Population and Individual Elephant Response to a Catastrophic Fire in Pilanesberg National Park

Abstract

In predator-free large herbivore populations, where density-dependent feedbacks occur at the limit where forage resources can no longer support the population, environmental catastrophes may play a significant role in population regulation. The potential role of fire as a stochastic mass-mortality event limiting these populations is poorly understood, so too the behavioural and physiological responses of the affected animals to this type of large disturbance event. During September 2005, a wildfire resulted in mortality of 29 (18% population mortality) and injury to 18, African elephants in Pilanesberg National Park, South Africa. We examined movement and herd association patterns of six GPS-collared breeding herds, and evaluated population physiological response through faecal glucocorticoid metabolite (stress) levels. We investigated population size, structure and projected growth rates using a simulation model. After an initial flight response post-fire, severely injured breeding herds reduced daily displacement with increased daily variability, reduced home range size, spent more time in non-tourist areas and associated less with other herds. Uninjured, or less severely injured, breeding herds also shifted into non-tourist areas post-fire, but in contrast, increased displacement rate (both mean and variability), did not adjust home range size and formed larger herds post-fire. Adult cow stress hormone levels increased significantly post-fire, whereas juvenile and adult bull stress levels did not change significantly. Most mortality occurred to the juvenile age class causing a change in post-fire population age structure. Projected population growth rate remained unchanged at 6.5% p.a., and at current fecundity levels, the population would reach its previous level three to four years post-fire. The natural mortality patterns seen in elephant populations during stochastic events, such as droughts, follows that of the classic mortality pattern seen in predator-free large ungulate populations, i.e. mainly involving juveniles. Fire therefore functions in a similar manner to other environmental catastrophes and may be a natural mechanism contributing to population limitation. Welfare concerns of arson fires, burning during "hot-fire" conditions and the conservation implications of fire suppression (i.e. removal of a potential contributing factor to natural population regulation) should be integrated into fire management strategies for conservation areas.

Full text

Population and Individual Elephant Response to a
Catastrophic Fire in Pilanesberg National Park
Leigh-Ann Woolley
1
*, Joshua J. Millspaugh
2
, Rami J. Woods
2
, Samantha Janse van Rensburg
1
, Robin L.
Mackey
1
, Bruce Page
1
, Rob Slotow
1
1Amarula Elephant Research Programme, School of Biological and Conservation Sciences, University of KwaZulu-Natal, Durban, South Africa, 2Department of Fisheries
and Wildlife Sciences, University of Missouri, Columbia, Missouri, United States of America
Abstract
In predator-free large herbivore populations, where density-dependent feedbacks occur at the limit where forage resources
can no longer support the population, environmental catastrophes may play a significant role in population regulation. The
potential role of fire as a stochastic mass-mortality event limiting these populations is poorly understood, so too the
behavioural and physiological responses of the affected animals to this type of large disturbance event. During September
2005, a wildfire resulted in mortality of 29 (18% population mortality) and injury to 18, African elephants in Pilanesberg
National Park, South Africa. We examined movement and herd association patterns of six GPS-collared breeding herds, and
evaluated population physiological response through faecal glucocorticoid metabolite (stress) levels. We investigated
population size, structure and projected growth rates using a simulation model. After an initial flight response post-fire,
severely injured breeding herds reduced daily displacement with increased daily variability, reduced home range size, spent
more time in non-tourist areas and associated less with other herds. Uninjured, or less severely injured, breeding herds also
shifted into non-tourist areas post-fire, but in contrast, increased displacement rate (both mean and variability), did not
adjust home range size and formed larger herds post-fire. Adult cow stress hormone levels increased significantly post-fire,
whereas juvenile and adult bull stress levels did not change significantly. Most mortality occurred to the juvenile age class
causing a change in post-fire population age structure. Projected population growth rate remained unchanged at 6.5% p.a.,
and at current fecundity levels, the population would reach its previous level three to four years post-fire. The natural
mortality patterns seen in elephant populations during stochastic events, such as droughts, follows that of the classic
mortality pattern seen in predator-free large ungulate populations, i.e. mainly involving juveniles. Fire therefore functions in
a similar manner to other environmental catastrophes and may be a natural mechanism contributing to population
limitation. Welfare concerns of arson fires, burning during ‘‘hot-fire’’ conditions and the conservation implications of fire
suppression (i.e. removal of a potential contributing factor to natural population regulation) should be integrated into fire
management strategies for conservation areas.
Citation: Woolley L-A, Millspaugh JJ, Woods RJ, Janse van Rensburg S, Mackey RL, et al. (2008) Population and Individual Elephant Response to a Catastrophic
Fire in Pilanesberg National Park. PLoS ONE 3(9): e3233. doi:10.1371/journal.pone.0003233
Editor: Robert Brooks, The University of New South Wales, Australia
Received June 4, 2008; Accepted August 27, 2008; Published September 17, 2008
Copyright: !2008 Woolley et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Funding received from Amarula (Distell (Pty) Ltd), PPC Cement, National Research Foundation. The funders had no role in study design, data collection
and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: leigh-ann@polka.co.za
Introduction
Successful conservation management of large mammals has the
ironic consequence of problems associated with overpopulation [1].
This is particularly so with fragmented, small populations or with
keystone species that, at high population densities, can impose
negative impacts on the system [2]. A key uncertainty thatemerges is
what limits such populations naturally, and whether such limitation
will occur at the same levels in human modified systems (e.g. with
fences or artificial water) compared to natural systems [1]. Some
species may be resource limited, displaying density dependent
responses [3]. Others may be top-down limited by predators [4].
Long-lived species may also be limited by environmental catastro-
phes, such as drought, flood, fire or disease, which can cause sudden
and, at times, significant shifts in population size and dynamics over
a very short time, if the effects of such catastrophic impacts on
demographics are of sufficient frequency and intensity [5]. Although
there is some theoretical and empirical evidence that drought may
limit elephant populations [6], there has been no evaluation of the
role that fire may play. Due to their rare occurrence, evaluation of
the impacts of such events on population dynamics and individual
responses is also rare.
