Modeling boll maturation period, seed growth, protein, and oil content of cotton (Gossypium hirsutum L.) in China

Field Crops Research 112 (2009) 131–140

Contents lists available at ScienceDirect

Field Crops Research journal homepage: www.elsevier.com/locate/fcr

Modeling boll maturation period, seed growth, protein, and oil content of cotton (Gossypium hirsutum L.) in China Wenfeng Li a, Zhiguo Zhou a, Yali Meng a,*, Naiyin Xu a, Michel Fok b a b

Key Laboratory of Crop Growth Regulation, Ministry of Agriculture, Nanjing Agricultural University, Nanjing 210095, Jiangsu Province, PR China The Center of International Cooperation on Agronomic Research for Development (CIRAD), TA 72/09, 34398 Montpellier Cedex 5, France

A R T I C L E I N F O

A B S T R A C T

Article history: Received 27 November 2008 Received in revised form 17 February 2009 Accepted 20 February 2009

The simulation of cottonseed (Gossypium hirsutum L.) growth is still an area of great uncertainty, especially in the process of cottonseed quality formation. A simple process-based model was developed to predict cotton boll maturation period and simulate cottonseed biomass accumulation, protein, and oil content. The cotton boll maturation period module took solar radiation and N nutrition factors into account in addition to temperature and variety maturity profile. Based on the hypothesis that the accumulation of biomass, oil, and protein are mainly sink-determined, the model was developed by considering parameters of cultivar characteristics, weather (temperature and solar radiation), and crop management variables (precisely N supply). The subtending leaf N concentration of cotton boll was simulated by a new semi-empirical model, and worked as the direct indicator of the N nutrition effect on cottonseed growth and development. The model was calibrated using data obtained in experiment conducted in Nanjing (the lower reaches of Yangtze River Valley) in 2005 and 2006. The model was then tested using two field experimental data sets. One was obtained in Nanjing, China in 2007, and the other in the Yellow River Valley (Xuzhou and Anyang) and the lower reaches of Yangtze River Valley (Huaian), China in 2005. The simulated values of boll maturation period by the model were very consistent with the observed values, with root mean square error (RMSE) lower than 3 days. The RMSE of cottonseed dry weight, protein content, and oil content predictions were 8.9 mg seed1, 2.19%, and 2.71%, respectively. The result showed that the model is sufficiently robust to predict the cotton boll maturation period, cottonseed dry weight, and quality in wide range of conditions. It is not only a necessary component of cotton growth model, but also provides a good platform for further study in modeling cottonseed protein and oil yield. ß 2009 Elsevier B.V. All rights reserved.

Keywords: Cottonseed Crop model Boll maturation period Biomass accumulation Protein content Oil content

1. Introduction Simulation models are increasingly used for the assessment of crop productivity and the impact on the environment that may result from given combinations of crop characteristics, weather, soil, and crop management. The study of crop growth model started from 1960s and cotton (Gossypium hirsutum L.) was one of the earliest crops studied. Since the model SIMCOT (Duncan, 1973), a lot of systemic cotton growth models had appeared. GOSSYM (Fye et al., 1984; Reddy and Baker, 1988, 1990; Boone et al., 1993) was one of the most famous models. It simulated the dynamics of development, growth, and yield production of cotton plants. Meanwhile, KUTUN developed by Mutsaers (1984) in Holland was also a more mechanistic model. The studies in Australia paid more attention to field management, such as

* Corresponding author. Tel.: +86 25 84396813; fax: +86 25 84396813. E-mail address: [email protected] (Y. Meng). 0378-4290/$ – see front matter ß 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.fcr.2009.02.009

SIRATAC (Hearn and daRoza, 1985) and OZCOT (Wells and Hearn, 1992; Hearn, 1994). With the development of eco-physiology studies in cotton, COTTAM (Jackson et al., 1988; Larson et al., 1996) was developed with more comprehensive inputs including climate condition, soil parameters, and management variables. Linking physiological and architectural models enhanced the cotton models in function, and the COTONS model developed in 1998 was a good example (Jallas et al., 2000; Hanan and Hearn, 2003). Moreover, many other models at that time showed their particular feature, such as COTCO2 (Wall et al., 1994), COTGROW (Pan et al., 1996), and the model developed by Reddy (1994). Now many researchers are still engaging in the area, with new advances made (Ko et al., 2005; Yang et al., 2008; Ravula et al., 2008). Cotton is not only the most important fibre crop in the world but also the second best potential source for plant proteins after soybean, the fifth oil-producing plant after soybean, palm-tree, colza, and sunflower (Sawan et al., 1988; Ahmad et al., 2007). Therefore, there is a need to understand the accumulating

