Modelling elephant population growth in

small, fenced, South African reserves

Robin L. Mackey , Bruce R. Page , Kevin J. Duffy & Rob Slotow *
1,2 1 2 1

1
Amarula Elephant Research Programme, School of Biological and Conservation Sciences, George Campbell Building,
University of KwaZulu-Natal, Durban, 4041 South Africa
2
Centre for Systems Research, Durban Institute of Technology, Durban, South Africa
Received 13 September 2004. Accepted 14 December 2005

Elephants have been, and continue to be, introduced to small conservation areas in South
Africa despite concerns that elephants may adversely affect the vegetation for other herbivores,
reduce the aesthetic value of the landscape for tourism, and fundamentally reduce local
biodiversity. The rate at which elephant populations grow is essential knowledge for
elephant management. We examined growth rates of elephant populations introduced to
eight small reserves in South Africa, determined demographic parameters (age at sexual
maturity, calving interval, sex ratio at birth, and age-specific probabilities of conception and
survival) for each population, and made short-term projections of future population growth
based on these parameters and the current population structure. Average annual population
growth rates following introduction ranged from 5 to over 10%, and are projected to be in the
range of 6.2 to 8.9% over the duration of the next 20 years. Model projections suggest that
populations can double in less than 10 years, that calving interval is the most important
demographic factor determining population growth rate, and that the structure of the
introduced population will greatly influence future growth. Without active management to
reduce or maintain population sizes in the next few years, small reserves throughout South
Africa will undoubtedly experience overpopulation of elephants within 10 to 20 years.
Population growth rates of elephants in new areas and on the short term can be much higher
than expected from large, established populations.
Key words: African elephant, conservation, elephant demography, population growth, population
model, South Africa.

INTRODUCTION large numbers to new areas.

The conversion of rangeland to conservation To effectively manage elephants, managers
areas in South Africa has resulted in the formation require knowledge about the rates at which their
of many small parks and game reserves. To promote populations may grow and insight into what size
tourism and/or recreate historical ecological and structure of introduced population will best fit
communities, more than 800 African elephants their short- and/or long-term management needs.
(Loxodonta africana) have been introduced to Without knowing how quickly elephant populations
approximately 60 reserves in South Africa between will grow or how many elephants the reserve can
1979 and 2001 (Garai et al. 2004). Subsequent to support on a long-term basis, managers have little
the introduction of elephants, vegetation commu- guidelines for their introductions (in terms of num-
nities and habitat have changed (e.g. Cumming ber of individuals, age structure, and sex ratio) and
et al. 1997; Johnson et al. 1999; Lombard et al. ongoing management needs. The size and struc-
2001; Wiseman et al. 2004). Such changes not ture of a reserve’s desired elephant population
only alter the appearance of the landscape (which will depend on the objectives of that reserve; for
may affect the tourism value of the area), but can example, a reserve being managed for the conser-
also include the extirpation of some plant species vation of biodiversity will want fewer elephants
(Cumming et al. 1997; Waithaka 1997; Lombard than a similarly sized reserve that stocks elephants
et al. 2001) and shift the plant community to domi- for tourism or consumptive use. Either way,
nance by species unpalatable to other herbivores though, the issue of high population growth rates
(Wiseman et al. 2004). Despite these issues, and the need for population management are
elephants continue to be introduced in relatively necessary. To date, introduction and management
*To whom correspondence should be addressed.
policies for elephants in small reserves have been
E-mail: slotow@ukzn.ac.za based on expectation from natural populations,
South African Journal of Wildlife Research 36(1): 33–43 (April 2006)
34 South African Journal of Wildlife Research Vol. 36, No. 1, April 2006

expectation from established areas, and a post- could obtain detailed elephant introduction data
ponement of management intervention until crisis (sex and age or size estimates for each introduced
necessitates action (Slotow et al. 2005). individual) and/or subsequent detailed population
Recent observations indicate that elephant census data. Sufficient data were compiled for
populations are growing exponentially at rates eight small reserves: Hluhluwe-iMfolozi Park,
much higher than previously assumed realistic. Mabula Game Reserve, Madikwe Game Reserve,
Hanks & McIntosh (1973) and Calef (1988) Phinda Game Reserve, Pilanesberg Game Re-
predicted maximum population growth rates of 4% serve, Pongola Game Reserve, Greater St Lucia
and 7%, respectively. However, Slotow et al. Wetland Park, and Welgevonden Private Game
(2005), in their examination of South Africa’s small, Reserve.
confined, re-introduced elephant populations,
calculated average annual growth rates over 15%. Population introductions
Consequently, management policies based on an The quantity and quality of data for introduced
assumption of future annual population growth in populations varied among reserves. Unless other-
the range of 4–7% may lead to populations that wise indicated (see below and Acknowledg-
very quickly approach or exceed the desired ments), introduction and census data were
elephant density for the reserve. This trend may be obtained from EMOA (Elephant Managers and
a very real and current problem in South African Owners Association, which collated data up to
reserves, as many elephant populations have 2001). The ages of introduced individuals were
increased much more quickly than managers had estimated from direct shoulder height measure-
expected, and management now needs to devise ments taken during the translocation process,
and implement policies to deal with surplus ele- using Laws et al.’s (1975) growth relationships.
phants in their reserves. Thus, accurate estimates Known introduction, removal, cull, and death data
of population growth and demographic parameters, for each of the eight populations are summarized
in addition to projections of future population in Table 1.
growth based on actual introduced populations,
will be invaluable for the management of existing Birth data
elephant populations and for devising introduction Birth data for the eight populations are of varying
plans for future reserves. quality. For the Mabula, Pongola and St Lucia
The aim of this study is to address conservation populations, detailed birth records enabled accu-
implications of elephant population growth, and rate aging and mother–offspring relationships of
the main objectives are: 1) calculate rates at which elephants born subsequent to the introductions.
small, confined, re-introduced South African ele- For populations in Phinda and Pilanesberg, birth
phant populations are increasing; 2) determine data of current subadults were based either on 1)
which demographic parameters influence popula- known birth years and mother–offspring relation-
tion growth; and 3) make projections of future ships, if available, or 2) field assignment of
elephant population growth if active management subadults to age classes, based on height and
of existent populations is not undertaken. From appearance relative to individuals of known ages,
detailed census data, historical growth rates are and likely mother–offspring relationships based on
calculated for elephant populations introduced to behaviour and groupings within the family units.
eight small reserves in South Africa. Using empiri- For the Welgevonden and Madikwe populations,
cal data from these populations in a stochastic ages of elephants born subsequent to introduction
model of population growth, future population were based solely on estimates of number of
growth and size are predicted for the eight reserves. births per year, as provided by EMOA; mother–
We also examine which demographic parameters offspring relationships were not known for these
(calving interval, age at sexual maturity, sex ratio populations. In the case of Hluhluwe-iMfolozi,
of newborns, probability of conception, probability the 2003 population structure was manually
of survival) have the greatest effect on population constructed using the introduced populations,
growth. mark and recapture estimates of 2003 population
size and an observed average of 2.5 subadults per
METHODS female (Slotow et al., unpubl. data), and estimates
This study of elephant population growth in South of demographic parameters (female age at sexual
Africa is limited to parks and reserves for which we maturity = nine years, calving interval = three
 Mackey et al.: Modelling elephant population growth in small South African reserves 35

