Variation in mammal browsing damage between eucalypt plantations in Tasmania, and attempts to associate the variation with environmental features

Mammal browsing damage to Eucalyptus nitens and E. globulus seedlings in 23 eucalypt plantations throughout Tasmania was surveyed over time by repeated visits. All sites were planted with no initial browsing protection and monitored intensively. The levels of browsing damage overtime varied considerably among plantations, but more than half were damaged at rates sufficiently low as to suggest reactive, low-intensity post-planting browsing control options would have been more appropriate in such sites rather than intensive, pre-emptive browsing control before planting. In order to investigate the possibility of identifying such sites before planting, three different predictive modelling approaches were taken (ordinary least squares regression, binary response logistic regression, and classification regression trees) using 97 variables derived from GIS data that described habitat within and around the plantations. Five variables significantly explained the variation in browsing pressure, but when models based on these variables were tested on a validation set of plantations that had been classified as either high or low browsing damage based on detailed records of their browsing control operations, they performed poorly. Principal components analysis using the five GIS variables showed that the 23 plantations used for modelling occupied a similar ordination space as the plantations used for validation. We conclude that browsing pressure is related to a wider range of variables than those we were able to identify from the 23 intensively-monitored plantations.