SciELO - Scientific Electronic Library Online

vol.46 issue3Enological Frontier of the Southern Cone of America: Cabernet-Sauvignon in Chile and Malbec in ArgentinaEffect of pre- and postpartum body development on the rebreeding of first-calf heifers author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand




Related links

  • On index processCited by Google
  • Have no similar articlesSimilars in SciELO
  • On index processSimilars in Google


Ciencia e investigación agraria

On-line version ISSN 0718-1620

Cienc. Inv. Agr. vol.46 no.3 Santiago Dec. 2019 

Research Paper

Animal Science

Calibration models for the nutritional quality of fresh pastures by near-infrared reflectance spectroscopy

Modelos de calibración para la cuantificación nutricional de praderas frescas mediante espectroscopía de infrarojo cercano

Iris Lobos1 

Cristian J. Moscoso1 

Paula Pavez1 

1Instituto de Investigaciones Agropecuarias INIA, Remehue. P.O Box 24-0, Osorno, Chile.


High levels of animal performance and health depend on high-quality nutrition. Determining forage quality both reliably and quickly is essential for improving animal production. The present study describes the use of near infrared reflectance spectroscopy (NIRS) for the quantification of nutritional quality (dry matter (DM), water-soluble carbohydrates (WSC), crude protein (CP), in vitro dry matter digestibility (DMD), organic matter digestibility (OMD), neutral detergent fiber (NDF) and the WSC/CP ratio) in samples from fresh pastures in southern Chile (39° to 40° S). Calibration models were developed with wet chemistry and NIRS spectral data using partial least squares regression (PLSR). The coefficients of determination in the validation set ranged between 0.69 and 0.93, and the error of prediction varied from 0.064 to 2.89. The evaluation of the model confirmed the high predictive ability of NIRS for DM and CP and its low predictive ability for DMD, OMD, NDF and the WSC/CP ratio. It was not possible to obtain a model for WSC because it would have required an increased number of samples to improve the spectral variability and the R2 value (> 80%).

Key words: calibration models; external validation; forage; NIRS; nutritive value


La nutrición influye en el rendimiento y salud animal, donde la determinación rápida y confiable de la calidad del forraje es esencial para mejorar la producción animal. El presente estudio describe el uso de la espectroscopía de reflectancia del infrarrojo cercano (NIRS) para la cuantificación de la calidad nutricional en términos de materia seca (MS), proteína cruda (PC), carbohidratos solubles en agua (CHOS), digestibilidad in vitro de la MS (DIV), digestibilidad de la materia orgánica (DMO), fibra detergente neutro (FDN) y la relación CHOS/PC en praderas frescas del sur de Chile (39 a 40° S). Los modelos de calibración se desarrollaron entre la química húmeda y los datos espectrales NIRS usando la regresión por mínimos cuadrados parciales (PLSR). El rango de coeficientes de determinación en el conjunto de validación varió entre 0,69 y 0,93 y el error de predicción entre 0,064 y 2,89. La evaluación del modelo obtenido confirmó la alta capacidad predictiva del NIRS para MS y PC, baja capacidad predictiva para FDN, DIV, DMO y la relación CHOS/PC. No fue posible obtener un modelo para CHOS, siendo necesario aumentar el número de muestras y así mejorar la variabilidad espectral y el valor de R2 (>80%).

Palabras clave: Calidad nutritiva; forraje; modelos de calibración; NIRS; validación externa


Animal production in southern Chile relies mainly on forages and pastures that are heterogeneous in terms of species, yield and management. Grassland classification of the principal dairy and beef production systems are divided into sown (11.0%) and naturalized with (46.3%) and without management (42.5%) (Moscoso and Urrutia, 2017). In terms of dairy and beef production, 73% of the milk produced in Chile and 44% of bovines come from the southern regions: Los Rios and Los Lagos (ODEPA, 2015). This highlights the importance of pastures and their utilization, with an inverse relationship between the use of pastures and production cost (Dillon et al., 2005). It is therefore of utmost importance to know the nutritive value of forages to improve grazing management decisions (Corson et al., 1999).

