Patterns of long-term climate variability and predation rates by a marine apex predator, the white shark Carcharodon carcharias

Abstract

Understanding potential responses of aquatic animals to climate variability is important, given the wide-ranging implications of current and future climatic change scenarios. Here, we used long-term data from natural predator?prey interactions between white sharks Carcharodon carcharias and Cape fur seals Arctocephalus pusillus pusillus in False Bay, South Africa, paired with environmental monitoring to examine potential relationships between temperature variability and shark predation rates on seals. Based on generalized linear modelling of a dataset of 941 shark attacks on seals collected over 15 years (1999?2013) during the austral winter (May?September) season, we found water temperature was included as a significant predictor of daily and monthly variability in predation rates. However, the signal of temporal variability over the season emerged as a more predominant predictor. Moreover, inter-annual variability in predation rate appeared linked to other environmental factors (wind, water visibility, and the occurrence of El Niño and La Niña events) rather than water temperature. These data suggest that water temperatures on an intra-annual scale might contribute to predation patterns in white sharks either directly or indirectly (e.g. due to associated changes in prey availability), but do not implicate water temperature as a primary driver in this scenario, or at an interannual scale. It is possible that (1) the metabolic demand of white sharks may be modulated against temperature variability by their partially endothermic nature, and (2) the predation patterns of white sharks on seals are the result of a complex interplay between ambient physical conditions and broader oceanographic, biological, and ecological factors.

Full text

MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 587: 129–139, 2018
https://doi.org/10.3354/meps12424 Published January 25
INTRODUCTION
The world’s aquatic habitats are experiencing sig-
nificant physiochemical shifts due to human-induced
climate change. Warming, deoxygenation, and acidi-
fication associated with rising atmospheric CO2can
impose physiological stressors on aquatic animals
that, in turn can increase extinction risk at the local
and global scales (Hoegh-Guldberg & Bruno 2010).
Therefore, understanding the potential effects on
species in response to climate variability is critical at
this time (Sunday et al. 2012), especially with respect
to higher trophic-level predators, such as many large
sharks, which can play an important role in regulat-
ing ecosystem structure and health (Ruppert et al.
2013, Barley et al. 2017).
While a variety of climate variables can impact the
biology and physiology of fish, arguably the most
significant on diurnal to interannual timescales is
ambient water temperature. For instance, a species’
© Inter-Research 2018 · www.int-res.com*Corresponding author: ras347@miami.edu
Patterns of long-term climate variability
and predation rates by a marine apex predator,
the white shark Carcharodon carcharias
R. A. Skubel1,2,*, B. P. Kirtman1, 3, C. Fallows4, N. Hammerschlag1, 2
1Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, FL 33149, USA
2Abess Center for Ecosystem Science & Policy, University of Miami, Miami, FL 33146, USA
3Cooperative Institute for Marine and Atmospheric Science, University of Miami, FL, 33149, USA
4Apex Expeditions, Cape Town, South Africa
ABSTRACT: Understanding potential responses of aquatic animals to climate variability is impor-
tant, given the wide-ranging implications of current and future climatic change scenarios. Here,
we used long-term data from natural predator−prey interactions between white sharks Carcharo-
don carcharias and Cape fur seals Arctocephalus pusillus pusillus in False Bay, South Africa,
paired with environmental monitoring to examine potential relationships between temperature
variability and shark predation rates on seals. Based on generalized linear modelling of a dataset
of 941 shark attacks on seals collected over 15 years (1999−2013) during the austral winter
(May−September) season, we found water temperature was included as a significant predictor of
daily and monthly variability in predation rates. However, the signal of temporal variability over
the season emerged as a more predominant predictor. Moreover, inter-annual variability in pre-
dation rate appeared linked to other environmental factors (wind, water visibility, and the occur-
rence of El Niño and La Niña events) rather than water temperature. These data suggest that
water temperatures on an intra-annual scale might contribute to predation patterns in white
sharks either directly or indirectly (e.g. due to associated changes in prey availability), but do not
implicate water temperature as a primary driver in this scenario, or at an interannual scale. It is
possible that (1) the metabolic demand of white sharks may be modulated against temperature
variability by their partially endothermic nature, and (2) the predation patterns of white sharks on
seals are the result of a complex interplay between ambient physical conditions and broader
oceanographic, biological, and ecological factors.
KEY WORDS: Predator ecology · Climate variation · El Niño · Shark · Seal · Temperature
Resale or republication not permitted without written consent of the publisher

Mar Ecol Prog Ser 587: 129–139, 2018
movements may be limited to water temperatures
within a fundamental thermal range defined by
physiological tolerance (Kearney & Porter 2009,
Allen-Ankins & Stoffels 2017) and, among elasmo-
branchs, individuals may behaviorally thermoregu-
late by moving among different water temperatures
to optimize hunting, digestion, and gestation (Hight
& Lowe 2007, Di Santo & Bennett 2011, Ketchum et
al. 2014, Papastamatiou et al. 2015). Experimental
exposure of sharks and other aquatic predators to
warmer water temperatures in the laboratory have
shown increased hunting effort and feeding rate
(Dowd et al. 2006, Miller et al. 2014, Pistevos et al.
2015); however, comparative studies in the wild
involving large mobile fish are lacking due to logisti-
cal and technological challenges. Moreover, most
experiments have been over relatively short dura-
tions (weeks to months) and have not been able to
provide insight into how long-term (years to decades)
climate variability may directly impact ecological
processes, such as predation intensity. Thus, it re -
mains unknown how long-term climate cycles and
temperature variability may impact the biology and
ecology of aquatic top predators.
As with other members of the lamnidae family,
white sharks Carcharodon carcharias are partially
endothermic elasmobranch fish, with stomach tem-
peratures measured up to 14.3°C higher than that of
the surrounding water (Goldman 1997). Given that
white sharks can modulate their body temperature to
some extent, within a narrow range (e.g. 23.4− 26.7°C,
Goldman 1997), recurrent predation patterns are
likely due to behavioral optimization of predation for
energy intake rather than for a need to maintain opti-
mal body temperature (as opposed to teleost fishes
(Pink et al. 2016)). False Bay, South Africa, is the
location of a long-term monitoring program on the
predation rates of white sharks on Cape fur seals
Arctocephalus pusillus pusillus (Martin et al. 2005,
2009, Hammerschlag et al. 2006, 2012, 2017, Fallows
et al. 2012, 2013, 2016, Martin & Hammerschlag 2012)
as well as a site of ongoing climate and temperature
monitoring (Dufois & Rouault 2012). Accordingly, the
system provides an opportunity to investigate for pos-
sible relationships between climate-driven tempera-
ture variability and predation rates by a marine apex
predator.
