Environmental biology of fishes, tập 90, số 1, 2011

Environ Biol Fish (2011) 90:1–17
DOI 10.1007/s10641-010-9706-x

Allochthonous and autochthonous carbon sources for fish
in floodplain lagoons of an Australian dryland river
Elvio S. F. Medeiros & Angela H. Arthington

Received: 14 January 2010 / Accepted: 5 August 2010 / Published online: 10 September 2010
# Springer Science+Business Media B.V. 2010

Abstract Dryland rivers associated with arid and
semi-arid land areas offer an opportunity to explore
food web concepts and models of energy sources in
systems that experience unpredictable flooding and
long dry spells. This study investigated the sources of
energy supporting three species of fish feeding at
different trophic levels within floodplain lagoons of
the Macintyre River in the headwaters of the MurrayDarling river system, Australia. Stable isotope analyses
revealed that fish consumers derived, on average,
46.9% of their biomass from zooplankton, 38.1% from
Coarse Particulate Organic Matter (CPOM) and 24.0%
from algae. Ambassis agassizii derived on average
57.6% of its biomass carbon from zooplankton and
20.4–27.8% from algae or CPOM. Leiopotherapon
unicolor derived most of its carbon from zooplankton

and CPOM (38.3–39.5%), with relatively high contributions from algae compared to the other species
(33.3%). An average of 48.4% of the biomass of
Nematalosa erebi was derived from zooplankton, with
CPOM contributing another 38.1%. Zooplankton was

the most important source of organic carbon
supporting all three fish species in floodplain
lagoons. Phytoplankton, and possibly, particulate
organic matter in the seston, are the most likely
energy sources for the planktonic suspension
feeders (zooplankton) and, consequently, the fish
that feed on them. These results indicate a stronger
dependence of consumers on autochthonous sources and on locally produced organic matter from
the riparian zone (i.e., the Riverine Productivity
Model), than on other resources.
Keywords Fish . Food web . Stable isotopes . Riverine
Productivity Model

E. S. F. Medeiros : A. H. Arthington
Australian Rivers Institute and eWater Cooperative
Research Centre, Griffith University,
Nathan QLD 4111, Australia
Present Address:
E. S. F. Medeiros (*)
Centro de Ciências Biológicas e Sociais Aplicadas,
Universidade Estadual da Paraíba—UEPB,
Campus V. Av. Monsenhor Walfredo Leal, no. 487,
Tambiá CEP 58020-540 João Pessoa—PB, Brazil
e-mail:

Introduction
Food webs in freshwaters can be supported from bottom
detritus and/or primary production in the pelagic zone.
In many river systems, the detrital food chain is
considered more important, based largely on the

delivery of allochthonous materials from headwaters
and/or the decomposition of aquatic macrophytes (the
River Continuum Concept, RCC; Vannote et al. 1980).


2

This model argues that downstream communities are
adapted to capitalize on upstream processing inefficiencies (leakage), therefore, emphasizing the influence of
nutrients and organic matter from upstream processes on
the structure and function of lowland reaches. However,
food webs in large floodplain rivers can be fueled by
terrestrial inputs and organic matter delivered laterally
from floodplains (the Flood Pulse Concept; Junk et al.
1989). During floods, aquatic organisms migrate to the
floodplain and exploit the newly available habitats and
their resources, whereas, as floodwaters recede, nutrients
and newly produced animal biomass are returned to the
main river channel. In both models, strong reliance on
allochthonous inputs has been emphasized. Other
studies suggest that autochthonous primary production
is an important, often major, contributor to metazoan
production in rivers (Thorp and Delong 2002; Bunn
et al. 2006). For example, studies on the Orinoco River
and its floodplain showed that phytoplankton and
periphyton were the main carbon sources for invertebrates and fish (Lewis et al. 2000; Lewis et al. 2001).
Similarly, Araujo-Lima et al. (1986) showed that
detritivorous fish in the Amazon River floodplain,
despite feeding mostly on detritus, derived most of their
carbon from phytoplankton production.

Nevertheless, these studies were developed in tropical
and temperate river systems with predictable flooding
patters and low flow variability (Puckridge et al. 1998).
Dryland rivers offer an opportunity to explore food web
concepts in systems that experience far less predictable
flooding patterns than mesic floodplain rivers. In the
arid regions of Australia, dryland rivers feature extensive floodplains and a network of anabranching tributaries that provide a vast terrestrial-water interface (Walker
et al. 1995; Bunn et al. 2003), but for most of the time
these systems exist as disconnected and highly turbid
waterholes and floodplain lagoons (Bunn and Davies
1999). During seasonal or extended dry periods these
isolated waterbodies act as refugia for obligate aquatic
organisms such as fish (Morton et al. 1995; Bunn and
Davies 1999; Arthington et al. 2005). When flooding is
resumed, fish left in isolated waterholes are able to
colonize new habitats and resources opened up on the
floodplain, and when floods recede, fish return again to
isolated waterholes. It is of interest to know how fish are
sustained in these remnant waterholes and what energy
sources support individuals that may recolonize the river
floodplain and channel network when flooding is
resumed (Winemiller 1996; Burford et al. 2008).

Environ Biol Fish (2011) 90:1–17

Fish consumers in Australian dryland floodplain rivers
use a range of trophic resources for nutrition, including
algae, aquatic invertebrates and zooplankton (Balcombe
et al. 2005). In Cooper Creek, a large arid-zone
floodplain river in western Queensland (Australia),

aquatic consumers have been shown to derive most of
their carbon from algae growing in a “bathtub ring”
around the shallow littoral zone of channel waterholes
and floodplain lagoons (Bunn and Davies 1999; Bunn
et al. 2003). This dependency on algae suggests the
Riverine Productivity Model (RPM, Thorp and Delong
1994) as a more relevant model for these dryland rivers
than the Flood Pulse Concept (FPC). The RPM highlights the importance of local in-stream production by
phytoplankton, benthic algae and/or aquatic plants and
suggests that the role of autochthonous sources has been
underestimated in previous models and assessments of
food webs in large river systems (Thorp and Delong
1994). An integrated model that synthesizes the above
mentioned views and their corollaries into an heuristic
approach for riverine ecosystems (the Riverine Ecosystem Synthesis, RES) is proposed by Thorp et al. (2006).
According to the RES, algal production is the primary
source of organic energy fueling aquatic metazoan food
webs in the floodplain of riverine systems during
flooding. Nevertheless, this model also recognizes the
importance of allochthonous organic matter to some
species and in seasons when interaction between
riparian vegetation and the aquatic environment is at
its maximum.
Like those of Cooper Creek, isolated waterbodies
on the floodplains of the Macintyre River in the
headwaters of the Murray-Darling river system are
highly turbid (Houldsworth 1995; Medeiros 2005).
The production of aquatic plants and algae in this
system is likely to be limited by low light availability
and by the absence of water flow and associated

nutrient pulses during dry periods. Given these
constraints on primary production in turbid waterbodies, we hypothesize that fish consumers in the
Macintyre River are dependent on allochthonous
organic carbon during dry periods, that is, carbon
derived from riparian vegetation. To test this hypothesis, we used stable isotope techniques to investigate
the importance of a range of energy sources to three
fish species feeding at different trophic levels, the
olive perchlet Ambassis agassizii Steindachner, 1867
(microcarnivore), spangled perch Leiopotherapon unicolor (Günther, 1859) (omnivore) and bony bream


Environ Biol Fish (2011) 90:1–17

Nematalosa erebi (Günther, 1868) (algivore/detritivore)
(Pusey et al. 2004; Medeiros and Arthington 2008a;
Medeiros and Arthington 2008b).
In this paper we document the types and origin of
energy sources and their relative importance in supporting the three species of fish in floodplain lagoons of the
Macintyre River. Specifically, we assess the importance
of allochthonous versus autochthonous carbon sources
in light of the current models proposed (RCC, FPC,
RPM) to explain energy flow in large river systems.

Materials and methods
Study area
This study was conducted on the floodplain of the
Macintyre River, a dryland river in the Border
Rivers catchment, located along the southern
Queensland and northern New South Wales border
and comprising a major portion of the headwaters

of the Barwon and Darling River systems
(McCosker 1996) (Fig. 1). In the study area, a
number of streams diverge from the Macintyre River
in the vicinity of the towns of Boggabilla and
Goondiwindi (DWR 1995) where the river passes
through a relatively well-defined floodplain containing numerous intermittent and semi-permanent billabongs on prior river channels (see McCosker 1996;
Medeiros 2005 for further details).
The study period (2002–2003) was relatively dry
compared to previous years, with average discharges of
the Macintyre River between 780 and 845 megaliters
per day (ML·day−1) (data from the Boggabilla gauging
station—416002). Major to moderate floods occurred
early and late in 2001, however only minor to
moderate floods were recorded for the study period,
with mean discharges of 14487 ML·day−1 on 31 March
2002 and 17412 ML·day−1 on 26 February 2003
(Fig. 2). Mean water levels rose up to 4 m during the
2002 flood and 4.6 m during the 2003 flood. Even
though such floods may cause inundation of low lying
areas adjacent to the main river channel, they were not
sufficient to inundate all study sites.
Study design
Seven study sites, including six floodplain lagoons
and one site in the main channel of the Macintyre

3

River (Fig. 1) were sampled on three occasions in
2002–2003. The first sampling event took place in the
dry season of 2002, that is, early in the summer of

that year (20–31 October). The second sampling
event occurred near the end of the 2002–2003
summer, soon after the wet season (10–20 March
2003), when some of the study sites experienced
minor flooding. The third sampling event occurred in
the winter of 2003 (15–25 July), during the dry
season. During the study period, two of the six
lagoons were flooded (South Callandoon East and
Rainbow lagoons, see Fig. 1) and one dried up
completely (Broomfield Lagoon). The remaining
sites, Serpentine, Punbougal and Maynes lagoons,
decreased in size and volume continuously throughout the study period but did not dry up completely
(Medeiros 2005).
Collection of primary sources and consumers
Major primary sources of terrestrial and aquatic
origin were collected from each study site on each
sampling occasion. Fallen leaves from major riparian
trees (mostly Eucalyptus spp.) were collected by hand
from the margins. Benthic detritus was collected with
dip and hand nets and wet-sieved in the field into
coarse (>1 mm to 1 cm) particulate organic matter
(CPOM).
Samples of algae (Rhizoclonium sp. and Cladophora sp.) were taken from the shallow littoral
margins both directly off the mud surface and from
submerged wood or rocks, and washed in the field to
remove any associated organic debris. Because of
the high levels of suspended sediment and the
presence of unidentified particulate matter in the
water, it was not possible to take clean samples of
phytoplankton. Zooplankton (calanoids and cladocerans) was sampled at dusk and dawn by towing a

250 μm plankton net just below the surface of the
water.
The three species of fish were sampled by hauling
a seine net along the littoral zone of the study sites.
Where possible, three replicate samples of each of the
potential food sources and consumers and five
replicate fish samples were collected from different
areas of habitat in each lagoon. All animal and plant
samples were immediately refrigerated, frozen within
four hours of collection and stored frozen prior to
processing for stable isotope analysis.


4

Environ Biol Fish (2011) 90:1–17
Border Rivers
catchment

QLD

NT

Brisbane

SA

NSW
Sydney

VIC

N

Broomfield Lg.
Goondiwindi
South Callandoon
Lg. East

Serpentine Lg.

