First Estimates of Mortality and Population Size of White Sharks on the South African Coast

Full text

GREATWHITE SHARKS
The
Biology
of
Carcharodon carcharias
Edited
by
A.
Peter
Klimley
Bodega
Marine
Laboratory
University
of
California,
Davis
Bodega
Bay,
California
David G. Ainley
H.
T.
Harvey
&
Associates
Alviso,
California
Academic Press
An Imprint oj
Elsevier
San Diego London Boston
New
York Sydney Tokyo Toronto

CHAPTER
36
First Estimates of Mortality and
Population Size of White Sharks on
the South African Coast
GEREMY CLIFF
Natal
Sharks Board
Umhlanga
Rocks,
South Africa
RUDY P. VAN DER ELST and
ANESH GOVENDER
Oceanographic Research
Institute
Durban,
South Africa
TRAIL K. WITTHUHN
Struis
Bay,
South Africa
ELINOR M. BULLEN
Oceanographic Research
Institute
Durban,
South Africa
Introduction
The white shark
Carchnrodou carcharias
occurs along
the entire South African coast, but the center of its
distribution is the temperate waters of the Western
Cape (Bass et al., 1975). The species' range extends
into Namibia and possibly southern Mozambique
(Compagno et al., 1989).
This species has bitten humans in Cape waters and
also in KwaZulu-Natal, where a gill-netting pro-
gram, introduced in 1952, has reduced the incidence
of shark attack (Davies, 1964; Wallett, 1983;
Cliff,
1991).
At present, 40 km of nets, maintained by the
Natal Sharks Board (NSB), catches an average of 1354
sharks per year, including 39 white sharks (see Chap-
ter 32, by Cliff et al.). Initially, all potentially danger-
ous sharks found alive in the nets were killed. Since
1988,
however, most live sharks, including danger-
ous ones, have been released and, whenever pos-
sible,
tagged (Cliff and Dudley, 1992b).
Other sources of fishing mortality of white sharks
are big-game angling, spear fishing, and incidental
catches in commercial fisheries in Western Cape wa-
ters.
None of these mortalities has been quantified,
although one angler captured 18 white sharks in False
Bay over a 4-year period (Wallett, 1983). Some spear
fishermen carry a "powerhead,'' that is, a rifle bullet
modified to fit over the end of the spear, which is
fired at any threatening shark. The considerable com-
mercial value of the jaws and teeth may lead to in-
creased targeting and possible overexploitation
(Compagno, 1991). In South Australia, fishermen and
divers claim that the species has undergone a serious
decline in recent years (Bruce, 1992).
In South Africa, research has been carried out on
white sharks caught in the shark nets (Bass et al.,
1975;
Cliff ct al., 1989, Chapter 32). In 1990, the Shark
Research Center of the South African Museum initi-
ated research into the movements, habitat use, be-
havior, abundance, and population structure of the
species (Compagno, 1991).
In the absence of information about the status of
the stock, government protection was granted to C.
carcharias in April 1991, making it illegal to "catch or
kill any white shark except on the authority of a per-
mit issued by the Director-General of Environment
Affairs'' (Compagno, 1991).
Clearly, there is an urgent need to investigate the
population dynamics of white sharks to assess the
validity and effectiveness of this preemptive legisla-
GREAT WHITE SHARKS
The Biology of Carcharodon carcharias 393 Copyright © 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.

394
GEREMY CLIFF ET
AL.
tion. The tagging of white sharks under the auspices
of a national marine fish tagging program, initiated
by the Oceanographic Research Institute (ORI) in
1984,
has provided this opportunity. The mark-and-
recapture study described in this chapter is a first
attempt at quantifying the stock of white sharks along
a large section of the South African coast.
Materials and Methods
Tagging and Shark Capture
Each dart tag had a stainless steel head and a plastic
streamer, at first red (Floy Tag), but lately yellow (Hall-
print) (van der Elst, 1990). The latter tag also was used
by Bruce (1992), and a tag of similar design has been
used extensively by the U.S. National Marine Fisheries
Service (Casey, 1985). Shark lengths used in the text
are precaudal measurements (PCLs). Details about the
white sharks tagged and recaptured are given in Table I.
