J. Great Lakes Res. 29 (Supplement 2):243–257 Internat. Assoc. Great Lakes Res., 2003 Stable Isotope Analyses of Food Web Structure and Fish Diet in Napoleon and Winam Gulfs, Lake Victoria, East Africa Linda M. Campbell 1,* , Robert E. Hecky 1 , and Sylvester B. (S.B.) Wandera 2 1 Department of Biology University of Waterloo 200 University Avenue West Waterloo, Ontario N2L 3G1 2 Fisheries Resources Research Institute P. O. Box 343 Jinja, Uganda ABSTRACT. The food web structures in Napoleon and Winam gulfs, Lake Victoria, were characterized using stable nitrogen and carbon isotope analyses. Similar biota in Napoleon Gulf had significantly lighter δ 15 N values and heavier δ 13 C values than similar biota in Winam Gulf, indicating different basal isotopic values. In both gulfs, Nile perch (Lates niloticus) was the top trophic predator while Nile tilapia (Oreochromis niloticus) was littoral and feeding at lower trophic levels. Rastrineobola argentea and Yssichromis laparograma had surprisingly high δ 15 N values, close to those of Nile perch, which were not consistent with the high isotopic values of their assumed zooplankton prey. Caridina nilotica, a freshwater shrimp, had a wide range of δ 13 C values but low δ 15 N values, consistent with their appearance in nearly all habitants in the lake, and their presence in the stomaches of most fish species. Nile perch showed an increase in δ 15 N and δ 13 C values with size, signifying that piscivory increases and their dietary reliance on invertebrates decreases as they mature. Stable isotope values for Napoleon Gulf biota which were adjusted for different basal values were not statistically different from those of Winam Gulf biota, sug- gesting that stable carbon and nitrogen isotopes fractionate consistently through trophic transfers in Lake Victoria. The stable isotope data illustrate a short food web, with the top predator Nile perch feed- ing on a restricted set of fish and macroinvertebrate species, including its own young. INDEX WORDS: Stable isotopes, Nile tilapia, Nile perch, Lake Victoria, aquatic food webs. INTRODUCTION Lake Victoria has experienced dramatic changes in recent times, including eutrophication and deoxy- genation, (Hecky 1993, Hecky et al. 1994), the extir- pation of native cichlid species (Witte et al. 1992), an increase in Nile perch (Lates niloticus), Nile tilapia (Oreochromis niloticus), and Rastrineobola argentea fisheries (SEDAWOG 1999) and the upsurge and de- cline of water hyacinth, Eichhornia crassipes (Twongo 1996; R.E. Hecky pers. comm.). The food web structure in the lake is shifting, based on the volume and composition of the fish catch in the lake as well as on the stomach contents of fish (Balirwa * Corresponding author. E-mail: [email protected]Current address: Canada Centre for Inland Waters, Environment Canada, 867 Lakeshore Road, Burlington, Ontario L7R 4A6. 243 1998, Ogutu-Ohwayo 1995, Wanink 1998). There has to date been little actual quantification of the trophic interactions within the system beyond diet studies for selected species which are limited in spa- tial and temporal coverage. An understanding of the current food web structure in the lake, including site- to-site variability, is essential to support current and future management initiatives. This information will also complement on-going studies attempting to quantify the impacts of changing fish community composition on the economic and social well being of the people living in Lake Victoria’s watershed. In this study, the food web structures in two gulfs of Lake Victoria, Winam (Kenya) and Napoleon (Uganda), are quantified using stable isotope rela- tionships. The two gulfs have similar population and human usage patterns, although the morphometery
Transcript
J. Great Lakes Res. 29 (Supplement 2):243–257Internat. Assoc. Great Lakes Res., 2003
Stable Isotope Analyses of Food Web Structure and Fish Diet in Napoleon and Winam Gulfs, Lake Victoria, East Africa
Linda M. Campbell1,*, Robert E. Hecky1, and Sylvester B. (S.B.) Wandera2
1Department of BiologyUniversity of Waterloo
200 University Avenue WestWaterloo, Ontario N2L 3G1
2Fisheries Resources Research InstituteP. O. Box 343Jinja, Uganda
ABSTRACT. The food web structures in Napoleon and Winam gulfs, Lake Victoria, were characterizedusing stable nitrogen and carbon isotope analyses. Similar biota in Napoleon Gulf had significantlylighter δ15N values and heavier δ13C values than similar biota in Winam Gulf, indicating different basalisotopic values. In both gulfs, Nile perch (Lates niloticus) was the top trophic predator while Nile tilapia(Oreochromis niloticus) was littoral and feeding at lower trophic levels. Rastrineobola argentea andYssichromis laparograma had surprisingly high δ15N values, close to those of Nile perch, which were notconsistent with the high isotopic values of their assumed zooplankton prey. Caridina nilotica, a freshwatershrimp, had a wide range of δ13C values but low δ15N values, consistent with their appearance in nearlyall habitants in the lake, and their presence in the stomaches of most fish species. Nile perch showed anincrease in δ15N and δ13C values with size, signifying that piscivory increases and their dietary relianceon invertebrates decreases as they mature. Stable isotope values for Napoleon Gulf biota which wereadjusted for different basal values were not statistically different from those of Winam Gulf biota, sug-gesting that stable carbon and nitrogen isotopes fractionate consistently through trophic transfers inLake Victoria. The stable isotope data illustrate a short food web, with the top predator Nile perch feed-ing on a restricted set of fish and macroinvertebrate species, including its own young.
INDEX WORDS: Stable isotopes, Nile tilapia, Nile perch, Lake Victoria, aquatic food webs.
INTRODUCTION
Lake Victoria has experienced dramatic changes inrecent times, including eutrophication and deoxy-genation, (Hecky 1993, Hecky et al. 1994), the extir-pation of native cichlid species (Witte et al. 1992), anincrease in Nile perch (Lates niloticus), Nile tilapia(Oreochromis niloticus), and Rastrineobola argenteafisheries (SEDAWOG 1999) and the upsurge and de-cline of water hyacinth, Eichhornia crassipes(Twongo 1996; R.E. Hecky pers. comm.). The foodweb structure in the lake is shifting, based on thevolume and composition of the fish catch in the lakeas well as on the stomach contents of fish (Balirwa
*Corresponding author. E-mail: [email protected] address: Canada Centre for Inland Waters, Environment Canada,867 Lakeshore Road, Burlington, Ontario L7R 4A6.
243
1998, Ogutu-Ohwayo 1995, Wanink 1998). Therehas to date been little actual quantification of thetrophic interactions within the system beyond dietstudies for selected species which are limited in spa-tial and temporal coverage. An understanding of thecurrent food web structure in the lake, including site-to-site variability, is essential to support current andfuture management initiatives. This information willalso complement on-going studies attempting toquantify the impacts of changing fish communitycomposition on the economic and social well beingof the people living in Lake Victoria’s watershed. Inthis study, the food web structures in two gulfs ofLake Victoria, Winam (Kenya) and Napoleon(Uganda), are quantified using stable isotope rela-tionships. The two gulfs have similar population andhuman usage patterns, although the morphometery
244 Campbell et al.
and water chemistry differ (Table 1). Winam Gulf isan isolated gulf nearly closed to the main lake, andits waters have a higher conductivity (177µmho/cm). Napoleon Gulf is open to the lake, is con-stantly flushed by the Nile River outflow, and haslower conductivity (101 µmho/cm) which is similarto Lake Victoria proper.
Characterization of trophic levels and food webstructure has traditionally been based on dietaryanalyses of fish stomach contents. While dietaryanalysis provides valuable taxonomic informationon fish diets, they can be complemented by theanalysis of stable nitrogen (δ15N) and carbon (δ13C)isotope ratios of biota to characterize food webstructure and trophic interactions (Peterson and Fry1987). δ13C and δ15N values integrate long-term di-etary patterns and can be used as numerical vari-ables for statistical analyses. Stable isotopemeasurements have been successfully used to esti-mate feeding patterns and food web structure inseveral African lakes, including Lakes Kyoga inUganda (Hecky and Hesslein 1995) and Malaw∆i insouthern Africa (Bootsma et al. 1996, Genner et al.1999, Hecky and Hesslein 1995, Kidd et al. 2001).