Goondiwindi
Weir

River site

Boggabilla

Boggabilla
Weir

Rainbow Lg.
Morella Watercourse

Macintyre River

Punbougal Lg.
Maynes Lg.
5 km

Flow

Fig. 1 Location of the lagoons studied in the floodplain of the Macintyre River and the Border Rivers catchment within the MurrayDarling River system

Sample preparation
In the laboratory, primary carbon sources (tree leaves,
particulate organic matter and algae) were rinsed
in distilled water, and, in the case of algae, any
remaining organic debris was removed. All samples
were oven-dried at 60°C for 36 to 48 h and then
ground to a powder-like consistency in a ring grinder.
For each of the zooplankton samples collected, half
was treated in 10% HCl for approximately two hours
to remove carbonates from exoskeletons in preparation for δ13C analysis. The remaining half of the
sample was not treated in acid and was used for
δ15N analysis (after Bunn et al. 1995). Samples of
muscle tissue were taken from individuals of each

fish species, from the region above the lateral line
and adjacent to the dorsal fin (after the fish was
scaled and skinned). Stable isotope signatures were
analyzed from single individuals of the three study
species. All animal samples were oven-dried at 60°C
for 24 to 48 h and then ground by hand with a mortar
and pestle. Lipids were not removed from tissue
samples.
Dried and ground samples were oxidized at high
temperature and the resultant CO2 and N2 were
analyzed for percentage C, N and stable isotope ratios
using a continuous-flow isotope-ratio mass spectrometer (Europa Tracermass and Roboprep, Crewe,

England) at the Stable Isotope Analysis laboratory at
Griffith University. Ratios of 13C/12C and 15 N/14 N


Environ Biol Fish (2011) 90:1–17

5

90000
2001

2002

2003

80000

Discharge (ML/day)

70000
60000
30 Nov 01

50000
40000
30000
20000

26 Feb 03

31 Mar 02

10000

wet

dry

Jan/01
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan/02
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct

Nov
Dec
Jan/03
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec

0

late dry

Month
−1

Fig. 2 Daily discharge (ML·day ) in the Macintyre River (recorded at the Boggabilla gauging station—416002) between 2001 and
2003. Arrows indicate sampling occasions. Dates of flooding events during the study period are also indicated

were expressed as the relative per million (‰)
difference between the sample and the conventional
standards (PeeDee Belemnite carbonate and N2 in air)
as follows:

δ X(‰) =

Rsample − Rs tan dard
Rs tan dard

× 1000

where X¼ 13 C or 15 N and R¼13 C =12 C or 15 N=14 N:
Measurement precision of the mass spectrometer was
approximately 0.1‰ for 13C/12C and 0.3‰ for 15 N/14 N.
Data analysis
Data were first analyzed using the stable isotope values
of sources and consumers for each sampling occasion
and site. The significance of differences among sites and
sampling occasions were tested using simple ANOVA
(Zar 1999) followed by post-hoc multiple comparisons
using Tukey’s HSD test (significance level of 0.05).
Pearson correlations between δ13C and δ15N signatures
of the three species of fish were used to add weight to
mixing model results. The contribution of the more
abundant primary sources to the biomass of fish was
calculated using the two- and three-source linear
mixing models of Phillips and Gregg (2001). The

two-source mixing model was based on δ13C signatures of zooplankton and either organic matter
(CPOM) or algae, depending upon which was collected. Where all three major sources were available, the
three-source linear mixing model was based on δ13C
and δ15N signatures of zooplankton, algae and organic
matter (CPOM) (Phillips 2001; Phillips and Gregg
2001). For three-source mixing models, when the

isotopic signatures for the mixture fall outside the
region constrained by the source isotopic signatures,
negative source proportions and variances may result,
therefore yielding an unfeasible solution. In such cases,
the contribution of the more abundant primary sources
to the biomass of fish was calculated using the twosource linear mixing model of carbon signatures of the
two sources closest to the consumer. This adjustment is
mentioned where appropriate. Field studies generally find a carbon enrichment of 0 to 2‰ and 3 to
5‰ for nitrogen, for each increase in trophic level
(Peterson and Fry 1987). To account for fractionation, 2 to 2.5‰ per trophic level from the nitrogen
isotopic signature of the fish was added to δ15N
values of food sources, and for δ13C an adjustment
of 0.1–0.3‰ per trophic level was made (DeNiro
and Epstein 1981; France 1996; Vander Zanden and
Rasmussen 2001; Bunn et al. 2003). For the mixing


6

models only, algal δ13C and δ15N values for Rainbow
Lagoon in October 2002 were estimated from algae
values from South Callandoon Lagoon on the same
sampling occasion, as both sites had similar algal
composition and it was not possible to separate algae
from benthic detritus collected from Rainbow Lagoon
on that occasion.

Results
Primary sources
Algae were significantly more 13C enriched than all

other primary sources (ANOVA, df=2,153; F=65.3;
p<0.05), despite considerable variability in δ13C values
across sites and sampling occasions (from −10.9 to
−31.9‰) (Table 1). Even though algal δ15N values
were relatively low, with algae from most sites being
15
N depleted (3.5 to 11.0‰), isotope values of 15N
algae were significantly higher than values for both
CPOM and riparian vegetation (ANOVA, df=2,153;
F=11.3; p<0.05). Values of δ15N were not significantly
different between CPOM and riparian vegetation
(ANOVA, df=2,153; F=11.3; p=0.81) (Table 2). However, riparian tree leaves were 13C depleted (−27.4 to
−29.7‰) in comparison with coarse organic matter
(−23.8 to −28.8‰) (ANOVA, df=2,153; F=65.3; p=
0.04), suggesting a potential contribution of algal
carbon to CPOM (Table 1).
Significantly lower algae δ13C averages were observed on the sampling occasions after flooding
(ANOVA, df=2,28; F=13.4; p<0.05). Nevertheless,
data available for non-flooded sites also showed
depleted algae 13C values for the later sampling
occasions (namely July 2003) (t-test, df=4.5, t=3.1;
p=0.03). Values of δ15N were not significantly
different across sampling occasions for both flooded
(ANOVA, df=2,28; F=1.7; p=0.20) and non-flooded
sites (t-test, df=4.9, t=-2.0; p=0.10).
Spatial variation in isotope values of algae was
significant for both 13C (ANOVA, df=6,36; F=6.2;
p<0.05) and 15 N (ANOVA, df=6,36; F=17.4; p<0.05);
however, such variation was not associated with the
occurrence of flooding. Serpentine Lagoon and the

Macintyre River were significantly depleted in 13C
compared to Broomfield and Maynes lagoons, which
were consistently carbon enriched (Tukey post-hoc
p<0.05) (Table 1). Serpentine Lagoon had highly

Environ Biol Fish (2011) 90:1–17

enriched algal 15N values (14.6 to 19.8‰) and these
values were significantly higher than algal 15N values
for all other study sites (Tukey post-hoc p<0.05).
Rainbow Lagoon also had enriched algal 15N values,
which were significantly higher than those from the
Macintyre River and South Callandoon Lagoon (Table 2).
Average isotope values of particulate organic matter
(CPOM) of flooded sites were very consistent across
sampling occasions with no significant differences for
both δ13C (ANOVA, df=2,23; F=0.9; p=0.40) and
δ15N (ANOVA, df=2,23; F=0.1; p=0.88). Nonflooded sites also showed no significant differences
in δ13C CPOM (ANOVA, df=2,27; F=0.7; p=0.50)
across sampling occasions, however average July 2003
δ15N values were significantly higher than values from
October 2002 (Tukey post-hoc p=0.02).
Spatial variation in isotope values of CPOM was
significant for 13C (ANOVA, df = 6,49; F = 10.5;
p<0.05). The flooded sites, Macintyre River and South
Callandoon Lagoon, were significantly 13C depleted
compared to the non-flooded sites of Broomfield,
Maynes and Punbougal lagoons (Tukey post-hoc
p<0.05). No significant variation in δ13C values was
observed between the later non-flooded sites (Tukey

post-hoc p=0.99). Nevertheless, the non-flooded Serpentine Lagoon was not significantly different in δ13C
values to any other flooded site (Tukey post-hoc
p>0.90). This site was also significantly 13C depleted
compared to the non-flooded sites, Maynes and
Punbougal Lagoon (Tukey post-hoc p<0.01). Furthermore, Rainbow Lagoon was significantly more 13C
depleted than Maynes and Punbougal Lagoons (Tukey
post-hoc p<0.02), but showed no significant differences to the other study sites (both flooded and nonflooded) (Tukey post-hoc p>0.32) (Table 1).
Values of δ15N CPOM showed high variation
across the study sites (ANOVA, df=6,49; F=25.5;
p<0.05). Similar to δ13C values, variations in 15 N
were not associated with flooding, indicating that
factors such as riparian vegetation composition across
study sites and decomposition levels of the CPOM
were more important in determining isotope signatures than the presence of flooding. The only nonsignificant differences in 15N observed were between
Broomfield Lagoon and the Macintyre River (Tukey
post-hoc p=0.87), Broomfield and Rainbow lagoons
(Tukey post-hoc p=1.00), Maynes and Punbougal
lagoons (Tukey post-hoc p = 0.55), Maynes and
South Callandoon lagoons (Tukey post-hoc p=1.00),


Environ Biol Fish (2011) 90:1–17

7

Table 1 Stable carbon isotope ratios (‰) of sources and
consumers from floodplain lagoons and one site on the
Macintyre River, on each of three sampling occasions, preSample type

South

Callandoon
Lagoon East

Primary sources
Riparian vegetation

Zooplankton

A. agassizii

L. unicolor

N. erebi

−13.7

−28.1

−29.7

−32.6

−28.7

–

–29.0

(± 0.5)

−28.4

(± 0.3)

(± 0.5)

(± 0.3)

−23.1

−28.1

−29.2

−36.5

–

(± 4.1)

(± 0.5)

(± 0.2)

(± 0.6)

−18.9

−28.4

−28.7

−37.3

−34.3

−32.5

−34.1

(± 0.7)

(± 0.7)

(± 0.3)

(± 0.1)

(± 0.2)

(± 1.1)

(± 0.3)

–

−27.5

−28.7

−32.9

−30.0

−27.0

−29.5

(± 0.1)

(± 0.3)

(± 0.3)

(± 0.3)

(± 0.5)

(± 1.8)

−26.1

−27.0

−29.2

−31.7

−28.2

−26.9

−28.0

(± 2.8)

(± 0.3)

(± 0.3)

(± 0.1)

(± 0.3)

−27.4

(± 1.2)

−17.6

−27.8

−29.0

−32.2

−29.1

−25.5

−29.6

(± 1.3)

(± 0.2)

(± 0.5)

(± 0.3)

(± 0.3)

−26.4

(± 1.9)

−15.5

−27.8

−28.0

–

−29.5

–

−27.1

(± 0.2)

(± 1.2)

(± 1.4)

−31.9

−28.8

−29.6

(± 0.4)

(± 0.3)

(± 0.6)

−27.9

−28.6

−28.9

(± 1.4)

(± 0.3)

(± 0.2)

Oct 2002

–

−25.3

−28.6

(± 0.6)

(± 0.3)

Mar 2003

–

−26.8

−28.5

(± 0.2)

(± 0.3)

−13.9

−24.4

−28.6

−15.2

(± 0.1)

(± 0.4)

Oct 2002
Mar 2003

Oct 2002

Jul 2003
Oct 2002
Mar 2003
Jul 2003
Punbougal
Lagoon

Jul 2003
Maynes
Lagoon

Oct 2002
Mar 2003

Serpentine
Lagoon

−27.6

−28.4

(± 2.5)

