Abstract

We investigated the functions of perch relocations within a communal night roost of wintering bald eagles (Haliaeetus leucocephalus) along the Nooksack River, Washington, during two winters. We tested seven predictions of two nonexclusive hypotheses: (1) bald eagles relocate within roosts to assess foraging success of conspecifics, and (2) bald eagles relocate to obtain thermoregulatory benefits of an improved microclimate. Additionally, we gathered descriptive information toward refinement of further alternative hypotheses. We rejected the hypothesis that relocations are a means of assessing foraging success. Counter to our expectations, immatures did not reposition to be closer to adults, and relocations were less frequent when food was less abundant. Our data support the hypothesis that eagles relocate within night roosts to obtain a favorable microclimate during winters with cold stress and food stress. Evening relocations were more frequent than morning relocations in both winters. Most evening relocations were to the center of the roost rather than to its edge in both winters, and relocation frequency to the center was greater during low temperatures. The microclimate hypothesis, however, explains only a limited number of relocations. It is likely, based on our findings, that relocations have multiple functions, including establishing and/or maintaining foraging associations, establishing and/or maintaining social dominance hierarchies when food is less abundant, and nonsocial functions.
 
 

     

     

The phenomenon of communal roosting is well documented for the bald eagle (Haliaeetus leucocephalus) throughout its winter range (Southern 1963; Platt 1976; Joseph 1977; Knight 1981; Anthony et al.1982; Crenshaw 1985; Buehler et al. 1991; Stalmaster and Kaiser 1997). Typically, roosts are used year after year, are located near areas with plentiful food resources, and contain special habitat features such as trees that are older and larger than the average trees in the stand.

The functional and adaptive significance of avian communal roosting is not fully understood; however, several hypotheses have been proposed and applied to bald eagles (Anthony et al. 1982; Stalmaster 1987). First, the food information center hypothesis (Ward and Zahavi 1973; Zahavi 1996) proposes that communal roosting allows individual birds to gather information on potential feeding sites. Knight and Knight (1983) found possible support for this hypothesis at the local level, when they documented a higher incidence of eagles following each other to and from roosts when food was scarce than when it was abundant. McClelland et al. (1982) evaluated the regional applicability of this hypothesis at communal roosts used before and during migration and suggested that the location of "geographically remote temporary food sources" is communicated from experienced to inexperienced eagles.

Second, the energetics hypothesis proposes that communal roosting provides thermoregulatory benefits via the physical structure of the site and/or the presence of other birds. Keister (1981) and Stalmaster (1981) documented that roosts provide a microclimate more favorable to thermoregulation than the microclimate of surrounding areas. However, support for the presence of other eagles providing thermoregulation is mixed. Edwards (1969) and Joseph (1977) found that eagles moved closer to one another at roosts on nights with temperatures below 0° C. Stalmaster (1981) and Stalmaster and Gessaman (1984), however, found no evidence of "huddling" behavior in eagles and thus concluded that the radiative effects of roosting eagles are probably negligible.

Finally, the social function hypothesis proposes that communal roosting serves a social function, for example, to form or renew pairbonds, especially during late winter and early spring (Allen and Young 1982). Curnutt (1992) documented the formation of a potential breeding pair (adult male and female in subadult plumage) at a year-round roost in Florida. Anthony et al. (1982) suggested that communal roosting "aids in the establishment of a social hierarchy or other social functions." Palmer (1988) suggested that pair formation among bald eagles may come about through extended association of pre-breeding individuals.

While the functional aspects of communal night roosting have been well addressed, the functions of the behaviors within the roost have received significantly less attention. Typically, three behaviors occur within roost sites: perch displacements, vocalizations, and perch relocations (Cooksey 1962; Edwards 1969; Jonen 1973; Lish 1973; Platt 1976; Joseph 1977; Grubb 1984; Sabine and Klimstra 1985). Lish (1973) speculated that perch displacements are for maintaining and establishing social dominance hierarchies. Joseph (1977) suggested that vocalizations are a means for spacing eagles within the roost. Perch relocations (hereafter, relocations) have been associated with entrance/exit activity at the roost and with immatures moving to central positions within the roost (Sabine and Klimstra 1985).

We pose and examine two nonexclusive hypotheses for the function of relocations within night roosts. First, bald eagles relocate within the roost to assess foraging success. Second, relocations within the roost are for obtaining an improved microclimate. The objectives of this study were to test the following predictions of these hypotheses. Predictions for each hypothesis are listed in increasing order of the strength of evidence they would provide if supported. For example, Prediction 1 is considered necessary, but not sufficient, to support Hypothesis 1.

