AI models predict where reefs can recover
Oyster and mussel reefs are important bio-engineers that enhance biodiversity and strengthen coastal ecosystems. However, many of these reefs have disappeared due to overfishing, disease and changes in the ecosystem. Under the European Nature Restoration Regulation, the Netherlands must determine where restoration of these habitats is possible. In this project, we use AI (Random Forest) and Bayesian spatial models to create maps showing where the restoration of oyster and mussel reefs is most likely to succeed.
Where are the opportunities for restoring shellfish beds?
Oyster and mussel beds are among the most valuable habitats in coastal seas. These shellfish act as bio-builders: they form reefs that increase biodiversity, filter water and create habitats for many other species. In Dutch waters, such reefs were historically widespread, but many have disappeared due to overfishing, disease and changes in the ecosystem.
At the same time, the European Nature Restoration Regulation obliges Member States to actively work towards the restoration of degraded ecosystems. For the Netherlands, this means that the National Restoration Plan (Nationaal Natuurherstelplan) must identify where the restoration of marine habitats is possible and effective. An important question is therefore: where are the conditions suitable for the restoration of oyster and mussel beds?
From data to restoration opportunities
To answer this question, Deltares has developed habitat suitability maps for three key shellfish species: the flat oyster (Ostrea edulis), the blue mussel (Mytilus edulis) and the horse mussel (Modiolus modiolus). These maps show where in the North Sea, the Wadden Sea and the South-Western Delta the likelihood of stable shellfish beds forming is greatest. To this end, we combined ecological and physical data.
Many of the map layers used originate from Deltares’ numerical models, which provide spatial information on the physical conditions in the various waters. Information on seabed composition came from a new spatially comprehensive sediment map by TNO, based on thousands of seabed samples and AI analysis.
In addition, Wageningen Marine Research provided monitoring data with locations where oysters and mussels have actually been observed. These observations form an essential basis for statistically modelling the relationship between species and their habitat. Using artificial intelligence (Random Forest) and Bayesian spatial models, we then analysed the relationship between environmental conditions and the occurrence of shellfish.
Random Forest helps to identify complex patterns in large datasets, whilst Bayesian spatial statistics explicitly account for uncertainties and spatial relationships. Combining both methods yields a robust picture of habitat suitability. The statistical habitat models were also compared with Dynamic Energy Budget (DEB) models for oysters and mussels.
These mechanistic DEB models describe how environmental conditions influence the energy balance of organisms and thereby determine their growth, survival and reproduction. This allowed us to verify whether areas with a high probability of occurrence are also locations where populations can actually grow and reproduce, which provides additional confidence in the identified restoration opportunities.
From model to practical application
The results of this research are spatial maps showing where the restoration of shellfish beds is most likely to succeed. These maps provide policymakers, nature organisations and marine planners with an objective basis for planning and prioritising restoration measures.
The project was commissioned by the Ministry of Agriculture, Fisheries, Food Security and Nature (LVVN) as part of the National Restoration Plan (Nationaal Natuurherstelplan). By combining advanced data analysis, AI and ecological models, Deltares supports policymakers in identifying locations where the restoration of oyster and mussel beds is both ecologically feasible and sustainable.
"These models enable us to better assess where the restoration of shellfish beds (such as flat oysters and horse mussels) is likely to succeed. This helps us to implement protected areas and restoration measures under the Nature Restoration Regulation, the Marine Framework Directive, the North Sea Nature Enhancement Programme, and nature-inclusive wind farm construction in a more targeted and effective manner.” The results of the models are made available via InformatieHuis Marien so that everyone can make use of them as quickly as possible; there has been immediate and significant interest, including from other North Sea countries." - Dr Maarten de Jong Ministry of Agriculture, Fisheries, Food Security and Nature