2. the location of historically productive yellow eel fisheries. As glass eels
and elvers are novel life stages in the Lake Ontario watershed, it was
not known what habitat(s) they would occupy prior to the onset of
this study. It is also uncertain whether the formerly productive habitat
for American eel in Lake Ontario is still productive, given the extensive
physical and biotic changes wrought by repeated invasions of aquatic
invasive species over the past few decades (e.g., Johannsson et al.,
2011; Mills et al., 2003; Stewart et al., 2010).
The yellow-stage American eel is benthic oriented and is character-
ized as a habitat generalist as they can be found at a wide range of
depths, temperatures and salinities, and over a variety of substrates
(Greene et al., 2009; Pratt et al., 2014). This characterization may hide
the fact that, as eels grow, they undergo ontogenetic shifts, and studies
have identified differences in the habitat use (depth, velocity and sub-
strate composition) of small and large eels (Machut et al., 2007; Meffe
and Sheldon, 1988). This is consistent with the fact that different sized
eels consume different prey types. Smaller American eels are limited
by gape to the types and sizes of prey items that they can consume
with smaller eels feeding primarily on smaller invertebrates, whereas
larger eels feed mainly on fish or large crustaceans (Stacey, 2013;
Wenner and Musick, 1975). In addition, eels are well adapted to live
in interstitial spaces within the substratum, and they can burrow in
mud substrates (Koehn et al., 1994; Tesch, 2003; Tomie et al., 2013).
It might be expected that shifts in substrate preference would be neces-
sary as eels grow simply because their physical requirements change as
they get larger. Seasonal variation in habitat associations of American
eel may occur as a result of seasonal changes in the abundance and dis-
tribution of food and macrophyte cover, as seasonal variation in feeding
has been observed for the European eel (Anguilla anguilla; Bouchereau
et al., 2009). Thus, the few studies of habitat associations of American
eel at specific sizes may have led to the impression that eels are habitat
generalists when in reality they may have specific habitat requirements
at various life stages.
The purpose of our study was to assess nearshore habitat selection
by American eel stocked in Lake Ontario and the upper St. Lawrence
River. Specific objectives were to determine whether habitat selection
was related to eel size, and whether selection varied seasonally. We hy-
pothesized that habitat preferences would shift both seasonally and
with increasing body size, given the observations of seasonal and onto-
genetic shifts in other eel studies (Bouchereau et al., 2009; Machut et al.,
2007; Meffe and Sheldon, 1988; Wenner and Musick, 1975). In addition,
we were interested in assessing whether both stocking locations
contained suitable habitat to support eels despite the trophic changes
that have devastated other fishes in Lake Ontario (Mills et al., 2003).
As eels are flexible in their prey selection and most of the trophic chang-
es have resulted in the increased benthification of the Lake Ontario eco-
system, which is where eels generally feed and reside, our expectation
was that suitable habitat would still be available for eels.
Methods
Study area
This study was conducted in the Bay of Quinte located in the eastern
part of Lake Ontario, and the upper St. Lawrence River (Fig. 1). Lake
Ontario is a mesotrophic lake; and, because of seasonal temperature
changes, the lake stratifies and supports populations of warm and
coldwater fishes, both of which are represented in the Bay of Quinte
and the upper St. Lawrence River. The Z-shaped Bay of Quinte is
64 km in length and has a surface area of 254 km2
. The upper portion
of the bay (Big Bay, Telegraph Narrows) is relatively shallow with a
mean and maximum depth of 3.2 m and 8 m, respectively, whereas
the middle portion of the bay (Long Reach, Hay Bay) is deeper with
a mean depth of 6.3 m and maximum depth of 17 m (Hurley and
Christie, 1977). The substrate changes from primarily gravel bars and
sandy bays in nearshore areas to bedrock and gravel further offshore,
and ultimately to mud in deeper sections (Dermott, 2001; Hurley and
Christie, 1977). Macrophytes are present along the shoreline of the en-
tire bay (Crowder and Bristow, 1986). The upper St. Lawrence River ex-
tends 180 km from the outlet of Lake Ontario to the Robert Moses-
Robert H. Saunders Power Dam, at Cornwall Ontario. Almost all of the
river's water is supplied by Lake Ontario. Water levels are controlled
by the hydroelectric facility. The river provides a wide range of habitat
types and a diverse plant and fish community. However, the river
does not thermally stratify; and, in the summer, the water is too
warm to support coldwater fishes (LaPan et al., 2002). Stocking and
sampling locations can be seen in Fig. 1, while stocking numbers,
dates and eel lengths and weights at the time of stocking are reported
in Table 1.
Eel sampling
Habitat associations were studied in conjunction with a stocking ex-
periment and monitoring program in the spring and fall of 2010 and
2011 (Table 2). Boat electrofishing was used in predetermined 100 m
transects that ran parallel to shore at depths of 1.5 m or shallower. Tran-
sects were selected to represent the substrate types present within the
study locations, and where habitat did not vary much along a given
transect. The same transects were sampled each year and season unless
macrophyte growth or water depth inhibited boat access, in which case
a new transect was sampled nearby or the transect was not sampled for
that season. In general, more transects were fishable during the fall
when water levels had stabilized (Table 2). The vessel used for this re-
search was a 4.3 m Smith-Root SR-14h boat equipped with a Model
5.0 GPP generator and operated with boom anodes and hull cathode ar-
rays. The generator was set to 2.5 A of DC current. Assessments were
conducted during calm nights to maximize eel detection probabilities.