If relocations are a means of assessing foraging success, we expect (1) relocations to occur more frequently in the evening than in the morning if eagles are using cues such as full crops or are trying to secure a position near successful foragers for the purpose of following them in the morning (Ward and Zahavi 1973). Furthermore, we expect (2) immature eagles to relocate more frequently and (3) to position themselves near adults either on the same limb or in the same tree. These predictions are based on the finding that immature eagles are less efficient than adults in finding food (Jonen 1973; Shea 1973; Fischer 1985). Finally, we expect (4) relocations to occur more frequently during periods of food scarcity as eagles (especially immatures) seek food resource information from roost-mates.

If relocations are a means of obtaining a more favorable microclimate within the roost, we expect, as in Prediction 1 for Hypothesis 1, (1) relocations to occur more frequently in the evening than in the morning, as eagles seek a protected microclimate for the night. We also predict (2) more relocations to the center of the roost than to its edge, assuming more favorable microclimates exist at the center. Further, we expect (3) that relocation frequency to the center of the roost to be greater during low temperatures and high winds. Finally, we expect (4) relocations to more protected trees (conifers) within the night roost to occur more frequently during periods of low temperatures and high winds. Stalmaster and Gessaman (1984) found that eagles sought protected microclimates by roosting in conifers, which provide "higher ambient temperatures and long wave radiation levels, and lower wind velocity." They found that coniferous roosts were sought 76.3% of the time in comparison to deciduous roosts (23.7%). Swingland (1977) found that rooks (Corvus frugilegus) responded to microclimatic differences within a roost by moving into these favorable microclimates as weather conditions deteriorated.

Criteria for accepting or rejecting Hypotheses 1 and 2 are conditioned by the fact that some predictions are necessary but not sufficient. For example, evidence that supports Predictions 1 and 2 (Hypothesis 1) offers necessary but not sufficient support for Hypothesis 1; additional evidence in support of Prediction 4 would also be needed. Additionally, relocations may have functions unrelated to foraging success or thermoregulation. We pose alternative nonexclusive hypotheses based on descriptive information obtained in this study.

Study area and methods

Observations were conducted at Cottonwood communal night roost, located at river mile 48.5 on the North Fork of the Nooksack River, Whatcom County, Washington (48° 54¢ N, 122° 08¢ W). The roost is a dense stand of mature black cottonwood (Populus trichocarpa) trees interspersed with conifers (< 20% of the roost area) approximately 2 hectares in size. Approximately 25 trees were used for roosting. The roost, located on a flood plain, is bordered on one side by the Nooksack River and on three sides by clearcuts and second growth forest. Due to the roost's size and structure, eagles in different parts of the roost could be concealed from each other. Major tree species along the river system include Douglas fir (Pseudotsuga menziesii), western hemlock (Tsuga heterophylla), western red cedar (Thuja plicata), black cottonwood, bigleaf maple (Acer macrophyllum), and red alder (Alnus rubra). Grand fir (Abies grandis), sitka spruce (Picea sitchensis), and northwestern paper birch (Betula papyrifera) are also found within the riparian habitats. Nonresident bald eagles congregate along the Nooksack River (December-March) to feed on spawned salmon carcasses (Oncorhynchus spp.) and sleep in communal night roosts within 1.5 km of the river. In addition to Cottonwood roost, eagles regularly used Van Zandt (cottonwood trees), Kenney Creek (conifer trees), and Bear Creek (conifer trees) communal night roosts at river miles 40.0, 42.0, 46.5, respectively.

We conducted observational surveys from 22 December 1980 to 9 February 1981 (hereafter, Winter 1) and 22 December 1983 to 15 January 1984 (hereafter, Winter 2). Observers used binoculars and viewed the roost from a distance of >500 m, a distance from which, we believe, observer presence would not alter eagle behavior. Evening roost observations began about 90 minutes before sunset and continued until dark (range 1430-1750 PST for Winter 1 and 1440-1710 PST for Winter 2). Morning counts began 30 minutes before sunrise and continued until most birds left the roost (range 0710-0850 PST for Winter 1 and 0720-0930 PST for Winter 2).

We recorded roosting behaviors (relocations, vocalizations, and perch displacements) during 14 evening surveys in both winters 1 and 2 and 11 and 15 morning surveys during winters 1 and 2, respectively. During evening surveys, in Winter 1 (Winter 2), we recorded 127 (38) immature relocations and 101 (72) adult relocations. During morning surveys, in Winter 1 (Winter 2), we recorded 15 (6) immature relocations and 17 (9) adult relocations. In Winter 2, we were able to age all eagles engaged in a relocation event; however, in Winter 1 we were unable to age 4 eagles during evening relocations and 2 eagles during morning relocations. We used unaged relocations in analyses that required the total number of eagles. A relocation event occurred when a roosting eagle left its perch and flew to a different perch within the roost. When a roosting eagle flew from the roost, this movement was not recorded as a relocation. A perch displacement event occurred when one eagle caused another eagle to leave its perch. Perch displacements constituted 5% (20 of 391) of all morning and evening relocations and were considered as relocations in the analyses. Vocalizations were recorded when an individual eagle called; if two eagles called simultaneously, then two vocalizations were recorded.