(± 1.2)

–

−23.8

(± 0.9)

−29.7
–

–29.9

–

(± 1.0)
−24.8

−25.7

−27.5

(± 0.9)
−26.0

–

–

–

–

–

–

–

–

–

–

–

–

(± 1.3)
−26.7
(± 0.8)
−24.2
(± 0.8)
−24.8
(± 0.4)

–

–

−27.7

−22.9

–

–

(± 0.3)

(± 0.2)

(± 0.2)

−24.1

−27.4

−27.3

–

–

−21.2

(± 0.1)

(± 0.6)

(± 0.2)

−26.7

−29.2

−25.9

−23.8

−22.3

−22.2

(± 1.9)

(± 1.1)

(± 0.2)

−28.4

−29.6

−19.1

(± 0.1)

(± 0.1)

(± 0.1)

−28.1

−28.0

−27.9

−29.3

(± 0.5)

(± 0.2)

(± 0.2)

(± 0.3)

−10.9

−25.7

−29.1

–

(± 0.2)

(± 0.1)

(± 0.4)

Oct 2002

−17.0

Oct 2002

(± 0.5)
–

–

–

Jul 2003

a

−11.6
(± 0.5)

Jul 2003

Mar 2003

Broomfield
Lagoona

Fish

CPOM

Mar 2003

Macintyre
River

Primary consumer

Algae

Jul 2003
Rainbow
Lagoon

flood (October 2002), post-flood summer (March 2003) and
post-flood winter (July 2003). Data correspond to mean values
(±s.d., for n=3–7 samples) or individual values (where n<3)

–

Broomfield Lagoon dried out after the first sampling occasion

−22.6
(± 1.5)

−18.3
(± 1.7)

(± 0.4)
−19.6

−21.3

−17.1

(± 0.6)

(± 0.3)

–

–

−17.9

–

–

−23.4

(± 0.2)
(± 0.8)

8

Environ Biol Fish (2011) 90:1–17

Table 2 Stable nitrogen isotope ratios (‰) of sources and
consumers from floodplain lagoons and one site on the
Macintyre River, on each of three sampling occasions, preSample type

South
Callandoon
Lagoon East

Primary sources

Macintyre
River

Riparian vegetation

Zooplankton

A. agassizii

L. unicolor

N. erebi

7.6

6.5

3.8

22.3

14.5

–

13.2

(± 0.5)

7.8

(± 0.8)

(± 0.1)

(± 0.9)

3.5

7.8

7.0

12.3

–

–

13.5

(± 1.5)

(± 0.4)

(± 4.3)

(± 0.1)

8.0

6.5

5.6

7.8

16.0

15.5

(± 1.9)

(± 0.8)

(± 0.5)

(± 0.1)

(± 0.2)

(± 0.3)

(± 0.5)

Oct 2002

–

3.8

5.9

16.2

16.5

15.4

14.9

(± 0.1)

(± 0.8)

(± 0.2)

(± 0.6)

(± 0.7)

(± 2.1)

Mar 2003

8.5

2.8

4.8

12.8

15.2

16.1

12.6

(± 2.4)

(± 0.4)

(± 1.4)

(± 0.1)

(± 0.2)

16.5

(± 0.9)

Jul 2003

11.0

5.1

6.3

14.5

15.2

12.6

13.0

(± 1.1)

(± 0.1)

(± 0.4)

(± 0.3)

(± 0.5)

13.0

(± 1.4)

7.5

4.9

4.7

–

12.7

–

9.7

(± 1.0)

(± 0.7)

(± 0.1)

Oct 2002
Mar 2003

Oct 2002

Jul 2003

Maynes
Lagoon

Serpentine
Lagoon

a

6.4

4.4

5.5

(± 0.4)

(± 0.3)

(± 1.5)

4.4

4.5

6.4

(± 3.6)

(± 0.1)

(± 0.1)

Oct 2002

–

6.5

5.5

(± 0.5)

(± 0.5)

Mar 2003

–

6.4

5.4

(± 0.1)

(± 0.5)

Jul 2003

7.4

6.8

8.1

7.6

(± 0.2)

(± 0.2)

Oct 2002

–

(± 1.2)

–

15.4

(± 1.1)
12.4

9.2

14.8

(± 0.8)
10.1

–

–

–

–

–

–

12.1

–

–

–

11.5

–

–

–

10.6

(± 1.1)
(± 1.8)
(± 0.8)
(± 1.3)
–

–

–

14.0

7.8

–

–

9.3

8.2
(± 0.4)

–

–

11.2

5.3
(± 0.6)

14.3
(± 0.2)

16.6

18.2
(± 0.6)

16.8

5.2

5.5

9.8

16.8

17.9

12.4

(± 0.1)

(± 0.7)

(± 0.1)

19.8

6.4

5.3

21.9

(± 1.4)

(± 0.3)

(± 0.5)

(± 0.4)
–

6.6

6.5

8.1

(± 1.5)

(± 1.1)

Mar 2003

–

7.1

7.6

(± 0.4)

(± 0.6)

(± 0.1)

Jul 2003

–

8.2
(± 0.2)

8.3
(± 0.4)

Oct 2002

14.6

6.1

(± 0.2)

Mar 2003

–

Oct 2002

(± 1.4)

13.3

(± 0.4)

Jul 2003
Broomfield
Lagoona

Fish

CPOM

Mar 2003

Punbougal
Lagoon

Primary consumer

Algae

Jul 2003
Rainbow
Lagoon

flood (October 2002), post-flood summer (March 2003) and
post-flood winter (July 2003). Data correspond to mean values
(±s.d., for n=3–7 samples) or individual values (where n<3)

8.2

4.0

6.3

(± 0.1)

(± 1.1)

(± 0.1)

Broomfield Lagoon dried out after the first sampling occasion

(± 1.4)
(± 1.7)

–

(± 0.4)

(± 0.6)

–

15.0
(± 0.3)

–

–

13.8
(± 0.1)


Environ Biol Fish (2011) 90:1–17

Punbougal and South Callandoon lagoons (Tukey
post-hoc p=0.81), Punbougal and Serpentine lagoons
(Tukey post-hoc p=0.44), and the Macintyre River
and Rainbow Lagoon (Tukey post-hoc p = 0.51)
(Table 2). Such results also support the idea of a
weak influence of flooding, since no differences in
15
N CPOM were observed across some flooded and
non-flooded sites.
Variation in isotope values of riparian vegetation of
flooded sites was not significant for both δ13C
(ANOVA, df = 2,24; F = 1.6; p = 0.22) and δ 15 N
(ANOVA, df=2,24; F=1.5; p=0.23). Non-flooded

sites also did not show significant variation in δ15N
for riparian vegetation (ANOVA, df=2,27; F=1.8; p=
0.18). However, isotope values of δ13C were significantly lower in October 2002 than in July 2003
(Tukey post-hoc p = 0.04) for non-flooded sites.
Spatial variation in δ13C was significant (ANOVA,
df=6,50; F=3.4; p<0.05), with Maynes Lagoon being
significantly enriched compared to South Callandoon,
Serpentine and Rainbow lagoons (Tukey post-hoc
p<0.04) (Table 1). Similar results were observed
across sites for δ15N values for riparian vegetation
(ANOVA, df=6,50; F=4.1; p<0.05), where Maynes
Lagoon was significantly enriched compared to South
Callandoon, Serpentine and Rainbow lagoons, and the
Macintyre River (Tukey post-hoc p<0.01) (Table 2).
Results for riparian vegetation corroborate the overall
trend for the sources studied, which indicate the
absence of effects of flooding on isotopic values for
both 13C and 15N. Since flooding was mild, waterhole
isotope values were driven by local characteristics
associated with riparian vegetation, algal composition
and previous flow history.
Zooplankton and fish consumers
Zooplankton δ13C values were temporally and spatially
variable (−19.1 to −37.3‰) yet consistently more 13C
depleted (−29.8±5.4‰) than most primary sources
(−25.8 ± 5.0‰) and consumers (−26.5 ± 4.2‰)
(ANOVA, df=6,358; F=25.9; p<0.05), with the
exception of riparian vegetation (Tukey post-hoc p=
0.89) and A. agassizii (Tukey post−hoc p=1.00)
(Table 1). Furthermore, δ15N values were significantly

enriched compared to the primary sources and depleted
compared to the omnivorous L. unicolor (ANOVA, df=
6,358; F=111.9; p<0.05) (Table 2). Values of δ15N for
zooplankton were not significantly different from the

9

micro-carnivorous A. agassizii (Tukey post-hoc p=0.07)
and the detritivorous N. erebi (Tukey post-hoc p=0.42).
Temporal comparisons revealed no significant differences in mean zooplankton δ13C values across sampling occasions for flooded sites (ANOVA, df=2,15;
F=1.2; p=0.32), whereas δ15N values were significantly higher in October 2002 at flooded sites
(ANOVA, df = 2,15; F = 13.4; p < 0.05). For nonflooded sites, March 2003 showed significantly
enriched zooplankton 13C (ANOVA, df=2,12; F=
34.9; p<0.05) compared to the other sampling occasions, whereas zooplankton δ15N values at non-flooded
sites were not significantly different across sampling
occasions (ANOVA, df=2,12; F=2.7; p=0.10).
Comparison of stable isotope values across sites
revealed that spatial variation in zooplankton values
was high with significant differences across sites for
both δ13C (ANOVA, df=3,29; F=29.3; p<0.05) and
δ15N (ANOVA, df=3,29; F=3.8; p=0.02). South
Callandoon and Rainbow lagoons, which were subject to flooding, were consistently 13C depleted
compared to the non-flooded Serpentine and Maynes
lagoons (Tukey post-hoc p<0.05) (Table 1). Comparing δ15N values of zooplankton across sites, South
Callandoon Lagoon (14.1±6.4‰) and Serpentine
Lagoon (15.3±5.3‰) had highly variable values, but
their δ15N values were not significantly different from
each other (Tukey post-hoc p = 0.93) nor from
Rainbow Lagoon values (Tukey post-hoc p>0.98).
The latter site showed less variable δ15N values (14.5±

1.5‰) (Table 2). Maynes Lagoon however, was
considerably 15N depleted (8.0±0.4‰) compared to
Serpentine and Rainbow lagoons (Tukey post-hoc
p<0.04). Even though, Tukey post-hoc comparisons
showed no significant differences in mean δ15N values
between Maynes Lagoon and South Callandoon Lagoon
(Tukey post-hoc p=0.06), this relationship needs careful
examination. Data for October 2002 were not available
for Maynes Lagoon and δ15N values of zooplankton in
South Callandoon were highly enriched in October 2002
and March 2003 compared to Maynes Lagoon (see
Table 2).
Comparison of variations in δ13C values between
primary sources (algae, CPOM and riparian vegetation), zooplankton and fish consumers revealed
significant differences (ANOVA, df=6,358; F=25.9;
p<0.05). The three species of fish had intermediate
mean δ13C isotope signatures (−17.1 to −34.3‰),
being significantly 13C depleted compared to algae


