-The inshore islands of the northern Kimberley are relatively un-altered examples of the adjacent mainland's substrates and vegetation communities. Being naturally sheltered from many anthropogenic disturbances evident on the mainland, these islands are likely to be important refuges for mammals. A systematic survey of 24 of the largest islands along the northern Kimberley coast revealed the presence of 18 non-volant mammal species (19 species are now known from these islands), most of which fall within the 'critical weight range' of 35 g-5.5 kg. We recorded previously undetected species on a number of islands including two new island populations of each of the threatened species Dasyurus hallucatus, Isoodon auratus and Mesembriomys macrurus, and one new population of the Kimberley endemic Zyzomys woodwardi. The persistence of two other Kimberley endemics on the islands, Petrogale burbidgei and Wyulda squamicaudata was also confi rmed. Small mammals (≤35 g) were rarely detected on the islands, with only two small rodents recorded (Pseudomys nanus and P. delicatulus). The most important factor determining mammal species richness was island area, closely followed by distance to the mainland, then proximity to a river mouth. Species composition was related to presence of preferred mainland habitats.

-A biological survey of the Pilbara biogeographic region was undertaken between 2002 and 2007 to provide a regional perspective on biodiversity patterns as a contribution to nature conservation planning. During this survey, 304 sites were sampled for small ground-dwelling mammals, birds, reptiles, spiders, ants, beetles and scorpions. A further 98 sites were sampled for wetland invertebrates, aquatic macrophytes and fringing riparian vegetation. Data for these two groups of sites were aggregated separately (i.e. terrestrial fauna and wetland biodiversity) and models of turnover in species composition within each data set were developed using generalised dissimilarity modelling (GDM). A wide range of environmental variables was assessed as predictors of compositional turnover -biotic (vegetation cover indices), climate, landform, hydrologic, regolith (soil and geology) and geographic distance. Generally, predictors associated with regolith were the most strongly supported in both the terrestrial fauna and wetland biodiversity models, followed by combined landform/ hydrologic variables, then climate/biotic variables. Geographic distance between sites was retained in the terrestrial fauna model only. The fi nal GDM models explained 46.1% and 58.5% of the deviance in the compositional turnover of terrestrial fauna and wetland biodiversity, respectively. Spatial representation of the coverage of survey sites showed that a large proportion of the core study area was well represented for both terrestrial fauna and wetland biodiversity. However, gaps in the proportional representation of both groups within the 2011 conservation reserve system were evident, particularly in the coastal region of the Pilbara (Roebourne subregion) and the Fortescue River valley (Fortescue subregion). With the addition of proposed reserves (in 2015) within these two subregions, the representation of terrestrial fauna and wetland biodiversity was substantially improved.