We recorded the following information during a relocation event: time, age class (eagles with white heads and tails were classed as adults; all others were classed as immatures), original and final roosting composition (tree type (conifer/cottonwood) and number and age of eagles in each tree), vocalizations during relocation (yes/no), displacement during relocation (yes/no and age class), and location within the roost (center/edge). We defined the central one-third of the roost as the center and the outer two-thirds as the edge.

We recorded the following information during scan-samples (Altmann 1974) at ten-minute intervals: number and age class of roosting eagles, roosting tree type (conifer/cottonwood), and location within the roost (center/edge). We recorded vocalizations continuously and totaled them every ten minutes. We recorded weather conditions (state of weather, temperature, wind speed and direction, cloud cover) at the beginning of each roost survey. We used a measure of food availability (salmon-to-eagle index) calculated for the study area and winters by Knight and Knight (1983) and Knight and Skagen (1988).

Statistical analyses

We used a Wilcoxon rank sum test to compare temperature, food availability, number of roosting eagles, number of relocations per eagle, and number of vocalizations per eagle between winters 1 and 2. Because we wanted to discern potential important trends between winters, we considered findings significant at P £ 0.10. Winters 1 and 2 differed in both food availability and temperature; therefore, we analyzed the data from the two winters separately. We considered findings significant at P £ 0.05 for a priori predictions.

Statistical tests of the food assessment hypothesis (Hypothesis 1)

The first prediction of both the foraging assessment and microclimate hypothesis is that relocations should occur more frequently in the evenings than in the mornings. In Winter 1, pairing an evening survey with its morning counterpart was not possible (surveys either occurred on the same day or within three days of each other); therefore, we used a Wilcoxon rank sum test. In Winter 2, however, pairing an evening survey with its morning counterpart was possible; therefore, we used a two factor mixed model (proc mixed), with time of day as a fixed effect and date as a random effect.

The second prediction of the food assessment hypothesis is that immatures should relocate more frequently than adults. For each age class, the relocation event was weighted using information from scan-samples (number and age of eagles) to calculate a relocation frequency (the total number of relocations per eagle per roost survey). When differences were normally distributed, we used a paired t-test.

The third prediction is that immatures would relocate more frequently to limbs or trees occupied by adult(s), rather than to those occupied by immature(s). To compare relocations to a limb occupied by another eagle (for both age classes), it is important to consider the proportion of birds present in each age class; therefore, we followed Hailman's (1975) procedure for using chi-squared tests.

We categorized immature relocations as to their final social context (in a tree occupied by immature(s) or in a tree occupied by adult(s)) to determine if immature eagles were relocating more frequently to trees occupied by adults. We refer to this type of relocation as occupied-tree relocation. Relocations to trees occupied by both immatures and adults were not included in this analysis, nor were solitary relocations. We used a chi-squared test (one-sample test for goodness of fit) to determine if immatures relocated more frequently to trees occupied by adults based on known proportions of each age class.

The final prediction of the food assessment hypothesis is that relocations should occur more frequently during periods of food scarcity as eagles seek food resource information from roost-mates. We used linear regression to examine the relationship between relocation frequency (total number of relocations per eagle per evening survey) and food availability.

Statistical tests of the microclimate hypothesis (Hypothesis 2)

The tests used for the first prediction of the microclimate hypothesis were discussed above in Prediction 1 (Hypothesis 1). The second prediction for the microclimate hypothesis is that more relocations would occur to the center of the roost than to its edge. We used a chi-squared test (one-sample test for goodness of fit) to determine if eagles relocated more to the roost center than to its edge; expected values were calculated based on proportional availability of center and edge.

The third prediction for this hypothesis is that relocation frequency to the center of the roost is greater during lower temperatures. For each age class and total eagles (immatures/adults/unaged), the number of relocations to the center of the roost was weighted by scan-sample data (number and age of eagles) to calculate a center-relocation frequency (number of relocations to the center per eagle per evening survey). We also used this procedure to calculate edge-to-center relocation frequency. We used linear regression to examine the relationship between both center-relocation frequency and edge-to-center relocation frequency and temperature.

The final prediction of the microclimate hypothesis is that relocations to more protected trees (conifers) within the roost would occur more frequently during periods of low temperatures and high winds. To meet the assumptions of linear regression (i.e., homoscedasticity of residuals), we transformed the count data (number of relocations into conifers per evening survey) using a square root-transformation and regressed them against temperature. We used SAS (SAS Institute 1988) for all statistical analyses. Critical assumptions were met for all statistical tests. Due to pairing of the data or surveys with missing temperature data, sample sizes may vary in different analyses.