The shark nets are distributed along 325 km of
KwaZulu-Natal coast (Fig. 1), Richards Bay being the
northernmost location. Mean monthly sea-surface
temperature (SST) ranges from 19.9°C in August to
25.3°C in February (Cliff and Dudley, 1991a). Specifi-
cations of the nets and the manner in which they are
deployed and serviced have been given by Wallett
(1983) and Cliff et al. (1988a).
In 1990, one of the authors (T.K.W.), a commercial
line fisherman, started tagging white sharks in the
vicinity of Struis Bay (Fig. 1), 6 km east of Cape
Agulhas, the southernmost point on the African con-
tinent. The mean monthly SST there ranges from
15°C in July to 21°C in January (Greenwood and
Taunton-Clark, 1992). The sharks were caught on rod
and 40-kg line, using small live sharks as bait. SST
was measured (by T.K.W.) when most white sharks
were tagged.
Estimating Population Size
A modified Petersen estimate (Seber, 1982) was
used to calculate the population size, N, at the end of
time interval / as
_ (M, + 1) (n, + 1)
(m,
+ 1) 1 (1)
where m, is the number of marked animals recap-
tured in the total catch, ri„ of white sharks, and M, is
the number of marked animals alive at the end of
interval /.
In this study, the tagging and recapture of white
sharks were undertaken concurrently over a 5-year
period. This period was split into five
1-year
periods,
and a Petersen estimate was obtained for each year.
The total population size of white sharks, N, was
estimated as
N= 2^'
(s-
1) (2)
where s is the number of samples.
In applying the Petersen estimate, it was assumed
that all tagging took place at the beginning of time
interval / and that all recaptures occurred at the end.
Allowance was made for the rerelease of tagged
sharks upon recapture and the loss of a constant frac-
tion of tagged animals, 1 - a, either through immedi-
ate tag shedding or death from capture stress. As-
suming that mo animals are marked at the start of a
time interval, / = 0, then the number effectively
tagged. Mo, is
Mo = amo (3)
The number surviving to the next time interval, / =
1,
is equal to
M, am^^e^ Z) (4)
TABLE I Details of Tagged and Recaptured White Sharks
Locality"
tagged
Ballito
Marina Beach
Struis Bay
Struis Bay
Struis Bay
Glenmore
Sex"
M
F
F
F
F
F
Precaudal length
(cm)
180
180
250
250
300
150
Recapture
locality
Algoa Bay
Trafalgar
Struis Bay
Struis Bay
Durban
Glenmore
Time free
(days)
27
1
373
357
527
366
Distance travelled
(km)
774
4
0
0
1409
0
Growth
(cm)
0
15
"Localities shown in Figure 1.
''Abbreviations: F, female; M, male.

36.
Population
Size
off South Africa
395
25°S-
30°S—1
fliito
pafalgar
plenmore
Hi NETTED REGION
as^'s—i
False Bay
^ Kenton on Sea
Algoa Bay
Struis Bay
-25°S
-30°S
—1—
15°E 20°E 25°E
—I—
30°E
-35°S
FIGURE 1 The southern African distribution of the white shark (after Compagno et ai, 1989). The shark-netted region in
KwaZulu-Natal and the localities at which white sharks were tagged or recaptured are indicated.
where Z is the instantaneous rate of total mortality,
taken to be constant. If, at this time, a further m^
animals are tagged and released or if recaptured ani-
mals are rereleased, then the number surviving to the
next time interval, M,, is
Ml = a(me
""
+ m, + W,) (5)
where W, is the number of marked animals re-
released. In general,
M, = M, + «(»/,,, + W,^,) (6)
In order to apply the Petersen estimate, we need to
know the total annual catch, Uj, which comprises a
known component, Kj (the sum of the catch in the
shark nets and that of T.K.W.), and an unknown com-
ponent, Uj (the sum of trawl net catches and those
made by spear fishermen and other big-game an-
glers),
such that
Uj
= Kj + Uj.
Estimating Mortality
Using the Baranov catch equation (Ricker, 1975),
the estimated number of tag recaptures, K„ that are
reported, assuming that all such recaptures are re-
ported, is
R,
= M, J (1 ') (7)
where F is the instantaneous fishing mortality rate,
taken to be constant throughout the experiment.
