HABITAT USE PATTERNS AND BEHAVIORAL ECOLOGY IN THE PACIFIC NORTHWEST. A Thesis. Presented to. The Faculty of Moss Landing Marine Laboratories

Transcription

1 HABITAT USE PATTERNS AND BEHAVIORAL ECOLOGY OF KILLER WHALES (Orcinus orca) IN THE PACIFIC NORTHWEST A Thesis Presented to The Faculty of Moss Landing Marine Laboratories and the Department of Biological Sciences San Jose State University In Partial Fulfillment of the Requirements for the Degree Master of Science in Marine Science By James Robert Heimlich-Baran May 1987

2 iii TABLE OF CONTENTS ACKNOWLEDGMENTS.... INTRODUCTION.... Natural History of the Species.... Killer Whales in the Pacific Northwest.... METHODS.... Mapping.... Behavioral Sampling.... Feeding.... Travel.... Rest..... Socializing.... Analytical Methods.... RESULTS.... Distribution of Resident Whale Observations.... Patterns of Resident Behavior and Habitat Use.... Feeding Areas.... Quadrants 1, 2, and Quadrants 6 and Quadrant Georgia Strait.... Puget Sound.... Summary.... Travel Areas.... Quadrant Quadrant Pooled Quadrants 7, 8, and Central Georgia Strait.... Summary Rest Areas.... Socializing Areas.... Quadrants 10 and Juan de Fuca Strait.... Summary.... Distribution of Transient Whale Observations.... Patterns of Transient Whale Behavio:r: and Habitat Use DISCUSSION Cooperative Foraging.... Residents and Transients.... LITERATURE CITED.... Page v

3 iv LIST OF TABLES Number Page 1. Regional Distribution of Resident Whale Observations Annual Distribution of Resident Whale Observations Monthly Distribution of Resident Whale Observations Counts of Behavioral Observations for Haro Strait Quadrants Counts of Behavioral Observations for Regions and Subregions LIST OF FIGURES Number Page 1. Bathymetric Map of the Study Area Map of Regions and Subregions of the Study Area Distribution Map of Resident Whale Encounters Bathymetric Map of Haro Strait Quadrants Behavior Distribution for All Areas Chi-square Values for Feeding Areas Chi-square Values for Travel Areas Chi-square Values for Rest Areas Chi-square Values for Socializing Areas Distribution Map of Transient Whale Encounters... 43

4 v ACKNOWLEDGMENTS These data were collected under the auspices of Moclips Cetological Society, with support from National Marine Fisheries Service, the Marine Mammal Commission, and the Packard Foundation. My foremost thanks go to Ken Balcomb (Center for Whale Research, Inc.), Camille Goebel, and Rick Chandler for getting me started with killer whales in Rich Osborne, Sara Heimlich-Baran, Nancy Haenel, and Fred Felleman were all intimately involved with the data and ideas presented in this thesis and I thank them all for their great help and friendship. Bernd Wursig has been my continuing friend, advisor, teacher, and reviewer over the last five years and I have learned immensely from him. My other committee members, Drs. Ken Norris and Michael Foster, also provided much help through several drafts of this thesis and I thank them sincerely. I also want to thank Moss Landing Marine Laboratories for providing a stimulating working environment over the last five years. Finally, I must thank Sara Heimlich-Baran for her on-going support and confidence.

5 1 INTRODUCTION This paper presents eight years of observations of habitat use patterns in Pacific Northwest killer whales (Orcinus orca). Killer whales appear to use different physiographic regions of their habitat in unique behavioral ways. I will show that: 1) resident whales tend to feed more in areas of high relief subsurface topography along salmon migratory routes, and may use these geographic features to increase their feeding efficiency; 2) transient whales primarily feed in shallow protected areas around concentrations of their prey, harbor seals; 3) whales travel across deep, featureless areas, moving from one feeding area to another; 4) whales rest depending on the previous sequence of behaviors, not in specific geographic locations; and 5) whales tend to play in open water areas or adjacent to feeding areas. These consistent relationships between behavior and habitat characteristics over an eight year period provide clear examples that these whales possess a high degree of familiarity with a complex and varying environment. These patterns of behavior most likely represent cultural mechanisms that have been learned through trial and error experiences leading to successful foraging strategies.

6 2 Before presenting these findings, I want to review a few salient points about the biology and behavior of killer whales, both worldwide and in the Pacific Northwest. The important ideas are that: 1) worldwide killer whale distribution is centered around areas of productivity; 2) feeding preferences are diverse worldwide, but localized populations specialize on seasonally abundant prey items; 3) killer whales in the Pacific Northwest live in stable pods which associate in three isolated communities; and 4) there are two distinct forms of killer whales, residents and transients, which show extensive differences in social structure, behavior, and possibly even genetics. This information will help to understand the habitat use patterns of Pacific Northwest killer whales. Natural History of the Species Killer whales, (Orcinus orca), are distributed throughout all oceans (Perrin 1982), but concentrations are generally greater at higher latitudes (Matkin and Leatherwood 1986; Norris and Prescott 1961; Perrin 1982), and are usually in areas of high productivity. The seasonal occurrence of killer whales has been correlated with the occurrence of prey species: southern elephant seal and penguin species in the south Indian Ocean (Candy et al 1978); the northern elephant seal at San Benites Island,

