ArticlePDF Available

Modelling Impacts of Poaching on the Sumatran Elephant Population in Way Kambas National Park, Sumatra, Indonesia

Authors:
  • WWF Indonesia, Jakarta
  • Zoological Society of London

Figures

Content may be subject to copyright.
31
Gajah 28 (2008) 31-40
Modelling Impacts of Poaching on the Sumatran Elephant Population
in Way Kambas National Park, Sumatra, Indonesia
Arnold F. Sitompul1,2, John P. Carroll3, James Peterson4 and Simon Hedges5
1Conservation Science Initiative, Medan, Indonesia
2Department of Natural Resources Conservation, University of Massachusetts, Amherst, MA, USA
3Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA, USA
4U.S.G.S. Georgia Cooperative Fish and Wildlife Research Unit, Warnell School of Forestry
and Natural Resources, University of Georgia, Athens, GA, USA
5Wildlife Conservation Society, International Programs, Bronx, New York, NY, USA
Introduction
Poaching has been known to have a large impact
on elephant populations in both Africa (e.g.
Douglas-Hamilton 1987; Poole & Thomsen
1989) and Asia (Sukumar 1989; Sukumar et al.
1998). There are fears that poaching of Asian
elephants has increased since CITES approved
an experimental one-off sale of ivory from
Botswana, Namibia, and Zimbabwe to Japan
in July 1999, following compliance with a
number of agreed conditions. Another one-off
sale from South Africa, Namibia, and Botswana
was approved in 2002 but that sale has not yet
taken place (CITES 2000; Milliken 2004). In
Sumatra, during the 1980s and 1990s, poaching
was not considered a major threat to elephants
(Blouch & Haryanto 1984; Blouch & Simbolon
1985; Santiapillai & Jackson 1990); however it is
feared that poaching activity has increased since
year 2000 (Sitompul et al. 2002; Hedges et al.
2005). While poaching activity is predicted to
continue increasing, accurate data on poaching is
very dif cult to obtain. Furthermore, there have
been no eld studies in Sumatra identifying the
impact of poaching on elephant abundance and
population trends.
Population modelling has been widely used in
wildlife ecology studies for many terrestrial large
mammals (e.g. Belovsky 1987; Berger 1990;
Rothley et al. 2005). Incorporating modelling
approaches as part of adaptive management
strategies, allows managers to develop more
effective conservation strategies (Cromsigt
et al. 2002) while reducing the uncertainty
about how the system responds to management
actions (Williams et al. 2002). Furthermore,
modelling allows managers to make an empirical
assessment of the species of interest and identify
and implement the management strategies that
are most likely to increase the probability of a
species persisting over a given time period.
However, developing detailed and accurate
population models for many species requires
extensive historical baseline data (i.e., population
size, age structure, sex-ratio, fecundity rate, and
natural survival and mortality rates). In Sumatra,
reliable baseline data for Sumatran elephant is
uncommon; however the results of a couple of
studies (Riley 2002; Hedges et al. 2005) provide
reliable data for the elephant population in Way
Kambas National Park. We believe that modelling
of elephant populations and poaching threats will
help managers identify key parameters to monitor,
and strategies to adopt, in order to minimize
extinction threats for Sumatran elephants.
In this paper, we estimate the potential impact
of poaching on the elephant population in Way
Kambas National Park (WKNP) using a stochastic
population model. We projected the population
trend under three different poaching scenarios:
no poaching, low poaching, and high poaching.
For each model, we predicted the population’s
age distribution, growth rate, and trends in
abundance estimates over 50 years. Finally, we
calculated the extinction probability for each
scenario and conducted sensitivity analyses to
identify the parameter that had the largest effect
on the model’s estimates.
32
Methods
Study area
Field data used in the model were collected in
Way Kambas National Park (WKNP), Sumatra,
Indonesia. WKNP is located in eastern part of
Lampung Province in south-eastern Sumatra
(4o62’–5o26’ S and 105o54’–105o90’ E), and
is 1235 km2 in area The entire park is < 50 m
above sea level and annual rainfall is 2000–
3000 mm. Vegetation types are typical tropical
lowland and swamp forest. Most of the park was
logged in the 1960s and 1970s, so most of the
forested area in the park is relatively degraded.
Nonetheless, the park has still been categorized
as the second highest priority for Sumatran
elephant conservation (Santiapillai & Jackson
1990). The park boundary is approximately 227
km long and 65% (148 km) of it is bordered by
34 villages. The elephant population in the park
was estimated to be 180 (95% CI = [144, 225])
in 2002 (Hedges et al. 2005). The government
of Indonesia established an Elephant Training
Centre (ETC) in the south-eastern area of the
park in the early 1980s; the purpose of this ETC
was to house “problem elephants” captured as
a result of human–elephant con ict and habitat
conversion in WKNP and other parts of Lampung
Province (Hedges et al. 2005). The “problem
elephants” were then tamed and trained at the
ETC for tourism purposes. The ETC in WKNP
is the largest such centre in Sumatra and during
2000–2002 was known to contain about 100
elephants (authors’ pers. obs.).
Methods
We developed a stage-based stochastic population
model to determine the impact of poaching in
the park based on known rates of illegal killing
of elephants in WKNP (Sitompul et al. 2002).
Population trajectories and maximum population
size under different scenarios were predicted
for elephants in WKNP using a Leslie matrix
projection model (Leslie 1945, 1948). The model
consisted of four different life-history stages:
calf, juvenile, subadult, and adult and operated
on an annual time step basis (Fig. 1). The calf
stage included any elephant <1 year old, juveniles
included ages 1–5 years, subadult elephants
included individuals >5–15 years old, and adults
included individuals >15 years old (Sukumar
1989). Each simulation began by assigning
individuals to one of the four life history stages:
calves were 8.04% of the population, juveniles
were 28.57%, subadults 50%, and adults 13.39%,
based on the demographic con guration of the
elephant population in WKNP in Reilly (2002).
