Diets of short-tailed shearwaters in the southeastern Bering Sea

In the late 1990s, the southeastern Bering Sea exhibited a number of anomalous conditions, including a major die-off of short-tailed shearwaters ( Pufﬁnus tenuirostris ), a trans-equatorial migrant that constitutes a major portion of the marine bird biomass in the southeastern Bering Sea. As part of a larger study of the ecological role of the inner or structural front over the southeastern Bering Sea shelf, in 1997–1999, we collected short-tailed shearwaters to determine diet composition. In spring 1997, we found that short-tailed shearwaters were consuming predominately the euphausiid Thysanoessa raschii , a diet expected on the basis of past studies. However, in subsequent years, short-tailed shearwater diets in spring contained increasingly larger proportions of ﬁsh, in particular, sandlance ( Ammodytes hexapterus ), as well as other species of euphausiids ( T. inermis in 1999). In summer and fall collections, short-tailed shearwater diets were more varied than in spring, and included both ﬁsh (age-0 gadids, 21–35% by weight) and a wider variety of euphausiid species (T. inermis and T. spinifera ). In summer and fall, crab zoea (August 1998) and copepods (August 1999) were eaten by shearwaters collected while feeding within the inner front. Diets in 1997–1999 were broader than those found in previous studies of short-tailed shearwaters over the inner shelf and Bristol Bay, which had documented diets composed almost solely of T. raschii . Our data are consistent with the hypothesis that euphausiids were less available to short-tailed shearwaters foraging over the middle and coastal domains of the southeastern Bering Sea in 1997–1999 than has previously been true. Our results are also consistent with hypothesis that the inner front can affect the availability of prey to shearwaters


