Environmental biology of fishes, tập 92, số 2, 2011

Environ Biol Fish (2011) 92:141–150
DOI 10.1007/s10641-011-9823-1

Patterns of shield darter, Percina peltata, distribution
in the Eastern Piedmont of Maryland, USA
Patrick Ciccotto & Scott Stranko

Received: 3 February 2010 / Accepted: 4 April 2011 / Published online: 27 April 2011
# Springer Science+Business Media B.V. 2011

Abstract Detailed analyses of habitat associations
with rare species are typically constrained by limited
sample size and the availability of habitat data. The
dense spatial coverage of stream sampling by the
Maryland Biological Stream Survey provides ample
data to quantitatively examine correlations between
habitat and rare species distributions. The shield
darter, Percina peltata, has a widespread distribution
on the Atlantic Slope of the United States, but is
uncommon throughout its range in Maryland. Associations of in situ physical habitat, water chemistry,
and alterations in landscape with shield darter
presence in the Eastern Piedmont physiographic
province in Maryland were examined. Shield darter
occurrence was associated with larger sized streams in
concordance with the species’ known ecology. Shield
darter distribution was further associated with stream
segments with deep riffle habitats with diverse
velocities, low concentrations of chloride and sulfate,
low levels of urbanization in upstream catchments,
and several pollution intolerant fish species. Although

P. Ciccotto (*)
University of Maryland Baltimore County,
1000 Hilltop Circle,
Baltimore, MD 21250, USA
e-mail:
S. Stranko
Maryland Department of Natural Resources,
Monitoring and Non-Tidal Assessment Division,
580 Taylor Avenue,
Annapolis, MD 21401, USA

the exact mechanism of the effects is not clear, results
indicate that the shield darter is sensitive to urban
development and habitat and water quality alteration
that typically accompanies urbanization. Shield darter
conservation in Maryland necessitates the protection
and restoration of minimally urbanized watersheds
where they are known to occur. The results from this
study indicate that habitat information on rare species
may be important in elucidating important habitat
associations that are not evident via examination of
community level data.
Keywords Shield darter . Habitat quality .
Urbanization

Introduction
The shield darter, Percina peltata, is a small (up to
90 mm) freshwater species of the family Percidae.
This species typically inhabits the riffles of warm,
streams and rivers with substrates ranging from

detritus and silt to gravel, cobble, and boulders.
Spawning typically occurs over gravel substrates
(New 1966; Jenkins and Burkhead 1994; Schmidt
and Daniels 2004).The diet of shield darter consists
mainly of aquatic insects. Based on its ecology, it is
considered intolerant to anthropogenic disturbances
(Southerland et al. 2005). The species has a large
Atlantic Slope distribution, ranging from the Hudson
and Susquehanna Rivers in New York south to the


142

Environ Biol Fish (2011) 92:141–150

James River in Virginia in the Eastern Piedmont and
onto the upper Coastal Plain (Rohde et al. 1994; see
inset Fig. 1). In Maryland, the shield darter is
currently uncommon throughout the state, ranging
from lower Susquehanna River tributaries to Potomac
River watersheds in the Eastern Piedmont. Shield
darter populations are also known to occur in watersheds of the Choptank River basin on the eastern
shore of the state as well as in lower portions of
Antietam Creek and Catoctin Creek in the Ridge and
Valley physiographic province in western Maryland.
The shield darter is currently on the Watch List in
Maryland (MDNR 2007). Despite the importance of
this species to mid-Atlantic aquatic biodiversity
(particularly Maryland), research on the effects of
habitat and water quality alteration has not been

conducted.
The family Percidae is the second most diverse
family of freshwater fishes in North America. Despite
their diversity however, almost half of the 191
described percid species are considered imperiled
(Jelks et al. 2008). Threats such as pollution,
fragmentation, and habitat-loss as a result of human-

induced landscape alteration appear to be consistent
with those facing other aquatic taxa (Allan and
Flecker 1993; Burkhead et al. 1997). Seven of the
nine native darters in Maryland are listed on the
state’s list of rare, threatened, and endangered species
(MDNR 2007). Habitat degradation and expanding
urbanization from the Baltimore and Washington D.C.
metropolitan areas threaten several of these species,
including the shield darter, which in Maryland is
found within close proximity to these urban centers.
Urbanization from Washington D.C. and Baltimore
has caused major anthropogenic disturbances to many
of the streams within Maryland’s portion of the shield
darter distribution by altering physical habitat and
chemical composition of lotic systems. Changes in
fish communities have been widely documented to
result from these alterations in urbanized watersheds
with the loss of endemic species and shifts from
intolerant to tolerant and/or non-native species in
coincidence (Klein 1979; Weaver and Garman 1994;
Wang et al. 1997, 2000; Paul and Meyer 2001; Tabit
and Johnson 2002; Walters et al. 2003; Morgan and

Cushman 2005).

Fig. 1 Map of site locations where shield darter were present
(black circle) and absent (white circle) used in 422 sites
sampled in the Easter Piedmont from 2000–2009 in Maryland.
Gray lines represent the boundaries of the Eastern Piedmont,

while the dashed black line represents the division between
watersheds draining directly into the Chesapeake Bay and those
draining into the Potomac River


Environ Biol Fish (2011) 92:141–150

Rare and imperiled species have often been
excluded in examinations of broad landscape-scale
alterations on stream quality because of the paucity of
records that can muddle statistical results. However,
understanding the influence of human alterations to
streams on these rare species is of particular importance for the conservation of the species themselves
and for efforts to maintain regional biodiversity. The
Maryland Department of Natural Resources, Maryland Biological Stream Survey (MBSS), has a unique
opportunity by providing relatively extensive records
of some imperiled fish species concomitantly with
fairly comprehensive stream habitat, chemical, and
landscape data. The purpose of this paper is to
determine correlations of shield darter presence and
absences in the Eastern Piedmont physiographic
province with fish assemblage, landscape alterations,
water chemistry, and physical habitat data to investigate the potential influence of human-related alterations to stream conditions within the Baltimore and

Washington D.C. metropolitan areas on shield darter
distribution in Maryland.

Methods
We screened sites from the MBSS database based on
inherent physiographic and stream size related variables to target areas where shield darters are likely to
occur. The purpose of screening sites by physiographic province and stream size variables was to identify
potential influences on shield darter distribution that
may be addressed from a management standpoint.
Natural phenomena that cannot be practically
addressed in conservation practices (e.g., biogeography or upstream watershed areas) were not included.
Water chemistry, habitat quality, and landscape alterations are potential influences on shield darter
distribution that are addressable and were thus the
focus of our second group of analyses. After the
potentially confounding effects of physiographic
province (Utz et al. 2009) and stream size were
removed, water chemistry, fish habitat, and fish
assemblage data were used to investigate the potential
influences on shield darter distribution.
We limited our analyses to the Eastern Piedmont
because 1) it is the “center” of shield darter
distribution within the state (i.e., contains the largest
proportion of sites where shield darter have been

143

observed) and 2) this area is undergoing an increase
in urban development, which is hypothesized to be
detrimental to shield darter populations. Fishes were
sampled from 422 randomly-selected stream sites

75 m in length from 2000–2009 in the Eastern
Piedmont to provide for the most statistically rigorous
methodology of assessing the patterns in shield
distribution. Due to the random selection of stream
sites, several stream reaches or watersheds were
subject to multiple sampling events. We treat these
as independent samples however, based on the
generally limited dispersals observed in stream fishes,
particularly in some adult Percina species (Freeman
1995; Warren and Pardew 1998; Skyfield and Grossman 2008). Fish collections at these sites consisted of
double-pass electrofishing at low flow conditions
during June–September. To minimize fish movement
into and out of the sites, block nets were placed at the
upstream and downstream ends of the segments. Fish
species were identified in the field before being
released.
We focused on larger sites (i.e., streams with wider,
deeper habitats and higher stream flows) where shield
darters are typically observed to best explain patterns
of shield darter occurrence with anthropogenic effects
without the potentially confounding influence of
stream size. Empirical observations of shield darters
in larger streams in Maryland are consistent with
observations of this species throughout its range
(Jenkins and Burkhead 1994; Rohde et al. 1994). To
quantify stream size, five measurements were taken in
the 75 m site. Average thalweg depth (the deepest part
of each transect), average thalweg velocity (water
velocity at the deepest part of each transect), and
average wetted width were averages of four wetted

measurements taken at 25 m intervals on the 75 m
sites. Maximum depth was the deepest area found
anywhere in the 75 m site. Stream discharge followed
U.S. Geological Survey (USGS) methods (Rantz
1982). The area of the upstream site catchment was
also calculated to quantify stream size for each
individual site. Site catchments were hand digitized
based on USGS quarter quad topographic lines and
their areas were calculated in ArcMap 9.1(ESRI
2005). Sites were categorized as shield darter present
or absent and we conducted a principal component
analysis (PCA) on the stream-size variables. For
significant axes (where the eigenvalue was greater
than the broken-stick eigenvalue, see McCune and


144

Environ Biol Fish (2011) 92:141–150

Grace 2002) we calculated 95% confidence intervals
around the mean axis scores of sites where shield
darter was absent in ordination space. These confidence intervals were used to eliminate sites where
shield darter are expected to be absent by removing
all sites below the confidence interval limit. The
remaining larger-sized sites were then analyzed to
identify patterns of shield darter occurrences with fish
community, water chemistry, physical habitat, and
landscape variables.
Water chemistry data were based on laboratory

analyses of grab samples, and included measurements
of pH, conductivity, dissolved organic carbon, chloride, sulfate, total nitrogen, total phosphorous, orthophosphate, ammonia, nitrate, and nitrogen dioxide.
Dissolved oxygen concentrations were measured in
situ with a Hydrolab Quanta. Physical habitat variables were measured using MBSS protocols (Kazyak
2001). Guidelines for physical habitat scores are
reported in Table 1. Instream habitat, pool/glide/eddy
quality, and riffle/run quality were each rated on a 0–
20 scale using well-defined, quality controlled visual
assessments. Instream habitat scores were assigned
based on perceived value of habitat to supporting the
fish community. Sites with a variety of habitat types
and substrate particle sizes received higher instream
habitat scores. Pool/glide/eddy quality scores were
assigned based on the variety and complexity of slowor still-water habitat at a site. Higher scores were
assigned to segments with undercut banks, woody
debris, and other potential types of cover for fishes.
Riffle/run quality scores were assigned based on
riffle/run depth, complexity, and importance as habitat

structure for aquatic fauna. Segments with deeper
riffle/run areas, stable substrates, and a variety of
water current velocities received the highest scores.
Land cover (percent urban, percent agriculture, and
percent forest) was calculated for all hand digitized
site catchments from the USGS 2001 National Land
Cover Dataset (Homer et al. 2007), such that this
measurement represents all land area draining to each
site from upstream.
In addition to the above habitat variables, we
wanted to compare the fish assemblage compositions

of sites where shield darters were present or absent.
For each site, we compiled total species richness and
abundances from the electrofishing surveys. Fish
species that were found in less than 5% of these sites
were omitted from the analysis to limit the effects of
rare species on the ordination. An outlier analysis
using the Sorenson distance among the sites was also
used and identified one site with an average distance
3.8 standard deviations away from the grand mean
distance based on fish data. This site was removed
from further analysis. We removed rare fish and
outliers in order to reduce “noise” in the statistical
analyses (Reynoldson and Rosenberg 1996; Rodriguez and Lewis 1997). More specifically, we did not
want rare fishes or sites with outlying fish communities to skew the ordinations that would skew
otherwise important patterns useful to the management of our focal species, the shield darter. Water
chemistry and physical habitat data were log10(x+1)
transformed, while land cover data underwent an
arsine square root transformation. Sites were again
grouped as present or absent for shield darter to assess

