Dichlorodiphenyltrichloroethane in the aquatic ecosystem of the Okavango Delta, Botswana, South Africa

7

Environmental Toxicology and Chemistry, Vol. 22, No. 1, pp. 7–19, 2003q 2003 SETAC

Printed in the USA0730-7268/03 $12.00 1 .00

DICHLORODIPHENYLTRICHLOROETHANE IN THE AQUATIC ECOSYSTEM OF THEOKAVANGO DELTA, BOTSWANA, SOUTH AFRICA

BONTLE MBONGWE,† MELISSA LEGRAND,‡ JULES M. BLAIS,† LYNDA E. KIMPE,† J. JEFFREY RIDAL,§ andDAVID R.S. LEAN*†

†Department of Biology, University of Ottawa, 30 Marie Curie Road, Ottawa, Ontario K1N 6N5, Canada‡Dietetics and Human Nutrition, McGill University, St. Anne de Bellevue, Quebec, Canada

§Saint Lawrence River Institute of Environmental Sciences, Windmill Point, 2 Belmont Avenue, Cornwall, Ontario K6H 4Z1, Canada

(Received 21 October 2001; Accepted 15 June 2002)

Abstract—Concentrations of DDT and its metabolites were measured in water, plants, invertebrates, and fish from lagoons in theOkavango Delta, Botswana (Africa), where DDT has been used for approximately 50 years. The sampling area was sectioned todistinguish spraying for malaria and for African sleeping sickness. Average concentrations of total DDT (sum of DDT and itsmetabolites) in the Okavango ranged from 0.009 ng/L in water to 18.76 ng/g wet weight in fish. These levels are approximately1% of those found in piscivorous fish from temperate North America. The dichlorodiphenyl ethylene (DDE) metabolite was themost abundant fraction of total DDT. Although total DDT concentrations were higher in areas treated for malaria than areas treatedfor sleeping sickness, these concentrations were likely driven by factors other than the historic application of the pesticide.Equilibration with air concentrations is the most likely explanation for these levels. Since the mean annual temperature exceedsthe temperature of vaporization of DDT, this research points to the need for reliable transport models. Our results showed that totalDDT concentration in fish was best explained by lipid content of the fish and trophic position inferred by d15N, regardless of DDTapplication history in those areas. The reservoir above Gaborone Dam, an area downstream of the Okavango but where DDT hadnot been used, was sampled to compare total DDT levels to the treated areas. The two species (a herbivorous threespot talapia andthe omnivorous sharptooth catfish) from Gaborone had levels higher than those found in the Okavango Delta, but these differencescan again be explained using trophic position inferred by d15N rather than by fish size or location.

Keywords—Okavango Botswana DDT Fish Aquatic ecosystem

INTRODUCTION

One of the dirty dozen chemicals now slated for globalrestrictions is DDT [1,2]. In Botswana, use of DDT dates tothe 1950s for control of the mosquito (Anopheles sp.), whichtransmits malaria, and the 1960s for control of the tsetse fly(Glossina morsitans centralis), which transmits African sleep-ing sickness [3]. Both malaria and sleeping sickness are prev-alent in the Okavango Delta (World Health Organization andMinistry of Health, Botswana, Africa, unpublished data).

Although records for the total DDT applied to the distinc-tive regions do not exist, unpublished reports from the WorldHealth Organization state that DDT (75% wettable powder orDDT emulsion) was applied to homes in the area of the Oka-vango from 1950 to 1970. The Ministry of Agriculture (frompesticide survey reports by P. Mosupi and B. Koosimile, PlantProtection Division, Ministry of Agriculture, Gaborone, Bot-swana, Africa) estimated DDT applications in the Okavangoto range between 5,000 and 10,000 kg annually in the early1990s. After experimenting unsuccessfully with Fenitrothion,a synthetic pyrethroid, in 1972, DDT applications resumedfrom 1973 until 1997 for malaria control in the form of indoorresidual spraying at a rate of 2 g/m2 once a year. Furthermore,regions within the Okavango sprayed for control of anopheleshad also been historically sprayed to control the tsetse fly. Incontrast, the tsetse regions had only been sprayed for controlof the tsetse. Following pressure from international organi-zations, DDT in Botswana was banned in 1997.

* To whom correspondence may be addressed([email protected]).

Persistent organic pollutants, like DDT, have the capacityto accumulate in organisms and magnify in food chains. Fac-tors known to contribute to this biomagnification include lipidcontent [4,5], some measure of body size such as length orweight [4,6], and trophic position [7,8].

Naturally occurring stable isotopes of nitrogen and carboncan be used to determine distinctive food web pathways lead-ing to the top trophic position. The ratio of isotope 15N to 14Nexpressed as d15N (fraction of 15N atoms per 1,000 atoms of14N) are used to characterize an organism’s trophic position[7–11]. Historically, food chain structure was established usingstomach content analysis, which Hyslop [12] described as pro-viding only a snapshot of the fish’s present diet. The advantageof the d15N technique, as opposed to the conventional stomachcontent analysis, is that the isotope signatures integrate dietaryhabits over a prolonged period of years [13] and thus is ableto accommodate slower-growing species [14]. Peterson andFry [15] showed that d15N is enriched by a factor of 3 to 5‰from prey to predator, thereby providing a continuous variablewith which to quantify biomagnification of organochlorines inaquatic food webs.

Isotopes of carbon (d13C) reflect an organism’s diet and havebeen used to trace the source of carbon from prey to predatorin complex food webs [16]. These authors found that the frac-tionation rate of carbon isotopes is relatively low (0–3‰) com-pared to that of nitrogen (3–5‰) isotopes [13]. If food preyoptions have a distinctive d13C signature, this isotope has theadvantage of providing a time-integrated measure of assimi-lated diet independent from stomach contents [17].

The objectives of this study are to compare levels of total

8 Environ. Toxicol. Chem. 22, 2003 B. Mbongwe et al.

Fig. 1. Map of the Okavango Delta showing the malaria (M) and tsetse or Xakanaka (T) areas where sampling took place.

DDT in the water and pelagic biota (plankton to fish) in areaswith different histories of DDT application; to establish a re-lationship between inferred trophic position (d15N), carbonsource (d13C), lipid content, length and weight, versus totalDDT tissue concentration for selected fish species; and to com-pare biomagnification factors, defined as the slope of log DDTconcentration versus d15N [8] with those of published studiesin temperate regions.

MATERIALS AND METHODS

Study site

The Okavango Delta is an alluvial fan located between 198and 208S and 228 and 248E in the northern part of Botswana.The fan has an approximate surface area of 25,000 km2 [18].The area of the wetland is approximately 12,000 km2 [19]. Itlies on the edge of the Kalahari Desert and is formed as aresult of the discharge of the Okavango River at the southernextremity of the East African Rift system [20]. The entireOkavango River system lies within the Kalahari Basin partly

filled with sediments of various types, notably, wind-blownbrown and white desert sands [19–24]. Maximum and mini-mum average temperatures in ambient air during the summerrange from 30.5 to 33.78C and 14.8 to 19.28C [22]. Watertemperatures vary between 178C in winter (July) to 358C inthe summer rainy season (January–February) [23,24].

To study the fate and persistence of DDT in the OkavangoDelta, we chose two regions with distinct historical patternsof DDT spraying [25]. The location of the Okavango and thesampling sites are shown on Figure 1. Xakanaka lagoon waschosen as a site where DDT was used exclusively for tsetsefly control from the 1960s to the late 1970s. Three other la-goons (Ikoga, Samuchima, and Guma) were chosen as siteswhere DDT was used for Anopheles mosquito control sincethe 1950s until 1997 [19]. A review of literature on DDT andother pesticides used in the delta suggests that these threelagoons were also treated historically for tsetse control. Itshould therefore be noted that distributions of DDT in Xak-anaka are the only ones that can be solely attributed to tsetse

DDT in Botswana, Africa Environ. Toxicol. Chem. 22, 2003 9

fly control activities, while the other three locations may beaffected by DDT used for both tsetse and Anopheles controlactivities.

The reservoir above Gaborone Dam, located in the capitalcity of Botswana, where DDT use has not been recorded, wasalso chosen as control site. Although the area has moderateindustrial activity, it is a nonagricultural area and thereforeexcludes the possibility for the reservoir to be contaminateddirectly by DDT use for the control of agricultural pests butdoes not exclude upstream or atmospheric sources. Throughoutthis paper, we refer to Gaborone as the control area, Xakanakaas the tsetse area, and the combined lagoons that received DDTfor both Anopheles and tsetse control as the malaria area.