Fire is commonly applied for ecosystem management in
savannas and arson fires occur regularly [7]. Whilst the impact
of fire on plant mortality has been extensively researched, there is
little research that has assessed the influence of fire on mortality in
animals or the welfare issues associated with fire in savanna
systems. Given elephants are highly intelligent and social
mammals, fire, or other severe disturbances, may also precipitate
behavioural or physiological responses. For example, high
elephant poaching caused heavily stressed elephants to form
larger groups than unstressed elephants [8].
The extremely hot, dry, windy (‘‘hot-fire’’) conditions experi-
enced towards the end of the dry season in Pilanesberg National
Park (PNP), a small (570 km
2
), fenced reserve in South Africa
facilitated the spread of an uncontrolled wild fire. The area had a
PLoS ONE | www.plosone.org 1 September 2008 | Volume 3 | Issue 9 | e3233

1–2 year fuel load, with the last pre-fire rains falling in May 2005.
Below average (,630 mm p.a.) annual rainfall of 554 mm was
recorded during the 2004/5 wet season, while 824 mm fell in
2003/4 and 411 mm in 2002/3. On 21 September 2005, ambient
midday air temperature was 34uC, while wind speed was generally
strong but variable in direction. The fire entered the western
boundary of the Park near Tlatlaganyane village on 20 September
2005 and within two days had moved across an area of
approximately 61 km
2
. This catastrophic fire resulted in the
mortality of 29 and injury to 18 elephants, unprecedented in PNP,
where few natural elephant mortalities had occurred prior to this
event [9,10]. This event provided us with the opportunity to assess
the potential influence of severe fires in which animals become
trapped on the behavioural, physiological and demographic
responses of the elephant population. We provide an assessment
of (1) the behavioural and physiological responses of the elephants
to this large disturbance event, and (2) the potential for rare,
stochastic mass-mortality events to limit population size. We
examined movement and herd association patterns of six GPS-
collared breeding herds, and evaluated physiological response
through faecal glucocorticoid metabolite (stress) levels [11,12]. We
investigated population size, structure and projected growth rates.
Results
Behavioural response
Daily displacement. There was no significant difference in
mean daily displacement over four days before versus after the fire
for all cows (t
5
=21.238, P= 0.271). However, injured herds
(CE03, CE88) and a herd in close proximity to the fire at time of
injury (CE32) moved significantly further per day after the fire
than before (t
2
=26.915, P= 0.020), while there was no significant
difference in mean daily displacement of uninjured herds (CE13,
CE61, CE81) over four days before versus after the fire (Figure 1).
There was no significant difference in daily displacement among
collared cows over a ten-day period before the fire (F
5, 54
= 0.305,
P= 0.908). However, there was a significant difference in daily
displacement among collared cows over the ten-day period, post-
flight, after the fire (F
5, 54
= 9.346, P,0.0005). Injured cows (CE03
and CE88) moved at a significantly slower rate in the ten days after
the fire (t
9
= 4.486, P= 0.002; t
9
= 2.756, P= 0.022 respectively),
compared with ten-day daily displacement before the fire (Figure 1).
Uninjured cows CE61, CE81, CE32 and CE13, did not show a
significant change in daily displacement in the ten day period before
versus after the fire (t
9
=20.450, P= 0.663; t
9
=20.930, P= 0.928;
t
9
= 1.084, P= 0.307; t
9
= 0.745, P= 0.476 respectively) (Figure 1).
There was no significant difference in the coefficients of variation
(CV) of daily displacement for the ten-day periods pre- and post-fire
(t
9
=22.064, P=0.094), but a general trend of increased variability
is evident for those herds involved in the fire (CE03, CE88) or those
close to the fire when injuries occurred (CE32) (Figure 1).
There was no significant difference in daily displacement among
collared cows over a three-month period before the fire (F
5,
540
= 1.709, P= 0.131). However, there was a significant difference
in daily displacement among collared cows over a three-month
period post-fire (F
5, 540
= 5.720, P,0.001). The daily displacement
over three months for injured cows CE03 and CE88, as well as
CE88’s new matriarch CE13, were not statistically different, while
daily displacement for CE88 and CE13 was not statistically different
from uninjured cows (CE81, CE61 and CE32) (Figure 1). Uninjured
cows CE81 and CE61 showed significant increase in their daily
displacement during three months post-fire (t
90
=23.664, P,0.001;
t
90
=23.830, P,0.001) (Figure 1). Severely injured cow CE03
showed a significant decrease in three-month daily displacement
post-fire (t
90
=23.240, P,0.0005) (Figure 1). Less severely injured
cow CE88, matriarch CE13 and uninjured cow CE32 showed no
significant difference in three month daily displacement before
versus after the fire (t
90
=21.337, P= 0.185; t
90
=20.747,
P= 0.457; t
90
=21.641, P= 0.104 respectively) (Figure 1). There
was a significant difference between pre- and post-fire CV in daily
displacement over three months (t
9
=22.984, P= 0.031), with a
general trend of increase in variability post-fire (Figure 1).
Home range. There was no significant difference in home
range size before and after the fire among all cows (50% kernel
size, t
5
= 0.505, P= 0.635; 95% kernel home range, t
5
=20.024,
P= 0.982). Only severely injured cow CE03 reduced the size of
her core home range (36.1 km
2
to 6.3 km
2
) and 95% home range
(305.9 km
2
to 71.6 km
2
) dramatically after the fire and her home
range shifted from the central areas of the Park to the south-
eastern wilderness area (Figure 2).
Cows spent significantly more time in the wilderness areas of the
Park in the three months after than in the three months before the
fire (t
5
=24.510, P= 0.006). Percentage overlap of home ranges
indicated a shift in home range location post-fire (Figure 2).
There was no significant difference in core (50% kernel) range
size over the 44 day period before versus after the fire (t
5
=
21.290, P= 0.267). However, the size of 95% kernel home range
differed significantly over this time period (t
5
=23.753, P=0.020),
with most cows having a larger 95% home range after the fire.
Home range size for the 44 day period after the first spring rains
was not significantly different to home range size before the rain
(50%: t
5
= 0.096, P= 0.928; 95%: t
5
= 0.654, P= 0.542). Percent-
age overlap of before rain 95% home range and after rain 95%
home range was 47.2%, 80.5%, 73.2%, 90.7%, 64.8% and 78.1%
for CE03, CE13, CE32, CE61, CE81 and CE88 respectively.
