Evaluating effects of deforestation, hunting, and El Nino events on a threatened lemur

B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
a v a i l a b l e a t w w w . s c i e n c e d i r e c t . c o m
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / b i o c o n
Evaluating effects of deforestation, hunting, and El Nin˜o
events on a threatened lemur
Amy E. Dunhama,*, Elizabeth M. Erhartb, Deborah J. Overdorffc, Patricia C. Wrightd
aDepartment of Organismic and Evolutionary Biology, Harvard University, Harvard University Herbaria, 22 Divinity Ave.,
Cambridge, MA 02138, United States
bDepartment of Anthropology, Texas State University, 601 University Drive, San Marcos, TX 78666, United States
cDepartment of Anthropology, University of Texas at Austin, Austin, TX 78712, United States
dDepartment of Anthropology, Stony Brook University, Stony Brook, NY 11794, United States
A R T I C L E I N F O
A B S T R A C T
Article history:
Madagascar ranks as one of the world’s top extinction hotspots because of its high ende-
Received 23 May 2006
mism and high rate of habitat degradation. Global climate phenomena such as El Nin˜o
Received in revised form
Southern Oscillations may have confounding impacts on the island’s threatened biota
7 September 2007
but these effects are less well known. We performed a demographic study of Propithecus
Accepted 12 October 2007
edwardsi, a lemur inhabiting the eastern rainforest of Madagascar, to evaluate the impact
Available online 26 November 2007
of deforestation, hunting, and El Nin˜o on its population and to re-evaluate present endan-
germent categorization under the IUCN. Over 18 years of demographic data, including
Keywords:
survival and fecundity rates were used to parameterize a stochastic population model
Demography
structured with three stage classes (yearlings, juveniles, and adults). Results demonstrate
El Nin˜o
that hunting and deforestation are the most signi?cant threats to the population. Analysis
Endangered species
of several plausible scenarios and combinations of threat revealed that a 50% population
Extinction risks
decline within three generations was very likely, supporting current IUCN classi?cation.
Management
However, the analysis also suggested that changing global cycles may pose further threat.
Population models
The average fecundity of lemurs was over 65% lower during El Nin˜o years. While not as
Population viability analysis
severe as deforestation or hunting, if El Nin˜o events remain at the current high frequency
Propithecus edwardsi
there may be negative consequences for the population. We suggest that it is most critical
Lemurs
for this species continued survival to create more protected areas, not only to thwart hunt-
Madagascar
ing and deforestation, but also to give this endangered lemur a better chance to recover
from and adapt to altered climate cycles in the future.
Ó 2007 Elsevier Ltd. All rights reserved.
1.
Introduction
received relatively little attention. Given the potential role of
human-induced climate change in altering the frequency of
While many efforts have been made to elucidate the effects of
ENSO events (Fedorov and Philander, 2000; Timmermann
deforestation and hunting pressures on the viability of wild-
et al., 1999), there is a critical need to asses the impact of such
life populations, the confounding effects of global climate cy-
factors on the viability of wildlife populations. This requires
cles such as El Nin˜o Southern Oscillations (ENSO) have
detailed information on the population characteristics of
* Corresponding author: Present address: Department of Ecology and Evolutionary Biology, Rice University, 6100 Main St., Houston, TX
77005, United States. Tel.: +1 703 5050281.
E-mail addresses: amyedunham@hotmail.com (A.E. Dunham), berhart@txstate.edu (E.M. Erhart), overdorff@mail.utexas.edu (D.J.
Overdorff), patricia.wright@ic.sunysb.edu (P.C. Wright).
0006-3207/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biocon.2007.10.006