Stable nitrogen isotopes are useful in determining
the relative trophic position of biota. Nitrogen iso-topes consistently fractionate in organisms: 14N isselectively eliminated while 15N is incorporatedinto body tissues. Consequently, with each succes-sive trophic transfer, δ15N values in the tissue ofbiota increase (become “heavier”). Many studiesfind that the average δ15N difference between ananimal and its food source is approximately 3 to4‰ (DeNiro and Epstein 1981, Vander Zanden andRasmussen 2001). This consistent change providesa powerful analytical tool to quantify relativetrophic position, which can also be correlated withcontaminant bioaccumulation or dietary changes infish (Cabana and Rasmussen 1994).
In contrast, stable carbon isotopes fractionate verylittle in biota, with around 1‰ enrichment in δ13Cper trophic level (Peterson and Fry 1987, VanderZanden and Rasmussen 2001). Because of these lowfractionation rates, the stable carbon isotope valuesof organisms reflect the average δ13C of their diets.δ13C values can vary at the base of the food web dueto differences in photosynthetic enzymatic fixation,growth rates, CO2 and pH levels (Hecky andHesslein 1995). Because free-floating pelagic algaeclose to the water surface have access to a large dis-
TABLE 1. Selected water quality parameters for Lake Victoria, Napoleon Gulf, and Winam Gulf. Super-scripts indicate the original data source of each parameter. Personal data (oxygen, Secchi depth, pH, con-ductivity, and temperature) were collected during the study period. Land use parameters and biologicaloxygen demand (BOD) are presented to give an idea of human impacts and possible pollution. Note thatall references from Scheren et al. 2000 include the whole shoreline in each country, not just specificgulfs.
Napoleon WinamParameters Gulf Gulf
Chlorophyll-a (mg/m3)—Dry season (June to Sept.) 22.1–51.4 1 9.3–21.0 2*
Present population density (people/km2) 246 (projected 7 2598
% of dry land occupied by large-scale farms 27 119
Estimated % of catchment that is cultivated 40.0%10 30.2%10
Estimated production of industries in catchment (tonnes*/yr) 151,82010 875,77010
Most likely total BOD loading (tonnes*/yr) 4,54010 7,51010
Estimated % of BOD loading from domestic sources 75.9%10 88.1%10
1 Muggide 1992; 2 Lung’ayia et al. 2000 (*extreme value of 71.5 excluded); 3 Ramlal et al. 2001; 4 Personal data; 5 Lehman and Branstrator 1994; 6 Gophen et al. 1995; 7 The Republic of Uganda 1999 Statistical Abstract. 2000; 8 Re-public of Kenya 2000 Economic Survey. 2000; 9 Republic of Kenya 1997; 10 Scheren et al. 2000*metric tons
Food Web Structure in Northern Lake Victoria 245
solved CO2 reservoir which fractionate only slightly,photosynthetic fixation can result in “lighter” (morenegative) δ13C values of –29‰ (Hecky and Hesslein1995). At the sediment-water interface growth in aboundary layer restricts the CO2 reservoir available,and the carbon-limited benthic algae are less isotopi-cally discriminating, resulting in “heavier” (morepositive) δ13C signatures which may be between –25to –10‰ (Hecky and Hesslein 1995, Bootsma et al.1996). Emergent macrophytes obtain their carbonfrom the atmosphere, and their δ13C values are influ-enced by their photosynthetic pathway (C-3 or C-4)and the need to conserve water (Hecky and Hesslein1995). C-4 plants such as the tropical aquatic hippograss Vossia spp. and papyrus Cyperus papyrus areless discriminatory against 13CO2, which is reflectedby their heavier δ13C values (typically –12 to –14‰).C-3 plants, such as water hyacinth, are more discrim-inatory against 13CO2 so their δ13C values tend to belighter (typically –26 to –28‰). As a result, there aredifferences in δ13C values between organisms withina food-web based on different sources of primaryproduction. These differences are passed up in thefood chain, indicating the origin of organic carbon inorganisms at higher trophic levels (Hecky andHesslein 1995). This difference can be used to quan-tify the relative importance of pelagic versus benthicalgal sources in an organism’s diet (Bootsma et al.1996), and to determine numerically the changes incarbon sources both over time and with growth.
METHODS
Napoleon Gulf is situated in southeasternUganda, and leads to the source of the Nile Riverlocated near the town of Jinja (Fig. 1). The town islightly industrialized, with the Owen Falls Dam hy-droelectric facility on the Nile River, a brewery, asugarcane processing plant, a textile factory andseveral fish processing plants. The population den-sity in districts around the gulf is estimated at 246people per square kilometer (Table 1). NapoleonGulf is eutrophic (Table 1), with a highly convo-luted shoreline and numerous bays which provide arange of aquatic ecosystems from wetland topelagic. Samples were collected in the vicinity ofJinja Bay and Buvuma Channel, which share simi-lar environmental characteristics (Fig. 1).
Winam (Nyanza) Gulf in western Kenya is ashallow mesotrophic to eutrophic gulf (Table 1)nearly closed off to the main lake (Fig. 1). Kisumu,the main town in the region, is heavily industrial-ized (paint, solvent, and plastics manufacturing aswell as sugar, food, and fish processing). Sewageeffluent, usually untreated, enters the Kibos Rivernear its mouth at Lake Victoria. The populationdensity for the areas bordering the gulf is estimatedat 259 people per square kilometer (Table 1). TheWinam Gulf shoreline is less convoluted than thatof Napoleon Gulf, and the waters are well mixedthroughout the gulf. Samples for this study weretaken from outside the Kisumu region toward the
FIG. 1. Location of Napoleon Gulf (Uganda) and Winam Gulf (Kenya) in northern Lake Victoria,and the location of Lake Victoria in Africa. Regions bordered by thick lines indicate the samplingarea.
246 Campbell et al.T
AB
LE
2.
Fis
h,
inve
rteb
rate
an
d or
gan
ic m
atte
r co
llec
ted
in t
his
stu
dy w
ith
th
e as
sign
ed c
odes
use
d in
th
e fi
gure
s. T
he
nu
mbe
r of
sam
ples
fro
m e
ach
loc
atio
n t
hat
wer
e an
alys
ed a
re l
iste
d fo
r ea
ch g
ulf
, al
ong
wit
h t
he
mea
n v
alu
es ±
stan
dard
dev
iati
on f
or s
tabl
en
itro
gen
(�
15N
) an
d ca
rbon
(�
13C
) is
otop
e va
lues
an
d pu
blis
hed
die
t in
form
atio
n (
Cop
ley
1958
, G
reen
woo
d 19
66,
Gre
enw
ood
1981
,va
n O
ijen
199
5, a
nd
Bal
irw
a 19
98).
Th
ose
sepa
rate
d by
sla
shes
indi
cate
die
tary
ch
ange
s oc
curr
ing
wit
h g
row
th a
nd
mat
uri
ty.
Nap
oleo
n G
ulf
Win
am G
ulf
Spe
cies
Cod
en
δ15 N
δ13 C
nδ1
5 Nδ1
3 CP
ubli
shed
die
t co
mpo
siti
on
Org
anic
mat
ter
Cyp
erus
pap
yrus
U2
0.2
– 0.
6–1
2.7
– –1
2.6
~~
~P
hoto
synt
hesi
s (c
omm
on n
ame:
pap
yrus
)E
ichh
orni
a cr
assi
pes
W2
2.7
– 3.
4–2
7.8
– –2
6.4
~~
~P
hoto
synt
hesi
s (c
omm
on n
ame:
wat
erhy
acin
th)
Sus
pend
ed f
loc
F2
2.7
– 3.
2–1
9.3
– –1
9.0
~~
~P
hoto
synt
hesi
sP
hyto
plan
kton
(ne
arsh
ore)
K3
4.3
±0.