Near-infrared methods are becoming widely used in animal science to predict several variables, including the chemical composition of forages, intake and growth rates (Coates, 2000; Stuth and Tolleson, 2000). They are also becoming more commonly used to identify animal species, gender and pregnancy (Tolleson and Stuth, 2002).

The traditional characterization of the nutritional quality of forages though wet chemistry procedures has the disadvantage of being slow and destructive, requiring specialized equipment, which is usually expensive (Deaville and Flinn, 2000). In contrast, using near infrared spectroscopy (NIRS) to predict the nutritive quality of forages could prove to be a quick and effective tool for nutrient composition determination of forages, which would help improve the strategic use of supplements and adjustments in ration formulations for dairy and beef bovines.

The NIR region is the wavelength range between 700-2500 nm in the electromagnetic spectrum. When a sample is analyzed, the radiant energy is absorbed selectively according to the specific vibration of the molecule, which produces an overtone in the spectrum. Prominent absorptions in NIRS forage spectra include water, aliphatic hydrocarbons, lipids and carbohydrates (Conzen, 2006). Unlike most conventional analytical methods, NIRS is fast and nondestructive; it does not use chemicals, does not generate chemical wastes that require disposal and is also multiparametric, provided that several parameters can be determined simultaneously in the same measurement (Eldin, 2011).

There are a number of major steps that are essential to obtain a satisfactory working calibration: sample selection, acquisition of spectra and reference data, pretreatment of spectral data, derivation of the regression model and validation of the model (Deaville and Flinn, 2000). The quality of the reference-method analysis has a crucial effect on the accuracy of NIR calibrations (Conzen, 2006).

This technology is broadly accepted as a fast and reliable method for evaluating the nutritional quality of pasture silages (Ibáñez and Alomar, 2008; Restaino et al., 2009), as well as dried and fresh pastures (Cozzolino and Labandera, 2002; Alomar et al., 2009; Burns et al., 2013; Lobos-Ortega et al., 2013; Moscoso and Balocchi, 2016). It has also been used to evaluate how forage sample preparation affects nutrient composition analysis (Alomar et al., 2003), green forage intake and digestibility in ruminants (Decruyenaere et al., 2009), and water-soluble carbohydrates (WSC) in stolon samples of white clover (Inostroza et al., 2017). NIRS has also been used to characterize and quantify isoflavones and phenolic acid contents in red and white clover (Krähmer et al., 2013).

Although NIRS is a quick and relatively low-cost technology, prediction equations for assessing nutritional quality have mostly been developed using leave-one-out cross validation (Garcia and Cozzolino, 2006; Ibáñez and Alomar, 2008; Alomar et al., 2009; Restaino et al., 2009; Burns et al., 2013). These equations do not typically utilize a independent test set to generate a robust model to predict an unknown pasture sample.

The objective of this study was to generate NIRS prediction models as a quick and effective tool to quantify the nutritional quality of fresh permanent pastures for dairy and beef production systems in southern Chile.

Materials and methods

The present study used 915 permanent pasture samples collected randomly between 2014 and 2016 from several farms located in southern Chile (39° to 40° S). These samples were grown on different soil types and under different soil fertility conditions, representing the variability of pastures used in dairy and beef production systems under grazing. Permanent pasture samples comprised different proportions of grasses, such as Lolium perenne L., Agrostis spp., Holcus lanatus L., Bromus spp., Dactylis glomerata L. and clovers (mostly Trifolium repens L. and T. pratense L.).

NIR Spectroscopy

Fresh samples were cut to 2-3 cm with hand shears. After this procedure, each fresh subsample was exposed to an electromagnetic scan over a spectral wavelength range of 700-2,500 nm (near infrared) using an MPA model FT-NIR (Bruker Optik GmbH, Germany). Energy in this spectral range was directed into the sample, and the reflected energy (R) was measured by the instrument. Each spectral measurement was obtained from 32 scans performed at a wavenumber resolution of 16 cm-1.

Chemical analysis

At the laboratory of Nutrición Animal y Medio Ambiente (INIA Remehue), all samples were thoroughly homogenized, oven-dried at 60°C for 48 h, and ground through a 1 mm sieve, after which their chemical composition was determined. Samples were analyzed for DM (method 934.01) and CP (method 984.13), according to the procedures outlined by AOAC (1995), and WSC according to MAFF (1986). In vitro dry matter digestibility (DMD) and organic dry matter digestibility (OMD) were determined by incubating the samples in a two-stage rumen liquid for 48 h at 39°C, followed by incubation in a pepsin acid medium for 24 h (Tilley and Terry, 1963). Finally, NDF was determined as described by Mertens (2002).

