Fishery Bulletin 88(1). 1990 



abundance estimate for eastern spinner dolphins was 

 579,600 animals with a CV of 0.332. 



Trends in abundance 



Assuming the CV(A^,j) for subsequent surveys will be 

 constant during the sampling period, and with a and 

 P errors equal to 0.10, a 10% annual decrease in abun- 

 dance of offshore spotted dolphins can be detected in 

 a minimum of 8 years. After 5 years, and assuming the 

 CV{N,j) of 0.255 will remain constant, a minimum an- 

 nual decline of 17.6% may be detected during which 

 a 62% decrease of the offshore spotted stock would 

 have occurred. 



Discussion 



A biased estimate may be acceptable to detect trends 

 in population changes if the bias is constant among an- 

 nual surveys and if it results in a more precise estimate. 

 For example, the estimate of school density that was 

 calculated by fitting a detection function to data pooled 

 over all strata is biased upwards because searching ef- 

 fort was not allocated to strata uniformly but propor- 

 tionately to historical estimates of density. The inshore 

 stratum, which historically had the largest density (Holt 

 et al 1987), received a greater proportion of the search- 

 ing effort (43% of total effort) compared with its rela- 

 tive size (30% of total area). However, the pooled esti- 

 mate is more precise than the stratified estimate, and 

 the bias should be consistent during subsequent years. 



Our estimate of target school size (99.13 animals. 

 Table 1) was half the estimate from aerial data collected 

 in 1979 (199.8 animals per school) (Holt 1985). Although 

 school size may have declined between 1979 and 1986, 

 our estimate is similar to a previous school size estimate 

 that used data collected 1979-83 aboard research 

 vessels (119.9 animals per school) (Holt 1985). In addi- 

 tion, we used only schools detected within 3.7 km 

 perpendicular distance of the trackline, while the pre- 

 vious aerial and ship studies used schools detected 

 within 11.1 km perpendicular distance (approximate 

 distance to horizon from ship). Our estimate using the 

 11.1 km perpendicular distance was 1 1 1 .97 animals per 

 school. The previous estimates from airplanes and ships 

 may have been biased upward because large schools 

 are more likely detected at greater distances than are 

 small ones. 



The inverse log weighting factor used to adjust biases 

 in the school-size and species proportion estimates 

 caused by detecting disproportionately more large 

 schools may have over- or undercompensated by an 

 unknown degree. Recent work by Drummer (1985) 

 investigating size biases may be utilized during com- 

 parisons of this data and subsequent years' data. 



Estimates of the relative proportions {P'ljk) of 

 coastal and offshore spotted dolphins and of eastern 

 and whitebelly spinner stocks in their respective areas 

 of overlap were based on data pooled over all strata 

 and included all schools occurring within 11.1 km (6 nm) 

 perpendicular distance from the trackline; however, 

 they were based on small sample sizes (29 spotted and 

 96 spinner schools). These estimates may change with 

 collection of additional survey data. However, the 

 proportion of eastern spinner dolphins to whitebelly 

 spinner dolphins in their overlap area was 0.703 dur- 

 ing 1986 (Table 1) and was 0.714 when all research 

 vessel data collected from 1976-86 were combined (207 

 schools). 



Our population abundance estimates are intended to 

 serve as the baseline estimates for relative comparisons 

 using data collected during subsequent surveys. When 

 our estimates of northern offshore spotted dolphins are 

 compared with previous estimates, ours are much 

 smaller than those made for data collected through 

 1979 (2,775,000 animals) (Holt and Powers 1982) and 

 for data collected through 1984 (2,533,300 animals) 

 (Holt 1985). Both the latter absolute and our current 

 base estimates may be biased because they share com- 

 mon data collection constraints— failure to detect all 

 trackline schools— which bias the density estimates 

 downward. The older estimates may also contain other 

 biases which result in relatively higher values. For 

 example, those estimates used a combination of data 

 collected aboard airplanes, research vessels and tuna 

 vessels, and survey coverage was pooled over several 

 years, seasons, and areas. However, some variables 

 used to calculate our estimates used small data sets that 

 may have resulted in less precise results. 



Because our population estimates are intended to 

 serve as the baseline estimate for relative comparisons 

 using data collected during subsequent surveys, con- 

 sistent bias during the sampling period will not jeopar- 

 dize the results. Therefore, several options will be 

 reviewed in analyzing subsequent years' data which 

 may reduce variability associated with the population 

 estimates. Sample sizes to calculate school sizes and 

 species proportions may be increased by utilizing all 

 schools detected at perpendicular distances from the 

 trackline out to the horizon. We will investigate spotted 

 dolphin abundance estimates using only schools of 

 spotted dolphins. A computerized binocular system 

 (Holt and Sexton 1987) may yield more precise esti- 

 mates of radial distance and sighting angles to dolphin 

 schools, and we have incorporated use of a ship-based 

 helicopter to obtain aerial photographs of dolphin 

 schools to calibrate observer estimates of school size. 

 Hopefully, some or all of these factors may reduce the 

 CViNii) levels to around ]2%, as anticipated l)y Holt 

 et al. (1987) in initial survey design. 



