Foraging destinations of three low-latitude albatross (Phoebastria) species

Transcription

1 J. Zool., Lond. (2001) 254, 391±404 # 2001 The Zoological Society of London Printed in the United Kingdom Foraging destinations of three low-latitude albatross (Phoebastria) species Patricia FernaÂndez 1, David J. Anderson 1 *, Paul R. Sievert 2 and Kathryn P. Huyvaert 1 1 Department of Biology, Wake Forest University, Winston-Salem, NC , U.S.A. 2 Department of Natural Resources Conservation, University of Massachusetts, Amherst, MA , U.S.A. (Accepted 6 September 2000) Abstract Satellite telemetry was used to identify the foraging distributions of three congeneric species of albatrosses that nest in the tropics/subtropics. Breeding waved albatrosses Phoebastria irrorata from the GalaÂpagos Islands travelled to the productive upwelling near the Peruvian coast and nearby areas during the rearing period in Black-footed albatrosses P. nigripes and Laysan albatrosses P. immutabilis nesting in the Hawaiian Islands and tracked during the 1997±98 and 1998±99 breeding seasons also performed long foraging trips, to continental shelf areas of North America. In both years, breeding black-footed albatrosses made long trips to the west coast of North America (British Columbia to California). In 1997±98, breeding Laysan albatrosses travelled primarily to the north of the Hawaiian Islands and reached the waters of the Aleutian Islands and the Gulf of Alaska. In 1998±99, Laysan albatrosses had a complete breeding failure, and no long trips by breeders were tracked as a result. These three species mixed short and long trips during the chick-rearing period, but not the brooding period nor incubation period. Waved albatrosses made only long trips during the incubation period. Analysis of movement patterns showed that the core feeding areas during long trips were located over the continental shelves of North and South America. The data on foraging biology of these species have implications for assessing bycatch risk in commercial sheries. Key words: albatross, Phoebastria, foraging, bycatch INTRODUCTION Recent advances in satellite telemetry have opened a window on the previously obscure life of albatrosses at sea. They have shown, for example, that breeding wandering albatrosses Diomedea exulans can cover between 3600 and km between visits to their chick, reaching speeds of up to 80 km/h over distances of up to 900 km/day (Jouventin & Weimerskirch, 1990). This technique has been applied to various studies of seabird foraging ecology, documenting resource partitioning between sympatric species (Waugh et al., 1999) and between sexes (Prince, Wood et al., 1992), changes in foraging behaviour through the breeding season (Weimerskirch, Salamolard et al., 1993; Arnould et al., 1996; Weimerskirch, Wilson & Lys, 1997), biotic and abiotic properties of the feeding site (Rodhouse & Prince, 1993; Weimerskirch, Doncaster et al., 1994; Cherel & Weimerskirch, 1995; Weimerskirch, Wilson, Guinet et al., 1995; Hull, Hindell & Michael, 1997), and *All correspondence to: D. J. Anderson. da@wfu.edu particularly distribution of albatrosses at sea in relation to bycatch risk in sheries operations (Croxall & Prince, 1996; Brothers et al., 1998; Prince, Croxall et al., 1998). Most of the albatross species that have been tracked by satellite nest on high-latitude islands in the Southern Ocean. Little is known about the foraging characteristics of the albatross species nesting in the tropics and subtropics; these species comprise the genus Phoebastria, the `North Paci c' albatross clade (see Robertson & Nunn, 1998). Of these species, the black-footed albatross P. nigripes and Laysan albatross P. immutabilis nest primarily in the Hawaiian Islands and are sympatric on most of their breeding islands. The shorttailed albatross P. albatrus, currently limited to a single breeding population on Torishima Island, Japan, and the waved albatross P. irrorata of the GalaÂpagos Islands are the remaining members of the genus. The objective of this study was to use satellite telemetry to identify the foraging areas of three of these species: black-footed, Laysan, and waved albatrosses. Ship-based sightings show that Laysan albatrosses are observed and recovered in greatest numbers in the north-western part of the Paci c during their breeding

2 392 P. FernA  ndez ET AL. season, whereas the number of black-footed albatrosses increases to the south-east and east, to the North American coast, during the same period (Shuntov, 1974). The distribution data used to reach these conclusions depended in part on the distribution of boats bearing observers, with the resulting potential for sampling bias. Satellite tracking overcomes this bias by having no geographical limits on observers. Satellite tracking also allows the study of movements of speci c individuals of known status throughout their entire trip, which is not possible using non-telemetric methods. For example, in a preliminary study satellite tracking was used to study movements of waved albatrosses during the incubation period in 1995 (Anderson, Schwandt & Douglas, 1998). All tracked birds `commuted' c km from the breeding site on Isla EspanÄola, GalaÂpagos to the cold upwelling area over the western South American continental shelf, off the coasts of Peru and Ecuador. While movement between GalaÂpagos and the continental shelf was direct and rapid, movement within the upwelling zone was slow with frequent turning. Simple distribution data would indicate that waved albatrosses occupy the entire area between GalaÂpagos and the continent, while the satellite tracking data revealed the area-speci c behaviour of the birds. MATERIALS AND METHODS Seven waved albatrosses breeding on Isla EspanÄ ola, GalaÂpagos (1822'S, 89839'W) were tracked between 4 June and 13 October 1996, during the brooding (hatching to chick age 18 days) and rearing (chick age > 18 days; Harris, 1973; Whittow, 1993a,b) periods as a complement to work in the previous breeding season during the incubation period, using the methods of Anderson et al. (1998). PTT100 (Platform Transmitter Terminals; Microwave Telemetry, Columbia, MD) transmitters were attached to dorsal contour feathers using epoxy glue. The signals from these transmitters were received by orbiting TIROS-N satellites, passed to Argos System ground stations (Service Argos, Largo MD), and forwarded to us by electronic mail. The two Hawaiian albatross species breeding on Tern Island (23852'N, 'W), French Frigate Shoals, north-west Hawaiian Islands were also studied. Tape attachment, rather