Table 1. The numbers of individuals introduced, removed and/or culled, and that died from reasons other than culling
for the eight elephant populations examined in this study. Sex of individuals is indicated by F (female) and M (male). To
simplify the table, individuals are broken down into two age groups (<10 years old and ≥10 years old). The data reflect
the total numbers accounted for from the initial introduction date through to the end of 2003.

Population1 Date of 1st Introduced Removed/culled Died

introduction <10 yrs ≥10 yrs <10 yrs ≥10 yrs <10 yrs ≥10 yrs
F M F M F M F M F M F M

Hluhluwe-iMfolozi 1981 1842 10 3 4 12 2 11 12 3

Mabula 1992 4 2 1
Madikwe 1992 70 45 90 28 2 2
Phinda 1992 34 27 13 7 15 6 1 3
Pilanesberg 1979 35 50 2 8 14 1 13 2
Pongola 1997 7 3 7 9 1 12 2
St Lucia 2001 3 11 18 7 1
Welgevonden 1994 10 17 19 4 6 2
1
Data sources. Hluhluwe-iMfolozi: Dominy et al. 1998, EMOA (Elephant Managers and Owners Association); Mabula: Pretorius (2004),
EMOA; Madikwe: EMOA; Phinda: H. Druce (pers. comm.), EMOA; Pilanesberg: T. Burke (pers. comm.), EMOA; Pongola: G. Shannon
(pers. comm.), EMOA; St Lucia: C. Dickson (pers. comm.); Welgevonden: J. Malan (pers. comm.), EMOA.
2
Sex unknown.

years) that were probable for the size, sex ratio, calculated for the other seven populations (i.e. 10
and adult:subadult ratio estimated for the 2003 years old). Because it was not possible to calcu-
population (Slotow et al., unpubl. data). late age-specific probabilities of conception for the
majority of the populations, we assumed all
Demographic parameters females between the ages of sexual maturity
Several demographic parameters were estimated and menopause will be reproductively active (i.e.
for each of the eight populations: 1) age at sexual probability of conception = 1).
maturity for females (Table 1), 2) calving interval Calving interval (average number of years
(Table 1), 3) age-specific probabilities of conception, between consecutive births for a mother) was
4) birth sex ratio, 5) age-specific probabilities of calculated for populations with detailed accounts
survival, 6) age at menopause, and 7) maximum of mother-calf relationships and accurate age
age. Parameter estimation required accurate estimates of offspring (Table 2). For populations
accounts of births and deaths, identification of without such data (Hluhluwe-iMfolozi, Madikwe,
mother–offspring relationships, and ages of mothers Welgevonden), calving interval was estimated by
and offspring. For populations in which these data manipulating calving interval in our population
were not available (Hluhluwe-iMfolozi, Madikwe, growth model until model projections of historical
St Lucia, Welgevonden), demographic parameters growth matched the known historical growth of
were derived through other means (see below). the population; admittedly, this is not the best
Female age at sexual maturity was determined approach, but without accurate birth data it was
from the mean and median age at which females in our best option.
the population gave birth (Table 2). To account for In populations where sex ratio of births since
the average 22 months gestation period, two years introduction could be calculated, the ratio was
was subtracted from this age to provide an estimate equal to or very close to 1:1. This ratio was applied
of age at sexual maturity. Because sufficient birth to all eight populations.
data were not available for the Hluhluwe-iMfolozi, Age at menopause was not possible to determine
Madikwe and Welgevonden populations, age at from the relatively young populations in our
sexual maturity was back-calculated from census dataset. Estimates from the literature range between
data and determined to be the value that, along 45 (Hanks & McIntosh 1973) and 60 years (Wu &
with calving interval estimates (see below), makes Botkin 1980). Owen-Smith’s (1988) and Hanks &
population projections that most closely match the McIntosh’s (1973) moderate estimate of 50 years
historical counts. The St Lucia population was was adopted for this study.
too new to calculate a confident age at sexual Very few deaths have been recorded in the eight
maturity; therefore, we assigned the average age elephant populations studied, indicating a high
36 South African Journal of Wildlife Research Vol. 36, No. 1, April 2006

Table 2. Average calving interval and female age at sexual maturity for eight elephant populations. ‘Estimates’ indicate
the integer value used in the model simulations.