W. Li et al. / Field Crops Research 112 (2009) 131–140

132

processes of oil and protein in cottonseeds. Cotton modeling has been well studied, but studies on simulating cotton seed growth have so far remained limited. To our knowledge, there is no report or communication regarding the modeling of the formation of cottonseed protein and oil. Both linear (Thaker et al., 1989) and cubic polynomial (Rabadia et al., 1999) models could be used to delineate cottonseed biomass accumulation. The nitrogen accumulation and the change of oil content after flowering had also been quantified (King and Leffler, 1979) by simple equations. However, none of them could express the changes in the accumulation under diverse environment conditions, nor explain the physiological mechanisms involved. Modelling of boll maturation period serves at the basis of the modelling of cottonseed growth and quality formation. Previous studies usually predicted crop development related to thermal time mainly determined by temperatures (Hearn and daRoza, 1985; Jamieson et al., 1998). Recently, Physiological Development Time (PDT) has been widely used in many crops such as wheat, rice, and cotton (Cao and Moss, 1997; Meng et al., 2003; Zhang et al., 2003), mainly taking into account variety maturity profile, temperature conditions, and photoperiod. With regard to cotton boll development, some studies have showed that the boll maturation period is also influenced by solar radiation and nitrogen nutrition (Chen and Yu, 2001; Reddy et al., 2004), in addition to genotype and temperature conditions. Abundant sunshine accelerates boll development as a subsidiary factor, especially when temperature conditions are sub-optimal (Chen and Yu, 2001; Ma et al., 2005). Excessive nitrogen nutrition has been found to delay cotton boll maturation (Tang et al., 2003; Reddy et al., 2004). So far, studies on modeling boll development have not integrated the factors of solar radiation and N nutrition. This paper presents the results of a first integrated model on cottonseed growth and development encompassing: (1) a module of cotton boll maturation period model taking into account the effects of cultivar characteristic, weather conditions, and crop management variables (precisely N supply); (2) module of cottonseed growth model dealing with cottonseed biomass accumulation in daily step; (3) module of cottonseed protein content and cottonseed oil content model based respectively on N accumulation and fat synthesis. 2. Materials and methods 2.1. Field data 2.1.1. Experiments for boll maturation period model module Field experiments were conducted at Jiangsu Agricultural Sciences Academy, Nanjing, Jiangsu, China (328020 N, 1188500 E) in 2006 and 2007. The soil at the experimental site was a yellowbrown loam with 25 mg kg1 organic matter, 12 g kg1 total N, 85.1 mg kg1 available N, 13.0 mg kg1 available P, and 91.6 mg kg1 available K contained in 20 cm depth of the soil profile. Two cultivars, Kemian 1(The growth period is 135 days) and NuCOTN 33B (The growth period is 120 days), were planted in 2006; Three cultivars,

Kemian 1, NuCOTN 33B and Dexia 1(The growth period is 100 days), were planted in 2007. Based on the research result that 240 kg ha1 is the optimum N rate in Nanjing (Xue et al., 2006), three N rates (0, 240, and 480 kg ha1) were applied in two equal amounts, before transplanting and at initial flowering stage. Cotton seeds were planted on 25 April in 2006 and 2007. When the seedlings had three true leaves, individual healthy and uniform plants were transplanted to the field at row spacing 0.9 m and plant spacing 0.3 m. The experiments were designed as randomized complete blocks with three replicates. Each plot was 6.3 m wide and 8.4 m long. In the experiments the boll flowering dates, opening dates, and boll position were recorded. 2.1.2. Experiments for cottonseed growth and quality formation model modules To study the cottonseed biomass accumulation, protein and oil content, a series of field experiments were carried out in Nanjing (328020 N, 1188500 E), Huaian (338180 N, 1198050 E), and Xuzhou (348120 N, 1178360 E) in Jiangsu Province, and Anyang (368040 N, 1148130 E) in Henan Province, China in 2005. The soil type and soil nutrient contents were listed out in Table 1. Two cultivars, Kemian 1 and NuCOTN 33B, were planted on 25 April and 25 May, at row spacing 0.9 m and plant spacing 0.3 m. Three N rates (0, 240, and 480 kg ha1) were applied in two equal amounts, before transplanting and at the initial flowering stage. The experiments were designed as randomized complete blocks with three replicates in Nanjing and four replicates in other locations. Crop management was in line with local cultivation practices followed for upland cotton production in the region. The white flowers at the first and second nodes of cotton plants were tagged on 15 July, 25 July, 10 August, 25 August, and 10 September. 10–15 pairs of boll tagged and its subtending leaf were picked each time at the 5th, 10th, 17th, 24th, 31th, 38th, 45th, and 52th day after flowering. The shells, seeds, and fiber were separated and then dried. The oil contents of cottonseeds were measured with Soxhlet extraction (Luque De Castro and Garcia-Ayuso, 1998), and the N concentrations of cottonseeds and subtending leaves were measured with Kjeldahl method (Feil et al., 2005). Cottonseed protein content = 6.25  N concentration. 2.2. Statistical analysis Field data collected in Nanjing in 2005 and 2006 were used to develop model and calibrate model parameters. The model was tested using independent field data collected in Nanjing in 2007 and in Xuzhou, Huaian and Anyang in 2005. Simulated values were compared with the observed values using the root mean square error (RMSE) calculated as: 2 30:5 n X ðY j  X j Þ2 5 4 RMSE ¼ n j¼1 where Yj = simulated value on the ith day, Xj = observed value on the ith day, n = number of pairs of simulated and observed value.

Table 1 Soil types and soil nutrient contents at the experimental sites in 2005. Site

Soil type

Total N content (g kg1)

Available N content (mg kg1)

Available P content (mg kg1)

Available K content (mg kg1)

Nanjing Xuzhou Huaian Anyang

Yellow brown Loam Alluvial soil Alluvial soil Sand loam

10.3 11.7 9.7 9.4

40.13 47.83 41.33 39.28

28.95 29.00 27.7 23.57

79.17 77.54 75.7 71.19

W. Li et al. / Field Crops Research 112 (2009) 131–140

133

where DTE means the daily temperature effect, SRE means the solar radiation effect. SRE is a linear function of solar radiation (SR, MJ m2) and can be calculated as:

3. Results 3.1. Model development This model was built to incorporate the current understanding of the basic eco-physiological processes involved in cottonseed growth and development. It included four modules: cotton boll maturation period module (BMPM), cottonseed biomass accumulation module (CSBAM), cottonseed protein content module (CSPCM), and cottonseed oil content module (CSOCM). It was developed by the hypothesis that the biomass accumulation, nitrogen accumulation, and fat synthesis are mainly sink-determined.