than glue, was used to x the transmitters to the mid-dorsal feathers of the mantle (FernaÂndez, 1999). In January 1998, 12 PTT100 units were mounted on 6 Laysan and 6 black-footed albatrosses during the brooding part (hatching to chick age 18 days) of the nesting cycle. The transmitters were removed from the rst set of birds after 10±16 days and re-mounted on a second set of 12 breeding adults. Five of these transmitters were recovered for deployment on a third set of breeders, as a result of which, we were able to track a total of 15 black-footed and 14 Laysan albatrosses during the 1997±98 breeding season. In the subsequent breeding season in January 1999, 8 transmitters were mounted on black-footed albatrosses and 8 on Laysan albatrosses, also during brooding. One set of 16 birds were tracked through their entire breeding effort, but transmitters recovered from 3 birds whose chicks died was mounted on 3 other individuals (2 blackfooted and 1 Laysan albatross). Therefore, in 1998±99, a total of 10 black-footed albatrosses and 9 Laysan albatrosses were tracked. Overall, a total of 48 albatrosses (25 black-footed and 23 Laysan) were tracked for 4±182 days (x = 57 days). No bird was tracked in both seasons. Transmitters in both seasons used 8 h on : 24 h off duty cycles. Previous satellite tracking studies of albatrosses have concluded that the PTT does not affect the performance of the bird measurably during the breeding season (Jouventin & Weimerskirch, 1990; Prince, Wood et al., 1992; Wiemerskirch & Robertson, 1994; Arnould et al., 1996; Anderson et al., 1998). Moreover, innovations in the design of the transmitters have reduced their size; the mass of our transmitters (32 g) was c. 1% of the bird's own mass, and was unlikely to impose a signi cant energetic cost on tagged birds (Anderson et al., 1998). Nevertheless, to identify any effect of transmitters on the birds' reproductive success, we monitored nest histories of 20 time-matched nests, located near the nests of tagged birds, that served as controls. We demonstrated that beak length accurately indicates gender by associating cloacal distension, at the time of egg laying, with beak length. Within Laysan albatross breeding pairs, 20/21 (95.2%) of males had longer beaks than their mates did. Similarly, 33/33 (100%) of black-footed albatross males had longer beaks than their mates did. Beak length of both members of a pair was used to determine the gender of the birds tracked in this study. In both the GalaÂpagos and Hawaii studies birds were chosen for tracking and for controls using only 2 criteria: they had a nest in the area designated for the study and they had hatched their egg. We are not aware of any other bias in our choice of study animals or assignment to tracked or control groups. Argos System performance on Tern Island was ground-truthed by comparing the reported locations of all transmitters during 5 days with a known, stationary location (by GPS). Anderson et al. (1998) conducted a similar exercise at the GalaÂpagos nesting site. To analyse the performance of the transmitters at different locations and years, we calculated the mean number of signals received per transmitter per day during each year of the study. These means were pooled and their average calculated; the variation in reception frequency is expressed with the standard error (se), as is commonly used to present the standard deviation of statistics (Sokal & Rohlf, 1995). Descriptions of the movement parameters (distance, maximum range, and days spent at sea) during the breeding season were based on complete round-trips only. However, all those trips in which the transmitter fell off or the battery died (for 1997±98, 2 black-footed albatrosses and 5 Laysan albatrosses; for 1998±99, 4 black-footed albatrosses, 5 Laysan albatrosses) were

3 Foraging destinations of three low-latitude albatross species 393 used to determine locations of foraging sites and distribution at sea during these 2 years. The distance travelled was calculated by summing the distance between every pair of consecutive locations within a trip, using the Great Circle Formula (Fitzpatrick & Modlin, 1986). Speed was calculated using 2 consecutive locations separated by at least 2 h. This method prevented spurious calculations of high speeds over short time intervals that resulted from the inherent error of satellite xes. We also excluded pairs of points separated by > 8 h; because of the duty cycle of the PTTs, pairs were separated by 8hor 24 h. Turning angles were calculated using 3 consecutive locations within a single trip. RESULTS Argos system performance Since the transmitters used a duty cycle of 8 h on : 24 h off, they operated only 25% of the total time that they were deployed. On Tern Island, we received an average of 3.03 (se= 0.17) locations per transmitter per day and a total of 5417 locations during the 1997±98 breeding season, and 4.10 (se= 0.46) locations per transmitter per day and a total of 5411 locations during the 1998±99 season. Due to the length and extent of the trips, the duty cycle of the transmitters did not limit the characterization of the foraging trips of these species. The Argos System assigns six different qualities (from the most accurate to the least: 3, 2, 1, 0, A, B) to locations based on an evaluation of transmitter signal strength. Ground-truthing data for the Hawaiian species (Fig. 1) showed that location quality B had a markedly higher degree of error than other quality classes did: thus we omitted these data from our analysis and used 3506 (85.6%) and 4636 (85.7%) of the 1997±98 and 1998±99 locations, respectively. This work includes a follow-up study on waved albatross foraging behaviour (Anderson et al., 1998). We analysed the data from the 1996 brooding period, when 465 locations were received. The average frequency of reception from waved albatrosses was low (1.28 (se= 0.08) locations per transmitter per day), because the girth of the Earth at the equator limits the number of satellite passes per day (Service Argos, 1988). Thus we used all the locations that were collected regardless of their location quality. As a result, 30% of the GalaÂpagos locations were of the lowest location quality with an average error of 17.8 km (Anderson et al., 1998). The accuracy of the Argos System in the GalaÂpagos study (Anderson et al., 1998) was similar to or better than that further north in the Paci c Ocean (Fig. 1). Effects of transmitters on breeding performance In 1997±98, control Laysan albatrosses edged 8/20 (40%) offspring compared to transmitter-bearing