The number of calves produced each time step
was a function of the number of adults and sub-
adults and fecundity. Stage-speci c maximum
annual fecundity rate was assumed to be constant
over time and estimated to be 0.225 for both
subadult and adult elephants, and was based on
long-term studies of Asian elephants in other
regions (Sukumar 1989). Stage-speci c natural
survival rate was assumed to be similar to
Asian elephants in India and averaged 0.85 for
the calf, 0.96 for the juvenile, 0.98 for the sub-
adult, and 0.85 for the adult life history stages.
We incorporated stochasticity into the model by
randomly generating annual survival rates from
a beta distribution with the mean speci ed above
and a standard deviation that was 10% of the
mean.
Figure 1. Model ow for population estimation
and demographics as a function of recruitment,
survival and poaching for elephants projected for
50 years in Way Kambas National Park.
33
For each simulation scenario, we ran 1000
replicate simulations for a 50 year time period,
and observed the nal population structure at
year 50. Mean and 95% con dence interval (95%
CI) of population size, population structure, and
population growth rate (λ) were calculated. In
addition, a quasi-extinction coef cient (EC) was
estimated as the proportion of the 1000 replicate
simulations that resulted in extinction before 50
years.
We evaluated the effect of poaching on elephant
populations using three different scenarios.
The rst scenario, which we called the control,
assumed that the elephant population in the park
was fully protected, resulting in no anthropogenic
removal of elephants (no poaching and elephant
capture due to con ict with human). The second
scenario assumed poaching occurred at a low rate
de ned as the mean number of elephants known
to have been removed from the population per
year due to poaching over the years 2000–2004.
The number of elephants poached in the park
was estimated from the total number of carcasses
with signs of poaching activity found in the
park in the 2000–2002 period (n=8 elephants)
plus 8 elephants that had been found killed by
poachers in the 2003–2004 period (Sitompul et
al. 2002; Hedges et al. 2005; WCS unpub. data).
We assumed only sub-adult and adult elephants
were poached. The third scenario assumed that
high poaching would occur in the park based on
continued human population growth and land use
trends in Lampung Province. High poaching was
de ned as a 2x increase on the previously de ned
low poaching rate described above. Because the
relationship between poaching and population
size is unknown, we modelled poaching rates as a
function of population size using four alternative
functions: (1) poaching was constant over time;
(2) poaching was a negative linear function of
population size; (3) poaching was an exponential
decay function of population size; and (4)
poaching was a logistic function of population
size. For the high poaching rate scenario,
poaching functions were kept the same as in the
low poaching rate scenario. For each poaching
function, the number of sub-adult and adult
elephants poached from the park was randomly
assigned using a Poisson distribution and the
scenario-speci c rate. Thus, the rate of poaching
per year, in the model, was assumed to be additive
to the stage-speci c natural mortality. We did not
include sex-speci c differences in poaching rate
because there was no information on such sex-
speci c differences for WKNP. There is evidence
that adult female elephants are also poached in
Sumatra and their toenails, genitalia, and other
body parts are collected for use in traditional
medicines (Sitompul et al. 2002).
Several other assumptions were required in
constructing the models. Natural mortality
rates used were derived from data on Indian
elephants, which might be different than
Sumatran elephants. However, it is unlikely that
they would be substantially different because
elephants in India and Sumatra have similar
life histories. Furthermore, we did not include a
carrying capacity function because the carrying
capacity of the study area is not well studied
(but is thought to be much higher than the
present population size) and because our primary
concern was preventing declining populations
and local extinction, the effect of density-
dependent factors as the population approached
carrying capacity was considered unimportant.
However, model scenarios projecting increases
in population will need re nement and some
measure of carrying capacity should be included
as those data become available. Finally, potential
genetic problems associated with small isolated
elephant populations (e.g. inbreeding depression)
were not included in our model.
Sensitivity analyses
The purpose of the sensitivity analyses was
to determine the relative in uence of each
parameter and alternative poaching model on
model estimates (Williams et al. 2002). Relative
sensitivity of model estimates can be evaluated by
varying model input parameters over a speci ed
range and examining the change in model
outputs. For this study, we evaluated the relative
sensitivity of the year 50 model estimates to each
parameter by calculating a Sensitivity Index (SI)
using regression analysis to calculate the slope
and uncertainty of each poaching function and
then multiplying the slope and uncertainty of
34
the parameter to calculate the SI following the
methods of Wiegand et al. (1998). We evaluated
the sensitivity of reproductive parameters of
sub-adult and adult elephants by varying the
reproductive rates from 0.19 to 0.25, with 0.01
increments. We also evaluated model sensitivity
to the survival rate parameter for the calf to
juvenile transition and the sub-adult to adult
transition by varying the survival parameter for
each life history stage from 0.75 to 0.90, with
0.05 increments. To understand the sensitivity of
the population model to the alternative poaching
functions, we varied poaching rate from the low
poaching scenario’s 50% to 200% of the estimate
values in 10% increments. The results of these
sensitivity analyses for the high poaching rate
scenario will be identical to the low poaching
rate scenario since the difference between
the low and high poaching rate scenarios is
simply the magnitude of the poaching rate. All
simulation modelling and sensitivity analyses
were conducted using SAS (SAS version 8.2).
Results
Projection of the WKNP elephant population
over a 50-year period showed the population
increasing from 180 elephants to 594 elephants
(95% CI = [570, 618]) if we assumed that
poaching stopped. The extinction coef cient for
the control population was 0.0 and population
growth rate (λ) was 1.02 (0.0001 SE). Under the
low poaching rate scenarios we also showed that
the elephant population would increase (Fig. 2).
The linear poaching function produced an elephant
population in year 50 of 422 (95% CI = [403,
441]). The extinction coef cient using the linear
function was also 0.0 and λ was 1.02 (0.0002 SE).
If poaching in the park behaves as an exponential
extinction function, the elephant population in
year 50 was estimated to be 325 (95% CI = [308,
342]). The extinction coef cient for this function
was 0.009 and λ was 1.01 (0.0002 SE). The
constant and logistic poaching functions in the
model produced estimates of elephant population
size of 253 (95% CI = [235, 271]) and 263 (95%
CI = [245, 281]), respectively. The extinction
coef cient with constant poaching was 0.099,
and logistic poaching resulted in an estimate of
0.086. The population growth rate with constant
poaching was 1.0 (0.0005 SE) and λ with logistic
poaching was 1.0 (0.0005 SE; Table 1). The age
distribution after 50 years for the control and low
poaching rate scenarios changed slightly from
one dominated by sub-adults towards one more
dominated by adults (Fig. 3).
