Munyandorero: Climate effects on Mtcropogonias undulatus 
59 
Time series of the predicted posterior medians with 95% central intervals for (A and B) the Na- 
tional Marine Fisheries Service-Northeast Fisheries Science Center fall index and (C and D) the 
Southeast Area Monitoring and Assessment Program fall index for the Atlantic Croaker stock 
off the U.S. Atlantic coast, derived from Bayesian state-space biomass dynamic models: without 
minimum winter estuarine temperature, MWET (left panels) and with MWET (right panels), 
1972-2008. Filled circles are observed data; solid, dotted, and dashed lines relate to model base 
runs and model sensitivity runs with the prior 17(0.01,0.15) for the intrinsic rate of population 
increase r or the scaling parameter tq, and including the southeastern (North Carolina-east 
Florida) shrimp trawl fishery bycatch, respectively. Models in which the prior for the coefficient 
controlling MWET effects (a) was centered on zero with a precision 0.02 produced similar cred- 
ible estimates. These estimated values were not plotted for clarity. 
prior a~N(0,0.02) (Fig. 4). The stock availability coef- 
ficient associated with the NEFSC index (hence, the 
NEFSC Bottom Trawl Survey catchability) was 8 times 
higher during 1994-2008 (AnefsC94-08 = 0.7-0. 8) than 
during 1972-93 (AnefsC72-93 =0.11). The stock avail- 
ability coefficient, global efficiency (=l.llx 10~ 3 ), and 
catchability by survey and period were insensitive to 
model configurations. 
The components of the DIC statistics (Table 2) 
for models without MWET were the NEFSC index, 
SEAMAP index, depletion time series, and process 
error variance. They included the coefficient a for 
models incorporating MWET. Model fits were largely 
dominated by biomass indices (98-99% of DIC sta- 
tistics, of which the NEFSC index amounted to 68- 
73%). The total estimates of po were positive and 
consistent across models, supporting the evidence 
that all models generally had the same number of 
parameters. 
Other DIC statistics for models of the study were 
greater for (base and sensitivity) M2 than for (base and 
sensitivity) Ml and, among model configurations, were 
the least for MlrU and M2rU. The former result for 
competing models, with or without MWET, was mainly 
due to the extra parameter a that clearly had no ex- 
planatory power. The DIC associated with a was 4.17 
for all M2s and represented 92-101% of ADIC (Table 
2). For the competing models with alternative priors 
for the parameter r or rp, lower DIC statistics for MlrU 
and M2rU resulted from the decrease in D and D for 
biomass indices — a decrease that largely contributed to 
ADIC. This finding indicates an improvement in fit as 
also evidenced by a slight increase in the correspond- 
ing pi). However, this situation was counterbalanced by 
