摘要 : Interest in production of sawlogs from eucalypt plantations in Australia has increased as logging in native forests has declined. Management for sawlog production commonly involves early-age selection and pruning of potential sawl... 展开 Interest in production of sawlogs from eucalypt plantations in Australia has increased as logging in native forests has declined. Management for sawlog production commonly involves early-age selection and pruning of potential sawlog trees, and thinning. Existing stand-level models developed using data from unthinned plantations, grown primarily for pulpwood, cannot provide accurate estimates of sawlog yield for thinned plantations, and further model development is required. We addressed this issue using an individual-tree distance-independent modelling approach. Data from thinning and pruning experiments at three sites in northern Victoria were analysed, and diameter increment models were estimated for three tested species, <i>Eucalyptus globulus</i> Labill., <i>E. grandis</i> Hill ex Maiden and <i>E. nitens</i> (Deane and Maiden) Maiden. A new model form with use of two indicator variables was designed in this study to compare the growth of sawlog trees selected in thinned treatments with their equivalent cohorts in unthinned treatments. The best variables identified by regression analysis included initial diameter of the target tree, stand age, site index and time since thinning, together with a measure of thinning intensity (the removed basal area ratio) and a measure of inter-tree competition (basal area for the larger trees). Structured in this way, the developed models are flexible for predicting the thinning responses of selected sawlog trees for a wide range of tree size, site and stand conditions, and thinning treatments, leading to more accurate prediction of sawlog yield and associated size class distribution information. This result implies that better decisions on whether to thin may be achieved in plantation management practice. To provide the models with sound statistical properties, nonlinear mixed-model techniques were used to correct for within- and between-subject auto-correlation and heterogeneity of errors. Further study to validate the developed models for a wide range of stand and site conditions, and thinning treatments, using independent data is recommended. 收起