. Maize is not solely impacted by the
wide (Zoghdan, 2021). It is grown as an essential grain crop amount of water or nitrogen but also by the timing of their
for consumption in many places and purposes both by application (Louise & James, 1996). While, in practice, most
humans and animals (Undie et al., 2012), and it provides producers apply seasonal crop N requirements in one appli-
substantial economic benefits to farming and related com- cation (mostly in the autumn), N should be applied in split
munities and industries. It is one of the primary carbohy- applications with the growing season to maximize N use
drate sources for humans and animals and is used as a fuel efficiency, prevent N losses and reduce N leaching potential
source in some countries. The adaptability of maize to vari- to the surface and groundwater sources. Split N application
ous climatic and soil conditions has made it widely culti- meets crop requirements when needed and plays a crucial
vated worldwide. In terms of cultivated land area, maize is role in enhancing plant growth, yield and crop water pro-
the third largest food crop globally, after wheat and rice. ductivity (CWP) (Irmak et al., 2022). Water is also a critical
However, to achieve successful maize production, knowl- input for high maize yield, as it is reduced by water stress,
edge of various management techniques is needed, such as especially during the most water-sensitive stages of tassel-
selecting the appropriate planting dates, choosing hybrids ling, silking and grain filling (Otegui et al., 1995).
suitable for particular locations, adjusting the planting Field experiments gather data that are used to
depth and density, providing adequate soil fertility and develop, calibrate and validate simulation models that
ensuring adequate water supply, which significantly con- can be used to predict the response to environmental var-
tributes to achieving high yields (Fathi et al., 2020). Maize iables (water, nutrients, soil, climate, etc.) as well as man-
is among the most sensitive grain crops to water and nutri- agement practices, differences in cultivars and genetic
ent availability, particularly nitrogen resources, for attributes. However, conducting field research to
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 3

determine these responses and their interactive impacts in vertisols with high clay mineral content. The simula-
on plant productivity is a difficult and critical task; being tion results indicate the necessity to enhance ET estima-
impacted by extreme climatic events that may require tion abilities in rainfed systems compared with irrigated
repeating field experiments to answer complicated ques- crops. Based on a metadata analysis carried out of all
tions may not be possible in some cases. One way to over- studies on ETc modelling for maize using CERES-Maize,
come these limitations is by using crop simulation the model often performs well in irrigated situations
models. These models are cost-effective and simplified (error <10%) but performs poorly in rainfed systems
strategies for testing various management techniques, (Basso et al., 2016). Kimball et al. (2019) noted that most
providing useful data where field analysis is infeasible or ETc simulations by different models, including CERES-
costly (Sandhu & Irmak, 2019a, 2019b). Numerous crop Maize, performed poorly under waterlogged and very dry
simulation models have been developed with varying conditions. The impact of agricultural ETc on crop pro-
levels of complexity, input requirements and user friend- ductivity, and therefore economic yield, is well known
liness, such as Crop-Syst (Stöckle et al., 2003), EPIC because of its interdependence with CO2 assimilation
(Cavero et al., 2000), CERES-Maize (Decision Support (Sinclair & Muchow, 2001). The CERES-Maize model is
System for Agrotechnology Transfer, DSSAT) (Jones capable of accurately simulating crop yield, ET and soil
et al., 2003), AquaCrop (Raes et al., 2009; Steduto water content in response to various irrigation regimes in
et al., 2009) and WOFOST (Boogaard et al., 1998). Crop dry and semi-arid climates (Anothai et al., 2013). Accord-
simulation models can aid in understanding the correla- ing to the study by Babel et al. (2019), both the AquaCrop
tions between various crop production factors, such as and CERES-Maize models under- and overestimated the
irrigation, N, weeds and disease, and their impact on soil water content, respectively, on the basis of results
yield and productivity (Hoogenboom et al., 2019). from the Himalayan region of India. The prediction of
Farmers can use model recommendations to input use grain yield using the AquaCrop model had a 1.0–2.8%
efficiency for specific farming conditions. Crop models prediction error during calibration based on irrigation,
predict a crop's growth, development and yield based soil water depletion and maize N levels. The model
on a combination of genetic coefficients and relevant calibrated for grain production with intense soil mois-
environmental factors. All crop models are a combina- ture depletion had a 4.4–12.3% prediction error
tion of mechanistic and empirical models. The former (Amiri & Abedinpour, 2020). With vigorous calibration,
are based on physical and physiological processes, while validation and performance analyses, crop growth models
the latter fit simulations to the measured data. Models can be effective tools in developing management strate-
entail dozens or hundreds of algorithms with mostly gies and forecasting current and future productivity
determined constants. Models simulate complex crop pro- metrics. In many cases, models can also be valuable
cesses involving growth, development, biomass accumu- tools for researching different and complicated scenar-
lation and yield production reactions to environmental ios for which conducting field research may not be fea-
variables (Monteith, 1996). To effectively analyse and sible. In most cases, crop models are underutilized in
interpret the results of models, their structures, modules practical applications in simulating maize phenology,
and assumptions must be understood to make conclu- yield, ETc and CWP response to water and N. Little
sions and strategies. The crop simulation models used in research has been done on comparing the performance
this study were the CERES-Maize model (radiation/ of different models under various irrigation and N man-
energy-driven) and the AquaCrop model (water-driven). agement conditions using long-term data from field
Both are adaptable to a variety of climates, soil types and research. Modelling and evaluation can be very benefi-
management conditions for simulating a variety of crop cial in developing effective management strategies to
response variables. In some studies, the CERES-Maize enhance crop productivity and minimize environmental
model showed inaccuracy in water and N dynamics, contamination. The practices should be communicated to
implying that the growth and soil water modules require producers for more efficient crop production. Many
improvement for accuracy (Kanda et al., 2018). According studies have been conducted in the field of exploiting
to Menefee et al. (2021), the coefficient of determination multi-year field data to assess the efficiency of varying
(R2) between CERES-Maize and measured ETc was low models in simulating maize plant traits. In light of
due to overestimation of simulated ETc over 3 years. They extensive previous research on calibrating and evaluating
found that the average per cent root mean squared error CERES-Maize and AquaCrop models by various scholars,
(RMSE) for daily ETc simulation across all 3 years was considerable gaps still persist in comprehending the
35% for the Priestley–Taylor method and 30% for the implications of various treatments on the evaluation of
FAO-56 method. Potential sources of overestimation these models, primarily in the context of shifting climatic
included estimation of evapotranspiration (ETc), crop factors and the necessity for sustainable agricultural prac-
coefficient estimation and soil water balance simulation tices. Our current study takes a holistic approach to
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 AMIRI ET AL.

