Population responses to natural and human-mediated

disturbances: assessing the vulnerability of the common

hippopotamus (Hippopotamus amphibius)
Rebecca Lewison*
Nicholas School of the Environment and Earth Sciences, Duke University, 135 DUML Rd., Beaufort, NC 28516, U.S.A.

decline over the next 60 years in response to a combi-

Abstract
nation of environmental fluctuations and human-medi-
Vulnerable wildlife populations can face a suite of ated threats.
anthropogenic activities that may threaten their persist-
Key words: catastrophe, environmental stochasticity,
ence. However, human-mediated disturbances are likely
habitat loss, hippopotamus, multi-matrix population
to be coincident with natural disturbances that also
model, population persistence, sensitivity to disturbance
influence a population. This synergism is often neglected
in population projection models. Here I evaluate the
effects of natural (rainfall fluctuation) and human dis-
turbances (habitat loss and unregulated hunting) using Résumé
a multi-matrix environmental state population model for
Les populations sauvages vulnérables peuvent être con-
the common hippopotamus (Hippopotamus amphibius). By
frontées à une suite d’activités humaines qui risquent de
evaluating each disturbance type (natural and human)
menacer leur persistance. Cependant, les perturbations
alone and then together, I explicitly consider the
causées par l’homme sont susceptibles de coı̈ncider avec
importance of incorporating realistic environmental
des perturbations naturelles qui influencent aussi une
variability into population projection models. The model
population. Cette synergie est souvent négligée dans les
population was most strongly affected by moderate
modèles de projection des populations. Ici, j’évalue les effets
habitat loss, which yielded the highest probability of
des perturbations naturelles (fluctuations des chutes de
crossing the risk thresholds over the 60 year time per-
pluie) et humaines (perte d’habitat et chasse non
iod, although these probabilities were relatively low
réglementée) en utilisant une chaı̂ne de matrices aléatoires
(£0.31). However, the likelihood of crossing the risk
pour l’état environnemental d’une population d’hippopo-
thresholds were two to five times as high when human-
tames (Hippopotamus amphibius). En évaluant chaque type
mediated and natural disturbances were considered
de perturbation (naturelle et humaine) seul et ensemble,
together. When these probabilities were calculated per
je considère explicitement l’importance qu’il y a d’intégrer
year of the simulation, the results suggested that even
une variabilité environnementale réaliste dans les modèles
relatively mild human disturbances, when considered in
de projection des populations. La population modèle était
conjunction with realistic natural disturbance, resulted
surtout affectée par une perte d’habitat modérée, qui
in a high probability (>0.50) of substantial declines
réunissait la plus grande probabilité de dépasser les seuils
within decades. The model highlights the importance of
de risque en une période de 60 ans, même si cette prob-
integrating realistic natural disturbances into population
abilité était relativement faible (£0,31). Cependant, la
models, and suggests that, despite locally abundant
probabilité de franchir les seuils de risque était 2 à 5 fois
populations, protected hippopotamus populations may
plus grande lorsque les perturbations d’origine humaine et
naturelle étaient considérées ensemble. Lorsque ces prob-
*Correspondence: IUCN Hippo Specialist Group, Biology Depart-
abilités étaient calculées par année, les résultats de la
ment, San Diego State University, San Diego, CA 92182, U.S.A.
Tel.: +1 619 594 8287; Fax: +1 619 594 5676; E-mail:
simulation suggéraient que même des perturbations
[email protected] humaines faibles, quand on les considérait en conjonction