Results

Winter 1 was characterized by warmer temperatures and more salmon carcasses relative to eagle numbers than Winter 2 (Table 1). Numbers of bald eagles night roosting at the Cottonwood roost fluctuated throughout both winters, with a peak of 39 eagles on 15 January 1981 and a peak of 36 eagles on 31 December 1983. Low numbers of roosting eagles were observed in late December for both winters, two adult eagles on 26 December 1980 and one adult eagle on 22 December 1983. There were fewer roosting immatures per evening survey in Winter 2 than in Winter 1 (Table 1); numbers of roosting adults were consistent in both winters.

We recorded a total of 232 evening relocations and 34 morning relocations in Winter 1 and 110 evening relocations and 15 morning relocations in Winter 2. The relocation frequency of eagles was significantly lower in Winter 2, due mostly to adults (Table 1).

Table 1. Mean + SD of temperature, food availability, number of roosting eagles, and roost behaviors during Winter 1 and Winter 2 at Cottonwood roost and vicinity. Sample size in the parentheses is the number of roost surveys. P values from the Wilcoxon rank sum test (two-tailed probabilities) between winters are reported.

Hypothesis 1. Relocations are a means of assessing foraging success

In both years, there were significantly more relocations in the evening than in the morning (Wilcoxon rank sum test: S = 33.5, n₁ = 7 surveys, n₂ = 7 surveys, P = 0.006 for Winter 1 and proc mixed: t = -4.86, df = 5, P = 0.002 for Winter 2), consistent with Prediction 1. Immatures did relocate more than adults in Winter 1 (paired t-test: t = 2.53, n = 10 surveys, P = 0.02), consistent with Prediction 2, but not in Winter 2 (paired t-test: t = 1.59, n = 9 surveys, P = 0.08). During the mornings in Winter 1, immatures did not relocate significantly more than adults (Wilcoxon signed-rank test: S = .5, n = 6 surveys, P = 0.5). We did not analyze morning relocation data in Winter 2, due to the small sample size (n = 3).

We found no support for Prediction 3, that immature eagles relocate more often to positions near adults than near immatures. In Winter 1, 22% of 232 evening relocations resulted in an eagle relocating to a limb occupied by another eagle (occupied-limb relocations). In Winter 2, 25% of 110 relocations were occupied-limb relocations. In Winter 1, immatures initiated occupied-limb relocations more than expected based on their proportions (c ² = 4.10, df = 1, P = 0.043). Immatures, however, did not relocate next to adults as predicted, but rather relocated to limbs occupied by other immatures more than expected (c ² = 4.78, df = 1, P = 0.029). Adults relocated next to immatures less than expected based on their proportions (c ² = 4.31, df = 1, P = 0.038). In Winter 2, immature and adult occupied-limb relocations were not significantly different from what was expected based on their proportions.

In Winter 1 (Winter 2), 74% (92%) of all immature relocations have information on the final social context. As with occupied-limb relocations (during Winter 1), we found that immatures relocated to trees occupied by other immatures more than expected based on their proportions (c ² = 4.1, df = 1, P = 0.04). In Winter 2, immature occupied-tree relocations were not significantly different from what was expected based on proportions of immatures and adults (c ² = .11, df = 1, P = 0.74).

We also found no support for Prediction 4, that relocation frequency is greater when food is scarce than when food is abundant. In Winter 1, there was no relationship between relocation frequency (total relocations per eagle per evening survey) and food abundance (salmon-to-eagle index) (linear regression analysis: F_{[1, 12]} = 0.036, P = 0.43) for either age class. In contrast to our expectations, in Winter 2 more relocations occurred during periods of higher food abundance (Figure 1). This positive correlation was significant for both immatures and adults. There was no relationship between salmon availability and maximum number of eagles present in the roost (linear regression analysis: F_{[1, 12]} = 0.40, P = 0.54).

Figure 1. Relationship between relocation frequency (number of relocations per eagle per evening survey) and salmon-to-eagle index (abundance of salmon carcasses relative to number of eagles) at Cottonwood roost, during Winter 2 (22 December 1983 to 13 January 1984) (F_{[1, 11]} = 33.78). This positive correlation was significant for both immatures (F_{[1, 7]} = 12.83, P = 0.005, r² = 0.65) and adults (F_{[1, 11]} = 8.43, P = 0.007, r² = 0.43).

