| For improving wood properties through genetic selection or silviculture prescription and optimal utilizing plantation wood resource, a large sampling population would be needed to rapidly, accurately, and comprehensively assess wood properties for the effective decision. However, traditional methods are difficult, expensive and time consuming. So it's an important program to explore a rapid and accurate technique for tree genetic improvement and wood science. Near infrared spectroscopy technique has been recognized as a powerful non-destructive analytical technique for rapid and accurate determination of various properties of increment core, solid wood or wood meal.. In recent years, there has been a growing interest in the development of near infrared spectroscopy applications in wood science; however, there have been few studies on the application of near infrared spectroscopy to wood science in China. Therefore, the first objective of this research was to predict wood properties of slash pine (Pinus elliottii Engelm) plantation using near infrared spectroscopy; the second were to detect early decay, assess the extent of decay (determined as weight loss), and determine residual mechanical strength of slash pine wood based on time of exposure to decay fungi by near infrared spectroscopy coupled with multivariate data analysis. The main results of this research are following, 1) The ability of using near infrared spectroscopy as a tool for predicting chemical compositions in slash pine wood is demonstrated. The results from this study showed that good correlations between NIR predicted and wet-chemistry method determined holocellulose, alpha-cellulose, and klason lignin in slash pine wood using multivariate data analysis, the correlation coefficient (r) between NIR predicted and lab determined wood compositions could be above 0.82. 2) This study adopted different pretreatments of NIR spectra to investigate the effect of spectra pretreatments on the prediction ability, the results showed that higher correlation was achieved with normal spectra and the first derivative. Comparisons of multiple linear regression (MLR), principal component regression (PCR), and partial least square regression (PLS) models were performed for predicting wood compositions. We found PLS technique performed the best to utilize near infrared spectroscopy as a tool for rapid predicting multiple wood properties. 3) Correlation analysis between wood crystallinity in slash pine and tree growth traits, and chemical compositions was also studied in this paper. The results showed that there was a significant correlation among the crystallinity and tree growth traits and chemical compositions, and the crystallinity was a very important properties of wood. The feasibility of near infrared spectroscopy to rapidly predict the x-ray diffraction (XRD) measured crystallinity of plantation wood was investigated in this paper. The results showed that NIR could be correlated with crystallinity of plantation wood, the correlation coefficient r between NIR predicted and XRD measured crystallinity of wood was above 0.84. It was noted that even we reduced the NIR spectra region, we can get good correlations between NIR spectra (i.e., 1 250 nm～2 050 nm and 2 050 nm～2 500 nm) and XRD measured crystallinity. However, the correlation between NIR spectra of wood in the short wavelength NIR region (780 nm～1 250 nm) and crystallinity was slightly decreased due to NIR spedctra in the short wavelength region showing subtle features. 4) The experimental variables studied included a brown-rot fungus and a white-rot fungus, and tree growth rates and six exposure periods in which weight loss and mechanical property reduction measurements of slash pine wood were recorded. The results of SAS analysis showed that there was a significant correlation between wood decay and exposure time, fungi species, and tree growth rates at the α=0.05 level. The weight loss, WML, MOR, and MOE losses of wood treated with brown-rot fungi were greater than those exposed to white-rot fungi, and the relationship between mechanical property reduction and weight loss due to brown-rot decay were very significant. The correlation of strength reduction with weight loss for brown-rot samples was greater than that of white-rot samples. There was a 34% reduction in work to maximum load (WML) and 28% decrease in MOR and 23% decrease in MOE of wood treated with brown-rot fungi even before the samples experienced was 10% weight loss (ie., during the early stage decay); The data showed that the ratio in the comparative loss in mechanical properties compared to weight loss was 7.2: 1 for WML, 6.4: 1 for MOR, and 3.8: 1 for MOE, for samples exposed to G. trabeum. The results for samples exposed to T. versicolor was 3.0 : 1 for WML, 2.2 : 1 for MOR, and 4.1 : 1 for MOE. 5) Three groups of wood samples, which randomly selected from control, white-rot fungi, and brown-rot fungi decay treated samples at weight loss as low as 10%, were recognized based on the principal component analysis (PCA) of near infrared spectroscopy data, the result showed that it's a potential applications of near infrared spectroscopy analysis for detecting and identifying early wood decay. Near infrared spectroscopy coupled with SIMCA (Soft independent modeling of class analog) and PLS discriminant analysis to detect early decay was investigated in this study. The result showed that the fungi species could be discriminated by SIMCA and PLS discriminant analysis from NIR spectra data. These techniques may be used to detect different types of biodegradation when it is subjected to fungi decay. 6) The correlation coefficient (r) for calibration models were 0.97 for weight loss and the quality of the test set was also very good with r value of 0.96 by PLS1 model for specimens sampling form 0%～30% of weight loss. When the specimens sampling from early stage decay (weight loss as low as 10%) were adopted to construct the PLS models, we also can find good results with r values of 0.94 and 0.93 for calibration and test, respectively. When wereduced the sampling range, e.g., weight loss as low as 5%, there has been good correlation between NIR spectra and weight loss. Even PLS model constructed with those specimens weight loss less 3%, the r values were 0.85 for calibration and 0.83 for test, respectively. 7) Regression of NIR test results and mechanical properties (i.e., WML, MOR and MOE) for those specimens sampling form 0%～30% of weight loss yielded a good correlation coefficient. The r values for the PLS2 calibration model were 0.90 for WML, 0.89 for MOR, and 0.83 for MOE, respectively, the r values were 0.88 for WML, 0.87 for MOR, and 0.83 for MOE. When the specimens sampling from early stage decay (weight loss less 10%) were adopted to construct the PLS2 models, the r values of test results were 0.85 for WML, 0.85 for MOR and 0.88 for MOE, respectively. When we reduced the sampling range (less 5% of weight loss), there was also good correlation between NIR spectra and mechanical properties. Even PLS2 model constructed with the specimens sampling from very early stage decay (i.e., weight loss less 3%), the quality of the test set was also very good with r values were 0.84 for WML, 0.84 for MOR and 0.80 and MOE, respectively. It is anticipated that this research can assist future efforts to 1) develop a model to predict Chinese plantation wood properties, 2) develop a comprehensive genetic tree improvement program for slash pine incorporating tree growth rate and decay resistance, and 3) develop some techniques coupled with NIR to for detecting wood decay (especially in early stage decay) and predicting residual mechanical strength.
|
PDF Full Text Request