66 
Fishery Bulletin 112(1) 
fluctuations are not necessarily random (Jensen, 2002; 
Sinclair and Crawford, 2005). Environmental varia- 
tions may themselves be driven by other, direct or in- 
direct anthropogenic or natural events, as would have 
happened for MWET (e.g., Connelly et al. 7 ; Fogarty et 
al. 8 ). Third, elusive relationships may have been due 
to the shortness of the time series for the regressed 
variables. 
Finally, surplus-production models are oversimplifi- 
cations of the population dynamics in the form of just 
2 or 3 parameters (Laloe, 1995; Keyl and Wolff, 2008). 
Other possible reasons for the blurring or weakening 
of the impacts of MWET on Atlantic Croaker produc- 
tivity could be the noisy nature of the tuning indices 
especially since 1990, the lack of fishing effort that 
precluded the partitioning of any roles between fishing 
intensities and MWET, and the functional relationship 
between the parameter r and MWET. In reality, this 
function is unknown, and alternative functional forms 
(e.g., Freon, 1988; Stenseth et al., 2002; Rose, 2004; 
Hatton et al., 2006) are conceivable and deserve test- 
ing as well. 
Overall, the trends generated by this study behaved 
like those trends obtained through runs of nonequilib- 
rium production models with A Stock Production Model 
Incorporating Covariates (ASPIC, vers. 5.34. 9, which 
is included in the NOAA Fisheries Toolbox, http://nft. 
nefsc.noaa.gov/ASPIC.html) software and Excel spread- 
sheets (ASMFC 1 ). Various implementations of BDMs 
also produced similar estimates of the initial depletion, 
MSY, and Rmsy> and conveyed a common message that 
the Atlantic Croaker stock was exposed to a relatively 
low risk of overfishing in the 2000s. But this analysis 
showed differences from ASMFC’s 1 implementations 
about the opposing stock status prior to the 2000s (Fig. 
6). Contrary to ASMFC’s 1 results, this study indicates 
frequent episodes of overfishing, often with high risks 
of being overfished, that marked the Atlantic Croaker 
stock during the period of records. Likewise, overfish- 
ing of the Atlantic croaker stock may have been high 
during 1993-2001 (along with high risks for this stock 
being overfished) but were low in most years before 
1993. 
The causes underlying these discrepancies would 
require dedicated experimental designs for BDM per- 
7 Connelly, W., L. Kerr, E. Martino, A. Peer, R. Woodland, and 
D. Secor. 2007. Climate and saltwater sport fisheries: 
prognosis for change. Technical Report Series No. TS-537-07 
of the University of Maryland Center for Environmental Sci- 
ence. Ref. No. [UMCESJCBL 07-119. Chesapeake Biological 
Laboratory, UMCES, Solomons, MD. [Available at: http:// 
www.seasonsend.org/pdfs/Saltwater%20Fisheries.pdf; ac- 
cessed May 2012.] 
8 Fogarty, M., L. Incze, R. Wahle, D. Mountain, A. Robinson, 
A. Pershing, K. Hayhoe, A. Richards, and J. Manning. 2007. 
Potential climate change impacts on marine resources of 
the northeastern United States. Northeast Climate Impacts 
Assessments (NECIA). [Available at: http://www.northeast- 
climateimpacts.org/pdf/miti/fogarty_et_al.pdf; accessed May 
2012 .] 
formance analyses, which were not the focus of this 
study. However, all other things being equal (i.e., no 
errors pertaining to fishery removals and parameter 
estimation), the conflicts in the performance of, for ex- 
ample, ASPIC and the BDMs used in this study, can 
be attributed to model uncertainty (Caddy and Mahon, 
1995; Harwood and Stokes, 2003), itself inherently em- 
bedded in the general scientific uncertainty (Ralston et 
al., 2011; Rothschild and Jiao, 2011). 
These conflicts may have been jointly or separately 
rooted in at least 3 major factors. The first factor was 
the difference in BDM structures (continuous formula- 
tion for ASPIC vs. discrete formulation in this study) 
and the way the corresponding estimation approaches 
(frequentist vs. Bayesian) dealt with uncertainty. The 
second factor related to the BDM behaviors resulting 
from the constrained starting values (ASPIC), nonuse 
of starting values (this study), and differing estimable 
parameters. The third and, perhaps, most important 
(Polacheck et al., 1993; Ono et al., 2012) factor was the 
error structures assumed including the specifications 
of the priors’ PDFs (observation error for ASPIC vs. 
observation and process errors in this study). Note that 
observation errors are year-specific, whereas process 
errors can propagate over time (Kimura et al., 1996). 
This study generated inconclusive, somewhat con- 
flicting results about MWET effects on the production 
dynamics of Atlantic Croaker. Specifically, these effects 
were associated with a coefficient without explanatory 
power or with various linear relationships that proved 
weak or negligible in justifying addition of a related 
parameter in BDMs. If BDMs are to be used for assess- 
ing the Atlantic Croaker stock, it appears reasonable to 
continue performing them without considering MWET. 
Unambiguously discerning the extent of MWET effects 
through BDMs will perhaps be possible when longer 
time series of relevant fishery data, winter estuarine 
temperature (or, preferably, direct estimates of kills 
caused by cold winter), and other environmental fac- 
tors will be gathered and accounted for together. 
Conclusions 
Given the well-established effects of the changes in 
winter water temperatures on the production dynamics 
of Atlantic Croaker along the U.S. Atlantic coast, the 
title of Keyl and Wolff’s (2008) article deserves para- 
phrasing: what can (assessment) models do to track 
such effects, modify the perception of the stock, and 
better guide management? The present study has at- 
tempted to answer this question through state-space 
BDMs with and without MWET. BDMs incorporating 
MWET were not statistically supported by the data 
and did not outperform BDMs without MWET. The 
retained BDMs without MWET were associated with 
process errors, surplus production, and instantaneous 
surplus production that indicated that MWET had pos- 
itive effects on Atlantic Croaker productivity. However, 
