Abstract: The chemical compositions and basic densities of 147 fast growing acacia samples were analyzed by using the traditional method, and the near-infrared (NIR) spectra were also collected. Partial least squares (PLS) method and cross-validation were used to confirm the best factor and build the calibration models for holocellulose, lignin, benzene-alcohol extract and basic density after the original spectra were pretreated. The best factors of the four models were 10, 8, 9 and 9. The independent verification of the calibration models showed the coefficients of determinations (R²_val) were 0.910 3, 0.950 5, 0.970 6 and 0.969 5, respectively. The root mean square errors of prediction (RMSEP) were 0.45%, 0.32%, 0.21% and 0.007 1 g/cm³, respectively. The relative percentage deviations (RPD) were 3.34, 4.50, 5.82 and 5.73, respectively. And the absolute deviations (AD) were -0.60%-0.68%, -0.50%-0.48%, -0.29%-0.33% and -0.009 7-0.009 1 g/cm³, respectively. The root mean square errors of prediction and the absolute deviations basically met the needs of error and the four calibration models could fulfil the rapid determination in pulping and paper making industry for the good predictive performance.

Key words: near-infrared spectroscopy, partial least squares(PLS), fast growing acacia, chemical composition, basic density