Domeier et al.: Tracking Thunnus thynnus orientalis with the aid of an automated algorithm 



293 



location algorithm to further our understanding of blue- 

 fin tuna movements in the eastern Pacific. 



The light sensors on archival and pop-up satellite 

 tags provide data on the time of sunrise and sunset, 

 allowing one to calculate the approximate geographic 

 position of an animal (Delong et al., 1992; Wilson et al., 

 1992; Hill, 1994; Bowditch, 1995; Sobel, 1995; Welch 

 and Eveson, 1999; Hill and Braun, 2001; Metcalfe, 

 2001;Smith and Goodman; 1 Gunn et al. 2 ). The accu- 

 racy of the light-based geolocation estimates have been 

 studied under controlled conditions (tags tethered to 

 a moored buoy) and field conditions (tags attached to 

 fish at a known location). Locations from tethered tags 

 have been reported to be accurate to within ±0.2-0.9° 

 in longitude and ±0.6-4.4° in latitude (Welch and Eve- 

 son, 1999, 2001; Musyl et al., 2001). Tagged tuna have 

 provided light-based geolocation estimates within ±0.5° 

 of longitude and ±1.5-2.0° latitude (means) of known 

 locations (Schaefer and Fuller, 2002; Gunn et al. 1 ). 



Light-based estimates are not precise and comparing 

 studies that have examined the accuracy of this method 

 is complicated by differences in tag hardware and geo- 

 location algorithms used by different researchers. Other 

 physical and biological factors complicate the issue fur- 

 ther. Day length is not a good predictor of latitude dur- 

 ing the spring and fall equinox, therefore estimates of 

 latitude at times surrounding the equinox contain more 

 error than at other times of the year (Hill and Braun, 

 2001). Latitude estimates are also more prone to error 

 the closer the animal is to the equator (Hill and Braun, 

 2001). Additional errors can be introduced into esti- 

 mates of both latitude and longitude by the behavior of 

 the tagged animal (e.g., diving), bio-fouling of the tag, 

 cloud cover, and wave action (Metcalfe, 2001). 



Poor resolution of latitude estimates continues to be 

 a problem for researchers using light-based geolocation 

 algorithms. Under ideal theoretical conditions the vari- 

 ability in latitude error cannot be less than 0.7° and the 

 expected variability in longitude will be a constant 0.32° 

 (Hill and Braun, 2001). Sibert et al. (2003) developed 

 an algorithm that applies a Kalman filter to light-based 

 geolocation estimates in an attempt to reduce the error 

 of these estimates. Although this approach smoothes 

 data, it does not incorporate external data (data not 

 collected by the tag) and therefore is still affected by 

 errors inherent in the use of light-based geolocation es- 



1 Smith, P., and D. Goodman. 1986. Determining fish 

 movements from an "archival" tag: precision of geographi- 

 cal positions made from a time series of swimming, tem- 

 perature and depth. NOAA. Tech. Memo. NMFS-SWFC-60, 

 13 p. Southwest Fisheries Science Center, La Jolla, CA 

 92038. 



2 Gunn, J. S„ T. W. Polacheck, T. L. O. Davis, M. Sherlock, and 

 A. Betlehem. 1994. The development and use of archival 

 tags for studying the migration, behavior and physiology of 

 southern bluefin tuna, with an assessment of the potential 

 for transfer of the technology to groundfish research. In 

 Proceedings of ICES mini-symposium on fish migration, 

 23 p. International Council for the Exploration of the Sea, 

 Palaegade 2-4, DK-1261 Copenhagen K. Denmark. 



timates of latitude. It has been suggested that sea-sur- 

 face-temperature (SST) and bathymetry data be used 

 to refine light-based geolocation estimates (Block et al., 

 2001). These techniques are particularly useful when 

 there is a north-to-south gradient of bathymetry or SST. 

 The use of bathymetry to refine latitude requires an as- 

 sumption that maximum diving depth is limited by the 

 bottom depth; certainly this assumption introduces a 

 new source of error. In addition, for animals that move 

 off the continental shelf, bathymetry would be useless. 

 The use of SST or bathymetry data to refine latitude 

 necessitates the arduous task of matching tag data with 

 another source of data. 



It was our opinion that the accuracy of tracking ma- 

 rine animals could be improved through the develop- 

 ment of an algorithm that automatically resolved lati- 

 tude estimates by matching SST measurements from 

 the tag to those taken from satellites. Here we present 

 such an algorithm; one that was designed to operate 

 in a geographic information system (GIS) environment, 

 allowing for rapid analysis and display of archival and 

 PSAT tag data. We demonstrate the algorithm and its 

 product through the analyses of data we collected from 

 Pacific bluefin tuna tagged in the eastern Pacific. 



Materials and methods 



Tagging in the field 



Pacific bluefin tuna were captured on rod and reel from 

 a recreational fishing vessel by using live bait and circle 

 hooks. Fishing took place 123 nmi southwest, 86 nmi 

 southwest, and 178 nmi south of San Diego in years 

 2000, 2001, and 2002, respectively. Fish were lifted into 

 the boat with a vinyl sling and then placed on a soft mat, 

 eyes were covered with a cloth, and the gills irrigated 

 with seawater. The fish were then measured (fork length 

 and girth), tagged, and immediately released. Sixteen 

 fish were tagged with Wildlife Computers Inc. (Redmond, 

 WA) pop-up satellite archival tags (PSATs), one fish was 

 tagged with a Microwave Telemetry Inc.! Columbia, MD) 

 PTT-100 PSAT, and seventeen fish were tagged with 

 Lotek Wireless Inc. (Newmarket, Ontario) LTD2310 

 nontransmitting archival tags. The two types of PSATs 

 either provided data once an hour (depth, water tempera- 

 ture, light level [Microwave Telemetry, Inc.]) or sum- 

 marized data that had been collected every two minutes 

 (Wildlife Computers, Inc.) — the difference being an arti- 

 fact of the two tag manufacturers. The Lotek archival 

 tags provided us with data every two minutes detailing 

 the swimming depth, water temperature, internal fish 

 temperature, and light level. Pressure sensor drift was 

 adjusted by the tag manufacturers' software for PSAT 

 tags and in the laboratory for the Lotek tags. 



The PSAT tags were rigged with 300-lb monofilament 

 leaders and a nylon dart. In 2000 and 2001 the dart 

 was a "bluefin-type" provided by Eric Prince (NMFS- 

 SEFSC); in 2002 a Pfleger Institute of Environmental 

 Research (PIER) "umbrella" dart was used (Fig. 1). 



