You are on page 1of 9

Small Ruminant Research 99 (2011) 178–186

Contents lists available at ScienceDirect

Small Ruminant Research


journal homepage: www.elsevier.com/locate/smallrumres

Measurement of luster in Suri alpaca fiber


C.J. Lupton a,∗ , A. McColl b
a
Texas AgriLife Research and Extension Center, 7887 US Highway 87 North, San Angelo, TX 76901, USA
b
Yocom-McColl Testing Labs, 540 West Elk Place, Denver, CO 80216, USA

a r t i c l e i n f o a b s t r a c t

Article history: A study was conducted to evaluate the ability of two instruments for measuring luster
Received 2 December 2010 in Suri alpaca fiber and to compare the results with subjectively assessed luster scores.
Received in revised form 3 March 2011
In addition, the study was designed to determine the correlations between luster and
Accepted 15 March 2011
other value-determining, objectively measured physical properties of Suri alpaca fiber. For
Available online 13 April 2011
this purpose, 205 Suri alpaca fleece samples were submitted by 49 breeders that repre-
sented a broad cross-section of Suri alpaca in terms of visually assessed luster, color, and
Keywords:
fiber characteristics. Raw staples and “washed-and-aligned” staples were independently
Alpaca
Suri and subjectively assessed for luster by 2 alpaca judges certified by the Alpaca Owners
Luster and Breeders Association (AOBA). The fiber samples were measured for luster and lumi-
Objective fiber measurements nance using a SAMBA Hair System and were scanned using a near-infrared reflectance
Near-infrared reflectance spectrometer spectrometer. Standard objective test methods were used to measure mean fiber diameter
and variability (at mid-staple and overall), prickle factor, comfort factor, mean fiber cur-
vature and variability, medullation (white, beige, and light fawn samples only), laboratory
scoured yield, mean staple length and variability, and mean staple strength and variabil-
ity. The precision of the luster measurement (MLUS) was estimated using CV based on 3
measurements per subsample and averaged 3.9% across all colors. In contrast, precision of
the luminance measurement (MLUM) was higher with an average CV of 2.3%. Agreement
between the subjective luster scores of 2 certified alpaca judges was low (r2 = 0.17 and
0.10 for raw and washed-and-aligned samples, respectively). Despite this, evidence was
obtained linking instrument-measured luster and luminance with subjectively assessed
luster scores. Two systems were proposed for expressing instrument-measured luster that
will permit comparisons between alpacas of different color. Small negative correlations
were present between mean fiber diameter and subjectively assessed luster scores. Small
positive correlations were present between mean fiber diameter and objectively measured
luster values. A strong correlation (r = −0.94) existed between instrument-measured luster
and log10 luminance. Prediction equations derived from near-infrared reflectance spectra
(NIRS) produced precise and accurate estimates of log10 luminance. The NIRS estimates of
instrument-measured luster were less precise and less accurate, and estimates of subjec-
tively assessed luster were neither accurate nor precise. The SAMBA Hair System is capable
of producing estimates of luster that are accurate enough to be useful to Suri alpaca breeders
who intend to select for this trait.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction

The alpaca (Vicugna pacos) is commercially the most


important fiber producer of the South American camelid
∗ Corresponding author. Tel.: +1 325 653 4576; fax: +1 325 658 4364. family. Two breeds of alpaca are recognized, the Huacaya
E-mail address: c-lupton@tamu.edu (C.J. Lupton). and the Suri. This study deals exclusively with the lustrous

0921-4488/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.smallrumres.2011.03.045
C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186 179