8–1
9.9
±0.
7~
~~
Pho
tosy
nthe
sis
Vos
sia
spp.
V2
0.5
– 0.
8–1
2.8
– –1
2.4
~~
~P
hoto
synt
hesi
s (c
omm
on n
ame:
hip
pogr
ass)
Inve
rteb
rate
sC
arid
ina
nilo
tica
C17
4.4
±0.
9–1
9.0
±1.
06
7.2
±0.
7–2
2.8
±0.
4B
enth
ic a
lgae
, alg
al d
etri
tus,
zoo
plan
kton
Sm
all
mus
sels
M~
~~
27.
1 –
7.8
–24.
05 –
–23
.4S
uspe
nded
par
ticu
late
s ne
ar b
enth
icsu
rfac
es
Eph
emer
opte
ra n
ymph
E5
2.6
±0.
3–1
8.9
±0.
210
6.7
±0.
7–2
3.9
±0.
4A
lgae
(co
mm
on n
ame:
may
flie
s, l
ake-
flie
s)O
dona
ta n
ymph
D7
5.3
±0.
8–1
8.6
±0.
85
6.0
±0.
4–2
7.0
±1.
1A
lgae
, zoo
plan
kton
, inv
erte
brat
es (
com
-m
on n
ame:
dra
gonf
lies
)P
roso
bran
ch s
nail
N1
5.6
–18.
9~
~~
Wat
er h
yaci
nth
epip
hyto
nZ
oopl
ankt
on (
near
shor
e)Z
37.
1 ±
0.3
–17.
6 ±
0.8
~~
~S
uspe
nded
par
ticu
late
sZ
oopl
ankt
on (
offs
hore
)Z
27.
7 –
7.8
–20.
6 –
–20.
4~
~~
Sus
pend
ed p
arti
cula
tes
Fis
hB
agru
s do
cmac
F.
B~
~~
112
.7–2
1.3
Lar
ge i
nver
tebr
ates
, Car
idin
a /
smal
l fi
shC
lari
as g
arie
pinu
sB
.G
~~
~1
8.78
–25.
02S
mal
l fi
sh, m
ollu
scs,
ins
ect
larv
ae, a
ndpl
ant
detr
itus
Hap
loch
rom
issp
p.H
29.
1 –
9.3
–18.
5 –
–18.
15
9.2
±1.
1–2
4.1
±0.
5P
resu
med
ins
ecti
vore
s (s
peci
es u
nkno
wn)
Lat
es n
ilot
icus
L.
L12
8.0
±1.
1–1
8.9
±0.
517
11.8
±1.
4–2
2.2
±1.
2C
arid
ina
/ R
. arg
ente
a, s
mal
ler
Nil
e pe
rch
/la
rger
fis
hO
reoc
hrom
is n
ilot
icus
L.
O11
5.9
±0.
8–1
8.1
±0.
813
9.8
±1.
1–2
2.3
±1.
9P
hyto
plan
kton
, det
ritu
s, C
arid
ina,
chi
rono
-m
ids,
mol
lusc
sIm
mat
ure
O. n
ilot
icus
O’
57.
0 ±
0.3
–17.
1 ±
0.7
46.
1 ±
0.5
–18.
6 ±
1.0
Phy
topl
ankt
on, b
enth
ic a
lgae
, det
ritu
sP
roto
pter
us a
ethi
opic
usH
.P
27.
3 –
7.7
–19.
5 –
–19.
33
10.6
±0.
4–2
3.8
±0.
5In
sect
s /
Gas
trop
oda,
Car
idin
a, s
mal
l fi
shR
astr
ineo
bola
arg
ente
aP
.R
88.
1 ±
0.6
–17.
0 ±
0.7
311
.3 ±
1.4
–23.
8 ±
0.5
Cop
epod
s /
chir
onom
ids
and
chao
bori
ds /
smal
l C
arid
ina
Schi
lbe
inte
rmed
ius
L.
S~
~~
99.
5 ±
0.7
–27.
4 ±
1.5
Odo
nata
/ s
mal
l fi
sh (
R. a
rgen
tea
and
youn
g N
ile
perc
h)
Syno
dont
is a
frof
isch
eri
H.
A~
~~
78.
8 ±
1.5
–27.
8 ±
3.0
Chi
rono
mid
s /
mol
lusc
sT
ilap
ia z
illi
G.
T1
8–1
9.2
~~
~P
hyto
plan
kton
, ben
thic
alg
ae, m
acro
phyt
esY
ssic
hrom
is l
apar
ogra
ma
G.
Y5
7.9
±0.
2–1
9.9
±0.
32
10.1
– 1
0.5
–26.
7 –
–24.
7Z
oopl
ankt
on o
r pe
lagi
c ch
iron
omid
s an
dch
aobo
rids
Food Web Structure in Northern Lake Victoria 247
mouth of the gulf, excluding Asembo and Homabays (Fig. 1).
The species sampled included fish, invertebrates,macrophytes, and phytoplankton (Table 2). Com-mercially important Nile perch (L. niloticus) andNile tilapia (O. niloticus) were collected from bothgulfs, along with two pelagic fish species (thecyprind R. argentea and the haplochromid Ys-sichromis laparograma), and the freshwater shrimp,Caridina nilotica. Other taxa were included when-ever possible. Fish were obtained by trawling theregion and supplemented by overnight gill net setsand fish purchased directly from the local fisher-men. All fish were dissected, and for the large fish,a 10-cm3 muscle sample was collected from the lat-eral muscle. Smaller fish were filleted. Near-shoreand offshore zooplankton in Napoleon Gulf werecollected with a Schindler trap and filtered througha mesh. Phytoplankton were collected with a 53 µmnet. Suspended floc were collected off the surfacesof mesh bags suspended at 0.5 below the water sur-face near the Jinja Pier. Macrophytes collected inNapoleon Gulf were used as terrestrial endpointspotentially contributing to the lake’s organicsources. Because samples were collected for bothstable isotope and mercury analyses (Campbell etal. 2003), trace-metal clean protocols were fol-lowed during sampling. It was not possible to main-tain the same sampling protocol and effort for eachsite due to logistical constraints, but every attemptwas made to collect a broad and diverse sample set.Samples were wrapped in hydrochloric acid-cleaned aluminium foil, double-wrapped in Ziploc®
bags, and frozen. The samples, transported toCanada on ice, were still frozen on arrival.
Small sub-samples of fish tissue and whole inver-tebrates were freeze-dried and ground into fine pow-der for δ15N and δ13C analyses using a MicromassVG-Isochrom Continuous Flow Isotope Ratio MassSpectrometer (CF-IRMS) at the Environmental Iso-tope Laboratory, University of Waterloo. The ratiosof the stable isotopes were then measured against thereference standards PeeDee belemnite for δ13C andthe nitrogen gas in ambient air for δ15N (Eqn. 1).The delta notation (δ), the difference (‰, or parts perthousand) between the isotopic ratio of the sampleand the standard, was calculated as:
δ13C or δ15Ν = {(Rsample – Rstandard)/(Rstandard)} × 1000 (1)
where R= 13CO2 / 12CO2 for δ13C or R= 15N2 / 14N2 for δ15N
Working standards used to determine inter- andintra-run variation and accuracy of the results in-cluded the International Atomic Energy Agency(IAEA) standards CH6 (δ13C = –10.4‰), N1 (δ15N= 0.36‰) and N2 (δ15N = 20.3‰), and the in-housestandards: EIL-70 (powdered lipid-extracted LakeOntario walleye; δ13C = –19.34‰, δ15N = 16.45‰)and EIL-72 (powdered Whatman cellulose fiber;δ13C = –25.4‰). Replicate Nile perch sampleswere included in every run to determine between-run variation. Standard deviations for the standardswere ± 0.3‰ for δ15N and ± 0.2‰ for δ13C, andstandard deviations of replicate samples were ±0.16‰ for δ15N and ± 0.24‰ for δ13C.