SREi ¼ 0:4 

SRi 30

(4)

Temperature is considered as the primary environmental factor in cotton boll development (Roussopoulos et al., 1998). According to the non-linear response curves of boll development to temperature (Plant et al., 1998), the relative temperature effect (RTE) is calculated by the Eq. (5), and the daily temperature effect (DTE) is calculated by Eq. (6).

8 0; > >  1þT o T=T o T b   > > > T c  T T c T o =T o T b < T  Tb  ; T T T  Tb RTEðTÞ ¼  o   T c T oc=T o Tob > > > T  Tb Tc  T > > ;  : Tc  To To  Tb

3.1.1. Cotton boll maturation period module (BMPM) This module predicts the developmental phases from flowering to maturity of cotton boll. The number of days having optimal climatic conditions needed to complete a development phase is defined here as physiological developmental time (PDT), which determines the maximal development rates. Physiological effect (PE) stands for the ratio of boll developmental rate in actual conditions to which in the optimum condition, which is calculated as a result of parameters of varietal maturity, weather, and crop management variables (precisely nitrogen supply). PEi ¼ VE  CETSi  ENbdi

(1)

where VE is a variety early maturity parameter, CETS means the comprehensive effect of temperature and solar radiation, and ENbd stands for the effect of nitrogen on boll development. In this paper, the subscript i always represents the days after flowering of the boll observed. VE stands for the relative development rate of the tested cultivar in the optimum condition, which can be estimated according to the ratio of the boll maturation period of the given early matured cultivar here to which of the tested cultivar. According to the observed values in present experiments, 35 days was assumed as the boll maturation period of the given cultivar under optimum growth condition. The comprehensive effect of temperature and solar radiation (CETS) can be calculated as: CETSi ¼ DTEi þ SREi  DTEi  SREi

(3)

(5)

To < T  Tc

 RTEðT min ÞÞ

(6)

where Tb, To, and Tc are the cardinal temperatures values (base, optimum, and ceiling temperature) for development. They are retained as 15, 30, and 35 8C, respectively (Chen et al., 2006). Tavg, Tmax, and Tmin mean the daily average, maximum, and minimum air temperature, respectively. The effect of nitrogen on boll development (ENbd) is defined by the actual subtending leaf nitrogen concentration (LNC, %) in relation to critical leaf N concentration (CLNC, %):

(2)

i¼1

Tb  T  To

DTE ¼ 0:5  RTEðT avg Þ  0:25  RTEðT max Þ  0:25

ENbdi ¼ 1 þ 0:3 

n X PEi PDT ¼

T > T c ; or; T < T b

CLNCi1  LNCi1 CLNCi1

(7)

According to the field experiment conducted in Nanjing in 2005, LNC decreases gradually following a negative exponential function with boll development (Fig. 1). CLNC is quantified according to the maximum LNC (Nmax, %), minimum LNC (Nmin, %), and boll development (indicated by PDT). Two semi-empirical formulas were developed respectively to indicate the change of LNC and CLNC with PDT: LNCCi ¼ ðNmax  Nmin Þ  expð0:02  PDTÞ þ Nmin

(8)

LNCi ¼ ELBPi  ESN  ðNmax  Nmin Þ  expð0:015  PDTÞ þ N min

(9)

where ELBP means the effect of boll position on LNC, and ESN means the effect of soil N supply on LNC. According to the responses of LNC to the boll position and soil N supply, the response functions were developed as follows:

8   ! > 24  PDT 2 > > > 1  0:3  1  ; > > 24 > > >    2 ! > > > > 1  0:3  1  GDD  380  1  24  PDT > ; > < 220 24 ELBPi ¼  2 ! > GDD  600 24  PDT > > > 1  0:5   1 ; > > 600 24 > > ! >   > > > 24  PDT 2 > > ; > : 1  0:5  1  24

GDD < 380 380  GDD < 600 (10) 600  GDD  1200 GDD > 1200

134

W. Li et al. / Field Crops Research 112 (2009) 131–140

growth rate (GR, g seed1 day1) is calculated as: GRi ¼ GRPi  CETSi  ENgri

(14)

where CETS stands for the comprehensive effect of temperature and solar radiation, written by Eq. (3). ENgr means the effect of nitrogen on cottonseed growth, calculated as: 8  < LNCi 0:5 ; LNCi < CLNCi (15) ENgri ¼ : CLNCi 1; LNCi  CLNCi Cottonseed dry weight (DW) is calculated as follows: DWi ¼ DWð0Þ þ

i X GRi

(16)

i¼1

Then the value of cottonseed index (CSI, g per 100 seeds) is obtained as 100 times of DW.

Fig. 1. Observed (symbols) and simulated (lines) subtending leaf N concentrations of cotton bolls of cv. Kemian 1 and cv. NuCOTN 33B at different flowering dates (from 15 July to 10 September) in Nanjing in 2005.

