15 



meeting the assumptions associated with generating abundance estimates from 

 mark-resighting analyses. 



Abundance Estimates and Trends 



The catalog-size index (Method 1) resulted in minimum population estimates 

 of 165 to 243 dolphins over the five years of the study, with an average of 203 (Table 

 3). The Method-1 estimates are known to be underestimates because they do not 

 take into account the unmarked dolphins. Methods 2, 3, and 4 attempted to correct 

 for this underestimation. 



Method 2 (mark-proportion method) calculated population-size estimates 

 from proportions of marked animals relative to revised minimum, revised 

 maximum, and final best group size estimates. The differences between minimum 

 and maximum population-size estimates were so small that we present only the 

 estimates based on the final best group size. The number of dolphins estimated by 

 Method 2 ranged from 226 to 422, with an average of 302 (Table 3). 



Method 3 (mark-resight method) provided annual point estimates from the 

 combined sightings made during two or three "complete surveys". The estimates 

 ranged from 238 to 385 across all years, with an average of 313 (Table 3). 



Method 4 (resighting-rate method) provided annual point estimates ranging 

 from 194 to 385 dolphins, with an average of 267 (Table 3). 



The abundance estimates were examined for trends across the five years of 

 the surveys. Population-size estimates varied from one year to the next (Figure 7). 

 The trends in abundance roughly followed variation in field effort, but the 

 relationship did not appear to be strong. Comparison of 95% CL for Methods 2 and 3 

 (Figure 8) indicate a significant difference in the abundance estimates from the first 

 three years compared to the last two years of the survey. 



Power Analysis 



The catalog-size index (Method 1) used a regression analysis of the five 

 annual estimates to remove the effect of a potential trend and calculated a CV of 0.15 

 from the residuals (although no trend was apparent, a test with only five data points 

 would be sensitive to outliers and would have low power). Given that alpha = 0.05, 

 power = 0.80, r = 1.00 or -0.50, and CV = 0.15, we can then calculate the minimum 

 number of surveys necessary to detect a trend. Three survey sessions would be 

 required to detect a decreasing trend and four for an increasing trend. 



A bootstrap variance procedure applied to Method 2 (mark-proportion 

 method) yielded CVs ranging from 0.04 to 0.06, with an average CV of 0.05. This 

 would allow an increasing or a decreasing trend to be detected in two surveys. 



