1022 



Fishery Bulletin 97(4), 1999 



The Jolly-Seber analysis indicated that the open 

 population model explained in a better way the data 

 for both Revillagigedo and coastal stocks (Table 5). 

 In the Revillagigedo stock the goodness of fit x' test 

 on model A (general model) did not reject the null 

 hypothesis (P=0.0667), but the result showed that 

 dataarebetter explained (P=0. 799) by model 2 (tem- 

 porary trap response model), assuming that survival 

 rates of newly captured animals differ from those of 

 previously captured animals. In other words, this test 

 indicates that some animals may have moved into 

 the capture site and once captured, they could have 

 left and even re-entered the sampling area. This in- 

 ference an be supported by the fact that data for this 

 stock were collected only in Socorro Island and that 

 some animals may have a preference for anther is- 

 land, such as Clarion, far away from the sampling 

 area (Fig. 1). For the coastal stock, the goodness-of- 

 fit x~ tests indicate that the model A (P=0.87.54 ) and 

 model 2 (P=0.8892) are almost equally good in ex- 

 plaining the data. Although sampling coverage was 

 wider for this stock, allowing a better collection of 

 photographs for moving anirfials, this problem still 

 remained, but had a smaller effect on the estimates. 



Having accepted that the Jolly-Seber model ex- 

 plained the data well for the two stocks, population- 

 size estimates were obtained with the software pro- 



gram RECAP. Results show that the modified esti- 

 mator (modified J-S model) provided more stable 

 estimates and narrower and more realistic confidence 

 intervals (Table 6). 



Because capture probabilities are larger than those 

 obtained by the closed population methods, and con- 

 sidering that the Jolly-Seber estimates may be nega- 

 tively biased given heterogeneity, population size are 

 likely overestimated by the closed population methods. 



In light of the above arguments, we conclude that 

 the best estimates of population size for the Mexi- 

 can stocks, using the data at hand, are those pro- 

 vided by the modified Jolly-Seber model of program 

 RECAP. Estimates obtained with closed population 

 methods are not recommended when data are pooled 

 for the whole extension of the study; however, it may 

 be worth obtaining estimates with pairs of years with 

 models testing for heterogeneity and capture prob- 

 abilities varying with time, although certain infor- 

 mation and precision may be sacrificed. 



Estimates for each year show an apparent trend 

 towards increase for both stocks. Such a trend was 

 considered an artifact of a reduction in the bias 

 caused by heterogeneity in capture probabilities af- 

 ter sample size increased with time. Consequently, 

 we consider that the best estimates for the stock sizes 

 were those obtained for the last years and are 1813 

 (95^^ CI: 918-2505 ) fbr the coastal stock m 1992, and 

 914 (95'7r CI: 590-1193) for the Revillagigedo stock 

 in 1991. 



