- Email: [email protected]
Evaluation of botanical and chemical composition of sheep diet by using faecal near infrared spectroscopy
Animal Feed Science and Technology 222 (2016) 1–6
Contents lists available at ScienceDirect
Animal Feed Science and Technology journal homepage: www.elsevier.com/locate/anifeedsci
Short communication
Evaluation of botanical and chemical composition of sheep diet by using faecal near infrared spectroscopy ∗ ˜ ˜ Blanco, Valeriano Nieves Núnez-Sánchez , Domingo Carrion, Francisco Pena Domenech García, Ana Garzón Sigler, Andrés Luis Martínez-Marín Departamento de Producción Animal, Universidad de Córdoba, Campus de Rabanales, Carretera Madrid-Cádiz km 396, 14071 Córdoba, Spain
a r t i c l e
i n f o
Article history: Received 12 May 2016 Received in revised form 19 September 2016 Accepted 21 September 2016 Keywords: Near infrared reflectance spectroscopy Faecal NIRS Faeces Diet composition Ewe
a b s t r a c t The aim of this study was to use near infrared reflectance spectroscopy (NIRS) to directly predict the chemical composition and forage content of diets consumed by ewes from the analysis of their faeces. Twenty four ewes (46 ± 3 kg BW) placed in individual cages were used. Each animal was randomly allocated to four diets fed in four successive 28-days periods, so that a total of 16 diets were fed (six ewes per diet). Diets were designed to have different contents of neutral detergent fibre and crude protein. Daily rations of 800 g (as-fed) were formulated with alfalfa hay (0–1000 g/kg) or cereal straw (0–800 g/kg) as forage base, different mixtures of maize, peas and lupins, offered as whole grains, and pelleted sunflower meal. Faeces samples were collected at the end of each feeding period, directly from the rectum. A total of 96 faeces samples were dried and ground. The samples were scanned in small ring cups in a FOSS-NIRSystems 6500 SY-II scanning monochromator. Faecal Modified Partial Least Squares calibration equations, selected in terms of standard error of cross validation (SECV) and coefficient of determination (r2 ), showed good predictive values for the prediction of dietary botanical ingredients (g/kg) alfalfa (41.9 and 0.98), cereal straw (48.6 and 0.97), maize (89.1 and 0.86), and forage (67.4 and 0.90) or concentrate (67.5 and 0.90) respectively, and also for chemical composition of fed diets (g/kg DM): ash (5.2 and 0.93), crude protein (9.6 and 0.89), ether extract (2.9 and 0.80), neutral detergent fibre (34.5 and 0.95) and non-fibre carbohydrates (38.7 and 0.94), respectively. Lower predictive capacity was obtained for peas (78.0 and 0.43), lupins (37.0 and 0.84) and sunflower meal (6.7 and 0.99), probably due to the limited variability of these ingredients in the diets. These results support the viability of faecal NIRS as a fast and reliable analytical method which allows to accurately predict the botanical composition, proportions of forage ingredients and chemical composition of diets consumed by ewes having free access to a variety of feed ingredients. Future works should be focused both to increase the variability of the diet ingredients and also to validate the models with external samples. © 2016 Elsevier B.V. All rights reserved.
Abbreviations: aNDF, neutral detergent fibre; CP, crude protein; DM, dry matter; EE, ether extract; MPLS, modified partial least squares; MSC, multiplicative scatter correction; NFC, non fibrous carbohydrates; NIRS, near infrared reflectance spectroscopy; r2 , coefficient of determination of cross validation; SEC, standard error of calibration; SECV, standard error of cross validation; SEP, standard error of prediction; SNV, standard normal variate; T, outlier samples. ∗ Corresponding author. ˜ E-mail address: [email protected] (N. Núnez-Sánchez). http://dx.doi.org/10.1016/j.anifeedsci.2016.09.010 0377-8401/© 2016 Elsevier B.V. All rights reserved.
2
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
1. Introduction Composition of the food ingested by animals is of paramount importance on the management of extensive farms and for the evaluation of the consumption habits of wild animals (Landau et al., 2006). Farmers need updated information on the nutritive value of forages and supplementary feeds consumed by the animals to manage their nutritional status. This is especially important in the new strategies aiming towards greater efficiency in feeding systems based on the free choice of the food ration, not only the forage but also components of the concentrate (Dixon and Coates, 2009). However, it is difficult to obtain and quantify this information directly, as it requires expensive animal handling that is not always possible. The composition of faeces has been long considered to contain information of the diet consumed by the animal. Hence, the evaluation of faeces has been considered as a good method to monitor many attributes of the fed diet, and some aspects of animal physiology (Dixon and Coates, 2009). Near infrared reflectance spectroscopy (NIRS) has been reported to be a fast, reliable and economic tool to accurately predict different attributes of raw materials and feeds (Williams and Norris, 2001), considered as well as an environmental-friendly analytical solution. NIR analysis of faeces could greatly facilitate the study of the quality and quantity of the diet selected by both domestic and feral grazing herbivores (Dixon and Coates, 2009; Dryden, 2003). In fact, this technology has been successfully used to predict chemical composition and digestibility of the ˜ diet through faeces analysis in pigs (Schiborra et al., 2015), rabbits (Núnez-Sánchez et al., 2012), and ruminants (Landau et al., 2006; Dixon and Coates, 2009). Few works have attempted to assess the ability of NIRS evaluation of faeces to predict diet composition in sheep (Li et al., 2007; Decandia et al., 2009), and botanical composition of the diet (Dixon and Coates, 2009; Landau et al., 2008; Glasser et al., 2008; Kneebone and Dryden, 2015). However, it would be interesting to assess the presence of concentrates in the diet, especially under low pasture availability, as reported for cows (Ottavian et al., 2015). In this case, a qualitative model was developed to discriminate animals among two levels of concentrate supplementation. Hence, it would be interesting to test the ability of faecal NIRS to quantify the amount of forage and concentrates percentages of fed diets, and also predict the proportion of the mixed ingredients. The aim of this work was to develop faecal NIRS calibrations to assess both the botanical and chemical composition of several diets consumed by ewes in a free access basis. 