The food web structure was graphically repre-sented by plotting δ15N against δ13C for all organ-isms collected from the two gulfs. Dietaryinformation for all fish and invertebrate species wascompiled from stomach content data and publishedsources to compare information derived from stableisotope analyses and stomach content data. Statisti-cal analyses were performed using SYSTAT version8.0 for Windows (SPSS Inc.). T-tests were done todetermine if the stable isotope values were signifi-cantly different in adult Nile perch and Nile tilapiabetween the two gulfs and to compare the stableisotope values of R. argentea and Y. laparograma.
Adult Nile tilapia and Nile perch δ13C and δ15Νvalues were regressed against their total length (TL)to determine if there were isotopic changes with in-creasing fish size. TL ranged from 9 to 90 cm foradult Nile perch and from 15 to 60 cm for adultNile tilapia. An extremely large Nile perch (TL =158 cm) from Winam Gulf was excluded as an out-lier. Grouped t-tests were used to demonstrate that“immature” Nile tilapia (≤ 5 cm) δ13C and δ15Νvalues were different from those of adult Niletilapia. Comparison of regression slopes and inter-cepts were completed in an ANCOVA model to de-termine if there were significant differences instable isotope changes with adult fish growth ineach gulf.
To determine the difference in basal δ13C andδ15Ν values between the two gulfs, the δ13C andδ15Ν values of selected taxa were compared, basedon their assumed trophic position. Estimating theextent of the basal isotopic differences provides ameans of adjusting the isotopic signatures of biotafrom one gulf to the basal values of another gulf.This could permit the direct comparison of foodweb structure and other aspects such as the trophictransfer of contaminants (Campbell et al. 2003).Taxa were selected by the following criteria: shar-
248 Campbell et al.
ing overlapping δ13C values, analogous taxa foundin both gulfs, and relative food web position in bothgulfs. The biota selected included Caridina, Niletilapia, P. aethiopicus (lungfish), and Nile perch.Taxa were assigned a numerical value representingtheir “relative” trophic position, based on theirknown dietary patterns, which gradually rangedfrom planktivory to piscivory (Table 3). Caridina,an invertebrate commonly consumed by many fishspecies, was assigned the lowest trophic position(1) and Nile perch, a top predator, was assigned thehighest trophic position (4). The δ13C and δ15Ν val-ues for each taxon were then regressed against itsassigned numeric trophic position. The differencesbetween the slopes and intercepts were tested inANCOVA and the difference between the interceptsfor each regression (δ13C and δ15Ν vs. trophic posi-tion) was used to “adjust” Napoleon biota stableisotope values to those of Winam Gulf.
RESULTS
Plotting δ15Ν values against δ13C values providesa visual characterization of the food web structureand can be used to assess predator-prey and cohortrelationships (Fig. 2). The range of δ13C and δ15Νvalues for Nile perch is wide, particularly in WinamGulf, with these values extending over as much as4‰ for δ13C and 5‰ for δ15Ν. The range is sup-ported by stomach content information, indicating abroad diversity of prey items ranging from Carid-ina to large fish (Table 2). In both gulfs, Nile perchstable isotope values overlap with many other fishspecies, including Nile tilapia and lungfish, al-though Nile perch tend to have higher δ15Ν values
and their δ13C values are midway along the rangeof available prey organisms (Fig. 2, Table 2). Theδ15Ν values for Nile tilapia, lungfish and mostpelagic haplochromids generally place these fishspecies between Nile perch and most invertebrates,including Caridina, Ephemeroptera, and Odonata(Table 2). The in δ13C values for these fish speciesare varied, with Nile tilapia showing the most posi-tive values, lungfish having median values and hap-lochromids having the most negative values. Thelow δ15Ν values for most invertebrates and algae(including phytoplankton and floc) places them in alower trophic position relative to fish (Fig. 2). In
TABLE 3. Trophic numbers assigned to eachtaxon common to both Gulfs, based on diet. Thetaxon codes are in Table 2.
Assigned trophic Reason for assigning
Taxon number trophic number
L 4 Top predator in Lake Victoria (fish and C. nilotica)
P 3 Feeds upon both fish and inverte-brates (C. nilotica, snails andmollusks)
O 2 Strong preference for phyto-plankton detritus, but will eat C.nilotica
C 1 Zooplankton and phytoplankton
FIG. 2. The relationship between �15N, indicat-ing trophic position and �13C, indicating dietarycarbon source, for 21 taxa from Winam andNapoleon gulfs. Each taxa is represented by acode defined in Table 2. Note that the x-axis scalesof the two figures are different, reflecting thewider range of �13C values in Winam Gulf.
Food Web Structure in Northern Lake Victoria 249
Winam Gulf, Schilbe intermedius and Syndodontisafrofischeri both have δ13C values that place themon the extreme negative end of the range of thefood web (Fig. 2). T-test analyses indicate that thehigh δ15Ν values of Y. laparograma and R. argen-tea (Fig. 2, Table 2) are not significantly differentthan δ15Ν values of Nile perch in both gulfs (Table4). The δ13C values for Nile perch are significantlydifferent from those for the pelagic fish species inWinam Gulf. In Napoleon Gulf, the δ13C values forY. laparograma and Nile perch are significantly dif-ferent from each other (Table 4).
The difference between the two gulfs is visuallydiscernible by the shift of Winam Gulf δ15N valuesto the right (Fig. 3A), and the lighter biotic δ13Cvalues (and also wider range as indicated by widerstandard deviation values) in similar species inWinam Gulf (Table 2, Fig. 3B). Furthermore, Nileperch and Nile tilapia in Napoleon Gulf have signif-icantly lighter δ15N and heavier δ13C values thanthe same fish species in Winam Gulf (Table 5). Re-gressing the δ13C and δ15N values of common taxaagainst their assigned numerical trophic position(Fig. 4) illustrates the similarities and differencesbetween the gulfs. The δ15N regressions (Fig. 4A,Eqns. 2 and 3) are significant for both gulfs (p, ≤0.001). The ANCOVA results indicate that the in-tercepts are significantly different (p, ≤ 0.000),while the slopes are similar (p, 0.604).
TABLE 4. Results of grouped t-tests for �15N and δ13C values for Napoleon andWinam Gulfs: between adult Nile perch (L), R. argentea (R) and Y. laparograma (Y)and between immature Nile tilapia (O′) and adult Nile tilapia (O). Mean values aregiven in Table 2. Significance is determined at p-value 0.05 and the alpha value is0.027.
FIG. 3. Mean ± s.d. of �15N (A) and �13C (B)values in fish, invertebrates, and plants inNapoleon and Winam gulfs, Lake Victoria. Notethat �15N values for biota in Winam Gulf tend tobe higher than for similar biota in Napoleon Gulf.Note that �13C values for biota in Winam Gulfextend over a broader range and are lighter thanthose than for similar biota in Napoleon Gulf.
250 Campbell et al.
δ13C values did not change significantly betweenassigned trophic values. The slopes of theδ13C:trophic position regressions are not signifi-cantly different (p, 0.394) while the intercepts aresignificantly different (p, ≤ 0.001).
The difference between the two intercepts forWinam and Napoleon gulfs (δ15N = 3.14; δ13C =4.02), provides a means of “adjusting” for the dif-ferent values at the base of the food-web of onegulf to the other.
Using the intercept differences to “adjust”Napoleon Gulf stable isotope values to Winam Gulfvalues, a new set of stable isotope data was gener-ated. Figure 5A compares the original stable iso-tope values of the two food-webs and 5B showshow the adjusted values for Napoleon Gulf becomesimilar to Winam Gulf. δ15N and δ13C values afteradjustment are not significantly different for bothNile perch and Nile tilapia (Table 5). The apparentdifferences in δ15N and δ13C values of the samespecies between the gulfs are largely eliminated bythe correction for basal signatures.
