74 
Fishery Bulletin 1 13(1) 
Month 
Figure 3 
Distribution of back-calculated hatching dates for recruits of European hake 
(Merluccius merluccius ) caught in June 2011 during the Mediterranean Interna- 
tional [bottom] Trawl Survey (MEDITS) in the Ligurian Sea and northern Tyr- 
rhenian Sea. 
Results 
At the 13 selected stations, 3123 specimens of Euro- 
pean hake were caught during the 2011 MEDITS in 
the Ligurian Sea and northern Tyrrhenian Sea. Sizes 
ranged from 4.0 to 41.0 cm TL; the length-frequency 
distribution of the specimens belonging to the first 
year class (up to 18 cm TL) is shown in Figure 2. The 
length-frequency distribution showed a single normal 
component with mean size of 9.0 cm TL (standard de- 
viation [SD] 2.0). 
Counts of otolith daily increments from microstruc- 
ture analysis ranged from 77 to 340. The number of in- 
crements within the primordium of the otolith ranged 
between 36 and 70, with an average of 50 (SD 5). The 
results of the rereading of 40 otolith slides used by Bel- 
cari et al. (2006) showed no discrepancies larger than 
10%, a proportion that is considered a reading preci- 
sion threshold (Arneri and Morales-Nin, 2000), provid- 
ing further support for the reading method used in the 
present study. 
Back-calculation of hatching dates provided esti- 
mates of the main hatching period to be from Decem- 
ber 2010 to March 2011, with a peak during January- 
February 2011 (Fig. 3). 
A subsample of 271 individuals was used to fit the 
growth model by means of GAM. The 271 individu- 
als were born during the main hatching period (De- 
cember 2010-March 2011) with a size range between 
4.5 and 13 cm TL and an age range between 77 and 
205 days (Fig. 4); therefore, they represented a single 
cohort. 
Data exploration highlighted collinearity between 
area, bottom temperature, SST, and scalar wind speed 
(Fig. 5), supported by r >0.8 and VIF values >3. There- 
fore, area, SST, and scalar wind speed were removed 
from the model, and a model containing age, bottom 
temperature, fish density, depth, chlorophyll-a, and the 
second-order interaction depth:density as explanatory 
variables was tested by means of analysis of variance 
and backward selection: 
Length = a + fi(Age) + fcfbottom T) + f^(density) 
+ f\(depth) + f$(c\\\-a) + fs(depth:density) + £p (2) 
Based on backward selection, the best model (Table 
2), with variables significant only at the 5% level and 
with the lowest AIC (689.3), contained only age and 
density as explanatory variables: 
Length = a + f\(Age) + f^density ) + £p (3) 
That final model explained 81% of the total devi- 
ance, with a generalized cross-validation score of 0.727. 
The multiple linear regression model fitted with the 
same explanatory variables had heterogeneity within 
the model residuals, as well as a higher AIC (704.2). 
The analysis of variance between these models was sig- 
nificant (F=9.221, P<0.05), indicating that the smooth- 
ers were not linear. 
The explanatory variable age had a linear effect, al- 
though with some fluctuations, especially in the first 
