334 



Fishery Bulletin 97(2), 1999 



Seals were viewed with 22x and 15-60x spotting 

 scopes. Scan surveys ( Altmann, 1974) were conducted 

 at ten minute intervals throughout the observation 

 period. During each half-hour period, the first scan 

 was a count of all seals, the second scan included 

 size structure (number of harbor seal pups and sub- 

 adults and adults), and the third scan was used to 

 quantify vigilance of seals (head up, alert but not 

 oriented toward a disturbance source) and sources 

 of disturbance (within 1.0 km of the haul-out site). 

 Counts of size structure included only those individu- 

 als that could be assigned a given category (e.g. seals 

 were not included if only a flipper was visible as in 

 the first scan). 



Counts from these scans were used to determine 

 how many seals entered the water following a dis- 

 turbance and when recovery had occurred. Recovery 

 was measured by the increase in number of harbor 

 seals on the haul-out site after harassment. Recov- 

 ery was divided into four categories: 1) full recovery 

 (number of seals ashore after harassment returned 

 to preharassment levels); 2) partial recovery (num- 

 ber of seals ashore increased after the harassment, 

 but did not reach preharassment levels); 3) no recov- 

 ery (number of seals ashore did not increase after 

 harassment); and 4) no chance to recover (number of 

 seals ashore never increased after harassment owing 

 to repeated disturbances or rising tide washing over 

 the haul-out site). Because animals were not marked, 

 full recovery did not imply that individuals returning 

 to shore were necessarily the same ones that were ha- 

 rassed, but partial and no recovery did indicate cer- 

 tain individuals did not return to the haul-out site. 



Terhune (1985) and da Silva and Terhune (1988) 

 reported that the number of vigilant harbor seals was 

 dependent on group size. To eliminate the potential 

 effect of group size on vigilance, the original data 

 were subsampled to produce subsets of equal group 

 sizes. Therefore, a single value is not presented for a 

 site because it varies with each comparison. 



Data were collected for every potential source of 

 disturbance that approached the haul-out site. 

 Sources were divided into the following categories; 

 airplanes, powerboats (including sailboats under 

 motor power), sailboats, kayaks and canoes, people, 

 bald eagles (Haliaeetus leucocephalus), unknown, or 

 other Vessel speed was classified as underway fast 

 (creating a breaking bow wake), underway slow 

 (nonbreaking bow wake), and drifting (motor not in 

 gear or turned off). Harbor seal reactions to a distur- 

 bance were categorized as 1) detection: >1 seal with 

 head raised and oriented toward potential distur- 

 bance source; 2) alarmed: >1 seal moved from its rest- 

 ing place, but did not enter the water; and 3) ha- 

 rassed: >1 seal entered the water. 



Positions of an approaching vessel were monitored 

 by using a Nikon NT2A or Pentax TH20D theodo- 

 lite. Bearings to the approaching vessel and seals 

 exhibiting disturbance reactions were recorded. The 

 distance from theodolite to vessel or seal was calcu- 

 lated by using the tangent of the vertical angle from 

 the theodolite and height of the theodolite above the 

 water. The distance between vessel and seals was 

 calculated by using the Law of Cosines that incorpo- 

 rates distances between theodolite and vessel and 

 theodolite and seals and by using the horizontal angle 

 between the vessel and seals (from theodolite). 

 Height of the theodolite above water was measured 

 directly or estimated from a cosine prediction of tide 

 height ( San Juan Current and Tide Tables, published 

 by Island Canoe, Bainbridge Island, Washington). 

 The tidal constituent used was Port Townsend, Wash- 

 ington, with a correction for Echo Bay, Sucia Island 

 (approximately 1 km from Ewing Island and 6 km 

 from Puffin Island). The observation point at Ewing 

 Island (for Clements Reef surveys) was near a verti- 

 cal rock ledge, which allowed the observer to mea- 

 sure theodolite heights above water level (using a 

 tape measure with float attached). Direct measure- 

 ment of height above water was accurate to approxi- 

 mately 0.1 m. Direct measurement was not possible 

 at Puffin Island; therefore, theodolite heights above 

 water level were based on tide height predictions, 

 which were accurate to approximately ±0.3 m. Error 

 in theodolite locations was less than 10 m; accuracy 

 was based on calibration with fixed objects (e.g. buoy 

 or island). 



Analyses 



Results of statistical analyses were considered sig- 

 nificant at a = 0.05. Mean percentage of pups among 

 the sites was compared by using analysis of variance 

 ( ANOVA) with arcsine transformation. A Tukey mul- 

 tiple comparison test was usedfor significant ANOVA 

 results (Zar, 1984; Day and Quinn, 1989). To test 

 whether pups were less tolerant of disturbance than 

 adults, the frequency of positive and negative changes 

 in the proportion of pups ashore before and after a 

 harassment was compared by using a chi-square 

 goodness-of-fit test (data were combined from all 

 sites). Percentage of vigilant seals was compared 

 between sites by using Mann- Whitney U tests (sepa- 

 rate two-sample tests were conducted owing to ran- 

 dom subsampling to control for group size). Differ- 

 ences in distance of disturbance among powerboat 

 approach speeds were tested with Mann-Whitney, t, 

 and Kruskal-Wallis tests. Significant differences in 

 distance of disturbance among categories of detec- 

 tion, alarm, and harassment were detected with 