Hilborn (1990) has shown that the sampling distribu-
tion of the tag recoveries can be approximated by the
Poisson distribution. The likelihood of the expected
number of recoveries, R„ given the observed number
of tag recoveries, r„ is
L{Rj
I Tj)
= exp (-•RA
" r,! (8)

396 GEREMY
CLIFF
ET At.
The total likelihood of observing all r„ given the cor-
responding R^ is therefore the product of all the indi-
vidual likelihoods:
TABLE II Mark-Recapture Statistics
off the South African Coast
R!'
L{R \r) = U exp (-K,) ^ (9)
For computational convenience, the negative of the
log likelihood was calculated, and this equation
formed the quantity to be minimized:
-In lLi(x,F,Z)] = -2 In (exp (-R,) ^ (10)
Given the number of animals that are marked and
recaptured in each time interval, three parameters (a,
f, and Z) need to be estimated to determine the size
of the white shark population, N. Initial trials indi-
cated that there was insufficient contrast in the data
to estimate all the parameters. We therefore fixed a
while allowing free estimation of F and Z. The model
was set up in a spreadsheet that was programmed to
perform a nonlinear optimization routine.
Estimating Parameter Variances and
Confidence Intervals
A bootstrap technique (Efron, 1981; Punt and But-
terworth, 1993) was used to estimate variances for the
parameters F, Z, N„ and N. A large number of arti-
ficially generated recapture data sets were randomly
drawn from a Poisson distribution using the proce-
dure described in the book by Press ct al. (1986). The
procedure requires, as input, the expected mean
number of recoveries in each interval (observed
mean, 1) (Table II). The number of rereleases of white
sharks in each time interval, /, was calculated as
0.6(G,), where G, is the random deviate from a Poi-
sson distribution of mean value = 1, and 0.6 repre-
sents the average number of recaptured white sharks
that have been rereleased (Table II). To each pseu-
dodata set a new set of parameters and derived quan-
tities (such as the N/s) are estimated. The standard
error of a parameter or derived quantity, Q, is then
obtained from
SE.Ql^VSS^ (11)
where Q" is the value of Q from the n\h data set and
Q is the mean of the Q"'s. The 95% confidence inter-
vals were calculated using the percentile method.
Year
1989
1990
1991
1992
1993
Total
Number
tagged
6
20
16
13
18
73
Recaptures,
nil
1"
1
1
3
0
6
Rei x^leases.
1"
0
1
2
0
4
Known
61
50
36
38
53
238
Catch
, Unknown,
50
50
10
10
10
130
"Not used in the analyses.
Results
Sharks Tagged
Between 1978 and 1993, 97 (15.7%) of 616 white
sharks caught in the nets were found alive. Initially,
these sharks were killed, but in 1989 the first of 22 live
sharks (11.8% of the white shark catch) was released
from the nets after being marked with a dart tag. Three
free-swimming sharks, one of which was tagged
twice, were marked while they fed on a dead whale off
Durban. Of the 25 sharks tagged, 13 were females and
11 were males. The males ranged from 130 to 370 cm
PCL, with a mode of 200 cm; the two largest males, 320
and 370 cm, were both feeding on the whale carcass.
The females ranged from 150 to 265 cm PCL, with a
mode of 180-200 cm (Fig. 2).
In the Struis Bay area, 46 were tagged (by T.K.W.).
They included 32 females (range, 150-450 cm; mode,
300 cm) and 8 males (range, 250-350 cm; mode, 350
cm) (Fig. 2). These sharks were tagged in water where
the average SST was 18.6°C (range, 16.2-21.8°C; N =
42).
A 180-cm unsexed specimen was tagged by a com-
mercial angler off Kenton-on-Sea, and a 158-cm fe-
male was tagged and released from a beach seine net
in False Bay. In total, between 1989 and 1993, 73 white
sharks were tagged along the south and east coasts of
South Africa (excluding sharks tagged by members of
the White Shark Research Institute, Cape Town).