---

288
B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
the species concerned, which covers a time-series long en-
rized by IUCN as endangered based on a suspected past
ough to capture demographic rates (such as survival and
reduction in population size of 50% over the last three gener-
fecundity) and their variability as well as patterns caused by
ations (39 years) due to decline of suitable habitat (IUCN,
global cycles.
2004). Recent studies have demonstrated that hunting is also
The Milne Edward’s Sifaka, (Propithecus edwardsi, formerly
a threat (Irwin et al., 2000; Lehman et al., 2006b).
P. d. edwardsi; Mayor et al., 2004) of Madagascar, is an excel-
We created population viability models with three stage
lent candidate for such analysis because it has been the sub-
classes (yearlings, juveniles, adults) with new estimates of
ject of two long-term studies (Erhart and Overdorff, 1998;
demographic parameters of P. edwardsi sifakas, based on
Wright, 1995) which together encompass eight El Nin˜o years.
two long-term studies within Ranomafana National Park.
Data collected from these studies and recent ?eld surveys
With these models we quantitatively evaluated the effects
(Irwin et al., 2005; Lehman et al., 2005a) provide detailed infor-
of deforestation, hunting, and ENSO on their population
mation on population vital rates and on the sizes and spatial
dynamics to address the following questions. (1) How do di-
structure of subpopulations. These data provide a suitable
rect anthropogenic threats including deforestation and hunt-
case study for evaluating the interacting effects of direct
ing impact the dynamics of the sifaka population? (2) Does
anthropogenic threats, such as deforestation and harvesting,
the global cycle, ENSO impact demographic rates including
as well as global climate cycles on the viability of small
survival and fecundity? (3) If so, how does it affect the popu-
populations.
lation trajectory and probability of decline? (4) What are the
The high frequency of ENSO in the last few decades has
sensitivities of the population to uncertainty in or possible fu-
raised questions about how human-induced climate change
ture changes in the three threats examined here? (5) Do re-
is affecting or will affect the frequency of ENSO (Fedorov
sults from these models based on long-term demographic
and Philander, 2000; Timmermann et al., 1999). In Madagascar
studies, support the endangerment level categorization under
and southern Africa, ENSO events have been documented to
IUCN criteria (IUCN, 2001)?
cause drought (Thomson et al., 2003) and vegetational
changes (Ingram and Dawson, 2005) which may negatively af-
2.
Methods
fect wildlife (Gould et al., 1999; Wright, 1999). The prospect of
increased ENSO events in Madagascar is daunting given the
2.1.
Species, study area, and population
severity of other anthropogenic threats to its biodiversity.
Madagascar is home to more endemic families and genera
P. edwardsi occur in the southeastern rainforests of Madagas-
than any other conservation ‘hotspot’ in the world, and high
car (Wright, 1995). They live in variable multimale–multife-
rates of habitat loss and other anthropogenic disturbances se-
male social groups ranging in size from 3 to 9 individuals
verely threaten native species (Myers et al., 2000). Over 80% of
(Pochron and Wright, 2003). Data for their population analysis
the forest cover has already been lost and the populations of
came from individuals living within Ranomafana National
many remaining species are small and fragmented (Ganzhorn
Park (RNP; 21°150S, 47°270E), a 41,300 ha reserve located in
et al., 2001; Whitmore, 2000). Current deforestation mainly
south-eastern Madagascar. The park spans elevations of
consists of slash-and-burn agriculture for rice cultivation.
600–1500 m and its vegetation consists primarily of submon-
The government of Madagascar has recently committed to
tane tropical rainforest (Wright and Andriamihaja, 2002).
increasing its protected areas from 1.7 million hectares to 6
Previous studies from RNP have reported age-speci?c
million hectares (Goodman and Benstead, 2005). With this
demographic rates for P. edwardsi based on lemurs living in
new opportunity for designing and creating protected areas,
the Talatakely trail system (Pochron et al., 2004; Wright,
knowledge about direct and indirect threats affecting species’
1995). This site was selectively logged for timber in 1986–
viabilities in Madagascar is crucial.
1989. The present analysis combined data from the Talatakely
Population viability analysis has become an important tool
study (by PCW) and from a study conducted at Vatoharanana,
for assessing such threats. Even for populations with limited
another site within RNP (by DJO and EME).
data it can be useful for evaluating the relative sensitivity of
In both areas the sifakas were habituated to human
species to various anthropogenic pressures and for focusing
observers and could be identi?ed by color markings in the
conservation priorities and estimating ef?cacy of manage-
form of neck collars and tags (see Glander et al., 1991 for cap-
ment efforts. However, few such studies have involved prima-
ture methods). The animals in both studies were routinely
tes, a diverse order with over one third of its species
surveyed and followed by observers. During the studies all
threatened with extinction (Mittermeier et al., 2005). In Mad-
births, deaths and dispersal events were recorded. Deaths
agascar, the situation is especially dire, holding more threa-
were recorded as deaths only when evidence warranted the
tened primates than in any other region (Mittermeier et al.,
classi?cation. Wright (1995) and Erhart and Overdorff (1998)
2005). P. edwardsi lemurs have been subject of long-term stud-
provide more complete details of their study methods.
ies providing detailed demographic data useful for such
From these long-term studies, demographic rates were
analyses.
estimated from a total of six groups of animals (two in Vato-
The long-term studies of P. edwardsi sifakas have shown
haranana, and four in Talatakely) which varied in size over
that they are a long-lived and slowly reproducing lemur (Po-
time. Both study sites are within the southern parcel of the
chron et al., 2004) that is primarily folivorous (Hemingway,
park. Talatakely is located directly south of the Ranomafana
1998). They live in groups of variable composition (Pochron
Road (Route National 25), which bisects the park and Vatoha-
and Wright, 2003) and are poor dispersers across matrix hab-
ranana is 4 km south of Talatakely. For each area, detailed
itat (Lehman et al., 2006a). P. edwardsi are presently catego-
demographic data was obtained using all individuals

---

B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
289
throughout the study periods with a sample size of 76 individ-
(see Fig. 1). Only 523.5 km2 of their range consists of protected
uals (48 individuals from 1986–2003 in Talatakely; 28 individ-
forest, found in Ranomafana National Park and Andringitra
uals from 1992–2003 in Vatoharanana). As this was the only
National Park. Although the core-protected area of Ranomafa-
long-term data available for this species we applied it to all
na National Park is 435 km2 (Wright, 1997), image analysis of
areas of their range in the model. However, because of the
recent forest cover maps provided by ANGAP (2003) reveals
protected status of the lemurs within the two study areas it
that 373.67 km2 is actually forested and available for lemur
was assumed that if bias exists in predictions based on loca-
habitat. The park is divided north and south by a road and
tion of study subjects, they are most likely to be conservative
small villages, which separate subpopulations three and four.
(lower extinction risk prediction).
These parcels are each connected to unprotected habitat
Sample sizes of individuals within age classes were small
extending north and south of the park. Andringitra is located
for individual study sites so statistical comparison of survival
south of Ranomafana and occupies the northwestern region
and fecundity rates was not meaningful (although 95% con?-
of subpopulation 5, containing 150 km2 of protected forest.
dence intervals were overlapping for these vital rates). The
The total population of P. edwardsi in all areas is estimated
average intrinsic rate of growth based on population time-ser-
to be approximately 28,600 (±4442) individuals. The number
ies data (the total number of individuals surveyed in each
of females was calculated for the model assuming a 1:1 sex
area, each year), however, suggested very similar growth rates
ratio.
within the two areas (Talatakely, k = 0.9843; Vatoharanana,
k = 0.9844). Data taken from the two sites were therefore com-
2.2.
Population models
bined to enhance sample size for the analysis.
The area of habitat occupied by sifakas was determined in
Because of the long life span (up to 32 years, King et al., 2005)
ARC-GIS by overlaying vegetation maps (ANGAP, 2003; Du Puy
and limited numbers of marked lemurs, we could not justify
and Moat, 1998) and previously reported range data taken
using an age-structured model. Therefore, we constructed a
from Irwin et al. (2005). For the purposes of this study, patches
female-only, stochastic, three stage population model in RA-
of forest <15 km2 isolated by P5 km of deforested matrix hab-
MAS METAPOP (Akc¸akaya, 2002b). The model was based on
itat, were not included in the model because we assumed
a transition matrix, A, structured by stages of development.
they were unlikely to hold viable populations in the long term.
The transition matrix was made up of transition probabilities
These patches were also mostly found in areas where hunting
(vital rates), aij, representing the average number of individu-
is presumed common (Irwin et al., 2000; Lehman et al., 2006b)
als that an individual in stage i at time t would contribute to
and the persistence of lemurs in these habitats is question-
stage j at time t + 1. An individual could contribute to stage j
able. While this simpli?ed our models, it does hold certain
through survival and growth to the next stage or through
assumptions. However, we do not expect that the possible
reproduction. For P. edwardsi lemurs, even fecundity may be
small additional numbers of animals in these patches would
interpreted as a probability since the maximum fecundity of
signi?cantly in?uence results of this study given the current
a given female is one.
demographic characteristics of this population. All major
Transition probabilities were calculated for females of
roads bisecting the habitat were noted as they may severely
three stage classes (yearlings, juveniles, and adults) (Fig. 2).
limit dispersal and no observation has been made of this spe-
These life history stages were assigned to ages based on sim-
cies crossing a road during the 18 years of study. Other work
ilarities in survival and fecundity values. Yearlings were
has suggested this species is very unlikely to cross deforested
de?ned as between 1 and <2 years of age, juveniles were 2–
matrix habitat (Dehgan, 2003; Lehman et al., 2006a). Popula-
3 years and adults were >3 years. Determining gender of year-
tion size was estimated by multiplying our calculated area
lings can be troublesome through observation alone. In cases
of occurrence by densities of P. edwardsi published by Irwin
when gender was unknown, birth and survival rates of female
et al. (2005).
yearlings were estimated under the assumption of equivalent
P.
edwardsi
was
estimated
to
have
approximately
mortality levels among sexes and a 1:1 sex ratio. There was no
4230.8 km2 of habitat remaining (Table 1). The geographic
signi?cant difference in the ratio of males to females surviv-
range of this species is divided by roads or matrix habitat into
ing to one year of age when offspring with known gender
?ve distinct areas that results in ?ve separate subpopulations
were compared (student’s t-test, t = 0.202, DF = 17, p = 0.842).
Table 1 – Estimated population sizes of P. edwardsi lemurs based on density estimates of Irwin et al. (2005)
Subpopulation
Total forested
Protected
Estimated #
Estimated # of lemurs
area (km2)
area (km2)
of lemurs
in protected areas
1
340.85
0
2304 ± 358
0
2
1264.20
0
8546 ± 1327
0
3
573.11
246.25
3874 ± 601
1596 ± 248
4
849.31
127.22
5741 ± 891
860 ± 134
5
1203.30
150.00
8134 ± 1263
1014 ± 157
Total
4230.77
523.47
28,600 ± 4442
3540 ± 550
Numbers indicate the identi?cation of the subpopulation (see Fig. 1).

