Introduction

Seabirds are important components of marine ecosystems, yet their ecology and spatial distribution during the non-breeding period is poorly known. Bird movements outside the breeding season can cover thousands of square km (e.g. Phillips et al. 2005; Egevang et al. 2010; Dunn et al. 2011; Mosbech et al. 2012), and the individuals experience a multitude of different habitats, feeding conditions and potential threats. The marine environment, in which the seabirds spend their whole life cycle, is changing due to over-harvesting, pollution, habitat modifications and global climate change (e.g. Halpern et al. 2008). Many seabird species in the North Atlantic have experienced large reductions in population size and breeding success during the last decades, probably caused by reduced prey abundance caused by climate alterations and overfishing (e.g. Frederiksen et al. 2004, 2007, 2008; Barrett et al. 2006).

One of the seabird species that have experienced the largest reductions in breeding numbers in Norwegian waters is the common guillemot (Uria aalge) whose numbers in most breeding colonies have declined with more than 95 % (e.g. Barrett et al. 2006), and the species is now classified as critically endangered (CR) on the Norwegian Red List (Kålås et al. 2010). The causes for this decline are partly unknown and probably related to both food availability and/or climate changes in the marine environment (e.g. Frederiksen 2010). However, the common guillemot is also one of the species that is most frequently affected by oil spills (e.g. Cadiou et al. 2004). Due to the status of the species, it is now critically important to have better knowledge on inter-breeding movements and wintering sites. This is particularly so because (1) while its breeding biology is well known, wintering ecology and wintering sites are poorly known, (2) climate changes are expected to be most pronounced in the Arctic (IPCC 2007), (3) mortality due to oil spills mostly occur during winter. Harris et al. (2007) found that most mortality of adult seabirds occurred outside the breeding season, and hence, knowing their wintering areas is essential in order to understand factors regulating the populations. This is particularly important for species that are declining and/or threatened.

Most information about inter-breeding movements and distribution of wintering seabirds are from ship-based surveys or from ringed birds. For seabirds, however, ring recoveries are biased by several factors, partly because many recoveries are from where carcasses have washed ashore and not necessarily where they died. In addition, only a small fraction of ringed birds are recovered. In Norway, only 2.4 % of the ringed common guillemots have been recovered (Bakken et al. 2003), indicating a huge effort, and potentially much disturbance in the breeding colonies, for a low return rate of data. During recent years, miniature global location sensors (GLS loggers or geolocators) that enable the movements of seabirds to be tracked over extended periods (up to several years) have been developed (e.g. Phillips et al. 2004; Bost et al. 2009; Dunn et al. 2011).

The main aim of this study was to map the inter-breeding movements of common guillemots breeding at a colony off the coast of Central Norway using GLS loggers. In particular, we aimed at identifying the key areas used during the non-breeding period when common guillemots may potentially be especially vulnerable to anthropogenic impacts. In order to evaluate possible seasonal differences in movement patterns, we also calculated the mean speed of the movements in 5-day periods.

Materials and methods

Logger deployment and retrieval

Field work was conducted at Sklinna Nature Reserve (65°13′N 10°58′E), a small archipelago 20 km off the mainland coast of Central Norway. On 30 June 2009, 25 chick-rearing adults (9 males, 14 females and 2 individuals of unknown sex) were caught using a nose pole. A GLS logger (Lotek LAT 2500, 128 Kb memory, no pressure sensor, 8 × 35 mm, mass of 3.6 g in air, 0.4  % of the body mass of the equipped birds) attached to a plastic leg ring was attached the tarsus on each of the birds. Body mass (accuracy ±10 g when BM <1,000 g and ±20 g when BM >1,000 g) was measured using Pesola spring balances. In order to sex the birds, 25 μl of blood was collected in a capillary tube after puncturing one of the veins on the leg or the web and immediately suspended in 1 ml of “Queen’s lysis buffer” (Seutin et al. 1991). During 1–7 July 2010, 10 of the loggers were retrieved (4 males, 5 females and 1 of unknown sex).

Sexing

DNA was extracted from the blood using GeneMole® (Mole genetics, www.molegenetics.com). This instrument performs an automatic DNA extraction on a 1:10 solution of blood in a lysis buffer and PBS (pH = 7.4), see http://www.molecookbook.com//index.php?qlink=84. After DNA extraction, sexing was performed using the methods described by Bantock et al. (2008).