Population Calving interval (yrs)3 Age at sexual maturity (yrs)

Mean ± 95% C.L. Estimate Mean ± 95% C.L.4 Median Estimate

Hluhluwe-iMfolozi1 3 9
Mabula 2.40 ± 0.12 3 12.33 ± 1.06 12 12
Madikwe1 5 11
Phinda 3.91 ± 0.35 4 10.25 ± 1.11 10 10
Pilanesberg 3.29 ± 0.18 3 9.18 ± 0.44 9 9
Pongola 3.09 ± 0.36 3 8.37 ± 1.06 8 8
St Lucia2 4 10
Welgevonden1 4 8
1
There were insufficient data to calculate these parameters from birth data for the Hluhluwe-iMfolozi, Madikwe, and Welgevonden popula-
tions. Therefore, calving interval and age at sexual maturity were calculated based on what would be required to produce the most recent
census population, given the introduction, removal, culling, and death data reported in Table 1.
2
The St Lucia population is too new (established 2003) for parameters to be calculated, therefore the averages of the other seven popula-
tions were used, and this should be taken into consideration when projections for this population are examined.
3
Calculated from all known inter-calf intervals of all females with more than one calf. Presented as mean ± 95% confidence limits. For
Mabula, the calving interval estimate used for simulations was rounded up to three years because a prolonged two-year interval was
deemed unrealistic.
4
Calculated from all known mother-calf relationships with first-born elephants of known age. To account for a 22-month gestation period, two
years was subtracted from age at first birth to estimate age at sexual maturity. Two years, rather than 22 months, was subtracted because
birth dates generally are not known to the month; using 1.83 years would give a false sense of accuracy.

probability of survival. Where deaths were recorded, growth in small reserves in South Africa, we used a
ages of individuals were rarely specified, and model based on Wu & Botkin’s (1980) model that
calculation of age-specific mortality or survival takes into account the life history of individuals in
was therefore not attainable. Following Whyte the population. Tables 1 and 2 contain summaries of
et al. (1998), we set probability of survival at 1.0 for parameters and values used in our model. All
all individuals below the maximum age. Maximum individuals are modelled in one-year age classes,
age was set at 60 years (Hanks & McIntosh 1973; and the model was designed to examine the popu-
Wu & Botkin 1980; Whyte et al. 1998); individuals lation in discrete one-year time intervals. Over the
reaching 60 years of age had a probability of course of her lifetime, a modeled female will transi-
survival of 0. tion through several biological states: immature;
sexually mature, but not pregnant; in first year of
Population growth pregnancy; in second year of pregnancy; in first,
Historical average annual elephant population second, or third year of lactational (post-parturition)
growth rates were calculated for each of the anoestrous; and dead. Male individuals similarly
eight populations using the standard equation for age and die, but are not transitioned through matu-
exponential population growth (Begon & Mortimer rity states. It is assumed that sufficient sexually
1986): mature males are always present in the population
r (i.e. conception is not limited by males). The sex of
% population growth = (e –1) × 100,
a newborn calf is randomly determined, based on
where r = (lnNt2–lnNt1)/t; Nt2 and Nt1 are population a constant 1:1 sex ratio.
size at the end and beginning of the time span in Transitions from one biological state to another
question, respectively; and t is the length of the are age- and/or state-specific. A female individual
time span in years. When possible, to prevent reaches sexual maturity at a given age, which is
estimates of natural growth rates from being constant over time for a given population but may
altered by introductions or removals of individuals, vary among populations. Once she is sexually
population growth rates were calculated for the mature, her probability of conceiving is 1.0.
period from the most recent introduction/removal Probability of conception remains 1.0 until age
up to the most recent population census. 50, after which menopause is assumed and the
probability of conception drops to 0. An individual’s
The model ability to conceive is affected by her biological
To make projections of elephant population state and by the population’s calving interval:
 Mackey et al.: Modelling elephant population growth in small South African reserves 37