Recaptures
Six sharks (8.2% of those tagged) were recaptured
(Table I). The mean distance traveled while at liberty

36. Population Size off South Africa 397
12
ioH
8
ID
O)
^ 6
d
F(N);
n=13
M(N);n=11
F(C);
n=32
M(C);
n=8
100 140 180 220 260 300 340 380 420 460
PCL (cm)
FIGURE 2 Size-frequency distribution of female (F) and male (M) white sharks tagged in
KwaZulu-Natal (N) and in the western Cape (C). Sharks of unrecorded sex are excluded, PCL,
Precaudal length.
was 365 km (SE = 244; range, 0-1409 km), and the
mean time at liberty was 275 days (SE = 86; range, 1-
527 days). Three of the 22 sharks released from the
shark nets were recaptured. One traveled 4 km in 1
day and was rereleased. Another, of 150 cm, was
recaptured in the same shark net installation at Glen-
more 366 days later; it had grown by 15 cm. It was
also released. A third shark, of 180 cm, traveled 774
km from Ballito to Algoa Bay in 27 days, a rate of at
least 28.7 km/day.
Three sharks tagged in the western Cape (by
T.K.W.) were recaptured, two at the same locality
(Struis Bay) after 357 and 373 days, respectively; both
sharks were released. The third shark was recaptured
in the Durban shark nets, having traveled 1409 km in
527 days.
Estimate of Mortality and Population Size
The number of sharks tagged, the number recap-
tured, the number of recaptures rereleased, and the
known and assumed unknown catches are shown
annually for the 5-year period 1989-1993 (Table III).
The shark that was recaptured 1 day after release was
excluded from the analysis, because the time at liber-
ty was too short to assume complete mixing of
marked and unmarked white sharks. Given a surviv-
al factor, a ^ 0.9 (see Discussion), the computed val-
TABLE III White Shark Population Estimates
off the South African Coast, 1989-1993
Year
1989
1990
1991
1992
1993
Mean
Survivors/'
M,
5.4
21.2
27.8
29.9
33.8
Estimated
Total Catch,
//,
111
100
46
48
63
Estimated
Population,''
^,
716 (147-591)'
1119(388-1685)
676 (350-1582)
377(439-1701)
2227 (718-2772)
1279 (839-1843)
CV
(70
24
43
63
123
34
24
CV, Coefficient of variation.
"See Eq. 8 in the text.
''See Eq. 1 in the text.
'95%
confidence limits of the population estimate are given in
parentheses.
ues for instantaneous rates of mortality were Z = 0.53
and f = 0.055 year~i. of these two parameters, Z
was a better estimate than f, as shown by the lower
coefficient of variation (Table IV). Assuming that the
average annual unknown catch was 50 for the years
1989 and 1990, and 10 in the years 1991-1993 (T. Fer-
reira, personal communication), then the annual pop-
ulation estimate ranged from 377 (1992) to 2227

398 GEREMY CUFF ET AL.
sharks (1993) (Table III). The overall estimate for the
5-year study period was 1279, with a coefficient of
variation of 24%.
Discussion
The sharks tagged in KwaZulu-Natal were smaller
than those in Cape waters (see Chapter 35, by Fer-
reira and Ferreira). Although two sharks >300 cm
PCL were tagged while feeding on a whale carcass off
Durban and four mature males were recently caught
in the shark nets (see Chapter 32, by Cliff et al.), it
would appear that white sharks >250 cm are not as
common as smaller individuals in the netted region.
The high recapture rate (13.6%) of white sharks
released from the nets is encouraging. Murru (1990)
regards gill nets as the most stressful means of catch-
ing elasmobranchs, and therefore survival must be
low. The lower recapture rate (6.5%) of sharks tagged
by T.K.VV. may be due to the ban on angling for this
species in 1991.
The overall white shark recapture rate of 8.2% was
similar to or slightly higher than that of other large
coastal sharks marked in the ORI tagging program,
van der Elst and Bullen (1992) reported the following
recaptures: 7.3% of 96 tiger sharks
Galeocerdo
cuvier,
6.8% of 88 bull sharks Carcharhinus leucas, 3.5% of
1122 raggedtooth sharks Carcharias taurus, and 1.6%
of 252 broadnose sevengill cowsharks Notorynchus
cepedianus.