2. Materials and methods 2.1. Samples and reference data determination The experiment was carried out in accordance with the Spanish regulation on experimental animals. Twenty four ˜ sheep (46 ± 3 kg BW), neither lactating nor pregnant, were used in the assay. Each animal was randomly allocated Segurena to four diets fed in four successive periods of 28 days. A total of 16 diets were fed (six ewes per diet). Diets (800 g/d, as-fed) were formulated to have variable contents of neutral detergent fibre (aNDF) and crude protein (CP), by using combinations of forage/non-forage ingredients from 100/0 to 20/80. Forage base was alfalfa (0–1000 g/kg) or cereal straw (0–800 g/kg). In order to get a good digestive and nutritional balance, different mixtures of cereals, legumes and sunflower meal as nonforage ingredients were used: maize (0–800 g/kg), peas (0–300 g/kg) and lupins (0–200 g/kg), offered as whole grains, and pelleted sunflower meal (0–200 g/kg). Visual observation of the troughs revealed that feed offered daily was completely consumed from day to day. Hence, the design of the assay allowed to know the real individual daily intake (g/d) of the different ingredients of fed diets. Individual faecal samples were collected for two consecutive days at the end of each feeding period, directly from the rectum. A pool sample was originated from mixing 2 days of faeces from each ewe. The total 96 individual faecal samples obtained were dried at 100 ◦ C for 24 h and ground to pass through a 1 mm sieve. After that, the samples were frozen at −20 ◦ C until NIR analyses were performed. Dry matter (DM), ash, crude protein (CP) and ether extract (EE) contents of feed ingredients were determined according to methods 930.15, 942.05, 984.13 and 920.39 of AOAC (2000), respectively, whereas aNDF was determined according to Van Soest et al. (1991). Non fibrous carbohydrates (NFC) were calculated by difference method, as recommended by NRC (2001); NFC = 100 − aNDF − ash − CP − EE. 2.2. NIR spectra collection Reflectance spectra from dried and ground faeces samples were obtained on a FOSS-NIRSystems 6500 SY-II scanning monochromator (FOSS-NIRSystems, Silver Spring, MD, USA) equipped with a spinning module. Faeces samples were scanned using small ring cups. Spectral absorbance values were recorded from 400 to 2498 nm, every 2 nm. Two capsules per sample were filled for NIR analysis and the average spectra of two subsamples were used for calibration. WINISI II software, version 1.50 (Infrasoft International LLC, State College, PA) was used for spectral data collection. 2.3. Data processing and calibration development WINISI IV software, version 4.8 (Foss, Denmark) was used for data processing and calibration development. Regression models were performed to predict botanical and chemical composition of fed diets. For that purpose, each faecal spectrum was paired with the information of the diet fed by the corresponding ewe: the proportion (g/kg) of the botanical ingredients
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
3
Table 1 Botanical and chemical composition of the diets used in the assay. Diet
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Mean SD Minimum Maximum
Diet ingredients (g/kg)
Diet composition (g/kg DM)
Alfalfa
Straw
Maize
Peas
Lupins
Sunflower meal
Forage
Concentrate
Ash
CP
EE
aNDF
NFC
1000 800 600 400 200 600 400 0 0 0 0 0 0 0 0 0 250.0 338.6 0 1000
0 0 0 0 0 0 0 800 800 600 600 600 400 400 400 200 300.0 309.8 0 800
0 200 400 600 800 200 300 0 0 0 200 200 400 400 400 600 293.8 240.7 0 800
0 0 0 0 0 200 300 0 0 200 200 0 200 0 0 0 68.8 107.8 0 300
0 0 0 0 0 0 0 0 200 200 0 200 0 200 0 200 62.5 95.7 0 200
0 0 0 0 0 0 0 200 0 0 0 0 0 0 200 0 25.0 68.3 0 200
1000 800 600 400 200 600 400 800 800 600 600 600 400 400 400 200 550.0 225.1 200 1000
0 200 400 600 800 400 600 200 200 400 400 400 600 600 600 800 450.0 225.1 0 800
104.6 86.4 68.2 50.0 31.8 71.7 55.3 88.0 74.7 63.9 60.1 60.3 45.7 46.0 59.3 31.7 62.4 19.9 31.7 104.6
174.8 156.8 138.8 120.8 102.8 170.0 167.7 90.6 102.4 143.0 87.8 111.8 97.3 121.2 109.5 130.6 126.6 29.0 87.8 174.8
11.9 16.7 21.6 26.5 31.3 16.6 18.9 10.6 19.5 19.5 15.9 24.5 21.0 29.6 20.8 34.7 21.2 6.7 10.6 34.7
460.6 390.7 320.7 250.8 180.8 343.4 284.8 729.3 685.0 572.0 540.8 549.3 405.1 413.5 457.8 277.7 428.9 155.6 180.8 729.3
248.0 349.4 450.8 552.2 653.6 398.2 473.3 81.8 118.8 201.8 295.6 254.4 431.2 390.0 353.0 525.6 361.1 156.6 81.8 653.6
SD: standard deviation; CP: crude protein; EE: Ether extract; aNDF: neutral detergent fibre; NFC: Non fibrous carbohydrates (100 − aNDF − ash − CP − EE).
(alfalfa, cereal straw, maize, peas, lupins and sunflower meal) and the chemical composition (g/kg DM) (ash, CP, EE, aNDF and NFC) of the diets. Furthermore, models to predict the forage and concentrate (g/kg) contents of the diet were also developed. Regression models for all the parameters were obtained using the Modified Partial Least Squares (MPLS) regression method (Shenk and Westerhaus, 1995a). Separate MPLS calibrations were performed for each parameter in the calibration set. When developing MPLS equations, the use of cross-validation is recommended in order to select the optimal number of factors and avoid overfitting (Shenk and Westerhaus, 1995a,b, 1996). Given the hierarchical structure of the samples, with 16 diets and 6 ewes for each diet, a leave-out-one-group cross-validation strategy was used for all the calibrations. Therefore, the sample set was divided into 16 cross-validation groups, in a way that calibrations were developed, each time, with samples from 15 diets, and were validated with the samples belonging to the remaining diet until all groups of samples have been used for validation. A set of calibrations for each of the parameters under study have been developed after combining, in different wavelength ranges (400–2500 nm and 1100–2500 nm), pretreatments for scatter correction of the data (algorithms SNV and Detrend, and MSC) and Norris–Williams spectral derivatives, to remove additive and multiplicative effects in the spectra. Spectral derivatives are named by using four digits: the first digit is the order of the derivative, the second is the gap over which the derivative is calculated, the third is the number of data points in a running average (or smoothing), and the fourth is the second smoothing. Four of the most usual derivative pretreatments (1, 5, 5, 1; 1, 10, 10, 1; 2, 5, 5, 1 and ˜ et al. (2012), following Shenk and Westerhaus 2, 10, 10, 1), have been used in this study, as described in Núnez-Sánchez (1995a,b, 1996) recommendations for agricultural products. Hence, sixteen regression equations per parameter/variable were developed by combining four spectral derivative math treatments and two scatter correction methods in two spectral regions. Two passes of elimination of T outlier samples (samples with abnormally high residuals of predicted versus reference values), were applied. A critical T value of 2.5 was set for T outliers. The statistics used to select the best calibration models were: the standard error of calibration (SEC), the standard error of cross validation (SECV) and the coefficient of determination of cross validation (r2 ). The standard error or prediction (SEP) is the best estimate of the prediction capability of the equation and should be ideally estimated by applying the model to an independent validation set. However, as the calibration set of this work is not sizeable, it was not possible to get that statistic. Hence, SECV has been used to evaluate the predictive capacity of the models, although, as previously mentioned, external validation of the equations should be done in the future with an independent set of samples. Therefore, the best NIR equation per parameter was selected based on the higher r2 value and lower SECV (Shenk and Westerhaus, 1995a,b, 1996). 3. Results and discussion 3.1. Botanical and chemical composition of diets The chemical composition of each diet was calculated from the chemical information of their ingredients, determined in a laboratory. The design of the diets, with different mixtures of forages and concentrates, resulted in a sizeable variability of both the botanical and the chemical composition range for all parameters, except for peas, lupins and sunflower meal, with
