1Movement patterns of predators should allow them to detect and respond to prey patches at different spatial scales, particularly through the adoption of area-restricted search (ARS) behaviour. Here we use fine-scale movement and activity data combined with first-passage time (FPT) analysis to examine the foraging strategy of northern gannets Morus bassanus in the western North Sea, and to test the following hypotheses: (i) birds adopt a hierarchical foraging strategy characterized by nested ARS behaviour; (ii) the locations and characteristics of ARS zones are strongly influenced by physical oceanography; (iii) the initiation of ARS behaviour is triggered by the detection and pursuit of prey; (iv) ARS behaviour is strongly linked to increased foraging effort, particularly within nested ARS areas.
2Birds on 13 of 15 foraging trips adopted ARS behaviour at a scale of 9·1 ± 1·9 km, and birds on 10 of these 13 trips adopted a second, nested ARS scale of 1·5 ± 0·8 km, supporting hypothesis 1 above. ARS zones were located 117 ± 55 km from the colony and over half were within 5 km of a tidal mixing front ~50 km offshore, supporting hypothesis 2 above.
3The initiation of ARS behaviour was usually followed after only a short time interval (typically ~5 min) by the commencement of diving. Gannets do not dive until after they have located prey, and so this pattern strongly suggests that ARS behaviour was triggered by prey detection, supporting hypothesis 3 above. However, ~33% of dives in mixed coastal water and 16% of dives in stratified water were not associated with any detectable ARS behaviour. Hence, while ARS behaviour resulted from the detection and pursuit of prey, encounters with prey species did not inevitably induce ARS behaviour.
4Following the initiation of ARS behaviour, dive rates were almost four times higher within ARS zones than elsewhere and almost three times higher in zones with nested ARS behaviour than in those without, supporting hypothesis 4 above and suggesting that the foraging success of birds was linked to their ability to match the hierarchical distribution of prey.
Foraging behaviour is an important part of the daily routines of many species and forms an essential link between prey availability and predator reproductive success. A central issue in this context is the variability in foraging behaviour across heterogenous landscapes used by predators (Nams 2005; Bailey & Thompson 2006). Prey often occur in patches and these are typically nested within larger-scale patches, forming nested patch hierarchies within the landscape (Wu & Loucks 1995; Fauchald & Erikstad 2002). Movement patterns of predators should therefore allow them to detect and respond to prey patches at different spatial scales (Fauchald 1999; Fritz, Saïd & Weimerskirch 2003). To understand the foraging behaviour of predators, it is thus crucial to determine their scales of interaction with the environment, particularly in nested scale systems where large-scale patterns tend to mask patterns at finer scales (Fauchald, Erikstad & Skarsfjord 2000).
In models of foraging behaviour in heterogenous landscapes, it is generally assumed that predators adjust their movement pattern in response to prey density by reducing their speed of travel or increasing turn rate once prey have been encountered or in response to environmental cues, adopting a so-called area-restricted search (ARS) behaviour (Kareiva & Odell 1987; Farnsworth & Beecham 1999). In a hierarchically structured system, predators should also display scale-dependent ARS patterns that match the spatial structure of their environment (Fauchald 1999). There is strong evidence of ARS behaviour in a wide range of species (Frair et al. 2005; Johnston, Westgate & Read. 2005), but fewer studies have examined nested ARS behaviour at different spatial scales (Pinaud & Weimerskich 2005; Fauchald & Tveraa 2006). Moreover, empirical evidence that ARS behaviour results from individuals detecting prey is derived mainly from studies of insects foraging over small-scale habitats (Wiens, Schooley & Weeks 1997). The conditions that lead longer-ranging predators to initiate ARS behaviour have received little attention, especially for vertebrates, and one recent study of a long-ranging predator found little association between ARS behaviour and prey encounter (Weimerskirch et al. 2007).
