Introduction
For animals operating in aerial and aquatic environments, movement costs and capacities are profoundly affected by air and water currents respectively1. This drives widespread and large-scale patterns of animal movement; with birds selecting tailwinds or updrafts as they migrate across ocean basins or between land masses2, and fish minimising counter-flows as they migrate upstream3. In fish, current selection is also a key driver of habitat use outside periods of travel4. Indeed, there is a rich history of research on fish space-use in relation to flow characteristics in rivers dating back to the 1960s5, which has demonstrated that water velocity, and in some cases velocity shear, determine abundance, habitat use (both within and between species), and interspecific competition in some species4. In contrast, there is little to no research on how local airflows affect the distribution of birds outside travelling, including the distribution of breeding colonies.
Reproductive success is closely linked to the physical characteristics of breeding sites in many taxa6,7 which can feed into broader species distribution models8. In colonial animals, breeding sites can represent the nexus of reproductive activity for tens of thousands of individuals9. There is therefore a clear need to establish what drives colony location in these taxa, in order to identify the availability of breeding habitat, and predict how areas differ in quality, now and in the future10-16. Over 95% of seabirds are colonial breeders17. Seabirds are also more at risk than other comparable groups of birds, with widespread decline in populations due to commercial fisheries, pollution, habitat change and the introduction of invasive predators18. In some cases this has led to entire breeding colonies being lost19,20. Here, conservation practitioners need to know where to focus restoration efforts e.g. by decoy deployment and acoustic attraction to re-seed breeding activity21. This is crucial given that there will always be a fitness cost associated with breeding in sub-optimal habitat22. However, while a wide range of studies have analysed breeding site characteristics in seabirds6,7,9,23, and compared them with available habitat24, we are unaware of any that have successfully applied predictions from one site to another (cf .25).
The tendency of seabirds to breed on offshore islands and/ or coastal cliffs has been attributed to the need to reduce exposure to terrestrial predators and be close to feeding areas9,26. Nonetheless, for cliff nesting species, it is clear that not all cliffs are equal, as colonies tend to be clumped, with great swathes of cliff habitat left empty7. Indeed, cliffs should vary in their accessibility to terrestrial predators (primarily through variation in slope angle), as well as the availability of suitable breeding ledges, with species varying in their need for different ledge characteristics according to their body size and nest building habit6,27.
There are also compelling reasons why flow characteristics should affect breeding habitat preferences, particularly for groups such as seabirds, which are exposed to strong flows. Wind can affect the risk of eggs/ birds being displaced from the nest28, as well as influencing exposure to rain (particularly in cliff nesting species) and heat stress (through evaporative heat loss), both of which can cause mortality 28,29,30. Wind also has a strong influence on flight capacity. In common guillemots (Uria aalge ) and razorbills (Alca torda ), 60% of attempts to land at their cliff nests were found to fail in a strong breeze31, suggesting there are advantages to breeding on more sheltered cliffs. Indeed, frigatebirds (Fregata magnificens ) nest in relatively wind still areas, despite the associated reduction in ability of birds to lose heat, which may reflect the difficulties that adults would experience in remaining on the nest and operating close to it in high winds due to their low weight and wing loading32. The importance of being able to maintain flight control close to the nest suggests that habitat selection could be influenced by several airflow characteristics, including the strength of the horizontal and vertical components, as well as the turbulence.
Despite the potential importance of airflows for these animals that breed in exposed locations, there is a complete lack of information on the flow characteristics associated with colony presence and absence (though see33). This is likely due to the difficulties of quantifying wind over complex, often steep terrain. It may also reflect our inability to see flow characteristics, in contrast to rivers where this can be evident from surface characteristics. Where the impact of wind has been assessed by proxy, wind fetch was found to have contrasting effects in the presence/absence of colonies of three Pygoscelis penguin species34, while colony aspect was not significant for species of the auks family23,31. Nonetheless, aspect may be a poor proxy for the precise wind conditions experienced at colonies, as airflow characteristics will be modified by the particular topography of the surrounding area. As a result, two cliffs with the same aspect and prevailing wind conditions can experience very different flow regimes31.
We use computational fluid dynamics (CFD) to provide the first assessment of whether and how local airflow conditions predicts the distribution of seabirds, taking colonies of common guillemots (Uria aalge ) breeding on Skomer island, UK, as our study system. We estimate a number of airflow characteristics, including the magnitude of the wind, the horizontal and vertical wind components, air pressure (as a predictor of exposure) and finally gustiness and turbulence, which may affect flight control close to the breeding cliffs31. Our specific objectives were to: (i) assess whether airflows associated with the prevailing wind direction predict breeding site selection (patterns of presence and absence), and habitat quality (colony density), (ii) quantify the airflow conditions that birds will be exposed to with changes in wind direction and (iii) test our model of habitat selection by predicting colony presence and absence on a neighbouring island. This test that is considerably stronger than standard cross-validation, but rarely performed25. Overall, our approach should provide insight into the conditions birds select and avoid in the prevailing wind, and the “penalty” they suffer in terms of the adverse conditions they are exposed to if the wind direction changes, either over the short-term, or as part of larger scale climatic shifts35,36.