Evaluating AquaCrop model for simulating production of amaranthus (Amaranthus cruentus) a leafy vegetable, under irrigation and rainfed conditions

Agricultural and Forest Meteorology 247 (2017) 300–310

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology journal homepage: www.elsevier.com/locate/agrformet

Research Paper

Evaluating AquaCrop model for simulating production of amaranthus (Amaranthus cruentus) a leafy vegetable, under irrigation and rainfed conditions

MARK

⁎

Z.A. Bello , S. Walker Department of Soil, Crop and Climate Sciences, University of the Free State, Bloemfontein 9300, South Africa

A R T I C L E I N F O

A B S T R A C T

Keywords: AquaCrop Calibration Crop modelling Amaranthus Calibration Validation Leafy vegetable

Amaranthus (Amaranthus spp.), a leafy vegetable in South Africa, has the potential to be cultivated as a crop, but is rarely cultivated because it easily grows naturally on any waste land. The crop tolerates adverse environmental conditions, but performs better with application of water and soil organic or inorganic fertilizers. The AquaCrop crop model was calibrated and validated for amaranthus under irrigation and rainfed conditions for this study. Field experiments were carried out during the 2008–09 and 2009–10 seasons under line source sprinkler system while pot experiment was carried out during the 2010–11 season. The pot and field data sets were used for parameterisation, calibration and validation of the model. The model was adequately calibrated for biomass and cumulative evapotranspiration (ET) for amaranthus under irrigation and rainfed conditions. However, pooled data across irrigation and rainfed conditions showed canopy cover (CC) was moderately simulated (root-mean-square error (RMSE) = 20.8%; model efficiency (ME) = 0.11; R2 = 0.577; d index of agreement (d) = 0.746; mean absolute percentage error (MAPE) = 43.4%). During validation, the model was able to adequately predict biomass and cumulative evapotranspiration (ET) for amaranthus for pooled data of irrigation (Full irrigation = W5 & Moderate irrigation = W3) and rainfed (W1) with RMSE of 1.96 t ha−1 and 75.64 mm, ME of 0.89 and 0.76, R2 of 0.92 and 0.91, d index of agreement of 0.91 and 0.91 and MAPE of 24.1 and 37.6% respectively. The prediction of soil water content by the model was moderate (RMSE = 50.62 mm; ME = 0.19; R2 = 0.30; d = 0.67; MAPE = 40.09) and needs improvement. It is recommended that datasets from other agro-ecological regions be used to improve calibration and validation for this crop.

1. Introduction In order to bridge the gap between the increasing world population and low food production, there is need to find a sustainable and secure way of increasing food production. One of the ways of achieving this is through diversification away from over-reliance on staple crops such as maize, wheat and rice (Mayes et al., 2011). Staple crops have lost their financial benefit due to static producer prices and continuous increase in input costs (Allemann, 2004). Future challenges facing staple crop production include among others, climate change, water shortage and availability of adequate land. Therefore, there is a need for efficient crop production systems and alternative crops. Other crops are needed as an alternative to established staple crops in terms of nutritional and financial values (Allemann, 2004). Many different types of crops are consumed by humans, but their importance and potential has not been well exploited. Alternative crops have the advantages of contributing to food security and alleviating poverty by providing a means of income

⁎

Corresponding author. E-mail addresses: [email protected], [email protected] (Z.A. Bello).

http://dx.doi.org/10.1016/j.agrformet.2017.08.003 Received 15 February 2017; Received in revised form 19 July 2017; Accepted 4 August 2017 Available online 23 August 2017 0168-1923/ © 2017 Elsevier B.V. All rights reserved.

generation. Apart from creating new markets, another important advantage of alternative crops includes sustainable production with low inputs (Anon, 1996; Bavec and Bavec 2006; Anon, 2009). Many of alternative crops have the ability to adapt to a wide range of adverse environmental conditions such as drought, high temperature and soil with low nutrient status (van Wyk, 2011). Amaranthus (Amaranthus spp.) is one of the underutilised crops out of the large rich plant biodiversity of South Africa (Cunningham et al., 1992; Jansen van Rensburg et al., 2007) which can be cultivated as an alternative crop. As an underutilized crop, it can serves as a source of livelihoods, stabilization of ecosystems and creating new markets (Anon, 2009) without displacing established staple crops (Allemann et al., 1996). Amaranthus is popular and consumed on a wide scale in South Africa due to its nutritional values and high palatability. Mostly, amaranthus is rarely cultivated because it grows easily naturally on any waste land and or during the first rains of summer (Jansen van Rensburg et al., 2007). Amaranthus becomes scarce after the first rains

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

due to the intensive harvest by the communities. The fact that there is little or no awareness of its ease of cultivation does not help the availability of this crop. Meanwhile, cultivation of amaranthus requires few input and less labour (Anon, 2009). In addition, amaranthus has high nutritional values (Cunningham et al., 1992; Allemann et al., 1996; Faber et al., 2010). The leaves of amaranthus have high protein, vitamin and mineral content (Makus and Davis, 1984). This is seen as a potential of the crop to serve as a source of vitamin A for nutritionally vulnerable communities in South Africa (Allemann et al., 1996; Faber et al., 2010). Amaranthus is an annual C4 plant that grows optimally under warm conditions (van den Heever and Coertze, 1996; Maboko, 1999: Schippers, 2000). Amaranthus is considered a promising crop for semi-arid regions, because of the ability to adapt to adverse environmental conditions (Cunningham et al., 1992; Allemann et al., 1996; Grubben, 2004; Maundu and Grubben, 2004). It can grow on a wide range of soils and can tolerate soil pH from 4.5 to 8.0 (Palada and Chang, 2003). The ability to tolerate salinity stress helps the plant to survive in semi-arid regions or on lands prone to high soil salinity (Omami, 2005). Irrespective of the ability of the crop to survive under adverse conditions, application of water and soil organic or inorganic fertilizer will increase fresh and dry mass production (Akparobi, 2009). Due to drought adaptation attributes, amaranthus is suitable for cultivation under South African climatic conditions. South Africa is a water scarce country with annual precipitation of around 500–600 mm (Nieuwoudt et al., 2004). To achieve high crop productivity in this water scarce region, understanding good water management with respect to types of crops to be cultivated, irrigation management and environmental sustainability are very important. These will help to develop good strategies to promote efficient water use in semi-arid regions. Crop modelling is another means of promoting efficient water use in the region. However, there is little literature information on crop modelling of alternative crops such as amaranthus. Walker et al. (2013) also observed that few of the currently available crop models have been calibrated for alternative crops. AquaCrop is a water driven crop model developed by the FAO for simulating crop yield responses to water (Raes et al., 2009; Steduto et al., 2009). It segregates the crop responses to water stress into four separate components, namely, canopy growth, canopy senescence, transpiration and harvest index (Steduto et al., 2009). The whole concept and underlying principles of AquaCrop model is described in Steduto et al. (2009). This model was developed to help agronomists, consultants, irrigation engineers, and farm managers to increase crop water productivity under rainfed and irrigated conditions (Raes et al., 2009). Under water limiting conditions, AquaCrop can simulate water requirements, water use efficiency and crop productivity. Apart from being easy to operate when compared to other models, it also requires a limited set of input parameters for predictions. Calibration and validation of the model for cereals and root staple crops has been reported, but not for leafy vegetables. In view of this, the objective of this study was to calibrate and validate the AquaCrop crop model for amaranthus, a leafy vegetable, under irrigation and rainfed conditions.