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415 407
408 R. Lewison

avec une perturbation naturelle réaliste, résultaient en une Southeastern Africa where hippo populations tend to occur
forte probabilité (>0,50) de déclin substantiel en quelques at high densities (Fig. 1). Information from The World
décennies. Le modèle souligne l’importance qu’il y a d’in- Conservation Union (IUCN) surveys suggests that in many
tégrer les perturbations naturelles réalistes dans les countries where common hippos are found, their popula-
modèles de population et suggère que, malgré des popu- tions are declining (Fig. 2). These declines have been
lations localement abondantes, les populations protégées attributed to two anthropogenic activities: habitat loss as
d’hippopotames peuvent décliner au cours des 60 wetlands are converted or impacted by agricultural
prochaines années en réaction à une combinaison de development (Smuts & Whyte, 1981; Jacobsen & Kleyn-
fluctuations environnementales et de menaces d’origine haus, 1993) and unregulated hunting for meat and ivory
humaine. from the common hippo’s large canine and incisor teeth
(Vega, 1995). Small-scale habitat loss is most commonly
the result of weirs or dams redirecting water from rivers
and lakes to crop areas (Cole, 1992; Jacobsen & Kleyn-
Introduction
haus, 1993; Viljoen, 1995; Viljoen & Biggs, 1998). Adult
Successful wildlife conservation requires timely manage- hippos are hunted for meat and ivory across a gradient of
ment and conservation actions. Because of the need for hunting intensities in which the intensity of the hunting
proactive management, conservation of long-lived ani- pressure reflects the nutritional needs and density of
mals, in particular, relies heavily on our ability to project coincident human populations (Weiler, De-Meulenaer &
future population trends. As a result, population models Vanden-Block, 1994; Vega, 1995; Shambaugh, Ogle-
are a sine qua non of species viability management, despite thorpe & Ham, 2001).
uncertainty associated with model components and Because hippos spends most daytime hours in or next
structure (Beissinger & Westphal, 1998; Ludwig, 1999; to water (Field, 1970), they exhibit notable demographic
Regan et al., 2003). The primary focus of many such sensitivity to annual rainfall variability. Field records and
models is the impact of current human activities on future research conducted over the past 50 years have quan-
population trends (Crowder et al., 1994; Doak, Kareiva & tified marked changes in hippo demographic rates in
Klepetka, 1994; Heppell, Crouse & Crowder, 2000; Karei- response to rainfall fluctuations (Moffett, 1958; Atwell,
va, Marvier & McClure, 2000). However, human-mediated 1963; Pienaar, Van Wyk & Fairall, 1966; Smuts &
disturbances are likely to be coincident with natural dis- Whyte, 1981). In low rainfall or drought years, hippo
turbances that also can influence population trends. populations can crash, because of a dramatic decline in
Although the conceptual importance of including conception and an abrupt increase in mortality, most
environmental variability, including catastrophes, has likely due to lack of forage, heat stress and increased
been identified as a necessary component of population vulnerability to disease (Sayer & Rakha, 1974; Smuts &
models (Mangel & Tier, 1994; Ludwig, 1996; Ripa & Whyte, 1981). The proportion of females likely to con-
Lundberg, 2000; Drake & Lodge, 2004; et al., 2004), ceive in a drought year has been estimated to drop from
realistic natural disturbances are not typically a primary 30% to <5% (Smuts & Whyte, 1981). Historical
element within population projection models (but see Liu, accounts from as early as 1929 from protected areas in
Menges & Quintana-Ascencio, 2005). This omission may many African countries (South Africa, Tanzania, Zam-
lead model results to be overly optimistic, i.e. when bia, Zimbabwe) provide support for this pronounced
anthropogenic activities are considered in conjunction response to low rainfall conditions (Sayers, 1930). Con-
with natural disturbances, even relatively moderate or versely, in response to higher than average rainfall,
mild human-mediated disturbances may lead to significant common hippo populations exhibit dramatic population
population changes. surges. During these surges, the population can return
Populations of common hippopotamus (Hippopotamus to predrought levels, often in 1 or 2 years, because of
amphibius) have declined in response to human dis- high forage availability and increased reproductive
turbances. Common hippopotamus (hippos) were once activity from sexually mature adults. In high rainfall
found throughout sub-Saharan Africa, although only years, individuals younger than the average age of
estimated historic distributions are available. The current sexual maturity also may become sexually active in
distribution is primarily concentrated in Eastern and response to favourable conditions and contribute to

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
 Population persistence of the hippopotamus 409

