Development of a fodder beet potential yield model in the next generation APSIM

Agricultural Systems 158 (2017) 23–38

Contents lists available at ScienceDirect

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

Development of a fodder beet potential yield model in the next generation APSIM

MARK

E.N. Khaembah⁎, H.E. Brown, R. Zyskowski, E. Chakwizira, J.M. de Ruiter, E.I. Teixeira The New Zealand Institute for Plant & Food Research Limited, Private Bag 4704, Christchurch, New Zealand

A R T I C L E I N F O

A B S T R A C T

Keywords: Fodder beet model Potential yield Physiological development Nitrogen partitioning

The growing importance of fodder beet (Beta vulgaris subsp. vulgaris var. alba L.) as stock feed in recent years has created the need to develop a crop model to help assess crop yield potential across environmental growth conditions. This paper describes the development of a biophysical model for simulating fodder beet growth and development. The model was developed using the Plant Modelling Framework (PMF) within the next generation Agricultural Production Systems sIMulator (PMF-APSIM). The model was parameterised/calibrated and validated using independent datasets from field experiments conducted in the Canterbury region of New Zealand. A sensitivity analysis was conducted to explore yield response to variation in the extinction coefficient and the air temperature. The results show that canopy-related variables (leaf appearance, leaf senescence, leaf area index and light interception) were the most accurately simulated. Dynamic dry matter (DM) and nitrogen (N) accumulation in different plant organs were simulated with intermediary accuracy. Reduced accuracy was mainly observed in the earliest (September) and latest (December) sowing dates. This suggests that responses to seasonal environmental drivers, such as day length and threshold temperatures, are areas that require further research. Similarly, more mechanistic representations of carbon and N partitioning to different plant organs may improve simulation accuracy. The sensitivity analysis showed that DM production was responsive to temperature and the extinction coefficient. This initial development and testing of the fodder beet model in APSIM has helped to identify key knowledge gaps in the understanding of the physiology of the crop and provides new directions for model development.

1. Introduction Fodder beet (Beta vulgaris subsp. vulgaris var. alba L.) is a forage crop of increasing importance for livestock production in the temperate climates of Europe (Albayrak and Yuksel, 2010; Turk, 2010), New Zealand and Australia (Chakwizira et al., 2014a; Matthew et al., 2011; Rawnsley et al., 2013). In New Zealand, a combination of high productivity and animal performance has made the forage economically attractive for winter grazing systems, with a 60% increase in sown area during the last 4 years (Chakwizira et al., 2016a). High yield (> 20 t DM/ha per year) and high energy content (~ 12.5 MJ/kg DM) of the forage (Chakwizira et al., 2014a), mostly due to carbohydrates accumulated in storage roots, enables elevated stocking rates and high animal performance (Edwards et al., 2014b). Recent studies also indicate potential environmental benefits of fodder beet use, when compared with other winter grazing forage options in New Zealand, due to the lower risk of nitrogen (N) leaching from deposited animal urine as the forage has a high carbon:nitrogen (C:N) ratio (Edwards et al.,

⁎

2014a; Malcolm et al., 2016). Fodder beet is closely related to sugar beet (Beta vulgaris subsp. vulgaris convar. vulgaris var. altissima). Plant growth and the formation of individual organs are governed by similar factors although there are morphological and physiological differences between the two crops (Černý et al., 1988). As stated by these authors, leaves are fewer, slightly undulated, and petioles are shorter and narrower in fodder beet than sugar beet. Also, the storage root of fodder beet is usually more prominent above the soil surface, has a larger proportion of the hypocotyl, and its cross-section shows a smaller number of rings of vascular bundles. In terms of research, sugar beet has received a lot of attention including development of simulation crop models (Baey et al., 2014; Lemaire et al., 2008; Qi et al., 2005; Spitters et al., 1989; Vandendriessche, 2000; Webb et al., 1997; Werker et al., 1999). The same cannot be said of fodder beet, but its recent comeback has been accompanied by increased research attention that makes the development of simulation model, a logical and important research step. The development of a fodder beet model provides the opportunity to

Corresponding author. E-mail address: [email protected] (E.N. Khaembah).

http://dx.doi.org/10.1016/j.agsy.2017.08.005 Received 28 February 2017; Received in revised form 21 August 2017; Accepted 21 August 2017 0308-521X/ © 2017 Published by Elsevier Ltd.

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Experiment 1 (2011–2012), Experiment 2 (2012–2013) and Experiment 3 (2013–2014) are detailed in Chakwizira et al. (2014a,b, 2016c). Briefly, experimental treatments consisted of a range of irrigation water and N fertiliser amounts supplied to fodder beet crops during growth from spring to a final harvest in late-autumn (May) to early winter (June). Data used to parameterise/calibrate and test the model were only from treatments that received full irrigation and full N fertiliser (i.e. potential yield under unconstrained crop growth conditions). Best management agronomic practices were applied in all experiments to minimise risks of yield loss by insects, pathogens and weeds. The experimental data used in this modelling study were (i) biomass and N concentration determined separately for individual fractions of the plant (leaf laminae, petioles, storage roots and dead plant material), (ii) canopy development (leaf appearance, leaf area index (LAI) and leaf senescence) and (iii) canopy light interception. Experiment 4 (unpublished) was conducted from 2014 to 2015 and involved fodder beet crops growing under unconstrained water and N supply conditions, and subjected to a range of sowing dates (19 September, 17 October, 17 November and 15 December in 2014). This provided the opportunity to test the model across a wider range of weather conditions. Experiment 4 data collection details are described below. Measurements in Experiment 4 included those taken in Experiments 1–3 and additional data on soil temperature from sowing to germination, plus individual leaf expansion rates. Soil temperature was recorded electronically using three thermistors attached to a HOBO Pro Series 8 data logger (Onset Computer, Bourne, MA, USA). Thermistors were placed 2 cm deep (depth at which seeds were drilled) in the soil at sowing and temperature recorded every 30 min. Thermal-time to emergence was estimated as degree days (°Cd) accumulated above a base temperature (Tb) of 0 °C (Chakwizira et al., 2016c) from sowing until achievement of 80% of the final population. Emergence counts were made daily within 2 m2 quadrats marked in plots. Leaf appearance and leaf senescence were measured fortnightly in five marked plants per plot throughout the growing season following the methodology described by (Chakwizira et al., 2016a). Leaf senescence was assumed to have occurred when 50% of the leaf laminae area was pale-yellow (e.g. Milford et al., 1985b). Leaf laminae expansion at individual leaf positions was observed only in the October-sown crops. Measurements were made weekly at the 12th, 16th and 20th node positions on three plants per replicate for two replicates, following the procedure outlined by Milford et al. (1985c). Leaf area (i.e. length ∗ breadth) was adjusted using a form factor (0.85) estimated from 131 leaves following the method described by Bryson et al. (1997).

identify strengths and knowledge gaps in the quantitative understanding the crop's physiology and guide future research. Crop biophysical models can be used to support decision making at multiple scales (e.g. field, farm and region) and to inform research (Teixeira et al., 2016) and policy decisions (Ewert et al., 2015; Holzworth et al., 2014). Biophysical models are valuable tools to investigate the interactions between crop genotypes, management and environmental conditions that determine the productivity and environmental aspects of agricultural systems (Teixeira et al., 2015, 2016). This is particularly important for crops such as fodder beet because recent studies stress the need to expand the range of economically relevant species currently represented in crop models for a more comprehensive assessment of regional and global changes to agricultural production (Ewert et al., 2015; Rosenzweig et al., 2013; Rotter et al., 2011). New model development efforts can now benefit from advanced software platforms that enable flexible and transparent representation of crop physiological processes (Holzworth et al., 2014). This is the case with the Plant Modelling Framework (PMF; Brown et al., 2014) recently developed within the Agricultural Production Systems sIMulator (APSIM) model. The objectives of this study are: (i) to develop a fodder beet simulation model within the PMF-APSIM and evaluate the model's predictive ability using independent data from experiments conducted in the Canterbury region of New Zealand and (ii) to identify strengths and research gaps in the quantitative understanding of fodder beet physiology to guide future research and improvement of the model. 2. Materials and methods 2.1. Next generation APSIM and PMF model overview APSIM (available at www.apsim.info) is a process-based model that simulates physical and biological processes in agricultural systems. It contains a suite of modules that enables simulation of a range of plant, animal, soil, climate and management interactions. The fundamental attributes of APSIM have been described by Keating et al. (2003). Development of the model began in the early 1990s and has evolved significantly over time as detailed by Holzworth et al. (2014). The next generation APSIM is a more modern, multi-platform, source codebase, based on the .NET framework (Holzworth et al., 2014). The Plant Modelling Framework (PMF) developed by Brown et al. (2014) is a new framework containing a library of plant organ and process sub-models that can be coupled, at runtime, to construct a model in much the same way that models can be coupled to construct a simulation (Holzworth et al., 2014).