10

(Tukey post-hoc p<0.05) and, in the case of L.
unicolor and N. erebi, were enriched compared to
zooplankton (Tukey post-hoc p<0.01). The δ13C
values for Ambassis agassizii and zooplankton were
not significantly different (Tukey post-hoc p=1.00).
Differences between CPOM and riparian vegetation
and the three species of fish were also not significant
(Tukey post-hoc p>0.05), except for Nematalosa

erebi, which was significantly 13C enriched compared
to riparian vegetation (Tukey post-hoc p < 0.01)
(Table 1). Such results indicate an important contribution from zooplankton and organic matter carbon of
terrestrial origin to the fish species studied.
Overall, differences between fish isotope values and
their potential food resources were significant for δ15N
(ANOVA, df=6,358; F=111.9; p<0.05). Ambassis
agassizii and N. erebi were significantly 15N enriched
compared to all potential carbon sources (Tukey posthoc p<0.01), but not compared to zooplankton (Tukey
post-hoc p>0.07), which is consistent with a high
dietary contribution from zooplankton. Leiopotherapon
unicolor was significantly 15N enriched compared to
all sources (Tukey post-hoc p<0.01), as well as to
zooplankton (Tukey post-hoc p<0.01) (Table 2).
Ambassis agassizii was significantly depleted in
13
C, compared to the other fish species (Tukey posthoc p<0.02), whereas L. unicolor and N. erebi had
similar δ13C values (Tukey post-hoc p=1.00). Regarding the δ15N values, N. erebi was significantly depleted
compared to the other fish species (Tukey post-hoc
p<0.05), whereas A. agassizii and L. unicolor had
similar δ15N values (Tukey post-hoc p=0.94).
The minor flooding observed during the study period
had little effect on stable isotope values for the three
species of fish. Analysis performed on sites where
sufficient data was available showed that A. agassizii
δ13C values were significantly lower in July 2003 in
flooded South Callandon Lagoon (t-test, df=5.4, t=
38.1; p<0.01), and significantly higher on both March
and July 2003 in the flooded Rainbow Lagoon
(ANOVA, df=2,12; F=54.1; p<0.01) (Table 1). Differences observed in δ15N values were also significant

for both sites, South Callandon Lagoon (t-test, df=4.3,
t=3.9; p<0.01) and Rainbow Lagoon (ANOVA, df=
2,12; F=13.4; p<0.01) (Table 2). Variations across
sites were also significant for δ13C (ANOVA, df=3,25;
F=14.1; p<0.01) and δ15N (ANOVA, df=3,25; F=7.3;
p<0.01). The non-flooded Serpentine Lagoon was
significantly 13C depleted compared to all other sites

Environ Biol Fish (2011) 90:1–17

(Tukey post-hoc p<0.01). The only significant differences in 13C for A. agassizii across the remaining sites
occurred between Rainbow and South Callandoon
lagoons (Tukey post-hoc p=0.03) (Table 1). The
flooded Macintyre River had significantly higher
δ15N values than all other study sites (Tukey posthoc p≤0.01), which were not significantly different
(Tukey post-hoc p>0.05) (Table 2).
Stable isotope values of L. unicolor across sampling occasions where data was available showed
significantly enriched δ13C values for March 2003 in
Serpentine Lagoon (t-test, df=7.4, t=-2.9; p=0.02)
(Table 1). Differences in δ15N values in Serpentine
Lagoon for October 2002 and March 2003 were not
significant (t-test, df=6.5, t=0.9; p=0.42) (Table 2).
In contrast, δ13C values for the flooded Rainbow
Lagoon were not significantly different across sampling occasions (ANOVA, df=2,6; F=3.9; p=0.08),
whereas δ15N values were significantly depleted only
in July 2003 (ANOVA, df=2,6; F=16.8; p<0.01)
(Tables 1 and 2).
Significant differences in δ13C (ANOVA, df=3,22;
F=171.4; p<0.01) and δ15N (ANOVA, df=3,22; F=
21.8; p<0.01) values of L. unicolor were observed

across sites. South Callandoon Lagoon was significantly 13C depleted (Tukey post-hoc p<0.01) and
Serpentine Lagoon was significantly 13C enriched
(Tukey post-hoc p<0.01). Rainbow Lagoon and the
river site were not significantly different in δ13C
values (Tukey post-hoc p=0.76). With regard to the
δ15N values, Serpentine Lagoon was significantly 15N
enriched compared to all other sites (Tukey post-hoc
p<0.01). No difference in δ15N values of L. unicolor
was observed among the other study sites (Tukey
post-hoc p>0.13).
Comparison of N. erebi δ13C values across sampling occasions showed significant differences for sites
subjected to flooding (ANOVA, df=2,63; F=3.7;
p=0.03), with averages for March 2003 and July
2003 being significantly different (Tukey post-hoc p=
0.02) (See Table 1). However, sites not flooded also
presented significant differences in δ13C across sampling occasions (ANOVA, df=2,52; F=8.7; p<0.01),
with N. erebi δ13C values for October 2002 and March
2003 being significantly different (Tukey post-hoc
p<0.01) (Table 1). Similar results were observed for
N. erebi δ15N values from non-flooded sites (ANOVA,
df=2,52; F=8.7; p<0.01) where October 2002 and
March 2003 sampling occasions were significantly


Environ Biol Fish (2011) 90:1–17

different (Tukey post-hoc p<0.01). Differences in δ15N
values of N. erebi from flooded sites were not
significant over time (ANOVA, df=2,63; F=2.4;
p=0.10) (Table 2).

Variation in isotope values of N. erebi across study
sites was significant for both δ13C (ANOVA, df=
6,114; F=91.3; p<0.01) and δ15N values (ANOVA,
df=2,114; F=17.2; p<0.01). Fish from South Callandoon and Rainbow lagoons were significantly 13C
depleted compared to all other sites (Tukey post-hoc
p<0.01), whereas individuals from Serpentine and
Maynes lagoons were significantly 13C enriched
(Tukey post-hoc p < 0.01). Differences between
Maynes and Broomfield lagoon N. erebi were not
significant (Tukey post-hoc p=0.13). The flooded
Macintyre River showed significant δ13C differences
only from Punbougal Lagoon (Tukey post-hoc p=
0.79) (Table 1). Values of N. erebi δ15N were highly
variable, being significantly enriched in the Macintyre
River (Tukey post-hoc p < 0.01). The remaining
flooded sites (South Callandon and Rainbow lagoons)
had significantly different δ15N values only from
Punbougal Lagoon and the Macintyre River (Tukey
post-hoc p<0.01). South Callandoon was also significantly 15N enriched compared to Maynes Lagoon
(Tukey post-hoc p=0.02).
Spatio-temporal relationships between δ13C values
for fish consumers and primary sources
Ambassis agassizii, L. unicolor and N. erebi showed a
negative relationship between δ13C and δ15N values
(Table 3), that is, for individual sampling occasions,
more 13C depleted individuals were more 15N
enriched. Therefore, the stable isotope signatures of
the three fish species were more similar to zooplankton than to algae or particulate organic matter and
riparian vegetation. Benthic and attached algae were
generally too 13C enriched and 15N depleted, relative

to fish isotopic signatures, to have contributed to the
nutrition of the three species of fish examined during
this study. Furthermore, 15N values for riparian
vegetation and particulate organic matter were too
15
N depleted for these allochthonous sources to be
supporting fish in lagoon food webs.
If zooplankton, or to a lesser extent algae or
particulate organic matter, were an important source
of organic carbon for fish consumers, it would be
expected that the variability in isotopic signatures of

11

these sources would track the variability in signatures
of fish in space and time. That is, the isotopic
signatures of each of the three species of fish should
be positively correlated with the isotopic signatures of
their food source across sampling occasions. Fig. 3
shows clearly that this was the case for zooplankton and
all three fish species. Zooplankton explained 98% of the
variation in δ13C across study sites for A. agassizii,
85% for L. unicolor and 81% for N. erebi. Organic
matter (CPOM) derived from the riparian zone had
relatively consistent δ13C values across sites and
explained only 13% of the variation in carbon isotopic
signatures for N. erebi, and less than 6% for A.
agassizii and L. unicolor. Benthic attached algal δ13C
was also weakly correlated with δ13C of fish consumers,
explaining only 6% of the variation in δ13C for N. erebi

and less than 1% of A. agassizii and L. unicolor (Fig. 3).
Contribution of autochthonous versus allochthonous
sources of carbon to consumer biomass
Estimates derived from mixing models for study sites
where zooplankton was present suggest an important
contribution of zooplankton carbon to the biomass of fish
consumers (Table 4). Averaged contributions (± s.d.)
across sites indicate that fish consumers derived 46.9%
(±20.2) of their biomass from zooplankton. Contributions of particulate organic matter and algae were
variable across sampling occasions but, on average,
contributed another 38.1% (±26.5) and 24.0% (±23.2),
respectively. Ambassis agassizii relied most on zooplankton, deriving an average of 57.6% (±22.2) of its
biomass carbon from this autochthonous resource.
Nevertheless, important contributions from CPOM were
apparent in South Callandoon Lagoon. This species
showed least variation across sites and sampling
occasions, with a shift from zooplankton to algae as
the main source of carbon in Rainbow Lagoon soon
after flooding (March 2003) and back to zooplankton
later in the same year. South Callandoon Lagoon was
also the site of most variation in the contribution of
zooplankton carbon to A. agassizii, given that prior to
flooding CPOM was the principal source of carbon
to the biomass of this species, followed by a shift to
zooplankton carbon. Leiopotherapon unicolor was the
least reliant on zooplankton carbon with an average
across sites of 38.3% (± 22.3) contribution to its biomass.
This species was also the most variable regarding the
sources of carbon to its biomass, with important

12

Environ Biol Fish (2011) 90:1–17

Table 3 Pearson’s coefficient of correlation (r2) for relationships between δ13C and δ15N values (‰) for Ambassis agassizii,
Leiopotherapon unicolor and Nematalosa erebi. Only relationships based on n≥3 are shown
Ambassis agassizii
Sampling
occasion

October
2002

March
2003

Leiopotherapon unicolor
July
2003

Nematalosa erebi

October
2002

March
2003

July

2003

October
2002

March
2003

July
2003

Rainbow

0.9120

0.4793

0.0017

0.1172

–

–

0.7869

0.0638

0.0351

South
Callandoon
Macintyre River

0.5748

–

0.0089

–

–

0.3093

0.9314

0.7445

0.6356

–

–

–

–

–

–

0.4862

0.6448

0.5761

Maynes

–

–

–

–

–

–

0.7391

0.9409

–

Punbougal

–

–

–

–

–

–

0.0052

0.0046

0.7462

Serpentine

–

–

–

0.8829

0.9169

–

–

0.4871

0.5739

Broomfield

–

–

–

–

–

–

0.5702

–

–

contributions from all three sources (algae, CPOM and
zooplankton) across different times and sites (see Table 4).
A shift similar to the one reported for A. agassizii may
also be observed for L. unicolor in Rainbow Lagoon
where reliance on CPOM prior to flooding (October
2002) changed to algae soon after flooding (March
2003) and back to CPOM later in the same year (July
2003). Nematalosa erebi showed strong reliance on
carbon biomass from zooplankton (48.9% ±20.2 averaged across sites and sampling occasions), despite
important contributions from CPOM in South Callandoon Lagoon. Similar to the other study species,
N. erebi showed a shift in carbon sources in Rainbow
Lagoon from zooplankton before flooding to algae soon
after flooding and back to zooplankton late after
flooding (Table 4).