The climate of the study region is influenced by sea-
surface temperature (SST) leakage from the Agulhas
current (known as Agulhas leakage), which transports
warm and saline waters from the Indian Ocean (Gor-
don & Haxby 1990, de Ruijter et al. 1999, van Sebille
et al. 2009, Beal et al. 2011). However, recent climate
modeling has revealed that El Niño-Southern Oscilla-
tion (ENSO) can influence interannual variability of
Agulhas leakage SST with a 2 yr time lag, which sub-
sequently affects regional climate variability (Putrasa-
han et al. 2016), including ENSO effects on water
temperatures in False Bay (Dufois & Rouault 2012).
Here, we compared changes in white shark predation
patterns, in terms of successful and unsuccessful pre-
dation attempts on seals, in relation to environmental
and temporal variability and ENSO events, over a
15 yr period. In our analysis, explanatory environ-
mental and temporal variables tested against shark
predation rates included water temperature and visi-
bility at the site, and the variables day of year (DOY),
month, and year to account for change over time
MATERIALS AND METHODS
Predation events
Seal Island (35° 8’ 6”S, 18° 35’00” E) in False Bay,
South Africa, is inhabited by approx. 60 000 Cape fur
seals (Fig. 1). White sharks Carcharodon carcharias
aggregate here during the winter to actively hunt
Cape fur seals Arctocephalus pusillus pusillus when
they leave and return from Seal Island (Martin et al.
2005, 2009). The seals forage in False Bay or up to
30 km (or greater) offshore, returning to Seal Island
at irregular intervals (Martin et al. 2005, Fallows et
al. 2012). Seals travel to and from the island via por-
poising at the surface (Fallows et al. 2012). Thus,
white shark attacks on seals occur at the water sur-
face, where they can be documented (Hammerschlag
et al. 2006, Martin & Hammerschlag 2012). Attacks
are concentrated on the southern side of the island,
close to shore (within 2 km). Thus, by positioning at
the south end of Seal Island where most predatory
activity occurs, a single vessel can survey at least
270° uninterrupted to a distance of at least 3.5 km
(Hammerschlag et al. 2006, Martin et al. 2009). Ob -
servations were made at Seal Island each month, by
1 boat, between 1999 and 2013. At least part of the
research team averaged some 200 d per year on the
water, although the majority of observation occurred
during austral winter months (May− September),
when most predatory activity occurs (Hammerschlag
et al. 2006). The research vessel arrived and began
observations at Seal Island at about 07:00 h (~1.5 h
before sunrise), sea conditions permitting. Although
multiple people were on the observation boat each
day, the data in this study was restricted to that gath-
ered by a trained observer (C. Fallows) onboard.
130

Skubel et al.: Long-term climate and shark predation trends
As described in Fallows et al. (2012), predatory
events were detected from the surface by one or
more of the following: (1) white shark breach with a
seal in its mouth or a seal leaping away from its
mouth; (2) a sudden change in the travel behavior of
seals, switching from directional porpoising to zigzag
evasive maneuvers with a shark in pursuit; (3) a
splash accompanied by a blood stain, oil slick, a dis-
tinctive odor, and by any of the following indicators
such as a floating seal head or entrails floating on the
surface or trailing from the gill openings of a white
shark in the immediate vicinity and/or highly local-
ized plunge-diving black-backed kelp gulls Larus
dominicanis vetula picking up and feeding on seal
entrails. Any subsurface kills could be detected by
the appearance of a bloodstain at the surface and
floating seal entrails (Martin et al. 2005). Observed
predatory events were recorded and classified as
unsuccessful, in which the seal escaped, or success-
ful, in which the seal was consumed (Hammerschlag
et al. 2006).
While observations were made year-round, there
were more frequent trips during the winter months
when white sharks aggregate at Seal Island to
actively hunt juvenile Cape fur seals (Hammerschlag
et al. 2006). To standardize data analyses when pre-
dation activity peaked and observational effort was
consistent, we restricted our analysis to predation
data collected between 1999 and 2013, from 07:00 h
to 09:30 h, for the months of May to September. Fol-
lowing Fallows et al. (2016), we considered this 120−
150 min observational period (approx. ±1 h of sun-
rise) per day as a sample. Upon arrival at the study
site, water temperature was recorded each day using
the vessel’s onboard temperature sensor (Furu no
model 1870). Wind speed observational estimates were
determined daily from windfinder.com for measure-
ments taken at Muizenberg (8 km from Seal Island),
then, once arriving at Seal Island, wind speed in
5 knot bins was estimated and recorded. Water visi-
bility was estimated daily using an anchor line sus-
pended into the water column as a reference.
SST and ENSO status
Monthly daytime SSTs for the study period were
retrieved from Pathfinder Version 5.0 (www.nodc.
noaa. gov/SatelliteData/pathfinder4km/) for 1999− 2002
and MODIS Aqua for 2003−2013 (NASA JPL) as
there was not one contiguous data set available for
the study period. Both Pathfinder and MODIS pro-
vided 4 km resolution daytime SST, which was pre-
ferred to a continuous coarser resolution source for
SST (e.g.ERSSTv4; Huang et al. 2015). For Pathfinder
SST, the quality of a measurement when considering
pixel clouding ranges from 0 to 7, with 4 the mini-
mum quality considered acceptable (Kilpatrick et al.