Fig. 1. (a) Daily mean temperature (Tavg), reference evapotranspiration (ETo) and (b) rainfall at the experimental site, Kenilworth, Bloemfontein for the two cropping seasons (2008/2009 and 2009/2010).

Experimental field of the same department (Department of Soil, Crop and Climate Sciences), Kenilworth, 20 km North West of the main campus, near Bloemfontein (latitude and longitude of 29.02°S, 26.15°E and altitude of 1354 m). Bloemfontein has a mean annual temperature of 15.9 °C, with an average maximum and minimum of 30.8 °C and 15.3 °C during January and 16.8 °C and −2.0 °C during July respectively. The mean annual rainfall is 559 mm and the maximum is received in February with 111 mm precipitation. The patterns of daily mean air temperatures at the experimental farm for the two seasons (2008/2009 & 2009/2010) were similar (Fig. 1a). The daily mean temperatures ranged between 15.2 °C and 28.6 °C for first season and ranged from 14.2 °C to 26.8 °C for the second season. Reference evapotranspiration (ETo) for the two seasons declined with months while the sharpest decline (1.1 mm) was found in January (Fig. 1a). The ETo for the month of December was higher in the 2009/2010 season than that of the 2008/2009 season. Rainfall distribution for the two seasons indicated that the 2009/2010 season was wetter (575 mm) than the 2008/2009 season (415 mm) (Fig. 1b).

2. Materials and methods 2.2. Experiments 2.1. Site descriptions and experimental procedures 2.2.1. Pot experiment The pot dimensions were 36 cm in diameter and 28 cm high with a volume of 28.5 L. Forty pots were filled with top soil from the experimental site where field trials were conducted. The soil was oven dried at 105 °C for 24 h to determine the initial water content of the soil. All the pots were filled and then saturated with water and left to drain, and weighed daily until a constant mass was observed and recorded. Differences between the dried soil mass and drained soil mass were taken as the water content at full water holding capacity. The pots were covered with quartz gravel to minimize soil surface evaporation and two pots were left bare to serve as reference for determining the

Two sets of experiments were performed for the study, namely pot and field experiments. The sets of experiments were used for calibration and validation of the AquaCrop crop model. Both sets of experiments were carried out with a genotype of amaranthus (Amaranthus cruentus ex Arusha) provided by the Agricultural Research Council Vegetable and Ornamental Plant Institute (ARC-VOPI, Roodeplaat). The pot experiment was carried out in a glasshouse facility of the Department of Soil, Crop and Climate Sciences, main campus, University of the Free State, Bloemfontein (latitude and longitude of 29.11°S, 26.19°E, and altitude of 1395 m). The field experiment was conducted at the 301

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

evaporation rate. It was observed that the evaporation was very small compared to plant transpiration. Therefore, evaporation was considered negligible and all water loss considered as transpiration. Seeds were sown in trays to raise seedlings and one seedling per pot was transplanted after three weeks to the pots when the seedlings were between 4 and 5 cm high. Transplanting was done during the late afternoon to avoid transplant shock. Irrigation was done immediately after transplanting to enhance good root and soil contact. The temperature of the greenhouse was regulated to 30 °C during the day and 16 °C at night. Management practices for the pot experiment were kept at optimum levels including fertilizer application, which was according to the recommendation for the field trial (van den Heever and Coertze, 1996). Weeds were manually controlled only at the reference pots since the other pot surfaces were covered with quartz gravel to minimize soil surface evaporation.

Table 2 Soil profile characteristics for the Bainsvlei soil as described by Chimungu (2009). Soil

Soil water content at

layers

Description

Thickness m

PWP mm

FC

SAT

TAW mm/m

Ksat mm/day

τ

1 2 3

loamy sand sandy loam sandy clay loam sand sandy loam

0.25 0.45 0.50

83.0 83.0 83.0

228.0 243.0 268.0

380.0 410.0 470.0

145.0 160.0 185.0

800.0 500.0 125.0

0.90 0.76 0.47

0.25 0.40

83.0 83.0

282.0 282.0

360.0 410.0

199.0 199.0

1500.0 500.0

1.00 0.76

4 5

PWP, Permanent wilting point; FC, field capacity; SAT, saturation; TAW, total available soil water; Ksat, saturated hydraulic conductivity; τ, drainage coefficient.

2.4. Measurements 2.2.2. Field experiment Amaranthus was cultivated in the field under a line-source sprinkler system as described by Hanks et al. (1980). The plots were laid out as a split-plot design with four replications. Treatments included five waterapplication levels, from fully irrigated plots closest to the line source (W5) to rainfed plots (W1) furthest from the irrigation line. Full irrigation (W5), moderate irrigation (W3) and rainfed (W1) were considered for this study. Rain gauges were used to quantify the amount of irrigation water according to the distance from the line-source sprinkler and to measure amount of rainfall received. This enabled quantification of the water availability per treatment. The rainfed plots were twice the size of the irrigated plots to avoid any border or lateral movement effect of water. Irrigation was applied during windless conditions, mostly at night. The total size of the plot was 23 m × 36 m while the plot for each treatment was 11 m × 3 m. Seedlings of the crop were raised in the glasshouse before transplanting into the field at the height between 4 and 5 cm. The recommended planting date for amaranthus in the Bloemfontein area is between October and November (van den Heever and Coertze, 1996). A day before transplanting, a 2:3:4 (30) NPK fertilizer was broadcasted at a rate of 20 kg N ha−1, 30 kg P ha−1and 40 kg K ha−1. Transplanting took place on 30 December for the 2008/ 2009 season while it was done on 11 November for the 2009/2010 season. Transplanting was delayed in the first season because of difficulty of obtaining the seed, which caused the delay in raising seedlings. Transplanting was done at the spacing of 100 cm between rows and 30 cm within row. Plants were monitored and irrigated until establishment was achieved four days after transplanting. Topdressing with 50 kg of limestone ammonium nitrate (LAN) was done 45 days after transplanting. Weeds were controlled manually when required.

