Prediction Of Wood Production And Quality Of Populus Xiaohei Based On Remote Sensing Technology

Tree crown constitutes the principal component of a tree; therefore it is frequently studiedfor the purposes of monitoring tree growth, predicting wood volume, and even evaluatingwood quality. Owing to their unique advantages such as large-scale observations, richness ininformation and repeatability, high spatial resolution remote sensing images provide an idealmeans to study forest tree crowns. The main purpose of this dissertation was to explore thefeasibility of evaluating wood production and quality of poplar plantation trees usinghigh-resolution remote sensing techniques, providing a new approach for assessment ofproduct value of plantation trees. Specifically, in the present study, Populus×xiaohei, one ofthe poplar clones was selected, and high-resolution QuickBird satellite image was used toextract tree crown information of individual trees. Subsequently, we attempted to establish thequantitative relationship of QuickBird-derived tree crown parameters and estimated treecharacteristics directly with growth characteristics, wood characteristics, stem volume,wetwood volume, juvenile wood volume and wood product (veneer, fiber) recovery ofindividual trees. The establishment of these models will allow the forest industry to analyzeand evaluate forest tree growth, wood quality and the potential wood product value recoveredat large scales, providing guidance for determining the optimum application of wood resources(veneer production or pulpwood). This research attempted to apply remote sensing techniquesinto wood processing and utilization, which is significant to forest management and orientedcultivation of wood resources. The main research results are summarized as follows:(1) Crown size had significant effects on tree growth attributes and increments (excludingtree height). Trees with larger crown grew vigorously, giving rise to higher diameter growthand more stem volume, while simultaneously resulting in lower live branch height, larger stemtaper and more juvenile wood and wetwood. There was a close relationship between crowncharacteristics and tree growth. On the other hand, crown size had little effects on wood basicdensity, modulus of rupture (MOR) and fiber morphological characteristics (fiber length, fiberwidth and the ratio of fiber length to width). while it imposed significant effects on averagering width, modulus of elasticity (MOE) and compression strength parallel to grain. The average ring width was positively and closely correlated with crown width, while MOE andcompression strength parallel to grain was negatively correlated with crown width. MOE andcompression strength parallel to grain were unable to be estimated well from the only variableof crown width (R~2=0.45). Therefore, for tree growth attributes that are closely related tocrown characteristic, i.e., diameter at breast height (DBH), stem volume, juvenile wood andwetwood area, it is possible to estimate these important growth attributes from crown widthderived from high resolution remote sensing data, while wood characteristics are hardlyestimated using remote sensing technique owing to their poor relationship with crown width.(2) A series of image analyzing and processing method such as geometric rectification,data fusion, shadow removal, and visual interpretation, were used successfully to accuratelyextract crown information from high resolution QuickBird satellite image. The extracted crownwidths were very close to the field-measured value, with the mean absolute error of 0.15 m andmean relative error of 5.74％. QuickBird image-based approach has the potential to be used toreplace or supplement the traditional field-based method for measurement of tree crowns.(3) Using either QuickBird image-derived crown width alone or estimated DBH and treeheight can successfully estimate individual stem volume of Populus×xiaohei with the R~2value raging from 0.87 to 0.92, depending on different model forms. Especially, stem volumecan be estimated directly from QuickBird image-derived crown width.(4) Wetwood represented a substantial part of the stem, accounting for 56.0％～65.2％ofstem volume. Wetwood was mainly observed in the lower part of the stem, and the wetwoodarea and proportion in the stem cross-section decreased from the tree base to the top,displaying a general conical shape within the stem. Tree spacing had a significant influence onwetwood growth and its proportion in the stem cross-section, and wider spacing resulted inmore wetwood. Estimation of wetwood volume from external tree characteristics (DBH, treeheight and live branch height) was possible using various regression models with fairly highcoefficients of determination ranging from 0.87～0.92. The variable of DBH was the bestexplanatory variable for the prediction of wetwood volume.(5) The distribution of juvenile wood within the stem was not constant. A larger juvenilewood area and higher proportion was found at the stem base. Along the stem upward, thejuvenile period shortened gradually, and meanwhile the juvenile wood area and its proportion in the stem cross-section decreased. The juvenile wood proportion for 27-year-old Populus×xiaohei ranged from 68.5％to 79.0％. There were significant differences in the volume andproportion of juvenile wood between different spacings. It was possible to estimate wetwoodvolume of individual tree using two-order polynomial models with DBH as the onlyexplanatory variable (R~2=0.95).(6) DBH estimated from QuickBird image-derived crown width could alone explain mostvariation in both wetwood and juvenile wood volume of individual trees, with R~2 being 0.78and 0.83, respectively. It is possible to estimate wetwood and juvenile wood volume ofindividual trees of Populus×xiaohei using high resolution QuickBird satellite data.(7) The veneer volume recovery and fiber yield of individual trees were accuratelyestimated using back-propagation (BP) artificial neural network with field-measured treecharacteristics (DBH, tree height and the ratio of tree height to DBH) as the input variables.The mean prediction accuracies for dry fiber weight, cellulose yield, holocellulose yield, ligninyield, extractives yield and veneer volume recovery were 94.53％, 94.61％, 94.97％, 93.22％,84.31％and 83.57％, respectively.(8) The mean prediction accuracies of BP artificial neural network exceeded 80％for dryfiber weight, cellulose yield, holocellulose yield, lignin yield and veneer volume recoverybased on QuickBird image-derived tree characteristics. It is possible to estimate veneer volumerecovery and fiber yield of individual trees using high resolution QuickBird satellite data.