2001); following convention, we only used SST meas-
urements with a quality of 4 or greater. The area
included in this analysis is indicated in Fig. 1. The
SST grid extended 5 km from the viewing area at the
south point of Seal Island, to encompass the 3.5 km
visual range of the observers. Monthly ENSO status
was obtained via the Multivariate ENSO Index (MEI)
maintained by the National Oceanic and Atmo -
spheric Administration’s Earth System Research Lab-
oratory (NOAA ESRL). The MEI is a bimonthly index
extending from 1950 to present, reflecting the first
unrotated principal component of 6 variables (sea
131
Fig. 1. Study area of Seal Island in False Bay, South Africa. The
area indicated by the dashed square is the 10 ×10 km grid over
which sea surface temperature was evaluated (−34.09° to
−34.18° N, 18.53° to 18.64° E). The observation area south of
Seal Island is indicated. Imagery is provided by ESRI

Mar Ecol Prog Ser 587: 129–139, 2018
level pressure, zonal and meridional components of
surface wind, sea surface temperature, surface air
temperature, and cloudiness fraction of the sky)
(Wolter & Timlin 1993, 1998). ENSO classification
was based on historically ranked MEI values (1
through 67), with the upper quartile (54−67) indica-
ting an El Niño and lower quartile (1−14) indicating a
La Niña.
Data analysis
For each day, kills (K) and unsuccessful attempted
predations (AP) by white sharks upon seals were
averaged over the sample period of 07:00−09:30 h to
calculate an hourly rate of kills and total predations.
This yielded a single value of kill rate (K h−1) and total
predation rate (K+AP h−1) for each day of observation
within the standardized sampling period. The broader
(K+AP) h−1 metric showed hunting effort, while the K
h−1metric indicated presumed seal consumption by
the animal, and putatively, some extent of metabolic
demand fulfilment. Given that each of Kh−1and
(K+AP) h−1 were assigned on a daily basis, it was not
feasible to directly compare the 2 metrics as a direct
measure of relative predation success. Rather, each
was considered independently as a measure of gen-
eral predation patterns. Daily, monthly, and annual
scale generalized linear models (GLMs) were used to
explore relationships of kill rates and total predation
rates with environmental variability. For the daily
GLM, explanatory variables tested were water tem-
perature (Tw, °C) and water visibility (Vis, m) meas-
ured at the site, and the variables day of year (DOY),
mo, and yr to account for change over time. For the
monthly GLM, explanatory variables tested were
monthly means of Tw, Vis, and wind speed (wind,
knots), and an indication of ENSO status via the MEI.
GLMs were constructed in Matlab using a stepwise
procedure with interactions, using the default method
of ‘sse’ to determine model criterion. Here, the func-
tion ‘stepwiseglm’ begins with an initial model con-
taining all terms, then uses forward and backward
stepwise regression to construct the final model,
comparing the explanatory power of these smaller or
larger models via the p-value of an F-test of change
in sum of squared error (SSE) with addition or
removal of terms. Terms themselves are added or
removed based on whether the term would have a
zero coefficient once added to the model (or the term
is removed from the model if it has a zero coefficient
when included, e.g. in the initial model or with suc-
cessive steps). Interaction terms are not added unless
both constituent terms are in the model, after which
point the single terms may be removed. For the
annual GLM, explanatory variables tested were
annual means of Tw, Vis, and wind, the standard
deviation of annual Twto look for effects of variabil-
ity, and whether an El Niño or La Niña event had
occurred that year from May−September (indicated
by a binary 0 for no, and 1 for yes). Putrasahan et al.
(2016) indicated a 2 yr lag in thermal impacts found
for the region; however, comparison of 2 yr- lagged
Twfrom our dataset with MEI did not show a signifi-
cant trend, so a second iteration of the annual GLM
run with 2 yr-lagged El Niño or La Niña event occur-
rence was not included in the results. Wind was not
included in the daily model as there was insufficient
data coverage, reducing the number of observations
in the model to 185 of 1085 observation days. Vari-
ables were judged as significant in the model if they
were significant at p < 0.05 (t-statistic versus the con-
stant model). All analyses were performed in Matlab
(Mathworks). Although SST was initially considered
in the monthly GLMs, models produced for both K
h−1and (K+AP) h−1 were over-parameterized and had
lower performance than those fitted without SST, so
SST was excluded from the final GLMs.
RESULTS
Over the study period from 1999−2013, SST and Tw
measured during the observational period showed
overall similar seasonal and interannual variability
(Fig. 2a). When the Furono record of Tw(monthly
mean) from the study site and remotely-sensed SST
for the region were compared, the 2 data sets demon-
strated consistent monthly variations in temperature
trends. However, a 2-way t-test comparing the
month ly means of the 2 data sets revealed statistical
difference (p = 9.04 ×10−5), reflecting differences in
the spatial and temporal coverage of the 2 data sets.
While Furono measured daily temperatures at Seal
Island, the remotely sensed SST encompassed a
broader regional area. Further, Twcaptured water
temperature at the time of predation observation
expeditions (morning) versus the more encompass-
ing daytime SST which was capable of capturing
warmer temperatures over the course of the entire
diel period. Although SST tended to be higher than
Tw, temporal trends were generally similar among
the 2 datasets, supporting the use of the daily water
temperature measurements from the predation ob -
servation expeditions. El Niño conditions based on
the MEI occurred in 2002, 2005, 2006, 2009, and
132

Skubel et al.: Long-term climate and shark predation trends
2012, while La Niña conditions occurred in 1999,
2007, 2010, and 2011 (Fig. 2b).
From 1999 through 2013 from May through Sep-
tember, a total of 1085 observation periods occurred
between 07:00 and 09:30 h, 941 of which featured
successful and/or unsuccessful predations of white
sharks Carcharodon carcharias upon Cape fur seals
Arctocephalus pusillus pusillus and 801 of which
featured successful attacks. A comparison of daily,
monthly, and annual (K+AP) h−1 and Kh−1 showed
consistently strong and significant Pearson correla-
tion coefficients (r = 0.875, 0.934, 0.957 and p = 0,
1.19 ×10−31, 2.32 ×10−8 for daily, monthly, and annual
scales respectively), while a t-test showed the 2 met-
rics to be significantly different (p = 7.73 × 10−43, 2.91
× 10−7, 7.73 × 10−43 for daily, monthly, and annual
scales respectively). Predation intensity peaked in the
austral winter, with the greatest K h−1 and (K+AP) h−1
from June through September (Fig. 2d). Years 2003,
2004, 2005, and 2007 exhibited the greatest magni-
tude peaks in (K+AP) h−1 during this season (Figs. 2a
& 3). Frequency distributions of Twfor all daily obser-
vational periods, and those days with either Kh−1 or
(K+AP) h−1 greater than zero, show that most preda-
tion effort, successful and otherwise peaked at
14−14.5°C (10.8% of all observations), although this
may have been an artefact of the distribution of
observation effort (Fig. 3).
