480 



Fishery Bulletin 105(4) 



B Type 2 



Figure 4 



Prior (dotted lines) and posterior distributions for over- 

 dispersion parameters (K) by data type (type 1, where 

 data were available for year, month, and PFMC area for 

 each individual tag recovery; type 2, where data were 

 available for only month and area; type 3, where data 

 were available for area only; and type 4, where data 

 were only available to indicate recovery inside or outside 

 the area of tagging for model 11. Priors are scaled to be 

 more visible on the plot. 



southerly movement on equilibrium harvests of adult 

 English sole should be explored because this movement 

 would have implications for current and future manage- 

 ment strategies. 



Given the presence of significant outliers, the degree 

 of support among models based on the posterior prob- 

 ability should be interpreted with caution. There are 

 many potential reasons for the outliers in observed re- 

 coveries present in the data. Although an attempt was 

 made to retain a structure commensurate with what 

 might be possible to replicate in a stock assessment, 

 the analysis may simply not be complex enough. Many 

 parameters, such as fishing mortality, reporting rate, 

 and tag loss rate, are assumed to be temporally and 

 spatially invariant because of a lack of available data. 

 Spatial differences in the location of the initial tagging 

 within the larger PMFC area were not accounted for 

 and reporting rates could be variable due to a mixture 

 of Canadian and U.S. fishing vessels with varied incen- 

 tives for returning tags. During much of the time period 

 over which tags were recovered, there was a substantial 

 fishery for mink food; this fishery may have had very 

 different handling and tag-recovery practices than those 

 of the concurrent fishery targeting fish for human con- 



sumption. The relative effects of violating these various 

 assumptions could be addressed in the future through 

 simulation testing. 



Spatially and temporally local increases in fishing- 

 induced mortality rates, resulting in additional recap- 

 tured tags, could have generated many of the positive 

 residuals from a model that does not allow fishing mor- 

 tality to vary within a single data set. Exploration of 

 how model parameters are influenced by localized in- 

 creases in fishing mortality should be explored through 

 future simulation analysis. Another potential use of this 

 method is to extend the hierarchal model of fishing mor- 

 tality rates, allowing them to vary over space, time (or 

 both) within a data set. Estimation of the coefficient of 

 variation of this distribution could also be explored. The 

 coefficient of variation of total U.S. catch over the years 

 in which tagging projects were conducted for English 

 sole is on the order of 0.5, indicating that interannual 

 variation in fishing mortality rate may be an important 

 factor absent from the analysis. Similarly, a hierarchi- 

 cal approach could be taken with regard to movement 

 parameters between specific areas. Geographic regions 

 could be defined on the basis of likely bathymetric fea- 

 tures such as submarine canyons or rocky headlands 