birds Error (km) (a) 1998 n = 20 (b) A B Location quality Fig. 1. Box plot of the error associated with each location quality provided by the Argos System (from the higher to the lower accuracy: 3, 2, 1, 0, A, B). Median, line; 25th and 75th percentiles, box; points, outliers; error bars also shown. Data from: (a) 1997±98; (b) 1998±99. that edged 14/14 (100%) offspring (Yates corrected w 2 = 10.49, d.f. = 2, P < 0.05). In 1998±99 all nine control and all nine tagged Laysan albatross nests failed. In 1997±98 control black-footed albatrosses edged 18/20 (90%) offspring compared to transmitter-bearing birds that edged 13/15 (87%) offspring (Yates corrected w 2 = 0.05, d.f. = 2, P=0.82). In 1998±99 control blackfooted albatrosses edged 7/8 (88%) offspring compared to transmitter-bearing birds that edged 5/8 (63%) offspring (Yates corrected w 2 = 0.33, d.f. = 2, P=0.56). These data offer no evidence of a negative effect of transmitters on breeding albatrosses. It was not possible to document the success of the waved albatrosses to the end of the breeding season, but our previous experience indicates no negative effect of transmitters (Anderson et al., 1998). Waved albatross movements 1996 rearing period n = 60 n = 142 n = 33 n = 94 n = 187 n = 140 n = 160 n = 59 n = 86 n = 37 n = 82 A total of 19 trips was recorded (in three of these trips the transmitter ceased operation) from seven individuals. The results from this study were similar to those of Anderson et al. (1998) during the incubation period when waved albatrosses travelled to the continental shelf off the western South American coast, reaching the waters of Peru and Ecuador. During the rearing period of 1996, they also travelled primarily to the same area

4 394 P. FernA Â ndez ET AL Isla Española Brooding period Chick-rearing period 0 Ecuador Frequency Perú Distance travelled (km) Fig. 3. Frequency distribution of distance travelled by waved albatrosses Phoebastria irrorata during foraging trips in the 1996 breeding season km Black-footed albatross movements during the breeding season Fig. 2. Distribution at sea of breeding waved albatrosses Phoebastria irrorata during the 1996 chick-rearing period. (Fig. 2). The maximum foraging range during this period was 1228 km from EspanÄ ola Island. They covered waters between 0 o ±118S and 788W±918W. However, we also documented short trips during the brooding period lasting < 5 days around the tropical waters of the GalaÂpagos Islands (Fig. 3). Later in the season short (< 5 days at sea) and long trips (> 7 days at sea) were also mixed (Fig. 3), as during the brooding period. The median of maximum ranges during the brooding period (n = 6) was 430 km and the median distance covered was 1245 km (Table 1). During the rearing period the median of maximum ranges was 730 km and the median distance covered per round trip was 1764 km, not signi cantly different from values during the brooding period (Table 1). Both the distance covered and the maximum range were positively correlated with trip duration (n = 10; for distance r s =0.85, P < 0.01; for maximum range r s =0.76, P =0.01). Short trips comprised 50% of all trips during both the brooding period (3/6) and the rearing period (5/10). 1997±98 breeding season During this period 77 complete trips were recorded from this species. Black-footed albatrosses travelled mostly to the north and north-east of the nesting site (Fig. 4) during the breeding season. They ranged from tropical to subarctic latitudes (188N±488N) and over a broad range of longitude in the eastern north Paci c (1218W±1728W). The time spent at sea and the distance travelled from the nesting grounds changed during the breeding season (Table 2). During the brooding period, when chicks are attended very closely, parents remained in the vicinity of the nesting site (Fig. 5) on trips lasting a maximum of 11 days and reached a median maximum range of 303 km and travelled a median distance of 692 km (Table 2). When we compared distance of consecutive trips, we found no signi cant correlation during either the brooding (r s = 0.11, n = 29, P = 0.57) or rearing periods (r s = 70.21, n = 29, P = 0.27); that is, trips neither became increasingly longer nor were alternated in a short±long fashion. Instead, dramatically longer trips began to occur at the transition from the brooding to the rearing period: the median trip distances during Table 1. Descriptive statistics of waved albatross Phoebastria irrorata foraging trips during the 1996 breeding season. The median of all trips was calculated using every trip recorded during this period. The median of all birds was calculated using each individual bird's median as a data point. None of the comparisons between brooding and chick-rearing period was signi cantly different at P < 0.05 level Distance travelled (km) Maximum range (km) Trip duration (days) Period Median, trips Median by bird Median, trips Median by bird Median, trips Median by bird (chick age) Min±max n Min±max n Min±max n Min±max n Min±max n Min±max n Brooding (0±18 days) 75± ± ± ±1228 1±21 1± Chick-rearing (> 18 days) 144± ± ± ±1296 1±27 3±

5 Foraging destinations of three low-latitude albatross species 395 Table 2. Descriptive statistics of black-footed albatross Phoebastria nigripes foraging trips during the 1998 breeding season. The median of all trips was calculated using every trip recorded during this period. The median of all birds was calculated using each individual bird's median as a data point. Bold numbers are signi cantly different Distance travelled (km) Maximum range (km) Trip duration (days) Period Median a, trips Median by birds Median b, trips Median, by birds Median c, trips Median by birds (chick age) Min±max n Min±max n Min±max n Min±max n Min±max n Min±max n Brooding (0±18 days) 195± ± ± ±559 1±11 1.5± Chick-rearing 3,424 4,117 1,174 1, (> 18 days) 153± ± ± ± ±28 4± a Mann±Whitney U = 263, n 1 = 42, n 2 = 35, P < b Mann±Whitney U=261, n 1 = 42, n 2 = 35, P < c Mann±Whiney U = 324, n 1 = 42, n 2 = 35, P < E 165 W 145 W 125 W N. America Tern Island Female Male 500 km Fig. 4. Distribution of male and female black-footed albatrosses Phoebastria nigripes in the north Paci c Ocean during the 1998 breeding season. these two periods were signi cantly different (Mann± Whitney U = 263, n 1 = 42, n 2 =35, P < 0.01) as were the maximum ranges and trip durations (maximum range: Mann±Whitney U = 261, n 1 = 42, n 2 = 35, P < 0.01; trip duration: Mann±Whitney U = 324, n 1 = 42, n 2 = 35, P < 0.01). During the rearing period, black-footed albatrosses mixed short trips near Hawaii with