Large ecosystems are often partitioned into spatial compartments (bio- or biophysical regions and/or ecoregions) in order to better understand the relative importance of ecosystem processes or for the purposes of managing human activities in relatively ecologically discrete areas. Regionalisation algorithms attempt to partition a broad spatial area into discrete spatial regions, each with relatively homogeneous and predictable ecosystem properties but with properties different from neighbouring regions. The Southern Ocean has been divided up into regions before, primarily based on frontal features. In this paper, we demonstrate a method developed for a regionalisation of the southern Indian Ocean in order to facilitate the development of ecosystem models for the area. Here, we extend this work to other areas of the Southern Ocean to see how well the approach might be applied more generally and therefore be of assistance to large scale ecological modelling and, perhaps, to CCAMLR in its work to develop a bioregionalisation of its Convention Area. This paper describes the steps and issues in undertaking a regionalisation and presents a statistical method for achieving a regionalisation of the Southern Ocean in an objective and consistent manner. Results are presented for each of the three CCAMLR Areas. We conclude that it is tractable to resolve the challenges facing the subdivision of the ocean into meaningful regions for the purposes of modelling and management.