Seabirds and marine oil incidents: is it possible to predict the spatial distribution of pelagic seabirds?

• Published in: Journal of Applied Ecology
• Main Authors: Fauchald, Per, Erikstad, Kjell Einar, Systad, Geir Helge
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2002
• Subjects: Ecology; Barents Sea; uria
• Online Access: http://dx.doi.org/10.1046/j.1365-2664.2002.00717.x; https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1046%2Fj.1365-2664.2002.00717.x; https://besjournals.onlinelibrary.wiley.com/doi/pdf/10.1046/j.1365-2664.2002.00717.x

Summary Data on the spatial distribution of seabirds at sea is commonly used in risk assessments of the possible impact of oil spills. The validity of such assessments depends on the stability of the observed spatial pattern through time. In this study we explored the year‐to‐year predictability in the spatial distribution of guillemots ( Uria spp.) from a 9‐year data set covering an area of approximately 1000 × 600 km 2 in the Barents Sea from late January to early March. Spatial correlograms were used to elucidate the strength and scale of the spatial patterns within years, and the concordance of these patterns between years. Broad‐scale oceanographic features were used in linear regressions to model the spatial pattern in guillemot distribution for each year. The ability of these models to predict their spatial distribution in other years was then evaluated. The analyses revealed two nested levels of patchiness. The large‐scale pattern, with a characteristic scale of 300 km, had a weak ( R 2 = 0·06) but significant spatial predictability between years. The predictability increased marginally when the data set was divided into two time periods ( R 2 equal to 0·07 and 0·17, respectively). Nested within the large‐scale pattern, the analyses revealed a small‐scale level of patchiness with no significant spatial year‐to‐year predictability. The broad‐scale oceanographic variables could explain from 14% to 42% of the variance in the spatial distribution of guillemots within each year. The models were on average significantly better than using a random model when predicting other years. We found no relationships between the fit of the models and the ability to predict guillemot distribution in other years. There was a large variation in the parameter estimates between years, resulting in a large range in predicted values. This study illustrates some of the difficulties associated with predicting the spatial distribution of mobile and patchy organisms. Our results indicate that the use of restricted survey data in ...