Ecological Interactions, Social Organization, and Extinction Risk in African Wild Dogs JOHN A. VUCETICH* AND SCOTT CREEL† *School of Forestry, Michigan Technological University, Houghton, MI 49931, U.S.A., email [email protected] †Department of Biology, Montana State University, Bozeman, MT 59717, U.S.A., email [email protected]

Abstract: The conservation of wild dogs depends on the persistence of small populations because African wild dog ( Lycaon pictus) populations are low in density, are limited by range loss, and are often restricted to parks containing fewer than 100 adults. Although major limiting factors for wild dog populations have been identified, including interspecific competition and diseases, such factors have not been translated into extinction risk. To assess wild dog extinction risks, we used individual-based simulations constructed from data from a 6-year field study in the Selous Game Reserve, Tanzania. Our simulations predicted that extinction risk for wild dogs was extremely sensitive to competition with lions. Extinction rates ( for periods as short as 20 years) rose sharply to near 1.0 when lion populations exceeded moderate densities (approximately 110– 140 lions/1000 km2). This prediction is remarkably consistent with, and highlights, ecological processes that may be responsible for recent patterns of extinction among wild dog populations. Infectious diseases that kill adults, such as rabies, also reduced population persistence if they increased mortality by $0.3 and occurred at average intervals of #10 years. In contrast, diseases killing only pups, such as canine parvovirus, had weaker effects on persistence. Although persistence declined sharply for mean litter sizes #6, persistence was unaffected by increasing mean litter size above its normal range (i.e., 8–12 in Selous). Increasing mean pack size from typical levels reduced extinction risk, but reproductive suppression may set an upper limit on pack size. Although the risk of extinction for 20- to 100-year time frames was appreciable for many realistic ecological and demographic conditions, even low immigration rates substantially increased persistence probabilities. Active management to mitigate the effects of interspecific competition, facilitate dispersal among populations, or augment population size appears essential for wild dog conservation. Interacciones Ecológicas, Organización Social, y Riesgo de Extinción de Perros Silvestres Africanos Resumen: La conservación de perros silvestres depende de la persistencia de poblaciones pequeñas debido a que las poblaciones del perro silvestre africano (Lycaon pictus) exhiben baja densidad, estan limitados por la périda del rango de distribución y se encuentran frecuentemente restringidos a parques que contienen poco menos de 100 adultos. A pesar de que los factores limitantes principales para las poblaciones de perros silvestres han sido identificadas (e.g., competencia interspecífica y enfermedades), estos factores no han sido traducidos en términos de riesgos de extinción. Para evaluar los riesgos de extinción e perros silvestres, utilizamos simulaciones basadas en individuos, construídas en base a datos de un estudio de campo de seis años en la Reserva Selous Game, en Tanzania. Nuestras simulaciones predicen que el riesgo de extinción para perros silvestres fué extremadamente sensitiva a la competencia con leones. Las tasas de extinción (por períodos tan cortos como 20 años) se incrementaron rápidamente hasta acercarse a 1.0 cuando las poblaciones de leones excediéron densidades moderadas (z110–140 leones/1000 km2). Esta predicción es remarcablemente consistente con, y resalta a procesos ecológicos que pueden ser responsables de patrones recientes de extinción entre poblaciones de perros silvestres. Enfermedades infecciosas que aniquilan adultos, como lo es la rabia, también reducen la persistencia poblacional si la mortalidad se incrementa por $0.3 y ocurre a intervalos promedio de #10 años. En contraste, las enfermedades que atacan solo a cachorros, como lo es el parvovirus canino, tuvo un efecto dénil en la persistencia. Aunque la persistencia disminuye marcadamente en camadas

Paper submitted July 17, 1998; revised manuscript accepted February 17, 1999.

