518 



Fishery Bulletin 101 (3) 



10.0 1 



-10.0 



-10.0 



0.0 



Longitudinal error (degree) 



10.0 



Figure 2 



Average errors of location estimates from each archival tag in the 

 benchmark tests. Smaller circles are the average errors of the tags 

 tested in air and the larger circles are the average errors for archival 

 tags implanted in pen-held fish. Bar shows a standard deviation. 



A more useful measure of in-water accuracy was pro- 

 vided by comparison between actual recapture locations 

 of 18 tags and the locations that those tags estimated one 

 or two days prior to capture (thus avoiding the disturbed 

 light data on the final day). Average differences of the 18 

 tags were -0.1 ±0.8° (range: -2.0 ±1.7°) in longitude and 

 -1.6 ±1.8° (range: -5.7 ±0,6°) in latitude. 



Because the tag's latitude estimate based on day length 

 was found to have limited reliability, we estimated latitude 

 using sea-surface temperature (SST) as recorded in the 

 summary file for each day. The temperature reference field 

 used was the SST map published by Japan Fisheries In- 

 formation Service Center, which gave average SST weekly 

 for the western Pacific Ocean (west of 160°E), and every 

 10 days for the eastern Pacific Ocean (east of 160°E). The 

 longitude value determined automatically by the tag was 

 used to choose a longitude on the SST map. Along that lon- 

 gitude line a point was sought where the map SST matched 

 the SST value recorded by the tag. If multiple points were 

 found to satisfy this criterion, the point that gave the most 

 plausible movement was selected, based on fish locations 

 on several adjacent days. If a location still could not be 

 determined, it was interpolated as a midpoint between the 

 adjacent two days' locations. 



One example of location re-estimation is shown in Figure 3. 

 After consulting with the SST maps, we used 1.4% of the 

 locations estimated automatically from 29 recaptured tags. 



and 79.7% of latitudes were changed by -(-0.3 (±2.8°) on av- 

 erage wdth the SST method. The remaining 18.9% of days did 

 not provide any reasonable location estimates for various rea- 

 sons, including anomalous longitudinal estimates, no match 

 points of SST along the estimated longitudinal line, or the 

 existence of a wide latitudinal area showing the same SST. 



Reliability of temperature and pressure sensors 



One hundred tags calibrated within half a year of manufac- 

 ture showed average errors of 0. 1 ±0. 1°C for both internal 

 and external temperature sensors. Nine tags recovered from 

 fish and tested more than one year after manufacture showed 

 average errors of 0.0 ±0.1°C for both sensors. It thus appears 

 that no deterioration of the temperature sensors occurred 

 because of release-recapture or the passage of time. 



No large error in pressure sensors was observed dur- 

 ing calibration of tags before release. However, 20 of 27 

 tags recovered from tagged and released fish were found 

 on recalibration to record substantially lower than actual 

 pressure. One example is shown in Figure 4. No further 

 deterioration of pressure sensors was observed when these 

 tags were kept in air for an additional half year There 

 was no way to know exactly when the sensor deteriora- 

 tion had occurred during the time the fish were in water 

 However, the frequency of records showing swimming at 

 m depth was remarkably higher in the second part of 