Population models with high poaching rate
scenarios showed a different trend to the low
poaching rate scenarios over the 50-year period.
In the high poaching rate scenarios, only linear
and exponential decay poaching patterns showed
that the elephant population in WKNP would
increase over the 50 years (Fig. 4). Population
size in year 50 for the linear and exponential
decay poaching functions was estimated to
be 274 (95% CI = [263, 285]) and 217 (95%
CI = [211, 226]), respectively. The extinction
coef cient for the linear and exponential
poaching functions was 0.0 and λ was 1.0
(0.0002 SE). For the exponential decay poaching
function, the extinction coef cient was 0.01 and
λ was 1.0 (0.0003, SE). In contrast, the constant
poaching and logistic poaching functions in the
high poaching scenarios showed that elephant
population in WKNP would decline dramatically
(Fig. 4). Final population size in year 50 for the
constant and logistic poaching functions was
41 (95% CI = [33, 49]) and 37 (95% CI = [30,
44]), respectively. The extinction coef cient
for constant poaching was 0.75 and for logistic
poaching it was 0.76. The population growth rate
was 0.97 (0.008 SE) for constant poaching and
0.97 (0.009 SE) for logistic poaching (Table 1).
The age distribution in the high poaching rate
scenarios showed similar patterns to the low
poaching scenarios, with more adult individuals
found at the end of each simulation (Fig. 5).
Sensitivity analyses
Sensitivity analyses for each natural parameter
revealed high levels of variation in the model. The
result of the sensitivity analyses for the sub-adult
and adult reproductive parameters showed that
small changes in the adult reproductive parameter
caused large changes in the nal population size.
For example, an increase of 6% in the adult
reproduction rate could cause a 76.01% change in
nal population size. In contrast, a 6% change in
35
0
100
200
300
400
500
600
700
0 5 10 15 20 25 30 35 40 45 50 55
Time
Control
Constant
Exponential
Linear
Logistic
Figure 2. Simulated population trends of Asian elephants for 50 years
under control and low poaching scenarios in Way Kambas National
Park,. Density dependent effects using low poaching level scenarios were
developed (constant, exponential, linear and logistic).
0
50
100
150
200
250
300
350
400
Calf
Juvenile
Subadult
Adult
Figure 3. Projection of the age structure of the elephant population in
Way Kambas National Park after 50 years of simulation, presented in the
current population (start) and in control and low poaching scenarios.
0
100
200
300
400
500
600
700
0 5 10 15 20 25 30 35 40 45 50 55
Time
Control
Constant
Exponential
Linear
Logistic
Figure 4. Simulated population trends over 50 years period under control
and high poaching scenarios in Way Kambas National Park. Density de-
pendent effects using high poaching level scenarios were developed (con-
stant, exponential, linear and logistic).
0
50
100
150
200
250
300
350
400
Start Control Constant Linear Exponential Logistic
Calf
Juvenile
Subadult
Adult
Figure 5. Projection of the age structure of the elephant population in Way
Kambas after 50 years of simulation, presented in the current population
(start) and in control and high poaching scenarios.
36
sub-adult reproduction rate only caused a 26.84%
change in nal population size (Fig. 6). For the
survival parameter, sensitivity analyses showed
that juvenile survival and young survival rates had
relatively similar impact on the nal population
size. An increase of 5% in survival of young and
juvenile elephants independently caused a change
of 29.25% and 29.87% in nal population size,
respectively (Fig. 7). However, the adult survival
parameter had a far more sensitive effect on the
nal population size compared to the sub-adult
survival parameter. Changing the adult survival
parameter 5% could cause an 86.54% change in
nal population size. In contrast, a 5% change
in the sub-adult parameter only caused a 37.46%
change in nal population size (Fig. 8).
Sensitivity analysis for the four poaching function
parameters showed clear differences in model
sensitivity (Fig. 9, Table 2). The logistic poaching
function appeared to have the greatest in uence,
which is shown by it having the lowest index
(SI = -2.626) followed by the constant poaching
function (SI = - 0.013). The linear, constant, and
exponential poaching functions appeared to have
relatively similar sensitivity in the model (Fig. 9).
The level of uncertainty of poaching parameter in
the model showed that the exponential parameter
had the lowest uncertainty compared to the other
three poaching parameters (Table 2).
Discussion
Our model clearly demonstrates that in the control
(no poaching) scenarios the elephant population
in the park will increase over time. Furthermore,
the low poaching rate scenarios also show the
elephant population increasing. These results
imply that the low poaching rates observed in the
past did not have a serious negative impact on the
elephant population in the park. The population
growth rate in the low poaching rate scenarios
remained about 1.0 or above and extinction
encounter rate after 1000 simulations was less
than 0.1. However, if we doubled the poaching rate
from the minimum known rate observed in 2000–
2002, as in the high poaching scenarios, we found
that the population could decline dramatically
for the logistic poaching and constant poaching
functions, with the extinction coef cients for
both functions increasing signi cantly up to
about 75%. For both the constant and logistic
poaching functions, the magnitude of poaching
pushed the population into negative growth rates.
In contrast, the linear and exponential poaching
functions did not differ much from the lower
poaching scenarios. In this situation, poaching
(linear and exponential functions) seemed to
have little effect on the population even though
the magnitude of the poaching increased two fold
from the low poaching scenarios. It is clear from
these results that further study of the WKNP
population, and other Asian elephant populations,
is necessary in order to decide which poaching
function best describes reality and therefore
allow us to better model population trajectories
under different scenarios.
The age distribution in the model showed that
the proportional representation of the different
age stages in the population shifted towards the
adult age stage for the low and high poaching rate
scenarios. The overall pattern of age distribution
for both poaching scenarios was the same, with
Table 1. Summary of model result representing nal population size; population growth rate and
extinction encounter using all possible scenarios in the model. ƒ = poaching function of population
size. N50= population at year 50; λ = population growth ate; EC= Extinction Coef cient.