Chemometric analysis

Partial least-squares regression (PLSR) with a test-set validation was used to calibrate the spectral data with the wet chemistry data. The PLSR algorithm selected the successive orthogonal factors that maximized the covariance between predictor (spectra) and response variables (laboratory data). The software OPUSTM version 6.0 (Bruker Optik GmbH, Germany) was used to create models by selecting wavelengths, mathematical pretreatments, PLSR factors, outlier determinations and PLSR regression, among others.

To develop and validate the regression models for each parameter, the samples were divided into two sets: a calibration set for modeling and a validation set for the developed model. To define the two groups, the samples were arranged according to the experimental value of each parameter and subsequently distributed alternately, for example, i) when the number of samples was more than 200, one sample was used for calibration and one for validation; ii) when the number of samples was less than 200, two were used for calibration and one for validation. In either case, both sets covered the whole range of chemical data (Conzen, 2006).

The PLSR method was performed on the spectral region, setting the upper limit at 10 PLS factors (Conzen, 2006). To optimize calibration accuracy, spectral data were subjected to a variety of preprocessing transformations using common scatter correction treatments and derivations of the original spectrum to transform spectral data in such a way that the signals would better adhere to Beer’s law, which states that absorbance and concentration are linearly correlated (Rinnan et al., 2009). The tested preprocessing transformations underwent vector normalization (VN), minimum-maximum normalization, multiplicative scatter correction (MSC), subtraction of a straight line, constant offset elimination, first and second derivative, or a combination of these options. The root mean square error of calibration (RMSEC) was used to calculate the analysis error of the calibration values (Equation 1).

RMSEC=i=1n(yily^i)2n-r-1 (1)

where ŷi is the NIR predictive value, yi is the chemical reference value of sample i, n is the number of samples and r is the number of PLS factors.

The validation set was used to identify the best preprocessing technique and select the optimal number of PLS factors for each model. The root mean square error of prediction (RMSEP), which represents an objective assessment of the overall error between modeled and reference values, was used to evaluate and compare the accuracy of the different PLS models developed (Equation 2).

RMSEP=i=1n(yiy^i)2n (2)

where ŷi is the NIR predictive value, yi is the chemical reference value of sample i from the prediction set of samples, and n is the number of samples.

Various PLS models were built correlating preprocessed NIRS spectral data with the chemical data of the calibration set. The best models were selected using the lowest RMSEP; therefore, we chose among the best five models the one that had the lowest number of PLS factors. The coefficient of determination of prediction (R2p) and the residual predictive deviation (RPD) were also used to evaluate the predictive ability of the models. The RPD is the ratio between the standard deviation (SD) of the reference values and the error of prediction; it is thus a qualitative measure for the assessment results. The smaller the error of prediction, compared to the variance of the reference values, the larger the RPD value, and therefore, the better the model. An RPD value below 2.4 characterizes a poor model, whereas RPD values approximately 2.5-3.0 could be used for screening purposes (Williams and Sobering, 1996; Conzen, 2006).

Results and Discussion

The mean and standard deviations of DM, WSC, CP, DMD, OMD and NDF were 15.1% (±4.10), 8.0% (±3.13), 23.4% (±5.57), 83.5% (±4.97), 73.2% (±5.66) and 44.1% (±6.21), respectively, and the standard error of the laboratory was 0.34, 0.22, 0.32, 0.57, 0.63 and 0.90 for DM, WSC, CP, DMD, OMD and NDF, respectively (Figure 1). These mean values are similar to those found by Demanet et al. (2015), who studied the nutritive quality of permanent pastures over 11 years and found a range between 15.2-23.2%, 13.4-36.7%, and 48.6-58.4% for DM, CP and NDF, respectively. In another study, in the evaluation of different pasture renovation strategies, Keim et al. (2014) found concentrations between 58.3% and 78.3% for OMD and 5% and 12.6% for WSC, similar to those observed in the present study.