Table 1 Physical habitat assessment guidelines used at sampling sites (see Kazyak 2001)
Habitat
Optimal
parameter 16–20
Instream
Habitat

Pool/
Glide/
Eddy

Quality

Sub-optimal
11–15

Greater than 50% of a variety of cobble,
30–50% of stable
boulder, submerged logs, undercut banks,
habitat; adequate
snags, rootwads, aquatic plants, or other
habitat
stable habitat
Complex cover and/or depth >1.5 m; both
Deep (>0.5 m) areas
deep (>0.5 m) and shallow (<0.2 m) present present, but only
moderate cover

Riffle/
Riffle/run depth generally >10 cm, with
Run
maximum depth greater than 50 cm;
Quality
substrate stable and variety of current
velocities

Riffle/run depth
generally 5–10 cm,
variety of current
velocities

Marginal
6–10

Poor
0–5

10–30% mix of stable
habitat; habitat
availability less than
desirable
Shallows (<0.2 m)
prevalent in pool/
glide/eddy habitat;
little cover

Less than 10% stable
habitat; obvious lack
of habitat

Riffle/run depth
generally 1–5 cm;
primarily a single
current velocity

Riffle/run depth <1 cm
or riffle/run substrates
concreted

Max depth <0.2 m in
pool/glide/eddy

habitat; or absent
completely


Environ Biol Fish (2011) 92:141–150

whether any of the explanatory variables were
associated with shield darter occurrences. To test for
significant differences in water chemistry, physical
habitat, and landscape variables between sites where
shield darter was present and absent, a Wilcoxon
rank-sum test was conducted for each variable in SAS
Enterprise Guide, using a Bonferroni correction (α=
0.05/18 abiotic variables = 0.003) (SAS 2006). Nonmetric multidimensional scaling (NMS) was then
used to examine patterns of shield darter occurrence
with fish assemblage and the 18 abiotic variables. The
NMS was run using a Sorenson distance with fish
presence/absence data, including shield darter, as the
primary matrix and abiotic variables as the secondary
matrix. Fish species were classified as stressor or
pollutant/disturbance intolerant, tolerant, or no tolerance type based on Southerland et al. (2005). We used
PC-ORD for the NMS (McCune and Mefford 2006).
Pearson’s correlation coefficients and associated P
values were calculated for each variable to the
ordination axes in SAS Enterprise Guide, again using
a Bonferroni correction (α=0.05/[18 abiotic variables*3 NMS axes]=0.0009).

Results
Shield darters were collected at 26 stream sites of the
422 sites sampled (6.2%) and 41 fish species were

included in the NMS ordination. Axis 1 of the PCA of
stream size variables of the initial 422 sites was the
only significant axis (eigenvalue of 3.69 compared to
Fig. 2 Plot of the first two
principal components (PCA
Axis 1 and 2) of stream size
variables for 422 sites where
shield was either present
(black circle) or absent
(white circle) in the Eastern
Piedmont. The lower limit
of the 95% confidence interval around the mean of
sites where shield darter was
absent is indicated by the
line within the plot

145

broken-stick eigenvalue of 2.45) and explained 61.5%
of the variance (Fig. 2). The highest loadings for the
first eigenvector were for average wetted width
(−0.48), average thalweg depth (−0.45), discharge
(−0.44), and catchment area (−0.44). The 95%
confidence interval was calculated around the mean
axis 1 ordination score (mean=0.25) of sites where
shield darter was absent. Sites above the lower
confidence limit (lower C.I.=0.10) were removed for
the Wilcoxon rank-sum tests and NMS to allow the
examination of potential chemical, habitat, and
landscape correlations with the distribution of this

species rather than stream size. A total of 263 sites
were above the confidence interval and removed for
the further analyses.
After the removal of a single outlier, 158 sites were
included in univariate and multivariate analyses.
Means and ranges of abiotic values between sites
where shield darters were present and absent are
presented in Table 2. Sulfate concentrations and riffle/
run quality scores were the only significantly different
(P<0.003) variables between present and absent sites.
Sulfate concentrations on average were lower at sites
where shield darters were present (mean=7.17 mg/L,
maximum=11.04 mg/L) compared to sites where
shield darters were absent (mean 11.28). Riffle/run
quality scores were, on average, higher at sites where
shield darters were observed (mean=17.1, minimum=
12) compared to sites where it was absent (mean=
13.8). Sites where shield darters were observed
typically had riffle/run scores in the optimal category
(see Table 2), categorized by stable substrates, a


146

Environ Biol Fish (2011) 92:141–150

Table 2 Mean and range of abiotic variables at sites where
shield darter was absent and present. The P values from the
Wilcoxon rank-sum tests between the present and absent sites

for each variables are also presented. Significant differences (P
<0.003) denoted by *

Absent (N=132)
Mean

Present (N=26)
Range

Mean

P value
Range

Water chemistry
pH
Conductivity (μmho/cm)
Dissolved organic carbon (mg/L)

7.66
362.5
2.1

6.48–9.50
70.3–12441.0
0.7–7.0

7.45
186.3

6.75–8.20

0.06

125.2–292.8

0.07
0.03

1.6

0.7–4.1

25.44

16.63–43.78

2.52–44.95

7.17

3.44–11.04

0.0004*

0.35–6.58

3.47

Chloride (mg/L)

66.40

5.81–3251.20

Sulfate (mg/L)

11.28

Total nitrogen (mg/L)

2.87

Total phosphorus (mg/L)

0.0353

Orthophosphate (mg/L)

0.0158

Ammonia (mg/L)

0.0341

Nitrite nitrogen (mg/L)

0.0118

Nitrate nitrogen (mg/L)

2.63

0.17–6.39

Dissolved oxygen (mg/L)

8.2

Instream habitat (0–20)
Pool/glide/eddy quality (0–20)
Riffle/run quality (0–20)

0.61

2.03–5.11

0.01

0.0205

0.0062–0.0490

0.40

0–0.5782

0.0045

0.0001–0.0231

0.03

0–0.5847

0.0153

0.0033–0.0752

0.16

0–0.0827

0.0090

0.0026–0.0167

0.91

3.21

1.86–4.82

0.01

4.5–10.7

8.3

4.6–9.8

0.18

15.0

5–19

16.4

12–19

0.01

14.6

8–18

14.9

11–18

0.83

13.8

0–19

17.1

12–19

<.0001*

0.006–0.682

Physical habitat

Land cover
Urban development (%)

14.7

0–88.9

3.9

0.8–20.2

0.03

Agriculture (%)

56.7

1.5–87.9

64.4

41.6–77.1

0.81

Forest (%)

27.0

8.4–80.5

30.2

13.7–41.7

0.01

variety of current velocities, depths greater than
10 cm, and maximum depths greater than 50 cm.
Although a significant difference in urbanization
percentages between sites where shield darter was
present and absent was not detected in the Wilcoxon
rank-sum test, sites where shield darter was present
had on average, low percentages of urban development in their upstream catchments (Table 2).
The NMS ordination found an optimum solution
at three dimensions. Final stress was 16.5 with a
final instability of 0.00121 at 200 iterations. The
variation explained for each axis was 34.3% for
axis 1, 20.1% for axis 2, and 28.9% for axis 3. The
strongest correlations of the abiotic variables to the
three ordination axes were found for riffle/run
quality (r=−0.46), nitrate (r=−0.39), and urban land
cover (r=0.39), all significantly (P<0.0009) correlated to axis 2 (Table 3). Other variables with

significant positive correlations to NMS axis 2

included conductivity (r=0.33), chloride (r=0.33),
dissolved organic carbon (r=0.33), and sulfate (r=
0.28). Total nitrogen and percent agriculture had
significant negative correlations to NMS axis 2 as
well (r=−0.37 and −0.33, respectively). Because
categorical presence or absence data was used for
fish communities, correlations could not be determined; however general patterns of fish assemblages
are evident (Fig. 3).
Biogeographical factors appear to explain the
ordination in fish assemblages along NMS axis 1,
with species towards the left, including greenside
darter, Etheostoma blennioides, fantail darter, E.
flabellare, and Potomac sculpin, Cottus girardi, being
either restricted or most common in the Potomac
River drainages, shifting towards species more common in the other western Chesapeake Bay drainages
towards the right of the ordination plot. NMS axis 2
illustrates changes in fish assemblage in relation to


Environ Biol Fish (2011) 92:141–150
Table 3 Correlations of
abiotic variables from the
non-metric multidimensional scaling based on Pearson’s r values on the three
significant axes. Significantly correlated variables
denoted with *(P<0.0009)

147
NMS Axis 1

NMS Axis 2

NMS Axis 3

r value

P value

r value

r value

−0.15

0.06

0.20

0.04

0.63

0.19

0.02

0.33

<0.0001*

0.10

0.22

−0.12

0.14

0.33

<0.0001*

−0.12

0.13
0.21

P value

P value

Water chemistry
pH
Conductivity
Dissolved organic carbon

0.01

Chloride

0.24

0.00

0.33

<0.0001*

0.10

Sulfate

0.03

0.73

0.28

0.0004*

0.01

0.91

−0.02

0.84

−0.37

<0.0001*

0.22

0.01

Total nitrogen
Total phosphorus

−0.07

0.40

0.05

0.50

0.05

0.54

Orthophosphate

−0.04

0.62

0.01

0.92

0.04

0.58

Ammonia

−0.14

0.09

0.13

0.10

0.03

0.73

Nitrogen dioxide

−0.01

0.93

0.14

0.08

0.11

0.16

Nitrate

−0.01

0.89

−0.39

0.004

−0.16

Dissolved oxygen

0.23

<0.0001*

0.20

0.01

0.04

−0.14

0.07

0.002

−0.003

0.97

0.60

0.15

0.05

Physical habitat
Instream habitat

−0.20

0.01

−0.25

Pool/glide/eddy quality

−0.01

0.91

0.04

Riffle/run quality

−0.06

0.49

−0.46

<0.0001*

0.05

0.56

0.39

<0.0001*

0.04

0.59

<0.0001*

Land cover
Urban development

0.26

0.001

Agriculture

−0.21

0.01

−0.33

Forest

−0.14

0.07

−0.11

the abiotic variables, notably riffle/run quality as well
as land cover and associated chemical variables.
Species above NMS axis 1 are species that are pool
specialists or generally do not prefer riffle habitats,
including several centrarchid species, brown bullhead,
Ameiurus nebulosus, and several larger system cyprinids. Below axis 1 are species that rely more on riffle
habitats, including shield darters. Within this lower
half of the plot, a disturbance gradient is present with
intolerant species clustering towards the bottom,
including shield darter, river chub, Nocomis micropogon, and common shiner, Luxilus cornutus, in areas
with higher percent agriculture in upstream catchments, and tolerant species, including white sucker,
Catostomus commersoni, tessellated darter, Etheostoma olmstedi, and eastern blacknose dace, Rhinichthys atratulus, clustering towards the top of the entire

riffle/run species assemblage coinciding with the
urban development vector (Fig. 3). The NMS plot
indicates shield darters are associated with pollution
intolerant species that associate with deeper riffle
habitats with a variety of water velocities at sites with

0.17

0.07

0.38

−0.16

0.04

low urban development and higher concentrations of
nitrate. Furthermore, correlations of variables to NMS
axis 2 suggest the shield darter and the intolerant
species it is associated with were collected at sites
with low concentrations of sulfate and chloride, low
conductivity, and higher total nitrogen concentrations.