long trips over pelagic waters, frequently reaching the continental shelf of North America (Table 2, Fig. 5). Of the eight individuals followed during the rearing period, six of them approached the continental shelf of North America (California to British Columbia) from 348N to 488N. Most of these trips followed a looping route. Remarkably, some of the individuals reached the same location several times (see Fig. 6), and in some of these trips they ew on a straight line to San Francisco Bay, suggesting that they may have been following boats; others foraged up and down the continental shelf. Males and females seemed to have similar foraging behaviour during this period, but de nitive conclusions cannot be made since females are over-represented in the sample (two males, six females; Fig. 4). One of the two males also approached the coast of North America, highlighting the importance of this zone as a foraging destination for both male and female black-footed albatrosses. Trip distance and maximum range were positively correlated with the duration of foraging trips (brooding period, distance r s = 0.76, n = 42, P < 0.01 and maximum range r s = 0.73, n =42, P < 0.01; rearing period, distance r s = 0.92, n = 35, P < 0.01 and maximum range, r s = 0.89, n = 35, P < 0.01). It is important to note that during the rearing period the maximum range reached a plateau of 4500 km when birds arrived at the west coast of North America; the minimum time for such a round-trip was 15 days. 1998±99 breeding season To determine if movement patterns were similar across years we intended to track the same birds in two different years, but during the 1998±99 breeding season we did not encounter most of the individuals tracked in the previous season, and the ones that were encountered had a high hatching failure. Therefore, during the 1998± 99 breeding season 10 new breeders were tracked ( ve birds of each sex). During this period, 46 complete trips were recorded, and in four other trips the transmitters stopped functioning, probably as a result of attachment failure. During two trips, the forager's chick died but the parents presumably did not know this fact and behaved in a similar way as during the preceding trips, including returning to the nest to end the trip. We included these two trips in the following analysis. In general, the direction of the trips was very similar to that of the previous year; black-footed albatrosses ranged from 198N±518N and 1238W±1758W (Fig. 7),

6 396 P. FernA Â ndez ET AL. 155 E 165 W 125 W 500 km Flight speed (km/h) 40 N 20 N Tern Island (a) Chick-brooding period (b) Chick-rearing period Turn angle ( ) (c) (d) Fig. 5. Black-footed albatross Phoebastria nigripes ight speeds and turning angles at known locations during: (a), (c) brooding periods; (b), (d) rearing periods. (a), (b) Speeds: 0±15 km/h, closed circles; > 15 km/h, open circles. (c), (d) Turning angles: 0±90 o, closed circles; > 90 o, open circles. approaching the west coast of North America. Their movements changed in some respects through the different stages of the breeding cycle. During the 1998±99 brooding period they travelled for longer periods (Tables 2 & 3) and reached more distant points than during the previous year's brooding period (for distance, Mann±Whitney U = 397, n 1 = 42, n 2 = 30, P < 0.01; maximum range, Mann±Whitney U = 418, n 1 = 42, n 2 = 30, P = 0.02; trip duration, Mann±Whitney U = 412, n 1 = 42, n 2 = 30, P = 0.01). Exceptionally for this stage, one female ew to the continental shelf, reaching the waters off Oregon and California. This trip started when the chick was 16 days old and the chick died before the female returned. During the brooding period, duration of foraging trips was positively correlated with the total distance covered and with maximum range (brooding period, distance r s = 0.90, n = 29, P < 0.01 and maximum range, r s = 0.88, n = 29, P < 0.01; rearing period, distance r s = 0.82, n = 16, P < 0.01 and maximum range, r s = 0.80, n = 16, P < 0.01). The distances of consecutive trips during the brooding period were positively correlated (r s = 0.56, n = 20, P = 0.01); that is, breeders did not alternate short and long trips. During the rearing period, both short and long trips occurred, as in the 1997±98 breeding season. Contrary to the previous year's results, these two types of trip were easily distinguishable in 1998±99: short trips lasted 0±3 days and long trips lasted 8±23 days. We documented no trips of 4±7 days. No statistically signi cant relationship existed between the distance covered during consecutive trips during the rearing period (r s =70.21, n =10, P = 0.56), but a suggestion exists in the negative correlation coef cient of alternating short and long trips. In contrast to the results for the brooding period, trips during the rearing period covered similar distances for similar durations in the two seasons (distance Mann±Whitney U = 322, n 1 = 37, n 2 = 18, P = < 0.85; maximum range Mann±Whitney U = 319, n 1 = 37, n 2 = 18, P = 0.80; trip duration Mann±Whitney U = 296, n 1 = 37, n 2 = 18, P = 0.51). Combining all trips, 5/16 (31%) trips during the brooding period were short trips. Post-breeding movements The nestling stage lasts 140 days (Whittow, 1993a), and after this period in 1997±98 one female was followed but the trip was recorded incompletely because the transmitter ceased operation. This female's destination was primarily along the coasts of Oregon and

7 Foraging destinations of three low-latitude albatross species 397 Table 3. Descriptive statistics of black-footed Phoebastria nigripes albatross foraging trips during the 1999 breeding season. The median of all trips was calculated using every trip recorded during this period. The median of all birds was calculated using each individual bird's median as a data point. Bold numbers in the same column are signi cantly different Distance travelled (km) Maximum range (km) Trip duration (days) Period Median a, trips Median by bird Median b, trips Median by bird Median c, trips Median by bird (chick age) Min±max n Min±max n Min±max n Min±max n Min±max n Min±max n Brooding (0 ±18 days) 142± ± ± ±1566 1±35 1.5± Chick-rearing (>18 days) 333± ± ± ±3121 1±23 2± a Mann±Whitney U = 148, n 1 = 30, n 2 = 16, P < b Mann±Whitney U = 152, n 1 = 30, n 2 = 16, P < c Mann±Whitney U = 126.5, n 1 = 30, n 2 = 16, P < E 165 W 145 W 125 W Tern Island Daytime locations Night time locations 300 km Fig. 6. Four consecutive long foraging trips of female black-footed albatross Phoebastria nigripes 51C, 26 February±1 June California, from the mouth of the Columbia River (47842'N, 'W) to Monterey Bay (37844'N, 'W). After spending 1 month (May) around the Columbia River mouth, she travelled to the south along the coast of California and spent 5 days outside San Francisco Bay. Birds were not followed after they nished breeding in 1998±99, since most of the PTTs ceased operation before black-footed albatross offspring edged. Laysan albatross movements during the breeding season 1997±98 breeding season Laysan albatrosses travelled primarily north and northwest of Tern Island (n = 54 trips; Fig. 8), reaching the Gulf of Alaska and the Aleutian Islands and covering from 208N to 608N and 1558E to 1258W. The characteristics of foraging trips varied through the breeding

