Factors influencing tree growth in tropical savanna: Studies of an abrupt Eucalyptus boundary at Yapilika, Melville Island, northern Australia

Most of the land surface of Melville Island, Australia's second largest island, is covered in Eucalyptus savanna. One exception is an area at Yapilika where a large tract of savanna is dominated by Acacia shrubs. An ordination analysis of 122 quadrats revealed that the boundary of Eucalyptus dominance did not correspond to a major change in floristic composition. Detailed transect studies at one site on the boundary showed that Eucalyptus trees were abruptly replaced by a band of Grevillea trees which gradually gave way to Acacia shrub dominance. There was a gradual change in the floristic composition of the savanna across the boundary. The distributional limit of Eucalyptus was found to be independent of any hydrological discontinuity. There was a slight decrease (<2.5 m) in altitude from Eucalyptus to Acacia savanna. The Acacia savanna soils were sandier and their surface soil had significantly lower concentrations of Ca and Mg and significantly greater concentration of Al compared with the Eucalyptus savanna soils. Eucalyptus seedlings planted in the three savanna communities were not found to be under drought stress (pre-dawn leaf xylem potentials of >-0.9 MPa) during the dry season. Over a 12 month period Eucalyptus tetrodonta and E. miniata seedling growth was not significantly different on the Acacia or Eucalyptus savanna, although this result may be due to the counteracting effects of greater soil fertility and tree competition in the Eucalyptus savanna and lower soil fertility in the treeless, and hence competition-free, Acacia savanna. This hypothesis is supported by the significantly greater growth of Eucalyptus seedlings on fertilized Acacia savanna soils. The limited production, dispersal and establishment of Eucalyptus seeds and the greater frequency of fires in the Acacia savanna probably explains the abrupt limit to Eucalyptus dominance along the edaphic gradient.