appraising the aforementioned models, encompassing and 0.34 m3 m3, respectively, and a saturation point of
diverse environmental variables as well as undisclosed 0.53 m3 m3 (Irmak, 2010). Maize roots typically reach a
managerial strategies. This investigation seeks to enhance depth of 1.20–1.50 m at the research site (Irmak, 2015a,
the precision of replication and prediction, which has 2015b). A GPS-guided linear (lateral)-move irrigation sys-
important ramifications for augmenting crop output and tem (Valmont Industries, Valley, Neb., USA) with variable
mitigating environmental degradation. In addition, this rate irrigation and fertigation capabilities was used to irri-
research offers a concrete application of these models in gate the field. Irrigations were managed based on the soil
simulating maize phenology, yield, ETc and CWP, which water status of FIT-252 kg N ha⁻1. In each irrigation, soil
have not been adequately tapped into prior practical water was replenished to approximately 90% of field capac-
applications. The objectives of this research were to cali- ity to reserve soil water storage capacity to accommodate
brate and validate the CERES-Maize and AquaCrop any potential rainfall meeting crop water requirements
models and evaluate their performance for simulating and reduce irrigation requirements, which can result in
maize phenology, growth, grain yield, ETc and CWP increased CWP. In each irrigation, 75% of the FITs received
response to different irrigation and N levels. 25% less irrigation water than the FITs, and there were no
irrigation applications in the rainfed treatment. Typically,
irrigations were applied between June and early to mid-
2 | MATERIALS AND METHODS September. Full irrigation was applied when soil water
depletion was approximately 40–50% of the water-holding
2.1 | Field research capacity. The soil water content (SWC) was measured
using a Troxler Model 4300 (Troxler Electronics Laborato-
This research used field research data generated at the Uni- ries, Research Triangle Park, NC) neutron probe at rooting
versity of Nebraska-Lincoln (UNL) South Central Agricul- depths of 0–1.50 m at 0.30 m increments on a weekly basis
tural Laboratory at advanced irrigation engineering, ETc, (sometimes twice a week) in all treatments in all growing
plant physiology and crop production research facilities. seasons. Phenological observations were made every 10–
The research data sets published by Rudnick and Irmak 14 days to record the growth and development stages of
(2014) and Rudnick et al. (2016) were used in the model maize across all treatments from emergence until physio-
calibration, validation and comparisons for performance logical maturity. Field records were kept for the days of
analyses. The research site is located at latitude 40 430 N emergence, silking, anthesis, maturity, maximum canopy
and longitude 98 80 W at an elevation of 552 m above cover (CC) and the onset of senescence. Days to flowering,
mean sea level near Clay Center, Neb., USA (Irmak, 2010, days to maturity, maximum CC and the start of senescence
2015a, 2015b). Additional information about soil and crop were also recorded. Based on neutron gauge measurements
management practices and other information are provided of SWC, a universal soil water balance method was used to
in Rudnick and Irmak (2014) and Rudnick et al. (2016), quantify seasonal crop ETc:
and those resources and procedures relating to this study's
aims will be presented here. Field experiments were con- ETc ¼ P þ I þ U  R  ΔS  D ð1Þ
ducted in 2011, 2012, 2013 and 2014 in a factorial-based
randomized block design with three replications of each where ETc is crop ETc (mm), P is precipitation (mm), I is
treatment and the same management procedures in the irrigation (mm), U is upward water flux (mm), R is run-
same field (Rudnick et al., 2016; Rudnick & Irmak, 2014). off (mm), ΔS is the change in soil water storage
The research was conducted using a split-plot design with (mm) between the beginning and end of the growing sea-
irrigation level as the main treatment (primary effect) and son and D is deep percolation from the crop root zone
in-season N fertilizer in the form of urea ammonium (mm). Because the groundwater depth was over 30 m in
nitrate (UAN 32%) as the subtreatment (secondary effect). the experimental field (Irmak, 2015a), upward water flux
The irrigation treatments investigated were full irrigation was considered negligible. The United States Department
treatment (FIT), 75% FIT and rainfed. The N fertilizer of Agriculture-Natural Resources Conservation Service
treatments were 0, 84, 140, 196 and 252 kg ha⁻1. The (USDA-NRCS, 1985) curve number method was used to
within-field locations for each plot were re-randomized calculate run-off from individual treatments separately. A
each year. The N treatments within each irrigation treat- daily soil water balance computer program was used to
ment were eight rows wide and 50 m long, with 0.76-m calculate deep percolation (Bryant et al., 1992; Djaman &
row spacing in a north–south planting direction. Experi- Irmak, 2012; Payero et al., 2009). The CWP was calcu-
ments were conducted on a Hastings silt loam (20% clay, lated for all treatments for each growing season. The
65% silt, 15% sand), which is a well-drained upland soil term ‘productivity’ refers to the ratio between the prod-
with a permanent wilting point and field capacity of 0.14 uct (e.g. yield) produced per unit of water (e.g. ETc) used.
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 5

This is a similar concept of the CWP term introduced by alignment with our field observations. Key parameters
Molden (1997), which has been gaining widespread addressed in the calibration process included those gov-
acceptance in the scientific community and literature erning CC dynamics and soil water dynamics. Addition-
(Fernandez et al., 1996). The CWP was computed as the ally, adjustments to biomass and yield were made by
ratio of grain yield to ETc: modifying the water productivity and transpiration crop
coefficient (KcTr,x). The calibration of the CERES-Maize
CWP ¼ Y =ETc ð2Þ model involved cultivar-specific observations and
employed DSSAT module calibration. Data on weather,
where CWP is expressed as kg m⁻3, Y is grain yield soil properties and management practices were integrated
(g m⁻2) and ETc is in mm. into the model runs. Crop development and phenology are
affected by cultivar characteristics and air temperature,
among other factors. Growing degree day (GDD) and cal-
2.2 | Model calibration and validation endar day (CDD) modes are available for running the
AquaCrop model. For this study, GDD was selected. A
Based on daily climate data, soil and management trial-and-error approach was used to evaluate the parame-
parameters specific to the site and field experiment data ters iteratively until the simulation values were close to
collected from 2011 to 2014, the CERES-Maize model what was observed in the field. Afterward, the parameters
version 4.7 was calibrated. All calibrations were con- affecting soil water dynamics were calibrated, and then
ducted based on the models' reference manuals. The biomass and yield were calibrated by adjusting the water
AquaCrop model, as parameterized by Hsiao et al. productivity and transpiration crop coefficient (KcTr,x).
(2009), offers globally applicable crop default settings. The phenological stages of crops are critical to estimating
However, the user needs to enter additional crop development accurately. To simulate canopy devel-
cultivar-specific crop information and calibrate the opment accurately, the first estimates of the growth stages
model, including conservative parameters, to achieve were derived from field observations, and model simula-
accurate results for a given situation. In this study, both tions were further adjusted. Time before blooming and
models were first run with default maize parameters with yield formation, time until the harvest is achieved and
poor results, indicating the need to calibrate the model time before flowering are factors that influence the repro-
for specific conditions; therefore, the models were cali- ductive growth phase. When the crop reached full CC,
brated with crop, soil, water and climate variables the blooming stage was initiated. The maximum rooting
observed from the field experiments in 2011 and 2012. depth was reached shortly before the onset of senescence,
According to Sandhu and Irmak (2019a), most variables and the rooting depth at the location was estimated to be
were simulated unsatisfactorily with the default parame- 1.5 m. For maximum KcTr,x, a default value of 1.05 was
ters of the AquaCrop model. After calibration, they used. The calibrated water productivity for the C4 crops
reported that the model adequately simulated most vari- was 31.7 kg m-3, which is within the recommended range.
ables. In this study, 2011 and 2012 data were calibrated All of the parameters mentioned above were calibrated
using the same set of conservative and calibrated parame- under conditions without water stress (i.e. FIT). Stress fac-
ters, and there was no separate calibration for each year. tors affecting canopy expansion, stomatal closure and
A full assessment of the model with many field observa- crop senescence were simulated and observed for condi-
tions is needed to understand its behaviour and the tem- tions of water-stressed treatments (i.e. rainfed and limited
poral variation induced by varied weather parameters, irrigation treatments). Salinity factors were not considered
irrigation, and crop and soil management variables. Thir- during the modeling since the research site did not have
teen parameters of the AquaCrop model were fine-tuned any salinity issues.
using a trial-and-error approach to align the simulated With the CERES-Maize model, cultivar-specific
grain yield, biomass and water productivity with parameter calibration was conducted using DSSAT mod-
observed data. Field experimental data from two irrigated ule calibration. A genetic coefficient was determined
years (2011 and 2012) involving various treatment combi- using cultivar-specific observations. Weather data (daily
nations (as shown in Table 6) were utilized for calibra- maximum and minimum air temperature, rainfall and
tion. The same treatments from the following 2 years of incoming short-wave radiation), soil surface information
data (2013 and 2014) were used for model validation. (slope, colour, drainage, presence of stones/gravel in the
Subsequently, the model underwent calibration using surface), soil profile information as a function of soil
site-specific data, encompassing information on crops, depth (bulk density, drained upper limit, lower limit, N,
soil properties, water and climate variables. Calibrated available phosphorus, soil organic matter, cation
parameters were iteratively adjusted to ensure close exchange capacity, soil pH, soil texture, level of root
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 AMIRI ET AL.