2.3. Data use 2.2. Field experiments Data from Experiment 1 and the October sowing date of Experiment 4 were used to parameterise/calibrate the model. The model was validated using data from Experiment 2, Experiment 3 and Experiment 4 (September, November and December sowing dates).

Data for parameterising/calibrating and validating the model were obtained from four different field experiments (Experiments 1–4) conducted in Canterbury, New Zealand between 2011 and 2015. A commercial fodder beet variety ‘Rivage’ (Agricom, Christchurch) was used in all experiments. Experiments 1, 2, and 4 were conducted at New Zealand Institute for Plant & Food Research Limited, Lincoln site (43° 37′ 33.0″ S, 172° 28′ 15.9″ E, 12 m a.s.l) on a deep well-drained Templeton silt loam soil (Udic Ustochrept, U.S. Soil Taxonomy). Experiment 3 was conducted at Ashley Dene (43° 38′ 45.5″ S, 172° 20′ 34.4″ E, 30 m a.s.l) which is located ~10 km from Lincoln and characterised by Balmoral shallow stony silt loam soil (Udic Haplustepts loamy skeletal, U.S. Soil Taxonomy) with low to moderate plant available water (Chakwizira et al., 2016a). All experiments were laid out in a Randomized Complete Block Design with 3–4 replicates. The sites are assumed to be subjected to the same climatic conditions because of their proximity. Mean monthly temperature and monthly total rainfall during the experimental period (2011–2015) as well as longterm (1970–2010) temperature and rainfall data are presented in Fig. 1.

2.4. Model structure and parameterisation/calibration The following sections describe key structural and parameterisation/calibration elements of the fodder beet model. A comprehensive representation of the model is provided in the supplementary file. 2.4.1. Schematic representation of the model The fodder beet model overview is represented in Fig. 2. The model simulates development and growth of fodder beet, assuming crop growth is only limited by temperature and radiation with other environmental conditions, management and growth resources being optimal, i.e. crops are sufficiently provided with water, nutrients and protection from diseases and pests. Parameters used in the model as well as sources are presented in Table 1. Additional parameters and 24

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Fig. 1. Monthly total rainfall (a) and average temperature (b) during the experimental period (2011–2015) of four experiments conducted at Lincoln (EXP 1, 2 and 4) and Ashley Dene (EXP 3) sites in Canterbury, New Zealand. LTM represents longterm mean. Data were obtained from ‘Broadfields’ weather station (National Institute of Atmospheric Research – NIWA 17603 (NIWA, 2016), located at the Lincoln experimental site.

250

Monthly mean total rainfall (mm)

(a)

EXP 1 (2011-12) EXP 2 (2012-13)

200

EXP 3 (2013-14) EXP 4 (2014-15) LTM (1970-2010)

150

100

50

0 Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Month of the year 20

Monthly mean temperature (°C)

(b)

15

10

5

0 Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Month of the year

accumulation is calculated for each period and combined to give daily values (Jones et al., 1986). The optimum temperature was set at 25–30 °C with a maximum of 45 °C (Fig. 3a), targets assumed similar to those of sugar beet (Ober and Rajabi, 2010; Sanghera et al., 2016). Targets of thermal-time accumulation determine the transition between phenological stages. Maximum and minimum daily air temperatures determine the rate of progression toward the next phase based on fixed TTsum targets, except for the germinating phase which depends on soil moisture and the emerging phase where the TTsum target is calculated based on the sowing depth, i.e. °Cd/mm of sowing depth (Table 2). Data from Experiment 4 indicated an average of 200 °Cd and a range of 142–250 °Cd across sowing dates. The average TTsum for the October sowing date (250 °Cd) was used in the model because most fodder beet crops in New Zealand are sown in October. Crops were drilled to a depth of 20 mm and therefore, an emergence rate of 12.5 °Cd/mm was initially assumed in the model to describe hypocotyl elongation during crop emergence.

specifications are detailed in sub-sections of the model description for ease of reference.

2.4.2. Phenological development Crop development is described by a progression through six phenology stages (Table 2). The transition between sequential phenological development phases in the model is driven mainly by temperature. The effect of temperature on phenological development is quantified using a thermal-time accumulation function with a base temperature (Tb) of 0 °C (Chakwizira et al., 2016c). In New Zealand, biennial fodder beet plants are sown in spring and grazed or harvested between autumn and winter. They are not normally carried through to flowering and therefore this modelling will only focus on the vegetative stages of development (i.e. from planted to vegetative). However, the “Bolt” and “Maturity” phases (Table 2) are included for continuity to seed production. A value of 3600°Cd (i.e. sum of °Cd from crop emergence to the end of the vegetative phase) is assumed for the duration of the “Maturity” phase. Accumulated thermal time (TTsum) for each phase (planted–vegetative) was empirically derived from the October sowing date data of Experiment 4. Temperature is extrapolated to 3-hourly values from daily maximum and minimum using a sinusoidal function, and thermal-time

2.4.3. Canopy development and light interception Individual leaf size is parameterised/calibrated for each leaf position, with values increasing from 80 cm2 at the first leaf position, reaching a maximum of 500 cm2 at node position 10 and decreasing to 25

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Fig. 2. Representation of the components and controlling environmental variables in the fodder beet model. RUEtotal = radiation use efficiency for total plant biomass, LAI = leaf area index, DM = dry matter, Rado = total incident solar radiation, and Radi = intercepted solar radiation.

100 cm2 beyond leaf position 50. Individual leaf sizes per node position were measured in the October sowing date of Experiment 4 and the pattern of the leaf size distribution for sugar beet crops (Milford et al., 1985c) was assumed. The period necessary for full leaf expansion, the duration of the fully expanded leaf and the onset and rate of leaf senescence were calculated in thermal-time bases for each leaf node position (Fig. 7b). The proportion of solar radiation intercepted by the crop increases exponentially to full canopy cover (Radi/Rado of 1.0; Eq. (1)) in response to increasing green LAI (Monsi and Saeki, 2005).

Radi = 1 − e−k × LAI Rado

Table 2 Phenological stages of development parameterised/calibrated for the fodder beet model. TTsum denotes accumulated thermal time in degree days (°Cd).

(1)

Stage

Start phase

End phase

Criteria

Planted

Sowing

Germination

Germinating Emerging Vegetative

Germination Pre-emergence Established

Bolt

Initial reproduction Start bolting

Pre-emergence Established Initial reproduction Start bolting

Plant available soil moisture > 0 mm at sowing depth. TTsum = 250 °Cd TTsum = 400 °Cd TTsum = 3200 °Cd

Maturity

where Radi is the radiation intercepted by the crop canopy, Rado is the global solar radiation i.e. the incident radiation above the canopy, k is the extinction coefficient. A k value of 0.74 based on previous studies (Chakwizira et al., 2016c) was used.