Each transect took about 5 min to complete. Netters used long handled
nets (6.4 mm mesh) to capture observed eels. Captured eels were
placed in a 100 L live well, and were processed at the end of each tran-
sect. Netters also enumerated eels that they saw but were unable to cap-
ture. We are confident that virtually all eels observed in this study were
stocked because stocked eels are distinguishable from naturally occur-
ring eels that migrate up the St. Lawrence River through the Moses
Saunders dam eel ladder by oxytetracycline hydrochloride (OTC-HCL)
markings (Pratt and Threader, 2011). In both sampling years, all cap-
tured eels that were sacrificed for origin assessment (n = 335) were
identified as stocked, so we believe that it is likely that the vast majority
of eels that were only enumerated or were captured and released were
also stocked.
Eel habitat assessment
Water chemistry, substrate type and macrophyte density were
assessed when each transect was completed. Measured parameters
included conductivity, water depth, water/air temperature and
dissolved oxygen concentration, which were taken from the mid-
point of the 100 m transect. Parameters were measured with an
ECTest waterproof conductivity meter and a YSI 550A temperature/
DO meter. Mean transect depths were recorded from the on-board
GPS-linked echosounder.
Substrate type was visually classified as a percentage, using
eight categories based on particle diameter: bedrock, boulder
(300–600 mm), rubble (100–300 mm), cobble (75–100 mm), gravel
(5–75 mm), sand (1–5 mm), silt (b1 mm) and organic material. Sub-
strate was assessed at the beginning and the end of each transect. The
dominant substrate type for each transect was used to provide a simple
contrast of habitat availability and eel density, while the percentage of
substrate types were used in the modeling analysis. Macrophyte density
was concurrently assessed, with percent cover of the transect classified
into one of four categories: none (0%); sparse (0–25%); moderate
(25–50%); and dense (50–100%).
882 M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
3. Data analysis
Eel density was calculated by assuming a 2.5 m effective fishing width
for each transect, and assuming that all eels were either captured or enu-
merated. A fully factorial ANOVA was used to test for differences in the
mean body length of captured eels between sampling locations (Bay of
Quinte, upper St. Lawrence River), seasons and years. The assumptions
of normality and homogeneity of variance for these data were assessed
using a Lilliefors test and Levenes test, respectively, and a Tukey HSD
test was used to conduct post-hoc comparisons among factors.
Species distribution and habitat data are often spatially autocorrelated,
which can create problems for statistical tests that assume independence
of error terms (Legendre and Legendre, 1998). The variance to mean ratio
(an index of spatial clustering; Krebs, 1989) was calculated for each
season. To assess the potential influence of spatial autocorrelation on
eel distribution, the association between eel presence (and counts)
and the geographic distance separating each sampling site was test-
ed using the Mantel test (999 permutations) (Fortin and Dale, 2005).
Distance matrices were constructed using Jaccard (presence) and
Euclidian (count) distance measures. Spatial analyses were complet-
ed using PASSaGE 2.0 (Rosenberg and Anderson, 2011).
Generalized additive models (GAM) were used to identify habitat
variables influential in predicting eel occurrence during spring and fall
sampling. Prior to modeling, a variable reduction procedure based on
principal component analysis (PCA) was used to improve the ratio of
observations to independent variables, and reduce multi-collinearity.
Principal components with eigenvalues ≥0.7 and that cumulatively ex-
plain ≥80% of the total variation were retained. For each component,
the variable with the highest loading was kept (Jolliffe, 1972; King and
Jackson, 1999). As variables were measured with different units, vari-
ables were first standardized by subtracting the mean and dividing by
the standard deviation (Legendre and Legendre, 1998). Visual observa-
tions where bedrock was the dominant substrate were rare (b5% of
transects) and not included in PCA.
Calculations were done in S-Plus (Mathsoft, 2002), and nonparamet-
ric functions were estimated using a spline smoothing function. The
best combination of independent variables was determined by evaluat-
ing the change in deviance resulting from dropping each variable from
the model in the presence of all other variables. The significance of
each variable on the probability of occurrence was tested with analysis
of deviance and likelihood ratios based on the binomial distribution
(Guisan et al., 2002). Relative importance of significant variables was
determined by calculating Akaike Information Criteria (AIC) (Burnham
and Anderson, 1998; Guisan et al., 2002).
Response curves were developed that describe the contribution of
the significant predictors of the probability of occurrence. Response
curves were based on partial residuals standardized to an average
value of 0. Influential variables exhibit a high range of values, and values
above and below the zero indicate, respectively, a positive or negative
association with the dependent variable (Granadeiro et al., 2004).
Fig. 1. Map of Lake Ontario and the upper St. Lawrence River, including specific American eel stocking locations (noted by ovals with white hatches) and sampling sites (noted by ovals
with dark gray shading).
Table 1
Stocking dates, location, number stocked and size of eel (when available) for American
eels stocked into the Lake Ontario watershed at two locations, the upper St. Lawrence Riv-
er (uSLR) and the Bay of Quinte (BQ), from 2006 through 2010. Means are presented with
±SE (n is given in parentheses); blank cells indicate that no data are available.
Stocking
date
Location Number
stocked
Mean total length
(mm)
Mean mass (g)
12 October 2006 uSLR 167,000 0.60 ± 0.004 (25)
5 June 2007 uSLR 294,000 59.2 ± 0.5 (49)
15 May 2008 uSLR 797,000 60.9 ± 0.6 (40) 0.17 ± 0.0006 (40)
29 May 2008 uSLR 518,000 60.4 ± 0.5 (40) 0.14 ± 0.0004 (40)
11 June 2008 BQ 686,000 56.5 ± 0.5 (40) 0.14 ± 0.006 (40)
2 June 2009 uSLR 650,000 59.1 ± 0.4 (246) 0.18 ± 0.04 (246)
2 June 2009 BQ 650,000 59.1 ± 0.4 (246) 0.18 ± 0.04 (246)
21 June 2010 uSLR 75,000
21 June 2010 BQ 68,000
Totals uSLR 2,501,000
BQ 1,404,000
883M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
4. Canonical correspondence analysis (CCA) was used to determine
whether size-specific eel habitat associations were present (ter Braak
and Šmilauer, 2002). CCAs were run separately for spring and fall sam-
pling seasons, with data pooled between years. Length-frequencies
were compared by separating eels into four total length (TL) categories
(b150–250, 251–350, 351–450, N450 mm). CCA assumes a nonlinear
Gaussian relationship. Monte Carlo permutation tests were used to de-
termine the statistical significance of the first axis and the sum of all ca-
nonical axes (ter Braak and Šmilauer, 2002).