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
abundance across depth), cation exchange capacity, soil 1 Xn
pH, soil texture, level of root abundance (dry biomass of RMSE ¼ i¼1 i
ð S  Oi Þ 2 ð3Þ
n
leaves, stem and grain), crop management data (prior
crop harvested specifications), initial soil moisture con- P P P
Si :Oi  Si : Oi
tent before planting, plant density, planting date, plant R2 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 2 P qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 2 P ffi ð4Þ
population density per m2, date of harvest, fertilizer (type S i  ð S i Þ2  Oi  ð Oi Þ 2
of fertilizer, date of application, amount applied, method
of application) and soil management (tillage) information RMSE
RMSEn ¼  100 ð5Þ
and data were all incorporated into the model runs. For o
the purpose of deriving genetic coefficients for CERES-
Maize cultivars, the generalized likelihood uncertainty P
n
ð Oi  S i Þ 2
estimation (GLUE) approach (He et al., 2020) was used. i1
IOA ¼ 1  n     ð6Þ
P 
GLUE was run 6000 times for each cultivar's parameters, S i  O  þ  O i  O Þ 2
and then human modifications were performed using i1
Boote et al.'s (1996) phenology and growth factor data
approach until acceptable simulation fits were obtained. ð S i  Oi Þ
Pe ¼  100 ð7Þ
To calibrate the models, 13 parameters of the AquaCrop Oi
model were adjusted to match the simulated grain yield,
ETc, water productivity, days to flowering and day where Si and Oi are the model-simulated and field-
maturity of the model with the observed data. The observed values, respectively; O is the mean of the
CERES-Maize outputs were also calibrated for six genetic observed data; and n is the number of the modelled and
coefficients to match the experimental values for grain observation data. An analysis of the differences between
yield, days to flowering, days to maturity and ETc. Field the simulated and observed outputs was conducted by
experiments of various treatment combinations from using the paired t-test. The performance of an RMSEn-
2011 and 2012 were used to calibrate the model. based simulation model is considered to be most satisfac-
tory/acceptable if it is below 10%, good enough if it is
between 10 and 20%, fair if it is between 20 and 30% and
2.2.1 | Model performance assessments poor if it is above 30%, according to Jamieson et al.
(1991).
Model performance comparisons with respect to field-
measured data can provide important information for
practitioners, users and other professionals as to which 3 | RESULTS A ND DISCUSSION
model would best fit their location, crop and soil manage-
ment conditions and climate and which model requires 3.1 | Plant phenology
what input variables and parameters and their availabil-
ity so that educated decisions can be made as to which The model calibration was carried out for both models
model to adopt for simulations. Based on the data col- using default parameters and the measured data. The
lected during field experiments, the model's performance input for the CERES-Maize model includes genetically
was compared with regard to varying levels of irrigation calibrated coefficients that are specific to the crop variety
and N application in 2011 and 2012 for calibration and being used (Table 1). Table 2 presents the input variables
2013 and 2014 for validation. Five variables (grain yield, and crop parameters that were utilized for the AquaCrop
ETc, water productivity, days to flowering and maturity) model simulations. Tables 3 and 4 represent the simula-
were simulated by both models with the calibrated tion of maize plant phenological characteristics in the
parameters, and the results were compared with field Ceres-Maize and AquaCrop models, respectively. Each N
experiment data for the corresponding years. To assess and irrigation treatment scenario was taken into
the simulation/performance accuracy and compare the consideration for this evaluation. Table 3 displays the
simulated and observed results, several statistical indica- performance statistics of the maize module in the
tors were used, including the RMSE (Equation (3)), the CERES-Maize model after calibration. The simulated
coefficient of determination (R2; Equation (4)), normal- duration from planting to flowering and physiological
ized RMSE (RMSEn; Equation (5)), index of agreement maturity closely matched the observed data set. Table 3
(IOA; Equation (6)), paired t-test, and model error or per shows that the simulated time periods from planting to
cent deviation from the field-measured data (Pe; flowering and from planting to physiological maturity
Equation (7)): closely matched the observed data set with an RMSEn of
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 7

TABLE 1 Genotype-specific parameters for maize varieties used in the CERES-Maize model.

Parameter/variable Plant development Value Unit



P1 Thermal time from seedling emergence to the end 255 C days
of juvenile phase
P2 Delay in development for each hour that day- 0.76 Days h⁻1
length is above 12.5 h

P5 Thermal time from silking to physiological 840 C days
maturity (growth aspects)
G2 Maximum kernel number per plant 650 Unitless
G3 Kernel growth rate during linear grain-filling 10.20 mg day⁻1
stage under optimum conditions

PHINT Thermal time between successive leaf tip 38.9 C days tip⁻1
appearance

T A B L E 2 Input data information/parameters for maize used in well. For the simulated crop variety, the RMSE, RMSEn,
the AquaCrop model simulations. IOA and R2 values were 2 days, 3%, 0.9 and 0.8, respec-
Variable/input data Value Unit tively, for the duration until flowering and 4 days, 3%,
 0.89 and 0.65, respectively, for the duration until maturity
Base temperature 0 C

(Table 4). The Pe for the duration until flowering ranged
Cut-off temperature 30 C
from 3 to +3%, while for the duration until maturity, it
Canopy growth coefficient (CGC) 0.16 % day⁻1 ranged from 6% to 0 (Table 4). Under different water
Canopy decline coefficient (CDC) at 0.135 % day⁻1 and N levels, the Pe considering all these variables was
senescence fairly low, indicating that the models were able to simu-
Maximum rooting depth (m) 1.5 m late days to flowering and physiological maturity ade-
Maximum canopy cover (CCx) 0.94 % quately. The calibration of both models employing
Normalized water productivity (WP) 32 kg/m3 default parameters and measured data has yielded prom-
ising outcomes. In particular, the phenological variables
Reference harvest index (HIo) 45 %
exhibited a remarkable congruence with the observed
Time from sowing to emergence 72 GDD
data set, denoted by the low root mean square error of
Time from sowing to start of flowering 900 GDD normalized RMSEn values for both days to flowering and
Time from sowing to start of senescence 1284 GDD days to maturity. The CERES-Maize model has demon-
Time from sowing to physiological 1560 GDD strated a notable efficacy in simulating phenological vari-
maturity ables. It is pertinent to emphasize that these simulations
Length of the flowering stage 156 GDD were executed within the specific confines of this study,
Abbreviation: GDD, growing degree day.
underscoring the model's adaptability to the experimental
context. In a similar vein, the AquaCrop model has effec-
tively delineated the temporal dynamics until flowering
3% for both days to flowering and maturity. The CERES- and maturity, albeit with slightly elevated RMSEn values
Maize model had R2 values of 0.95 and 0.90 for days to in comparison to the CERES-Maize model. The outcomes
flowering and days to maturity, respectively (Table 3). posit that with judicious calibration, AquaCrop holds
The IOA for days to maturity was 0.91, while for days to potential for furnishing precise estimations of these criti-
flowering, it was 0.80. The percentage error (Pe) for phe- cal phenological phases.
nological phases ranged from 4 to +3% for days to flow-
ering and from 4 to +4% for days to maturity (Table 3).
The strong concurrence between the simulated and 3.2 | Grain yield
observed values indicated that, under the conditions of
this study, CERES-Maize was highly effective in simulat- The results of comparisons of the models under different
ing phenological variables. The simulations that utilized treatments are presented in Tables 5 and 6. In these
the calibration data set exhibited a strong correlation tables, the calibration values were obtained from the
with the measured values. AquaCrop effectively average data of 2011 and 2012, while the validation
simulated the duration until flowering and maturity as values were obtained from the average data of 2013 and
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 AMIRI ET AL.

T A B L E 3 Simulated and measured


Days to flowering Days to maturity
days after planting to flowering and
Year Measured Simulated Pe (%) Measured Simulated Pe (%) physiological maturity for growing
2011 75 76 1 130 132 2 seasons of 2011–2014 for the CERES-
Maize model.
2012 74 71 4 130 125 4
2013 76 78 3 136 142 4
2014 80 81 1 146 152 4
Mean 76 77 0 136 138 2
N 4 4
RMSE 2 4
RMSEn 3 3
IOA 0.80 0.91
2
R 0.95 0.90

Abbreviations: IOA, index of agreement; N, number of observations; Pe, per cent error; RMSE, root mean
squared error; RMSEn, normalized root mean squared error.

T A B L E 4 Simulated and measured


Days to flowering Days to maturity
days after planting to flowering and
Year Measured Simulated Pe (%) Measured Simulated Pe (%) physiological maturity for growing
2011 75 75 0 130 129 1 seasons of 2011–2014 for the AquaCrop
model.
2012 74 76 3 130 130 0
2013 76 75 1 136 131 4
2014 80 78 3 146 137 6
Mean 76 76 0 136 132 3
N 4 4
RMSE 2 4
RMSEn 3 3
IOA 0.89 0.77
2
R 0.79 0.65

Abbreviations: IOA, index of agreement; N, number of observations; Pe, per cent error; RMSE, root mean
squared error; RMSEn, normalized root mean squared error.

