19 



given time, but they may be important ecologically by providing a means of genetic 

 exchange between populations, as demonstrated for the Sarasota dolphin community 

 (Duffield and Wells 1991). 



Comparison of Abundance Estimation Methods 



Methods 2, 3, and 4 produced similar estimates of population size (Table 2) even 

 though the sampling units and calculations differed. All three of these methods have 

 similar assumptions: a closed population, an equal probability of sighting all animals, 

 random samples of dolphins resighted, and permanent and reliable marks on the 

 dolphins. 



To detect a trend in abundance, the method with the lowest bias, greatest 

 precision, and easiest implementation in the field would be preferred. The accuracy 

 of the estimates depends greatly on the adherence to the assumptions above. The 

 problem of heterogeneity of sighting probabilities can cause a negative bias in the 

 estimate of N (e.g., Hammond 1986), and has been shown to occur in mark-resight 

 studies on bottlenose dolphins in Sarasota Bay (Wells and Scott 1990). To examine the 

 effects of heterogeneity on the different methods, a greater understanding of the 

 community structure of the area is necessary. Method 3, the mark-resight method, 

 attempted to reduce the potential effect of heterogeneity by balancing the coverage 

 of the regions within the study area, under the assumption that multiple 

 communities of dolphins having restricted home ranges could be over- or under- 

 sampled if coverage is not equal for all regions. Piecing together segments surveyed 

 over a period of several weeks, however, could lead to biases if the assumption of 

 population closure was violated. This assumption, based on the dolphin communities 

 of Sarasota Bay, could be tested when the movements and ranges of Tampa Bay 

 dolphins are better known. 



The precision of the estimates is largely a result of the size and number of the 

 samples and the proportion of marked dolphins in the population (M/N). Three of 

 the above methods illustrate a range of compromises that can be made between the 

 first two factors. The mark-proportion method (Method 2) sampled individual 

 dolphin schools as units; this led to a large number of replicates, but the small size of 

 these schools (mean school size » 5.85 + 6.012 SD, n = 480) led to relatively high 

 variation in the proportion of marked dolphins in the groups. Alternatively, the 

 resighting-rate method (Method 4) used the entire survey season as a sampling unit, 

 yielding large sample sizes per season (about 200-600 dolphins), but at the expense of 

 replicate sampling. The mark-resight method (Method 3) used one to three "complete 

 surveys" of the area as a sampling unit, and about 100-380 dolphins per field season, 

 with sample sizes of about 20-170 dolphins per survey. The CVs calculated from 

 Methods 2 and 3 were both acceptably low, although they cannot be compared 

 directly because of the difference in variance methods (Method 2 = non-parametric 

 bootstrap; Method 3 = binomial). 



All of these methods may be prone to a negative bias due to heterogeneity of 

 sighting probabilities, but this would be particularly true for Methods 2 and 4 if care 

 was not taken to survey all areas at least some time during the six-week period. The 

 similarity of the estimates from Methods 2, 3, and 4 suggest that, in practice, the 

 effect of this potential bias due to unequal effort in different regions was relatively 

 small. Estimates from Methods 2 and 4 averaged 6.0% and 8.0% lower than those of 

 Method 3, but a Wilcoxen paired-sample test revealed no significant differences 

 between any of these methods. 



