Foraging-trip analysis

Foraging trips were defined as all trips targeting the open sea, the mainland, or other areas, but which clearly led away from the colony, as confirmed by spatial analyses using ArcGIS 10.0 (ESRI). No signs of foraging activities were detected by visual observations near the colony, neither in the dunes nor on the beaches. All statistical analyses were performed using the open-source software R 3.0.0 [23]. The significance level was set at p < 0.05.

The foraging range was defined as the direct distance between the nest and the most distant point of a foraging trip. Foraging range and total distance flown were calculated using the R-package trip 1.1–15 [24]. The straightness index was derived from [25] (see also [26]), with (foraging range)×2×total distance flown^-1. This results in a maximum value of 1 for a completely straight flight from the colony to the most distant point and back, and gives increasingly smaller values for less directed, more tortuous flights.

We set up four different models to analyse the effects of trip destination (i.e. sea or inland) on: (1) foraging-trip duration; (2) foraging range; (3) total distance flown; and (4) straightness index (response variables). Mixed trips (i.e. trips targeting both sea and land) and trips to Spiekeroog island were excluded from the analysis, leaving 99 foraging trips for analysis. Explanatory variables were log-transformed when necessary. All models were calculated with a Gaussian error distribution. Linear mixed models (LMMs; [27]) were analysed using functions within the R packages lme4 [28] and lmerTest [29]. All LMMs included trip identity nested within the individual as a random term to account for pseudo-replication.

Human activities influencing the foraging behaviour of gulls (fishing, agriculture) may vary over a weekly cycle because of human activity schedules, and the relative uses of land and sea areas were therefore analysed for each day of the week. The proportions of positions over the mainland and over the sea were calculated for each individual on each weekday, and then averaged (including standard errors) over all eight individuals.