Data analysis

The GLS loggers basically record time, temperature and light intensity. Geographical positions are then estimated from changes in light intensity over time. Sunset and sunrise are estimated from thresholds in the light curves, and this enables estimations of latitude through daylength and longitude from the time of midday with respect to Greenwich Mean Time (e.g. Phillips et al. 2004).

Data from the loggers were processed in LAT Viewer Studio (Lotek wireless, Newmarket, Ontario) using the template fit option (Ekstrom 2004). Latitudes are unreliable around the vernal and autumnal equinox, and data for 18 September–4 October and 1–24 March were therefore removed. Latitudes were also unobtainable for periods when birds were in regions of constant darkness. Thus, in this study, all latitudinal data from 5 October–28 February were refined using the loggers’ records of sea surface temperature (SST), which were compared with SST data from NASA’s Terra and Aqua satellites and two polar-orbiting TIROS satellites using the algorithms integrated in LAT Viewer Studio. Satellite data were downloaded from http://whiteshark.stanford.edu/public/lotek_sst. The algorithm in LAT Viewer Studio was run with the start position manually set to 68°N, 10°E, as all equipped birds moved north the months following the breeding season. The loggers’ SST data were, additionally, compared (visually) with a map of sea surface temperatures from along the Norwegian coast and the Barents Sea during January 2010 (downloaded from http://modis.gsfc.nasa.gov/).

Positions obtained from GLS loggers have an inherent average error of approximately 186 km (Phillips et al. 2004). In order to reduce the influence of outliers when calculating distances and allocating positions to areas of interest, positions were smoothed using a 3-position moving average based on spherical trigonometry (e.g. Frederiksen et al. 2012). Distances between successive smoothed (and unsmoothed) positions were calculated. We estimated kernel utilization distributions over all 10 individuals for autumn (August–September), winter (October-February) and spring (March–April). From these utilization distributions, 25 and 50 % isopleths were derived. This removed any remaining outliers despite all previously mentioned filtering and calibrations. Kernel densities (Gaussian, bivariate normal) were estimated using the geospatial modelling environment (GME) extension (http://www.spatialecology.com/gme/index.htm). Mapping was done in ArcMap 10 (ESRI, Redlands, California).

Kernel utilization distributions are smoothed using the so-called bandwidth, which is represented as the radius of a circle within which points are counted around each cell. It is therefore important to choose the right bandwidth, to avoid over- or under-smoothing (GME manual downloaded from the address above). Although least-squares cross-validation (LSCV) is recommended as the bandwidth selection method (Seaman et al. 1999), newer methods such as the plug-in and solve-the-equation (STE) bandwidth methods are preferred in the statistical literature (Jones et al. 1996). Using simulated point patterns, Gitzen et al. (2006) found that although the relative differences were small, the plug-in and STE approaches provided good alternatives to LSCV. However, the choice of a bandwidth selection method may vary depending on the study goals, sample size and patterns of space use by the study species. Gitzen et al. (2006) recommended LSCV for clumped distributions, and plug-in and STE approaches for estimating smoother distributions. Given the extensive range use of our study species, the plug-in method was considered to be most appropriate.

To obtain an estimate of the bandwidth, we projected the longitudinal data into UTM Zone 33 (WGS 1984) to render X/Y coordinates on a metric scale. We then increased the inaccuracy in 5 km steps from 0 to 200 km. For each step, the X and Y coordinates of the original data set were adjusted by adding randomly derived deviations from each coordinate using a normal distribution with accuracy as standard deviation around a mean of zero. From this adjusted data set, unique distances between all pairs of positions were calculated, which was used as input to the bandwidth calculations. This process was iterated 100 times to estimate mean and range for each (accuracy) step. In addition, the predicted trend was plotted based on regressing bandwidth against the square of accuracy.

Assuming absolute accuracy (i.e. 0 km), a mean bandwidth of 11.2 km (accuracy range, 11.2–21.4 km) was obtained. We therefore used the rounded value of 10 km as a measure of bandwidth.

Swimming speed during the autumn migration was calculated for every 5-day period (numbered consecutively from 1 January so that e.g. 5–9 June is 5-day period 32, 10–14 June is 5-day period 33) from when the birds left the colony until they arrived at the autumn staging site.

