HOLT ET AL.: MONITORING DOLPHIN ABUNDANCE 



DISCUSSION 



Our analyses indicate that our ability to detect 

 changes in the size of spotted dolphin populations 

 in the eastern tropical Pacific is not very great 

 without substantial long-term ship time. This is 

 not surprising given the vast area of ocean inhab- 

 ited by the dolphins and the low sighting rate 

 from ships. We feel our results represent a gener- 

 ally accurate picture based on available data. 

 However, the analyses must be qualified by not- 

 ing that the data used to generate these results 

 were accumulated during all seasons over 5 

 years. Data collected in the future will come from 

 surveys conducted at the same time each year and 

 may be less variable. In addition, more precise 

 data gathering techniques or data fitting models 

 may become available. If so, these factors would 

 yield greater ability to detect lower rates of de- 

 crease, greater power, and lower required number 

 of years. On the other hand, the estimates of ex- 

 pected variance have dealt with survey precision 

 (measurement error) only. If environmental vari- 

 ability is important, data collected in future 

 surveys may be more variable than we have cal- 

 culated. In long-lived animals with many year 

 classes contributing to reproduction, however, en- 

 vironmental variability will tend to be less impor- 

 tant than survey imprecision (Gerrodette in 

 press). 



The selection of appropriate alpha and beta 

 errors levels depends on one's perspective. An 

 alpha error would occur if we concluded that a 

 decrease in dolphin abundance was occurring 

 when in fact it was not. It is therefore of interest 

 to the tuna industry to minimize this type of er- 

 ror. A beta error would occur if we concluded that 

 no decrease in dolphin abundance was occurring 

 when in fact it was. It is in the interest of conser- 

 vation groups to minimize this type of error. As is 

 well known in statistical theory, however, there 

 is a trade off between the two types of error, a 

 decrease in one leads to an increase in the other. 

 In our analyses we have balanced the two types of 

 error by making alpha and beta equal. We have 

 also used a range of equal alpha and beta levels 

 (0.05, 0.10, and 0.20) to illustrate how choice of 

 error level can affect sampling design. Higher tol- 

 erance of error leads to lower rates of decrease 

 which could be detected in shorter times, but, of 

 course, one is less sure of the conclusions reached. 

 Thus the choice of acceptable alpha and beta lev- 

 els to use in detecting changes in spotted dolphin 

 population size is a management decision based 



primarily on social rather than statistical crite- 

 ria. 



At least two ships are required to provide repre- 

 sentative coverage of the survey area. Although 

 use of a third ship provides better coverage, it 

 does not substantially improve detection of popu- 

 lation decreases. For alpha and beta levels of 

 0.05, a 5% per year decrease can be detected in 

 9 years with use of three ships or 10 years with 

 use of two ships (Table 8). For other alpha and 

 beta levels, use of the third ship only increases 

 our ability to detect specific decreases by about 

 1 year. Given the annual cost of each ship, it 

 would be more cost-effective to conduct the sur- 

 veys for an additional year using only two ships. 

 Another strategy is to conduct surveys less fre- 

 quently than annually. Gerrodette (in press) pro- 

 vides a numerical example of this approach. For 

 parameter values appropriate to spotted dol- 

 phins, conducting surveys less frequently than 

 annually (every second or third year, for example) 

 could save substantial ship time, but more years 

 would elapse before a trend was detected. 



If a 5% annual decrease in population size oc- 

 curred, the number of spotted dolphins killed 

 would have to be large. Assuming a spotted dol- 

 phin population of 2.5 million animals (Table 1) 

 and disregarding natural mortality and reproduc- 

 tion, approximately 125,000 animals would be 

 killed each year. The estimates of all dolphins 

 taken by the fishery during each of the last few 

 years are only about 40,000 animals per year 

 (Hammond and Tsai 1983). It may be unreason- 

 able to expect annual decreases at the 5% annual 

 level; rather decreases of 3% or 1% per year would 

 be more reasonable. If so, two ships would require 

 at least 14 years to detect the decline (Table 8). 



Nonetheless, the number of dolphins actually 

 killed may exceed 40,000 animals per year be- 

 cause dolphin mortality aboard the unsampled 

 trips of U.S. and non-U. S. registered vessels, 

 which is assumed to be similar to that on the 

 sampled trips, may in fact be substantially 

 higher. In addition, the effects of chasing and cap- 

 turing dolphins several times per year are not 

 estimated in our analyses. 



Techniques and data are presented in our paper 

 to determine the optimal number of ships and 

 number of years required to detect decreases in 

 spotted dolphin populations in the eastern tropi- 

 cal Pacific. However, these techniques are appli- 

 cable to investigate the amount of effort and time 

 required to monitor changes in any appropriate 

 population index for any species where sufficient 



445 