Maturity

TTsum = 800 °Cd TTsum = 3600 °Cd

2.4.4. Crop growth and biomass partitioning In a typical situation, daily DM production is assumed to be proportional to the daily intercepted radiation. Intercepted light is converted into total crop biomass at a maximum RUE for the total plant

Table 1 List of parameters, abbreviations and values used in fodder beet model. DM denotes dry matter and N denotes nitrogen. Parameter

Abbreviation

Unit

Value

Source

Leaf area growth pattern Leaf growth and lag duration Minimum N uptake Maximum N uptake Organ N concentration Optimum growth temperature Maximum growth temperature Radiation extinction coefficient Plant height Phyllochron 1 (0–3 leaves) Phyllochron 2 (≥3 leaves) Radiation use efficiency Specific leaf area Plant population

– – – – _ – – k – – – RUE SLA –

– – kg ha− 1 day− 1 kg ha− 1 day− 1 Percent (%) °C °C m2 (ground) m− 2 (leaf) m °C days leaf− 1 °C days leaf− 1 g (DM) MJ− 1 intercepted radiation mm2 (leaf) g− 1 (leaf DM) Plants m− 2

50,000 – 0.1 5.8

Experiment 4, Milford et al. (1985a,b,c) Teixeira et al. (2011) Assumed Armstrong et al. (1986) Chakwizira et al. (2016b) Ober and Rajabi (2010), Sanghera et al. (2016) Ober and Rajabi (2010), Sanghera et al. (2016) Chakwizira et al. (2016c) Experiment 4 Assumed Experiment 4 Brown et al. (1987) Experiment 4 Experiment 4

26

25–30 45 0.74 0.5 100 50 1.4 16,000 9

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

(a)

1.0

25

Relative leaf size (fractional)

Thermal-time accumulation (oCd)

30

20

15

10

5

0.8

(b)

Maximum leaf size

Senescence phase

Expansion phase

0.6

0.4

Leaf appearance

0.2

Leaf senesced

0.0

0 0

10

20

30

40

50

Thermal-time

Temperature (oC)

Fig. 3. Daily thermal-time accumulation in response to temperature (a) and schematic representation of leaf life span in response to thermal-time (b).

calculated from potential LAI expansion and the mean specific leaf area of 16,000 mm2/g estimated from the October sowing date of Experiment 4. Allocation of DM to other organs is calculated based on simple partitioning coefficients empirically associated with plant vegetative development represented by number of emerged leaves (Table 4). Dry matter allocation to fibrous roots, not measured in the trial, was assumed fixed at 5% at all times.

(RUEtotal) of 1.4 g DM/MJ of total solar radiation found for sugar beet at optimal air temperature (Brown et al., 1987). The maximum RUEtotal is multiplied by a temperature factor to take into account the effect of daily temperature and their fluctuations on photosynthesis (e.g. Mahbod et al., 2014). It is important to point out here that the RUE of 1.21 g DM/MJ for the October sowing date of Experiment 4 was not used because the temperature range during the experiment was mostly below the optimum. The RUEtotal was 1.05, 1.02 and 1.14 g DM/MJ for the September, November and December sowing date treatments, respectively. Crop biomass and N are separated into three types of pools: structural, non-structural and metabolic (Table 3). The constitution of each pool and movement of biomass among these pools is summarised by Brown et al. (2014). Briefly, the structural biomass always remains in the plant organ and is lost when the organ senesces. Metabolic biomass may be reallocated to other organs upon organ senescence or retranslocated (moved to another organ at any time when supplies do not meet the structural and metabolic DM demands of growing organs). Non-structural biomass may be remobilised at any time and reallocated or retranslocated only when all organs have received their structural and metabolic allocations. Each day, total DM supply is calculated as the sum of (i) newly photosynthesised DM (ii) DM available for reallocation from senescing organs and (iii) DM remobilised from non-structural DM pools. The amount of DM that each organ makes available for daily DM supply is calculated as the product of the amount of non-structural DM and the maximum rate of daily re-translocation of each organ (Table 3). Reallocation is determined by the rate of organ senescence. For each organ, the reallocation factor specifies the proportion of senesced metabolic and non-structural biomass to be reallocated (Table 3). Dry matter partitioning among organs is carried out by the “arbitrator module” (Brown et al., 2014). All organs will be allocated the amount of DM demanded if the total supply is greater or equal to total demand. If total supply is less than demand, then DM is allocated in proportion to organ's relative demands. Dry matter demand for the leaf lamina is

2.4.5. Nitrogen uptake and partitioning The daily demand for N (kg/ha) is determined by the sum of N demand by individual organs in a plant and extrapolated to the plant population. Each organ's structural N demand is calculated as the product of potential daily biomass growth and the organ's minimum N concentration (Table 5). Calculation of metabolic N demand follows a similar logic but uses the critical N concentration of the organ. The maximum, minimum and critical N concentration of organs were determined in a previous study (Chakwizira et al., 2016b). Non-structural N demand is required to bring the organ to its maximum N concentration. Nitrogen is supplied from the following three pools arranged in decreasing order of priority: N re-allocation (i.e. N moved before senescence of organs), N root uptake (i.e. daily soil N availability) and N re-translocation (N moved from storage pools of other organs). If N available from all three pools is insufficient to fulfil daily N demand, biomass growth is limited to ensure the minimum N concentration of the organ is not exceeded. Early growth N uptake is assumed to be 0.1 kg N/ha/day, and this increases with root growth reaching a maximum of 5.8 kg N/ha/day. The maximum N uptake per day is assumed from sugar beet data (Armstrong et al., 1986). Nitrogen uptake is also linearly limited by low soil water content. Re-allocation and retranslocation daily rates are determined in the same way as for DM guided by the coefficients specified in Table 3. 2.5. Analysis of model performance The model was evaluated for its accuracy of predicting total crop biomass accumulation, biomass components (leaf, petiole and storage

Table 3 Structural and non-structural fraction per organ and parameterisation/calibration of re-allocation and re-translocation of dry matter (DM) and nitrogen (N) per organ. Organ

Leaves Petiole Storage root Roots

Structural

0.40 0.95 0.95 0.80

Non-structural

0.20 0.00 0.00 0.00

Metabolic

0.20 0.05 0.05 0.20

27

Re-allocation

Re-translocation

DM

N

DM

N

0.95 0.00 0.00 0.00

0.95 0.95 0.05 0.00

0.95 0.05 0.05 0.00

0.01 0.10 0.05 0.00

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

variation in the observed data explained by the predicted data.

Table 4 Dry matter partitioning fractional coefficients in relation to phenological stages. Values were estimated from the October sowing date of Experiment 4. Leaf node position

Petiole

Storage root

3. Model efficiency (EF; Eq. (3)) n

Fibrous roots

EF = 1– 1 6 15 30 51

0.10 0.15 0.24 0.10 0.10

0.10 0.35 0.70 0.80 0.80

0.05 0.05 0.05 0.05 0.05

Minimum

Maximum

Critical

2.5 1.0 1.0 0.5

5.0 4.0 3.0 1.0

2.5 – – –

2.6. Sensitivity analysis The sensitivity of the model was tested by the response of total DM (output) to the modulation of one input parameter (extinction coefficient) and one weather parameter (air temperature). The parameters were selected to represent test sensitivity of canopy structure, plant organ growth and integration of growth and development components of the model. Values of the extinction coefficient were changed by ± 20% at steps of 10%. Sensitivity to temperature was calculated by responses to changes of daily average temperature by −2, −1, + 1 and + 2 °C. Evaluation of the model's sensitivity to temperature is important because temperature is one of the main drivers of crop production through its influence on development and growth processes (Monod et al., 2006).

