Field work was conducted on the island of Heligoland (54∘11′ N, 7∘55′ E) in the southeastern North Sea (Fig. 1). A total of 14 adult northern gannets that were either incubating or rearing chicks were caught on 12–13 May, 17–18 June, or 22–23 July 2015. All birds were equipped with data loggers. A total of 10 gannets each received a Bird Solar GPS logger (e-obs GmbH, Munich, Germany) and the other four birds were equipped with both a CatLog-S GPS logger (Catnip Technologies, Hong Kong SAR, China) and a precision temperature–depth (PTD) logger (Earth and Ocean Technologies, Kiel, Germany). All loggers were attached to the base of the four central tail feathers using TESA^® tape (Beiersdorf AG GmbH, Hamburg, Germany; Fig. 2). Data obtained from these loggers covered durations of 0.4–10.9 weeks.

The total masses of the attached devices (including sealing, base plate, and tape) were about 48 g (Bird Solar) and 64 g (CatLog-S plus PTD), representing 1.5 and 1.9 %, respectively, of the mean gannet body mass of 3286 g (Wanless and Okill, 1994). This is well below the potential threshold of 3 % (Phillips et al., 2003; see Vandenabeele et al., 2012). Although attachments to the tail may have a negative influence on flight behaviour (Vandenabeele et al., 2014), most pairs successfully incubated their eggs and/or raised their chicks, similar to non-handled nests, with no visible effects on bird behaviour.

These devices recorded date, time, and position (latitude, longitude) and were set at an interval of 5 min. Dimensions varied slightly according to battery type but were about 50×35×8 mm. The plate was encased by a heatable plastic housing. Data were retrieved by recapturing the bird and downloading data from the device.

The northern gannet is the largest seabird species in the North Atlantic. It has a body mass of 2.3–3.6 kg and breeds in colonies of up to several tens of thousands of pairs. Northern gannets spend their entire life at sea, except for breeding on land. They usually start breeding at 5–6 years old, and may live to 20 years or older. They lay one egg that is incubated for 6 weeks, followed by a chick-rearing period of about 13 weeks. Only one adult of the pair is usually at the nest at any one time during incubation or chick guarding, while the other is at sea (Nelson, 2002; Bauer et al., 2005). Apart from short flights to collect nesting material or due to disturbance/interactions at the nest site, gannets carry out foraging trips to collect food for themselves and their offspring. They usually forage using so-called plunge dives, which are initiated when flying (and searching) at a few to several tens of metres above the sea surface (Nelson, 2002; Garthe et al., 2014). Two different dive types can be distinguished in this species (Garthe et al., 2000). U-shaped dives occur when the birds remain at a largely constant depth for a period after plunging into the water, with little vertical movement, before returning to the sea surface. In contrast, V-shaped dives are usually short and shallow, with the ascent almost immediately following the descent.

Northern gannets have recently been studied intensively by satellite telemetry and data loggers in various places (e.g. Hamer et al., 2001; Pettex et al., 2012; Wakefield et al., 2013), thus allowing comparisons among regions and populations.

Figure 3a and b show the flight patterns of two adult northern gannets that were typical of 13 of the 14 individuals tracked in 2015. Birds flew in all directions from the island to search for food, but most flights targeted areas to the (N)NW (north–northwest) of Heligoland. Foraging trips (defined in this paper as absences from the nest site of at least 20 min and of at least 2.0 km direct distance) were remarkably variable in duration and distance; most trips lasted 1–15 h and extended from 3 to 80 km from the breeding colony on Heligoland. One individual's behaviour differed from that of the other gannets by repeatedly flying far north to forage in the Skagerrak (Fig. 3c). These long-distance foraging trips were almost identical in their structures (n=3; duration of 44.3–59.3 h, most distant location 375–388 km away, total distance of 971–1019 km flown) and were interspersed with “normal” foraging trips into the German Bight.

