Modelling wood property variation among Tasmanian Eucalyptus nitens plantations

Characterisation of forest resources is increasingly focused on tree wood properties as they are important drivers of economic value. We used the Random Forests machine learning algorithm to model variability among plantations of Eucalyptus nitens in key wood properties affecting the value of solid-wood products. Using breast-height disk samples from 46 even-aged stands of different ages across the island of Tasmania, Australia, we modelled site variation in SilviScan-3^TM derived estimates of wood density, microfibril angle and modulus of elasticity. Regional-scale models were developed with respect to plantation age, environmental and climatic variables. Wood density and MOE increased, and MFA decreased with age. Wood density and MOE decreased and MFA increased with elevation. Increasing elevation is associated with increasing annual precipitation and decreasing temperature, but the variation in wood properties was mainly associated with precipitation. Wood density decreased and MFA increased with annual precipitation. MOE was positively related to wood density and negatively related to MFA, thus as expected, increased with annual precipitation. Using the Random Forests climate models we demonstrate the potential for predicting and mapping wood properties relevant for solid wood products at a scale that is relevant to strategic planning by the forest industry. Such models allow quantification of the impact of different growing conditions on the wood properties of harvested logs and therefore on wood products. When coupled with genetic, growth and economic models, these wood property models have the potential to assist in characterisation of standing forest resources and future estate planning.