Hostname: page-component-586b7cd67f-l7hp2 Total loading time: 0 Render date: 2024-11-22T16:02:37.832Z Has data issue: false hasContentIssue false

Feasibility study on the potential of visible and near infrared reflectance spectroscopy to measure alpaca fibre characteristics

Published online by Cambridge University Press:  01 July 2007

M. Gishen
Affiliation:
Shingleback Ridge Alpacas, PO Box 69, Hahndorf, SA 5245, Australia The Australian Wine Research Institute, PO Box 197 Glen Osmond SA 5064, Australia
D. Cozzolino*
Affiliation:
The Australian Wine Research Institute, PO Box 197 Glen Osmond SA 5064, Australia

Abstract

Visible (Vis) and near infrared (NIR) reflectance spectroscopy is a rapid and non-destructive technique that has found many applications in assessing the quality of agricultural commodities, including wool. In this study, Vis and NIR spectroscopy combined with multivariate data analysis was investigated regarding its feasibility in predicting a range of fibre characteristics in raw alpaca wool samples. Mid-side samples (n = 149) were taken from alpacas from a range of colours and ages at shearing time over 4 years (2000 to 2004) and subsequently analysed for fibre characteristics such as mean fibre diameter (MFD) and standard deviation (and coefficient of variation), spin fineness, curvature degree (and standard deviation), comfort factor, medullation percentage (by weight and number in white samples only) using traditional reference laboratory testing methods. Samples were scanned in a large cuvette using a FOSS NIRSystems 6500 monochromator instrument in reflectance mode in the Vis and NIR regions (400 to 2500 nm). Partial least squares (PLS) regression was used to develop a number of calibration models between the spectral and reference data. Mathematical pre-treatment of the spectra (second derivative) as well as various combinations of wavelength range were used in model development. The best calibration model was found when using the NIR region (1100 to 2500 nm) for the prediction of MFD, which had a coefficient of determination in cross-validation (R2) of 0.88 with a root mean square standard error of cross validation (RMSECV) of 2.62 μm. The results show the NIR technique to have promise as a semi-quantitative method for screening purposes. The lack of grease in alpaca wool samples suggests that the technique might find ready application as a rapid measurement technique for preliminary classing of shorn fleeces or, if used directly on the animal, the technology might offer an objective tool to assist in the selection of animals in breeding programmes or shows.

Type
Full Paper
Copyright
Copyright © The Animal Consortium 2007

Introduction

The alpaca (Lama pacos) is commercially the most important fibre producer of the South American camelid family (Lupton et al., Reference Lupton, McColl and Stobart2006). Alpacas were introduced into Australia during the 1860s but the industry failed to establish and by the 1880s, the farmed alpaca population was all but extinct (McGregor, Reference McGregor2002). With the re-introductions of alpacas to Australia in the late 1980s, members of the alpaca industry have made numerous claims regarding the use of their fibre (McGregor, Reference McGregor2002). If alpaca fibre production is to become a viable alternative industry to sheep wool production, it is essential that easy and cheap methods be available for the alpaca’s breeders, wool producers and textile industry.