Sampling

Most of the sampling was conducted in July and August1999 with a few additional fish collected in December 1999.Water, particulate matter, plant material, and fish were collectedfrom the malaria region. Similar collections were made at Xak-anaka with the addition of zooplankton. From Gaborone, onlytwo species of fish (the herbivorous threespot tilapia and thecarnivorous sharptooth catfish) and zooplankton were collect-ed. In total, 4 water, 3 suspended particulate, 2 plant material,2 zooplankton, and 44 fish samples were gathered.

Fish species used in this study have been described bySkelton [26]. They include piscivorous fish such as the tigerfish(Hydrocynus vittatus), the omnivorous sharptooth catfish(Clarias gariepinus), and the herbivorous redbreast tilapia(Tilapia rendalli). Other species include the threespot tilapia(Oreochromis andersonii), the brownspot largemouth (Ser-ranochromis thumbergi), the nembwe (Serranochromis ro-bustus), the bulldog (Marcusenius macrolepidotus), spottedsqueakers (Synodontis nigromaculatus), finetooth squeaker(Synodontis vanderwaal) and the striped robber (Brycinus la-teralis). Refer to Table 1 for individual fish species collectedin each region and to Table 2 for mean values.

Most of the fish were bought from fishermen within a fewhours of collection. Standard length and weight were measuredon site before gut removal, then frozen in solvent-rinsed alu-minum foil and sealed in Whirlpakt bags (NASCO Industries,Fort Atkinson, WI, USA). Zooplankton was collected by hor-izontally towing a zooplankton net (150 mm). The contentswere transferred to 100-ml solvent-rinsed bottles and frozen.No enumeration of zooplankton was done. Some samples ofaquatic plant material were collected from each study site andfrozen in Whirlpack bags.

Stomach content analysis

Fish stomachs were removed in the field. Most were im-mediately frozen, and about 10% (randomly chosen) were pre-served in 4% formalin. All frozen contents were wrapped insolvent-rinsed aluminum foil, stored in cooler boxes filled withice in the field, and then stored at 2608C in the laboratoryuntil analysis. Thirty-seven of 44 fish stomach contents (frozenand formalin preserved) were numerically identified with theaid of a dissecting microscope. Categories for identifiablestomach contents (noted by its presence and not the totalamount of each item found) were periphyton, terrestrial plantmatter, terrestrial insects, aquatic insects, mollusks, aquaticseeds, and fish (Table 3). Stable isotope analysis was conductedon selected frozen fish stomachs from each of the three areasof interest (Table 4). When possible, individual componentsof the stomachs were analyzed. For instance, when a whole

fish was intact in the stomach (not yet digested), individualanalysis was run. For plant material or when samples weredigested, we opted for bulk analysis.

Stable isotope analysis

Zooplankton, fish muscle (excluding skin), and plant tissuewere freeze-dried in the laboratory for stable isotope analysisbut not for chemical analysis. Both wet and dry (freeze-dry)weights were determined. Samples were ground with a mortarand pestle, and a 200-mg subsample was transferred into tincapsules. All samples were analyzed by the G.G. Hatch StableIsotope Laboratories, Department of Earth Sciences, Univer-sity of Ottawa (Canada) for carbon and nitrogen, with an au-tomated CE Instrument (Model EA—1110; elemental C andN analyzer) coupled to a Finnigan Mat deltaPLUS IRMS witha Conflow II interface (Finnigan Mat, Bremen, Germany). Foreach run of 10 samples, two isotopic standards for carbon weresupplied by the International Atomic Energy Association([IAEA], Vienna, Austria) and are also available from the Na-tional Institute of Standards and Testing (NIST) and the U.S.Geological Survey. The standards are IAEA-USGS 24 graphiteand IAEA-CH-6 (NIST-8542), and two standards for nitrogen(IAEA-N-1 and IAEA-N-2) were used to correct for drift invalues between runs if present. A precision of #0.3‰ (basedon analysis of replicates) was obtained using this analyticalprocedure, generally within 0.2‰ for nitrogen and 0.31‰ forcarbon.

Stable nitrogen and carbon isotopes are expressed in deltanotation (d) and have units of parts-per-thousand differencefrom the standard. d13C and d15N are calculated as follows:

15 13 3d N or C per mil 5 [(R /R ) 2 1] 3 10sample standard

where R is 15N/14N or 13C/12C.

Water sample collection for total DDT analysis

Subsurface water samples were collected in 19-L airtightstainless-steel containers. Prior to collection, the cans wererinsed three times with wash-grade acetone/hexane. Water wasextracted on site by slowly pumping it through a 142-mmWhatman GF/F glass-fiber filters and through a solid-phaseextraction cartridge (type ENV1) supplied by Chromatograph-ic Specialties (Brockville, ON, Canada) using a multiplate fil-tration system [27]. Water was allowed to flow at about 250ml/min. The ENV1 cartridges (200 mg) were cleaned in theEcotoxicology Laboratory of Jules Blais (University of Ot-tawa, Ottawa, ON, Canada) prior to shipment to Botswana byeluting them with 15 ml of dichloromethane (allowed to flowby gravitation) and dried with a vacuum desiccator. Cartridgeswere then wrapped with aluminum foil (precleaned with wash-grade acetone/hexane solvents) and sealed in Whirlpack bags(NASCO). Prior to extraction, ENV1 cartridges were condi-tioned with 5 ml of trace-grade methanol followed by anotherelution of the column with deionized water (DIW) to get ridof methanol. Samples were then spiked with polychlorinatedbiphenyl-30 (PCB) and octyl-chloro-naphthalene (OCN) sur-rogates. The ENV1 cartridges and the glass-fiber filters (GF/F)were then rewrapped in solvent-rinsed aluminum foil, sealedin Whirlpack bags, and frozen immediately until shipment tothe University of Ottawa, where they were extracted for anal-ysis. Field and laboratory blanks of ENV1 cartridges werecarried along with the extraction apparatus in the field but notused for sample collection.


Table 1. Length (mm), weight (g), d15N (‰), d13C (‰), lipid (%), p,p9-DDE (ng/g), SDDT (ng/g) and total DDT (tDDT) (ng/g) in water (ng/L), and pelagic biota (particulate matter, plant material, zooplankton, and fish) from Gaborone, the three malaria lagoons, and Xakanaka (Africa).

A dash represents a value below detection limit or a parameter not measured

Location Region Common name Sample ID d15N d13C Length Weight Lipidp,p9-DDEa SDDT tDDT

GaboroneGaboroneGaboroneGaboroneGaborone

ControlControlControlControlControl

ZooplanktonThreespot tilapiaThreespot tilapiaThreespot tilapiaSharptooth catfish

GABZOOGAB7GAB6GAB5GAB3

4.309.83

10.1210.3513.62

229.50219.00219.30217.60221.80

—230240210410

—316359233846

—0.40.10.62.7

0.2551.320.172.34

22.99

—0.09—

0.210.07

0.3691.660.192.81

23.45GaboroneGaboroneGaboroneGumaGuma

ControlControlControlMalariaMalaria

Sharptooth catfishSharptooth catfishCarpParticulate matterNembwe

GAB2GAB1GAB4GusupGU3

13.9014.2713.43

0.107.67

219.30219.30219.80225.00226.30

350420520—

230

6001,0002,500

—307

0.92.52.1—0.6

13.7537.43

7.440.040.45

0.062.02—

0.020.11

13.9837.43

7.500.060.61

GumaGumaGumaGumaGuma

MalariaMalariaMalariaMalariaMalaria

NembweNembweRedbreast tilapiaStriped robberStriped robber

GU5GU2GU4GU6GU1

7.949.004.296.276.44

225.10223.50224.70221.80225.80

220290200

8580

333500255

1410

0.71.40.9

10.413.9

0.771.451.750.324.45

0.060.1—

0.081.64

1.21.811.750.477.13

IkogaIkogaIkogaIkogaIkoga


WaterWaterParticulate matterWater lilyBrownspot largemouth

IKW1IKW2IksupIkolyIK4

——

1.511.696.92

——

225.50223.10227.10

————

360

————

707

———0.11.1

——

0.080.060.48

0.0190.020.110.110.54

0.0220.0240.190.171.02

IkogaIkogaIkogaIkogaIkogaIkoga

MalariaMalariaMalariaMalariaMalariaMalaria

Redbreast tilapiaThreespot tilapiaThreespot tilapiaThreespot tilapiaThreespot tilapiaThreespot tilapia

IK3IK2IK8IK1IK10IK9

4.326.166.206.787.557.80

223.50223.70226.30227.20226.40227.10

220255200240220230

370484215329364395

0.40.41.02.42.42.3

0.050.580.061.973.991.2

0.060.370.011.423.010.98

0.110.950.073.397.312.18

IkogaIkogaIkogaIkogaSamuchima


Sharptooth catfishSharptooth catfishSharptooth catfishWestern bottlenoseBrownspot largemouth

