CORRELATIONS BETWEEN SEABIRDS AND OCEANIC FRONTS AROUND THE PRIBILOF ISLANDS, ALASKA

Located on the extensive continental shelf of the Bering Sea, the Pribilof Islands, Alaska are the site of one of the largest breeding colonies of seabirds in the northern hemisphere. During summer these islands are surrounded by a front that separates vertically homogeneous waters from well stratified waters farther seaward. We studied the front with hydrographic data and the bird distributions with concurrent counts during summer 1977 and spring, summer and fall 1978. Murres (Uris lomvia and U. aalge) sitting on the water aggregated near the front during summer 1977 and probably during summer 1978. Other species, such as northern fulmars (Fulmarus giacialis) and auklets (Aethia pusilla and A. cristatella) were unaffected by the front. We hypothesize that the aggregation of the murres was related to an enhanced availability of their food near the front.


Introduction
Oceanic fronts have long been recognized by oceanographers, and there has been a belief that fronts affect the abundance of marine organisms. With the notable exception of upwelling fronts, however, evidence for these effects has been mostly qualitative or anecdotal. Recently quantitative investigations of biological effects near shelf and shelfbreak fronts have appeared (e.g. Savidge, 1976;Fournier et al., 1977Fournier et al., , 1979Pingree et al., 1978;Iverson et al., 19796;Simpson et al., 1979;Ainley& Jacobs, 1981;Bowman et al., 1981). Such studies show that oceanic fronts indeed have significant impact on the marine ecology of some regions.
Distributions of seabirds, which occupy an upper level in the trophic web, can supply one clue to the interaction between biological and physical processes. Seabird distributions have been hypothesized to be related to the oceanic-scale variations in the availability of food (Bourne, 1963;Ashmole, 1971) as well as to other variables (Gould, 1971  (1979) demonstrated a relationship between water masses and bird communities in the Indian Ocean, and Hunt & Schneider (in preparation) have shown significant relationships between bird communities and water masses in the southeastern Bering Sea. Quantitative relationships between medium-scale  km or so) physical features and seabird concentrations have rarely been established, in part because this requires concurrent ornithological and oceanographic measurements.
During 1977 and 1978 we made coordinated hydrographic measurements and bird counts in the vicinity of the Pribilof Islands in the southeastern Bering Sea ( Figure I). These islands are relatively isolated: the nearest land is the Aleutian Islands about 300 km distant. The Pribilofs sit on the 500 km wide continental shelf about 30 to IOO km north of the shelfbreak.
The islands are breeding sites for herds of up to 1.4 x 10~ northern fur seals (Cullorhinus ursinus;Baker et al., 1970) and they have colonies numbering 2.75 x 10~ seabirds (Hickey & Craighead, 1977): northern fulmars, black-and red-legged kittiwakes (Rissa troductyla and R. brewirostris), thick-billed and common murres, least and crested auklets, parakeet auklets (Cyclo~~hynchus psittaculu), horned puffins (Frutemdu cornidutu) and tufted puffins (Lunda cirrhutu). Dark-bellied shearwaters (Pufinusgriseus and P. tenuirostris) move into the Bering Sea from the southern hemisphere during summer. Although many of these species are seasonal residents, their food requirements must represent a large part of upper trophic web predation in the Bering shelf ecosystem.
Satellite infrared images (Plate I) show that zones of relatively cold water surround each of the Pribilof Islands during summer. Those zones appear similar to the region of well mixed waters found inshore of the front which parallels the 50 m isobath on the southeastern Bering shelf (Schumacher et al., 1979), w h ere a system of three fronts separates the shelf into three hydrographic domains (Coachman et GZ., 1980;Kinder & Schumacher, 1981). This frontal system has been shown to have dramatic consequences for the trophic web over the shelf east of the Pribilofs (Iverson et al., r979a;Schneider & Hunt, 1982). As we show in this paper, the zones of cold water surrounding the islands are set off from surrounding waters by fronts similar to those described by Schumacher et al. (1979). Our hypothesis is that if the physical processes associated with the fronts strongly affect trophic relationships, then a significant correlation should exist between the physical variables that define frontal locations and the distribution of seabirds. We first describe the fronts surrounding the islands, and then we show the small but significant correlation between the bird distributions and the fronts.