Figure 1 Boxplots displaying the mean, quartiles, outliers and the range of distribution for dry matter (DM), water-soluble carbohydrates (WSC), in vitro dry matter digestibility (DMD), organic matter digestibility (OMD) and neutral detergent fiber (NDF) in fresh pasture samples from southern Chile. Boxplots are the 25th, 50th and 75th percentiles, and the bottom and top whiskers are the 10th and 90th percentiles, respectively. 

Chemical variations could be the direct result of variability in pasture components (botanical composition), growth stage, climatic and management factors, and soil type, among others (Hopkins, 2000). The wide range of values for nutrient composition parameters included two important conditions desirable for an optimal calibration, as described by Murray (1988): the amplitude of composition and its homogeneous distribution across the range of composition. Figure 1 shows a boxplot displaying quartiles and ranges of distribution for the nutrition quality of the sampled fresh forage.

NIRS calibration and validation

The predictive capacity of NIRS for all of the studied parameters is provided in Table 1. After testing several mathematical treatments, the best calibration was selected according to the R2, RMSEP and RPD statistics from the set validation. The coefficients of determination for the validation set were higher than 0.69, the errors of prediction were lower than 2.9, and RPD values above 1.8 were observed (Table 1). The relation between the NIRS prediction and the composition obtained by the reference methods is shown in Figure 2.

Table 1 Descriptors of the NIRS calibration and validation sets to estimate the chemical composition of fresh forage samples. 

Parameters Calibration set Validation set
DM (%) 433 8.14-28.00 0.94 1.05 4.0 432 8.36-27.30 0.93 1.13 3.7
CP (%) 456 10.25-33.81 0.87 2.04 2.8 459 8.39-34.25 0.84 2.22 2.5
WSC (%) 144 2.80-16.60 0.74 1.68 2.0 138 2.90-17.10 0.69 1.74 1.8
DMD (%) 109 65.30-93.20 0.77 2.98 2.1 45 69.30-93.00 0.76 2.41 2.2
OMD (%) 113 56.90-86.70 0.79 3.03 2.2 53 58.60-83.00 0.78 2.61 2.2
NDF (%) 113 28.80-54.30 0.85 2.45 2.6 49 30.30-53.90 0.78 2.89 2.2
WSC/CP 100 0.01-0.73 0.76 0.06 2.0 47 0.10-0.72 0.74 0.06 2.0

DM: dry matter; CP: crude protein; WSC: water-soluble carbohydrates; DMD: in vitro dry matter digestibility; OMD: organic matter digestibility; NDF: neutral detergent fiber; N: number of samples; R²c: coefficient of determination for calibration; RMSEC: root mean square error of calibration; RPD: residual prediction deviation; RMSEP: root mean square error of prediction; R2p: coefficient of determination in prediction.

Figure 2 Correlation of wet chemistry and the predicted NIRS for a) dry matter (DM), b) water-soluble carbohydrates (WSC), c) crude protein (CP), d) in vitro dry matter digestibility (DMD), e) organic matter digestibility (OMD), f) neutral detergent fiber (NDF) and g) the WSC/CP ratio in fresh pasture samples from southern Chile. The diagonal line represents the equal response line. 

The calibration models obtained for DM and CP were the most robust because they obtained RPD values ≥ 2.5, which is the critical value used for the prediction of unknown samples, and R2 values higher than 0.84 (Williams, 2004). On the other hand, the models generated for DMD, OMD, NDF and WSC/CP ratio were less precise (RPD ≥ 2.0), even though the R2 and RPD parameters for NDF were greater than 0.85 and 2.6, respectively, in the calibration set (Table 1). Finally, it was not possible to obtain adequate prediction models for WSC. This may be a consequence of the high water content of the fresh pastures. The water absorption bands in the NIRS spectral data are strong and can limit the detection capacity of other constituents (Thyholt and Isaksson, 1997).

The predictive ability of these models can be explained in part by the low standard error of the laboratory given that the reference data were obtained by gravimetric and chemical analysis, techniques that present high precision in comparison to that of other techniques that include biological elements (Ibañez and Alomar, 2008).