---

290
B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
1
2
3
4
N
5
25 50 75 100
Km
Fig. 1 – Map of the distributional range of the Milne Edwards sifaka. Lines represent roads or other barriers to dispersal.
Subpopulations used for this study are numbered.
Fecundity was modeled here as the number of female off-
pling effort (Akc¸akaya, 2002a). The weighted transition proba-
spring per adult female surviving to one year of age. This
bilities were calculated as
avoids correlational problems associated with using fertility
PY Nit
(birth rate per female) and infant survival separately, since
a
t¼1
ij ¼ P
ð1Þ
Y
X
there is often a negative relationship between them. For
t¼1
jtþ1
P. edwardsi, the interbirth interval for mothers with offspring
where Nit is equal to the number of individuals of stage i at
that survive their ?rst year is greater than for females whose
time t, and Xjt+1 is equal to the number of individuals contrib-
offspring do not survive (Pochron et al., 2004).
uted by them to stage j at the next time step. This contribu-
Despite the long duration of the study, estimating the aver-
tion can be through survival, survival and maturity, or
age age of ?rst reproduction is dif?cult as females often emi-
fecundity depending on the transition described. Weighted
grate from natal groups before reproducing. Although in only
measures of overall variance were calculated for each transi-
one instance was a female observed from the time of birth to
tion probability with the following equation as described in
reproduction, the average age of ?rst reproduction has been
Kendall (1998)
estimated to be about four years (Pochron et al., 2004). Most
PY Ntðp À pÞ2
juvenile females transfer from their natal group between
varðpÞ ¼
t¼1
t
P
ð2Þ
Y
N
three and four years of age and are thought to begin breeding
t¼1
t
shortly after. Accurate age determination of immigrating fe-
where pt is the observed transition rate (vital rate) at time step
males is presently not possible but may be achievable in the
t, and Nt is the number of individuals at time step t in the
future with the re?nement of techniques presently being
stage for which the transition rate is calculated. We then cal-
developed using teeth casting (King et al., 2005). No three-
culated average demographic variance (from Akc¸akaya,
year-old females were observed to reproduce during the years
2002a) as
of 1989–2003 (N = 7).
PY p ð1 À p Þ
t¼1 t
t
Weighted measures of transition probabilities (Kendall,
demVarðpÞ ¼
P
ð3Þ
Y
Nt
1998) and environmental variance (Akc¸akaya, 2002a) were
t¼1
used because of the variation in sampling effort (some groups
This average demographic variance Eq. (3) was then sub-
were not monitored all years). Weighted methods reduce bias
tracted from total observed variance Eq. (2) to obtain an esti-
when variation in sample size results from variation in sam-
mate of environmental variance.