Discussion
Stable carbon and nitrogen isotope signatures for
consumers and primary food resources in floodplain
lagoons of the Macintyre River are similar to those
reported for other Australian rivers (Bunn and Boon
1993; Boon and Bunn 1994; Bunn and Davies 1999;
Bunn et al. 2003) although consumers were comparatively enriched in nitrogen. Spatial differences were
observed in δ13C and δ15N of consumers and sources
across individual lagoons and were mostly associated
with variations in the δ13C and δ15N values of
zooplankton and benthic attached algae. These spatial
variations are likely to be the result of differences in the

available sources of carbon and nitrogen assimilated by

primary producers across the study sites, especially in
relation to the effects of the flooding and drying history
of each lagoon on nutrient sources derived from the
floodplain, or direct nutrient inputs into isolated lagoons
by cattle coming to drink. In contrast, relatively
unchanged isotopic signatures of riparian tree leaves
reflect the similarities in riparian vegetation between
study sites and the stability of terrestrial plant isotope
signatures (Bunn and Boon 1993).
The importance of flooding has been recognized as
a mechanism for dispersion of nutrients produced in
different areas of a river catchment, such as the
headwaters or the floodplain (Thorp et al. 1998). In
the present study, it was not possible to ascertain if
temporal variations in isotope values of carbon
sources, and consequently consumers, were the result
of processes associated with flooding (e.g. delivery of
nutrients) or were due to temporal factors not related
to flooding, since high variability was observed for
both flooded and non-flooded sites. Furthermore,
there was no clear pattern of variation in stable
isotope values of sources and consumers across
sampling occasions prior to and after flooding. The
fact that the study period was relatively dry compared
to previous years, meant only minor flooding of sites
closer to the main channel of the Macintyre River,
whereas sites located further out on the floodplain
were not flooded at all and decreased in volume with
time. In floodplain river systems, flooding plays an
important role in ensuring the connectivity of marginal habitats with the main channel, thereby affecting

Environ Biol Fish (2011) 90:1–17
0
-5

Zooplankton

N. erebi
y = 0.9557x + 3.3136
R2 = 0.8076

-10

A. agassizii
y = 0.7836x - 3.9719
R2 = 0.9777

-15

13

δ C (‰)
Consumer

Fig. 3 δ13C values of
A. agassizii (white triangle),
L. unicolor (white square)
and N. erebi (white diamond) versus δ13C of primary sources (zooplankton,
algae and organic matter)

across study sites and sampling occasions where all
three primary sources were
available. Correlations
strength (R2) and equations
for regression between a
given source and each species of fish are also shown.
Diagonal lines represent the
1:1 relationship

13

-20

L. unicolor
y = 0.5785x - 8.774
R2 = 0.855

-25
-30
-35
-40
0
-5

Algae

-15

N. erebi
y = 0.1288x - 23.159

R2 = 0.061

-20

13

δ C (‰)
Consumer

-10

-25
-30
A. agassizii
y = 0.033x - 28.348
R2 = 0.0018

-35

L. unicolor
y = 0.0209x - 26.338
R2 = 0.0014

-40
0
-5

CPOM

-10

L. unicolor
y = 1.1418x + 5.6994
R2 = 0.0572

13

δ C (‰)
Consumer

-15

N. erebi
y = 1.0378x + 2.9111
2
R = 0.1324

-20
A. agassizii
y = 1.0293x + 0.6453
R2 = 0.0215

-25
-30
-35
-40
-40

-35

-30

-25

-20

-15

-10

-5

0

13

δ C (‰)
Sources

nutrient inputs and isotopic values of consumers and
sources of carbon (Hein et al. 2003). This lack of
connectivity of floodplain lagoons with the Macintyre
River may have had an affect on nutrient inputs and
levels over time, leading to the overall absence of

strong temporal patterns of isotope values for both
sources and consumers.
In general, fish had relatively similar δ13C isotopic
signatures, indicating similar sources of organic
carbon, despite differences in their feeding habits

68.3 (± 1.9)

–

–

–

Maynes

–

–

–
–

–
–

Unfeasible solution
Unfeasible solution

65.0 (± 2.1)
59.8–70.1

34.1 (± 30.7) 48.4(± 18.9)

29.8–40.1

35.0 (± 2.1)

42–73.8

57.9(± 6.7)

23.0 (± 4.3)

0–100
13.2–32.8

35.5 (± 8.0)

–

–

Zoopl.

24.5 (± 3.5)

23.0

34.9 (± 7.1)

–

–

–

–

–

–

–

39.1 (± 10.1) 60.9 (± 10.1)

47.5 (± 23.9) 40.8 (± 29.6)

–

17.8 (± 13.2) 82.2 (± 13.2)

77.0

5.2 (± 8.1)

–

–

16.4 (± 13.5) 49.4 (± 13.4)

73.4 (± 3.6)

28.7 (± 13.1) 59.2 (± 16.9)

CPOM

24.0 (± 23.2) 38.1 (± 26.5) 46.9 (± 20.2)

–

–

–

11.6 (± 6.1)

–

–

–

37.7 (± 17.3) 59.9 (± 1.3)

–

–

77.0 (± 4.3)

14.8–56.1

2.2 (± 0.1)

12.1 (± 4.9)

Algae

59.0 (± 19.7) 34.2 (± 9.5)

9.8–43.9

26.9 (± 5.4)

58.6–80.6

69.6 (± 4.9)

Zoopl.

67.2–86.8

0–54.3

1.0 (± 19.2)

–

–

1.0 (± 14.1)

50.7–90.9

70.8 (± 7.2)

2.4–33.7

18.1 (± 7.2)

CPOM

Average

Estimations are based on two- and three-source mixing models using δ13 C and δ15 N. (−) indicates no data available and (a ) indicates source which was removed from three-source
mixing model to enable feasible solution. b Averages (± s.d.) include only the feasible solutions. Note that only Rainbow and South Callandoon lagoons were flooded during the
study period

–

20.4 (± 23.4) 27.8 (± 24.3) 57.6 (± 22.2) 33.3 (± 19.7) 39.5 (± 21.6) 38.3 (± 22.3) 24.5 (± 25.2)

–

Averageb

–

a

–

34.3–64.5

49.4 (± 5.9)

6.6 (± 6.9)
0–24.4

–

35.5–65.7

50.6 (± 5.9)

6.6 (± 2.9)
0–19.2

–

–

75.0(± 3.1)
35.7–100

Unfeasible solution

Unfeasible solution

a

0–100

–

63.1–73.3

26.7–36.9

a

0–64.1

–

64.6–81.2

72.9 (± 3.0)

–

0–96.2

Serpentine

68.2 (± 1.6)

31.8 (± 1.6)

52.2–63.8

–

18.8–35.4

27.1 (± 3.0)

–

0–64.4

Maynes

a

32.0 (± 3.2)

23.8–40.2

9.9 (± 1.6)

6.4–13.5

18.4 (± 3.6)

–
58.0 (± 2.4)

–

–

–

–

91.5 (± 3.2)

–

–

–

–

0–100

8.5 (± 3.2)

–

0–49.5

July 2003 (after flooding, winter)

–

South
Callandoon

–

–

34.1 (± 10.5) 39.9 (± 12.1)
–

40.1 (± 15.9) 58.6 (± 30.2) 14.6 (± 17.9) 26.8 (± 16.1) 61.2 (± 32.3)

–

26.5 (± 7.9)
–

0–8.4

2.3 (± 2.6)

3.2–21.4

12.3 (± 4.2)

Algae

0–100

–

39.4 (± 6.0)
–

–

32.8–46.6

39.7 (± 2.5)

Zoopl.

0–100

Serpentine

Rainbow

55.1 (±10.5)
–

–

–

43.4 (± 3.5)
34.2–52.5

12.6–21.2

16.9 (± 1.8)

CPOM

N. erebi

59.8 (± 29.9) 0.1 (± 17.8)

South
–
Callandoon

Rainbow

March 2003 (after flooding, summer)

Maynes

21.6 (± 8.4)

–

–

Serpentine

23.2 (± 5.1)

22.0 (± 4.1)
4.3–39.7

75.9 (± 5.5)

62.9–73.7

52.3–99.5

24.6 (± 2.8)

17.5–31.7

7.1 (± 1.3)

4.1–10.1

South
2.1 (± 1.7)
Callandoon 0–6.4

Rainbow

October 2002 (prior to flooding)

Algae

Zoopl.

Algae

CPOM

L. unicolor

A. agassizii

Table 4 Percent contribution (±s.e.) and lower-upper 95% confidence interval of major energy sources (zooplankton, algae and organic matter) to the study species of fish on
sampling occasions where zooplankton was available

14
Environ Biol Fish (2011) 90:1–17


Environ Biol Fish (2011) 90:1–17

(Pusey et al. 2004; Medeiros 2005). Of the major
primary sources likely to be supporting fish consumers, littoral algae were too 13C enriched to be
contributing significantly to fish biomass. Mixing
model results suggest that two major food resources
supported fish consumers—zooplankton and particulate organic matter. The nitrogen isotopic signatures
of zooplankton and particulate organic matter were
markedly different, and the fact that fish had strong
negative relationships between δ13C and δ15N is also
consistent with a 15N enriched zooplankton endpoint.
Mixing model results from the present study and
dietary data for the study species are consistent in
identifying zooplankton is a major source of energy
supporting Macintyre River fish consumers. Medeiros
and Arthington (2008a,b) showed that zooplankton
form an important component of the diet of A. agassizii,
L. unicolor and N. erebi in all or some stages of their
life history. Ambassis agassizii fed consistently on
zooplankton in Macintyre River floodplain lagoons,
which is in accordance with results of the present

study. Leiopotherapon unicolor consumed zooplankton and aquatic insects and N. erebi diets were
comprised of zooplankton and detritus (Medeiros and
Arthington 2008b). These differences in dietary food
sources are supported by the mixing model results
reported here, where L. unicolor displayed relatively
high variability in isotopic signatures reflecting contributions from zooplankton, algae and CPOM. Given
that this species fed consistently on detritivore/herbivore aquatic insects (as well as zooplankton—Medeiros
and Arthington 2008b) it is not surprising that carbon
contributions varied across different sources as shown
by mixing model analysis. Nematalosa erebi also
presented mixing model results indicative of a second
important source of biomass carbon, namely CPOM.
This species fed mostly on detritus, which consisted of
a mixture of fine particulate organic matter and, to a
lesser extent, algae (Medeiros and Arthington 2008b).
Fisher et al. (2001) recognized zooplankton as
important agents of energy transfer to small sized fish
and, in turn, to higher piscivorous consumers. Based
on stable isotopic data, Fisher et al. (2001) showed
that the major source supporting zooplankton communities in floodplain habitats was an autochthonous
primary producer, namely phytoplankton. In the
present study, high concentrations of suspended
sediment in the water prevented the collection of
clean phytoplankton samples. Nevertheless, we sug-