IK7IK11IK6IK5SA7

8.078.188.975.317.84

222.80225.00223.90223.80223.10

440470550270245

1,4521,5002,000

550420

0.60.2

10.55.70.4

1.42.527.983.92.09

0.9916.713.91.1

2.393.54

15.648.333.86

SamuchimaSamuchimaSamuchimaSamuchimaSamuchima


Brownspot largemouthRedbreast tilapiaRedbreast tilapiaThinface largemouthTigerfish

SA4SA1SA3SA5SA2

8.354.044.858.178.16

223.00220.30220.20224.20224.20

330280280290290

500400600500200

0.61.60.50.61.9

2.322.682.362.232.57

0.81.580.451.353.92

4.264.274.113.977.48

SamuchimaXakanakaXakanakaXakanakaXakanaka

MalariaXakanakXakanakXakanakXakanak

TigerfishWaterWaterParticulate matterTswii (rooted)

SA6XGwXkwXGsupTswii

9.29——

1.881.05

224.70——

225.50224.30

500————

2,400————

11.6———0.1

9.920.0090.040.010.08

6.71—

0.080.020.01

18.760.0090.120.030.09

XakanakaXakanakaXakanakaXakanakaXakanaka

XakanakXakanakXakanakXakanakXakanak

ZooplanktonBrownspot largemouthBulldogFinetooth squeakerRedbreast tilapia

XzooXGA4EX9EX10EX3

3.217.345.367.801.30

222.30221.60220.80221.00218.80

—190240210165

—109220192164

0.40.4

10.03.20.4

0.030.163.754.750.05

0.090.041.441.6—

0.180.26.176.810.05

XakanakaXakanakaXakanakaXakanakaXakanaka

XakanakXakanakXakanakXakanakXakanak

Redbreast tilapiaRedbreast tilapiaRedbreast tilapiaSharptooth catfishSharptooth catfish

EX5EX7EX4EX6EX2

1.831.882.066.858.46

219.80221.70220.50220.00221.20

157175159365535

127219139636

1,000

0.10.20.23.90.3

0.050.010.162.331.34

0.03—

0.010.12—

0.080.020.244.641.34

XakanakaXakanakaXakanakaXakanaka

XakanakXakanakXakanakXakanak

Spotted squeakerThinface largemouthTigerfishTigerfish

EX11EX1EX8XGA3

7.945.047.578.19

222.80223.90222.80222.80

205125280395

194139458680

5.00.20.32.5

5.490.120.232.42

2.84——

0.71

9.260.140.273.47

a DDE 5 dichlorodiphenyl.

Sample extraction for organochlorine compound analysis

To extract organochlorines, ENV1 cartridges were unfroz-en first to remove excess water. They were then eluted with15 ml of dichloromethane (DCM) followed by another 10 mlto remove any remaining compounds from the column. Theeluant containing organochlorine pesticides was rotary evap-orated down to 2 ml and separated into three subfractions ona Florisil column (Supelco, Mississauga, ON, Canada). Theglass-fiber filters used for the collection of suspended partic-

ulate material from water were cut into small pieces, mixedwith sodium sulfate to remove water, and extracted with DCMon a Soxhlet extractor. Extracts were then treated as describedin the following.

Fish samples were prepared for gas chromatographic anal-ysis for organochlorine compounds using similar methods de-scribed by Kidd et al. [13]. Briefly, the dorsal muscle (in-cluding skin) and whole small fish samples of 5 to 10 g werehomogenized with sodium sulfate, placed on cellulose extrac-


Tab

le2.

Mea

nle

ngth

(mm

),w

eigh

t(g

),d1

5N(‰

),d1

3C(‰

),li

pid

(%),

p,p9

-DD

E(n

g/g)

,S

DD

T(n

g/g)

and

tota

lD

DT

(tD

DT

)(n

g/g)

inw

ater

(ng/

L),

pela

gic

biot

a(p

arti

cula

tem

atte

r,pl

antm

ater

ial,

zoop

lank

ton,

and

fish

)fr

omG

abor

one

(Gab

.),

pool

edm

alar

iala

goon

s(M

al.)

,an

dX

akan

aka

(Xak

.)(A

fric

a).

Ada

shre

pres

ents

ava

lue

belo

wde

tect

ion

lim

itor

apa

ram

eter

not

mea

sure

d

Lat

insp

ecie

sC

omm

onna

me

Loc

alit

yn

d15N

d13C

Len

gth

Wei

ght

%L

ipid

p,p9

-DD

Ea

SD

DT

tDD

T

Serr

anoc

hrom

isth

umbe

rgi

Serr

anoc

hrom

isan

gust

icep

s

Bro

wns

pot

larg

emou

th

Thi

nfac

ela

rgem

outh

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

3 1 1 1

7.70

7.34

8.17

5.04

224

.40

221

.60

224

.20

223

.90

312.

0019

0.00

290.

0012

5.00

542.

2010

8.80

500.

0013

8.50

0.70

0.40

0.60

0.20

1.63

0.16

2.23

0.12

0.81

0.04

1.35 —

3.05

0.20

3.97

0.14

Serr

anoc

hrom

isro

bust

us

Til

apia

rend

alli

New

bwe

Red

brea

stti

lipi

a

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

3 4 4

8.20

4.38

1.77

225

.00

222

.18

220

.20

247.

00

245.

0016

4.00

380.

00

406.

1816

2.00

0.90

0.85

0.23

0.89

1.71

0.07

0.09

0.67

0.01

1.21

2.56

0.10

Ore

ochr

omis

ande

rson

ii

Mar

cuse

nius

mac

role

pido

tus

Thr

eesp

otti

lipi

a

Bul

ldog

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

3 5 1

10.1

06.

90

5.36

218

.63

226

.14

220

.80

226.

6722

9.00

240.

00

302.

4735

7.42

219.

50

0.37

1.70

10.0

0

1.28

1.56

3.75

0.15

1.16

1.44

1.55

2.78

6.17

Cap

rinu

sca

rpio

Syno

dont

isva

nder

waa

l

Car

p

Fin

etoo

thsq

ueak

er

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

1 1

13.4

3

7.80

219

.80

221

.00

520.

00

210.

00

2,50

0.00

191.

60

2.10

3.20

7.44

4.75

— 1.60

7.50

6.81

Syno

dont

isni

grom

acul

atus

Bry

cinu

sla

tera

lis

Spo

tted

sque

aker

Str

iped

robb

er

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

1 2

7.94

6.36

222

.80

223

.80

205.

00

82.5

0

193.

80

12.2

5

5.00

12.1

5

5.49

2.39

2.84

0.86

9.26

3.80

Hyd

rocy

nus

vitt

atus

Mor

myr

usla

cerd

a

Tig

erfi

sh

Wes

tern

bott

leno

se

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

2 2 1

8.73

7.88

5.31

224

.45

222

.80

223

.80

395.

0033

7.50

270.

00

1,30

0.00

569.

20

549.

80

6.75

1.40

5.70

6.25

1.33

3.90

5.32

0.71

3.90

13.1

21.

87

8.33

Cla

rias

gari

epin

us

Wat

er

Sha

rpoo

thca

tfish

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

3 3 2 2 2

13.9

38.

417.

66 — —

220

.13

223

.90

220

.60

— —

393.

3348

6.67

450.

00

— —

815.

331,

650.

6781

3.05

— —

2.03

3.77

2.10 — —

24.7

23.

971.

84 — 0.02

5

0.72

2.90

0.06

0.02

2

24.9

57.

192.

99

0.02

30.

0645

Sus

pend

edpa

rtic

ulat

e

Pla

ntm

ater

ial

Gab

.M

al.

Xak

.G

ab.

Mal

.X

ak.

2 1 1 1

0.80

11.

88

1.69

41.

045

225

.25

225

.5

223

.12

24.3

— — — —

— — — —

— — 0.1

0.04

0.06

0.01

0.06

0.08

0.06

50.

02

0.11

0.01

0.12

50.

03

0.17

0.09

Zoo

plan

kton

Gab

.M

al.

Xak

.

1 1

4.30

3

3.21

3

229

.5

222

.3

— —

— —

— 0.4

0.25

5

0.03

0.04

9

0.09

0.37

0.18

aD

DE

5di

chlo

rodi

phen

yl.


Tab

le3.

Sto

mac

hco

nten

tid

enti

fica

tion

anal

ysis

for

fish

from

the

cont

rol,

mal

aria

,an

dX

akan

aka

area

s(A

fric

a).