Methods
During August 1977 and April-May, August and September 1978 we made bird counts and hydrographic measurements during four cruises of about 5 days each. Bird counts were taken during daylight hours (unless prohibited by fog) and hydrographic measurements were made at night. We took a total of 1378 bird transects (649 in summer), did IZZ hydrographic (CTD) profiles (58 in summer) and did 224 temperature (XBT) profiles (99 in summer).
Bird counts were made by the transect method (Burnham et al., 1980) modified for use at sea. Counts were made while the ship was underway and all birds within 300 m of the ship were recorded during each ro-min count. At typical ship speeds, each counting period corresponded to about I km2 of ocean surface. Species were identified to the lowest possible taxonomic level based on distinctive field marks. Ship-following individuals were noted to prevent duplicate counts. Sea surface salinity and temperature were measured during each ro-min transect.
During hours of darkness, when counting was impossible, we did hydrographic sections in the vicinity of the islands. We used a continuous conductivity-temperature-depth profiling instrument (CTD) which was accurate to better than 0.0~ "C and 0.01%~ in temperature and salinity when averaged to I m values. During summer and fall of 1978 the CTD profiles were supplemented by expendable bathythermographic (XBT) drops, which yielded tempetatures within 0.2 "C.
The limited hydrographic measurements were used to define the frontal structure and location, We regressed bird density against five environmental variables; distance to shelfbreak, distance to land, water depth, sea surface salinity, or sea surface temperature. Statistical analysis was done using the Statistical Package for the Social Sciences (SPSS: Nie et al., 1975).

Fronts around the Pribilofs
During the three seasons covered by our surveys, spring, summer and fall, we found that the water columns were always well mixed near the islands and much less so farther from the islands. The transition between these well mixed and well stratified waters is abrupt (when viewed from a shelf-wide perspective) and it is therefore called a front (Schumacher et al., 1979). This front is most strongly developed during summer and its surface manifestation is clearest then, so we emphasize summer data.
T. N. Kinder et al. A CTD section taken north from St Paul Island on 5 August 1977 illustrates the frontal structure ( Figure 2). Away from the island the water column was strongly stratified in two Iayers: a 20-m thick upper layer, a 50-m thick lower layer and a IO m thick pycnocline. Differences in temperature, salinity and density (a,) between the layers were 8.1 "C, o*z%,, and 1.2 kg rnw3. At the station closest to shore, however, the 37-m deep water column was nearly homogeneous. Maximum vertica1 temperature, salinity and density differences were 0.08 "C, 0.01%~ and 0.01 kg m-3.
This section is typical for summer: isotherms, isohalines and isopycnals are congruent. There is strong (often two-layer) stratification present away from the islands in deeper water, and weaker stratification present dose to the islands in shallower water. When temperatures or density differences were plotted on a chart of the area, lines of equal stratification surrounded each island. While 40 stations (August 1977) were inadequate to clearly delineate the entire pattern, the satellite infrared image taken on I August (Plate I) confirmed the inferred temperature pattern. This image is an exceptionally cloud-free summer view of the Bering shelf, and it shows two large areas of cooler (lighter gray) water surrounding each island. In the original image the patch surrounding St. Paul is lighter (i.e. colder) than that around St, George. We interpret these patches as the upward mixing of the colder lower layer water (Figure 2). The waters near St. George are less well mixed than those around St. Paul, probably because of the generally deeper water close to St. George ( Figure I): the capacity of the mixing process is inversely proportional to water depth (Schumacher et al., 1979;Simpson & Pingree, 1978).
The spring (27 April-3 May) and autumn (23-27 September) 1978 data showed changes from the summer (August 1977 and1978) structure. In both spring and autumn the pattern of stratification away from the islands and well mixed waters close to the islands obtained. In spring, however, stratification was due to relatively warm and salty water from the slope region intruding near the bottom, similar to the situation described by Coachman & Charnel1 (1979).
In autumn, the surface signature of the front had disappeared and the upper 20-30 m was horizontally uniform, probably as a result of the onset of seasonal cooling. These spring and autumn structures fit into an annual pattern (Kinder & Schumacher, 1981), but we are only able to show a relationship between bird numbers and hydrographic data for summer.  Correlation between bird counts and sea surface temperature Sea surface temperature, which was measured during each bird count, was a valid measure of stratification during the two August cruises. In 1977 the regression of vertical density difference accounted for 90% of the surface temperature variance [ Figure 4(a)]. During 1978, when temperature (XBT) data were much more abundant than density (CTD) data, vertical temperature difference accounted for 81% of the surface temperature variance [ Figure 4(b)]. Thus, sea surface temperature was a useful variable for investigating the relationship between summertime stratification and bird counts. Data from the fall and spring cruise were excluded because of poor correlation between surface temperature and stratification.
Bird density is inversely related to distance from the breeding colonies due simply to geometric spreading as birds commute to and from the islands. Sea surface temperature (the explanatory variable) was also inversely related to distance from the island, because of the colder surface temperatures near the island. We controlled for the possible spurious correlation due to geometric spreading by introducing distance as a covariate in a one-way Deviations from the grand mean with distance as a covariate shows higher murre concentrations (on the water) near the seaward side of the front ( Figure  6). The midpoints of 1.0' C wide temperature bins are plotted.
(N = 214 counts in water depths < I 50 m.) analysis of variance in bird numbers due to sea surface temperature. This is equivalent to regressing bird numbers against distance, then analysing the residuals from regression with respect to sea temperature. Results of this analysis are expressed as deviations from the grand mean, because of the adjustment for distance in the analysis. To be valid, this procedure requires that the effects of sea surface temperature and distance from the island be additive (independent) in their effects on bird numbers. For the August 1977 data three bird groups showed non-significant interaction effects in a two-way analysis of variance in bird numbers : murres (P = o-67), auklets (P = 0.37) and fulmars (P = 0.62). Because the effect of geometric spreading was additive, it could be controlled statistically in subsequent analysis of these groups. Murres on the water showed significantly higher density at the outer part of the front (IO-II "C, compare Figures 2 and 5). For all murres, including flying birds, density drops rapidly with distance from land, and the relation between water temperature and bird density is not significant (P = 0.66). A plot of all murres encountered along a single track parallel to the hydrographic section in Figure 2 showed the same result: decreasing numbers out to the front with a peak at 50-60 km, corresponding to the location of the front (compare Figures 2 and 6). Taken together, the statistical and graphical analyses indicate that murres fly outward from the colonies, decreasing their density by geometrical spreading, then aggregate on the water near the front.
Analyses of bird density relative to surface temperature, with distance controlled, were also made for auklets and northern fulmars. Auklets were confined to the immediate vicinity of the breeding colony and showed no significant relation between density and the water temperatures (P = 0.10).
Fulmars, which forage at great distances from the breeding colonies, showed no significant relation between density and water temperature (P = o-15). Of the three bird groups that showed independence between distance and temperature, only murres on the water had a significant association with surface temperature (P = 0.014).
The 1977 August results were checked by performing the same analysis on the August 1978 data. Again, there was a peak in density of murres on the water near the outer edge of the front (Figure 7). The probability of obtaining this result (P = 0.03) is the product of  Figure  7. Analysis of variance for murres on the water, August 1978 (cf. Figure  5). The lower limits of 0.5 "C wide temperature bins are plotted.
(N = 232 counts in water depth <200 m.) the probability of obtaining a mean as extreme as that we observed (P = 0.17) and the probability of having the largest mean coincide with the front (one chance in five, since five temperature classes were used). We also compared the effects of several environmental variables on bird counts for all four cruises (Table I). Multiple linear regression of bird density vs. distance to land, distance to the shelfbreak, water depth, sea surface temperature and sea surface salinity were used. This analysis was used because it is well understood and straightforward, even though the relationship of bird density on surface temperature, for example, was not linear. Significant results were obtained for each cruise, with a median of 8% explained variance. There is a trend for higher correlation in the August cruises than in the spring and fall cruises, especially among murres and northern fulmars. As anticipated, different species responded differently to the variables (e.g. murres, auklets and fulmars in August 1977 described above), so that grouping all birds together reduced explained variance. There were statistically significant correlations between bird densities and environmental variables, including frontal variables, but the explained variance was low. Obviously the front, while significant for murres, is only one factor in explaining seabird distributions around the Pribilofs.