Regressing stable isotope data against fish sizeenables the statistical interpretation of dietary shiftsduring fish growth. Nile perch length is positivelycorrelated with δ15N in both Napoleon ( p, ≤ 0.001)and in Winam ( p, 0.013; Table 6, Fig. 6A), indicat-ing that Nile perch increase their trophic level asthey mature. Dietary shifts are also indicated bysignificant correlations between Nile perch lengthand δ13C in Napoleon ( p, 0.027) and in Winam ( p,0.018; Table 6, Fig. 6B). ANCOVA analyses indi-cate that the slopes for δ15N and δ13C vs. Nileperch TL are not significantly different between thetwo gulfs ( p, 0.903 and 0.702, respectively) but theintercepts are significantly different with p, ≤ 0.001for both gulfs. Note the outlier points in Figures 6A
TABLE 5. Results of grouped t-tests for comparison of �15N and �13C values of Nileperch and Nile tilapia between Napoleon and Winam gulfs. The original �15N and �13Cvalues of both fish species are compared between the two gulfs, and the adjusted �15Nand �13C values for Napoleon Gulf (see text) are compared with original Winam Gulfvalues. Mean values are given in Table 2. Significance is determined at p-value = 0.05and the alpha value is 0.027.
FIG. 4. �15N (A) and �15C (B) values of selectedfish and Caridina versus their assigned numerictrophic position.
Food Web Structure in Northern Lake Victoria 251
and B which represent the 158-cm long Nile perchfrom Winam Gulf—its δ13C value and particularlyits δ15N value drops below the trend-line for theother Nile perch.
Adult Nile tilapia show a small but significantcorrelation between δ15N and length ( p, 0.022 and0.028 in Napoleon and Winam Gulf respectively;Table 6). However, the correlation between Niletilapia δ13C and length is not significant inNapoleon Gulf ( p, 0.903) but is significant inWinam Gulf ( p, 0.032; Table 6). The slopes for
δ15N vs. TL are not significantly different betweenthe gulfs ( p, 0.086) but the intercepts are ( p,0.972). Due to the lack of significance for δ13C vs.TL in Napoleon Gulf, an ANCOVA for δ13C vs. TLbetween the gulfs was not carried out. ImmatureNile tilapia stable isotope values are usually segre-gated from adult Nile tilapia in both gulfs; adultNile tilapia usually have significantly heavier δ13Cvalues and their δ15N values are significantly differ-ent (higher in Winam and lower in Napoleon) fromthose of immature Nile tilapia (Table 4).
DISCUSSION
The basal stable isotope values in Winam andNapoleon gulfs are different. Most of the externalnutrient input of total nitrogen to the main body ofLake Victoria (including Napoleon Gulf) comesfrom biological N-fixation, especially in productiveinshore waters (Muggide 2001), and carbon fromthe atmosphere through gas exchange. In a well-mixed lake, the basal stable isotope values wouldbe relatively uniform, set by δ15N and δ13C valuesderived from terrestrial and atmospheric processesas well as from biological processes within the lake.However, in a lake as large as Lake Victoria, basalstable isotope values derived from algal photosyn-thesis can vary greatly from one region to another.The differences can arise from variation in recy-cling patterns of nutrients, which will affect theδ15N of the inorganic nitrogen sources (Fogel andCifuentes 1993), and from variability in the rates ofalgal growth rates and CO2 availability affectingthe δ13C of phytoplankton (Hecky and Hesslein1995). There are two ways that basal δ15N and δ13Cvalues can be affected without external inputs. Oneis phytoplankton-driven and the other is nutrient-driven. If phytoplankton growth is low, then inor-ganic N compounds and CO2 can be in excessrelative to phytoplankton demand, leading to heav-ier δ15N and lighter δ13C basal values entering thefood web (Keough et al. 1998, Schindler et al.1997). This is the typical scenario for offshore sta-ble isotope values produced in a deep light-limitedmixed layer, and seems to be occurring in WinamGulf with its higher turbidity (~20 NTU), leading tolower Secchi depths and lower chlorophyll concen-trations (Table 1). Alternatively, when the algal bio-mass is high, CO2 and N demand increasesrequiring N-fixation and bicarbonate uptake to meetalgal demand. This would lead to lighter δ15N(from increased N-fixation) and heavier δ13C (fromincreased use of bicarbonate) basal values for the
FIG. 5. The relationship between �15N and �13Cvalues for biota from the Napoleon and Winamgulfs before (A) and after (B) calibration of theNapoleon values to Winam values using the inter-cepts in Figure 4. The dotted points represent S.intermedius and S. afrofischeri from WinamGulf, which have very light �13C values and verylikely do not make up a significant portion of Nileperch diets.
252 Campbell et al.
food web in near-shore environments with highalgal biomass (Muggide 2001). Thus, increasedabundance of rapidly growing phytoplankton innear-shore regions such as Napoleon Gulf can leadto lighter nitrogen and heavier basal carbon isotopicsignatures. Possible external influences include in-creased animal waste and sewage which can alsolead to heavier δ15N and lighter δ13C values due tothe accumulation and increased availability of the15N and 13C compounds in the aquatic environment(Cabana and Rasmussen 1996, Kwak and Zedler1997). This also may be occurring in Winam Gulfwith its large-scale farms and heavy industry and isreflected in its higher biological oxygen demand es-timates (Table 1) relative to Napoleon Gulf. Differ-ences in phytoplankton ecology and pollution levelsmay contribute to the offset in basal food web dif-ference in δ13C of 4.0 and δ15N of 3.1 betweenWinam and Napoleon gulfs.
Once the Napoleon Gulf stable isotope data were“adjusted” to Winam Gulf data values, there wereno significant differences between the δ15N or δ13Cvalues for Nile perch or Nile tilapia between thetwo gulfs. This indicates that the fractionation ofstable isotopes through trophic transfers in fish ineach gulf remains consistent regardless of the basalisotopic differences between the two gulfs. Thismakes it possible to compare contaminant biomag-nification patterns using stable isotopes between thetwo gulfs (Campbell et al. 2003). In this study a rel-atively arbitrary method of assigning trophic posi-
tion was used, which was based upon known foodrelationships in Lake Victoria as supported by sta-ble isotope data, to obtain the intercept valuesneeded to “adjust” the food web stable isotope data.However, the calculated basal isotopic differences(δ15N: 3.1; δ13C: 4.0) are independently verified bythe similarity of TL-δ15N / δ13C regression inter-cept differences (Table 6, Fig. 6) between the twogulfs. For example, the intercept differences forδ15N and δ13C are 3.2 and 4.0 respectively for Nileperch and for Nile tilapia, the TL-δ15N intercept is2.6. In addition, the food web structure of Winamand Napoleon gulfs are similar, so such a rough es-timate is justified as long as its limitations are rec-ognized.
It has been suggested that large Nile perch maybe consuming young of the same species because ofthe decline in the availability of hapolochromineprey (Hughes 1992, Ogutu-Ohwayo 1994). The rel-atively low δ15N values of Nile perch in this studywould seem to support the cannibalization hypothe-sis. A Napoleon Gulf stomach content study foundthat Nile perch that are 60 to 100 cm tend to con-sume more Nile perch (usually < 20 cm; > 54 %frequency of occurrence) relative to other preyitems (Ogutu-Ohwayo 1994). In Napoleon Gulf, theδ15N difference (∆δ15N) between Nile perch ≥ 60cm (δ15N, 10.5‰, n = 1) and smaller Nile perch (< 20 cm; δ15N, 6.6 ± 0.1‰; n,3) is 3.9‰. InWinam Gulf, the ∆δ15N value is 3.8‰ (60 to 100cm, 13.1 ± 0.4‰, n = 6; < 20 cm, 9.3 ± 0.4‰, n = 6
TABLE 6. Regressions of �15N and �13C for Nile perch and Nile tilapia againsttotal length (TL) for Winam and Napoleon gulfs. For each regression, the sample size(n), mean ± s.d. and range of TL, as well as the intercept, slope and adjusted r2, arelisted. A 158-cm Nile perch from Winam Gulf was excluded as an outlier. For eachpair of regressions, ANCOVA analyses (�, 0.027) indicate that the intercepts (p, >0.05) but not the slopes (p, < 0.001) are significantly different. The exception is �13Cvs TL regression for Napoleon Gulf Nile tilapia, which is not significant.
