Li et al.: A comparison of 4 age-structured stock assessment models 
lolol) 
O99 
999 
ROB BOR POR BPO 
O99 
SOO 
POR BOR 
Soo 
leloto) 
0 
0 
0 
C11 
> 
C12 
9990 999 999 9992 S9So SSX o°oF*% 
ROR BOR BOR BOR BOR BOK AOA 
2 ae 
= a 
xt 
163 
ep) 
zx 
Figure 6 
Violin plot of relative error (RE) across years and iterations for spawning stock biomass (SSB), 
recruitment (Rf), fishing mortality rate (F), SSB,,4;, (ratio of SSB to SSB at maximum sustainable 
yield [MSY]), and F,,,;, (ratio of F to F at MSY) for each of 4 estimation models under cases 0-12 
(CO-12). The models, evaluated in this study for use in stock assessments, include the Assessment 
Model for Alaska (AMAK), the Age Structured Assessment Program (ASAP), the Beaufort Assess- 
ment Model (BAM), and Stock Synthesis (SS). 
differences in estimates might have been forced unwit- 
tingly, because of parameter misspecification. 
In addition to differences in selectivity parameterization, 
we also identified differences in how age is modeled in the 
EMs. If ages modeled in SS start with age 0 (the default) 
and not age 1, a mismatch between SS and the other EMs 
would have been induced. In this case, the mismatch would 
manifest only in scaling of recruitment to account for natu- 
ral mortality at age 0. Therefore, identification of common 
features and comparison of source codes are particularly 
important in cross-testing a set of assessment models. The 
comparison framework developed (Fig. 1) and the approach 