---

B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
291
Fc
Pc
Yearlings
Adults
Pa
Juveniles
Pb•Tc
Pb•(1-Tc)
Transition Probabilities
Estimated
Parameter
Weighted mean
Environmental Variation
Pa
0.862
0.293
Pb•(1-Tc)
0.353 0.120
Pb•Tc
0.294
0.100
Pc
0.946 0.032
Normal years
Fc
0.250 0.028
ENSO years
Fc
0.086 0.038
Fig. 2 – Schematic representation of life history stages of lemurs. Dashed line represents production of new individuals. Solid
line represents movement of individuals within or between stages. Px is the probability of survival of stage x to the next time
step. Tx is the probability of moving to stage x. Fx is the fecundity of stage x.
The juvenile stage had a low sample size and data pro-
well for ranking the importance of probability based parame-
duced a variance that may have been overestimated (would
ters because it does not scale prospective variance in the
have resulted in severe truncations). Thus, the environmental
parameter to the size of mean as a standard elasticity analy-
variance for juveniles was calculated with the coef?cient of
sis does (Link and Doherty, 2002) but instead uses an arcsine
variation of the yearling stage, which was assumed to have
square-root scale, where the absolute magnitude of a change
a more similar response to environmental variability than
has meaning independent of the value of the parameter. This
the adult stage. However, until further data is available, this
method has advantages over standard elasticity analysis
assumption is a possible source of error in the model. Envi-
(deKroon et al., 1986) because it does not rank complimentary
ronmental stochasticity was modeled with a lognormal distri-
rates unequally (i.e. survival and mortality) and is thus more
bution. RAMAS Metapop uses binomial distributions to model
appropriate for model parameters based on probabilities (Link
demographic stochasticity (Akc¸akaya, 2002b).
and Doherty, 2002). We calculated VSS as
The following assumptions were implicit in the model.
?????????????????
p
!
o log k
hð1 À hÞ ok
There was no signi?cant correlation between fecundity and
VSS ¼
??
p ? ¼
ð4Þ
o½2 sinÀ1ð hÞ?
k
oh
survival rates observed so they were modeled independently.
Survival rates were modeled as correlated among themselves.
where h represents the demographic parameter being tested
Environmental stochasticity was correlated between patches.
(Link and Doherty, 2002).
Because the probability of dispersal across major roads is so
low, we assigned an annual probability of 0.001 between
2.3.
Population threats
adjacent populations separated by roads. We assumed that
dispersal between population 2 and 3 was 0 because of the ex-
2.3.1.
El Nin˜o Southern Oscillations
tent of deforested habitat separating the two areas; P. edwardsi
ENSO was de?ned as a year that falls during an El Nin˜o event
is very unlikely to cross deforested matrix habitat (Dehgan,
with an annual average standard deviation of Southern Oscil-
2003).
lation Index 6À1.5 and with at least three continuous months
We performed a variance-stabilized sensitivity analysis
with monthly averages of the same. From 1986 to 2003, there
(VSS) based on an arcsine square-root transformation (Link
were eight ENSO years (1987, 1991–1995, 1997, and 2002). To
and Doherty, 2002) to determine which population parame-
examine effects of ENSO, we compared vital rates with non-
ters had the greatest impact on population growth. VSS works
ENSO years. Fecundity was the only vital rate that demon-

---

292
B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
strated an effect of ENSO based on comparisons with 95%
individuals (adults), therefore we altered adult mortality by
con?dence intervals (Table 2). Fecundity was found to be
varying degrees to understand the effects of different hunting
65.6% lower in ENSO years.
pressures. Hunting was imposed only in areas north of
This notable difference resulted in a bimodal distribution
Ranomafana in areas 1, 2 and 3, where hunting is known to
of fecundity values for the study species. To rectify this, we
occur (Fig. 1).
modeled ENSO separately as a stochastic event in which
fecundity was decreased by 65.6% with a certain annual prob-
3.
Results
ability. If the parameter in question becomes unimodal when
extreme events are excluded, the model can be improved by
3.1.
Model results
modeling the extreme events separately (Akc¸akaya, 2000).
Thus, we were able to use a lognormal distribution for simu-
The geometric mean of the ?nite growth rate of the popula-
lating stochasticity in fecundity by excluding the fecundity
tion estimated from a time-series of numbers of P. edwardsi fe-
values for ENSO years and modeling them separately. The an-
males of all stages surveyed in Ranomafana National Park
nual ENSO probability calculated over the time of the study
between 1987 and 2003 was 0.993. The basic stage-based mod-
period was 0.389. This may be unusually high so we examined
el which assumed density dependence, static habitat condi-
effects of lower ENSO frequency as well (range, 0.14–0.389).
tions, and ENSO frequency modeled as observed in the data,
produced a similar estimate with an average growth rate of
2.3.2.
Habitat decline
0.997. This result suggests the sifaka population within the
Deforestation may be the largest threat to the sifaka popula-
protected area is more stable than suggested by a previous
tion and has been estimated to occur in Madagascar rainfor-
demographic study which predicted a quickly declining pop-
est at a rate of about 1.4–4.7% per year (Achard et al., 2002;
ulation (Pochron et al., 2004).
FAO, 2003), predominately due to slash-and-burn agricultural
We ranked population parameters in order of importance
practices. In this study, the effects of deforestation (de?ned
to population growth trajectories with a VSS sensitivity anal-
here as habitat loss) were modeled by imposing a carrying
ysis. The results suggest that the growth rate is most sensitive
capacity that declined linearly with habitat area. A ceiling
to changes in adult survival (0.1824), closely followed by
model of density dependence (manifested as a simple upper
fecundity (0.1253). Growth was least sensitive to juvenile
threshold) was employed for all models since these lemurs
and yearling survival (0.0693, 0.0289, respectively) .
are known to have limited habitat, and it is unrealistic for this
No model scenarios demonstrated a risk of extinction for P.
species to assume exponential growth. Previous work has
edwardsi over the next 100 years but all models suggested the
suggested that ceiling models provide adequate predictions
population is at risk of decline. Estimated probabilities of de-
about the probabilities of population decline, and are less
cline risk over the span of three generations (36 years) varied
prone to underestimate risk than more complex non-linear
greatly depending on threat category and degree of threat ap-
forms of density dependence (Sabo et al., 2004). The popula-
plied to the model (Table 4).
tion was assumed to be at carrying capacity at the start of
each simulation.
3.2.
Effects of ENSO
2.3.3.
Hunting
ENSO had strong effects on fecundity levels, reducing rates by
Hunting of lemurs in eastern Madagascar is generally for sub-
65.6% on average. All models included ENSO events occurring
sistence and for large leaping species such as Propithecus often
at various frequencies (Table 4), and when an event occurred
involves hunting with blowguns, darts and slingshots (Leh-
in a simulation it reduced fecundity for all populations. High-
man and Ratsimbazafy, 2001). Hunting is common in subpop-
er frequencies of ENSO events reduced average ?nal popula-
ulations north of Ranomafana National Park, (Irwin et al.,
tion size (Fig. 3a), and increased risk of decline (Table 4).
2000; Lehman et al., 2006c), but its severity has not yet been
The sensitivity of outcomes to uncertainty in ENSO frequency
quanti?ed. To examine what different hunting levels might
was low, estimated to maximum of 3.2% difference in the risk
have on the population we modeled several scenarios of
of declining by 50% over three generations. When ENSO
hunting in these areas. Hunters are most likely to select large
Table 3 – Levels of threat used in models to quantify
Table 2 – Vital rates and 95% con?dence intervals
impact on decline risk
observed in data for ENSO and normal years (non-ENSO)
Parameter
Level measure
Levels of threat examined
Vital rates
ENSO years
Normal years
Low
Medium
High
(7 years)
(10 years)
Average
95% CI
Average
95% CI
ENSO
Annual probability
0.140
0.264
0.389a
Deforestation
Annual proportion
0.014
0.029
0.047
Fecundity
0.086
0.070–0.124
0.250
0.238–0.290
rateb
of habitat lost
Yearling survival
0.800
0.680–1.000
0.867
0.631–1.000
Hunting rateb
Annual proportion
0.02
0.06
0.10
Juvenile survival
0.667
0.428–1.000
0.667
0.390–1.00
of adults taken
Adult survival
0.942
0.913–1.000
0.950
0.941–1.00
a Rate observed in data.
Effect of ENSO was only apparent for fecundity.
b Only applied to areas outside of national parks.