ESN ¼ 1  expð0:015  ðN soil  NAP  NRÞÞ

(11)

where GDD is degree days accumulated above the threshold on 15 8C from the first flowering date in the plant to present flowering date; Nsoil is the seasonal soil N supply of no fertilizer field (kg ha1), NAP is the N fertilizer rate(kg ha1), and NR is the N fertilizer use efficiency. 3.1.2. Cottonseed biomass accumulation module (CSBAM) In general, crop growth models simulated biomass accumulation driven by photosynthesis, and simulated the partitioning in organs with allocation coefficient. The source-sink relationship for cottonseed growth is difficult to quantify, so the present module simulates the cottonseed biomass accumulation by the way of logistic function and assumes that the potential growth is mainly sink-determined. The potential growth rate (GRP g day1) is calculated as follows: GRPi ¼ RGRi  DWi1   DWi RGRi ¼ r  1  DWmax

(12) (13)

3.1.3. Cottonseed protein content module (CSPCM) Cottonseed protein consists of the structural and storage protein. The structural protein mainly forms at the initial stage and the storage protein forms following. The field data shows a similar pattern of cottonseed protein content change during cottonseed growth between different cultivars or different flowering dates (Fig. 2). The change process is divided into two phases during cottonseed growth. At initial stage, the cottonseed protein content (PC, %) is comparatively high and declines gradually with PDT, when the nitrogen accumulation is mainly used for structural protein. The decline in cottonseed N concentration is a consequence of N dilution. At about 12 PDT, PC declines to the minimum. Then nitrogen accumulation enters into the second phase when PC is increasing gradually and the nitrogen is mainly used for storage protein. Total nitrogen accumulation (NA g seed1) is calculated as NAi ¼ DWð0Þ  NCð0Þ þ

i X ðNstri þ N stoi Þ

(17)

i¼1

where NC(0) (%) means the initial N concentration of cottonseed, Nstr (g seed1 day1) means daily accumulation of structural N, and Nsto (g seed1 day1) means daily accumulation of storage N. Nstr is simulated as a result of the cottonseed growth rate (GR, g day1) and the N concentration (NCstr, %) in daily cumulative biomass (Eq. (18)). The NCstr is calculated as a function of PDT and N supply (Eq. (19)). N stri ¼ GRi  NCstri

NCstri

8 > <

  ! PDTi 2  ENpri ; 0:016 þ 0:024  1  ¼ 12 > : 0:016;

(18)

PDT < 12 PDT  12 (19)

where ENpr stands for the effect of N supply on cottonseed N uptake, which is a result of LNC and CLNC:

  LNCi 2 ENpri ¼ (20) where RGR (g seed1 day1) is potential relative growth rate, and CLNCi DW (g seed1) is the dry weight per seed. The initial cottonseed dry weight (DW(0), g seed1), the maximum dry weight (DWmax, In the second phase, the potential accumulation rate of storage N (NPsto, g seed1) is determined by the demand of protein and the g seed1), and r are cultivar-specific parameters. Considering the effect of weather and crop management boll development. NPsto and the actual accumulation rate of measures (precisely N supply) on cottonseed growth, the actual storage N (Nsto) are calculated by: 8 0; PDT < 9 > >   < PDTi  9 2 (21) NPstoi ¼ ðDWi  NCmax  DWi1  NCi1 Þ  0:104  ; 9  PDT  25 > 16 > : ðDWi  NCmax  DWi1  NCi1 Þ  0:104; PDT > 25

W. Li et al. / Field Crops Research 112 (2009) 131–140

135

Fig. 3. Observed cottonseed oil content of cv. Kemian 1 and cv. NuCOTN 33B at different flowering dates (from 15 July to 10 September) in Nanjing in 2005. Fig. 2. Observed cottonseed protein content of cv. Kemian 1 and cv. NuCOTN 33B at different flowering dates (from 15 July to 10 September) in Nanjing in 2005.

Nstoi ¼ NPstoi  CETSi  ENpri  EBP

(22)

where NCmax (%) means the maximum N concentration of cottonseed, and NC (%) means the actual N concentration. N accumulation is closely linked to biomass accumulation, so actual storage N accumulation is also influenced by the impact factors which were used in CSBAM. EBP means the effect of boll position, which can be calculated by degree days accumulated from the first flowering date in the plant to present flowering date. EBP linearly increase from 0.9 to 1 with the increasing of degree days from 0 to 400. Cottonseed N concentration (NC) is the ratio of N accumulation to cottonseed biomass. The protein content (PC) is calculated as 6.25 times of N concentration. 3.1.4. Cottonseed oil content module (CSOCM) Cottonseed oil content is mainly under genetic control, but it can be affected to some extent by environmental conditions. The materials required for fat synthesis mainly derived from carbohydrates accumulated in cottonseed. Nitrogen is an essential nutrient for the synthesis of fat which requires both N and carbon skeletons (Egelkraut et al., 2004). The change process of cottonseed oil content is divided into two phases during the cottonseed growth (Fig. 3). In the initial stage, oil content decline slightly, as a consequence of the rapid biomass accumulation and slow fat synthesis. With the acceleration in fat synthesis, the oil content stops declining at about 12–15 PDT and then increases gradually following logistic curve. In the module, it is assumed that daily potential oil accumulation (DPOA) is mainly determined by the fat synthesis capability (FSC) and oil demand. The oil demand can be expressed by sink activity (SA). Daily actual oil accumulation (DAOA) is influenced by

temperature and nitrogen, and also associated with boll position. They are written: DPOAi ¼ SAi  FSCi

(23)

DAOAi ¼ DPOAi  EToil  ENoili  EBP

(24)

where EToil means the effect of temperature, ENoil the effect of nitrogen, and EBP the effect of boll position on fat synthesis. With the increasing of oil content, SA decreases gradually till the time when oil content gets the maximum. FSC also changes with oil accumulation and seed development. SA and FSC can be written as SAi ¼

DWi  OCm  TOAi1 DWi  OCm  DWð0Þ  OCð0Þ

(25)