Marine prey resources occur at a wide range of spatial and temporal scales of distribution and abundance, reflecting scale-dependent interactions between ocean currents, bathymetry and other physical and biological processes that promote the growth and retention of plankton, leading to zones of enhanced productivity (Hunt & Schneider 1987; Sims & Quayle 1998). Several recent studies have found relationships between large-scale movements of marine predators and major oceanographic features such as continental shelf slopes and major frontal zones (Hyrenbach, Fernández & Anderson 2002; Pinaud & Weimerskich 2007), while some studies have also revealed nested search behaviour over scales of 10s to 100s kms in long-ranging pelagic predators with foraging ranges of 1000s kms (Pinaud & Weimerskich 2005; Fauchald & Tveraa 2006). Few studies, however, have examined fine-scale movements of species with narrower foraging ranges that do not include such large-scale oceanographic features. At this smaller scale, aggregation of predators is found in predictable regions of enhanced productivity such as bathymetric gradients, fronts and tidal eddies (Hunt et al. 1998; Hamer et al. 2000) but evidence of nested search behaviour is very limited. Becker & Beissinger (2003) recorded scale-dependent habitat selection in a near-shore predator, the marbled murrelet Brachyramphus marmoratus (Gmelin), but their study was based on marine survey data and so they were not able to examine search strategies of individual birds. Moreover, the relationship between prey encounter and search behaviour has not previously been examined for any marine predators with this scale of foraging range.
Northern gannets Morus bassanus (Linn) (hereafter gannets) are medium-ranging foragers, capable of travelling > 1000 km on a single round-trip to obtain food for themselves and their offspring (Garthe et al. 2007; Hamer et al. 2007), although typical ranges of trips (< 200 km) are much shorter than those of long-distance pelagic foragers (Weimerskirch & Cherel 1998; Magalhães, Santos & Hamer 2008). Adults exploit a wide range of species and sizes of prey, obtained using several distinct capture techniques (vertical plunge-diving, underwater pursuit, scooping from the surface and scavenging discards from fishing vessels; Garthe, Benvenuti & Montevecchi 2000; Hamer et al. 2000; Lewis et al. 2003). Birds at one of the largest colonies of gannets (~48 000 breeding pairs; Wanless, Murray & Harris 2005), at Bass Rock, south-east Scotland (56°6′N, 2°36′W), were recorded to forage repeatedly over a narrow range of bearings before switching to a separate range of bearings in a markedly different direction from the colony, providing strong evidence that individuals learned and remembered the directions to feeding sites and used that knowledge on subsequent foraging trips (Hamer et al. 2001, 2007). However, it was not known whether or not adults also adopted ARS behaviour at sea or if so, how this related to oceanography or to what extent it was associated with detection of prey.
The North Sea is a semi-enclosed shelf sea where the interaction between bathymetry, tidal currents, solar irradiation and wind patterns creates a mosaic of mixed, stratified and frontal regions (Scott et al. 2006). The main frontal zone in the region of the Bass Rock is a tidal mixing front located about 50 km offshore, at a distance between the shelf break and the coast where the water is shallow enough for tidal mixing to reach the surface (Fig. 1; Skov et al. 2008). Inshore of the front, the water is mainly mixed while beyond the front, the water is stratified during the summer (Daunt et al. 2006). Primary and secondary production are typically concentrated at frontal regions (Sims & Quayle 1998) and there is evidence that this front may be a focus for the foraging activity of marine predators in the region, including gannets (Daunt et al. 2006; Skov et al. 2008). Marine survey data indicated that gannets captured prey at shallower depths inshore of the tidal mixing front than offshore (Camphuysen, Scott & Wanless 2006), but it was not known how this difference related to the movement patterns or overall foraging strategies of birds.
This paper uses fine-scale movement and activity data from global positioning system (GPS) loggers combined with recently developed first-passage time (FPT) analysis (Fauchald & Tveraa 2003) to examine the spatial scales of foraging and nested search strategy of gannets in relation to the physical oceanography of the western North Sea. FPT is based on the time taken by a foraging animal to cross a circle of given radius, providing a scale-dependent index of search effort and allowing identification of the spatial scales at which effort is concentrated by increases in sinuosity and/or decreases in speed of movement. Detection and pursuit of prey by gannets is indicated by the initiation of diving behaviour (Garthe et al. 2000; Lewis et al. 2002) and here, we test the following hypotheses: (i) birds adopt a hierarchical foraging strategy characterized by nested ARS behaviour; (ii) the locations and characteristics of ARS zones are strongly influenced by physical oceanography; (iii) the initiation of ARS behaviour is triggered by the detection and pursuit of prey; (iv) ARS behaviour is strongly linked to higher foraging effort, particularly within nested ARS areas.
