768 



Fishery Bulletin 100(4) 



previous day were considered improbable and excluded 

 from further analyses. For the 22 bigeye tuna at liberty 

 for at least 30 days, this filtering procedure retained an 

 average of 48.8% {95%CI=7.8%, range=15.3% to 76.5%) of 

 the daily geolocation estimates. This procedure provided 

 data sets of geolocation estimates throughout the period 

 of time at liberty for each fish, about which we have a high 

 level of confidence based on the expected accuracy and 

 precision of these data (accuracy being the closeness of 

 geolocation estimates to the actual position and precision 

 the closeness or dispersion of repeated estimates IKendall 

 andBuckland, 19821 ). 



Each set of filtered geolocation estimates for individual 

 bigeye tuna was integrated into a Geographic Information 

 System (GTS). The animal movement analyst extension 

 (AMAE) (Hooge and Eichenlaub, 1997; Hooge, et al.^) and 

 the Arc View (Environmental Systems Research Institute, 

 Inc., Redlands, CA) GIS progi'am were used for mapping 

 movement paths and for performing various spatial analy- 

 ses of the data. The site-fidelity test in AMAE, employing 

 1000 random walks, was used to test the null hypothesis of 

 random movement. The test is a modified Monte Carlo ran- 

 dom walk, starting at the location of release, constrained 

 by the coastline, and uses the actual sequence of distances 

 between geolocation estimates to determine walk points. 

 The fixed kernel home range model in the AMAE, which 

 incorporates a least-squares cross-validation smoothing 

 function, was used to assess probabilistic home ranges of 

 fish for which the null hypothesis was rejected in the site- 

 fidelity test. The 95% and 50% utilization distributions 

 (probability contours) were chosen to describe the areas 

 probably used (95%) and the probable core areas (50% ) of 

 activity, respectively, of individual fish. 



For fish at liberty for 30 d or more, the behavior for 

 each day at liberty was classified as unassociated (with 

 a floating object I type-1, unassociated type-2, or associ- 

 ated with a floating object. Type-1 behavior is defined as 

 that in which the fish primarily occupies the mixed layer 

 during the night, descends at dawn below the thermocline 

 to depths greater than 100 m and remains below 100 m 

 throughout the day (aside from vertical forays into the 

 mixed layer with a frequency of no more than 12 per day) 

 and then ascends back into the mixed layer at dusk. This 

 basic pattern has been described by Holland et al. (1990) 

 and Dagorn et al. (2000), based on sonic tracking studies of 

 bigeye tuna. Data from periods immediately after release 

 at FADs until the time the FADs were removed and the 

 data immediately preceding recapture at FADs were eval- 

 uated to determine criteria for classifying behavior of fish 

 associated with floating objects. The percentages of time 

 that 19 bigeye tuna spent at depths greater than 50 m, 

 within 24-h intervals, for a total of 74 days, ranged from 



2 Hooge, P. N., W. Eichenlaub, and E. K. Solomon. 2001. Using 

 GIS to analyze animal movements in the marine environment. 

 In 2001, spatial processes and management of marine popula- 

 tions (G. H. Kruse, N. Bez, A. Booth, M. W. Dorn, S. Hills, R. N. 

 Lipcius, D. Pelletier, C. Roy. S. J. Smith, and D. Withercll, eds.), 

 20 p. Univ. Alaska Sea Grant, P.O. Box 755040, Fairbanks, AK 

 99775-5040. 



0.0% to 29.1%. Sixty-nine percent of the 24-h intervals in- 

 dicated that 16% or less time was spent below 50 m. The 

 days when the fish were below 50 m for 16'X of the time or 

 less were classified as behavior associated with a floating 

 object, and the days in which that time was greater than 

 16% were classified as unassociated type-2 behavior. For 

 each fish, the numbers and durations of each behavioral 

 event were determined throughout the period at liberty. A 

 statistical test for a difference between the spatial distri- 

 butions of the geolocation estimates for unassociated and 

 associated behavior was conducted (Syrjala, 1996). 



Dives in excess of 500 m were classified as deep dives, 

 and the time, duration, and maximum depth of each event 

 was determined for each bigeye tuna throughout periods 

 at liberty. For those dives in which the pressure sensor 

 reached its limit, at approximately 1000 m, depth was 

 estimated from the ambient temperature value recorded 

 by the tags by using historical annual mean temperatures 

 at depths for the area of the deep diving event (Levitus 

 and Boyer, 1994). The light level and ambient temperature 

 data recorded by tags from bigeye tuna making dives in 

 excess of 500 m indicated that light was essentially nonex- 

 istent below 500 m and that the ambient temperature was 

 less than 8.5°C. 



The archival tag data sets (for each recaptured bigeye 

 tuna at liberty for at least 30 d) were separated into peri- 

 ods of nighttime and daytime by using light-level records. 

 Nighttime was classified as the period between the time of 

 the first record after dusk when there was no recognizable 

 light from the sun until the time of the last record (before 

 dawn) of no recognizable light from the sun. The individu- 

 al data sets for night and day were used in evaluations of 

 diel differences in behavior and habitat selection. The data 

 for each night were classified by the moon phase (U. S. Na- 

 val Observatory, 2001 1. The classification scale used was 1 

 through 29 for the entire moon phase cycle. We analyzed 

 changes in average daytime and nighttime depths for 20 

 individual bigeye tuna in relation to the visible disk area 

 of the moon. 



Results 



Behavior 



Evaluation of the depth and temperature records for 

 bigeye tuna carrying archival tags resulted in the discrim- 

 ination of four distinct behaviors: 1) unassociated type-1, 

 2) unassociated type-2, 3) associated with a floating object, 

 and 4 ) deep diving. For the 23 fish at liberty for 30 or more 

 days, behavior was classified for each day at liberty, and 

 the duration of each behavior was determined (Table 2). 



The behaviors of fish released or recaptured at FADs 

 were distinct and discernible from the behaviors after the 

 FAD was removed until the fish associated with the FAD 

 at which it was recaptured (Fig. 1). Fish released at FADs 

 showed a consistent swimming depth, predominantly less 

 than 50 m during the night and day. When the FAD was 

 removed fish showed an erratic up-and-down swimming 

 behavior, followed by a consistent pattern of shallow (less 