Results
A total of 564 electrofishing transects were fished over the two year
study, with effort split relatively evenly between the two years
(Table 2). More transects were fished in the fall sampling period than
in the spring sampling period in both years, and more effort was consis-
tently put into sampling the Bay of Quinte (Table 2). A total of 1017 eels
were either captured or observed over the study (Table 2). Eel densities
were similar and varied similarly in both locations (Fig. 2). Both study
locations showed an increase in density from spring to fall 2010, a sim-
ilar density in the spring of 2011, and then a decrease in density in the
fall of 2011 to levels well below densities in the fall of 2010 (Fig. 2).
Mean eel length was significantly influenced by sampling location,
season and year (Table 3). Eels were significantly larger in the fall, re-
gardless of location (Fig. 3a). Eels captured from the upper St. Lawrence
River were significantly larger than those captured from the Bay of
Quinte in 2010 and 2011 (Fig. 3b). The length range of captured eels
was similar between stocking locations, ranging from 99 to 687 mm in
the Bay of Quinte and from 116 to 668 mm in the upper St. Lawrence
River. In the Bay of Quinte, the majority of eels captured were in the
151–200 mm and 201–250 mm size classes, whereas in the upper St.
Lawrence River, the majority of eels captured were in the 301–350
and N450 mm size classes (Figs. 4a, b).
Habitat preferences
Sampled sites in the Bay of Quinte were predominantly gravel and
cobble substrate, while sand and silt substrates comprised the majority
of sites sampled in the upper St. Lawrence River (Figs. 5a, b; Table 4).
The Bay of Quinte had largely rocky substrate (e.g., gravel, rubble, cobble
and boulder), but less silt substrate, than the upper St. Lawrence River.
Eels were found in all substrate types. Eel observations and captures
in the Bay of Quinte were highest in boulder, bedrock, cobble and
sand substrate, and lowest in organic sediment and gravel substrate
(Fig. 5a). In the upper St. Lawrence River, observations and captures
were highest in cobble, rubble and gravel substrate and organic sedi-
ment, and lowest in boulder, bedrock and sand substrate (Fig. 5b).
When separated by size class, smaller eels (b250 mm total length)
were found in a variety of substrate types in both sampling locations
(Table 4). Larger eels (N350 mm) were found in more complex sub-
strate (gravel, cobble and boulder) in the Bay of Quinte, while the larg-
est eels (N450 mm) were primarily detected in sand, silt and organic
substrate in the upper St. Lawrence River (Table 4).
Eel distribution was clumped during both seasons (variance to mean
ratio: 5.1 in spring; 5.2 in fall; Chi-square test p b 0.0001 in both cases).
However, eel distribution and abundance were not correlated to the dis-
tances separating sampling sites during either season (Mantel test:
p N 0.13).
The habitat variables used in the GAM with spring data were water
temperature and the percentages of silt, boulder, organic material, rub-
ble, cobble, and gravel in the substrate (Table 5). Five of the seven var-
iables were significant, and PC loadings associated with these
variables were generally greater than |0.6|. Likelihood of eel observation
and capture increased at sites with high percentages of organic (N60%)
and silt (N40%) substrates, but declined at gravel-dominated (N50%)
sites (Fig. 6). Eel observations and captures were positively associated
with water temperatures between 13 and 16 °C, and negatively associ-
ated with higher temperatures. The trend associated with the boulder
response curve was not interpretable.
Table 2
The number of electrofishing transects and the total number of eels captured and observed by season and year for sampling in the Bay of Quinte and the upper St. Lawrence River.
Year Season Sampling dates Location
Bay of Quinte Upper St. Lawrence River
# of transects # of eels # of transects # of eels
Captured Observed Total Captured Observed Total
2010 Spring 6–18 May 80 56 67 123 49 22 43 65
Fall 13–26 September 88 116 83 199 60 44 87 131
2011 Spring 4–13 May 79 68 113 181 60 60 81 141
Fall 12–20 September 87 50 47 97 61 43 47 80
Fig. 2. Eel density estimates from spring and fall sampling in 2010 and 2011 from the Bay
of Quinte and the upper St. Lawrence River. Error bars represent ± standard error.
Table 3
Results of a factorial analysis of variance on American eel length in Lake Ontario and the
upper St. Lawrence River for the spring and fall of 2010 and 2011 sampling periods.
Source df F p
Location 1 98.9 b0.001
Season 1 40.1 b0.001
Year 1 52.6 b0.001
Location × season 1 1.1 0.31
Location × year 1 1.7 0.21
Season × year 1 8.4 b0.001
Location × season × year 1 0.08 0.78
Residual 446
884 M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
5. Habitat variables used in the GAM run with fall data were percent
macrophyte cover and the percentage of boulder, silt, rubble and gravel
in the substrate. PC loadings associated with most of these variables
were greater than |0.55|; however the percentage of rubble was the
only habitat variable that significantly influenced eel occurrence
(Table 5). There was a greater likelihood of eel capture at sites with
more rubble (N50% coverage) (Fig. 6).
