Content uploaded by Albrecht Götz
Author content
All content in this area was uploaded by Albrecht Götz on Aug 21, 2014
Content may be subject to copyright.
African Journal of Marine Science 2011, 33(1): 107–113
Printed in South Africa — All rights reserved
Copyright © NISC (Pty) Ltd
AFRICAN JOURNAL OF
MARINE SCIENCE
ISSN 1814–232X EISSN 1814–2338
doi: 10.2989/1814232X.2011.572357
African Journal of Marine Science is co-published by NISC (Pty) Ltd and Taylor & Francis
An alternative method for estimating the status of resident reef fish stocks,
based on differential fishing effort across a marine reserve boundary
A Götz1*, SE Kerwath2, CG Attwood3 and WHH Sauer4
1 Elwandle Node, South African Environmental Observation Network (SAEON), Private Bag 1015, Grahamstown 6140,
South Africa
2 Branch Fisheries, Department of Agriculture, Forestry and Fisheries, Private Bag X2, Rogge Bay 8012, South Africa
3 Marine Research Institute, Zoology Department, University of Cape Town, Private Bag Rondebosch 7700, South Africa
4 Department of Ichthyology and Fisheries Science, Rhodes University, PO Box 94, Grahamstown 6140, South Africa
* Corresponding author, e-mail: albrecht@saeon.ac.za
Manuscript received August 2010; accepted December 2010
The stock status of roman Chrysoblephus laticeps was estimated in the Goukamma, a temperate
South African marine protected area (MPA). Standardised catch per unit effort (CPUE) from a control-
led angling survey on both sides of the MPA border was employed to extrapolate the CPUE at zero
fishing mortality. Converted into biomass, the estimate (61% of unexploited biomass) lay midway
between those of two biomass-per-recruit (B/R) models for the same population based on angling
and diving surveys (i.e. 52% and 69% of unexploited biomass respectively). The extrapolated CPUE at
zero fishing mortality (4.4 fish angler-hour–1) in this study compared well with the mean CPUE of 4.6
fish angler-hour–1 determined during a concomitant survey in the core area of the nearby Tsitsikamma
MPA — the oldest, and one of the largest, MPAs in Africa. Extrapolations of CPUE have the potential
to deliver reliable and consistent estimates of stock status and could offer a practical alternative to
conventional B/R models.
Keywords: biomass-per-recruit models, Chrysoblephus laticeps, CPUE extrapolation, fisheries management, marine
protected area, stock status
The world’s fisheries provide employment to a large
number of people and are an important source of food for
an exponentially increasing human population (Garibaldi et
al. 2004). Marine resources are threatened by a number of
factors, but the impact of fishing exceeds all other threats
and has reduced the majority of fish stocks to well below
maximum productivity levels (Jackson et al. 2001, Carr et
al. 2003, Worm et al. 2009, Froese and Proelß 2010).
There are widespread concerns about the ability of
conventional measures to manage fisheries sustainably
(Bohnsack 1993, Ludwig et al. 1993, Worm et al. 2009). This
has led to intensive investigations into conceptually different
strategies such as marine protected areas (MPAs), which
have long been advocated as a viable alternative (Bohnsack
and Ault 1996, Clark 1996, Roberts 1997, Guénette and
Pitcher 1999, Gell and Roberts 2003).
A major drawback of conventional fisheries management
strategies is their dependence on reliable biological reference
points (Caddy and Mahon 1995), not least because of the
difficulty of separating fishing-induced mortality from natural
mortality. For studies on highly resident fish, MPAs may
offer an alternative means of estimating biological reference
points, because they generally provide areas with differential
fishing effort.
This study uses data from a survey of a population of
roman Chrysoblephus laticeps — a resident temperate
reef fish endemic to South Africa — in and around a small
MPA. An alternative method, based on the assumption of a
linear relationship between standardised catch per unit effort
(CPUE) and abundance, is applied to estimate the stock
status of this roman population. The result is evaluated by
comparison with results obtained from conventional biomass-
per-recruit (B/R) models.
