Mr Clément Violet1, Mr Thomas Benoit1, PhD Aurélien Boyé1, PhD Graham Edgar2, PhD Rick Stuart-Smith2, PhD Martin Marzloff1
1Ifremer DYNECO LEBCO, Plouzané, France, 2Institute for Marine and Antarctic Studies, Hobart, Australia
Temperate coastal reef habitats increasingly exhibit abrupt regime shifts, suddenly switching to fundamentally different and persistent ecosystem states. These changes alter the delivery of ecosystem services and may be hard to reverse, but predicting and anticipating their occurrence is challenging. Indeed, increasing anthropogenic pressures over the past decades have eroded the resilience of coastal reef habitats, even small changes in environmental conditions can now tip these ecosystems into new states. Thus, a better understanding of the drivers and risk of regime shifts is necessary to improve management and conservation outcomes.
Identifying early warning signals of regime shifts requires long-time series at high frequency. Such datasets are rare and inadequate to guide management policy at large scale. One alternative would be to better understand how environmental and anthropogenic drivers influence the distribution of alternative regime states.
Here, we applied machine learning to predict reef states using environmental and anthropogenic data. Reef states were characterised from a typology defined using habitat coverage data from temperate sites sampled by the Reef Life Survey program. With this model, we identified the main factors associated with the different states as well as key thresholds fostering transition between states. Then, by projecting the expected states across the world’ s temperate coasts, we were able to define (1) expected dominant reefs states in data-poor regions, (2) identify zones that are likely to host alternative stable states and thus with higher risks of regime shifts, and (3) transition areas prone to environmentally driven regime shifts.
Presentation Slides – Clement Violet
