Identifying abrupt biogeographic shifts in a complex model ocean ecosystem

Ocean ecosystems are changing, and will continue to change. Some of these changes are gradual and straightforward to characterize; other changes have been or may be sudden, unpredictable, enigmatic, and catastrophic. It is therefore valuable to discern which components or properties of ecosystems are predisposed to sudden shifts, when and where they are likely to occur, what statistical metrics are most effective at identifying them, and most importantly whether early warning statistical signals are capable of predicting them. A number of metrics have been proposed for identifying abrupt shifts (e.g. change point analysis) and a number of different early warning signals have been proposed for their prediction (e.g. increased autocorrelation). Metrics of identification will not necessarily agree on what counts as an abrupt shift, and identified shifts may be associated with different early warning signals. To this end, we explored the capacity of these metrics to identify and predict abrupt changes in future projections of a state-of-the-art complex ocean ecosystem model, focusing on biogeographic transitions. The model includes biogeochemical functional and trophic groups (e.g. diatoms, mixotrophs) and a range of size classes from 0.6 to >200 µm in diameter. This virtual ecosystem is perturbed with a “business as usual” scenario that results in ~3°C sea surface temperature warming, sea ice retreat, increased stratification, and an altered overturning circulation. Phytoplankton type turnover by the end of the century is greatest in the Arctic and edges of oligotrophic gyres. This turnover reflects a shift in phytoplankton community composition, with diatoms disappearing in the Arctic and being replaced by coccolithophores and dinoflagellates. We discuss the attribution of these shifts to specific mechanistic drivers, as well as their the ecological and biogeochemical consequences.