Fig 1 Estimated (a) former (c. 1959) and

(b) current (c. 1993) distribution of com-
mon hippopotamus (from IUCN, 1993)

population growth (Sayer & Rakha, 1974; Smuts & Methods

Whyte, 1981).
Given their sensitivity to natural disturbance regimes, Model structure and parameters
and the human-mediated threats that have been linked to
I constructed the model using a Leslie–Lefkovitch matrix
population declines, hippos are a good case study to
based on the known demographic demarcations of hippos
consider the synergistic effects of natural and human
(Sayer & Rakha, 1974). The age distribution was divided
disturbances on population growth. Here, I present an
into six age classes (Table 1). The annual probability of an
assessment of the sensitivity of a model hippo population
individual growing into the next age class was calculated as:
to both natural and mild or moderate human distur-


ri Ti

T 1
bances. I measured population sensitivity to these dis-  rki i
k
turbances by identifying population thresholds that ci ¼
k¼1 ð1Þ
ri Ti
k 1
signify increased risk to the persistence of the population
and calculated the likelihood of crossing these thresholds where Ti is time in stage i, ri is survivorship in stage i, and
over the time horizon of the model. The goal of this k is rate of population change, assumed to be 1 (Caswell,
exercise was to evaluate population sensitivity to realistic 2001).
disturbances and project likely responses of a common Model parameters were taken from multi-year studies
hippo population within African protected areas. across multiple populations. This included field data col-

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
410 R. Lewison

multiple studies (Laws, 1968; R. Lewison, unpublished

data). Annual survival incorporates the probability of
surviving and remaining in the same age class (for classes
that extend beyond one year), and the probability of sur-
viving and growing into the next age class, P and G
respectively. Fecundity was based on standardized
female : offspring estimates averaged across years in KNP
and on the percent of culled females whose reproductive
organs showed evidence of conception from five cropping
regimes (Bere, 1959; Laws & Clough, 1966; Sayer &
Rakha, 1974; Marshall & Sayer, 1976; Smuts & Whyte,
1981). Thus, fecundity was the product of the proportion
of females in the population and the proportion of females
likely to have conceived during a given year.
Demographic stochasticity was also included. At each
1 year time step, matrix elements were sampled from a beta
distribution (using a Monte Carlo method) to incorporate
temporal variability and estimation error in the parame-
ters. The model also implicitly accounted for density-
Fig 2 Estimated common hippo population trends (adapted from dependent responses as the population exhibits higher
IUCN, 1993)
fecundity and recruitment following a population crash
(i.e. when the population is low), assuming habitat condi-
Table 1 Age classes for common hippos based on known demo- tions improve. The model assumed a postbreeding census,
graphic demarcations i.e. the yearly sampling follows birthing events. Because
hippos are polygynous and assumed to have an even sex
Age class Description Duration (year)
ratio, the model only accounted for female individuals.
1 Newborn 1 The initial age distribution was taken from standardized
2 First year calf 1 age-class ratios based on census counts conducted in four
3 Juvenile 5
African protected areas (Masai Mara Game Reserve, Ken-
4 Subadult 2
5 Adult 21
ya; Katavi NP, Tanzania; Queen Elizabeth NP, Uganda;
6 Senescent adult 5 Luangwa NP, Zambia). An initial population size of 5000
females was based on median population sizes in 56 pro-
tected areas from the 1993 IUCN Action Plan (IUCN,
lected between 1996 and 1998 in Katavi National Park 1993). The population was assumed to be closed to
(KNP), in southwestern Tanzania (R. Lewison, unpub- immigration, which realistically represents most protected
lished data) and from published studies from the 1950s to hippo populations. The model was run in @Risk software
1980s from unthreatened populations in Uganda, Zambia, (Palisade Decision Tools, Palisade Corporation), which is
Zimbabwe and South Africa (Bere, 1959; Pienaar et al., an Excel add-in programme.
1966; Mackie, 1976; Marshall & Sayer, 1976; Smuts &
Whyte, 1981; O’Connor & Campbell, 1986). These un-
Environmental stochasticity: modelling rainfall variability
threatened populations were not, however, subject to the
human-mediated disturbances considered here. Thus the To incorporate environmental stochasticity, I used the
demographic values derived are not subject to ‘double described demographic data in a multi-matrix environ-
dipping’, i.e. the demographic rates do not already incor- mental state model. In this type of model, multiple matrices
porate the effects of the human disturbances and thus will are used to represent a range of environmental conditions
not tend to overestimate potential declining trends (Brook, (sensu Beissinger, 1995). In this case, the model was
2000). Survivorship estimates were based on life tables composed of four matrices representing four rainfall states
derived from 350 carcasses found in protected areas across – drought, low, average and high (Table 2). The four