Near infrared (NIR) spectroscopy coupled with chemometric methods such as principal components analysis (PCA) and partial least squares (PLS) regression have seen a rapid increase in industrial use (Cowe and McNicol, Reference Cowe and McNicol1985; Monin, Reference Monin1998; Cleve et al., Reference Cleve, Bach and Schollmeyer2000; McClure, Reference McClure2003). NIR spectroscopy has been used for both at-line and on-line applications in the textile industry, including the measurement of residual wool wax, moisture, synthetic/natural fibre blend ratios and the treatment levels of various chemical processes (Connell and Brown, Reference Connell and Brown1978; Connell, Reference Connell1983; Coleman et al., Reference Coleman, Lupton, Pfeiffer, Minikhiem and Hart1999; Church and O’Neill, Reference Church and O’Neill1999; Hammersley and Townsend, Reference Hammersley and Townsend2004; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005). Spectroscopy (both the visible (Vis) and NIR regions) has become a more attractive analytical technique for measuring quality parameters in foods and agricultural products with decreasing instrument prices, simultaneous determination of several quality parameters, the ability to replace expensive and slower reference techniques and the lack of need for sample preparation and minimal data analysis (McCaig, Reference McCaig2002; McClure, Reference McClure2003). More recent developments have exploited the improvements in NIR instrumentation (McClure, Reference McClure2003). Extensions of the wavelength range into the Vis region enable the measurement of colour in wool (Hammersley, Reference Hammersley1992; Hammersley and Townsend, Reference Hammersley and Townsend1994 and Reference Hammersley and Townsend2004; Hammersley et al., Reference Hammersley, Townsend, Graystone and Ranford1995; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005). Spectroscopy in the NIR region will provide information about the relative proportions of C–H, N–H and O–H bonds that are the primary constituents of the organic molecules (Murray, Reference Murray1993; Deaville and Flinn, Reference Deaville and Flinn2000). Many reports are available in the literature related to the use of NIR to predict several parameters in sheep wool samples such as residual grease, moisture content, colour, fibre diameter, medullation and vegetable matter content (Slack-Smith et al., Reference Slack-Smith, Fong and Douglas1979; Larsen and Kinnison, Reference Larsen and Kinnison1982; Hammersley, Reference Hammersley1992; Hammersley and Townsend, Reference Hammersley and Townsend1994 and Reference Hammersley and Townsend2004; Hammersley et al., Reference Hammersley, Townsend, Graystone and Ranford1995; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005).

In this study, the potential use of Vis and NIR spectroscopy combined with multivariate data analysis was investigated for its feasibility in predicting a range of fibre characteristics in raw alpaca wool samples.

Material and methods

Wool samples

Mid-side (approx. 100 g) samples (n = 149) were taken from alpacas (L. pacos) (male and female) of a range of colours and ages at the time of shearing over 4 years (2000 to 2004) and divided for subsequent analysis for fibre characteristics using traditional laboratory testing methods. Fibre characteristics measured included, mean fibre diameter (MFD) with its standard deviation (SDFDM) and MFD coefficient of variation (CVFD) of fibres coarser than 30 μm (%F > 30 μm), spin fineness (SPINF), curvature (degree and standard deviation), comfort factor, medullation percentage (by weight and number in white samples only) using the OFDA100 method at commercial fibre-testing service providers (OFDA, IWTO-47-95, 1995) (Sommerville, 2002). In addition, total fleece weight (TFW) (kg) and staple length (SL) (mm) were also measured. SL was estimated on 12 staples measured to the nearest 0.5 cm according to methods described elsewhere (AS 3535-1988, 1988) (Sommerville, 2002).

Visible and near infrared spectroscopy

Samples were scanned over the Vis and NIR regions (400 to 2500 nm) in a large cuvette (260 mm × 55 mm × 33 mm) using a FOSS NIRSystems6500 monochromator instrument (FOSS NIRSystems, Silver Spring, MD, USA) in reflectance mode. Each spectrum was collected at 2 nm intervals (1050 data points) and reflectance (R) data was stored as log (1/R). The spectrum of each sample was the average of 32 successive scans.

Multivariate analysis

PCA was used to study patterns in the NIR raw spectra of the wool samples analysed. Calibration models between reference values and spectral data were developed using PLS with full cross validation. Spectral data was transformed into a NSAS format and exported into The Unscrambler (1996) software for multivariate analysis. PCA was performed before PLS regression models were developed (Martens and Naes, Reference Martens and Naes1989; Naes et al., Reference Naes, Isaksson, Fearn and Davies2002). PCA was used to derive the first principal components (PCs) from the spectral data to examine the possible grouping of samples and to detect possible spectral outliers before using the data set to develop the PLS regression models (Martens and Naes, Reference Martens and Naes1989). No mathematical treatments or spectral transformations were applied when PCA was performed. After PCA analysis, each spectrum was smoothed by a 19-point Savitzky–Golay (SG) second-order filtering operation and transformed using the second derivative (SG) in order to avoid noise and complexity of the spectrum (The Unscrambler, 1996). The optimum number of latent variables/terms in the PLS calibration models were determined by cross validation and defined by the PRESS function (prediction residual error sum of squares) (Naes et al., Reference Naes, Isaksson, Fearn and Davies2002). Due to the limited number of samples available, calibration models were developed and evaluated using full cross validation (Martens and Naes, Reference Martens and Naes1989; Martens and Dardenne, Reference Martens and Dardenne1998; Naes et al., Reference Naes, Isaksson, Fearn and Davies2002). The resulting calibration equations between the chemical analyses and the Vis and NIR data were evaluated based on the coefficient of determination in calibration (R 2) and the root mean square of the standard error in cross validation (RMSECV). The ratio of standard deviation and RMSECV, namely residual predictive value (RPD) was used to test the accuracy of the calibration models (Williams, Reference Williams2001).