Scenarios ƒ N50 95%CL λ95%CL EC
Control 594 23.59 1.02 0.0002 0
Low-poaching constant 253 17.87 1.01 0.001 0.099
linear 422 19.03 1.02 0.0004 0
exponential 325 16.63 1.01 0.0006 0.009
logistic 263 17.80 1.00 0.0009 0.086
High-poaching constant 41 7.86 0.97 0.016 0.75
linear 274 11.08 1.00 0.0005 0
exponential 217 9.40 1.00 0.0007 0.01
logistic 37 7.09 0.97 0.018 0.76
37
Figure 6. Response on predicted elephant population size in 50 years
simulation for various combinations of adult reproduction rate (y-axis) and
sub adult reproduction rate (x-axis). Line in different color represents el-
ephant population size for speci c adult and sub adult reproduction rate.
Figure 7. Response on predicted elephant population size in 50 years
simulation for various combinations of juvenile survival rate (y-axis) and
calf survival rate (x-axis). Line in different color represents elephant popu-
lation size for speci c juvenile and calf survival rate.
Figure 8. Response on predicted elephant population size in 50 years
simulation for various combinations of adult survival rate (y-axis) and sub
adult survival rate (x-axis). Line in different colour represents elephant
population size for speci c adult and subadult survival.
0
0.5
1
1.5
2
2.5
3
0 0.5 1 1.5 2 2.5
Rate of change
Constant
Exponential
Linear
Logistic
Figure 9. Response of population size in 50 years simulation to the rate of
change on the poaching function parameter performed in the model. Dif-
ferent colour line represents different poaching function in the model.
38
the highest proportion of the population formed by
the adult stage followed by the calf, juvenile, and
sub-adult stages. If we examine the relationship
between population growth and age structure after
simulation, we nd that for the low poaching rate
scenarios the population is predicted to grow after
50 years. A similar pattern was also found for the
exponential and linear poaching functions in the
high poaching rate scenario. If the population is
growing, that means the population growth rate
is equal to or more than one. In this situation we
would expect the age distribution at the end of
simulation year to be dominated by the younger
age classes. However, our models did not predict
this, suggesting that improved survival of sub-
adult and adult elephants in the population over
a relatively short projection period (50 years)
relative to an elephant’s lifespan provided our
populations with much greater numbers of older
individuals. As a result, there was not enough new
recruitment to shift the age distribution towards
the younger age classes.
Sensitivity analyses
Our sensitivity analyses showed that variation
in reproduction parameters for adults had the
greatest impact on model variability. Relatively
small changes in adult reproduction rate could
cause a signi cant impact on nal population
size. Therefore, reproduction rate of adult
elephants needs to be determined accurately if
models such as ours are to be useful management
tools and to allow the demographic condition
of populations of interest to be assessed. If we
assumed reproduction rate in the population to
be deterministic, and compared the sensitivity of
the survival rate, we found the model was more
sensitive to the adult survival parameter compared
to the subadult survival parameter. Sukumar
(1989) suggested that among adult elephants,
female survival rate had a more signi cant effect
on the population than did male survival rate. His
study suggested that if adult male elephants have
low survival, the population could still grow if
female survival rate was high. Similar results
have also been demonstrated for other long-lived
species such as grizzly bears in Yellowstone
National Park (Eberhardt et al. 1994).
Sensitivity analyses for the poaching parameter
revealed a clear sensitivity to poaching function in
the model and this was re ected in the sensitivity
index value for the parameter. Sensitivity analyses
showed the logistic poaching function was the
most sensitive poaching function. This is most
likely because the number of elephants poached
per year was maintained at the maximum level and
at the same time randomization was incorporated
into the function. Clear differences can be found
if we compare the sensitivity of the logistic to the
constant poaching function: the constant poaching
function tended to be less sensitive, even though
the number of elephants poached per year was
maintained at the maximum level, presumably
because no randomization was incorporated into
this poaching function.
Management implications
Our model suggests that the elephant population
in WKNP will not decline over the next 50
years provided poaching rates remain at the low
level observed in 2000–2002. While this result
is encouraging, there is a possibility that the
2000–2002 poaching rate data used in this study
underestimated real poaching rates in the park at
that time because they were based on the number
of elephant remains found without dedicated
carcass searches. There is, therefore, a possibility
that the number of elephants killed because of
poaching was higher than our estimate, and our
models suggest this if this were so the increased
poaching could push the population toward
negative growth. Moreover, even if the 2000–
2002 data were representative of actual poaching
rates at that time an evidence-based adaptive
management approach to protecting the park’s
Table 2. Sensitivity analysis of the poaching
parameter. Poaching was speci ed as function
of population size. β0= parameter value;
α(β,β0)= slope; Δ(β)= approximate uncertainty
in the parameter; SI(β,β0)= sensitivity index of
parameter β within point β0.
Poaching β0α(β,β0)Δ(β) SI(β,β0)
Constant 2.848 -0.258 0.049 -0.013
Exponent. 2.630 -0.105 0.012 -0.001
Linear 2.780 -0.161 0.000 0.000
Logistic 4.050 -1.802 1.457 -2.626
39
elephants would require monitoring of poaching
rates to determine, for example, whether law
enforcement targets were being achieved.
Therefore a poaching monitoring program
(e.g. systematic carcasses searching) should be
established as a priority for management of the
park’s elephant population. This could perhaps
involve the use of detection dogs (sniffer dogs) to
improve carcass detection ef ciency, as elephant
carcasses are surprisingly dif cult to nd in
forested environments. In addition to improving
detection rates, the limited number of arrests in
relation to elephant poaching and the existence
of local ivory markets clearly also need to be
addressed (Hedges et al. 2005). Interestingly,
reducing poaching could also reduce human–
elephant con ict around WKNP because
research in Africa has shown that poachers
hunting elephants in forests can drive them into
closer proximity to surrounding farmland thus
increasing crop depredation rates (e.g. Nchanji
2005).