GLMs constructed for daily Kh−1 and (K+AP) h−1
had significant yet weak fits (adjusted r2for the
model and p-value for the F-statistic = 0.173 and 6.97
× 10−20 and 0.222 and 1.14 × 10−26 for Kh−1 and (K+AP)
h−1, respectively). Model fit for monthly GLMs of K
h−1 and (K+AP) h−1 had similarly weak fits, with lower
significance (adjusted r2for the model and p-value
for the F-statistic = 0.245 and 0.00221 and 0.139 and
133
Fig. 2. Monthly trends in (a) daily water temperature (Tw) and sea surface temperature (SST) measured during predation ob-
servation expeditions, (b) multivariate ENSO index (MEI) and El Niño/La Niña status, (c) daily wind speed and visibility (Vis)
measured during predation observation expeditions, and (d) daily kills per hour (Kh−1) and total predations (kills plus
attempted predations per hour, (K +AP) h−1) measured during predation observation periods. Standard deviation is indicated
by the shaded area around the main trendline in panels a, c and d

Mar Ecol Prog Ser 587: 129–139, 2018
0.0291 for Kh−1 and (K+AP) h−1, respectively). Both
daily and monthly GLMs indicated the importance of
temporal variability through the inclusion of DOY
and monthly terms, and of temperature variability
through the inclusion of Tw. The daily GLM of Kh−1
indi cated Tw*Month as the highest magnitude coeffi-
cient (positive), with the model of (K +AP) h−1 in -
cluding Tw*DOY as a lower magnitude (positive)
coefficient (Table 1 a,b). Vis*Month was included in
daily models of both Kh−1 and (K+AP) h−1 models as
a negative coefficient. In the monthly GLMs, Month
was the largest magnitude coefficient for both Kh−1
and (K+AP) h−1 (Table 1c,d). For both monthly Kh−1
and (K+AP) h−1, Tw*Year and Wind*Vis were in -
cluded as negative coefficients. Annual GLMs had
the best fit, as expected for the smaller number of
observations fitted to a similar quantity of explana-
tory variables (adjusted r2for the model and p-value
for the F-statistic = 0.950 and 0.0112 and 0.805 and
0.00965 for K h−1 and (K+AP) h−1, respectively). The
annual GLM of K h−1 included interaction terms with
both La Niña and El Niño occurrence, while the
annual GLM of (K+AP) h−1 included interactions with
only El Niño oc currence (Table 1 e,f). The highest-
magnitude positive term in both annual GLMs was
EN*Wind, while EN*Vis was a high-magnitude
negative coefficient. No Twterms were included in
the annual model. All GLMs exhibited a normal
distribution.
DISCUSSION
Within a given year, the distribution
of Kh−1 and (K+AP) h−1 appears to
correspond with temporal trends
in water temperature. However, al -
though Twat Seal Island was a coeffi-
cient in the daily and monthly models
of kill and total predation rates, the
absence of Twfrom the annual model
indicates that temperature variability
at Seal Island may play an intra-
annual rather than inter-annual role
with res pect to white shark Carcharo-
don carcharias predation. The inter-
action terms of El Niño and La Niña
with wind speed and water visibility
were suggestive of inter-annual vari-
ability in these environmental param-
eters with possible ties to ENSO
cycling, which may have played a
greater role than water temperature
in influencing inter-annual predation
trends.
Clark et al. (1996) found that fish assemblages in the
surf zone of False Bay were influenced by water tem-
perature and wind speed, among other factors. In this
study, wind speed was implicated as an explanatory
variable in monthly and annual GLMs, including as
an interaction with El Niño and La Niña occurrence in
the annual models. Taken together with the results of
Clark et al. (1996), teleost distribution may be vacillat-
ing with wind speed and water temperature, with
possible bottom-up impacts at the level of white shark
predation patterns, or white sharks may be respond-
ing to similar environmental drivers as the fish moni-
tored by Clark et al. (1996). In another study, Weltz et
al. (2013) monitored visual sightings of white sharks
at beaches in False Bay (Fish Hoek and Muizenberg),
finding that probability of shark sighting increased as
SST rose from 14 to 18°C. Although we did not
directly measure sightings, frequency distribution of
daily Twrecorded for days when shark predation on
seals were observed show that attacks peaked at 14−
14.5°C and that very few occurred above 18°C (Tw>
18°C: 6.4% of days for (K+AP) h−1 > 0, 6.2% of days for
Kh−1 > 0, and 5.1% of all observation days). El Niño
events are characterized by changes in wind strength
on global and local scales (Rasmusson & Wallace
1983); here, there may be a contemporaneous effect
from El Niño from anomalies in the Pacific changing
global circulation patterns driving global atmospheric
circulation and thereby wind speed. However, it is
134
Fig. 3. Frequency distribution of water temperature measured during preda-
tion observation periods, for all days, for days when (K+AP) h−1 > 0, and for
days when Kh−1 > 0. See Fig. 2 for abbreviations

Skubel et al.: Long-term climate and shark predation trends
speculative to presume that our measurements cap-
tured this ENSO-modulated variability in wind speed,
particularly given that meteorological teleconnections
differ between ENSO events (Gershunov & Barnett
1998). Although the annual GLMs indicated that El
Niño and La Niña occurrence had explanatory power
for inter-annual variability in predation rates, the ab -
sence of Twfrom these annual models suggests that,
at a local scale, ENSO-influenced water temperatures
had minimal effects on white shark predation. How-
ever, it is possible that ENSO-mediated changes in
water temperature are impacting predation-related
factors external to the study area surrounding Seal Is-
land; for instance, large-scale changes in biological
135
Coefficient SE tp
(a) Daily Kh−1~ 1 + Tw*Vis + Tw*Month + Vis*Month + DOY*Month
Adj. r2= 0.173, obs. = 538, F= 15, p = 6.97E−20
(Intercept) −2.48 5.88 −0.422 0.673
Tw*Vis −0.0288 0.0164 −1.76 0.0798
Tw*Month 0.216 0.0514 4.20 3.10E−05
Vis*Month −0.0400 0.0195 −2.05 0.0406
DOY*Month −0.00964 0.00158 −6.08 2.24E−09
