Using an algorithmic model to reveal individually variable movement decisions in a wintering sea duck.

1. Many migratory birds are assumed to remain fairly stationary during winter. However, recent research indicates that mid-winter movements are evident in a variety of bird species, and the factors causing individuals to move are poorly understood. 2. We examined the winter movements of 95 individual king eiders (Somateria spectabilis, L.) tracked with satellite transmitters in the Bering Sea between 2002 and 2006 to explore whether environmental factors such as day length, location, sea ice, and habitat quality could explain the occurrence of winter movements longer than 50 km. 3. We used a novel algorithmic random forest model to assess the importance of variables predicting whether a bird remained or departed from a wintering site. 4. We found extremely high individual variability in winter movement decisions by king eiders, and the individual bird was the most important variable followed by location, date, and sea ice concentration. 5. We conclude that individual strategies exist that interact with environmental conditions to form multiple movement patterns. 6. While a minor proportion of winter movements may be forced by environmental conditions, we propose that many winter movements may be of an exploratory nature where individuals aim to acquire information about alternative wintering sites that may enhance their survival probability at some point in time when environmental fluctuation renders their preferred wintering site unsuitable.