15

gest that phytoplankton was the most likely carbon
source for zooplankton based on other evidence for
Australian dryland rivers. For example, despite the

high turbidity levels in floodplain waterholes of the
arid Cooper Creek catchment, western Queensland,
Bunn et al. (2003) suggested that marked diel
variation in dissolved oxygen saturation in the open
surface water indicates high rates of phytoplankton
production. Similarly, there were indications of
phytoplankton production in some of the Macintyre
study sites, based on diel variation in dissolved
oxygen saturation in surface water and field observations of algal slicks (Medeiros 2005).
The composition of seston, or particulate organic
matter suspended in the water column, also provides
further evidence that phytoplankton is the major
source of energy for zooplankton. Huryn et al.
(2001) found significant temporal changes in isotopic
signatures in several size classes of seston and argued
that such changes were the result of changes in the
ratio of terrestrial to aquatic (phytoplankton) sources
of organic carbon related to seasonal changes in water
flow. This could explain the variations in isotopic
signatures of zooplankton observed during the present
study in relation to sites and sampling occasions.
Assuming that zooplankton consumed fine and ultrafine sestonic organic matter and phytoplankton, variation in the proportion of these two sources of organic
carbon in the seston across lagoons and sampling
occasions may have led to the observed variation in
isotopic signatures of zooplankton. Because isotopic
signatures of particulate organic matter (CPOM)
showed little change across sites in the Macintyre
floodplain, a significant contribution of allochthonous
carbon from this source would be more likely to
generate relatively constant isotopic signatures of

zooplankton across sampling occasions (cf. Hadwen
2003; Hadwen and Bunn 2004), which clearly was not
the case in the present study.
In other studies (see Araujo-Lima et al. 1986;
Forsberg et al. 1993), it has been found that despite
having stomach contents dominated by detritus, fish
have shown little contribution of organic carbon
derived from tree leaves, macrophytes or periphyton.
Instead, these detritivorous species receive a large
fraction of their carbon from phytoplankton (AraujoLima et al. 1986; Forsberg et al. 1993). In the
floodplain lagoons of the Macintyre River, sestonderived phytoplankton probably provides the main


16

energy base for the three fish species, most likely via
zooplankton in the case of A. agassizii and L. unicolor,
and either via zooplankton or sestonic detrital matter
for N. erebi. The large detrital component found in the
stomachs of N. erebi during the study period (Medeiros
2005) is most likely to be material incidentally
consumed along with the nutritious phytoplankton
from the seston. Organic matter consumed by detritivorous fishes is typically composed of a mixture of
phytoplankton and other detrital matter (see Forsberg
et al. 1993), with the phytoplankton component being
either physically separated and ingested (Bowen 1984)
or simply selectively digested and assimilated. The
more labile nature of the autochthonous component of
detrital material renders it a more suitable and readily
digestible source of nutrition for consumers than the

more refractory allochthonous organic matter (Thorp
et al. 1998). Thus it is reasonable to suggest that the
phytoplankton of floodplain lagoons would be easier to
assimilate than detritus, and was therefore an important
source of energy for the fish, especially N. erebi.
In conclusion, results of the present study do not
support the predictions of the RCC and FPC, as most
of the evidence indicates a stronger dependence of
consumers on autochthonous sources than on
allochthonous particulate organic matter. Thorp and
Delong (1994) proposed in the RPM that a combination of local autochthonous production (by phytoplankton, benthic algae or macrophytes) and direct
inputs from the riparian zone (tree leaves and
particulate organic matter) during periods not limited
to flood pulses are the major sources of carbon
assimilated by consumers in large rivers. In the
present study, there is an indication that the RPM
would be a more appropriate model for floodplain
lagoons of the Macintyre River, but further research is
needed to clarify the influence of phytoplankton and
bacterially processed carbon on the food web. Even
though the fish species in our study had similar ranges
of variation in δ13C isotopic values, there are
indications from δ15N and the mixing models that
all three species (particularly N. erebi) relied on a
combination of zooplankton (and, presumably, ultimately
phytoplankton) and organic matter of riparian origin as
their main energy sources (cf. Pusey et al. 2004). In that
case, the integration of RPM into the broader RES
model would better represent the sources of carbon for
fish in isolated floodplain lagoons of the Macintyre

catchment. Predictions from the RES would also suggest

Environ Biol Fish (2011) 90:1–17

that flooding might have an important effect on sources
of carbon for fish in these floodplain lagoons. However,
the role of flooding was uncertain during the present
study since most of the lagoons were not flooded,
leading to the overall lack of clear differences in 13C and
15
N values across sources and consumers between
flooded and non-flooded sites and sampling occasions.
Acknowledgments Fish were collected under Queensland
and New South Wales Fisheries Permit Nos. PRM00234H,
PRM03315D and P01/0089, and Griffith University Research
Ethics Protocol No. AES/02/01/aec. The authors thank Glenn
Wilson (Northern Basin Laboratory, MDFRC, Goondiwindi)
for contributions to fieldwork design and methods; René
Diocares (Stable Isotope Analysis laboratory, Griffith University)
for technical advice and sample analysis. We are grateful to Stuart
Bunn (ARI, Griffith University) for valuable advice on food web
analysis and Wade Hadwen (ARI, Griffith University) for helpful
comments on an earlier draft of this manuscript. Elvio Medeiros
thanks the Brazilian Foundation for Post-Graduate Education
(CAPES) for his PhD scholarship (BEX 1475/99-1). Financial and
administrative support from Griffith University is also gratefully
acknowledged. This paper is a contribution to project 1.F.102 of
the eWater Cooperative Research Centre.

References

Araujo-Lima CARM, Forsberg BR, Victoria R, Martinelli L
(1986) Energy sources for detritivorous fishes in the
Amazon. Science 234:1256–1258
Arthington AH, Balcombe SR, Wilson GA, Thoms MC,
Marshall JC (2005) Spatial and temporal variation in fish
assemblage structure in isolated waterholes during the
2001 dry season of an arid-zone floodplain river, Cooper
Creek, Australia. Mar Freshwater Res 56:25–35
Balcombe SR, Bunn SE, Davies PM, McKenzie-Smith FJ (2005)
Variability of fish diets between dry and flood periods in an
arid zone floodplain river. J Fish Biol 67:1552–1567
Boon PI, Bunn SE (1994) Variations in the stable isotope
composition of aquatic plants and their implications for
food web analysis. Aquat Bot 48:99–108
Bowen SH (1984) Detritivory in neotropical fish communities. 59–
66. In: Zaret TM (ed) Evolutionary ecology of neotropical
freshwater fishes. Dr. W. Junk Publishers, The Hague, p 173
Bunn SE, Boon PI (1993) What sources of organic carbon drive
food webs in billabongs? A study based on stable isotope
analysis. Oecologia 96:85–94
Bunn SE, Davies PM (1999) Aquatic foodwebs in turbid, aridzone rivers: preliminary data from Cooper Creek, Western
Queensland. 67–76. In: Kingsford RT (ed) A free-flowing
river: The ecology of the Paroo River. National Parks and
Wildlife Service, Hurstville, p 320
Bunn SE, Loneragan NR, Kempster MA (1995) Effects of acid
washing on stable isotope ratios of C and N in panaeid
shrimp and seagrass: implications for food-web studies using
multiple stable isotopes. Limnol Oceanogr 40:622–625

Environ Biol Fish (2011) 90:1–17
Bunn SE, Davies PM, Winning M (2003) Sources of organic
carbon supporting the food web of an arid zone floodplain
river. Freshw Biol 48:619–635
Bunn SE, Balcombe SR, Davies PM, Fellows CS, McKenzieSmith FJ (2006) Productivity and aquatic food webs of
desert river ecosystems. In: Kingsford RT (ed) Ecology
of desert rivers. Cambridge University Press, Cambridge,
pp 76–100
Burford MA, Cook AJ, Fellows CS, Balcombe SR, Bunn SE
(2008) Sources of carbon fuelling production in an arid
floodplain river. Mar Freshwater Res 59:224–234
DeNiro MJ, Epstein S (1981) Influence of diet on the
distribution of nitrogen isotopes in animals. Geochim
Cosmochim Acta 45:341–351
DWR (1995) Water resources of the Border Rivers system in
northern New South Wales, New South Wales Department
of Water Resources
Fisher SJ, Brown ML, Willis DW (2001) Temporal food web
variability in an upper Missouri River backwater: energy
origination points and transfer mechanisms. Ecol Freshwater Fish 10:154–167
Forsberg BR, Araujo-Lima CARM, Martinelli LA, Victoria RL,
Bonassi JA (1993) Autotrophic carbon sources for fish of
the central Amazon. Ecology 74:643–652
France RL (1996) Carbon-13 conundrums: limitations and
cautions in the use of stable isotope analysis in stream
ecotonal research. Can J Fish Aquat Sci 53:1916–1919
Hadwen WL (2003) Effects of nutrient additions on dune lakes
on Fraser Island, Australia. PhD Thesis, Griffith University, Brisbane, Australia. p 150
Hadwen WL, Bunn SE (2004) Tourists increase the contribution of autochthonous carbon to littoral zone food webs in
oligotrophic dune lakes. Mar Freshwater Res 55:1–8

Hein T, Baranyi C, Herndl GJ, Wanek W, Schiemer F (2003)
Allochthonous and autochthonous particulate organic
matter in floodplains of the River Danube: the importance of hydrological connectivity. Freshwater Biol 48:
220–232
Houldsworth B (1995) Central and north west regions water
quality program. 1994/95 report on nutrients and general
water quality monitoring. Department of Land and Water
Conservation Technical Services Division, Parramatta. p 95
Huryn AD, Riley RH, Young RG, Arbuckle CJ, Peacock K,
Lyon G (2001) Temporal shift in contribution of terrestrial
organic matter to consumer production in a grassland river.
Freshwater Biol 46:213–226
Junk WJ, Bayley PB, Sparks RE (1989) The flood-pulse
concept in river-floodplain systems. In: Proceedings of the
International Large River Symposium (LARS). Dodge DP
(ed) Canadian Journal of Fisheries and Aquatic Sciences
Special Publication 106. pp 110-127
Lewis WM Jr, Hamilton SK, Lasi MA, Rodriguez M, Saunders
JF III (2000) Ecological determinism on the Orinoco
floodplain. Bioscience 50:681–692
Lewis WM Jr, Hamilton SK, Rodriguez MA, Saunders JF III,
Lasi MA (2001) Foodweb analysis of the Orinoco
floodplain based on production estimates and stable
isotope data. J N Am Benthol Soc 20:241–254

17
McCosker RO (1996) Environmental scan of the Border Rivers
catchment. Department of Natural Resources & Department of
Land and Water Conservation, New South Wales/Queensland
Medeiros ESF (2005) Trophic ecology and energy sources for

fish on the floodplain of a regulated dryland river:
Macintyre River, Australia. PhD Thesis, Griffith University, Brisbane, Australia. p 247
Medeiros ESF, Arthington AH (2008a) Diel variation in food
intake and diet composition of three native fish species in
floodplain lagoons of the Macintyre River, Australia. J
Fish Biol 73:1024–1032
Medeiros ESF, Arthington AH (2008b) The importance of
zooplankton in the diets of three native fish species in
floodplain waterholes of a dryland river, the Macintyre
River, Australia. Hydrobiol 614:19–31
Morton SR, Short J, Barker RD (1995) Refugia for biological
diversity in arid and semi-arid Australia. Biodiversity
Series, Paper no. 4, Report to the Biodiversity Unit of the
Department of Environment, Sport and Territories, Canberra, Australia. p 171
Peterson BJ, Fry B (1987) Stable isotopes in ecosystem studies.
Annu Rev Ecol Syst 18:293–320
Phillips DL (2001) Mixing models in analysis of diet using
multiple stable isotopes: a critique. Oecologia 127:166–170
Phillips DL, Gregg JW (2001) Uncertainty in source partitioning using stable isotopes. Oecologia 127:171–179
Puckridge JT, Sheldon F, Walker KF, Boulton AJ (1998) Flow
variability and the ecology of large rivers. Mar Freshwater
Res 49:55–72
Pusey BJ, Kennard MJ, Arthington AH (2004) Freshwater
Fishes of North-eastern Australia. CSIRO Publishing,
Melbourne, p 684
Thorp JH, Delong MD (1994) The riverine productivity model:
an heuristic view of carbon sources and organic processing
in large river ecosystems. Oikos 70:305–308
Thorp JH, Delong MD (2002) Dominance of autochthonous
autotrophic carbon in food webs of heterotrophic rivers.