Our white shark recapture rate (8.2%) was lower
than that in South Australia, where 13.6% of 22 white
sharks tagged were recaptured, with traveling dis-
tances of 18-220 km and times at liberty of 30-78 days
(Bruce, 1992). The higher recapture rate and lower
mean time at liberty (60 days) suggest that South Aus-
tralian white sharks may incur a higher fishing mor-
tality, which is understandable, as no legislation pro-
hibits angling for this species. Only two of a small but
TABLE IV Parameters in White Shark Mortality
Estimates off the South African Coast
Parameter
a.
F
Z
Value
0.9 (fixed)
0.055
0.530
CV (%)
46
12
Statistic
L (95%)
0.015
0.420
R (95%)
0.100
0.660
CV, Coefficient of variation; L, left (lower) 95% confidence limit
of estimate; R, right (upper) 95% confidence limit of estimate.
unspecified number of white sharks tagged on the
East Coast of the United States have been recaptured,
one of which traveled 614 km during 1.3 years at
liberty (Casey et al., 1991).
In the present study, four recaptures were made in
the same area; three of the four occurred close to 1
year later. This pattern of sharks returning to fixed
localities at yearly intervals has been suggested by
research in California (Ainley et al, 1985; Klimley and
Anderson, Chapter 33). In South Australia, there is
also a high degree of site fidelity, with 36% of the 58
marked sharks resighted, all at their original locations
(Strong ^f a/., 1992).
The large distances (774 and 1409 km) covered by
two of the recaptured sharks indicate that the animals
are also highly mobile in South African waters. Al-
though these long-distance movements were farther
than those reported in Australia or the United States,
they, too, were coastwise. The low incidence of white
sharks near islands in the Pacific and Indian oceans
indicates that there may be limited transoceanic
movement, and the warm tropical surface waters
may act as a barrier to regular transequatorial move-
ment of white sharks. Although the population of
white sharks along the southern African coast may
not be geographically isolated, there may be little re-
cruitment from other centers of abundance, high-
lighting the need for local legislation to prevent pos-
sible overexploitation. Genetic studies are being
conducted to compare white sharks from different
regions (see Chapter 6, by Martin).
Several assumptions, some of which were dis-
cussed by Ricker (1975), were made in using the Pe-
tersen estimate to determine the size of the white
shark population.
1.
Instantaneous mortality rates, Z and f, are con-
stant. The assumption of constant natural mortality
may not be unreasonable, given that the duration of
the study is relatively short compared to the life span
of white sharks. Furthermore, the sharks tagged com-
prised few very small animals, which may be more
prone to predation, and no very large sharks, which
were approaching the end of their life span. On the
other hand, fishing mortality would be affected by
any change in fishing effort during the study period.
The largest components of this effort are the shark
nets,
which remained constant, and big-game an-
gling, whose effort was heavily curtailed by the pro-
tective legislation introduced in 1991. There were no
reports of any change in effort in the trawl fishery
and by spear fishermen. The extent of the reduction
in f is unknown.
2.
There is a 10% instantaneous tag shedding and
tag-induced mortality. Tags inserted into sharks

36. Population Size off South Africa 399
caught in the nets were checked to ensure that they
were firmly embedded before the animals were re-
leased. Due to the stress of capture on a baited line or
in a gill net, some mortality is likely. Telemetry may
provide an assessment of capture mortality. In this
study, we assumed that 90% of tagged animals sur-
vived the stress of capture, hence a = 0.9.
3.
There is no long-term tag shedding. The good
condition of the tag in a shark recaptured after 527
days indicates good tag retention. Tag shedding is
therefore thought to be low. A large bull shark re-
tained its dart tag for 11 years in captivity before be-
ing released. Double-tagging experiments on white
sharks may provide more information on long-term
tag shedding rates.
4.
All recaptures of tagged sharks are reported. All
recaptures in the shark nets should be reported; how-
ever, following the introduction of protective legisla-
tion, recaptures by big-game anglers may be unre-
ported for fear of prosecution. Some spear fishermen
are known to shoot sharks that threaten them. Any
tagged sharks killed in this way may pass unnoticed
by the diver, and hence may not be reported.
5.
The distribution of tagged fish or the fishing
effort is random. Despite the high degree of site fidel-
ity discussed earlier, the average time to recapture of
0.754 year is ample time to allow for the mixing of
tagged and untagged sharks. The recapture of a shark
1 day after release was excluded from the popula-
tion analyses, because there was insufficient time for
mixing. Fishing effort is not random, as the nets are
permanently installed at fixed localities, while the
efforts of T.K.W. are concentrated in the Struis Bay
area.