1172 Conservation Biology, Pages 1172–1182 Volume 13, No. 5, October 1999

Vucetich & Creel

Extinction Risk in African Wild Dogs

1173

con una media #6, la persistencia no fue afectada al incrementarse el tamaño de la camada por arriba de su rango normal (i.e., 8–12 en Selous). Un incremento promedio en el tamaño de la manada por arriba de su tamaño típico redujo el riesgo de extinción, sin embargo, la supresión reproductiva puede establecer un limite máximo en el tamaño de la manada. A pesar de que el riesgo de extinción para bloques de tiempo de 20 a 100 años fue apreciable para varias condiciones ecológicas y demográficas realistas, aún una tasa de inmigración baja incrementa sustaincialmente la probabilidad de persistencia. Un manejo activo para mitigar los efectos de la competencia interspecífica facilita la dispersión entre poblaciones, el aumento en al tamaño poblacional parece ser esencial para la conservación de perros silvestres.

Introduction The ecology of large carnivores presents difficult problems for conservation (Schaller 1996). Almost by definition, conflicts with humans and livestock are a serious issue for animals that are large and carnivorous (Clark et al. 1996). Real or potential conflicts with human activities in multiple-use areas restrict large carnivores to reserves and adjacent areas in much of the world. Such reserves must be large and ecologically intact to accommodate large carnivores because they exhibit lower population densities than smaller species or those at lower trophic levels (Blackburn & Gaston 1994) and because large carnivores range widely (Woodroffe & Ginsberg 1998). African wild dogs (Lycaon pictus) illustrate these problems clearly. The wild dog is a large (20–28 kg), pack-living carnivore that preys mainly on ungulates. They are the only living member of the genus Lycaon, which diverged from the genus Canis (wolves and jackals) about the time this genus appeared 3–5 million years ago ( Wayne et al. 1997 ). Wild dogs formerly inhabited most of subSaharan Africa, with the exception of desert and rainforest, but now have a patchy distribution and are largely confined to protected areas (Fanshawe et al. 1991, 1997). Historically, wild dogs were never known to exhibit high densities (Selous 1908), and population densities now vary from about 5 adults/ 1000 km2 (Serengeti National Park) to 40 adults/1000 km2 (Selous Game Reserve). Densities of 17–20 adults/ 1000 km2 are typical in large parks dominated by open woodland (Fuller et al. 1992; Maddock & Mills 1994). A park of 10,000 km2 is likely to hold ,200 wild dogs, and field studies have documented few populations exceeding 100 adults (Maddock & Mills 1994; McNutt 1995; Fanshawe et al. 1997). The Selous Game Reserve contains the largest remaining population, with approximately 880 adults and yearlings (Creel & Creel 1998). Because wild dog conservation depends on the viability of populations with a few hundred individuals, the quantitative assessment of factors affecting extinction risk for such populations is a high priority. Field studies have identified factors that affect the density

and demography of wild dogs, including habitat loss, human-caused mortality, infectious diseases, and competition with lions and hyenas ( Kruuk 1972; Malcolm 1979; Gascoyne et al. 1993; Creel et al. 1995, 1997b; van Heerden et al. 1995; Creel & Creel 1996; M. G. L. Mills & Gorman 1997; Gorman et al. 1998). Some aspects of human effects on wild dog populations are at least conceptually well understood. For example, habitat loss reduces population size, and human-caused mortality is likely to alter survivorship, both of which would increase the risk of extinction. Moreover, despite the strong negative human effects on several sympatric species of large carnivores, most of these species (except wild dogs) often maintain high densities ( Hofer et al. 1993; Creel & Creel 1996). This observation suggests that some aspect of wild dog ecology plays an important role in their vulnerability to local extinction. Because the relationship between extinction risk and ecological factors such as interspecific competition is potentially important but less well understood ( Vucetich et al. 1997 ), our goal is to assess whether intrinsic ecological factors may cause wild dogs to be extinction prone. The extinction simulation program VORTEX (Lacy & Kreeger 1992) has been used to explore the role of chance in the decline of a population of 20–30 wild dogs in Serengeti National Park (Burrows et al. 1994; Ginsberg et al. 1995). VORTEX was also used, in a broader context, to examine the effects of population size, fragmentation, inbreeding, and changes in mortality on extinction risk in wild dogs (Ginsberg & Woodroffe 1997). These simulations suggest that population size (which was varied from 20 to 100 adults) was the most important variable affecting persistence. Extinction risk was more sensitive to changes in the mortality of adults than juveniles. Fragmentation reduced persistence for small populations (i.e., ,50 dogs), unless the population received immigrants. Although these conclusions seem reasonable, results from generalized population viability analysis (PVA) programs (e.g., VORTEX) can be misleading. For example, four generalized population viability programs (including VORTEX) yielded widely divergent estimates of extinction risk for a single grizzly bear population, despite

