Vous êtes sur la page 1sur 3

Rodrigues Junior, F.A. et al. Proc. Int. Soc. Sugar Cane Technol., Vol.

28, 2013 ______________________________________________________________________________________

USING LEAF NUTRIENTS TO MODEL SUGARCANE QUALITY PARAMETERS By F.A. RODRIGUES JUNIOR1, P.S.G. MAGALHES1,2, H.C.J. FRANCO2 and D.G.P. CERRI1
1

School of Agriculture Engineering University of Campinas UNICAMP, Campinas/SP, 13083-875 Brazil 2 Brazilian Bioethanol Science and Technology Laboratory CTBE, Campinas/SP, CEP 13083-970 Brazil graziano@feagri.unicamp.br

KEYWORDS: Precision Agriculture, Multiple Regressions, Spatial Variability. Abstract THE USE OF MODELS to describe the relationship between quality parameters and leaf nutrients of sugarcane may be important for the development of on-the-go sensors. The objective of this work was to fit mathematical models to sugarcane Brix, pol and fibre using leaf nutrients as predictors. This work was carried out in Araras, Brazil, during the first and second ratoons of a crop. Macronutrient contents were analysed from leaves collected before harvest. The variables used in the models were chosen using a stepwise procedure, and models fitted through multiple regressions. The Brix, pol and fibre models gave an R2 of 0.18, 0.08 and 0.15 for the first ratoon and 0.21, 0.21 and 0.29 for the second ratoon, respectively. These models were statistically significant at the 5% level of probability, but showed low coefficients of determination. However, the values of the root mean square errors were low as well. Introduction To improve the efficiency of ethanol production, it is necessary not only to improve yield, but also to increase the quality of the sugarcane. Quality, according to Mutton and Mutton (2005) and Mutton (2008), is influenced by a complex of many factors influenced by soil fertility and fertilisation management, adequate control of weeds, pests and diseases, as well as climatic and weather conditions. Crop yield forecasting, often important for management purposes, is usually done by biometry or satellite imagery. However, there are several factors that could induce reduced biomass accumulation and, consequently, lead to underestimation of crop yield. Four main factors are lodging, low leaf nitrogen, inhibition of photosynthesis by high sugar content in mature or winterripened sugarcane, and high respiratory demand (van Heerden et al., 2010). Kumar and Verma (1997) applied multiple regression analyses to leaf nutrient levels, sugarcane yield and juice quality parameters. They observed that the amounts of N, P, K, Zn and Cu explained 93% of the variation in sucrose levels and that the leaf quantities of N, P, K and Cu explained 95% of the variation in commercial sugar content. Furthermore, they claimed that, under the conditions of the experiment, leaf nutrient analysis could be used to predict sugar and cane yield. The use of correlation models between quality parameters of sugarcane and leaf nutrients may be important in understanding the relationship among those parameters and may help to rationalise inputs, especially through variable rate fertilisation, and to increase productivity. There
1

Rodrigues Junior, F.A. et al. Proc. Int. Soc. Sugar Cane Technol., Vol. 28, 2013 ______________________________________________________________________________________

are several models available for yield estimation in response to management. APSIM (Keating et al., 2003) allows prediction of the long-term consequences of farming practices on the soil resource, but we were looking for a simpler method to allow on-the-go prediction of yield and sugar quality. Our main goal was to analyse correlations among leaf nutrients and sugarcane quality parameters of Brix, pol and fibre. We used multivariate analysis (stepwise procedure) to select the important variables and we present a mathematical model to explain the variation of those parameters. Materials and methods We used data from a sugarcane field (10 ha) in Araras, So Paulo State, collected during two consecutive growth years: December 2009 (first ratoon) and July 2011 (second ratoon). Sugarcane, variety SP80-3280, was planted in 2007 and was mechanically green-harvested during each of the cropping seasons. No fertilisers were applied to the cane. The area was divided into a regular 30-m grid (totalling 117 sample points) using Pathfinder Office software (Trimble). Sampling points were located using a GPS GeoExplorer 3 (Trimble) device. Plant samples were collected at each point just prior to the harvest to determine the sugarcane quality parameters. We collected 10 stalks at random in 2 m lengths of the same row. The third leaves of those plants were collected for determination of the leaf macronutrients nitrogen (N), phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg) and sulfur (S) by standard laboratory methods. Conventional descriptive statistical analyses of the samples were performed as a first approach for the evaluation of the parameters. Skewness and kurtosis indices, together with normality test (Kolmogorov-Smirnov test), were calculated to evaluate the quality of the data distribution (Table 1). Prior to the modelling process, we selected 70 of the points to run the variable-selection process and create the models. The models were then applied to the entire dataset of 117 points. Variables were selected using forward stepwise multiple regression with a probability coefficient of 0.25 for inclusion of variables into the model. Independent variables in the selection and modelling process were the levels of leaf macronutrients. The independent variables were used to model the quality parameters for the same crop cycle. After the selection process, the models were fitted by means of multiple regressions using the standard least squares method for Brix, pol and fibre.
Table 1Descriptive statistics for Brix, pol, fibre and leaf macronutrients.
Parameter Brix Pol Fibre N P K Ca Mg S Brix Pol Fibre N P K Ca Mg S Min 19.3 14.3 11.5 7.3 1.0 8.3 2.7 0.8 1.1 16.4 11.3 10.3 7.2 7.9 0.8 2.8 1.0 1.1 Med 20.8 15.9 12.4 9.6 1.7 14.4 4.1 1.3 1.6 19.2 14.5 11.5 10.8 12.5 1.3 3.6 1.5 1.4 Max 22.0 16.9 15.3 18.8 2.5 19.6 6.9 3.3 3.2 21.4 16.4 14.7 18.7 16.9 1.8 8.4 2.4 1.7 Mean 2009 20.7 15.9 12.4 9.6 1.7 14.4 4.2 1.4 1.7 2011 19.1 14.4 11.7 10.8 12.4 1.3 3.8 1.5 1.4 SD 0.5 0.5 0.4 1.2 0.3 2.2 0.8 0.4 0.3 0.8 0.8 0.8 1.4 1.7 0.1 0.8 0.2 0.1 CV 2.6 3.3 3.6 12.2 18.6 15.5 19.6 24.9 18.7 4.1 5.8 7.2 1.9 13.9 13.9 22.2 15.3 9.7 sk 0,36 0.59 2.07 4.00 0.30 0.05 0.67 1.87 1.55 0.28 0.59 0.93 1.41 0.04 0.15 1.89 0.37 0.29 k 0.31 0.41 12.68 30.56 0.04 0.05 0.03 6.43 4.81 0.76 1.17 1.14 7.55 0.10 0.03 5.83 0.54 0.03 p-value 0.02 >0.15 >0.15 <0.01 <0.01 >0.15 <0.01 <0.01 <0.01 >0.15 >0.15 <0.01 <0.01 >0.15 <0.01 <0.01 0.02 <0.01