Site Object Regression n TL (cm) Intercept Slope r2adj
Napoleon Nile perch δ15N vs TL 12 9.0–62.0 5.92 0.058 0.71δ13C vs TL (35.9 ± 16.4) –19.73 0.022 0.40
Winam Nile perch δ15N vs TL 16 10.4–87.8 9.11 0.053 0.72δ13C vs TL (47.9 ± 22.0) –23.70 0.029 0.30
Napoleon Nile tilapia δ15N vs TL 11 15.5–41.8 4.45 0.100 0.46δ13C vs TL (22.1 ± 7.9) –18.24 0.010 0.00
Winam Nile tilapia δ15N vs TL 13 25.5–52.1 7.08 0.100 0.17δ13C vs TL (41.3 ± 8.2) –28.70 0.140 0.29
Food Web Structure in Northern Lake Victoria 253
respectively). These ∆δ15N values of 3.8 to 3.9‰are remarkably similar to the ∆δ15N value of 3.7‰found for cannibalizing Arctic char in a remotenorthern lake (Hobson and Welch 1995), and is inagreement with Nile perch stomach-content evi-dence (Hughes 1992, Ogutu-Ohwayo 1994). These∆δ15N values represent only one trophic transfer ifa mean enrichment of 3.5‰ per transfer is assumed.An alternative hypothesis is that Nile perch dependsheavily upon Caridina. Nile perch from 20 to 60cm have a high frequency of occurrence (29 to55%) of Caridina in stomach contents (Ogutu-Ohwayo 1994), and the mean ∆δ15N is 3.8‰ be-tween Caridina and Nile perch in that size class.These structures show that the dominant species in
Lake Victoria, the Nile perch, has a wide dietaryspectrum with a heavy dependence on invertebratesas juveniles and increasing piscivory as it ages andgrows. The result of this broad spectrum and om-nivory is a relatively short food chain length with a∆δ15N of only about 4‰ between Caridina andNile perch and no more than 8‰ between Caridinaand the largest Nile perch. Assuming a mean en-richment of 3.5‰ per transfer, these ∆δ15N valuesrepresent only one to two trophic transfers betweenCaridina and the top predator.
Changes in stable isotope values with fish growthmay occur due to shifts in diet with increased fishsize. Nile perch in both gulfs have similar size-dependent dietary shifts and trophic position as evi-denced by the similar slopes of the stable-isotope:TL regressions. Nile perch’s diet selectionis limited only by the size of the prey their gapesize allows. Gape size consistently increases withthe size of Nile perch (gape size ≅ 0.114 (TL)),which allows larger Nile perch to incorporate big-ger prey items (Witte and van Densen 1995). This issupported by stomach content studies indicatingthat the size and type of prey changes consistentlywith the growth of Nile perch (Mkumbo andLigtvoet 1992, Ogari and Dadzie 1988, Ogutu-Ohwayo 1994). Although the scatter in the stableisotope values indicates some degree of individualvariability in feeding behavior within a size-classcohort, this does not negate the general trend of in-creasing δ15N and δ13C values with size in Nileperch. Differences in δ15N and δ13C values amongNile perch of similar sizes are more influenced bybasal stable isotope values than by dietary differ-ences. The higher δ15N values of Winam Gulf Nileperch cannot be taken as an indicator of highertrophic position or different dietary patterns relativeto Napoleon Gulf Nile perch.
Large Nile perch ≥ 100 cm incorporate more Niletilapia (34%) relative to other prey items than thesmaller size classes, and an equivalent amount ofNile perch (32%; Ogutu-Ohwayo 1994). This repre-sents a potential decrease in trophic position forvery large Nile perch because their intake of Nileperch decreases with size. This is supported by theobservation that the δ15N value (12.1‰) of thelargest Nile perch (158 cm) sampled in Winam Gulfis lower than the mean δ15N value (13.1‰) for the60 to 100 cm class (Fig. 6A). The stomach contentsof this giant individual were found to containmostly Nile tilapia with a few C. gariepinus. Ac-cording to Ogutu-Ohwayo (1994), Nile perch < 100cm consume prey with a mean length of 16.7 ± 3.0
FIG. 6. �15N (A) and �13C (B) values in Nileperch versus their total length. The dotted outlierin each graph represents a very large Nile perch(TL = 158 cm) which was not included in theregression calculations.
254 Campbell et al.
cm. Nile tilapia in this size class have a mean δ15Nvalue of 8.5 ± 0.5‰ (n, 3) leading to a ∆δ15N of3.6‰ for this giant Nile perch, similar to the ∆δ15Nvalues listed above. However, Nile perch ≥ 100 cmare increasingly rare in L. Victoria (Ogutu-Ohwayo1999), and current populations likely have limitedimpacts on the aquatic ecosystem.
Detailed stomach-content analyses of NapoleonGulf Nile tilapia from < 15 cm to > 35 cm primarilyhave shown that their diet consists of detritus andchironomids, and remains relatively constant,changing only with in-situ ecological, seasonal, ordiel variation (Balirwa 1998). This has the effect ofweakening any correlation between trophic leveland fish size. In both gulfs, positive regressions be-tween δ15N values and TL indicate that adult Niletilapia do increase their δ15N values as they grow.In addition, immature Nile tilapia (less than 5 cm)have similar or significantly higher δ15N valuesthan adult Nile tilapia in both gulfs. This may bedue to a shifting quality of detritus; larger Niletilapia feed almost exclusively on detritus and chi-ronomids away from the shoreline while smallerNile tilapia feed upon a mix of detritus and inverte-brates (including chironomids, mollusks, andCaridina) nearer to the shore (Balirwa 1998). Basedon C/N ratios, it has been hypothesized that detritusfurther away from shore (C/N = 15:1 dry weight)have higher nutritional value compared to near-shore detritus (C/N = 40:1; Balirwa 1998). The lackof correlation between δ13C values and TL for adultNile tilapia in Napoleon Gulf suggests that theirdiets only shift in terms of improving quality, notcomposition. In Winam Gulf, there is a significantcorrelation of δ13C and length in Nile tilapia, sug-gesting that their dietary carbon sources shift duringgrowth. In Winam Gulf, as Nile tilapia becomelarge enough to avoid predation by Nile perch, theymay gradually move into the more open waters ofthe gulf and either move to a higher trophic positionor consume food with higher δ13C values.
The stable isotope values of R. argentea and Y.laparograma in both gulfs seem inconsistent withthe simple food chain (algae → zooplankton →pelagic fish and Caridina → Nile perch) suggestedby Kaufman (1992) among others. Prior to the hap-lochromine collapse, Nile perch preferred hap-lochromine fish to R. argentea , but after thecollapse, R. argentea began to appear in Nile perchstomachs (Ogutu-Ohwayo 1994). R. argentea and Y.laparograma, however, have relatively high δ15Nvalues, comparable to top trophic Nile perch. There
is no evidence of the 3 to 4‰ trophic fractionationbetween these predator and its putative prey, whichprecludes R. argentea and Y. laparograma as regu-lar dietary items for Nile perch. If R. argentea wereimportant to Nile perch diets, this should make Nileperch average δ15N values around 11.7‰ inNapoleon Gulf and 14.9‰ in Winam Gulf, which isnot the case. This may mean that these species forma smaller proportion of the Nile perch somaticgrowth than is indicated by dietary studies (Hughes1992, Ogari and Dadzie 1998, Ogutu-Ohwayo1994) or that the situation has changed significantlyfrom the time of the earlier studies.