Materials and methods
Field work on the Bass Rock took place between mid-July and mid-August 2003. One chick-rearing adult from each of 13 nests with hatching dates ±2 weeks from the annual mode was captured at the nest using a 6-m telescopic pole with a brass noose. A GPS logger with an external temperature and depth probe (Earth & Ocean Technologies, Kiel, Germany) weighing 70 g (< 3% of adult mass) was attached to the feathers on the back of each bird using waterproof tape (Tesa AG, Hamburg, Germany), such that the wings of the bird protected the device during plunge diving. The temperature and depth sensor was located at the end of an external wire, secured with an L-shaped piece of thermoplastic and waterproof tape so that it hung beneath the tail feathers. Attachment of tags took ~15 min and after release, birds returned to the nest almost immediately. After release, birds were tracked for 1–4 days each over a period of 1 month during mid chick rearing, after which time the bird was recaptured and the tag removed. Loggers had no discernible impacts on trip durations or body masses of birds in comparison to untagged controls (Hamer et al. 2007).
GPS loggers provided location data at intervals of 3 min (accuracy within 20 m for 90% of all fixes) while the external probe measured temperature (resolution 0·01 °C) and pressure (resolution within 0·1 m) every 2 s. Locations of birds at sea were examined in arc-view gis. We used the furthest recorded location from the colony to separate outward and return portions of each trip, and we calculated total distance travelled as the sum of distances between consecutive locations at sea. To identify zones of ARS, we used the adehabitat and ade4 packages in software r 2·7·0 (R Development Core Team 2007) to apply FPT analysis, following Fauchald & Tveraa (2003). To ensure that points along foraging tracks were equally represented (Pinaud 2008), we interpolated locations to obtain a uniform distance interval of 1 km. FPT was then calculated for every 1 km for a radius r from 1 to 100 km.
A plot of log (FPT) against r (using log-transformation to ensure that the variance was independent of the magnitude of the mean; Fauchald & Tveraa 2003) allowed identification of the ARS scales of each foraging trip by peaks in the variance. Inclusion of overnight periods when birds remain sitting on the water (Hamer et al. 2000; Lewis et al. 2002) could have dramatically inflated the variance in FPT at small spatial scales (Weimerskirch et al. 2007), and so this analysis was restricted to hours of daylight (4·00–23·00 BST) during foraging trips. By plotting FPT values at appropriate ARS scales as a function of time elapsed since leaving the colony, we identified where and when birds entered and left ARS zones. We then estimated the area of each ARS zone by using the maximum distance between any two points within each zone as a measure of diameter. Next, we repeated the FPT analysis at a finer spatial scale (every 100 m for a radius from 100 m to 10 km) to search for nested ARS behaviour within each larger-scale ARS zone.
The timing, duration and maximum depth attained during dives were analysed using mt-dive 4·0 (Jensen Software, Loehe, Germany). Apparent dives < 0·3 m were excluded because many of these were probably nonforaging movements associated with bathing (Garthe et al. 2000). Durations and depths of dives were highly correlated (rS = 0·8, n = 420, P < 0·01) and so only depths are presented here. Other activities at sea were identified from a combination of location and temperature data: time on the sea surface was indicated by more-or-less constant temperature and was confirmed, for periods exceeding 3 min, by checking that the bird had not moved from its previous location. Flight was indicated by rapid changes in location, accompanied by large fluctuations in temperature. These data were used to locate dives at sea and to calculate the proportion of time spent in different activities.