Finally, this model did not incorporate habitat
degradation or destruction in and around the
park. However, illegal killing of elephants
and other wildlife is known to be correlated
with road building, agricultural encroachment,
and other forms of habitat degradation and
destruction that facilitate human access into
wildlife-inhabited areas (Duckworth & Hedges
1998), and so elephant population management
in WKNP and elsewhere on Sumatra should also
focus on reducing habitat destruction, especially
encroachments into elephant habitat.
Acknowledgements
The study was conducted as a collaboration
between the Wildlife Conservation Society and
the Indonesian Ministry of Forestry’s Directorate
General of Forest Protection and Nature
Conservation (PHKA). The project was funded
by the Wildlife Conservation Society and the US
Fish & Wildlife Service (through Asian Elephant
Conservation Fund grants 1448-98210-00-G496,
98210-1-G806, and 98210-2-G292), the National
Geographic Society, and WWF-US. Data analysis
was supported by the Warnell School of Forestry
and Natural Resources, the University of Georgia,
USA. We thank Clint Moore and Michael Conroy
for valuable advice on modelling. Finally we
thank Margaret Kinnaird, Tim O’Brien, Josh
Ginsberg, and Martin Tyson for support and
advice during the project.
References
Belovsky, G.E. (1987) Extinction model and
mammalian persistence. In: Viable Populations
for Conservation. Soule, M.E. (ed) Cambridge
University Press. New York. pp 35-57.
Berger, J. (1990) Persistence of different-sized
populations: An empirical assessment of rapid
extinctions in bighorn sheep. Conservation
Biology 4: 91-98.
Convention on International Trade in Endangered
Species of Wild Fauna and Flora (CITES) (2000)
Experimental trade in raw ivory of populations in
Appendix II: interpretation and implementation
of the convention. Conservation of and trade in
elephants [Doc 11.31.1]. In: Eleventh Meeting of
the Conference of the Parties. CITES Secretariat,
Geneva, Switzerland.
Blouch, R.A. & Haryanto (1984) Elephants in
Southern Sumatra. Unpublished report, IUCN/
WWF Project 3033, Bogor, Indonesia.
Blouch, R.A. & Simbolon, K. (1985) Elephants
in Northern Sumatra. Unpublished report, IUCN/
WWF Project 3033, Bogor, Indonesia.
Cromsigt, J.P.G.M., Hearne, J., Heitkönig,
I.M.A. & Prins, H.H.T. (2002) Using models
in the management of black rhino populations.
Ecological Modeling 149: 203-211.
Douglas-Hamilton, I. (1987) African elephants:
population trends and their causes. Oryx 21: 11-
24.
Duckworth, J.W. & Hedges, S. (1998) Tracking
Tigers: A Review of the Status of Tiger, Asian
Elephant, Gaur, and Banteng in Vietnam,
Lao, Cambodia, and Yunnan (China), with
Recommendations for Future Conservation
Action. WWF Indochina Programme, Hanoi,
40
Vietnam.
Eberhardt, L.L., Blanchard, B.M. & Knight,
R.R. (1994) Population trend of the Yellowstone
grizzly bear as estimated from reproductive and
survival rates. Canadian Journal of Zoology 72:
1-4.
Hedges, S., Tyson, M.J., Sitompul, A.F., Kinnaird,
M.F., Gunaryadi, D. & Aslan (2005) Distribution,
status and conservation needs of Asian elephant
(Elephas maximus) in Lampung Province,
Sumatra, Indonesia. Biological Conservation
124: 35-48.
Leslie, P.H. (1945) On the use matrices in
population mathematics. Biometrika 33: 183-
212.
Leslie, P.H. (1948) Some further notes on the
use of matrices in population mathematics.
Biometrika 35: 213-245.
Milliken, T. (2004) African Elephants and the
Thirteenth Meeting of the Conference of the
Parties to CITES, Bangkok, Thailand 2004. A
TRAFFIC Brie ng Document. TRAFFIC East/
Southern Africa, Harare, Zimbabwe.
Nchanji, A.C. (2005) Elephant poaching weapons
and new experiences from the Banyang-Mbo
wildlife sanctuary, Cameroon. Pachyderm 39:
33-42.
Poole, J.H. & Thomsen, J.B. (1989) Elephants
are not beetles: implications of the ivory trade
for the survival of the African elephant. Oryx 23:
188-198.
Reilly, J. (2002) Growth in the Sumatran
elephant (Elephas maximus sumatranus) and age
estimation based on dung diameter. Journal of
Zoology 258: 205–213.
Santiapillai, C. & Jackson, P. (1990) The Asian
Elephant: An Action Plan for its Conservation.
IUCN/SSC Asian Elephant Specialist Group,
Gland, Switzerland.
SAS Institute. Proc Regression, Version 8.02.
SAS Institute, Cary, North Carolina, USA.
Sitompul, A.F., Hedges, S. & Tyson, M.J. (2002)
Elephant Deaths in Sumatra as a Result of
Poaching or Elephant Capture Operations, 1
January 2000 to 1 November 2002. Unpublished
Report, Wildlife Conservation Society – Indonesia
Program, Bogor, Indonesia.
Sukumar, R. (1989) The Asian Elephant.
Cambridge University Press, Cambridge, UK.
Sukumar, R., Ramakrishnan, U. & Santosh,
J.A. (1998) Impact of poaching on an Asian
elephant population in Periyar, southern India: A
model of demography and tusk harvest. Animal
Conservation 1: 281-291.
Wiegand. T., Naves, J., Stephan, T. & Fernandez,
A. (1998) Assessing the risk of extinction for
the brown bear (Ursus arctos) in the Cordillera
Cantabrica, Spain. Ecological Application 68:
539-570.
Williams, B. K., Nichols, J.D. & Conroy, M.J.
(2002) Analysis and Management of Animal
Populations. Academic Press, California, USA.
Corresponding author’s e-mail:
asitompu@forwild.umass.edu
... Furthermore, elephant poaching for ivory is a serious problem in several parts of Asia. Elephant scavenging for ivory, stowaways, shrubbery meat, and other items has drastically reduced their population across a large area (Kyaw and Cho2004; Sitompul et al., 2008). ...