An RPD value greater than three was considered adequate for analytical purposes in most of NIR applications for agricultural products (Williams, Reference Williams2001; Fearn, Reference Fearn2002).

Results and discussion

Figure 1 shows the mean and standard deviation spectrum in the NIR region of the alpaca wool samples. The mean spectrum of wool samples showed absorption bands at 1200, 1430 and 1930 nm related with O–H stretch overtones, mainly related with moisture of wool samples (Murray, Reference Murray1986; Osborne et al., Reference Osborne, Fearn and Hindle1993; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005). Absorption bands around 1700 nm associated with either S–H first overtones or C–H stretch overtones of lipids and fatty acids and at 2120 and 2300 nm they were related to C–H deformation and combination tones associated with amino acids (Murray, Reference Murray1986; Osborne et al., Reference Osborne, Fearn and Hindle1993; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005). Standard deviation between samples was observed around 1400 and 1900 nm, related with the O–H overtones (water content) (Murray, Reference Murray1986; Osborne et al., Reference Osborne, Fearn and Hindle1993; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005). PCA is a statistical data analytical technique that has recently seen application in the textile and food industries (Naes et al., Reference Naes, Isaksson, Fearn and Davies2002) and involves identifying the components accounting for variance within a data set, averaging the spectral set and then comparing this to each spectrum. Additionally, it creates a synthetic spectrum which accounts for the largest part of the variance within the spectral set that produces the first loading vector or principal component (PC). The scaling factor that represents the amount of the loading vector in each of the spectra in the data set is known as the score. Multiplying the loading vector by the score for each spectra and subtracting this from the original spectra produces a new spectral set (Naes et al., Reference Naes, Isaksson, Fearn and Davies2002). This allowed us to interrogate the spectra in order to investigate for relevant patterns in the set of wool samples analysed. Figure 2 shows the score plot of the first two PCs of the alpaca wool samples analysed by NIR spectroscopy. The first two PC accounted for 76% of the variance of the spectra of the wool samples, PC1 57% and PC2 19%, respectively. The PC score plot shows that no effect of the year was observed in the wool samples analysed, but a separation between wool samples related to fibre colour was evident by the observation of the spectra.

Figure 1 Near infrared mean spectrum (line) and standard deviation (dotted line) of alpaca wool samples analysed (raw spectra, 500 to 2500 nm).

Figure 2 Principal component score plot of alpaca wool samples analysed by visible and near infrared spectroscopy, labelled by year.

Table 1 shows the mean, standard deviation and range of the fibre characteristics analysed in the wool samples. The broad range in chemical composition did not reflect any year effect (ANOVA, data not presented). Table 2 shows the Pearson correlations (P > 0.05) between the fibre parameters measured in alpaca wool samples. It was interesting to note that MFD was inversely correlated with SL (r = −0.10) and curvature (i.e. crimp) (r = −0.51). High and positive correlations were found between SPINF and MFD (r = 0.99) and SPINF and SDFDM (r = 0.92). It was also observed that animal age was correlated positively with MFD, and negatively with SL.

Table 1 Descriptive statistics for the wool parameters measured in alpaca wool samples

Abbreviations are: MFD = mean fibre diameter, SDFDM = standard deviation of fibre diameter, CVFD = coefficient of variation in fibre diameter, SPINF = spin fineness, CURVEDEG = curvature, CURVESD = standard deviation of curvature, PMEDNUM = percent medullation by number, PEMWET = percent medullation by weight, TFW = total fleece weight, SL = staple length, AGEATSHEAR = age of animal at shear, s.d. = standard deviation, Min = minimum, Max = maximum, CV = coefficient of variation = (s.d./mean) × 100, R = coefficient of correlation in calibration, RMSECV = root mean square of the standard error of cross validation, RPD = SD/RMSECV, LV = latent variables/number of PLS terms used to develop the calibration models.