Hypothesis 2. Relocations occur within a roost to obtain an improved microclimate

As described above in Prediction 1 (Hypotheisis 1), there were significantly more evening than morning relocations in both winters. Our data also support Prediction 2, that there would be more relocations to the center of the roost than to its edge. The destination of a relocation was categorized as center or edge in 87% (201 of 232) and 98% (108 of 110) of all evening relocations in winters 1 and 2, respectively. In Winter 1, 74% (179 of 201) of relocations were to the center of the roost and, in Winter 2, 95% (103 of 108) of relocations were to the center. In both winters, more relocations to the center of the roost were observed than expected for Winter 1 and for Winter 2).

We also found support for Prediction 3, during the colder of the two winters. There were more relocations to the center of the roost during low temperatures (Figure 2); this correlation was significant for adults and immatures. In Winter 1, however, there was no relationship between adult relocation frequency and temperature (linear regression analysis: F_{[1, 6]} = 0.01, P = 0.46); in fact, immatures relocated to the center of the roost more during periods of high temperatures (linear regression analysis: F_{[1, 6]} = 7.8, P = 0.02, r² = 0.57).

Figure 2. Relationship between center-relocation frequency (number of relocations to the center of the roost per eagle per evening survey) and temperature at Cottonwood roost, during Winter 2 (22 December to 13 January 1984) (F_{[1, 9]} = 7.3). This negative correlation was significant for adults (F_{[1, 9]} = 6.3, P = 0.02, r² = 0.41) and immatures (F_{[1, 6]} = 3.7, P = 0.05, r² = 0.39).

During both winters, we quantified the origin as well as the destination of a subset of the relocations within the roost for both age classes. Four relocation categories were distinguished: edge to edge (EE), edge to center (EC), center to edge (CE), and center to center (CC). 82% and 97% of the evening relocations of eagles of known age were placed into the four relocation categories in Winter 1 and Winter 2, respectively. During both winters, in this subset of the relocation observations, CC relocations were the most frequent type (45%, n = 186 for Winter 1 and 78%, n = 107 for Winter 2), followed by EC relocations (28% and 17%), EE relocations (21% and 0%), and CE relocations (6% and 5%).

In Winter 1, immatures accounted for 55% of the EC relocations (n = 53) and adults for 45%. In Winter 2, adults accounted for 72% of the EC relocations (n = 18) and immatures for 28%. In Winter 1, there was no relationship between EC relocation frequency and temperature (linear regression analysis: F_{[1, 6]} = 0.07, P = 0.40). However, there were more EC relocations during lower temperatures in Winter 2 (linear regression analysis: F_{[1, 5]} = 27.0, P = 0.002, r² = 0.84). We did not assess this trend for the age classes individually due to small sample sizes.

We found no support for Prediction 4, that relocations to more protected trees (conifers) within the night roost occur more frequently during periods of low temperatures and high winds, during Winter 1. Relocation information pertaining to tree type, however, was not recorded during Winter 2. The conifer component of the Cottonwood roost comprises a small portion of available trees (< 20% of the roost area). Of the 232 evening relocations, 28 (12%) were to conifers, 184 (79%) to cottonwoods, and 20 (9%) were not categorized. Relocations to conifers per survey were not correlated with lower temperatures (linear regression analysis: F_{[1, 9]} = 0.03, P = 0.44). We found no evidence that eagles relocate to conifers more frequently on windy evenings with low temperatures, although our sample size of evenings with low temperatures and winds exceeding 10-16 kph was small (n = 2). However, relocations to conifers were associated with maximum number of eagles present in the roost (linear regression analysis: F_{[1, 9]} = 6.13,

P = 0.035, r² = 0.41).

Additional descriptive information

We sought to describe aspects of roosting behavior that may provide insight on the possible social function of relocations within night roosts. We cover four main aspects: original and final social context of relocating eagles, perch displacements, timing of relocations relative to sunset, and vocalizations.

During both winters, we categorized the original and final social context of relocating eagles as solitary (trees with no eagles) or social (trees occupied by one or more eagles) to determine if eagles were relocating to be closer to other eagles. In Winter 1 (Winter 2), 47% (81%) of all evening relocations have information on both original and final social context. During both winters combined, the context of relocations for both age classes tended to be social; two-thirds of original and final immature relocations (n = 92) and slightly more than half (n = 105) of original and final adult relocations were social. There was no tendency, however, for eagles to secure more social positions after a relocation; the original and final states were independent of one another in both Winter 1 (immatures: c ² = 1.5, df = 1, P = 0.225; adults: c ² = 1.9, df = 1, P = 0.172) and Winter 2 (immatures: c ² = 1.4, df = 1, P = 0.241; adults: c ² = .001, df = 1, P = 0.977).

Displacement of an eagle from its roost perch during a relocation event was infrequent in Winter 1, occurring in 2% (5 of 266) of all morning and evening relocations. However, in Winter 2, perch displacement increased to 12% (15 of 125) of all morning and evening relocations. For both winters combined, information regarding the age of the initiator was determined for 60% (12 of 20) of the perch displacements. Immatures initiated 58% (n = 12) of known-age perch displacements with the remaining 42% initiated by adults. The age of the recipient was determined for 90% (18 of 20) of the perch displacements. Adults were recipients during 75% (15 of 20) of the perch displacements.