2014. Table 5 shows the CERES-Maize grain yield simu- varied from 5.3 to 11.6 t ha⁻1 for different treatments. The
lation, which was calibrated with an RMSEn best of 5% CERES-Maize simulated grain yield values varied from
and validated with an RMSEn of 10%. The AquaCrop 6.2 to 11.4 t ha⁻1 and the AquaCrop grain yield simula-
model had an RMSE of 0.8 t ha⁻1, RMSEn of 8% and R2 tions varied from 6.1 to 11.9 t ha⁻1. The model error for
of 0.83 with an IOA of 0.96 during the calibration. How- grain yield varied from 8 to +16% for CERES-Maize
ever, during validation, these statistics were 1.03 t ha⁻1, and from 5 to +16% for AquaCrop. Despite AquaCrop
12%, 0.94 and 0.98 for the same variables, respectively. underestimating the yield more than CERES-Maize, its
Table 6 provides further details on the comparison of the simulations were more accurate based on Figures 1–4.
two models. During the calibration period, the measured Both models simulated yields well, with most data points
grain yield varied from 8.1 to 14.4 t ha⁻1. The CERES- distributed around the 1 : 1 line. Both models performed
Maize simulated grain yield values closely followed the strongly when simulating the grain yield for maize under
measured values and varied from 8.83 to 13.3 t ha⁻1. In different irrigation and N management conditions.
the calibration, the AquaCrop model-estimated grain Although the CERES-Maize model had a lower error dur-
yield varied from 7.1 to 14.2 t ha⁻1. The CERES-Maize ing calibration, the AquaCrop model had higher accu-
model calibration error for grain yield varied from 5 to racy. During validation, both models had similar errors,
+13%, while the AquaCrop model error varied from 17 but AquaCrop tended to slightly underestimate the yield.
to +7%. During validation, the measured grain yield The study found that the calibration and validation of the
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 9

T A B L E 5 Evaluation (calibration and validation) results of CERES-Maize and AquaCrop model simulations of grain yield (t ha⁻1), crop
evapotranspiration (ETc, mm) and crop water productivity (CWP, kg m⁻3) for the 2011–2014 growing seasons.

Variable Model N Xobs Xsim RMSE RMSEn R2 IOA


Model calibration (2011–2012) Grain yield CERES-Maize 27 10.6 10.5 0.6 5 0.91 0.97
AquaCrop 27 10.6 10.2 0.8 8 0.83 0.96
ETc CERES-Maize 27 484 482 43 9 0.43 0.85
AquaCrop 27 484 533 89 17 0.53 0.55
CWP CERES-Maize 27 2.3 2.3 0.1 6 0.70 0.84
AquaCrop 27 2.3 2.0 0.2 7 0.65 0.69
Model validation (2013–2014) Grain yield CERES-Maize 30 10.6 10.5 0.6 10 0.91 0.97
AquaCrop 30 8.7 9.0 1.0 12 0.94 0.98
ETc CERES-Maize 30 8.7 8.8 0.9 10 0.94 0.96
AquaCrop 30 457 469 38 8 0.75 0.86
CWP CERES-Maize 30 457 475 34 8 0.66 0.71
AquaCrop 30 1.9 1.9 0.3 17 0.67 0.67
2
Abbreviations: IOA, index of agreement; N, number of measured/simulated data; R , adjusted linear correlation coefficient between simulated and measured
values; RMSE, absolute root mean squared error; RMSEn (%), normalized root mean squared error; Xobs, mean of measured values; Xsim, mean of simulated
values.

TABLE 6 Observed and simulated grain yield (t ha⁻1) of all treatments for the growing seasons of 2011–2014.

Calibration (2011–2012) Validation (2013–2014)

AquaCrop CERES-Maize AquaCrop CERES-Maize


Grain yield Grain yield
Irrig. Trt Nitrogen Trt Obs. Sim. Pe (%) Sim. Pe (%) Obs. Sim. Pe (%) Sim. Pe (%)
Rainfed 0 8.3 7.0 15 8.9 7 5.6 6.1 10 6.5 16
84 10.2 8.5 17 9.3 0 7.2 8.0 10 8.0 10
140 11.3 9.4 17 10.1 5 8.6 9.0 4 8.9 4
196 11.4 10.3 10 10.7 4 9.2 9.1 2 9.9 8
252 11.7 10.5 10 11.0 4 9.6 9.1 5 10.5 9
75% FIT 0 8.2 8.1 2 8.8 8 5.3 6.1 15 6.2 17
84 9.4 10.1 7 9.9 13 7.1 8.2 16 7.6 7
140 12.2 11.8 3 11.4 1 9.6 9.7 2 8.8 8
196 13.1 13.5 3 12.6 1 10.6 11.2 6 9.9 7
252 13.2 13.7 4 13.1 2 11.5 11.6 1 10.9 6
FIT 0 8.1 8.2 1 8.9 9 5.8 6.1 5 6.6 13
84 10.5 10.4 2 10.0 10 8.1 8.3 2 8.0 2
140 12.5 12.2 2 11.4 0 9.8 9.9 1 9.2 6
196 13.8 14.0 2 12.6 3 11.4 11.4 0 10.4 8
252 14.4 14.2 2 13.3 2 11.6 11.9 3 11.4 1

Abbreviations: FIT, full irrigation treatment; Pe, per cent model error.

CERES-Maize model from 2011–2012 and 2013–2014, and N methods for maize. The study supports previous
respectively, were consistent, and the model showed evidence about the CERES-Maize model evaluation for
strong correlations between simulated and observed maize. Sandhu and Irmak (2019a, 2019b) observed that
grain yield. Therefore, the data indicated that the models the AquaCrop model performed well during calibration
can accurately simulate grain yield for different irrigation and validation for irrigated treatments but performed
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 AMIRI ET AL.

F I G U R E 1 Simulated and measured grain yield (t ha⁻1) for rainfed and irrigated treatments under different nitrogen (N) fertilizer
treatments for the model calibration. FIT, full irrigation treatment.

F I G U R E 2 Simulated and measured evapotranspiration (ETc, mm) for rainfed and irrigated treatments under different nitrogen
(N) treatments for the model calibration. FIT, full irrigation treatment.

F I G U R E 3 Simulated and measured water productivity (WP) for rainfed and irrigated treatments under different nitrogen
(N) treatments for the model calibration. FIT, full irrigation treatment.
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 11

F I G U R E 4 Simulated and measured grain yield (t ha⁻1) for rainfed and irrigated treatments under different nitrogen (N) fertilizer
treatments for the model validation. FIT, full irrigation treatment.