---

B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
293
Table 4 – Model results under various scenarios of declining carrying capacity (K), hunting pressures, and ENSO events
Parameter
Low
Medium
High
Maximum difference
Probability of 50% decline in three generations (36 years)
ENSO probability
0.036
0.049
0.068
0.032
Deforestation ratea
0.353
1.000
1.000
0.647
Hunting ratea
0.104
0.386
0.724
0.620
Combined parameters
0.365 (best case)
1.000
1.000 (worst case)
0.635
Probability of 80% decline in three generations (36 years)
ENSO probability
0.000
0.000
0.000
0.000
Deforestation ratea
0.000
0.000
0.301
0.301
Hunting ratea
0.000
0.001
0.003
0.003
Combined parameters
0.000 (best case)
0.030
0.427 (worst case)
0.427
Maximum difference shows the sensitivity of results to uncertainty of each threat. See Table 3 for an explanation of parameter level.
a ENSO set at medium probability level (0.264).
events were modeled at their highest frequencies (equal to
50% decline over three generations between harvesting 2%
frequencies observed in the data), and other risks were held
versus 10% of adults was 62% (Table 4). High values predicted
at zero, there was no estimated risk of declining by 80% over
a 72% chance of 50% decline, while low values predicted 10%
the next three generations. However, there was a slight risk of
chance of decline. Chance of declining by 80% during the
declining below 50% under all ENSO scenarios modeled (Table
same time period was under 1% for all values of hunting
4).
intensity modeled.
3.3.
Effects of deforestation
3.5.
Combined threats
In our models, realistic levels of deforestation of sifaka habi-
The sifakas we studied are likely to be experiencing simulta-
tat had a strong effect on the population dynamics (Fig. 3b,
neously all of the threats we examined separately here. We,
Table 4). The recent, most conservative estimated rate of
therefore, created models with mixed threats to look at
forest loss (1.4% per year, FAO, 2003) resulted in a 35.6% lower
probabilities of decline for best and worst case scenarios
?nal average abundance than when modeled without defor-
(Table 4). Best case scenario for mixed-threat models was a
estation (Fig. 3b). The highest estimate of deforestation rate
33.8% chance of declining by 50% over the next three
used in the model (4.7% per year, Achard et al., 2002) resulted
generations. Under the worst case scenario, the model
in 76.7% lower ?nal average abundance than when habitat
predicted a 100% chance of declining by half in three genera-
was modeled as static. Uncertainty in the rate of forest loss
tions. Only the worst case scenario predicted any risk for an
was shown to have a strong effect on the decline risk out-
80% decline, which was estimated as 33.1%. The models are
comes of the model. At intermediate and higher levels of
most sensitive to uncertainty in hunting rates and deforesta-
deforestation the risk of decline by 50% in three generations
tion (Table 4).
was 100%. The model employing low values of forest loss re-
sulted in a risk of 35% (Table 4). Under these scenarios the risk
4.
Discussions and conclusions
of P. edwardsi declining by 80% or more in just three genera-
tions (to qualify as critically endangered) was low or absent
Although there were limitations to the data set and our model
(Table 4). Only when high values were modeled was there
had several implicit assumptions, we were able to make some
any risk of such decline (30.1% chance).
clear conclusions from this population viability analysis.
While obvious anthropogenic disturbances such as deforesta-
3.4.
Effects of hunting
tion and hunting pressures clearly threaten Madagascar’s
lemur populations, our analyses also suggest that changes
In the models, hunting effects were applied only to P. edwardsi
in the frequency of the global cycle of ENSO through climate
populations existing north of Ranomafana National Park be-
change could have confounding effects on populations by
cause that is where hunting has been documented (Lehman
reducing fecundity. In this study the growth rate of the popu-
et al., 2006b) and hunting of P. edwardsi in the southern areas
lation of P. edwardsi, was found to be sensitive to changes in
is restricted due to local taboos (Wright et al., 2005). While all
fecundity, second only to adult survival rates. With increasing
studies of hunting of this species have been qualitative, it is
global warming, changes in ENSO frequencies and intensities
still useful to look at a range of potential hunting values to in-
may become more frequent (Fedorov and Philander, 2000) and
form managers of how hunting rates may affect the viability
more of a problem for the population. In this study adult and
of this species. The highest level of hunting used was 10%
juvenile mortality was not affected by the global cycles,
of adults harvested annually, which resulted in a 45% lower
however, if dry spells continue to increase in frequency and
?nal average abundance than in models without hunting
duration, it is possible that all stages will be affected; a
(Fig. 3c). The sensitivity of the outcome of the model to uncer-
scenario which was not modeled here, but which could have
tainty in hunting is great. The difference in probability of a
devastating impacts on the population.