Therefore, all except severely injured cow CE03 had similar 95%
home range location before versus after the rain. This suggests that
the change in season post-fire was not the reason for the change in
95% home range we observed.
Herd fission/fusion. The time spent associating with other
herds pre- and post-fire was significantly different for uninjured
versus injured cows (t
4
=23.675, P= 0.021). For the first two
months post-fire, fission behaviour was exhibited by injured cows,
with CE03 and CE88 spending only 10.3 % and 34.7 % of their
time associating with other herds respectively, compared with 91.2
% and 62.2 % respectively before the fire. In the third month post-
fire, CE03 exhibited increased fusion behaviour, with association
time increasing from 10.3 % to 43.8 % and CE88 joined uninjured
collared cow CE13 (permanent association to August 2008).
Uninjured cows generally exhibited greater fusion behaviour after
the fire.
Physiological stress response
While pre- versus post-fire measurement, in general, had no
significant effect on stress hormone levels (F
1, 133
= 0.261,
P= 0.610), there was a significant difference among elephant
age-sex classes (i.e. juvenile, adult bull, adult cow) (F
2, 133
= 16.155,
P,0.001). There was also a significant interaction between pre-
versus post-fire and elephant age-sex class (F
2, 133
= 4.240,
P= 0.016). Before the fire, adult cow and juvenile stress levels
were not significantly different, but were both significantly lower
than adult bull stress levels (Figure 3). Cow stress levels increased
significantly post-fire but juvenile and bull stress levels were
unchanged by the fire (Figure 3).
Evaluation of stress hormone levels before versus after the first
Spring rain fell showed that there was no significant change in
stress hormone levels between wet and dry 44 day periods (F
1, 98
=
0.015, P= 0.902).
Elephant Response to Fire
PLoS ONE | www.plosone.org 2 September 2008 | Volume 3 | Issue 9 | e3233

Demography
Five family units and six independent adult bulls suffered burn
injuries (47 individuals), of which 29 mortalities occurred (17.6%
of pre-fire population total) (Table 1). Five juveniles between six
and ten years of age, 11 adult females and two adult males
recovered from their burn injuries (Table 1). Fifteen infants (#3
years old), seven weaned calves (4 to 10 years old), four adult
females and three adult males died, either as a direct result of their
burn injuries, or from euthanasia implemented by Park authorities
due to the severity of their injuries (Table 1). Initial post-fire
assessment resulted in the euthanasia of two of the four adult
females and all three adult males, with veterinarians deciding on
strict euthanasia criteria which included .50% burns to total body
surface area, marked oedema, eschars, severe supparative oozing
and severe impairment of mobility due to burn lesions. Seventeen
of the injured juveniles were taken to a holding facility off-site and
their wounds treated. Only two of these elephants survived and
were released back into the Park. Of the fifteen juveniles that died,
ten were euthanazed, with euthanasia criteria including .50%
burns to total body surface area, large skin surface area with open
Figure 1. Behavioural response to fire giving mean daily displacement (695% CL) of each collared cow over: (A) 4 days, (B) 10 days
and (C) 3 months, before and after the fire. Coefficients of variation (CV (%)) of daily displacement before or after the fire are given above or
below the upper or lower limit of CL bars. Pre-fire CV’s are located above if the value of pre-fire mean695% CL is located above (i.e. is greater than)
post-fire mean695% CL. Where mean695% CL pre- and post-fire are equal, pre-fire CV’s are located below CL bar.
doi:10.1371/journal.pone.0003233.g001
Elephant Response to Fire
PLoS ONE | www.plosone.org 3 September 2008 | Volume 3 | Issue 9 | e3233

tissue, comparative behavioural records indicating severe pain and
distress, collapse without recovery after revival, as well as low
blood protein and calcium. Euthanasia was only considered in
cases where recovery was impossible (criteria for recovery see [13])
and thus mortality can be considered representative of natural fire
mortality.
Figure 2. Behavioural response to fire showing core home range and 95% home range kernels for: uninjured adult cows (A) CE13, (B)
CE81, (C) CE61, (D) CE32 and injured adult cows (E) CE88 and (F) CE03 three months (a) before and (b) after injury in a fire on 21 September 2005 in
Pilanesberg National Park. The percentage overlap between 95% home ranges before and after the fire was: (A) 80.5, (B) 72.7, (C) 49.7, (D) 86.0, (E)
74.1 and (F) 52.5 respectively.
doi:10.1371/journal.pone.0003233.g002
Figure 3. Physiological response to fire indicated by glucocorticoid metabolite (stress) levels (mean695% CL) for adult bulls, adult
cows and juveniles before and after the fire. Sample size (n) is shown above each category.
doi:10.1371/journal.pone.0003233.g003
Elephant Response to Fire
PLoS ONE | www.plosone.org 4 September 2008 | Volume 3 | Issue 9 | e3233

Age structure, classified according to 10-year and 4-year age
classes, was significantly different after the fire than before (10-year
age classes: G
3
= 70.637, P,0.001; 4-year age classes:
G
3
= 71.598, P,0.001). Model projections over 30 years showed
no change in projected population growth rates achieved using
demographic data before the fire, as well as demographic data
after fire mortalities were accounted for (6.5% p.a.). It took four
years for the projected population to recover to the pre-fire
population size of 165 individuals (Figure 4). In the absence of fire,
the population was projected to grow to 303, 577, and 1079
individuals in 10, 20, and 30 years respectively (Figure 4). Taking
into account the effects of fire on the population structure, the
population was projected to reach 255, 485, and 903 individuals in
the same timeframes (Figure 4). For this type of mortality event to
reduce long-term population growth rate to 0%, it would be
required at a frequency of every three to four years (Figure 4).
Discussion
A large disturbance event causing catastrophic injury and
mortality has consequences that can significantly affect the
functioning and behaviour of an elephant population. In response
to a catastrophic fire in PNP, injured elephant cows showed an
initial short-term (lasting about four days) flight response post-fire,
a longer-term (over about ten days to three months) decrease in
daily displacement, a shift in home range, social withdrawal,
seclusion to non-tourist areas, and significantly higher stress levels.