Conservation Biology Volume 13, No. 5, October 1999

1174

Extinction Risk in African Wild Dogs

efforts to implement the programs as identically as possible (L. S. Mills et al. 1996). Inconsistencies arise because all PVAs depend on assumptions about demography, ecological interactions, and social structure, which vary among PVA programs. Because the structure of a population model has important consequences for its dynamics (Caswell et al. 1997), the match between PVA assumptions and a given species’ population biology strongly affects the plausibility of viability estimates (L. S. Mills et al. 1996; Pascual et al. 1997). Using a generalized PVA model to predict viability when the population’s biology does not fit the PVA model assumptions is analogous to drawing inferences from a statistical test for data that do not meet its assumptions. Although generalized PVA programs provide valuable insight when applied to populations whose biology matches the model assumptions, generalized programs are likely to be inappropriate when, for example, population dynamics are strongly affected by social structure or complex ecological interactions (Boyce 1992; Caughley 1994; Vucetich et al. 1997). Both social structure and complex ecological interactions are important factors in determining the dynamics of wild dog populations. Wild dog social structure—formation and maintenance of stable packs—influences population dynamics through the effect of pack size on foraging success (Creel 1997) and reproductive success (Creel et al. 1997a). The ecological interactions of wild dogs, in the form of interspecific competition with lions and spotted hyenas, affect foraging success, energy balance, habitat selection, and population density (Fanshawe & Fitzgibbon 1993; Creel & Creel 1996; Mills & Gorman 1997; Gorman et al. 1998). Given the importance of these complexities, an individual-based viability model (DeAngelis & Gross 1992) that allows for the incorporation of these ecological and demographic factors represents a more appropriate approach to modeling wild dog population viability. We assessed extinction risk using an individual-based model that mimics the ecological and social characteristics of wild dog populations. The model’s logical structure (e.g., pack membership, rules for dispersal), causal relationships (e.g., pack size and reproduction), and parameter values (e.g., age- and sex-specific survival) are based on demographic and ecological data from a wild dog population in the Selous Game Reserve. Because lion density correlates negatively with wild dog population density (Creel & Creel 1996), and because lion competition and predation are likely to be important components of extinction risk for wild dogs (Mills & Gorman 1997), we structured the model to focus on the effects of interspecific competition with lions. Because infectious diseases may also affect extinction risk (Malcolm 1979; Gascoyne et al. 1993; Creel et al. 1995; Woodroffe et al. 1997), we also analyzed the sensitivity of extinction risk to infectious disease.

Conservation Biology Volume 13, No. 5, October 1999

Vucetich & Creel

Study Population and Parameter Estimation Means and variances for model parameters were based on observations of 366 wild dogs in 11 packs on a 2600km2 site in the north of the Selous Game Reserve, Tanzania (lat 78359S, long 388159E), between June 1991 and October 1996. Over this time period, population density varied from 49 dogs/1000 km2 to 63 dogs/1000 km2. We typically had about 140 individuals under study at one time. We radio-collared two individuals in each pack (Creel et al. 1997a), which allowed us to relocate the dogs for demographic monitoring and behavioral observations. We identified individual wild dogs by variations in their patchy brown, black, and white coats, with the aid of a collection of reference photographs. From field notes we compiled a database with one record per individual for every month in which that individual was seen or inferred to be alive from later observations (for a total of 7317 dog-months). This database contained information on age, reproduction, dispersal status, and social status. From this database we derived a second database summarizing information at the level of packs (number of adults of each sex, number of yearlings of each sex, number of breeding females, number of pups of each sex born, number of pups of each sex surviving at 1 year). With user-written programs, we determined means and variances for age- and sex-specific survival, dispersal risk (method of Waser et al. 1994), pack size and composition, sex ratios, and the parameters for a regression of reproductive success on pack size. We constructed a table summarizing 35 dispersal events (71 female dispersers and 56 male dispersers) and used the table to derive dispersal rules for the simulation.