Xre

pres

ents

pres

ence

ofth

eit

em

Loc

atio

nC

omm

onna

me

Fis

hID

d15N

d13C

tDD

TF

ish

Per

iphy

ton

Det

ritu

sA

quat

icse

eds

Mol

lusk

sA

quat

icin

sect

sT

erre

tria

lin

sect

sP

lant

Non

iden

tifi

ed

Gab

oron

eG

uma

Gum

aG

uma

Gum

a

Bro

wns

pot

larg

emou

thN

embw

eN

embw

eN

embw

eR

edbr

east

tila

pia

GA

B4

GU

2G

U3

GU

5G

U4

13.4

36.

449.

007.

674.

29

219

.80

225

.80

223

.50

226

.30

224

.70

7.50

7.13

1.81

0.61

1.75

X X X XX

X

X X XX

XG

uma

Gum

aIk

oga

Ikog

aIk

oga

Str

iped

robb

erS

trip

edro

bber

Bro

wns

pot

larg

emou

thR

edbr

east

tila

pia

Sha

rpto

oth

catfi

sh

GU

1G

U6

IK4

IK3

IK6

7.94

6.27

6.92

4.32

8.97

225

.10

221

.80

227

.10

223

.50

223

.90

1.2

0.47

1.02

0.11

15.6

4X

XX X

XX X X X

X X

Ikog

aIk

oga

Ikog

aIk

oga

Ikog

a

Sha

rpto

oth

catfi

shS

harp

toot

hca

tfish

Thr

eesp

otti

lapi

aT

hree

spot

tila

pia

Thr

eesp

otti

lapi

a

IK7

IK11

IK1

IK2

IK8

8.07

8.18

6.78

6.16

6.20

222

.80

225

.00

227

.20

223

.70

226

.30

2.39

3.54

3.39

0.95

0.07

X X

XX X X

X X XX

X

X

X

Ikog

aIk

oga

Ikog

aS

amuc

him

aS

amuc

him

a

Thr

eesp

otti

lapi

aT

hree

spot

tila

pia

Wes

tern

bott

leno

seB

row

nspo

tla

rgem

outh

Bro

wns

pot

larg

emou

th

IK9

IK10

IK5

SA

4S

A7

7.80

7.55

5.31

8.35

7.84

227

.10

226

.40

223

.80

223

.00

223

.10

2.18

7.31

8.33

4.26

3.86

X X

XX X X

XX X

X X X X

X

X

Sam

uchi

ma

Sam

uchi

ma

Sam

uchi

ma

Sam

uchi

ma

Sam

uchi

ma

Red

brea

stti

lapi

aR

edbr

east

tila

pia

Thi

nfac

ela

rgem

outh

Tig

erfi

shT

iger

fish

SA

1S

A3

SA

5S

A2

SA

6

4.04

4.85

8.17

8.16

9.29

220

.30

220

.20

224

.20

224

.20

224

.70

4.27

4.11

3.97

7.48

18.7

6

X X X

XX

X XX

Xak

anak

aX

akan

aka

Xak

anak

aX

akan

aka

Xak

anak

a

Bul

ldog

Fin

etoo

thsq

ueak

erR

edbr

east

tila

pia

Red

brea

stti

lapi

aR

edbr

east

tila

pia

EX

9E

X10

EX

3E

X4

EX

5

5.36

7.80

1.30

2.06

1.83

220

.80

221

.00

218

.80

220

.50

219

.80

6.17

6.81

0.05

0.24

0.08

XX X X X

X X X X

XX

X

XX

akan

aka

Xak

anak

aX

akan

aka

Xak

anak

a

Red

brea

stti

lapi

aS

harp

toot

hca

tfish

Sha

rpto

oth

catfi

shS

pott

edsq

ueak

er

EX

7E

X2

EX

6E

X11

1.88

8.46

6.85

7.94

221

.70

221

.20

220

.00

222

.80

0.02

1.34

4.64

9.26

X X

X X

X X XX

Em

pty

stom

ach

X XX

Xak

anak

aX

akan

aka

Xak

anak

a

Thi

nfac

ela

rgem

outh

Tig

erfi

shT

iger

fish

EX

1E

X8

XG

A3

5.04

7.57

8.19

223

.90

222

.80

222

.80

0.14

0.27

3.47

X X

Em

pty

stom

ach

XX X


Table 4. Comparison of d15N and d13C (‰) for selected fish species in relation to their prey items from the stomach

Fish species Common name Fish ID d15N d13C Prey item d15N d13C

Clarias gariepinusClarias gariepinusHydrocynus vittatusMormyrus lacerdaSerranochromis robustus

Sharptooth catfishSharptooth catfishTigerfishWestern bottlenoseNembwe

EX6GAB2SA2IK5GU5

8.4613.9

8.165.317.94

221.2219.3224.2223.8225.1

FishFishFishInsectsFish

5.037.575.953.737

223.54217.65225.27217.65228.59

Serranochromis thumbergiSerranochromis thumbergiSerranochromis thumbergiSynodontis nigromaculatusTilapia rendalliTilapia rendalli

Brownspot largemouthBrownspot largemouthBrownspot largemouthSpotted squeakerRedbreast tilapiaRedbreast tilapia

SA4SA4SA7EX11EX5GU4

8.358.357.847.941.834.29

223.1223.1223.1222.8219.8224.7

FishLampreyFishInsectsAlgaePeriphyton

7.877.237.763.181.514.36

228.25227.76228.25224.47221.32225.29

tion thimbles (precleaned by Soxhlet for 6 h), and extractedwith DCM using a Soxhlet apparatus for 16 h. Surrogate re-covery standards of polychlorinated biphenyl 30 (PCB30) andOCN were added to the samples prior to extraction. The extractwas rotary evaporated to approximately 2 ml. Extractable lip-ids were determined gravimetrically with one-tenth of the ex-tract, and the remaining lipids were removed by gel permeationchromatography on a column of biobeads SX-3 with Hexane/DCM (55:45). The gel permeation chromatography eluant con-taining organochlorine pesticides was collected and separatedinto three subfractions on a Florisil column (8 g, 1.2% de-activated with high-performance liquid chromatography–gradewater). The first fraction was eluted with 37 ml of hexane, thesecond with 38 ml of hexane-DCM (85:15), and the third with50 ml of dichloromethane DCM. The first fraction containedall PCBs and most of the DDE, while fraction 2 contained allhexachlorocychlohexane, chlordane, and DDT compounds.

Gas chromatography

All Florisil eluates (fish, plant, zooplankton, particulates,and water) were rotary-evaporated down to approximately 2ml and solvent exchanged to isooctane, which was then furtherconcentrated to a final volume of 1 ml under a steady streamof ultra-high-purity nitrogen. The final eluate was then ana-lyzed by capillary gas chromatography with 63Ni electron cap-ture detection on a 30-m 3 0.25-mm DB-5 column using Hew-lett-Packard gas chromatography 6890 (Palo Alto, CA, USA)and helium carrier gas 1.3 ml/min, initial flow of 2.0 ml/min,constant pressure of 22.5 psi, and N2 makeup gas. The tem-perature program used held samples at 808C for 2 min, 108C/min ramp to 1108C, and 38C/min to 2808C with a 5-min hold.Polychlorinated biphenyl and organochlorine standards wererun once every 10 injections and reagent blanks every 10samples.

Extraction efficiencies for all quantified DDT and metab-olites were determined by analysis of spiked standards (En-vironment Canada, CCIW, Burlington, ON). Average (6stan-dard deviation [SD]) surrogate recoveries for zooplankton,fish, and plant material were 86 6 16% and 73 6 13% forPCB30 and OCN, respectively. Surrogate recoveries for waterwere 76 6 6.2% and 98 6 13% for OCN and PCB30, re-spectively. All data were reported without correction of ex-traction efficiencies. Organic contaminant peaks were identi-fied and quantified using Chemstation (Hewlett-Packard, 1998)software based on their appearance at the same retention time(60.3 s) as the standard. Reagent blanks were generally clearfor DDTs and their metabolites.

Quality control and assurance

Procedural blanks containing 10 g of sodium sulfate (pre-cleaned at 6008C for 16 h) were extracted with DCM on aSoxhlet apparatus and subjected to the same extraction pro-cedures as the samples. These were run every 10 samples.Field and laboratory blanks were also run for ENV1 cartridgesused for the extraction of water samples.

Data treatment and statistical analysis

The term total DDT (tDDT) will refer to the sum of theconcentrations of compounds including p,p9-DDT, o,p9-DDT,o,p9-DDE (dichlorodiphenyl ethylene), p,p9-DDE, o,p9-DDD(dichlorodiphenyl diethylene), and p,p9-DDD. The term SDDTrefers to the sum of the concentrations of p,p9-DDT and o,p9-DDT isomers. In the Results section, we present only p,p9-DDE because it was the most abundant fraction of all theisomers, SDDT and tDDT. Concentrations of each compoundwere calculated on a wet-weight and lipid-weight basis. Datawere analyzed using Systatt 7.0.1 [28].