pregnant individuals cannot conceive, nor can that the populations will experience during the
individuals within a specified duration (one, two, or projection time. The model makes projections
three years) of lactational anoestrous. For a popu- based on the assumption that environmental
lation with a calving interval of three years, an conditions will not change significantly, on average,
individual will not be able to conceive until three during the next 20 to 25 years.
years after her previous conception (i.e. for two To test the robustness of our projections, sensitivity
years while she is pregnant, and for one year of analyses were performed on the demographic
lactational anoestrous). Each population was parameters and probabilities included in the model
assigned a constant calving interval of three, four, (i.e. age at sexual maturity, calving interval, sex
or five years. ratio of newborns, probability of conception, and
During each one-year increment of the simula- probability of survival). Parameters were varied
tion, the following sequence of transitions and sufficiently for the sensitivity analysis to encompass
events occurred. First, all individuals were aged by realistic ranges of values.
one year. Females in lactational anoestrous during The algorithms are coded in FORTRAN 90,
the previous one-year interval were then trans- compiled with IBM’s xlf 8.1 beta for Mac OS X.
ferred to the next year of anoestrous, or to the non- Simulations ran on the BSD UNIX layer of Mac OS
pregnant or first year of pregnancy state, accord- X 10.2.8.
ing to the population’s calving interval and proba-
bility of conception. Females who were in their RESULTS
second year of pregnancy in the previous one-year In general, elephant populations will more than
interval were transitioned to their first year of double by 2015 if there are no changes in the
lactational anoestrous, and their calves were current demographic parameters that control
assigned a sex and added to the 0–1 year age population growth (Table 3, Fig. 1). Projected
class. Females previously in their first year of preg- growth rates were calculated for the time period of
nancy were moved into their second year of 2003 to 2025, and ranged from 6.2% to 8.9%
pregnancy. The survival probability function was per year (Table 3). The Madikwe and Phinda
applied, and individuals that had reached the populations had the slowest projected growth rates
maximum age of 60 years were removed. Young while the Pongola population had the greatest.
females reaching the age of sexual maturity in this Projected annual population growth rates were
interval were moved into either the nonpregnant or less variable between populations than were the
first year of pregnancy state, as determined by the pre-projection (‘Historical’) growth rates which
probability of conception function. were calculated from census data (Table 3).
This sequence was repeated every year for 21 to Historical growth rates were calculated over a
24 years (to give projections up to the year 2024 minimum time span of five years and reflect the
(i.e. 20 years from the current date)). At the end of average annual rates at which the actual popula-
each year, the population was censused. Total tions were increasing prior to the dates at which
number of elephants, individuals per one-year age the simulations commenced. Historical growth
class, and individuals per state (male, immature rates ranged from 4.7% in Hluhluwe-iMfolozi to
female, nonpregnant mature female, female in 11.8% in Pongola, and perhaps higher in Mabula
first year of pregnancy, female in second year of (Table 3).
pregnancy, and females in first, second, or third Population density varied greatly among the
year of lactational anoestrous) were calculated. reserves (Table 3). Densities of the real populations
2
Means and standard deviations of these yearly ranged between 0.07 elephants/km in St Lucia
2
counts were calculated from 500 replicates of the and 0.64 elephants/km in Pongola. The Pongola
simulation. population also had the highest projected density
2
Because the intention of our study was to make at 4.43 elephants/km in 2025. And, St Lucia was
relatively short-term projections of elephant popu- projected to remain the least dense increasing to
2
lation growth, we didn’t include environmental 0.35 elephants/km in 2025.
change and time-varying parameters in the model. Sex ratios and the percentage of reproductive
Instead, we assumed that the conditions experi- females in the population (females from age at
enced over the past several years (i.e. the time sexual maturity to 50 years old) were quite vari-
span for which demographic parameters were able among populations prior to the simulations
calculated) are representative of the conditions (Table 4), due to sex biases in introductions and/or
38 South African Journal of Wildlife Research Vol. 36, No. 1, April 2006

Table 3. Recent historical and projected population size (2015 and 2025), density (2025), and average annual growth
rates (2025) for eight South African elephant populations.

Historical population Population projections

2015 2025
Population Size1 Density Annual growth Size4 Size4 Density Annual growth
(/km2)1 rate (%)2 (/km2) rate (%)5

Hluhluwe-iMfolozi 387 0.43 4.7 946 ± 0.1 2063 ± 1.4 2.29 7.7 (7.6–7.7)
Mabula 10 0.12 15.3 (4.8)3 26 ± 0.0 54 ± 0.2 0.64 7.6 (7.3–7.9)
Madikwe 323 0.52 6.8 833 ± 0.1 1503 ± 0.9 2.42 6.2 (6.1–6.3)
Phinda 71 0.48 8.2 150 ± 0.0 281 ± 0.4 1.94 6.4 (6.3–6.5)
Pilanesberg 159 0.32 10.6 412 ± 0.1 903 ± 1.0 1.84 7.9 (7.9–7.9)
Pongola 47 0.64 11.8 140 ± 0.0 328 ± 0.7 4.46 8.9 (8.5–9.3)
St Lucia 39 0.07 n/a 98 ± 0.0 188 ± 0.4 0.35 6.8 (6.4–7.1)
Welgevonden 77 0.23 7.1 225 ± 0.2 454 ± 1.1 1.38 7.1 (7.0–7.1)
1
The dates for the ‘historical’ population size and density estimates are: 2000 for Welgevonden, 2001 for Madikwe, and 2003 for all others.
Mabula, Pongola, Phinda, St Lucia, and Pilanesberg estimates are from field counts. HIP is estimated from a mark and recapture study.
Welgevonden and Madikwe estimates are calculated from the introduction, birth, death, and removal data.
2
For timespan of : Hluhluwe-iMfolozi ,1990–1996 (based on Dominy et al.’s (1998) statement of 49 births during this period, and excluding
the 1994 introduction of 12 immature individuals); Madikwe, 1993–2001; Phinda, 1994–2003 (takes into account a total of two males and
one female culled in 1996 and 1998; estimate may be lower than the potential growth rate due to the removal of the potentially reproductive
female); Pilanesberg, 1998–2003; Pongola, 1997–2003; St Lucia, growth rate not available because all individuals were introduced from
2001– 2003; Welgevonden, 1994–2000.
3
15.3: for timespan of 1998(year of first birth)–2003. 4.8: 1992(year of introduction)–2003. The true long-term growth is likely intermediate to
these estimates.
4
Mean elephant population size ± standard error, calculated from 500 replicates of the simulation.
5
Annual population growth rate and 95% confidence limits calculated from slope of ln(population size) vs time for duration of 2003–2025.