2.4.1. Pot experiment Data collected from the pot experiment included weekly total aboveground biomass from 4 plants per treatments, daily transpiration and stomatal conductance of the crop under well watered and stress conditions. Harvested plants were oven dried at 65 °C for 36–48 h to determine dry mass every week. Pots were divided into two groups of 10 pots per treatment; well-watered and water-stressed. Well-watered pots were weighed three times a week and then water added to refill to full water holding capacity as the difference in mass. Following the weighing, the stressed pots were not rewatered to full water holding capacity until the water depletion was below 30%. All pots were weighed irrespective of the treatment at around 17:00 h and then rearranged within the treatments after weighing to maintain random distribution in the greenhouse. Any difference in the mass of the pots over a 2 day interval was taken as water uptake of the plants, which represents the water transpired (Eq. (1)). The water uptake was converted from mass to volume in reference to the surface area of the pot. Transpired water, T (mm) = PWn − PWf

where PWn is the initial mass of the pot on a given date and PWf is the mass of the pot at the end of the interval. Plant water status was monitored through the measurement of stomatal conductance. Stomatal conductance of the crop was measured with a leaf porometer (Decagon Devices, Inc.). The measurements took place twice a week around midday between 12:00 and 14:00 h under cloudless conditions. Four fully expanded and fully exposed leaves per treatment were sampled for these measurements. All leaves were sampled randomly and at the same upper level on the stem of the plant. 2.4.2. Field experiments Data collected from the field studies included leaf area, aboveground biomass and radiation interception at weekly intervals under irrigated and rainfed treatments. Leaf area was measured with the aid of a leaf area meter (LI3000; LI-COR, Lincoln, NE, USA) while radiation interception was measured with the aid of line quantum sensor (LI191R; LI-COR, Lincoln, NE, USA). However, leaf area was only measured in the 2009–10 season due to logistic reasons. Leaf area was converted to LAI using Eq. (2);

2.3. Experimental data The pot experiment datasets were used for parameterization and calibration (2010/2011 season) of the model while the field experiment datasets were used for calibration (2009/2010 season) and validation (2008/2009 season) of the model (Table 1). Table 1 Summary of source of datasets for calibration and validation of AquaCrop model. Experiment

Pot experiment

Field trial

Purpose for datasets Season Irrigation Water stress Rainfed Transplanting date

Parameterization 2010/2011 X X

Calibration 2009/2010 X

Validation 2008/2009 X

X 2009/11/11

X 2008/12/30

2010/12/11

(1)

LAI = leaf area × plant population

(2)

Phenological development was monitored during the 2008–09 and 2009–10 seasons. Developmental stages of amaranthus were classified according to the Biologische Bundesanstalt, Bundessortenamt and Chemical Industry (BBCH) system of coding phenological growth stages of plants. The BBCH scale is a system designed for uniform coding of stages of development of mono- and dicotyledonous plant species, which are phenologically similar (Meier et al., 1993). Harvesting of amaranthus plants started 14 days after transplanting, 4 plants per treatment, to

X indicates datasets used.

302

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Table 3 Selected crop parameters and values for calibration and validation of AquaCrop for amaranthus. Crop parameters Type of Crop Carbon cycle T base Tupper Method of planting Initial cover CCo Plant density Time to CCx CCx CGC CDC Time for decline Time to recover Time to Zr (max) Time to senescence Time to harvest Zr (max) Zr (min) Expansion Kctr Aging Green canopy cover WP* HIo Soil water stresses Canopy expansion Water stress coefficient upper threshold Water stress coefficient lower threshold Stomatal closure Water stress coefficient upper threshold Early canopy senescence Water stress coefficient upper threshold

Descriptions

Input

Base temperature (°C) Upper temperature (°C) Sowing/Transplanting Cover size transplanted seedling (cm2 plant−1) Initial canopy cover (%) Plants ha−1 Planting to CCx (day) Maximum canopy cover (%) Canopy growth coefficient (% day−1) Canopy decline coefficient (% day−1) Canopy decline (day) transplants recovery (day) from plant to max rooting depth (days) from plant to start senescence (days) from plant to maturity/harvest (days) Max effective rooting depth (m) Min. rooting depth (m) Avg.root zone expansion (cm/day−1) Coefficient for transpiration Reduction with age (% day−1) Effect of canopy in late season (%) Water productivity (ton ha−1) Reference harvest index (%)

leafy vegetable C4 7 30 Transplanting 20 67 33 333 55 95 14.7 8.0 37 4 60 90 100 1.75 0.3 2.7 0.8 0.15 60 28 85 Moderately tolerant to water stress 0.25 0.6 6 Moderately tolerant to water stress 0.65 6 Moderately tolerant to water stress 0.65 5 Very sensitive to water logging

Ks p (upper) Ks p (lower) shape factor Ks p (upper) shape factor Ks p (upper) shape factor

Aeration stress

determine the fresh and dry mass. This was continued every week until senescence set in. Plants were oven dried at 65 °C for 36–48 h to determine dry mass every week. Total aboveground biomass at weekly intervals was regarded as the yield considering amaranthus as a leafy vegetable crop. Soil-water content was monitored weekly in field trials during both seasons with the aid of neutron moisture meter (Model 503DR; Campbell Pacific Nuclear (CPN), Concord, CA, USA) in the 1.8 m soil profile. The soil water measurement was from each treatment plot at 0.3 m intervals down to 1.8 m depth. The ET was estimated by using the water-balance equation (Eq. (3));

evaporative demand was estimated from the minimum and maximum air temperatures and ETo. AquaCrop requires the temperatures for calculation of growing degree-days (GDD), which influences crop growth and phenological development (Raes et al., 2009).