Discussion
We have demonstrated the existence of shelf fronts around the Pribilof Islands and higher densities of sitting (probably feeding) murres near these fronts during summer. Correlations between the fronts and other populous species of seabirds were not significant. An important question is: what causes this preference for the front ?
The link between the birds and the water column is probably trophic because the murres were on the water and probably feeding. Murres feed on juvenile walleye pollock (Theragra chalcogramma) in size ranges of 2-15 cm (Hunt et al., 1981). These pollock in turn typically feed on large-bodied zooplankton (Clarke, 1978). Murres can dive to greater than 40 m depth (Tuck, 1960), that is, down to typical pycnocline depths in the stratified waters. The relationship between the front and the murres probably includes not only pollock and their prey, but the next lower level(s) in the trophic web as well. An important clue may be that a diving bird, such as the murres, concentrated at the front while a non-diving bird, such as the fulmar, did not.
Murres might preferentially feed near the front because of higher productivity or because of greater food availability. Arguments for increased productivity near fronts (e.g. Pingree et al., 1975) often have been based on nutrient depletion in the upper layer, but nutrient concentrations in our study area apparently remain high even in the upper layer of the stratified water (Figure 3). Availability of food might be enhanced by horizontal convergences in the cross-frontal circulation patterns such as appear in the model of James (1978). Such convergences, either surface or especially subsurface, would concentrate organisms that attempt to actively or passively maintain their depth. Alternately, the pycnocline itself may provide the stability necessary for increased production (e.g. Pingree et al., 1978) or for the accumulation of food particles.
In order to understand the concentration of seabirds near the front, a knowledge of the activities of the phytoplankton, zooplankton and pollock near the front is required. As a first step we suggest an integrated sampling program that includes simultaneous sampling of the physical environment (including frontal circulation) and the various levels of the trophic web. This would be extension of work now being done over the shelf east of the Pribilofs (Iverson et al., 1979~).