Using ensemble learning to model climate associated variation in wood properties of planted Eucalyptus nitens in north-western Tasmania

There is increasing interest in the phenotypic plasticity of tree species to variation in climate as it may affect the economic value of current and future plantations. This study applied ensemble learning methods to explore the plastic response of Eucalyptus nitens pulpwood selection traits of growth (DBH, diameter at 1.3 m height), wood density and Kraft pulp yield to variation in elevation, geography and associated climate variables. To help explain the pulp yield response, we also modelled underlying biological traits—cellulose, lignin and extractives. The study was based on data from 84 harvest-age plots of common genetic origin across the pulpwood plantation estate in north-western Tasmania, Australia. DBH and wood density were obtained from resistance profile traces, collected on standing trees using a drilling resistance tool. In addition, outerwood cores were taken at 1.3 m stem height for (i) calibrating the resistance measures for wood density, and (ii) for assessment of pulp yield and wood chemistry based on near infrared spectroscopy. Modelling of the variation in plot means using the random forest algorithms showed growth and wood properties were influenced by the growing period climate of the plot. Of the climate variables studied, growth was mainly influenced by temperature, while wood density was mainly affected by rainfall-related variables. Wood density varied independently of growth and decreased with increasing annual rainfall and elevation of the plots. Pulp yield had the poorest fit statistics, it was influenced by a mix of climatic and geographic variables, and appeared independent of variation in growth and wood density. The plot variation in pulp yield was best explained by the modelled trends in the underlying biological traits of cellulose and lignin. Using these models, the plastic response of the key pulpwood traits to climate was mapped across the E. nitens plantation estate to help predict current plantation attributes and guide future choices of sites for plantation establishment.