ET = P + I − ΔSW − D − R

⎛ 1 − e− 1.3 CC = ⎜ ⎜ 1 + e− LAI 1.3 ⎝

2.5.2. Crop data To create a crop file, the observations of crop development and phenology from the field were used. AquaCrop identifies crop canopy development as canopy cover (CC). With Eq. (4), field measured leaf area index (LAI) was converted to CC (Garcia-Vila et al., 2009):

( (

(3)

where P is precipitation (mm), I is irrigation (mm), D is deep percolation (mm), R is runoff (mm), ΔSW is the change in SWC (mm). Deep percolation (D) and runoff (R) were assumed as zero since the amount of irrigation was controlled.

LAI

) ⎞⎟ × 100 ) ⎟⎠

(4)

An extinction coefficient of 0.74 was used to derive an equation from a general relationship between intercepted solar radiation at midday and LAI. Extinction coefficient of the crop was derived from Bello (2013) unpublished data. It was not possible to measure some parameters, therefore, model default values were used for these parameters. Information from the literature was used for root depth for the crop file in the absence of observation of root development (Place et al., 2008; Johnson and Henderson, 2002). Information on soil characteristics of the experimental site from previous studies (Chimungu, 2009) was used for creating a soil-profile characteristics file. Soil type of the whole profile coupled with the physical characteristics such as soil-water content (SWC) at saturation, field capacity, permanent wilting point, and saturated hydraulic conductivity (Ksat) formed part of the information for this study (Table 2). The model generated total available soil water

2.5. Model parameters and input data 2.5.1. Climatic data A climate file was created with the daily weather data from the Automatic Weather station (AWS) at the study site, Kenilworth Experimental Site courtesy of Agricultural Research Council- Institute of Soil, Climate and water (ARC-ISCW). Daily weather data relevant for the climate file included the minimum and maximum air temperatures, rainfall amount, wind speed, maximum and minimum relative humidity, and solar radiation. Reference evapotranspiration (ETo) was calculated by using the FAO ETo calculator (http://www.fao.org/nr/ water/eto.html) as recommended in the AquaCrop model. Atmospheric 303

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Fig. 2. Comparison of simulated (Sim) and observed (Obs) canopy cover (CC) under irrigation (W5 & W3) and rainfed (W1) treatments during the 2009/2010 season used for calibration of the AquaCrop model for amaranthus. (a) Full irrigation, W5 (b) Moderate irrigation, W3 (c) Rainfed, W1 (d) Pooled data of across irrigation and rainfed conditions.

2.6. Model calibration and validation

Table 4 Statistic indices to evaluate calibration and validation of AquaCrop for amaranthus in terms of root mean squares (RMSE, RMSEs, RMSEu), model efficiency (ME), coefficient of determination (R2), index of agreement (d) and mean absolute percent error (MAPE) between simulated and observed values of canopy cover (CC), biomass production, soil water content (SWC) and cumulative evapotranspiration (ET).

AquaCrop version 3.1 (Raes et al., 2009; Steduto et al., 2009) was used for simulating amaranthus response to different water regimes. Calibration entails adjusting certain model parameters to make the model match the measured values at the specific location (Farahani et al., 2009). During the calibration, simulation periods were linked to the growing cycle of the crop, starting with the initial SWC measured in the field. Conservative parameters were selected as default value, which are generally applicable for a given species or specific cultivar. Default values were selected for some parameters that were not measured during the experimental studies. The CC per transplanted seedling, water-extraction pattern and average root-zone expansion were the parameters not measured during the experimental studies. Observations such as phenological stages of the crop from the field trials were used in parameterisation of the model. Calibration was performed with the field datasets for well-watered, moderate irrigation and rainfed conditions. This was carried out as was described by Steduto et al. (2009) and Raes et al. (2009) for the calibration of AquaCrop. Parameters such as canopy growth and canopy decline coefficients were generated by the model from the observed values. The observations from the field that were used as inputs for the crop development parameters were plant density, time to recover, maximum CC and time to harvest. Calibration of the model for the crop started with the green CC data from 2009 to 10 season, because leaf area was not measured during the previous season (2008–09). During the calibration process, the importance of the coefficient of transpiration (Kctr) was proven, because it is proportional to CC (Karunaratne et al., 2011). The CC of the well-watered treatments was calibrated before the rainfed CC, which was assumed to represent severe water-stress conditions. Simulations were run and the Kctr was reduced until a good fit was achieved for CC under the irrigated and

Calibration 2009–10 Parameters

RMSE

RMSEs

RMSEu

ME

R2

d

MAPE

Canopy cover CC (%) Biomass (t ha−1) Cumulative ET (mm)

20.82 1.87 34.11

10.79 1.04 22.42

17.80 1.55 25.73

0.11 0.80 0.97

0.12 0.90 0.96

0.75 0.96 0.99

43.39 20.78 38.50

Validation 2008–09 Biomass(t ha−1) SWC (mm) Cumulative ET (mm)

1.96 50.62 75.64

1.82 38.88 71.57

0.74 40.11 72.67

0.89 0.19 0.76

0.92 0.30 0.91

0.91 0.67 0.91

24.12 40.09 37.62

RMSEs: systematic root mean squares, RMSEu: unsystematic root mean squares.

from the field capacity and permanent wilting point values, and drainage coefficient was generated from the Ksat values.

2.5.3. Field management The AquaCrop irrigation file was created with the data of the actual amount of irrigation water applied and dates of irrigations during the field trials. The field management was set at optimum field conditions of non-limiting soil fertility, without surface mulches and no temperature stress. Stepwise, datasets from the irrigated plots were used for calibrating the crop for non-water-stress conditions, whereas datasets for the pot experiment under water-stress were used for parameterization of the model for water-stress conditions. 304

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Fig. 3. Comparison of simulated (Sim) and observed (Obs) biomass under irrigation (W5 & W3) and rainfed (W1) treatments during the 2009/2010 season used for calibration of the AquaCrop model for amaranthus. (a) Full irrigation, W5 (b) Moderate irrigation, W3 (c) Rainfed, W1 (d) Pooled data of across irrigation and rainfed conditions.