Material and methods
Study area
The Goukamma ‘no-take’ MPA, situated on the south coast
of South Africa, extends 14 km along the shoreline and one
nautical mile (1.852 km) offshore (Figure 1). Established
10 years before the start of this study, it includes about
40 km2 of hard and soft substrates. The study area included
Introduction
Götz, Kerwath, Attwood and Sauer
108
the MPA and surrounding areas extending 27 km along
the coast and 5 km offshore. The subtidal reefs around
Goukamma are targeted by a multispecies commercial and
recreational boat-based linefishery. The bulk of the catch
constitutes several endemic sparid species, of which roman
is the main one targeted (Götz 2005).
CPUE model
Fish abundance can be affected by fishing, density-
dependent movement of fish, and reduced recruitment.
CPUE is often used as an indicator for reef fish abundance
(e.g. Bennett and Attwood 1991, Millar and Willis 1999). As
no detailed pre-exploitation data on CPUE of reef fish are
available for South Africa, the reduction of CPUE due to
reduced recruitment cannot be determined. In theory, the
core area of a large MPA in existence for a suitable length
of time could be used to obtain an estimate of CPUE in a
contemporary unexploited population. However, due to its
large size, control sites (exploited sites outside the MPA),
which are necessary to determine the relative reduction in
CPUE caused by exploitation, would be too far away and
not likely to be comparable, owing to habitat and oceano-
graphic differences (Tobler 1970, Booth 2004). On the other
hand, CPUE determined in a small MPA, which may have
suitable control sites nearby, can be influenced by illegal
fishing and by density-dependent emigration of post-recruit
fish (‘spill-over’ effects: Figure 2a).
In this study, the effect of post-recruit spill-over is negli -
gible due to the small home range (<100 m linear extent)
and high residency (>91%) of roman (Kerwath et al. 2007a,
2007b, 2008). However, if illegal fishing occurs inside the
Goukamma MPA, not only the level of exploitation outside,
but also within the MPA, would have to be determined. Such
information would enable an estimate of a contemporary
‘pristine CPUE’, by regressing through CPUE values in
areas with differential fishing mortality (Figure 2b). We
understand the term ‘pristine CPUE’ as a surrogate for fish
abundance in an area with no extractive fishing pressure,
determined by catch and release (research) angling with a
negligibility low post-release mortality.
Exploitation in the study area
A systematic survey of fishing boat numbers and positions
inside and outside the Goukamma was undertaken between
2000 and 2004. Using a small power boat, the research team
approached each fishing boat in the study area to determine
their GPS positions. Results from this survey, which covered
76 boat-count days and 154 recorded fishing boat positions,
are reported in Götz (2005) and illustrated in Figure 3. On
a given research day, this spatial survey of exploitation
preceded the biological survey work, so as to avoid fishing
boats leaving their initial positions due to our presence in
the study area. Many fishers fish as close as possible to the
border of the MPA and occasionally undertaking deliberate
AFRICA
South
Africa
SOUTH
AFRICA
5 km
Cape Town Port Elizabeth
Plettenberg Bay
Swartvlei Lagoon
Knysna Lagoon
Goukamma
Knysna Heads
Study area
GMPA
Buffels Bay
WESTERN CAPE
WESTERN CAPE
AT L AN T IC
OCEAN
INDIAN OCEAN
INDIAN OCEAN
34° S
20° E 24° E
N
WE
S
Figure 1: Map of the study area, indicating the Goukamma MPA (GMPA)
African Journal of Marine Science 2011, 33(1): 107–113 109
boundary violations. Such fishing boats were often encoun-
tered inside the MPA, but were mostly limited to its outer
edge. Therefore, for this analysis, the study area was divided
into two zones: a ‘core area’ in which fishing boats were least
often recorded (0.3 boats km–2) and a ‘border area’ where
fishing boats were most often encountered (2.7 boats km–2).
Both areas were comparable in size (~27 km2), bathymetry
and oceanographic conditions (i.e. current, temperature and
turbidity; Götz et al. 2009a). The reef complex was continuous
across the two areas and of equivalent depth and profile.