The error of calibration for DM was lower than that obtained by Alomar et al. (2009), who reported values of standard error of cross validation (SECV) of 7.5, although in their study they obtained higher coefficients of determination and RPD values (0.98 and 7.15, respectively). The calibration model error obtained for CP was lower than that found by Gislum et al. (2004) in red fescue and perennial ryegrass and by Lobos-Ortega et al. (2013) and Alomar et al. (2009) in dry and fresh permanent pastures, respectively. The models generated for DMD and NDF were lower than those reported by Decruyenaere et al. (2009) in pastures, legumes and a forage mix and were similar to those reported by Alomar et al. (2009) in fresh pastures. Nonetheless, the calibration descriptors in our study were higher than those obtained by Restaino et al. (2009) in pasture silage. The calibration error obtained in this study for OMD was lower than that reported by Corson et al. (1999) in pasture samples with a similar concentration range. The coefficient of determination obtained in this study for WSC was lower than that reported by Alomar et al. (2009) and Lobos-Ortega et al. (2014) in fresh forage samples and by Inostroza et al. (2017) in dried white clover. Finally, the results for R2c and the RPD values for the WSC/CP ratio were lower in the present study than those shown by Rivero et al. (2014).

The predicted DM and CP showed a strong relation with the reference values (Figure 2), as shown by the concentration of individual observations around the equal response line (Figure 2a, 2c). For the WSC parameter, the low predictive ability is confirmed by the lower correlations among the reference value and spectrum (Figure 2b). These results suggest that the water content can often mask NIR signals and generate a limited predictive model (Reeves, 2000). Although errors are often slightly higher for components such as CP, DMD, OMD and NDF in fresh samples when compared to those of dried and ground samples, this is balanced by the ability to scan a greater number of samples in less time, thus avoiding changes in the composition due to oven-drying procedures (Alomar et al., 2009). In addition, the minimum sample preparation helps to reduce the time and cost of the analysis. The RMSEP and RMSEC values are very close (Table 1), indicating that the obtained calibration model retained its predictive capability for the samples that did not belong to the calibration set (González-Sáiz et al., 2007).

The main conclusions are as follows. The evaluation of a calibration model obtained for fresh forage confirmed the high predictive ability of NIRS for DM and CP, a low predictive quality for DMD, OMD, NDF and WSC/CP ratio, while it was not possible to obtain an adequate prediction model for WSC. It is therefore necessary to increase the number of fresh pasture samples used for calibration to improve the spectral variability and R2 value (>80%). These results suggest that it is possible to use NIR spectroscopy to optimize pasture management decisions for grazing high quality pastures to maximize animal production.


This research was funded by the Fondo Nacional de Desarrollo Científico y Tecnológico (FONDECYT) Initiation grant n°11140296, Fondo para la Innovacion y la Competitividad de la Region de Los Lagos N°30234274 (FIC Regional) and the INIA Forage and Pasture programme (502087-70). We also thank Cristian Cares, Cristian Fierro, Marcelo Villagra and Alfredo Torres for their help in the field and with the laboratory procedures and Natalie Urrutia for helping to improve the English writing of the manuscript.


Alomar, D., R. Fuchslocher, and M. de Pablo. 2003. Effect of preparation method on composition and NIR spectra of forage samples. Animal Feed Science and Technology 107:191–200. [ Links ]

Alomar, D., R. Fuchslocher, J. Cuevas, R. Mardones, and E. Cuevas. 2009. Prediction of the composition of fresh pastures by near infrared reflectance or interactance-reflectance spectroscopy. Chilean Journal of Agricultural Research 69:198–206. [ Links ]

AOAC. 1995. Official methods of analysis of AOAC International, 16th ed. Association of Official Analytical Chemists, Arlington, USA. [ Links ]

Burns, G.A., T.J. Gilliland, D. Grogan, S. Watson, and P. O’kiely. 2013. Assessment of herbage yield and quality traits of perennial ryegrasses from a national variety evaluation scheme. The Journal of Agricultural Science 151:331–346. [ Links ]

Coates, D. 2000. Faecal NIRS-what does it offer today’s grazier?. Tropical grasslands 34:230–240. [ Links ]