fiber from the distinct-stapled Suris. Approximately 28,000 lent luster) ranged from 0.29 to 0.41 [Lupton, 2006]). Most
Suri alpacas are registered in the U.S. at this time repre- people associated with Suri alpacas recognize this lack of
senting about 18% of the total registered alpaca population consistency among subjective luster assessments, particu-
(Alpaca Registry Inc., 2010). In South America, Suris are larly on the live animal. Despite the recognized importance
estimated to compose 7% of the entire alpaca population of luster in Suri and alpaca generally, numerous scientists
(Hoffman, 2003b). (e.g., Wuliji et al., 2000; Aylan-Parker and McGregor, 2002;
Luster in alpaca is the shiny appearance of the fibers McGregor, 2006; Lupton et al., 2006) omitted assessment
when they are illuminated. Several attempts have been of luster from their alpaca fleece studies. Multiple reasons
made over the years to quantify luster in animal fibers may be responsible for these omissions including: high
(e.g., Elliot, 1986; Orwin and Woods, 1983; Van Rensburg cost of goniophotometric measurements, no access to such
and Maasdorp, 1985). A more recent study was concerned measurements, or lack of confidence in subjective assess-
with “silkiness” in mink pelts, a property that includes ment of luster.
“glossiness”, and which appears to be very similar to lus- Recently, Bossa Nova Technologies (BNT, Venice, CA)
ter in alpaca and mohair (Rasmussen and Dyck, 2000). Van introduced an instrument, the SAMBA Hair System, based
Rensburg and Maasdorp (1985) used a goniophotometer on polarization imaging technology that is capable of
to evaluate mohair luster in their studies but were not suc- differentiating and measuring the different light scatter-
cessful in establishing this technique for routine analysis ing sources (shine, chroma, and diffuse) in a “staple” (in
of this mohair attribute. The goniophotometer undoubt- this context, a structure of straightened parallel fibers) of
edly provides accurate luster information on single fibers. hair (Clemenceau, 2007). Luminance and luster values are
Unfortunately, the cost of the instrument and the cost of reported using a dedicated computer and numerous differ-
measuring many individual fibers per sample (to obtain ent formulas. The instrument is being used in the cosmetic
statistically significant results for the sample as a whole) industry to substantiate claims and to evaluate the efficacy
have caused this test method to be prohibitively expensive of products such as hair shampoos, conditioners, and shine
for routine testing of individual animals and commercial sprays and has been shown to be effective in measuring
consignments. Researchers at the Textile Research Insti- luster in human hair (Lefaudeux et al., 2008). Consequently,
tute at Princeton (Keis and Kamuth, 2004) reviewed past several sectors of the animal fiber industry are optimistic
work and summarized their recent results with objective that the SAMBA instrument can also be technically and
measurement of human hair luster. economically adapted for objective measurement of lus-
In their attempts to identify which alpaca fiber charac- ter in Suri alpaca, mohair, and other lustrous fibers. In
teristics are important to textile processors, 2 prominent fact, in a preliminary study sponsored by the Suri Network
U.S. alpaca breeders and authors (Hoffman, 2003a; Safley, and reported by Bossa Nova Technologies, 71 Suri alpaca
2004) interviewed leading processors and fiber experts in samples were measured for luster with the SAMBA Hair
Arequipa, a major center for alpaca processing in Peru. A. System (Anon., 2006). The objectively measured luster of
Burgos, head fleece buyer with Grupo Inca was quoted clean samples was shown to be correlated with subjec-
by Safley (2004) as saying “Fineness is the first consider- tive assessments within a specific color range. The SAMBA
ation. Suri is valued for its excellent handle and luster.” Hair System is also capable of concurrently producing an
Hoffman (2003a) presented 3 tables listing the premium estimate of color (luminance value). In order to be able to
alpaca fiber characteristics according to 3 processors (D. compare luster of differently colored fleeces, it was deemed
Michell, largest alpaca fleece buyer in the world for Michell desirable that an estimate of luster that is independent
and CIA; L. Chavez, general manager with Inca Tops; and G. of color be developed. The SAMBA Hair System was also
Sarfaty, general manager of Prosur). All 3 men included lus- used to measure luster in a few samples of cleaned mohair
ter (or sheen) as one of 6 premium fiber characteristics, and once sample preparation had been optimized, promis-
the others being fineness, handle, medullation, density, ing results were obtained (Lefaudeux, 2007; Lefaudeux
and cleanliness. The processors also pointed out (Safley, et al., 2007). In a recent study, Lupton (2006) concluded
2004) that luster is essential for the types of products that a near-infrared reflectance spectrometer (NIRS) was
manufactured from Suri alpaca, i.e., predominantly fabrics not accurately predicting luster in mohair that had also
composed of worsted yarns that are tailored into men’s and been subjectively assessed and measured for luster using a
ladies coats and blazers and are typically brushed in the fin- goniophotometer. The suspected reason for this result was
ishing process to further enhance the natural luster. Thus likely the quality of the luster data used for the NIRS cali-
we conclude that textile manufacturers consider luster to bration. Only relatively few fibers (∼25 per sample) were
be an important characteristic of Suri alpaca. measured (at great expense) in order to establish luster of
Subjective assessments of fleece luster on live ani- a particular calibration sample. This number of fibers mea-
mals are commonly made by breeders and show judges sured is now considered to be inadequate. It appears the
throughout the world but are complicated by several fac- SAMBA instrument may have the potential to provide more
tors including the observer’s individual perception, amount accurate data with which to calibrate the NIRS that will
of direct sunlight or other light source on the fleece, amount subsequently be evaluated to predict unknown samples.
and type of dirt in the fleece, color, etc. and agreement of Like the SAMBA Hair System, most NIRS instruments are
subjective luster scores even between trained and expe- also capable of providing data that will permit an objective
rienced individuals is typically not high (e.g., coefficients description of color.
of determination among 3 judges scoring luster in 89 raw Based on the above information, we arranged for a
mohair fleeces using a scale of 0 (no luster) to 5 (excel- short-term lease of the SAMBA instrument for the purpose
180 C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186