We recorded many dives and, in some cases, several ARS zones per trip. To examine individual variation in foraging behaviour and to take account of potential pseudoreplication of data, we therefore used generalized linear models (GLM) with trip identity included as a random effect. We also initially included sexes of birds but there was no difference between sexes in ARS behaviour or dive rate (P > 0·3 in all cases) and so this variable was dropped from all models. We used the position of the tidal mixing front during July 2008 (Fig. 1) to indicate the boundary between mixed coastal water and seasonally stratified water (see Skov et al. 2008 for method of estimating position of front). Data were tested for normality and, where necessary, log-transformed before statistical tests. Mean values are given ±1 SD unless otherwise stated.
The mean duration of foraging trips was 21·5 ± 6·7 h (range = 11·1–34·9 h, n = 15 trips from 13 birds), with a mean distance covered per trip of 439·8 ± 234·4 km (range = 159–984 km) and a mean foraging range of 155·2 ± 65·3 km (range = 68–276 km; Fig. 2). On average, birds spent 41·6 ± 11·7% (range = 22·0–60·4 %, n = 14) of each foraging trip in flight, with an average flight speed of 44·7 ± 7·2 km h−1 (range = 33·4–60·2 km h−1, n = 14; one logger did not record activity data). Trips encompassed both mixed coastal water and seasonally stratified water east of the tidal mixing front: nine trips were restricted largely to coastal or frontal water and six trips extended well over stratified water beyond the front (Fig. 2). Most trips had more or less direct flight paths, leaving and returning to the colony along similar routes, but two trips had more elliptical paths (Fig. 2).
area-restricted search behaviour
ARS zones detected by the FPT method were found on all but two foraging trips, both extending over seasonally stratified water, which had very low dive rates (see below) suggesting little success in locating prey. For these two birds, the variance in log FPT decreased linearly with increasing spatial scale, suggesting random foraging with no distinct ARS behaviour (Fauchald & Tveraa 2006). All other trips had a clear peak in the variance of log FPT, indicating the scale of ARS behaviour, at radii of 4·1–57·4 km (back-transformed mean log radius = 9·1 ± 1·9 km; logs used to normalize data). There was a significant positive relationship between ARS scale and maximum foraging range (GLM using log radius; F1,10 = 14·2, n = 13 trips, P < 0·01). However, there was no difference in the scale of ARS of trips that extended over stratified water and those that did not (F1,10 = 1·1, P = 0·3).
We identified up to four ARS zones per trip (mean = 1·9 ± 0·9). A further six apparent zones that closely matched hours of darkness and contained no dives (see below) were excluded from the data. Birds spent time overnight in 46% of 25 recognized ARS zones but all such occasions also included at least 2 h of daylight and began (and sometimes also ended) with a sequence of dives, and so were not the result of birds resting on the water overnight. ARS zones were typically characterized by increased sinuosity of foraging tracks in addition to long passage times (Fig. 3) and were all on the outward phases of trips before reaching the maximum foraging range (Fig. 3), at a mean distance of 117·0 ± 55·0 km from the colony (range = 35–275 km). ARS zones occurred over both mixed coastal water and stratified water, with over half (58%) within 5 km of the tidal mixing front (Fig. 4). The estimated areas of ARS zones (22·6 ± 14·1 km2) differed consistently among individuals (GLM; F12,11 = 3·4, P < 0·05) but did not differ between mixed coastal water (including the region of the tidal mixing front) and stratified water beyond the front, and were not related to distance from the colony (P = 0·9 in both cases).
nested ars behaviour
Nested fine-scale ARS behaviour was detected within 15 of 25 ARS zones (60%) during 10 of 13 trips, with a mean radius of 1·5 ± 0·8 km (range = 0·5–3 km). There was no difference between mixed coastal and stratified water or among birds in the proportion of ARS zones showing nested search behaviour (P > 0·5 in each case; four apparent nested ARS zones completely occupying hours of darkness were excluded from these data). Where such behaviour was present, there were up to four nested search areas within each ARS zone (mean = 1·8 ± 0·9) and birds spent 49 ± 45 min (range 12–180 min) in each nested ARS search area.
ars and diving behaviour
The first dive of each trip was 53·9 ± 45·7 km from the colony (range = 3·6–148·0 km, n = 420) and 67·0% of dives were on the outward phase of the trip before reaching the maximum foraging range. Birds did not dive overnight. The two birds with no detectable ARS behaviour (see above) had very low dive rates compared to birds that had ARS (0·4 ± 0·3 dives h−1 and 1·7 ± 1·5 dives h−1, respectively; t-test using unequal variance estimate; t8·5 = 2·5, P < 0·05).