8 PDT < 6; or; PDT > 36 < 0;   PDT  6 (26) FSCi ¼ : TOAi1 pSIN p ; 6  PDT  36 30 where OCm means the maximum oil content (%), OC(0) means the initial oil content(%), and TOA means the total oil accumulation per seed (g seed1). OCm, OC(0) and p are cultivar-specific parameters. Fat synthesis is a complex biological process strongly influenced by temperature. The temperature response function here is different with the one for cottonseed growth: 8 0; T < 10; or; T > 35 > > > > T  10 > > ; 10  T < 20 > < 30 T  15 (27) EToil ðTÞ ¼ ; 20  T < 30 > > > 30  15 > > > T  30 > : ; 30  T  35 35  30 EToili ¼ 0:5  EToil ðT avg Þ þ 0:25  ðEToil ðT max Þ þ EToil ðT min ÞÞ

(28)

136

W. Li et al. / Field Crops Research 112 (2009) 131–140

Fig. 4. Comparison between simulated and observed boll maturation period of cv. Dexia 1, cv. Kemian 1, and cv. NuCOTN 33B in Nanjing in 2007.

where EToil(T) means instantaneous effect of temperature, and EToil means daily relative effect of temperature. The nitrogen effect on oil synthesis (ENoil) is calculated by the LNC in relation to CLNC: 8 LNCi > > LNCi  0:85  CLNCi < 0:85  CLNC ; i   (29) ENoili ¼ > LNCi 1 > :1  ; 0:85  CLNCi < LNCi 0:85  0:85 CLNCi

Then the total oil accumulation (TOA, g seed1) can be written as:

TOAi ¼ DWð0Þ  OCð0Þ þ

i X DAOAi

(30)

i¼1

The oil content (OC, %) is the ratio of the TOA to cottonseed biomass.

Fig. 5. Comparison between simulated (lines) and observed (symbols) cottonseed biomass accumulation of cv. Kemian 1 and cv. NuCOTN 33B at different flowering dates and sowing dates (25 April (a) and 25 May (b)) under three N rates (^, - - - 0 kg N ha1; &, – – – 240 kg N ha1; and ~, — 480 kg N ha1) in Xuzhou, Anyang, and Huaian, China.

W. Li et al. / Field Crops Research 112 (2009) 131–140

137

Table 2 Parameter values used in the model. Parameter

LNCmax LNCmin DWm DW(0) r NCm NC(0) OCm OC(0) p

Explanation

The maximum leaf N concentration The minimum leaf N concentration The maximum cottonseed dry weight The initial cottonseed dry weight Reproductive potential parameter for cottonseed The maximum cottonseed N concentration The initial cottonseed N concentration The maximum cottonseed oil content The initial cottonseed oil content Reproductive potential parameter for cottonseed oil

3.2. Model calibration Values for the model parameters were determined from the literature or by calibration on experimental data. The parameters for BMPM were calibrated with the data from the field experiment conducted in Nanjing in 2006, and the other modules with the data from the field experiment conducted in Nanjing in 2005. Based on the observed data of boll maturation period in present study, the varietal early maturity parameters (VE) were estimated as 0.972 for cv. Dexia 1, 0.933 for cv. Kemian 1, and 0.919 for cv. NuCOTN 33B. Table 2 listed out the parameter values for CSBAM, CSPCM,

Value

Unit

Kemian 1

NuCoTN 33B

4.853 2.1 0.12 0.005 0.17 4.5 4.0 18.0 4.0 0.5

4.853 2.1 0.13 0.005 0.155 4.2 4.0 22.0 4.0 0.5

(%) (%) (g seed1) (g seed1) (day1) (%) (%) (%) (%) (day1)

and CSOCM. LNCmax and LNCmin were set by the reference of Pan et al. (1996), and the others were set directly by the field data or by the model calibration using field data. In the study, only two or three cultivars were used for the model calibration. Adjusting the cultivar-specific parameters in the model is needed to make the model suitable for more other cultivars. 3.3. Model validation The accuracy of the boll maturation period module was tested using the independent data from a field experiment conducted

Fig. 6. Comparison between simulated (lines) and observed (symbols) cottonseed protein content of cv. Kemian 1 and cv. NuCOTN 33B at different flowering dates and different sowing dates (25 April (a) and 25 May (b)) under three N rates (^, - - - 0 kg N ha1; &, – – – 240 kg N ha1; and ~, — 480 kg N ha1) in Xuzhou, Anyang, and Huaian, China.

138

W. Li et al. / Field Crops Research 112 (2009) 131–140

with different cultivars and different N rates in Nanjing in 2007. The observed boll maturation period (BMP) widely ranged from 40 to 65 days for different flowering dates. Results showed good coherence between the predicted BMP values and the observed BMP values with RMSE of 2.25 days for cv. Dexia 1, 2.61 days for cv. Kemian 1, and 2.75 days for cv. NuCOTN 33B (Fig. 4). CSBAM, CSPCM, and CSOCM were tested using the data sets from the field experiments in three locations (Huaian, Xuzhou, and Anyang) belonging to two cotton growing areas in China. Cottonseed biomass showed a wide range of values in accumulation rate under diverse environmental conditions. Cottonseed growth rate in Xuzhou was higher than other sites, in cv. Kemian 1 was higher than cv. NuCOTN 33B (Fig. 5). CSBAM showed good consistency with the RMSE of 9.5 mg seed1 for cv. Kemian 1, 8.2 mg seed1 for cv. NuCOTN 33B, and 8.9 mg seed1 on average. The change of protein content during the cottonseed growth was obviously different at different flowering dates (Fig. 6). The cottonseed protein content variation in most of the bolls flowering before 25 August appeared in V-shape curve, while for those bolls which flowered later expressed little or no increasing trend at the late stage of development. The CSPCM also worked well with RMSE of 2.05% for cv. Kemian 1, 2.33% for cv. NuCOTN 33B, and 2.19% on average. Cottonseed oil content changed with a S-shaped curve during the cottonseed growth and the maximum oil contents were different at different treatments (Fig. 7). The CSOCM could well describe the oil content with RMSE of 2.45% for cv. Kemian 1, 2.95% for cv. NuCOTN 33B, and 2.71% on average. Fig. 7 shows that the CSOCM did slightly poorly in the later stage of boll development for