Chapter
As the human population has gradually increased in recent years, human elephant conflict concerns have become increasingly serious. For space and resources, many animals are increasingly competing with humans. These problems arise as a result of human encroachment on natural habitats for the sake of agriculture or poaching. The agonizing death of an elephant in Kerala was recently caused by a culture of mistreatment. The death of a pregnant wild elephant in Kerala after allegedly consuming a tainted natural product has prompted outpourings of grief on social media, as well as analysis and shock from people throughout the country. The tale has taken several forms, from net distorting to providing a common tone to the incident. In any event, it has pushed the issue of animal mistreatment in India to the forefront. Thousands of elephants are killed each year in India when their paths cross those of humans, but the image of a critically injured elephant standing emotionlessly in a Palakkad canal while life ebbed out of it will be etched in the mind. It makes no difference if the bomb-caught pineapple that lost its life was intended for elephants or other animals, for similar traps litter the disrupted sceneries that encircle timberlands across the country. In any event, the tragic fate that befell this animal is an unwelcome sign of the growing conflicts between people and animals that are unavoidably going to arise as commercial pressures eat away at essentially reduced living area. The present literature condenses the review of the behavior, ecology, threats to the Asian Elephants and the current scenario of Human elephant conflict and different aspects of mitigation and conservation strategies.
... The park boundary is approximately 227 km long and has 38 village lands occupying about 65% (148 km) of its length, with a human population density (in 2020) of 196 individuals/km 2 (BPS, 2020a(BPS, , 2020bSitompul, Carroll, Peterson, & Hedges, 2008). Permanent agriculture, primarily paddy fields and areas with cultivated bananas, cassava, and maize dominate the landscape surrounding the park as the main source of income for the local communities. ...
Article
Full-text available
Understanding coexistence between humans and threatened wildlife is a central focus in conservation. Way Kambas National Park in Sumatra Island, Indonesia, harbors one of the largest populations of the critically endangered Sumatran elephant (Elephas maximus sumatranus). The people who live alongside this population are affected by intensive crop foraging. Our study investigated the factors which influenced attitudes toward elephants. We then evaluated the implications of reported attitudes for future willingness to live with elephants. We surveyed 660 respondents in 22 villages around the park. People generally reported positive attitudes toward elephants (smartness 95%, usefulness 62%, importance 57%, and pleasantness 53%), apart from where human safety was concerned (safety 11%). Each dimension of attitude was explained by different factors including age, gender, knowledge of elephants, and distance to crop foraging locations. Most respondents (62%) expressed no willingness to coexist with elephants. Such willingness was lower when elephants were perceived to be more dangerous, but higher if beliefs in the benefits of elephants were greater. Efforts to improve crop foraging mitigation practice and to increase people's awareness of elephant benefits may promote support for their conservation. Through this study, we advocate the integration of social science to promote human–wildlife coexistence strategies, an approach that is currently limited in Indonesia.
... Furthermore, elephant poaching for ivory is a serious problem in several parts of Asia. Elephant scavenging for ivory, stowaways, shrubbery meat, and other items has drastically reduced their population across a large area (Kyaw and Cho2004; Sitompul et al., 2008). ...
Book
Medicinal plants are described in most of historcal literatures. We are using them from the initial of human civilization in different modes and traditional ways. The practices and raw meterials are wealth of our modern civilization and need to conserve them for future. The pandemic COVID-19 again bring attention towards the medcinal plants and their traditional therapeutic systems. Ambka Prasad Research Foundation (APRF) has taken an initiative to gather the information on medicinal values of plants nationally in the form of a series of the edited book entitled "Medico-Biowealth of India". This is the fourth volume of the series and here authors discussed about the ethnomedicinal plants used against diarrhea; aerial parasitic plants; plants used in diabetes; medicinal values of some common mangrove plants etc. The content of the book is very useful and wish a grand sucess. I congrutulate to all authors and my co-editors.
... In the context of the Sumatran elephant, Sitompul et al. (2008) develop simulation models and scenarios for the impact of elephant hunting in WKNP using elephant hunting data in 2000-2002. If the hunting data has been duplicated in that period, the elephant population could decline dramatically, with the extinction coefficient significantly increased to around 75%. ...
... Pada konteks Gajah Sumatera, simulasi model dan skenario dampak perburuan gajah di TNWK dengan menggunakan data perburuan gajah di tahun 2000-2002, memprediksikan jika data angka perburuan pada periode tersebut digandakan maka populasi gajah dapat menurun secara dramatis dengan koefisien kepunahan meningkat secara signifikan hingga sekitar 75% (Sitompul et al., 2008). Analisis ini tidak memasukan faktor degradasi habitat di dalam dan sekitar kawasan TNWK. ...
Preprint
Full-text available
Studi tentang persepsi masyarakat tentang ancaman di Taman Nasional Way Kambas (TNWK), Lampung merupakan rangkaian kegiatan untuk mendukung kolaborasi pengelolaan TNWK terutama dalam pengembangan strategi pemberdayaan masyarakat dalam penanganan ancaman perburuan, illegal fishing dan kebakaran hutan. Lokasi studi difokuskan di Desa Braja Harjosari dan Desa Rantau Jaya Udik II yang merupakan “pilot model” kemitraan TNWK dengan masyarakat. Studi yang dilakukan selama bulan Juni dan Juli 2020, mendokumentasikan pengetahuan, perilaku dan praktek-praktek masyarakat terkait dengan hal-hal yang positif dalam penanganan ancaman di TNWK melalui serangkaian focus group discussion di Desa Braja Harjosari dan Desa Rantau Jaya Udik II. Pengumpulan data persepsi masyarakat dilakukan dengan metode kuesioner terstandar (standardized), melibatkan 267 responden rumah tangga di kedua desa tersebut dengan menggunakan 4 (empat) aspek penilaian persepsi yaitu aspek sosial ekonomi, lingkungan, aspek legitimasi dan aspek akseptabilitas. Analisis dalam studi persepsi ini “mungkin” berbeda dan belum banyak dilakukan di kawasan TNWK dan desa penyangga terutama dalam pendekatan pemodelan statistik untuk analisis persepsi (Correlation Somers' D) dan pendugaan faktor-faktor yang mendorong masyarakat dalam bertindak negatif di kawasan TNWK (Regresi Logistik Biner) seperti pada aktivitas ilegal seperti perburuan, pemancingan, dan pembakaran hutan. Analisis-analisis dalam studi persepsi menunjukkan beberapa temuan penting terkait dengan ancaman yang terjadi di TNWK yang nantinya akan berguna dalam merancang strategi penanganannya.
