Relationship between three cotton trash measurements and near infrared spectral response

Authors: Liu, Yongliang; Thibodeaux, Devron; Gamble, Gary

Submitted to: Journal of Cotton Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/4/2012
Publication Date: 12/20/2012
Citation: Liu, Y., Thibodeaux, D.P., Gamble, G.R. 2012. Relationship between three cotton trash measurements and near infrared spectral response. Journal of Cotton Science. 16:272–281.

Interpretive Summary: Presence of non-lint materials (trashes) in commercial cotton bales at various amounts degrades the market values and further influences the end-use qualities. In order to ensure a fair trading, USDA’s AMS has introduced the high volume instrument (HVI) measurement as an universal standard index. Trash contents are generated by one of three HVI modules and represent the trash portion only on samples’ surface. In additional to HVI’s geometric method, gravimetric-based Shirley analyzer (SA) and advanced fiber information system (AFIS) have also been utilized to yield the trash contents. With the increasing acceptance of HVI readings in the domestic and international trading, there is a continued interest in the relationship between HVI trash and SA trash from cotton fiber customers and regulators. Due to the complexity of not only trash type, size, and its weight distribution but also the nature of HVI and SA tests, it is understandable that there has few study available trying to bridge two types of trash readings and, apparently, this is a challenge. During the 2011 Beltwide Cotton Conference in Atlanta, Dr. James Knowlton (USDA, AMS) talked his latest trip to China in the promotion of HVI measurement and mentioned one concern on the relationship between HVI trash and gravimetric method. The authors voluntarily performed this study by analyzing the data collected over a 4-year span at the house. This research will have a direct impact on domestic and foreign customers and regulatory agencies, and also trading organizations. The outcome helps cotton fiber engineers, researchers, ginners, and trading regulators in relating the HVI trash to Shirley analyzer trash.

Technical Abstract: Presence of non-lint materials (trashes) in commercial cotton bales at various amounts degrades the market values and further influences the end-use qualities. In order to ensure a fair trading, USDA’s AMS has introduced the high volume instrument (HVI) measurement as an universal standard index. Trash contents are generated by one of three HVI modules and represent the trash portion only on samples’ surface. In additional to HVI’s geometric method, gravimetric-based Shirley analyzer (SA) and advanced fiber information system (AFIS) have also been utilized to yield the trash contents. With the increasing acceptance of HVI readings in the domestic and international trading, there is a continued interest in the relationship between HVI trash and SA trash from cotton fiber customers and regulators. Due to the complexity of not only trash type, size, and its weight distribution but also the nature of HVI and SA tests, it is understandable that there has few study available trying to bridge two types of trash readings and, apparently, this is a challenge. This study first investigated the correlations between two HVI trash readings, and revealed a general conversion of HVIcount=104.5*HVIarea among low trash samples (HVIarea = 0.40). Then, correlations between the HVI and SA trash and also against AFIS trash were examined, and a stronger relationship between HVI and SA trash than between HVI and AFIS trash was observed. Next, the samples were sub-grouped subjectively according to the ratios of HVIarea/SAvisible (or HVIcount/SAvisible), and from the plots with the least intercepts, it was proposed two general conversions of SAvisible = 6.82*HVIarea and SAvisible = 0.069*HVIcount. In order to verify the likely conversion, NIR spectra were correlated with HVIarea readings. Considering the heterogeneous distribution of trashes in fibers and different sampling specimens between NIR spectral and HVI reference measurement, a 90% confidence interval was applied to exclude outlier samples from the calibration and validation sets. The recalibrated models revealed different response to validation samples in various subsets. Remarkably, the model from samples with the HVIarea/SAvisible ratio of 0.12-0.18 suggested the highest r2, RPD, and means/SEP, indicating the most appropriate references for the samples in this subset and echoing the earlier finding from simple descriptive approach of least intercept. Unquestionably, conversion constant might change with relative amount of trash size and type and also their weight distribution.