To investigate effects of scale on basin communities in New Zealand’s fjords, quantitative patterns of deep benthic faunal composition and abundance are presented. In Doubtful Sound, basin communities show high, small-scale patchy distributions. Diversity varies between basins significantly. Basins fauna at the entrance may represent a subset of species found on near-by continental shelf. Although inner fjord basin fauna was highly reduced, some unique elements were found, possibly more representative of deep ocean habitats, such as species of the families Nuculanidae, Neilonellidae, and Malletidae (Bivalvia), Aplacophora, Trochochaetidae (Polychaeta), and Pogonophora. These taxa may represent an imprint of the first basin communities established after post-glacial sea level rise flooded the fjord basins, and reinforce the notion that fjord faunas are a unique assemblage in the Forsterian biogeographical area.

Physical factors (sediment characteristics, carbon and nitrogen content) varied between basins, and showed high within-basin spatial structure (spatial autocorrelation) at the 200 m spatial scale, while biotic variables were structured over >1000 m spatial scales. This lack of similar spatial structure between biotic and abiotic variables suggests that basin-wide environmental factors may mask small-scale patterns over basin landscapes.

Fjord-wide gradients were examined, where strong gradient patterns in small scale (sediment variables), local scale (disturbance, between basin connectivity, surface production), and fjord-wide scale (connectivity to regional pool) explained 36% of variation in basin community structure throughout Doubtful Sound. Partitioning of variation showed that local (basin-wide) factors were important in structuring communities, while small scale factors were less important. Small and local scale factors were correlated. Fjord-wide factors were least important in structuring communities.

To test the hypothesis that processes of between-basin connectivity play an important role in maintaining local community structure, the predictions of a metapopulation model were examined. Based on results above, basin communities meet the assumptions required before a metapopulation model can be useful. The relationships between species patch (basin) occupancy and abundance, and factors associated with extinction and colonisation were examined. Abundance, occupancy, colonisation and extinction all showed patterns predicted by a metapopulation model. Community maintenance in the deep-basins of Fiordland may operate in a similar way to landbridge islands, where basin communities (islands) contained a full compliment of species from the regional pool (mainland) when they were first formed during the last post-glacial sea level rise. Over time, competitive and deterministic processes have modified communities such that a bimodal species occupancy distribution is predicted by the core-satellite species hypothesis. Using only the polychaete fauna (to meet assumptions), a weak bimodal distribution was observed.

New Zealand fjords are unique in that they represent an overlap between shallow estuarine conditions of strong horizontal and vertical physical gradients, and deep-sea systems, where organic input is poor, and the benthic habitats are comparatively stable. The metapopulation model provides a framework for future management of the fjord ecosystem, and demonstrates that metapopulation models can be examined in the marine environment when the appropriate spatial scales are defined.