removals. In all cases but Hluhluwe-iMfolozi and annual population growth rate, the model is fairly
Phinda, females were more abundant than males. insensitive to small changes (i.e. 1–3%) in mortal-
Over the duration of the simulation, sex ratios ity rates of juvenile age classes (Table 7). Projec-
converged towards equality, which is likely a result tions are not particularly sensitive to changes in a
of the assumed equal mortality rates for females female’s probability of conception; reductions in
and males. The percentage of reproductive females probability of conception from 1.0 to 0.9 and 0.7
in the population, which ranged between 19.9% yield mean declines in projected growth rates of
and 46.1% prior to the simulations, converged to only 0.2 ± 0.1 and 0.6 ± 0.1 percentage points, re-
a narrower range of 23.6% to 28.0% of the total spectively. To have a strong effect on growth pro-
population by 2025. jections, the declines in probability of conception
Sensitivity analyses indicate the response of the would have to be sufficient to increase average
projected elephant population sizes to changes in calving interval by a year or more. Finally, the
the demographic parameters of the model, or the model projections are robust to variation in new-
robustness of model projections to changes or born sex ratio, with changes in proportion of new-
errors in demographic parameter estimation. borns being male changing from 0.5 to 0.4 and 0.6
Population projections are most sensitive to resulting in population growth rates changing by
changes in calving interval and probability of –0.2 ± 0.1 and 0.2 ± 0.1 percentage points.
survival. Changes in calving interval produce rela-
tively large responses, with an increase from three DISCUSSION
to four years resulting in a reduction in average The high growth rates we have observed and
annual population growth rate (calculated over 22 projected, along with observations of van Jaars-
years) of 1.1 ± 0.1% (Table 5). The effect of a one- veld et al. (1999) and Slotow et al. (2005), indicate
year change in calving interval lessens at higher that South Africa’s small, confined, re-introduced
calving intervals. A greater than three year change elephant populations have been growing quickly,
in age at sexual maturity is required to yield a and will feasibly continue to increase in the short-
similar 1% change in annual growth rate (Table 6). and medium-term at rates previously thought
While an error or change in the population’s overall unrealistically high. Our projections predict that
annual mortality rate of 1% yields a 1% change in elephant populations can more than double in
40 South African Journal of Wildlife Research Vol. 36, No. 1, April 2006

Table 5. Sensitivity analysis for the average interval (yrs) at which females in the population have calves. Average
annual growth rates of eight elephant populations are shown for simulations in which calving interval was held
constant at one interval in the range of 3–6 years. The mean difference in growth rates for each calving interval from
4–6 years vs 3 years is presented in the bottom row. The bold values indicate the baseline projections for each
population.

Population Projected average annual growth rate for 2003–2025 (%)1

Calving interval (yrs)
3 4 5 6

Hluhluwe-iMfolozi 7.7 (7.6–7.7) 6.6 (6.5–6.7) 5.8 (5.6–6.0) 5.2 (5.0–5.4)

Mabula 7.6 (7.3–7.9) 6.6 (6.3–6.8) 5.8 (5.5–6.0) 5.3 (5.1–5.6)
Madikwe 7.9 (7.7–8.1) 6.9 (6.8–7.1) 6.2 (6.1–6.3) 5.6 (5.4–5.8)
Phinda 7.4 (7.4–7.5) 6.4 (6.3–6.5) 5.6 (5.4–5.7) 5.0 (4.8–5.2)
Pilanesberg 7.9 (7.9–7.9) 6.8 (6.7–6.9) 6.0 (5.8–6.2) 5.3 (5.0–5.5)
Pongola 8.9 (8.5–9.3) 7.6 (7.4–7.8) 6.7 (6.5–7.0) 6.0 (5.7–6.2)
St Lucia 8.0 (7.6–8.3) 6.8 (6.4–7.1) 6.0 (5.7–6.2) 5.4 (5.2–5.6)
Welgevonden 8.3 (8.2–8.4) 7.1 (7.0–7.1) 6.1 (6.0–6.2) 5.5 (5.4–5.6)
Difference from 3 yrs –– –1.1 ± 0.1 –1.9 ± 0.2 –2.6 ± 0.2
(mean ± S.D.)
1
Annual population growth rate and 95% confidence limits calculated from slope of ln(population size) vs time for duration of 2003–2025.

Reproductive females, which varied between 20 decrease/increase in the future as the age-
and 46% in the populations at their most recent structure changes and stabilizes. It follows then
census, are predicted to converge to 24 to 28% of that when examining the growth rate of a relatively
the population by the year 2025. These conver- new population in a reserve and attempting to esti-
gence values are within the range generally mate future relative changes in growth, one must
deemed normal for a stable age distribution (Calef take into consideration the age and sex structure
1988; Wittemyer 2001). This may be a useful tool of that population.
for managers interested in potential future rates of Sensitivity analyses indicate that our model
population growth. If the proportion of reproductive projections are fairly robust to reasonable levels
females is higher/lower than 24 to 28% in a popula- of change in the demographic parameters. The
tion, then growth rates should be expected to results also provide some insight into which demo-

Table 6. Sensitivity analysis for age at which female elephants reach sexual maturity. Average annual growth rates of
eight elephant populations are shown for simulations in which age at sexual maturity was held constant at one age in
the range of 8–13 years. The mean difference in growth rates for each age at sexual maturity from 9–13 years vs
8 years is presented in the bottom row. The bold values indicate the estimated age at sexual maturity for each
population.

Population Projected average annual growth rate for 2003–2025 (%)1

Age at sexual maturity (yrs)
8 9 10 11 12 13

Hluhluwe-iMfolozi 8.1 (8.0–8.1) 7.7 (7.6–7.7) 7.3 (7.2–7.4) 7.0 (6.9–7.1) 6.7 (6.6–6.8) 6.4 (6.3–6.6)
Mabula 9.0 (8.6–9.3) 8.5 (8.3–8.8) 8.2 (7.9–8.5) 7.9 (7.6–8.2) 7.6 (7.3–7.9) 7.3 (7.0–7.7)
Madikwe 6.9 (6.7–7.0) 6.6 (6.5–6.8) 6.4 (6.3–6.5) 6.2 (6.1–6.3) 6.0 (5.9–6.1) 5.8 (5.6–5.9)
Phinda 6.9 (6.9–7.0) 6.6 (6.5–6.7) 6.4 (6.3–6.5) 6.1 (6.0–6.2) 5.8 (5.8–5.9) 5.6 (5.5–5.7)
Pilanesberg 8.3 (8.3–8.3) 7.9 (7.9–7.9) 7.5 (7.5–7.6) 7.2 (7.2–7.3) 6.9 (6.8–7.0) 6.6 (6.6–6.7)
Pongola 8.9 (8.5–9.3) 8.4 (8.1–8.8) 8.1 (7.7–8.4) 7.7 (7.4–8.1) 7.4 (7.0–7.7) 7.1 (6.8–7.4)
St Lucia 7.3 (7.0–7.6) 7.1 (6.8–7.3) 6.8 (6.4–7.1) 6.6 (6.2–6.9) 6.4 (6.1–6.7) 6.2 (5.9–6.5)
Welgevonden 7.1 (7.0–7.1) 6.8 (6.7–6.8) 6.5 (6.4–6.6) 6.2 (6.1–6.3) 6.0 (5.9–6.1) 5.8 (5.6–5.9)
Difference from – –0.4 ± 0.1 –0.7 ± 0.1 –0.9 ± 0.2 –1.2 ± 0.2 –1.5 ± 0.3
8 yrs (mean ± S.D.)
1
Annual population growth rate and 95% confidence limits calculated from slope of ln(population size) vs time for duration of 2003–2025.
 Mackey et al.: Modelling elephant population growth in small South African reserves 41