CPUE assessment
The fishery-independent CPUE assessment was conducted
by a research team, at pre-selected stations, from a small
power boat (Götz et al. 2007). Stations were stratified
across area (core/border), season (summer/winter) and time
Before
exploitation
(19th century)
Far from
MPA
(21st century)
Large MPA
(21st century) Small MPA
(21st century) Close to
small MPA
(21st century)
Large MPA
(21st century)
Small MPA
(21st century)
Close to small MPA
(21st century)
Reduced recruitment
Reduced recruitment
Spill-over loss
Illegal fishing
Reduced recruitment
Spill-over gain
Full exploitation
Reduced recruitment
Full exploitation
Measured difference
No data Incomparable
sites
Incomparable
sites
Comparable
sites
CPUECPUE
Illegal fishing
Full exploitation
No spill-over effects, reduced recruitment
Spill-over effects, reduced recruitment
Highest
CPUE under
reduced recruitment
Extrapolation Extrapolation
(a)
(b)
Figure 2: (a) Hypothetical rates of CPUE at sites with different rates of exploitation and (b) their usefulness for stock assessment
Götz, Kerwath, Attwood and Sauer
110
of day (morning/afternoon). Researchers were trained in the
controlled angling procedure to reduce the effects of skill
differences and the composition of the angling team was
varied as little as possible. Standard hook sizes and baits
were used and the standard effort per station was one hour.
Two baited hooks were used per line. All fish captured were
measured and returned. The depth at sampling stations
was determined using an echo-sounder; temperature and
turbidity measurements were taken using a bathy-turbidity-
thermograph.
Generalised linear models (GLMs) were applied to
separate the effects of measured parameters on the roman
catch. The fit of different models was assessed using the
Akaike information criterion (AIC; Akaike 1973) to find the
optimal combination of parameters. The ‘Wald’ statistic (W)
and its p-level were applied to test the significance of each
regression coefficient (Harrell 2001). The catch of roman
in the study area was modelled including the following
parameters (after Götz et al. 2008):
(1)
where
β
i values were the estimated coefficients (McCullagh
and Nelder 1995). Effort was not included in the GLM model
because equal effort was applied at controlled angling
stations.
CPUE extrapolation
The average survey CPUE was calculated at the two levels
of effort corresponding to the two zones, core and border.
These two points were used to infer the CPUE value at
zero fishing effort, by calculating the intercept of the linear
regression fitted to the two points. The standardised CPUE,
expressed as number of roman per angler-hour, together
with the actual fishing effort (fishing boats km–2), were used
to regress the theoretical CPUE value of an unexploited
population:
CPUE (unexploited) = CPUE (exploited) + a × fishing effort
(2)
where ‘a’ is the slope of the linear regression based on the
data from the two areas (core and border).
Confidence limits for CPUE values were estimated by
applying a conditioned parametric bootstrap procedure
(Efron and Tibshirani 1986). For this purpose, two vectors of
random Poisson variables were generated 1 000 times from
the predicted CPUE values for the core and border areas
respectively. The length of each vector was governed by the
number of observations available per area. The linear regres-
sion was then refitted to the means of the two vectors, and
the slope, intercept and corresponding CPUE at zero fishing
effort were noted. The percentile method (Buckland 1984)
was used to estimate 95% confidence intervals from the
resulting bootstrap vectors; the 2.5% and 97.5% percentiles
were chosen to obtain the lower and upper 95% confidence
intervals respectively.
To compare results from the CPUE extrapolation to those
from B/R models, numbers of roman were converted to
mass:
(3)
where N is the number of roman in the sample. The length
measurements of the individual fish in the cumulative catch
from the respective areas (core and border Ntotal = 807) were
converted to mass according to the length-mass relationship
based on a retained sample (Nsample = 287) taken at the end
of the study period within a 10-day interval.
B/R model
Of the 287 fish in the retained sample, the otoliths of 250
were readable (Nsample(core) = 119, Nsample(border) = 168) and
used for age and growth calculations (Götz 2005). The von
Bertalanffy growth equation was applied to model length-
at-age:
(4)
where Lt is length at time t, L∞ is the theoretical asymptotic
length, K is the body growth coefficient, and t0 the age of
fish at zero length. To estimate growth parameters, general-
ised non-linear least squares procedures were applied.