Discussion
Physical habitat, water chemistry, watershed landscape, and fish assemblage data here provide insight
into the potential conservation of the state imperiled
shield darter. Shield darters in the Eastern Piedmont
of Maryland are associated with larger-sized streams
and rivers with high quality habitats and low concentrations of contaminants. The association with riffle
habitats with diverse velocities, depths, and substrates

has been documented in several Percina species,
including the closely related gilt darter, P. evides, and
blackside darter, P. maculata (Thomas 1970; Hatch
1985; Skyfield and Grossman 2008). While the


148

Environ Biol Fish (2011) 92:141–150

Fig. 3 Joint plots of the results of the non-multidimensional
scaling of 18 abiotic variables with 41 fish species, with
emphasis on shield darter (in box). Abiotic variables (in italics)
whose correlation strength exceeded a Pearson’s r2 of 0.15 are

shown as vectors. Species are coded as pollution intolerant
(black square), pollution tolerant (white circle), or as not
having a tolerance category (+)

response of shield darters to pollutants such as
sulfate and chloride are not specifically documented, the concentration and/or ratios of these
contaminants have been documented to negatively
influence the survival and reproduction of fishes in
general (Burnham and Peterka 1975; Saiki et al.
1992) and can trigger the avoidance of polluted areas
by sensitive species (Giattina et al. 1981). The cooccurrence of several pollution intolerant species
such as river chubs and common shiners with shield
darters further suggests that shield darter is sensitive
to altered habitat quality.
Diverse, stable riffle habitats and unpolluted waters

are typically associated with low levels of urbanization (Klein 1979). The preponderance of evidence
linking physical alterations to stream channels with
urbanization in previous studies (Herlihy et al. 1998;
Paul and Meyer 2001; Walters et al. 2003; Chadwick
et al. 2006; Stranko et al. 2008) suggests a potential
pattern of habitat alteration with urban development,
although causal mechanisms cannot be determined

from the present analyses. Urban development has
been linked to alterations in water chemistry in
Maryland (Morgan et al. 2007; Walsh et al. 2005),
and is again supported here with the positive
correlations of sulfate, chloride, and conductivity with
NMS axis 2 in the direction of the urbanization vector
(Table 3). Elevated sulfate and chloride concentrations may be linked to industrial and municipal waste
as well as road salting in these more urban developed
areas (Eisen and Anderson 1979; Kaushal et al.
2005).
Although the underlying mechanisms responsible
are not well understood, several other darter species
have also demonstrated vulnerability to the impacts of
urbanized conditions. In a survey of Wisconsin
streams, the blackside darter was not found in streams
with greater than 10% imperviousness in urban
environments (Wang et al. 2000). Runoff from
anthropogenic activities appears to be the main threat
to the goldline darter, P. aurolineata (Powers and
Mayden 2004). Etowah darter Etheostoma etowahae

Environ Biol Fish (2011) 92:141–150

and bronze darter P. palmaris occurrences in the
Etowah River basin in Georgia were negatively
correlated with increasing impervious surface area
associated with urban landscapes (Wenger et al.
2008). The low abundance of the tolerant johnny
darter E. nigrum in Tuckahoe Creek, Virginia was
suggested to be the result of siltation and consequent
reduction of benthic prey due to shifts from agriculture and forest to urban land-use patterns (Weaver
and Garman 1994).
The correlation of nitrate with NMS axis 2 likely
reflects increasing agricultural land use as the percentage of urban land cover decreases. Despite the fact that
agriculture also alters watersheds, there was no
negative association of shield darter or other intolerant
species occurrences with increasing percentages of
agriculture and nutrient (nitrate) runoff associated with
farming practices. This could be, in part, because most
of the land in Maryland was cleared for farming
100 years ago (Brush 2009), potentially eliminating
stream species that could not tolerate extensive
sedimentation from agricultural run-off (Harding et al.
1998). Conversely, urban land cover has only recently
begun spreading away from the major metropolitan
centers of Baltimore and Washington D.C. impacting
species such as shield darter.
Indexes of Biological Integrity (IBIs) have proven
useful for identifying thresholds of watershed urbanization in the impairment of aquatic ecosystems.
Impervious cover of 10–15% associated with urban
development in a stream catchment is generally

accepted as the maximum threshold for preserving
biological integrity (Klein 1979; Wang et al. 2000;
Stranko et al. 2008). In Maryland streams, high
catchment urbanization is linked to higher numbers
of tolerant species and lower IBI scores (Morgan and
Cushman 2005); however such analyses often omit
rare species such as the shield darter. While these
methods are effective in addressing overall stream
health, they may often not be applicable to rare
species conservation due to the impacts of rare
species in statistical analyses as outliers. Species
specific approaches, such as the analyses used here
for shield darters, allow for individual life history
traits and ecologies to be addressed in species
conservation. Causal mechanisms, i.e., the actual
impacts of sulfate and chloride concentrations on
shield darter development and physiology, should be
the next step in understanding the specific threats to

149

this species. The patterns identified here though
indicate the conservation of shield darter, a species
in potential decline in Maryland, necessitates the
protection of larger sized streams in the Eastern
Piedmont through the preservation of high quality
riffle habitats and the limitation of pollutants associated with urbanized watersheds.
Acknowledgments We thank Matt Kline, Anthony Prochaska, Jay Kilian, Marty Hurd, Christopher Millard, and the
many MBSS sampling crew members and volunteers for data
collection. We also thank Jay Kilian, Michael Kashiwagi,

Matthew Ashton, Andrew Becker, and Christopher Swan for
valuable assistance and advice and Rich Raesly for critical
review. This study was funded in part by State Wildlife Grant
funds provided to the state wildlife agencies by U.S. Congress
and administered through the Maryland Department of Natural
Resources’ Natural Heritage Program.

References
Allan JD, Flecker AS (1993) Biodiversity conservation in
running waters: identifying the major factors that threaten
destruction of riverine species and ecosystems. Biosci
43:32–43
Brush GS (2009) Historical land use, nitrogen, and coastal
eutrophication: a paleoecological perspective. Estuaries
Coasts 32:18–28
Burkhead NM, Walsh SJ, Freeman BJ, Williams JD (1997)
Status and restoration of the Etowah River, an imperiled
southern Appalachian ecosystem. In: Benz GW, Collins
DE (eds) Aquatic fauna in peril: the southeastern perspective. Southeast Aquatic Research Institute Special Publication 1, Decatur, pp 375–444
Burnham BL, Peterka JJ (1975) Effects of saline water from
North Dakota lakes on survival of fathead minnow
(Pimephales promelas) embryos and sac fry. J Fish Res
Board Can 32:809–812
Chadwick MA, Dobberfuhl DR, Benke AC, Huryn AD,
Suberkropp K, Thiele JE (2006) Urbanization affects
stream ecosystem function by altering hydrology, chemistry, and biotic richness. Ecol Appl 16:1796–1807
Eisen C, Anderson MP (1979) The effects of urbanization on
ground-water quality—a case study. Ground Water
17:456–462
ESRI (Environmental Systems Research Institute) (2005)

ArcMap version 9.1. ESRI, Redlands, CA
Freeman MC (1995) Movements of two small stream fishes in a
large stream. Copeia 1995:361–367
Giattina JD, Cherry DS, Cairns JC Jr, Larrick SR (1981)
Comparison of laboratory and field avoidance behavior of
fish in heated chlorinated water. Trans Am Fish Soc 110:526–
535
Harding JS, Benfield EF, Bolstad PV, Helfman GS, Jones EBD
III (1998) Stream biodiversity: the ghost of land use past.
Proc Natl Acad Sci USA 95:14843–14847


150
Hatch JT (1985) Distribution, habitat, and status of the gilt darter
(Percina evides) in Minnesota. J Minn Acad Sci 51:11–16
Herlihy AT, Stoddard JL, Johnson CB (1998) The relationship
between stream chemistry and watershed land cover data in the
mid-Atlantic region, U.S. Water Air Soil Pollut 105:377–386
Homer CC, Dewitz J, Fry J, Coan M, Hossain N, Larson C,
Herold N, McKerrow A, VanDriel JN, Wickham J (2007)
Completion of the 2001 National Land Cover Database for
the Conterminous United States. Photogramm Eng Remote Sens 73:337–341
Jelks HL, Walsh SJ, Burkhead NM, Contreras-Balderas S,
Díaz-Pardo E, Hendrickson DA, Lyons J, Mandrak NE,
McCormick F, Nelson JS, Platania SP, Porter BA, Renaud
CB, Schmitter-Soto JJ, Taylor EB, Warren ML Jr (2008)
Conservation status of imperiled North American freshwater and diadromous fishes. Fisheries 33:372–407
Jenkins RE, Burkhead NM (1994) Freshwater Fishes of Virginia.
American Fisheries Society, Bethesda, pp 800–802
Kaushal SS, Groffman PM, Likens GE, Belt KT, Stack WP,

Kelly VR, Band LE, Fisher GT (2005) Increased salinization of fresh water in the northeastern United States. Proc
Natl Acad Sci USA 102:13517–13520
Kazyak PF (2001) Maryland biological stream survey sampling
manual. Maryland Department of Natural Resources, Monitoring and Non-tidal Assessment Division, Annapolis, MD
Klein RD (1979) Urbanization and stream quality impairment.
Water Resour Bull 15:948–963
McCune B, Grace JB (2002) Analysis of ecological communities. MjM Software, Gleneden Beach, pp 119–120
McCune B, Mefford MJ (2006) PC-ORD multivariate analysis
of ecological data, version 5.10. MjM Software, Gleneden
Beach
MDNR (Maryland Department of Natural Resources) (2007)
Rare, threatened, and endangered animals of Maryland.
Maryland Department of Natural Resources, Wildlife and
Heritage Service, Natural Heritage Program. Annapolis, MD
Morgan RP, Cushman SF (2005) Urbanization effects on stream
fish assemblages in Maryland, USA. J North Am Benthol
Soc 24:643–655
Morgan RP, Kline KM, Cushman SF (2007) Relationships
among nutrients, chloride, and biological indices in urban
Maryland streams. Urban Ecosyst 10:153–166
New JG (1966) Reproductive behavior of the shield darter,
Percina peltata peltata, in New York. Copeia 1966:20–28
Paul MJ, Meyer JL (2001) Streams in the urban landscape.
Annu Rev Ecol Syst 32:333–365
Powers SL, Mayden RL (2004) Threatened fishes of the world:
Percina aurolineata Suttkus & Ramsey, 1967 (Percidae).
Environ Biol Fish 63:16
Rantz SE (1982) Measurement and computation of streamflow:
Volume 1. Measurement of stage and discharge. U.S.
Geological Survey Water Supply Paper 2175, 1982,

