Benson, Vegetation of Agnes Banks 
47 
VEGETATION ANALYSIS 
Methods 
During mapping and description of the plant communities, considerable inter¬ 
gradation between them was recognized. This was particularly noticeable in the 
understorey species, whose topographic positions appeared to reflect a gradient of 
drainage change, both across the five communities, and within particular communi¬ 
ties. There was also evidence of frequent fire, (charcoal remains on trees, dead 
shrub remains, localized patches of species regarded as indicators of frequent fire, 
e.g. Pimelea lin folia ) and the question of its importance in determining the distribution 
of shrub species was raised. 
The data from the 24 sampling sites recorded in 1976 were therefore subjected 
to both classification and ordination analyses. Because of the concentration of 
most sampling in two communities, the E. sclerophylla-Angophora bakeri-Banksia 
serrata woodland and the E. parramattensis low open-woodland, most discussion is 
confined to the relationships within and between these two. However, some idea of 
the relationships of the two communities each represented by only one site may be 
indicated by the position of those sites in the classification or ordination analyses. 
Presence/absence data of species recorded from more than 10% of the sites, 
i.e. more than 2 of the 24, (Table 1) were analysed using programs from the TAXON 
library on the CS1RO CS1RONET Control Data Cyber 76 computer. Programs 
used were the agglomerative polythetic classification program MULTBET (Lance 
& Williams, 1967), the principal coordinate ordination program GOWER (Gower, 
1966) and the diagnostic programs GROUPER and GOWECOR. Of 131 species 
recorded from the sites, 70 were used in the analyses. Both normal and inverse 
MULTBET analyses were carried out and the results plotted in a two-way table 
of site groups against species groups. 
Results 
Following convention the MULTBET analyses arc discussed as though divisive, 
i.e. down the hierarchy. In the normal analysis the first major division is between 
well-drained and poorly-drained sites, the former corresponding generally to sites 
in the Eucalyptus sclerophylla-Angopliora bakeri-Banksia serrata woodland and the 
latter to sites in the Eucalyptus parramattensis low open-woodland. Species con¬ 
tributing to the dry-site groups are Leptospermum attenuatum, Bossiaea rhombifolia, 
Ricinocarpos pinifolius and Isopogon anemonifolius. These are generally known to 
be species of well-drained or deep sandy situations. Species contributing to the 
wet-site groups are Leptospermum flavescens, Drosera spatlmlata, Baeckea diosmifolia 
and Gonocarpus micrantlnts, species normally found in poorly-drained or moist 
situations. 
Lower divisions in the analysis could not be interpreted satisfactorily. 
The first axis of the GOWER ordination of sites (accounting for 26.8% of the 
variation) is interpreted as reflecting a moisture/drainage gradient (Figure 7). 
Species correlated with ordination axes were obtained from the diagnostic program 
GOWECOR. Positively correlated with the first axis were Acacia ulicifolia (0.83), 
Leptospermum attenuatum (0.83), Bossiaea rhombifolia (0.81), Ricinocarpos pinifolius 
(0.76) and Hove a linearis (0.75), species of well-drained or deep sandy situations. 
Negatively correlated were Leptospermum flavescens (—0.85), Baeckea diosmifolia 
( — 0.81), Gonocarpus micranthus (—0.76), Leptocarpus tenax (—0.75) and Drosera 
spathulata (-0.74), species normally found in poorly-drained or moist situations. 
The response of the species to the postulated moisture/drainage gradient is 
shown in Table 5. Here the MULTBET species groups, arranged approximately 
according to the first GOWECOR species axis, are tabulated against the sites 
arranged in first GOWER site axis order. Though some species are restricted to 
either the well-drained or poorly-drained ends of the gradient, many species overlap. 