1. Coefficient of determination (R2), 2. Absolute root mean squared error (RMSE) and relative RMSE (rRMSE) i.e. RMSE expressed as a percentage of observed mean (Eq. (2)). n

∑i = 1 (oi –pi )2 n

; rRMSE = ⎛ ⎝

(3)

Statistical calculations and graphs were constructed using the statistical program R (R Core Team, 2012).

root), N content of plant organs and underlying processes that influence biomass accumulation (leaf appearance, LAI expansion and light interception). Predicted and observed values were compared and the accuracy of predictions measured by four criteria:

RMSE =

(oi – o)2

4. Student's t-test (Pt) of means assuming unequal variance (e.g. Gaydon et al., 2017).

N concentration (%)

Leaf lamina Petiole Storage root Fibrous roots

n ∑1 = 1

Values of EF can be negative or positive with a maximum of 1. A positive value indicates that the simulated values describe the trend in the measured data better than the mean of observations. A negative value indicates that the simulated values describe the data less well than a mean of the observations (Smith et al., 1997). A model efficiency level > 0.6 is generally considered very efficient (Smith et al., 1996).

Table 5 Metabolic and structural fractions of N in individual organs of fodder beet plants. Values estimated. Plant organ

∑i = 1 (pi –oi )2

RMSE ⎞ ∗100 o ⎠

(2) 3. Results

where n is the number of observations, oi is the observed value for observation i, pi is the predicted value for observation i and o is the average of observed values. A model is considered to reproduce experimental data best when RMSE between predicted and observed data is similar or less than the standard deviation of the experimental observations (Gaydon et al., 2017). The rRMSE gives an indication of the

3.1. Crop emergence Crop emergence estimates for Experiment 4 was well-estimated for the later-sown crops, but under-estimated in the early-sown crops (Fig. 4). Fig. 4. Comparison of simulated and observed number of days from sowing to emergence for fodder beet sown on four different dates in Lincoln, New Zealand. Each symbol represents the mean (n = 4).

28

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

good fit and non-significant difference between observed and predicted values. The high accuracy was also supported by the RMSE values which were either lower or similar to the SD of the experimental data (Table 6). For the number of appeared and green leaves, simulated values reproduced observed values, although there was a tendency to over-predict in early crop development (Fig. 5a, b). Results also indicated greater bias in the number of green leaves in late crop development (Fig. 5b). Predictions of LAI and light interception were slightly less accurate than the number of appeared or green leaves (Table 6). For light interception, the model tended to over-predict and underpredict in early and late crop development, respectively (Fig. 5d). For plant biomass, R2 values (0.60–0.86; Table 6) indicated moderate to good agreement between predicted and observed values, and the t-test indicated similarity (Pt > 0.12) between predicted and observed values for all biomass variables. The EF values (0.42–0.86) indicated that the simulated values captured the trend in the observed data better than the mean of observation. Among all biomass variables, the petiole DM was the least accurately predicted (Table 6). Graphical representation indicated slight over-prediction of the total DM in early crop development and increased bias in late crop development (Fig. 6a). A similar pattern was depicted for the storage root DM (Fig. 6b). Prediction of leaf DM was the most accurate among the three components (storage root DM, leaf DM and petiole DM) and was also

Table 6 Model parameterisation/calibration statistical analysis of observed and predicted data of variables evaluated in the study. n = number of measurements, R2 = coefficient of determination; RMSE = root mean squared error; SD = standard deviation; EF = modelling efficiency; Pt = Student's t-test at the 95% (P < 0.05) confidence level. Parameter

n

R2

RMSE

SD

EF

Pt

Appeared leaves Green leaves LAI Light interception Total DM Storage root DM Leaf DM Petiole DM Total N Storage root N Leaf N Petiole N

25 25 14 22 14 14 14 14 13 13 13 13

0.99 0.92 0.78 0.59 0.86 0.86 0.82 0.60 0.83 0.77 0.81 0.40

2.0 2.7 0.6 0.20 348.1 386.7 31.3 53.7 5.5 4.4 1.6 0.8

4.4 4.2 1.6 0.04 275.7 228.1 34.9 33.1 5.3 4.0 1.3 0.7

0.99 0.89 0.77 0.59 0.86 0.77 0.80 0.42 0.53 0.66 0.32 0.10

0.39 0.49 0.40 0.44 0.44 0.24 0.46 0.12 0.08 0.23 0.42 0.10

3.2. Model parameterisation/calibration The predicted values were in good agreement with the observed data for all canopy-related traits evaluated in this study (Table 6). The high values of EF (0.59–0.99) and Pt (0.39–0.49) (Table 6) indicated a

Fig. 5. Model parameterisation/calibration results of predicted and observed data for canopy-related traits: (a) Number of appeared leaves, (b) number of green leaves, (c) leaf area index (LAI), and (d) light interception (fractional). The solid line is a 1:1 relationship and the dotted line is the linear relationship between observed and predicted with a 95% confidence interval.

29

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Fig. 6. Model parameterisation/calibration results showing predicted versus observed DM (g/m2) in different organs of fodder beet crops evaluated in field trials over three seasons in Canterbury, New Zealand. The solid line is a 1:1 relationship and the dotted line is the linear relationship between observed and predicted with a 95% confidence interval.

experimental uncertainty. Graphical representation of predicted and observed data indicated that the pattern for light interception closely resembled that for the number of green leaves (Fig. 8b, d). The prediction of plant DM indicated varying levels of agreement among plant components (Table 7). Predicted and observed values for the total DM were strongly correlated (R2 = 0.87), and both Pt and EF values indicated a good agreement (Table 7). There was a tendency to over-predict total DM, especially in the late stages of crop development (Fig. 9a). Overall, the total DM was over-estimated by only 4.3%. The RMSE was 44% greater than the SD value (Table 7), implying that the predicted total DM fell outside the limits of the experimental uncertainty. For storage root DM which comprised most of total DM, predicted and observed data correlated well (R2 = 0.87) and the model fit was accurate as indicated by EF and Pt values of 0.86 and 0.46, respectively (Table 7). The storage root DM produced toward the end of the season was under-estimated (Fig. 9b) resulting in an overall underestimation of 10.5%, and a 47% discrepancy between the SD and RMSE (Table 7). The EF and Pt values indicated leaf DM was accurately predicted although with less accuracy than the storage root DM (Table 7). However, predicted leaf DM data accounted for most of the variation (75%) in the observed leaf DM data. Both EF and Pt indicated a good agreement between predicted and measured data for petiole DM (Table 7). However, both R2 (0.5) and the graphical representation of

the only variable with comparable values of RMSE and SD (Table 6). The prediction of N uptake was accurate and closely mirrored that of plant biomass especially in relation to the R2 values and also how the variables ranked in prediction accuracy (Table 6). However, contrary to biomass prediction, all statistical analyses indicated good agreement between predicted and observed values (Table 6). Graphical representation of data indicated similar patterns for biomass and N uptake (Figs. 6 & 7). 3.3. Model validation The model validation prediction accuracy indicators mostly reflected the parameterisation/calibration results for the respective evaluated variables (Tables 6 & 7). All analyses indicated accurate prediction of all canopy-related traits, with the number of appeared leaves being the most accurately predicted (Table 7). For the number of green leaves, graphical comparison indicated substantial bias in late crop development, mainly contributed by the earliest (September) and latest (December) sowing dates (Fig. 8b). Both LAI and light interception were predicted with moderate accuracy, as shown by respective R2 values of 0.67 and 0.62 (Table 7). The EF values were also moderate for LAI and light interception, but SD and RMSE values were comparable (Table 7), indicating that the errors were within the limits of

30

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Fig. 7. Model parameterisation/calibration results of the relationship between predicted and observed N uptake (g/m2) by different organs of fodder beet crops evaluated in field trials over three seasons in Canterbury, New Zealand. The solid line is a 1:1 relationship and the dotted line is the linear relationship between observed and predicted with a 95% confidence.