Washington DC
Reynoldson TB, Rosenberg, DM (1996) Sampling strategies
and practical considerations in building reference data
bases for prediction of invertebrate community structure.
In: Bailey RC, Norris RH, Reynoldson TB (eds) Study
design and data analysis in benthic macroinvertebrate
assessment of freshwater ecosystems using a reference site
approach. North American Benthological Society, 44th
Annual Meeting, pp 1–31

Environ Biol Fish (2011) 92:141–150
Rodriguez MA, Lewis WM Jr (1997) Structure of fish
assemblages along environmental gradients in floodplain
lakes of the Orinco River. Ecology 67:109–128
Rohde FC, Arndt RG, Lindquist DG, Parnell JF (1994)
Freshwater fishes of the Carolinas, Virginia, Maryland,
and Delaware. The University of North Carolina Press,
Chapel Hill, pp 194–196
Saiki MK, Jennings MR, Wiedmeyer RH (1992) Toxicity of
agricultural subsurface drainwater from the San Joaquin
Valley, California, to juvenile Chinook salmon and striped
bass. Trans Am Fish Soc 121:78–93
Institute SAS (2006) SAS Enterprise Guide, version 4. SAS
Institute, Cary, North Carolina
Schmidt RE, Daniels RA (2004) Description of the larval stages
of the shield darter, Percina peltata (Pisces: Percidae), in
New York. Zootaxa 774:1–7
Skyfield JP, Grossman GD (2008) Microhabitat use, movements
and abundance of gilt darters (Percina evides) in southern
Appalachian (USA) streams. Ecol Freshw Fish 17:219–230

Southerland MT, Rogers GM, Kline MJ, Morgan RP, Boward
DM, Kazyak PF, Klauda RJ, Stranko SA (2005) New
biological indicators to better assess the conditions of
Maryland streams. Maryland Department of Natural
Resources, Monitoring and Non-tidal Assessment Division, Annapolis, MD
Stranko SA, Hildebrand RH, Morgan RP II, Staley MW, Becker
AJ, Roseberry-Lincoln A, Perry ES, Jacobson PT (2008)
Brook trout declines with land cover and temperature
changes in Maryland. N Am J Fish Manage 28:1223–1232
Tabit CR, Johnson GM (2002) Influence of urbanization on the
distribution of fishes in a southeastern upper piedmont
drainage. SE Nat 1:253–268
Thomas DL (1970) An ecological study of four darters of the
genus Percina (Percidae) in the Kaskaskia River, Illinois.
Illinois Natural History Survey. Biol Notes 70
Utz RM, Hildebrand RH, Boward DM (2009) Identifying regional
differences in threshold responses of aquatic invertebrates to
land cover gradients. Ecol Indic 9:556–567
Walsh CJ, Roy AH, Feminella JW, Cottingham PD, Groffman
PM, Morgan RP II (2005) The urban stream syndrome:
current knowledge and the search for a cure. J North Am
Benthol Soc 24:706–723
Walters DM, Leigh DS, Beardon AB (2003) Urbanization,
sedimentation, and the homogenization of fish assemblages
in the Etowah River Basin, USA. Hydrobiologia 494:5–10
Wang L, Lyons J, Kanehl P, Bannerman R, Emmons E (2000)
Watershed urbanization and changes in fish communities
in southeastern Wisconsin streams. J Am Water Resour
Assoc 36:1173–1189
Wang L, Lyons J, Kanehl P, Gatti R (1997) Influence of

watershed land use on habitat quality and biotic integrity
in Wisconsin streams. Fisheries 22:6–12
Warren ML Jr, Pardew MG (1998) Road crossings as barriers to
small-stream fish movement. Trans Am Fish Soc 127:637–644
Weaver LA, Garman GC (1994) Urbanization of a watershed
and historical changes in a stream fish assemblage. Trans
Am Fish Soc 123:162–172
Wenger SJ, Peterson JT, Freeman MC, Freeman BJ, Homans
DD (2008) Stream fish occurrence in response to
impervious cover, historic land use, and hydrogeomorphic
factors. Can J Fish Aquat Sci 65:1250–1264


Environ Biol Fish (2011) 92:151–157
DOI 10.1007/s10641-011-9824-0

A size-dependent migration strategy in Atlantic salmon
smolts: Small smolts favour nocturnal migration
Anton T. Ibbotson & William R. C. Beaumont &
Adrian C. Pinder

Received: 9 February 2010 / Accepted: 4 April 2011 / Published online: 4 May 2011
# Springer Science+Business Media B.V. 2011

Abstract Migration theory states that migration
behavioural strategies should be optimised to maximise
fitness. Many studies have shown that in downstream
migrating Atlantic salmon Salmo salar L. smolts,
mortality from predation is high and negatively size
dependent. The most common predators are birds and

piscivorous fish that are mainly daylight feeders. Given
the high mortality during this stage we should expect
to observe smolts to follow predator avoidance
strategies that may be affected by body size. We tested
the hypothesis that small smolts have a higher
tendency to exhibit predator avoidance strategies (i.e.
nocturnal versus diurnal migration) than larger smolts.
The number and size of out-migrating/downstreammigrating wild Atlantic salmon smolts was recorded as
they passed through a glass-sided channel during
April-May, 1996–1999. In all years, the mean size of
nocturnal migrating smolts was significantly lower
than the mean size of diurnal migrating smolts.
Analysis of the size of smolts, during early and late
A. T. Ibbotson (*) : W. R. C. Beaumont
Salmon and Trout Research Centre,
Game and Wildlife Conservation Trust,
East Stoke,
Wareham BH20 6BB Dorset, UK
e-mail:
A. C. Pinder
Aquatic Ecology Laboratories,
APEM Ltd,
East Stoke,
Wareham BH20 6BB Dorset, UK

stages of the migration period showed size-dependent
nocturnal migration behaviour up to the end of April.
After this, no such size dependent migration pattern
was observed. However, small smolts (<100 mm) were
absent during this period. We suggest that nocturnal

migration is an adaptive behaviour that small Atlantic
salmon smolts have to avoid predation by large
daylight feeding visual piscivorous predators (e.g. pike
Esox Lucius L. and fish eating birds).
Keywords Atlantic salmon . Salmo salar . Smolt .
Nocturnal migration . Diel migration . Size dependent
behaviour

Introduction
Many animals undertake extensive migrations to take
advantage of more productive feeding grounds or to
reach breeding sites (Baker 1978). Whilst the benefits
can include increased growth rates and higher survival
of juveniles (Baker 1978), migration mortality can also
be high, e.g., up to 85% in a long distance migrant,
like a passerine (Sillett and Holmes 2002) and between
65 and 73% in smolts of a wild steelhead trout
(Oncorhynchus mykiss) population (Melnychuk et al.
2007). Theoretically, natural selection should favour
those behavioural strategies that reduce mortality and
maximise fitness (Berthold 2001; Hedenstrom 2008).
Examples of varying migration behaviour include
differential patterns of migration based on differences


152

in timing, sex and migration distance (Dierschke et al.
2005; Johnson et al. 2007; Coppack and Pulido 2009).
In birds, optimality analyses and models (e.g. Alerstam

and Hedenstrom 1998) have been used to predict that
natural selection favours individuals exhibiting traits
for increased migration speed rather than predator
avoidance (Hedenstrom 2008). Similar predictions for
fishes are more difficult as migratory behavioural traits
are less documented and thus models are not as well
developed.
Atlantic salmon Salmo salar L. juveniles exhibit
differential patterns of migration during their life
history. For example, a portion of the parr population
migrates downstream in the autumn (Pinder et al.
2007). Seaward migration of smolts occurs in the
spring, with older and larger smolts showing a
tendency to migrate in the earlier part of the migration
period (Bohlin et al. 1996; Jutila and Jokikokko
2008), although smolt body size shows no consistent
pattern with time of migration in the River Frome
(pers. obs.). Furthermore, juvenile salmonids and
migrating smolts exhibit diel changes in feeding
activity patterns (Godin 1981; Alanärä and Brännäs
1997; Reebs 2002; Carlsen et al. 2004). In general
this is linked to ambient temperatures, with warmer
conditions favouring daytime activity and colder
conditions nocturnal activity (Fraser et al. 1993),
although these activity patterns are complicated by
physiological state-dependent individual differences
and the availability of food (Metcalfe et al. 1999).
Parr approaching smoltification and migration tend
more towards diurnal activity in winter conditions,
especially if they are small, despite the increased risks

of predation during daylight (Metcalfe et al. 1998).
The increased body size achieved by the more efficient
daylight feeding may decrease size-dependent predation risks during subsequent migration. The smolts
tend to nocturnal migration during the first few weeks
but then switch to favouring daylight migration at
higher temperatures towards the end of the migration
period (Ibbotson et al. 2006).
The spring smolt migration of juvenile salmonids
represents the high-risk element of a strategy that
affords the individual survivors with higher growth
rates, larger final sizes and increased reproductive
opportunities in the form of higher fecundities, larger
egg size and better defence of spawning territories
(Klemetsen et al. 2003). Since many salmonid species,
and populations, exhibit both seaward migrating and

Environ Biol Fish (2011) 92:151–157

river resident strategies (Jonsson et al. 2001; Klemetsen
et al. 2003), there must be a fine balance between the
advantages and disadvantages of these contrasting
strategies. The juvenile salmon smolts, usually ranging
from 80 to 250 mm (fork length), leave their low
predation natal streams in early spring and migrate
downstream through habitats that are unfamiliar and
where predation rates from birds and piscivorous fish
are known to be high (Mann 1982; Larsson 1985;
Kennedy and Greer 1988; Bostrom et al. 2009).
With such a high cost of migration there should be
a strong evolutionary pressure favouring behavioural

strategies that can reduce mortality risk during this
migration. Birds and fish, most of which are daytime
visual hunters, are the most common predators of
smolts (Thorpe and Moore 1996) and predation on
smolts has been shown to be negatively sizedependent (Feltham 1995) although not always
(Hvidsten and Lund 1988). Therefore we could
hypothesise that small smolts would have a higher
tendency to exhibit predator avoidance strategies,
specifically nocturnal versus diurnal migrations, than
larger smolts. Such an observation would demonstrate
the importance of subtle influences on natural
selection at key life-history stages and lead to a
greater understanding of the component parts that
make up an entire stock. We tested this hypothesis by
measuring smolts at a real-time automatic smolt
counter on the River Frome, southern England during
the spring, downstream, seaward migration period,
1996–1999.