---

294
B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
attributed to malnourishment during nursing and/or weaning
a
16000
stages, particularly for infants of older mothers (King et al.,
15000
2005). Malnourishment may also increase predation risks of
14000
older weaned infants because it may cause them to become
weak and/or separated from the group and thus easy targets
13000
for predators (Gould et al., 1999). Drought conditions during
12000
ENSO years may also reduce fecundity by lowering fertility
11000
rates for P. edwardsi. A recent study of several sifaka species
10000
found that low body weights of adults were associated with
9000
prolonged drought seasons (Lehman et al., 2005b), which
Final average population size
may cause reduced birth rates as in other primates (Bercov-
8000
0.1
0.15
0.2
0.25
0.3 0.35
0.4
itch et al., 1999).
ENSO frequency
While ENSO cycles were shown to have important effects
on P. edwardsi, the most pressing and obvious threat to the
16000
b
population of P. edwardsi sifakas is the present rate of defores-
tation. The models demonstrate that present rates of forest
14000
loss will seriously affect the long-term viability of this popu-
12000
lation and will likely result in a major decline in population
size over just three generations (36 years). If deforestation is
10000
not halted, the eventual range of this species will become
restricted to the protected areas of Ranomafana and Andrin-
8000
gitra National Parks where the population size would be re-
6000
duced to an estimated 3500 individuals and split among
three isolated fragments including the north and south par-
4000
cels of Ranomafana National Park, and the eastern half of
Final average population size
Andringitra National Park. We suggest that under this sce-
2000
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
nario, the small subpopulation sizes and increased forest
Rate of habitat loss (%/yr)
edge area, which facilitates further anthropogenic distur-
bance, would then severely threaten the viability of this
c
16000
species.
Although our model included several subpopulations, we
14000
were unable to assign habitat quality to a landscape model
12000
because of a lack of data. Habitat quality and elevational con-
straints are likely to affect survival and fecundities in the sub-
10000
populations and are an important area for future study and
for re?ning population models. A closely related lemur spe-
8000
cies, Propithecus diadema has actually been shown to respond
to degraded forest and edge with higher densities (Irwin,
6000
Final average population size
2006) presumably because of increased available forage. The
4000
effects of this on vital rates of P. edwardsi are unclear. While
0
0.02
0.04
0.06
0.08
0.1
we recognize that habitat quality may have important conse-
Hunting rate (proportion of
quences for source/sink dynamics not examined here, we
adults taken per year)
suspect that variability within the population, unless ex-
Fig. 3 – The effects of (a) ENSO, (b) deforestation, and (c)
treme, will have only a marginal effect on the results of this
hunting on the ?nal average population abundance of
study. The geographic range maps of this species is also in
females after running models for three generations
need of revision and habitat area may already be smaller than
(36 years). Original population size was 14,000 females.
presumed here because of habitat preferences and hunting
pressures (Lehman et al., 2006c).
A run of the model in its present form does not show any
ENSO events caused periods of reduced fecundity (averag-
negative effect of the habitat’s present level of fragmentation
ing 65.6% lower) that may be linked to reduction in precipita-
when compared with a model of a similar but unfragmented
tion levels. While rainfall data for this species’ range available
population. Effects of fragmentation may become more of a
to us was too patchy both temporally and spatially to com-
problem in the future however, when subpopulations become
pare differences between ENSO and non-ENSO years, ENSO
smaller and/or habitat becomes more fragmented. Under-
is known to cause drought (Thomson et al., 2003) and changes
standing how vital rates are linked with habitat quality and
in vegetational indices (Ingram and Dawson, 2005) in
threats in the different areas would allow further exploration
Madagascar’s rainforests. Rainfall level during critical times
of this question and may provide different results depending
of the year has already been shown to affect lemur vital rates
on the source sink dynamics of the population.
(Gould et al., 1999; Jolly et al., 2002), including infant survival
Hunting, even at low intensity levels has a strong effect on
of P. edwardsi sifakas (King et al., 2005). This effect has been
decline risk of P. edwardsi sifakas. Our models demonstrate