We calculated relocation frequency (total number of relocations per eagle per evening survey) relative to sunset for both winters. In general, relocations increased as sunset approached and peaked 10-20 minutes after sunset. This trend was found in both years and for both age classes. In Winter 1, when food was more abundant, both age classes displayed a secondary peak of relocations 50-60 minutes prior to sunset (Figure 3); this secondary peak of relocations was reduced in Winter 2.

Figure 3. Relocation frequency (number of relocations per eagle per survey) of immatures (open circles) and adults (closed circles) relative to sunset at Cottonwood roost for Winter 1 and Winter 2. Minutes before and after sunset are presented in ten-minute intervals.

On average, relocations tended not to be accompanied by vocalizations. In 59% (95 of 162) and 67 % (76 of 114) of assessed morning and evening relocations, there were no accompanying vocalizations for winters 1 and 2, respectively. In Winter 1, the number of morning vocalizations per eagle per survey was higher than evening vocalizations in winters 1 and 2 and morning vocalizations in Winter 2 (Table 1). The number of vocalizations per eagle per evening survey was not related to food availability within winter (linear regression analysis: F_{[1, 11]} = 0.21, P = 0.66 for Winter 1 and F_{[1, 9]} = 0.54, P = 0.48 for Winter 2).

Discussion

Hypothesis 1. Relocations are a means of assessing foraging success

Our data in general do not support this hypothesis. We found support for two predictions that we consider necessary but not sufficient for accepting this hypothesis. Prediction 1 was supported during both winters; relocations at Cottonwood roost were more frequent in the evenings than the mornings. We note, however, that this prediction is also consistent with Hypothesis 2. Prediction 2 was supported by our data for one of the two winters. During Winter 1, as predicted, immature eagles relocated more frequently than adults in the evening but did not do so in Winter 2, when food was less abundant. These findings for Winter 1 are consistent with Sabine (1981) who also found that most relocations were attributed to immatures, but not with Cooksey (1962) who reported no correlation of frequency of relocations and age of eagles.

Our evidence does not support Prediction 3. Immatures in both winters did not reposition to either limbs or trees occupied by adults more than expected. In fact, immatures in Winter 1 positioned themselves next to immatures more than expected based on their proportions.

Prediction 4, that relocations are more frequent during food scarcity, is our strongest test of Hypothesis 1, and we consider support of this prediction to be sufficient for acceptance of the hypothesis. Our evidence not only did not support the prediction, but was in the opposite direction of our expectations during the winter of lower food abundance (Winter 2). During periods of food scarcity in Winter 2, eagles relocated less frequently, perhaps to conserve limited energy.

Hypothesis 2. Relocations occur within a roost to obtain an improved microclimate

Our evidence suggests that eagles may relocate to obtain an improved microclimate during winters with lower temperatures. Of the four predictions for this hypothesis, we consider two, Predictions 1 and 2, to be necessary but not sufficient for accepting the hypothesis, and Predictions 3 and 4 as stronger tests of this hypothesis. Consistent with Prediction 1, as discussed above, evening relocations were more frequent than morning relocations in both winters, a finding also consistent with Hypothesis 1. In support of Prediction 2, there were more evening relocations to the center of the roost than to its edge in both winters. Further, Prediction 3, that relocation frequency to the center of the roost is greater during low temperatures, was supported in the colder winter (Winter 2). For both age classes in Winter 2, movements to the center of the roost, including edge to center, were associated with low temperatures. In Winter 1, we found no support for this prediction in adults and evidence that immatures relocated to the center of the roost more during high temperatures, a finding in the opposite direction of our prediction. Predictions 2 and 3 are based on the assumption of a more favorable microclimate in the center of the roost, an assumption we did not test. However, we feel that this is a reasonable assumption based on documented within-roost microclimatic variations (Swingland 1977).

During Winter 1, we found no support for Prediction 4, that relocations to conifers would increase during colder temperatures; rather, relocations to conifers were associated with maximum number of eagles present in the roost, suggesting a possible relationship between relocations to conifers and available perches in the roost. Unfortunately, we do not have relocation information pertaining to conifers for Winter 2.

Also consistent with this hypothesis is the peak of relocations occurring 10-20 minutes after sunset, which may occur as eagles seek more protected microclimates for the night. We feel that this hypothesis warrants further testing.