All but two large-scale ARS zones (92%) included dives and the mean interval between entering ARS and commencing diving was 7·0 ± 7·6 min. However, 27·1 ± 24·3% of dives per trip were not in ARS zones (Fig. 5), and the mean interval between the final dive outside an ARS zone and the subsequent initiation of ARS behaviour was 56·9 ± 4·8 min. The proportion of dives within ARS zones was 17% higher in stratified water beyond the tidal mixing front than in mixed coastal water (Fig. 5; GLM using arcsine-square root transformed proportions for each trip; F1,6 = 15·7, P < 0·01) and also differed significantly among trips (F11,6 = 5·1, P < 0·05). However, intervals between initiation of ARS behaviour and diving events did not differ among trips or between water bodies, whether considering dives before or after birds entered ARS (P ≥ 0·5 in all cases).
Following the initiation of ARS behaviour, the mean dive rate during hours of daylight within ARS zones (3·8 ± 4·6 dives h−1) was almost four times that outside ARS zones (1·0 ± 0·9 dives h−1; Fig. 6; GLM; F1,20 = 13·2, P < 0·01). Moreover, 82·7 ± 27·9% of dives within large-scale ARS zones were in nested small-scale ARS areas, and daytime dive rates were almost three times higher in zones with such nested ARS behaviour than in those without (Fig. 6; F1,10 = 8·4, P = 0·01). Dive rate also varied consistently among birds (F11,10 = 5·2, P < 0·01) but did not differ between ARS zones in mixed coastal water and in stratified water beyond the tidal mixing front (F1,10 = 0·03, P = 0·9).
characteristics of dives
Birds dived up to 19 m below the sea surface but 58% of dives were to depths < 2 m. The frequency distribution of dive depths was bimodal with two peaks in depth usage at 0·3–1 m and 6–7 m, indicating a clear separation between shallow dives (< 2 m) and deep dives (≥ 2 m). Overall, birds used deep and shallow dives with similar frequencies (mean proportion of shallow dives per trip = 55·2 ± 25·0%, n = 14 trips), but there was marked variation between individuals in this respect (range = 14–94%), which was related to where birds foraged but not to ARS behaviour: the mean proportion of shallow dives per trip was nine times higher in mixed coastal water (81·6 ± 26·5%) than in stratified water beyond the tidal mixing front (9·0 ± 8·7%; Fig. 4; GLM; F1,22 = 24·7, P < 0·001) but was no different within and outside ARS zones (F1,22 = 0·07, P = 0·8).
Shallow dives were significantly shallower within ARS zones than elsewhere (mean = 0·67 ± 0·39 m, n = 167 and 0·78 ± 0·37 m, n = 66, respectively; GLM; F1,216 = 5·1, P < 0·05) but did not differ between water bodies or times of day (P = 0·1 in both cases). Deep dives were significantly deeper in stratified water (5·2 ± 2·7 m, n = 140) than in mixed coastal water (4·3 ± 1·7 m, n = 47; F1,172 = 7·2, P < 0·01) but did not differ with respect to ARS behaviour (F1,172 = 0·3, P = 0·6). Deep dives were also deeper during the main part of the day (7:00–19:00 h BST; mean = 5·5 ± 2·9 m, n = 114) than in the early morning and evening (3:00–7:00 h and 19:00–23:00 h; mean = 4·2 ± 1·2 m, n = 73; F1,172 = 7·7, P < 0·01). Depths of both shallow and deep dives also differed significantly among trips (F13,216 = 6·6, P < 0·001 and F10,172 = 8·4, P < 0·001, respectively).