8 398 P. FernA Â ndez ET AL. 155 E 165 W 125 W 155 E 165 W 125 W N N Tern Island Laysan albatross Black-footed albatross 500 km Female Male Tern Island 500 km Fig. 7. Hawaiian albatross distribution at sea during the brooding (Phoebastria immutabilis and P. nigripes) and rearing period (black-footed albatross P. nigripes), season. During the brooding period, Laysan albatrosses remained relatively close to the nesting site on Tern Island: the median distance travelled during this stage was 1079 km. The foraging range covered during this period was similar to that of black-footed albatrosses (Figs 5 & 9). The distances covered on consecutive trips were positively correlated (r s = 0.79, n = 22, P < 0.01), indicating a tendency to increase trip length as the brooding period progressed. Trip distance, maximum range, and trip duration increased as the breeding season progressed; comparisons of brooding period and rearing period values revealed signi cant differences in each parameter (Table 4). During the rearing period, Laysan albatrosses mixed short and long trips, which were clearly separable as short (1±4 days) and long (12±29 days). The median distance travelled during the short foraging bouts (n = 6) was 879 km and the median maximum range reached was 431 km. Therefore, during this stage short trips had Fig. 8. Distribution at sea of male and female Laysan albatrosses Phoebastria immutabilis during the 1998 nestling period. a shorter range than during the brooding period (Mann±Whitney U = 44, n 1 = 32, n 2 =6, P = 0.04). During the rearing period we did not nd a signi cant relationship between the distances of consecutive trips (r s =70.21, n = 15, P = 0.44), indicating that while Laysan albatrosses mix long and short trips during the rearing period, they do not alternate those trips regularly. The duration of Laysan albatross trips was correlated with both maximum range and distance travelled in both brooding and rearing period (brooding period, distance r s = 0.80, n = 32, P < 0.01 and maximum range, r s = 0.80, n = 32, P < 0.01; rearing period, distance r s = 0.89, n = 22, P < 0.01 and maximum range r s = 0.73, n =22, P < 0.01). Combining all trips, 6/22 (27%) trips during the brooding period were short trips. During the long trips, Laysan albatrosses ew continuously over great distances, primarily straight north of the nesting site, in both looping and linear courses. The median distance covered during these trips was Table 4. Descriptive statistics of Laysan albatross Phoebastria irrorata foraging trips during the 1998 breeding season. The median of all trips was calculated using every trip recorded during this period. The median of all birds was calculated using each individual bird's median as a data point. Bold numbers in the same column are signi cantly different Distance travelled (km) Maximum range (km) Trip duration (days) Period Median a, trips Median by bird Median b, trips Median by bird Median c, trips Median by bird (chick age) Min±max n Min±max n Min±max n Min±max n Min±max n Min±max n Brooding (0±18 days) 291± ± ± ±2380 1±22 2± Chick-rearing (> 18 days) 114± ± ± ±4010 1±29 9.5± a Mann±Whitney U = 168, P < 0.01, n 1 = 32, n 2 = 22. b Mann±Whitney U = 174, P < 0.01, n 1 = 32, n 2 = 22. c Mann±Whitney U = 165, P < 0.01, n 1 = 32, n 2 = 22.

9 Foraging destinations of three low-latitude albatross species E 165 W 125 W 500 km Flight speed (km/h) 40 N 20 N Tern Island (a) (b) Chick-brooding period Chick-rearing period Turn angle ( ) (c) (d) Fig. 9. Laysan albatross Phoebastria immutabilis ight speeds and turning angles at known locations during: (a), (c) brooding periods; (b), (d) rearing periods. (a), (b) Speeds: 0±15 km/h, closed circles; > 15 km/h, open circles. (c), (d) Turning angles: 0±90 o, closed circles; > 90 o, open circles km and the median of maximum ranges was 1694 km (n = 16). In contrast to black-footed albatrosses, only one Laysan albatross commuted repeatedly to a speci c place. This male returned to the neritic waters around two Aleutian Islands (Umnak Island (53810'N, 'W) and Unalaska Island (53830'N, 'W)) on four consecutive trips. This bird did not make short trips between these long trips. Males and females showed similar movements during foraging trips. During the brooding period, both males and females travelled in the vicinity of the Hawaiian Islands (n = 6 individuals of each sex). During the rearing period, three females and four males were tracked; the genders behaved in like manner, travelling primarily to the north of the nesting site and over similar distances (median distance for males = 8602 km; for females = 6124 km; Mann±Whitney U = 39, n 1 = 12, n 2 = 10, P = 0.17; Fig. 8). 1997±98 post-breeding movements One female was still being tracked after its chick edged in July. This female travelled north of Tern Island along the Japan current (35845'N, 'W) and later moved farther north-west, near the Kuril Islands (45855'N, 'E). 1998±99 breeding season This breeding season was characterized by high breeding failure in the entire Laysan albatross population of Tern Island and other islands in the Hawaiian archipelago (A. Asquith, pers. comm.). None of the birds tracked in 1998±99 (n = 9) raised a chick to edging and most offspring died early in the breeding season. For the analysis of the data of breeding birds, only complete trips initiated when the parent's chick was alive were used. The six birds tracked during the brooding period had a median trip duration of 3.5 days at sea (Table 5). Trip duration was positively related to trip distance and maximum range (distance, r s = 0.82, n = 10, P < 0.01; maximum foraging range, r s = 0.82, n = 10, P < 0.01). Male Laysan albatrosses travelled a median distance of 1799 km (n = 10); only one female was tracked during this season and the mean distance travelled was 2569 km (n = 2 trips).