However, behavioural responses were not limited to injured
individuals alone. Uninjured cows also showed altered physiolog-
ical and behavioural responses post-fire. These cows had
significantly raised stress levels, a general increase in daily
displacement, more variability in daily distance moved, withdraw-
al to non-tourist areas and a herd fusion response. Injured herds
therefore may have signalled their distress to uninjured herds.
Table 1. Demographic response to fire: elephant mortality
and survival recorded after a fire in Pilanesberg National Park
on 21 September 2005, giving the herd of origin and an
estimate of elephant age are given (as of December 2005).
Herd Collared cow Age of individuals in different categories *
Injured and died Injured and survived
Gold CE 57 1, 2,3, 4, 8 4,8, 8, 10,25,30,35
Monica CE 98 1, 2,4 10,12,30
Red CE 07 2, 2, 3, 35 8, 12
Sheena CE 88 2, 2, 3, 4, 4, 6, 30,30 8, 20
Yellow CE 03 1, 1,2, 4, 5, 30 15,42
Adult bulls 12, 15, 20 12, 15
*
Females are indicated in bold.
doi:10.1371/journal.pone.0003233.t001
Figure 4. The effect of fire on the future Pilanesberg National Park elephant population: (A) comparative modelled projection over a 30
year period using population data before and then after the fire in September 2005; (B) effect of three year and four year fire frequency on population
size over a 300 year period using population data before the fire, and the mortality parameters associated with this fire event.
doi:10.1371/journal.pone.0003233.g004
Elephant Response to Fire
PLoS ONE | www.plosone.org 5 September 2008 | Volume 3 | Issue 9 | e3233

Elephant family groups that show a high frequency of association
have been known to act in a co-coordinated manner, due to the
complex social behaviour and long-range communication used by
elephants [14,15]. The stress of injury, together with social
disruption due to the loss of and injury to family members is likely
to have affected the behaviour of injured breeding herds. This is
additional to the increased vulnerability of injured juveniles to
predation, or the compromised ability of injured adult cows to
protect their young, which would have increased stress levels. The
incidence of elephant calf predation has been found to increase
during times of drought when nutritional stress and dehydration
facilitates the circumstances where calves can lag behind the herd
and become vulnerable to predators [16]. Injured calves were seen
alone in PNP after the fire (pers. obs.), increasing their
vulnerability to predation.
The physiological and behavioural responses apparent in the
PNP population post-fire are consistent with elephant reactions to
stressful conditions. Breeding herds showed raised stress levels and
a fusion response to cow immobilizations and high-volume tourist
activity in PNP [9]. Working elephants in a safari operation had
high stress hormone levels associated with transportation and
episodic loud noises, such as lightning and thunderstorms and
human-induced activities, with baseline levels of faecal glucocor-
ticoid metabolites for adult elephants in PNP of approximately
25 ng.g
21
[12]. Heavily stressed elephants, responding to high
levels of poaching, formed larger groups than unstressed elephants
[8]. Stress responses to culling in Kruger National Park were initial
flight, taking elephants outside of home ranges [17], as well as the
movement of elephants into and out of culling regions in response
to culling events [18,19]. As these studies indicate, elephants are
stressed by human-induced and natural disturbances. Stressed
animals alter their behaviour in an attempt to eliminate the
stressor. Thus, shifts in home range and seclusion to non-tourist
areas are predictable, adaptable responses to disturbance.
Therefore, a fire event resulting in elephant mortality has the
potential to induce severe behavioural and physiological stress
responses (see [20] review of trauma effects on neuroendochrino-
logical development of elephants, and subsequent non-normative
behaviour). Whereas drought may cause elephant mortalities over
an extended period of time [21], fire mortality occurs within a
short time period after the event. Long-term elephant behavioural
response to fire mortality may therefore persist, due the dramatic
and traumatic nature of the event [20].
The demographic impact of fire on the PNP elephant
population predominantly involved the mortality of juveniles
(76% of total mortality). Among large herbivore populations where
predators are absent, high temporal variation in juvenile survival is
often seen, with fairly constant adult survival [22–24]. In systems
where large predators are present, both adult and juvenile survival
responds to environmental variability, due to interactions between
resource availability, population size and predation pressure [4].
Without constant predation pressure, the natural mortality
patterns often seen in African elephant populations during
stochastic events, such as droughts, follows that of the classic
mortality pattern seen in predator-free large ungulate populations,
which mainly involves juveniles [21,25,26]. Fire therefore
functions in a similar manner to other environmental catastrophes.
Population structure prior to the fire was significantly different
post-mortality, due to predominant mortality in the juvenile age-
class. In effect, the loss of a high proportion of juveniles serves not
only to lower population size, but also to increase calving interval
where the lost calf creates a gap between siblings. However,
among large herbivore populations, population growth is most
sensitive to adult mortality, especially that of prime-aged females
[22–24,29]. The mortality of only four adult females from the PNP
population as a result of the fire meant that projected population
growth rate remained unchanged. Therefore, the ability of this
type of stochastic, catastrophic mortality exhibited in the PNP fire
to limit population size or growth would require higher or more
frequent mortality, would need to include a higher proportion of
adult females [19,29], or cause demographic delays such as a
decline in conception rates, increased inter-calving interval, or
increased age at sexual maturity [10,30].
In order to reduce the PNP elephant population growth rate to
zero, this type and level of mortality event would be required at a
frequency of approximately three to four years. This gives an
indication of the resilience of elephant populations to environ-
mental perturbation. The demographic response of populations to
episodic mortality is influenced by the life-history characteristics of
the species. Elephant life-history is typical of large-bodied
ungulates in that these mammals have long generation times,
low fecundity and high adult survival [30,31]. In elephant
populations, a unique combination of life-history traits prolong
demographic response to environmental disturbance [30] and
maximum population growth rate tends to be maintained until the
very limit where forage resources can no longer support the
population before density dependent feedbacks occur [1,32,33].