Oikos 96:543–550
Thorp JH, Delong MD, Greenwood KS, Casper AF (1998) Isotopic
analysis of three food web theories in constricted and
floodplain regions of a large river. Oecologia 117:551–563
Thorp JH, Thoms MC, Delong MD (2006) The riverine
ecosystem synthesis: biocomplexity in river networks
across space and time. River Res Applic 22:123–147
Vander Zanden MJ, Rasmussen JB (2001) Variation in δ15N
and δ13C trophic fractionation: implications for aquatic
food web studies. Limnol Oceanogr 46:2061–2066
Vannote RL, Minshall GW, Cummins KW, Sedell JR, Cushing
CE (1980) The river continuum concept. Can J Fish Aquat
Sci 37:130–137
Walker KF, Sheldon F, Puckridge JT (1995) A perspective on
dryland river ecosystems. Regul Rivers 11:85–104
Winemiller KO (1996) Factors driving temporal and spatial
variation in aquatic floodplain food webs. 298–312. In: Polis
GA, Winemiller KO (eds) Food webs: integration of patterns
and dynamics. Chapman and Hall, New York, p 472
Zar JH (1999) Biostatistical analysis. Prentice Hall, Upper
Saddle River, p 663


Environ Biol Fish (2011) 90:19–27
DOI 10.1007/s10641-010-9713-y

Settlement and pigmentation of glass eels
(Anguilla rostrata Lesueur) in a coastal lagoon
Daniel F. Luers & Joseph W. Love &
Gretchen Bath-Martin

Received: 16 August 2009 / Accepted: 6 August 2010 / Published online: 2 September 2010
# Springer Science+Business Media B.V. 2010

Abstract American eel (Anguilla rostrata Lesueur) is
born in the Atlantic Ocean, but larvae redistribute
throughout diverse habitats of North American estuaries and freshwater streams. We hypothesized that
early stage A. rostrata differed in abundance among
sites within a single coastal lagoon, Newport Bay
(Maryland) for two sampling seasons (March–May in
2007 and 2008). Catch per unit effort (CPUE) of early
D. F. Luers : J. W. Love (*)
Department of Natural Sciences,
University of Maryland Eastern Shore,
Princess Anne, MD, USA
e-mail:
G. Bath-Martin
National Oceanic and Atmospheric Administration,
National Marine Fisheries Service,
Southeast Fisheries Science Center,
Beaufort, NC, USA
Present Address:
J. W. Love
Maryland Department of Natural Resources,
Division of Inland Fisheries,
Annapolis, MD 21401, USA
Present Address:
D. F. Luers
National Oceanic and Atmospheric Administration,
National Marine Fisheries Service,

Pacific Islands Regional Office,
Honolulu, HI, USA

stage A. rostrata was usually similar between years at
a site, except that it was higher at one site in 2008
than in 2007. The CPUE varied among sites within
Newport Bay, but not significantly so because of
high, intra-annual variance in CPUE at a site. As
reported for New Jersey coastal estuaries, variation in
the CPUE tended to be higher in brackish water
habitats. Intra-annual variation in CPUE from March
until May was partially explained by oxygenation and
salinity. The CPUE of settling eels was greatest
when water was well-oxygenated (dissolved oxygen
>8 mg·L−1) and mildly brackish (0.3–3.2 ppt). While
larval supply to a coastal estuary may annually
influence the magnitude of potential settlers, temporal
differences in habitat conditions within Newport Bay
also influenced settlement patterns. Differences in
habitat conditions can affect the pigmentation rates of
settling early stage eels. We measured rates of
pigmentation, which corresponded with age of the
fish. However, rates did not differ among sites or vary
with habitat conditions. Pigmentation levels from
March to May increased at a rate of about 0.02%
per week. Monitoring programs for early phase
American eel should consider the steepness of habitat
gradients within estuaries and habitat covariates when
assessing population status.
Keywords Fish . Maryland . Salinity . Recruitment .

Larval . Occupancy . Detection


20

Introduction
Anguillid eel populations are declining (Haro et al.
2000; Stone 2003), possibly because of anthropogenic changes in watersheds, overfishing, parasitism, resource availability, or changes in oceanic
currents (Castonguay et al. 1994; Haro et al. 2000;
Bonhommeau et al. 2008). While reproduction
occurs in the Sargasso Sea (Facey and Van Den
Avyle 1987), northern and western currents (Power
and McCleave 1983) lead larval American eel
(Anguilla rostrata Lesueur) to the east coast of the
United States after eight to twelve months (ASMFC
2000; Wang and Tzeng 2000) and to Mid-Atlantic
coastal systems in spring (Love et al. 2009a). Larvae
then metamorphose from leptocephalus larvae into
glass eels (Facey and Van Den Avyle 1987; Powles
and Warlen 2001), which use selective tidal transport and sense chemical odorants to migrate into
estuaries and freshwater streams (Creutzberg 1959,
1961; McCleave and Kleckner 1982; Sorenson
1986; Sola 1995; Sola and Tongiorgi 1996). Once
glass eels reach estuaries and freshwater streams,
they settle into the benthos (Haro and Krueger
1988), where their early life history may be affected
by predators and harvest by humans, resources, and
possibly parasites or disease. Factors affecting the
distribution and growth of juvenile American eel in
coastal estuaries are not well-known, despite their

commercial harvest in such systems.
We examined distribution and growth in relation to
the environment for the early life history of American
eel. The life history stage studied here included glass
eels without pigment to fully pigmented, age 0 elvers
(i.e., stage 7; Haro and Krueger 1988), hereafter
collectively termed juveniles. At large spatiotemporal
scales, ingress of juveniles is correlated with precipitation (Sullivan et al. 2006). Settlement of juveniles
within the estuary differs among locations (Sullivan et
al. 2009) and the majority preferred brackish habitats
in a laboratory study (Edeline et al. 2005). The size of
juvenile populations in brackish, lower estuary habitats may be highly variable over time (Sullivan et al.
2009). Freshwater contingents that tend to be more
active may periodically migrate upstream and leave
brackish water contingents behind (Edeline et al.
2005). Thus, settlement and abundance in freshwater
habitats should be lower and less variable over time
than that in brackish habitats.

Environ Biol Fish (2011) 90:19–27

Differences in settlement habitat may affect pigmentation rate, which increases with age and water
temperature (Dou et al. 2003; Briand et al. 2005).
Pigmentation provides disruptive coloration and
countershading that presumably reduces an eels’
conspicuousness to predators and increases survivorship. Traditionally, pigment is measured by staging
pigment into categories (Haro and Krueger 1988).
Such methods complicate the analysis of pigmentation rates when samples are represented by only two
or three stages. Here, we generated a new technique
for measuring pigmentation level of American eel and

addressed whether pigmentation rates varied with
time, water temperature, and other habitat conditions.
Our objectives were: 1) to determine if salinity was
an important factor in the settlement of juvenile
American eel; and 2) to determine if pigmentation
levels or rates varied with time, while accounting for
variation in water temperature and site effects. We
hypothesized that: 1) juvenile American eel abundance differed among habitats and temporal variability was lowest in freshwater; and 2) pigmentation
levels were positively correlated with time and water
temperature.

Methods
Sampling methods
Juveniles were sampled weekly from 27 April to 29
June (2007) and from 1 May to 26 June (2008). Five
sites were sampled for glass eels in tributaries of
Newport Bay (coastal bays watershed, Maryland)
(Fig. 1). Newport Bay is a shallow water lagoon,
which is separated from the Atlantic Ocean by two
barrier islands. Two inlets allow for tidal exchange
with the Atlantic Ocean, Ocean City Inlet (OCI) and
Chincoteague Inlet (CI). Five upstream tributaries
were sampled: Ayers Creek, Trappe Creek, Trappe
Creek at Trappe Bridge (hereafter, Trappe Bridge),
Hayes Landing Bridge, and Marshall Creek Dock
(Fig. 1). Study sites were selected based on their
accessibility. Site locations were plotted in relation to
Newport Bay using ArcGIS (Version 9.0 Environmental Systems Research Institute).
Juveniles were surveyed using artificial habitat
collectors constructed following Silberschneider et al.

(2001). The collectors were constructed by attaching


Environ Biol Fish (2011) 90:19–27

21

Fig. 1 Map of Newport
Bay, a sub-watershed of the
coastal lagoons of Maryland
(U.S.A., see inset). Five
sites were sampled: Marshall Creek; Hayes Landing;
Trappe Creek at Trappe
Bridge; Trappe Creek; and
Ayers Creek

72 cm high and 0.95 cm diameter frayed polyethylene
rope to a 38 cm diameter plastic planter bottom.
Fifteen frayed ropes were attached to the concave
surface of the planter. The frayed ropes were folded in
half and attached to the center point using plastic zip
ties. Approximately 500 g of weight was attached to
the bottom of the unit to promote in-stream stability.
To reduce artificial appearances, each collector was
placed in a stream for at least three weeks prior to the
commencement of sampling for conditioning.
Eel collectors were set for a week before removing
them from the water and shaking them 30 times into a
15 L plastic tub before returning them to the water.
The contents of the tub were sieved with a 1 mm

mesh metal sifting grate. All juveniles (≤75 mm) were
counted; up to 60 per week were preserved in 10%
neutral buffered formalin for pigmentation analysis.
Preserved specimens were photographed using a
Nikon Coolpix 5000 digital camera. Each digital
image of a glass eel was analyzed using TPSdig
(Version 2.0). The depth of pigmentation (PD) was
measured from the insertion of the dorsal fin to the
termination of pigment along the left, lateral side
of the body. Total length (TL) and body depth
(BD) at the first dorsal fin were measured
following Hubbs and Lagler (2004). The percent
pigmentation level (%P) was calculated as a ratio of
PD to BD: %P = PD/BD * 100.
To evaluate the effectiveness of our pigmentation
method, we subset our data and staged 71 glass eels
following Haro and Krueger (1988), who categorized
glass eels into 7 stages according to pigmentation

intensity. We used analysis of variance (ANOVA) to
determine if our calculations of %P (dependent
variable) differed among stage classifications.
To test our hypotheses that salinity and temperature
were related to abundance and %P, respectively, we
measured these variables during each sampling event.
Because these variables can co-vary with other habitat
conditions and time, we measured additional habitat
variables to parse out the effects of each. The
measured variables were chosen because they were
thought to be the ones that would most importantly

affect abundance and %P. During each sampling event,
we recorded: site water temperature (°C), salinity (ppt),
water transparency (%), dissolved oxygen (DO;
mg·L−1), chlorophyll a (μg·L−1), and stream depth
(cm). Water transparency and depth were measured
using a graded Secchi disk by lowering the Secchi disk
to the maximum level of light penetration and to the
substrate, respectively. The % water transparency was
computed as a ratio of the depth of transparency of
water to the depth of the water column. Chlorophyll a
was measured using a Flurometer (Turner Designs).
The remaining variables were measured using a
Yellow Springs Instruments device (Model 85).
Analysis
Abundance was converted to catch per unit effort
(CPUE), which was total abundance at a site divided
by the number of traps used at the site. CPUE was
then square-root transformed, which did not normalize the variance (Shapiro–Wilks test for normality, P<


