Witzell et al.: Origin of Caretta carctta at Hutchinson Island 



627 



Table 3 



Maximum likelihood estimates of contribution by source populations to immature loggerhead turtles from the St. Lucie power 

 plant (n=99). Estimates were generated by using UCON software (Masuda et a!., 1991 ). Standard errors and 95% confidence inter- 

 vals were generated from 100 bootstraps of both the stock and mixture by using GIRLSEM. The source populations NWFL, Brazil 

 and Greece, were removed from the analysis. SFL = south Flordia; and NEFL-NC = northeast Florida to North Carolina. 



Source population 



Contribution 



Standard error 



95% Confidence interval 



SFL 



NEFL-NC 



Yucatan 



0.6965 

 0.0984 

 0.2050 



0.2211 

 0.1151 

 0.1296 



0.6527-0.7403 

 0.07.56-0.1212 

 0.1793-0.2307 



lation indicates that this nesting population is not the sole 

 contributor to this foraging population. 



Maximum likelihood analysis 



The initial maximum likelihood analysis yielded esti- 

 mates of contribution that were associated with high 

 standard errors (Table 2). The estimate for NWFL had an 

 extremely high standard error This high standard error is 

 an indication of sampling error associated with the source 

 population (Epifanio et al., 1995). Because of the large 

 standard errors, an attempt was made to reduce the com- 

 plexity of solutions that were involved in the maximum 

 likelihood algorithm's search. Removing a potential source 

 population decreases the potential number of "answers," 

 consequently reducing the standard errors about the 

 mean estimates. 



The maximum likelihood analysis was repeated, remov- 

 ing the smallest and most distant source populations 

 (>AVFL, Brazil, and Greece) as potential contributors 

 (Table 3). In addition to the statistical assumption that 

 the large variation associated with several estimates is 

 partially due to sampling error, we also made some bio- 

 logical assumptions. By removing the source populations 

 we assumed that they were not contributing to this forag- 

 ing area at levels sufficient to detect in this ML analysis. 

 The nesting effort in NWFL may include 100-200 turtles 

 annually (Meylan et al., 1995), as compared to tens of 



thousands of turtles that nest in southern Florida. Hence, 

 the Florida panhandle is an important nesting area, but is 

 probably too small to detect with precision in ML analyses. 

 The removal of Greece as a potential source population 

 was based on the results of the first analysis that indi- 

 cated that this source population did not contribute at de- 

 tectable levels (Table 1). Brazil was also removed from the 

 analysis because there were no indications (either from 

 the observed haplotypes or from the initial ML analysis) 

 that this population contributed individuals to the forag- 

 ing population at Hutchinson Island. 



The results from the final ML analysis provided the 

 most realistic estimate of the genetic composition of 

 this area (Table 3). This analysis indicated that the SFL 

 subpopulations is a major contributor to the Hutchinson 

 Island foraging population. In addition, the Yucatan nest- 

 ing population also appears to contribute a significant 

 percentage of individuals to this area. Although the small 

 northeast Florida to North Carolina (NEFL-NC) rookery 

 appears to contribute some individuals, the standard er- 

 ror is still large; therefore, the estimate is not precise. 

 Individuals from the NEFL-NC nesting population are 

 present, but an accurate estimate (finer than 7-12%) is 

 not possible at this time. 



The results of the analysis of the St. Lucie foraging pop- 

 ulation are provisional and there are several limitations to 

 the maximum likelihood analysis. A major assumption of 

 this type of analysis is that all potential source populations 