Dives of gannets breeding on Heligoland and foraging in the (south) eastern North Sea were generally shallow, with more than half of the dives only reaching depths of 1–3 m (Fig. 4). The maximum dive depth was 11.4 m, and the median dive depth was 2.2 m (n=4 individuals, n=2577 dives). Most dives were V-shaped, though the measuring interval of 3 s did not allow a precise determination of the proportions of U- and V-shaped dives (see Garthe et al., 2000). The measuring interval of 3 s might also mask the deepest parts of some dives and may thus underestimate dive depth in general. We therefore re-analysed a random sample of 100 dives from Garthe et al. (2014), where gannets exhibited a similar high percentage of V-shaped dives, using both 1 and 3 s intervals (10 individuals, 10 dives each). Scaling down to 3 s missed 10 % of the dives, while the median-detected dive depth was only slightly smaller (4.3 vs. 4.5 m). These subtle differences demonstrate the validity of 3 s measuring intervals to determine the dive-depth pattern.

These dives were shallow compared with previous studies conducted in the northwestern North Sea (Lewis et al., 2002), the English Channel (Grémillet et al., 2006), the northwest Atlantic (Garthe et al., 2000), and the Gulf of Saint Lawrence, Canada (Garthe et al., 2007). Although the sample sizes of individuals are small, this does not hold true for the number of days the birds were tagged and the number of dives. Dive depths recorded from Heligoland gannets in 2015 remained much shallower even when subsampling small data sets from a large database from eastern Canada (S. Garthe et al., unpublished data).

Gannets showed a clear diurnal rhythm in their diving activity (Fig. 5). Dives were almost completely restricted to the daylight period, with the remaining dives occurring around dawn and dusk. No dives were recorded between 22:09 and 05:21 Central European Summer Time (CEST). This pattern fits well previous studies (Garthe et al., 2000, 2003).

Dive positions were analysed for the fixed habitat parameters including distance to colony (Heligoland), water depth, and distance to nearest land (except for Heligoland; Fig. 6). Almost two-thirds of the dives were carried out at a distance of less than 50 km from the colony, with proportions declining further away from the colony. However, at the largest distances, proportions of dives increased again, strongly indicating that gannets may have specifically targeted such distant foraging areas (see also Sect. 3.5, Fig. 9). As related to water depth, gannets from Heligoland were diving most often in waters of 20–40 m depths, less often in shallower, and rarely in deeper waters (Fig. 6). For foraging, gannets mostly stayed away from the coast, with the highest proportions at a distance of 40–60 km. This pattern differs completely from studies in eastern Canada where gannets were found to concentrate their diving efforts on the coastal zone (Garthe et al., 2007). Both Heligoland (located approximately 43 km north of the East Frisian Islands) and Funk Island (Newfoundland, Canada; located approximately 50 km away from the coast) have a similar placement and, in consequence, the location of the colonies cannot explain the observed difference in coastal focus. However, the near-coastal waters in the Canadian study sites are characterised by much more marine conditions and larger water depths compared to the Wadden Sea coast with extended shallow waters in the German Bight.

Two Gaussian linear mixed models were used to analyse the impact of the habitat parameters including distance to colony, water depth, and distance to nearest land on (1) the dive depth and (2) the dive duration of the tagged birds (using R version 3.3.2; R Development Core Team, 2016; package “lme4”, function “lmer” by Bates et al., 2015). Both response variables were log transformed to approach normality. The three habitat parameters were used as numeric explanatory variables. Dive ID nested within bird ID was taken as a random factor to avoid pseudo-replication due to multiple measurements per bird. As the function lmer does not provide p values, we further used the function “drop1” to find the relevant habitat parameter explaining the variance of dive depth and/or dive duration. This function tests every term in the model as if it was the last entering the model. In turn, every term in the model is omitted, and the reduced model is then (automatically) compared to the full model by a likelihood ratio test under 1 degree of freedom (i.e. a so-called marginal frequentist F test; Korner-Nievergelt et al., 2015). Though the sample size of individuals was low, the temporal coverage (12–18 days) and the number of dives per individual (381–773) were high. In this data set, dive depth could not be explained statistically by any of the three fixed habitat variables, while dive duration could be explained by water depth (Table 1). It is to be expected that larger data sets that will be collected in the future may exhibit more significant relationships to these and other habitat parameters.

The altitudes of flying birds are important in relation to their migratory movements, prey-searching behaviour, and potential overlap with technical installations at sea.