Introduction
Each year millions of short-tailed shearwaters (Puffinus tenuirostris) migrate from their breeding colonies in Australia to spend the austral winter in the North Pacific (Shuntov, 1964;Tanaka and Kajihara, 1979;Guzman, 1981;Gould et al., 1982).As many as 16 million of these shearwaters forage in the Bering Sea between April and October (Schneider and Shuntov, 1993), where they have been recorded in flocks of tens to hundreds of thousands of birds (Guzman, 1981;Hunt et al., 1981b;Gould et al., 1982).The evolution of this annual trans-equatorial migration implies that extraordinary amounts of prey must be readily available to these birds in the Bering Sea.
The southeastern Bering Sea shelf is up to 500 km wide (Fig. 1).The inner and mid-shelf waters are differentiated by hydrographic structure associated with characteristic bottom-depth ranges (Coachman, 1986;Schumacher and Stabeno, 1998;Stabeno et al., 2001).The coastal domain (depth o50 m) typically is either weakly stratified or well mixed as a result of a combination of tidal and wind mixing.The middle shelf domain (depth between 50 and 100 m) is twolayered in summer because mixing energy is not sufficient to stir the entire water column in the presence of a positive buoyancy flux.A structural front (the inner front) 5-30 km wide separates the middle and coastal domains (Schumacher et al., 1979;Kachel et al., 2002).It is present along the entire southeastern shelf to north of Nunivak Island.A similar feature exists around the Pribilof Islands (Kinder et al., 1983).Advection across the structural front is minimal (Coachman, 1986;Kachel et al., 2003).
Between May and September, major concentrations of short-tailed shearwaters occur in an arc from the eastern passes of the Aleutian Islands into Bristol Bay and northwestward over the nearshore waters to Nunivak Island (Bartonek and Gibson, 1972;Guzman, 1981;Hunt et al., 1981b;Gould et al., 1982).These large aggrega-tions appear to be associated with the 50-m isobath throughout the eastern and western Bering Sea, as well as in the shallower passes of the Aleutian Islands (Schneider and Shuntov, 1993).
Previous studies have shown that in Bristol Bay and adjacent areas of the southeastern Bering Sea, short-tailed shearwaters forage primarily on euphausiids (in particular, Thysanoessa raschii and T. inermis) (Ogi, 1973;Ogi et al., 1980;Hunt et al., 1981a;Schneider et al., 1986;Troy and Bradstreet, 1991).At the Pribilof Islands in July-August 1989, short-tailed shearwaters foraged almost exclusively on adult T. raschii (Hunt et al., 1996).In the Bristol Bay region, Ogi et al. (1980) estimated that short-tailed shearwaters consume at least 30,000 metric tons of euphausiids annually, a consumption roughly equivalent to the consumption of euphausiids by sockeye salmon during the month that these salmon reside in Bristol Bay before ascending rivers to spawn (Nishiyama, 1974).
Short-tailed shearwaters forage by pursuit diving from the surface, plunging from above the surface, seizing prey at the surface, and by hydroplaning, a foot-propelled movement at the surface in which the head and back may be beneath the surface while the wings are held out of the water (Ogi et al., 1980;Morgan, 1982).Short-tailed shearwaters also may obtain prey by filter feeding (Morgan and Ritz, 1982;Lovvorn et al., 2001).Maximum dive depths have been estimated to be between 20 and 40 m (Skira, 1979;Hunt et al., 1996).Thus, short-tailed shearwaters must seek euphausiids either in near-surface swarms, or in epibenthic layers less than 40 m deep.Recent modeling studies suggest that these shearwaters must use filter-feeding in dense nearsurface euphausiid aggregations if they are to meet their daily energy requirements with this type of prey (Lovvorn et al., 2001).
In the late 1990s, the southeastern Bering Sea exhibited a number of anomalous conditions, including unusually warm surface temperatures in 1997 and a recurring coccolithophore bloom that increased turbidity in the water column (Vance et al., 1998;Hunt et al., 1999;Napp and Hunt, 2001;Stabeno et al., 2001;Stockwell et al., 2001).In the Bering Sea in fall 1997, hundreds of thousands of short-tailed shearwaters starved, and in fall 1998, short-tailed shearwaters were emaciated, though moribund birds were not encountered (Baduini, 2000;Baduini et al., 2001a).In spring 1999, short-tailed shearwaters in Bristol Bay had body masses intermediate between those in 1997 and 1998, and in summer 1999, body masses were greater than those measured in either fall 1997 or fall 1998 (Baduini et al., 2001b).In comparison to past studies of short-tailed shearwater foraging ecology in the southeastern Bering Sea, in August and September 1997, short-tailed shearwater diets included a wider variety of prey than expected (Baduini et al., 2001a).Baduini et al. hypothesized that the more diverse diets than expected were a response to the unavailability of adult euphausiids, the previously dominant prey of short-tailed shearwaters over the middle and inner shelf domains.In this paper, we examine the spatial and temporal variability of short-tailed shearwater diets in the southeastern Bering Sea, and relate changes to factors that may have affected the availability of euphausiid prey.