6. Recruitment or emigration is negligible. Recruit-
ment of newborn white sharks to the fishery may be
balanced by a combination of natural mortality and
inaccessibility of particularly large specimens. Tag-
ging occurred between Richards Bay, KwaZulu-Na-
tal,
and Struis Bay, Western Cape, or only part of the
shark's southern African range. There will be consid-
erable movement of sharks into and out of the tag-
ging region, resulting in a gradual reduction in the
proportion of tagged animals in the study area.
The Petersen estimate of 1279 sharks (Table III) ap-
plies only to the region between Richards Bay and
Struis Bay and excludes all the northern and much of
the western Cape coasts, where the many colonies of
South African fur seal Arctocephalus piisillus pusilliis
(Oosthuizen and David, 1988) may attract a large
number of white sharks. The estimate is fairly insen-
sitive to decreased values of the unknown compo-
nent of the annual catch, which, when halved, results
in a 14% decline in the population, that is, to 1098
sharks. The coefficient of variability (24%) of the esti-
mate is low, considering the small sample sizes.
In this study, F - 0.055, Z = 0.53, and Z - f - 0.48
year"^ which represents the sum of M and U, where
M is the instantaneous rate of natural mortality and U
is the sum of the instantaneous rate of emigration and
long-term tag shedding. M is unknown, but is likely
to be low for this apex predator, with its apparent
slow growth and low fecundity. In the porbeagle
shark Lamna nasus, another member of the family
Lamnidae, M = 0.18 year^^ (Aasen, 1963). M is un-
likely to exceed this value in white sharks of the size
range tagged in this study. This results in U = 0.3
year '. As mentioned above, long-term tag loss is
likely to be low, and emigration, whereby tagged
white sharks move out of the study region, is the
major component of U.
A possible yardstick in assessing the validity of
protective legislation is to ensure that fishing mortal-
ity, now mainly that due to shark nets, does not ex-
ceed natural mortality. In this study, f is considerably
lower than the sum of M and U. Many of the sharks
caught in this study were released alive; conse-
quently, the real F may be lower than 0.055 year '
and, in our opinion, does not represent overfishing of
white shark stocks. Improved estimates of mortality
and emigration are needed, however, before relax-
ation of the current protective legislation can be con-
sidered.
Summary
The ORI Tagging Programme tagged 73 white
sharks C. carcharias in South African waters between
January 1989 and December 1993. Anglers in temper-
ate Cape waters tagged 48 (66%) of the sharks; the
remainder were tagged by the NSB. Cape specimens
were larger than those from KwaZulu-Natal; most of
the sharks were 150-400 cm PCL. Six of the sharks
(8.2%) were recaptured within a mean of 275 days
(range, 1-527 days) and a mean distance traveled of
365 km (range, 0-1409 km). A modified Petersen esti-
mate was used to determine the size of the white
shark population for each of the 5 years of the study.
Allowance was made for capture-induced mortality
and the rerelease of tagged sharks that were recap-
tured. Fishing mortality was assumed to be constant,
despite the introduction of protective legislation in
1991.
The overall estimate was 1279 sharks (95% con-
fidence limits, 839-1843) for the region Richards Bay
in KwaZulu-Natal to Struis Bay in Western Cape.
Mortality rates were estimated as f = 0.055 year '
(95%
confidence limits, 0.015-0.10) and Z = 0.53

400 GEREMY
CLIFF
ET AL.
year^^ (95% confidence limits, 0.42-0.66). Improved
estimates of mortality are needed before any relax-
ation of the protective legislation can be considered.
Acknowledgments
We are indebted to the members of the ORI Tagging Pro-
gramme, particularly the field staff of the NSB, who have tagged
white sharks. The financial support of Stellenbosch Farmers Win-
ery and the Southern Africa Nature Foundation, sponsors of the
tagging program, is gratefully acknowledged. A. E. Punt, L. Beck-
ley, S. F. J. Dudley, and V. Peddemors commented on the manu-
script. The senior author (G.C.) thanks the National Audubon Soci-
ety and the Steinhart Aquarium, California Academy of Sciences,
for financial assistance in attending the symposium.