(b) Daily (K+AP )h
−1~ 1 + Tw*Vis + Tw*DOY + Vis*Month + DOY*Month
Adj. r2= 0.245, obs. = 538, F= 20.2, p = 1.14E−26
(Intercept) −26.1 9.60 −2.71 0.00687
Tw*Vis −0.0572 0.0283 −2.02 0.0439
Tw*DOY 0.00916 0.00293 3.12 0.00188
Vis*Month −0.0637 0.0333 −1.91 0.0571
DOY*Month −0.0256 0.00272 −9.56 4.49E−20
(c) Monthly Kh−1 ~ 1 + Month + Tw*Vis + Tw*Year + Wind*Vis
Adj. r2= 0.245, obs. = 64, F= 3.55, p = 0.00221
Month 0.308 0.0963 3.20 0.00228
Tw*Vis −0.143 0.0705 −2.023 0.0475
Tw*Year −0.125 0.0502 −2.49 0.0157
Wind*Vis −0.0339 0.0139 −2.44 0.0180
(d) Monthly (K+AP )h
−1 ~ 1 + Month + Tw*Year + Wind*Vis
Adj. r2= 0.139, obs. = 64, F= 2.45, p = 0.0291
(Intercept) −4235 2286.4 −1.85 0.0693
Month 0.428 0.181 2.37 0.0214
Tw*Year −0.149 0.0801 −1.87 0.0674
Wind*Vis −0.0729 0.0255 −2.86 0.00597
(e) Annual Kh−1 ~ 1 + Wind*Vis + Wind*EN + Wind*LN + Vis*EN + Vis*Year + EN*Year
Adj. r2= 0.950, obs. = 15, F= 25.1, p = 0.0112
(Intercept) 1361 196 6.94 0.00614
Wind*Vis −0.723 0.0916 −7.91 0.00423
Wind*EN 1.60 0.248 6.46 0.00753
Wind*LN 0.499 0.109 4.56 0.0198
Vis*EN −3.39 0.996 −3.41 0.0423
Vis*Year 0.115 0.0162 7.07 0.00582
EN*Year −0.725 0.286 −2.54 0.0849
(f) Annual (K+AP )h
−1 ~ 1 + Wind*Vis + Wind*EN + Vis*EN + Vis*Year
Adj. r2= 0.805, obs. = 15, F= 8.22, p = 0.00965
(Intercept) 1276 555 2.30 0.0613
Wind*Vis −1.25 0.179 −6.96 0.000438
Wind*EN 2.41 0.446 5.40 0.00166
Vis*EN −2.14 0.494 −4.32 0.00496
Vis*Year 0.100 0.0439 2.29 0.0624
Table 1. For daily, monthly, and annual generalized linear models of kills + attempted predations per hour, (K+AP) h−1, and
kills per hour, Kh−1: coefficient estimates, SE, t-statistic, and p-value of each variable included in the final model, adjusted r2,
number of observations (obs.), F-statistic versus the initial model, and p-value of the F-statistic. Abbreviated variables are
water temperature (Tw), visibility (Vis), day of year (DOY), El Niño (EN), and La Niña (LN).

Mar Ecol Prog Ser 587: 129–139, 2018
productivity under El Niño conditions (Barber &
Chavez 1983).
Greater visibility appeared to have a negative
effect on both kill and total predation rates across
temporal scales; Robbins (2007) found that increased
visibility was inversely related to white shark pres-
ence at Neptune Island (Australia). In the present
study, low visibility may lead to greater abundance,
and thus greater chance of observing an attack, or
the lower visibility may provide a higher chance of
success for the attacking shark as an individual may
be less visually apparent to the targeted seal (Martin
& Hammerschlag 2012). Given the high energetic re -
quirements of a predation attempt, white sharks are
observed to launch attacks on seals only when envi-
ronmental conditions are optimal for success, increas-
ing efficiency of energy expenditure (Hammerschlag
et al. 2006). These conditions include targeting seals
during a specific range of low-light levels as well as
staging attacks from identifiable deep-water loca-
tions around the Island that provide sharks with a
visual and tactical advantage over seals (Hammer-
schlag et al. 2006, Martin et al. 2009).
Considering the 2 metrics of predation patterns,
both (K+AP) h−1 and Kh−1 showed similar temporal
trends (Fig. 2d) and frequency distribution with re -
spect to water temperature (Fig. 3), as well as similar
final GLMs. However, kill rates tended to have
higher model fit (adjusted r2), which may have been
due to this metric showing slightly greater corre-
spondence to variability in the explanatory variables
considered, while total predation rates may be more
reflective of shark presence as a whole. However,
without direct evidence of physiological response to
changes in environmental condition (e.g. telemetry
of internal temperature as per Goldman (1997) or
swimming speed as per Semmens et al. (2013)) it is
not possible to draw inferences from our measure-
ments of kill rates to conclusions regarding metabo-
lism or feeding.
Change in ambient Twcan influence metabolic de -
mand (e.g. metabolic Q10), with an estimated Q10 of
between 2 and 3 presumed for the white shark as per
Carlson et al. (2004). Change in metabolic rate with
temperature increase has not been directly measured
for this species, although studies have measured
body temperature as an indicator (Carey et al. 1982,
McCosker 1987, Goldman 1997) or body mass and
swimming speed (Semmens et al. 2013) as field-
based approaches to estimating metabolic rate.
White shark oxygen consumption has previously
been measured in transit to an aquarium (Ezcurra et
al. 2012); however as temperature was not manipu-
lated, its direct impact on metabolic demand could
not be ascertained. Complicating estimates of change
in metabolic rate with respect to temperature in the
white shark, (1) the large body size of this species
allows for thermal inertia with movement through
heterogeneous water temperatures, and (2) heat
shedding is facilitated through manipulation of the
hepatic sinus which bypasses the suprahepatic rete
(Carey et al. 1981). As white sharks are among the
largest elasmobranchs (Compagno 1984, Schmidt-
Nielsen 1984) and have a lower surface area to vol-
ume ratio for gaining or losing heat, they may be less
sensitive to temperature variability than smaller spe-
cies given greater capa city for thermal inertia. Simi-
larly, metabolic Q10 has been found to be higher with
lower body mass, suggesting higher sensitivity to
temperature with smaller-bodied species and the
inverse for larger species (Du Preez et al. 1988);
given that the white shark’s ability to regulate body
temperature, the Q10 value may have less signifi-
cance for metabolic rate in response to ambient tem-
perature change. In our study, daily and monthly
GLMs suggested that in creasing Twcorresponded
with increasing (K+AP) h−1 and Kh−1, but the influ-
ence of water temperature was not supported on an
inter-annual scale. Taken together with the heat reg-
ulation capacity of white sharks, it is not directly evi-
dent that the predation patterns observed here were
a result of rising metabolic demand with temperature.