bolls flowering on 10 September, where the simulated values were lower than the observed ones. 4. Discussion Accurately predicting boll maturation period is the basis to simulate cottonseed growth and quality. The growth-degree-day was a common and simple method when one refers to temperature only but lacks accuracy. To improve the accuracy of prediction, cotton model COTCO2 (Wall et al., 1994) adapted physiology time to predict cotton development considering more environmental factors. Complex calculation procedure and excessive number of model parameters were its main defects. PDT driven by temperature and variety maturity profile has been used in modeling BMP (Zhang et al., 2003). Ma et al. (2005) further exploited this path by introducing sunshine hours in boll maturation period model. Nevertheless PDT model remains imperfect in considering the environment factors as sunshine hours, but cannot perfectly reflect the sunlight condition. In this study, the solar radiation and N nutrition factors are taken into account besides temperature and varietal maturity. Hence, the present model is more comprehensive and systemic. This module predicted BMP more precisely with a lower RMSE of 2.54 days on average. Cottonseed index, protein and oil content are the main criteria of cottonseed quality and they differ between cultivars (Gotmare et al., 2004; Mert et al., 2005). Environmental conditions also strongly influenced the seed quality (Turner et al., 1976). Some cultivar-specific parameters and response functions are

Fig. 7. Comparison between simulated (lines) and observed (symbols) cottonseed oil content of cv. Kemian 1 and cv. NuCOTN 33B at different flowering dates and sowing dates (25 April (a) and 25 May (b)) under three N rates ( , - - - 0 kg N ha1; &, – – – 240 kg N ha1; and ~, — 480 kg N ha1) in Xuzhou, Anyang, and Huaian, China.

W. Li et al. / Field Crops Research 112 (2009) 131–140

incorporated in the present model to illustrate the effects of cultivars, weather, soil, and crop management measures (such as N supply). This paper combined with the physiological knowledge and developed a mechanistic simulation model of cottonseed quality formation. The model was based on the hypothesis that the potential cottonseed biomass accumulation, N accumulation, and oil synthesis are sink-determined. The hypothesis of ‘‘sinkdetermination’’ has also been used in modeling organ growth or nutrients accumulation for many other crops, such as tomato for dry matter distribution (Heuvelink, 1996), wheat for protein content (Martre et al., 2006) and for nitrogen uptake (Jamieson and Semenov, 2000). Our study addressed the cottonseed potential growth rate by logistic function which has been widely used in modeling crop biomass accumulation (Bange and Milroy, 2004). The major features of the logistic method are simple structure, understandability, and stability in running. Logistic model was a suitable method in modeling cottonseed biomass accumulation. Integrated with response functions of environment conditions, the model predicted cottonseed dry weight with RMSE of 8.9 mg. Combined with the physiological knowledge in cottonseed N uptake, the present model simulated the change of cottonseed protein content and maintained a high accuracy with RMSE of 2.19%. The model simulated oil potential accumulation through oil demand and synthesis capability. The result in modeling oil content is validated with RMSE of 2.71%. Adequate N supply can raise cottonseed biomass and protein content, and lower oil content (Sawan et al., 1985, 2001, 2006). The subtending leaf of cotton boll is the main source organ for the boll growth, and the N concentration in subtending leaf directly influence the cottonseed growth and development. It is hence more relevant to link the cottonseed growth and development to the N concentration in subtending leaf rather than the whole plant N concentration or N rate. Our study simulated the effect of N nutrition using the subtending leaf N concentration as a direct indicator, and developed a semi-empirical model to illustrate the change of actual and critical N concentration in subtending leaf during the cottonseed growth. The quantitative studies in critical N concentration of crops were rich (Pan et al., 2006; Keating et al., 1999). Xue et al. (2006) also qualified the cotton plant critical N concentration. But few reports concerned the subtending leaf critical N concentration for cottonseed growth and development. By using different response functions, the N nutrition effect was considered in each module of the model. In comparison to previous studies which have quantified the relationship of cottonseed N concentration and N rates by linear equations (Egelkraut et al., 2004), the present model appears to be more reasonable. The field data were obtained from four locations in 2005, and provided a wide range of environmental conditions and crop characteristics for validation purposes. The validation led to good consistency between the simulated and the observed values, indicating that the model is robust. However, the simulated values of cottonseed oil content were inferior to observed ones in the late stage of development for late flowering bolls. This phenomenon occurs in November when the meteorological conditions were much less favorable. This is an indication that the model cannot work well to simulate cottonseed oil content under extreme environmental conditions. The verification of the model on a broader set of data remains needed, through additional field experiments in more regions with more cultivars. Efforts to finetune the model should be concentrated on the formation of cottonseed quality in more variable environmental conditions. 5. Conclusion The model we have developed are capable of predicting boll maturation period and simulating cottonseed growth, protein and

