MATHER ET AL.: TAGGED BLUEFIN TUNA 



correct. It is assumed that all instantaneous shed- 

 ding, mortality, and emigration rates are 

 constant — within years and among years. Since 

 fishing was concentrated during the summer sea- 

 son, the assumption of constant fishing mortality 

 is not valid. Also fishing effort probably varied 

 over the years of the study. Available measures of 

 fishing effort are thought to be inaccurate but 

 were used in the last part of the analysis. The 

 validity of the assumption of constant rate of 

 emigration is not known, but tag returns suggest 

 that transatlantic migrations are sporadic. This 

 suggests that the assumption of constant emigra- 

 tion is not valid. While we recognize that some of 

 our assumptions probably are invalid, it is our 

 judgment that the effect of the violations on our 

 results is not serious. 



For the first part of the analysis we assumed 

 that c, 77, and p equal one. In the first analysis 

 minimum variance unbiased estimates of the total 

 annual survival rates were computed by the 

 method developed by Chapman and Robson ( 1960) 

 for the recovery data by year pooled in a variety of 

 different ways over release categories. Confidence 

 intervals were computed fors, the fraction surviv- 

 ing per year, andZ, the associated instantaneous 

 mortality rate. A chi-square test was used to de- 

 termine if the number of recaptures in the first 

 recovery period is compatible with the survival 

 pattern exhibited by the rest of the data, i.e., if the 

 hypothesis of constant F and X is true. This test 

 was applied sequentially, i.e., the second year was 

 defined as the first recapture category and the test 

 repeated until either all recapture years were 

 eliminated or a survival rate was obtained from 

 some subset of the data. For all releases the tag- 

 ging year was taken as the first recapture category 

 at the start of the analysis. The results of the 

 survival rate computations are shown in Table 4 

 for the following data groupings: 



1) Over all years 



2) Over three adjacent release years 



3) Individual years 



4) July releases for three adjacent 



release years 



5) Individual months within years 



6) Release groups as defined in Table 1. 



An obvious feature of this analysis is that in 

 many of the recapture series, the numbers recov- 

 ered the first year or two were higher than ex- 

 pected from the entire recapture series. This re- 



sult is somewhat surprising because it might be 

 expected that the number recaptured the first year 

 would be underrepresented because of less expo- 

 sure to the fishery. Three possible factors that could 

 have caused the higher than expected recaptures 

 in the first year or two after release are: 



1 ) Tagged fish were released into an area where 



fishing activity was concentrated. 



2) The proportion of the population migrating 



into the fishing area decreased as the fish 

 became older; thus the availability of the 

 tagged fish in the fishing area may have 

 fallen off rapidly enough in later years to 

 have caused a disproportionate number 

 of recaptures in the first and second years 

 after release. 



3) The method of estimation assumes constant 



fishing mortality rates and other loss 

 rates; variability in recovery effort could 

 have caused the number of recaptures per 

 year to deviate from a simple exponential 

 decline with time. 



Several aspects of the data emerge from Table 4. 

 The estimates of survival rates are low but highly 

 erratic; restricting the releases to finer time- 

 location grids did not improve the stability of the 

 estimates as might be expected. Since no time 

 trend in survival is evident, pooling overyears is a 

 useful device to average out some of the fluctu- 

 ations in the data. In one sense this is a substi- 

 tute for use of recapture effort statistics and work- 

 ing with the number of recoveries per unit of 

 recovery effort. The recaptures per year were 

 combined over years using various weighting fac- 

 tors to develop adjusted numbers recaptured per 

 year. None of these weighting schemes offered an 

 improvement in the use of the simple unweighted 

 average percent recapture per year at liberty for 

 the years 1964-68. The proportion surviving per 

 year as estimated from the simple average of the 

 percentages was 0.188. This value is well within 

 the confidence interval of the s -value of 0.231 es- 

 timated from the actual numbers pooled over all 

 years. For the latter estimate, however, we did not 

 use the recoveries made during the first 2 yr at 

 liberty. 



Inclusion of the first two recapture periods in 

 the Chapman-Robson analysis, particularly for 

 the last set of release groups which are fairly 

 homogeneous, had the general effect of reducing 

 the survival estimates. The numbers of recaptures 



907 