The collected fish, plant material, particulate matter, andwater from the malaria lagoons were all pooled to make oneregion rather than three lagoons. The rationale follows that thethree lagoons are proximate from one another, that DDT ap-plication is similar in all three lagoons, and that it completesthe food web for the malaria region so that it is comparablewith the Xakanaka (tsetse) region. If values among sites arenot different (as shown by analysis of variance [ANOVA]),data were combined.

To attain the first objective of this paper, ANOVA was usedto compare mean DDT levels in similar species between areaswith different histories of DDT application. When the as-sumptions of normality, homoscedasticity, and independenceof residuals were not met, the nonparametric ANOVA equiv-alent called Kruskall–Wallis was used.

Percentage lipid, weight, and length were log10 transformed.Analysis of covariance was used to compare relationships be-tween the independent variables—lipid, d15N, d13C or bodysize, and the dependent variable, tDDT. The purpose of thisanalysis was to assess whether the effect of the independentvariables was equivalent among the two regions of the Oka-vango (malaria and Xakanaka). We cannot include the Ga-borone region in this analysis because only two species of fishwere analyzed. This is insufficient for a food web analysis butsufficient for the first objective (to compare specific speciesamong regions). Homogeneity of slopes was first tested fol-lowed by that of the intercepts. If the slopes or intercepts weresignificantly different (p , 0.05), then separate regression


Fig. 2. Total DDT (tDDT) concentration expressed per gram wetweight (top) and per gram lipid weight plotted against d15N frompooled data of the Okavango (bottom) (see Table 1). Regressionequations for these graphs are provided in Table 5.

analysis for each location was performed. Data from the tworegions were pooled if the slopes among lagoons were notdifferent. Single regression analysis of all the pooled data fromthe Okavango was used to determine which variable (lipid,d15N, d13C, or body size) best predicted the concentrations ofDDT and its metabolites across all lagoons (all pooled datafrom the Okavango). A stepwise multiple regression analysiswas run to determine which combination of the variables statedpreviously best predicted concentrations of DDT and its me-tabolites in the Okavango.

RESULTS

DDT concentrations in water and biota

Total DDT (tDDT) concentrations ranged from 0.009 ng/Lin water to 18.76 ng/g wet weight in fish of the OkavangoDelta (Table 1). When tDDT concentration in water from themalaria region was compared to that of Xakanaka, no signif-icant difference (p 5 0.533) was observed despite higher ab-solute values found in Xakanka (three times higher but allvalues near detection limit). However, fish species from themalaria area, in general, had higher tDDT levels than the samefish found in Xakanaka. For example, mean concentrations(6SD) of tDDT concentrations in Clarias species were 7.196 7.34 ng/g in the malaria areas, while it was 2.99 6 1.13ng/g in Xakanaka. The same was observed with T. rendalli,where the mean concentration of tDDT in malaria and Xak-anaka were 2.56 6 2.00 ng/g and 0.1 6 0.1 ng/g, respectively.Along with higher tDDT levels, we also found higher d15N inthe malaria region compared to Xakanaka (Table 2).

In contrast to our prediction that concentrations of tDDTin Gaborone reservoir (the control area) would be low, valuesranged from 0.37 in zooplankton to 37.43 (ng/g wet wt) infish (Table 1). Results of the ANOVA between the malaria,Xakanaka, and Gaborone showed no significant difference (p5 0.072) for tDDT in the sharptooth catfish (C. gariepinus).However, when pooling data from the Okavango (malaria plusXakanaka) and comparing them to Gaborone, we note a sig-nificant difference in mean concentration of tDDT for thisspecies (ANOVA, p 5 0.018), with higher absolute levels inGaborone. A comparison of mean d15N values for C. garie-pinus between Gaborone and Okavango showed higher av-erage (6SD) values for the former (13.93 6 0.33‰ comparedto 8.11 6 0.78‰). Oreochromis species from Gaborone res-ervoir also had higher mean d15N values for the same speciesthan in the Okavango (10.10 6 0.26‰ and 6.9 6 0.76‰,respectively), but tDDT levels for Oreochromis were not sig-nificantly different (p . 0.05) with Okavango having a valueof 2.78 ng/g and Gaborone 1.55 ng/g.

Analysis of covariance for tDDT on d15N between the tworegions of the Okavango showed no significant difference inthe slope and intercept, thus allowing us to pool the data tomake one area, the Okavango. The pooled data from the Oka-vango (malaria plus Xakanaka) showed that tDDT concentra-tions (ng/g wet wt) did relate with trophic level as inferredfrom d15N (Fig. 2). Significant relationships were observedbetween tDDT concentration normalized to both wet weightand lipid weight and d15N (Table 5 and Fig. 2). The relationshipbetween wet-weight concentrations of tDDT and d15N wasstronger (r2 5 0.49) than the lipid-normalized concentrations(r2 5 0.16). The d15N and log percent lipid were also related(r2 5 0.31, p 5 0.0001; Fig. 3).

Analysis of covariance for tDDT on d13C between the tworegions of the Okavango showed no significant difference in

the slope and intercept, thus again allowing us to pool the datato make one area, the Okavango. The pooled data showed thattDDT concentrations (ng/g wet wt) did not correlate with d13C(p 5 0.655).

Log tDDT was significantly related to log percent lipid (r2

5 0.57, p 5 0.0001; Table 5 and Fig. 4). Slopes of regression(analysis of covariance) for log-transformed tDDT concentra-tions versus percent lipid were not significantly different be-tween the malaria and the Xakanaka region (p . 0.05). Percentlipid was a significant predictor of log tDDT (p , 0.05) inthe pooled data from the malaria and the Xakanaka area.

Slopes of regression and intercepts for log tDDT versus loglength and log weight were not significantly different amongregions (p . 0.05). Pooled data from the two regions showedthat log tDDT in fish muscle were significantly related to logfish length (r2 5 0.36, p 5 0.0001, and r2 5 0.37, p , 0.0001,respectively). Log tDDT concentrations were also significantlyrelated to fish weight (r2 5 0.30, p , 0.05, and r2 5 0.29, p, 0.05, respectively; Table 5).

The effect of d15N, d13C, percent lipid, and body size (lengthand weight) on tDDT concentration using the pooled data fromthe Okavango was estimated using multiple regression analysisand showed that the best predictors for tDDT were log lipidand d15N. This is despite the issue that d15N and lipid arethemselves correlated (multicollinearity) yet still explain moreof the total variance in tDDT when they are considered to-gether:

21log tDDT(ng g wet wt)

155 0.78(60.14)log lipid 1 0.14(60.04)d N 2 0.78(60.22),

2r 5 0.67, p 5 0.0001 (1)


Table 5. Linear relationship between d15N, d13C (‰), and log % lipid, then between d15N (‰), log % lipid, length (mm), log weight (g), andthe log concentrations of total DDT (tDDT) (ng/g wet wt and lipid normalized) in biota from the Okavango Delta (n 5 43)

Y variable Slopea X variable Intercept (b0)a SEest r2 p

Y 5 bX 6 b0

% Lipid% LipidLog Y 5 b1X1 6 b0

tDDT (ng/g wet wt)

0.13 (60.03)20.04 (60.05)

0.21 (6 0.03)

d15Nd13C

d15N

20.91 (60.22)21.1 (61.04)

21.86 (60.22)

0.490.59

0.59

0.310.03

0.49

,0.0001.0.05

,0.0001tDDT (ng/g wet wt)tDDT (ng/g wet wt)tDDT (ng/g wet wt)tDDT (ng/g lipid)Log length (mm)b

Log weight (g)b

1.01 (60.15)2.98 (60.68)1.25 (60.34)0.09 (60.04)0.05 (60.01)0.09 (60.02)

Log lipidLog lengthLog weightd15Nd15Nd15N

10.15 (60.09)27.05 (61.65)23.10 (60.88)11.57 (60.24)12.12 (60.06)12.05 (60.15)

0.530.630.660.550.120.28

0.570.370.30.160.420.31

,0.0001,0.0001,0.05,0.05,0.0001,0.0001

a Values given as means 6 standard error.b For fish only.

Fig. 3. Lipid content of the Okavango food web components plottedagainst d15N (see Table 1). Regression equation provided in Table 5.

Fig. 4. Log total DDT (tDDT) plotted against log percent lipid forpooled data from the Okavango (see Table 1). Regression equationprovided in Table 5.

