Prediction of Extractive Content of Eucalyptus globoidea Heartwood Using Near Infrared Spectroscopy
By Satoru Kuwabara, November 2017.
Download SWP-T040 (pdf)
A dissertation submitted in partial fulfilment of the requirements for the Degree of Bachelor of Forestry Science BForSc (Hons)
Natural durability of wood is highly sought for a number of wood products due to the emerging concern over and tightening regulations on the use of toxic preservatives. In New Zealand, various wood properties of several naturally durable Eucalyptus species have been investigated in order to provide potential alternatives to CCA (Copper Chromium Arsenic) treated timber of radiata pine, the widely planted commodity species. Extractive content (EC) in wood plays an important role in determining the natural durability. In this study, near infrared (NIR) spectroscopy was used to predict the extractive content of E. globoidea heartwood. The EC of the wood powder samples ranged between 0.54 and 13.51%. A number of Partial Least Squares (PLS) regression models were developed to predict EC from the NIR spectra of the samples. Several spectra pre- processing techniques were applied. The significance Multivariate Correction (sMC) variable selection was also applied to improve the model accuracy. Of all models, 1st derivative NIR spectra with sMC variable selection gave the best results (R2v = 0.97; RMSEv = 0.52%). Spectral bands around 4,700 cm-1 (2112 nm), a spectral region assigned to the bond vibration of chemical components characteristic of lignin and extractives, explained much of the variance of the EC. Calibration of NIR for the EC of E. globoidea heartwood was successful. The model can potentially be used in the future to accurately and rapidly predict the EC for a breeding programme which aims to improve the natural durability of this species.
No posts yet