0.5

rainfed treatments. Biomass production was the next parameter for calibration. A conservative parameter, water productivity (WP), was initially derived from the pot experiment and adjusted with consecutive simulations to get a good fit for biomass production. Conservative parameters are the model parameters that do not change with time, management practices, and/or geographic locations once they are calibrated for a specific crop (Raes et al., 2009). WP is biomass per unit of cumulative transpiration, which tends to be constant for a given climatic condition (de Wit, 1958; Hanks, 1983; Tanner and Sinclair, 1983). The reference harvest index (HIo) was set at default of 85%, which was within the range for leafy vegetables. This was to determine yield as the product of HIo and biomass. Fine-tunings and adjustments of parameters were done until good matches between simulated and measured parameters were obtained. For the stress section, responses due to salinity, fertility and temperature were not considered during the calibration and validation of the model for amaranthus. Only biomass, soil water content and cumulative ET parameters were used to evaluate the performance of the model in terms of the validation.

n RMSE = ⎡n−1 ∑ (Si − Oi )2⎤ ⎢ ⎥ i=1 ⎣ ⎦

(5) 0.5

n 2 RMSEs = ⎡n−1 ∑ (Sˆi − Oi ) ⎤ ⎢ ⎥ i=1 ⎣ ⎦

(6)

0.5

n 2 RMSE u = ⎡n−1 ∑ (Si − Sˆi) ⎤ ⎢ ⎥ i=1 ⎣ ⎦

n

d=1−

∑i =1 n

∑i =1

MAPE =

∑i =1 n ∑i =1

100 n

(Si − Oi )2

( Si − O + Oi − O )2

n

ME = 1 −

n

∑ i=1

(7)

(8)

(Oi − Si)2 (Oi − O )2

(9)

Oi − Si Si

(10)

where n is the number of observations; Si and Oi are the simulated and observed values for the corresponding parameter; Ŝi is derived from Ŝi = a + bOi where a and b are the intercept and slope, respectively, of a least-squares regression between the simulated and observed values; Ō is the mean of the observed variable. The RMSE assumes the same unit as the parameter under observation. The model shows good performance when the value of RMSE is close to zero. The model goodnessof-fit increases as RMSEs and RMSEu values approach zero and RMSE values, respectively. The values of the index of agreement (d) range from 0 to 1. The closer the d value to 1, the better the agreement between the simulated and measured values. The ME is a measure of

2.7. Statistical analyses Coefficient of determination, R2; root-mean-square error (RMSE, Eq. (5)), together with its systematic (RMSEs, Eq. (6)) and unsystematic (RMSEu, Eq. (7)) components, index of agreement (d, Eq. (8)) (Willmott, 1982) and model efficiency (ME, Eq. (9)) were the four statistical methods used to evaluate goodness-of-fit for the calibration and validation of the model for each of the crop parameters, mean absolute percentage error (MAPE, Eq. (10)) was used to evaluate prediction accuracy: 305

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Fig. 4. Comparison of simulated (Sim) and observed (Obs) cumulative evapotranspiration (ET) under irrigation (W5 & W3) and rainfed (W1) treatments during the 2009/2010 season used for calibration of AquaCrop model for amaranthus. (a) Full irrigation, W5 (b) Moderate irrigation, W3 (c) Rainfed, W1 (d) Pooled data of across irrigation and rainfed conditions.

for all the treatments after transplanting to help with the establishment of the crop. This might have helped the plants to develop larger CC than the simulated CC. Hence, the underestimation of CC by the model. At 30 days after transplanting, the simulated CC was very low in all the treatments, but was a good match by the end of the season in both irrigated treatments. Overall, there was a moderately good agreement between the simulated and observed CC with RMSE of 20.82%, RMSES of 1.79, RMSEU of 17.80, ME of 0.11, R2 of 0.577 and d of 0.746 (Table 4). The prediction of CC was about 55% accurate with the MAPE of 43.4%. Mabhaudhi et al. (2014) also reported constraint in simulating taro (Colocasia esculenta L. Schott) CC under rainfed conditions with AquaCrop. This was due to the continuous shedding canopy growth pattern of the crop. Biomass production of amaranthus was well simulated by AquaCrop for most of the treatments (Fig. 3). Contrary to CC simulations, the better fit was found in the W1 treatments. There was overestimation of biomass production in the W5 treatment. It was suspected this might be due to the effect of nutrient leaching in the field situation due to large amount of water applied as the nutrient stress was not considered during the calibration of the model. During the two seasons of the study, least irrigation treatments (W2, not part of the chosen treatments for this study) produced the highest yield compared to other irrigation treatments (Bello, 2013). Previous studies also pointed out that to prevent nutrient leaching and yield reduction in amaranthus, over irrigation should be avoided (Palada and Chang, 2003; Neluheni et al., 2007). The RMSE (1.87 t ha−1), RMSES (1.04), RMSEU (1.55), ME (0.80), R2 (0.90) and d (0.96) also showed the good overall performance of the AquaCrop to simulate biomass production of amaranthus (Table 4). The prediction of biomass was with high accuracy of MAPE of 20.78%. Low RMSES and RMSEU values that approached RMSE and

robustness of a model where the closer the ME is to 1, the more robust the model. 3. Results and discussion 3.1. Crop parameters and values Table 3 presents the crop parameters and values resulting from the calibration of the model for amaranthus using the 2009–10 season data. Crop parameters that depend on management include plant density (33,333 plants ha−1), time to recover after transplanting (4 days) and maximum CC (95%), while conservative parameters include water productivity normalized for ETo and CO2, WP*, (28 t ha−1) and crop transpiration coefficient, Kctr, (0.8). The effect of soil water stress on canopy expansion, stomatal closure and early canopy senescence was set at moderately tolerant to stress for calibrating the model for this crop. 3.2. Model calibration During the calibration process, there was a good match between the observed and the simulated CC of well-watered treatment (W5) and moderate irrigation treatment (W3) (Fig. 2). However, CC of plants from the rainfed plots (W1) was under estimated by the model throughout the growing season. Reasons for this may be due to the small value of the initial cover size of transplanted seedling (CCo = 0.67%) generated by the model, which posed a major concern during the calibration process. In terms of CC, the model might have made provision for transplanting shock since plants tend to grow slowly because of recovery time after transplanting. Also, there was irrigation 306

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Fig. 5. Validation results and comparison of simulated (Sim) versus observed (Obs) amaranthus biomass under irrigation (W5 & W3) and rainfed (W1) treatments during the 2008/2009 season. (a) Full irrigation, W5 (b) Moderate irrigation, W3 (c) Rainfed, W1 (d) Pooled data of across irrigation and rainfed conditions.