Methods
Between June 1997 and August 1999, we studied the food habits of short-tailed shearwaters during six research cruises to the southeastern Bering Sea (Table 1).To determine food habits, we shot foraging shearwaters or birds sitting on the water that were assumed to have been foraging in the recent past.On each cruise, collections were made in each of our sampling grids (Fig. 1, Table 1) where foraging short-tailed shearwaters were seen.Although we were not able to collect birds from every foraging group encountered, we attempted to collect birds from each area of each grid where birds were foraging.At each collection site, between 3 and 8 birds were shot.
After collection, proventriculus (the fore-stomach of birds where chemical digestion begins) contents were removed immediately, weighed, and preserved in 80% ethanol.Wet weight of alcoholpreserved specimens, their displacement volume, and direct counts were used to determine the diets of individual birds.Proventriculus contents containing more than 50 g (wet weight) biomass of zooplankton were split using a Folsom splitter, and a subsample of the prey present was identified and enumerated.Although squid beaks occurred occasionally, we did not include them in the present analysis because we were unable to determine whether they were recently ingested, or had been carried in the alimentary canal for some time, and thus represented prey ingested outside the study area.
To assign short-tailed shearwater collections to habitat regions with respect to water depth, we identified four habitat categories: offshore (water depth=z > 55 m); the region of the inner front (45 m ozo55 m); the central region of the inner domain (25 m ozo45 m); and the inshore region (zo25 m).The lengths of these regions varied greatly depending on the slope of the bottom.At Slime Bank and Port Moller, grids were only about 40 km long, whereas at Cape Newenham and Nunivak Island, grids were up to 120 km long.Thus, division into four habitat categories reflected the ease with which short-tailed shearwaters might have been able to reach epibenthic prey, rather than distance from shore.We identified the frontal region separately because we wished to know if it was a feature in determining the types of prey taken.
In assessing diet composition for different time-Â location categories (e.g., year, season, cruise, grid or habitat), we determined the proportion of a prey type in a particular bird in terms of both the proportion by displacement volume, and for zooplankton, the proportion by number of prey items present.We then averaged proportions across birds in a category.This approach avoided the possibility that a few birds with particularly large amounts of prey would have disproportionate influence on our assessment of the diet composition of short-tailed shearwaters in our study.
To test for differences between seasons, years and locations, we performed Kruskal-Wallis one-way analyses of variance.Diet data were aggregated into 63 groupings of birds to avoid pseudo-replication.Each group of birds was composed of all the birds shot at a given time and place, and the proportions of prey taken by the individuals in the group were averaged to provide the proportion of prey from that sample.This decision criterion for independent sample units was based on the assumption that birds foraging in a given flock were not independent samples of short-tailed shearwater diets.Most birds taken from a flock contained the same foods.However, occasionally in a flock of birds feeding on a particular prey, we would find an individual that contained not only the prey being taken by others in the flock, but an additional type of prey as well.Thus, our estimate of independent sampling units is conservative.In the tables we present the numbers of birds sampled in each category, as is common practice, whereas in the text, where statistical tests on differences in diets are presented, we use the more conservative sample size based on grouped birds.

Results
The biomass of prey present in the proventriculus of short-tailed shearwaters varied among cruises (Table 2).The mass of prey present in stomachs was generally greater in spring than in fall (Mann-Whitney U ¼ 15293; n ¼ 288 birds, po0:001).The mean displacement volumes of proventriculus contents varied among prey types (euphausiids, 33 ml, to age-0 gadids, 56 ml), as did the maximum volumes recorded (euphausiids, 155 ml, sandlance, 180 ml and age-0 gadids, 198 ml) (Table 3).The maximum for euphausiids was almost double the maximum stomach load of 89 g recorded by Ogi et al. (1980).
The types of prey consumed varied with respect to the location within the grids where foraging short-tailed shearwaters were collected (Table 6).Foraging short-tailed shearwaters that had eaten sandlance were collected primarily from waters shoreward of the inner front (Kruskal-Wallis statistic=11.617,df=3, n ¼ 63 groups, p ¼ 0:009).In contrast, age-0 gadids were found almost exclusively in short-tailed shearwaters collected in middle domain waters deeper than 55 m (Kruskal-Wallis statistic=7.832,df=3, n ¼ 63 groups, p ¼ 0:050).Likewise, there was striking spatial segregation in the locations where various kinds of zooplankton were encountered in foraging short-tailed shearwaters (Table 7).In particular, T. raschii was found primarily in birds collected inshore of the front (Kruskal-Wallis statistic=13.273,df=3, n ¼ 63 groups, p ¼ 0:004).Smaller zooplankton, such as crab zoea and copepods were consumed by shearwaters collected while foraging in the frontal region.