The Model We constructed an individual-based model to simulate the behavior of an African wild dog population, in which lion population dynamics constitute a strong influence on wild dog dynamics. The fate of each wild dog in the population was traced from birth to death, and extinction occurred when either sex fell to zero individuals. Each individual was characterized by its pack membership, sex, and age. During each year of the simulation, each wild dog had an age-dependent probability of survival (0.59 for pups, 0.78 for yearlings, and 0.80 for adults). A litter was added annually to each pack that held at least one male and one female (i.e., packs composed of a single sex did not produce litters until joined by members of the opposite sex). Litter size was selected randomly from a Poisson distribution whose mean value was determined by pack size. Specifically, the mean litter size was calculated as 1.87 1 (0.51 3 pack size), based on least-squares re-

Vucetich & Creel

gression of litter size on pack size in Selous (t33 5 2.24; p 5 0.03). To produce agreement in age structure between simulated populations and the Selous population, we adjusted the intercept to 2.30 and the slope to 0.63. This adjustment was necessary to make the model internally consistent by producing simulated age structures and pack sizes similar to those observed in the study population. Although the regression is significant, our data on age structure are more precise than the regression coefficients: the adjusted slope and intercept are within the 95% confidence intervals of the least-squares estimates. The sex ratio of each litter was selected from a binomial distribution in which the average proportion of males was 0.57. This proportion was estimated from the study population and is similar in other free-ranging and captive populations (Malcolm 1979; Fuller et al. 1992). Although the mechanisms that determine the number of packs in a population are not fully understood, two patterns suggest an appropriate method for modeling the number of packs. First, lion density has a strong negative correlation with wild dog population density: dogs/km2 5 exp(0.65 2 [45.99 3 lions/km2]) (Creel & Creel 1996). Second, pack size is not detectably related to lion density ( p 5 0.45; n 5 6 populations; simple correlation). Based on these patterns, we predicted the number of packs for each year of the simulation (except the first) by dividing the predicted number of wild dogs (based on lion density according to the immediately preceding expression) by the mean pack size in the Selous population (4.7 yearlings 1 9.0 adults). Mean pack size of the Selous population was used only to determine pack formation or failure; simulated pack size rose independently of this process. As described above, pack size increased according to the relationship between pack size and fecundity and decreased according to age- and sex-specific mortality rates. If the predicted number of packs (rounded to the nearest integer) exceeded the current number of packs in the simulated population, then a new pack was established by one male and one female from the pack(s) with the largest number of males and females, respectively. If the predicted number of packs was less than the actual number, then the smallest pack in the simulated population failed, and each wild dog in that pack dispersed into randomly selected packs. If any pack’s size exceeded 12, all of the yearlings of one sex (chosen randomly) dispersed from that pack into the smallest pack. This dispersal pattern is similar to that observed in Selous and Botswana (McNutt 1996; S. Creel and N. M. Creel, unpublished data). Consistent with observations from Selous, where dispersal mortality was estimated to fall between 0.38 and 0.83 (n 5 127 dispersers; method of Waser et al. 1994), simulated dispersal was accompanied by a 0.5 chance of mortality. Collectively, the rules for recruitment, mortality, and dispersal determined the

