Manderson et al.: Residence time and habitat duration for predators in a small mid-Atlantic estuary 
151 
Table 3 
Results from the final generalized additive mixed models of effects of body size, estuarine temperature, and freshwater dis- 
charge (FW) on the residence time of ultrasonieally tagged Striped Bass (Morone saxatilis), Bluefish (Pomatomus saltatrix; 
age-0 and age-l+), and Weakfish ( Cynoscion regalis ) released into the Navesink River, New Jersey, in 2006 and 2007 (see 
Figs. 4 and 5 for deviance plots). Individual fish was included as a random effect (i.e. , intercept) in all models. Temporal 
autocorrelation in detections was considered as a first-order, autoregressive process that occurred within each fish The 
independent variables included in initial models were year as a factor, as well as release day, body length, temperature , and 
freshwater discharge , all of which were first considered with cubic smoothing splines (s) with a maximum of 4 degrees of 
freedom. Tensor product smooths (t2) were used to model interactions. Variables were included as linear effects if expected 
degrees of freedom (EDF) of splines were close to 1, and they were eliminated from models when they did not contribute 
to a reduction in Akaike’s information criterion (AIC). Body length (total length in millimeters) was considered as a class 
variable in modeling the temperature response of age-l+ Bluefish because the lengthx temperature interaction was strongly 
discontinuous. 
Species and Parametric coefficient 
Estimate 
SE 
Z-value 
P-value 
AIC 
Striped Bass 
Intercept 
-2.778 
0.177 
-15.73 
<0.0001 
4354 
Approximate significance of nonparametric terms 
EDF 
X 2 
t2 ( Temperature , body length) 
16.678 
228.65 
<0.0001 
4111 
s(log(FW Discharge + 1)) 
3.898 
201.90 
<0.0001 
3888 
s(Release day) 
Coefficient of multiple determination [f? 2 ]=0.15 
2.925 
29.66 
<0.0001 
3844 
Age-1+ Bluefish 
Intercept 
-1.739 
0.376 
-4.626 
<0.0001 
4301 
Year 
-1.055 
0.473 
-2.230 
0.0257 
4309 
Approximate significance of nonparametric terms 
EDF 
X 2 
s (Temperature)'.Length <500 mm 
3.927 
286.79 
<0.0001 
3700 
s (TemperatureY.Length >500 mm 
2.970 
193.74 
<0.0001 
s(log(FVV Discharge + 1)) 
3.893 
418.689 
<0.0001 
3122 
s (Release day) 
R 2 = 0.264 
1.958 
9.362 
0.0088 
3120 
Age-0 Bluefish 
Intercept 
0.861 
0.550 
1.566 
0.1170 
1974 
log(FVF Discharge + 1) 
0.922 
0.171 
5.385 
<0.0001 
1900 
Approximate significance of nonparametric terms 
EDF 
X 2 
s (Temperature) 
R 2 = 0.235 
3.345 
330.9 
<0.0001 
1128 
Weakfish 
Intercept 
-16.980 
4.714 
-3.602 
0.0179 
3731 
Year 
2.484 
1.049 
2.367 
0.0257 
3729 
Release day 
0.079 
0.022 
3.555 
0.0004 
3727 
Approximate significance of terms 
EDF 
X 2 
t2 (Temperature, Body length ) 
12.413 
580.5 
<0.0001 
2060 
s(log(FW Discharge + 1)) 
3.865 
131.0 
<0.0001 
1932 
R 2 =0.368 
charge fell below ~2 m 3 s" 1 than when discharge was 
higher (Table 3, Fig. 5). This effect was evident when 
we included year as a factor and when we modeled 
years separately. As a result, the response to low dis- 
charge did not appear to be related to interannual 
differences in sample size or freshwater discharge. 
Striped Bass were also likely to leave the tributary 
during episodes of high freshwater discharge ( >50 m 3 
s _1 ; Fig. 5A). Age-0 Bluefish and Weakfish were best 
modeled with linear discharge terms, indicating that 
the animals were not likely to leave the estuary dur- 
ing periods when freshwater discharge was high (Fig. 
5, B and C). 
There were significant differences in patterns of 
residency among individual fishes (random intercept; 
Table 3). Furthermore, the year effect was significant in 
GAMMs for Striped Bass, Weakfish, and age-l+ Bluefish, 
consistent with descriptions in the previous section, un- 
der Patterns of egress. Release date was significant in the 
models for Striped Bass, Weakfish, and age-l+ Bluefish. 