of conducting an independent evaluation of the instrument specific color. MLUS is defined as: 100 × [Sin /(D + Sout )] × [1/Wvisual ], where
for measuring luster in Suri alpaca fiber. Because near- Sin = integral inside the specular peak in the specular profile, Sout = integral
in the wings of the peak in the specular profile, D = integral in the dif-
infrared reflectance spectroscopy (NIRS) was considered to
fused profile, and Wvisual = mean value of the bandwidth along the region
have potential for estimating luster and color in alpaca, this of interest from the real image.
technology was also evaluated in the current project for Spectra were measured in the visible and near-infrared ranges using
its ability to estimate luster and luminance measurements a near-infrared reflectance spectrometer (NIRS, Feed and Forage Analyzer
made by the SAMBA instrument. Both the SAMBA and NIRS Model 6500M, Foss North America, Eden Prairie, MN) for each raw sample
and each washed-and-aligned sample that had been previously measured
instruments are potentially capable of providing measure- using the SAMBA Hair System. Reflected energy (log 1/R) was measured
ments at considerably lower cost than the current standard on duplicate samples and averaged over 25 scans and recorded at 2-nm
goniophotometric measurement. Because it is well estab- intervals from 400 to 2498 nm. WinISI II software (version 1.04, Infrasoft
lished (Orwin and Woods, 1983; Lupton, 2006) that mean International, Port Matilda, PA) was used to transform spectral data and
calculate calibration equations and expected levels of precision and accu-
fiber diameter and distribution influence fiber luster, cor-
racy for NIRS estimates of SAMBA-measured and subjectively assessed
relations between alpaca luster and other commercially luster, luminance, and several of the other objectively measured fiber
important physical properties were also investigated. properties.
Thus the specific objectives of the project were to: (1) The following objective measurements were conducted on each sam-
determine the precision of the SAMBA Hair System for ple using the referenced international or national standard test methods.
Relaxed mean staple length (MSL) was determined using 10 staples and
measuring luster and luminance of washed and aligned
the Agritest Staple Length Meter (IWTO-30-98). Mean staple strength
Suri alpaca fibers; (2) develop an estimate of luster that (MSS) was determined using 10 staples and an Agritest Staple Breaker
is independent of color; (3) determine the correlations Model 2 (IWTO-30-98). Other washed and conditioned subsamples were
between SAMBA luster measurements and subjective lus- minicored and 2 mm snippets were evaluated for mean fiber diameter
(MFD), prickle factor (PF), comfort factor (CF), total medullation (MED)
ter assessments, lab scoured yield, mean fiber diameter
and objectionable (OBJ) fibers (natural white, beige, and light fawn sam-
(and variability), mean fiber curvature (and variability), ples only), and mean fiber curvature (MFC) using an Optical Fiber Diameter
mean staple length (and variability) and mean staple Analyser 100 (OFDA 100; IWTO-47-00 and IWTO-57-98). Because the
strength (and variability); (4) calibrate a near-infrared SAMBA Hair System measures luster and luminance close to the mid-point
reflectance spectrometer using the SAMBA measurements of staples or tresses, one of each of the washed-and-aligned subsamples
was guillotined at its mid-point to produce 2-mm snippets and these were
and determine how accurately the NIRS instrument can
also measured with the OFDA 100.
predict SAMBA-measured luster and luminance. Each raw sample and each washed-and-aligned sample was indepen-
dently assessed for luster on a scale of 1–20 by 2 AOBA-certified alpaca
2. Materials and methods judges. The same light source (fluorescent cool white) and location were
used to illuminate and view all samples. The judges assessed the same
2.1. Sample description washed-and-aligned samples that were measured using the SAMBA Hair
System.
Two hundred and five Suri alpaca raw fiber samples were obtained
from numerous North American locations with the assistance of the Suri 2.3. Statistical analysis
Network and 49 of its member breeders. Individual breeders were asked
to describe their submitted samples using 10 color categories and to sub- The means procedure (PROC MEANS) of SAS (SAS Institute, Inc., Cary,
mit samples having ranges in subjectively assessed luster and mean fiber NC) was used to calculate means, standard deviations, coefficients of
diameter. Individual samples (>200 g raw material, 10 cm minimum sta- variation, minima and maxima for the 50 characteristics measured and
ple length) were received at Yocom-McColl Testing Laboratories, Inc. in an assessed in this study. Paired t-tests were used to identify significant
undisturbed natural staple configuration (i.e., not folded over) contained differences in measured luster and luminance between subsamples and
in sealed Ziploc bags. between fiber judges. Stepwise multiple regression analyses (PROC REG)
were used to determine to what extent other measured traits contributed
2.2. Methods of test to measured luster and scored luster. And the CORR procedure of SAS was
used to calculate correlation coefficients between selected measured and
Subsamples were drawn for: measurements of staple length and scored characteristics.
strength; lab scoured yield, luster and luminance; and measurements
made with the OFDA 100 instrument (BSC Electronics, Ardross, Western 3. Results
Australia). A protocol for cleaning Suri alpaca staples clamped at the base
of staple was developed at Yocom-McColl Testing Laboratories that used
standard washing conditions (in terms of detergent and sodium carbon- 3.1. Description of samples and summary statistics
ate concentrations, water temperatures, etc. as specified in ASTM D 584
(2003)) and resulted in clean, relatively undisturbed staples that were The distribution of samples by color and the ranges
air-dried (to avoid potential changes in color and luster caused by hot-air in subjectively assessed luster score and objectively mea-
drying) in a flat configuration under standard conditions of temperature
and relative humidity (20 ◦ C and 65%, respectively). The protocol also
sured mean fiber diameter are shown in Table 1. The
involved hand carding the staples before washing (to remove loose dirt abbreviations and units used throughout this report are
and separate individual fibers) and after drying and conditioning. This listed in Table 2. Table 3 summarizes all the data assessed,
procedure concurrently provided duplicate measures of lab scoured yield measured, or calculated in the study in terms of number of
(LSY) and duplicate subsamples with parallel, untangled fibers that were
measurements (N), mean values, standard deviations (SD),
mounted and fastened onto black cardboard and subsequently placed over
a cylinder and measured with the SAMBA Hair System. To determine the coefficients of variation (CV), and minimum and maximum
precision of the SAMBA Hair System measurements of luster and lumi- values.
nance, two subsamples were each measured 3 times on 97 samples. The
SAMBA Hair System actually outputs 25 variables, 5 of which are estimates 3.2. Subjective assessments of alpaca luster
of luster calculated using different formulas (BNT, Reich-Robbins, Textile
Research Institute, Stamm, and Guiolet [Lefaudeux et al., 2008]). Only 2
of those variables are considered in this report, i.e., BNT luster (MLUS) Mean luster scores, measures of variability, minimum
and luminance (MLUM), the latter providing a number that is unique for a and maximum values for both judges on raw and washed-
C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186 181

Table 1
Distribution of Suri alpaca samples by color and ranges for subjective luster score and mean fiber diameter.