Extinction Risk in African Wild Dogs

1175

size and composition of each simulated pack in each year. This approach allowed pack size to be influenced by the fecundity and mortality of the individuals within each pack, whereas the number of packs was influenced by ecological factors thought to affect population density. The appropriateness of these rules was supported by close concordance in size and composition between the real and simulated packs. Because calculating the number of packs required simulated information about lion density, we modeled lion population dynamics according to a diffusion process (Dennis et al. 1991) in which the average lion population size, total range of lion population size, and annual lion population growth rate were directly specified as independent variables. Year-to-year variability in population size was estimated from data from a naturally regulated lion population (i.e., from Ngorongoro Crater, Tanzania; Packer et al. 1991; C. Packer, personal communication). We used this population because it provided the best information available for estimating yearto-year variability in a lion population. Each simulation began with a population of 98 adult and yearling wild dogs divided into seven packs, where each pack initially consisted of 9 adults, 5 yearlings, and 8 pups (mimicking the Selous population). We chose 98 wild dogs as a baseline model because most wild dog populations are about this size or smaller, and initial simulations suggested that, in the absence of ecological or demographic problems, a population of approximately 100 dogs will persist for 100 years with a probability near one. Thus, a population of 98 wild dogs provides a reasonable standard of comparison for evaluating extinction risk in simulated populations subjected to challenges such as increased interspecific competition or disease. The initial sex ratio was determined by randomly assigning each wild dog a sex, where the probability of being male was 0.57. With all parameter values fixed, we calculated the probability of persisting to 20, 50, and 100 years (denoted PP20, PP50, PP100, and generally as PPx ) based on the trajectories of 20,000 simulated populations for each scenario. We evaluated the sensitivity of population persistence by varying one parameter at a time.

Sensitivity Analysis and Results Lion Density We modeled probabilities of persistence in wild dog populations for average lion densities ranging from 50 lions/1000 km2 to 200 lions/1000 km2. This covers the range of lion densities typically observed in nature (Fig. 1). When average lion densities were below 100 lions/1000 km2, the probability of persistence for wild dog populations was near 1.0 for periods of 20–100 years. As average lion density increased from 100 to 140 lions/1000

Conservation Biology Volume 13, No. 5, October 1999

1176

Extinction Risk in African Wild Dogs

Figure 1. The influence of lion density on the probability of a population of 98 wild dogs persisting 20, 50, and 100 years. Arrows indicate lion densities measured in several ecosystems. Heavy arrows indicate ecosystems in which wild dogs have disappeared.

km2, PPx dropped precipitously (Fig. 1). For average lion densities exceeding 140 lions/1000 km2, PPx was virtually zero, even for periods as short as 20 years. To examine the effect of mean lion density with minimal variation in lion density, we held lion density within 7% of the average density for these simulations. To examine the effect of variation in lion density on wild dog population viability, we calculated PPx for constant average lion densities but allowed the range (i.e., maximum–minimum) to vary (Fig. 2). These simulations were repeated for moderate (100 lions/1000 km2) and high (131 lions/1000 km2) lion densities. At moderate lion densities, PPx was insensitive to changes in lion variation for ranges ,30 lions/1000 km2 and was only modestly affected by variation for ranges .30 lions/1000 km2. At high lion densities, PPx dropped precipitously as the range increased beyond 20 lions/1000 km2. Under these conditions, with a high mean and variance, lion density occasionally reached levels that quickly decimated wild dog populations. Although the lion population may subsequently decline, wild dogs may never recover from the population bottleneck. Disease We modeled the influence of three diseases likely to affect wild dog populations: rabies (Gascoyne et al. 1993; Kat et al. 1995), canine distemper virus (CDV; Alexander et al. 1996; Roelke-Parker et al. 1996), and canine par-