7 3 California (Norris and Prescott 1961); herring in the northeast Atlantic (Jonsgaard and Lyshoel 1970); and salmon in the Pacific Northwest (Heimlich-Baran 1986). Thus, the distribution of killer whales often appears to be closely tuned to the distribution of their prey. Killer whales are noted for their predation on other marine mammals (Eschricht 1866; Scammon 1874). However, observations of attacks and stomach contents have indicated that killer whales feed on a diverse array of marine vertebrates and invertebrates, ranging from baleen whales, small toothed whales, and pinnipeds to fish, sea turtles, birds and cephalopods (Caldwell and Caldwell 1969; Hancock 1965; Jonsgaard and Lyshoel 1970; Nishiwaki and Handa 1958; Rice 1968; Smith et al 1981; Steltner et al 1984; Straneck et al 1983; Tarpy 1979). Because of the wide range of prey items, killer whales have been considered by some to be generalized, opportunistic feeders (Yablokov et al 1975) A closer examination suggests that individual populations of killer whales may actually specialize on preferred prey for their given area, perhaps even shifting prey preferences in response to seasonal variations in prey abundances and catchability (Felleman 1986). Most detailed accounts of feeding behavior describe groups of killer whales hunting in a coordinated fashion (see review in

8 4 Wursig 1986) similar to that described for a wolf pack (Mech 1970) and other social carnivores (Kruuk 1972; Schaller 1972). This behavior is dependent on a high degree of communication and coordination within the group (Wursig 1986) and is an important means to increase feeding efficiency on either individual, large marine mammals or large schools of small fish (Felleman 1986). Killer Whales of the Pacific Northwest Killer whales of the inland marine waters of Washington and British Columbia travel in long-term, cohesive groups called pods. All individuals of this population have been photographed and identified by recognition of pigmentation patterns and scars on the dorsal fin and back (Balcomb and Bigg 1986; Balcomb et al 1980, 1982; Bigg 1982; Bigg et al 1976). Capture photographs from as early as 1965 show that many of the same groups of individuals have been traveling together for over twenty years. There are three apparently isolated communities of whales that inhabit the Washington and British Columbia coasts during the spring to fall: 1) a northern ''resident'' community inshore of the northern half of Vancouver Island, totalling 12 pods of 135 whales; 2) a southern resident community inshore of the southern half of Vancouver Island, totalling 3 pods of 79 whales; and 3) a

9 5 "transient" community, totalling 15 pods of 47 whales, whose members were seen sporadically throughout the entire region (Bigg 1982). There are numerous lines of evidence which show distinct behavioral differences between residents and transients, in terms of distribution, seasonal occurrence, acoustic dialects, feeding behavior, and associations with tidal currents, as well as suggestions of genetic differences such as dorsal fin morphology and pigmentation (Balcomb and Bigg 1986; Balcomb et al 1980; Bigg 1982; Felleman 1986; Ford and Fisher 1982, 1983). Duffield (1986) has proposed that such genetic differences should be apparent in the highly variable chromosomal markers which characterize Orcinus, but this work is still in progress. The information to date shows the communities to be geographically or socially isolated stocks, implying reproductive isolation between residents and transients. This paper will examine the habitat use patterns of resident and transient killer whales in the inland marine waters of Washington and southern British Columbia. The region is characterized by a highly varied subsurface topography (Figure 1). I will show how the whales behave differently in different parts of their environment. This

10 6 ISLAND 49'N DEPTH CONTOURS PACIFIC OCEAN M M M )300 M kilometers miles 24'W Figure 1. Bathymetric map of the study area.

11 7 will show how complex, learned patterns of behavior can be important in the behavioral ecology of these whales. METHODS Killer whales were observed in the inland marine waters of Washington and British Columbia between 47.0 degrees and 49.3 degrees north latitude and degrees and degrees west longitude from 1 April 1976 to 31 December Research effort was centered on Haro Strait in the San Juan Islands (Figure 2). Whales were photographed and identified from an identification catalog developed in conjunction with M.A. Bigg. This analysis will only deal with confirmed observations of southern resident and transient pods. The main effort in data collection was to maximize the time with whales, not to determine geographical distribution with randomly collected transect data. Sighting effort was present year-round, but winter data collection was limited to only a few observers. The primary assumption of this thesis was that whale behaviors were sampled randomly. The times when whales were observed (although not random with respect to location and season) were representative of the full repertoire of behaviors which occur during the spring to fall season in the study area.

12 8 ISLAND 49'N Belling!lam SAN JUAN ISLANDS PACIFIC OCEAN kilometers w miles Figure 2. area. Map of regions and subregions of the study

13 9 The field methods involved 1) accurate mapping of whale routes and 2) categorizing the predominant behavior of the whales. These parameters were updated every 15 minutes. Analytical methods involved: 1) chi-square testing for significant differences between observed and expected behavior distributions; 2) heterogeneity chi-square testing to determine the validity of pooling observations of adjacent quadrants; and 3) subdivision of chi-square tables to isolate the main behavioral contributors to significance. I will briefly explain each of these methods. Mapping Whales were tracked from 5-6 m motorboats or from m sailboats. Each day of tracking an identified pod was termed an encounter day. Exact whale locations were determined from compass triangulations on nearby shorelines and marker buoys using a Morin Opti2 hand bearing compass. The study area was divided into a grid of 441 quadrants. Each quadrant was approximately 4.6 km square. Whale routes were plotted on this grid and quadrant locations were recorded using a 15 min interval scan sampling method (Altmann 1974). The quadrant which contained the majority of whales for the majority of the time period was recorded. This occasionally required locations of individual whales