until the end of the season in all the treatments, but the observed and simulated trends were similar (Fig. 6). Although, the initial observed and simulated SWC was the same, there was overestimation in W5 and W3 from 30 days after transplanting till the end of the season. The model underestimated SWC up to 37 days after transplanting and thereafter, there was overestimation of SWC till the end of the season. The discrepancies between the simulated and observed may be due to the fact that information from the literature was used to calibrate the model for effective root depth, because no data was collected during the field studies for this parameter for amaranthus. The values of RMSE, RMSES, RMSEU, ME, R2, d index of agreement and MAPE for the performance of AquaCrop during the validation of the model for SWC were 50.616 mm, 38.88, 40.11, 0.19, 0.30, 0.67 and 40.09% respectively (Table 4). The low value of ME suggested that the model was not robust in simulating SWC. However, the model performed moderately well in simulating cumulative ET (Fig. 7). The model over predicted at the earliest stage and under predicted cumulative ET at the later stage of crop growth. Out of all the treatments, the best agreement between the simulated and observed cumulative ET was found in the rainfed plots (W1). AquaCrop performed well in simulating cumulative ET with RMSE of 75.64 mm, RMSES of 71.57, RMSEU of 72.67, ME of 0.76, R2 of 0.91, d index of agreement of 0.91and MAPE of 37.62% (Table 4). Table 5 compares results between observed and simulated biomass, SWC and cumulative ET by the model during the validation process. Biomass and cumulative ET were underestimated while SWC was overestimated. The least underestimation was found in the rainfed treatment for both biomass and cumulative ET with the values of 6.35 and 15.09% respectively. Meanwhile, there was biomass underestimation of 70% by the AquaCrop for W3 treatment. The difference between the biomass of

high d index of agreement confirmed the good agreement between the simulated and observed biomass. In terms of cumulative ET, simulation was good in all the treatments (Fig. 4). There was RMSE of 34.11 mm, RMSES of 22.42, RMSEU of 25.73, ME of 0.97, R2 of 0.963, d of 0.989 and MAPE of 38.50% to prove the good performance of the model in simulating seasonal cumulative ET for amaranthus. In AquaCrop, cumulative ET is partitioned into evaporation and transpiration, which are at their maximum rate when the soil is wet. Therefore, evapotranspiration is highly dependent on evaporative demand of the atmosphere, crop development and independent of the soil water content (Raes et al., 2009). In a study by Katerji et al. (2013), AquaCrop underestimated daily ET for maize since actual ET reflects transpiration losses other than evaporation and simulated ET corresponds mainly to soil evaporation. Therefore, the model overestimated water stress effect on ET, which resulted in low simulated daily ET. Apart from simulation of CC, with the values of ME, the model was robust in simulating biomass and cumulative ET for amaranthus. Bello and Walker (2016) also reported that AquaCrop model is more robust in simulating ET than CC for pearl millet (Pennisetum glaucum).

3.3. Model validation AquaCrop was able to accurately simulate biomass production for the well-watered (W5) and rainfed (W1) treatments (Fig. 5). There was an underestimation of biomass produced at the end of the season for the W3 treatment. On average, the trend of biomass production with time was well predicted and this was supported statistically by RMSE of 1.96 t ha−1, RMSES of 1.82, RMSEU of 0.74, ME of 0.89, R2 of 0.916, d index of agreement of 0.91 and MAPE of 24.12% (Table 4). AquaCrop overestimated soil water content around 40 days after transplanting 307

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Fig. 6. Validation results and comparison of simulated (Sim) and observed (Obs) soil water content (SWC) in amaranthus plots under irrigation (W5 & W3) and rainfed (W1) treatments of the 2008/ 2009 season. (a) Full irrigation, W5 (b) Moderate irrigation, W3 (c) Rainfed, W1 (d) Pooled data of across irrigation and rainfed conditions.

W3 and W1 treatments was close to 5 t ha−1. This could mean that the irrigation at this level of the treatment (W3) might be close to the optimum since full irrigation is not directly beneficial to the crop. During the field studies, the least and moderately irrigated plots produced higher fresh and dry mass of amarathus for both seasons compared to the other treatments (Bello, 2013). The overestimation of SWC by the model was between 15.82 and 19.08%. Generally, calibration and validation of AquaCrop for amaranthus was satisfactory, although the R2 of simulated versus observed SWC for the validation was low with moderate d index of agreement. Simulation of SWC has been identified as one of the limitations of AquaCrop model (Bello and Walker, 2016). The tendency of the model to over predict SWC, similar to that found during the validation process for amaranthus, was reported by Farahani et al. (2009) and Hussein et al. (2011). They reported that the model was able to give a good prediction of the trend of SWC with time due to irrigation events with absolute values deviating from measured values in cotton field experiments (Hussein et al., 2011). Their reports also support that results of AquaCrop simulation of cumulative ET were very good irrespective of the outcome of the simulation of SWC. There were reports that AquaCrop also overestimated SWC for other crops such as maize and winter wheat (Zhang et al., 2013; Ahmadi et al., 2015). Irrespective of the performance of the model in simulating SWC, good performance of AquaCrop to predict biomass and yield of different crops has been reported for different crops. It should also be noted that since yield is reported in terms of biomass for leafy vegetables, biomass should be considered as yield of amaranthus. Studies have proved that AquaCrop performs well in simulating yields of crops, whether in grains or biomass. Geerts et al. (2009) reported calibration and validation of grain quinoa, a crop similar to amaranthus, with good agreement between the simulated and

observed values of CC and biomass in different agro-climatic regions under different management conditions. They reported that simulated versus observed biomass from eight quinoa fields used for calibration provided R2 of 0.91 while simulated versus observed values from 14 fields used for validation of the model for the crop provided R2 of 0.88. These are comparable to the values of R2 of 0.900 and 0.916 achieved during the calibration and validation of the model for amaranthus for this study near Bloemfontein. AquaCrop is good in predicting cumulative ET. Studies that reported good performance of AquaCrop in predicting ET include Farahani et al. (2009), Garcia-Vila et al. (2009) and Hussein et al. (2011) for cotton, Greaves and Wang (2016) for maize, Katerji et al. (2013) for tomato, Bello and Walker (2016) for pearl millet and Montoya et al. (2016) for potato. In general for abovementioned studies as well as the present study, AquaCrop underestimated ET for the crops even though there was good fit of agreement. It is therefore possible for the model to overestimate crop productivity parameters such as water use efficiency (WUE) since ET is part of the components of determination. Farahani et al. (2009), Garcia-Vila et al. (2009), Hussein et al. (2011) and Greaves and Wang (2016) reported overestimation of WUE due to underestimation of ET by the AquaCrop model.