This study provides strong evidence that gannets used a nested search strategy to locate prey, in support of hypothesis 1 above. Birds typically concentrated their search effort at a spatial scale of about 10 km and within these large-scale search areas, they often searched more intensively at a scale of about 1 km. This hierarchical search strategy was similar to that seen over much larger spatial scales (100 s to 10 s km) in long-ranging petrels and albatrosses (Pinaud & Weimerskich 2005; Fauchald & Tveraa 2006). In gannets, the largest scale of ARS behaviour during foraging trips was positively related to the maximum range of each trip, indicating that birds searched for prey over larger spatial scales during trips that extended further from the colony, as also found in yellow-nosed albatrosses Thallasarche carteri (Rothschild) (Pinaud & Weimerskich 2005) and Antarctic petrels Thalassoica antarctica (Gmelin) (Fauchald & Tveraa 2006). Moreover, combining data from this study with previous data for other species, there was a significant positive relationship between the mean distance of ARS zones from the colony and mean ARS scale (Fig. 7; F1,7 = 17·7, P < 0·01, β = 0·05 ± 0·01, R2 = 0·72), suggesting a positive relationship between foraging distance and ARS scale both within and among species.
Several recent studies of the movement patterns of predators in relation to the distribution of prey patches on a single scale have suggested that predators employ a form of random walk termed a Lévy flight pattern, characterized by large numbers of short movements interspersed with occasional much longer displacements (Viswanathan et al. 1999). More recently, mathematical simulations have suggested that this pattern of movement maximizes encounter rates with randomly distributed patches of prey and may be ubiquitous among marine predators (Sims et al. 2008). However, a hierarchical search strategy will also generate this type of pattern, with a large number of short movements within small-scale prey patches and occasional long displacements between large-scale patches (Fauchald 1999; Fauchald & Tveraa 2006). In the current study, most gannets did not choose displacement distances at random but used a nonrandom nested search strategy, as also found in other species (e.g. Pinaud & Weimerskich 2005). Moreover, birds at this colony do not choose foraging trajectories at random but forage repeatedly over a narrow range of bearings on successive trips (Hamer et al. 2001). In view of the ubiquity of hierarchical spatial structures in nature (Kotliar & Wiens 1990; Wu & Loucks 1995) we agree with the suggestion by Fauchald & Tveraa (2006) that nested search strategies are likely to occur in a wide range of species in both terrestrial and marine ecosystems, and this merits further study.
The importance of physical oceanography in driving spatial heterogeneity in pelagic marine ecosystems is well known (Hunt & Schneider 1987; Hofmann & Murphy 2004). In this study, ARS zones of gannets were strongly associated with the tidal mixing front offshore from the colony, supporting hypothesis 2 above. This pattern is consistent with a previous analysis of habitat suitability based solely on dive locations (Skov et al. 2008) and suggests that similar biophysical processes to those that concentrate predators over much larger scales within pelagic marine ecosystems are also important at smaller spatial scales within coastal seas. Birds in 2003 preyed extensively on sandeels (mainly Ammodytes marinus Raitt; > 50% of the diet by mass; Hamer et al. 2007), which are concentrated in the region of the front and frequently driven close to the surface by predatory fish, cetaceans and pursuit-diving seabirds (Camphuysen et al. 2006). In other years, birds preyed less on sandeels and their foraging activity was focused more on other oceanographic features including sandbanks and deeps at greater distance from the colony (Hamer et al. 2000, 2007). Despite marked differences between years in trip durations, foraging ranges and total distances travelled, gannets did not appear to alter their overall search strategy or sinuosity of foraging paths (i.e. the extent of deviation from a straight-line course) between years in relation to trip duration or foraging range (Hamer et al. 2007). These data are consistent with the suggestion that patterns of search behaviour are a feature of individual species and foraging habitats, not greatly affected by annual variation in environmental conditions (Pinaud & Weimerskich 2007), but further data are needed to examine the nested search strategies of gannets under different environmental conditions.