Table 2 Pearson correlation between fibre characteristics measured in alpaca wool samples

For abbreviations see Table 1 footnote. High correlations are in bold.

Table 3 shows the NIR calibration statistics for the fibre characteristics measured on the alpaca wool samples. The best calibration model was found when using the NIR region for the prediction of MDF, which had a coefficient of determination in cross-validation (R 2) of 0.88 with a standard error of prediction of 2.62 μm. The value for the RPD was 2, meaning that the technique has promise as a semi-quantitative method for screening purposes. The R 2 and RMSECV for the other parameters measured in alpaca wool were 0.75 (RMSECV: 0.86 μm) for SDFDM; 0.84 (RMSECV: 2.81) for the SPINF; 0.74 (RMSECV: 3.39) for curvature (CURVEDEG); 0.46 (RMSECV: 3.3) for the standard deviation of curvature (CURVESD); 0.76 (RMSECV: 15.7 μm) for the per cent medullation by number (PMEDNUM); 0.86 (RMSECV: 14.42% w/w) for the per cent medullation by weight (PEMWET); 0.70 (RMSECV: 0.61 kg) for the TFW and 0.75 (RMSECV: 31.9 mm) for SL, respectively. The RPD obtained for those calibrations were lower than 3, indicating that they only can be used for rough screening. Other workers have found that fibre diameter in greasy wool samples was poorly predicted with NIR, while clean wool showed good relationships (Hammersley and Townsend, Reference Hammersley and Townsend1994 and Reference Hammersley and Townsend2004; Hammersley et al., Reference Hammersley, Townsend, Graystone and Ranford1995; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005). The lack of grease in alpaca wool samples suggests that the technique might find ready application as a rapid measurement technique for preliminary classing of shorn fleeces or, if used directly on the animal, the technology might offer an objective tool to assist in the selection of animals in breeding programs or shows.

Table 3 Near infrared calibration statistics (cross validation) for the wool parameters measured in alpaca wool samples

For abbreviations see Table 1 footnote.

Figure 3 shows the PLS loadings for the optimal calibrations models developed. It was observed that PLS loadings for MDF were opposite to per cent medullation by number (PMEDNUM). Highest loadings were observed around 1400 and 1900 nm related to O–H tones (mainly water), around 1700 nm related to C–H first overtones, mainly related with fatty acids and grease in the wool samples, and around 2070 and 2200 nm related to C–H combinations tones (Murray, Reference Murray1986; Miller, Reference Miller2001; Cozzolino et al., Reference Cozzolino, Montossi and San Julian2005).

Figure 3 Partial least squares (PLS) loadings of the optimal calibrations for mean fibre diameter (MFD), total fleece weight (TFW), curvature (CURVEDEG) and per cent medullation by number (PMEDNUM) measured by near infrared spectroscopy (1100 to 2500 nm). Optimal number of PLS loadings in brackets.

It is well known that in practice, qualitative or semi-quantitative NIR analyses tend to be less demanding and more straightforward to develop and maintain than quantitative methods, and they can provide information that is very useful in the qualitative assessment of incoming raw material as in the case of wool (Brimmer and Hall, Reference Brimmer and Hall2001). In the industrial manufacturing environment, these types of simple qualitative checks help ensure that products are kept within specification, which reduces the amount of off-specification material produced or increases the number of samples being analysed. The use of those calibrations might have great interest when the objective is to measure wool fibre diameter for screening purposes, animal selection or for use in breeding programmes, where the accuracy in fiber diameter is not as important as having a rapid and low cost method, when a high throughput sample system is required for on-farm analysis. However, this kind of approach will be used only to determine whether the wool has a low, medium or high fibre diameter.