Color number Color Number Luster score range (1–20) Mean fiber diameter range (␮m)

1 White 52 11.0–16.5 18.0–36.5


2 Beige 10 12.5–15.0 22.3–32.6
3 Light fawn 29 11.5–17.0 18.5–36.9
4 Medium fawn 14 12.0–16.0 16.8–29.6
5 Dark fawn 5 13.0–15.0 20.7–33.5
6 Light brown 15 12.5–15.0 18.1–39.0
7 Medium brown 24 12.5–15.5 19.4–39.3
8 Dark brown 17 12.5–15.0 20.6–35.6
9 Bay black 6 13.0–14.5 24.5–34.2
10 True black 33 12.5–16.0 22.1–36.9

and-aligned samples are presented in Table 3. Judge 1 nation for several simple linear regressions are presented
had overall lower scores on the raw samples (11.5 versus in Table 4.
12.1, P < 0.05), greater variability (16.6 versus 10.4% for CV,
P < 0.05), and a wider range in scores (5–16 versus 9–16). 3.3. Precision and accuracy of SAMBA Hair System
Mean scores, variability, and ranges were much closer for measurements
the washed samples (14.1 versus 1.0, 9.1 versus 9.5%, and
10–17 versus 11–17, respectively). Coefficients of determi- Precision estimates were based on 97 samples that
were measured 6 times each. Means of 6 subsamples had
CVLUS > 10% (Fig. 1) but most CV values were between 0%
Table 2 and 10%, the average value being 3.9% (Table 5). In contrast,
Key to abbreviations (and units) used in text and tables.

MLUS Mean luster using Bossa Nova Technology


Table 3
(BNT) formula
Summary statistics for properties measured on Suri alpaca samples.
SDLUS Standard deviation of 3 measurements on the
same subsample Property N Mean SD CV Min Max
CVLUS Coefficient of variation of 3 measurements on
the same subsample MLUS 205 7.4 7.3 99.4 0.6 29.1
MLUM Mean luminance MLUM LOG10 MLUM 205 1599.8 1889.7 118.1 4.5 8680.5
SDLUM Standard deviation of 3 measurements on the LOG10 MLUM 205 2.6 0.9 34.9 0.6 3.9
same sample J1R 205 11.5 1.9 16.6 5 16
CVLUM Coefficient of variation of 3 measurements on J2R 205 12.1 1.3 10.4 9 16
the same sample J1W 205 14.1 1.3 9.1 10 17
LOG10 MLM Log10 mean luminance J2W 205 14.0 1.3 9.5 11 17
J1R (1–20) Judge 1 luster score on raw sample MLSR (1–20) 205 11.8 1.3 11.4 7.5 15.5
J2R (1–20) Judge 2 luster score on raw sample MLSW (1–20) 205 14.1 1.1 7.5 11.0 17.0
J1W (1–20) Judge 1 luster score on washed-and-aligned AGE 205 2.4 1.7 68.8 0.6 9.7
sample MFD (␮m) 205 26.0 4.9 18.7 16.8 39.3
J2W (1–20) Judge 2 luster score on washed-and-aligned SDFD (␮m) 205 6.7 1.5 22.9 3.9 11.1
sample CVFD (%) 205 25.9 3.9 15.2 16.8 37.1
MLSR (1–20) Mean luster score of raw sample PF (%) 205 23.6 21.0 88.8 0.8 84.7
MLSW (1–20) Mean luster score of washed-and-aligned CF (%) 205 76.4 21.0 27.4 15.3 99.2
sample MFC (◦ /mm) 205 12.9 2.7 21.1 6.8 26.7
AGE (yr) Age to nearest year SDFC (◦ /mm) 205 18.4 4.1 22.2 8.6 33.3
MFD (␮m) Mean fiber diameter (␮m) CVFC (%) 205 143.4 20.0 14.0 93.1 218.5
SDFD (␮m) Standard deviation of fiber diameter (␮m) MED (%)a 91 15.6 14.2 90.7 0.5 59.5
CVFD (%) Coefficient of variation of fiber diameter (%) OBJ (%)a 91 6.2 3.5 56.9 0 18.7
PF (%) Prickle factor, fibers >30 ␮m (%) MIDMFD (␮m) 205 27.2 5.0 18.4 18.0 39.4
CF (%) Comfort factor, fibers ≤30 ␮m (%) MIDSDFD (␮m) 205 6.7 1.7 24.8 3.5 12.7
MFC (◦ /mm) Mean fiber curvature (◦ /mm) MIDCVFD (%) 205 24.8 4.3 17.3 14.2 37.1
SDFC (◦ /mm) Standard deviation of fiber curvature (◦ /mm) MIDPF (%) 205 28.4 23.4 82.3 0.5 85.6
CVFC (%) Coefficient of variation of fiber curvature (%) MIDCF (%) 205 71.6 23.4 32.7 14.5 99.5
MED (%) Medullated fibers (white, beige, and light fawn MIDMFC (◦ /mm) 205 10.4 2.7 26.3 5.8 23.4
samples only) (%) MIDSDFC (◦ /mm) 205 12.8 5.4 42.0 5.6 36.9
OBJ (%) Objectionable fibers (white, beige and light MIDCVFC (%) 205 120.3 23.4 19.4 79.2 200.6
fawn samples only) (%) MIDMED (%)a 89 12.8 13.5 105.6 0.1 63.0
MID An abbreviation preceded by MID indicates MIDOBJ (%)a 89 4.5 4.1 90.2 0 25.0
this variable was measured at the mid-point of LSY (%) 205 77.8 7.4 9.5 56.7 94.8
the staple MSL (mm) 205 177.2 47.6 26.9 108.4 332.2
LSY (%) Lab scoured yield (%) SDSL (mm) 205 13.2 12.5 94.7 1.7 128.7
MSL (mm) Mean staple length (mm) CVSL (%) 205 7.3 5.4 73.6 0.8 45.5
SDSL (mm) Standard deviation of staple length (mm) MSS (N/ktex) 205 35.9 10.9 30.3 8.6 65.7
CVSL (%) Coefficient of variation of staple length (%) SDSS (N/ktex) 205 7.5 3.9 57.1 2.1 26.4
MSS (N/ktex) Mean staple strength (N/ktex) CVSS (%) 205 23.8 16.8 70.5 4.1 105.8
SDSS (N/ktex) Standard deviation of staple strength (N/ktex) Key: See Table 2.
CVSS (%) Coefficient of variation of staple strength (%) a
White, beige, and light fawn samples only.
182 C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186

Table 4
Coefficients of determination for simple linear regressions between sub-
jective luster in raw and washed and aligned Suri alpaca scored by two
judges.

