Schmitt et al.: Modeling the predation dynamics of invasive Ictalurus furcatus in Chesapeake Bay 
279 
Figure t 
Map of eastern Virginia showing the 542 locations (black circles) in 4 major 
tidal rivers where blue catfish (Ictalurus furcatus) were collected from April 
through October in 2013-2016 for analysis of the contents of their stomachs 
(/z=14,488). From north to south, rivers sampled were the Rappahannock, 
Mattaponi, Pamunkey, and James Rivers. The Mattaponi and Pamunkey 
Rivers converge to form the York River. 
water temperature, salinity, and coordinates were recorded 
for each sampling event. Fish weight (in grams) and TL (in 
millimeters) were also recorded, and stomach contents 
were placed on ice and later frozen. 
Laboratory methods 
Prior to examination, stomachs were thawed, and stom¬ 
ach contents were blotted dry with a paper towel (Schmitt 
et al., 2017). Prey items were then weighed, counted, 
and identified to the lowest possible taxon. Digested fish 
remains that lacked morphological distinctiveness were 
identified by using DNA barcoding techniques. The use of 
DNA barcoding enabled us to identify 70-80% of fish prey 
that were unidentifiable by gross morphology, excluding 
instances in which only bones or scales remained. Our 
DNA barcoding methods are described 
in Moran et al. (2016) and Schmitt et al. 
(2017, 2019). 
Modeling diet drivers for blue catfish 
Populations of blue catfish extend from 
tidal fresh water into mesohaline waters 
in Chesapeake Bay, where species 
assemblages change along the salinity 
gradient (Wagner and Austin, 1999; Jung 
and Houde, 2003). Seasonality affects 
the availability of some prey resources, 
such as adults of Alosa species, which 
enter tidal rivers during spring to spawn 
(Waldman, 2013), or blue crab, which 
migrate seasonally (Aguilar et al., 2005). 
Moreover, blue catfish exhibit ontoge¬ 
netic trophic niche shifts, with differ¬ 
ently sized fish consuming different 
prey (Schmitt et al., 2017). We therefore 
hypothesized that the diet of blue catfish 
would vary with season, salinity, and 
catfish size. 
We explored overall patterns in the 
diet of blue catfish by using canonical cor¬ 
respondence analysis (CCA; ter Braak, 
1986). This analysis is a form of multi¬ 
variate ordination in which a matrix of 
response variables is “regressed” (con¬ 
strained) on a matrix of independent 
variables; it is the multivariate analog 
of multiple linear regression. It is often 
used for analyzing relationships between 
species assemblages and multidimen¬ 
sional environmental data (ter Braak 
and Verdonschot, 1995), but CCA has 
also been used for assessing feeding pat¬ 
terns (Clifton and Motta, 1998; Jaworski 
and Ragnarsson, 2006). Because we were 
interested in general diet patterns, we 
first grouped all diet items into 6 broader 
categories: fish species, mollusks, crus¬ 
taceans, other invertebrates, vegetation, and other (e.g., 
anthropogenic debris, terrestrial mammals, birds, and 
other rare items). Each CCA was based on the binary 
presence-absence of diet items (i.e., frequency of occur¬ 
rence) because it is less biased than other diet measures 
and is preferred for assessing feeding patterns (Baker et al., 
2014; Buckland et al., 2017). Predictor variables included 
salinity zone, TL (rounded to the nearest 100 mm), and 
season (also coded as 3 dummy variables). 
We assessed whole-model and variable-wise statistical 
significance with F-tests, and significance was assessed 
by using an alpha threshold of 0.05. Magnitude of rela¬ 
tionship groupings of individual fish and constraining 
variables or diet items were assessed on the basis of load¬ 
ing scores (an analog of correlation coefficients, centered 
at 0 and ranging from -1 to 1). Because previous studies 