---

B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
295
that the life history characteristics of the sifakas make their
available, we suggest that the listing of this species as ‘Endan-
population sensitive to perturbations in adult survival rates.
gered’ is appropriate for the present time.
Hunting, which is likely to target larger animals, has a strong
Our sensitivity analysis suggested that changes in adult
potential to limit the viability of the population. This result
mortality and fecundity will have the largest effects on the
underscores the critical need to quantify poaching of P. ed-
population growth rate. We suggest, therefore, a focus on
wardsi so we can fully understand risks facing their popula-
conservation strategies that include reduction of hunting
tion and implement effective management strategies for
pressures and maintaining intact habitat, which are both
protecting their long-term viability.
important to adult survival and fecundity. Future work lead-
A survey of lemurs by Lehman et al. (2005a, 2006b) con-
ing to the understanding of speci?c adult mortality factors
ducted within the range of subpopulation one and two, sug-
and their relationship to habitat quality is also important.
gests declines due to hunting may have already occurred in
For P. edwardsi it is particularly important to aim conserva-
the northern areas. In seven of eight different sites they sur-
tion strategies that protect adult breeders and quality habi-
veyed, no P. edwardsi individuals were sighted. P. edwardsi
tat so they can successfully reproduce. We suggest a
was only observed in the southern-most site. Although village
critical conservation effort for maintaining their population’s
informants explained that sifakas were common in the for-
continued survival is to increase protected area in the re-
ests within the past one to ten years, it is possible that heavy
gion, not only to thwart hunting and deforestation, but also
hunting from blowguns, darts and slingshots have already
to give this endangered lemur a better chance to recover
dramatically reduced numbers in these areas (Lehman and
from and adapt to altered climate cycles such as ENSO in
Ratsimbazafy, 2001).
the future.
The presence of hunting in the southern region of the pop-
Given our preliminary results, we suggest four types of
ulation (populations 4 and 5) has not been documented and
data that would most improve the model and our under-
was, therefore, not included in the models. In the southern
standing of the viability of this species. The ?rst two data
region, hunting of sifakas is known to be taboo, and is not
types needed include better estimates and spatial informa-
practiced (Wright et al., 2005). However, opportunistic hunting
tion about the deforestation rate and hunting rates faced by
may occur in some areas. Also, attitudes about hunting in the
this species. These parameters resulted in the greatest
south may change. With a growing economy in and around
amount of uncertainty in our model predictions. The third
Ranomafana, immigrants are settling from other areas of
and fourth include dispersal information and data on the
Madagascar who may not share the same taboos. For these
relation between habitat quality and its effect on vital rates
reasons it is critical to also monitor hunting pressures in this
for each subpopulation. The latter two would allow an explo-
southern region. If hunting occurs in subpopulations 4 and 5
ration of the source/sink dynamics of the population and
the rates of decline for the population are expected to be even
would provide information on the bene?ts of corridors or
greater than described by our results.
road bridges to link fragmented habitat.
A recent demographic study by Pochron et al., 2004 used
age-speci?c life tables to estimate that the population may
Acknowledgements
be declining with a ?nite growth rate of 0.9371. If accurate,
this level of decline suggests that the population may be fac-
In Madagascar we thank the National Association for the
ing >80% reduction over the next three generations, qualify-
Management of Protected Areas (ANGAP), Department of
ing the population to be listed as critically endangered
Water and Forests, and the Ministry of the Environment,
under IUCN criteria (IUCN, 2004). This is in contrast to our
Water and Forests. Many thanks to B. Andriamihaja and A.
study which suggested that the population has nearly no
Feistner and staffs of MICET, Centre ValBio, and ICTE for logis-
chance of declining by 80% or more over three generations
tical help. Funding was provided from the David and Lucile
even under the worst case scenario. The demographic study
Packard Foundation, Douroucouli Foundation, the Wenner-
of Pochron et al. (2004), however, was not meant as a popula-
Gren Foundation, the John D. and Catherine T. MacArthur
tion viability study, and the ?nite growth rate was based on a
Foundation, National Geographic Society, National Science
model with 22 separate age classes. Conclusions drawn from
Foundation, Earthwatch Institute, and Stony Brook University.
such detailed models with limited data (e.g. some age classes
Thanks to research technicians P. Rasabo, the late G. Rakoton-
represented by one or a few individuals) are likely to lead to
irina, R. Rakotovao, R. Ratsimbazafy, P. Talata, L. Ralisoa, and
unreliable conclusions with a high level of uncertainty. In spe-
A. Telo for their expert assistance following animals. Thanks
cies such as P. edwardsi where long generation times preclude
also to R. Akc¸akaya for helpful advice on modeling and V. Ru-
even long-term studies from supplying suf?cient data for age-
dolf, M. Irwin, and two anonymous reviewers for comments
structured models, it may be appropriate to reduce model
on the manuscript.
complexity to increase sample size for vital rate estimates
as we did in our models.
Despite considerable uncertainty in the threat parameters
R E F E R E N C E S
used in our models, several scenarios supported the present
IUCN listing of the sifakas as ‘Endangered’ (Table 4); results
suggest a high likelihood of a reduction in population size
Achard, F., Eva, H.D., Stibig, H., Mayaux, P., Gallego, J.,
by 50% or more over the next three generations (IUCN,
Richards, T., Malingreau, J., 2002. Determination of
2004). Even the best case scenario suggested a probability of
deforestation rates of the world’s humid tropical forests.
declining to this level, of nearly 40%. Given the present data
Science 297, 999–1002.

