Abstract
Examines the relative capacities of a set of models to account for patterns of distributions of 92 species of forest and woodland birds along a 250 km transect in SE Australia. Multivariate analyses of habitat ordinations (nonmetric, multidimensional scaling) were used to characterize the habitat breadth and habitat position of each of the avian species. Breadth is the range of structural types of habitat that a species occupies. Position is a measure of the "representativeness' of the typical habitat type occupied by a species relative to the entire sample of habitat types. There was a significant inverse correlation between breadth and position. The models were used to try to account for this inverse relationship and related statistics, including the amount of co-occurrence between species. Three models based on the geographic proximity of sites provided reasonable representations of the observed patterns, except for co-occurrence, which all models consistently underestimated. Many distributional patterns observed for this avifauna could be accounted for by using simple site-selection algorithms, and by assuming that species behaved independently. The major source of variation unaccounted for by the models was a consistent underestimation of levels of co-occurrence between avian species, which may mean that further modelling may require interdependence rather than independence of species.
Original language | English |
---|---|
Pages (from-to) | 360-374 |
Number of pages | 15 |
Journal | Ecology |
Volume | 71 |
Issue number | 1 |
Publication status | Published - 1990 |