poorly for rainfed treatments. In their investigation of particular, during the senescence stage, Heng et al. (2009)
maize output in Canada, Liu et al. (2011) used the and Katerji et al. (2013) suggested that the AquaCrop
CERES-Maize model to evaluate changes in maize yield algorithm(s) are less acceptable for modelling the length,
and N dynamics. The model produced estimates of yearly timing and intensity of water stress. Our comparative
maize yields that were poorly correlated with the mea- analysis of the models under diverse treatment conditions
sured data, but the general agreement was as strong as has provided intriguing insights. While both models have
what other researchers have observed. Near-surface soil demonstrated their aptitude for simulating grain yield,
mineral N and cumulative soil nitrate loss in the CC-F disparities in accuracy have emerged. During the cali-
treatment were relatively well approximated by the bration phase, the CERES-Maize model exhibited
model, with RMSEn values of 62 and 29%, respectively. reduced RMSEn values, signifying its superiority in the
In contrast, the model consistently overestimated the soil context of these specific experimental conditions. Never-
nitrate loss when applied to the CC-NF treatment (the theless, during the validation phase, both models mani-
treatments selected for this study included continuous fested analogous errors, with AquaCrop marginally
maize with fertilization (CC-F) and continuous maize underestimating yield. This study corroborates anteced-
without fertilization (CC-NF) treatments). These findings ent research, affirming the prowess of the CERES-Maize
could be used to better understand the long-term impacts model in maize yield estimation and accentuating the
of fertilizer management strategies on maize output and precision of AquaCrop under select scenarios. Signifi-
soil characteristics. Malik et al. (2019) evaluated the cantly, our findings advocate that augmented N levels
impact of optimal management practices on irrigation have culminated in heightened precision and the atten-
water requirements and N losses by leaching using sev- uation of performance errors in both models when sim-
eral scenarios. They calibrated and validated the CERES- ulating grain yield. The models have demonstrated
Maize model for maize under various N availability proficiency in simulating grain yield under varying irri-
levels. Among all plots, the model simulated grain yield gation and N management modalities.
reasonably well, with an RMSE lower than 708 kg ha⁻1
and a high Willmott Agreement Index. The prediction
of residual nitrate in soil was acceptable. As a result of 3.3 | Crop ETc
well-managed irrigation, nitrate leaching would be
reduced by 97%, and the total quantity of seasonal irri- Overall, modelling irrigated systems with the CERES-
gation water would be reduced by 31%, without notice- Maize model showed better performance than with
ably lowering grain output. Greaves and Wang (2016) rainfed systems, indicating that the model needs further
studied the effectiveness of the AquaCrop model in improvements in simulating water-limiting systems. This
reproducing CC, biomass, yield, ETc and CWP in a was also reported by Basso et al. (2016) and Menefee
tropical and deficit irrigation setting. They recom- et al. (2021). In the analysis, it is essential to consider the
mended AquaCrop as a reliable tool for assessing the intricacies of crop ETc simulation, particularly during
efficiency of suggested irrigation management options water-deficit scenarios. The AquaCrop model generally
for maize, although caution should be exercised when performed better than CERES-Maize in simulating irri-
analysing the results under high-stress conditions. In gated conditions. During calibration, CERES-Maize had
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 AMIRI ET AL.

an RMSE of 43 mm for simulated ETc, while during vali- and 2014. The CERES-Maize model performance was
dation it had an RMSE of 34 mm. The results for RMSEn, better than that of AquaCrop in estimating ETc for both
R2 and IOA were 9%, 0.43 and 0.85, respectively. The vali- calibration and validation. Higher N levels produced
dation statistics for the same variables were 8%, 0.66 and higher accuracy and reduced the performance error for
0.71, respectively (Table 5). During calibration, the Aqua- both models in simulating ETc, and ETc was not as accu-
Crop simulations had higher RMSE (89 mm) and RMSEn rately simulated as the grain yield. The findings of this
(17%) values than CERES-Maize, while during validation, investigation differed from the study by Paredes et al.
R2 and IOA were 0.53 and 0.55, respectively (Table 5). (2014), but they were consistent with the overall R2 of
During AquaCrop validation for ETc, RMSE, RMSEn, R2 0.81 obtained with the AquaCrop model for simulating
and IOA were 38 mm, 8%, 0.75 and 0.86, respectively seasonal ETc. Adeboye et al. (2019) evaluated the perfor-
(Table 5). In calibration, CERES-Maize simulated ETc mance of AquaCrop using the regression coefficient (b),
ranged from 436 mm for rainfed with no N treatment to R2, RMSE, RMSEn, Nash and Sutcliffe efficiency coeffi-
504 mm for FIT with 252 kg N ha⁻1, while AquaCrop- cient (EF), Willmott's IOA (d-index) and percentage bias
simulated ETc ranged from 477 mm for rainfed with no (PBIAS). AquaCrop simulated CC adequately for two sea-
N fertilizer to 561 mm in FIT with 84 and 140 kg N ha⁻1 sons (0.95 ≤ b ≤ 0.99; R2 ≥ 0.95 and RMSE ≤ 11.6%). The
(Table 7). The CERES-Maize model error (Pe) for ETc model captured the variability in soil water storage well
ranged from 6% (FIT with 196 kg N ha⁻1) to 8% (rainfed with 0.95 ≤ b ≤ 1.01; R2 ≥ 0.74 and RMSE ≤ 10.2 mm.
with 140 kg N ha⁻1). In the AquaCrop model, the model The model simulated ETc with R2 = 0.66 and
error ranged from 4% (FIT with 252 kg N ha⁻1) to 23% RMSE = 65 mm. Dry above-ground biomass was simu-
(75% FIT with 0 N) (Table 7). During AquaCrop valida- lated by the model with an RMSE ≤ 0.25 t ha⁻1 and a
tion, the measured ETc ranged from 399 mm for rainfed d-index ≥ 0.98. With b = 1.01, R2 = 0.99, RMSE of
with no N fertilizer treatment to 497 mm for FIT and 75% 0.03 t ha⁻1, d-index = 1.0 and a maximum difference
FIT with 196 kg N ha⁻1 in 2013 and 2014. The CERES- between the simulated and predicted grain yields as low
Maize model-simulated ETc values ranged from 445 mm as 3%, grain yield was likewise accurately predicted. The
for rainfed with no N treatment to 491 mm for FIT with CWP was simulated with R2 = 0.78 and RMSE = 0.08 kg
252 kg N ha⁻1, and ETc varied from 430 mm for rainfed m⁻3. They emphasized that the AquaCrop model's general
(252 kg N ha⁻1) to 502 mm in FIT (196 kg N ha⁻1) in 2013 performance in replicating CC, dry above-ground

TABLE 7 Observed and simulated ET (mm) of all treatments for the growing seasons of 2011–2014.

Calibration (2011–2012) Validation (2013–2014)

AquaCrop CERES-Maize AquaCrop CERES-Maize


ETc ETc
Irrig. Trt Nitrogen Trt Obs. Sim. Pe (%) Sim. Pe (%) Obs. Sim. Pe (%) Sim. Pe (%)
Rainfed 0 411 477 16 436 6 399 431 8 445 12
84 441 492 11 461 4 407 452 11 462 14
140 430 491 14 462 8 433 452 4 466 8
196 439 493 12 462 5 424 438 3 468 10
252 462 494 7 463 0 433 430 1 468 8
75% FIT 0 454 560 23 467 3 440 450 2 462 5
84 486 552 14 494 2 457 482 5 479 5
140 496 552 11 496 0 474 493 4 485 2
196 517 547 6 498 4 497 496 0 487 2
252 503 548 9 499 1 486 493 2 489 1
FIT 0 472 577 22 471 0 462 450 3 468 1
84 517 561 9 496 4 471 484 3 484 3
140 523 561 7 498 5 493 499 1 487 1
196 531 555 5 499 6 497 502 1 490 1
252 533 556 4 504 5 491 501 2 491 0

Abbreviations: ETc, evapotranspiration; FIT, full irrigation treatment; Pe, per cent model error.
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 13

biomass and grain yield is sufficient but performs poorly had an IOA value of 0.69 and a validation IOA value of
when modelling ETc and CWP for rainfed conditions. 0.67. During CERES-Maize model calibration, the mea-
The results of the models' performance in irrigated sys- sured WP varied from 1.72 kg m⁻3 for FIT (0 N) to 2.6 kg
tems revealed that, predominantly during the calibration m⁻3 for rainfed (196 kg N ha⁻1) in 2011 and 2012, while
phase, CERES-Maize exhibited superior proficiency com- the simulated WP varied from 1.9 kg m⁻3 for FIT and 75%
pared to AquaCrop. However, during the validation FIT (0 N) to 2.7 kg m⁻3 for FIT (252 kg N ha⁻1), which
phase, AquaCrop showed commendable performance, closely matched the measured data. With AquaCrop, the
elucidating its capacity to effectively emulate irrigated WP varied from 1.4 kg m⁻3 for FIT (0 N) to 2.4 kg m⁻3 in
systems. These results are in accordance with previous FIT (252 kg N ha⁻1) (Table 8), which also agreed well
research endeavours. It is worth noting that the fidelity of with the measured data. The model error for water pro-
ETc simulations did not attain the same level of precision ductivity varied from 11% (in rainfed with 140 kg N
as grain yield. Both models have a margin for improve- ha⁻1) to 15% in FIT (84 kg N ha⁻1) with the CERES-Maize
ment in the simulation of ETc, as discerned from the dis- model, while it varied from 26% (in rainfed with 0 N) to
parities between simulated and observed values. close to zero in FIT (84 kg N ha⁻1) (Table 8) with Aqua-
Crop. During validation, the measured water productivity
varied from 0.9 kg m⁻3 for 75% FIT (0 N) to 2.0 kg m⁻3 for
3.4 | CWP 75% FIT (252 kg N ha⁻1) in 2013 and 2014, while the sim-
ulated water productivity values varied from 1.2 kg m⁻3
CWP is an effective indicator of the response of crop pro- for 75% FIT (0 N) to 2.3 kg m⁻3 for 75% FIT (252 kg N
duction systems to water, nutrients and other manage- ha⁻1) in the CERES-Maize model. The AquaCrop model
ment practices. In terms of calibration and validation for had simulated water productivity values that varied from
CWP assessment, the CERES-Maize model had lower 1.3 kg m⁻3 for 75% FIT and FIT (0 N) to 2.2 kg m⁻3 in FIT
RMSE and RMSEn than AquaCrop. During the calibra- (252 kg N ha⁻1) in 2013 and 2014 (Table 8). Compared
tion, the CERES-Maize model had a higher R2 (0.70) than with the CERES-Maize model, the AquaCrop model had
AquaCrop (0.65) (Table 5). For calibration and validation, higher simulated water productivity values (Figures 3–6).
the CERES-Maize model generated IOA values of 0.84 The lowest observed and simulated values of water pro-
and 0.83, respectively. The AquaCrop model calibration ductivity were obtained with the 0 N treatment, while the