Relocations may have functions unrelated to foraging success or thermoregulation

Because our study rejects Hypothesis 1 and supports Hypothesis 2 for a limited number of relocations, we pose alternative social and nonsocial functions for consideration. We found that nearly 60% of the relocations with origins and destinations categorized as center or edge of roost were center to center movements. The center to center relocations are not explained by the microclimate hypothesis unless several microclimates exist at the center of the roost. Rather, eagles may be relocating within the center of the roost to refine their positions relative to access (ease of flight out of the roost), view, or branch structure (Stalmaster 1987).

Our data suggest that relocations may aid in establishing or maintaining foraging associations among immature eagles when food is relatively abundant, such as in Winter 1. There are benefits of group foraging in eagles when food is concentrated and abundant. For example, birds may gain experience in food finding and piracy techniques that would benefit them when food is scarce (Knight and Skagen 1988), and immatures foraging in groups are able to displace solitary adults attempting to maintain feeding territories, thereby allowing access to food (SKS and RLK personal observation). Such foraging associations may be established or maintained by means of relocations to limbs or trees occupied by other eagles, such as in Winter 1 when immature eagles positioned themselves next to other immatures more than expected based on their proportions. The increase in morning vocalizations heard in Winter 1 is also consistent with this explanation. During food scarcity, however, we might expect efforts to maintain foraging associations to subside if costs of group foraging due to competition increase and outweigh the benefits; we found no evidence suggesting the maintenance of foraging associations in Winter 2. Other observed occupied-limb relocations (immature-adult, adult-adult, and adult-immature) as well as occupied-tree relocations are not addressed by this explanation and suggest further social aspects to these types of relocations.

While our data indicate that perch displacements are not an important outcome of relocations, some (2-12%) relocations did result in displacements, especially in the winter of lower food abundance. The constraint of lower food availability during Winter 2 may have increased the need to establish and maintain social dominance hierarchies via relocations. In Winter 2, immatures initiated 60%, and adults were the recipients in 75% of the displacements. The initiating immatures may have been females displacing the smaller adult males, thereby establishing the hierarchy. Knight and Skagen (1988) found that the larger body size of females corresponded with dominance at feeding sites more so than did age. Consistent with this speculation, Burger et al. (1977) found that larger species almost always won in perch displacements at mixed species communal roosts.

The need to conserve energy during periods of food scarcity may constrain the occurrence of relocations within night roosts. In our study, eagles relocated less often in the winter of lower food availability, Winter 2, than in the winter of relative food abundance, Winter 1. Furthermore, eagles relocated less frequently when food was most scarce during Winter 2. The timing of relocations relative to sunset also differed between the two winters, exhibiting a secondary peak of relocations an hour before sunset when food was abundant but not when food was scarce. Both age classes exhibited this reduction in early relocations, but it was most accentuated in adults.

In summary, our study indicates that eagles at the Cottonwood roost may relocate within the night roost to obtain a favorable microclimate during winters with cold stress and food stress. Our data do not support the hypothesis that relocations are a means of assessing foraging success of conspecifics. Although the microclimate hypothesis may explain the function of some relocations, it does not explain most of them. It is likely, based on our findings, that relocations may have several nonsocial and social functions, including refining position in the roost for improved access, view, or branch structure, establishing and/or maintaining foraging associations, and establishing and/or maintaining social dominance hierarchies when food is scarce. Furthermore, we suggest that relocation behavior is constrained by the need to conserve energy when food is scarce.

Acknowledgments

We are grateful to the following individuals for providing field assistance: L. Anderson, R. Anderson, K. Boehlein, S. Holdeman, A. Knight, N. Marr, B. Orians, G. Orians, E. Rose, K. Rose, B. Rohwer, S. Rohwer, B. Wood, and L. Young. We thank R. and E. Anderson for housing and logistical support. Funding for field work was provided by the Jefferson E. Davis Fund of the University of Wisconsin, the Frank M. Chapman Memorial Fund of the American Museum of Natural History, and the O. C. Marsh Fund of the National Academy of Sciences. Support during manuscript preparation was provided by Colorado State University and the Biological Resources Division of U. S. Geological Survey.

References

Allen, J.L., and Young, L.S. 1982. An Annotated Bibliography of Avian Communal Roosting. Olympia, Washington: Washington Department of Game.

Altmann, J. 1974. Observational study of behavior: sampling methods. Behaviour, 49: 227-267.

Anthony, R.G., Knight, R.L., Allen, G.T., McClelland, B.R., and Hodges, J.I. 1982. Habitat use by nesting and roosting bald eagles in the Pacific Northwest. In: Forty-Seventh North American Wildlife and Natural Resources Conference, Portland, Oregon, March 26-31, 1982. Washington DC: Wildlife Management Institute pp. 332-342.

Buehler, D.A., Mersmann, T.J., Fraser, J.D., and Seegar, J.K.D. 1991. Winter microclimate of bald eagle roosts on the northern Chesapeake Bay. Auk, 108: 612-618.

