-Sixty-three quadrats each of 16 ha were chosen to represent the geographical extent and diversity of terrestrial environments in a 75 000 km 2 area of the Carnarvon Basin, Western Australia. A total of 626 plant and 456 animal species were recorded from the quadrats, an average of 120 (s.d. = 22.1) species per quadrat.After species that occurred at only one quadrat and species for which the sampling methods were unreliable (e.g. snakes and raptors) were removed from the data-set, 730 species remained, an average of 108.6 per quadrat (s.d. = 20.1). These comprised 81 herpetofauna, 13 small ground mammals, 85 birds, 9 scorpions, 12 centipedes, 122 ground-dwelling spiders and 408 plants.The data were compiled into a single matrix comprising the presence or absence of the 731 species at the quadrats. When the species were classified according to their co-occurrences, thirteen assemblages were distinguished. Each assemblage could be characterised in terms of the Australia-wide habitat preferences of its component species. Also, quadrat similarity matrices were generated for each of the seven types of organism sampled and 1000 random matrices. These were output as linear similarity vectors so that the differences in their biodiversity patterns could be quantified as a single matrix of correlation coefficients.Analyses revealed that: 1. Geographical patterns in species composition derived from the combined matrix correlated with processes operating at both biogeographical and local (ecological) scales: the compositional structure of each assemblage was related to a different set of climatic plus soil and/or topographic attributes. Poisson error models with logarithmic links fitted the gradient in species richness of each assemblage across the study area. Similar environmental attributes emerged whether an assemblage's composition or its richness was analysed. Since these attribute-sets were also consistent with the assemblages' Australia-wide characterisations, they are unlikely to be artifacts of quadrat positioning or study area extent. 2. Each of the seven ecologically different types of organism had a distinct influence on the biodiversity model; cross-taxon congruence levels were low. To be representative, a Carnarvon Basin reserve system should sample the geographical range of the various climatic, soil and topographic gradients identified by the analyses. It should also be designed using a biodiversity model that incorporates a wide range of organisms.

-The Pilbara region has one of the most diverse reptile assemblages in the world and includes many typical arid zone species as well as many saxicoline endemics. We present the results of a four-year survey of the region during which pitfall trap lines were used to sample 297 quadrats for a total of two weeks in spring and in autumn. The quadrats were located across the region and were chosen to sample the main substrate types. Environmental variables were measured to characterise the soil as well as the geomorphic, vegetative and climatic setting of each quadrat. We used the program PATN to expose patterns in species composition among quadrats, and in relation to the environmental variables.A strong signal for community structure emerged that was linked to surface type. Broad rock, clay, sand and loam categorisations seemed to explain most assemblage composition, with climatic variables showing only weak infl uence. However, many taxa occurred across several surface types, refl ecting the close juxtaposition of habitats within the region, wide dispersion of some taxa, and perhaps the infl uence of recent environmental history including anthropogenic changes.Examination of biogeographic characteristics of the species groups indicated strong endemism for taxa that preferred rocky substrates. In addition, many saxicoline Pilbara taxa have ranges that extend south into rocky areas of the Gascoyne, western Murchison and adjacent areas. Sandy habitats similar to those found in the Pilbara are widespread across Australia's arid zone, and accordingly few endemic taxa showed a preference for sandy surface types, and taxa with such preferences have distributions extending far beyond the Pilbara. The sandy areas of the coastal plain in the Roebourne subregion contained many taxa from Dampierland and the Great Sandy Desert to the north and north-east, as well as some taxa from the rich sand-adapted fauna of the Carnarvon region to the south. Some endemic taxa are associated with clay in the region, although clay surfaces were not as extensive as the rocky, loamy and sandy surfaces. Species associated with the fi ne-textured soils of the mulga woodlands in the southern part of the region appear to be an extension of herpetofaunal communities of regions to the south.Herpetofauna composition at the regional scale appears to have persisted, and patterns discerned in our analysis appear robust, despite the changes that have accompanied European colonisation in the Pilbara, including pastoralism, changed fi re regimes, weeds, introduced animals, mining and infrastructure. However, the existing reserve system does not adequately capture the region's herpetofaunal diversity, especially the endemic taxa in the Fortescue and Roebourne subregions, and there are likely to have been changes in community composition at the local level.

-Cround-dwelling frogs, reptiles and small mammals were sampled at 252 quad rats chosen to represent the geographical extent and diversity of uncleared terrestrial environments across the WestE'rn Australian wheatbelt. These sitE's were not overtlv affected by secondarv salinisation, but did include sites that were 'natu~ally' saline. We recorde~i a total of 144 species from 74 genera and 15 families. There was an average of 10.4 species per quadrat with a range from one to 19. Vertebrate species richness was highest on dissection valley floors and sandy depositional surfaces of the 'old plateau' but lowest on saltflats. Total species richness was positively correlated with high levels of sand, with low levels of soil nutrients and with good soil drainage. When frogs, reptiles and mammals were considered separately, temperature and rainfall attributes were also shown to exhibit correlations with species richness. Patterns in species composition could be explained in terms of climatic and substrate variables, including salinity. Two distinct faunas were identified -one concentrated in the semi-arid northern and inland parts of the study area, and one concentrated in the more mesic south and south-east. Further patterning could be discerned within this dichotomy, including the existence of a small group of species that arc associated with saline areas. Correlations between climatic and substrate variables could be discerned even at very low levels in the classification analysis, suggesting strong deterministic patterns in vertebrate species composition across the study area. Integrated management programs over entire catchments will be necessary in order to maintain conservation values, but there is doubt that the impacts~)f salinisation and fragmentation can be mitigated quickly enough to allow the small, ground-dwelling vertebrate fauna to withstand the effects of these processes in the Western Australian wheatbelt.