TABLE 8 Observed and simulated water productivity (CWP, kg m⁻3) of all treatments for the growing seasons of 2011–2014.

Calibration (2011–2012) Validation (2013–2014)

AquaCrop CERES-Maize AquaCrop CERES-Maize


CWP CWP
Irrig. Trt Nitrogen Trt Obs. Sim. Pe (%) Sim. Pe (%) Obs. Sim. Pe (%) Sim. Pe (%)
Rainfed 0 2.0 1.5 26 2.0 1 1.2 1.4 17 1.4 17
84 2.1 1.7 18 2.0 4 1.3 1.7 35 1.7 35
140 2.5 1.9 22 2.2 11 1.3 1.9 41 1.9 41
196 2.6 2.1 18 2.3 9 1.6 1.8 16 2.1 15
252 2.5 2.2 12 2.4 4 1.7 1.7 2 2.2 2
75% FIT 0 1.8 1.4 20 1.9 5 0.9 1.3 45 1.2 44
84 1.8 1.8 3 2.0 12 1.1 1.6 45 1.5 45
140 2.3 2.0 11 2.3 1 1.5 1.9 27 2.0 27
196 2.5 2.3 6 2.5 3 1.6 2.1 31 2.1 31
252 2.6 2.4 6 2.6 3 2.0 2.2 10 2.3 10
FIT 0 1.7 1.4 17 1.9 9 1.0 1.3 36 1.2 37
84 1.8 1.8 0 2.0 15 1.4 1.6 21 1.7 20
140 2.2 2.1 7 2.3 5 1.5 1.9 28 2.0 27
196 2.5 2.3 5 2.5 3 1.9 2.2 17 2.3 16
252 2.5 2.4 5 2.7 4 1.9 2.2 17 1.9 17

Abbreviations: CWP, crop water productivity; FIT, full irrigation treatment; Pe, per cent model error.
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 AMIRI ET AL.

F I G U R E 5 Simulated and measured evapotranspiration (ETc, mm) for rainfed and irrigated treatments under different nitrogen
(N) treatments for the model validation. FIT, full irrigation treatment.

F I G U R E 6 Simulated and measured water productivity (WP, kg m⁻3) for rainfed and irrigated treatments under different nitrogen
(N) treatments for the model calibration. FIT, full irrigation treatment.

highest were obtained with higher N levels. Under differ- year, the model production estimates were 92, 97, 98 and
ent irrigation and N levels, Ji et al. (2014) evaluated the 101% for 1982, 1983, 1984 and 1985, respectively, of the
CERES-Maize model for wheat crops in the Guanzhong official US government estimates. They suggested that
Plain of north-west China and reported RMSEn values the model can successfully be used for large area yield
less than 21%, indicating reasonable deviations between and production estimation in the United States with min-
simulated and observed CWP. An RMSEn value below imal regional calibration. The findings of the current
20% was achieved in this current study, which allowed study suggest the adequacy and accuracy of both models
for a high level of accuracy in the simulation of water in simulating maize CWP. The results revealed that the
productivity. As a result of strong CWP simulations, the CERES-Maize model, when scrutinized through the lens
CERES-Maize model has shown adequate efficiency and of RMSE and RMSEn metrics, outperformed AquaCrop,
robustness in simulations across all treatments. Hodges accentuating its pre-eminence in modelling CWP. Never-
et al. (1987) investigated the ability of the CERES-Maize theless, AquaCrop has delivered commendable perfor-
model to estimate annual fluctuations in maize produc- mance, especially in the validation phase, underscoring
tion for the US Cornbelt and ran the model for 51 weather its reliability in CWP assessment. The models have eluci-
stations in the 14 states, which account for 85% of US dated that across varied irrigation and N levels, CWP
maize production at the time of their study. The model values have undergone pronounced variations. The inter-
was calibrated for the region by deriving varietal coeffi- play between N levels and CWP values has been brought
cients for each station based on minimal growth stage to the forefront, with elevated N levels translating to an
data and yields for the 1982 season. For the calibration augmentation in CWP. Different scenarios were used to
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 15

compare the models in this study, including Scenario A: AquaCrop model. However, AquaCrop has assumed the
optimal irrigation and optimal N consumption, Scenario mantle of supremacy in the domain of CWP. This mantle
B: low irrigation and optimal N consumption, Scenario is particularly evident in terms of R2 and IOA values,
C: optimal irrigation and low N consumption and Sce- which underscore its unparalleled precision in CWP sim-
nario D: low irrigation and low N consumption. In ulations. The critical selection between CERES-Maize
Scenario A, characterized by optimal levels of irrigation and AquaCrop hinges upon the exacting demands of a
and N input, both models consistently demonstrated given study. CERES-Maize's exceptional calibration per-
robust performance across all variables. These findings formance renders it an attractive choice under certain cir-
suggest that the models exhibit notable effectiveness cumstances. Nevertheless, AquaCrop's all-encompassing
when applied in agricultural settings characterized by superiority in precision, particularly in the context of ETc
high input levels. Scenario B, characterized by limited and CWP, renders it a compelling choice for specialized
irrigation but optimal N inputs, provided insights into applications. The performance of both models varied
the models' capacity to navigate scenarios marked by depending on the specific treatment combinations and
nutrient-rich yet water-scarce conditions. Notably, the time periods. Both models performed well under limited
CERES-Maize model displayed heightened sensitivity to and full irrigation conditions. However, their accuracy
water stress under such conditions, while the AquaCrop needs to be improved for rainfed conditions.
model exhibited a more balanced response. Scenario C,
which emphasized sufficient irrigation coupled with A C KN O WL ED G EME N T S
restricted N inputs, underscored the pivotal role of nutri- The work presented in this paper is a part of a long-term
ent management. The performance of the AquaCrop research that investigates the fundamentals of coupled
model was discernibly affected by the availability of N, irrigation water and nitrogen management strategies on
whereas the CERES-Maize model demonstrated resil- grain yield, water productivity, evapotranspiration, evap-
ience to variations in N inputs. In Scenario D, which is oration and transpiration dynamics, yield production
representative of scenarios characterized by low levels of functions, soil-water dynamics and yield response factors,
irrigation and N inputs, both models encountered notable and other productivity variables and environmental rela-
challenges. In this context, AquaCrop exhibited a height- tionships for different cropping systems in the Irmak
ened sensitivity to limitations in both water and nutrient Research Laboratory’s advance irrigation engineering,
availability, while the CERES-Maize model maintained a evapotranspiration, plant physiology and climate science
moderate level of performance. research facilities. The authors thank Daran Rudnick
(former MS and Ph.D. student of Professor Suat Irmak)
and Matthew Drudik (former MS student and Research
4 | C ON C L U S I ON Technologist of Professor Suat Irmak) for their work and
contributions in field research and data collection. The
This study sought to comprehensively evaluate maize trade names or commercial products are provided solely
growth and productivity variables by rigorously using the for the information of the reader and do not constitute a
CERES-Maize and AquaCrop models. The evaluation recommendation for use by the authors or their
was conducted across a spectrum of scenarios encom- institutions.
passing various irrigation and N application rates. Within
the scope of this investigation, the models were meticu- DA TA AVAI LA BI LI TY S T ATE ME NT
lously juxtaposed against field-observed data. The results The corresponding author, Professor Suat Irmak, can pro-
showed that both the CERES-Maize and AquaCrop vide data used in this study with a reasonable request.
models effectively demonstrated their ability to faithfully
replicate critical phenological milestones, most notably ORCID
the timing of flowering and maturity. The commendable Suat Irmak https://orcid.org/0000-0003-0583-2861
congruence between their simulations and real-world
observations attests to the efficacy of the models in RE FER EN CES
encapsulating the pivotal phases of maize growth. This Adeboye, O.B., Schultz, B., Adekalu, K.O. & Prasad, K. (2019) Per-
formance evaluation of AquaCrop in simulating soil water stor-
phase witnessed the model's most compelling perfor-
age, yield, and water productivity of rainfed soybeans (Glycine
mances. While both models have shown potential, Aqua-
max L. merr) in Ile-Ife, Nigeria. Agricultural Water Manage-
Crop's consistency in offering higher levels of accuracy in ment, 213, 1130–1146. Available from: https://doi.org/10.1016/j.
most of the simulated scenarios is particularly notewor- agwat.2018.11.006
thy. In the CWP assessment, the CERES-Maize model Amiri, E. & Abedinpour, M. (2020) Simulating maize yield response
offered lower RMSE and RMSEn values than the to depletion of available soil water and nitrogen management
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 AMIRI ET AL.