10 400 P. FernA Â ndez ET AL. Table 5. Descriptive statistics of breeding and non-breeding Laysan albatross Phoebastria nigripes foraging trips during the 1999 breeding season. The median of all trips was calculated using every trip recorded during this period. The median of all birds was calculated using the bird's median as a data point Distance travelled (km) Maximum range (km) Trip duration (days) Period Median, trips Median by bird Median, trips Median by bird Median, trips Median by bird (chick age) Min±max n Min±max n Min±max n Min±max n Min±max n Min±max n Brooding (0±18 days) 416± ± ± ±1497 2±7 3± Non-breeding birds 363± ± ± ±3847 2±36 4± E 165 W 125 W 0.25 (a) Black-footed Laysan Frequency Flight speed (km/h) (b) Black-footed Laysan Tern Island 500 km Fig. 10. Distribution at sea of Laysan albatrosses Phoebastria immutabilis after their breeding failure, February±June Laysan albatross foraging patterns and the areas covered differed from those of the previous year. Even when the data from the individuals that failed during breeding (and so were free to move away from the breeding island) was included in the analysis, we found that they did not travel as far north in 1998±99 as in 1997±98. The farthest northern point reached was 46847'N, 'W and none of the birds reached the waters around the Aleutian Islands in 1998±99. In addition, they extended their longitudinal foraging range markedly (Fig. 10). However, the trip durations, distances, and maximum ranges did not differ between years for parents of living nestlings during the brooding period (trip distance Mann±Whitney U = 135, n 1 = 32, n 2 = 11, P = 0.25; maximum range Mann±Whitney U = 129, n 1 = 32, n 2 =11, P = 0.19; trip duration Mann± Whitney U = 135.5, n 1 = 32, n 2 = 11, P = 0.25). After the death of their chicks, most of the birds spent a great part of their foraging time along the east coast of Japan (around 1408E). One female extended her foraging range to the east to 348N, 1328W Turning angle ( ) Fig. 11. Frequency distributions of: (a) ight speeds; (b) turning angles, of black-footed Phoebastria nigripes and Laysan albatrosses Phoebastria immutabilis during the 1998 breeding season. Dotted lines, criteria used to identify speed and turning angle classes in Figs 5 & 9. Analysis of foraging behaviour to determine core feeding sites To separate foraging from travelling behaviour using our movement data, we assumed that the birds engage in area-restricted searches (Curio, 1976) after locating patchily distributed prey, both reducing speed and turning at sharper angles when foraging in comparison with travelling. Both parameters can be estimated from our data, and they are negatively correlated as expected under this assumption (waved albatrosses, 1996 rearing period: r s =70.24, n = 221, P < 0.01; black-footed albatrosses, 1997±98 brooding and rearing periods: r s =70.38, n = 272, P < 0.01; Laysan albatrosses,

11 Foraging destinations of three low-latitude albatross species ±98 brooding and rearing periods: r s =70.32, n = 786, P < 0.01). In addition, we have found a negative correlation between speed of movement and frequency of landing on the water (a prerequisite for foraging) in Laysan and black-footed albatrosses (FernaÂndez & Anderson, 2000). Flight speed and turning angle were used as indicators of activity to identify foraging sites. Speeds 15 km/h (58% of the total recorded) were classi ed as low speed (Fig. 11a), and turning angles 908 were classi ed as sharp (Fig. 11b). Figures 5 & 9 show the spatial distribution of ight speed and turning angle of the two Hawaiian species during the 1997±98 brooding and rearing period; waved albatrosses in 1996 did not provide enough information for this analysis. During the brooding period both species were concentrated in areas around the nesting site where both classes of speed and angle of turn were observed. During the rearing period, area-restricted searches were performed most often over the North American continental shelf, in species-speci c areas. We have already shown that these same areas appeared to be the destinations of the longest trips. As a body, these data indicate that areas over the continental shelves are the principal foraging sites of these species during the rearing part of the breeding cycle. DISCUSSION General features of the foraging movements Waved albatross This study extends the results of Anderson et al. (1998), indicating that waved albatrosses travel to the cold upwelling region off the coast of Peru and Ecuador throughout the breeding season. During the rearing period, trips were found to have both linear and looping courses. Also, they performed short trips around the GalaÂpagos Islands during the brooding period and later in the season they mixed them with long trips. Anderson et al. (1998) did not nd differences in behaviour of males and females during the incubation period. In this study, the sex of the individuals was not identi ed and therefore conclusions about differences in behaviour between sexes cannot be made. Black-footed albatross Breeding black-footed albatrosses directed their foraging ights to waters located east and north-east of the nesting site. These results agree with Shuntov's (1974) data based on boat observations. During both years of study this species travelled to the west coast of North America during long trips. In both 1997±98 and 1998±99, we received reports of our tagged birds (the antenna visible on the dorsal surface) following ships hundreds of kilometres from Tern Island (P. Banko & J. Dutton, pers. comm.), which is a common behaviour in this species (Whittow, 1993a; Veit et al., 1996). Six out of 11 foraging trips that we documented included long, straight tracks from oceanic waters to San Francisco Bay, perhaps indicating further shipfollowing. Laysan albatross This species travelled primarily to the north during the 1997±98 breeding season, reaching the Aleutian Islands and the Gulf of Alaska. In the majority of the trips recorded Laysan albatrosses travelled to northern pelagic waters, but the northerly distribution was more restricted during 1998±99, a year with very low reproductive success at the population level. Following offspring death, but still during the breeding season, the birds extended their foraging range to the west, reaching the east coast of Japan where the cold waters of the Oyashio current mix with the warmer waters of the Kuroshio current. Previous studies at sea have identi ed this area as an important feeding site for Laysan albatrosses during the non-breeding season (Fisher & Fisher, 1972; Shuntov, 1974). It is noteworthy that we did not record a single individual approaching the coast of California, where Laysan albatrosses have been seen mainly over deeper water offshore, especially from October to February (Whittow, 1993b); these observations may be of non-breeders or pre-breeders. Changes in foraging through the breeding season We established that most, but not all, breeding individuals of these three Phoebastria species mix short and long trips during the nestling period. When parents brood alternately they make short trips close to the colony. Beginning at approximately chick age 18 days, many make both short and long trips. This dual strategy has been found not only in the family Diomedeidae but also in other families of the order Procellariiformes. Weimerskirch, Chastel et al. (1994) found that long trips of wandering albatrosses were initiated when adult body condition was poor, and short trips were initiated when the chick had not been fed recently. Tveraa et al. (1998) found that the body condition of parent Antarctic petrels plays a major role in allocation of resources between reproduction and survival. The mixture of short and long trips re ects a trade-off experienced by breeders between offspring nutrition and parental condition. Differences in foraging behaviour between species Tern Island albatross populations covered much of the Paci c Ocean north of 238N during their foraging activities. Based on 131 complete trips from both species, we found that they occupy largely nonoverlapping areas of the ocean during the foraging trips