Results

There were no differences in the body mass of the guillemots at deployment and removal of the GLS loggers approximately 1 year after, indicating that the attachment of the loggers did not have any negative influence on the equipped individuals; Wilcoxon signed-ranks test, males Z = 1.07, p = 0.285, df = 3; females Z = 1.10, p = 0.273, df = 3 (Table 1). In 2010 and 2011, at least 18 of the 25 birds equipped with loggers were observed (8 individuals, no logger retrieval) or re-caught (10 individuals), giving a total return rate of 72 %. Although a small sample size, this is lower than the annual survival (return rate) of adults not equipped with geolocators at the same site (mean adult survival = 91.8 %, S.H. Lorentsen unpublished data).

Table 1 Summary of data on 10 common guillemots equipped with GLS loggers. For information on dates see text. Autumn (August–September), Winter (October–February) and Spring (March–April) indicates main staging area during those periods
Full size table

Eight of 10 birds (80 %) moved directly to the Barents Sea after breeding (Table 1). The two remaining individuals also moved northwards, but did not enter the Barents Sea. All individuals started to move northwards shortly after the chicks had left for the sea, which at Sklinna normally occurs between 10 and 25 July. The mean swimming speed varied much between the individuals but increased slightly from 0.5 to 2 km/h during the first month (until the end of August, 5-day period 46 = 21–25 August, Fig. 1). The swimming speed was then quite low for a few weeks (<0.5 km/h) before it increased steeply during the first half of September (5–20 September, 5-day period 49–51).

Fig. 1
figure 1

Daily swimming speed (mean ± SD, calculated by diving the daily distance moved by 24) by common guillemots from the end of the breeding period (late July) until late autumn (end of September)

Full size image

During the autumn (August–September), the area from central Norway to the Barents Sea was used (Fig. 2). However, all individuals that entered the Barents Sea did so in late August–early September. The individuals that did not enter the Barents Sea stayed in the area outside Central Norway and Lofoten, whereas the individuals entering the Barents Sea used the central parts as far east as 45°E.

Fig. 2
figure 2

GLS positions (yellow dots), kernel density surface plots (red and yellow) and isopleths (50 % = red lines, 25 % = green lines) for common guillemots during autumn (August–September, upper panel), winter (October–February, mid panel) and spring (March–April, lower panel). Geographical names used in the text are indicated in the lower panel

Full size image

Three of the individuals that moved to the Barents Sea during autumn moved southward during late autumn/early winter and stayed outside Lofoten/Central Norway. Half of the individuals stayed in the Barents Sea during winter (mid-October–end of February). In the Barents Sea, the main wintering areas judged from the kernel isopleths were outside the coast of Finnmark and the eastern part of the Kola Peninsula. The other half of the individuals spent the winter in the area from off the coast of Lofoten in Northern Norway to Central Norway (Fig. 2). Most of the individuals had returned to the breeding colony in March. While the northwards movement in autumn was quite directed and fast, changes in latitude during the other time periods suggested much lower movements and also showed a greater variance in latitude (Fig. 3).

Fig. 3
figure 3

Latitudinal (mean ± SD) changes in average distribution between 5-day periods (40 = 20–24 July, 73 = 26–31 December) for common guillemots from Sklinna, Central Norway. The reference line indicates the latitude of the study colony. Time periods around vernal and autumnal equinoxes, and the breeding season, are filtered out

Full size image

Discussion

Individuals equipped with geolocators had a lower return rate than birds without geolocators. There were however no differences in the body mass of the equipped individuals at deployment and retrieval of the loggers suggesting that the loggers did not have any negative influence on the equipped individuals. The cause of the lower return rate of the equipped individuals is not obvious, but might be related to inter-colonial differences in breeding sites of the individuals caught for logger deployment. The common guillemots at Sklinna nests among huge rocks, and most of the birds that were equipped were caught below a rock where they had few escape possibilities, whereas the rest were caught below rocks with more escape possibilities. Unfortunately, we did not take notes of where we caught the actual individuals within the colony as we wanted to make the catching events as quick as possible in order to reduce the disturbance. Thus, birds that were caught below rocks with several exits were much more difficult to re-trap, which might explain the differences in return rates between individuals with and without loggers. Hence, many of the birds with loggers that we were not able to catch were observed close to breeding sites with escape possibilities.