14 10 to be averaged to account for the general movement of the entire group. Behavioral Sampling Behavioral sampling was initially conducted on a continuous basis using the methodology described for these killer whales by Osborne (1986). However, for this analysis, a 15 min sampling regime was overlaid on top of these data in order to generate frequency counts of equal intervals, thus standardizing the data and making them comparable with the location data. Sampling involved recording the dominant behavior of all whales present. In an attempt to objectively base behavior categories on recognizable surface activities, Osborne (1986) devised a method of combining quantifiable parameters visible at the surface such as group size, pod composition, spacing of individuals and subgroups, group speed, and directionality of movement. Osborne (1986) analyzed these combinations and defined eight functional behavioral categories: foraging, percussive foraging, milling, travel, percussive travel, rest, play, and intermingling. It is important to understand each of these behavioral categories in order to understand how killer whales behave differently in different parts of their

15 11 environment. Much of the following behavior description is a summary of Osborne (1986). Feeding Foraging, percussive foraging, and milling were all considered to be indicative of feeding and each is thought to represent a unique feeding strategy. Foraging was defined as group directional travel interspersed with occasional breaks of apparently random, non-directional milling. Whales often travelled in groups wider than long, which is an efficient searching formation (Norris and Dahl 1980a). Percussive foraging had characterist_ics similar to foraging, travel interspersed with milling; but percussive foraging was performed in much tighter coordinated flank formations and was interspersed with surface splashing behaviors such as lob-tailing and pectoral fin slapping. Percussive behaviors have been shown to constitute herding behavior of dusky dolphins accumulating prey into higher densities, as well as serving a communicatory function to maintain group coordination (Wursig and Wursig 1980). t1illing was defined as long-term non-directional swimming by all group members. It appeared to be individual feeding on congregated schools of fish. These behaviors could occur in the order listed in the progression of feeding

16 12 from searching to collecting to capture or as prey densities increased, but this has not been tested to date. Travel Travel was defined as continuous directional swimming by all group members. Percussive travel was this same behavior interspersed with splashing behaviors and often characterized by high speed (greater than nine km/hr). Animals often travelled in line-abreast formation. This behavior may also serve some coordinated herding function, but it is not interspersed with feeding-related milling, suggesting that prey may be herded over large distances, perhaps aggregated from diverse small groups, before feeding begins. Percussive behaviors during traveling were often observed in peripheral subgroups with calves, and may simply represent calf play. On the other hand, percussive behaviors may also be serving some social signaling function in dispersed travel groups. Rest was defined as slow directional swimming or motionless hovering, and was characterized by short, synchronous respiration cycles. Killer whales were silent during this behavior (Hoelzel and Osborne 1986), as has been reported for spinner dolphins (Norris and Dohl 1980b)

17 13 Socializing The final two behaviors may be clumped as socializing behaviors (including "sexual'' behaviors). These behaviors seemed to occur year-round, and were not limited to any breeding season. Play consisted of a wide ranging group of individual behaviors usually identified by repeated breaching out of the water by most group members. Intermingling behavior was a unique behavior rarely observed and was characterized by whales from more than one pod congregated in tight milling clumps in body contact with each other. This behavior was apparently socializing behavior between two or more pods that occurred when pods met after being apart for some time, and may function similarly to ''rallying'' groups of spinner dolphins grouping up to feed (Norris and Dohl 1980b) or to ''greeting ceremonies" of wild dogs (Estes and Godard 1967). Analytical Methods Whale routes and behavior categories for each encounter were sorted by quadrant and behavior, thus pooling all observations of a given behavior in a given quadrant throughout the study period. In some cases sample sizes for individual quadrants were too small; observations were then clumped into eight major regions (Figure 2). In order to examine differences between the distribution of

18 14 behaviors in specific heavily used quadrants (or regions) and the distribution of behaviors for the total study area, chi-square goodness of fit testing was applied (Zar 1974). The intrinsic null hypothesis was that the behavior distribution for each quadrant should be the same as that for the study area as a whole (i.e. that behaviors were distributed uniformly throughout the area). The occurrence of unique behavior distributions for specific quadrants would imply that the whales were using specific areas for particular behavioral purposes. Adjacent quadrants were tested with heterogeneity chi-square analysis to examine the statistical validity of pooling observations (Zar 1974). Chi-square tables were then subdivided to locate which behaviors were responsible for overall significant differences (Zar 1974). The behaviors with the largest chi-square values were sequentially removed until the remaining behaviors were not significantly different from the overall behavior distribution. In many cases, more than one behavior was responsible for significant chi-squares. These could be ranked in order of importance. Also, significance in chi-square analyses can be due to observations either greater than or less than expectation. For this analysis, I will focus on those behaviors which occurred more than expected.