139

oil content driven by the inputs of cultivar parameters, weather, and crop management variables (N fertilization). Our study improved cotton boll maturation model by quantifying the response to solar radiation and N supply. This model structure was based on the processes of biomass accumulation, N uptake and fat synthesis in cottonseed. The responses of cottonseed growth and quality formation to weather or crop management were quantified respectively. The field data set conducted in wide range of climate conditions with different cultivars cotton planted on different dates was used for model validation. The result showed that the model is robust enough to accurately predict cotton boll maturation period, cottonseed biomass, protein and oil content, with the RMSE of 2.54 days, 8.9 mg, 2.19% and 2.71% on average, respectively. Its main advantages lie upon comprehensive integration of inputs as compared to existing models, reasonable model algorithms, higher accuracy, and application to a diverse range of environment conditions. The model can be not only used independently to predict cottonseed quality formation, but also used to predict the cottonseed, protein and oil yield by integration with a cotton growth model. Acknowledgments This work was funded by the National Natural Science Foundation of China (Nos. 30771277 and 30771279). References Ahmad, S., Anwar, F., Hussain, A.I., Ashraf, M., Awan, A.R., 2007. Dose soil salinity affect yield and composition of cottonseed oil? J. Am. Oil Chem. Soc. 84, 845– 851. Bange, M.P., Milroy, S.P., 2004. Growth and dry matter partitioning of diverse cotton genotypes. Field Crop Res. 87, 73–87. Boone, M.Y., Porter, D.O., McKinion, J.M., 1993. Calibration of GOSSYM: theory and practice. Comput. Electron. Agric. 9, 193–203. Cao, W., Moss, D.N., 1997. Modelling phasic development in wheat: a conceptual integration of physiological components. J. Agric. Sci. 129, 163–172. Luque De Castro, M.D., Garcia-Ayuso, L.E., 1998. Soxhlet extraction of solid materials: an outdated technique with a promising innovative future. Anal. Chim. Acta 369, 1–10. Chen, G., Yu, Y., 2001. Preliminary study on the temperature-light effects on boll development. Cotton Sci. 13, 63–64 in Chinese, with English abstract. Chen, B., Cao, W., Zhou, Z., 2006. Simulation and validation of dry matter accumulation and distribution of cotton boll at different flowering stages. Sci. Agric. Sin. 39, 487–493 in Chinese, with English abstract. Duncan, W.G., 1973. SIMCOT: a simulation of cotton growth and yield. In: Murphy, C., Hesketh, J.D., Strain, B. (Eds.), Modelling the Growth of Trees. Oak Ridge National Laboratory, Oak Ridge, pp. 115–118. Egelkraut, T.M., Kissel, D.E., Cabrera, M.L., Gascho, G.J., Adkins, W., 2004. Nitrogen concentration in cottonseed as an indicator of N availability. Nutr. Cycl. Agroecosyst. 68, 235–242. Feil, B., Moser, S.B., Jampatong, S., Stamp, P., 2005. Mineral composition of the grains of tropical maize varieties as affected by pre-anthesis drought and rate of nitrogen fertilization. Crop Sci. 45, 516–523. Fye, R.E., Reddy, V.R., Baker, D.N., 1984. The validation of GOSSYM: Part 1—Arizona conditions. Agr. Syst. 14, 85–105. Gotmare, V., Singh, P., Mayee, C.D., Deshpande, V., Bhagat, C., 2004. Genetic variability for seed oil content and seed index in some wild species and perennial races of cotton. Plant Breed. 123, 207–208. Hanan, J.S., Hearn, A.B., 2003. Linking physiological and architectural models of cotton. Agric. Syst. 75, 47–77. Hearn, A.B., 1994. OZCOT: a simulation model for cotton crop management. Agric. Syst. 44, 257–299. Hearn, A.B., daRoza, G.D., 1985. A simple model for crop management 1 applications for cotton 2 (Gossypium hirsutum L). Field Crops Res. 12, 49–69. Heuvelink, E., 1996. Dry matter partitioning in tTomato: validation of a dynamic simulation model. Ann. Bot. 77, 71–80. Jackson, B.S., Arkin, G.F., Hearn, A.B. (Eds.), 1988. The cotton simulation model ‘‘COTTAM’’: fruiting model calibration and testing. Trans. ASAE 31, 846–854. Jallas, E., Martin, P., Sequeira, R., Turner, S., Cretenet, M.G.E., 2000. Virtual COTONS1, the firstborn of the next generation of simulation model. Virtual Worlds 1834, 235–244. Jamieson, P.D., Semenov, M.A., 2000. Modelling nitrogen uptake and redistribution in wheat. Field Crops Res. 68, 21–29. Jamieson, P.D., Semenov, M.A., Brooking, I.R., Francis, G.S., 1998. Sirius: a mechanistic model of wheat response to environmental variation. Eur. J. Agron. 8, 161–179.

