Ability of near Infrared Spectroscopy to Monitor Air-dry Density Distribution and Variation of Wood

Process control of wood density with near infrared spectroscopy (NIR) would be useful for pulp mills that need to maximize pulp yield without compromising paper strength properties. If models developed from the absorbance at wavelengths in the NIR region could provide density histograms, fiber supply personnel could monitor chip density variation as the chips enter the mill. The objectives of this research were to a) develop density histograms from actual density versus density histograms developed through NIR modeling, and b) determine the precision of density models developed from absorbance in the NIR region with a recommendation for the sample size needed to estimate the standard deviation of density at a given precision. Models for density were developed from calibration samples (n= 170) and then validated with 93 randomly held aside samples. The samples were systematically removed from 10 longleaf pine trees of equal age, but different growth rates. The histogram patterns for actual density almost paralleled the histogram patterns developed from predictive models. Subsequently, the validation data set was randomly categorized into groups of three, and the standard deviations of density were measured. For three measurements per data point, the predicted standard deviation covaried with the actual standard deviation of density with an R' = 0.61 and 0.55 for the calibration and validation data set, respectively. A sample size of 30 was recommended to estimate the standard deviation of density with a precision of 0.01 gicm3.