As the ocean warms, marine species are on the move, but where will they go, and which areas will remain safe havens? Our biodiversity data science group just published a new study exploring the future of 16 cetacean species in the Northeast Atlantic, applying advanced artificial intelligence to project how these vital marine mammals will respond to a changing climate.

Knowledge gap

When predicting climate change impacts on biodiversity, scientists typically focus on single dimensions of change, such as total habitat loss or poleward range shifts. However, a highly mobile species might lose much of its current range while successfully expanding into new areas, or it might shrink in extent but maintain a highly stable core habitat. Evaluating these dimensions separately fails to capture the true climate vulnerability of marine ecosystems and limits our ability to make effective conservation decisions.

Main approach

We modelled the habitat suitability of 16 distinct whale and dolphin species across the Northeast Atlantic from 2030 to 2100. By evaluating a low-emissions and a high-emissions climate scenario, we simultaneously quantified changes in habitat extent, spatial redistribution, and the long-term persistence of suitable conditions. Integrating these different dimensions creates a much more comprehensive framework for understanding climate-driven habitat change.

Technological challenge

Predicting ecological shifts decades into the future requires computational models that can transfer reliably to novel, unseen environmental conditions without overfitting to present-day data. We tackled this by using Bayesian Additive Regression Trees (BART), an advanced machine learning algorithm that flexibly models complex, non-linear species-environment relationships while controlling for high model uncertainty. Additionally, to process massive but heterogeneous occurrence datasets—ranging from citizen science sightings to standardized fisheries observations—we developed rigorous environmental thinning protocols to ensure robust, unbiased model calibration.

Main finding

Our models reveal that climate change is projected to reorganize suitable habitats rather than uniformly reduce their extent. While some species, like the killer whale, face severe habitat contraction of up to 48%, others may expand or rapidly shift poleward. Crucially, despite this extensive redistribution, we identified specific macro-areas—most notably the northwest African shelf, the Iberian shelf-break, the Bay of Biscay, and the Norwegian shelf—that consistently act as multi-species refugia. These areas maintain highly stable environmental conditions despite widespread ocean warming.

Main implications

These findings directly support the UN 2030 Agenda for Sustainable Development by providing an actionable, forward-looking blueprint for marine conservation. The varying responses of different species indicate that no single management strategy will work for all.

For species facing persistent habitat contraction, we must rapidly safeguard existing habitats and minimize local anthropogenic stressors.

For climate-tracking species, management efforts must focus on maintaining ecological connectivity and coordination across international borders as populations shift poleward.

Ultimately, the persistent multi-species refugia we identified represent immediate, strategic priorities for establishing adaptive marine protected area (MPA) networks capable of protecting biodiversity in a rapidly changing ocean.