Conzen, J.P. 2006. Multivariate calibration. A practical guide for developing methods in the quantitative analytical chemistry. 2nd English ed. Bruker Optics, Ettlingen, Germany. [ Links ]

Corson, D., G. Waghorn, M. Ulyatt, and J. Lee. 1999. NIRS: Forage analysis and livestock feeding. Proceedings of the New Zealand Grassland Association. 127–132. [ Links ]

Cozzolino, D., and M. Labandera. 2002. Determination of dry matter and crude protein contents of undried forages by near-infrared reflectance spectroscopy. Journal of the Science of Food and Agriculture 82:380–384. [ Links ]

Deaville, E., and P. Flinn. 2000. Near-infrared (NIR) spectroscopy: an alternative approach for the estimation of forage quality and voluntary intake. In: Givens, D., E. Owen, R. Axford, and H. Omed (eds.). Forage evaluation in ruminant nutrition. CABI, Wallingford, UK. 301–320. [ Links ]

Decruyenaere, V., P. Lecomte, C. Demarquilly, J. Aufrere, P. Dardenne, D. Stilmant, and A. Buldgen. 2009. Evaluation of green forage intake and digestibility in ruminants using near infrared reflectance spectroscopy (NIRS): Developing a global calibration. Animal Feed Science and Technology 148:138–156. [ Links ]

Demanet, R., M.L. Mora, M.Á. Herrera, H. Miranda, and J.M. Barea. 2015. Seasonal variation of the productivity and quality of permanent pastures in Andisols of temperate regions. Journal of Soil Science and Plant Nutrition 15:111–128. [ Links ]

Dillon, P., J. Roche, L. Shalloo, and B. Horan. 2005. Optimising financial return from grazing in temperate pastures. Satellite meeting of the XXth International Grassland Congress. Utilisation of grazed grass in temporal animal systems, Cork, Ireland. 131–147. [ Links ]

Eldin, A., 2011. Near Infra Red Spectroscopy. In: Akyar, I. (ed.) Wide Spectra of Quality Control. InTech, Rijeka, Croatia. 237–248. [ Links ]

Garcia, J., and D. Cozzolino. 2006. Use of Near Infrared Reflectance (NIR) Spectroscopy to Predict Chemical Composition of Forages in Broad-Based Calibration Models. Agricultura Técnica 66:41–47. [ Links ]

Gislum, R., E. Micklander, and J.P. Nielsen. 2004. Quantification of nitrogen concentration in perennial ryegrass and red fescue using near-infrared reflectance spectroscopy (NIRS) and chemometrics. Field Crops Research 88:269–277. [ Links ]

González-Sáiz, J.M., C. Pizarro, I. Esteban-Díez, O. Ramírez, C.J. González-Navarro, M.J. Sáiz-Abajo, and R. Itoiz. 2007. Monitoring of Alcoholic Fermentation of Onion Juice by NIR Spectroscopy: Valorization of Worthless Onions. Journal of Agricultural and Food Chemistry 55:2930–2936. [ Links ]

Hopkins, A., 2000. Grass: Its Production and Utilization, 3rd ed. Blackwell Science Ltd, Oxford, U.K. 440. [ Links ]

Ibáñez, L., and D. Alomar. 2008. Prediction of the chemical composition and fermentation parameters of pasture silage by near infrared reflectance spectroscopy (NIRS). Agricultura Técnica 68:352–359. [ Links ]

Inostroza, L., I. Lobos, H. Acuña, C. Vasquez, and G. Tapia. 2017. NIR-Prediction of water soluble carbohydrate in white clover and its genetic relationship with cold tolerance. Chilean Journal of Agricultural Research 77:218–225. [ Links ]

Keim, J.P., I.F. López, and O.A. Balocchi. 2014. Sward herbage accumulation and nutritive value as affected by pasture renovation strategy. Grass and Forage Science 70:283–295 [ Links ]

Krähmer, A., G. Gudi, N. Weiher, M. Gierus, W. Schütze, and H. Schulz. 2013. Characterization and quantification of secondary metabolite profiles in leaves of red and white clover species by NIR and ATR-IR spectroscopy. Vibrational Spectroscopy 68:96–103. [ Links ]