12 402 P. FernA Â ndez ET AL. (Figs 4 & 8), differing in both latitude (blackfooted albatross median = 288N; Laysan albatross median = 418N) and longitude (black-footed albatross median = 1618W; Laysan albatross median = 1678W). Water characteristics of the Paci c Ocean may partly explain the spatial distribution of the Hawaiian species. The north-western part of the Paci c contains subarctic waters with a cold intermediary layer, and is typically colder than the north-eastern part (Bigg, 1996). Laysan albatross behaviour is particularly interesting since nesting on a subtropical island and making trips to the boreal and arctic waters far north signify great tolerance of temperature change. Black-footed albatrosses started making long trips at older offspring ages than did either Laysan or waved albatrosses. Studies of food allocation in albatrosses (Weimerskirch, Cherel et al., 1997) have found that long trips are energetically poor for the offspring but pro table for the parents, while short trips are the opposite. Therefore, the fact that Laysan and waved albatrosses initiated long trips as soon as the chick could be left alone, may indicate lower foraging ef ciency during short trips for these species. Black-footed albatrosses are known to supplement their diet with sheries offal (Whittow, 1993a; Gould et al., 1998), and may have a higher rate of prey catch during short trips among the Hawaiian Islands-based commercial shery than the other two species do during short trips. The foraging ranges of the two species contracted during the short chick-brooding period, with both species performing short trips near Tern Island. However, the spatial overlap of these two species during this stage was only partial since black-footed albatrosses hatch their eggs 2 weeks earlier than Laysan albatrosses (Whittow, 1993a) and so were switching to longer trips at about the time of Laysan albatross egg hatching. Therefore, Laysan albatrosses used the waters close to the nesting site later than did black-footed albatrosses. Comparison between years The foraging destinations of the two Hawaiian species during the 1998±99 brooding period were generally similar to those of the previous year and the fact that new individuals were used during that year supports the idea that an easterly foraging pattern of black-footed albatrosses and a northerly pattern for Laysan albatrosses are probably observed in most breeding seasons for the populations at Tern Island. However, in 1998± 99, Laysan albatross behaviour changed later in the season when the parents lost their chicks, becoming failed breeders. None of the individuals tracked travelled to the Aleutian Islands and their distribution resembled the description of non-breeder behaviour based on boat observations. The poor reproductive success of both species, especially Laysan albatrosses, during 1998±99 and the expansion of the foraging range during the brooding period may have been related to changes in the weather of the North Paci c Ocean. The change from El NinÄo (ENSO) conditions, with unusual warm water temperatures at the Equatorial Paci c during 1997±98, to La NinÄ a conditions in 1998±99, which have an opposite effect on climate, may have affected the reproductive success of these species. ENSOs are associated with mass mortality in equatorial seabirds. Schreiber & Schreiber (1989) estimated that during the 1982±83 ENSO 75±90% of the 10 million birds nesting on Christmas Island (108S, 1058E) died as a consequence of this phenomenon. However, the same study did not nd any obvious effects on the reproductive output of Laysan and black-footed albatross nesting on Midway Atoll, Hawaii (288N, 1798W). Shea & Ricklefs (1996) also found that most other Hawaiian seabirds were not demonstrably affected by the 1982±83 ENSO event, and they concluded that the ENSO effect on tropical seabirds did not extend to subtropical North Paci c populations. In the rst year of this study, during the 1997±98 ENSO, the reproductive success of blackfooted albatrosses did not decrease in comparison to other years ( edging success = 68.3%, mean of previous years = 68.8% (1984±97); A. Viggiano, pers. comm.), but Laysan albatrosses did experience a decline ( edging success = 32.8%, the third worst year in 16 years of population studies at the island from 1984 to 1999). Laysan albatross chicks that died showed signs of malnutrition and the chicks that survived grew slowly in comparison to black-footed albatross chicks. For blackfooted albatrosses, we suspect that our data from 1997±98 represent foraging under typical conditions; the situation is less clear for Laysan albatrosses. During 1998±99, unusually warm weather occurred at Tern Island during normally cool months, followed by signi cant storms during the period when the chicks were very young (F. Juola & L. Carsten, pers. comm.). In addition, La NinÄ a conditions included a higher water temperature around the coast of Alaska and a coincident decrease in marine productivity (SeaWiFs, 1999). As a result, the breeding success of both species was low in comparison to previous years, especially for Laysan albatrosses, whose reproductive success (9.0%) for 1998±99 was the lowest recorded for the Tern Island population in 16 years of studies of population dynamics at the island (1984±99; A. Viggiano, pers. comm.). Between-year comparisons revealed longer foraging trips in 1998±99, both in distance and time, only by black-footed albatrosses and only during the brooding period. Laysan albatrosses did not perform longer trips in 1999 even though they were experiencing a dramatic breeding failure. It is possible that the foraging characteristics of Laysan albatrosses revealed by satellite telemetry in both seasons are typical of periods of food stress, since reproductive success of this species was low in both seasons. Further studies integrating measures of chlorophyll concentration in the North Paci c Ocean during the period of study with our documented foraging movements will be useful to explain the changes in foraging behaviour and the low reproductive success in this species.