Variables r2

J1R J1W 0.25


J2R J2W 0.44
J1R J2R 0.17
J1W J2W 0.10
MLSR MLSW 0.42

Key: See Table 2.


Fig. 2. CV of luminance versus average luminance for 3 measurements
made on one subsample.

the SAMBA Hair System. Pearson correlation coefficients


(and P values) were calculated between 35 of these prop-
erties. The more important correlations are presented in
Table 6.

3.5. Near-infrared reflectance spectroscopy

Tables 7 and 8 summarize the calibration, cross valida-


Fig. 1. CV of BNT luster versus average BNT luster for 3 measurements tion, and prediction statistics for several Suri alpaca traits
made on one subsample. determined using the WINISI software. No independent
Table 5
samples were predicted using these equations so no gen-
Means, SD, and CV of triplicate measurements made on two subsamples uine validation statistics are presented.
and probability values from paired t-tests.

Property N Subsample 1 Subsample 2 P 4. Discussion


MLUS 97 4.22 4.11 0.3770
SDLUS 97 0.21 0.23 0.7454
4.1. Samples and summary statistics
CVLUS 97 4.17 3.63 0.2010
MLUM 97 2714.7 2779.3 0.1563 Colors were not represented equally; white, true black,
SDLUM 97 37.4 45.4 0.4685 and light fawn samples predominated. However, represen-
CVLUM 97 1.94 2.59 0.3339
tation of the other colors was considered to be adequate
Key: See Table 2. for the purpose of this project. Ranges in luster score
and mean fiber diameter (Table 1) were similar for each
the average CVLUM was 2.3% with no individual CV > 10% color. The MLUS values (Table 3) ranged from 0.6 to 29.1
(Fig. 2). with smaller values being associated with white fiber and
larger values with black fiber. Conversely, MLUM values
3.4. Correlations and multiple regression analyses increased from black to white and exhibited a wide range,
4.5–8680.5. For plotting purposes, it was convenient to use
In all, we measured or calculated 50 physical properties the log10 of luminance rather than the luminance value
for each of 205 samples of Suri alpaca. This total included itself. Mean fiber diameter ranged from 16.8 to 39.3 ␮m
only 2 of the many variables measured and calculated by providing a wide range for this trait. All other measured

Table 6
Pearson correlation coefficientsa , b , c between selected Suri alpaca fiber traits.

MFC MED OBJ LSY MSL MSS MIDMFD MLSR MLSW MLUS MLUM

MFD −0.6341 0.6249 0.3786 0.3419 −0.3367 0.4025 0.9054 −0.3676 −0.2133 0.3468 −0.1875
MFC −0.3453 −0.0231 −0.2971 0.3428 −0.3279 −0.5512 0.1911 0.0378 −0.1912 0.1591
MED 0.5527 0.1210 −0.1513 0.4075 0.5472 −0.4330 −0.4627 −0.3076 0.2420
OBJ 0.1357 −0.0331 0.3556 0.3110 −0.2603 −0.2871 −0.0811 0.2277
LSY −0.1748 0.1936 0.2779 −0.1533 −0.0768 −0.00467 0.1362
MSL −0.1331 −0.3316 0.0984 0.1087 −0.0784 0.0564
MSS 0.3673 −0.2579 −0.1358 −0.0025 −0.0645
MIDMFD −0.3484 −0.1935 0.3706 −0.2748
MLSR 0.6480 0.0460 −0.0568
MLSW 0.1521 −0.1300
MLUS −0.6577

Key: See Table 2.


a
N = 205 for all correlations except those that include MED and OBJ in which cases N = 91.
b
Correlation coefficients >0.15 or <−0.15 are significant (P < 0.05) when N = 205.
c
Correlation coefficients >0.21 or <−0.21 are significant (P < 0.05) when N = 91.
C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186 183

Table 7
NIRS calibration and cross validation statistics for selected constituents in raw Suri alpaca samples.

Constituent N Calibration Prediction


2
Mean SD SEC R SECV SEP Slope R2

MLUS 187 6.39 6.54 1.50 0.95 1.56 2.36 1.03 0.90
MLUM 190 1317.56 1470.67 241.37 0.97 404.72 881.05 1.13 0.80
LOG10 MLUM 194 2.64 0.90 0.14 0.98 0.15 0.19 1.00 0.96
MLSR 198 11.86 1.26 1.26 0.22 1.16 1.22 0.93 0.18
MLSW 199 14.04 1.00 1.00 0.01 1.00 1.05 0.87 0.01

Key: N, number of samples used in calibration model; SD, standard deviation; SEC, standard error of calibration; R2 , multiple coefficient of determination
for calibration; SECV, standard error of cross validation; SEP, standard error of prediction; R2 , coefficient of determination for the prediction.

Table 8
NIRS calibration and cross validation statistics for selected constituents in washed-and-aligned Suri alpaca samples.