Table 7. Sensitivity analyses for probability of survival. Average annual growth rates of eight elephant populations are
shown for simulations in which probability of survival was altered for the age classes indicated; survival probability was
held constant at 1.0 for remaining ages. Mean difference in growth rates for each probability of survival from 0.99 to
0.90 vs 1.0 (the baseline value, shown in bold) are also presented.

Population Probability Projected average annual growth rate for 2003–2025 (%)1
of survival
Ages (years)
0–60 0–2 3–5 6–9 10–60

Hluhluwe-iMfolozi 1.00 7.7 (7.6–7.7)

0.99 6.7 (6.6–6.7) 7.5 (7.5–7.6) 7.6 (7.5–7.6) 7.5 (7.4–7.6) 7.1 (7.1–7.2)
0.97 4.7 (4.6–4.7) 7.3 (7.2–7.3) 7.3 (7.2–7.4) 7.1 (7.1–7.2) 6.1 (6.0–6.1)
0.95 2.6 (2.5–2.6) 7.0 (6.9–7.1) 7.0 (7.0–7.1) 6.8 (6.7–6.9) 5.0 (5.0–5.1)
0.90 –2.8 (–2.9– –0.27) 6.3 (6.3–6.4) 6.4 (6.3–6.5) 5.9 (5.9–6.0) 2.6 (2.5–2.6)
Mabula 1.00 7.6 (7.3–7.9)
0.99 6.5 (6.2–6.8) 7.5 (7.1–7.8) 7.4 (7.1–7.8) 7.4 (7.1–7.8) 6.9 (6.6–7.3)
0.97 4.7 (4.3–5.0) 7.2 (6.9–7.5) 7.2 (6.9–7.6) 7.1 (6.8–7.5) 5.9 (5.6–6.2)
0.95 2.8 (2.4–3.2) 7.0 (6.7–7.3) 7.0 (6.7–7.3) 6.9 (6.6–7.2) 4.7 (4.3–5.0)
0.90 –1.8 (–2.2– –1.5) 6.4 (6.2–6.6) 6.5 (6.2–6.8) 6.2 (5.9–6.6) 2.3 (1.9–2.6)
Madikwe 1.00 6.2 (6.1–6.3)
0.99 5.2 (5.1–5.3) 6.1 (6.0–6.2) 6.1 (6.0–6.2) 6.0 (5.9–6.2) 5.5 (5.4–5.7)
0.97 3.1 (3.0–3.3) 5.9 (5.8–6.0) 5.8 (5.7–6.0) 5.7 (5.6–5.8) 4.3 (4.2–4.4)
0.95 1.1 (0.9–1.2) 5.6 (5.5–5.8) 5.6 (5.5–5.7) 5.4 (5.3–5.6) 3.1 (3.0–3.2)
0.90 –4.4 (–4.5– –4.3) 5.1 (5.0–5.2) 5.1 (5.0–5.2) 4.7 (4.6–4.9) 0.2 (0.1–0.4)
Phinda 1.00 6.4 (6.3–6.5)
0.99 5.4 (5.3–5.5) 6.2 (6.1–6.3) 6.2 (6.1–6.3) 6.2 (6.1–6.3) 5.7 (5.6–5.8)
0.97 3.3 (3.2–3.4) 6.0 (5.9–6.1) 6.0 (5.9–6.1) 5.9 (5.8–6.0) 4.6 (4.5–4.6)
0.95 1.3 (1.2–1.4) 5.8 (5.7–5.9) 5.8 (5.7–5.9) 5.6 (5.5–5.7) 3.4 (3.4–3.5)
0.90 –3.8 (–3.9– –3.7) 5.2 (5.1–5.3) 5.2 (5.1–5.3) 4.9 (4.8–5.0) 0.8 (0.7–0.9)
Pilanesberg 1.00 7.9 (7.9–7.9)
0.99 6.9 (6.9–6.9) 7.8 (7.7–7.8) 7.8 (7.7–7.8) 7.7 (7.7–7.7) 7.4 (7.3–7.4)
0.97 4.9 (4.8–4.9) 7.5 (7.4–7.5) 7.5 (7.5–7.5) 7.3 (7.3–7.4) 6.3 (6.2–6.3)
0.95 2.7 (2.7–2.7) 7.2 (7.2–7.2) 7.2 (7.2–7.3) 7.0 (6.9–7.0) 5.3 (5.2–5.3)
0.90 –2.4 (–2.5– –2.4) 6.5 (6.5–6.6) 6.5 (6.5–6.6) 6.1 (6.1–6.2) 2.9 (2.8–2.9)
Pongola 1.00 8.9 (8.5–9.3)
0.99 7.9 (7.5–8.3) 8.8 (8.4–9.2) 8.7 (8.4–9.1) 8.7 (8.3–9.1) 8.4 (8.1–8.8)
0.97 5.8 (5.4–6.2) 8.5 (8.1–8.9) 8.4 (8.0–8.8) 8.3 (7.9–8.7) 7.5 (7.2–7.9)
0.95 3.8 (3.4–4.2) 8.2 (7.8–8.6) 8.1 (7.7–8.5) 7.9 (7.5–8.3) 6.6 (6.2–7.0)
0.90 –1.6 (–2.0– –1.2) 7.4 (7.0–7.8) 7.3 (6.9–7.7) 7.0 (6.6–7.4) 4.5 (4.2–4.9)
St Lucia 1.00 6.8 (6.4–7.1)
0.99 5.8 (5.5–6.1) 6.6 (6.3–7.0) 6.7 (6.3–7.0) 6.6 (6.3–6.9) 6.1 (5.8–6.5)
0.97 3.7 (3.4–4.1) 6.4 (6.1–6.7) 6.4 (6.1–6.7) 6.4 (6.0–6.7) 4.9 (4.6–5.3)
0.95 1.7 (1.3–2.0) 6.2 (5.9–6.5) 6.2 (5.9–6.5) 6.0 (5.7–6.4) 3.8 (3.5–4.1)
0.90 –3.4 (–3.8– –3.1) 5.6 (5.3–5.8) 5.6 (5.3–6.0) 5.4 (5.0–5.8) 1.1 (0.7–1.4)
Welgevonden 1.00 7.1 (7.0–7.1)
0.99 6.0 (6.0–6.1) 6.9 (6.9–7.0) 6.9 (6.8–7.0) 6.9 (6.8–6.9) 6.5 (6.4–6.6)
0.97 3.9 (3.9–4.0) 6.6 (6.6–6.7) 6.6 (6.6–6.7) 6.6 (6.4–6.5) 5.5 (5.4–5.5)
0.95 1.9 (1.8–1.9) 6.3 (6.3–6.4) 6.3 (6.3–6.4) 6.2 (6.1–6.2) 4.4 (4.3–4.5)
0.90 –3.5 (–3.6– –3.4) 5.7 (5.6–5.8) 5.7 (5.6–5.7) 5.2 (5.1–5.3) 2.3 (2.2–2.3)