Body size and habitat occupation
CCA ordination for the first canonical axis from spring data
(eigenvalue = 0.39, F = 27.6, p = 0.002) and all canonical axes was sig-
nificant (F = 3.3, p = 0.002). In the spring, small eels (b150–250 mm)
were associated with coarse substrates (gravel, cobble, rubble and boul-
der) and little or no macrophytes. Medium-sized eels (251–350 mm)
were associated with silt substrates, warmer temperatures and dense
macrophyte cover. Large eels (351–450 mm and N451 mm) were asso-
ciated with deeper water, silt substrate, and moderate macrophyte
density (Fig. 7, top panel).
CCA ordination for the first canonical axis from fall data
(eigenvalue = 0.40, F = 34.8, p = 0.002) and all canonical axes
(F = 4.1, p = 0.002) was significant. Patterns in habitat association
were similar between spring and fall. Small eels were associated with
coarse substrates and little or no macrophytes (Fig. 7, bottom panel).
Medium-sized eels were associated with silt substrate, warm water
temperatures and dense macrophyte cover. Large eels were associated
with deeper water, silt substrates and moderate macrophyte density
(Fig. 7, bottom panel).
Discussion
Yellow-stage eels are characterized as habitat generalists, as they in-
habit a wide variety of temperatures, depths and salinities with little
consistent preference for habitat type, cover or substrate (Greene
et al., 2009; Helfman et al., 1987; Wiley et al., 2004). We expected that
habitat preferences would shift both seasonally and with increasing
body size, given the observations of ontogenetic shifts in other eel stud-
ies (e.g., Machut et al., 2007; Meffe and Sheldon, 1988; Wenner and
Musick, 1975). In our study, which focused on nearshore areas b1.5 m
deep, habitat shifts were observed with increasing body size, but we
found little support for seasonal habitat shifts in eels of the same size
class. Stocked American eels resided in a variety of habitats, but the im-
portance of coarse substrates (gravel, rubble, cobble) appeared to di-
minish, while the importance of finer substrates (sand, silt) increased
as eels grew larger. This is likely due to a combination of physical
space requirements, habitat availability and prey preference changing
with increasing body size, as eels need to balance their requirements
for a suitable refuge, finding prey and dealing with intraspecific interac-
tions. The availability of habitats differed between main stocking loca-
tions, and neither location had an ideal mix of coarse substrates (for
smaller eels) and fine substrates (for larger eels) at the depths sampled
in this study. This indicates the importance of dispersal if stocking is to
continue in the Great Lakes to ensure that eels can reach their preferred
habitat. Whether these patterns exist outside of the Great Lakes or out-
side of the depths surveyed in this study remain an open question.
The notion that yellow-stage American eels are habitat generalists is
easy to understand. Eels are found in both lotic and lentic waters, from
the high-water mark to at least 33 m in depth (Geer, 2003, from a mix of
Fig. 3. The mean length of captured American eel between: a) year and season and b) year
and location. Error bars represent ± 1 SE.
Fig. 4. Length distribution of eels captured in the Bay of Quinte and the upper St. Lawrence
River in a) 2010 and b) 2011. Data were pooled from spring and fall sampling.
885M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
6. freshwater and saline sites). Yellow eels are primarily benthic, and use a
variety of substrates (rock, sand, mud), and bottom debris such as
woody debris and submerged macrophytes for protection and cover
(Greene et al., 2009). They appear to demonstrate little consistent pref-
erence for habitat type, cover, substrate, or water temperature (Wiley
et al., 2004). Yet in reality, only limited research has been undertaken
into American eel habitat relationships, particularly for lakes, and
studies to date have usually found some habitat segregation by size.
For example, smaller eels in streams were associated with faster water
velocities and larger eels with slow, deeper habitats (Meffe and
Sheldon, 1988). Machut et al. (2007) also found smaller stream resident
eels associated with cobble and gravel substrate, and larger eels associ-
ated with larger cobble and boulder substrate. American eels can also
be found in soft sediments (Ford and Mercer, 1986), and in sediment
with aquatic macrophyte growth (LaBar and Facey, 1983). Our find-
ings were generally consistent with the results presented above,
with fewer, larger eels found in deeper water, and smaller eels grad-
ually progressing through increasingly larger substrate particle sizes
that had no associated macrophyte cover, until reaching a size at
which they shifted into sand or softer sediments that were associat-
ed with moderate density macrophyte cover (most often beds of the
colonial algae, Chara spp.). The exception to this was high densities
in bedrock substrates in the Bay of Quinte, but it is difficult to give
this finding much weight given that only 4 predominantly bedrock
substrate transects were fished in the bay over the course of the
study. Similar to our study, observations that small shortfinned eels
(Anguilla australis) were associated with gravel and mud substrates
and larger eels were associated with sand substrate were made in
two New Zealand lakes (Jellyman and Chisnall, 1999). Ontogenetic
shifts in habitat occur when fish outgrow resources, such as food,
Fig. 5. Number of transects (black bars), eels observed and captured (gray bars) and the
density of observed and captured eels (★) by substrate type from sampling both years
and seasons combined in the a) Bay of Quinte and b) upper St. Lawrence River.
Table 4
Number of eels captured by size class (mm) and substrate type, and the ratio of eels captured per transect sampled, from sampling both years and seasons combined in the Bay of Quinte
and upper St. Lawrence River sampling locations.
