Predicting eucalypt distributions in Tasmania : an application of generalised linear modelling

This thesis develops a systematic approach to the routine prediction of Eucalyptus species' distributions in Tasmania from compiled ecological data comprising over 15 500 observations. The method of logistic regression, being an application of generalised linear modelling, was used to correlate species' occurrence with environment. Preliminary analyses tested sampling adequacy in terms of ecological variability and species' ranges, and derived environmental indices that could be directly related to plant physiological processes. Subsequent realised mche models were derived for the distribution of E. globulus in eastern regions of Tasmania considering biotic and abiotic attributes as predictors and relative dominance as a response in addition to occurrence. Different aspects of the ecology of this species were explored by considering response variables defined by vegetation class or pure and mixed stand occurrences of E. globulus and related species from the series Viminales. The results of these predictive models were displayed as nested, univariate responses along environmental gradients, representing a direct gradient analysis that facilitated their interpretation in terms of ecological theory and plant physiological processes.