Constituent N Calibration Prediction


2
Mean SD SEC R SECV SEP Slope R2

MLUS 188 6.61 6.62 1.67 0.94 1.72 2.45 1.03 0.89
MLUM 194 1320.76 1417.28 267.72 0.96 309.18 1058.00 1.14 0.71
LOG10 MLUM 196 2.62 0.92 0.12 0.98 0.13 0.167 1.00 0.97
MLSR 197 11.81 1.30 1.27 0.05 1.29 1.33 0.62 0.03
MLSW 200 14.04 1.00 0.99 0.02 1.00 1.05 0.83 0.01

Key: See Table 6.

properties exhibited ranges in values that were compared and MLUS. For MLSW the following variables entered the
to those reported earlier for Huacaya alpaca (Wuliji et al., regression model in the indicated order (partial r2 and P
2000; McGregor, 2006; Lupton et al., 2006). Generally, values follow the abbreviation): MLSR (0.4503, <0.0001),
Suri alpaca had lower mean fiber curvature (less crimp MIDMED (0.0353, 0.0172), and MFC (0.0472, 0.0043) for a
frequency) compared to Huacaya (12.9 versus 33.2◦ /mm model R2 = 0.5328. No other variable met the 0.05 signifi-
[values in parenthesis from this study and Lupton et al., cance level for entry into the model. In other words, 53% of
2006]), comparable medullation (15.6 versus 17.5%), lower the variability in subjectively scored luster in the washed-
lab scoured yield (77.8 versus 89.8%), longer staple length and-aligned samples was accounted for by variability in
(177.2 versus 116.3 mm), and lower staple strength (35.9 these 3 traits. In contrast, for MLUS the following vari-
versus 50.2 N/ktex). With the exception of staple length, ables entered the regression model in the indicated order:
means and ranges tabulated in this study (MFD, CF, MFC) LOG10 MLUM (0.5913, <0.0001), MLUM (0.1030, <0.0001),
are also similar to those reported in an earlier Suri alpaca MIDMED (0.0601, <0.0001), COLORNO (0.0385, 0.0002),
study (Lupton, 2009). MLSW (0.0240, 0.0014), MIDSDFC (0.0207, 0.0017), SDFC
(0.0188, 0.0016) and MIDCVFD (0.0750, 0.0387) for a model
R2 = 0.8640. No other variable met the 0.05 significance
4.2. Subjective assessment of alpaca luster level for entry into the model. It is interesting to note that
a measure of medullation (MIDMED) made a small but sig-
Subjective luster scores (MLSR, average values for 2 nificant contribution to both MLSW and MLUS.
judges) on the raw samples ranged from 7.5 to 15.5 with a
mean value of 11.8 (Table 3) and predictably increased after
the samples had been washed, dried, and aligned resulting 4.3. Precision and accuracy of SAMBA Hair System
in a mean value of 14.1 and a range of 11.0–17.0 (MLSW measurements
values). The problem of consistency in subjective assess-
ments of luster, as previously alluded to, is exemplified Figs. 1 and 2 illustrate that the measurement of lumi-
by the following results. Coefficients of determination for nance is more precise than that of luster. Small differences
raw versus washed-and-aligned scores by the same judge in the alignment of fibers in the staples and smoothness
(Table 4) were low (r2 = 0.25 and 0.44) and r2 values for the of the staple surface have been reported to affect luster
regressions between the scores of individual judges on raw measurements (Gao et al., 2009), hence the great care nec-
and clean samples were even lower (0.17 and 0.10, respec- essary in sample preparation. Apparently, these differences
tively). The r2 value for averaged scores on raw versus do not affect color measurement to the same degree. Paired
washed-and-aligned was 0.42. Thus, even if we assumed t-tests indicated no differences (P > 0.16, Table 4) between
the instrument measurements were accurate, it would be subsamples for MLUS, SDLUS, CVLUS, MLUM, SDLUM, and
unreasonable to expect a high level of correlation between CVLUM. Accuracy of the SAMBA measurements was not
the objective SAMBA measurements and the subjectively determined because no absolute or standard methods were
assessed scores. used in this trial to compare them with. In fact, com-
We also conducted stepwise multiple linear regression parison of SAMBA-measured luster on staples containing
analyses to determine which combination of all the mea- many fibers having a wide distribution of fiber diameter
sured properties would be used to best predict MLSW with measures of luster based on multiple goniophoto-
184 C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186

Fig. 3. BNT luster versus log10 luminance for 205 Suri alpaca samples. Fig. 5. BNT luster versus log10 luminance showing 95% confidence limits
for the luster measurement.

metric measurements of single fibers may not have been


appropriate and certainly would not have been practical or to subjectively assessed luster scores. The practical ques-
affordable. We were actually more interested in determin- tion then arises, how can the BNT luster measurements
ing how well SAMBA-measured luster related to the judges be used to produce estimates of luster that are indepen-
subjective evaluations. dent of color? Two possibilities include the following. If the
assumption is made that the samples used in this study are
truly representative of all colors and levels of luster found in
4.4. Instrument measurement of luster
Suri alpacas, then it is possible to calculate 95% confidence
limits about the regression line (Fig. 5) and conclude that
Interpretation of BNT luster values is complicated by the
samples that fall between the 95% confidence limits have
observation that MLUS is highly dependant on luminance
intermediate or normal luster, while those above could be
(Fig. 3). Consequently, direct comparison of BNT luster val-
termed high and those below could be termed low luster,
ues between animals of different color is meaningless. An
thus providing a 3-point scoring system. The present sub-
earlier study (reported by Vonderhaar, 2009) conducted by
jective system (show ring) for scoring luster in Suri alpaca is
BNT personnel identified different trend lines for visually
a 20-point system so breeders and judges might find the 3-
low, medium, and high luster when Stamm luster was plot-
point system too restrictive. So alternatively, knowing the
ted against luminance. A similar relationship was noted
range of MLUS for a specific luminance (Fig. 6), luster scores
in this study when BNT luster (MLUS) was plotted against
could be assigned within that range for a specific lumi-
log10 luminance (MLUM) for 25 samples with extreme high
nance. Because the MLUS range decreases with increasing
scores, 25 with extreme low scores and 25 with intermedi-
luminance, moving from black to white, the smallest ranges
ate luster scores (Fig. 4). So, despite the aforementioned
are in the white area of the spectrum. Using the formulas
problems with consistency of subjective assessments of
in Fig. 6, the range in MLUS for a value of LOG10 MLUM = 3.5
luster, we were also able to demonstrate clear evidence that
was calculated to be 0.8163–2.4932. Using the formulas in
SAMBA luster and luminance measurements are related
Fig. 5 (based on 97 samples only), the 95% confidence limits
at this same luminance value are of the order ±0.05. So even