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
 Population persistence of the hippopotamus 411

Table 2 Demographic values for each matrix representing one of Table 3 Transition probabilities used in the rainfall submodel
the four rainfall states (a) drought, (b) low, (c) average, and (d)
high. The average matrix (c) represented a stable population with To
k1 From Drought Low Average High

Age class Annual survival Annual fecundity Drought 0.00 0.50 0.50 0.00
Low 0.00 0.13 0.75 0.12
(a) Average 0.03 0.16 0.75 0.06
1 0.45 0 High 0.25 0.00 0.50 0.25
2 0.60 0
3 0.73 0 Probabilities were determined from the Katavi National Park
4 0.78 0 (KNP) data shown in Fig. 3, after assigning each annual rainfall
5 0.85 0.10 records to one of the four rainfall states based on its deviation from
6 0.42 0.02 the overall mean rainfall value.
(b)
1 0.55 0
(Fig. 3), annual rainfall was assigned to one of the four
2 0.70 0
3 0.75 0 rainfall states based on the mean rainfall value. Drought
4 0.85 0.23 was defined as rainfall that was greater than or equal to two
5 0.92 0.31 standard deviations below the mean. The high and low
6 0.47 0.14 rainfall states were defined as rainfall that was greater or
(c) equal to one standard deviation above or below the mean,
1 0.67 0 respectively. Although the rainfall data come from a spe-
2 0.74 0
cific field site, the pattern is likely to be representative of
3 0.85 0
4 0.88 0.26
other African protected areas with bimodal rainfall.
5 0.95 0.36
6 0.50 0.15
Disturbance scenarios
(d)
1 0.76 0 I evaluated the model population’s sensitivity to three
2 0.78 0 relatively mild or moderate human disturbance scenarios –
3 0.88 0.04
small-scale habitat loss, and two levels of hunting pressure
4 0.95 0.35
(Table 4). A mild form of habitat loss occurs when water is
5 0.97 0.43
6 0.52 0.25 diverted from rivers and lakes to agricultural areas via
small dams and weirs. When water is diverted, although
the rainfall pattern per se remains unchanged, the rainfall
matrices were calibrated with demographic data from field that the hippos experience is reduced, i.e. water diversion
studies collected between 1959 and 1998 (Bere, 1959; is in effect less rainfall. Because the most dramatic demo-
Pienaar et al., 1966; Mackie, 1976; Marshall & Sayer, graphic response to low water availability is an abrupt
1976; Smuts & Whyte, 1981; O’Connor & Campbell, decline in conception, I represented this habitat loss by
1986; Lewison, unpublished data). Inter-matrix variability decreasing fecundity by 5%. I considered unregulated
in adult survival is lower than variability of juvenile sur- hunting that occurred at two levels – low and moderate
vival (Gaillard, Festa-Bianchet & Yoccoz, 1998). hunting pressure. I assume hippo populations adjacent to
At each time step in the model, a rainfall submodel low human density areas experience a low level of
selected one of the four matrices to reflect one of four unregulated hunting pressure, whereas hippo populations
rainfall states. The submodel used a discrete, first-order adjacent to more densely populated areas would experi-
Markov chain with stationary transition probabilities to ence higher hunting pressure. Unregulated hunting was
represent stochastic variation in rainfall (Stewart, 1994). modelled by decreasing adult survival (age classes 4 & 5)
This assumes that the rainfall state at time t + 1 depends by 0.05% and 1%, to represent low and moderate hunting
only on the rainfall state at time t. The transition proba- pressure respectively.
bilities (Table 3) among the rainfall states were derived The model population’s sensitivity to each of the dis-
from a 50-year-rainfall record collected in and around KNP turbances was assessed by the proportion of simulation

