Ecological association of birds and vegetation structure in wet tropical forests of Australia

Publisher Website
Google Scholar
Abstract
Birds and environments of tropical rainforests were studied in 18 selected areas from sea level to 1520 m within the humid tropical region of northeastern Australia. A total of 122 sites (20 m × 20 m each) were established within the 18 areas, and bird data (observation and mist‐netting) and environmental data (pro forma survey and foliage measurements) were collected at each site. The bird data were analysed numerically in two ways: abundance of 146 species in 18 areas and presence and absence of 108 species at 122 sites. Both classification and ordination analyses produced a major discontinuity between the tableland sites and the lowland sites, each divisible further into various types of vine forest and non‐vine forest vegetation. The few rare species did not contribute to the general pattern. The environmental data were reduced to a set of those attributes that showed variation among the sites. They included topographic and edaphic features, physiognomic and structural features of vegetation, fruiting and defoliating habits of plants, foliage height diversity and evenness, and canopy cover. The primary pattern in the data set of 135 attributes by 122 sites was vine forest versus nonvine forest, and altitudinal divisions appeared as a secondary pattern, in both classification and ordination.Canonical correlation coefficients between bird vectors and environmental vectors were reasonably high for the first four canonical vectors, which correlated well with major Gower vectors of birds and environments. The back‐correlation to the raw data produced bird species and matching environmental attributes with sites sharing them for both positive and negative ends of each canonical vector. In general, the analysis extracted elements of both fauna and environment that characterized the vine forest as opposed to the non‐vine forest and the tableland sites as opposed to the lowland sites. Canonical correlation analysis is limited by the unverifiable nature of derived correlations. Such correlations may suggest certain associations to be tested by independent means (hypothesis‐generating function), but may also formalize self‐evident or fortuitous associations.