... Gajah menempati habitat yang luas pada beberapa tipe ekosistem mulai dari pesisir, savana, rawa, sampai pegunungan. Menurut Sitompul et al. (2008) gajah cenderung menggunakan canopy medium dan canopy terbuka dimana canopy tertutup sering digunakan gajah pada malam hari. Abdullah et al. (2012) menyatakan bahwa gajah menggunakan hutan sekunder sebagai daerah mencari makan dan menggunakan hutan primer sebagai tempat berlindung, beristirahat dan melakukan perkawinan. ...
Article
Forest land converted into palm oil plantations have caused habitat fragmentation of elephant and land degradation. These lead to land use conflict between human and elephant. The conflict often caused the elephant killed and destructed agricultural land. The study was aimed to estimate potential use and carrying capacity of elephant habitat. Data collection of undergrowth vegetation were analyzed using twelve square plots of 1 x 1m, the distance between the plot of 50 m, tree vegetation of seedlings size 1 x 1 m, saplings 5 x 5 m, and trees 20 x 20 m, the distance between the plot of 200 m and of 1000 m lenght. Vegetation used as elephants feed were observed using purposive sampling and systematically procedure. The analysis showed that biomass of plants producing elephant fodder in Tambang Besi were of Cyperus rotundus (3600.26 kg/ha), Cynodon dactylon (346.74 kg/ha), Melaleuca leucadendron (255.21 kg/ha), and Melastoma malabatricum (156.40 kg/ha). While, the highest biomass in Tebing Penigasan plot is Cyperus rotundus (3575 kg/ha), and in Barak Gajah Plot is Isachne globusa (4013.33 kg/ha). The carrying capacity of elephants habitat of Tambang Besi, Tebing Penigasan, and Barak Gajah plots are 0.78, 0.29, and 0.41 individual/ha/day, respectively.
... Such herds can also suffer from inbreeding depression which refers to a genetic disturbance on the genetic traits of Asian elephants caused by close breeding with members in the same herd. This is because the formation of pocket herd or isolated population of elephants due to herds displacements into remaining patches of forest has reduced the social communication and interaction between elephants and subsequently increased the possibility of inbreeding depression (Ahlering et al., 2011; Debata et al., 2013; Santiapillai & Sukumar, 2006; Sitompul et al., 2008; Sukumar & Santiapillai, 2006). As a result, the good genes might be depleted if the situation became adverse. ...
Article
ASEAN is undergoing a paradigm shift from Government-to-Government (G2G) to Community-to-Community (C2C) relationships with the emphasis on integration and collaboration. The relatively recent developments of Information and Communication Technology (ICT), especially Social Networks, Web 2.0, mobile technology, big data and its related technologies have become the main drivers of this paradigm shift. Big data optimizes capabilities process, high growth and diversified data that create value and knowledge for community within ASEAN in facing ASEAN Economic Community. This study assesses the ICT readiness of each of the ASEAN members to take advantage of the ICT development to build C2C integration and collaboration in facing big data era. The study reveals the score value of each country member to portray ICT readiness for the components of infrastructure development, human capital, people empowerment, innovation, and economic transformation. The paper shows the stages in ICT initiatives in the context of ASEAN and recommend ICT development for each country to eliminate the digital gap between members. It also proposes model of big data for ASEAN member country in supporting economic transformation.
... The mitigation methods used include electric fences, anti-elephant ditches and walls, and improving elephant habitats to increase their natural feed availability and keep them away from farm lands. In some instances village settlements have been relocated, but (Sitompul et al. 2008) has shown that the population would continue to increase over the next 50 years, even with a moderate degree of anthropogenic removal of elephants. ...
Article
Full-text available
In 1985 many African elephant populations, which had been monitored for a decade or more, were either in rapid decline or down to a fraction of their former size. The author examines regional trends and information on key populations with reference to the critical factors affecting the survival of the African elephant, most significantly poaching and the illegal trade in ivory How to Cite This Article Link to This Abstract Blog This Article Copy and paste this link Highlight all http://dx.doi.org/10.1017/S0030605300020433 Citation is provided in standard text and BibTeX formats below. Highlight all BibTeX Format @article{ORX:4958396,author = {Douglas-Hamilton,I.},title = {African elephants: population trends and their causes},journal = {Oryx},volume = {21},issue = {01},month = {1},year = {1987},issn = {1365-3008},pages = {11--24},numpages = {14},doi = {10.1017/S0030605300020433},URL = {http://journals.cambridge.org/article_S0030605300020433},} Click here for full citation export options. Blog This Article Blog This Article : Highlight all African elephants: population trends and their causes I. Douglas-Hamilton (1987). Oryx , Volume 21 , Issue01 , January 1987, pp 11-24 http://journals.cambridge.org/action/displayAbstract?aid=4958396 The code will display like this African elephants: population trends and their causes I. Douglas-Hamilton January 1987 Oryx, ,Volume21, Issue01, January 1987, pp 11-24 http://journals.cambridge.org/abstract_S0030605300020433 I. Douglas-Hamilton (1987). African elephants: population trends and their causes. Oryx, 21, pp 11-24. doi:10.1017/S0030605300020433. Metrics 0Comments
Article
Full-text available
In the mid 1980s, Asian elephant (Elephas maximus) populations were believed to persist in 44 populations on the Indonesian island of Sumatra. Twelve of these populations occurred in Lampung Province, but our surveys revealed that only three were extant in 2002. Causal factors underlying this decline include human population growth, changes in land use, and human–elephant conflict. Nevertheless, our surveys in the Province’s two national parks, Bukit Barisan Selatan and Way Kambas, produced population estimates of 498 (95% CI = [373, 666]) and 180 (95% CI = [144, 225]) elephants, respectively. The estimate for Bukit Barisan Selatan is much larger than previous estimates; the estimate for Way Kambas falls between previous estimates. The third population was much smaller and may not be viable. These are the first estimates for Southeast Asian elephant populations based on rigorous sampling-based methods that satisfied the assumptions of the models used, and they suggest that elephant numbers in these parks are of international importance. While our results suggest that Sumatra’s remaining elephant populations may be larger than expected, they also suggest that the future for these animals is bleak. Human–elephant conflict was reported around all three areas in Lampung and their elephant populations are currently threatened by habitat loss and poaching. Local solutions are possible, but will require much greater commitment by all stakeholders.