22

0.0001). We used a non-parametric test (Kruskal–
Wallis, or K–W) to determine if CPUE differed
among sites for each year of the study. Pigmentation
level was averaged across individuals collected during
a sampling event and transformed by the arcsine
square-root. The distribution of variance for transformed values did not differ from a normal distribution (P=0.89). We used an analysis of covariance
(ANCOVA) to determine if %P varied among sites
(independent effect), while accounting for variance

due to week of sampling (covariate) and BD
(covariate). For this analysis, data were pooled
between years and we included an effect to test for
the interaction of week and site to determine if
pigmentation rate differed among sites.
To determine if environmental conditions influenced CPUE or %P, we used a modeling approach to
relate CPUE and %P (response variable) to environmental variables (predictor variables). For CPUE,
environmental predictors included salinity, DO, water
temperature, water clarity, and depth. For %P,
environmental predictors were the same as above,
but also included BD and time (covariates to %P). We
used separate nonparametric multiplicative regression
(NPMR) models that are similar to multiple regression models, but are nonparametric and fit local,
rather than global models to the dataset (McCune
2006). We used a locally fitting model with a
minimum, average neighborhood window size of 5%
of the sample size; sparse data points were interpolated within this window using a Gaussian weighting
function. Thus, the interpolation smoothed the response surface across environmental predictors. We
sought the most parsimonious model that maximized
the amount of variance explained in CPUE or %P
with the fewest number of predictors. The number of
predictors allowed in the model were moderately
restricted by the modeling procedure (improvement
criterion = 5), which allowed for selective retention of
predictors that improved the model by 5%.
Each environmental variable retained in the final
model was assigned a tolerance value (τ), which is a
smoothing parameter that indicates the local or global
importance of the variable in the model. Low values of
τ suggest more global importance than local importance, and relatively low tolerance to variation in the

predictor variable. It has ecologically been defined as

Environ Biol Fish (2011) 90:19–27

the tolerance of the species to the variable (McCune
and Mefford 2004). The fit of the model to the dataset
was evaluated using an xr2 (a cross-validated coefficient of determination for nonparametric models) and a
Monte-Carlo randomization test that tested the null
hypothesis that the fit of the model is not better than
one achieved from randomized datasets. In this case,
the response variables are shuffled and then related to
the predictor variables. The proportion (p) of 100
randomized runs that produced a model with an equal
or better fit was calculated and used to evaluate the
significance of the model. For NPMR analysis, we
used Hyperniche (Version 1.10, McCune and Mefford
2004). For all other analyses, we used SYSTAT
(Version 11 SYSTAT Software, Inc.).

Results
Trappe Bridge and Trappe Creek sites were freshwater, small and cooler streams than the other sites
(Table 1). Marshall Creek, Hayes Landing, and Ayers
Creek sites were the deepest and most saline of the
sites (Table 1).
We collected and photographed 955 juvenile eels
throughout the entire study period. Catch ranged from
0 to 42 per trap and averaged 3.2±2.4 SD. Catch per
unit effort did not differ, on average, among sites in
2007 (K–W test statistic = 4.92, p=0.30). In 2008,
however, average CPUE was higher at Hayes Landing

than other sites (K–W test statistic = 21.02, p<
0.0001; Fig. 2). Variance in CPUE was lower for
freshwater sites, Trappe Bridge and Trappe Creek,
than other sites (Fig. 2).
Salinity (τ=0.13), DO (τ=0.17), and water clarity
(τ=0.97) explained approximately 23% of variance in
the CPUE data (xr2 =0.23; p<0.05). The CPUE was
highest in oligohaline habitats (0.3–3.2 ppt) that were
oxygen rich (near 8 mg·L−1) (Fig. 3). Water clarity
had less influence on CPUE than salinity or DO.
Pigmentation level generally increased with stage
number (Fig. 4), with significant increases occurring
from stages 2 through 5 (Tukey–Kramer’s Honestly
Significant Difference Test, P<0.05). Pigmentation
level did not differ among stage groups 5 through 7,
which may differ in distinction by intensity rather
than depth of pigment along the body.


Environ Biol Fish (2011) 90:19–27

23

Table 1 Habitat conditions for sites (abbreviations in Fig. 1)
surveyed for glass eels (Anguilla rostrata) in the Newport Bay
sub-watershed (Maryland, U.S.A.) from Ocean City Inlet
(Spring 2007 and 2008). Averages (across all sampling

events, ± SD) for environmental variables at each site were:
temperature (TEMP, in °C), dissolved oxygen (DO, in

mg·L−1), salinity (SAL), water clarity (WC, in cm), and depth
(DEP, in cm)

TEMP

DO

SAL

WC

DEP

Marshall Hall Creek

22.0±4.0

6.4±3.8

9.6±7.0

40.9±14.5

118.9±25.8

Hayes Landing

22.6±4.8

7.4±2.6

5.2±8.1

40.3±15.0

95.5±18.5

Trappe Bridge

19.2±3.5

6.0±1.7

1.2±4.3

54.4±19.9

79.0±25.1

Trappe Creek

19.4±3.4

5.4±1.9

0.5±1.2

40.3±11.4

61.6±18.7

Ayers Creek

23.4±4.6

7.6±2.9

7.8±7.6

39.9±12.6

146.7±21.9

Pigmentation level and rate were similar among sites.
Neither site (F4,54 =1.60, p=0.19) nor the interaction of
site and week (F4,54 =1.74, p=0.16) explained variation
in %P. Pigmentation level significantly increased by
0.02%/week (F1,54 =14.66, p<0.0001), while accounting for BD (F1,54 =15.54, p<0.0001). Environmental
variables explained less variance in %P than time (τ=
0.11) and BD (τ=0.20)(xr2 =0.58, p<0.05) (Fig. 3d, e).
Water clarity (τ=1.08) was the only environmental
variable retained in the final NPMR model, but had little
influence on %P (Fig. 3f).

Discussion
Abundance of juvenile American eels differed among
sites and can depend on salinity levels at the sites
(Tosi et al. 1990; Edeline et al. 2005). Residence of
Fig. 2 Catch per unit effort
(CPUE) of juvenile

American eel (Anguilla
rostrata) collected in
Newport Bay (Maryland,
U.S.A.). Five sites were
sampled: MC, Marshall
Creek; HL, Hayes Landing;
TB, Trappe Creek at Trappe
Bridge; TC, Trappe Creek;
and AC, Ayers Creek.
Different superscript letters
above standard deviation bars
indicate significantly different means among sites
determined with analysis of
variance (see legend).
Asterisks indicate the significant difference in CPUE
between years at a site

American eel in brackish water can be highly variable
within an estuary (Sullivan et al. 2009). Two of the
three brackish sites surveyed had eel CPUE’s with
relatively large standard deviation bars, and therefore
more variable populations over time. Migration of
anguillid eels from brackish water to upstream
habitats begins when water temperature exceeds
10°C (Overton and Rulifson 2009; Sullivan et al.
2009) and during high spring tides (Sorensen and
Bianchini 1986). Some young eels may spend up to a
year in the estuary before entering freshwater, thereby
maximizing their growth and contribution to the
spawning population (Jessop et al. 2002).

Abundance was highest when habitats were oligohaline (0.3–3.2 ppt) and declined with increasing
salinity or extremely low salinities (≤0.1 ppt). A
relatively small proportion of glass eels constitute a
freshwater contingent (Edeline et al. 2005), which


24

Environ Biol Fish (2011) 90:19–27

Fig. 3 Catch per unit effort (CPUE) and pigmentation levels of
American eel (Anguilla rostrata) vary with environmental
gradients in Newport Bay (Maryland, U.S.A.). Partial model
CPUE (open symbols) and predicted model CPUE values
(shaded symbols) that were produced by nonparametric

multiplicative regression are plotted against: a dissolved
oxygen (DO), b salinity, and c water clarity. Pigmentation
values are plotted against: d water temperature, e body depth,
and f water clarity

possibly explains the low catches in these habitats.
Abundance may also be low during high levels of
stream discharge (Edeline et al. 2005; Acou et al.
2009; Overton and Rulifson 2009) or because odors
stimulate dispersal (Creutzberg 1961; Sorenson 1986;
Tosi et al. 1988; Sola 1995). Such odors are emitted
by substrates such as wood debris (Chisnall et al.
2002; Silberschneider et al. 2004). We found little
evidence to support a relationship between substrate

composition and CPUE. Interestingly, few anguillid
eels were collected at Trappe Bridge and Trappe Creek,
which were also associated with the highest levels of
woody debris and snag habitat (JWL pers. obs.).

Settlement levels were similar across a broad range
of DO values, but were highest in habitats with DO
levels near or exceeding 8 mg·L−1 (i.e., normoxia).
While extremely hypoxic environments may deter
settlement (Baker and Mann 1992), these extreme
values were rarely encountered in this study. Pigmented or yellow-phase American eels are not
efficient at aerial respiration, but tend to have a high
tolerance for hypoxemia (Hyde et al. 1987). Even so,
they exhibit a strong preference for oxygen-rich water
(Facey and Van Den Avyle 1987).
Pigmentation levels were not related to environmental conditions and did not differ among sites. Our


Environ Biol Fish (2011) 90:19–27

25

Fig. 4 Pigmentation of
juvenile American eel
(Anguilla rostrata) averaged
for each stage defined by
Haro and Krueger (1988).
Different superscript letters
above standard deviation
bars indicate significantly

different means determined
following an analysis of
variance model (see inset)

novel method of measuring pigmentation generally
followed that expected from staging, especially for
the majority of eels surveyed here (stages 2–5), and
provided a quantitative means for determining how
environmental variation was correlated with differences in %P. Others have found that warmer water
temperatures were positively correlated with growth,
metabolism, and pigmentation in Atlantic and Pacific
anguillids (Haro and Krueger 1988; Dou et al. 2003;
August and Hicks 2006). While we found that %P
was higher in warm water, water temperature increased over time and independent regressions of
temperature and pigmentation for each week of the
study did not show any significant relationships
(JWL, unpubl. data). Sample sizes for these independent regressions were small (number of sites = 5) and
a broader spatial distribution of sites may be
necessary.
Pigmentation stage in anguillids is generally
thought to be independent of growth, with pigmentation occurring while negative growth is
taking place (Haro and Krueger 1988; Dou et al.
2003). For the stages analyzed here, we found that %
P increased with size of the fish (i.e., BD) and time,
suggesting that pigmentation during stages measured here could be a suitable measure for growth
of juvenile American eel. In contrast to earlier
studies that staged American eel according to
pigment level (Haro and Krueger 1988), our
method yielded a continuous variable suitable for

general linear and nonlinear modeling. One problem
with the method, however, was that intensity of the
pigment could not be considered. Quantifying
intensity would likely have been useful for distinguishing among later developmental stages of pigmentation. Another problem with the method could
be that the type of preservative to fix tissues may
affect the measured level of pigmentation. This
problem, however, should not affect results of our
study as the same preservative was used to fix all
samples.
Elucidating the early life history of American eels
is an important step in managing the species. Here,
we highlight the importance of oligohaline habitats
for young American eel. The species utilizes a wide
range of salinities over the lifetime of an individual,
particularly estuaries and oligohaline habitats where
the majority of life is spent. Current climatic patterns,
such as El Niño events, may affect the salinity
gradients differently among estuaries of eastern North
America (Chigbu et al. 2009). Saltwater intrusion to
oligohaline habitats can adversely affect the distributions of freshwater-dependent species (Love et al.
2008) and coastal fish assemblages (Love et al.
2009b). Thus, climate change may not only affect
the survivorship of anguillid lepotcephalus larvae
(Bonhommeau et al. 2008), but also the distributions
of juveniles settling in estuaries, and consequently
population-level traits such as sex ratios (Krueger and
Oliveira 1999).