The Bird Solar GPS loggers calculated height above the ellipsoid when connecting with satellites during positional fixes, and altitude measurements were therefore corrected for geoid height (39.1 m at colony location). Altitude estimates are improved when connection time to the satellite is increased (e.g. Corman and Garthe, 2014), and we therefore used pulses of GPS fixes over 11–15 s and analysed the last and assumed best altitude measurement from each pulse. Figure 7 shows non-smoothed altitude measurements for one foraging trip of 22.7 h. Colony attendance was derived from positional fixes and known nest position; on-water periods were determined from ground speed (< 3 km h-1) and positional fixes. Although values fluctuated slightly even for fixed places, such as the nest site in the colony and the sea surface, the measurements appeared reasonable and showed that most flight activity occurred at an altitude between the sea surface and 40 m, with maximum values in this study for when birds were commuting to/from the colony. Measurements in the colony and on water can be used to calibrate altitude measurements because of their relatively well-known heights.

Flight heights of gannets have also been determined by radar measurements (e.g. Krijgsveld et al., 2011), visual observations (e.g. Johnston et al., 2014), and pressure sensors (Garthe et al., 2014; Cleasby et al., 2015). Overall, flight heights of gannets tend to be low, with relatively few flights above 50 m and very few recorded above 100 m, though no comprehensive analysis has yet been published.

Animal movements can be tracked using motion sensors. Many data loggers contain accelerometers that ideally cover all three axes (x, y, z), and frequent measurements allow behavioural differentiation at a fine scale (e.g. Sakamoto et al., 2009).

Figure 8 shows an example of accelerometer measurements of a northern gannet at a 0.25 min interval over 24 h. Recordings start when the bird is on its nest, with very little activity, obviously sleeping. After about 3 h, the bird remains on its nest but its activity increases, coinciding with dawn. A few hours later, the bird leaves the nest and flies off, followed by a period of about 10 h of mostly flying, interrupted by a few shorter swimming periods. Towards the end of the recording period, the bird settles down on the sea surface and remains floating there overnight (Fig. 8). Such information is important in many ways. It may help identifying the relevance of certain sea areas, i.e. whether areas are used for foraging or just for resting, or for long(er)-distance movements. Quantifying birds' activities is a well-established tool to measure energy expenditure. Such energy budgets may, e.g. help unravelling seabird movement strategies as has been shown by Garthe et al. (2012) for northern gannets wintering in different regions of the northeast Atlantic. Analyses of these kinds will be done for more birds in a separate study.

To determine the relevance of certain sea areas and to improve our understanding of the flight patterns of the birds, it is necessary to know when and where the birds are feeding. Because gannets almost always obtain food by plunge diving, observing dives provides a good proxy for determining feeding areas. Figure 9 shows the flight tracks and dive locations of the gannet that flew repeatedly towards the Skagerrak. It shows that the gannet was foraging intensively in the Skagerrak, while longer passages on outbound and inbound flights were long-distance flights without much foraging activity.

The ability to track seabird movements at small spatial and temporal scales makes it possible to study the potential impacts of human activities at sea comprehensively. A total of 12 offshore wind farms have been built and became operational in the German Bight between 2008 and November 2016, and a further 5 are currently under construction. Another 15 wind farms have been given consent, and several tens more have been applied for. The impact of wind farms on seabirds, which is currently a hot topic in conservation biology and environmental policy (e.g. Furness et al., 2013; Masden et al., 2015), can thus be studied comprehensively in German North Sea waters. In 2014, gannets were tracked near existing wind farms for the first time. All three individuals largely avoided the wind farm area north of Heligoland (Garthe et al., 2017).

The flight tracks of the gannets shown in Fig. 3 were projected on top of the wind farms that were operational or under construction during the tracking period. The three gannets mostly stayed away from the wind farms and passed around them much more frequently than flying through them (Fig. 10). Wind farms further from Heligoland were not entered, but gannets visited the areas around them. Focusing on the three wind farms north of Heligoland, 5 of the 14 gannets tracked in 2015 did not enter them, 4 only flew into the wind farms once, while 4 visited them occasionally and 1 frequently.