Therefore, stochastic mortality alone has the potential to limit
short-term population size, but is unlikely to affect population
growth over the long-term. However, in combination with density-
dependent effects, elephant populations may be limited by
environmental catastrophes and the stochasticity brought about
by temporal variation in resources. Thus when populations are
close to carrying capacity, and background mortalities are higher
and fecundities lower than observed here, less intense mortality
would be required to achieve a stabilizing effect, and longer
intervals between periodic catastrophes could still result in fire-
induced mortalities influencing demographics substantially. Cat-
astrophic fires are likely to be rare events with expected return in
the order of decades [27]. Therefore in isolation these events will
not provide population regulation, but in combination with other
stochastic environmental events and density-dependent feedbacks,
they may play a role in population limitation. Thus removal of fire
from the system in some actively managed nature reserves may not
only be detrimental to the vegetation [7,28], but also the dynamics
of herbivore populations where fire mortality may be avoided.
These fire events should not be considered as negative catastro-
phes but instead as integral to the savanna system, with the
potential to make infrequent but positive contributions to the
regulation of abundant herbivore populations.
Burning during the late dry season, under ‘‘hot-fire’’ conditions
when fires can be very intense, can result in catastrophic mortality
of large mammal species. Arson fires during these times have the
potential to impact not only the vegetation of the area, but also
raise welfare concerns over any animals affected by the fire due to
the significant stress responses and behavioural changes which
may occur. The conservation status and abundance of species will
influence fire management requirements. The contribution of fire
mortality to abundant game species (e.g. blue wildebeest
(Connochaetes taurinus), impala (Aepyceros melampus)) population
dynamics may be less problematic than to that of threatened
species such as the black rhino (Diceros bicornis), which would
conversely require extreme awareness of the need to prevent
potentially intense catastrophic fires to ensure minimum mortality
impacts. If herbivore populations are fairly stable, even a rare
catastrophic fire could cause a shift in population dynamics that, in
combination with the current factors causing regulation, may
cause a population decline. Therefore, the integration of the
Elephant Response to Fire
PLoS ONE | www.plosone.org 6 September 2008 | Volume 3 | Issue 9 | e3233

conservation implications of intense, hot-fire suppression (i.e.
removal of a potential contributing factor to natural population
regulation), the welfare concerns of arson fires and burning during
‘‘hot-fire’’ conditions into fire management strategies for conser-
vation areas is important.
Materials and Methods
Study site
Pilanesberg National Park (PNP; 25u249S, 27u089E; 570 km
2
),
North West Province, South Africa, is located within the transition
zone of Kalahari Thornveld in the west and Bushveld in the east
[34]. The habitat consists mainly of savanna ranging from
broadleaf/Acacia thickets to open grassland. There are several dams
within the Park, one major perennial river system and many
ephemeral tributaries and streams. The region has summer rainfall
of approximately 630 mm p.a. Geologically, PNP is an extinct
volcanic crater formed over 1 200 million years ago and is an
example of an alkaline ring complex [35]. The weathering of this
complex has created a rugged, hilly landscape, with steep slopes and
deep valleys (Figure 5). PNP is open to tourists, but has large
‘‘wilderness’’ areas where there is no tourist access, limited
management tracks, and is rarely traversed by people (Figure 5).
This wilderness zone comprises approximately half of the total area
of PNP. Elephant were introduced to PNP between 1981 and 1998
[36]. As of early September 2005, the PNP population totalled 165
individually identified elephants, of which 37 were independent adult
bulls and 128 were part of 18 relatively stable matriarchal family
groups. All individuals in the population were known from unique
ear notches and tusk configuration. Pre-fire, the population was
made up of 86 juveniles under 10 years of age (56 males, 30 females),
23 10–20 year old adults (10 males, 13 females), as well as 29 adult
females and 27 adult males between the ages of 20 and 42 (oldest
elephants in population). There had been no mortality from old age,
with the first expected to occur in around 15 years time [37].
Behavioural response
We investigated both short-term and long-term responses of the
elephants to a fire event that caused elephant injury on the afternoon
of 21 September 2005. Short-term responses were investigated over
a four and ten-day period pre- and post-fire. Longer-term responses
were assessed over three months pre- and post-fire. Prior to the fire,
GPS-collars had been fitted to six elephant cows, belonging to
different breeding herds within the PNP population. The movement
of these elephants (CE03, CE13, CE32, CE61, CE81 and CE88) was
assumed to depict the movement behaviour of the breeding herd to
which they belonged [14]. Location points were taken at similar
times of the afternoon for each cow every day. Members of two of
these breeding herds were injured in the fire (CE03 and CE88).
CE03 was severely injured, with more than 45% total body surface
area (TBSA) burned and CE88 was less severely injured, sustaining
burn injuries to approximately 20% TBSA [13]. Analyses pre-fire
included data before fire injury occurred, while post-fire analyses
included data post-injury.
All statistical analyses in this paper were performed in SPSS
15.0 (SPSS Inc., Chicago, Illinois, USA) with a= 0.05. In the case
of parametric tests, assumptions were tested and satisfied. The
work was approved by the Animal Ethics Committee of the
University of KwaZulu-Natal.
Daily displacement. The distance moved by the collared
elephants each day (24 hour fixes) was calculated using polylines in
the Animal Movement Extension [38] to ArcView 3.2 (ESRI Inc.,
Redlands, California, USA). We considered this shortest line
between the two readings as an index of daily displacement, and
refer to this value as daily displacement hereafter.
To test whether there was an initial flight directly after the fire,
mean daily displacement of all cows over four days before and after
the fire was compared using a paired samples t-test. Four day mean
daily displacement of injured cows (CE03, CE88) as well as uninjured
cow in close proximity to fire during injury (CE32) was tested with a
paired-samples t-test and the same was done for four day mean daily
displacement of uninjured cows (CE13, CE61, CE81).
The mean daily displacement during a ten day period before the
fire was compared among cows using one-way ANOVA, and the
same was done for mean daily displacement for a ten day period,
post-flight, after the fire (i.e. day 5–14). A paired-samples t-test was
performed on each cow’s daily displacement ten days before and
post-flight, after the fire. Variability in displacement was assessed
using coefficient of variation (CV) for each cow’s daily displace-
ment over ten days, pre- and after flight, post-fire; these were
contrasted using a paired-samples t-test. We performed the same
contrasts of daily displacement from a three month period directly
before the fire (21 June 2005–21 September 2005) and after the
fire (22 September 2005–22 December 2005).
