Near Infrared Reflectance Spectroscopy for Improving Wool Interlotting

The value of wool is determined primarily by mean fiber diameter (MFD) and clean yield (clean wool fibers present, CWFP) of greasy wool. In the U.S., wool from different growers is often combined into a single lot prior to sale in a process called interlotting. The purpose of interlotting is to improve the efficiency of marketing by providing larger presumably uniform lots of wool. Each bale in the inter-lot is weighed and cored according to an American Society for Testing and Materials (ASTM) standard, and the combined core samples are analyzed to obtain objective measurements for MFD, CWFP and other metrics. Because of the large sample size and high price for obtaining results from official methods, marketers use visual appraisal to determine which bales to combine to make a sale lot. The objective of this study was to determine if near infrared reflectance spectroscopy (NIRS) can be used to provide sufficiently accurate determination of MFD and CWFP to improve interlotting. This study obtained 1485 core samples from wool lots representing a broad cross-section of U.S. wool from Yocom-McColl Testing Laboratories, Denver CO together with the associated CWFP and MFD data as determined by ASTM methodology. The MFD range was 18 – 34 μm and CWFP ranged from 39 – 62 percent. A Feed and Forage Analyzer Model 6500M, (Foss North America, Eden Prairie, MN) fitted with a transport mechanism and using a customized sample holder (scanning area = 82 cm2) was used to obtain duplicate spectra (at 2 nm intervals in the range 400 to 2498 nm). When the Mahalanobis distance between duplicate scans exceeded 1, the samples were eliminated; otherwise duplicate spectra were averaged before further analysis. The select procedure of WinISI II software and constituent data for CWFP and AFD were used to subset samples into calibration and validation data sets. Calibration sample sets were structured to represent the spectral diversity of the entire data with a rectangular distribution of the constituents and thus different sample sets were used for each constituent. For both constituents the calibration set contained about one-third of the samples with the validation set containing the remaining two-thirds of the samples. Calibration was done using modified partial least squares regression of spectra from 1108 – 2492 nm after first derivative with four point gap and smooth pretreatments and no scatter correction. Validation statistics for MFD were: r2 = 0.92; root mean square error (RMSE) = 0.82 μm; slope = 0.97; bias = -0.14 μm; and ratio error range (RER) = 19.3. Validation statistics for CWFP were: r2 = 0.84; RMSE = 1.68 %; slope = 0.96; bias = 0.06 %; and RER = 13.2. The 95% confidence intervals (CI) for AFD and CWFP from this study are about ±1.6 μm and ±2.8%, respectively, and are superior to visual appraisal. For comparison the 95% CI for laboratory estimates of MFD and CWFP are about ±0.5 μm and ±1.0%, respectively. These results indicate that NIRS has the potential to improve the efficiency of interlotting.