Location Substrate Eel size class (mm) # of transects
b150 150–249 250–349 350–449 N450
Count Ratio Count Ratio Count Ratio Count Ratio Count Ratio
Quinte Organic 0 0.00 0 0.00 0 0.00 0 0.00 0 0.00 2
Silt 0 0.00 0 0.00 0 0.00 0 0.00 0 0.00 1
Sand 1 0.11 5 0.56 3 0.33 0 0.00 0 0.00 9
Gravel 15 0.13 15 0.13 14 0.12 0 0.00 6 0.05 116
Cobble 8 0.06 125 0.92 41 0.30 3 0.02 4 0.03 136
Rubble 4 0.17 7 0.30 2 0.09 1 0.04 0 0.00 23
Boulder 1 0.04 12 0.46 13 0.50 5 0.19 0 0.00 26
Bedrock 0 0.00 3 0.75 0 0.00 0 0.00 0 0.00 4
Upper St. Lawrence River Organic 1 0.06 9 0.56 7 0.44 2 0.13 1 0.06 16
Silt 2 0.05 5 0.12 14 0.33 1 0.02 3 0.07 42
Sand 1 0.01 15 0.12 23 0.19 12 0.10 24 0.20 121
Gravel 0 0.00 0 0.00 1 0.50 1 0.50 0 0.00 2
Cobble 2 0.08 8 0.32 10 0.40 3 0.12 1 0.04 25
Rubble 0 0.00 5 0.63 3 0.38 1 0.13 0 0.00 8
Boulder 0 0.00 0 0.00 0 0.00 0 0.00 0 0.00 7
Bedrock 0 0.00 0 0.00 0 0.00 0 0.00 0 0.00 2
Table 5
Results of generalized additive models developed for American eel habitat associations.
Significant model variables are highlighted in bold. Significance levels are: *(p b 0.01)
and **(p b 0.001). “Deviance increase” is the result of dropping selected variable from
the model. Percent increase (in parentheses) was calculated as (deviance increase /
(null deviance − model deviance)). AIC calculated as (deviance of full model less one
covariate) + 2((df of null model) − (df of full model less one covariate)).
Parameter Spring Fall
Null deviance 370.1 408.2
df (null model) 266 294
Model deviance 279.2 371.4
df (full model) 238.8 275.5
Deviance
increase
AIC Deviance
increase
AIC
Gravel 21.9 (31.7)** 348.0 Rubble 17.2 (88.0)* 362.0
Temperature 19.6 (27.5)** 345.8 Boulder 8.4 (29.4) 352.8
Silt 17.6 (24.0)* 343.7 Silt 7.5 (25.5) 351.8
Organic 17.0 (23.0)* 343.4 Macrophyte 3.9 (11.8) 350.4
Boulder 9.7 (12.0)* 335.6 Gravel 3.11 (9.2) 347.6
Cobble 5.9 (6.9) 332.0
Rubble 4.9 (5.7) 331.0
886 M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
7. in their current location and move to new habitats where optimal
resources are available (Wootton, 1998). Larger eels become pisciv-
orous (Cairns, 1942) and likely move to new locations where larger
prey species are more abundant. The observed ontogenetic shifts re-
sulted in the stocked American eels utilizing a vast array of habitats,
fitting the general view of the species as a habitat generalist, but the
likely reality is that individuals use specific habitat features that
change with increasing length.
While our GAM analysis identified differences in the best explanato-
ry variables of eel habitat use between seasons, the CCA analysis identi-
fied similar patterns in habitat use of the same-sized eels between
spring and fall samples. Seasonal changes in habitat use are often related
to feeding behavior. The importance of water temperature in the spring
GAM is likely explained by the fact that eels undergo torpor at low water
temperatures (Walsh et al., 1983), and emerge in the spring needing to
feed. In the closely related European eel, Nyman (1972) documented an
optimal temperature range for promoting eel swimming and feeding
behavior of 13–17 °C, which fits the pattern in our spring temperature
relationship with an increase in eel density at 13 °C, followed by a
decline in eel density at temperatures over 16 °C. European eels in a
French lagoon were more active during the warm season than the
cold season, which was posited to be due to increased prey availability
as there was a reduction in prey availability and diversity in the colder
months (Bouchereau et al., 2009). Additionally, the importance of vari-
ous substrate classes in the spring model and rubble in the fall model
may reflect the limited availability of macrophytes for feeding or
avoiding predators. When feeding, eels use macrophytes as camouflage
when stalking prey and to minimize their own predation risk (Koehn
et al., 1994). Macrophytes also positively influenced habitat selection
of shortfinned eels (Chisnall, 1996; Jellyman and Chisnall, 1999). It is
possible that smaller eels utilize protective substrates for protection
and possibly feeding when macrophytes are limited, and this supposi-
tion is supported by our densities, which were generally highest in
more complex substrates, and lowest in bedrock and sand substrates
that provide little protection or cover. The vast majority of eels observed
in this study emerged from the substrate after stimulation by the elec-
trofishing boat; a few (typically larger) eels were seen free-swimming
on the bottom prior to contact with the electrical field.
Fig. 6. Generalized additive model fits of the estimated nonparametric function (spline) and 95% confidence intervals for the significant predictors of American eel site occupancy in the
spring and fall. These functions describe the effect of each habitat variable on the probability of eel presence. Significant predictors are boulder, gravel, organic, silt, and water temperature
in the spring, and rubble in the fall.
887M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
8. Biotic interactions, including competition, may structure size class
segregation in eel assemblages (Domingos et al., 2006; ICES, 2009).
Larger eels are known for their aggressive behavior and territoriality,
and can be cannibalistic (Lookabaugh and Angermeier, 1992). In labora-
tory experiments, Barila and Stauffer (1980) observed aggressive biting
behavior among American eels, attributed to territorial defence. Their
study suggests that aggressive behavior by larger, more dominant indi-
viduals may force smaller, subdominant eels to select less optimal hab-
itat types. This observation may be reflected in our CCA analysis, as
there were no apparent seasonal effects, and habitat patterns in eel
habitat use were more dependent on eel size than season. In both spring
and fall, smaller eels were found in shallower water in protective sub-
strates that were not associated with macrophytes, while larger eels
were found deeper in silt substrates in association with moderate to
dense macrophyte beds.