Conservation Biology Volume 13, No. 5, October 1999

Vucetich & Creel

vovirus (CPV; Creel et al. 1997c). These simulations were characterized by an annual probability (ranging from 0 to 100%) that the disease would affect individual mortality for that year. Each disease affected mortality in a different manner: rabies affected all age classes equally, CDV primarily affected pups and yearlings and to a lesser extent adults, and CPV affected only pups. Conforming to this pattern, simulations for each disease were further characterized by specified (but arbitrary) increases in mortality (Fig. 3). We varied the effects on survival over wide ranges because estimates of true disease-induced mortality are poor. Because exposure or vulnerablity to these diseases may be greater when the density of interspecific competitors is high (Grenfell & Dobson 1995), we modeled the influence of disease at moderate and high lion densities. The qualitative effect of disease on wild dog populations was similar for moderate and high lion densities (Fig. 3). If the effects of disease and high lion density are additive as we have modeled them, however, extinction risk could be high in the presence of disease and high lion density. As modeled, rabies generally had a much greater effect on extinction risk than CDV or CPV. When CDV or CPV reduced survival by #0.20, extinction risk increased little unless outbreaks occurred frequently. In general, the simulated effect of CPV was substantially less important than the effect of rabies. Better data, however, are needed on mortality consequences of CDV before conclusions can be drawn. Immigration Because extinction risk was high when lion density was high or when infectious disease was severe, we investi-

Figure 2. The influence of temporal variation in lion density on the probability of a population of 98 wild dogs persisting 20, 50, and 100 years in environments with moderate (100 adult lions/1000 km2) and high (131 adult lions/1000 km2) lion density.

Vucetich & Creel

Extinction Risk in African Wild Dogs

1177

Figure 4. The influence of immigration on the probability of a population of 98 wild dogs persisting 50 years in environments with moderate (100 adult lions/1000 km2) and high (131 adult lions/1000 km2) lion density. Each immigrant group consisted, on average, of four same-sexed individuals.

Figure 3. The influence of viral diseases on the probability of a population of 98 wild dogs persisting 50 years in environments with moderate (100 adult lions/1000 km2) and high (131 adult lions/1000 km2) lion density. gated how immigration might mitigate extinction risk. Rather than modeling the dynamics of a neighboring source of immigrants, we simply provided immigrants under the following rules. Consistent with observed dispersal behavior of wild dogs (Frame et al. 1979; McNutt 1995; Waser 1996), each immigrating group consisted of same-sex individuals, where the sex was randomly selected (i.e., the probability of being a male is 0.57). Based on observations in Selous, the size of each immigrating group was selected randomly from a Poisson distribution with a mean of four. We determined the number of groups immigrating each year randomly by drawing from a Poisson distribution with a mean of one. Moreover, immigration occurred only during randomly chosen years to vary immigration rates from an average of 0.1 groups per year to 1 group per year (Fig. 4). Immigrating groups were incorporated into the population in one of three ways. The first option was to form a new breeding pack by joining a group composed of only opposite-sexed dogs, if such a group existed. The second option was to join a pack composed of both sexes but fewer members of the same sex, if such a group existed. In this case, the same-sex pack members would be displaced by the immigrants. The displaced

dogs then dispersed as a group into another pack (following the same rules for immigration), accompanied by a 0.5 risk of death added to normal age- and sex-specific mortality. If options one and two were not available, the third option was to establish a new and independent singlesex group. The new pack (like other packs) was subject to disbanding (within the next simulated year) if lion density was too high to allow an additional pack. When lion density was moderate (100 lions/1000 km2), the probability of persistence to 50 years was high in the absence of immigration, so immigrants had little effect on PP50. When lion density was high (131 lions/ 1000 km2), however, an average of one immigrant group every other year increased PP50 from 0.60 to 0.80, and an average of one immigrant group per year further increased PP50 to 0.98. Thus, a modest amount of connectedness among simulated populations can substantially increase the probability of persistence for wild dogs facing unfavorable ecological and demographic conditions (compare Figs. 1 & 4).

Litter Size Because pack size and litter size are correlated, we modeled the sensitivity of extinction risk to changes in average litter size by altering the y-intercept of the regression of litter size on pack size. In this manner, we estimated extinction risk when litter size varied from approximately 5 to 12 pups per litter. For context, mean litter size was 8.6 6 0.7 SE in Selous, 10.0 in Serengeti