140

W. Li et al. / Field Crops Research 112 (2009) 131–140

Keating, B.A., Robertson, M.J., Muchow, R.C., Huth, N.I., 1999. Modelling sugarcane production systems I. Development and performance of the sugarcane module. Field Crops Res. 61, 253–271. King, E.E., Leffler, H.R., 1979. Nature and patterns of proteins during cotton seed development. Plant Physiol. 63, 260–263. Ko, J., Maas, S.J., Lascano, R.J., Wanjura, D., 2005. Modification of the GRAMI model for cotton. Agron. J. 97, 1374–1379. Larson, J.A., Mapp, H.P., Verhalen, L.M., Banks, J.C., 1996. Adapting a cotton model for decision analyses: a yield response evaluation. Agric. Syst. 50, 145–167. Ma, F., Cao, W., Zhang, L., Zhu, Y., Li, S., Zhou, Z., Li, C., Xu, L., 2005. A physiological development time-based simulation model for cotton development stages and square and boll formation. Chin. J. Appl. Ecol. 16, 626–630 in Chinese, with English abstract. Martre, P., Jamieson, P.D., Semenov, M.A., Zyskowski, R.F., Porter, J.R., Triboi, E., 2006. Modelling protein content and composition in relation to crop nitrogen dynamics for wheat. Eur. J. Agron. 25, 138–154. Meng, Y., Cao, W., Zhou, Z., Liu, X., 2003. A process-based model for simulating phasic development and phenology in rice. Sci. Agric. Sin. 36, 1362–1367 in Chinese, with English abstract. Mert, M., Aki, Y., Gen, O., 2005. Genotypic and phenotypic relationships of lint yield, fibre properties and seed content in a cross of two cotton genotypes. Acta Agric. Scand. B: Soil Plant Sci. 55, 76–80. Mutsaers, H.J.W., 1984. KUTUN: a morphogenetic model for cotton (Gossypium hirsutum L). Agric. Syst. 14, 229–257. Pan, X., Han, X., Shi, Y., 1996. A cotton growth and development simulation model for culture management—COTGROW. Sci. Agric. Sin. 29, 94–96 in Chinese, with English abstract. Pan, J., Zhu, Y., Jiang, D., Dai, T., Li, Y., Cao, W., 2006. Modeling plant nitrogen uptake and grain nitrogen accumulation in wheat. Field Crops Res. 97, 322–336. Plant, R.E., Kerby, T.A., Zelinski, L.J., Munk, D.S., 1998. A qualitative simulation model for cotton growth and development. Comput. Electron. Agric. 20, 165–183. Rabadia, V.S., Thaker, V.S., Singh, Y.D., 1999. Relationship between water content and growth of seed and fibre of three cotton genotypes. J. Agron. Crop Sci. 183, 255–261. Ravula, P.P., Grisso, R.D., Cundiff, J.S., 2008. Cotton logistics as a model for a biomass transportation system. Biomass Bioenerg. 32, 314–325. Reddy, V.R., 1994. Modeling cotton growth and phenology in response to temperature. Comput. Electron. Agric. 10, 63–73. Reddy, V.R., Baker, D.N., 1988. Estimation of parameters for the cotton simulation model GOSSYM: cultivar differences. Agric. Syst. 26, 111–122.

Reddy, V.R., Baker, D.N., 1990. Application of GOSSYM to analysis of the effects of weather on cotton yields. Agric. Syst. 32, 83–95. Reddy, K.R., Koti, S., Davidonis, G.H., Reddy, V.R., 2004. Interactive effects of carbon dioxide and nitrogen nutrition on cotton growth, development, yield, and fiber quality. Agron. J. 96, 1148–1157. Roussopoulos, D., Liakatas, A., Whittington, W.J., 1998. Controlled-temperature effects on cotton growth and development. J. Agric. Sci. 130, 451–462. Sawan, Z.M., El-Kasaby, A.T., Sallouma, B.M., 1985. Effect of plant density, nitrogen fertilization and growth regulators on cottonseed yield and seedling vigour. J. Agron. Crop Sci. 154, 120–128. Sawan, Z.M., El-Farra, A.A., El-Latif, S.A., 1988. Cottonseed, protein and oil yields, and oil properties as affected by nitrogen and phosphorus fertilization and growth regulators. J. Agron. Crop Sci. 161, 50–56. Sawan, Z.M., Hafez, S.A., Basyony, A.E., 2001. Effect of nitrogen fertilization and foliar application of plant growth retardants and zinc on cottonseed, protein and oil yields and oil properties of cotton. J. Agric. Crop Sci. 186, 183–191. Sawan, Z.M., Saeb, A., Hafez, A.E.B., Alkassas, A.R., 2006. Cottonseed, protein, oil yields and oil properties as affected by nitrogen fertilization and foliar application of potassium and a plant growth retardant. World J. Agric. Sci. 2, 56–65. Tang, X., Tian, C., Lu, Z., 2003. Relationship between the spread quantity of nitrogenous fertilizer and the accumulated sunshine requirement of cotton in its growing stages. Arid Zone Res. 20, 226–229 in Chinese, with English abstract. Thaker, V.S., Saroop, S., Vaishnav, P.P., Singh, Y.D., 1989. Genotyopic variations and influence of diurnal temperature on cotton fibre development. Field Crops Res. 22, 129–141. Turner, J.H., Ramey, H.H., Worley, J.S., 1976. Influence of environment on seed quality of four cotton cultivars. Crop Sci. 16, 407–409. Wall, G.W., Amthor, J.S., Kimball, B.A., 1994. COTCO2: a cotton growth simulation model for global change. Agric. For. Meteorol. 70, 289–342. Wells, A.T., Hearn, A.B., 1992. OZCOT: a cotton crop simulation model for management. Math. Comput. Simulat. 33, 433–438. Xue, X., Zhou, Z., Zhang, L., Wang, Y., Guo, W., Chen, B., 2006. Development and application of critical nitrogen concentration Dilution Model for cotton after flowering. Acta Ecol. Sin. 26, 1781–1791 in Chinese, with English abstract. Yang, Y., Ouyang, Z., Yang, Y., Liu, X., 2008. Simulation of the effect of pruning and topping on cotton growth using COTTON2K model. Field Crops Res. 106, 126–137. Zhang, L., Cao, W., Zhang, S., Luo, W., 2003. Simulation model for cotton development stage based on physiological development time. Cotton Sci. 15, 97–103 in Chinese, with English abstract.