590 



Fishery Bulletin 104(4) 



In addition to the expected outcome of higher vari- 

 ance as data were removed, we also found a slight ten- 

 dency toward overestimating stock productivity that 

 resulted from eliminating multiple years of data. When 

 fishery-dependent length and age information were re- 

 moved, the model tended to predict slightly higher bio- 

 mass, lower fishing mortality, and faster population 



075 



1 2 3 4 5 6 



Simulated duration of moratorium (years) 



Figure 6 



Proportion of projected red porgy Pagrus pagriis biomass estimates 

 for the year 2016 under Amendment 12 for which the biomass was 

 greater than Bjjj,^. The mean likelihood for 50 model runs for each 

 moratorium duration is shown (with 95% confidence intervals) and 

 the range of the 50 model runs is shown in gray. For a management 

 option to be recommended, there must be at least a 50% likelihood 

 (horizontal line) of the biomass being above JB,,t,.y. Fishery-dependent 

 length and age data were removed from stock assessment model runs 

 for the number of years specified. 



5 



1 00 n 



0-75 



0.50 



S 0.25 



0.00 



Figure 7 



Stock projections (50 model runs for each moratorium duration) under 

 Amendment 12 fishing mortality levels. Fishery-dependent length 

 and age data were removed from stock assessment model runs for 

 the number of years specified. For a management option to be recom- 

 mended, there must be at least a 50% likelihood (horizontal line) of 

 the population biomass being greater than B^g^ before the year 2016 

 (vertical line). 



recovery. This finding was also supported by Chen et 

 al. (2003), for whom data removal led to a more opti- 

 mistic estimation of the current fishery status and stock 

 productivity. This bias would increase the potential for 

 overharvesting because the population may not be as 

 productive as the model results suggest. The results of 

 our study are specific to the South Atlantic red porgy 

 stock and may not be applicable to other 

 stocks or species. Stock status indicators 

 and productivity estimates are affected by 

 life history parameters, population size, 

 historical and current fishing pressure, 

 and data availability and these will dif- 

 fer among species. Additionally, use of 

 an alternative stock assessment model 

 could also affect results. However, if this 

 bias is present for other species and stock 

 assessment models, it could have serious 

 implications for managers because it im- 

 plies that the less information we have 

 on a population, the more productive we 

 would estimate (incorrectly) the popula- 

 tion to be. 



In general, removing red porgy length 

 and age data had only minor effects on 

 status indicator estimates, projections, 

 or management decisions. The signal of 

 a declining red porgy population was so 

 strong and relatively consistent among 

 data sources that removing a portion 

 of this information did not greatly af- 

 fect results. All model runs showed an 

 overfished population but one that is not 

 currently undergoing overfishing, and 

 management decisions were also rela- 

 tively consistent. In all cases, projections 

 showed that rebuilding to B^,^y by the 

 year 2016 would occur with no fishing 

 mortality or a moratorium, and fishing 

 mortality representative of Amendment 9 

 would be insufficient to rebuild the stock. 

 Differences in management decisions did 

 exist for model runs conducted under 

 Amendment 12 fishing mortality levels; 

 some data-poor models incorrectly identi- 

 fied Amendment 12 as a suitable option 

 for rebuilding the stock. For a fish popula- 

 tion closer to an overfished status or to a 

 condition where overfishing is in progress, 

 or where data sources were contradictory, 

 data removal could affect stock assess- 

 ment results and management decisions 

 to a greater extent. 



Because red porgy is a well-studied 

 species for which there are 30 years of 

 information from multiple sources, we 

 removed a relatively small proportion of 

 the data used in the stock assessment for 

 our model runs. Conducting these types 

 of simulations on a lesser-studied spe- 



