A behavioural approach to key area identification in seabirds for threat mitigation and spatial management

Identifying key areas of animal distribution using individual movement data is fundamental for conservation planning, threat mitigation, and spatial management. Methodologies which define these areas based on measures of high density and abundance may overlook spatial heterogeneity in behaviour-specific distributions. This is particularly relevant for behaviours that occur at lower densities but are associated with increased exposure to specific environmental threats. We used a dataset of 566 GPS tracked individuals and 14 colonies of a vulnerable species of seabird, the black-legged kittiwake (Rissa tridactyla), to compare two methods for delineating key areas. The first method applies kernel density estimates, based on 50% ('core area') utilisation distributions, to all movement data during an at-sea trip. This reflects a widely used density-based approach to identify high-use spatial areas. The second method incorporates hidden Markov modelling to classify movement data into three dominant behaviour states: resting, foraging, and transiting, to identify behaviour-specific high-use areas. We then compare population-level estimates of key areas based on each method using the BirdLife International Key Biodiversity Area framework. We also explore how the selection of an intermediate (70%) and home range (95%) utilisation distribution influences the capture of different behaviours.

We found that individual-level kernel density estimates based on core areas of all movement data fail to adequately capture the core distribution of transiting, a widespread and dispersed behaviour. Moreover, population-level estimates of key areas derived from transiting behaviour are significantly larger than those identified using all tracking data, suggesting that conventional methods likely underestimate exposure to threats encountered during transit. Conversely, key areas for resting and foraging behaviour are more spatially constrained than those derived from all movement data, implying that behaviour-specific analyses may improve the precision of conservation planning. Both individual and population-level key area estimates based on larger utilisation distributions (i.e. 75% and 95%) better capture the distribution of transiting behaviour as these larger distributions probabilistically encompass a greater fraction of observed movement trajectories.

These results highlight the importance of labelling movement data by behavioural state to enhance the utility of GPS data for conservation applications. By incorporating behavioural state differentiation into spatial analyses, threat exposure assessments can be refined to focus conservation resources more effectively. Furthermore, this approach has direct implications for environmental impact assessments, particularly in the context of expanding marine industries such as offshore renewable energy developments.

The online version contains supplementary material available at 10.1186/s40317-025-00427-z.