Quantifying habitat use and preferences of pelagic seabirds using individual movement data: a review

• Published in: Marine ecology. Progress series (Halstenbek) 2009-09, Vol.391, p.165-182
• Main Authors: Wakefield, Ewan D., Phillips, Richard A., Matthiopoulos, Jason
• Format: Article
• Language: English
• Subjects: Animals; Aquatic habitats; Breeding; Ecological modeling; Foraging; Habitat preferences; Marine; Marine ecology; Oceans; Sea birds; Seas; THEME SECTION: Spatiotemporal dynamics of seabirds in the marine environment
• ISSN: 0171-8630; 1616-1599
• DOI: 10.3354/meps08203

Colonial seabirds are relatively easy to observe, count, measure and manipulate, and consequently have long been used as models for testing ecological hypotheses. A combination of animal tracking and satellite imagery has the potential to greatly inform such efforts, by allowing seabird–environment interactions to be observed remotely. We review how this can be achieved by applying innovative statistical techniques to quantify habitat use and preferences. Seabird movements are now observable at scales of meters using GPS loggers, and up to several years using lightbased geolocation, while satellite remote sensing systems, at resolutions of km, are capable of characterizing the millions of km² of habitat that are accessible to seabirds. Physical forcing and biological processes result in a hierarchical, patchy distribution of prey. Hence, analyses of seabird movements should be conducted at appropriate scales. Variation in habitat accessibility should also be considered: this declines with distance from the colony during the breeding season, when seabirds are central place foragers, and may be limited in the nonbreeding period by migration corridors that are defined by wind patterns. Intraspecific competition can further modify spatial usage, leading to spatial segregation of birds foraging from different colonies. We recommend that spatial usage be modeled as a function of habitat preference, accessibility and, potentially, competition. At the population level, this is currently best achieved using an empirical approach (e.g. using mixed-effects generalized additive models). At the individual level, more mechanistic models (e.g. state–space models) are more appropriate and have the advantage of modeling location errors explicitly.