4
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
Table 2 Faecal NIR calibrations to estimate botanical and chemical composition of diets. Diet composition
Spectral range
Mathematical derivation treatment
Scatter correction
PLS terms
N
MEAN
SD
SECV
r2
Alfalfa(g/kg) Cereal straw (g/kg) Maize (g/kg) Peas (g/kg) Lupins (g/kg) Sunflower meal (g/kg) Forage (g/kg) Concentrate (g/kg) Ash (g/kg DM) CP (g/kg DM) EE (g/kg DM) aNDF (g/kg DM) NFC (g/kg DM)
1100–2500 400–2500 400–2500 400–2500 1100–2500 400–2500 1100–2500 1100–2500 400–2500 400–2500 400–2500 400–2500 400–2500
2,5,5,1 2,10,10,1 1,5,5,1 1,10,10,1 1,5,5,1 1,5,5,1 2,5,5,1 2,5,5,1 2,10,10,1 1,5,5,1 2,5,5,1 2,10,10,1 2,10,10,1
SNV + DT SNV + DT SNV + DT SNV + DT MSC MSC MSC SNV + DT MSC SNV + DT SNV + DT MSC SNV + DT
10 11 10 6 11 12 11 11 9 6 9 10 9
92 95 93 94 89 92 96 96 94 91 94 92 92
245.7 298.9 289.2 66.0 65.2 17.4 550.0 450.0 62.6 126.3 21.3 431.5 357.5
333.0 303.0 235.7 104.3 94.3 56.7 219.1 219.1 19.5 28.9 6.6 153.1 154.3
41.9 48.6 89.1 78.0 37.0 6.7 67.4 67.5 5.2 9.6 2.9 34.5 38.7
0.98 0.97 0.86 0.43 0.84 0.99 0.90 0.90 0.93 0.89 0.80 0.95 0.94
Mathematical derivation treatments: the first digit is the number of the derivative, the second is the gap over which the derivative is calculated, the third is the number of data points in a running average or smoothing and the fourth is the second smoothing. MSC: multiplicative scatter correction; SNV + DT: standard normal variate and detrend; N: number of samples in the calibration set; Mean: mean of the calibration set; SD: standard deviation of the calibration set; SECV: standard error of cross-validation; r2 : coefficient of determination of cross validation; CP: crude protein; EE: Ether extract; aNDF: neutral detergent fibre; NFC: Non fibrous carbohydrates.
only two or three levels of inclusion (Table 1). The use of cereal straw in the present work has resulted in a wide range of variability for aNDF. 3.2. NIR calibration results A calibration set of 96 faecal samples from ewes fed 16 different diets was used. MPLS regression models were developed with the complete set of samples 96 samples. 3.2.1. Botanical composition Excellent precision has been obtained for all the parameters, as they have r2 values close to or above 0.9 according to Shenk and Westerhaus (1996), with the exception of the model to predict the percentage of peas consumed by the ewes (r2 = 0.43). A possible explanation of this poor result could be that, in this case, only 5 diets (30 samples) contained this ingredient and always in low percentages (20–30%). The study of the T outliers in the calibration process, and the predicted values obtained after the application of the peas model to the average spectrum of faecal samples from each diet (data not shown) revealed that, in two cases, the model was not able to discriminate this ingredient, from maize and from lupins, respectively. Values of bias, slope and R2 obtained in validation (63.1, 1.2 and 0.67 in peas and 37.5, 0.96 and 0.85 in lupins), were poorer compared with the rest of the parameters. However, although only 2 diets (12 faecal samples) included sunflower meal, the models showed good predictive results of this ingredient (Table 2). A possible explanation could be that the contribution of this ingredient to the diets provided not only chemical (Table 1), but also clearly physical changes to the diets. Absorbance data of NIR spectra depend upon both physic and chemical properties of the product (Williams and Norris, 2001). That particularity in the faecal spectra from those diets could have contributed to an easier quantification of this ingredient, when present. However, it must be pointed out that, despite these favorable statistics, the model to predict the inclusion of sunflower meal in the diet would only be usable for diets with similar levels of inclusion to the ones used here (0 or 200 g/kg). Extensively, the same limitations should be considered for the ingredients peas and lupins. The results obtained in this study suggest that diet residues present in ewes faeces had sufficient spectral information to accurately predict dietary botanical components (Dixon and Coates, 2009). Similar SECV and r2 results were obtained in goats to predict the dietary percentages of hay, corn stover (comparable to cereal straw) and the percentage of concentrate from the faecal NIR analysis (Landau et al., 2004, 2008). Ottavian et al. (2015) developed a faecal NIR qualitative model in dairy cows to discriminate between two levels of concentrate supplementation. However, none of the previous reviewed reports have attempted the quantitative prediction of the ingredients of the concentrates in the diet. The design of the study permitted to know the exact individual intake (g/d) of the different ingredients of the diet, as well as the calculation of dry matter intake (DMI). As an attempt to check the predictive reliability of the calibrations to predict botanical ingredients, a graph comparing the real daily intake (g/d) of DM, forage, concentrates and the different ingredients with the intakes calculated from the predictions of the dietary percentages of each diet has been included (Fig. 1). The results confirm the reliability of the models to predict all the above mentioned parameters except peas and lupins. The results obtained in this study suggest that diet residues present in ewes faeces had sufficient spectral information to accurately predict dietary botanical components (Dixon and Coates, 2009). Similar results were obtained in goats to predict the dietary percentages of hay, corn stover (comparable to cereal straw) and the percentage of concentrate from the faecal NIR analysis (Landau et al., 2004, 2008). Ottavian et al. (2015) developed a faecal NIR qualitative model in dairy cows to discriminate between two levels of concentrate supplementation. However, none of the previous reviewed reports have attempted the quantitative prediction of the ingredients of the concentrates in the diet.
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
5
Fig. 1. Scatter plot of actual versus estimated individual daily intakes (DI, expressed in grams per day (g/d)) of different dietary ingredients. Estimated individual daily intakes were calculated by dividing the daily amount of feed ingested (800 g) by the NIR predicted dietary fractions of the corresponding diet.