Conservation Biology Volume 13, No. 5, October 1999

1178

Extinction Risk in African Wild Dogs

Vucetich & Creel

National Park, and 11.9 in Kruger National Park (Fuller et al. 1992; Maddock & Mills 1994; van Heerden et al. 1995; Creel et al. 1997a). Under moderate lion densities, reducing mean litter size had little effect on extinction risk until average litter size dropped below 6–7 pups (Fig. 5). Likewise, an increase in mean litter size from 8 to 12 pups per litter produced only a small reduction in extinction risk with high lion density. As mean litter size fell below 6 pups per litter, extinction risk increased substantially for both moderate and high lion densities. Pack Size Because reproductive success is strongly correlated with pack size, we directly examined the influence of pack size on extinction risk by changing the threshold size of the pack at which individuals dispersed. Our model best mimicked the Selous population for a dispersal threshold of 12 adults. Under moderate lion density, a 12-adult threshold was associated with average pack sizes that resulted in a nearly 100% chance of persistence to 50 years, so an increase in mean pack size would do little to decrease extinction risk (Fig. 6). In contrast, extinction risk increased sharply as the dispersal threshold dropped below 10 adults per pack. To illustrate, PP50 dropped from 0.96 to 0.74 as the dispersal threshold decreased from 10 to 6 adults. Under high lion density, 12 wild dogs is near (slightly above) the inflection point in the PP50 curve, so an increase or a decrease in mean pack size would substantially increase or decrease the population’s probability

Figure 6. The influence of average pack size on the probability of a population of 98 wild dogs persisting 50 years in environments with moderate (100 adult lions/1000 km2) and high (131 adult lions/1000 km2) lion density.

of persisting. Above approximately 20 dogs per pack, the curve flattens and population persistence increases little. These results suggest a fairly strong interaction among lion density, pack size, and the persistence of wild dog populations.

Discussion

Figure 5. The influence of average litter size on the probability of a population of 98 wild dogs persisting 50 years in environments with moderate (100 adult lions/1000 km2) and high (131 adult lions/1000 km2) lion density.

Conservation Biology Volume 13, No. 5, October 1999

Our model translates recent insights into wild dog ecology into estimates of extinction risk. Although the simulations represent an appropriate interpretation of available data, the quantitative estimates of risk should be treated with caution. Extinction risk may have been higher or lower if important population processes were left unincorporated. Although parameter estimates were based on 6 field-years of observation, these years may not be representative of population trends over the next 20–100 years. Nevertheless, these simulations provide valuable insight into the relative risk of extinction associated with various ecological and demographic processes. Our finding that lion competition is associated with dramatic increases in extinction risk is corroborated by empirical observations in Kruger (M. G. L. Mills & Gorman 1997), Serengeti (Kruuk 1972; Malcolm 1979), and Selous (Creel & Creel 1996) that interspecific competition and predation by lions increase extinction risk for wild dog populations. Specifically our predictions are consistent with the disappearance of wild dogs from Serengeti and Ngorongoro (Estes & Goddard 1967; Packer et

