Near Infrared Spectroscopy Based Larch Wood Carbon Content Prediction

As an important role in terrestrial ecosystems, forest ecosystems not only has the ability to improve and maintain the ecological environment, but also as the main role of the global carbon balance. The forest ecosystem highlights the important principal position to Carbon sequestration capacity. Nowadays,under the background of global warming, how to quickly and accurately determination of the carbon content of the forest has become an essential research topic. With the adoption of near-infrared spectroscopy (NIR) and Chemometric methods on larch samples, this study modeled the three sections(cross, radial and tangential section)of carbon content, the different organs(root, bark, trunk, branch and leaf)of carbon content and the overall of carbon content. The major results of this dissertation were as follows:(1)Use NIR technique combined with PCA and PLS predict Larch carbon storage. The correlation coefficient for calibration set and prediction set on using PCA are0.8030and0.7475, respectively, the SEC and SEP are0.2908and0.3721, respectively. The correlation coefficient for calibration set and prediction set on using PLS are0.7780and0.6818, respectively, the SEC and SEP are0.3066and0.4079, respectively.Compared with PCA and PLS, the model of correlation can be good reflected by PCA method.(2) Use NIR technique combined with PCA predict the three sections (cross, radial and tangential section)of carbon content, results showed that models on cross section are the best with correlation coefficient of0.8030and0.7475for the calibration set and prediction set, respectively. The correlation coefficient for calibration set and prediction set on radial section are0.7782and0.6932, respectively. The correlation coefficient for calibration set and prediction set on tangential section are0.7912and0.7287, respectively. The square error of calibration (SEC) were0.3782,0.3212and0.2520, respectively, while the square error of prediction (SEP) were0.3782,0.3212and0.2520, respectively. Compared carbon storage in cross, radial and tangential section: cross carbon storage>tangential carbon storage>radial carbon storage.(3) With PCA method, predict the different organs(root, bark, trunk, branch and leaf)of carbon content, results showed that models on trunk part are the best with correlation coefficient of0.8331and0.7912for the calibration set and prediction set, respectively, the SEC and SEP are0.1426and0.1493, respectively.The correlation coefficient for calibration set and prediction set branch part are0.8270and0.7697, respectively, the SEC and SEP are0.2086and0.3618, respectively.The correlation coefficient for calibration set and prediction set on bark part are0.8199and0.6902, respectively, the SEC and SEP are0.1645and0.2880, respectively. The correlation coefficient for calibration set and prediction set on root part are0.7905and7733, respectively, the SEC and SEP are0.2591and0.2921, respectively. The correlation coefficient for calibration set and prediction set on leaf part are0.6432and0.6703, respectively, the SEC and SEP are0.2752and0.3726, respectively.Compared carbon storage in root, bark, trunk, branch and leaf part:trunk carbon storage>branch carbon storage>bark carbon storage>root carbon storage>leaf carbon storage.(4) Using different forms of larch part of the trunk to establish carbon content model, compared models on solid sample and powder sample, the powder sample has higher the better correlation coefficient for the calibration set and prediction set is powder sample model, on correlation coefficient for the calibration set is higher3.6%, and for prediction set is higher5.6%.(5) With moving smoothing for averaging25point, First derivative, Second derivative, Multiplicative Scatter Correction and Standard Normalized Variate methods, predict the the best model on trunk and other organs(root, bark, branch and leaf)of carbon content, results showed the best method is MSC, the correlation coefficient for calibration set and prediction set on trunk part are0.8715and8454, respectively.