Discussion
In the present study, approximately 70% by weight of short-tailed shearwater prey was zooplankton, 13% sandlance, 8% age-0 gadids, and 8% unidentified fish.Of the zooplankton consumed, 76% by number were T. raschii, 20% were other species of or unidentified euphausiids, 2% were crab zoea and 2% were copepods.Our summer and fall data were collected primarily in July and August in the coastal domain, and in middle domain waters close to the inner front, seasons and places similar to those in which the earlier studies had been conducted.Thus, we expected to find that the euphausiid T. raschii would have made up almost 100% of the shearwaters' diet, as was found by Ogi et al. (1980)-99.4%euphausiids, Schneider et al. (1986)-82% in 1981, 100% in 1982, and Hunt et al. (1996)-100% in 1989.Diets from spring and fall 1997, and to a lesser extent June 1998, conformed to this expectation, although the euphausiids used in August/September 1997 included considerable numbers of T. inermis and T. spinifera, species that had not previously been important constituents of short-tailed shearwater diets in the region.What was unexpected was the heavy reliance on fish in August/September 1998, and throughout the spring and summer of 1999.Likewise, the use of crab zoea in fall 1998, and copepods in summer 1999, was not anticipated.
In other shelf regions of the Bering Sea from Cape Navarin around the northern shelf to the Pribilof Islands, short-tailed shearwater diets contained a wide mix of prey, and T. raschii was not the most important prey species.For example, Ogi et al. (1980), collected 69 samples in June and July of 1973, 1976and 1977 around the outer rim of the eastern Bering Sea (their ''shelf area'').In these samples, the hyperiid amphipod Parathemisto (Themisto) libellula was the most important prey (66.7% by weight, Ogi et al., 1980, Table 1), followed by euphausiids (20.5%) and fish (12.6%).Excluding two stations that they classified as atypical, the distribution of prey types by weight was: amphipods, 88.1%; fish, 11.2%; euphausiids 0.5%; squid, 0.1% (Ogi et al., 1980, Table 5).Fish species included both walleye pollock and capelin (Mallotus villosus).Ogi et al. concluded that shorttailed shearwaters had ''flexible feeding habits that probably varied according to the distribution of planktonic and neustonic organisms in the surface layer''.Thus, we interpret the use of prey other than euphausiids in our study as evidence that this type of prey was less accessible to short-tailed shearwaters foraging in our study areas than was true in the past.
The increase in the use of prey other than T. raschii by short-tailed shearwaters after spring 1997 could be the result of a decrease in the availability but not the abundance of T. raschii, a decrease in the abundance of T raschii, or the result of an increase in the availability of alternative prey.The short-tailed shearwaters turned to a number of different species of alternative prey, some of which, such as copepods and crab zoea, are smaller or with lower energy density than adult T. raschii.It seems unlikely that these alternative prey were more profitable to take than T. raschii had been.As an alternative, we hypothesize that after June 1997, the availability of T. raschii decreased in the inner and middle shelf domains where we were sampling short-tailed shearwater diets.
A decrease in the availability of T. raschii could result from one or more of the following reasons: (1) a change in horizontal distribution (e.g., a (3) a change in the ability of the short-tailed shearwaters to locate euphausiid swarms because of the extensive coccolithophore blooms over the middle and inner shelf (Stockwell et al., 2001); or (4) a decrease in abundance, possibly because of an increase in the biomass of species that may compete with short-tailed shearwaters for euphausiid prey.
For the most part, we lack sufficient data, particularly the historical data, necessary to test these hypotheses.Elsewhere (Baduini et al., 2001a), we have argued that the warm surface water in fall 1997 may have inhibited surface swarming of adult euphausiids.While this explanation may be relevant for 1997, it is not adequate to explain why there was an apparent lack of nearsurface swarms in 1998 and 1999, when surface temperatures were cooler.Alternatively, the presence of the coccolithophore bloom (Sukhanova and Flint, 1998;Vance et al., 1998;Napp and Hunt, 2001) may have reduced the ability of shorttailed shearwaters to find swarms of euphausiids from the air (Baduini et al., 2001a).The presence of dense coccolithophore blooms could reduce the ability to find swarms and thus diminish the region of ocean available for foraging (see also Coyle and Pinchuk, 2002).However, this explanation does not account for the increased use of alternative prey, such as sandlance, early in 1999 in areas distant from the incipient bloom.