Difference from 1.00 0.99 –1.0 ± 0.0 –0.1 ±0.0 –0.1 ± 0.0 –0.2 ± 0.1 –0.6 ± 0.1
(mean ± S.D.) 0.97 –3.0 ± 0.1 –0.4 ± 0.0 –0.4 ± 0.0 –0.5 ± 0.1 –1.7 ± 0.2
0.95 –5.1 ± 0.1 –0.7 ± 0.1 –0.6 ± 0.1 –0.8 ± 0.1 –2.8 ± 0.2
0.90 –10.3 ± 0.3 –1.3 ± 0.1 –1.3 ± 0.1 –1.6 ± 0.2 –5.2 ± 0.5
1
Annual population growth rate and 95% confidence limits calculated from slope of ln(population size) vs time for duration of 2003–2025.
42 South African Journal of Wildlife Research Vol. 36, No. 1, April 2006

graphic parameters may be most important for thought realistic. In good conditions, elephant
determining rates of population growth. Projected populations may double in as little as 10 years, and
population sizes were most sensitive to changes in many reserves may be faced with a great surplus
calving interval and survival probabilities. A delay of elephants. These results have application to the
in the onset of female sexual maturity affects other populations, such as some of those in East
population growth to a lesser extent. Changes in Africa that are recovering from poaching (e.g.
age-specific probabilities of conception and the Laikipia and Samburu, which increased at aver-
sex ratio of newborns have little effect on elephant age annual rates of 4.9% and 10.3%, respectively,
populations. Hanks & McIntosh (1973) also con- between 1996 and 2002; Blanc et al. 2005). They
cluded that calving interval is more important for will also have similar population structure and low
regulating population growth and size than is the densities, and so should experience similar high
age at sexual maturity. The insight gained about growth rates. Management policies need to be
the relative importance of the different demo- devised and implemented in the very near future to
graphic parameters, and the populations’ respon- prevent over-population of reserves (particularly
ses to changes in these parameters, need to be those that are small and confined) by elephants
considered when devising elephant population and subsequent potential long-term damage to the
management strategies. habitat and loss of biological diversity in
Our model incorporates no density-dependence. the reserves. Further work needs to be done to
Laws (1969) suggested that age at first reproduc- examine and compare different management
tion and calving interval are prolonged in popula- strategies for controlling elephant populations in
tions with higher densities. However, our observa- the near future and for determining reserve-
tion of continued exponential increase in elephant specific estimates of elephant carrying capacity.
populations that already meet, exceed, or are
quickly approaching carrying capacity or equilib- ACKNOWLEDGEMENTS
rium density (as estimated by Fowler & Smith Robin Mackey was supported by National
1973; Cumming et al. 1997, and van Aarde et al. Research Foundation grants to Kevin Duffy and
1999) suggests that density-dependent elephant Rob Slotow, and from support from Distell (Pty)
population growth is not occurring in the popula- Ltd. The following people provided data on elephant
tions we studied. van Jaarsveld et al. (1999) populations (all UKZN staff or students unless
observed a general lack of density-dependence in otherwise indicated): Hluhluwe-iMfolozi (Luca
their study of several South African elephant popu- Turelli), Mabula (Yolanda Pretorius), Pongola
lations that were growing at annual rates of up (Graeme Shannon), Phinda (Heleen Druce),
15%. Additionally, recent studies indicate that the Pilanesberg (Gus van Dyk (NWP&TB)), Magriet
growth rate of the elephant population at Addo van Niekerk and Tarryne Burke), St Lucia (Chantal
Elephant National Park is not declining (K. Gough, Dickson), and Welgevonden (Johan Malan,
pers. comm.) despite an exceedingly high popula- SANPARKS). We thank Sam Ferreira, Caroline
2
tion density of >2.5 elephants/km (Slotow et al. Reid, Graeme Shannon, and Norman Owen-Smith
2005). Thus, although some of the demographic for comments that improved this manuscript.
parameters (e.g. mean calving interval) may
change over the long-term, it is unlikely that any REFERENCES
ARMBRUSTER, P. & LANDE, R. 1993. A population
substantial density-dependent change would
viability analysis for African elephant (Loxodonta
occur during the timescale of our projections africana): how big should reserves be? Cons. Biol. 7:
and/or significantly alter our model predictions. 602–610.
Our model provides projections based on BEGON, M. & MORTIMER M. 1986. Population ecology:
current and recent historical conditions and data. A unified study of animals and plants, 2nd edn.
Blackwell Scientific Publications, Oxford.
Severe changes in such conditions (e.g. multiple- BLANC, J.J., BARNES, R.F.W., CRAIG, G.C.,
year droughts) during the projection time scale DOUGLAS-HAMILTON,I., DUBLIN, H.T., HART, J.A.
may result in population sizes and growth rates & THOULESS, C.R. 2005. Changes in elephant num-
different from those we predict. bers in major savanna populations in eastern and
southern Africa. Pachyderm 38: 19–28.
CALEF, G.W. 1988. Maximum rate of increase in the
CONCLUSION African elephant. Afr. J. Ecol. 26: 323–327.
Elephant populations in small reserves in South CUMMING, D.H.M., FENTON, M.B., RAUTENBACH,
Africa are growing at rates higher than previously I.L., TAYLOR, R.D., CUMMING, G.S., CUMMING,
 Mackey et al.: Modelling elephant population growth in small South African reserves 43