19 15 RESULTS Before presenting results on patterns of behavior and habitat use, it is important to examine the seasonal and geographic distribution of the data base. This places some constraints on the overall applicability of the habitat use results. Distribution of Resident Whale Observations Resident whales were observed on 239 vessel encounters for hrs (mean of 4.1 hrs/enc) during the study period from 1976 through They were tracked for 6660 km through 177 (40%) of the 441 study quadrants. The number of encounters per quadrant varied from one to 141 (Figure 3) and shows the distribution of the data base during the study. The Haro Strait region accounted for almost two-thirds of the total observations (Table 1). This region consisted of 13 quadrants ranging from the southern shores of San Juan Island north along the U.S. - Canadian border for 56 km to southern Georgia Strait (Figures 2 and 4). This Strait is also a migratory route for a majority of Fraser River salmon returning to spawn (Groot et al 1984). Whales were never observed in Hood Canal or the Whidbey Island Basin (although there were occasional sighting reports from those regions). The remainder of the

20 16 ISLAND 49'N RESIDENT WHALES NO. OF ENCOUNTER DAYS PACIFIC OCEAN Figure 3. encounters. D 0 ~ 1-5. ~ 6-10 ) kilometers 'W miles Distribution map of resident whale 47'N

21 17 (, /. >, \~;;? r< Or ebb lalond -::;_~:?\;)~, DEPTH CONTOUR~... ' ' - ' ( :; '.. )' ' Figure 4. Bathymetric map of Haro Strait quadrants.

22 18 observations were spread nearly equally throughout the other four regions (Table 1). Table 1. Regional distribution of resident whale observations. =========================================================== REGION HOURS % Haro Strait Georgia Strait Juan de Fuca Strait San Juan/Gulf Islands Puget Sound TOTAL Whales were seen in all years of the study period (Table 2). Two-thirds of the observations are from the first three years of the study because of greater effort. Behavior distributions from the individual years were not significantly different from the overall behavior distribution and were homogeneous (Heterogeneity chi-square test, P >.05), so data from all years were pooled for further analysis.

23 19 Table 2. observations. Annual distribution of resident whale =========================================================== YEAR HOURS % TOTAL Whales were seen during all months of the study (Table 3). However, three-fourths of the observations were in the months of July through September. Although sighting effort was not thoroughly quantified for these vessel observations, there was characteristically a decrease in effort during winter because boats were usually hauled out of the water. However, additional sighting sources (such as shore observations and public sighting reports, Heimlich-Baran 1986) show similar seasonal distribution patterns. In spite of limited winter observations, all of these sources have documented the occurrence of small-scale peaks in killer whale distribution which are closely timed to the winter occurrence of juvenile and resident salmon as well as migrating steelhead trout (Felleman et al In press). This implies a sufficient level of winter coverage

24 20 to conclude a predominantly spring to fall distribution for southern resident whales in the study area. Table 3. Monthly distribution of resident whale observations. =========================================================== MONTH HOURS % January February March April May June July August September October November December TOTAL Patterns of Resident Behavior and Habitat Use The percentage distribution of behaviors for the entire study period is shown in Figure 5. The three feeding behaviors comprised 47% of the whales' time. The two travel behaviors totalled 25% of the time, while rest and the two socializing behaviors occurred with approximately equal frequencies (13% and 15%, respectively). This frequency distribution of behaviors, because of the large and geographically diverse sample from which it came, was accepted as representative for these

25 21 FEEDING TRAVEL REST SOCIALIZING 22 N = 985 hrs. 20 PERCENT OF TOTAL OBSERVATIONS Figure 5. Behavior distribution for all areas. Behaviors listed are foraging (F), percussive foraging (PF), milling (M), travel (T), percussive travel (PT), rest (R), play (P), and intermingling (IM). \

26 22 whales. All other behavior distributions from specific subsamples were compared to it with goodness of fit testing. I will present the habitat use results for each of the specific behaviors. There were similarities among the areas which shared significant occurrences of particular behaviors. All behavior counts are presented in Table 4 for the 13 Haro Strait quadrants, and in Table 5 for the regions and subregions of the study area. Table 4. Counts of behavioral observations for Haro Strait quadrants. Each count represents a 15 min scan sample. Behaviors listed are foraging (F), percussive foraging (PF), milling (M), travel (T), percussive travel (PT), rest (R), play (P), and intermingling (IM). =========================================================== QUADj~ F PF M T PT R p IM TOTAL 1 & TOTAL

27 23 Table 5. Counts of behavioral observations for regions and subregions. Each count represents a 15 minute scan sample. Behavior abbreviations are given in Table 4. =========================================================== AREA F PF M T PT R p IM TOTAL Georgia 84 Delta 60 Central 19 South Hare San Juan/Gulf Juan de Fuca North South Puget TOTAL Feeding Areas Feeding behavior was observed significantly more than expected in three areas of Hare Strait: quadrants 1, 2, and 3, quadrants 6 and 17, and quadrant 12, and in two major regions of the study area, Georgia Strait and Puget Sound. Quadrants 1. 2, and 3. Whales were observed in quadrants 1, 2, and 3 for hours, or 22% of the total sample period (Table 4). Quadrants 1 and 2 were homogeneous in behavior distribution and could be pooled 2 (heterogeneity X = , df = 14, P >. 05). There were significantly greater observations of foraging behavior and