3.2.2. Chemical composition Faecal spectra were also used to predict the chemical composition of the diets consumed by the animals. The equations showed good precision, with r2 values close to or above 0.9 for ash, CP, aNDF and NFC (Table 2), while lower prediction ability in terms of r2 (0.80) has been obtained for the prediction of EE. In all cases, better results have been obtained using the complete spectral range, whereas the scatter and derivative pre-treatments did not offer differences. Contrarily to botanical composition, many studies have attempted to predict the chemical composition of the consumed diet from the NIR analysis of the faeces, mainly for goats and cows (Landau et al., 2006; Dixon and Coates, 2009). With respect to sheep, Li et al. (2007) obtained less accurate results, with higher SECV values for CP prediction, while Decandia et al. (2009) reported more precise equations, with lower SECV values, for the prediction of CP, aNDF and NFC. Therefore, the variability of the diets in this assay permitted to verify that NIR analysis of their faeces was able to estimate and differentiate the percentages of the forage (alfalfa hay or cereal straw) and concentrates, even in low amounts, on the diet ingested by ewes. Moreover, the precise and accurate prediction of the nutritional characteristics of fed diets, from rich to poor aNDF and CP contents, would be useful to assess the nutritional status and behavior of animals with free access to different feedstocks. 4. Conclusion The use of confined animals fed diverse diets composed by mixtures of forage and concentrates provided a varied database in terms of CP and aNDF contents. These results support the viability of faecal NIRS as a fast and reliable analytical method to monitor with high precision and accuracy the botanical composition and chemical attributes of the diet freely consumed by ewes, thus permitting, if necessary, the implementation of appropriate changes to the supplementation regime. Therefore, the spectra database created with a wide variability in terms of ingredients and chemical composition could be used in the future to assess the feeds chosen by animals with free access to similar diets. Future works should be focused both to increase the variability of the ingredients used in the diets and also to validate the models obtained here with external samples. Conflict of interest None declared.
6
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
Acknowledgements The authors thank José Javier Pérez Hernández and Inmaculada Andújar Ramírez for their support in the management of the animals and technical assistance in the NIR analyses. The study was performed using NIR equipment and facilities belonging to SCAI (NIR/MIR Unity), University of Cordoba (Spain). References AOAC, 2000. Official Methods of Analysis, seventeenth ed. Association of Official Analytical Chemists, Gaithersburg, MD. Decandia, M., Giovanetti, V., Boe, F., Scanu, G., Cabiddu, A., Molle, G., Cannas, A., Landau, S., 2009. Faecal NIRS to assess the chemical composition and the nutritive value of dairy sheep diet. Nutritional and Foraging Ecology of Sheep and Goats. Papachristou, T.G., Parissi, Z.M., Ben Salem, H., Morand-Fehr, P. (Eds). Options Méditerranéennes, Series A, No. 85, 135–139. Dixon, R.M., Coates, D.B., 2009. Review: near infrared spectroscopy of faeces to evaluate the nutrition and physiology of herbivores. J. Near Infrared Spectrosc. 17, 1–31. Dryden, G. McL., 2003. Near Infrared Reflectance Spectroscopy: Applications in Deer Nutrition. Rural Industries Research and Development Corporation, Canberra, Publication No. W03/007. Glasser, T., Landau, S., Ungar, E.D., Perevolotsky, A., Dvash, L., Muklada, H., Kababya, D., Walker, J.W., 2008. A fecal near-infrared reflectance spectroscopy-aided methodology to determine goat dietary composition in a Mediterranean shrubland. J. Anim. Sci. 86, 1345–1356. Landau, S., Glasser, T., Dvash, L., Perevolotsky, A., 2004. Fecal NIRS to monitor the diet of Mediterranean goats. S. Afr. J. Anim. Sci. 34, 76–80. Landau, S., Glasser, T., Dvash, L., 2006. Monitoring nutrition in small ruminants with the aid of near infrared reflectance spectroscopy (NIRS) technology: a review. Small Rumin. Res. 61, 1–11. Landau, S., Giger-Reverdin, S., Rapetti, L., Dvash, L., Dorléans, M., Ungar, E.D., 2008. Data mining old digestibility trials for nutritional monitoring in confined goats with aids of near-infrared spectrometry. Small Rumin. Res. 77, 146–158. Li, H., Tolleson, D., Stuth, J., Bai, K., Mo, F., Kronberg, S., 2007. Faecal near infrared reflectance spectroscopy to predict diet quality for sheep. Small Rumin. Res. 68, 263–268. Kneebone, D.G., Dryden, G.McL., 2015. Prediction of diet quality for sheep from faecal characteristics: comparison of near-infrared spectroscopy and conventional chemistry predictive models. Anim. Prod. Sci. 55, 1–10. ˜ Núnez-Sánchez, N., Martínez Marín, A.L., Pérez Hernández, M., Carrion, D., Gómez Castro, G., Pérez Alba, L.M., 2012. Faecal near infrared spectroscopy (NIRS) as a tool to assess rabbit’s feed digestibility. Livest. Sci. 150, 386–390. Ottavian, M., Franceschin, E., Signorin, E., Segato, S., Berzaghi, P., Contiero, B., Cozzi, G., 2015. Application of near infrared reflectance spectroscopy (NIRS) on faecal samples from lactating dairy cows to assess two levels of concentrate supplementation during summer grazing in alpine pastures. Anim. Feed Sci. Technol. 202, 100–105. Schiborra, A., Bulang, M., Berk, A., Susenbeth, A., Schlecht, E., 2015. Using faecal near-infrared spectroscopy (FNIRS) to estimate nutrient digestibility and chemical composition of diets and faeces of growing pigs. Anim. Feed Sci. Technol. 210, 234–242. Shenk, J.S., Westerhaus, M.O., 1995a. Analysis of Agriculture and Food Products by Near Infrared Reflectance Spectroscopy. Monograph. NIRSystems. Shenk, J.S., Westerhaus, M.O., 1995b. Routine Operation, Calibration, Development and Network System Management Manual. NIRSystems, Inc., 12101 Tech Road, Silver Spring MD 20904, USA. Shenk, J.S., Westerhaus, M.O., 1996. Calibration de ISI way. In: Davies, A.M.C., Williams, P. (Eds.), Near Infrared Spectroscopy: The Future Waves. NIR Publications, Chichester, West Sussex, UK, pp. 198–202. Van Soest, P.J., Robertson, J.B., Lewis, B.A., 1991. Methods for dietary fibre, neutral detergent fibre and non-starch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74, 3583–3597. Williams, P.C., Norris, K., 2001. Near-Infrared Technology in the Agricultural and Food Industries, second ed. American Association of Cereal Chemists, Inc., St. Paul, Minnesota, USA.