4. Conclusions This study calibrated and validated AquaCrop for a leafy vegetable, amaranthus, the first for a leafy vegetable crop type. The AquaCrop model was able to adequately simulate canopy cover (CC) under irrigation, but with some challenges under rainfed conditions. Biomass production and cumulative evapotranspiration (ET) for amaranthus under irrigation and rainfed conditions were well simulated by the 308

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

Fig. 7. Validation results and comparison of simulated (Sim) and observed (Obs) amaranthus cumulative evapotranspiration (ET) under irrigation (W5 & W3) and rainfed (W1) treatments of the 2008/ 2009 season. (a) Full irrigation, W5 (b) Moderate irrigation, W3 (c) Rainfed, W1 (d) Pooled data of across irrigation and rainfed conditions.

Table 5 Comparison between observed (Obs) and simulated (Sim) validation results for biomass, soil water content (SWC) and cumulative ET for amaranthus under irrigations (W5, W3) and rainfed (W1) treatments in the 2008–10 season. Treatments

Biomass

SWC −1

Obs (t ha W5 W3 W1

8.90 12.74 7.88

)

Sim (t ha 7.47 7.47 7.41

−1

)

Cumulative ET

Deviation (%)

Obs (mm)

Sim (mm)

Deviation (%)

Obs (mm)

Sim (mm)

Deviation (%)

−19.15 −70.53 −6.35

338.48 317.96 259.35

418.30 377.70 319.60

19.08 15.82 18.85

437.22 372.58 307.64

285.40 282.50 267.30

−53.20 −31.89 −15.09

Acknowledgments

model. Due to the moderate performance of the model in simulating soil water content (SWC), it was therefore highlighted that more work is needed to be done to improve the water balance part of the model in terms of calibration and validation process. This can be done using the actual values from the field for all the parameters concerning rooting depths. The model code should be improved to accommodate better initial canopy cover of transplanted seedlings. In order to improve the model based on these challenges, there is a need to use datasets from other agro-ecological regions to improve the calibration and validation for amaranthus. Irrespective, the study showed that the model could predict the performance of underutilised crops under different agronomic conditions. Moreover, the model was able to adequately simulate growth and yield of this crop with limited number of inputs. The model has the potential to be used as a decision support tool to increase water productivity and to study different scenarios and management conditions of amaranthus cultivations.

The authors thank the Water Research Commission of South Africa (WRC) for funding this study under WRC K5/1771 Water-Use of Drought Tolerant Crops; Professor A. T. Modi, Dr Y. G. Beletse, Dr T. Mabhaudhi, Mr Cinisani Tfwala, Mr. Charles Tharaga and other members of the project team for their criticism and contributions; and all of the staff of the Experimental Site and Department of Soil, Crop and Climate Sciences, University of the Free State. The authors would like to thank Mrs Liesl van Westhuizen for proof reading and editing of the manuscript. References Ahmadi, S.H., Mosallaeepour, E., Kamgar-Haghighi, A.A., Sepaskhah, A.R., 2015. Modeling maize yield and soil Water content with AquaCrop under full and deficit irrigation managements. Water Resources Manage. 29, 2837–2853. Akparobi, S.O., 2009. Effect of farmyard manures on the growth and yield of Amaranthus cruentus. Agric. Trop. Subtr. 42, 1–4. Allemann, J., 2004. In: Status of Underutilised Species in South Africa. South and Eastern African Network on Underutilized Crops (SEANUC). 8–9 March, 2004, Lusaka,

309

Agricultural and Forest Meteorology 247 (2017) 300–310

Z.A. Bello, S. Walker

water stress conditions in the Mediterranean region. Agric. Water Manage. 130, 14–26. Mabhaudhi, T., Modi, A.T., Beletse, Y.G., 2014. Parameterisation and evaluation of the FAO-AquaCrop model for a South African taro (Colocasia esculenta L. Schott) landrace. Agric. For. Meteorol. 192, 132–139. Maboko, S.M., 1999. Vegetable Amaranth Improvement for South Africa. www.newcrops. uq.edu.au/.newslett/ncn11169.htm (Accessed 15 September 2008). Makus, D.J., Davis, D.R., 1984. A mid-summer crop for fresh greens or canning: Vegetable amaranth. Arkansas Farm Res. 33, 10. Maundu, P.M., Grubben, G.J.H., 2004. Amaranthus graecizans L. In: Grubben, G.J.H. (Ed.), Amaranthus Cruentus L. PROTA, Wageningen, The Netherlands, pp. 76–78. Mayes, S., Massawe, F.J., Alderson, P.G., Roberts, J.A., Azam-Ali, S.N., Hermann, M., 2011. The potential for underutilized crops to improve security of food production. J. Exp. Bot. 63, 1075–1079. Meier, U., Bachman, L., Buhtz, H., Hack, H., Klose, R., Marlander, B., Weber, E., 1993. Phenological stages of the beta-beet (Beta vulgaris L. ssp.). Coding and the BBCH-scale (with pictures). Nachrichtenbl. Deut. Pflanzenschutzd. 45, 37–41. Montoya, F., Camargo, D., Ortega, J.F., Córcoles, J.I., Domínguez, A., 2016. Evaluation of Aquacrop model for a potato crop under different irrigation conditions. Agric. Water Manage. 164, 267–280. Neluheni, K., Du Plooy, C.P., Mayaba, N., 2007. Yield response of leafy amaranths to different irrigation regimes. In: Proceedings of the African Crop Science Conference. El-Minia, Egypt, 27–31 October, 8. pp. 1619–1623. Nieuwoudt, W.L., Backeberg, G.R., du Plessis, H.M., 2004. The value of water in the South African economy: some implications. Agrekon 43, 1–22. Omami, E.N., 2005. Response of Amaranth to Salinity Stress. Ph.D. Thesis in Horticulture. University of Pretoria, Pretoria, South Africa. Palada, M.C., Chang, L.C., 2003. Suggested Cultural Practices for Vegetable Amaranth. Asian Vegetable Research & Development Center, AVRDCpp. 3–552. Place, G., Bowman, D., Burton, M., Rufty, T., 2008. Root penetration through a high bulk density soil layer: differential response of a crop and weed species. Plant Soil 307, 179–190. Raes, D., Steduto, P., Hsiao, T.C., Ferreres, E., 2009. AquaCrop – the FAO crop model for predicting yield response to water: II. Main algorithms and software description. Agron. J. 101, 438–447. Schippers, R.R., 2000. African Indigenous Vegetables. An Overview of the Cultivated Species. Natural Resources Institute/ACP – EU Technical centre for agricultural and rural cooperation, Chatham, UK (pp. 214). Steduto, P., Hsiao, T.C., Raes, D., Fereres, E., 2009. AquaCrop – The FAO crop model to simulate yield response to water: I. Concepts and underlying principles. Agron. J. 101, 426–437. Tanner, C.B., Sinclair, C.R., 1983. Efficient water use in crop production research or research? In: Taylor, H.M., Jordan, W.R., Sinclair, T.R. (Eds.), Limitations to Efficient Water Use in Crop Production. ASA, CSSA, SSSA, Madison, WI, USA, pp. 1–27. van den Heever, E., Coertze, A.F., 1996. Indigenous Leaf Crops A1–Amaranthus (marog). Agricultural Research Council Leaflet, ARC –VOPI, Pretoria, South Africa (pp. 2). van Wyk, B.E., 2011. The potential of South African plants in the development of new food and beverage products. South Afr. J. Bot. 77, 857–868. Walker, S., Bello, Z.A., Beletse, Y.G., Mabhaudhi, T., Modi, A.T., 2013. Calibration of AquaCrop model to predict water requirements of traditional African vegetables. Acta Hortic. 1007, 943–950. Willmott, C.J., 1982. Some components on the evaluation of model performance. Bull. Am. Meteorol. Soc. 63, 1309–1313. Zhang, W., Liu, W., Xue, Q., Chen, J., Han, X., 2013. Evaluation of the AquaCrop model for simulating yield response of winter wheat to water on the southern Loess Plateau of China. Water Sci. Technol. 68, 821–828.