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
412 R. Lewison

2500

2000
Rainfall (mm)

1500

1000

Fig 3 Rainfall data from Katavi National

500
Park, Tanzania. Data were compiled from
local (International Research Institute for
0 Climate Prediction, Rukwa Region Land
1950
1953
1956
1959
1962
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
1998
Planning Office) and Park records. Data for
1993–1995 were incomplete

Table 4 The human disturbances included in the model

Disturbance Cause Modelled as

Small-scale habitat loss Water diverted from lakes and rivers for agriculture 5% reduction in fecundity
Low hunting pressure Unregulated hunting targeting adults 0.05% reduction in annual adult survival
Moderate hunting pressure Unregulated hunting targeting adults 1% reduction in annual adult survival

Table 5. The impact of disturbance scenarios on the model hippo population with and without consideration of typical rainfall variability

Without rainfall variability With rainfall variability

P (30) P (40) P (50) P (30) P (40) P (50)

Habitat loss 0.31 0.08 0 0.61 0.49 0.37

Moderate hunting pressure 0.14 0 0 0.58 0.47 0.34
Low hunting pressure 0 0 0 0.41 0.28 0.17

P (30), P (40), P (50) are the probabilities of the model population declining to 30%, 40% and 50% of its original size of 5000 individuals.
The probability values were calculated as the percentage of the simulation iterations that crossed each threshold over 60 years.

runs in which the population dropped below three popu- chosen by the IUCN as the standard length of time over
lation size thresholds – 3500, 3000 and 2500, individuals. which population declines should be assessed ( IUCN,
Population thresholds are often used to represent quasi- 2001). Each disturbance simulation was run for 1000
extinction or minimum viable population levels (Brook & iterations.
Kikkawa, 1998). Here the thresholds represent a boundary
into a larger risk category – a population that has declined
Results
30%, 40%, or 50% from its original size. These thresholds
were based on IUCN Red List specifications that rely on
Impact of disturbances
similar population reductions in the criteria for endan-
gered, threatened, or vulnerable status for an entire pop- I measured the impact of the disturbances on the model
ulation (IUCN, 2001). Each simulation was run for 60 hippo population as the proportion of simulation runs in
time steps, or years. This approximates three generations, which the population dropped below one of three-risk
where generation length is defined as the average age of thresholds (Table 5). Across the entire simulation time
parents of the current cohort. This time horizon has been period, habitat loss alone yielded the largest likelihood of

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
 Population persistence of the hippopotamus 413

1.00

Probability of population declining = 2500 (50% decline)

Low hunting pressure
Moderate hunting pressure
0.80
Habitat loss

0.60

0.40

Fig 4 Projections of the likelihood of the 0.20

model population dropping to 50% of its
original size, calculated for each year of
0.00
the model. The dashed line marks a prob- 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58
ability of 0.50 Year

