Davis and Berkson Effects of a simulated fisfiing moratoiium on tfie stock assessment of Pagrus pagrus 



589 



that remained constant among model runs included 

 natural mortality, maturity at age, and selectivity. 

 Projections for each model run were repeated 100 

 times for each simulation. The resulting projections 

 allowed us to investigate how data reduction would 

 affect management decisions. For each moratorium 

 length, we estimated the probability of rebuilding 

 success as the proportion of the stochastic runs in 

 which the biomass reached B„.sv before the rebuild- 

 ing deadline of year 2016. For a management option 

 to be recommended, the SAFMC wanted at least a 

 507r likelihood of the population biomass exceeding 

 Bn,_^y before 2016. 



Results 



Variability among the simulations increased when 

 multiple years of fishery-dependent length and age 

 data were removed (Fig. 4, and Fig. 5, A and B). 

 When only one or two years of fishery-dependent 

 length and age data were removed, variability and 

 status indicator estimates (such as B/B,;^y and 

 FIFj^gY^ did not differ substantially from simula- 

 tions where no data were removed. However, when 

 three to six years of fishery-dependent data were 

 omitted, variability increased, especially for Fl 

 ^MSY <Figs. 4 and 5B). These trends in variability 

 were also evident in population projections, where 

 the range and confidence interval width of pro- 

 jected outcomes increased as data were removed 

 (Fig. 6). 



Additionally, removal of data led to higher es- 

 timates of population productivity (i.e., higher 

 biomass and faster population rebuilding). When 

 greater than three years of data were removed, 

 projections tended to predict a slightly faster re- 

 covery (Figs. 6 and 7). Data availability would have 

 affected management decisions for Amendment 12 

 because many data-poor models would have incor- 

 rectly identified Amendment 12 as a suitable man- 

 agement option (Figs. 6 and 7). 



Discussion 



Removing data from fishery-dependent sources for 

 multiple years increased the variability of stock 

 status indicators and projections. This variability 

 was amplified during population projections, when 

 the likelihood of rebuilding became increasingly 

 variable as data were removed. Projections are, by 

 nature, uncertain, but this uncertainty increased 

 with even small reductions in information. Similar 

 increases in variance were found by Chen et al. 

 (2003) when information was removed from a New 

 Zealand abalone (Haliotic iris) stock assessment 

 model. When Chen et al. (2003) removed fishery- 

 independent length data, standard deviation of 

 parameter estimates increased 33%. 



08 

 07 



nCVof S2001/Bms>' 

  CV of F2000/FMSY 



12 3 4 5 6 



Simulated duration of moratorium (years) 



Figure 4 



Coefficient of variation (standard deviation among 50 simu- 

 lations for each moratorium duration divided by the meani 

 for red porgy iPagrus pagrus) for B2oii/^M.sy ^"^d f 2000'^A/sy 

 Fishery-dependent length and age data were removed from 

 the stock assessment model runs for the number of years 

 specified. 



06. ^ 



05 - 

 04 

 03 

 02 



62001 /Sm 



0.7 



06- 



05 



04 



03 



02 



01  



00 



B 



F2000/F„sv 



r*n 



12 3 4 5 



Simulated duration of moratorium (years) 



Figure 5 



Mean estimates and 959; confidence intervals of (A) B.,,,,,/ 

 B^igy and (B) ^2ooo^-^,usv f*"' ^0 model runs for each mora- 

 torium duration for red porgy iPagrus pagrus). Fishery- 

 dependent length and age data were removed from stock 

 assessment model runs for the number of years specified. 