Conclusion

This preliminary study showed that NIR spectroscopy has potential as a rapid analytical tool for determining MFD of alpaca wool. Thousands of wool samples are generated by animal selection and breeding programmes which commonly look to develop progress in several traits simultaneously. Low-cost, rapid methods are required for such programmes. The technique might find ready application as a rapid measurement technique for preliminary classing of shorn fleeces or, if used directly on the animal, as an objective tool to assist in the selection of animals in breeding programmes or at animal shows. Reflectance spectroscopy makes NIR spectroscopy an ideal tool for screening. Further research is required to improve and validate the calibration generated using experimental conditions if the technology is to be used by the industry.

Acknowledgements

The authors acknowledge The Australian Wine Research Institute for providing access to the visible and near infrared spectrophotometer. Comments and suggestions made in the manuscript by the editorial reviewers are also acknowledged.

References

AS 3535-1988. 1998. Wool measurements of the resistance to compression. Standards Association of Australia, Melbourne, pp 8.Google Scholar
Brimmer, PJ, Hall, JW 2001. Methods development and implementation of near infrared spectroscopy in industrial manufacturing support laboratories. In Near-infrared technology in the agricultural and food industries (ed. PC Williams and KH Norris), pp. 187199. American Association of Cereal Chemists, St Paul, MN, USA.Google Scholar
Church, JS, O’Neill, JA 1999. The detection of polymeric contaminants in loose scoured wool. Vibrational Spectroscopy 19, 285293.CrossRefGoogle Scholar
Cleve, E, Bach, E, Schollmeyer, E 2000. Using chemometrics methods and NIR spectrophotometry in the textile industry. Analytica Chimica Acta 420, 163167.CrossRefGoogle Scholar
Coleman, SW, Lupton, CJ, Pfeiffer, FA, Minikhiem, DL, Hart, SP 1999. Prediction of clean mohair, fiber diameter, vegetable matter, and medullated fiber with near-infrared spectroscopy. Journal of Animal Science 77, 25942602.Google Scholar
Connell, JP 1983. Predicting wool base of greasy wool by near infrared reflectance spectroscopy. Textile Research Journal 53, 651655.Google Scholar
Connell, JP, Brown, OH 1978. The yield testing of wool by reflectance spectroscopy. Journal of Textile Institute 69, 357363.Google Scholar
Cowe, IA, McNicol, JW 1985. The use of principal components in the analysis of near infrared spectra. Applied Spectroscopy 39, 257265.CrossRefGoogle Scholar
Cozzolino, D, Montossi, F, San Julian, R 2005. The use of visible (VIS) and near infrared (NIR) reflectance spectroscopy to predict fibre diameter in both clean and greasy wool. Animal Science 80, 333338.CrossRefGoogle Scholar
Deaville, ER, Flinn, PC 2000. Near infrared (NIR) spectroscopy: an alternative approach for the estimation of forage quality and voluntary intake. In Forage evaluation in ruminant nutrition (ed. DI Givens, E Owen, RFE Axford and HM Omed), pp. 301320. CABI Publishing, Wallingford, UK.CrossRefGoogle Scholar
Fearn, T 2002. Assessing calibrations: SEP, RPD, RER and R2. NIR News 13, 1214.CrossRefGoogle Scholar
Hammersley, MJ 1992. NIR analysis of wool. In Handbook of near-infrared analysis (ed. Burns DA and EW Ciurczak), pp. 475494. Marcel Dekker Inc., New York.Google Scholar
Hammersley, MJ, Townsend, PE 1994. Exploiting the shorter wavelengths: wool measurements including colour. In Leaping ahead with near infrared spectroscopy (ed. GD Batten, PC Flinn, LA Welsh and Blakeney), pp. 465469. Royal Australian Chemical Institute, Melbourne, Australia.Google Scholar
Hammersley, MJ, Townsend, PE 2004. Applications in the analysis of wool. In Near infrared spectroscopy in agriculture (ed. CA Roberts, J Workman and JB Reeves), agronomy monograph no. 44, pp. 663671. American Society of Agronomy, Crop Science Society of America, Soil Science Society of America.Google Scholar