Food web relationships and DDT biomagnification

Stable isotopes, d13C and d15N, of the Okavango food webcomponents provided information on the predator–prey rela-tionships and carbon flow of these food webs (Fig. 5). Therange in d15N is 0.096‰ in particulate matter to 9.29‰ in thetigerfish (Hydrocynus vittatus) (Table 1). The d15N signaturesdefine trophic levels of H. vittatus, C. gariepinus, and Syn-odontis sp. as the top predators with the highest d15N valuesof 6.85 to 9.29‰. Hydrocynus vittatus has the highest d15Nvalues, ranging from 7.5 to 9.29‰, followed by C. gariepinus(6.85–8.97‰) and Synodontis sp. (7.80–7.94‰). These iso-tope data suggest that those predatory species feed mainly onother smaller fish, such as B. lateralis and O. andersonii,which have mean d15N of 6.4 and 6.8‰, respectively. Thiswas confirmed by stomach content identification analysis ofthese species (Table 3) and is also confirmed by Skelton [26].Serranochromis sp. had a wider range of d15N, from 5.04 to9.0‰, reflecting a wider range of prey items. For instance, S.robustus preys on Synodontis sp., while Serranochromis an-gusticeps preys on Brycinus sp. and S. thumbergi on insects,snails, and fish. The next trophic level is occupied by O. an-dersonii, B. lateralis, Mormyrus lacerda, and M. macrole-pidotus with d15N values ranging from 5.3 to 7.8‰ (Fig. 5).They feed mainly on invertebrates. Tilapia rendalli was shownin this study to be a herbivore based on stomach contents, withmean (6SD) d15N of 3.1 6 1.4‰. Tilapia occupied the sameapproximate trophic level as zooplankton (3.21‰; Fig. 5).Plant material and particulate organic matter, with mean (6SD)

d15N values of 1.4 6 0.3‰ and 1.2 6 0.6‰, respectively,occupied the lowest trophic levels in the isotope model.

The d13C signatures of biota from the Okavango Deltaranged from 227.20‰ for O. andersonii to 218.84‰ for T.rendalli. Though the range was narrower to what may be foundin temperate North America, it permitted some separation ofcarbon sources among fish species (Fig. 5). The stable isotoped13C signature of H. vittatus (223.61 6 0.94‰) was generallysimilar to that of B. lateralis (223.8 6 2.9‰) and M. lacerda(223.84‰). The d13C of Synodontis sp. (221.91 6 1.2‰) wassimilar to that of T. rendalli (21.12 6 2.0‰) and M. macro-lepidotus (20.8‰). This was also supported by the isotopevalues for fish items obtained from the stomach of Hydrocynussp., where the d13C value was 225.27‰ (Table 4). The d13Cvalue for plant material of 223.7‰ was consistent with thatof T. rendalli’s gut content (223.3 6 2.8‰). This confirmsthe stomach content analysis, which showed plant material asa major diet for the redbreast tilapia.

The d15N and d13C of predator and prey were measured inselected fish stomachs (Table 4). Stomach content items forT. rendalli were dominated mostly by aquatic plant stems,leaves, periphyton, and algae (Table 3). This was consistentwith dietary studies done previously on this species [26]. Thestable isotope d15N values for this species were relatively lowand ranged from 1.3 to 4.85‰. One fish from this species hadaquatic insects in its stomach, which may help explain the highd15N value in this particular fish (4.5‰). Stomach items forthe catfish (C. gariepinus), a highly omnivorous fish, showeda consistent diet across the regions of small fish, plant material,and aquatic and terrestrial insects. The tigerfish (H. vittatus),


Fig. 5. d15N plotted against d13C. Symbols are as follow: BR 5 Bry-cinus spp., HV 5 Hydrocynus sp., MO 5 Mormyrus spp., CL 5Clarias spp., MA 5 Marcusenius spp., SER 5 Serranochromis spp.,TL 5 Tilapia spp., OR 5 Oreochromis spp., SY 5 Synodontis spp.,PM 5 plant material, POM 5 particulate organic matter, and ZOO5 zooplankton.

known to become exclusive piscivores when they attain alength of 90 to 100 mm [26], showed these feeding charac-teristics in this study as well, with the smallest tigerfish havingaquatic insects and fish in its diet and the largest and oldestindividuals exclusively feeding on fish. This study confirmedthese feeding habits with the d15N values increasing from thesmallest (7.57‰) to largest (9.29‰) size, suggesting a shiftin diet during its development. Figure 5 provides a clear rep-resentation of the food web structure in the Okavango Delta.The suspended particulate material is slightly lower d13C thanzooplankton with the fish similar to the suspended particulatecarbon. This suggests that the fish derive much of their energyfrom the particulate material and not the zooplankton. Theincrease in d15N when moving from one trophic level to thenext is approximately 4.5 units.

DISCUSSION

DDT concentrations in water and biota

Concentrations of DDT were not related as much to thearea of use as they were to biological factors such as trophicposition (as inferred by d15N) and lipid content. Fish fromSamuchima Lagoon of the malaria region were generally fatter,hence the high DDT concentrations compared to other lagoonsin the delta (Table 1). It can also be seen from Table 1 thatspecies such as the catfish and tigerfish, both of which hadhigher d15N values, were present in Samuchima; thus, the prev-alence of carnivores in Samuchima may have affected the highDDT concentrations observed there.

Concentrations of DDT in fish collected from Gaborone(the control area) were significantly higher than concentrationsin fish from the Okavango Delta (Table 2) but consistent withlevels expected from higher d15N. A word of caution, however:Sewage has higher d15N values than other natural water sam-ples [9], and we first thought that this may explain the elevatedvalues at this site. Arguing against this possibility is the ob-servation that zooplankton values at Gaborone and Xakanakaare similar (Table 1). Interestingly, SDDT (p,p9-DDT, o,p9-DDT) in Gaborone was on average 29 times lower than theDDE metabolite, suggesting an aged source of DDT in thisarea. The catchment area for Gaborone Dam, which includes

water from areas outside Gaborone, may also draw fish fromlocations where DDT may have been used and as such maycontribute to the DDT residues. The d15N values for the twofish species collected in Gaborone, C. gariepinus and O. an-dersonii, were higher than values found in the Okavango Delta(Table 2), and this alone can explain the higher levels. Thepercent lipid for these species was lower in Gaborone com-pared to the Okavango.

Kidd et al. [29] studied fish from Lake Malawi and showedlow average tDDT residue values in similar species to thoseof the Okavango. The Clarias and Oreochromis species inLake Malawi had average levels as low as 1.4 and 0.5 ng/gwet weight, respectively, making them three and five timeslower than the levels in the Okavango Delta. Synodontis spe-cies, however, had tDDT levels that were seven times higherthan the levels of the same species from the Okavango Delta[29]. However, average tDDT values (wet-wt basis) for Tilapiain the Zimbabwean Lake McIlwaine were 75 ng/g [30], makingthem over 50 times higher than the same species in the Oka-vango. In that same study [30], concentrations of tDDT incatfish and tigerfish had average values of 97 and 128 ng/gwet wt), respectively, making their concentrations 19 and 17times higher than the concentrations for the same species inthe Okavango Delta.

Comparisons of DDT concentrations between the Okavan-go and lakes in colder regions also show similar food webstructure and DDT relationships; that is, the slopes of theregression are similar. Despite similar slopes, intercepts of thecold regions seem to be exceedingly higher. High levels ofDDT in these northern areas are surprising considering thelack of DDT use since the 1970s. In 1990, De Vault et al. [1]showed that tDDT concentrations were on average 1900 ng/gin 1976 and 840 ng/g in 1990 in Laurentian Great Lakes. Lakesin the Yukon Territory also have levels of tDDT that are higherthan the Okavango Delta [8,11]. A likely explanation, in ad-dition to the migration of volatile organochlorines from warmto cold regions, is that fish in arctic lakes tend to live longerand grow more slowly and as such may accumulate morecontaminants because of the longer food chains than fish intropical regions [31].

Quantities of DDT used in Botswana are relatively lowcompared to those used in other countries. In Zimbabwe, forinstance, 3 3 105 kg/year of DDT were sprayed in 1982, about1.8 3 105 kg in 1984, 1.7 3 105 kg in 1986, and 0.8 3 105

kg in 1988 [32,33]. These quantities are on average 18 timeshigher than quantities used in Botswana. In colder regions,quantities of DDT previously used are not comparable to thoseused in Botswana. In New Brunswick, Canada, DDT quantitiesused for the control of the spruce budworm from 1952 to 1967were on average 3.83 3 105 kg/year ([34]; http://www.ec.gc.ca/ceqg-rcqe/English/pdf/tissuepsummaryptable.htm). Thesequantities greatly exceed the maximum of 0.1 3 105 kg/yearof DDT used for malaria control in Botswana.