crossing the population-risk thresholds, although even the drought will increase as global temperatures rise (Bwango
largest probabilities of decline for this disturbance were low et al., 2000). Although it is clear that including more in-
(P £ 0.31). However, the same disturbances when con- tense or severe disturbances would lead to greater popu-
sidered in conjunction with rainfall variability resulted in lation declines, it is important to note that even relatively
risk likelihoods that were two to five times higher (Table 5). mild to moderate human disturbances, when coupled with
When rainfall variability is included, the cumulative realistic environmental variability, led to substantial pop-
probabilities of crossing even the smallest risk threshold ulation changes. Because population models must balance
across the entire time period, i.e. 30% population decline, biological realism and model complexity, incorporating
were fairly high for all three disturbances (P ¼ 0.41–0.61). realistic environmental variability may be challenging.
To estimate when declines were likely to occur during However, without considering natural disturbance
the simulations, I calculated the per-year probability of the regimes, population projections for environmentally sen-
model population crossing the 50% risk threshold, i.e. sitive species may be overly optimistic in assessing the ef-
declining to <2500 individuals. The results suggest that by fects of anthropogenic activities.
year 40, even a large hippo population may decline to The use of population risk thresholds as indicators of
<50% of its original size in response to some mild or vulnerability has been applied to population models before
moderate disturbances and environmental variability (Gerber, DeMaster & Kareiva, 1999; Gerber & DeMaster,
(Fig. 4). 1999). Shifting the focus from population viability to
vulnerability acknowledges the uncertainty inherent in
population models and redefines population models as
Discussion
tools to determine relative, rather than absolute, threats
Although the projected trajectories are not powerful as (Beissinger & Westphal, 1998; Groom & Pascual, 1998;
absolute predictors of future trends in hippo populations, Regan, Colyvan & Burgman, 2002). Delineating risk
these findings point to the importance of integrating thresholds can also facilitate classification of increased
human disturbances with natural disturbance regimes. vulnerability and population recovery (Gerber & DeMaster,
There is evidence that hippo populations in some areas 1999). This approach also provides a means for land-
face far more severe disturbances than those included here. managing agencies to identify risk thresholds relevant to
For example, recent field surveys found that common management, i.e. population sizes thresholds below, which
hippo populations in Democratic Republic of Congo, once intervention or management is likely to be less effective or
home to the largest hippo populations among African too costly.
countries, have declined >95% during recent years from The hippo population modelled here is designed to
very intense hunting pressure (Shoumatoff, 2000; Hillman characterize wild hippo populations in African regions,
Smith et al., 2003). In addition, some climate models particularly East, Southeast and South Africa where the
predict that in many areas of East Africa, the frequency of largest numbers of common hippo are currently found.