28 24 a lack of observations of percussive travel (X 2 = 40.9, df = 7, P <.001) in these quadrants (Figure 6) These quadrants are located at the southern mouth of Haro Strait along the shores of San Juan Island, and contain some of the most varied bathymetry of the region (Figure 4). Quadrant 3 had significantly high amounts of milling behavior and a lack of percussive travel and play (X , df = 7, P <.001; Figure 6). These three quadrants are the beginning of a deep trough (200 m) which opens up from the shallower (55 m deep) waters of Juan De Fuca Strait. There are numerous seamounts in quadrants 1 and 2 rising up over 100 m from the seafloor, and complex tidal currents which cause extensive mixing of the waters (Thomson 1981). Quadrant 3 consists of a steep slope that drops to over 200 m deep less than a kilometer offshore of San Juan Island (Figure 4). Quadrants 6 and 17. The second area of increased feeding in Haro Strait was in adjacent quadrants 6 and 17 (Figure 4). These areas were primarily rest areas, but their role as feeding areas will be discussed here. Resident whales were observed in these two quadrants for hours (Table 4). Percussive foraging was observed significantly more than expected in quadrant 6, although ranked second to rest (x 2 = 99.4, df = 7, P <.001; Figure

29 25 QUADRANTS 1 & 2 N hrs. sou <EXP 10 L L-~~~--~----~ <EXP SOUAR <EXP CHI- CHI CHI SQUARE <EXP 10 L ~~---L~~----~ Figure 6. Chi-square values for feeding areas. Behavior abbreviations are given in Figure 5. Subdivision chi-square significance: *, P < 0.05, behaviors responsible for overall chi-square significance.

30 26 8). Milling, also ranked second to rest, occurred 2 significantly more than expected in quadrant 17 (X = 28.0, df = 7, P <.001; Figure 8). Quadrant 6 is characterized by the sheer cliffs along the eastern shoreline which drop to depths of 270 m a few hundred meters from shore (Figure 4). At the northern border of the quadrant there is a 20 m deep kelp-covered sill rising up almost 200 m in elevation and jutting 300m into the strait. Quadrant 17 is ringed by islands and shallow (< 100 m deep) channels on three sides and drops down to 200 m deep on the western edge along quadrant 6. There is a large kelp-covered seamount rising from 170m to 10m in the center of the quadrant, as well as a number of other smaller reefs (Figure 4). Quadrant 12. The third area of feeding in Haro Strait was in quadrant 12, located on the boundary between Haro Strait and Georgia Strait (Figure 4). Whales were observed during 18.5 hrs on 33 encounters (Table 4). Percussive foraging occurred significantly more than expected, although it ranked third in importance after the socializing behaviors of play and intermingling (X 2 = 26.7, df = 7, P <.001; Figure 9). The bathymetry of this area is similar to that of quadrant 6. There is a prominent kelp-covered sill which juts across the center of the quadrant. This reef is steep on both the Haro and Georgia

31 27 Strait sides, generating strong upwelling currents at both ebb and flood tides (Thomson 1981). The tidal currents are markedly strong and produce numerous tide rips, eddies and whirlpools at periods of peak tidal flow (Thomson 1981). Georgia Strait. In Georgia Strait (Figure 2), percussive foraging occurred significantly more than expected, although it only ranked fourth in importance. Increased observations of percussive travel and intermingling, and a lack of rest were the main contributors to the significant chi-square (x 2 = 108.9, df = 7, P <.001; Figure 7). Whales were observed in Georgia Strait for hours (Table 5). Three-fourths of the observations of percussive foraging in Georgia Strait were in the Fraser River delta (although this subregion only accounted for 57% of the total observations). Subdivision of Fraser River chi-square values (X 2 = 99.9, df = 7, P <.001) showed that a lack of observations of milling, travel, rest, and play were responsible for the significant differences (Figure 6). Observations of percussive foraging and percussive travel were much greater than expected, although not significantly so. The trend of feeding in Georgia Strait appeared to be important around the Fraser River delta. The Fraser River mouth forms a broad 35 km long delta along the eastern shore of the

32 28 Strait. There are steep slopes off the mouth of the river which drop over 100 m in less than 2 km (Figure 1). This is well known as a feeding and lingering area for salmon migrating upstream (Groot et al 1984). Perhaps the steep slopes are used to further aggregate or capture the salmon. In any case, percussive foraging was the predominant feeding behavior in Georgia Strait. Puget Sound. Milling was the only behavior in Puget Sound (Figure 2) which occurred significantly more than expected; percussive foraging, travel, and intermingling 2 were observed much less than expected (X = 80.2, df = 7, P <.001; Figure 6). Whales were observed in Puget Sound for hrs (Table 5). It is difficult to point to any specific bathymetric features of Puget Sound (Figure 1) that may account for the frequent occurrences of milling behavior. If milling represents an independent mode of feeding on abundant prey, either prey densities in the region were always high (probably of a migratory fish such as salmon) and no cooperative herding was required, or the whales fed on non-schooling fish (such as cod and rockfish) which were best hunted individually. Of course, these options are not mutually exclusive and the whales undoubtedly feed on a variety of prey. However, since I have already shown that the movements of these whales into

