Summary
Regionalisation algorithms attempt to partition a broad spatial area into discrete spatial regions, each with relatively homogeneous environmental or biotic conditions. Marine regionalisations commonly make use of data that can be represented naturally in a network context, such as current flow fields and animal movements. We investigated a network-based method for regionalisation, in which each site in the study area is represented by a vertex in the network. Pairs of vertices that represent spatially-neighbouring sites are connected by edges, with edge weights reflecting the similarity of the sites in terms of their environmental or biotic characteristics. Network clustering algorithms can then be used to partition the network into discrete regions. We demonstrate the approach on the East Antarctic sector of the Southern Ocean, using nutrient, surface current flow, and bathymetry data. These results are preliminary but suggest that this may be a useful approach to regionalisation. We discuss a number of questions that were raised during this work, such as handling uncertainties in the region boundaries and the advantages of a network-based approach to regionalisation.