time to the systematic lack of fit of the model identified in Figs. 8 & 9. There substantial variation in measured data, but the model satisfactorily captured the trend of LAI across sowing dates (Fig. 10). The model over-estimated LAI in early crop development for the Septemberand October-sown crops (Figs. 10a, b). In late development, the model under-estimated LAI for these crops. By contrast, estimation of LAI in later-sown crops was accurate except for two time points after peak LAI in the December-sown crops (Fig. 10c, d). The pattern of light interception (Fig. 11) reflected that of LAI (Fig. 10) i.e. over-estimation in early crop life and over-estimation in late development of for the earliest sown crops (Fig. 11a, b). There was slight over-estimation of light interception in early crop development of the November-sown crops (Fig. 11c). The LAI monitoring started later than this date and therefore the corresponding LAI for this date was not measured. Light interception was well-estimated in the December-sown crops, but the under-estimation corresponding to the under-estimated LAI was apparent (Fig. 11d). The model satisfactorily captured the pattern of biomass accumulation of fodder beet crops (Fig. 12). The tendency toward over-estimation in early crop development of the earliest-sown crops (Fig. 12a, b) reflects the over-estimated of LAI and light interception during this period. The seasonal dynamics of the total amount of N taken up by the

data (Fig. 9d) indicated that petiole DM was the least accurately predicted yield component. Graphs of LAI, light interception and total DM of Experiment 4 treatments were constructed to examine the contribution of sowing

Table 7 Model validation analysis of predicted and predicted variables evaluated in the study. n = number of measurements, R2 = coefficient of determination; RMSE = absolute root mean squared error; SD = standard deviation; EF = modelling efficiency; Pt = Student's t-test at the 95% (P < 0.05) confidence level. Parameter

n

R2

RMSE

SD

EF

Pt

Appeared leaves Green leaves LAI Light interception Total DM Storage root DM Leaf DM Petiole DM Total N Storage root N Leaf N Petiole N

93 93 36 58 36 36 36 36 18 18 18 18

0.99 0.86 0.69 0.67 0.87 0.89 0.71 0.50 0.92 0.84 0.60 0.46

2.1 2.4 0.7 0.1 346.0 288.0 40.0 57.3 3.3 4.6 1.8 0.8

4.3 3.0 0.6 0.1 197.0 152.0 28.5 32.0 5.0 3.7 1.3 0.8

0.98 0.85 0.64 0.65 0.84 0.86 0.66 0.49 0.90 0.70 0.47 0.41

0.55 0.84 0.64 0.58 0.75 0.46 0.95 0.64 0.66 0.41 0.83 0.49

31

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Fig. 8. Model validation results of predicted and observed data for canopy-related traits: (a) Number of appeared leaves, (b) number of green leaves, (c) leaf area index (LAI), and (d) light interception (fractional). The solid line is a 1:1 relationship and the dotted line is the linear relationship between observed and predicted with a 95% confidence interval.

4. Discussion

total biomass (leaf laminae, petioles and storage roots) was accurately predicted (Table 7), although there was a tendency toward over-estimation (Fig. 13a). Predicted and observed data correlated strongly and the variation of the measured total N explained by the predicted total N was high (88%; Table 7). The good fit was supported by other accuracy indicators (Table 7). For the individual plant components, predicted data explained most of the variation (71–75%) in the observed data, and both EF and Pt values indicated a good agreement (Table 7). This was also supported by RMSE values which were either less or similar to SD values (Table 7). Graphical representation (Fig. 13b, d) indicated under-estimation of storage root N and petiole N, particularly in late crop development. Predicted and measured petiole N values were moderately correlated (R2 = 0.46; Table 7).

In this study, a biophysical model to represent the physiology of fodder beet crops was developed using the PMF-APSIM. The accuracy of model simulations was evaluated using independent datasets derived from field experiments conducted in the Canterbury region of New Zealand between 2011 and 2015. Insights from this analysis are important indicators of key areas of research focus to improve the quantitative understanding of fodder beet physiology. The use of a constant TTsum requirement of 12.5 °Cd/mm sowing depth, derived from the October sowing date trial (Experiment 4) gave satisfactory simulations of the number of days from sowing to crop emergence for later-sown crops (November and December), but an underestimation for the earlier-sown crops (September and October). The TTsum for the October sowing date was 100 °Cd greater than the 150 °Cd observed for sugar beet crops (Lemaire et al., 2008). The measured TTsum varied among sowing dates (142–250 °Cd) and also within sowing dates. Possible causes of variation include differences in sowing depth and variation in soil moisture. Germination is sensitive to soil moisture, and while care was taken to ensure there was adequate soil moisture during the trials, it is possible that crops experienced sub-

3.4. Model sensitivity The sensitivity analysis indicated that changes to the extinction coefficient resulted in less than proportional changes (− 10–7%) in the total DM (Table 8). The total DM increased with increased average air temperature, while decreasing the temperature had an opposite but greater effect (Table 8).

32

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Leaf area index (LAI, mm2 mm-2)

Fig. 9. Model validation results of predicted versus observed biomass (g/m2) accumulated in different organs of fodder beet crops evaluated in field trials over three seasons in Canterbury, New Zealand. The solid line is a 1:1 relationship and the dotted line is the linear relationship between observed and predicted with a 95% confidence interval.

(a)

(b)

(c)

(d)

Time (days after sowing) Fig. 10. Measured (symbols [mean ± SD]) and predicted (lines) leaf area index of fodder beet crops sown on four different dates in 2014 at Lincoln, New Zealand.

33

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Light interception (fraction of total radiation; 0–1)

(a)

(b)

(c)

(d)

Time (days after sowing)

Total biomass (g DM m-2)

Fig. 11. Measured (symbols [mean ± SD]) and predicted (lines) proportion of light intercepted by fodder beet crops sown on four different dates in 2014 at Lincoln, New Zealand.

(a)

(b)

(c)

(d)

Time (days after sowing) Fig. 12. Measured (symbols [mean ± SD]) and predicted (lines) dry matter yield of fodder beet crops sown on four different dates in 2014 at Lincoln, New Zealand.

34

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Fig. 13. Model validation results showing the relationship between predicted and observed N uptake (g/m2) by different organs of fodder beet crops evaluated in field trials over three seasons in Canterbury, New Zealand. The solid line is a 1:1 relationship and the dotted line is the linear relationship between observed and predicted with a 95% confidence.

light interception. Also, this study revealed progressive reduction in leaf size after peak LAI, consistent with observations in sugar beet (Milford et al., 1985c). For the September-sown crops subjected to the longest growing season, the small leaf size at these advanced stages of crop development coupled with the saturating nature of light interception, implies a less than proportional impact of biases in green leaf number simulations on light interception and biomass production (e.g. White et al., 2000), The requirement of 588–680 °Cd (Tb = 0 °C) from emergence to full leaf expansion observed in this study compares well with the 300–650 °Cd range observed in sugar beet crops (Milford et al., 1985c). However, measurement of individual leaf area in this study was limited to three leaf positions and one dataset. Potential errors in light interception due to incorrect individual leaf sizes would require leaf area measurements at all node positions. Furthermore, it is well-established that leaf size in fodder beet is highly sensitive to the availability of water and N (Chakwizira et al., 2016c), suggesting the need to assess crop response to conditions of constrained growth. Despite both EF and Pt indicating a good fit between simulated and measured biomass, comparison of RMSE and SD indicated that simulations of total DM and DM in different organs were outside the bounds of experimental uncertainty (Gaydon et al., 2017). Similar to canopy variables, the bias was mostly associated with the out-of-season sowing times especially the September sowing date. Future model improvements could explore the inclusion of parameters adapted to changes in

