Development of calibrations to predict Mean Fineness in llama (Lama glama) fleeces with Near Infrared Spectroscopy

Abstract

Llama fiber has the potential to become the most valuable textile resource in the Puna region of Argentina. The ability of NIRS technology to predict average fineness of llama fiber was evaluated. Analyses were performed on 169 carded and non-carded samples combined with spectral preprocessing techniques in the Vis-NIR ranges. Spectral preprocessing consisted of wavelength selection, as well as multiplicative and derivative pretreatments. Modified partial least squares regression was used for predictive models. Predictability was evaluated by Coefficient of Determination (R2), Standard Error of Cross Validation (SECV), Standard Error of External Validation (SEV) and Residual Predictive Value (RPD). The best model was obtained by applying the first derivative (R2=0.67; SECV=1.965; SEV=2.235 and RPD=1.91). ANOVA showed differences between treatments. The models obtained could be used in genetic selection programs.