138 



Fishery Bulletin 104(1) 



Homogeneity of allele and genotype distributions 

 among samples was examined with exact tests; signifi- 

 cance of probability values was assessed by a Markov- 

 chain method, as implemented in Genepop and using the 

 same Markov-chain parameters as above. The degree of 

 differentiation between pairs of samples was estimated 

 as Weir and Cockerham's (1984) 9. as implemented in 

 F-Stat. Sequential Bonferroni correction (Rice, 1989) 

 was applied for all multiple tests performed simultane- 

 ously. Spatial (geographic) differences among samples 

 was assessed from multilocus data by estimating the 

 likelihood that any given individual could be assigned 

 to the sample locality from which it was drawn. The 

 Bayesian method of Rannala and Mountain (1997), as 

 implemented in Geneclass vers. 2.0 (Piry et al., 2005), 

 was used to "assign" sampled individuals to a locality; 

 the probability that an individual belonged to a given 

 locality was calculated by using the resampling algo- 

 rithm in Paetkau et al. (2004) and was based on 1000 

 simulated individuals. A locality was excluded as a 

 potential origin of a given individual if the probability 

 of the individual belonging to that locality fell below a 

 threshold level of 0.05. 



Temporal changes in allele frequencies between the 

 two cohorts were used to estimate variance effective 

 size (Afj.y.) at each locality. This "temporal" method (Wa- 

 ples, 1989) estimates effective size from the temporal 

 variance in allele frequencies over the time interval 

 between sampling, thus providing a contemporaneous 

 estimate of A^,.. The pseudo-maximum-likelihood method 

 described in Wang (2001) was used to obtain estimates 

 and 95% confidence intervals of N^y by using the pro- 

 gram MLNE available at http://www.zoo.cam.ac.uk/ 

 ioz/software.htm#MLNE. The 95% confidence intervals 

 were obtained as the range of support associated with 

 a drop of two logarithm units of the likelihood func- 

 tion, as inferred from the likelihood distribution (Wang, 

 2001). We used the analytical method developed by 

 Jorde and Ryman (1995, 1996) to account for effects of 

 overlapping generations on temporal-method estimates 

 of N,. In a population with overlapping generations, the 

 magnitude of temporal allele-frequency change is depen- 

 dent in part on age-specific survivorship (/,) and birth 

 rate (&,). Survivorship was calculated by assuming an 

 equal probability (S) of surviving from one year class to 

 the next and equal probability of survival of males and 

 females. The value of -S (0.56 for Texas and 0.604 for 

 Louisiana and Alabama) was estimated by using age- 

 structure data of red snapper to calculate age-specific 

 survivorship </^=S' ') for each age class ;. Birth rate was 

 estimated by calculating mean individual (wet) weight 

 at each age class, as an indicator of relative gamete con- 

 tribution. Individual weights averaged across males and 

 females within each age class were determined by using 

 von Bertalanffy equations (Fischer et al., 2004) for red 

 snappers at each locality; this mean value was then 

 multiplied by /, to obtain the proportional contribution 

 of each age class to offspring (p, ); p, values were then 

 summed over k age classes. Mean individual weights at 

 each age class were divided by 



k 



i=l 



to produce a standardized birth rate (6,), corrected to 

 reflect a nongrowing population with stable age struc- 

 ture, i.e.. 



Both age-structure and individual (wet) weight data 

 were from the commercial and recreational catch of 

 red snapper in the northern Gulf were provided by D. 

 Nieland of Louisiana State University. Resulting life- 

 history tables were used to calculate a correction factor 

 (C) for overlapping generations by using 100 iterations 

 of Equation 5 in Jorde and Ryman (1996). The value C 

 can be defined as a correction term that is determined by 

 the particular values of/, and 6, of the population under 

 study. G, the mean generation length in years, was calcu- 

 lated by using Equation 10 in Jorde and Ryman (1996). 

 Values of C and G obtained for each locality were sub- 

 sequently used to correct estimates of TV, by N^,^. = N^, x 

 [C/G], where N^, is the pseudo-maximum-likelihood 

 estimate of variance effective size obtained by follow- 

 ing Wang (2001). C and G values, respectively, for the 

 three localities were 10.1 and 6 (Texas), 12.1 and 6.1 

 (Louisiana), and 10.5 and 6.8 (Alabama). 



Results 



Summary statistics (number of alleles, allelic richness, 

 gene diversity; and results of tests of HW equilibrium) 

 for each sample are given in Appendix Tables 1 and 

 2. Number of alleles among all samples ranged from 

 4 to 7 at Prs260 to 20-23 at P;-s248, and averaged 

 (±SD) 11.67 ±5.15 (1995 cohort) and 11.30 ±5.02 (1997 

 cohort). Allelic richness generally paralleled the number 

 of alleles. Gene diversity among all samples ranged 

 between 0.178-0.238 (Lca20) and 0.898-0.915 (Prs257), 

 and averaged (±SD) 0.597 ±0.224 (1995 cohort) and 

 0.602 ±0.217 (1997 cohort). No significant difference 

 in allelic richness (P=0.35) or gene diversity (P=0.07) 

 was detected. 



Four of 114 tests of conformity to Hardy- Weinberg 

 equilibrium expectations were significant following 

 Bonferroni correction. These included two tests in the 

 1995 cohort {Prs275 in the Texas sample and Prsl37 

 in the Alabama sample) and two tests in the 1997 co- 

 hort (Lcn22 in the Texas sample and Prs229 in the 

 Louisiana sample). F^^, values over all loci for all four 

 samples ranged between 0.008 and 0.029 (Appendix 

 Tables 1 and 2). A total of 21 of 1026 (pairwise) tests 

 of genotypic disequilibrium were significant (P<0.05) 

 after Bonferroni correction. All 21 involved different 

 pairs of loci (i.e., only one out of six possible tests for a 

 given pair combination was significant) except for Lca64 

 and Prs328 in the 1995 cohort from Alabama and the 

 1997 cohort from Texas, and Lca64 and P7-s248 in both 

 cohorts sampled from Texas. 



