van Etten & Fox: Vegetation classification - Hamerlsey Ranges 
communities. Procrustes rotation between the two 
ordinations demonstrated a very good level of fit (0.15) 
given the number of sites (Belbin 1994), indicating very 
little difference in general patterns displayed between 
using cover values and IVs as the initial data. 
The three-dimensional ordination of all sites is 
displayed as two separate 2-dimensional plots (Fig 3). 
Sites belonging to a certain plant community generally 
grouped together in the ordination. Despite this general 
trend, the two outstanding features of the ordination 
plots were: 1) the considerable spread of sites within each 
community (reflects relatively high variability in species 
composition between sites within a plant community); 
and 2) the overlap of the ordination space occupied by a 
particular community with that of other communities. In 
other words, few discrete communities occurred. Indeed, 
a small number of communities clearly graded into other 
communities. There was also a small number of outliers 
(sites which are placed some distance from their 
respective community). 
Hummock grassland at upland sites (communities 1A, 
IB, 1C, 4A, 4B) were separated from the lowland 
communities more or less along the first vector (Fig 3). 
Generally the sequence of communities along this vector 
reflected topographic position with the mountain top 
community (4A) uppermost, followed by other upland 
communities, then pediment and bajada communities, 
with lowland communities at the base of this vector. 
Species strongly and significantly correlated to this 
vector were A. aneura (r = -0.79), T. basedowii (+0.74), E. 
leucophloia (+0.70), T. triandra (-0.67), C. fallax (-0.62), £. 
xerothermica (-0.51), A. adoxa (+0.47), S. glutinosa subsp. 
glutinosa (+0.41), E. victrix (-0.40), M. villosa (-0.39), E. 
'lanceolata' (-0.36), E. benthamii (-0.36), R. eremaea (-0.36) 
and C. hamersleyana (+0.35). Of the upland communities, 
1A, IB & 1C were reasonably separate from each other 
along the first and third vectors (Fig 3b), however 
community 4B was highly variable and overlapped with 
each of these communities (Fig 3b), as well as 
communities 3A & 3B (Fig 3a). Community 4A (E. 
kingsmillii low mallee over hummock grass on mountain 
tops) formed a fairly distinct group along the first vector, 
but showed considerable variation along the third vector 
where it overlapped with communities 4B, IB and 1C. 
The slight overlap with 1C was most likely by virtue of 
site E104 which is a mountain top site dominated by T. 
brizoides rather than T. wiseana, as is typical of community 
4A. Pediment communities (2A and 2B) were not readily 
discernible from one another, were extremely variable 
and overlapped considerably with hummock grassland 
communities of uplands. They occupied the ordination 
space between the upland hummock grassland 
communities and the woodland communities on flats and 
other drainage features, although there was appreciably 
less overlap with the lowland communities (Fig 3). 2B 
was generally positioned between 2A and upland sites. 
Communities on volcanics and calcareous soils (3A, 
3B & 4B) were generally separated from those derived 
from ironstone formations along vector 3 (Fig 3b). 
Hummock grassland with mallee (community 3A) 
formed a distinguishable, albeit variable, group at the left 
of the plots (Fig 3). This group overlapped considerably 
with the community 4B group; both groups shared a 
dominance of T. wiseana, but have different dominant 
trees. Community 4B therefore occupied the ordination 
space between the calcareous and iron-formation 
hummock grasslands. The snakewood woodland 
community on calcareous soils (community 3B) formed a 
more discrete and less variable group than 3A, although 
there was some overlap. 
In terms of lowlands, community 6B (mulga woodland 
with mixed T. melvillei and tussock grass understorey) 
occupied the ordination space between communities 5 
(bajada mulga with T. melvillei) and 6A (mulga with 
tussock grass), and overlapped slightly with both (Fig 3). 
Mulga woodland on steep scree slopes (community 4C) 
formed a fairly discrete and confined group in the middle 
of the ordination space between the other mulga 
woodland communities and hummock grassland 
communities dominated by T. iviseana. The two drainage 
line communities (7A & 7B) separated along the first 
vector (Fig 3a). Community 7A overlapped considerably 
with mulga woodland communities, while 7B overlapped 
more with hummock grassland on calcareous and 
volcanic soils (3A & 3B; Fig 3a), as well as community 2A 
(Fig 3b). Sites on drainage and creek-lines seemed to be 
placed (in the ordination space) near to the communities 
which they abut in real (i.e. geographic) space. 
Community 8 (E. victrix woodland in broad drainage 
basins) formed the most distinct group, although it was 
placed near mulga woodlands. 
Differential profile 
The distributions of communities along the Marandoo 
topo-sequence are shown in the differential profile (Fig 
4) with the degree of overlap between adjacent 
communities demonstrated in terms of species 
composition (vertical axis) and geographic distance 
(horizontal axis). A major change in species composition 
occurs between community 5 (T. melvillei hummock 
grassland) and communities 2A/2B (T. basedowii 
hummock grassland). The transition between 1A and 
2A/2B was considerably more gradual both spatially and 
floristically, whilst that between 5 and 6B was relatively 
sharp, although it involved only a small degree of change 
in species composition (Fig 4). Other transitions were 
difficult to describe given inadequate sampling in and 
around the transition zones. 
Discussion 
This paper identifies and describes 16 plant 
communities within an approximately 8,000 km 2 section 
of the central Hamersley Ranges. Indicator species are 
also identified for each community. A small number of 
communities are distinct in terms of floristic composition, 
structure, species richness and diversity, whilst most are 
only subtly different and intergrade with other 
communities in respect of these parameters. The pattern 
of species distribution across the landscape is one of 
several dominant to sub-dominant species found across a 
wide range of communities (such as T. wiseana, T. 
basedowii, A. aneura, E. leucophloia, 5. glutinosa, P. 
obovatus), with many of the other species uncommon and 
restricted to one or a few communities. A similar pattern 
of species distribution has been found in other 
mountainous deserts, such as Turkana, Northern Kenya 
(Patten & Ellis 1995), Oman (Ghazanfar 1991), central 
75 