33 29 Puget Sound corresponded to peak salmon fishing catch (Heimlich-Baran 1986), I would suggest that high salmon densities account for the occurrence of milling behavior in Puget Sound. Summary. In conclusion, feeding areas of the resident killer whales are primarily characterized by high relief bathymetry and the presence of shallow reef areas. These areas are most likely favored feeding areas because of increased prey availability. Seamounts and kelp-covered reefs are known to provide excellent fish habitat (Simenstad et al 1979), especially for residential fish such as rockfish (Matthews and Barker 1983). Numerous fish species feed here on smaller fish and zooplankton supported by high productivity associated with upwelling generated by the bathymetric features (Boje and Tomczack 1978). Kelp-covered reefs may also provide refuge from predators. Finally, the intensified tidal current regime around these reefs may be attractive to migrating salmon. Salmon traveling through the San Juan Islands tend to move in main current areas and orient along axes of tidal currents (Stasko et al 1976), such as are found adjacent to these reefs. Felleman (1986) has shown direct correlations between changes in tidal state and killer whale direction

34 30 changes and behavioral transitions, emphasizing that the whales are very aware of the tides also. It also appears that these habitats are conducive to collecting prey into higher densities. Some of the effects may be due to physically constraining the movements of migratory fish. Seamounts and sills will restrict the movements of fish and force them to travel closer together, thus making them easier for killer whales to catch. Felleman (1986) suggests that the narrow straits define the home ranges of resident killer whales primarily because of this salmon collecting effect. However, I also feel that the whales are actively using these features to herd prey together. The behavioral evidence for this is primarily numerous incidental observations in the vicinity of these features involving flank formations and percussive behaviors. Flank formations aid cetacean predators in scanning large areas for patchily distributed prey resources, while the percussive behaviors cause grouping responses and flight in fish prey (Norris and Dohl 1980a; Wursig 1986). The common physiographic element of all these bathymetric features is a steep, underwater slope rising up to within 10 meters of the surface. Whales would be able to drive fish towards such barriers in order to concentrate this prey into denser groups. This could

35 31 increase capture rates and thus make feeding more efficient. Along the shores of quadrants 6 and 12, the steep slopes abruptly end at shallow, kelp-covered sills jutting out from the shore. Fish driven along this slope would be forced into the kelp. Whales were often observed traveling less than 10 m from shore and then entering the kelp, getting it tangled on their fins, apparently as they attempted to feed on fish. The percussive behaviors so prominent in quadrant 12 are good examples of apparent herding in one of these areas. The pattern of percussively traveling through the western portion of northern Hare Strait (quadrants 7, 8, and 9; see below) may further serve to drive fish where they may be concentrated at the reef in quadrant 12. There were also numerous observations of whales entering the kelp beds of quadrant 17, possibly feeding. However, as mentioned above, kelp may also offer fish some refuge from killer whale predation. As supplemental evidence, all of these areas are well knmvn to local salmon fishermen as productive fishing grounds. Resident whales have shown little interest in other marine mammals. Other marine mammals were observed in the vicinity of the resident killer whales in 30 of the 459 encounters from 1976 to In these 30 observations, two resulted in observed attacks, once on a harbor porpoise

36 32 and once on a Dall porpoise. The harbor porpoise disappeared and the Dall was last observed with half of its fluke missing. On two other occasions the killer whales appeared to follow a minke whale, but no attack was observed. In the remaining 26 cases, the whales showed no interest in the other animals. Pinnipeds were observed in the water on 12 instances, usually within 50 m. This suggests that resident whales are not interested in preying on marine mammals. Resident killer whales off northern Vancouver Island also appear to favor salmon as prey, while marine mammals are ignored (Jacobsen 1986; Jefferson 1987; Spong et al 1970, 1971). Travel Areas Travel behaviors occurred significantly more than expected in three areas of Haro Strait: quadrant 4, quadrant 6, and pooled quadrants 7, 8, and 9. The subregion of the central Georgia Strait also showed a significant increase in observations of both traveling behaviors, as did the overall region of Georgia Strait. Quadrant 4. Quadrant 4 was characterized by increased observations of travel and reduced amounts of milling, play, and percussive travel (x 2 = 52.8, df = 7, P <.001; Figure 7). Whales were observed here for 124 hrs over 144

37 QUADRANT 4 N = hrs. 33 > EXP SQUAR <EXP 10 > EXP <EXP GEORGIA STRAIT N = hrs, > EXP <EXP 10 > EXP CHI CHI- Clll- CHI- <EXP 10 Figure 7. Chi-square values for travel areas. Behaviors and statistical significance are described in Figure 6.

38 34 encounters (Table 4). Quadrant 4 is a deep (260 m) and narrow (2 km wide) channel with a few reefs along the western border (Figure 4). It is located between two feeding areas in Haro Strait mentioned above. Quadrant 6. Percussive travel was observed significantly more than expected in the quadrant 6, although it ranked third to increased observations of rest and percussive foraging (X 2 = 99.4, df = 7, P <.001; Figure 8). Whales were observed here for 53.5 hours (Table 5). The bathymetry of this area has already been described in an attempt to understand the occurrence of feeding behaviors here. If whales were traveling 1-2 km offshore they would be in a featureless, deep water area, similar to other traveling areas. For feeding, it appeared that whales would have to move close to shore in order to utilize the more varied areas of the quadrant (e.g., the steep slopes along shore and the prominent sill to the north) as was done in other feeding areas. Unfortunately, all observations were pooled for this quadrant and no record was kept of the whales' distance from shore. Thus, I am unable to explore this hypothesis further. Pooled Quadrants 7, 8, and 9. These three quadrants were all found to have homogeneous behavior distributions

