• Indonesia;
  • parrot;
  • Wallacea;
  • habitat predictions;
  • logistic regression;
  • partialling;
  • ROC plots


Aim To investigate patterns of habitat use in nineteen populations of parrots on three Wallacean islands, and examine the degree to which habitats used by a species can be predicted across islands.

Location The Wallacean islands on Buru, Seram and Sumba.

Methods A logistic regression method was used to relate the presence/absence of parrot species at census plots to vegetation axes derived from ‘gross’ habitat parameters such as altitude, tree sizes, and percentages of vegetation coverage at different strata. Receiver Operating Characteristic Plots were then used to test predictions of habitat use by conspecific populations on the nearby and distant islands, and by different parrot species on the same island.

Results Parrots on Buru and nearby Seram tended to be associated with open-canopied lowland forests, while on the distant and low-lying island of Sumba, the preference was for closed-canopy forests at higher altitudes. There were usually close affinities in habitat usage by conspecific populations on the two nearby islands. One exception was where the species occurred on the two islands as separate subspecies. Habitat preferences on Sumba, however, were so distinct that paradigms developed for Buru and Seram failed to predict distributions on this distant island.

Main conclusions For populations on more distant islands, it will usually be better to predict habitat use by considering the habitat used by other parrot species occurring on that island, than a conspecific population from another region. Conservation research programmes which concentrate parrot study on just one or two islands in every region of Wallacea may be more successful than those which ‘run through’ several islands in a group, but which miss out whole regions.