Schmitt et al.: Modeling the predation dynamics of invasive Ictalurus furcatus in Chesapeake Bay 
281 
Table 1 
Results of the canonical correspondence analyses used to identify key drivers of the diet of blue catfish (Ictalurus furcatus) collected 
in 4 tributaries to Chesapeake Bay in Virginia during 2013-2016. Whole-model and variable-wise statistical significance (P<0.05) 
were evaluated with F-tests. Predictor variables include salinity zone, season, and predator total length. Separate models were 
developed for each river. 
James River Rappahannock River Pamunkey River Mattaponi River 
Variable 
df 
F 
P 
df 
F 
P 
df 
F 
P 
df 
F 
P 
Whole model 
4 
17.9 
<0.001 
4 
21.6 
<0.001 
4 
21.7 
<0.001 
4 
14.9 
<0.001 
Salinity zone 
1 
32.7 
<0.001 
1 
11.8 
<0.001 
1 
46.7 
<0.001 
1 
1.9 
0.081 
Season 
3 
21.6 
<0.001 
2 
49.2 
<0.001 
2 
13.3 
<0.001 
2 
43.1 
<0.001 
Total length 
1 
8.9 
<0.001 
1 
7.6 
<0.001 
1 
7.8 
<0.001 
1 
11.3 
<0.001 
James River 
Salinity 
(-0.62) 
Spring 
(0.72) 
CCA1 (61.8%) 
w 
T3 
D 
IQ 
On 
CO 
c 
3 
3 
CD 
-\ 
"T 
o 
O) 
Rappahannock River 
Spring Salinity 
(-0.66) (0.28) 
CD 
~o 
=3 
CQ 
•^J 
O 
C/) 
c 
3 
3 
CD 
'T 
o 
'<£) 
Pamunkey River 
Spring Salinity 
(-0.48) (0.82) 
sp 
CO 
2 
o 
o 
CCA1 (67.1%) 
# Predator 
• 
Invert. 
Fish 
*• 
0 
Vog. ^ ^ 
8 
m 
CCA1 (83.4%) 
Omnivore 
Mattaponi River 
Spring Salinity 
(-0.41) (0.15) 
co 
c 
3 
3 
CD 
-* 
"3 
CO 
o 
CO 
■O 
—s 
o’ 
CQ 
T 
o 
CD 
CD 
Figure 2 
Canonical correspondence analysis (CCA) plots used to identify key drivers of the diet of blue catfish (Ictalurus furca¬ 
tus) collected in 4 tributaries to Chesapeake Bay in Virginia during 2013-2016. Each point represents an individual 
fish and has been jittered to reduce overlap of individuals with the same combination of diet items. Gray points repre¬ 
sent individuals containing vegetation in their stomachs (i.e., omnivores), and black points represent predatory fish. 
The amount of variation in multivariate diet responses described by each axis is reported in each axis label. Loading 
scores of independent constraining variables (e.g., season and salinity zone) are presented on the outsides of the plots, 
instead of traditional arrowed vectors within plots. Constraining variable loading scores on a given axis should be 
interpreted as directional within plot halves. Only loading scores >0.4 are presented (for all axis loading scores, see 
Table 2). Text within plots indicates loading scores of diet items, such as other invertebrates (Invert.), crustacean 
(Crust.), and vegetation (Veg.). Although points have been jittered, they are positioned in the correct quadrants as 
close as possible to their original coordinates. 
but blue catfish shifted toward piscivory in higher salin¬ 
ity areas, especially in the James and Pamunkey Rivers 
(Fig. 2). Not surprisingly, herbivory was strongly associ¬ 
ated with spring and summer in all rivers. 
Perhaps most importantly, patterns were not consistent 
among rivers. For example, herbivory was strongly associ¬ 
ated with summer in all rivers except the Rappahannock 
River, where it was more prevalent in spring. Moreover, 
distinct length- and season-related breaks were observed 
in diets of individual blue catfish from some rivers (e.g., 
the Mattaponi River), but much more overlap occurred in 
other rivers (e.g., the James River; Fig. 2). All CCA axis 
loadings of diet items and constraining variables are pre¬ 
sented in Table 2. 
Predation models for species of concern 
Our GAMs demonstrate that predation by blue catfish 
on species of concern varies by river, salinity, month, and 
predator TL (Figs. 3-5). All GAMs were globally significant 