Food web relationships and DDT biomagnification

An increase in d15N with increasing DDT concentrationsindicated that trophic level plays an important role in the move-ment of contaminants through the Okavango food web. Resultsfrom stable isotope d15N explained 42% of the variability oftDDT concentrations in biota from the Okavango (pooled datafrom this study). Fish at a higher trophic position, such as H.vittatus, Clarias sp., and Synodontis sp., had the highest tDDTconcentrations. Previous studies have also shown that trophic


Fig. 6. Log total DDT (tDDT) plotted against d15N (top) where logtDDt 5 0.40 (d15N) 2 2.7, r 2 5 0.58, for the Yukon and log tDDt5 0.23 (d15N) 2 1.3, r 2 5 0.42, for Okavango and log percent lipid(bottom) where log tDDt 5 0.21 (log lipid) 1 0.7, r 2 5 0.66, forthe Yukon and log tDDt 5 0.23 (log lipid) 1 0.16, r 2 5 0.53, forOkavango. Data for the Okavango food web are provided in the pre-sent study, and data for the Yukon food web from Lake Laberge arefrom [31].

position as inferred by d15N is a strong predictor of DDTconcentrations [4,8,31,35]. A rather weak relationship betweenlipid-normalized tDDT and d15N was also observed, indicatingthat trophic position has a significant effect over and abovethat of lipid alone. Nevertheless, it is clear that lipid normal-ization of organochlorine data profoundly reduces the influ-ence of trophic position.

Trophic level was also significantly related to log percentlipid in other related studies in temperate regions [8,10,31,36],and in all cases, d15N versus contaminant concentration wasweaker on a lipid-adjusted concentration basis relative to wet-weight concentrations. Kiriluk et al. [10] found a strong re-lationship between d15N and wet-weight concentrations ofDDE, mirex, and total PCB than they did in lipid normalizedconcentrations. Kidd et al. [36] found a significant relationshipbetween d15N on both the log wet-weight and lipid-correctedconcentrations of DDT, toxaphene, and hexachlorocyclohex-ane in biota from Lake Laberge, Yukon Territory. They con-cluded that the slope of these regressions may be related tothe biomagnification capability of specific lipophilic contam-inants. Kidd et al. [36] concluded that d15N explained a greaterproportion of the variation in contaminant concentrations thandid percent lipid.

Species with high d15N values, such as H. vittatus (tigerfish)and C. gariepinus (catfish), had higher mean tDDT concen-trations than T. rendalli. Indeed, mean concentrations of tDDTin redbreast tilapia (T. rendalli) are six times lower than inthe tigerfish and the catfish. The relationship between d15Nsignatures of fish muscle and log tDDT concentrations (ng/gwet wt) was compared among lakes in the Yukon Territoryand lagoons in the Okavango Delta (Fig. 6, top). No significantdifferences were observed in the slopes (p . 0.05), indicatingsimilar biomagnification factors for DDT concentrations intropical and temperate regions.

Log percent lipid successfully predicted tDDT levels in theOkavango Delta as it has in several other studies undertakenin temperate [4,11,37,38] and tropical [4,30] regions. In LakesMcIlwaine and Kariba, Zimbabwe, southern Africa, fish withthe highest lipid content, such as the tigerfish, had the highesttDDT concentrations [4,30,35,39]. Douthwaite et al. [39] alsofound that fattier fish in the Okavango Delta contained moreresidues of the pesticide endosulfan. It was observed that vis-ceral adipose had both the highest lipid and DDT residue lev-els. Furthermore, Larson et al. [40] found lipid content to besignificantly related to persistent pollutants in salmon popu-lations (Salmo salar) in the southern Baltic Sea. They foundthat tDDT and SHCH were significantly related to lipid onmuscle tissue (p , 0.05, r2 5 0.53). A significant relationshipbetween percent lipid and tDDT was also observed by Bentzenet al. [5] in their study investigating the role of lipid and foodweb structure in lake trout (p , 0.05, r2 5 0.62). The rela-tionship (the slope of the regression) between percent lipidand DDT concentrations was also shown to be of importancein fish from lakes of the Yukon Territory [11]. The interceptsof these regression models are, however, higher than those ofthe Okavango, reflecting higher DDT concentrations in theselocations. The regression slopes of log tDDT on log lipid con-centrations between the Okavango lagoons and the Yukonlakes were compared (Fig. 6, bottom). Clearly, the subarcticlakes in the Yukon, despite having no historical uses of DDT,are more than 10 times more concentrated in DDT than fishfrom the Okavango Delta. This could be evidence of a mi-gration of DDT toward subarctic environments [41,42].

Body size as determined by length and weight of fish, weresignificantly related to d15N. The results are corroborated by thefindings of Bootsma et al. [43], who attributed an increase ofd15N with size in pelagic fish (Engraulicyprus sardella) in LakeMalawi to a shift in trophic level. Because larger individual fishtend to feed at higher trophic levels, they have higher d15N values


and consequently higher tDDT levels. The ability of body sizeto predict DDT concentrations was, however, low compared toboth lipid and d15N. On the whole, the best model (multipleregression) for predicting tDDT in the Okavango Delta was theone using d15N and percent lipid (Eqn. 1).

Stable isotope results were consistent with known feedinghabits of the species analyzed in this study. Tigerfish (H. vit-tatus) is a top predator and is known to feed mainly on otherfish ([26], this study). Both stable isotope results and stomachcontent analyses confirmed this feeding habit. The sharptoothcatfish (C. gariepinus) is known for its scavenging habits,eating any organic food from plants to fish and small mammals[26]. This was confirmed in the present study, where stomachcontent analysis showed a variety of food items, such as plants,insects, detritus, and small fish, reflecting an omnivorous feed-ing behavior. Tilapia rendalli relied heavily on plant materialand algae. A few insects were also found in some of the stom-achs of this species. These findings are consistent with otherstudies undertaken on the feeding habits of this species usinglimited gut content analysis [26,44]. Oreochromis sp. con-sumed invertebrates and small insects, while the diet of Syn-odontis sp. was mainly insects and fish [26, this study]. Thediet of Serranochromis sp. was dominated by fish and mol-lusks. These findings corroborate those of Merron and Bruton[44] and Skelton [26]. Clearly, a snapshot of the fish’s stomachcontents (provided by the stomach content analysis) is notsufficient to determine its trophic position, particularly if thefish was caught with an empty stomach. Thus, stable isotopeanalysis is a more reliable technique to determine, over anextended period of time, the fish’s trophic position.

Concentrations of DDT and its metabolites in water weremuch higher in the Okavango (39 6 55 pg/L) than the levelsfound in the neighboring southeastern African Lake Malawi(9.2 6 13 pg/L) [45], but both are much less than that observedin the Great Lakes [1]. This shows that differences in levelsin fish from tropical regions are lower than in temperate re-gions, as are levels of DDT in the water, despite local use.

CONCLUSIONS

Levels of DDT in fish and water from the Okavango werelow compared to levels in nearby locations, such as Lake Kar-iba in Zimbabwe and in lakes of temperate and subarctic NorthAmerica. Factors such as quantities of DDT used, high tropicaltemperatures in Botswana, and the fast tendency for DDT todegrade and evaporate may have affected the low levels ofDDT in the Okavango Delta.

This study demonstrated that d15N, percent lipid, and bodysize (fish length and weight) can be used to predict DDT con-centrations in biota from the Okavango Delta. The use of d15Nand lipid concentration in this study helped determine the re-lationship of the Okavango Delta food web and DDT biom-agnification. Larger fish, with a higher percentage of lipid, hadthe highest DDT concentrations. Percent lipid and d15N servedas the best predictors of concentrations of DDT and its me-tabolites (Eqn. 1).

Acknowledgement—This research was supported by National Sci-ences and Engineering Research Council (NSERC) Research Grantsto J. Blais and D. Lean and by the Canadian Commonwealth schol-arship to B. Mbongwe. The authors thank the Ministry of Agriculture,Botswana, in particular the Fisheries Division (P. Motsu was mosthelpful). The authors thank K. Kidd for helpful suggestions duringthis project. This manuscript benefited from the insightful reviews

provided by the two anonymous reviewers. We thank them for theircontribution.

REFERENCES

1. DeVault DP, Whittle M, Rang S. 1994. Toxic contaminants in theGreat Lakes. EPA 905-D-94-0001e. U.S. Environmental Protec-tion Agency, Washington, DC.

2. World Wildlife Fund. 1998. Resolving the DDT dilemma: Pro-tecting biodiversity and human health. Washington, DC.

3. Merron GS. 1992. Tsetse fly control and the environmental im-plications for fish in the Okavango Delta, Botswana. BotswanaNotes and Records 24:49–56.

4. Rasmussen JB, Rowan DJ, Lean DRS, Carey JH. 1990. Foodchain structure in Ontario Lakes determines PCB levels in Laketrout (Salvelinus namaycush) and other pelagic fish. Can J FishAquat Sci 47:230–238.