Lobos-Ortega, I., P. Gou, S. Hube, R. Saldaña, and M. Alfaro. 2013. Evaluation of potential nirs to predict pastures nutritive value. Journal of Soil Science and Plant Nutrition 13:463–468. [ Links ]

Lobos-Ortega, I., A. Torres, and R. Saldaña. 2014. Cuantificación del contenido de carbohidratos solubles en praderas del sur de Chile por espectroscopia NIR. Revista Argentina de Producción Animal 34 Supl. 1:541. [ Links ]

MAFF, 1986. Method 14. Carbohydrate, solubles, in herbage. In: Bailey, S. (ed.) The Analysis of Agricultural Materials: a manual of the analytical methods used by the Agricultural Development and advisory service. HMSO, London, UK. 43–45. [ Links ]

Mertens, D.R. 2002. Gravimetric determination of amylase-treated neutral detergent fiber in feeds with refluxing in beakers or crucibles: collaborative study. Journal of AOAC International 85:1217–1240. [ Links ]

Moscoso, C.J., and O. Balocchi. 2016. Water-soluble carbohydrate and nitrogen concentrations after defoliation in perennial ryegrasses (Lolium perenne L.) in spring. Agrosur 44:23–29. [ Links ]

Moscoso, C.J., and N. Urrutia. 2017. Overview of the forage land-use in southern Chile in a 30-year period. Crop, Forage, & Turfgrass Management 3:1–4. [ Links ]

Murray, I. 1988. Aspects of the interpretation of Near Infrared Spectra. Food Science and Technology Today 2:135–140. [ Links ]

ODEPA, 2015. Boletín de la Leche. [ Links ]

Reeves, J.B. 2000. Use of near infrared reflectance spectroscopy. In: J.P.F. D’Mello (ed). Farm animal metabolism an nutrition. CABI Publishing, Wallingford, UK. 185–207. [ Links ]

Restaino, E.A., E.G. Fernandez, A. La Manna, and D. Cozzolino. 2009. Prediction of the nutritive value of pasture silage by near infrared spectroscopy (NIRS). Chilean Journal of Agricultural Research 69:560–566. [ Links ]

Rinnan, Å., L. Nørgaard, F.V.D. Berg, J. Thygesen, R. Bro, and S.B. Engelsen. 2009. Data Pre-processing. In: Sun, D. (ed.). Infrared Spectroscopy for Food Quality Analysis and Control. Academic Press, San Diego, USA. 29–50. [ Links ]

Rivero, J., P. Loaiza, O. Balocchi, and D. Alomar. 2014. Prediccion de la relación carbohidratos solubles/proteína cruda de praderas de ballica perenne mediante NIRS. Revista Argentina de Producción Animal 34, Supl. 1:561. [ Links ]

Stuth, J.W., and D.R. Tolleson. 2000. Monitoring the nutritional status of grazing animals using near-infrared spectroscopy. Compendium on Continuing Education for the Practicing Veterinarian 22:108–115. [ Links ]

Thyholt, K., and T. Isaksson. 1997. Near infrared spectroscopy of dry extracts from high moisture food products on solid support - A review. Journal of Near Infrared Spectroscopy 5:179–193. [ Links ]

Tilley, J.M.A., and R.A. Terry. 1963. A two stage technique for the in vitro digestion of forage crops. Grass and Forage Science 18:104–111. [ Links ]

Tolleson, D.R., and J.W. Stuth. 2002. Exploring Grazing Animal Physiology via Fecal NIRS. Proceedings of an NIRS Workshop. University of Alberta, Canada, [ Links ]

Williams, P.C., and D.C. Sobering. 1996. How do we do it: a brief summary of the methods we use in developing near infrared calibrations. In: Davies, A.M.C., and P. Williams (eds.). Near infrared spectroscopy: The future waves, Chichester, UK. 185–188. [ Links ]

Williams, P. 2004. Near-infrared technology - getting the best out of light : a short course in the practical implementation of near-infrared spectroscopy for the user. Nanaimo, Canada. 12–13. [ Links ]

Received: September 11, 2018; Accepted: September 09, 2019

Corresponding author:

Creative Commons License All the contents of this journal, except where otherwise noted, is licensed under a Creative Commons Attribution License