Materials and methods
Since 1995, an automatic smolt counter has operated
during April–May on the River Frome (8.6 km above
the mean high water mark, at the River Laboratory,
East Stoke). The detailed operation of the counter has
been described (Welton et al. 2002), but essentially
the smolts are diverted from the main River Frome
with the aid of an acoustic bubble screen into the East
Stoke millstream that flows through a fluvarium. Fish
are then counted without handling as they pass
through resistivity counting tubes. Identification of

fish was made with the use of side viewing infra-red
sensitive CCTV cameras (illuminated with infra-red
light) positioned immediately below the counting
tubes. For 4 years (1996–1999), all fish successfully


Environ Biol Fish (2011) 92:151–157

filmed in the glass sided channel were measured on
screen and true fork lengths calculated from a linear
relationship between screen size and real size. Periods
of high turbidity were excluded due to the difficulty
in collecting video footage of migrating smolts during
those periods. This resulted in a total of 6 days being
missed over the 4 years.
In each year, differences in the cumulative percentage frequency of lengths of nocturnal and diurnal
migrating smolts were tested using the KolmogorovSmirnov two-sample test. To avoid the difficulty of
designating dawn and dusk conditions to night and
day, daylight periods were considered to occur
between half an hour after sunrise and half an hour
before sunset; and dark hours between half an hour
after sunset and half an hour before sunrise. Thus, 1 h
at sunrise and 1 h at sunset (crepuscular periods) were
excluded from the analysis. However, smolt migration
rarely takes place during these crepuscular periods
(Ibbotson et al. 2006) and so few smolts were omitted
from the analysis.
The smolt run of each year, was divided into three
separate periods (P1–P3 representing 30 March–14
April; 15 April–30 April; 1 May onwards), to negate

the temperature and timing influence on the diel
behaviour of smolt migration (Ibbotson et al. 2006).
A general linear model was used to analyse the
effects of time of day, period and year together with
their interactions on the mean size of migrating smolts,
with smolt length (log) as the dependent variable and
time of day (night or day), period (P1–P3) and year
(1996–1999) as factorial independent variables. We
used paired comparisons to examine the three-way
interaction between time of day, period and year.

Results
In each year, the cumulative percentage frequency of
lengths of nocturnal migrating smolts showed that
these were smaller than daylight migrating smolts
(p<0.001 in each year 1996–1999). For all years, the
median size was considerably smaller for nocturnal
smolts (137 vs. 146 mm in 1996; 138 vs. 145 mm in
1997; 154 vs. 162 mm in 1998; 140 vs. 150 mm in
1999; Fig. 1).
Mean smolt size varied between years (F3,10396 =
51.3, p < 0.00001) and period (F 2,10396 = 17.1,
p<0.00001) across all years. The significant interac-

153

tion term (F6,10396 =49.7, p<0.00001) between these
two showed there was no consistent change in smolt
size with period within each year. That is, in 1996 and
1997, smolt size tended to increase in P3, but in 1998

and 1999, smolt size did not change throughout the
three periods. Whilst paired comparisons between
years showed that mean sizes in most years were
significantly different, only smolt size in P3 was
significantly different from the other periods. It should
be noted that over the 4 years 68% of the smolts
migrated within the periods P1 and P2.
Across all years and periods, the mean size of
nocturnal smolts (143 mm ±1.0) was significantly
smaller than diurnal smolts (151 mm ± 1.9) (F1,10396 =
52.7, p<0.00001). However, a significant interaction
term between time of day and year (F3,10396 =7.3,
p=0.00007) indicates that this effect was not always
consistent. That is, the mean size of nocturnal smolts
was significantly smaller in every year apart from
1997, when there was no significant difference
between the size of nocturnal and diurnal smolts. In
all three periods nocturnal smolts were significantly
smaller than diurnal smolts (Fig. 2). A significant
interaction term between time of day and period
(F2,10396 =3.2, p=0.042) was due to the slightly larger
mean size of diurnal smolts in P1 as opposed to P2,
and the increasing mean size of nocturnal smolts with
period (Fig. 2).
The mean sizes of nocturnal smolts were
significantly smaller than diurnal smolts in 8 (4 in
P1; 2 in P2 and 2 in P3) out of 12 paired occasions
(3 periods x 4 years), explaining the significant
(F6,10396 =4.2, p<0.0004) three-way interaction term.
Nocturnal smolts were not significantly larger than

diurnal smolts on any occasion.
The probability of a smolt migrating at night
(adjusted for the comparative numbers of dark and
light hours in each period) was heavily dependent on
smolt size in P1 and P2 (test for linear trend in
proportions, P1, z = 3.74, p < 0.01; P2, z = 6.77,
p<0.001), but not in P3 (z=1.14, p=0.13; Fig. 3).

Discussion
This observational study of a wild Atlantic salmon
smolt population, 1996–1999 demonstrates that during
the downstream spring smolt run, nocturnal migrating
smolts are smaller than the diurnal migrating smolts.


154

Environ Biol Fish (2011) 92:151–157

Fig. 1 Cumulative percentage length frequency of
nocturnal (dashed line) and
diurnal (solid line) smolts
passing a fixed point on the
River Frome, Dorset,
England, during April–May,
1996–1999

Whilst, the proportion of smolts migrating during the
day increases in the later stages of the smolt migration
period (Ibbotson et al. 2006), smaller smolts always

exhibit a higher probability of migrating at night than
the larger smolts.
All smolts used in this study were deflected into a
side channel and then on to the smolt counter using an

Fig. 2 Mean size (mm, fork lengths with 95% CI) of nocturnal
and diurnal smolts migrating down the River Frome during
three time intervals (P1: 30 March–14 April; P2: 15 April–30
April; P3: 1 May–30 May), 1996–1999. In all periods,
nocturnally migrating smolts were significantly smaller than
diurnally migrating smolts (p<0.001; t-test on log-transformed
lengths)

acoustic bubble screen (Welton et al. 2002) and thus
we need to consider whether the results described
here reflect fish behaviour past these structures or real
migration patterns on unaffected river stretches. Trials
of the efficiency of the acoustic bubble screen used in
this study (Welton et al. 2002) indicated that it was
more effective at diverting fish during the hours of
darkness and less effective during daylight hours. In
order to explain the observation that larger smolts had

Fig. 3 Proportion of nocturnal smolts, by 20 mm size groups,
migrating down the River Frome, England, during three
different periods (P1: 30 March–14 April; P2: 15 April–30
April; P3: 1 May–30 May), 1996–1999. Lines represent
significant trends (p<0.05; test for linear trend in proportions)

Environ Biol Fish (2011) 92:151–157

a higher probability of migrating during the day, the
acoustic bubble screen would have to have a higher
efficiency for deflecting large fish during daylight
hours. During all the trials (Welton et al. 2002) no
size selection of the screens was observed. However if
there were size selection one might expect the larger
more confident fish to pass through the acoustic
bubble screen more easily with higher deflection rates
for the smaller fish. It is conceivable that the lower
efficiency of the acoustic bubble screen during
daylight may be the result of larger fish migrating
during this part of the 24-hour cycle rather than its
operation resulting in the observations we made. It is
known that smolts avoid channels that are covered
(Kemp et al. 2008) preferring to migrate through
uncovered channels when they have the choice.
Whilst there has been an observed hesitance of smolts
to enter the darkened fluvarium this appears to only
last a short period of a few minutes and does not
result in the accumulation of large numbers of smolts
ahead of the counters. This may be because the smolts
have no choice of using an uncovered channel at this
site. Indeed the numbers of smolts migrating through
the counters during daylight is greater than those
migrating during the night in some years (Ibbotson et
al. 2006). We therefore conclude that the structures
used to deflect and count smolts on the River Frome
are unlikely to be the cause of the observations made

in this study.
Nocturnal migration by small smolts is most likely
a strategy to reduce predation risk, which can be very
high in some populations. For example, goosanders
have been estimated to take between 3 and 16% of the
North Esk Atlantic salmon population (Feltham
1995), while Larsson (1985) found predation to
account for 50% of the smolt run in Baltic salmon.
Kennedy and Greer (1988) found that cormorants
Phalacrocorax carbo (L.) started feeding about 1 h
after sunrise and in the River Bush could remove
between 51% and 66% of the annual smolt run. In the
River Frome, it has been estimated that 30% of the
Atlantic salmon smolt population, that is the subject
of this study, is consumed by pike Esox lucius (L.)
(Mann 1982) . Smolts often experience negative
size-dependent predation rates (West and Larkin
1987) with smaller smolts more vulnerable to
predation by birds (Feltham 1990; Feltham 1995)
and fish (Bakshtansky and Nesterov 1976; Duffy and
Beauchamp 2008).

155

As the smolt migration progresses, smolts show an
increasing tendency to migrate in daylight hours. This
has been attributed, in other studies, to increases in
temperature (Thorpe et al. 1994; Ibbotson et al. 2006)
and associated improved ability to avoid endothermic
predators (Bakshtansky and Nesterov 1976; Thorpe et

al. 1994) and presumably the large number of
ectothermic predators e.g. pike that inhabit the River
Frome (Mann 1982). Similarly, nocturnal foraging
activity has been observed to occur in Atlantic salmon
parr at low temperatures as opposed to daylight
sheltering behaviour, with a shift to daylight foraging
in warmer temperatures (Heggenes et al. 1993).
However, other studies of salmonids have found that
temperature is not the only factor affecting diel rhythms
in salmonids (Alanärä and Brännäs 1997; Reebs 2002).
Our study highlights the importance of body size as a
factor in determining smolt migration behaviour.
One study has observed the same individual smolts
migrating both night and day later in the smolt run
(Moore et al. 1995), possibly responding to a need for
smolts to reach their destination at the optimum time
(Hansen and Jonsson 1989). This may explain the
observation of this study that from the 1 May onwards
(P3) the probability of migrating at any particular time
of day was not dependent on fish size. This cannot be
the case during the early part of the migration period
when the size differences between the smolts migrating
during darkness and daylight indicate these form two
different migrating components of the smolt population.
Prior to migration as smolts, small wild Atlantic
salmon parr have been shown to grow fast in the
months leading up to migration (Nicieza and Brana
1993) and in the laboratory small pre-smolts favoured
diurnal feeding to increase growth rate and thus body
size at the time of migration (Metcalfe et al. 1998).

This has been explained as a strategy to reduce future
predation risk at the expense of current predation risk,
which may be at a lower rate (Metcalfe et al. 1998).
This study demonstrates that at the onset of downstream
migration, smaller smolts have a higher tendency to
exhibit nocturnal migration. We suggest this reflects a
response to a greater predation risk for these fish.
Migration during daylight may aid visibility and
inspection of obstacles as fish travel through new
habitats and the balance between these advantages and
a lower risk of predation from visual predators may
result in a greater probability of daylight migration in the
larger smolts. Since, this study shows that individual


156

migration strategies depend on size, smolt migration
behaviour can be considered as much more complex
than can be supposed from the typical conceptual model
of a shift from nocturnal to diurnal migration as
temperatures increases (Thorpe et al. 1994; Ibbotson et
al. 2006). Further, salmonids may represent an example
of a group of animals where natural selection favours
behavioural traits that increase predator avoidance
during migration. Importantly, such subtle variations
in animal migration behaviour may be much more
common than supposed and we cannot discount the
role these may play in determining survival rates.
Any potential interruption or altering of natural

migration patterns e.g. operation of traps, weirs, bypass channels, outflows and non-generating periods in
hydropower plants could be detrimental to the
survival of some portions of salmonid smolt populations. For example, the common operation of mobile
research traps at night to take advantage of peak
migration periods may selectively sample small fish,
and hinder their ability to make progress downstream,
during the dark hours when they are less vulnerable to
predators. There is already some evidence that the
operation of traps may have a negative effect on survival
(Crozier and Kennedy 2002) and behaviour (Riley et al.
2007). Stocking practice, especially at the smolt stage
needs to be sensitive to the migration needs of the
stocked fish. Since these may be dependent on the
status of individual fish we would advise stocking into
open pens where the fish can choose when to migrate
instead of imposing choice on the fish by stocking
during daylight or at night (Roberts et al. 2009). We
conclude that migrating salmonids exhibit subtle
behavioural traits that may be important in reducing
natural mortality. Preserving the opportunity to exhibit
these should be an important part in the design and
planning of any river management.
Acknowledgements The authors would like to thank the many
students that spent hours watching videos of salmon smolts. The
collection of videos was completed during employment of the
authors at the Centre For Ecology and Hydrology.