Burger, J., Gladstone, D., Hahn, D.C., and Miller, L.M. 1977. Intra- and interspecific interactions at a mixed species roost of ciconiiformes in San Blas, Mexico. Biol. Behav. 2: 309-327.

Cooksey, B.F.Jr., 1962. A winter population of the bald eagle (Haliaeetus leucocephalus) in northeastern Oklahoma. M.S. thesis, Kansas State College, Pittsburg.

Crenshaw, J.G. 1985. Characteristics of bald eagle communal roosts in Glacier National Park, Montana. M.S. thesis, University of Montana, Missoula.

Curnutt, J.L. 1992. Dynamics of a year-round communal roost of bald eagles. Wilson Bull. 104: 536-540.

Edwards, C.C. 1969. Winter behavior and population dynamics of American eagles in western Utah. Ph.D. thesis, Brigham Young University, Provo.

Fischer, D.L. 1985. Piracy behavior of wintering bald eagles. Condor, 87: 246-251.

Grubb, T.G. 1984. Winter activity of bald eagles (Haliaeetus leucocephalus) at Navajo Lake, New Mexico. Southwest. Nat. 29: 335-341.

Hailman, J.P. 1975. Analysis of aggression in White-throated Sparrow types of different proportions. Bird-Banding, 46: 236-240.

Jonen, J.R. 1973. The winter ecology of the bald eagle in west-central Illinois. M.S. thesis, Western Illinois University, Macomb.

Joseph, R.A. 1977. Behavior and age class structure of wintering northern bald eagles (Haliaeetus leucocephalus alascanus) in western Utah. M.S. thesis, Brigham Young University, Provo.

Keister, G.P.Jr. 1981. Characteristics of winter roosts and populations of bald eagles in the Klamath Basin. M.S. thesis, Oregon State University, Corvallis.

Knight, S.K. 1981. Aspects of food finding and avoidance behavior of wintering bald eagles. M.S. thesis, Western Washington University, Bellingham.

Knight, S.K., and Knight, R.L. 1983. Aspects of food finding by wintering bald eagles. Auk, 100: 477-484.

Knight, R.L., and Skagen, S.K. 1988. Agonistic asymmetries and the foraging ecology of bald eagles. Ecology, 69: 1188-1194.

Lish, J.W. 1973. Status and ecology of bald eagles and nesting of golden eagles in Oklahoma. M.S. thesis, Oklahoma State University, Stillwater.

McClelland, B.R., Young, L.S., Shea, D.S., McClelland, P.T., Allen, H.L., and Spettigue, E.B. 1982. The bald eagle concentration in Glacier National Park, Montana: origin, growth, and variation in numbers. Living Bird, 19: 133-155.

Palmer, R.S. (ed), 1988. Handbook of North American birds, vol. 4 (part 1). Yale University Press, New Haven Connecticut.

Platt, J.B. 1976. Bald eagles wintering in a Utah desert. American Birds, 30: 783-788.

Sabine, N. 1981. Ecology of bald eagles wintering in southern Illinois. M.S. thesis, Southern Illinois University, Carbondale.

Sabine, N., and Klimstra, W.D. 1985. Ecology of bald eagles wintering in southern Illinois. Trans. Ill. Acad. Sci. 78: 13-24.

SAS Institute, 1988. SAS/STAT^® user's guide, release 6.03 ed., SAS Institute Inc., Carey, North Carolina.

Shea, D.S. 1973. A management-oriented study of bald eagle concentrations in Glacier National Park. M.S. thesis, University of Montana, Missoula.

Southern, W.E. 1963. Winter populations, behavior, and seasonal dispersal of bald eagles in northwestern Illinois. Wilson Bull. 75: 42-55.

Stalmaster, M.V. 1981. Ecological energetics and foraging behavior of wintering bald eagles. Ph.D. thesis, Utah State University, Logan.

Stalmaster, M.V. 1987. The Bald Eagle. Universe Books, New York.

Stalmaster, M.V., and Gessaman, J.A. 1984. Ecological energetics and foraging behavior of overwintering bald eagles. Ecol. Monogr. 54: 407-428.

Stalmaster, M.V., and Kaiser, J.L. 1997. Winter ecology of bald eagles in the Nisqually River drainage, Washington. Northwest. Sci. 71: 214-223.

Swingland, I.R. 1977. The social and spatial organization of winter communal roosting in rooks (Corvus frugilegus). J. Zool. Lond. 182: 509-528.

Ward, P., and Zahavi, A. 1973. The importance of certain assemblages of birds as "information centres" for food-finding. Ibis, 115: 517-534.

Zahavi, A. 1996. The evolution of communal roosts as information centers and the pitfall of group selection: a rejoinder to Richner and Heeb. Behav. Ecol. 7: 118-119.