Parameter variance was calculated through parametric
bootstrapping by re-sampling 500 bootstrap replicates,
and the first-order bias-corrected percentile method (Efron
and Tibshirani 1993) provided confidence intervals. The
same non-linear least squares procedure was fitted to the
age–length samples, using the Schnute growth equation, in
order to determine which equation modelled growth more
realistically. The fit of the two models was compared using
a likelihood ratio test (after Götz et al. 2008). Data collected
in the core and border areas were fitted separately and
differences in growth parameters between the two areas
were tested using a likelihood ratio test (Draper and Smith
1966).
log(catch) =
β
0+
β
1(depth) +
β
2(temperature) +
β
3(turbidity)
+
β
4(season) +
β
5(time of day) +
β
6(area) +
ε
[ ]
1
1
weight
CPUE [ angler-hour ]
CPUE[ angler-hour ]
N
i
i
g
g N
N
−
−
⎛⎞
⎜⎟
⎜⎟
= ⎜⎟
⎜⎟
⎜⎟
⎝⎠
×
∑
0
()
(1 e )
Kt t
t
LL −−
∞
=−
N
WE
S
Position of fishing boats (n = 154)
Core
Border
Study area
Core (27 km2) = core area (0.3 boats km−2)
Border (27 km2) = border area (2.7 boats km−2)
2 km
Figure 3: Map of part of the study area with positions of fishing
boats (crosses) indicating the core and border areas. The relative
fishing effort per area is shown (after Götz et al. 2009b)
African Journal of Marine Science 2011, 33(1): 107–113 111
Age-specific selectivity was modelled using a logistic
model (Butterworth et al. 1989), described as:
(5)
where St is the selectivity of the gear on a fish of age t, t50 is
the age-at-50% selectivity, and
δ
is the rate of change.
Total mortality rates (Z) were estimated from catch-
curves (Ricker 1975) derived from the CPUE assessment
and from underwater length frequency data gathered during
a diving survey over the same period (Götz et al. 2008).
The length frequency distributions were converted to age
frequency distributions according to the age-at-length key
based on otolith readings (Götz et al. 2008). The slope of a
straight line fitted to points on the descending limb provided
estimates of Z. Similarly, natural mortality (M) estimates
were derived from the angling and underwater diving data
gathered in the core area of the MPA. Although the core
area was lightly fished, we considered these estimates of
M as more acceptable than those derived from the Pauly
relationship. Fishing mortalities (F) were then calculated
by subtraction (i.e. F = Z – M). Mortality estimates from the
CPUE assessment and the diving survey were tested for
differences using a homogeneity-of-slopes model. Finally,
B/R was calculated as:
(6)
where Wt is the begin-year mass of a fish at age t, defined
as:
(7)
where a and b are the parameters of the mass–length
relationship.
Results
CPUE model
The core and border areas were targeted over a period of
41 months, with a total of 191 controlled angling stations
yielding a total catch of 807 roman. Within the core area,
roman CPUE was high (4.3 fish angler-hour–1), but signifi-
cantly lower in the fully exploited border area (3.5 fish
angler-hour–1) (Table 1; p < 0.001). Even at the higher catch
rates empty hooks were retrieved consistently, indicating
that a possible saturation at high fish abundances did not
occur when using this method to estimate abundance based
on a linear relationship to CPUE. The CPUE estimate for
unexploited conditions in the area, attained through extrapo-
lation, was 4.4 fish angler-hour–1 (Figure 4).
Expressed in terms of mass, the CPUE for roman caught
in the core and border areas were 4 162 and 2 713 g angler-
hour–1 respectively, due to the higher CPUE and mean size
(Götz et al. 2008) of roman in the core area. The CPUE at zero
fishing effort was estimated as 4 414 g angler-hour–1, so in
terms of biomass reduction, the stock of roman in the exploited
border area was reduced to 61% of its unexploited level.
B/R model
The fit of the von Bertalanffy growth model was not signif-
icantly different to that of the Schnute growth model (p =
0.570); therefore, the more commonly-used von Bertalanffy
equation was applied. Von Bertalanffy growth curves and
parameters were not significantly different (p(L∞; K; t0) = 0.93;
0.79; 0.52) between samples from the core and border
areas and were therefore combined for further calcula-
tions. Ages ranged between 2 and 19 years with a theoret-
ical asymptotic length (L∞) of 512.86 mm (fork length) and a
body growth coefficient (K) of 0.086 years–1. The age of fish
with zero length (t0) was determined as −1.77 years. Age-at-
50% selectivity for the linefishery was estimated at 7.60
years with a rate of change (
δ
) of 1.77 y–1.