We recognize that our observations and conclusions on American
eel habitat use are limited because of the depth limitations imposed
by our sampling gear. Boat electrofishing was able to successfully enu-
merate eels b100 mm to N450 mm, but it was restricted to sampling
shallow depths. Jellyman and Chisnall (1999) found using bottom
trawls that higher densities of shortfinned eels less than 300 mm total
length were primarily found in depths of 0.6–1.2 m, while the majority
of larger eels were captured further offshore. Similarly, a size-associated
depth pattern has been documented in Australian eels (Anguilla
reinhardtii), as Neveu (1981) found larger eels in deeper sections of a
marsh. We are uncertain what habitat is available for eels, and what
the use of those habitats by eels may be, outside of the 1.5 m depth con-
tour surveyed in this study. In addition, the capture efficiency of eels by
electrofishing is relatively low, around 50% for small eels (b10 cm), im-
proving to around 90% for eels 400 mm in length (Graynoth et al., 2008).
This means that our reported density estimates are lower than actual
densities, and as our eels were growing over the course of the study,
that their capture efficiency likely improved over time. However, as
long as capture efficiency remains similar among habitat types, then
we believe that the habitat use patterns presented in this study should
not be affected by the capture technique used as capture efficiency
would change equally over time in all habitats.
While the presence of relatively high densities of eels in both stock-
ing locations is indicative that suitable habitat is available for early sur-
vival and growth, neither the Bay of Quinte or the upper St. Lawrence
River stocking locations had an ideal mix of substrates (increasingly
coarse for small-medium sized eels, and silt for larger eels). Our study
indicates that habitat use for American eel during freshwater residence
in lakes is largely influenced by substrate type. The Bay of Quinte near-
shore was ideal for smaller eels, principally composed of cobble
substrate. However, the softer substrates preferred by larger eels,
at least at the depths we could sample, were rare. Conversely, sand
and silt substrates dominated the near-shore in the upper St. Law-
rence River, but coarser substrates were limited in this part of the
watershed. Eels of all sizes utilize nearshore habitats, and heteroge-
neous nearshore habitat in the vicinity of stocking areas needs to
be protected and available to accommodate the range of sizes and
ages of eels.
The ontogenetic shifts demonstrated by American eel in our study
and the apparent habitat limitations in our two main stocking areas
point to the importance of not restricting movements in this species
and ensuring that dispersal opportunities exist. American eel stocked
as part of this study has dispersed around the Lake Ontario basin
(Pratt and Threader, 2011), and it is likely that these movements were
related to high densities and limited habitat availability in the stocking
areas. Interestingly, radio-tagged American eels in Lake Champlain, an-
other large lake in the St. Lawrence River watershed, ranged more wide-
ly (on the order of kilometers) than eels in streams and rivers (on the
order of meters and hectares) (Dutil et al., 1988; Helfman et al., 1983;
LaBar and Facey, 1983; Morrison et al., 2003), potentially indicative
that eels in lakes need larger home ranges than those in lotic waters.
Providing adequate dispersal opportunities is also likely important for
maintaining the historic 100% female sex ratio in the watershed
(Casselman, 2003), as there is substantial support in the literature
that sex determination in temperate eels, including the American eel,
is mediated by habitat and density (reviewed by Davey and Jellyman,
2005).
In summary, the majority of the literature has identified North
Atlantic eels as a generalist species; highly adaptable to various environ-
ments and capable of adapting to habitats of different salinities, temper-
atures and geographic locations. It is easy to understand how eels are
classified as generalists because of their varied habitat use, but for
certain sized eels, certain habitats seem preferred, and those habitat
features change with increasing size.
Acknowledgments
Stan Yavno, Anna Rooke, and Jenilee Gobin provided advice on study
design and statistical analysis. A special thanks to Joshua Stacey for his
long days in the field and lab assistance. Fisheries and Oceans Canada
provided the funding and field crew (Lisa O'Connor, William Gardner
and Marla Thibodeau) for the project, and Ron Threader at Ontario
Power Generation was the primary advocate for the broader eel stock-
ing experiment. Finally, we thank the efforts of two anonymous re-
viewers who greatly improved this manuscript.
Fig. 7. CCA ordination for the Bay of Quinte and the upper St. Lawrence River American eel
length frequencies for fish captured in spring (top panel), and fall (bottom panel) of 2010
and 2011. Length classes (mm) are represented by triangles. The lengths of the arrows
represent the relative importance of the associated environmental variable.
888 M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889
9. References
Barila, T.Y., Stauffer, J.R., 1980. Temperature behaviour responses of the American Eel,
Anguilla rostrata (Lesueur) from Maryland. Hydrobiologia 74, 49–51.
Bouchereau, J.L., Marques, C., Pereira, P., Guelorget, O.M., Vergne, Y., 2009. Feeding behav-
iour of Anguilla anguilla and trophic resources in the Ingril Lagoon (Mediterranean,
France). Cah. Biol. Mar. 50, 319–332.
Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference: A Practical Informa-
tion-theoretic Approach. Springer-Verlag, New York.
Cairns, D., 1942. The life history of the two species of New Zealand freshwater eel. N. Z.
J. Sci. Technol. 23, 545–567.
Casselman, J.M., 2003. Dynamics of resources of the American eel, Anguilla rostrata: de-
clining abundance in the 1990s. In: Aida, K., Tsukamoto, K., Yamauchi, K. (Eds.), Eel
Biology. Springer-Verlag, Tokyo, pp. 255–274.
Chisnall, B.L., 1996. Habitat associations of juvenile shortfinned eels (Anguilla australis) in
shallow Lake Waahi, New Zealand. N. Z. J. Mar. Freshw. Res. 30, 233–237.
Côté, C.L., Gagnaire, P.-A., Bourret, V., Verreault, G., Castonguay, M., Bernatchez, L., 2013.
Population genetics of the American eel (Anguilla rostrata): FST = 0 and North
Atlantic Oscillation effects on demographic fluctuations of a panmictic species. Mol.