Fig. 4. BNT luster versus log10 luminance for 25 extreme high and low and Fig. 6. BNT luster versus log10 luminance showing minimum and maxi-
medium subjective scores. mum measured ranges of luster.
C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186 185

at this particular luminance where the observed range in cross validation statistics that used the same samples for
MLUS is relatively small, the SAMBA Hair System is capa- validation as were used for calibration. This was achieved
ble of differentiating between samples having numerically by removing a sample or group of samples and develop-
small differences in luster. Taking into account the actual ing a calibration model with the remaining samples. The
score ranges observed in this study and the precision of samples that were eliminated were then predicted and
the luster measurement, a score range of 1–5 (rather than residuals were calculated. This procedure was repeated
1–20) might be more appropriate. until all the samples had been predicted and also used in
When scores of 1–3 (based on the system described the calibration model development (Williams, 2005). The
above) were assigned to the 97 samples represented in derived calibration equation was used to predict the var-
Fig. 5, correlations between instrument-based scores and ious constituents thus producing another set of statistics
judges scores of raw and washed-and-aligned samples (SEP, slope, and R2 ). The slope of the line between pre-
were 0.1652 and 0.3043, respectively, higher than the cor- dicted and actual lab measurements approaches 1 as the
relations between MLSR and MLSW versus MLUS (0.0460 predictions become more accurate. R2 approaches 1 as the
and 0.1521, Table 6). predictions become more precise. It is apparent from both
Tables 6 and 7 that NIRS predictions of LOG10 MLUM are
4.5. Correlation analyses both accurate and precise. The NIRS predictions of MLUS
and MLUM are less accurate and precise, though potentially
Several of the correlations listed in Table 6 appear to still useful. And the NIRS predictions of MLSR and MLSW are
be intuitive and some (those between MFD, MFC, LSY, neither accurate nor precise.
MSL, and MSS) are similar in magnitude and sign to those
reported earlier for Huacaya alpaca fiber (Lupton et al.,
2006). Small negative correlations existed between MFD 5. Conclusions
and the scores made on raw and washed-and-aligned
fibers, finer samples receiving higher luster scores. How- 1) Precision estimates for the SAMBA Hair System mea-
ever, the opposite was true for the correlation between surements of luster and luminance were calculated and
MFD and instrument-measured luster in which a small pos- should be satisfactory for routine measurement of these
itive correlation was observed. Knowing that the SAMBA traits in Suri alpaca.
Hair System measures luster close to the middle of the 2) Two approaches for expressing instrument-measured
tress, we also measured MFD at mid staple. The MIDMFD luster that are independent of color in Suri alpaca have
and MFD values (the latter representing fibers through- been proposed.
out the staple) were highly correlated (r = 0.9054) and 3) Correlation between instrument-measured luster and
the MIDAFD versus MLUS correlation was slightly greater subjective scores was low. However, significant rela-
than the corresponding correlation with MFD (0.3706 tionships between the 2 were demonstrated for samples
versus 0.3468). This is interpreted to mean that approx- judged to have high, medium, and low luster. Cor-
imately 14% (r2 = 0.1373) of the variability in MLUS can relations between BNT luster and several objectively
be explained by the variability in MFD. It was also inter- measured fiber characteristics were calculated.
esting to note that both the subjective scores and the 4) Prediction equations derived from near-infrared
instrument measurements of luster were negatively cor- reflectance spectra produced precise and accurate
related with medullation. So the human perceptions as estimates of log10 luminance. Estimates of BNT luster
well as the instrumental measurements were affected by were less precise and less accurate, and estimates of
medullation in the same direction, though the magnitude subjectively assessed luster were neither accurate nor
was larger in the case of the scores (r = −0.4 versus −0.3 precise.
for scores and instrument measurements, respectively).
Very few of the other correlations are noteworthy. One
exception is the correlation between raw and washed- 6. Implication
and-aligned scores for which r = 0.6480, an indication that
a significant (P < 0.0001) degree of consistency is present The SAMBA Hair System is capable of producing pre-
between scores made on fiber in the two different physical cise estimates of luster that in conjunction with pedigree
forms. The other exception is the relatively high linear cor- information could be used to calculate the heritability of
relation between MLUS and MLUM (r = −0.6577). Though this trait in Suri alpacas and other species where luster is
not shown in Table 6, it is noteworthy that the corre- important. If luster proves to be a heritable trait, these mea-
lation between MLUS and LOG10 MLUM is considerably surements will be useful to Suri alpaca breeders who intend
higher (r = −0.9419). This linear correlation coefficient is to select for this trait.
only slightly smaller than the r value calculated from the
data represented in Fig. 3 (r = −0.9701) that is associated
with the polynomial relationship between these two traits. Acknowledgements