Article
Full-text available
The status of the brown bear (Ursus arctos) in Spain has suffered a dramatic decline during the last centuries, both in area and numbers. Current relict populations are suspected to be under immediate risk of extinction. The aim of our model is to attain an understanding of the main processes and mechanisms determining population dynamics in the Cordillera Cantabrica. We compile the knowledge available about brown bears in the Cordillera Cantabrica, northern Spain, and perform a population viability analysis (PVA) to diagnose the current state of the population and to support current management. The specially constructed simulation model, based on long-term field investigations on the western brown bear population in the Cordillera Cantabrica, includes detailed life history data and information on environmental variations in food abundance. The method of individual-based modeling is employed to simulate the fate of individual bears. Reproduction, family breakup, and mortalities are modeled in annual time steps under the influence of environmental variations in food abundance, mortality rates, and reproductive parameters. In parallel, we develop an analytical model that describes the mean behavior of the population and that enables us to perform a detailed sensitivity analysis. We determine current population parameters by iterating the model with plausible values and compare simulation results with the 1982-1995 time pattern of observed number of females with cubs of the year. Our results indicate that the population suffered a mean annual decrease of ~4-5% during the study period, 1982-1995. This decrease could be explained by a coincidence of high poaching pressure with a series of climatically unfavorable years during the period 1982-1988. Thereafter, population size probably stabilized. We estimate that the population currently consists of 25 or 26 independent females and a total of 50-60 individuals. However, our viability analysis shows that the population does not satisfy the criterion of a minimum viable population if mortalities remain at the level of the last few years of 1988-1995. The 'salvation' of at least one independent female every three years is required. The population retains relatively high reproductive parameters, indicating good nutritive conditions of the habitat, but mortality rates are higher than those known in other brown bear populations. The most sensitive parameters, adult and subadult mortality of females, form the principal management target. Our model shows that the series of females with cubs contains valuable information on the state of the population. We recommend monitoring of females with cubs as the most important management action, both for collecting data and for safeguarding the most sensitive part of the population.
Article
The scientific community now agrees that, more than anything else, it is the killing of African elephants for the ivory trade that has caused the very dramatic declines in elephant populations witnessed over the past decade. Based on samples of ivory trade data, recent population modelling and field data, the authors discuss the implications of the ivory trade for the future survival of viable populations of African elephants.
Article
The aim of this study was to investigate age-related growth in the Sumatran elephant Elephas maximus sumatranus and to use the derived relationship to determine the age structure of the wild elephant population in Way Kambas National Park (WKNP), Sumatra. Shoulder height, forefoot circumference and diameter of dung bolus were found to be related to age of captive Sumatran elephants using the Von Bertalanffy growth function. All length measurements were highly correlated with age in the Sumatran elephant and provide growth models for determining the age structure of wild populations. Female captive elephants reached their growth plateau earlier than male elephants who continued growing throughout the ages observed. There was no clear evidence of a secondary growth spurt in male elephants. The growth model relating dung diameter to age was used to predict the age structure of the wild elephant population in WKNP from dung measured along random line transects. The wild elephant population in WKNP is young and dominated by sub-adults (between 5 and 15 years of age). There are marked differences between the age structure of the population as revealed in the current survey and that reported from previous studies, suggesting that changes have occurred within the population in the intervening period. The use of dung diameter to predict age offers a robust field technique for use in situations where direct observations are limited, and the use of other age estimation methods is impractical. It is easily coupled with dung counts for estimating the size, age structure and biomass of elephant populations, and has considerable potential for investigating the effects of poaching on age structure and identifying where priority action should be directed in human–elephant conflict situations.
Article
Ivory poaching has been a serious problem in Asian elephant (Elephas maximus) populations. Reliable records of elephants killed and ivory harvested are generally unavailable. We have used a simulation modelling approach to estimate the numbers of male elephants killed and the quantities of ivory harvested over a 20 year period (1974–94) in the Periyar Reserve of southern India. The age-structured Leslie matrix projection model was modified for this purpose by considering three segments (female, tusked male and tuskless male), relating fecundity to adult sex ratio and iteratively simulating tusked male mortality rates so as to match the observed elephant population structure at Periyar. Four different scenarios of poaching all gave very similar results. We estimate conservatively that 336–388 tuskers have been poached and 3256–3334 kg of ivory harvested by poachers over the 20 year period. The maximum harvest came from the 10–20 year age class. Trends in various demographic parameters such as population numbers, tusked male to tuskless male ratios and female fecundity are described. The implications of ivory poaching and the extremely skewed sex ratios for the conservation and management of the elephant population at Periyar are discussed.
Article
Theory and simulation models suggest that small populations are more susceptible to extinction than large populations, yet assessment of this idea has been hampered by lack of an empirical base. I address the problem by asking how long different-sized populations persist and present demographic and weather data spanning up to 70 years for 122 bighorn sheep (Ovis canadensis) populations in southwestern North America Analyses reveal that: (1) 100 percent of the populations with fewer than 50 individuals went extinct within 50 years; (2) populations with greater than 100 individuals persisted for up to 70 years; and (3) the rapid loss of populations was not likely to be caused by food shortages, severe weather, predation, or interspecific competition These data suggest that population size is a marker of persistence trajectories and they indicate that local extinction cannot be overcome because 50 individuals, even in the short term, are not a minimum viable population size for bighorn sheep.