The high δ15N values of R. argentea and Y. la-parograma might be partially explained by the highδ15N values of their commonly-invoked prey, zoo-plankton. The consistently high δ15N values of zoo-plankton, regardless of their δ13C values, seems tosupport this hypothesis. For example, the ∆δ15N(2.7 to 4.7‰) between zooplankton and phyto-plankton / suspended floc indicates that the sampledzooplankton is about one trophic position above ofphytoplankton. However, the ∆δ15N of the pelagicfish species and zooplankton is only 0.9‰, whichindicate that other invertebrate species may be moreimportant to R. argentea and Y. laparograma thanzooplankton. In Mwanza Gulf, R. argentea isknown to opportunistically feed upon lake-fly lar-vae (chironomids and chaoborids) and Caridina insurface waters (Wanink 1998). For example, the∆δ15N value for pelagic fish and Caridina inNapoleon and Winam Gulfs are approximately 3.1to 3.6‰, while for lake-flies, the ∆δ15N values arebetween 3.6 to 5.3‰. These are more plausible val-ues, which provides strong lake-wide support forthe hypothesis that these pelagic fish are currentlynot limited to feeding on zooplankton, but are op-portunistic predators upon invertebrates.
In Napoleon Gulf, Y. laparograma has signifi-cantly lighter δ13C values indicating an offshorepelagic diet relative to R. argentea, which has heav-ier δ13C values. This is similar to the range of δ13Cvalues in zooplankton from offshore and nearshoresites in Napoleon Gulf. As seen in study, offshorezooplankton usually have lighter δ13C values,which may be due to lighter δ13C values in offshorephytoplankton (Hecky and Hesslein 1995, Ramlal2002), and this is likely reflected in other plankti-vore invertebrates such as lake-flies. Y. laparo-grama and R. argentea in Napoleon Gulf maycoexist by ecological segregation into near-shoreand offshore habitats, which is consistent with theirδ13C values.
Food Web Structure in Northern Lake Victoria 255
In Winam Gulf, S. intermedius and S. afrofischeriexhibit δ13C values which are lighter than mostfish, and seem to form a group separate from otherfish and invertebrates (Fig. 4). S. intermedius feedon Odonata larvae and small pelagic fish which par-tially explains their light δ13C values. Stomach con-tents indicate that S. afrofischeri feed on mollusksand chironomids in the benthic environment (Table1), and often feed on plankton by swimming upsidedown at the water surface. Mollusks and chirono-mids are often considered pelagic feeders becauseof their reliance on phytoplankton and pelagic or-ganisms (Cabana and Rasmussen 1996), which ex-plains the lighter δ13C values of this fish species.Since these two fish species constitute a food chainthat does not lead to Nile perch, and are not foundin stomach contents (Ogutu-Ohwayo 1994), theycan be eliminated from studies on Nile perchenergetics.
In Lake Victoria today, Nile perch is dependenton a variety of macroinvertebrates and its ownyoung for its growth through maturity and even intoadulthood (Fig. 7). Young Nile perch and Caridinaare isotopically consistent as prey items for largeradults and there is an increasing preference for pis-civory with increasing size. Nile tilapia inNapoleon Gulf have isotopic signatures consistentwith a strong preference for detritus throughout its
life history, while in Winam Gulf, there are indica-tions of shifting dietary patterns. R. argentea and Y.laparograma cannot be contributing to somaticgrowth of Nile perch (Fig. 7) as the pelagic fishhave δ15N signatures indistinguishable from Nileperch (and in Napoleon Gulf, significantly differentδ13C values). Zooplanktivory appears to be rela-tively unimportant to many of the fish species inLake Victoria, including R. argentea and Y. laparo-grama. Several common fish species such as S. in-termedius and S. afrofischeri, which coexist withNile perch in the modern lake, are trophically iso-lated from Nile perch. This suggests that only a nar-row group of fish species is available to sustaingrowth of larger Nile perch.
Overall, δ15N and δ13C values are different at thebase of the food webs in Winam and Napoleongulfs. However, the food web structures are verysimilar. The analysis of δ15N and δ13C values inbiota is a powerful tool to study trophic positionand dietary sources in Lake Victoria. One caveat tokeep in mind—it should be understood that theδ15N and δ13C values of biota are relative to thebasal stable isotope values. Comparisons of trophicposition and food web structure between differentlocations should be undertaken with an understand-ing of the possible differences in basal stable iso-tope values. There remains much to be understoodabout the dynamics of fish and invertebrates in theLake Victoria ecosystem, particularly during recentdramatic limnological and biotic shifts occurring inthe lake. However, food web structures in twowidely separated gulfs in Lake Victoria with differ-ing water qualities and anthropogenic effects aresimilar, demonstrating that food web studies in onegulf may be extrapolated to another gulf.
ACKNOWLEDGMENTS
The staff and scientists at Fisheries ResourcesResearch Institute, the European Union Lake Victo-ria Fisheries Research Project (both in Jinja,Uganda), and the Kenya Marine Fisheries ResearchInstitute (Kisumu, Kenya) aided with co-ordinatinglogistics and providing laboratory and storagespace. Dr. D. G. Dixon provided financial supportand office space at the University of Waterloo. W.Mark assisted with stable isotope analyses at theEnvironmental Isotope Laboratory, University ofWaterloo. Dr. J. O’Hara-Hines (Department of Sta-tistics and Actuarial Sciences, University of Water-loo) provided statistical consultation. Dr. K. Geheb(Lake Victoria Fisheries Research Project) provided
FIG. 7. A schematic food-web diagram of thefood web structure based on stable isotope datafrom Napoleon and Winam Gulfs, Lake Victoria.The thin solid arrows demonstrate the energy flowfrom important dietary items to consumers. Dottedarrows indicate occasional dietary sources that arenot isotopically important but are found in stom-ach contents of the predator fish. Dietary shiftsfrom young to adult to very large adult Nile perchand for Nile tilapia are indicated.
256 Campbell et al.
population references for Table 1. Financial supportwas provided by a NSERC PGS-B Graduate Schol-arship and International Development ResearchCouncil Doctorate Awards to LMC and NSERC re-search grants to REH and D. G. Dixon.
REFERENCESBalirwa, J.S. 1998. Lake Victoria Wetlands and the
Ecology of the Nile tilapia, Oreochromis niloticusLinné. Ph.D. thesis. Wageningen Agricultural Univ.,Wageningen, NL.
Bootsma, H.A., Hecky, R.E., Hesslein, R.H., and Turner,G.F. 1996. Food partitioning among Lake Malawinearshore fishes as revealed by stable isotope analy-ses. Ecology 77:1286–1290.
Cabana, G., and Rasmussen, J.B. 1994. Modelling foodchain structure and contaminant bioaccumulationusing stable nitrogen isotopes. Nature 372:255–373.
———, and Rasmussen, J.B. 1996. Comparison ofaquatic food chains using nitrogen isotopes. Proc.Nat. Acad. of Sci. 93:10844–10847.
Campbell, L.M., Hecky, R.E., Nyaundi, J., Muggide, R.,and Dixon, D.G. 2003. Distribution and food-webtransfer of mercury in Napoleon and Winam gulfs,Lake Victoria, East Africa. J. Great Lakes Res. 29(Suppl. 2):267–282.
Copley, H. 1958. Common Freshwater Fishes of EastAfrica. London, UK: H. F. & C. Witherby Ltd.
DeNiro, M.J., and Epstein, S. 1981. Influence of diet onthe distribution of nitrogen isotopes in animals.Geochim. et Cosmochim. Acta 45:341–351.
Fogel, M.L., and Cifuentes, L.A. 1993. Isotope fractiona-tion during primary production. In Organic Geochem-istry, eds. M.H. Engel and S.A. Macko, pp. 73–98.New York, New York: Plenum Press.
Genner, M.J., Turner, G.F., Barker, S., and Hawkins, S.J.1999. Niche segregation among Malawi cichlidfishes? Evidence from stable isotope signatures. Ecol-ogy Letters 2:185–190.
Gophen, M., Ochumba, P.B.O., and Kaufman, L.S. 1995.Some aspects of perturbation in the structure and bio-diversity of the ecosystem of Lake Victoria (EastAfrica). Aquat. Living Res. 8:27–41.