min minimum; med median; SD standard deviation; CV coefficient of variation; sk skewness; k kurtosis; p-value for normality test. Brix, pol and fibre in (%); N, P, K, Ca, Mg 1 and S in (g kg ).

Rodrigues Junior, F.A. et al. Proc. Int. Soc. Sugar Cane Technol., Vol. 28, 2013 ______________________________________________________________________________________

Results and discussion The forward stepwise procedure resulted in models between the dependent variables Brix, pol or fibre and the six initial leaf macronutrients (Table 2) in each of the two years. All of the models were statistically significant at the 5% level, but explained only 18 and 21% of the Brix variation, 8 and 21% of the pol variation, and 15 and 29% of the fibre variation for the first and second ratoon, respectively. However, all the root mean square errors (RMSE) ranged between 0.34 and 0.68%, being no more than 1% of the respective variables. Except for K in the regressions for Brix and N, S and Ca in the regressions for fibre, the variables were not repeated in each of the two regressions for the same variable. Leaf sampling was done in both seasons following the peak of the maturity and with about the same growth period. However, the first-ratoon sampling was done in November at the beginning of the rainy season, 14 months after harvesting, and second-ratoon sampling was done in July in the middle of the dry season, 15 months after harvesting. This was because the first ratoon was harvested only at the end of March 2010 as a standover crop.
Table 2Model coefficients for Brix, pol and fibre.
Coef. 21.5 0.582 0.037 0.337 0.446 R Fcalc RMSE
2

Brix Interc. P* ns K ns Mg ns S 0.185 0.009* 0.446

1st ratoon (2009) Coef. Pol Coef. Leaf data 2009 16.6 Interc. 12.4 0.411 P* 0.130 0.145 0.399 0.080 0.015* 0.454

Fibre Interc. N* Ca* S* 0.155 0.010* 0.349

Coef. 21.8 1.984

Brix Interc. K*

2nd ratoon (2011) Coef. Pol Leaf data 2011 18.3 Interc. 1.712 K* ns 1.186 S

Coef. 10.2 0.116 1.724 0.274 0.440 2.073

Fibre Interc. ns N K* Ca* ns Mg S* 0.295 0.0001* 0.619

R Fcalc RMSE

0.215 0.0001* 0.683

0.211 0.0001* 0.675


2

Where: Brix, pol and fibre in (%); coef. Model coefficient; Interc. Model intercept; R coefficient of ns determination; Fcalc F test; RMSE root mean square error; * statistically significant at 5%; not statistically significant.

Conclusions Stepwise regression allowed the identification of the most relevant leaf attributes that may influence crop-quality parameters. Although the coefficients of determination were below 30% for all quality parameters (Brix, pol, fibre), the root mean square error was very low (< 1%). This indicates that the identified macronutrients play an important role in determining sugarcane quality parameters and they may provide the basis for on-the-go sensors REFERENCES Keating, B.A., Carberry, P.S., Hammer, G.L. et al. (2003). An overview of APSIM, a model designed for farming systems simulation. Europ. J. Agron., 18: 267288. Kumar, V. and Verna, K.S. (1997). Relationship between nutrient element content of the index leaf and cane yield and juice quality of sugarcane genotypes. Commun. Soil Sci. Plant Anal., 28: 10211032. Mutton, M.J.R. (2008). Reflexos da qualidade da matria-prima sobre a fermentao etanlica. In: Workshop produo de etanol: qualidade da matria prima, 14., 2008, Lorena. Painel 2... Lorena: Escola de Engenharia de Lorena University of Sao Paulo, p. 15. Mutton, M.J.R. and Mutton, M.A. (2005). Identificao de perdas de acares no setor agrcola. STAB Acar, lcool & Subprodutos, Piracicaba, 23(4): 4246. Van Heerden, P.D.R., Donaldson, R.A., Watt, D.A. and Singels, A. (2010). Biomass accumulation in sugarcane: unravelling the factors underpinning reduced growth phenomena. J. Exper. Bot., 61: 28772887.
3

Vous aimerez peut-être aussi