Table 8 Simulated percentage change in the final total DM yield (DMY) with increase or decrease in changes extinction coefficient and air temperature. Parameters

DMY (%)

Extinction coefficient

Air temperature

–20%

−10%

10%

20%

− 2 °C

− 1 °C

1 °C

2 °C

−9.8

−4.3

3.7

6.5

− 9.6

− 4.4

2.9

4.8

optimal moisture levels during germination. This indicates the need for more careful evaluation of soil moisture at the depth of sowing and an evaluation of the rules to trigger germination in relation to soil moisture. Data on soil moisture effect on root development are important for deriving soil water factors that introduce the response to water stress, which is important in modern agriculture. The accurate simulation of the number of appeared leaves across seasons and sowing dates indicates that a phyllochron value of 50 °Cd, after the third leaf position, can be applied across contrasting environments. Although the number of green leaves was simulated accurately, the decline in model accuracy at advanced stages of crop development, mainly associated with earliest and latest sowing dates, indicates the value of the sensitivity of leaf growth and development to other seasonal drivers, such as day length (e.g. Milford and Lenton, 1976). Addressing this may resolve the similar pattern observed for 35

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

uptake and utilisation of these resources (Teixeira et al., 2014). The sensitivity analysis revealed that the magnitude of changes in total DM yield was greater for lower values of k and air temperature, but responses were less than proportional. These patterns of response are sensible and expected due to the non-linearity of plant processes represented in biophysical models (e.g. Porter and Semenov, 2005). As a next step, a wider sensitivity testing of the model could be done across a range of environmental conditions, including future climatic scenarios in which impact assessments are often conducted using biophysical models. In the development of this fodder beet model, a number of model physiological parameters were assumed similar to the values measured in sugar beet (Brown et al., 1987; Milford et al., 1985c; Ober and Rajabi, 2010; Sanghera et al., 2016). Detailed experiments are required to test these assumptions. In addition, there are well-known morphological and physiological differences between sugar beet and fodder beet. For example, the biphasic phyllochron pattern observed in sugar beet (Lemaire et al., 2008; Milford et al., 1985a,b) has not been observed in fodder beet (Chakwizira et al., 2016c). An understanding of the implications of such differences on yield and contributing plant processes e.g. light interception and photosynthesis, is important for future improvement of fodder beet models. Finally, this model has been parameterised/calibrated and validated using data from a single location and climate. Further work is required to evaluate the model against datasets from different environments (i.e. climates, soil types and managements) and with different fodder beet cultivars. This will reveal genotype × environment interactions which must be captured to improve the robustness of the model.

seasonal environmental factors (e.g. day length) such as performed for simulating sugar beet crops (Qi et al., 2005). Although the RUEtotal of 1.4 g DM/MJ used in this study agrees well with the 1.3–1.9 g DM/MJ range reported for sugar beet in New Zealand (Martin, 1986) and France (Damay and Le Gouis, 1993; Lemaire et al., 2008), an improvement in RUE response to temperature may also contribute to seasonal yield responses. Time course graphs identified sowing date-specific differences LAI and light interception which affected total DM production. In particular, under-estimation of LAI and light interception was evident in the late stages of development of the early-sown crops. It was also clear that, for these early-sown crops, the after-peak LAI declined more rapidly than measured light interception. Since DM production is closely related to light intercepted by plants which in turn depends primarily on the LAI, it is possible that light interception measurements in the experiment were over-estimated. Over-estimation of light interception may arise from the difficulty of excluding dead or senesced leaves when measuring transmitted light in the field. This problem was not identified in later-sown crops possibly because senesced leaves had not accumulated due to the short growing period. Application of correction factors to obtain more accurate green leaf area in the late stages of crop development has been used to overcome this (e.g. White et al., 2000), but these factors need to be empirically determined for fodder beet. The accuracy of individual organ simulations was variable, indicating that biomass partitioning rules require further improvement. In particular, petiole biomass was consistently over-estimated. Coefficients used to partition biomass into different plant organs were derived from Experiment 4 data and empirically described as a function of leaf appearance. Although empirical partitioning parameters have often been shown to be robust under unconstrained conditions (Justes et al., 1994), more mechanistic partitioning rules may be required to represent the effect fluctuating availability of water and N on crop growth. Future improvements in partitioning rules may include adjustments to the allometric relationships between plant organs (Ratjen and Kage, 2016; Ratjen et al., 2016) or more mechanistic approaches like use of “functional balance” among plant organs to optimise specific physiological outcomes such as maximum relative growth rates (Yin and Schapendonk, 2004). For example, there was a consistent trend for over-estimating total DM biomass in crops that grew for longer periods (September sowing dates). While this may be due to the cumulative effect from over-estimation in early stages of crop development, it may also have been caused by the storage of large amounts of carbohydrates over late autumn to winter period. In such conditions, that resemble the dormancy period of perennial species, the rate of DM decay in storage organs (i.e. senescence plus maintenance respiration) may become relatively higher in relation to C inputs from photosynthesis (Moot et al., 2015; Teixeira et al., 2007). Testing of the model with long-term datasets for these conditions will be needed to identify the most appropriate ways to represent these dynamics in future fodder beet models. Nitrogen taken up by the plant and its allocation to different organs were simulated with similar accuracy to biomass allocation. This similarity is partially an artefact of N uptake being the product of biomass and N concentration in the organ. Again, this study used simple empirical N partitioning rules in response to vegetative development after parameterising N concentration ranges for individual plant organs. This approach is in alignment with the allometric pattern of decay similar to other C3 species recently quantified for fodder beet crops (Chakwizira et al., 2016a). In PMF-APSIM, the cascading nature of the partitioning rules in the “arbitrator module” (Brown et al., 2014), starting from storage roots and finishing with petioles, implies an accumulation of errors for organs with low priority such as petiole. Our results suggest that a revised parameterisation/calibration using a datasets from a wide range of N nutrition trials and/or more detailed representation of N demand in the model may be required for accurate simulation of N uptake. This is particularly critical for conditions of simultaneous N and water stress conditions where co-limitations affect

5. Conclusions This study was undertaken to develop a fodder beet model within PMF-APSIM and evaluate its predictive accuracy using independent data. This exercise has identified strengths and weaknesses of the current physiological understanding of fodder beet growth and development processes. For unconstrained water and N growth conditions, the most robust aspects of the model were related to phenological and vegetative development. Canopy-related variables (LAI and light interception) were simulated with intermediate accuracy, which enabled representation of the time-dynamics of biomass growth in the whole plant. The model gave realistic estimates of total biomass production and total N uptake, but the greater biases observed in off-season sowing dates suggests that other seasonal drivers of crop growth and development should be investigated. Simulation of biomass and N allocation to different organs indicated systematic biases such as for petioles, highlighting the need for improved representation of biomass and N partitioning rules and parameters. Future testing of the model against field datasets with contrasting N and water supply is recommended to evaluate the robustness of the quantitative processes represented in the model. Data for parameterisation/calibration and validation were collated from a single region. Given the sensitivity of the model to variable growing temperature, testing the model across different environments is required to assess its geographical robustness.