Both a strength and limitation of this study is that
our results are based on comparing long-term obser-
vational data sets and not controlled laboratory ex -
periments. Although it appears that predation trends
across daily, monthly, and annual timescales are re -
lated to environmental and seasonal (temporal) vari-
ability, there are several other factors that could have
contributed, at least partly, to the documented changes
in predation rates and likely contributed to the vari-
ability unexplained by the GLM. These factors could
include, for example, variability in seal, shark and
teleost abundance. Although seal population size has
been shown to be relatively consistent over the study
period (Kirkman et al. 2007), there may be unac-
counted marginal changes impacting predation.
White shark population declines have been specu-
lated for the region based on mark-recapture and
genetic analysis (Andreotti et al. 2016), although
interpreted with some level of skepticism (Irion et al.
2017). Further, white shark sightings have declined
at Seal Island in False Bay over a 9 yr monitoring
period (Hewitt et al. 2018). If predation rates were
reflective of white shark population trends alone,
such as those suggested in the aforementioned stud-
136

Skubel et al.: Long-term climate and shark predation trends
ies, it would likely manifest as prolonged long-term
declines in predation rates, rather than the alternat-
ing increases and decreases observed here. Changes
in teleost prey cannot be ruled out, however were not
monitored in this investigation. Future studies includ-
ing measurement of prey abundance may be reveal-
ing of ecosystem-level interactions among predators
and other constituents of the regions they occupy.
One of the major challenges in projecting the climate
response of a predatory species is unknowns sur-
rounding respective spatial shifts in predator− prey
distribution (Winder & Schindler 2004); namely, will
there be changes in overlap (Hunsicker et al. 2013),
or will predators remain synchronized with their prey
species (Sergeant et al. 2014). Given the role of wind
and visibility in the GLMs, we recommend further
study exploring how local meteorology may impact
spatio temporal patterns of predator and prey abun-
dance, and predation behavior. Further, the strong
signal of temporal variability in intra-annual preda-
tion patterns may be tied to broader seasonal shifts in
diet of white sharks, which may be explored via
methods such as stable isotope analysis and oppor-
tunistic stomach content analysis.
Laboratory experiments with Port Jackson sharks
Heterodontus portusjacksoni exposed to water tem-
perature increases to projected 2100 levels increased
their feeding but not growth (compared to a control
treatment), as feeding simply supplemented the rise
in metabolic demand associated with higher temper-
atures (Pistevos et al. 2015). However, oceanographic
consequences of climate change are not limited to a
temperature increase. Indeed, with increasing CO2
emissions comes increased ocean acidification (IPCC
2013), which has been shown to negatively impact
olfactory ability and hunting behavior of some sharks
in controlled settings (Dixson et al. 2015, Pistevos et
al. 2015). If the hunting ability of white sharks is im -
paired by ocean acidification, but their predation
patterns are not directly impacted by water tempera-
tures (i.e. through metabolic demand), they may be
well placed to balance their energy budget in
warmer and more acidic waters. It may be that cli-
mate change will lead to changes in the availability
of preferred or alternative prey for white sharks, indi-
rectly impacting predation rates, rather than directly
impacting shark metabolism and hunting. For in -
stance, spatiotemporal changes in water temperature
may lead to shifts in prey availability due to such fac-
tors as temperature-regulated phenological events
(e.g. timing of migrations or reproductive events), or
different thermal sensitivities between predator and
prey species leading to changes in relative attack
and escape speeds (Grigaltchik et al. 2012, Creel et
al. 2016), contributing to asynchronous trophic rela-
tionships. These questions highlight areas for di rected
future experimentation when considering climate-
driven effects on predator–prey dynamics.
In summary, our analysis of white shark predation
rates on seals paired with environmental monitoring
over 15 years suggests that water temperatures on an
intra-annual scale might contribute to predation pat-
terns in white sharks either directly or indirectly, but
do not implicate water temperature as a primary
driver of predation rates in this scenario, or on an
interannual scale. In contrast, inter-annual variability
in predation rate appeared linked to other environ-
mental factors (wind, water visibility, and the occur-
rence of El Niño and La Niña events) which may
drive both prey abundance and predation efficacy.
Quantifying ecosystem-scale change in the wild to a
reliable and useful extent is not a feasible goal in
many natural systems due to known and unknown
complexities, and the limits of scientific measure-
ment in capturing these complexities. However, pro-
longed monitoring of species with known predator−
prey relationships, such as the white shark and Cape
fur seal Arctocephalus pusillus pusillus in False Bay,
may provide an opportunity for ongoing investiga-
tions of how climate variability might impact the eco -
logy of an ecosystem and for generating hypotheses
to be tested in controlled experimental studies.
Acknowledgements. We thank all the assistants, partici-
pants, and filmmakers throughout the years that supported
this work, including the guests and crew of Apex Shark
Expeditions that made data collection possible. This work
was conducted under permits from the South African
Department of Environmental Affairs and with permission
from the University of Miami Institutional Animal Welfare
and Care Committee.
LITERATURE CITED
Allen-Ankins S, Stoffels RJ (2017) Contrasting fundamental
and realized niches: two fishes with similar thermal per-
formance curves occupy different thermal habitats.