References
Alanärä A, Brännäs E (1997) Diurnal and nocturnal feeding
activity in Arctic char (Salvelinus alpinus) and rainbow

trout (Oncorhynchus mykiss). Can J Fish Aquat Sci
54:2894–2900

Environ Biol Fish (2011) 92:151–157
Alerstam T, Hedenstrom A (1998) The development of bird
migration theory. J Avian Biol 29:343–369
Baker R (1978) The evolutionary ecology of animal migration.
Holmes and Meier Publishers Inc, Teaneck, 1024 pp
Bakshtansky EL, Nesterov VD (1976) Hunting activities of
pike and their possible effect on the diurnal pattern of
downstream run of fingerlings of Atlantic salmon.
Trubui Vcecoynsnovo nauchno-uclebovatelkovo instituta morskovo ruybnovo chosyaustva i okeanografii
113:39–42
Berthold P (2001) Bird migration: a novel theory for the
evolution, the control and the adaptability of bird
migration. J Ornithol 142:148–159
Bohlin T, Dellefors C, Faremo U (1996) Date of smolt
migration depends on body-size but not age in wild searun brown trout. J Fish Biol 49:157–164
Bostrom MK, Lunneryd SG, Karlsson L, Ragnarsson B (2009)
Cormorant impact on trout (Salmo trutta) and salmon
(Salmo salar) migrating from the river Dalalven emerging
in the Baltic Sea. Fish Res 98:16–21
Carlsen KT, Berg OK, Finstad B, Heggberget TG (2004) Diel
periodicity and environmental influence on the smolt
migration of Arctic charr, Salvelinus alpinus, Atlantic
salmon, Salmo salar, and brown trout, Salmo trutta, in
northern Norway. Environ Biol Fish 70:403–413
Coppack T, Pulido F (2009) Proximate control and adaptive
potential of protandrous migration in birds. Integr Comp
Biol 49:493–506

Crozier WW, Kennedy GJA (2002) Impact of tagging with
coded wire tags on marine survival of wild Atlantic
salmon (Salmo salar L.) migrating from the R. Bush,
Northern Ireland. Fish Res 59:209–215
Dierschke V, Mendel B, Schmaljohann H (2005) Differential
timing of spring migration in northern wheatears Oenanthe
oenanthe: hurried males or weak females? Behav Ecol
Sociobiol 57:470–480
Duffy EJ, Beauchamp DA (2008) Seasonal patterns of
predation on juvenile Pacific salmon by anadromous
cutthroat trout in Puget Sound. Trans Am Fish Soc
137:165–181
Feltham MJ (1990) The diet of Red-Breasted Mergansers
(Mergus Serrator) during the smolt run in NE Scotland–
the importance of salmon (Salmo Salar) smolts and parr. J
Zool 222:285–292
Feltham MJ (1995) Consumption of Atlantic salmon smolts and
parr by goosanders: estimates from doubly-labelled water
measurements of captive birds released on two Scottish
rivers. J Fish Biol 46:273–281
Fraser NHC, Metcalfe NB, Thorpe JE (1993) Temperaturedependent switch between diurnal and nocturnal foraging
in salmon. Proc R Soc Lond B 252:135–139
Godin JGJ (1981) Circadian-Rhythm of swimming activity in
juvenile Pink Salmon (Oncorhynchus Gorbuscha). Mar
Biol 64:341–349
Hansen LP, Jonsson B (1989) Salmon ranching experiments in
the River Imsa–effect of timing of Atlantic salmon (Salmo
salar) smolt migration on survival to adults. Aquaculture
82:367–373
Hedenstrom A (2008) Adaptations to migration in birds:

behavioural strategies, morphology and scaling effects.
Philos Trans R Soc B-Biol Sci 363:287–299


Environ Biol Fish (2011) 92:151–157
Heggenes J, Krog OMW, Lindås OR, Dokk JG, Bremnes T
(1993) Homeostatic behavioral-responses in a changing
environment–Brown Trout (Salmo trutta) become nocturnal during winter. J Anim Ecol 62:295–308
Hvidsten NA, Lund RA (1988) Predation on hatchery-reared
and wild smolts of Atlantic salmon, Salmo salar L., in the
estuary of River Orkla. J Fish Biol 33:121–126
Ibbotson AT, Beaumont WRC, Pinder A, Welton S, Ladle M
(2006) Diel migration patterns of Atlantic salmon smolts
with particular reference to the absence of crepuscular
migration. Ecol Freshw Fish 15:544–551
Johnson JR, Knouft JH, Semlitsch RD (2007) Sex and seasonal
differences in the spatial terrestrial distribution of gray
treefrog (Hyla versicolor) populations. Biol Conserv
140:250–258
Jonsson B, Jonsson N, Brodtkorb E, Ingebrigtsen PJ (2001)
Life-history traits of Brown Trout vary with the size of
small streams. Funct Ecol 15:310–317
Jutila E, Jokikokko E (2008) Seasonal differences in smolt
traits and post-smolt survival of wild Atlantic salmon,
Salmo salar, migrating from a northern boreal river. Fish
Manage Ecol 15:1–9
Kemp PS, Gessel MH, Williams JG (2008) Response of
downstream migrant juvenile Pacific salmonids to accelerating flow and overhead cover. Hydrobiologia 609:205–217
Kennedy GJA, Greer JE (1988) Predation by cormorants,
Phalacrocorax carbo (L.), on the salmonid populations of

an Irish river. Aquac Fish Manage 19:159–170
Klemetsen A, Amundsen PA, Dempson JB, Jonsson B, Jonsson
N, O’Connell MF, Mortensen E (2003) Atlantic salmon
Salmo salar L., brown trout Salmo trutta L. and Arctic
charr Salvelinus alpinus (L.): a review of aspects of their
life histories. Ecol Freshw Fish 12:1–59
Larsson P-O (1985) Predation on migrating smolt as a
regulating factor in Baltic salmon, Salmo salar L,
populations. J Fish Biol 26:391–397
Mann RHK (1982) The annual food consumption and prey
preferences of pike (Esox lucius) in the River Frome,
Dorset. J Anim Ecol 51:81–95
Melnychuk MC, Welch DW, Walters CJ, Christensen V (2007)
Riverine and early ocean migration and mortality patterns
of juvenile steelhead trout (Oncorhynchus mykiss) from
the Cheakamus River, British Columbia. Hydrobiologia
582:55–65
Metcalfe NB, Fraser NHC, Burns MD (1998) State-dependent
shifts between nocturnal and diurnal activity in salmon.
Proc R Soc Lond B Biol Sci 265:1503–1507

157
Metcalfe NB, Fraser NHC, Burns MD (1999) Food availability
and the nocturnal vs. diurnal foraging trade-off in juvenile
salmon. J Anim Ecol 68:371–381
Moore A, Potter ECE, Milner NJ, Bamber S (1995) The
migratory behaviour of wild Atlantic salmon (Salmo salar)
smolts in the estuary of the River Conwy, North Wales.
Can J Fish Aquat Sci 52:1923–1935
Nicieza AG, Brana F (1993) Relationships among smolt size,

marine growth, and sea age at maturity of Atlantic Salmon
(Salmo salar) in Northern Spain. Can J Fish Aquat Sci
50:1632–1640
Pinder AC, Riley WD, Ibbotson AT, Beaumont WRC (2007)
Evidence for an autumn downstream migration and the
subsequent estuarine residence of 0 + year juvenile
Atlantic salmon Salmo salar L., in England. J Fish Biol
71:260–264
Reebs SG (2002) Plasticity of diel and circadian activity
rhythms in fishes. Rev Fish Biol Fish 12:349–371
Riley WD, Moore A, Russell IC, Davidson IC, Cove RJ, Ives
MJ (2007) Impact of trapping and tagging on the timing of
continued seaward migration of wild Atlantic salmon,
Salmo salar, smolts. Fish Manage Ecol 14:287–290
Roberts LJ, Taylor J, Gough PJ, Forman DW, de Leaniz CG
(2009) Night stocking facilitates nocturnal migration of
hatchery-reared Atlantic salmon, Salmo salar, smolts. Fish
Manage Ecol 16:10–13
Sillett TS, Holmes RT (2002) Variation in survivorship of a
migratory songbird throughout its annual cycle. J Anim
Ecol 71:296–308
Thorpe, J.E. and A. Moore. 1996. The migratory behaviour of
juvenile Atlantic salmon. pp. 39–46. In: H. Ueda & H.A.
Bern (ed.) Physiology and ecology of fish migration.
Proceedings of the International Symposium on Fish
Migration, Lake Toya, Hokkaido, Japan.
Thorpe JE, Metcalfe NB, Fraser NHC (1994) Temperature
dependence of switch between nocturnal and diurnal smolt
migration in Atlantic salmon. In: MacKinlay DD (ed)
High Performance Fish. International Fish Physiology

Symposium, Vancouver, pp 83–86
Welton JS, Beaumont WRC, Clarke RT (2002) The efficacy of
air, sound and acoustic bubble screens in deflecting
Atlantic salmon, Salmo salar L., smolts in the River
Frome, UK. Fish Manage Ecol 9:11–18
West CJ, Larkin PA (1987) Evidence for size-selective
mortality of juvenile sockeye-salmon (Oncorhynchus
nerka) in Babine Lake, British Columbia. Can J Fish
Aquat Sci 44:712–721


Environ Biol Fish (2011) 92:159–166
DOI 10.1007/s10641-011-9825-z

Environmental correlates of diet in the swordtail characin
(Corynopoma riisei, gill)
Niclas Kolm & Göran Arnqvist

Received: 8 March 2010 / Accepted: 4 April 2011 / Published online: 27 April 2011
# Springer Science+Business Media B.V. 2011

Abstract In the sexually dimorphic swordtail characin (Corynopoma riisei, Gill), males are equipped
with an opercular flag-ornament that has been
suggested to function as a food-mimic since females
bite at the ornament during courtship. However,
virtually nothing is known about the diet in wild
populations of this species. In this study, we first
investigated composition of and variation in the diet
of C. riisei across 18 different populations in
Trinidad, using gut content analyses. We then related

variation in gut content to habitat features of
populations to investigate the potential link between
environmental conditions and prey utilization. Our
results showed that the dominating food type in the
gut was various terrestrial invertebrates, both adults
and larvae, but we also document substantial variation
in prey types across populations. Furthermore, a
canonical correlation analysis revealed a relationship
between environmental characteristics and diet: populations from wider and more rapidly flowing streams
with more canopy cover tended to have a diet based
more on ants and mosquitoes while populations from
narrow and slow flowing streams with little canopy
cover tended to have a diet based more on springtails,
mites and mayfly larvae. Our results add novel
N. Kolm (*) : G. Arnqvist
Department of Ecology & Evolution, Animal Ecology,
Uppsala University,
Norbyv. 18 D,
75236, Uppsala, Sweden
e-mail:

information on the ecology of this interesting fish
and suggest the possibility of local adaptation reflecting differences in prey availability across natural
populations.
Keywords Sexual dimorphism . Gut content . Sensory
exploitation . Sensory drive . Local adaptation . Habitat