Acknowledgements This research was completed as part of the Forages for Reduced Nitrate Leaching programme with principal funding from the New Zealand Ministry of Business, Innovation and Employment (DNZ1301; RD1422). The programme is a partnership between DairyNZ Ltd, AgResearch, Plant & Food Research, Lincoln University, Foundation for Arable Research and Landcare Research. The authors acknowledge the contribution of The New Zealand Institute for Plant & Food Research Field Crops technical team. 36

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

Tsinghua University Press, pp. 13. Mahbod, M., Zand-Parsa, S., Sepaskhah, A.R., 2014. Adjustment of radiation use efficiency of winter wheat by air temperature at different irrigation regimes and nitrogen rates. Arch. Agron. Sci. 60, 49–66. Malcolm, B.J., Cameron, K.C., Edwards, G.R., Di, H.J., de Ruiter, J.M., Dalley, D.E., 2016. Nitrate leaching losses from lysimeters simulating winter grazing of fodder beet by dairy cows. N. Z. J. Agric. Res. 59, 194–203. Martin, R.J., 1986. Radiation interception and growth of sugar beet at different sowing dates in Canterbury. N. Z. J. Agric. Res. 29, 381–390. Matthew, C., Nelson, N.J., Ferguson, D., Xie, Y., 2011. Fodder beet revisited. Agron. N. Z. 39–48. Milford, G.F.J., Lenton, J.R., 1976. Effect of photoperiod on growth of sugar beet. Ann. Bot. 40, 1309–1315. Milford, G.F.J., Pocock, T.O., Riley, J., 1985a. An analysis of leaf growth in sugar beet. I. Leaf appearance and expansion in relation to temperature under controlled conditions. Ann. Appl. Biol. 106, 163–172. Milford, G.F.J., Pocock, T.O., Riley, J., 1985b. An analysis of leaf growth in sugar beet. II. Leaf appearance in field crops. Ann. Appl. Biol. 106, 173–185. Milford, G.F.J., Pocock, T.O., Riley, J., Messem, A.B., 1985c. An analysis of leaf growth in sugar beet. III. Leaf expansion in field crops. Ann. Appl. Biol. 106, 187–203. Monod, H., Naud, C., Makowski, D., 2006. Uncertainty and sensitivity analysis for crop models. In: Wallach, D., Makowski, D., Jones, J.W. (Eds.), Working With Dynamic Crop Models. Elsevier, Amsterdam, The Netherlands, pp. 55–99. Monsi, M., Saeki, T., 2005. On the factor light in plant communities and its importance for matter production. Ann. Bot. 95, 549–567. Moot, D.J., Hargreaves, J., Brown, H.E., Teixeira, E.I., 2015. Calibration of the APSIMLucerne model for ‘Grasslands Kaituna’ lucerne crops grown in New Zealand. N. Z. J. Agric. Res. 58, 190–202. NIWA, 2016. Climate database-NIWA. http://cliflo.niwa.co.nz/. Ober, E.S., Rajabi, A., 2010. Abiotic stress in sugar beet. Sugar Tech. 12, 294–298. Porter, J.R., Semenov, M.A., 2005. Crop responses to climatic variation. Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 360, 2021–2035. Qi, A., Kenter, C., Hoffmann, C., Jaggard, K.W., 2005. The Broom's Barn sugar beet growth model and its adaptation to soils with varied available water content. Eur. J. Agron. 23, 108–122. R Core Team, 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Ratjen, A.M., Kage, H., 2016. Modelling N and dry matter partitioning between leaf and stem of wheat under varying N supply. J. Agron. Crop Sci. 202, 576–586. Ratjen, A.M., Neukam, D., Kage, H., 2016. A simple drought-sensitive model for leaf: stem partitioning of wheat. J. Agron. Crop Sci. 202, 300–308. Rawnsley, R.P., Chapman, D.F., Jacobs, J.L., Garcia, S.C., Callow, M.N., Edwards, G.R., Pembleton, K.P., 2013. Complementary forages – integration at a whole-farm level. Anim. Prod. Sci. 53, 976–987. Rosenzweig, C., Jones, J.W., Hatfield, J.L., Ruane, A.C., Boote, K.J., Thorburn, P., Antle, J.M., Nelson, G.C., Porter, C., Janssen, S., Asseng, S., Basso, B., Ewert, F., Wallach, D., Baigorria, G., Winter, J.M., 2013. The Agricultural Model Intercomparison and Improvement Project (AgMIP): protocols and pilot studies. Agric. For. Meteorol. 170, 166–182. Rotter, R.P., Carter, T.R., Olesen, J.E., Porter, J.R., 2011. Crop-climate models need an overhaul. Nat. Clim. Chang. 1, 175–177. Sanghera, G.S., Singh, R.P., Kashyap, L., Tyagi, V., Sharma, B., 2016. Evaluation of sugar beet genotypes (Beta vulgaris L.) for root yield and quality traits under subtropical climates. J. Krishi Vigyan 5, 67–73. Smith, J., Smith, P., Addiscott, T., 1996. Quantitative methods to evaluate and compare Soil Organic Matter (SOM) Models. In: Powlson, D.S., Smith, P., Smith, J.U. (Eds.), Evaluation of Soil Organic Matter Models: Using Existing Long-term Datasets. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 181–199. Smith, P., Smith, J.U., Powlson, D.S., McGill, W.B., Arah, J.R.M., Chertov, O.G., Coleman, K., Franko, U., Frolking, S., Jenkinson, D.S., Jensen, L.S., Kelly, R.H., KleinGunnewiek, H., Komarov, A.S., Li, C., Molina, J.A.E., Mueller, T., Parton, W.J., Thornley, J.H.M., Whitmore, A.P., 1997. A comparison of the performance of nine soil organic matter models using datasets from seven long-term experiments. Geoderma 81, 153–225. Spitters, C.J.T., Van Keulen, H., Van Kraalingen, D.W.G., 1989. A simple and universal crop growth simulator: SUCROS87. In: Rabbinge, R., Ward, S.A., Van Laar, H.H. (Eds.), Simulation and Systems Management in Crop Protection. Pudoc, Wageningen, pp. 147–181. Teixeira, E.I., Moot, D.J., Pollock, K.J., Brown, H.E., 2007. How does defoliation management affect yield, canopy forming processes and light interception in lucerne (Medicago sativa L.) crops? Eur. J. Agron. 27, 154–164. Teixeira, E.I., George, M., Brown, H.E., Fletcher, A.L., 2011. Quantifying maize leaf expansion and senescence. Agronomy New Zealand 59–65. Teixeira, E.I., George, M., Herreman, T., Brown, H., Fletcher, A.L., Chakwizira, E., de Ruiter, J.M., Maley, S., Noble, A., 2014. The impact of water and nitrogen limitation on maize biomass and resource-use efficiencies for radiation, water and nitrogen. Field Crop Res. 168, 109–118. Teixeira, E.I., Brown, H.a., Meenken, E.D., Ewert, F., 2015. Evaluating methods to simulate crop rotations for climate impact assessments – a case study on the Canterbury plains of New Zealand. Environ. Model. Softw. 72, 304–313. Teixeira, E.I., Johnstone, P., Chakwizira, E., de Ruiter, J.M., Malcolm, B., Shaw, N.,