Contents lists available at ScienceDirect
Animal Feed Science and Technology journal homepage: www.elsevier.com/locate/anifeedsci
Short communication
Evaluation of botanical and chemical composition of sheep diet by using faecal near infrared spectroscopy ∗ ˜ ˜ Blanco, Valeriano Nieves Núnez-Sánchez , Domingo Carrion, Francisco Pena Domenech García, Ana Garzón Sigler, Andrés Luis Martínez-Marín Departamento de Producción Animal, Universidad de Córdoba, Campus de Rabanales, Carretera Madrid-Cádiz km 396, 14071 Córdoba, Spain
a r t i c l e
i n f o
Article history: Received 12 May 2016 Received in revised form 19 September 2016 Accepted 21 September 2016 Keywords: Near infrared reflectance spectroscopy Faecal NIRS Faeces Diet composition Ewe
a b s t r a c t The aim of this study was to use near infrared reflectance spectroscopy (NIRS) to directly predict the chemical composition and forage content of diets consumed by ewes from the analysis of their faeces. Twenty four ewes (46 ± 3 kg BW) placed in individual cages were used. Each animal was randomly allocated to four diets fed in four successive 28-days periods, so that a total of 16 diets were fed (six ewes per diet). Diets were designed to have different contents of neutral detergent fibre and crude protein. Daily rations of 800 g (as-fed) were formulated with alfalfa hay (0–1000 g/kg) or cereal straw (0–800 g/kg) as forage base, different mixtures of maize, peas and lupins, offered as whole grains, and pelleted sunflower meal. Faeces samples were collected at the end of each feeding period, directly from the rectum. A total of 96 faeces samples were dried and ground. The samples were scanned in small ring cups in a FOSS-NIRSystems 6500 SY-II scanning monochromator. Faecal Modified Partial Least Squares calibration equations, selected in terms of standard error of cross validation (SECV) and coefficient of determination (r2 ), showed good predictive values for the prediction of dietary botanical ingredients (g/kg) alfalfa (41.9 and 0.98), cereal straw (48.6 and 0.97), maize (89.1 and 0.86), and forage (67.4 and 0.90) or concentrate (67.5 and 0.90) respectively, and also for chemical composition of fed diets (g/kg DM): ash (5.2 and 0.93), crude protein (9.6 and 0.89), ether extract (2.9 and 0.80), neutral detergent fibre (34.5 and 0.95) and non-fibre carbohydrates (38.7 and 0.94), respectively. Lower predictive capacity was obtained for peas (78.0 and 0.43), lupins (37.0 and 0.84) and sunflower meal (6.7 and 0.99), probably due to the limited variability of these ingredients in the diets. These results support the viability of faecal NIRS as a fast and reliable analytical method which allows to accurately predict the botanical composition, proportions of forage ingredients and chemical composition of diets consumed by ewes having free access to a variety of feed ingredients. Future works should be focused both to increase the variability of the diet ingredients and also to validate the models with external samples. © 2016 Elsevier B.V. All rights reserved.
Abbreviations: aNDF, neutral detergent fibre; CP, crude protein; DM, dry matter; EE, ether extract; MPLS, modified partial least squares; MSC, multiplicative scatter correction; NFC, non fibrous carbohydrates; NIRS, near infrared reflectance spectroscopy; r2 , coefficient of determination of cross validation; SEC, standard error of calibration; SECV, standard error of cross validation; SEP, standard error of prediction; SNV, standard normal variate; T, outlier samples. ∗ Corresponding author. ˜ E-mail address: [email protected] (N. Núnez-Sánchez). http://dx.doi.org/10.1016/j.anifeedsci.2016.09.010 0377-8401/© 2016 Elsevier B.V. All rights reserved.
2
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
1. Introduction Composition of the food ingested by animals is of paramount importance on the management of extensive farms and for the evaluation of the consumption habits of wild animals (Landau et al., 2006). Farmers need updated information on the nutritive value of forages and supplementary feeds consumed by the animals to manage their nutritional status. This is especially important in the new strategies aiming towards greater efficiency in feeding systems based on the free choice of the food ration, not only the forage but also components of the concentrate (Dixon and Coates, 2009). However, it is difficult to obtain and quantify this information directly, as it requires expensive animal handling that is not always possible. The composition of faeces has been long considered to contain information of the diet consumed by the animal. Hence, the evaluation of faeces has been considered as a good method to monitor many attributes of the fed diet, and some aspects of animal physiology (Dixon and Coates, 2009). Near infrared reflectance spectroscopy (NIRS) has been reported to be a fast, reliable and economic tool to accurately predict different attributes of raw materials and feeds (Williams and Norris, 2001), considered as well as an environmental-friendly analytical solution. NIR analysis of faeces could greatly facilitate the study of the quality and quantity of the diet selected by both domestic and feral grazing herbivores (Dixon and Coates, 2009; Dryden, 2003). In fact, this technology has been successfully used to predict chemical composition and digestibility of the ˜ diet through faeces analysis in pigs (Schiborra et al., 2015), rabbits (Núnez-Sánchez et al., 2012), and ruminants (Landau et al., 2006; Dixon and Coates, 2009). Few works have attempted to assess the ability of NIRS evaluation of faeces to predict diet composition in sheep (Li et al., 2007; Decandia et al., 2009), and botanical composition of the diet (Dixon and Coates, 2009; Landau et al., 2008; Glasser et al., 2008; Kneebone and Dryden, 2015). However, it would be interesting to assess the presence of concentrates in the diet, especially under low pasture availability, as reported for cows (Ottavian et al., 2015). In this case, a qualitative model was developed to discriminate animals among two levels of concentrate supplementation. Hence, it would be interesting to test the ability of faecal NIRS to quantify the amount of forage and concentrates percentages of fed diets, and also predict the proportion of the mixed ingredients. The aim of this work was to develop faecal NIRS calibrations to assess both the botanical and chemical composition of several diets consumed by ewes in a free access basis. 