It is also possible that the abundance of euphausiids in the inner domain has decreased since the 1980s.During our surveys and transits between sampling grids, we frequently observed both minke (Balaenoptera acutorostrata) and humpback whales (Megaptera novaeangliae) foraging in shallow coastal waters, often in association with foraging short-tailed shearwaters that were consuming euphausiids (G.Hunt, Unpublished data).Each spring from 1997 to 1999, we encountered between 9 and 12 humpback whales foraging together in the shallow waters of inner Bristol Bay.In addition to potential competition from baleen whales, the number of sockeye salmon returning to the rivers of Bristol Bay has increased greatly since the 1970s (Kruse, 1998).However, data presented by Coyle and Pinchuk (2002) show that there is an overlap between the 95% confidence intervals around MOCNESS samples of euphausiid abundance in the 1990s and the estimates of euphausiid abundance from MOC-NESS tows from the 1980s.Therefore, they conclude that there has been no decrease in the abundance of euphausiids over the shelf since the early 1980s.Alternatively, foraging whales and salmon may disrupt the patch structure of the euphausiids, and decrease their availability to the shearwaters, but we lack data with which to test this hypothesis.
Short-tailed shearwater diets showed small-scale variation in composition that reflected the different habitats present in the study areas.In particular, consumption of age-0 gadids was found mostly in birds shot while foraging at or offshore of the inner front.These data are consistent with the observation that these fish were confined primarily to the stratified waters of the middle domain and the frontal region (Coyle and Pinchuk, 2002).In contrast, sandlance were taken mostly in shallow waters inshore of the front.Observations from the ship suggested that short-tailed shearwaters foraging on sandlance did so primarily when the fish were schooling near the surface.Thus, the inshore distribution of shearwaters that had consumed sandlance most likely reflects where sandlance are found in near-surface schools, rather than the depth range to which short-tailed shearwaters can dive.
The distribution of zooplankton obtained from foraging short-tailed shearwaters also indicated that there was considerable small-scale variation in the distribution of zooplankton species.The euphausiid T. raschii was obtained primarily from short-tailed shearwaters foraging inshore of the frontal region, whereas T. inermis and T. spinifera were found primarily at and offshore of this region.Although Pinchuk and Coyle (2002) found little evidence that the inner front influenced the distribution or abundance of copepods, our findings are similar to those reported by Smith (1991) and Coyle and Pinchuk (2002, Fig. 6) for euphausiids species, and are consistent with the hypothesis that the structural front can influence the distribution of fauna (Coachman, 1986;Kachel et al., 2002).

Fig. 1 .
Fig. 1.Southeastern Bering Sea shelf, showing the locations of the collection sites (marked with an X), the main study grids in relation to depth contours (in m) and the coastline.The inner domain is shoreward of the 50-m isobath, and the middle domain is between the 50-and 100-m isobaths.The structural front is roughly co-incident with the 50-m isobath.Numbers refer to collection sites mentioned in Table 1: (1) Nunivak Island Grid; (2) Cape Newenham Grid; (3) Round Island Site; (4) Dillingham Sites; (5) Port Heiden Site; (6) Port Moller grid; (7) Nelson Lagoon Sites; (8) Amak Island Site; and (9) Slime Bank Grid.

Table 1
Timing and distribution of sampling effort

Table 2
Mean net body mass (gross body minus mass of proventriculus contents) and mean volume of proventriculus contents of short-tailed shearwaters in the main study grids

Table 5
Types of zooplankton (percent by number) in the diets of short-tailed shearwaters in the Bristol Bay region of the southeastern Bering Sea

Table 7
Types of zooplankton (percent by number) taken as a function of location within the grids