References Albayrak, S., Yuksel, O., 2010. Effects of nitrogen fertilization and harvest time on root yield and quality of fodder beet (Beta vulgaris var. crassa Mansf.). Turk. J. Field Crops 15, 59–64. Armstrong, M.J., Milford, G.F.J., Pocock, T.O., Last, P.J., Day, W., 1986. The dynamics of nitrogen uptake and its remobilization during the growth of sugar beet. J. Agric. Sci. 107, 145–154. Baey, C., Didier, A., Lemaire, S., Maupas, F., Cournède, P.-H., 2014. Parametrization of five classical plant growth models applied to sugar beet and comparison of their predictive capacity on root yield and total biomass. Ecol. Model. 290, 11–20. Brown, K., Messem, A., Dunham, R., Biscoe, P., 1987. Effect of drought on growth and water use of sugar beet. J. Agric. Sci. 109, 421–435. Brown, H.E., Huth, N.I., Holzworth, D.P., Teixeira, E.I., Zyskowski, R.F., Hargreaves, J.N.G., Moot, D.J., 2014. Plant modelling framework: software for building and running crop models on the APSIM platform. Environ. Model. Softw. 62, 385–398. Bryson, R.J., Paveley, N.D., Clark, W.S., Sylvester-Bradley, R., Scott, R.K., 1997. Use of infield measurements of green leaf area and incident radiation to estimate the effects of yellow rust epidemics on the yield of winter wheat. Eur. J. Agron. 7, 53–62. Černý, V., Hruška, L., Petr, J., 1988. Yield Formation in Fodder Beet, Yield Formation in the Main Field Crops. Elsevier Science Publishers B.V, Amsterdam, The Netherlands, pp. 317–320. Chakwizira, E., de Ruiter, J.M., Maley, S., 2014a. Growth, nitrogen partitioning and nutritive value of fodder beet crops grown under different application rates of nitrogen fertiliser. N. Z. J. Agric. Res. 57, 75–89. Chakwizira, E., de Ruiter, J.M., Maley, S., Dellow, S.J., George, M.J., Michel, A.J., 2014b. Water use efficiency of fodder beet crops. In: Proceedings of New Zealand Grassland Association, pp. 125–134. Chakwizira, E., de Ruiter, J.M., Maley, S., 2016a. Growth and nitrogen partitioning of fodder beet crops (Beta vulgaris L.) grown under varying amounts of water and nitrogen fertiliser rates in shallow soils. Agron. N. Z. 85–98. Chakwizira, E., de Ruiter, J.M., Maley, S., Teixeira, E., 2016b. Evaluating the critical nitrogen dilution curve for storage root crops. Field Crop Res. 199, 21–30. Chakwizira, E., Dellow, S.J., Teixeira, E.I., 2016c. Quantifying canopy formation processes in fodder beet (Beta vulgaris subsp. vulgaris var. alba L.) crops. Eur. J. Agron. 74, 144–154. Damay, N., Le Gouis, J., 1993. Radiation use efficiency of sugar beet in northern France. Eur. J. Agron. 2, 179–184. Edwards, G.R., de Ruiter, J.M., Dalley, D.E., Pinxterhuis, J.B., Cameron, K.C., Bryant, R.H., Di, H.J., Chapman, D.F., 2014a. Urinary nitrogen concentration of cows grazing fodder beet, kale and kale-oat forage systems in winter. In: Proceedings of the 5th Australasian Dairy Science Symposium 2014, pp. 144–147. Edwards, G.R., de Ruiter, J.M., Dalley, D.E., Pinxterhuis, J.B., Cameron, K.C., Bryant, R.H., Di, H.J., Malcolm, B.J., Chapman, D.F., 2014b. Dry matter intake and body condition score change of dairy cows grazing fodder beet, kale and kale-oat forage systems in winter. In: Proceedings of the New Zealand Grassland Association, pp. 81–88. Ewert, F., Rötter, R.P., Bindi, M., Webber, H., Trnka, M., Kersebaum, K.C., Olesen, J.E., van Ittersum, M.K., Janssen, S., Rivington, M., Semenov, M.A., Wallach, D., Porter, J.R., Stewart, D., Verhagen, J., Gaiser, T., Palosuo, T., Tao, F., Nendel, C., Roggero, P.P., Bartošová, L., Asseng, S., 2015. Crop modelling for integrated assessment of risk to food production from climate change. Environ. Model. Softw. 72, 287–303. Gaydon, D.S., Balwinder, S., Wang, E., Poulton, P.L., Ahmad, B., Ahmed, F., Akhter, S., Ali, I., Amarasingha, R., Chaki, A.K., Chen, C., Choudhury, B.U., Darai, R., Das, A., Hochman, Z., Horan, H., Hosang, E.Y., Kumar, P.V., Khan, A.S.M.M.R., Laing, A.M., Liu, L., Malaviachichi, M.A.P.W.K., Mohapatra, K.P., Muttaleb, M.A., Power, B., Radanielson, A.M., Rai, G.S., Rashid, M.H., Rathanayake, W.M.U.K., Sarker, M.M.R., Sena, D.R., Shamim, M., Subash, N., Suriadi, A., Suriyagoda, L.D.B., Wang, G., Wang, J., Yadav, R.K., Roth, C.H., 2017. Evaluation of the APSIM model in cropping systems of Asia. Field Crop Res. 204, 52–75. Holzworth, D.P., Huth, N.I., deVoil, P.G., Zurcher, E.J., Herrmann, N.I., McLean, G., Chenu, K., van Oosterom, E.J., Snow, V., Murphy, C., Moore, A.D., Brown, H., Whish, J.P.M., Verrall, S., Fainges, J., Bell, L.W., Peake, A.S., Poulton, P.L., Hochman, Z., Thorburn, P.J., Gaydon, D.S., Dalgliesh, N.P., Rodriguez, D., Cox, H., Chapman, S., Doherty, Teixeira, Sharp, J., Cichota, Vogeler, I., Li, F.Y., Wang, E., Hammer, G.L., Robertson, M.J., Dimes, J.P., Whitbread, A.M., Hunt, van Rees, McClelland, T., Carberry, P.S., Hargreaves, J.N.G., MacLeod, N., McDonald, C., Harsdorf, J., Wedgwood, S., Keating, B.A., 2014. APSIM – evolution towards a new generation of agricultural systems simulation. Environ. Model. Softw. 62, 327–350. Jones, C.A., Kiniry, J.R., Dyke, P.T., 1986. CERES-maize: A Simulation Model of Maize Growth and Development. Texas A & M University Press, Texas, USA. Justes, E., Mary, B., Meynard, J.M., Machet, J.M., Thelier-Huche, L., 1994. Determination of a critical nitrogen dilution curve for winter wheat crops. Ann. Bot. 74, 397–407. Keating, B.A., Carberry, P.S., Hammer, G.L., Probert, M.E., Robertson, M.J., Holzworth, D., Huth, N.I., Hargreaves, J.N.G., Meinke, H., Hochman, Z., McLean, G., Verburg, K., Snow, V., Dimes, J.P., Silburn, M., Wang, E., Brown, S., Bristow, K.L., Asseng, S., Chapman, S., McCown, R.L., Freebairn, D.M., Smith, C.J., 2003. An overview of APSIM, a model designed for farming systems simulation. Eur. J. Agron. 18, 267–288. Lemaire, S., Maupas, F., Cournede, P.-H., de Reffye, P., 2008. A Morphogenetic Crop Model for Sugar-beet (Beta vulgaris L.), Crop Modeling and Decision Support.

37

Agricultural Systems 158 (2017) 23–38

E.N. Khaembah et al.

sugar beet crop. Ann. Bot. 80, 427–436. Werker, A.R., Jaggard, K.W., Allison, M.F., 1999. Modelling partitioning between structure and storage in sugar beet: effects of drought and soil nitrogen. Plant Soil 207, 97–106. White, M.A., Asner, G.P., Nemani, R.R., Privette, J.L., Running, S.W., 2000. Measuring fractional cover and leaf area index in arid ecosystems: digital camera, radiation transmittance, and laser altimetry methods. Remote Sens. Environ. 74, 45–57. Yin, X., Schapendonk, A.H.C.M., 2004. Simulating the partitioning of biomass and nitrogen between roots and shoot in crop and grass plants. J. Life Sci. 51-4, 407–426.

Zyskowski, R., Khaembah, E., Sharp, J., Meenken, E., Fraser, P., Thomas, S., Brown, H., Curtin, D., 2016. Sources of variability in the effectivenes of winter cover crops for mitigating N leaching. Agric. Ecosyst. Environ. 220, 226–235. Turk, M., 2010. Effects of fertilization on root yield and quality of fodder beet (Beta vulgaris var. crassa Mansf.). Bulgarian J. Agr. Sci. 16, 212–219. Vandendriessche, H.J., 2000. A model of growth and sugar accumulation of sugar beet for potential production conditions: SUBEMOpo I. Theory and model structure. Agric. Syst. 64, 1–19. Webb, C.R., Werker, A.R., Gilligan, C.A., 1997. Modelling the dynamic components of the

38