Fabrizio et al.: Home range and seasonal movements of Centropristis striata 
87 
keeping with this approach, we do not report P-values 
here but instead interpret the “importance” of factors 
on the basis of their contribution to the model. 
Analysis of activity 
We similarly modeled the effect of sex, time, release 
group, and environmental factors on the mean activ- 
ity index of those fish that exhibited movement, but 
here we used a general linear mixed model (GLMM) 
fitted with the MIXED procedure in SAS (vers. 9.3; Lit- 
tell et al., 2002). To meet the homogeneity-of-variance 
assumption, the activity index was log e transformed. 
Individual fish were treated as a random factor in the 
model, allowing us to estimate variation among indi- 
viduals. As before, we modeled the effects of the fol- 
lowing fixed factors on the activity of fish: continuous 
linear, quadratic, and cubic time (standardized to elim- 
inate collinearity), time period (dawn, day, dusk, and 
night), sex, season, release group, mean water tempera- 
ture at station Bl, and mean temperature and salinity 
differences between stations H7 and Bl (standardized). 
A repeated-measures approach was used here as well, 
with fish nested within time period as the subject. We 
used the following repeated-measures mixed model to 
fit the data: 
^i(m)jkln ~ A ^i(m) Aj + At + Y\ + ^ m + Di Gi + K 
+ a + interactions + £i(m)jklmn> 
where Tj( m )jkln = the activity index of the i th fish of the 
j th sex of the k th tagging group in 
the 1 th season nested in the m th time 
period of the n th day (time); 
/u = the expected activity index; 
«i(m) = the random effect of the i th fish nested 
in the m th time period; 
interactions refers to 2- and 3-way interactions 
between the fixed effects; 
e i(m)jklmn = the random unexplained error; and 
other terms are as defined before. 
We considered 2- and 3-way interactions in this 
model because such interactions greatly reduced the 
AIC; however, inclusion of more than two 3-way inter- 
actions or higher-order interactions resulted in severe 
loss of precision, and such models were abandoned. 
With the GLMM, we assumed that the response and 
the random error are normally distributed and that the 
variance of the response is homogeneous across the lev- 
els of the factors included in the model. We modeled 
heterogeneity in the covariance structure among dawn, 
day, dusk, and night in the GLMM through the use of 
the group option in the MIXED procedure and speci- 
fied the Kenward-Roger method for calculating degrees 
of freedom (Kenward and Roger, 1997). Model building 
followed the approach in Zuur et al. (2007) and Bolker 
et al. (2008); the random structure of the model ( i.e. , 
the appropriate variance-covariance structure and the 
importance of the random factor) was identified with 
REML-based estimates of AIC. Compound symmetry, 
autoregressive with lag 1, autoregressive moving aver- 
age with lag 1, and power covariance structures (Lit- 
tell et al., 2006) were used to model the correlations 
among the repeated responses. Next, using the random 
structure identified in the previous step, we calculated 
AIC values for models that contained different fixed 
effects, using maximum likelihood; those interactions 
that reduced AIC values were considered important 
and retained. Finally, REML was used to compute the 
final model parameters. 
Results 
Acoustic data were obtained from 70 of the 72 deployed 
receivers and most detections occurred during the sum- 
mer (from 30 May to 7 September; B4 and 12 were not 
recovered). After September 2003, we had reduced 
acoustic coverage of the study area because we were 
unable to retrieve 12 receivers in 2004; most of these 
receivers were located in areas that were infrequently 
occupied by Black Sea Bass (Fabrizio et al., 2013). Be- 
tween 30 May and 14 December 2003, when the last de- 
tection was recorded at the site, we obtained 1,252,573 
detections. Some of these detections were removed from 
consideration because they were redundant, occurring 
at the same time on adjacent receivers. Single detec- 
tions, which may have resulted from acoustic or envi- 
ronmental interference, and occurrences of less than 
5 detections during a 24-h period were also removed, 6 
as were detections from unknown transmitters. Home- 
range and seasonal movement analyses were based on 
the resulting set of 1,007,787 detections. 
Home range of Black Sea Bass during inshore residency 
Home-range size of individual Black Sea Bass varied 
greatly (13.7-736.4 ha, n=109 fish), and fish of un- 
known sex tended to exhibit the greatest variation in 
home-range size (Fig. 2). About 95% of Black Sea Bass 
of unknown sex used areas <488.8 ha (Fig. 2; n= 78). 
In contrast, 95% of males maintained home ranges 
<278.7 ha (n=31), and no males occupied home ranges 
>332.4 ha. On average, fish of unknown sex used 137.2 
ha (SE 17.19) home ranges, and males used 120.9 ha 
(SE 15.29) home ranges. These mean home ranges rep- 
resent about 2.6% and 2.2% of the total area of the 
HARS. Home ranges of individual fish overlapped. 
Home-range size of Black Sea Bass varied by sex 
and depended on duration of occupancy (significant 
interaction, F= 4.20, P=0.04). Using sex-specific mod- 
els, we explored the effects of duration of occupancy, 
fish length, and release date on home ranges of males 
6 Pincock, D. G., and F. W. Voegeli. 2002. A quick course in 
underwater telemetry systems, 31 p. VEMCO Ltd, Bedford, 
Nova Scotia, Canada. 