Freshw Sci 36: 635−652
Andreotti S, Rutzen M, van der Walt S, Von der Heyden S
and others (2016) An integrated mark-recapture and
genetic approach to estimate the population size of white
sharks in South Africa. Mar Ecol Prog Ser 552: 241−253
Barber RT, Chavez FP (1983) Biological consequences of El
Niño. Science 222: 1203−1210
Barley SC, Meekan MG, Meeuwig JJ (2017) Species diver-
sity, abundance, biomass, size and trophic structure of
fish on coral reefs in relation to shark abundance. Mar
Ecol Prog Ser 565: 163−179
Beal LM, De Ruijter WPM, Biastoch A, Zahn R (2011) On the
role of the Agulhas system in ocean circulation and cli-
137

Mar Ecol Prog Ser 587: 129–139, 2018
mate. Nature 472: 429−436
Carey FG, Teal JM, Kanwisher JW (1981) The visceral tem-
peratures of mackerel sharks (Lamnidae). Physiol Zool
54: 334−344
Carey FG, Kanwisher, JW, Brazier O, Gabrielson G, Casey JG,
Pratt Jr HL (1982) Temperature and activities of a white
shark, Carcharodon carcharias. Copeia 1982: 254– 260
Carlson JK, Goldman KJ, Lowe CG (2004) Metabolism,
energetic demand, and endothermy. In: Carrier JC,
Musick JA, Heithaus MR (eds) Biology of sharks and
their relatives, CRC Press, Boca Raton, FL, p 203−224
Clark BM, Bennett BA, Lamberth SJ (1996) Temporal varia-
tions in surf zone fish assemblages from False Bay, South
Africa. Mar Ecol Prog Ser 131: 35−47
Compagno LJV (1984) FAO species catalogue. Vol. 4.
Sharks of the world. An annotated and illustrated cata-
logue of sharks species known to date. Part 1. Hexanchi-
formes to Lamniformes. FOA Fish Synopsis 4: 249
Creel S, Creel NM, Creel AM, Creel BM (2016) Hunting on
a hot day: effects of temperature on interactions between
African wild dogs and their prey. Ecology 97: 2910−2916
de Ruijter WPM, Biastoch A, Drijfhout SS, Lutjeharms JRE
and others (1999) Indian−Atlantic interocean exchange:
Dynamics, estimation and impact. J Geophys Res 104:
20885
Di Santo V, Bennett WA (2011) Is post-feeding thermotaxis
advantageous in elasmobranch fishes? J Fish Biol 78:
195−207
Dixson DL, Jennings AR, Atema J, Munday PL (2015) Odor
tracking in sharks is reduced under future ocean acidifi-
cation conditions. Glob Change Biol 21: 1454−1462
Dowd WW, Brill RW, Bushnell PG, Musick JA (2006) Esti-
mating consumption rates of juvenile sandbar sharks
(Carcharhinus plumbeus) in Chesapeake Bay, Virginia,
using a bioenergetics model. Fish Bull 104: 332−342
Du Preez HH, Mclachlan A, Marais JFK (1988) Oxygen con-
sumption of two nearshore marine elasmobranchs,
Rhino batos annulatus (Muller & Henle, 1841) and Mylio-
batus aquila (Linnaeus, 1758). Comp Biochem Physiol
Part A Physiol 89: 283−294
Dufois F, Rouault M (2012) Sea surface temperature in False
Bay (South Africa): towards a better understanding of its
seasonal and inter-annual variability. Cont Shelf Res 43:
24−35
Ezcurra JM, Lowe CG, Mollet HF, Ferry LA, O’Sullivan JB
(2012) Oxygen consumption rate of young-of-the-year
white sharks, Carcharodon carcharias, during transport
to the Monterey Bay Aquarium. In: Domeier ML (ed)
Global perspectives on the biology and life history of the
white shark. CRC Press, Boca Raton, FL, p 17−26
Fallows C, Martin RA, Hammerschlag N (2012) Compar-
isons between white shark-pinniped interactions at Seal
Island (South Africa) with other sites in California
(United States). In: Domeier ML (ed) Global perspectives
on the biology and life history of the white shark. CRC
Press, Boca Raton, FL, p 105−117
Fallows C, Gallagher AJ, Hammerschlag N (2013) White
sharks (Carcharodon carcharias) scavenging on whales
and its potential role in further shaping the ecology of an
apex predator. PLOS ONE 8: e60797
Fallows C, Fallows M, Hammerschlag N (2016) Effects of
lunar phase on predator-prey interactions between white
shark (Carcharodon carcharias) and Cape fur seals
(Arcto cephalus pusillus pusillus). Environ Biol Fishes 99:
805−812
Gershunov A, Barnett TP (1998) Interdecadal modulation
of ENSO teleconnections. Bull Am Meteorol Soc 79:
2715−2726
Goldman KJ (1997) Regulation of body temperature in the
white shark, Carcharodon carcharias. J Comp Physiol B
167: 423−429
Gordon AL, Haxby WF (1990) Agulhas eddies invade the
south Atlantic: evidence from Geosat altimeter and ship-
board conductivity-temperature-depth survey. J Geo-
phys Res 95: 3117
Grigaltchik VS, Ward JW, Seebacher F (2012) Thermal
acclimation of interactions: differential responses to tem-
perature change alter predator-prey relationship. Proc R
Soc B 279: 4058−4064
Hammerschlag N, Martin RAR, Fallows C (2006) Effects of
environmental conditions on predator−prey interactions
between white sharks (Carcharodon carcharias) and
Cape fur seals (Arctocephalus pusillus pusillus). Environ
Biol Fishes 76: 341−350
Hammerschlag N, Martin RA, Fallows C, Collier R, Law -
rence R (2012) Investigatory behavior towards surface
objects and non-consumptive strikes on seabirds by
white sharks (Carcharodon carcharias) at Seal Island,
South Africa. In: Domeier ML (ed) Global perspectives
on the biology and life history of the great white shark.
CRC Press, Boca Raton, FL, p 91–103
Hammerschlag N, Meÿer M, Seakamela SM, Kirkman S,
Fallows C, Creel S (2017) Physiological stress responses
to natural variation in predation risk: evidence from white
sharks and seals. Ecology 98: 3199– 3210
Hewitt AM, Kock AA, Booth AJ, Griffiths CL (2018) Trends
in sightings and population structure of white sharks,
Carcharodon carcharias, at Seal Island, False Bay, South
Africa, and the emigration of subadult female sharks
approaching maturity. Environ Biol Fishes 101:39– 54
Hight BV, Lowe CG (2007) Elevated body temperatures of
adult female leopard sharks, Triakis semifasciata, while
aggregating in shallow nearshore embayments: evi-
dence for behavioral thermoregulation? J Exp Mar Biol
Ecol 352: 114−128
Hoegh-Guldberg O, Bruno JF (2010) The impact of climate
change on the world’s marine ecosystems. Science 328:
1523−1528
Huang B, Banzon VF, Freeman E, Lawrimore J and others
(2015) Extended reconstructed sea surface temperature
version 4 (ERSST. v4). Part I: upgrades and intercompar-
isons. J Clim 28:911–930
Hunsicker ME, Ciannelli L, Bailey KM, Zador S, Stige LC
(2013) Climate and demography dictate the strength of
predator-prey overlap in a subarctic marine ecosystem.