Although the F-estimate derived from standardised
angling (0.25 y–1) was different from that derived from diving
survey data (0.16 y–1; Götz et al. 2008), both estimates were
applied due to uncertainty associated with estimates of M.
50
(/)
1
(1 e )
tt
t
S
δ
−− −
=+
()
1
max
/
t
SF M
tkt
k t
R
tR
t
B R W e
−
+
=
−
⎡⎤
∑
⎢⎥
=⎢⎥
⎢⎥
⎢⎥
⎣⎦
∑
Wt=a(Lt)bwith Lt=L∞(1−e–K(t–t0))
Effect df Wp
Intercept 1 15.16 <0.001**
Season 3 70.53 <0.001**
Area 1 16.21 <0.001**
Depth 1 18.08 <0.001**
Temperature 1 15.44 <0.001**
Turbidity 1 27.62 <0.001**
Area × Season 3 10.89 0.012*
Area × Depth 1 3.41 0.067, ns
Area × Temperature 1 3.42 0.066, ns
Area × Turbidity 1 3.90 0.063, ns
** p < 0.01
* p < 0.05
ns: not significant
Table 1: Results of the GLM analysis, covering the effects of
season, area, depth, temperature and turbidity on the catch of
roman during CPUE assessments. The parameter ‘time of day’ was
discarded during a preceding AIC best-subset analysis (after Götz
et al. 2008)
y = −0.3x + 4.4
Core area (Tsitsikamma MPA)
3.5
4
0.5 1 1.5 2 2.5
4.4
4.6
FISHING EFFORT (boats km−2)
CPUE (fish angler-hour−1)
4.5
Core area (Goukamma MPA)
Border area (Goukamma MPA)
0
Figure 4: Extrapolation of standardised CPUE for roman, and
fishing effort measured in the core and border area, to obtain an
estimate of CPUE at zero fishing mortality. Standardised CPUE for
roman in the nearby Tsitsikamma MPA is indicated. Bars denote
95% confidence intervals calculated from 1 000 bootstraps
Götz, Kerwath, Attwood and Sauer
112
Based on Z-estimates obtained from the two datasets,
the B/R models indicate a reduction of the roman biomass
in the border area to 52% and 69% of its unexploited level
respectively (Figure 5).
Discussion
Traditional single species and multispecies management
require reliable information on biological reference points
such as stock status. In the present study, a linear regres-
sion was used to determine the standardised CPUE of an
unexploited population (4.4 fish angler-hour–1). A concom-
itant study by Smith (2005), who used the same procedure
and statistical analysis used here for an assessment in the
Tsitsikamma MPA, the oldest and one of the largest no-take
MPAs in Africa (established in 1964 and encompassing an
area of 320 km2), estimated a CPUE value of 4.6 fish angler-
hour–1 from the core area of the MPA. This value compares
favourably to the theoretical value for an unfished popula-
tion, determined in the present study. An explanation for
the slightly higher roman abundance during Smith’s (2005)
assessment might be the positive effect of reserve age on
fish abundance (Claudet et al. 2008). This is plausible consid-
ering that the Tsitsikamma MPA has been protected almost
four times longer than Goukamma MPA.
The survey reported on here was designed to reduce
common limitations of CPUE and per-recruit assessments.
Controlled angling protocols were fishery-independent and
highly standardised; site selection was unbiased within an
oceanographically homogeneous study area (Götz et al.
2009a) and the data analysis model accounted for environ-
mental factors. Furthermore, estimates of Z and M were
based on two independent methods using catch-curves
(controlled angling and diving survey) and original length-
at-age keys for the B/R assessments were constructed.
Interestingly, the biomass reduction estimate based on the
simple linear regression (61% of unexploited) was midway
between the estimates of the two B/R models (52% and 69%
of unexploited).
A shortcoming of this study was that the CPUE regres-
sion was based on only two data points, but this can be
improved in cases where the spatial scale of exploitation
allows for a finer subdivision of the studied area. Irrespective
of this, considerable research angling effort is necessary to
determine the state of the stock in and around a MPA when
using this alternative method. This effort could, however,
be integrated into existing monitoring programmes, within
MPAs, that aim to evaluate their effectiveness.
Despite these limitations, the method introduced here
demonstrates the potential of MPAs as reference areas
for the stock assessment of resident fish with similar life
histories as roman. If developed further, CPUE regression
methods, such as the one reported here, should deliver
consistent stock status estimates that could offer a practical
alternative to conventional assessment strategies.