---

296
B I O L O G I C A L
C O N S E R V A T I O N
1 4 1 ( 2 0 0 8 ) 2 8 7 – 2 9 7
Akc¸akaya, H.R., 2000. Population viability analysis with
Preliminary results and implications for lemur conservation.
demographically and spatially structured models. Ecological
Lemur News 5, 20–21.
Bulletins 48, 23–38.
IUCN, 2001. IUCN Red List Categories and Criteria: Version 3.1.
Akc¸akaya, H.R., 2002a. Estimating the variance of survival rates
IUCN Species Survival Commission, Gland, Switzerland and
and fecundities. Animal Conservation 5, 333–336.
Cambridge, UK.
Akc¸akaya, R.H., 2002b. RAMAS Metapop: Viability analysis for
IUCN, 2004. 2004 IUCN Red List of Threatened Species. A Global
stage-structured metapopulations. Version 4.0. Applied
Species Assessment. in: IUCN (World Conservation Union),
Biomathematics, Setauket, NY.
Gland and Cambridge, UK.
ANGAP, 2003. Carte de localisation du Parc National de
Jolly, A., Dobson, A., Rasamimanana, H.M., Walker, J.,
Ranomafana. Map by Association Nationale pour la Gestion
O’Connor, S., Solberg, M., Perei, V., 2002. Demography of
des Aires Prote´ge´es/FTM, Ranomafana.
Lemur catta at Berenty Reserve, Madagascar: Effects of troop
Bercovitch, F.B., Lebron, M.R., Martinez, H.S., Kessler, M.J., 1999.
size, habitat and rainfall. International Journal of
Primigravidity, body weight, and costs of rearing ?rst offspring
Primatology 23, 327–353.
in rhesus macaques. American Journal of Primatology 46,
Kendall, B.E., 1998. Estimating the magnitude of environmental
135–144.
stochasticity in survivorship data. Ecological Applications 8,
Dehgan, A. 2003. The behavior of extinction: Predicting the
184–193.
incidence and local extinction of lemurs in fragmented
King, S.J., Arrigo-Nelson, S.J., Pochron, S.T., Semprebon, G.M.,
habitats of southeastern Madagascar. Ph.D.
Godfrey, L.R., Wright, P.C., Jernvall, J., 2005. Dental senescence
Dissertation.University of Chicago Publisher, Department of
in a long-lived primate links infant survival to rainfall.
Ecology and Evolution, 248 pp.
Proceedings of the National Academy of Sciences 102,
deKroon, H.A., Plaisier, J.M., van Groenendael, J.M., Caswell, H.,
16579–16583.
1986. Elasticity; the relative contribution of demographic
Lehman, S., Ratsimbazafy, J., Rajaonson, A., Day, S., 2005a. Decline
parameters to population growth rate. Ecology 67, 1427–1431.
of Propithecus diadema edwardsi and Varecia variegata
Du Puy, D.J., Moat, J.F., 1998. Vegetation mapping and
variegata (Primates: Lemuridae) in south-east Madagascar.
classi?cation in Madagascar (using GIS): implications and
Oryx 40, 1–4.
recommendations for the conservation of biodiversity. In:
Lehman, S.H., Ratsimbazafy, R.H., 2001. Biological assessment of
Huxley, C.R., Lock, J.M., Cutler, D.F. (Eds.), Chorology, taxonomy
the Fandriana-Marolambo forest corridor, Madagascar. Lemur
and ecology of the Floras of African and Madagascar. Royal
News 6, 8.
Botanic Gardens, Kew, pp. 97–117.
Lehman, S.M., Mayor, M., Wright, P.C., 2005b. Ecogeographic size
Erhart, E.M., Overdorff, D.J., 1998. Infanticide in Propithecus
variations in sifakas: A test of the resource seasonality and
diadema edwardsi: An evaluation of the sexual selection
resource quality hypotheses. American Journal of Physical
hypothesis. International Journal of Primatology 19, 73–81.
Anthropology 126, 318–328.
FAO, 2003. State of the World’s Forests (SOFO) – SOFO 2003. Food
Lehman, S.M., Rajaonson, A., Day, S., 2006a. Edge effects and their
and Agricultural Organization of the United Nations, Rome,
in?uence on lemur density and distribution in Southeast
Italy.
Madagascar. American Journal of Physical Anthropology 129,
Fedorov, A.V., Philander, S.G., 2000. Is El Nin˜o changing? Science
232–241.
288, 1197–2002.
Lehman, S.M., Ratsimbazafy, J., Rajaonson, A., Day, S., 2006b.
Ganzhorn, J.U., Lowry II, P.P., Schatz, G.E., Sommer, S., 2001. The
Decline of Propithecus diadema edwardsi and Varecia variegata
biodiversity of Madagascar: one of the world’s hottest hotspots
variegata (Primates: Lemuridae) in south-east Madagascar.
on its way out. Oryx 35, 346–348.
Oryx 40, 108–111.
Glander, K.E., Fedigan, L.M., Fedigan, L., Chapman, C., 1991. Field
Lehman, S.M., Ratsimbazafy, J., Rajaonson, A., Day, S., 2006c.
methods for capture and measurement of three monkey
Ecological correlates to lemur community structure in
species in Costa Rica. Folia Primatologica 57, 70–82.
southeast Madagascar. International Journal of Primatology
Goodman, S.M., Benstead, J.P., 2005. Updated estimates of
27, 1023–1040.
biotic diversity and endemism for Madagascar. Oryx 39,
Link, W.A., Doherty, P.F.J., 2002. Scaling in sensitivity analysis.
73–77.
Ecology 82, 3299–3305.
Gould, L., Sussman, R.W., Sauther, M.L., 1999. Natural disasters
Mayor, M.I., Sommer, J.A., Houck, M.L., Zaonarivelo, J.R., Wright,
and primate populations: The effects of a 2-year drought on a
P.C., Ingram, C., Engel, S.R., Louis, E.E., 2004. Speci?c status of
naturally ocurring population of ring-tailed lemurs (Lemur
Propithecus spp. International Journal of Primatology 25,
catta). International Journal of Primatology 20, 69–83.
875–900.
Hemingway, C.A., 1998. Selectivity and variability in the diet of
Mittermeier, R.A., Valladares-Pa´dua C., Rylands A.B., Eudey
Milne-Edwards’ Sifakas (Propithecus diadema edwardsi):
A.A., Butynski T.M., Ganzhorn J.U., Kormos R., Aguiar J.M.,
Implications for folivory and seed-eating. International
Walker S., 2005. Primates in peril: The world’s 25 most
Journal of Primatology 19, 355–377.
endangered primates 2004–2006. Report by IUCN/SSC
Ingram, J.C., Dawson, T.P., 2005. Climate change impacts and
Primate Specialist Group (PSG), International Primatological
vegetation response on the island of Madagascar.
Society (IPS), Conservation International (CI), Washington,
Philosophical Transactions of the Royal Society A 363, 55–59.
DC. p. 45.
Irwin, M.T. 2006. Ecological impacts of forest fragmentation on
Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fanseca, G.A.B.,
diademed sifakas (Propithecus diadema) at Tsinjoarivo, eastern
Kent, J., 2000. Biodiversity hotspots for conservation priorities.
Madagascar: Implications for conservation in fragmented
Nature 403, 853–858.
landscapes. Ph.D. thesis. Stony Brook University Publisher,
Pochron, S., Tucker, T., Wright, P.C., 2004. Demography, life
Department of Physical Anthropology, 410 pp.
history and social structure in Propithecus diadema edwardsi
Irwin, M.T., Johnson, S.E., Wright, P.C., 2005. The state of lemur
from 1986 to 2000 of Ranomafana National Park,
conservation in Southeastern Madagascar: Population and
Madagascar. American Journal of Physical
habitat assessments for diurnal lemurs using surveys, satellite
Anthropology.
imagery and GIS. Oryx 39, 204–218.
Pochron, S., Wright, P., 2003. Variability in adult group
Irwin, M.T., Smith, T.M., Wright, P.C., 2000. Census of three
compositions of a prosimian primate. Behavioural Ecology and
eastern rainforest sites north of Ranomafana National Park:
Sociobiology 54, 285–293.

---

Document Outline

• Evaluating effects of deforestation, hunting, and El Ni ntilde o events on a threatened lemur
  • Introduction
  • Methods
    • Species, study area, and population
    • Population models
    • Population threats
      • El Ni ntilde o Southern Oscillations
      • Habitat decline
      • Hunting
  • Results
    • Model results
    • Effects of ENSO
    • Effects of deforestation
    • Effects of hunting
    • Combined threats
  • Discussions and conclusions
  • Acknowledgements
  • References

---