PLOS ONE 8: e66025
IPCC (2013) Climate Change 2013: the physical science
basis. Contribution of Working Group I to the Fifth As -
sessment Report of the Intergovernmental Panel on Cli-
mate Change. Cambridge University Press, Cambridge
Irion DT, Noble LR, Kock AA, Gennari E and others (2017)
Pessimistic assessment of white shark population status
in South Africa: comment on Andreotti et al. (2016). Mar
Ecol Prog Ser 577:251–255
Kearney M, Porter W (2009) Mechanistic niche modelling:
combining physiological and spatial data to predict spe-
cies’ ranges. Ecol Lett 12: 334−350
Ketchum JT, Hearn A, Klimley P, Espinoza E, Peñaherrera
C, Largier JL (2014) Seasonal changes in movements and
habitat preferences of the scalloped hammerhead shark
138

Skubel et al.: Long-term climate and shark predation trends
(Sphyrna lewini) while refuging near an oceanic island.
Mar Biol 161: 755−767
Kilpatrick KA, Podestá GP, Evans R (2001) Overview of the
NOAA/NASA advanced very high resolution radiometer
Pathfinder algorithm for sea surface temperature and
associated matchup database. J Geophys Res 106: 9179
Kirkman S, Oosthuizen W, Meÿer M, Kotze P, Roux J,
Underhill L (2007) Making sense of censuses and dealing
with missing data: trends in pup counts of Cape fur seal
Arctocephalus pusillus pusillus for the period 1972−2004.
Afr J Mar Sci 29: 161−176
Martin RA, Hammerschlag N (2012) Marine predator prey
contests: ambush and speed versus vigilance and agility.
Mar Biol Res 8: 90−94
Martin RA, Hammerschlag NS, Collier R, Fallows C (2005)
Predatory behaviour of white sharks (Carcharodon car-
charias) at Seal Island, South Africa. J Mar Biol Assoc UK
85: 1121−1135
Martin RA, Rossmo DK, Hammerschlag N (2009) Hunting
patterns and geographic profiling of white shark preda-
tion. J Zool (Lond) 279: 111−118
McCosker JE (1987) The white shark, Carcharodon car-
charias, has a warm stomach. Copeia 1987:195–197
Miller LP, Matassa CM, Trussell GC (2014) Climate change
enhances the negative effects of predation risk on an
intermediate consumer. Glob Change Biol 20: 3834−3844
Papastamatiou YP, Watanabe YY, Bradley D, Dee LE, Weng
K, Lowe CG, Caselle JE (2015) Drivers of daily routines
in an ectothermic marine predator: hunt warm, rest
warmer? PLOS ONE 10: e0127807
Pink M, Abrahams MV, Krkošek M (2016) Temperature and
its impact on predation risk within aquatic ecosystems.
Can J Fish Aquat Sci 73: 869−876
Pistevos JCA, Nagelkerken I, Rossi T, Olmos M, Connell SD
(2015) Ocean acidification and global warming impair
shark hunting behaviour and growth. Sci Rep 5: 16293
Putrasahan D, Kirtman BP, Beal LM (2016) Modulation of
SST interannual variability in the Agulhas leakage
region associated with ENSO. J Clim 29: 7089−7102
Rasmusson EM, Wallace JM (1983) Meteorological aspects
of the El Niño. Science 222: 1195−1202
Robbins RL (2007) Environmental variables affecting the
sexual segregation of great white sharks Carcharodon
carcharias at the Neptune Islands South Australia. J Fish
Biol 70: 1350−1364
Ruppert JLW, Travers MJ, Smith LL, Fortin MJ, Meekan MG
(2013) Caught in the middle: combined impacts of shark
removal and coral loss on the fish communities of coral
reefs. PLOS ONE 8: e74648
Schmidt-Nielsen K (1984) Scaling: Why is animal size so
important? Trends Neurosci 8: 241
Semmens JM, Payne NL, Huveneers C, Sims DW, Bruce BD
(2013) Feeding requirements of white sharks may be
higher than originally thought. Sci Rep 3: 1471
Sergeant CJ, Armstrong JB, Ward EJ (2014) Predator−prey
migration phenologies remain synchronised in a warm-
ing catchment. Freshw Biol 60: 724−732
Sunday JM, Bates AE, Dulvy NK (2012) Thermal tolerance
and the global redistribution of animals. Nat Clim Chang
2: 686−690
van Sebille E, Biastoch A, Van Leeuwen PJ, De Ruijter WPM
(2009) A weaker Agulhas current leads to more Agulhas
leakage. Geophys Res Lett 36: L03601
Weltz K, Kock AA, Winker H, Attwood C, Sikweyiya M
(2013) The influence of environmental variables on the
presence of white sharks, Carcharodon carcharias at two
popular Cape Town bathing beaches: a generalized
additive mixed model. PLOS ONE 8: e68554
Winder M, Schindler DE (2004) Climate change uncouples
trophic interactions in an aquatic ecosystem. Ecology 85:
2100−2106
Wolter K, Timlin MS (1993) Monitoring ENSO in COADS
with a seasonally adjusted principal component index.
Proc 17th Clim Diagnostics Work, NOAA/NMC/CAC,
NSSL, Oklahoma Climate Survey, CIMMS and the
School of Meteorology, University of Oklahoma, p 52−57
Wolter K, Timlin MS (1998) Measuring the strength of ENSO
events: How does 1997/98 rank? Weather 53: 315−324
139
Editorial responsibility: Alistair Hobday,
Hobart, Tasmania, Australia
Submitted: January 18, 2017; Accepted: November 20, 2017
Proofs received from author(s): January 22, 2018