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
414 R. Lewison

Because the model is based on demographic data from Brook, B.W. & Kikkawa, J. (1998) Examining threats faced
several protected areas, and incorporates typical features of by island birds: a population viability analysis on the
hippo populations from several countries, the model Capricorn silvereye using long-term data. J. Appl. Ecol. 35,
491–503.
population is representative of large, wild hippo popula-
Bwango, A., Wright, J., Elias, C. & Burton, I. (2000) Reconciling
tions. The model’s primary assumption is that hippo national and global priorities in adaptation to climate change:
demographic rates are not constant over time. Previous with an illustration from Uganda. Environ. Monit. Assess. 61,
research strongly supports the assertion that these rates 145–159.
fluctuate with rainfall (see Introduction). Despite their Caswell, H. (2001) Matrix population models: construction, analysis,
large size, longevity, and flexible demographic rates, wild and interpretation, 2nd edn. Sinauer Associates, Massachusetts.
hippo populations are likely to exhibit sensitivity in 722 pp.
Cole, M. (1992) Zimbabwe’s hippos threatened by drought. New
response to the combined effects of responses to annual
Sci. 134, 139.
rainfall fluctuations and anthropogenic factors. Given our
Crowder, L.B., Crouse, D.T., Heppell, S.S. & Martin, T.H. (1994)
current understanding of hippo ecology and available Predicting the impact of turtle excluder devices on loggerhead
demographic data, substantial population declines for sea turtle populations. Ecol. Appl. 4, 437–445.
common hippos are likely if wetland habitat loss continues, Doak, D., Kareiva, P. & Klepetka, B. (1994) Modeling population
particularly if hunting pressure also intensifies. For hippos, viability for the desert tortoise in the Western Mojave Desert.
and for other environmentally sensitive species, consider- Ecol. Appl. 4, 446–460.
Drake, J.M. & Lodge, D.M. (2004) Effects of environmental
ing the impacts of both natural and human-mediated
variation on extinction and establishment. Ecol. Lett. 7, 26–
disturbances is likely to be important to ensure population
30.
persistence. Field, C.R. (1970) A study of the feeding habits of the hippopot-
amus in the Queen Elizabeth National Park, Uganda, with some
management implications. Zool. Afr. 5, 71–86.
Acknowledgements
Gaillard, J.M., Festa-Bianchet, M. & Yoccoz, N.G. (1998) Popu-
Many thanks to R. Brock, T. Caro, P. Hodum, J. Neale, lation dynamics of large herbivores: variable recruitment with
M. Johnson, M. Mangel, S. Riley, B. Sacks, and S. Heppell constant adult survival. Trends Ecol. Evol. 13, 58–63.
Gerber, L.R. & DeMaster, D.P. (1999) A quantitative approach to
for helpful and insightful comments on manuscript drafts.
endangered species act classification of long-lived vertebrates:
Funding for field work was provided by the Frankfurt application to the North Pacific humpback whale. Conserv. Biol.
Zoological Society, Chicago Zoological Society, and British 13, 1203–1214.
Airways Conservation Programme. Special thanks to Gerber, L.R., DeMaster, D.P. & Kareiva, P.M. (1999) Gray whales
Acting Director General G. Bigurube of Tanzania National and the value of monitoring data in implementing the U.S.
Parks, M. Borner of Frankfurt Zoological Society, Principal Endangered Species Act. Conserv. Biol. 13, 1215–1219.
Chief Park Warden of Katavi National Park Mr A. Kyam- Groom, M.J. & Pascual, M.A. (1998) The analysis of population
persistence: an outlook on the practice of viability analysis. In:
bile, Katavi National Park ecologist Mr Lejora, and the staff
Conservation Biology 2nd edn. (Eds P. L. Fiedler and P. M.
at Katavi National Park. This publication is dedicated to
Kareiva). Chapman and Hall, New York.
the memory of Mike Rose, Gary Polis, Shigeru Nakano, Heppell, S.S., Crouse, D.T. & Crowder, L.B. (2000) Using matrix
Takuya Abe, and Masahiko Higashi. models to focus research and management efforts in conserva-
tion. In: Quantitative Methods for Conservation Biology (Eds S.
Ferson and M. Burgman). Springer-Verlag: Berlin.
References Hillman Smith, A.K., Merode, E., Smith, F., Ndey, A., Mushenzi, N.,
Atwell, R.I.G. (1963) Surveying Luangwa hippo. Puku, 29–49. & Mboma, G. (2003) Virunga National Park – North Aerial
Beissinger, S.T. (1995) Modeling extinction in periodic environ- Census of March 2003. Available at: http://assets.panda.org/
ments. Ecol. Appl. 5, 618–631. downloads/virungaaerialcount.pdf (accessed on 24 December
Beissinger, S.T. & Westphal, M.I. (1998) On the use of demo- 2006).
graphic models of population viability in endangered species Jacobsen, N.H.G. & Kleynhaus, C.J. (1993) The importance of wiers
management. J. Wildl. Manage. 62, 821–841. as refugia for hippopotamus and crocodiles in the Limpopo
Bere, R.M. (1959) Queen Elizabeth National Park, Uganda: the River, South Africa. Water S.A. 19, 301–306.
hippopotamus problem and experiment. Oryx 5, 116–124. Kareiva, P., Marvier, M. & McClure, M. (2000) Recovery and
Brook, B.W. (2000) Pessimistic and optimistic bias in population management options for spring/summer chinook salmon in the
viability analysis. Conserv. Biol. 14, 564–566. Columbia River Basin. Science 290, 977–979.

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415
 Population persistence of the hippopotamus 415