under drip irrigation with the FAO AquaCrop model. Russian Heng, L.K., Hsiao, T., Evett, S., Howell, T. & Steduto, P. (2009) Vali-
Agricultural Sciences, 46(6), 602–608. Available from: https:// dating the FAO AquaCrop model for irrigated and water defi-
doi.org/10.3103/S1068367420060038 cient field maize. Agronomy Journal, 101(3), 488–498. Available
Anothai, J., Soler, C.M.T., Green, A., Trout, T.J. & Hoogenboom, G. from: https://doi.org/10.2134/agronj2008.0029xs
(2013) Evaluation of two evapotranspiration approaches simu- Hodges, T., Botner, D., Sakamoto, C. & Haug, J.H. (1987) Using the
lated with the CSM–CERES–Maize model under different irri- CERES-Maize model to estimate production for the U.S. Cornbelt.
gation strategies and the impact on maize growth, development Agricultural and Forest Meteorology, 40(4), 293–303. Available
and soil moisture content for semi-arid conditions. Agricultural from: https://doi.org/10.1016/0168-1923(87)90043-8
and Forest Meteorology, 176, 64–76. Available from: https://doi. Hoogenboom, G., Porter, C.H., Boote, K.J., Shelia, V.,
org/10.1016/j.agrformet.2013.03.001 Wilkens, P.W., Singh, U. et al. (2019) The DSSAT crop model-
Babel, M.S., Deb, P. & Soni, P. (2019) Performance evaluation of ing ecosystem. In: Advances in crop modelling for a sustainable
AquaCrop and DSSAT-CERES for maize under different irriga- agriculture. Burleigh Dodds Series in Agricultural Science.
tion and manure application rates in the Himalayan region of Cambridge, UK: Burleigh Dodds Science Publishing, pp. 173–
India. Agricultural Research, 8(2), 207–217. Available from: 216 https://doi.org/10.19103/AS.2019.0061.10
https://doi.org/10.1007/s40003-018-0366-y Hsiao, T.C., Heng, L., Steduto, P., Rojas-Lara, B., Raes, D. &
Basso, B., Liu, L. & Ritchie, J.T. (2016) A comprehensive review of Fereres, E. (2009) AquaCrop—the FAO crop model to simulate
the CERES-wheat, -maize and -rice models' performances. yield response to water: III. Parameterization and testing for
Advances in Agronomy, 136, 27–132. Available from: https:// maize. Agronomy Journal, 101(3), 448–459. Available from:
doi.org/10.1016/bs.agron.2015.11.004 https://doi.org/10.2134/agronj2008.0218s
Boogaard, H. L., van Diepen, C. A., Rötter, R. P., Cabrera, J. M. Irmak, S. (2010) Nebraska Water and Energy Flux Measurement,
C. A., & van Laar, H. H. (1998). User's guide for the WOFOST Modeling, and Research Network (NEBFLUX). Transactions of
7.1 crop growth simulation model and WOFOST Control Cen- the ASABE, 53(4), 1097–1115. Available from: https://doi.org/
ter 1.5. Tech. Doc. 52. DLO-Win and Staring Centre, Wagenin- 10.13031/2013.32600
gen, the Netherlands. Irmak, S. (2015a) Interannual variation in long-term center pivot–
Boote, K.J., Jones, J.W. & Pickering, N.B. (1996) Potential uses and irrigated maize evapotranspiration and various water produc-
limitations of crop models. Agronomy Journal, 88(5), 704–716. tivity response indices. I: grain yield, actual and basal evapo-
Available from: https://doi.org/10.2134/agronj1996.0002196 transpiration, irrigation-yield production functions,
2008800050005x evapotranspiration-yield production functions, and yield
Bryant, K.J., Benson, V.W., Kiniry, J.R., Williams, J.R. & response factors. Journal of Irrigation and Drainage Engineer-
Lacewell, R.D. (1992) Simulating corn yield response to irriga- ing, 141(5), 04014068. Available from: https://doi.org/10.1061/(
tion timings: validation of the EPIC model. Journal of Produc- asce)ir.1943-4774.0000825
tion Agriculture, 5(2), 237–242. Available from: https://doi.org/ Irmak, S. (2015b) Inter-annual variation in long-term center pivot–
10.2134/jpa1992.0237 irrigated maize evapotranspiration and various water produc-
Cavero, J., Farre, I., Debaeke, P. & Faci, J.M. (2000) Simulation of tivity response indices. II: irrigation water use efficiency, crop
maize yield under water stress with the EPICphase and CROP- WUE, evapotranspiration WUE, irrigation-evapotranspiration
WAT models. Agronomy Journal, 92(4), 679–690. Available use efficiency, and precipitation use efficiency. Journal of Irri-
from: https://doi.org/10.2134/agronj2000.924679x gation and Drainage Engineering, 141(5), 1–11. Available from:
Djaman, K. & Irmak, S. (2012) Soil water extraction patterns and https://doi.org/10.1061/(asce)ir.1943-4774.0000826
crop, irrigation, and evapotranspiration water use efficiency of Irmak, S., Mohammed, A.T. & Kukal, M.S. (2022) Maize response
maize under full and limited irrigation and rainfed settings. to coupled irrigation and nitrogen fertilization under center
Transactions of the ASABE, 55(4), 1223–1238. Available from: pivot, subsurface drip and surface (furrow) irrigation: growth,
https://doi.org/10.13031/2013.42262 development and productivity. Agricultural Water Manage-
Fathi, A., Barari Tari, D., Fallah Amoli, H. & Niknejad, Y. (2020) ment, 263(2022), 107457. Available from: https://doi.org/10.
Study of energy consumption and greenhouse gas (GHG) emis- 1016/j.agwat.2022.107457
sions in corn production systems: influence of different tillage Jamieson, P.D., Porter, J.R. & Wilson, D.R. (1991) A test of the com-
systems and use of fertilizer. Communications in Soil Science puter simulation model ARCWHEAT1 on wheat crops grown
and Plant Analysis, 51(6), 769–778. Available from: https://doi. in New Zealand. Field Crops Research, 27(4), 337–350. Available
org/10.1080/00103624.2020.1729373 from: https://doi.org/10.1016/0378-4290(91)90040-3
Fernandez, J.E., Moreno, F., Murillo, J.M., Cayuela, J.A., Ji, J., Cai, H., He, J. & Wang, H. (2014) Performance evaluation of
Fernandez-Boy, E. & Cabrera, F. (1996) Water use and yield of CERES-Wheat model in Guanzhong Plain of Northwest China.
maize with two levels of nitrogen fertilization in SW Spain. Agricultural Water Management, 144, 1–10. Available from:
Agricultural Water Management, 29(2), 215–233. Available https://doi.org/10.1016/j.agwat.2014.04.016
from: https://doi.org/10.1016/0378-3774(95)01192-7 Jones, J.W., Hoogenboom, G., Porter, C.H., Boote, K.J.,
Greaves, G.E. & Wang, Y.M. (2016) Assessment of FAO AquaCrop Batchelor, W.D., Hunt, L.A. et al. (2003) The DSSAT cropping
model for simulating maize growth and productivity under def- system model. European Journal of Agronomy, 18(3–4), 235–
icit irrigation in a tropical environment. Water, 8(12), 557. 265. Available from: https://doi.org/10.1016/s1161-0301(02)
Available from: https://doi.org/10.3390/w8120557 00107-7
He, J., Yang, K., Tang, W., Lu, H., Qin, J., Chen, Y. et al. (2020) The Kanda, E.K., Mabhaudhi, T. & Senzanje, A. (2018) Coupling hydro-
first high-resolution meteorological forcing dataset for land pro- logical and crop models for improved agricultural water
cess studies over China. Scientific Data, 7(1), 25. Available management—a review. Bulgarian Journal of Agricultural Sci-
from: https://doi.org/10.1038/s41597-020-0369-y ence, 24(3), 380–390.
15310361, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ird.2917, Wiley Online Library on [25/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AMIRI ET AL. 17