13 Foraging destinations of three low-latitude albatross species 403 Foraging zones This is only the second study (after Anderson et al., 1998) to use satellite telemetry to provide data on the foraging behaviour of tropical/subtropical albatrosses. The changes in movement patterns (decreased ight speed and sharper turning angles over periods of several hours) during foraging trips suggest that the core feeding areas used by these species during long trips are frequently continental shelves. Veit & Prince (1997) suggested that albatrosses use a form of area-restricted search, in which individuals follow a straight line when they are searching for food; once prey is located they adopt a circular motion, enhancing the probability of nding additional food patches. This switch in foraging behaviour has been directly related to food ingestion using stomach temperature sensors (Weimerskirch, Wilson & Lys, 1997), corroborating the idea that this behaviour is associated with feeding activity rather than a response to environmental characteristics of shelves (i.e. changes in wind patterns near land). Therefore, the method used in this study to detect core feeding areas using differences in speed and angle of turn during foraging trips is probably valid. Continental shelves produce oceanographic fronts; these are zones of enhanced physical and biological activity (Le FeÁvre, 1986; Schneider, 1990) that often host a concentration of marine birds (Hunt & Schneider, 1987). Two hypotheses have been proposed to explain why bird abundance is higher at fronts. The rst states that enhanced primary production at continental shelves increases prey supply through increased animal growth, reproduction, or immigration. The second is that prey patches develop at fronts either through behavioural responses of prey to thermal or salinity gradients, or through interaction between prey behaviour and circulatory systems (Schneider, 1990). For the continental shelves off the west coast of North and South America, both effects probably contribute to high marine productivity and therefore the preference of albatrosses for feeding in these areas. This work has shown that the foraging behaviour of these three species of albatross resembles that of a larger species, the wandering albatross, mixing short trips (especially early in the season) and long-distance trips. The trips recorded followed the two basic foraging modes described for other species of albatross (Weimerskirch, 1998). In `commuting ights', birds ew directly to a speci c foraging site and spent the majority of the time feeding at this location, and in `searching ights', birds moved continuously throughout their foraging trips. These two ying modes were used during both short and long trips. However, short trips were characterized by a predominance of searching patterns. Boat-following behaviour has led to an increased concern for the interaction of these species with longline sheries. Gould et al. (1998) showed that blackfooted albatrosses, and to a lesser degree Laysan albatrosses, forage on offal discards from sheries operations, and they suggested that the foraging biology of this species is mainly determined by the spatial distribution of these activities in the North Paci c Ocean. Therefore, it is important to study the distribution of long-line sheries and their impact on blackfooted albatross populations, as well as aspects of diet (Harrison et al., 1983) and diel vs nocturnal foraging behaviour (FernaÂndez & Anderson, 2000). The data provided by this study are invaluable in this respect in identifying areas to which Hawaiian albatrosses direct their foraging ights. Acknowledgements We thank L. Carsten, F. Juola, A. Viggiano, and S. Wang for eld assistance, and D. Hyrenbach and two anonymous reviewers for comments on a previous draft. Funding was provided by National Science Foundation grants DEB and DEB and National Geographic Society grant to DJA, the US Fish and Wildlife Service, and Wake Forest University. REFERENCES Anderson, D. J., Schwandt, A. & Douglas, H. D. (1998). Foraging ranges of the waved albatrosses in the eastern tropical Paci c Ocean. In Albatross biology and conservation: 180±185. Robertson, G. & Gales, R. (Eds). Chipping Norton: Surrey Beatty. Arnould, J. P. Y., Briggs, D. R., Croxall, J. P., Prince, P. A. & Wood, A. G. (1996). The foraging behaviour and energetics of wandering albatrosses brooding chicks. Antarct. Sci. 8: 229± 236. Bigg, G. R. (1996). The oceans and climate. Cambridge: Cambridge University Press. Brothers, N., Gales, R., Hedd, A. & Robertson, G. (1998). Foraging movements of the shy albatross Diomedea cauta breeding in Australia; implications for interactions with longline sheries. Ibis 140: 446±457. Cherel, Y. & Weimerskirch, H. (1995). Seabirds as indicators of marine resources: black-browed albatrosses feeding on ommastrephid squids in Kerguelen waters. Mar. Ecol. Prog. Ser. 129: 295±300. Croxall, J. P. & Prince, P. A. (1996). Potential interactions between wandering albatrosses and longline sheries for Patagonian tooth sh at South Georgia. CCAMLR Sci. 3: 101±110. Curio, E. (1976). The ethology of predation. New York: Springer- Verlag. FernaÂndez, P. (1999). Foraging biology and reproductive rate in albatrosses (family Diomedeidae). MSc thesis, Wake Forest University. FernaÂndez, P. & Anderson, D. J. (2000). Nocturnal and diurnal foraging activity of Hawaiian albatrosses detected with a new immersion monitor. Condor 102: 577±584. Fisher, H. I. & Fisher, J.R. (1972). The oceanic distribution of the Laysan albatross, Diomedea immutabilis. Wilson Bull. 84: 7±27. Fitzpatrick, G. L. & Modlin, M. J. (1986). Direct line distances. United States edn. Metuchen, NJ: Scarecrow Press. Gould, P., Ostrom, P., Walker, W. & Pilichowski, K. (1998). Laysan and black-footed albatrosses: trophic relationships and driftnet sheries associations of non-breeding birds. In Albatross biology and conservation: 199±207. Robertson, G. & Gales, R. (Eds). Chipping Norton: Surrey Beatty.