FISHERY BULLETIN: VOL. 83, NO. 2 



in response to the ship. There are, however, other 

 factors (such as glare and sea state) which are 

 seldom constant long enough to allow for accumu- 

 lation of a reasonably precise frequency distribu- 

 tion, such that the effects due to school movement 

 would not be overwhelmed by the effects due to 

 sighting conditions. 



The results of this experiment suggest that 1) 

 dolphin schools occasionally react to the approach 

 of a survey vessel prior to their detection by ship- 

 board observers and 2) the expected rarity of the 

 event implies that a considerable amount of addi- 

 tional data would be required to quantify its effect. 



Any directed movement prior to detection biases 

 the frequency distribution of perpendicular dis- 

 tances and may bias the function, fix), fit to these 

 data. In the absence of information regarding 

 movement, Burnhametal. (1980) suggested choos- 

 ing a function which is relatively insensitive to 

 data contaminated by movement, i.e., a function 

 that monotonically decreases with distance from 

 the transect line. Their simulations suggest that 

 in situations where "undetected movement is rela- 

 tively minor, then use of an estimator based on a 

 monotonically decreasing function will minimize 

 bias in D," (Burnham et al. 1980:130). The small 

 sample size of the present experiment was suffi- 

 cient to qualify undetected movement as relatively 

 minor but not sufficient to quantify its effect on 

 the distribution of perpendicular distances. 



Although the work reported here was conducted 

 in the same geographic area (Clipperton Island, 

 lat. lO'N, long. 110 W) as the Au and Ferryman 

 (1982) observations, the two experiments are not 

 strictly comparable. Au and Ferryman used the 

 ship and helicopter to search for schools and col- 

 lected data on their reaction to the ship without 

 regard to the effect on survey operations; in four of 

 the eight schools they studied, the ship was turned 

 toward the school during tracking. They were in- 

 terested in describing the behavior of dolphin 

 schools and combining the description with a 

 search model to quantify survey bias. The present 

 experiment did not assume that the two processes 

 (reaction and detection) were independent and 

 was less ambitious because there was no intention 

 to generalize dolphin behavior Indeed, the results 

 presented here may only be relevant to this area 

 and for these sighting conditions. Both the reac- 

 tion distance and the sighting distance may be 

 affected by environmental conditions and may 

 vary between geographic areas with the degree of 

 animal naivete. 



The comparisons of aerial and shipboard results 



suggest that school-size estimates may be more 

 reliable than those of species composition. Al- 

 though neither observation platform can be con- 

 sidered to yield estimates without error, they do 

 provide unique vantage points with very different 

 views of the dolphin school. All shipboard observ- 

 ers, after exposure to observation conditions in the 

 helicopter, agreed that they could more confi- 

 dently estimate school size from the air than from 

 a vessel. The helicopter provides an opportunity to 

 observe the entire school over an extended period 

 of time, making it easier to estimate that portion 

 of the school which is submerged and not com- 

 pletely visible. Species proportions are more diffi- 

 cult to estimate and it is not clear which platform 

 is better; indeed, in the case of school 22, all four 

 shipboard observers reported similar proportions 

 which were quite different than that estimated 

 from the air One explanation may be that it is 

 more difficult to identify animals in plan view 

 than in profile view; alternately, the fluid charac- 

 ter of school structure may combine with the lim- 

 ited view of the school from a ship to preclude 

 accurate estimates of species proportions; a third 

 possibility is that both are inaccurate because of 

 species-specific behaviors which make the ani- 

 mals less visible from above and/or the side. 



Estimates of the density of dolphin schools are 

 multiplied by the area of the survey, the average 

 school size and the species proportions to estimate 

 species abundances (Holt and Fowers 1982). Be- 

 cause they affect the abundance estimates di- 

 rectly, biases in the latter two parameters may be 

 more serious than the effect of school movement 

 prior to detection. As an example, consider the six 

 schools compared during this experiment: the av- 

 erage number of S. atten uata per school, estimated 

 by shipboard observers, was 2T^'( greater than that 

 estimated from the helicopter data, the shipboard 

 estimate of S. longirostris was 34% less than the 

 helicopter estimates, and the estimate of S. coeru- 

 leoalba was the same for both platforms (Table 3). 

 Although these differences should only be consid- 

 ered as variability between two estimates, they 

 illustrate the direct dependence of abundance es- 

 timates on accurate estimates of species propor- 

 tions. Avoidance affects density estimates less 

 dramatically; its affect on /'(O) may be somewhat 

 offset by using a function that is relatively insen- 

 sitive to predetection movement. 



The application of line-transect methods re- 

 quires that along the transect line all schools are 

 seen with certainty. Any departures from the as- 

 sumption of perfect detection, either because of 



192 