Home range. Each cow’s 24-hourly locations for a period of
three months pre- and post-fire were mapped in ArcView 3.2. We
calculated Kernel home ranges (core home range enclosed by the
50% probability contour and 95% home range enclosed by the
95% probability contour) in animal movement extension SA 2.1
Figure 5. Map of Pilanesberg National Park incorporating 20 m
contours, tourist roads, management tracks, the site where
elephants were injured in a fire on 21 September 2005 and the
approximate extent of the fire.
doi:10.1371/journal.pone.0003233.g005
Elephant Response to Fire
PLoS ONE | www.plosone.org 7 September 2008 | Volume 3 | Issue 9 | e3233

[38], using least-squares cross-validation (LSCV) smoothing. Both
50% core and 95% home ranges were compared before versus
after the fire using paired-samples t-tests. Percentage of overlap
between the 95% home range before the fire and after the fire was
calculated for each cow according to the following equation [39]:
% overlap~½(Aab=Ab|Aab =Aa)"1
=
2|100,
where A
ab
is the area of overlap between home range before the
fire (A
b
) and home range after the fire (A
a
). Percentage overlap
data, together with the percentage of locations of each cow in
either the wilderness or tourist zones of PNP for three months pre-
and post-fire, were used to establish whether a shift in home range
had occurred subsequent to the fire and to ascertain if the
elephants avoided the tourist zone after the disturbance. A paired-
samples t-test was used to compare the percentage location of
collared cows in the wilderness zone pre- and post-fire.
An increase in home range size has been reported for elephants
in semi-arid environments during the wet season, due to increased
access to areas with ephemeral water sources in the wet season
[40,41]. The first spring rains fell in PNP on 4 November, 2005. In
order to examine whether there was a change in home range size
and location after the rain (and therefore establish whether any
change could be attributed to a seasonal shift in home range
alone), kernel home range was calculated for 44 day periods before
the fire (9 August–21 September 2005), after the fire but before the
rain (22 September–4 November 2005), as well as after the rain (5
November–18 December 2005). Areas of 50% and 95% home
ranges for each collared cow over these time periods were
compared (paired-samples t-test).
Approximately 70% of PNP was burnt during the 2005 dry
season. Thus during the late dry season (post-fire and before the
rains), the availability of forage was similar throughout the Park in
terms of fire-impacted vegetation. We therefore did not consider
the post-burn condition of the vegetation as a bias to elephant
movement decisions over the study period.
Herd fission/fusion. To determine whether the breeding
herds showed a ‘fission’ or ‘fusion’ response (i.e. whether breeding
herds came together or dispersed, respectively) following the fire,
we compared the number of matriarchs (where one matriarch
indicates the presence of one herd) seen together in the three-
month period pre- and post-fire. A herd’s grouping tendency was
represented by the percentage fusion, calculated as the number of
sightings of a particular herd with other breeding herds, as a
percentage of the total number of sightings of that herd. For each
herd, percentage fusion was calculated before the fire and after the
fire. Because injured and uninjured herds showed opposite fusion
trends, a t-test on the difference between pre- and post-fire
percentage fusion was used to compare injured and uninjured
collared herds. Injured cow fusion response over three months
post-fire was further broken down into the first two months post-
fire, and the month following that, to examine any change in
association pattern on partial recovery from injury.
Physiological stress response
Severe, persistent stress can cause glucocorticoid levels to increase
and remain elevated [42]. The measurement of glucocorticoid
metabolite levels in elephant faeces has proven a useful non-invasive
way of investigating stress levels in African elephants [12,43,44]. A
total of 171 fresh (i.e. ,6 hours since deposition) elephant faecal
samples were collected randomly from the PNP population in the
three-month periods before and after the fire. Some samples were
collected from known individuals at the time of deposition. Those
that were not, were classified according to the approximate age of the
elephant from which they came [45], where age was estimated from
dung bolus size. Average dung bolus diameter greater than 16 cm
was considered to belong to adult bulls; all adult cows in the PNP
population, with the exception of one, were below 30 years of age at
the time of sample collection, corresponding to dung bolus diameters
of approximately 14 cm. For anonymous samples with a bolus
.16 cm in diameter, the sample was considered to originate from an
adult bull if it was from a site where a single track indicated the
presence of a large, solitary elephant. Samples were assigned to cows
if they were collected from a site where tracks indicated breeding
herd activity and if bolus diameters were between 10–14 cm.
Samples with a dung bolus diameter ,10 cm were considered to
belong to juveniles.
Faecal glucocorticoid metabolite levels were measured using
methods involving the use of a corticosterone I
125
radioimmuno-
assay (RIA) kit (MP Biomedicals, Costa Mesa, California, USA)
[46,47]. This assay has been validated and used for elephants
[12,47].
Differences in glucocorticoid levels among adult males, adult
females and juveniles, pre- and post-fire, were compared using
two-way ANOVA. Repeated measures ANOVA was not used
because samples before and after the fire were not necessarily, and
likely improbably, from the same individuals.
The effect of season on stress levels was examined to establish
whether any change in stress hormone level after the fire was
consistent with the onset of the first spring rains and thus a
seasonal change in stress hormone level. An analysis (two-way
ANOVA) of stress hormone levels was carried out over 44 day
periods post-fire (i.e. before the rain 22 September–4 November
2005 and after the rain 5 November–18 December 2005). Wet
and dry season stress levels for different elephant states (juveniles,
adult cows and adult bulls) were compared.
Demography
The effect of the fire on PNP elephant population size and age
structure was assessed by accounting for all mortalities (categorized
according to age and sex) and comparing population size and age
structure before the fire with that after the fire. A G-test was used
to assess if age structure was different, by separating total count
data into 10-year and 4-year age classes and comparing the
number of elephants in each age class pre- and post-fire.