Ecol. 22, 1763–1776.
Crowder, A., Bristow, M., 1986. Aquatic macrophytes in the Bay of Quinte. Can. Spec. Publ.
Fish. Aquat. Sci. 86, 114–127.
Davey, A.J.H., Jellyman, D.J., 2005. Sex determination in freshwater eels and management
options for manipulation of sex. Rev. Fish Biol. Fish. 15, 37–52.
Dermott, R., 2001. Annual Report of the Lake Ontario Management Unit. Ontario Ministry
of Natural Resources, Picton, Ontario, Canada.
Domingos, I., Costa, J.L., Costa, M.J., 2006. Factors determining length distribution and
abundance of the European eel, Anguilla anguilla, in the River Mondego (Portugal).
Freshw. Biol. 51, 2265–2281.
Dutil, J.-D., Giroux, A., Kemp, A., Lavoie, G., Dallaire, J.P., 1988. Tidal influence on move-
ments and on daily cycle of activity of American eels. Trans. Am. Fish. Soc. 117,
488–494.
Ford, T.E., Mercer, E., 1986. Density, size distribution and home range of American eels,
Anguilla rostrata, in a Massachusetts salt marsh. Environ. Biol. Fish. 17, 309–314.
Fortin, M.-J., Dale, M., 2005. Spatial analysis. A Guide for Ecologists. Cambridge University
Press, Cambridge.
Geer, P.J., 2003. Distribution, relative abundance, and habitat use of American eel Anguilla
rostrata in the Virginia portion of the Chesapeake Bay. Am. Fish. Soc. Symp. 35,
101–115.
Granadeiro, J.P., Andrade, J., Palmeirim, J.M., 2004. Modelling the distribution of shore-
birds in estuarine areas using generalized additive models. J. Sea Res. 52, 227–240.
Graynoth, E., Francis, R.I.C.C., Jellyman, D.J., 2008. Factors influencing juvenile eel (Anguilla
spp.) survival in lowland New Zealand streams. N. Z. J. Mar. Freshw. Res. 42, 153–172.
Greene, K.E., Zimmerman, J.L., Laney, R.W., Thomas-Blate, J.C., 2009. Atlantic coast diadro-
mous fish habitat: a review of utilization, threats, recommendations for conservation,
and research needs. Atlantic States Marine Fisheries Commission Habitat Manage-
ment Series No. 9 (Washington, D.C.).
Guisan, A., Edwards Jr., T.C., Hastie, T., 2002. Generalized linear and generalized additive
models in studies of species distributions: setting the scene. Ecol. Model. 157,
89–100.
Helfman, G.S., Stoneburner, D.L., Bozeman, E.L., Christian, P.A., Whalen, R., 1983. Ultrasonic te-
lemetry of American eel movements in a tidal creek. Trans. Am. Fish. Soc. 112, 105–110.
Helfman, G.S., Facey, D.E., Hales Jr., L.S., Bozeman Jr., E.L., 1987. Reproductive ecology of
the American eel. Am. Fish. Soc. Symp. 1, 42–56.
Hurley, D.A., Christie, W.J., 1977. Depreciation of the warmwater fish community in the
Bay of Quinte, Lake Ontario. J. Fish. Res. Board Can. 34, 1849–1960.
ICES, 2009. Report of the Study Group on Anguillid Eels in Saline Waters (SGAESAW),
16–18 March 2009, Sackville, Canada; 3–5 September 2009, Gothenburg, Sweden.
ICES CM/DFC, p. 06.
Jellyman, D.J., Chisnall, B.L., 1999. Habitat preferences of shortfinned eels (Anguilla
australis), in two New Zealand lowland lakes. N. Z. J. Freshw. Res. 33, 233–248.
Jessop, B.M., 2010. Geographic effects on American eel (Anguilla rostrata) life history char-
acteristics and strategies. Am. Fish. Soc. Symp. 67, 326–346.
Johannsson, O.E., Bowen, K.L., Holeck, K.L., Walsh, M.G., 2011. Mysis diluviana population and
cohort dynamics in Lake Ontario before and after the establishment of Dreissena spp.,
Cercopagis pengoi, and Bythotrephes longimanus. Can. J. Fish. Aquat. Sci. 68, 795–811.
Jolliffe, I.T., 1972. Discarding variables in a principal components analysis. I: Artificial data.
J. R. Stat. Soc. Ser. C 21, 160–173.
King, J.R., Jackson, D.A., 1999. Variable selection in large environmental datasets using
principal components analysis. Environmetrics 10, 67–77.
Koehn, J.D., O'Connor, N.A., Jackson, P.D., 1994. Seasonal and size-related variation in mi-
crohabitat use by a southern Victorian stream fish assemblage. Aust. J. Mar. Freshwat.
Res. 45, 1353–1366.
Krebs, C.J., 1989. Ecological Methodology. Harper and Row Publishers, New York.
LaBar, G.W., Facey, D.E., 1983. Local movements and inshore population sizes of American
eels in Lake Champlain, Vermont. Trans. Am. Fish. Soc. 112, 111–116.
LaPan, S.R., Mathers, A., Stewart, T.J., Lange, R.E., Orsatti, S.D., 2002. Fish-community objec-
tives for the St. Lawrence River. Great Lakes Fish. Comm. Spec. Pub. 2002 (27 pp.).
Legendre, P., Legendre, L., 1998. Numerical Ecology. Elsevier Science BV, Amsterdam.
Lookabaugh, P.S., Angermeier, P.L., 1992. Diet patterns of American eel, Anguilla rostrata,
in the James River drainage, Virginia. J. Freshw. Ecol. 7, 425–434.
