46 
Fishery Bulletin 11 6(1) 
Table 4 
Interannual performance of 2 models used to incorporate an environmentally forced recruit¬ 
ment of European anchovy (Engraulis encrasicolus) in the Gulf of Cadiz, Spain: the dual-time 
resolution model of this study and the monthly resolution model of Ruiz et al. (2009). Values 
are the root mean squared errors between monthly means of estimations from the models and 
observations from available data (1988-2004). The model of Ruiz et al. (2009) did not estimate 
catches. Maximum and minimum values for observed catches are 0 and 136 millions of fish, 
respectively, and maximum and minimum values for observed catch per unit of effort (CPUE) 
are 0.138 and 1.148 tons/fishing trip, respectively. 
Current Model from 
model Ruiz et al. (2009) Difference 
Catch (millions of fish) 5.235 
CPUE (tons/fishing trip) 0.276 0.2734 0.0028 
Acoustic survey 1993 (millions of fish) 572.9414 185.3100 387.6314 
Acoustic survey 2004 (millions of of fish) 107.4575 180.7900 -73.3325 
The dual resolution also facilitates the embedding of 
environmental forcing into generalist models, such as 
the GPDM for fishery management (Mantyniemi et al., 
2015). This integration results in coherence between 
the model outputs and observed data beyond the initial 
stages of the life cycle, see for instance Figures 5 and 
Supplementary Figure 2 (online only) that show coherence 
between the size structures produced by the model and 
those reported by ICES. The coherence seems to indi¬ 
cate the suitability of the von Bertalanffy model used 
for Equations 4 and 5, and the bounds settled on for 
growth parameters provided by Bellido et al. (2000) 
(see Table 2). Other features observed in the popula¬ 
tion size-structure also seem to be well reproduced by 
the model. Therefore, the absence of large sizes and 
the high concentration of individuals between 10 and 
14 cm TL are consistent with the suggestion that fish¬ 
ing pressure hampers adult survival beyond a year and. 
the population relies on the youngest fish (BOTTOP 
control, Ruiz et al. (2007)). Nevertheless, discrepancies 
between model and data frequently occur in the first 
two length classes (10-12 cm TL and 12-14 cm TL, 
Suppl. Fig. 2) (online only). This result probably reflects 
the need to ameliorate the transition matrix G in fu¬ 
ture versions of the model. In this respect, the inclu¬ 
sion of another length class in the model (e.g., 8-10 
cm TL, representing mostly immature fish) seems rea¬ 
sonable according to the recent report of landings data 
(ICES 6 ). Further, the use of a seasonal growth model 
incorporating temperature effects may improve the ob¬ 
served results. 
It is worth underscoring that the model did not pre¬ 
viously use any of the in situ data for estimates of ear- 
ly-stage abundance presented in Figure 4 or with the 
population size structure reported by ICES (and used 
in the contrast of model versus observation in Figure 5 
and Supplementary Figure 2) (online only). Therefore, the 
Bayesian integration of environmental forcing in tradi¬ 
tional formulations of fishery management results in 
outputs that are coherent with field observations. This 
coherence holds over the whole life cycle, even when 
the observational data have not being previously used 
in the configuration of the model. 
The main feature of the time series, the collapse 
and recovery between 1994 and 1996, is described 
in previous works and points to the combined effect 
of wind and discharges over recruitment (Ruiz et al., 
2006, 2009). The resulting drastic variation in land¬ 
ings is also well resolved by the model. The posteriors 
of A and p in Figure 2 indicate, nevertheless, a higher 
role of the wind regime in driving the dynamics of the 
population, consistent with the analysis presented by 
Rincon et al. (2016). Consistent with a recruitment- 
driven fishery, the model transfers these major changes 
in recruits (Fig. 3A) into large fluctuations of landings 
(Fig. 3B). 
These types of dynamics can be reproduced only by 
models incorporating the mechanics of environmental 
forcing on recruitment. Classic models, such as virtual 
population analysis or extended survivor analysis, may 
be valid for assessing species with longer life spans. 
The impact the environment may have on these species 
results in abundance or size-structure changes that are 
slow enough to be detected by the models used in year¬ 
ly evaluations of the stock, even when those models do 
not incorporate environmental forcings. However, these 
models are not valid for recruitment-dependent fisher¬ 
ies such as that targeting the European anchovy in the 
Gulf of Cadiz because stock changes are too fast. More¬ 
over, knowledge of the cause and effect of environment 
on recruitment (beyond simple statistical correlations) 
is necessary to reach some degree of predictability in 
cases where severe and quick changes in recruitment 
pose heavy challenges for modeling. 
The model presented here confirms the critical role of 
the environment in shaping recruitment and landings 
of the European anchovy in the Gulf of Cadiz. Com¬ 
pared with previous models, its dual-time resolution 
