Wetzel et at: The effect of reduced data on monitoring overfished fish stocks 
199 
Full data Eliminated data 
Figure 5 
Relative error (RE) of spawning biomass and the estimates of steepness 
for the full and eliminated data scenarios for the time-invariant case; 
results are divided by whether the simulated stock was estimated to be 
rebuilt (65 simulations [gray]) or not (35 simulations [white]) for the elim¬ 
inated data scenario. The black lines in the gray boxes denote the median 
of the estimates, the gray boxes cover the 25-75% simulation interval, 
and the boxplot whiskers indicate the 95%> simulation interval for each 
assessment year. 
biomass were considerably less than the true values in 
the first assessment year (Fig. 5B ) for the 35 simula¬ 
tions in which the stocks were estimated not to rebuild 
by the end of the management period. The underesti¬ 
mates of spawning biomass (Fig. 5B) were driven by 
estimates of steepness that were much less than the 
true value in the first assessment (Fig. 5D). In the 
absence of new data, the underestimates of steepness 
resulted in the estimation method perceiving a less 
productive stock that required an extended period to 
rebuild to the target biomass. However, with full data 
present, estimated quantities (spawning biomass and 
steepness) improved for this subset of simulations and 
were median unbiased by the end of the management 
period (Fig. 5, A and C). 
The median number of years estimated for the simu¬ 
lated stocks to recover to the target biomass for the 
full data scenario was longer than the median time re¬ 
quired to rebuild the stock within the operating model 
simulations (Table 2). In contrast, both the reduced 
and eliminated data scenarios had shorter median re¬ 
covery times than those of the operating model (Table 
2). The contrast in estimated recovery times across the 
data scenarios was related to the average catch ob¬ 
tained during rebuilding along with the bias and vari¬ 
ability of estimates. The median error associated with 
relative spawning biomass for the full data scenario 
was less than zero, and there was low among-simula- 
tion variability (compared with those of the other data 
scenarios) for all assessment years, which resulted in 
estimates that predicted constant rebuilding but at a 
slower rate than the true rate of the simulated stock 
in the operating model (Fig. 3D). In contrast, the re¬ 
duced data scenario had higher variability over time 
(i.e., within-simulation) across the estimates of error 
associated with relative spawning biomass (Fig. 3E). 
The variability of estimates between assessments re¬ 
sulted in simulated stocks that were estimated to be 
recovered to the target stock size when the populations 
in the operating model were not yet recovered because 
of estimation error driven by the limited number of 
composition samples during rebuilding. 