Acknowledgements — This project was funded in part by the
National Research Foundation and the Marine Living Resources
Fund. The Elwandle Node of the South African Environmental
Observation Network (SAEON) provided additional assistance.
The fieldwork would not have been possible without the support
offered by Cape Nature. Thanks are also due to Henning Winker
for assistance with some of the statistical procedures and for his
useful comments.
References
Akaike H. 1973. Information theory and an extension of the maximum
likelihood principle. In: Petrov N, Csaki F (eds), Proceedings of the
second international symposium on information theory. Budapest:
Akademiai Kiado. pp 267–281.
Bennett BA, Attwood CG. 1991. Evidence for recovery of a
surf-zone fish assemblage following the establishment of a
marine reserve on the southern coast of South Africa. Marine
Ecology Progress Series 75: 173–181.
Bohnsack JA. 1993. Marine reserves: they enhance fisheries,
reduce conflicts and protect resources. Oceanus 36: 63–71.
Bohnsack JA, Ault JS. 1996. Management strategies to conserve
marine biodiversity. Oceanography 9: 73–82.
Booth AJ. 2004. Spatial statistics and aquatic Geographic
Information Systems. In: Nishida T, Kailola PJ (eds), Proceedings
of the second international symposium on GIS/spatial analyses
in fishery and aquatic sciences. 3–6 September 2002. Brighton:
University of Sussex. pp 3–44.
Buckland ST. 1984. Monte-Carlo confidence intervals. Biometrics
40: 811–817.
Butterworth DS, Punt AE, Borchers DL, Pugh JB, Hughes GS. 1989.
A manual of mathematical techniques for linefish assessment.
South African National Scientific Programmes (SANSP) Report
Unfished
Fished
52%
Unfished
Fished
69%
20
40
60
60
80
100
120
140
160
2 4 6 8 10 12 14 16 18
20
40
80
100
AGE (y)
EXPECTED B/R (g y−1)
B/Rmax = 78%
B/Rmax = 90%
(a)
(b)
Figure 5: Expected B/R for roman, calculated from M-estimates
derived from (a) the CPUE assessment and (b) the diving survey
dataset. Biomass for the unfished (grey shaded area) and fished
(white area) populations at (a) F = 0.25 and (b) F = 0.16 are
presented. Biomass is indicated as a percentage of unexploited stock
African Journal of Marine Science 2011, 33(1): 107–113 113
No. 160. Pretoria: Foundation for Research Development; CSIR.
Caddy JF, Mahon R. 1995. Reference points for fisheries manage-
ment. FAO Fisheries Technical Paper No. 347. Rome: Food and
Agricultural Organization.
Carr MH, Neigel JE, Estes JA, Andelman S, Warner RR, Largier JL.
2003. Comparing marine and terrestrial ecosystems: implications
for the design of coastal marine reserves. Ecological Applications
13: 90–107.
Claudet J, Osenberg CW, Benedetti-Cecchi L, Domenici P, García-
Charton JA, Pérez-Ruzafa A, Badalamenti F, Bayle-Sempere
J, Brito A, Bulleri F, Culioli JM, Dimech M, Falcón JM, Guala
I, Milazzo M, Sánchez-Meca J, Somerfield PJ, Stobart B,
Vandeperre F, Valle C, Planes S. 2008. Marine reserves: size
and age do matter. Ecology Letters 11: 481–489.
Clark CW. 1996. Marine reserves and the precautionary manage-
ment of fisheries. Ecological Applications 6: 369–370.
Draper NR, Smith H. 1966. Applied regression analysis. New York:
Wiley.
Efron B, Tibshirani RJ. 1986. Bootstrap methods for standard
errors, confidence intervals, and other measures of statistical
accuracy. Statistical Science 1: 54–77.
Efron B, Tibshirani RJ. 1993. An introduction to the bootstrap.
London: Chapman & Hall.
Froese R, Proelß A. 2010. Rebuilding fish stocks no later than
2015: will Europe meet the deadline? Fish and Fisheries 11:
194–202.
Garibaldi L, Lowther A, Csirke J, Crispoldi A, Smith A, Kelleher
K, Doulman D, Hishamunda N, Subasinghe R, Bartley DM,
Vannuccini S, Laurenti G, Josupeit H. 2004. The state of world
fisheries and aquaculture. Rome: FAO Publishing.