4.6. NIRS estimates of luster and luminance The authors wish to express their appreciation to the
Alpaca Research Foundation and the Suri Network for
In the absence of a true validation using independent financial support of this project and to the cooperating
samples, a cross validation technique was used to generate breeders who shared in the cost of the testing.
186 C.J. Lupton, A. McColl / Small Ruminant Research 99 (2011) 178–186

References Lefaudeux, N., Clemenceau, P., Lechocinski, N., 2007. Visual appear-
ance measurements on mohair samples. In: Bossa Nova Technologies
Alpaca Registry Inc., 2010. http://www.Alpacaregistry.com/public/reports/ Report, Venice, CA, May 22, p. 9.
alpacaworld (accessed 12.11.10.). Lefaudeux, N., Lechocinski, N., Clemenceau, P., Breugnot, S., 2008. New lus-
Anon., 2006. Shine measurement of hair care product applied on blond, ter formula for the characterization of hair tresses using polarization
light brown, and dark brown samples with Samba Hair System. In: imaging. J. Cosmetic Sci. 60 (2), 153–169.
Bossa Nova Technologies Report, Venice, CA, October, p. 14. Lupton, C.J., 2006. Objective measurement of mohair luster. In: Final
A STM D 584-96, 2003. Standard test method for wool content of raw wool Report to the National Sheep Industry Improvement Center. Texas
– laboratory scale. Annual Book of ASTM Standards. Sec. 7, vol. 07.01. Agric. Exp. Sta. Texas A&M Univ. System, p. 10.
American Society for Testing and Materials, West Conshohocken, PA, Lupton, C.J., 2009. Characteristics of North American Suri alpacas.
pp. 187–192. In: The Suri Network, http://surinetwork.org/pages/Luster Study
Aylan-Parker, J., McGregor, B.A., 2002. Optimising sampling techniques Reports/Characteristics%20of%20North%20AmericanSuri%20Alpacas.
and estimating sampling variance of fleece quality attributes in pdf (accessed 21.10.10.).
alpacas. Small Ruminant Res. 44, 53–64. Lupton, C.J., McColl, A., Stobart, R.H., 2006. Fiber characteristics of Huacaya
Clemenceau, P., 2007. Objective measurement of luster in alpaca Suri sam- alpaca. Small Ruminant Res. 64, 211–224.
ples. In: Bossa Nova Technologies Report, Venice, CA, August 7, p. McGregor, B.A., 2006. Production, attributes and relative value of alpaca
9. fleeces in southern Australia and implications for industry develop-
Elliot, K.H., 1986. The Ability of the WRONZ Bulkometer to Rank Wools for ment. Small Ruminant Res. 61, 93–111.
Luster. Wool Research Organization of New Zealand Rep. No. R135, p. Orwin, D.F.G., Woods, J.L., 1983. The effects of within-fibre diameter vari-
11. ability and other fiber characteristics on the luster of wool. J. Text. Inst.
Gao, T.A., Pereira, A., Zhu, S., 2009. Study of hair shine and hair surface 3, 118–130.
smoothness. J. Cosmet. Sci. 60, 187–197. Rasmussen, P.V., Dyck, J., 2000. Silkiness in brown mink pelts character-
Hoffman, E., 2003a. The Complete Alpaca Book. Bonny Doon Press, Santa ized with optical methods. J. Anim. Sci. 78, 1697–1709.
Cruz, CA, pp. 252–253. Safley, M., 2004. Some Views on Evaluating Suri Alpaca, http://www.ideal-
Hoffman, E., 2003b. The Complete Alpaca Book. Bonny Doon Press, Santa alpaca.com/article/-174.htm (accessed 02.03.11.).
Cruz, CA, p. 279. Van Rensburg, N.J.J., Maasdorp, A.P.B., 1985. A study of the luster of mohair
IWTO-30, 1998a. Determination of Staple Length and Staple Strength. fibres. In: Proc. 7th Int. Wool Text. Res. Conf., vol. 1 , Tokyo, Japan, pp.
International Wool Textile Organisation, Brussels, Belgium. 243–252.
IWTO-57, 1998b. Determination of Medullated Fibre Content of Wool and Vonderhaar, B., 2009. Characteristics of North American Suri alpacas.
Mohair Samples by Opacity Measurements Using an OFDA. Interna- In: The Suri Network, http://surinetwork.org/pages/Luster Study
tional Wool Textile Organisation, Brussels, Belgium. Reports/Vonderhaar’s%20Power%20Point.ppt (accessed 10.10.10.).
IWTO-47, 2000. Measurement of the Mean and Distribution of Fibre Williams, P., 2005. Near-Infrared Technology – Getting the Best Out of
Diameter of Wool Using an Optical Fibre Diameter Analyser (OFDA). Light. PDK Projects Inc., Nanaimo, BC, Canada, pp. 5–11, 5–12.
International Wool Textile Organisation, Brussels, Belgium. Wuliji, T., Davis, G.H., Dodds, K.G., Turner, P.R., Andrews, R.N., Bruce,
Keis, K., Kamuth, Y.K., 2004. TRI Princeton Business Briefing: Global Cos- G.D., 2000. Production performance, repeatability and heritability
metics Manufacturing, pp. 1–5. estimates for live weight, fleece weight, and fiber characteristics of
Lefaudeux, N., 2007. Mohair measurement. In: Bossa Nova Technologies alpacas in New Zealand. Small Ruminant Res. 37, 189–201.
Report, Venice, CA, April 20, p. 10.

You might also like