M.S., DUNLOP, J.M., FORD, A.G., HAVORKA, MOSS, C.J. 2001. The demography of an African
M.D., JOHNSTON, D.S., KALCOUNIS, M., elephant (Loxodonta Africana) population in Ambo-
MAHLANGU, Z. & PORTFORS, C.V.R. 1997. Ele- seli, Kenya. J. Zool. Lond. 255: 145–156.
phants, woodlands and biodiversity in southern OWEN-SMITH. R.N. 1988. Megaherbivores: The
Africa. S. Afr. J. Sci. 93: 231–236. influence of very large body size on ecology.
DOMINY, N.J., FERGUSON, N.S.& MADDOCK, A..1998. Cambridge Studies in Ecology. Cambridge University
Modelling elephant (Loxodonta africana africana) Press, Cambridge.
population growth in Hluhluwe-Umfolozi Park to PRETORIUS, Y. 2004. Stress in the African elephant on
predict and manage limits. S. Afr. J. Wildl. Res. 28: Mabula Game Reserve, South Africa. M.Sc. thesis,
61–67. University of KwaZulu-Natal, Durban.
FOWLER, C.W. & SMITH, T. 1973. Characterizing stable SLOTOW, R., GARAI, M.E., REILLY, B., PAGE, B. &
populations: an application to African elephant popu- CARR, R.D. 2005. Population ecology of elephants
lations. J. Wildl. Mgmt. 37: 513–524. re-introduced to small fenced reserves in South
FRITZ, H. & DUNCAN, P. 1994. On the carrying capacity Africa. S. Afr. J. Wildl. Res. 35: 23–32.
for large ungulates of African savanna ecosystems. VAN AARDE, R., WHYTE, I. & PIMM, S. 1999. Culling
Proc. R. Soc. Lond. B 256: 77–82. and the dynamics of the Kruger National Park African
GARAI, M.E., SLOTOW, R., CARR, R.D. & REILLY, B. elephant population. Anim. Conserv. 2: 287–294.
2004. Elephant reintroductions to small fenced VAN JAARSVELD, A.S., NICHOLLS, A.O. & KNIGHT,
reserves in South Africa. Pachyderm 37: 28–36. M.H. 1999. Modelling and assessment of South
HANKS, J. & McINTOSH, J.E.A.. 1973. Population African elephant Loxodonta africana population
dynamics of the African elephant ( Loxodonta persistence. Env. Mod. Assess. 4: 155–163.
africana). J. Zool., Lond. 169: 29–38. WAITHAKA, J. 1997. Management of elephant popula-
JOHNSON, C.F., COWLING, R.M. & PHILLIPSON, tions in Kenya – What have we learnt so far? Pachy-
P.B.1999. The flora of the Addo Elephant National derm 24: 33–36.
Park, South Africa: are threatened species vulnera- WHYTE, I., VAN AARDE, R. & PIMM, S.L. 1998.
ble to elephant damage? Biodiv. Cons. 8: 1447–1456 Managing elephants of Kruger National Park. Anim.
LAWS, R.M. 1969. The Tsavo research project. Conserv. 1: 77–83.
J. Reprod. Fertil. Suppl. 6: 495–531. WISEMAN, R., PAGE, B.R. & O’CONNOR, T.G. 2004.
LAWS, R.M., PARKER,I.S.C. & JOHNSTONE, R.C.B. Woody vegetation change in response to browsing in
1975. Elephants and their habitats. The ecology of Ithala Game Reserve, South Africa. S. Afr. J. Wild.
elephants in north Bunyoro, Uganda. Oxford Univer- Res. 34: 25–37.
sity Press, Oxford. WITTEMYER, G. 2001. The elephant population of
LOMBARD, A.T., JOHNSON, C.F., COWLING,R.M. & Samburu and Buffalo Springs National Reserves,
PRESSEY, R.L. 2001. Protecting plants from Kenya. Afr. J. Ecol. 39: 357–365.
elephants: botanical reserve scenarios within the WU, L.S-Y. & BOTKIN, D.B. 1980. Of elephants and men:
Addo Elephant National Park, South Africa. Biol. a discrete, stochastic model for long-lived species
Cons. 102: 191–203. with complex life histories. Am. Nat. 116: 831–849.

Corresponding editor: G.I. Kerley