2. Materials and methods 2.1. Samples and reference data determination The experiment was carried out in accordance with the Spanish regulation on experimental animals. Twenty four ˜ sheep (46 ± 3 kg BW), neither lactating nor pregnant, were used in the assay. Each animal was randomly allocated Segurena to four diets fed in four successive periods of 28 days. A total of 16 diets were fed (six ewes per diet). Diets (800 g/d, as-fed) were formulated to have variable contents of neutral detergent fibre (aNDF) and crude protein (CP), by using combinations of forage/non-forage ingredients from 100/0 to 20/80. Forage base was alfalfa (0–1000 g/kg) or cereal straw (0–800 g/kg). In order to get a good digestive and nutritional balance, different mixtures of cereals, legumes and sunflower meal as nonforage ingredients were used: maize (0–800 g/kg), peas (0–300 g/kg) and lupins (0–200 g/kg), offered as whole grains, and pelleted sunflower meal (0–200 g/kg). Visual observation of the troughs revealed that feed offered daily was completely consumed from day to day. Hence, the design of the assay allowed to know the real individual daily intake (g/d) of the different ingredients of fed diets. Individual faecal samples were collected for two consecutive days at the end of each feeding period, directly from the rectum. A pool sample was originated from mixing 2 days of faeces from each ewe. The total 96 individual faecal samples obtained were dried at 100 ◦ C for 24 h and ground to pass through a 1 mm sieve. After that, the samples were frozen at −20 ◦ C until NIR analyses were performed. Dry matter (DM), ash, crude protein (CP) and ether extract (EE) contents of feed ingredients were determined according to methods 930.15, 942.05, 984.13 and 920.39 of AOAC (2000), respectively, whereas aNDF was determined according to Van Soest et al. (1991). Non fibrous carbohydrates (NFC) were calculated by difference method, as recommended by NRC (2001); NFC = 100 − aNDF − ash − CP − EE. 2.2. NIR spectra collection Reflectance spectra from dried and ground faeces samples were obtained on a FOSS-NIRSystems 6500 SY-II scanning monochromator (FOSS-NIRSystems, Silver Spring, MD, USA) equipped with a spinning module. Faeces samples were scanned using small ring cups. Spectral absorbance values were recorded from 400 to 2498 nm, every 2 nm. Two capsules per sample were filled for NIR analysis and the average spectra of two subsamples were used for calibration. WINISI II software, version 1.50 (Infrasoft International LLC, State College, PA) was used for spectral data collection. 2.3. Data processing and calibration development WINISI IV software, version 4.8 (Foss, Denmark) was used for data processing and calibration development. Regression models were performed to predict botanical and chemical composition of fed diets. For that purpose, each faecal spectrum was paired with the information of the diet fed by the corresponding ewe: the proportion (g/kg) of the botanical ingredients
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
3
Table 1 Botanical and chemical composition of the diets used in the assay. Diet
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Mean SD Minimum Maximum
Diet ingredients (g/kg)
Diet composition (g/kg DM)
Alfalfa
Straw
Maize
Peas
Lupins
Sunflower meal
Forage
Concentrate
Ash
CP
EE
aNDF
NFC
1000 800 600 400 200 600 400 0 0 0 0 0 0 0 0 0 250.0 338.6 0 1000
0 0 0 0 0 0 0 800 800 600 600 600 400 400 400 200 300.0 309.8 0 800
0 200 400 600 800 200 300 0 0 0 200 200 400 400 400 600 293.8 240.7 0 800
0 0 0 0 0 200 300 0 0 200 200 0 200 0 0 0 68.8 107.8 0 300
0 0 0 0 0 0 0 0 200 200 0 200 0 200 0 200 62.5 95.7 0 200
0 0 0 0 0 0 0 200 0 0 0 0 0 0 200 0 25.0 68.3 0 200
1000 800 600 400 200 600 400 800 800 600 600 600 400 400 400 200 550.0 225.1 200 1000
0 200 400 600 800 400 600 200 200 400 400 400 600 600 600 800 450.0 225.1 0 800
104.6 86.4 68.2 50.0 31.8 71.7 55.3 88.0 74.7 63.9 60.1 60.3 45.7 46.0 59.3 31.7 62.4 19.9 31.7 104.6
174.8 156.8 138.8 120.8 102.8 170.0 167.7 90.6 102.4 143.0 87.8 111.8 97.3 121.2 109.5 130.6 126.6 29.0 87.8 174.8
11.9 16.7 21.6 26.5 31.3 16.6 18.9 10.6 19.5 19.5 15.9 24.5 21.0 29.6 20.8 34.7 21.2 6.7 10.6 34.7
460.6 390.7 320.7 250.8 180.8 343.4 284.8 729.3 685.0 572.0 540.8 549.3 405.1 413.5 457.8 277.7 428.9 155.6 180.8 729.3
248.0 349.4 450.8 552.2 653.6 398.2 473.3 81.8 118.8 201.8 295.6 254.4 431.2 390.0 353.0 525.6 361.1 156.6 81.8 653.6
SD: standard deviation; CP: crude protein; EE: Ether extract; aNDF: neutral detergent fibre; NFC: Non fibrous carbohydrates (100 − aNDF − ash − CP − EE).
(alfalfa, cereal straw, maize, peas, lupins and sunflower meal) and the chemical composition (g/kg DM) (ash, CP, EE, aNDF and NFC) of the diets. Furthermore, models to predict the forage and concentrate (g/kg) contents of the diet were also developed. Regression models for all the parameters were obtained using the Modified Partial Least Squares (MPLS) regression method (Shenk and Westerhaus, 1995a). Separate MPLS calibrations were performed for each parameter in the calibration set. When developing MPLS equations, the use of cross-validation is recommended in order to select the optimal number of factors and avoid overfitting (Shenk and Westerhaus, 1995a,b, 1996). Given the hierarchical structure of the samples, with 16 diets and 6 ewes for each diet, a leave-out-one-group cross-validation strategy was used for all the calibrations. Therefore, the sample set was divided into 16 cross-validation groups, in a way that calibrations were developed, each time, with samples from 15 diets, and were validated with the samples belonging to the remaining diet until all groups of samples have been used for validation. A set of calibrations for each of the parameters under study have been developed after combining, in different wavelength ranges (400–2500 nm and 1100–2500 nm), pretreatments for scatter correction of the data (algorithms SNV and Detrend, and MSC) and Norris–Williams spectral derivatives, to remove additive and multiplicative effects in the spectra. Spectral derivatives are named by using four digits: the first digit is the order of the derivative, the second is the gap over which the derivative is calculated, the third is the number of data points in a running average (or smoothing), and the fourth is the second smoothing. Four of the most usual derivative pretreatments (1, 5, 5, 1; 1, 10, 10, 1; 2, 5, 5, 1 and ˜ et al. (2012), following Shenk and Westerhaus 2, 10, 10, 1), have been used in this study, as described in Núnez-Sánchez (1995a,b, 1996) recommendations for agricultural products. Hence, sixteen regression equations per parameter/variable were developed by combining four spectral derivative math treatments and two scatter correction methods in two spectral regions. Two passes of elimination of T outlier samples (samples with abnormally high residuals of predicted versus reference values), were applied. A critical T value of 2.5 was set for T outliers. The statistics used to select the best calibration models were: the standard error of calibration (SEC), the standard error of cross validation (SECV) and the coefficient of determination of cross validation (r2 ). The standard error or prediction (SEP) is the best estimate of the prediction capability of the equation and should be ideally estimated by applying the model to an independent validation set. However, as the calibration set of this work is not sizeable, it was not possible to get that statistic. Hence, SECV has been used to evaluate the predictive capacity of the models, although, as previously mentioned, external validation of the equations should be done in the future with an independent set of samples. Therefore, the best NIR equation per parameter was selected based on the higher r2 value and lower SECV (Shenk and Westerhaus, 1995a,b, 1996). 3. Results and discussion 3.1. Botanical and chemical composition of diets The chemical composition of each diet was calculated from the chemical information of their ingredients, determined in a laboratory. The design of the diets, with different mixtures of forages and concentrates, resulted in a sizeable variability of both the botanical and the chemical composition range for all parameters, except for peas, lupins and sunflower meal, with