Introduction
Fishes show large variation in feeding habits both
between and within species (Helfman et al. 2009) and

local adaptation to different food types has been
suggested to be an important generator of diversity in
this group. For instance, niche separation according to
different food sources is considered an important
factor behind the adaptive radiation of the East
African cichlids (Clabaut et al. 2007; GonzalezVoyer et al. 2009). Ecological factors, for instance
canopy coverage and stream velocity, can have strong
effect on prey type availability in fish populations
(Dussault and Kramer 1981; Grether et al. 2001).
Therefore, identification and quantification of ecological factors and their relationships to food availability
is an important step towards understanding the
evolutionary mechanisms behind local adaptations
and population diversity (Coyne and Orr 2004).
The swordtail characin (Corynopoma riisei) is a
member of the large family of tetras (Characidae:
subfamily Glandulocaudinae), and is endemic to


160

rivers and streams in Trinidad and northern Venezuela
(Weitzman and Menezes 1998). This species displays a
remarkable and unique form of sexual dimorphism:
males carry an opercular flag-shaped ornament on each
side, which is normally held inconspicuously close to
the body (Nelson 1964). Fertilization is internal in this
subfamily and the male erects and displays this
opercular flag during courtship in front of the female
who bites at it prior to actual mating (Nelson 1964;
Amcoff et al. 2009). Due to this peculiar behavior,

Wickler (1968) hypothesized that the opercular flag is
mimicking the natural prey of the female swordtail
characin, which he believed to be aquatic invertebrates
such as Daphnia or other “water fleas” (see Wickler
1968). If this ornament indeed is a functional foodmimick, this would support a mechanism of ‘sensory
exploitation’. Under such a scenario, the opercular flag
has evolved to exploit female tendencies to be attracted
to objects that look like prey items (Arnqvist and
Rowe 2005; Amcoff and Kolm in review). However,
data on the actual diet of females across natural
populations is clearly needed to shed light on whether
there is scope for this ornament to have evolved as a
mimick of Daphnia or some other food item in the
natural diet of females.
Wickler’s (1968) hypothesis regarding the natural
prey of C. riisei is concordant with a more recent
general description of the Trinidadian tetras by Kenny
(1995), which suggests they are “omnivorous or
macrophagous, surface, mid-water or bottom
feeders”. To date, however, there is unfortunately no
detailed published information on the feeding ecology
or diet of C. riisei and it is not known whether there is
variation in available prey types or in food utilization
across populations. In the present study, our aim was
to investigate variation both in gut content of female
C. riisei and in ecological variables (stream width,
water depth, current velocity, level of canopy cover
and water turbidity) across 18 natural populations in
Trinidad. We then tested for a relationship between
food types and environmental factors to investigate

the potential for local adaptation to certain food types
in this species.

Environ Biol Fish (2011) 92:159–166

16
18

1

17

2
3
15

N
10
9

13

8

10 km

14

12
11

4
5

6

7

Fig. 1 Map over the major rivers and streams on Trinidad
(after Kenny 1995). Filled and numbered circles represent the
locations of the 18 populations sampled

Table 1) across the entire distribution of the species
in Trinidad (Kenny 1995). Specimens were caught
using a two-person push seine. Since C. riisei does
not occur in mountain streams (Kenny 1995), the sites
were located in the lowland or foothills and were
chosen to represent a range of different habitat types.
We aimed at collecting 10 sexually mature females at
each site (note that for two populations we only
managed to collect 9 females making the total sample
n=178). Whole fish were immediately preserved in
95% alcohol for subsequent analyses of gut content.
At each site, we also collected the following environmental variables: mean stream width, mean water
depth, current velocity, the amount of canopy cover
and water turbidity. Stream width and stream depth
was measured in meters and the other variables were
quantified independently by two observers, using a
ranking system (Table 1).
The stomach content of 178 females from the 18

populations was carefully inspected under a stereo
microscope, and all food/prey items were identified
with appropriate taxonomic resolution (i.e. to the level
of class, order, suborder or family) and counted. The
diet was then analyzed using data on the numerical
proportion of each prey type in each gut.
Statistical analyses

Materials and methods
In May, 2005, we collected samples of C. riisei from
18 different populations in 8 drainages (Fig. 1,

Due to the multitude of prey types recorded (i.e.,
variables), we used a combination of principal
component analysis and canonical correlation analysis
to maximize the power of the analyses while limiting


Environ Biol Fish (2011) 92:159–166

161

Table 1 Variation in environmental variables across populations. Current velocity was ranked on a scale from 1–3, canopy cover was
ranked on a scale from 0–2 and turbidity was deemed either clear (0) or turbid (1)
Population: river

Width (m)

Depth (m)

Current velocity

Canopy cover

Turbidity

1: Turure River

8.0

0.5

3

2

0

2: Tributary of Oropuche River

1.5

1.0

1

0

1

3: Caigual River

2.0

1.0

2

1

1

4: Tributary of Cunapo River

1.5

1.0

2

2

1

5: Coora River

2.0

0.5

3

1

1

6: Curamata River

4.0

2.0

1

0

1

7: Tributary of Inniss River

1.5

0.5

1

0

1

8: Guaracara River

2.0

1.0

2

2

1

9: Tributary of Guaracara River

2.5

0.3

2

1

0

10: Savonetta River

1.5

0.5

1

2

0

11: Poole River

8.0

2.0

2

1

1

12: Navet River

2.0

0.3

2

1

1

13: Tributary of Bois Neuf River

1.5

0.3

2

0

0

14: Canque of Nariva River

4.0

2.0

2

1

0

15: Sangre Grande River

2.0

1.0

2

1

0

16: Aripo River

4.0

1.0

2

1

0

17: El Mano River

2.5

1.0

2

1

0

18: Guanapo River

3.0

1.5

2

1

0

the number of inferential tests performed (Quinn and
Keough 2002). Firstly, we present descriptive data
both on proportions of all food types from the gutcontent analysis and on only the items that made
up >1% of the diet across populations. We also tested
for variation in food composition across populations
using Manova with all food items that made up >1%
of the diet as dependent variables and population as
our factor. Secondly, for the analysis of covariation
between mean diet and ecological variables across
populations, it was necessary to first reduce both the
environmental variables and the gut-content data
(mean per population) using principal component
analyses. For ecological variables, we retained the
first two principal components, which were the only
factors with eigenvalues larger than one (Jackson
1993). Together, these two factors accounted for 64%
of the total variation in the environmental variables
across populations. The loadings (i.e., the correlations

with the original variables) were as follows (loadings on
PC1 and PC2 within brackets): mean stream width
(0.62, 0.67), mean water depth (−0.03, 0.92), current
velocity (0.81, −0.15), canopy cover (0.72, −0.21) and
water turbidity (−0.46, 0.22). This means that postitive

values on PC1 describes shallow, fast flowing, clear
streams with substantial canopy cover while PC2
describes broad, deep, slow flowing, more turbid
streams with less canopy cover. For diet, we retained
the first six principal components from a PCA based on
the covariance matrix, which together accounted for
>95% of the variation in gut contents (see Table 2 for
the loadings of these principal components on the
original variables). In the final step of the analysis, we
used canonical correlation analyses to test for a
relationship between the two principal components of
the ecological variables and the five principal components describing variation in diet. All analyses were
run using Statistica 8.0 (StatSoft, Inc 2008).

Results
Each stomach contained on average 13.1 (SE=0.9)
distinct food items. Our data thus provides a
reasonably rich foundation for a description of the
diet of C. riisei. We found in total 41 different food
types in the gut content across populations (see
Table 3). When we only considered food types that


162

Table 2 Correlations (factor loadings) between principal components of gut
content (proportional data
on the 15 most common
items) and the original variables. Correlations >0.6 are
highlighted in bold

Environ Biol Fish (2011) 92:159–166
Variable

PC1

Trichoptera (larvae)

−0.19

0.19

−0.35

0.04

0.10

−0.14

Nematocera (larvae)

−0.32

−0.81

0.47

Nematocera (pupae)

−0.40

−0.24

−0.35

Coleoptera (larvae)

−0.55

−0.30

−0.32

0.11

−0.06

0.42

0.17

−0.16

0.10

0.04

−0.50

0.02

Ephemeroptera (larvae)

Fish eggs

PC2

PC3

PC4

PC5

PC6

0.20

−0.36

−0.40

−0.23

−0.76

−0.43

0.03

0.10

−0.08

−0.19

0.43

−0.23

Acari

−0.65

−0.02

0.01

0.14

−0.02

0.57

Collembola

−0.65

−0.04

−0.67

0.29

0.16

−0.13

Formicidae

0.99

−0.03

−0.03

0.08

0.05

−0.01

Hymenoptera

−0.06

0.15

−0.14

−0.01

−0.27

0.09

Nematocera (adults)

−0.33

0.65

0.61

0.29

0.08

−0.06

Homoptera

0.11

−0.10

−0.17

−0.09

0.04

0.17

Heteroptera

−0.05

0.13

−0.29

−0.58

0.19

0.05

Coleoptera (adults)

−0.04

0.47

−0.00

−0.76

0.41

−0.09

0.11

0.15

0.07

−0.38

−0.50

−0.27

Plant seeds

made up >1% of the average proportion of gut content
across populations, 15 remained. Out of these,
Formicidae (ants) strongly dominated the gut content
(>45%, Table 3) while Nematoceran larvae
(a Dipteran suborder, including mosquitoes and
midges), Coleoptera (beetles) and Collembola
(springtails) all contributed with >5% of the average
gut content across populations (Table 3). Other
invertebrates and their juvenile life stages (e.g.,

aquatic larvae and pupae), but also unidentified plant
seeds and fish eggs, made up the remainder of the diet
among these 15 most common food types (Table 3).
We found substantial variation in food composition
across populations (Fig. 2) (Manova, Wilk’s λ=0.003,
p<0.001). Out of the invertebrates found in the gut,
terrestrial invertebrates dominated the diet (mean% of
gut content (±SE): terrestrial groups: 77% (±0.03);
aquatic groups: 23% (±0.03); t-test of the null
hypothesis of a mean proportion of 50%: t=8.7,
n=18, p<0.0001).
The sampled populations displayed substantial
variation in environmental features, ranging from
turbid, slow flowing, and narrow streams with little
canopy cover to clear, fast flowing, and wide
streams with a well developed canopy cover
(Table 1; Fig. 3). As shown in the ordination plot
of the relationship between PC1 and PC2 of the
environmental variables (Fig. 3), there was some
grouping of populations into three clusters along