MacGregor, R.B., Mathers, A., Thompson, P., Casselman, J.M., Dettmers, J.M., LaPan, S.,
Pratt, T.C., Allen, W.A., 2008. Declines of American Eel in North America: complexities
associated with bi-national management. In: Schechter, M.G., Taylor, W.W., Leonard,
N.J. (Eds.), International Governance of Fisheries Ecosystems: Learning From the Past,
Finding Solutions for the Future. American Fisheries Society, Bethesda, MD,
pp. 357–381.
MacGregor, R., Casselman, J.M., Allen, W.A., Haxton, T., Dettmers, J.M., Mathers, A., LaPan,
S., Pratt, T.C., Thompson, P., Standfield, M., Marcogliese, L., Dutil, J.-D., 2009. Natural
heritage, anthropogenic impacts, and biological issues related to the status and sus-
tainable management of American eel: a retrospective analysis and management
perspective at the population level. Am. Fish. Soc. Symp. 69, 713–740.
Machut, L.S., Limburg, K.E., Schmidt, R.E., Dittman, D., 2007. Anthropogenic impacts on
American Eel demographics in Hudson River tributaries, New York. Trans. Am. Fish.
Soc. 136, 1699–1713.
Mathsoft, 2002. S-Plus 6 for Windows User's Guide. Insightful Corporation, Seattle.
Meffe, G.K., Sheldon, A.L., 1988. The influence of habitat structure on fish assemblage
composition in southeastern blackwater streams. Am. Midl. Nat. 120, 225–240.
Mills, E.L., Casselman, J.M., Dermott, R., Fitzsimons, J.D., Gal, G., Holeck, K.T., Hoyle, J.A.,
Johannsson, O.E., Lantry, B.F., Makarewicz, J.C., Millard, E.S., Munawar, I.F.,
Munawar, M., O'Gorman, R., Owens, R.W., Rudstam, L.G., Schaner, T., Stewart, T.J.,
2003. Lake Ontario: food web dynamics in a changing ecosystem (1970–2000).
Can. J. Fish. Aquat. Sci. 60, 471–490.
Morrison, W.E., Secor, D.H., Piccoli, P.M., 2003. Estuarine habitat use by Hudson River
American eels as determined by otolith strontium: calcium ratios. Am. Fish. Soc.
Symp. 33, 87–99.
Neveu, A., 1981. Seasonal variations in daily food intake of eels under natural conditions.
Acta Oecol. 2, 99–116.
Nyman, L., 1972. Some effects of temperature on eel (Anguilla) behaviour. Inst. Freshw.
Res. 52, 90–102 (Drottningholm, Sweden).
Pratt, T.C., Mathers, A., 2011. 2010 update on the status of American Eel (Anguilla rostrata)
in Ontario. DFO Can. Sci. Advis. Sec. Res. Doc. 2011/050 (vi + 18 pp.).
Pratt, T.C., Threader, R.W., 2011. Preliminary evaluation of a large-scale American eel con-
servation stocking experiment. N. Am. J. Fish Manag. 31, 619–628.
Pratt, T.C., Bradford, R.G., Cairns, D.K., Castonguay, M., Chaput, G., Clarke, K.D., Mathers, A.,
2014. Recovery potential assessment for the American Eel (Anguilla rostrata) in
eastern Canada: functional description of habitat. DFO Can. Sci. Advis. Sec. Res.
Doc. 2013/132 (v + 49 pp.).
Rosenberg, M.S., Anderson, C.D., 2011. PASSaGE: pattern analysis, spatial statistics and
geographic exegesis. Version 2. Methods Ecol. Evol. 2, 229–232.
Stacey, J.A., 2013. The Life History Strategy, Growth, Body Condition, and Diet of Stocked
American Eel (Anguilla rostrata) in the Upper St. Lawrence River and Lake Ontario
(Thesis (M.Sc.)), Trent University, Peterborough, ON (132 pp.).
Stewart, T.J., Johannsson, O.E., Holeck, K., Sprules, W.G., O'Gorman, R., 2010. The Lake
Ontario zooplankton community before (1987–1991) and after (2001–2005)
invasion-induced ecosystem change. J. Great Lakes Res. 36, 596–605.
ter Braak, C.J.F., Šmilauer, P., 2002. CANOCO reference manual and CanoDraw for
windows user's guide: software for canonical community ordination (version 4.5).
Microcomputer Power.
Tesch, F.W., 2003. The Eel. 3rd edition. Blackwell Science, Oxford.
Tomie, J.P.N., Cairns, D.K., Courtenay, S.C., 2013. How American eels Anguilla rostrata con-
struct and respire in burrows. Aquat. Biol. 19, 287–296.
Walsh, P.J., Foster, G.D., Moon, T.W., 1983. The effects of temperature on metabolism of
the American eel Anguilla rostrata (LeSueur): compensation in the summer and tor-
por in the winter. Physiol. Zool. 56, 532–540.
Wenner, C.A., Musick, J.A., 1975. Food habits and seasonal abundance of the American eel,
Anguilla rostrata, from the lower Chesapeake Bay. Chesap. Sci. 16, 62–66.
Wiley, D.J., Morgan, R.P., Hilderbrand, R.H., Raesly, R.L., Shumway, D.L., 2004. Relations be-
tween physical habitat and American Eel abundance in five river basins in Maryland.
Trans. Am. Fish. Soc. 133, 515–526.
Wootton, R.J., 1998. Ecology of Teleost Fishes. second ed. Kluwer Academic, Aberystwyth,
United Kingdom.
Zhu, X., Zhoa, Y., Mathers, A., Corkum, L., 2013. Length frequency age estimations of
American eel recruiting to the upper St. Lawrence River and Lake Ontario. Trans.
Am. Fish. Soc. 142, 333–344.
889M.H.M. Lloyst et al. / Journal of Great Lakes Research 41 (2015) 881–889