The abundance of birds, reptiles and frogs was recorded at 370 quadrats and the abundance of
mammals and the basal area of woody plants was recorded at these and a further 10 quadrats in
Stage I11 of Kakadu National Park. Patterns in the distribution of these species were related to two
environmental mapping schemes. The more specific and localised of these was a 1 : 100 000 habitat
map for Kakadu National Park, established by Schodde et al. (1987), based on environmental
attributes considered to be significant for the distribution of wildlife species: floristics, vegetation
structure, substrate and landscape position. The more generalised scheme was that of Wilson et a[.
(1991), which mapped vegetation communities at a 1:1 000 000 scale across the Northern Territory.
The distributions of assemblages of plant, all vertebrate, bird and reptile species were strongly
associated with the distributions of both Schodde habitats and Wilson vegetation units. The associations
were less good, but still highly significant, for distributions of native mammal and frog assemblages.
Patterns in the similarity of species composition between the different habitats or vegetation units
varied between different animal and plant groups. Similarity in vertebrate species composition was high
between most sandstone habitats. Floristic similarity was high between woodland habitats on different
substrates. Melaleuca open forests were distinct from other vegetation units in their composition of bird,
frog, reptile and plant species. Mammal species composition divided the vegetation units into an upland
rocky group and a lowland group.
The distribution and abundance of most individual animal species were significantly related to the
habitat or floristic unit divisions. This association was clearer for species recorded from at least
20 quadrats than for those recorded from 6-19 quadrats. For the latter group of species, association
was more apparent with the Schodde habitat scheme than with the Wilson vegetation classification.
The proportion of native mammals that showed significant associations with either classification was
smaller than that for birds and reptites.
The generally significant associations between distributions of individual species and the mapping
of defined habitats suggests that the Schodde scheme offers a useful template for predicting species
distributions within Stage 111 of Kakadu. However, the restriction of this habitat mapping to the
Kakadu area renders this scheme inapplicable for the prediction of distributions beyond Kakadu, and
therefore handicaps the assessment of the wildlife value of Kakadu National Park in a regional context.
The Wilson vegetation map can be used to extrapolate distributions beyond Kakadu, but because that
~lassification scheme includes many vegetation units that were not sampled within the Kakadu area the
predicted distribution in this case will be very incomplete.