Laws, R.M. (1968) Dentition and ageing in the hippopotamus. Sayers, G.F. (1930) The Handbook of Tanganyika, 1st edn. Macmil-
East Afr. Wildl. J. 6, 19–52. lan and Co., London.
Laws, R.M. & Clough, G. (1966) Observation of reproduction in Shambaugh, J., Oglethorpe, J. & Ham, R. (2001) The Trampled Grass:
the hippopotamus (Hippopotamus amphibius). Symp. Zool. Soc. Mitigating the Impacts of Armed Conflict on the Environment.
London 15, 117–140. Biodiversity Support Program, Washington, DC.
Liu, H., Menges, E.S., & Quintana-Ascencio, P.F. (2005) Population Shoumatoff, A. (2000) Four World Heritage Sites in Danger in
viability analyses of Chamaecrista keyensis: effects of fire season Eastern Congo: Biodiversity Conservation in the Vortex of Civil War.
and frequency. Ecol. Appl. 15, 210–221. Report to the United Nations Foundation. Available at: http://
Ludwig, D. (1996) The distribution of population survival times. www.dispatchesfromthevanishingworld.com/dispatch2/
Am. Nat. 147, 506–526. D2P2a.html (accessed on 24 December 2006).
Ludwig, D. (1999) Is it meaningful to estimate a probability of Smuts, G.L. & Whyte, I.J. (1981) Relationships between
extinction. Ecology 80, 298–310. reproduction and environment in the hippopotamus in the
Mackie, C. (1976) Feeding habits of the hippopotamus on the Kruger National Park. Koedoe 24, 169–185.
Lundi River, Rhodesia. Arnoldia 7, 1–16. Stewart, W.J. (1994) Introduction to the Numerical Solution of
Mangel, M. & Tier, C. (1994) Four facts every population biologist Markov Chains. Princeton University Press, Princeton, NJ.
should know about persistence. Ecology 75, 607–614. Vega, I. (1995) The Hippo Threatened Due to Ivory Trade. Quercus
Marshall, P.J. & Sayer, J.A. (1976) Population ecology and III, Mayo.
response to cropping of a hippo population in Eastern Zambia. J. Viljoen, P.C. (1995) Changes in number and distribution of hip-
Appl. Ecol. 13, 391–403. popotamus (Hippopotamus amphibius) in the Sabie River, Kruger
Moffett, J.P. (1958) Handbook of Tanganyika, 2nd edn. Govern- National Park, during the 1992 drought. Koedoe 38, 115–121.
ment of Tanzania, Dar es Salaam, Tanzania. Viljoen, P.C. & Biggs, H.C. (1998) Population trends of hippopot-
O’Connor, T.G. & Campbell, B.M. (1986) Hippopotamus habitat ami in the rivers of Kruger National Park, South Africa. In:
relationships on the Lundi River, Gonarezhou National Park, Behavior and Ecology of Riparian Mammals (Eds N. Dunstone and
Zimbabwe. Afr. J. Ecol. 24, 7–26. M. L. Gorman ). Cambridge Press, London.
Pienaar, U. de V., Van Wyk, P. & Fairall, N. (1966) An experi- Weiler, P., De-Meulenaer, T. & Vanden-Block, A. (1994) Recent
mental cropping scheme of hippopotami in the Letaba River of Trends in the International Trade of Hippopotamus Ivory. Traffic
the Kruger National Park. Koedoe 9, 1–33. Bulletin – IUCN Wildlife Trade Monitoring Unit 15, 47–49.
Regan, H.M., Colyvan, M. & Burgman, M.A. (2002) A taxonomy of World Conservation Union – IUCN. (1993) Pigs, Peccaries and
treatment of uncertainty for ecology and conservation biology. Hippos. (Eds W. L. R. Oliver ). World Conservation Union, Gland,
Ecol. Appl. 12, 619–628. Switzerland.
Regan, H.M., Akcakaya, H.R., Ferson, S., Root, K.V., Carroll, S. & World Conservation Union – IUCN. (2001) IUCN Red List Categ-
Ginzburg, L.R. (2003) Treatments of uncertainty and variability ories and Criteria Version 3.1. Prepared by the IUCN Species
in ecological risk assessment of single-species populations. Hum. Survival Commission. Gland, Switzerland.
Ecol. Risk Assess. 9, 889–906.
Ripa, J. & Lundberg, P. (2000) The route to extinction in variable (Manuscript accepted 7 December 2006)
environments. Oikos 90, 89–96.
Sayer, J.A. & Rakha, A.M. (1974) The age of puberty of the hip- doi: 10.1111/j.1365-2028.2006.00747.x
popotamus (Hippopotamus amphibius) in the Luangwa River in
eastern Zambia. East Afr. J. Wildl. 12, 227–232.

 2007 The Author. Journal compilation  2007 Blackwell Publishing Ltd, Afr. J. Ecol., 45, 407–415