4
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
Table 2 Faecal NIR calibrations to estimate botanical and chemical composition of diets. Diet composition
Spectral range
Mathematical derivation treatment
Scatter correction
PLS terms
N
MEAN
SD
SECV
r2
Alfalfa(g/kg) Cereal straw (g/kg) Maize (g/kg) Peas (g/kg) Lupins (g/kg) Sunflower meal (g/kg) Forage (g/kg) Concentrate (g/kg) Ash (g/kg DM) CP (g/kg DM) EE (g/kg DM) aNDF (g/kg DM) NFC (g/kg DM)
1100–2500 400–2500 400–2500 400–2500 1100–2500 400–2500 1100–2500 1100–2500 400–2500 400–2500 400–2500 400–2500 400–2500
2,5,5,1 2,10,10,1 1,5,5,1 1,10,10,1 1,5,5,1 1,5,5,1 2,5,5,1 2,5,5,1 2,10,10,1 1,5,5,1 2,5,5,1 2,10,10,1 2,10,10,1
SNV + DT SNV + DT SNV + DT SNV + DT MSC MSC MSC SNV + DT MSC SNV + DT SNV + DT MSC SNV + DT
10 11 10 6 11 12 11 11 9 6 9 10 9
92 95 93 94 89 92 96 96 94 91 94 92 92
245.7 298.9 289.2 66.0 65.2 17.4 550.0 450.0 62.6 126.3 21.3 431.5 357.5
333.0 303.0 235.7 104.3 94.3 56.7 219.1 219.1 19.5 28.9 6.6 153.1 154.3
41.9 48.6 89.1 78.0 37.0 6.7 67.4 67.5 5.2 9.6 2.9 34.5 38.7
0.98 0.97 0.86 0.43 0.84 0.99 0.90 0.90 0.93 0.89 0.80 0.95 0.94
Mathematical derivation treatments: the first digit is the number of the derivative, the second is the gap over which the derivative is calculated, the third is the number of data points in a running average or smoothing and the fourth is the second smoothing. MSC: multiplicative scatter correction; SNV + DT: standard normal variate and detrend; N: number of samples in the calibration set; Mean: mean of the calibration set; SD: standard deviation of the calibration set; SECV: standard error of cross-validation; r2 : coefficient of determination of cross validation; CP: crude protein; EE: Ether extract; aNDF: neutral detergent fibre; NFC: Non fibrous carbohydrates.
only two or three levels of inclusion (Table 1). The use of cereal straw in the present work has resulted in a wide range of variability for aNDF. 3.2. NIR calibration results A calibration set of 96 faecal samples from ewes fed 16 different diets was used. MPLS regression models were developed with the complete set of samples 96 samples. 3.2.1. Botanical composition Excellent precision has been obtained for all the parameters, as they have r2 values close to or above 0.9 according to Shenk and Westerhaus (1996), with the exception of the model to predict the percentage of peas consumed by the ewes (r2 = 0.43). A possible explanation of this poor result could be that, in this case, only 5 diets (30 samples) contained this ingredient and always in low percentages (20–30%). The study of the T outliers in the calibration process, and the predicted values obtained after the application of the peas model to the average spectrum of faecal samples from each diet (data not shown) revealed that, in two cases, the model was not able to discriminate this ingredient, from maize and from lupins, respectively. Values of bias, slope and R2 obtained in validation (63.1, 1.2 and 0.67 in peas and 37.5, 0.96 and 0.85 in lupins), were poorer compared with the rest of the parameters. However, although only 2 diets (12 faecal samples) included sunflower meal, the models showed good predictive results of this ingredient (Table 2). A possible explanation could be that the contribution of this ingredient to the diets provided not only chemical (Table 1), but also clearly physical changes to the diets. Absorbance data of NIR spectra depend upon both physic and chemical properties of the product (Williams and Norris, 2001). That particularity in the faecal spectra from those diets could have contributed to an easier quantification of this ingredient, when present. However, it must be pointed out that, despite these favorable statistics, the model to predict the inclusion of sunflower meal in the diet would only be usable for diets with similar levels of inclusion to the ones used here (0 or 200 g/kg). Extensively, the same limitations should be considered for the ingredients peas and lupins. The results obtained in this study suggest that diet residues present in ewes faeces had sufficient spectral information to accurately predict dietary botanical components (Dixon and Coates, 2009). Similar SECV and r2 results were obtained in goats to predict the dietary percentages of hay, corn stover (comparable to cereal straw) and the percentage of concentrate from the faecal NIR analysis (Landau et al., 2004, 2008). Ottavian et al. (2015) developed a faecal NIR qualitative model in dairy cows to discriminate between two levels of concentrate supplementation. However, none of the previous reviewed reports have attempted the quantitative prediction of the ingredients of the concentrates in the diet. The design of the study permitted to know the exact individual intake (g/d) of the different ingredients of the diet, as well as the calculation of dry matter intake (DMI). As an attempt to check the predictive reliability of the calibrations to predict botanical ingredients, a graph comparing the real daily intake (g/d) of DM, forage, concentrates and the different ingredients with the intakes calculated from the predictions of the dietary percentages of each diet has been included (Fig. 1). The results confirm the reliability of the models to predict all the above mentioned parameters except peas and lupins. The results obtained in this study suggest that diet residues present in ewes faeces had sufficient spectral information to accurately predict dietary botanical components (Dixon and Coates, 2009). Similar results were obtained in goats to predict the dietary percentages of hay, corn stover (comparable to cereal straw) and the percentage of concentrate from the faecal NIR analysis (Landau et al., 2004, 2008). Ottavian et al. (2015) developed a faecal NIR qualitative model in dairy cows to discriminate between two levels of concentrate supplementation. However, none of the previous reviewed reports have attempted the quantitative prediction of the ingredients of the concentrates in the diet.
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
5
Fig. 1. Scatter plot of actual versus estimated individual daily intakes (DI, expressed in grams per day (g/d)) of different dietary ingredients. Estimated individual daily intakes were calculated by dividing the daily amount of feed ingested (800 g) by the NIR predicted dietary fractions of the corresponding diet.