Most of the common guillemots (80 %) in this study moved directly from Sklinna to the Barents Sea after the breeding period. The remaining two individuals (20 %) also moved northwards but stopped in the Lofoten area. The movements were directional and with an increasing speed as the journey progressed, until late August when the daily movements dropped to <0.5 km/h before it increased from mid-late September. The drop in movements in late August–September coincides with the moult of their secondaries (Cramp 1985). Thus, the individuals that entered the Barents Sea in late August–early September probably were in, or had completed, their moult when they arrived.

Five of the eight individuals that entered the Barents Sea stayed there during the whole winter, whereas three moved out of the Barents Sea and wintered along the Norwegian coast from Lofoten to Central Norway. All individuals were back to the waters outside the breeding colony in March, about 2 months before egg laying (S.H. Lorentsen unpublished data).

The food of wintering common guillemots in the Barents Sea is not known. In the breeding season, the main food for common guillemots breeding at Hornøya in the eastern Barents Sea was capelin (Mallotus villosus), sandeel (Ammodytes sp.) and herring (Clupea harengus) (Barrett 2002; Bugge et al. 2011). It is strongly suggested that the food choice of common guillemots (and other seabirds) wintering in the Barents Sea are studied in order to understand its role in the marine ecosystem and to be able to access possible anthropogenic threats.

The GLS loggers record time, temperature and light intensity, and geographical positions are then estimated from changes in light intensity over time. However, when the birds carrying the loggers are in regions of constant darkness, latitudes are unobtainable. Light levels are low in the Barents Sea (Bear Island) from 7th November until 4th of February as the sun is below the horizon all the day. We therefore compared the logger’s records of sea surface temperature (SST) for the period 5 October–28 February with SST data from satellites. Due to cloud cover satellite, SST data were not available for all days in this period, but in general, most of the geographical positions were adjusted northwards to the Barents Sea after this procedure. Also, a visual comparison of the loggers’ SST data with a map of sea surface temperatures from along the Norwegian coast and the Barents Sea during January 2010 showed a close conformity of logger and satellite SST data.

Due to the inherited errors in the positions obtained from the GLS loggers (on average 186 km, Phillips et al. 2004), we used a 3-position moving average based on spherical trigonometry to reduce the influence of outliers when calculating distances and geographical positions. Still, we experienced some outliers, especially during spring and autumn (cf. Fig. 2). Although these outliers could be removed by manual inspection, we included them in the kernel analyses that, consequently, might be too wide especially outside the 50 % contour. Kernel density estimation does also not take into account topographic features such as the coast line. Thus, although the distributions may also to a certain extent encompass inland areas, we would like to stress that this study was performed, and gives insight into seasonal inter-breeding movements, on a rather large spatial scale. Also, the number of winter plots in the southern area might be misleading compared with the number of birds that wintered in the Barents Sea and outside the Barents Sea. This is probably related to the SST-based correction of plots for birds in the Barents Sea as satellite-based SST data could not be obtained for many days during winter due to extensive cloud cover. Thus, proportionally more of the plots from birds in the Barents Sea had to be removed as they could not be corrected than plots from birds wintering farther south.

The Barents Sea is known to be a major wintering area for common guillemots of unknown origin (e.g. Barrett and Golovkin 2000 and references therein). This study shows that this includes birds from central Norway. Also puffins (Fratercula arctica) (Anker-Nilssen and Aarvak 2009) and black-legged kittiwakes (Rissa tridactyla) (B. Moe pers. comm.) winters in the Barents Sea. This certainly has implications for the management authorities. For instance, both Norwegian and Russian authorities now open the area for oil exploration, and an accident causing a major oil spill could have huge impacts on moulting and wintering common guillemots. Much of the area in the Barents Sea–Lofoten area is classified as vulnerable with respect to specific environmental pressures such as oil pollution and other anthropogenic factors (e.g. von Quillfeldt et al. 2009).

This study suggests that the Barents Sea and the Norwegian coast from Central Norway and northwards are important staging and wintering grounds for common guillemots breeding in Central Norway. However, analyses of ring recoveries show that common guillemots from Central and Northern Norway also migrates south to the Skagerrak area during winter (Bakken et al. 2003). This might suggest variation (individual and/or annual) in inter-breeding movements and, thus, call for multi-year geolocator studies at a number of breeding colonies. We strongly hope that such studies are initiated in the near future, and that they also incorporate studies of the food choice of the common guillemot during the inter-breeding period.