We found that the initiation of ARS behaviour was typically followed after only a short time interval (~5 min) by the commencement of diving. Gannets do not dive until after they have located prey, and so this pattern strongly suggests that ARS behaviour was triggered by prey detection, supporting hypothesis 3 above. However, ~33% of dives in mixed coastal water and 16% of dives in stratified water were not associated with any detectable ARS behaviour, suggesting that while ARS behaviour was triggered by the detection and pursuit of prey, encounters with prey species did not necessarily induce ARS behaviour. This absence of ARS was possibly the result of unsuccessful dives, or of birds exploiting large prey such as mackerel Scomber scombrus Linn that quickly satisfied their nutritional requirements, or of short-lived feeding opportunities such as isolated food items, as found in wandering albatrosses Diomedea exulans Linn (Weimerskirch et al. 2007). Lewis et al. (2001) suggested that feeding opportunities of gannets close to large colonies may be restricted by interference competition with conspecifics, as a result of prey taking evasive action in response to gannets plunge diving. This could explain the lower proportion of dives within ARS zones in mixed coastal water close to the colony, where competition with conspecifics within prey patches is likely to have been higher (Lewis et al. 2001). The extent of feeding within ARS zones may also have been underestimated in mixed coastal waters, where birds take some prey by scooping from the surface without diving (Camphuysen et al. 2006).
Following the initiation of ARS behaviour, dive rates were almost four times higher within ARS zones than elsewhere and almost three times higher in zones with nested ARS behaviour than in those without. These data indicate that the adoption of ARS behaviour was strongly linked to enhanced foraging success, supporting hypothesis 4 above, particularly within nested small-scale ARS zones where birds may have used the behaviour of other individuals to enhance prey detection (Hamer et al. 2000; Grünbaum & Veit 2003).
Gannets attained a maximum depth of 19 m during dives in this study, which was similar to that recorded previously in the North Sea and elsewhere (Lewis et al. 2002; Garthe et al. 2007). The overall dive rate on trips was also similar to that recorded at the colony in 2001 (Lewis et al. 2002), although the latter study recorded relatively few shallow dives, coinciding with a much lower occurrence of sandeel in the diet that year (Hamer et al. 2006) and so probably reflecting less feeding in mixed coastal waters where dives were shallower. Dives were mainly in the outward portion of trips in this study and ARS areas were all in the outward half, indicating that birds returned fairly directly to the colony without much search effort, as also suggested by Hamer et al. (2007) from a coarse-scale analysis of speeds of travel at sea. Lewis et al. (2004a) suggested that birds partitioned foraging effort fairly evenly between initial, middle and final sections of trips in terms of time. However that study was not able to examine locations of dives at sea and the adoption of ARS behaviour results in birds attaining maximum range much later in the trip than would otherwise be expected. The absence of prey-searching behaviour during the return leg of foraging trips contrasts with the pattern recorded in Cape gannets Morus capensis (Lichtenstein) feeding in the Benguela current ecosystem, where adults frequently stopped to feed on the way back to the colony during trips with an average duration of < 10 h (Ropert-Coudert et al. 2004). This difference probably reflected the much longer foraging ranges and trip durations of northern gannets, which resulted in pressure for adults to return to the nest as quickly as possible to relieve the partner at the nest (Hamer et al. 2007).
We have shown in this study that gannets increased their search effort at nested hierarchical spatial scales, but with differences among individuals in this respect, which were closely associated with differences in dive rates during trips. Prey capture rate was not estimated in this study but, assuming that at least some dives result in prey capture, this pattern strongly suggests that the foraging success of predators in a hierarchical system could be linked to the ability of individuals to track the system, as also found in yellow-nosed albatrosses (Pinaud & Weimerskich 2005). Differences in foraging success during trips are likely to have consequences for individual fitness (Garthe, Grémillet & Furness 1999; Lewis et al. 2004b), emphasizing the importance of scale-dependent individual approaches to understanding the links between prey distribution and predator population processes in heterogenous landscapes.
We thank Sir H. Hamilton-Dalrymple for access to the Bass Rock, J. Croxall and B. Nelson for assistance in developing the project and B. Nelson and the Marr family for continuing logistic support and advice. We thank S. Lewis for invaluable assistance with deployment of loggers, and David Pinaud and one anonymous reviewer for helpful comments on an earlier draft of the manuscript. This work was funded by a grant from the European Union (Interactions between the marine environment, predators and prey: implications for sustainable sandeel fisheries (IMPRESS), Project Q5RS-2000-30864).