The potential for fire to limit the population was considered
using a probabilistic age and state model [48]. The model was
used to calculate population size over 30 years, a time-period
relevant for conservation management decision-making, using
population data (1) before the fire, and (2) after the fire. The model
was also used to determine how often a fire of this nature would
need to occur for long-term population growth rate to be reduced
to zero over a period of 300 years, to allow for ample reproductive
generations and growth. Population growth rate was calculated
using projected population size from demographic data before the
fire and after the fire, according to the following standard equation
for exponential population growth:
% population growth~(er#1)|100,
where r = (ln Nt
2
–ln Nt
1
)/tand Nt
1
and Nt
2
are population size at
the beginning and end of the time interval in question,
respectively; and tis the length of the time span in years.
The model incorporated aspects of the life history of individuals
and the following important demographic parameters according to
acceptable values from the literature: maximum expected lifespan
of 60 years [18,37], female age at sexual maturity of ten years
Elephant Response to Fire
PLoS ONE | www.plosone.org 8 September 2008 | Volume 3 | Issue 9 | e3233

[10,31], average calving interval for the population of four years
[31,49], age at menopause of 50 years [37,50] and a 1:1 sex ratio
of newborns [31,49]. These parameters were a slightly conserva-
tive estimate of those estimated from past elephant demographic
patterns in PNP [10].
The model used was a probabilistic matrix model, where
numbers of individuals of different ages were transitioned through
specific biological states, i.e. males; sexually immature females;
sexually mature, non-pregnant females; pregnant females (in first
or second year of pregnancy); females in the first, second, third or
fourth year post-parturition. Males were aged but not transitioned
through specific states. All parameters other than average calving
interval were input as probabilities for each age and state,
determined by comparison of input probability value with one
obtained from a random number generator that produced a
normal distribution of values between 0 and 1. The statistical
variation introduced by the probabilistic approach was determined
by repeating each simulation 500 times and the means and
standard deviations were calculated from these replicate simula-
tions. The population was recorded at the end of each year of a
simulation.
Acknowledgments
We thank Pilanesberg National Park management (North-West Parks and
Tourism Board) for logistical support and permission to undertake this
study. Special thanks to Hans Meier for the use of ParkControl!software,
facilitating elephant location in the field, and to Karen Trendler for
valuable information on euthanasia criteria and juvenile elephant response
to severe burn wounds. Brian van Wilgen contributed valuable comments
to the manuscript.
Author Contributions
Conceived and designed the experiments: LAW RS. Performed the
experiments: LAW. Analyzed the data: LAW RLM. Contributed reagents/
materials/analysis tools: LAW JM RJW SJvR BP. Wrote the paper: LAW
RS.
References
1. Owen-Smith N, Kerley GIH, Page B, Slotow R, van Aarde RJ (2006) A
scientific perspective on the management of elephants in the Kruger National
Park and elsewhere. S Afr J Sci 102: 389–394.
2. Sinclair ARE (2003) Mammal population regulation, keystone processes and
ecosystem dynamics. Phil Trans Roy Soc Lond B 358: 1729–1740.
3. Coulson T, Catchpole EA, Albon SD, Morgan BJT, Pemberton JM, et al. (2001)
Age, sex, density, winter weather and population crashes in Soay sheep. Science
292: 1528–1531.
4. Owen-Smith N, Mason DR, Ogutu JO (2005) Correlates of survival rates for 10
African ungulate populations: density, rainfall and predation. J Anim Ecol 74:
774–788.
5. Mangel M, Tier C (1994) Four facts every conservation biologist should know
about persistence. Ecology 75: 607–614.
6. Woolley L-A, Mackey RL, Page BR, Slotow R (2008) Modelling the effect of
age-specific mortality on African elephant population dynamics: can natural
mortality provide regulation? Oryx 42: 49–57.
7. van Wilgen BW, Govender N, Biggs HC, Ntsala D, Funda XN (2004) Response
of savanna fire regimes to changing fire-management policies in a large African
national park. Conserv Biol 18: 1–8.
8. Andersen KF, Eltringham SK (1997) Some preliminary observations on possible
stress in the elephants of Mikumi National Park, Tanzania. Afr J Ecol 35:
278–282.
9. Burke T (2005) The effect of human disturbance on elephant behaviour,
movement dynamics and stress in a small reserve: Pilanesberg National Park.
MSc thesis. Durban: University of KwaZulu-Natal.
10. Mackey RL, Page BR, Duffy KJ, Slotow R (2006) Modeling elephant population
growth in small, fenced, South African reserves. S Afr J Wildl Res 36: 33–43.
11. Foley CAH, Papageorge S, Wasser SK (2001) Non-invasive st ress and
reproductive measures of social and ecological pressures in free-ranging African
elephants (Loxodonta africana). Conserv Biol 15: 1134–1142.
12. Millspaugh JJ, Burke T, van Dyk G, Slotow R, Washburn BE, et al. (2007) Stress
response of working African elephants to transportation and Safari Adventures.
J Wildlife Manage 71: 1257–1260.
13. Shakespeare A, Strydom S (2006) A method for determining the extent of
thermal burns in elephants. J S Afr Vet Assoc 72: 70–74.
14. Moss CJ, Poole JH (1983) Relationships and social structure of African
elephants. In: Hinde RA, ed. Primate social relationships: an integrated
approach. Oxford: Blackwell Scientific Publications. pp 315–325.
15. McComb K, Moss CJ, Sayialel S, Baker L (2000) Unusually extensive networks
of vocal recognition in African elephants. Anim Behav 59: 1103–1109.
16. Loveridge AJ, Hunt JE, Murindagomo F, Macdonald DW (2006) Influence of
drought on predation of elephant (Loxodonta africana) calves by lions (Panthera leo)
in an African wooded savannah. J Zool 270: 523–530.
17. Whyte IJ (1993) The movement patterns of elephant in the Kruger National
Park in response to culling and environmental stimuli. Pachyderm 16: 72–80.
18. Whyte I, van Aarde RJ, Pimm SL (1998) Managing the elephants of Kruger
National Park. Anim Conserv 1: 77–83.
19. van Aarde RJ, Whyte IJ, Pimm S (1999) Culling and the dynamics of the Kruger
National Park elephant population. Anim Conserv 2: 287–294.