Hammersley, MJ, Townsend, PE, Graystone, GF, Ranford, SL 1995. Visible/near infrared spectroscopy of scoured wool. Textile Research Journal 65, 241246.CrossRefGoogle Scholar
Larsen, SA, Kinnison, JL 1982. Estimating quality components of natural fibres by near infrared reflectance. Clean wool base and average wool fibre diameter. Textile Research Journal 52, 2531.Google Scholar
Lupton, CJ, McColl, A, Stobart, RH 2006. Fiber characteristics of the Huacaya Alpaca. Small Ruminant Research 64, 211224.Google Scholar
McCaig, TN 2002. Extending the use of visible/near infrared reflectance spectrophotometers to measure colour of food and agricultural products. Food Research International 35, 731736.CrossRefGoogle Scholar
McClure, WF 2003. 204 years of near infrared technology: 1800–2003. Journal of Near Infrared Spectroscopy 11, 487518.CrossRefGoogle Scholar
McGregor, BA 2002. Comparative productivity and grazing behaviour of Huacaya alpacas and Peppin Merino sheep grazed on annual pastures. Small Ruminant Research 44, 219232.Google Scholar
Martens, H, Dardenne, P 1998. Validation and verification of regression in small data sets. Chemometrics and Intelligent Laboratory Systems 44, 99121.Google Scholar
Martens, H, Naes, T 1989. Multivariate calibration. John Wiley and Sons Ltd, New York.Google Scholar
Miller, CE 2001. Chemical principles of near infrared technology. In Near infrared technology in the agricultural and food industries (ed. PC Williams and KHY Norris), pp. 1937. American Association of Cereal Chemist, St Paul, MN.Google Scholar
Monin, G 1998. Recent methods for predicting quality in whole meat. Meat Science 49 (Suppl. 1), S231S243.CrossRefGoogle Scholar
Murray, I 1986. The NIR spectra of homologous series of organic compounds. In NIR/NIT conference (ed. J Hollo, KJ Kaffka and JL Gonczy), pp. 1328. Akademiai Kiado, Budapest.Google Scholar
Murray, I 1993. Forage analysis by near infrared spectroscopy. In Sward herbage measurement handbook (ed. A Davies, RD Baker, SA Grant and AS Laidlaw), pp. 285312. British Grassland Society, Maidenhead, UK.Google Scholar
Naes, T, Isaksson, T, Fearn, T, Davies, T 2002. A user-friendly guide to multivariate calibration and classification. NIR Publications, Chichester, UK.Google Scholar
Osborne, BG, Fearn, T, Hindle, PH 1993. Near infrared spectroscopy in food analysis, second edition. Longman Scientific and Technical, Harlow, UK.Google Scholar
Slack-Smith, T, Fong, D, Douglas, SAS 1979. The potential application of near infra red reflectance to estimate the alcohol extractable mater content of scoured wool. Journal of Textile Institute 70, 1.Google Scholar
Sommerville P. 2002. Fundamental principles of fibre fineness measurements. In Technologies for measuring the fineness of wool fibres, part 3, pp. 1–5. Australian Wool Testing Authority Ltd., Melbourne, Australia.Google Scholar
The Unscrambler 1996. User’s guide, Version 6.0. CAMO AS, Trondheim, Norway.Google Scholar
Williams, PC 2001. Implementation of near infrared technology. In Near infrared technology in the agricultural and food industries (ed. PC Williams and KH Norris), pp. 145171. American Association of Cereal Chemist, St Paul, MN.Google Scholar
Figure 0

Figure 1 Near infrared mean spectrum (line) and standard deviation (dotted line) of alpaca wool samples analysed (raw spectra, 500 to 2500 nm).

Figure 1

Figure 2 Principal component score plot of alpaca wool samples analysed by visible and near infrared spectroscopy, labelled by year.

Figure 2

Table 1 Descriptive statistics for the wool parameters measured in alpaca wool samples

Figure 3

Table 2 Pearson correlation between fibre characteristics measured in alpaca wool samples

Figure 4

Table 3 Near infrared calibration statistics (cross validation) for the wool parameters measured in alpaca wool samples

Figure 5

Figure 3 Partial least squares (PLS) loadings of the optimal calibrations for mean fibre diameter (MFD), total fleece weight (TFW), curvature (CURVEDEG) and per cent medullation by number (PMEDNUM) measured by near infrared spectroscopy (1100 to 2500 nm). Optimal number of PLS loadings in brackets.