Greenwood, P.H. 1966. The Fishes of Uganda. Nairobi,Kenya: The Uganda Society (Kampala).
———. 1981. The Haplochromine Fishes of the EastAfrican Lakes. Müchen, Germany: Kraus-ThomsonOrganization GmbH.
Hecky, R.E. 1993. The eutrophication of Lake Victoria.Internat. Ver. Theor. Ang. Limnol. Ver. 25:39–48.
———, and Hesslein, R.H. 1995. Contributions of ben-thic algae to lake food webs as revealed by stable iso-tope analyses. N. Amer. J. Benthol. Soc. 14:631–653.
———, Bugeny, F.W.B., Ochumba, P., Talling, J.F.,Mugidde, R., Gophen, M., and Kaufman, L. 1994.
Deoxygenation of the deep water of Lake Victoria,East Africa. Limnol. Oceanogr. 39:1476–1481.
Hobson, K.A., and Welch, H.E. 1995. Cannibalism andtrophic structure in a high Arctic lake: insights fromstable-isotope analysis. Can. J. Fish. Aquat. Sci.52:1195–1201.
Hughes, N.F. 1992. Nile perch, Lates niloticus, predationon the freshwater prawn, Caridina nilotica, in theNyanza Gulf, Lake Victoria, East Africa. Environ.Biol. Fish. 33:307–309.
Kaufman, L. 1992. Catastrophic change in species-richfreshwater ecosystem: The lessons of Lake Victoria.BioScience 42:846–858.
Keough, J.R., Hagley, C.A., Ruzycki, E., and Sierszen,M. 1998. 13C composition of primary producers androle of detritus in a freshwater coastal ecosystem.Limnol. Oceanogr. 43:734–740.
Kidd, K.A., Bootsma, H.A., Hesslein, R.H., Muir,D.C.G., and Hecky, R.E. 2001. Biomagnification ofDDT through the benthic and pelagic food webs ofLake Malawi, East Africa: Importance of trophic leveland carbon source. Environ. Sci. Technol. 35:14–20.
Kwak, T.J., and Zedler, J.B. 1997. Food web analysis ofsouthern California coastal wetlands using multiplestable isotopes. Oecologica 110:262–277.
Lehman, J.T., and Branstrator, D.K. 1994. Nutrientdynamics and turnover rates of phosphate and sulfatein Lake Victoria, East Africa. Limnol. Oceanogr.39:227–233.
Lung’ayia, H.B.O., M’harzi, A., Tackx, M., Gichuki, J.,and Symoens, J.J. 2000. Phytoplankton communitystructure and environment in the Kenyan waters ofLake Victoria. Freshw. Biol. 43:529–543.
Mkumbo, O.C., and Ligtvoet, W. 1992. Changes in thediet of Nile perch, Lates niloticus (L), in the MwanzaGulf, Lake Victoria. Hydrobiologia 232:79–83.
Muggide, R. 1992. Changes in phytoplankton primaryproductivity and biomass in Lake Victoria (Uganda).M.Sc. thesis. Dept. Botany, University of Manitoba,MB.
———. 2001. Nutrient Status and Planktonic NitrogenFixation in Lake Victoria, Africa. Ph.D. thesis. Dept.Biology, University of Waterloo, ON.
Ogari, J., and Dadzie, S. 1988. The food of Nile perch,Lates niloticus (L.) after the dissapperance of the hap-lochromine cichlids in the Nyanza Gulf of Lake Vic-toria (Kenya). J. Fish Biol. 32:571–577.
Ogutu-Ohwayo, R. 1994. Adjustments in Fish Stocks andin Life History Characteristics of Nile Perch, Latesniloticus L. in Lakes Victoria, Kyoga and Nabugabo.Ph.D. thesis. Dept. Zoology, University of Manitoba,MB.
———. 1995. Diversity and stability of fish stocks inLakes Victoria, Kyoga and Nabugabo after establish-ment of introduced species. In The Impact of SpeciesChanges in African Lakes, eds. T.J. Pitcher and P.J.B.
Food Web Structure in Northern Lake Victoria 257
Hart, pp. 59–82. London, Great Britain: Chapman &Hall.
———. 1999. Deterioration in length-weight relation-ships of Nile perch, Lates niloticus L. in Lakes Victo-ria, Kyoga and Nabugabo. Hydrobiologia 403:81–86.
Peterson, B.J., and Fry, B. 1987. Stable isotopes inecosystem studies. Ann. Rev. Ecol. Syst. 18:293–320.
Ramlal, P.S. 2002. Sources, Transport and Sinks ofOrganic Matter in Two African Great Lakes. Ph.D.thesis. Dept. Biology, University of Waterloo, ON.
———, Kling, G.W., Ndawula, L.M., Hecky, R.E., andKling, H.J. 2001. Diurnal fluctuations in PCO2, DIC,oxygen and nutrients at inshore sites in Lake Victoria,Uganda. In The Great Lakes of the World (GLOW):Food-web, health and integrity, eds. M. Munawar andR.E. Hecky, pp. 65–80. Leiden, The Netherlands:Backhuys Publishers.
Republic of Kenya. 1997. District Development Plans1997–2001 for Kisumu, Homa Bay, Rachuonyo,Siaya, Suba, Busia and Migori (7 separate reports).Office of the Vice-President and Ministry of Planningand National Development. Nairobi, Kenya.
———. 2000. Economic Survey. Central Bureau of Sta-tistics, Ministry of Finance and Planning. Nairobi,Kenya.
The Republic of Uganda. 2000. The Republic of Uganda1999 Statistical Abstract. Uganda Bureau of Statis-tics. Entebbe, Uganda.
Scheren, P.A.G.M., Zanting, H.A., and Lemmens,A.M.C. 2000. Estimation of water pollution sources inLake Victoria, East Africa: Application and eleabora-tion of the rapid assessment methodology. J. Environ.Manage. 58:235–248.
Schindler, D.E., Carpenter, S.R., Cole, J.J., Kitchell, J.F.,and Pace, M.L. 1997. Influence of food web structureon carbon exchange between lakes and the atmos-phere. Science 277:248–250.
SEDAWOG. 1999. The survey of Lake Victoria’s fishers.Socio-Economic Data Working Group (SEDAWOG),Lake Victoria Fisheries Research Project. Jinja,Uganda. Technical Document No. 5. LVFRP/TECH/99/05.
Twongo, T. 1996. Growing impact of water hyacinth onnearshore environments on Lakes Victoria and Kyoga(East Africa). In The limnology, climatology, andpaleoclimatology of the East African lakes, eds. T.C. Johnson and E.O. Odada, pp. 633–642. Amster-dam, The Netherlands: Gordon and Breach Pub-lishers.
van Oijen, M.J.P. 1995. Key to Lake Victoria fishesother than haplochromine cichlids. In Fish Stocks andFisheries of Lake Victoria: A Handbook For FieldObservation, eds. F. Witte and W.L.T. van Densen,pp. 209–300. Cardigan, UK: Samara Publishing Ltd.
Vander Zanden, M.J., and Rasmussen, J.B. 2001. Varia-tion in 15N and 13C trophic fractionation: Implicationsfor aquatic food web studies. Limnol. Oceanogr.46:2061–2066.
Wanink, J.H. 1998. The pelagic cyprinid Ratrineobolaargentea as a crucial link in the disrupted ecosystemof Lake Victoria. Ph.D. thesis. Sociale Wetenschap-pen. Univ. of Leiden, NL.
Witte, F., and van Densen, W.L.T. 1995. Fish Stocks andFisheries of Lake Victoria: A Handbook for FieldObservations. Cardigan, UK: Samara PublishingLimited.
———, Goldschmidt, T., Goudswaard, P.C., Ligtvoet,W., van Oijen, M.J.P., and Wanink, J.H. 1992.Species extinction and concomitant ecologicalchanges in Lake Victoria. Netherlands J. Zool.42:214–232.
Submitted: 15 August 2001Accepted: 20 October 2002
