Summary
We describe the use of structured graphs for the visual exploration of biodiversity data. A graph consists of a set of nodes, representing entities of interest, linked by edges showing relationships between those entities. For biodiversity data, the nodes are typically chosen to represent taxa (in which case the edges often represent co-occurrence or trophic interactions), or sites (in which case the edges might show similarities between sites in terms of species composition). In addition to a framework for visualising and exploring patterns and relationships, the structural characteristics of a graph can be used to gain insights into system-level properties that might not be easily gained from more traditional methods of analysis. For example, the consequences of the loss of a major species group in response to environmental change can be explored through simulated perturbation to the graph structure. We demonstrate the approach using a database of marine plankton samples acquired using a continuous plankton recorder towed behind ships in the Southern Ocean. Data collected from south of Tasmania to the Antarctic continent were used to construct a graph showing community structure based on co-occurrence of taxa. The graph showed good agreement with previously published results from more traditional cluster and ordination analysis methods. Physical environmental data such as sea surface temperature were used in the display of the graph (node and edge properties such as colour and shape) to allow visual exploration of the observed community structures with respect to environmental variables. A preliminary investigation of trophic relationships was undertaken by integrating seabird sighting data from surveys carried out on the same voyages but stored in a separate database. A bipartite graph was used to depict the relationships between the plankton and seabird observations and showed that seabird species such as prions, blue petrels, and storm petrels were associated with Antarctic krill Euphausia superba, in broad agreement with published predator-prey studies.