20. Bradshaw GA, Schore AN (2007) How elephants are opening doors:
developmental neuroethology, attachment and social context. Ethology 113:
426–436.
21. Dudley JP, Craig GC, Gibson D St C, Haynes G, Klimowicz J (2001) Drought
mortality of bush elephants in Hwange National Park, Zimbabwe. Afr J Ecol 39:
187–194.
22. Saether B (1997) Environmental stochasticity and population dynamics of large
herbivores: a search for mechanisms. TREE 12: 143–149.
23. Gaillard J-M, Festa-Bianchet M, Yoccoz NG (1998) Population dynamics of
large herbivores: variable recruitment with constant adult survival. TREE 13:
58–63.
24. Gaillard J-M, Festa-Bianchet M, Yoccoz NG, Loison A, Toigo C (2000)
Temporal variation in fitness components and dynamics of large herbivores.
Annu Rev Ecol Syst 31: 367–393.
25. Wittemyer G, Daballen D, Rasmussen H, Kahindi O, Douglas-Hamilton I
(2005) Demographic status of elephants in the Samburu and Buffalo Springs
National Reserves, Kenya. Afr J Ecol 43: 44–47.
26. Leggatt K (2003) The effect of drought and low rainfall on elephant populations.
Proceedings of the Norwegian Bonnic Programme Workshop. Kasane:
Botswana. 49 p.
27. van Wilgen BW, Biggs HC, O’Regan SP, Mare N (2000) A fire history of the
savanna ecosystems in the Kruger National Park, South Africa, between 1941
and 1996. S Afr J Sci 96: 167–178.
28. Brockett BH, Biggs HC, van Wilgen BW (2001) A patch mosaic burning system
for conservation areas in southern African savannas. Int J Wildland Fire 10:
169–183.
29. Owen-Smith N, Mason DR (2005) Comparative changes in adult vs. juvenile
survival affecting population trends of African ungulates. J Anim Ecol 74:
762–773.
30. Wittemyer G, Rasmussen HB, Douglas-Hamilton I (2007) Breeding phenology
in relation to NDVI variability in free-ranging African elephant. Ecography 30:
42–50.
31. Moss CJ (2001) The demography of an African elephant (Loxodonta africana)
population in Amboseli, Kenya. J Zool 255: 145–156.
32. Fowler CW, Smith T (1973) Characterizing stable populations: an application to
the African elephant population. J Wildlife Manage 37: 513–523.
33. Gough KF, Kerley GIH (2006) Demography and population dynamics in the
elephants Loxodonta africana of Addo Elephant National Park, South Africa: any
evidence of density dependent regulation? Oryx 40: 434–441.
34. Acocks JPH (1988) Veld types of South Africa. Memoirs of the Botanical Survey
of South Africa. South Africa: Botanical Research Institute, Department of
Agriculture and Water Supply.
35. Boonzaaier WW, Collinson RFH (2000) Pilanesberg National Park management
plan. Pilanesberg National Park Management Series. Rustenburg, South Africa:
Contour Project Managers and Collinson Consulting.
36. Slotow R, van Dyk G (2004) Ranging of older male elephants introduced to an
existing small population without older males: Pilanesberg National Park.
Koedoe 47: 91–104.
37. Hanks J, McIntosh JEA (1973) Population dynamics of the African elephant
(Loxodonta africana). J Zool 169: 29–38.
38. Hooge PN, Eichenlaub B (1997) Animal movement extension to ArcView
version 1.1. Alaska Biological Science Center. Anchorage: U.S. Geological
Survey.
39. Kernohan BJ, Gitzen RA, Millspaugh JJ (2001) Analysis of animal space use and
movements. In: Millspaugh JJ, Marzluff JM, eds. Radio tracking animal
populations. San Diego: Academic Press. pp 125–166.
40. Ottichilo WK (1986) Population estimates and distribution patterns of elephants
in the Tsavo ecosystem, Kenya in 1980. Afr J Ecol 24: 53–57.
41. Lindeque M, Lindeque PM (1991) Satellite tracking of elephants in northwestern
Namibia. Afr J Ecol 29: 196–206.
42. Creel S, Creel NM, Monfort SL (1997) Radiocollaring and stress hormones in
African wild dogs. Conserv Biol 11: 544–548.
43. Stead S, Meltzer DGA, Palme R (2000) The measurement of glucocorticoid
concentrations in the serum and faeces of captive African elephants (Loxodonta
africana) after ACTH stimulation. J S Afr Vet Assoc 71: 192–196.
Elephant Response to Fire
PLoS ONE | www.plosone.org 9 September 2008 | Volume 3 | Issue 9 | e3233

44. Ganswindt A, Palme R, Heistermann M, Borragan S, Hodges JK (2003) Non-
invasive assessment of adrenocortical function in the male African elephant
(Loxodonta africana) and its relation to musth. Gen Comp Endocr 134: 156–166.
45. Morrison TA, Chiyo PI, Moss CJ, Alberts SC (2005) Measures of dung bolus size
for known-aged African elephants (Loxodonta africana): implications for age
estimation. J Zool 266: 89–94.
46. Schwarzenberger F, Mo¨stl E, Bamberg E, Pammer J, Schmehlik O (1991)
Concentrations of progestagens and oestrogens in the faeces of pregnant
Lipizzan, Trotter and Thoroughbred mares. J Reprod Fertil Sup 44: 489–499.
47. Wasser SK, Hunt KE, Brown JL, Cooper K, Crockett CM, et al. (2000) A
generalized faecal glucocorticoid assay for use in a diverse array of non–domestic
mammalian and avian species. Gen Comp Endocr 120: 260–275.
48. Wu LS-Y, Botkin DB (1980) Of men and elephants: A discrete, stochastic model
for long-lived species with complex life-histories. Am Nat 116: 831–849.
49. Whitehouse AM, Hall-Martin AJ (2000) Elephants in Addo Elephant National
Park, South Africa: reconstruction of the population’s history. Oryx 34: 46–55.
50. Woodd AM (1999) A demographic model to predict future growth of the Addo
elephant population. Koedoe 42: 97–100.
Elephant Response to Fire
PLoS ONE | www.plosone.org 10 September 2008 | Volume 3 | Issue 9 | e3233