148 
Fishery Bulletin 112(2-3) 
,a> 
10 -L, i i . > 
May Jun Jul Aug Sep Oct 
150 200 250 300 
Day of the year 
Figure 2 
For a study of residence times and habitat duration of Striped 
Bass (Morone saxatilis), Bluefish (Pomatomus saltatrix), and 
Weakfish ( Cynoscion regalis) in 2006 and 2007, mean daily 
(A) temperatures were measured at the RCM 9 sensor moor- 
ing in the Navesink River and (B) freshwater (FW) discharge 
rates (y-axis on log scale) were measured at a U.S. Geological 
Survey stream flow station in the Swimming River, For loca- 
tions of the mooring and flow station, see Figure 1. 
Because the presence-absence data were serially 
correlated in time for each individual fish, errors 
were modeled as a first-order autoregressive pro- 
cess nested within each individual fish. Release 
date and year also were considered as model 
covariates. 
Preliminary models were made with smoothing 
splines, and covariates were chosen through the 
use of manual backward selection, analysis of par- 
tial deviance, and Akaike’s information criterion 
(AIC) (Wood, 2006). To avoid over-fitting smooths, 
we set gamma to 1.4 and the basis dimension (k) 
to 5, limiting the maximum degrees of freedom of 
the smooths to 4. A covariate was removed from 
the model if its smoother was statistically insig- 
nificant, the change in AIC was >0 when the vari- 
able was removed, or 2 standard error confidence 
bands in the deviance plots included zero through- 
out variable domain. Covariates with equivalent 
degrees of freedom =1 were tested as linear ef- 
fects before they were eliminated on the basis of 
these criteria. We used tensor product smooths, 
which are appropriate when covariates are mea- 
sured on different scales, to test 2-way interac- 
tions between body sizes and other significant 
covariates (Wood, 2006). Because the response 
of age-l-i- Bluefish to temperature was strongly 
discontinuous across body sizes (i.e., lengths) at 
-500 mm TL, we pooled individuals into 2 body- 
size classes (300-500 mm TL, >500 mm TL) and 
treated body size as a factor covariate. 
We used log-rank tests for differences in “residency” 
curves between the species-and-age classes and years 
(Harrington and Fleming, 1982). 
We examined relationships between the presence of 
individuals in each age class of each tagged species in 
the estuary and environmental variation with logis- 
tic generalized additive mixed models (GAMM) in the 
gamm4 library in R (Aarts et ah, 2008; Wood, 2012). 
We limited final analyses to body size and the environ- 
mental variables of water temperature and freshwater 
discharge, which are important drivers of the estua- 
rine habitat suitability. Other measured environmen- 
tal variables were correlated with temperature and 
freshwater discharge and had lower explanatory power 
in preliminary models. In addition, the time series for 
salinity and oxygen in the estuary were incomplete. Fi- 
nally, complex preliminary GAMMs with more than a 
few variables also failed to converge. 
We analyzed water temperatures measured at the 
RCM 9 mooring and daily freshwater discharge (cubic 
meters per second) measured in the Swimming River 
because they were the most complete and accurate 
time series. We log transformed freshwater discharge 
values, which were strongly leptokurtic. Individual 
fish was considered as the random effect in all models. 
Results 
Patterns of temperature and freshwater discharge 
Spring and summer of 2006 were hotter and drier than 
those seasons in 2007 (Fig. 2, A and B). In 2006, spring 
warming rates were slightly higher and, in late July- 
early August, temperatures exceeded 30.0°C (2006 
max=30.2°C; 2007 max=28.0°C). During the autumn, 
however, temperatures were cooler in 2006 than in 
2007. Discharge in the Swimming River was high dur- 
ing the spring of both years. Freshwater discharge was 
low (<2 m 3 s -1 ) and discharge events were relatively 
rare from mid-July through August 2006. In 2007, pe- 
riods of low discharge occurred briefly (2-3 d) once in 
July and twice in August. Discharge was low through- 
out much of the fall of 2007, in contrast to several epi- 
sodes of high river discharge that were produced by 
frequent rains during the autumn of 2006. 
Patterns of release 
The species-and-size classes were available for collec- 
tion and release during different periods of time (Table 
1). Striped Bass were released in May and June. We 
released age-l+ Bluefish from May to July, but large 