Zambia. . www.underutilized-species.org/Documents/.../status_ups_sa.ppt (Accessed 16 February 2010). Allemann, J., van den Heever, E., Viljoen, J., 1996. Evaluation of Amaranthus as a possible vegetable crop. App. Plant Sci. 10, 1–4. Anon, 1996. Report on the State of the World’s Plant Genetic Resources for Food and Agriculture. FAO, Rome, Italy pp. 511. Anonymous, 2009. Global Facilitation Unit for Underutilized Species. International Centre for Underutilized Crops. http://www.underutilized-species.org/ (Accessed 20 February 2010). Bavec, F., Bavec, M., 2006. Organic Production and Use of Alternative Crops. Taylor & Francis: CRC Press, New York, U.S.A. Bello, Z.A., 2013. Characterization and Modelling of Water Relations by Amaranthus and Pearl Millet. PhD Thesis. University of the Free State, Bloemfontein, South Africa. Bello, Z.A., Walker, S., 2016. Calibration and validation of AquaCrop for pearl millet (Pennisetum glaucum). Crop Pasture Sci. 67, 948–960. Chimungu, J.G., 2009. Comparison of Field and Laboratory Measured Hydraulic Properties of Selected Diagnostic Soil Horizons. M.Sc. Thesis in Soil Science. University of the Free State, Bloemfontein, South Africa. Cunningham, A.B., de Jager, P.J., Hansen, L.C.B., 1992. The Indigenous Plant Use Programme. Foundation for Research Development, Pretoria. de Wit, C.T., 1958. Transpiration and crop yields. Agric. Res. Rep. 64, 1–88. Faber, M., Oelofse, A., Van Jaarsveld, P.J., Wenhold, F.A.M., Jansen van Rensburg, W.S., 2010. African leafy vegetables consumed by households in the Limpopo and KwaZulu-Natal provinces in South Africa. South Afr. J. Clin. Nutr. 23, 30–38. Farahani, H.J., Izzi, G., Oweis, T.Y., 2009. Parameterization and evaluation of the AquaCrop model for full and deficit irrigated cotton. Agron. J. 101, 469–476. Garcia-Vila, M., Ferres, E., Matoes, L., Orgaz, F., Steduto, P., 2009. Deficit irrigation optimization of cotton with AquaCrop. Agron. J. 101, 477–487. Geerts, S., Raes, D., Garcia, M., Miranda, R., Cusicanqui, J.A., Taboada, C., Mendoza, J., Huanca, R., Mamani, A., Condori, O., Mamani, J., Morales, B., Osco, V., Steduto, P., 2009. Simulating yield response of quinoa to water availability with AquaCrop. Agron. J. 101, 499–508. Greaves, G.E., Wang, Y., 2016. Assessment of FAO AquaCrop model for simulating maize growth and productivity under deficit irrigation in a tropical environment. Water 8, 1–18. Grubben, G.J.H., 2004. Amaranthus cruentus L. In: Grubben, G.J.H., Denton, O.A. (Eds.), PROTA 2: Vegetables/Legumes (CD – Rom). PROTA, Wageningen, The Netherlands, pp. 71–72. Hanks, R.J., 1983. Yield and water-use relationships. In: Taylor, H.M., Jordan, W.R., Sinclair, T.R. (Eds.), Limitations to Efficient Water Use in Crop Production. ASA, CSSA, and SSSA, Madison, WI, pp. 393–411. Hanks, R.J., Sisson, D.V., Hurst, R.L., Hubbard, K.G., 1980. Statistical analysis of results from irrigation experiments using the line source sprinkler system. Soil Sci. Soc. Am. J. 44, 884–888. Hussein, F., Janat, M., Yakoub, A., 2011. Simulating cotton yield response to deficit irrigation with the FAO AquaCrop model. Span.J. Agric. Res. 9, 1319–1330. Jansen van Rensburg, W.S., van Averbeke, W.S., Abbert, R., Faber, M., van Jaarsveld, P., van Heerden, I., Wenhold, F., Oelofse, A., 2007. African leafy vegetables in South Africa. Water SA 33, 317–326. Johnson, B.L., Henderson, T.L., 2002. Water use patterns of grain amaranth in the Northern Great Plains. Agron. J. 94, 1437–1443. Karunaratne, A.S., Azam-Ali, S.N., Izzi, G., Steduto, P., 2011. Calibration and validation of FAO-AquaCrop model for irrigated and water deficient bambara groundnut. Exp. Agric. 47, 509–527. Katerji, N., Campi, P., Mastrorilli, M., 2013. Productivity, evapotranspiration, and water use efficiency of corn and tomato crops simulated by AquaCrop under contrasting

310