PC1 but a more continuous variation among populations along PC2. Based on the factor loadings,
populations 2, 6 and 7 were from slow flowing,
narrow streams with little canopy cover while
population 1 were from a rapid flowing, wide stream
with plenty of canopy cover. The remaining majority
of populations were all relatively intermediate in
terms of PC1 but more variable in PC2 (the latter
being related primarily to stream depth).
The canonical correlation analysis between the set

of six principal components of gut content variation
and the two principal components of environmental
variation showed a significant correlation between
food types and environment across populations
(canonical r=0.90, χ2 =24.7, d.f.=12, p=0.016). This
relationship was caused by the first pair (i.e. the first
root) of canonical variables since its removal produced a non-significant model (canonical correlation
with first pair of canonical variables removed:
canonical r=0.54, χ2 =4.4, d.f.=5, p=0.50). Inspection of the canonical loadings on this first pair of
canonical variables revealed that this pattern of
covariation was mainly the result of females from
rapid flowing, wide streams with more canopy cover
having a diet more dominated by Formicidae (ants)
and adult Nematocera (mosquitos and midges) but
with relatively few Acari (mites), Collembola (springtails) and Ephemeroptera larvae (mayflies). Hence,
females from wider streams with stronger currents


Environ Biol Fish (2011) 92:159–166
Table 3 Diet composition
and the environmental
origin of food items (aquatic
or terrestrial) across the
surveyed 18 populations

163

Food type

Average proportion (%)

Formicidae

45.7

Environmental origin
Terrestrial

Nematocera (larvae)

8.7

Aquatic

Coleoptera (adult)

7.1

Terrestrial

Collembola

6.0

Terrestrial

Coleoptera (larvae)

4.4

Aquatic

Acari

3.7

Aquatic

Heteroptera

2.3

Terrestrial

Nematocera (adult)

2.3

Terrestrial

Ephemeroptera (larvae)

2.0

Aquatic

Hymenoptera

1.7

Terrestrial

Trichoptera (larvae)

1.7

Aquatic

Homoptera

1.6

Terrestrial

Nematocera (pupae)

1.5

Aquatic

Plant seeds

1.4

–

Fish eggs

1.3

Aquatic

Brachycera/Cyclorrhapha (adult)

0.8

Terrestrial

Brachycera/Cyclorrhapha (larvae)

0.8

Terrestrial

Coleoptera (larvae)

0.8

Terrestrial

Thysanoptera

0.7

Terrestrial

Araneae

0.7

Terrestrial

Psocoptera

0.6

Terrestrial

Mallophaga

0.5

Terrestrial

Isopoda

0.5

Aquatic

Neuroptera (larvae)

0.4

Aquatic

Gastropoda

0.4

Aquatic

Simulidae (larvae)

0.4

Aquatic

Pseudoscorpionida

0.2

Terrestrial

Copepoda

0.2

Aquatic

Eggs (indeterminate)

0.2

–

Brachycera/Cyclorrhapha (larvae)

0.2

Aquatic

Chilopoda

0.2

Terrestrial

Coleoptera (adult)

0.1

Aquatic

Odonata (larvae)

0.1

Aquatic

Lepidoptera (larvae)

0.1

Terrestrial

Ostracoda

0.1

Aquatic

Lepidoptera (adult)

0.1

Terrestrial

Corixidae

0.1

Aquatic

Orthoptera

0.1

Terrestrial

Dermaptera

0.1

Terrestrial

Isoptera

<0.1

Terrestrial

Trichoptera (adult)

<0.1

Terrestrial

and greater canopy cover tended to eat more adult
terrestrial insects while females from slower flowing,
deeper and narrow streams with less canopy cover
tended to eat more juvenile and aquatic invertebrates.

Discussion
Our study is the first detailed account of the diet of
C. riisei, and it is clear from our results that terrestrial


164

Environ Biol Fish (2011) 92:159–166

Fig. 2 Visual representation
of variation in gut content of
food types that contributed
individually to >1% of the
total average gut content
across the sampled populations. The sum of the
remaining food types are
represented as ‘sum of other’.

See Table 1 for description of
the names and ecological
characteristics of each
population

100 %

Formicidae
Nematocera (larvae)

90 %

Proportion of total gut content

Collembola

80 %

Acari
Nematocera (adult)

70 %

Coleoptera (adult)

60 %

Ephemeroptera (larvae)
Coleoptera (larvae)

50 %

Heteroptera

40 %

Homoptera
Hymenoptera

30 %

Plant seeds

20 %

Fish eggs

10 %

Nematocera (pupae)
Trichoptera (larvae)

0%
1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18

Sum of other

Population

insects, particularly ants, dominate the diet of this
interesting fish. We also found substantial variation in
environmental variables as well as a clear link
between environmental variables and mean diet
across populations.
The aquatic invertebrates that dominate the fauna
of these streams, such as non-biting midges (Chironomidae) and mayflies (Ephemeroptera) (e.g., Turner
et al. 2008), did not dominate the diet of C. riisei.
3
2
2
1
PC1 (35.7 %)

7

6

13
12
10
4
15
8
17
9
18
5
16
3

0
14

11
-1
-2
-3
-4
-4

1
-3

-2

-1

0

1

2

3

PC2 (28.0 %)

Fig. 3 Ordination plot of correlation between PC1 and PC2 of
environmental variables. Populations group in three different
clusters along PC1 while they form a more continuous
distribution along PC2. Postitive values on PC1 describes
shallow, fast flowing and clear streams with substantial canopy
cover while PC2 describes broad, deep, slow flowing and more
turbid streams with less canopy cover (see text for details).
Values within brackets represent the proportion of total
variation among the environmental variables described by each
principal component

Instead, terrestrial insects formed the majority of the
gut content and there was little support for Wickler’s
(1968) hypothesis that aquatic invertebrates such as
Daphnia or other water fleas are important food

sources in this species. This is in fact unsurprising,
since C. riisei rarely if ever occurs in stagnant waters
(Kenny 1995; Table 1) where cladocerans are common (Viroux 2002). Together with the limited
representation of aquatic invertebrates in the diet, this
shows that C. riisei is mainly a surface-feeder. This is
further supported by a recent analysis of body
morphology in C. riisei which showed this species
share many of the characteristics normally associated
with surface feeding (e.g. an upwards pointing mouth)
although substantial variation in body morphology
was also found among populations (Arnqvist and
Kolm 2010).
We note that gut content analysis, as a basis for the
estimation of diet, can suffer from biases (Hyslop
1980). Most importantly, evacuation times of different
food items can vary substantially and be affected by
the amount of other food items (Rindorf and Lewy
2004). Our chosen method of gut content analysis, the
numerical method, has been suggested to be suitable
when food items are of similar size and when food
items occur in discrete units (Hyslop 1980). These
preconditions are fulfilled in C. riisei. Moreover,
sampling during the period of peak feeding has been
suggested to minimize bias since even prey items with
short evacuation times will thus remain present in the
gut content (Hyslop 1980). We sampled only during
day-time and since C. riisei is a diurnal feeder
(Nelson 1964) it is unlikely that any particular food
items were consistently left undetected. Finally, to test

Environ Biol Fish (2011) 92:159–166

whether the documented differences in diet across
different populations were caused simply by differences in the relative evacuation times of the utilized
prey, we correlated the average number of food items
per stomach to the mean scores of the principal
components of proportional gut content variation
across populations. However, these quantitative and
qualitative measures of gut content were not significantly correlated (Pearson correlations: p>0.24 in all
cases).
Even though this study forms only a snap-shot of
the diet during a limited temporal period, several
differences are apparent between C. riisei and other
species that occur in similar habitats in Trinidad. For
instance, the well studied tetra, Astyanax bimaculatus,
is omnivorous, feeding on terrestrial invertebrates,
aquatic invertebrates (including zooplankton) but also
on plant material and algae (Esteves 1996). Further,
the guppy, Poecilia reticulata, feeds mainly on
benthic algae and aquatic insect larvae (Dussault and
Kramer 1981). The diets of these species thus contrast
with the diet of C. riisei, which was dominated by
terrestrial and aquatic insects while it contained
virtually no vegetable matter, algae or zooplankton.
Hence, in comparison to these other common species
in similar environments in Trinidad, C. riisei appears
to occupy a partly distinct food niche.
Environmental features varied across the sampled
populations, suggesting that the habitat requirements

of C. riisei are not narrow (Fig. 3). Our results thus
support previous views of C. riisei being common
throughout Trinidad except for more brackish waters
and mountain streams with high velocity (Nelson
1964; Kenny 1995). The observed patterns of more
terrestrial insects, mainly ants, in populations with
more canopy cover and stronger currents and more
larvae-shaped invertebrates in slower flowing streams
with less canopy cover, point towards stable ecological differences that could set the stage for local
adaptation to habitat-specific prey-types in C. riisei
(Loreau 2000). We note that canopy-dwelling ants
frequently jump or fall from branches and leaves
(Yanoviak and Dudley 2006), resulting in a substantial “terrestrial drift” of ants in neotropical environments (Longino and Colwell 1997; Yanoviak et al.
2005). Apparently, C. riisei has capitalized on this
fact.
In light of these results we can make some
interesting inferences regarding the evolution of the

165

remarkable sexual dimorphism and courtship behaviors in C. riisei. Firstly, we reject the hypothesis that
the male opercular flag ornament is a water flea
mimic (Wickler 1968), since cladocerans are exceedingly rare both in the typical lotic environment of
C. riisei and in the actual diet. Instead, our data
suggests that the strong feeding response of females
towards the opercular flag during courtship (Amcoff
et al. 2009) occurs because the ornament is mimicking either an ant or some more larval-shaped
invertebrate. Secondly, the environmental influence
of diet in female C. riisei suggests, in addition to the
possibility of local adaptation to food types, that there

is also opportunity for the shape of the opercular flag
ornament to evolve to match the population-specific
diets of females. If female diet is temporally stable,
population-specific feeding preferences could cause
divergence among populations in male ornament
characteristics and even lead to speciation. This is
an important avenue for future research on this
fascinating fish.
Acknowledgements We thank Denise Dowlath, Trina Halfhide and Devon Ramoo for assistance in the field. Mary
Alkins-Koo and Dawn Phillip kindly provided advice
regarding field collections and Stan Weitzman generously
shared his knowledge of the swordtail characin. The field
collections were undertaken with permission from the
Wildlife Section of the Forestry Division of Trinidad. This
study was funded through grants from the Swedish Research
Council.

References
Amcoff M, Kolm N (in review) An experimental test of sensory
exploitation in the swordtail characin (Corynopoma riisei).
Ethology
Amcoff M, Arnqvist G, Kolm N (2009) Courtship signalling
with a labile bilateral signal: males show their best side.
Behav Ecol Sociobiol 63:1717–1725
Arnqvist G, Kolm N (2010) Population differentiation in the
swordtail characin (Corynopoma riisei): a role for sensory
drive? J Evol Biol 23:1907–1918
Arnqvist G, Rowe L (2005) Sexual conflict. Princeton
University Press, Princeton
Clabaut C, Bunje P, Salzburger W, Meyer A (2007) Geometric

morphometric analyses provide evidence for the adaptive
character of the Tanganyikan cichlid fish radiations.
Evolution 61:560–578
Coyne J, Orr A (2004) Speciation. Sinauer, Sunderland
Dussault G, Kramer D (1981) Food and feeding behavior of the
guppy, Poecilia reticulata (Pisces: Poeciliidae). Can J Zool
59:684–701