5. Bentzen E, Lean DRS, Taylor WD, Mackay D. 1996. The roleof foodweb structure on lipid and bioaccumulation of organiccontaminants by lake trout (Salvelinus namaykush). Can FishAquat Sci 53:2397–2407.

6. Berg H, Kiibus M, Kautsky N. 1992. DDT and other insecticidesin the Lake Kariba ecosystem, Zimbabwe. Ambio 21:444–450.

7. Cabana G, Rasmussen JB. 1994. Modeling food chain structureand contaminant bioaccumulation using stable nitrogen isotopes.Nature 372:255–257.

8. Kidd KA, Schindler DW, Muir DCG. 1998. Effects of trophicposition and lipid on organochlorine concentrations in fishes fromsub-arctic lakes in Yukon Territory. Can J Fish Aquat Sci 55:869–881.

9. Spies RB, Kruger H, Ireland R, Rice DW. 1989. Stable isotoperatios and contaminant concentrations in a sewage distorted food-web. Mar Ecol Prog Ser 54:157–170.

10. Kiriluk RM, Servos MR, Whittle DM, Cabana G, Rasmussen JB.1995. Using stable nitrogen and carbon isotopes to characterizethe biomagnification of DDE, mirex, and PCB in Lake Ontariopelagic food web. Can J Fish Aquat Sci 52:2660–2674.

11. Kidd KA. 1996. Use of stable nitrogen isotope ratios to char-acterize foodweb structure and organochlorine accumulation insubarctic lakes in Yukon Territory. PhD thesis. University of Al-berta, Edmonton, AB, Canada.

12. Hyslop EJ. 1980. Stomach contents analysis: A review of methodsand their application. J Fish Biol 17:411–429.

13. DeNiro MJ, Epstein S. 1981. Influence of diet on the distributionof nitrogen isotopes in animals. Geochim Cosmochim Acta 45:341–351.

14. Hesslein RH, Holland KA, Ramlal P. 1993. Replacement of sulfur,carbon and nitrogen in tissues of growing broad whitefish (Cor-egonus spp.) in response to change in diet traced by d34S, d13C,and d15N. Can J Fish Aquat Sci 50:2071–2076.

15. Peterson BJ, Fry B. 1987. Stable isotopes in ecosystem studies.Annu Rev Ecol Syst 18:293–320.

16. DeNiro MJ, Epstein S. 1978. Influence of diet on the distributionof carbon isotopes in animals. Geochim Cosmochim Acta 42:495–506.

17. Fry B, Sherr EB. 1984. d13C measurements as indicators of carbonflow in marine and freshwater ecosystems. Contributions to Ma-rine Science 27:13–47.

18. Food and Agriculture Organization of the United Nations. 1977.Investigation of the Okavango delta as a primary water resourcefor Botswana. FAP-AG-DP/BOT/71/506. Technical Report.Rome, Italy, p 284.

19. McCarthy TS. 1992. Physical and biological processes controlingthe Okavango Delta—A review of recent research. BotswanaNotes and Records 24:57–86.

20. Scholz CH, Koczynti TA, Hutchisons DG. 1976. Evidence forincipient rifting in Southern Africa. Geophysical Journal of theRoyal Astronomical Society 44:135–144.

21. Hutchins DG, Hutton SMR, Jones CR. 1976. The geology of theOkavango Delta. Proceedings, Symposium on the Okavango Del-ta and Its Future Utilization, Botswana Society, Gaborone, Bot-swana, September 21, 1975, pp 13–20.

22. Ellery K, Ellery WN, Rogers KH, Walker BH. 1991. Water depthand biotic insulation: Major determinants of back-swamp plantcommunity composition. Wetland Ecology and Management 1:149–162.

23. Cronberg G, Geiske A, Martins E, Prince Nengu J, StenstromIM. 1995. Hydrological studies of the Okavango delta and the


Kwando/Linyati/Chobe River, Botswana. I. Surface water qualityanalysis. Botswana Notes and Records 27:151–226.

24. McCarthy TS, Bloem A, Larkin PA. 1998. Observations on thehydrology, and geohydrology of the Okavango Delta. Geology101:101–117.

25. World Health Organization. 1989. DDT and its derivatives: En-vironmental aspects. Environmental Health Criteria 83. Interna-tional Program on Chemical Safety, Geneva, Switzerland.

26. Skelton PH. 1993. A Complete Guide to the Freshwater Fishesof Southern Africa. Southern Book, Halfway House, Zimbabwe.

27. Helm AP, Diamond ML, Semkin R, Bidleman TF. 2000. Degra-dation as a loss mechanism in the fate of a-hexachlorohexane inarctic watersheds. Environ Sci Technol 34:812–818.

28. Wilkinson L. 1997. SYSTATt: The system for statistics. SYSTAT,Evanston, IL, USA.

29. Kidd KA, Bootsma HA, Hesslein RH, Muir DCG, Hecky RE.2000. Biomagnification of DDT through the benthic and pelagicfood webs of Lake Malawi, East Africa: Importance of trophiclevel and carbon source. Environ Sci Technol 35:14–20.

30. Mhlanga AT, Madziva TJ. 1990. Pesticide residues in LakeMcIlwaine, Zimbabwe. Ambio 19:368–372.

31. Kidd KA, Schindler DW, Muir DCG, Lockhart WL, Hesslein RH.1995. High concentrations of toxaphene in fishes from a subarcticlake. Science 269:240–242.

32. Larsson P, Berglund O, Backe C, Bremle G, Eklov A, JarnmarkC, Person A. 1995. DDT fate in tropical and temperate regions.Naturwissenscchaften 82:559–561.

33. Matthiessen P. 1985. Contamination of wildlife with DDT insec-ticide residues in relation to tsetse fly control operations in Zim-babwe. Environ Pollut Ser B Chem Phys 10:82–211.

34. Canadian Council of Ministers of the Environment. 2001. Ca-nadian tissue guidelines for DDT for the protection of wildlifeconsumers of aquatic biota. Publication 12999. Edmonton, AB.

35. Douthwaite RJ, Tingle CCD, eds. 1994. DDT in the Tropics: TheImpact on Wildlife in Zimbabwe of Ground Spraying for Tsetse

Fly Control. Natural Resources Institute, Overseas DevelopmentAdministration, Chatham, UK, pp 25–37.

36. Kidd KA, Schindler DW, Muir DCG, Lockhart WL, Hesslein RH.1995. Correlation between stable nitrogen isotope ratios and con-centrations of organochlorines in biota from a freshwater food-web. Sci Total Environ 160/161:381–390.

37. Larsson P, Hamrin S, Okla L. 1991. Factors determining the up-take of persistent pollutants in eel population (Anguilar anguilaL.). Environ Pollut 69:39–50.

38. Larsson P, Lennart O, Collvin L. 1993. Reproductive status andlipid content as factors in PCB, DDT and HCH contamination ofa population of pike. Environ Toxicol Chem 12:855–861.

39. Douthwaite RJ, Fox PJ, Mathiessen P, Russel-Smith A. 1981.Environmental impact of aerosols of endosulfan, applied for tsetsefly control in the Okavango Delta, Botswana. Final Report 00745.Endosulfan Monitoring Project. Overseas Development Admin-istration, London, UK, p 141.

40. Larsson P, Backe C, Bremle G, Eklov E, Lokla L. 1996. Persistentpollutants in salmon population (Salmo salar) of the southernBaltic Sea. Can J Fish Aquat Sci 53:62–69.

41. Wania F, Mackay D. 1993. Global fractionation and cold con-densation of low volatility organochlorine compounds in polarregions. Ambio 22:10–18.

42. Wania F, Mackay D. 1996. Tracking the distribution of persistentorganic pollutants. Environ Sci Technol 30:390A–396A.

43. Bootsma HA, Hecky RE, Hesslein RH, Turner GF. 1996. Foodpartitioning among Lake Malawi nearshore fishes as revealed bystable isotope analyses. Ecology 77:1286–1290.

44. Merron GS, Bruton MN. 1990. The physiological and toxicolog-ical effects of aerial spraying with insecticides on fish stocks ofthe Okavango Delta of Botswana. Project 3914, Interim Report2. World Wildlife Federation, Gland, Switzerland.

45. Karlsson H, Muir DCG, Teixiera CF, Burnison DA, StrachanWMJ, Hecky RE, Mwita J, Bootsma HA, Grift NP, Kidd KA,Rosenberg B. 2000. Persistent chlorinated pesticides in air, waterand precipitation from the Lake Malawi Area, Southern Africa.Environ Sci Technol 34:4490–4495.

Documents

Dichlorodiphenyltrichloroethane in the aquatic ecosystem of the Okavango Delta, Botswana, South Africa