Gell FR, Roberts CM. 2003. The fishery effects of marine reserves
and fishery closures. Washington, DC: World Wildlife Fund.
Götz A. 2005. Assessment of the effect of Goukamma Marine
Protected Area on community structure and fishery dynamics.
PhD thesis, Rhodes University, South Africa.
Götz A, Kerwath SE, Attwood CG, Sauer WHH. 2007. Comparison
of the effects of different linefishing methods on catch composi-
tion and capture mortality of South African temperate reef fish.
African Journal of Marine Science 29: 177–185.
Götz A, Kerwath SE, Attwood CG, Sauer WHH. 2008. Effects
of fishing on population structure and life history of roman
Chrysoblephus laticeps (Sparidae). Marine Ecology Progress
Series 362: 245–259.
Götz A, Kerwath SE, Attwood CG, Sauer WHH. 2009a. Effects
of fishing on a reef community in South Africa II: benthic
invertebrates and algae. African Journal of Marine Science 31:
253–262.
Götz A, Kerwath SE, Attwood CG, Sauer WHH. 2009b. A change
of the seaward boundary of Goukamma Marine Protected Area
could increase conservation and fishery benefits. South African
Journal of Science 105: 330–331.
Guénette S, Pitcher TJ. 1999. An age-structured model showing
the benefits of marine reserves in controlling overexploitation.
Fisheries Research 39: 295–303.
Harrell FEJ (ed.). 2001. Regression modelling strategies: with
applications to linear models, logistic regression, and survival
analysis. New York: Springer Verlag.
Jackson JBC, Kirby MX, Berger WH, Bjorndal KA, Botsford LW,
Bourque BJ, Bradbury RH, Cooke R, Erlandson J, Estes
JA, Hughes TP, Kidwell S, Lange CB, Lenihan HS, Pandolfi
JM, Peterson CH, Steneck RS, Tegner MJ, Warner RR.
2001. Historical overfishing and the recent collapse of coastal
ecosystems. Science 293: 629–638.
Kerwath SE, Götz A, Attwood CG, Cowley PD, Sauer WHH. 2007a.
Movement pattern and home range of roman Chrysoblephus
laticeps. African Journal of Marine Science 29: 93–103.
Kerwath SE, Götz A, Attwood CG, Sauer WHH. 2007b. Area
utilization and activity patterns of roman Chrysoblephus laticeps
(Sparidae) in a small marine protected area. African Journal of
Marine Science 29: 259–270.
Kerwath SE, Götz A, Attwood CG, Sauer WHH. 2008. The
effect of marine protected areas on an exploited population of
sex-changing temperate reef fish: an individual-based model.
African Journal of Marine Science 30: 337–350.
Ludwig D, Hilborn R, Walters C. 1993. Uncertainty, resource
exploitation, and conservation: lessons from history. Science
260: 17–36.
McCullagh P, Nelder JA (eds). 1995. Generalized linear models.
London: Chapman & Hall.
Millar RB, Willis TJ. 1999. Estimating the relative density of snapper
in and around a marine reserve using a log-linear mixed-effects
models. Australian and New Zealand Journal of Statistics 41:
383–394.
Ricker WE. 1975. Computation and interpretation of biological
statistics of fish populations. Bulletin of the Fisheries Research
Board of Canada 191: 1–382.
Roberts CM. 1997. Ecological advice on the global fisheries crisis.
Trends in Ecology and Evolution 12: 35–38.
Smith MKS. 2005. Towards a new approach for coastal governance
with an assessment of the Plettenberg Bay linefisheries. MSc
thesis, Rhodes University, South Africa.
Tobler WR. 1970. A computer movie simulating urban growth in the
Detroit region. Economic Geography 46: 234–240.
Worm B, Hilborn R, Baum JK, Branch TA, Collie JS, Costello C,
Fogarty MJ, Fulton EA, Hutchings JA, Jennings S, Jensen OP,
Lotze HK, Mace PM, McClanahan TR, Minto C, Palumbi SR,
Parma AM, Ricard D, Rosenberg AA, Watson R, Zeller D. 2009.
Rebuilding global fisheries. Science 325: 578–585.