39 35 (heterogeneity x 2 = 22.0, df = 21, P >.05), so they were pooled for analysis. Increased observations of percussive travel and a lack of observations of foraging and intermingling characterized this area (X 2 = 55.9, df = 7, P <.001; Figure 7). Whales were observed in this area for a total of 75.0 hrs during 62 encounters (Table 4). Percussive travel was observed during 21 of the 62 encounters, characterized by high speed travel to the northeast. The topography of this region is that of a deep, featureless canyon, sloping from maximum depths of 365 m in quadrant 7 to 125 m in quadrant 9 (Figure 4). Central Georgia Strait. The subregion of central Georgia Strait (Figure 2) showed a significant increase in observations of both traveling behaviors along with a lack 2 of observations of rest (X = 74.9, df = 7, P <.001; Figure 7) (Table 5) Whales were observed here for hours The overall region of Georgia Strait was also characterized by increased occurrences of percussive travel, in addition to percussive foraging and intermingling (X 2 = 108.9, df = 7, P <.001; Figure 7). This region is a 150 m deep, featureless basin between the Fraser River delta and northern Haro Strait (Figure 1). As soon as whales entered central Georgia Strait from Haro Strait they would invariably begin traveling rapidly,

40 36 ''porpoising" out of the water. They would continue this behavior until they reached the opposite shore of the strait at the feeding area of the Fraser River delta. Summary. In summary, traveling areas are relatively deep water areas with low relief bathymetry. There are no apparent habitat distinctions between the two types of traveling (with or without percussive behaviors). However, the occurrence of percussive behaviors while traveling in northern Hare Strait and central Georgia Strait may represent herding towards adjacent feeding areas or some form of social signaling. Osborne (1986) showed a sequential relationship between the percussive behaviors of foraging and travel. Feeding may not occur in the travel areas either because prey are not abundant enough or because the whales are unable to catch prey without the aid of high relief bathymetric features. Also, since traveling areas appear to be located between feeding areas, the whales appear to have favored locations for feeding and use these other areas to get from one meal to another. Rest Areas Rest occurred significantly more than expected in two adjacent quadrants of Hare Strait, 6 and 17 (Figure 8). Rest was the primary behavior in quadrant 17 2 (X = 28.0, df

41 37 QUADRANT 6 N = 53.5 hrs. > EXP 1 <EXP ' QUADRANT 17 N = 20.0 hrs:-. > EXP 1 CHI- CHI- <EXP 10 Figure 8. Chi-square values for rest areas. Behaviors and statistical significance are described in Figure 6.

42 38 = 7, p <.001). It was observed on 10 of the 32 encounters. Quadrant 6 also had significant observations of rest (X 2 = 99.4, df = 7, P <.001). Both these areas also had increased occurrences of feeding. In addition, quadrant 6 was characterized by increased observations of percussive travel. The occurrence of multiple behaviors which occurred significantly more than expected shows that these were multi-purpose quadrants. The habitat requirements for rest are still unclear. Quadrant 17 has shallow, protected channels out of the main tidal flow, which may be conducive to resting. Condy et al (1978) found that killer whales rested in sheltered coves along Marion Island in the Indian Ocean. On the other hand, quadrant 6 is primarily a deep, open water area, offering no shallow protected waters. Rest has been observed in strong currents. During one encounter in quadrant 4 resting was observed while the pod oriented into a strong ebb current. Theodolite tracks showed that the currents were so strong the whales were actually drifting backwards in the current. Of course, the whales would have no awareness of currents unless there were cues from stationary reference points such as shore or the bottom. There is currently not enough information about the whales' requirements for rest to make any conclusions. It could be

43 39 that the important criterion for rest to occur is the previous sequence of behaviors, not the habitat. Osborne (1986) has shown that rest tended to significantly follow feeding. This may explain the occurrence of multiple significant behaviors in these quadrants. Whales may have rested in this area simply because they went there to feed. Socializing Areas The social behaviors of play and intermingling occurred significantly more than expected in two quadrants of Haro Strait: quadrants 10 and 12 (Figure 4) and in the Juan de Fuca Strait region (Figure 2). Quadrants 10 and 12. Behavior in quadrant 10 was characterized by significantly greater observations of play an d a 1 ac k o f o b servatlons. or - ml '11' lng ( X 2 = 18. 4, df = 7, p <.025; Figure 9). Whales were observed in this quadrant for hrs during 40 encounters (Table 4). Quadrant 12 was primarily characterized by increased observations of both intermingling and play (along with percussive foraging; chi-square listed above; Figure 9). These socializing behaviors occurred on 30% of the 43 encounters in these two quadrants. These quadrants share the physiographic characteristics of feeding areas (as described for quadrant 12). The entire region of northern

44 QUADRANT 10 N=17.2hrs. 40 >EXP 10 <EXP 10 QUADRANT 12 N = 18.5hrs. > EXP 10 <EXP 10 > EXP 1 CHI- CHI- < EXP 10 Figure 9. Chi-square values for socializing areas. Behaviors and statistical significance are described in Figure 6.

