Simulating and analysing seabird flyways: An approach combining least-cost path modelling and machine learning

Seabird migration is driven by general wind circulation and productive ocean regions. As a result, bird migration takes place along distinct corridors or "flyways” that have evolved by earth’s large-scale atmospheric circulation patterns. These flyways form a link between climate and bird migration, and by simulating their pattern we might better understand the present corridor and predict the potential future impacts of climate change. However, few studies have focused on modelling flyways (especially for multiple bird strategies, populations, seasons, species and oceans), with most of them simulating trajectories of individual birds.

We use climatic data in combination with a least-cost-path modelling approach to simulate and describe multiple seabird flyways. By combining bird tracking data and machine learning, we are able to infer whether the flyways used by the birds optimise time and/or energy. We focussed on five seabird flyways of arctic terns and sooty shearwaters, both spring and autumn migration either over the Atlantic or the Pacific Ocean. We will show that a bird's effort is influenced by tailwinds, crosswinds and food availability, and we use this to calculate how close to the theoretical optimal migration (time- or energy-minimising) these birds actually fly. Our findings show that it is possible to recreate observed flyways using environmental data and that these simulations can generate predictions about the effect of future climate change.