3.2.2. Chemical composition Faecal spectra were also used to predict the chemical composition of the diets consumed by the animals. The equations showed good precision, with r2 values close to or above 0.9 for ash, CP, aNDF and NFC (Table 2), while lower prediction ability in terms of r2 (0.80) has been obtained for the prediction of EE. In all cases, better results have been obtained using the complete spectral range, whereas the scatter and derivative pre-treatments did not offer differences. Contrarily to botanical composition, many studies have attempted to predict the chemical composition of the consumed diet from the NIR analysis of the faeces, mainly for goats and cows (Landau et al., 2006; Dixon and Coates, 2009). With respect to sheep, Li et al. (2007) obtained less accurate results, with higher SECV values for CP prediction, while Decandia et al. (2009) reported more precise equations, with lower SECV values, for the prediction of CP, aNDF and NFC. Therefore, the variability of the diets in this assay permitted to verify that NIR analysis of their faeces was able to estimate and differentiate the percentages of the forage (alfalfa hay or cereal straw) and concentrates, even in low amounts, on the diet ingested by ewes. Moreover, the precise and accurate prediction of the nutritional characteristics of fed diets, from rich to poor aNDF and CP contents, would be useful to assess the nutritional status and behavior of animals with free access to different feedstocks. 4. Conclusion The use of confined animals fed diverse diets composed by mixtures of forage and concentrates provided a varied database in terms of CP and aNDF contents. These results support the viability of faecal NIRS as a fast and reliable analytical method to monitor with high precision and accuracy the botanical composition and chemical attributes of the diet freely consumed by ewes, thus permitting, if necessary, the implementation of appropriate changes to the supplementation regime. Therefore, the spectra database created with a wide variability in terms of ingredients and chemical composition could be used in the future to assess the feeds chosen by animals with free access to similar diets. Future works should be focused both to increase the variability of the ingredients used in the diets and also to validate the models obtained here with external samples. Conflict of interest None declared.
6
N. Nú˜ nez-Sánchez et al. / Animal Feed Science and Technology 222 (2016) 1–6
Acknowledgements The authors thank José Javier Pérez Hernández and Inmaculada Andújar Ramírez for their support in the management of the animals and technical assistance in the NIR analyses. The study was performed using NIR equipment and facilities belonging to SCAI (NIR/MIR Unity), University of Cordoba (Spain). References AOAC, 2000. Official Methods of Analysis, seventeenth ed. Association of Official Analytical Chemists, Gaithersburg, MD. Decandia, M., Giovanetti, V., Boe, F., Scanu, G., Cabiddu, A., Molle, G., Cannas, A., Landau, S., 2009. Faecal NIRS to assess the chemical composition and the nutritive value of dairy sheep diet. Nutritional and Foraging Ecology of Sheep and Goats. Papachristou, T.G., Parissi, Z.M., Ben Salem, H., Morand-Fehr, P. (Eds). Options Méditerranéennes, Series A, No. 85, 135–139. Dixon, R.M., Coates, D.B., 2009. Review: near infrared spectroscopy of faeces to evaluate the nutrition and physiology of herbivores. J. Near Infrared Spectrosc. 17, 1–31. Dryden, G. McL., 2003. Near Infrared Reflectance Spectroscopy: Applications in Deer Nutrition. Rural Industries Research and Development Corporation, Canberra, Publication No. W03/007. Glasser, T., Landau, S., Ungar, E.D., Perevolotsky, A., Dvash, L., Muklada, H., Kababya, D., Walker, J.W., 2008. A fecal near-infrared reflectance spectroscopy-aided methodology to determine goat dietary composition in a Mediterranean shrubland. J. Anim. Sci. 86, 1345–1356. Landau, S., Glasser, T., Dvash, L., Perevolotsky, A., 2004. Fecal NIRS to monitor the diet of Mediterranean goats. S. Afr. J. Anim. Sci. 34, 76–80. Landau, S., Glasser, T., Dvash, L., 2006. Monitoring nutrition in small ruminants with the aid of near infrared reflectance spectroscopy (NIRS) technology: a review. Small Rumin. Res. 61, 1–11. Landau, S., Giger-Reverdin, S., Rapetti, L., Dvash, L., Dorléans, M., Ungar, E.D., 2008. Data mining old digestibility trials for nutritional monitoring in confined goats with aids of near-infrared spectrometry. Small Rumin. Res. 77, 146–158. Li, H., Tolleson, D., Stuth, J., Bai, K., Mo, F., Kronberg, S., 2007. Faecal near infrared reflectance spectroscopy to predict diet quality for sheep. Small Rumin. Res. 68, 263–268. Kneebone, D.G., Dryden, G.McL., 2015. Prediction of diet quality for sheep from faecal characteristics: comparison of near-infrared spectroscopy and conventional chemistry predictive models. Anim. Prod. Sci. 55, 1–10. ˜ Núnez-Sánchez, N., Martínez Marín, A.L., Pérez Hernández, M., Carrion, D., Gómez Castro, G., Pérez Alba, L.M., 2012. Faecal near infrared spectroscopy (NIRS) as a tool to assess rabbit’s feed digestibility. Livest. Sci. 150, 386–390. Ottavian, M., Franceschin, E., Signorin, E., Segato, S., Berzaghi, P., Contiero, B., Cozzi, G., 2015. Application of near infrared reflectance spectroscopy (NIRS) on faecal samples from lactating dairy cows to assess two levels of concentrate supplementation during summer grazing in alpine pastures. Anim. Feed Sci. Technol. 202, 100–105. Schiborra, A., Bulang, M., Berk, A., Susenbeth, A., Schlecht, E., 2015. Using faecal near-infrared spectroscopy (FNIRS) to estimate nutrient digestibility and chemical composition of diets and faeces of growing pigs. Anim. Feed Sci. Technol. 210, 234–242. Shenk, J.S., Westerhaus, M.O., 1995a. Analysis of Agriculture and Food Products by Near Infrared Reflectance Spectroscopy. Monograph. NIRSystems. Shenk, J.S., Westerhaus, M.O., 1995b. Routine Operation, Calibration, Development and Network System Management Manual. NIRSystems, Inc., 12101 Tech Road, Silver Spring MD 20904, USA. Shenk, J.S., Westerhaus, M.O., 1996. Calibration de ISI way. In: Davies, A.M.C., Williams, P. (Eds.), Near Infrared Spectroscopy: The Future Waves. NIR Publications, Chichester, West Sussex, UK, pp. 198–202. Van Soest, P.J., Robertson, J.B., Lewis, B.A., 1991. Methods for dietary fibre, neutral detergent fibre and non-starch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74, 3583–3597. Williams, P.C., Norris, K., 2001. Near-Infrared Technology in the Agricultural and Food Industries, second ed. American Association of Cereal Chemists, Inc., St. Paul, Minnesota, USA.