45 41 Haro Strait contained behaviors of high activity level, usually concluding once whales crossed the shallow sill of quadrant 12 and entered Georgia Strait. Thus, "play" behaviors in this area may just be different forms of percussive behavior used to drive fish against the sill for feeding. Juan de Fuca Strait. Behavior in Juan de Fuca Strait (Figure 2) was characterized by increased observations of play and a lack of observations of percussive travel (X 2 = 39.2, P <.001; Figure 9). Whales were observed here for hrs. Juan de Fuca Strait is one of the major open water areas of the region. Summary. The only characteristics which seemed to correspond to the occurrence of socializing behaviors were areas of open water or areas characterized by feeding and other percussive behaviors. There is no apparent reason why intermingling should be tied to specific habitat characteristics since it was primarily dependent on the meeting of separated pods (Osborne 1986), which could occur anywhere. ''Play'' behavior, as categorized by breaching, may have a wide range of functions, including noise-generating herding behavior or a form of communication between whales. Further conclusions on

46 42 relationships of play with habitat will have to wait until more is known about the functions of play behavior in these whales. Distribution of Transient Whale Observations Transient whales were observed on 18 encounters for hrs (mean of 2.4 hrs/encounter) from 1976 to They were observed in 29 of the 441 quadrants of the study area (Figure 10). Ten of those 29 quadrants (34%) were areas in which resident whales were never observed in almost 1000 hrs of observation, although the 10 quadrants were all adjacent to quadrants which resident whales frequented. This provides clear evidence of different geographic distributions between transient and resident whales. The seasonal distribution of the 18 transient whale encounters shows no clear seasonal trends. There were two peaks of observation. Five encounters occurred in the months of ~larch, April, and May. The remaining 13 encounters were in the months of August, September, and October. This is the period of harbor seal pupping in the region (Everitt et al 1979). Transients were never seen in the months of June and July, a period of increasing resident whale occurrence. Of course, this is still a very

47 43 ISLAND 49'N British Columbia PACIFIC OCEAN Washington TRANSIENT WHALES NO. OF ENCOUNTER DAYS Do 1-2 ~ 3-4 ) kilometers miles 24'W Figure 10. encounters. Distribution map of transient whale

48 44 limited data set and conclusions must be considered tentative. Patterns of Transient Behavior and Habitat Use The behavior of the transient whales was characterized by increased feeding behavior when compared to that of the 2 resident whales (X = 307.4, df = 7, P <.001). Feeding represented 81% of the transient whale observations (compared to 47% for resident whales). Travel behavior occurred 12% of the time, rest was observed 7% of the time, and socializing behaviors were never observed. Four of the 18 encounters (2.5 hrs) included direct observations of feeding on harbor seals, Phoca vitulina. These four observations represented clear examples of marine mammal predation. Areas of seal concentrations were directly approached and searched until prey were found. There were no clear observations of cooperative hunting, although groups of whales would occasionally attack the same seal. On all attacks, the actual kill appeared to be prolonged, perhaps as an instructional activity for young or inexperienced whales, as Lopez and Lopez (1985) have reported for killer whales in Argentina. The sample size of transient whale behavior was too small to allow statistical testing for significant geographic variation. However, the quadrants in which

49 45 transient whales were observed feeding were primarily shallow nearshore quadrants in areas away from major tidal channels, which in most cases corresponded to harbor seal haul-out areas (Everitt et al 1979). Harbor seals appear to be an important item in the diet of transient whales. Transient whales undoubtedly also feed on fish. The occurrences of one pod of transient whales over three years in one bay east of the San Juan Islands have been timed to salmon runs in the Nooksack River. Felleman (1986) reported on a preliminary analysis of sonar observations of large, multiple, bottom targets (usually indicative of bottomfish or schooling baitfish) in the vicinity of feeding transient whales. This contrasts with observations of large single targets (indicative of salmon) associated with resident pods (Felleman 1986). Transients have been observed apparently feeding in kelp, similar to resident whales. Most fish species inhabiting kelp are generally resident and non-migratory (Matthews and Barker 1983). Their site specificity makes them a similarly predictable resource as are seals resident to haul-out sites (Felleman 1986). More observations are needed of transient whale behavior before further conclusions can be made.

50 46 DISCUSSION The habitat use patterns of the resident killer whales in Haro Strait appear to be centered around the location of three feeding areas. These areas all share the high relief bathymetry and kelp-covered reefs conducive to the capture of migratory salmon and which correspond to suggested areas of fish abundance. The whales travel through featureless, deep-water areas between these feeding areas. The distribution of rest may occur primarily within the sequence pattern of behaviors described by Osborne (1986), tending to follow feeding. Play behaviors may occur adjacent to feeding areas as additional forms of noise-generating herding behavior or for whales to maintain contact in open water areas. Thus, the location of food resources and habitats suited to prey capture appear to be the prime determining factor in the behavioral ecology of these whales. Research in other areas has found similar correlations between delphinid behavior and habitat. Killer whales in the northern resident community have favored ''rock rubbing areas'' where they linger for up to 1.5 hrs rubbing their bodies on a shallow pebble substrate (Jacobsen 1986). Evans (1974, 1975) found correlations between the distribution of common dolphins and the location of underwater seamounts