Katerji, N., Campi, P. & Mastrorilli, M. (2013) Productivity, evapo- Drainage Engineering, 140(12), 04014039. Available from:
transpiration, and water use efficiency of corn and tomato https://doi.org/10.1061/(ASCE)IR.1943-4774.0000778
crops simulated by AquaCrop under contrasting water stress Rudnick, D., Irmak, S., Ferguson, R., Shaver, T., Djaman, K.,
conditions in the Mediterranean region. Agricultural Water Slater, G. et al. (2016) Economic return vs crop water productiv-
Management, 130, 14–26. Available from: https://doi.org/10. ity of maize for various nitrogen rates under full irrigation, lim-
1016/j.agwat.2013.08.005 ited irrigation, and rainfed settings in south central Nebraska.
Kimball, B.A., Boote, K.J., Hatfield, J.L., Ahuja, L.R., Stockle, C., Journal of Irrigation and Drainage Engineering, 142(6), 1-12.
Archontoulis, S. et al. (2019) Simulation of maize evapotranspi- 04016017. Available from: https://doi.org/10.1061/(ASCE)IR.
ration: an inter-comparison among 29 maize models. Agricul- 1943-4774.0001023
tural and Forest Meteorology, 271, 264–284. Available from: Sandhu, R. & Irmak, S. (2019a) Performance of AquaCrop model in
https://doi.org/10.1016/j.agrformet.2019.02.037 simulating maize growth, yield, and evapotranspiration under
Liu, H.L., Yang, J.Y., Drury, C.F., Reynolds, W.D., Tan, C.S., rainfed, limited and full irrigation. Agricultural Water Manage-
Bai, Y.L. et al. (2011) Using the DSSAT-CERES-Maize model to ment, 223, 105687. Available from: https://doi.org/10.1016/j.
simulate crop yield and nitrogen cycling in fields under long- agwat.2019.105687
term continuous maize production. Nutrient Cycling in Agroeco- Sandhu, R. & Irmak, S. (2019b) Assessment of AquaCrop model in
systems, 89(3), 313–328. Available from: https://doi.org/10. simulating maize canopy cover, soil-water, evapotranspiration,
1007/s10705-010-9396-y yield and water productivity for different planting dates and
Louise, B.B. & James, B.P. (1996) America's garden book. New York: densities under irrigated and rainfed conditions. Agricultural
Macmillan USA, p. 768. Water Management, 224, 105753. Available from: https://doi.
Malik, W., Isla, R. & Dechmi, F. (2019) DSSAT-CERES-maize org/10.1016/j.agwat.2019.105753
modelling to improve irrigation and nitrogen management Sinclair, T.R. & Muchow, R.C. (2001) System analysis of plant traits
practices under Mediterranean conditions. Agricultural Water to increase grain yield on limited water supplies. Agronomy
Management, 213, 298–308. Available from: https://doi.org/10. Journal, 93(2), 263–270. Available from: https://doi.org/10.
1016/j.agwat.2018.10.022 2134/agronj2001.932263x
Menefee, D., Rajan, N., Cui, S., Bagavathiannan, M., Schnell, R. & Steduto, P., Hsiao, T.C., Raes, D. & Fereres, E. (2009) AquaCrop-the
West, J. (2021) Simulation of dryland maize growth and evapo- FAO crop model to simulate yield response to water: I. Concepts
transpiration using DSSAT-CERES-Maize model. Agronomy and underlying principles. Agronomy Journal, 101(3), 426–437.
Journal, 113(2), 1317–1332. Available from: https://doi.org/10. Available from: https://doi.org/10.2134/agronj2008.0139s
1002/agj2.20524 Stöckle, C.O., Donatelli, M. & Nelson, R. (2003) CropSyst, a crop-
Molden D. 1997. Accounting for water use and productivity. System- ping systems simulation model. European Journal of Agronomy,
wide initiative on water management (SWIM). Rep. 1. Interna- 18(3–4), 289–307. Available from: https://doi.org/10.1016/
tional Irrigation Management Institute, Sri Lanka. p. 1–16. s1161-0301(02)00109-0
Monteith, J.L. (1996) The quest for balance in crop modeling. Undie, U.L., Uwah, D.F. & Attoe, E.E. (2012) Effect of intercropping
Agronomy Journal, 88(5), 695–697. Available from: https://doi. and crop arrangement on yield and productivity of late season
org/10.2134/agronj1996.00021962008800050003x maize/soybean mixtures in the humid environment of south
Otegui, M.E., Andrade, F.H. & Suero, E.E. (1995) Growth, water southern Nigeria. The Journal of Agricultural Science, 4(4), 37–
use, and kernel abortion of maize subjected to drought at silk- 50. Available from: https://doi.org/10.5539/jas.v4n4p37
ing. Field Crops Research, 40(2), 87–94. Available from: https:// USDA-National Agricultural Statistics Service (NRCS). (1985)
doi.org/10.1016/0378-4290(94)00093-R “Hydrology.” Section 4, National engineering handbook, Soil
Paredes, P., Rodrigues, G.C., Alves, I. & Pereira, L.S. (2014) Parti- Conservation Service (SCS). Washington, DC.
tioning evapotranspiration, yield prediction and economic Zoghdan, M.G. (2021) Evaluation of maize water productivity by
returns of maize under various irrigation management strate- FAO-AquaCrop model using integration of mulching, nitrogen
gies. Agricultural Water Management, 135, 27–39. Available fertilization and irrigation in saline and non-saline soils.
from: https://doi.org/10.1016/j.agwat.2013.12.010 Menoufia Journal of Soil Science, 6(7), 197–211. Available from:
Payero, J.O., Tarkalson, D.D., Irmak, S., Davison, D. & Petersen, J.L. https://doi.org/10.21608/mjss.2021.192510
(2009) Effect of timing of a deficit-irrigation allocation on corn
evapotranspiration, yield, water use efficiency and dry mass.
Agricultural Water Management, 96(10), 1387–1397. Available
from: https://doi.org/10.1016/j.agwat.2009.03.022 How to cite this article: Amiri, E., Irmak, S. &
Raes, D., Steduto, P., Hsiao, T.C. & Fereres, E. (2009) AquaCrop— Tari, D.B. (2024) Comparison of the AquaCrop and
the FAO crop model to simulate yield response to water: CERES-Maize models for simulating maize
II. Main algorithms and software description. Agronomy Jour- phenology, grain yield, evapotranspiration and
nal, 101(3), 438–447. Available from: https://doi.org/10.2134/
water productivity under different irrigation and
agronj2008.0140s
Rudnick, D. & Irmak, S. (2014) Impact of nitrogen fertilizer on maize
nitrogen levels. Irrigation and Drainage, 1–17.
evapotranspiration crop coefficients under fully-irrigated, limited Available from: https://doi.org/10.1002/ird.2917
irrigation, and rainfed settings. Journal of Irrigation and

You might also like