Vucetich & Creel

al. 1991; Creel & Creel 1996) within a decade of lion densities increasing from low (,100 lions/1000 km2) to high (.140 lions/1000 km2) (Fig. 1). Our simulations and these observations provide compelling evidence that competition with lions is critically important to wild dog conservation because the probability of persistence (PPx ) declines steeply as lion density increases, well within the observed range of lion density. The sensitivity of wild dog population viability to lion density, however, presents a significant challenge for wild dog conservation because accurate and precise population estimates for wild dogs and their competitors (lions and hyenas) are difficult to obtain (Rodgers 1974; Creel & Creel 1996; M. G. L. Mills 1996; Ogutu & Dublin 1998). Because spotted hyenas, like lions, also compete with wild dogs (Kruuk 1972; Creel & Creel 1996; Gorman et al. 1998), it would be desirable to model extinction risk for a wild dog population in competition with spotted hyenas as well as lions. Unfortunately, information on the population dynamics of hyenas is limited enough to prevent us from modeling their population dynamics as we did for lions. Because lion and hyena densities are often positively correlated across ecosystems (Stander 1991), however, disentangling the effects of lions and hyenas on wild dogs will be difficult (Creel & Creel 1996), and modeling these effects independently would probably overestimate extinction risk. This problem clearly highlights the difficulty of estimating extinction risk when multiple ecological interactions operate (Boyce 1992). Whereas lions primarily affect wild dogs through predation, the hyenas’ effect is mainly through food stealing (Gorman et al. 1998). Explicitly incorporating either of these effects into a demographic model is difficult because predation (like most other causes of death) is rarely observed and food loss affects demography indirectly. In our model, the effect of competition with hyenas could perhaps be considered by converting hyenas into “lion-equivalents.” Another complex and potentially important process that we were unable to explicitly account for is the synergism between lion density, transmission rates of disease between species (e.g., CDV: Alexander et al. 1996; Roelke-Parker et al. 1996), and infection rates among wild dogs. Under some circumstances, however, potentially synergistic effects may be indirectly indicated by comparing the PPx curves for moderate and high lion densities in Fig. 3 or by shifting to a higher annual probability of an epidemic for the high lion density curve. In this context, viral diseases had a similar qualitative effect on persistence, regardless of lion density (Fig. 3). Because the PPx curves are nonlinear, however, an interaction effect may vary from weak to strong. Although rabies and CDV are widely thought to progress from domestic dogs to wild carnivores in Africa (Gascoyne et al. 1993; Kat et al. 1995; Roelke-Parker et al. 1996), current data are inconclusive about transmission from do-

Extinction Risk in African Wild Dogs

1179

mestic dogs to one wild species and then to another. Because these interactions may be important, better data on routes and rates of interspecific transmission may be required for accurate assessments of extinction risk (Dobson 1995). Although the lethality of CDV for adult wild dogs represents another uncertainty (Alexander et al. 1996; RoelkeParker et al. 1996; Creel et al. 1997c), we modeled CDV as fatal for juveniles and yearlings but not for adults. If more extensive data indicate that CDV is typically lethal for adult wild dogs, then our CDV simulation will underestimate its effect, and the rabies simulation would provide a better model. For now, modeling three general classes of viral disease is useful: those killing all ageclasses (rabies), yearlings and pups (CDV, with the above caveat), and just pups (CPV). Although our approach to modeling disease by manipulating age-specific mortality is similar to previous modeling attempts (i.e., Woodroffe et al. 1997), we arrived at some conclusions that are strikingly different. For example, one previous attempt (using VORTEX) modeled CDV and rabies as catastrophes that occurred in 3% of years and reduced persistence by 50%, and it revealed essentially no effect of disease on 50-year extinction risk for populations larger than 20 dogs (i.e., Woodroffe et al. 1997). Viral outbreaks in wild dog populations, however, are known to substantially exceed rates of 0.03/ year (e.g., Serengeti with outbreaks of rabies and/or CDV at intervals of ,10 years; Schaller 1972; Malcolm 1979; Gascoyne et al. 1993; Roelke-Parker et al. 1996). Therefore, we modeled the effect of disease for annual probabilities of outbreak ranging from 0 to 1. In our simulations, rabies and CDV decreased in persistence when disease-induced adult mortality exceeded 0.3, and outbreaks occurred at intervals of ,10 years. The Serengeti population of wild dogs was exposed to similar conditions immediately prior to their disappearance (Malcolm 1979; Gascoyne et al. 1993; Burrows et al. 1994). In contrast to previous conclusions, we conclude that rabies and CDV might have substantial effects on 50-year extinction risk for a population of approximately 100 (cf. Fig. 3, with Fig. 5.2 of Woodroffe et al. 1997). Our simulation results, however, were also remarkably similar to some previous conclusions about the role of disease. For example, a previous attempt to model CPV showed that juvenile mortality was reduced in all years and concluded that “in all but the smallest populations [20 dogs], varying juvenile mortality in the region of 50–70% has little effect on population persistence. Above 70%, however, small increases in juvenile mortality generate large changes in population persistence” (Woodroffe et al. 1997:85). Suggesting that the result is robust, our results closely match this conclusion: our simulations, where annual probability of an outbreak is 1.0 (Fig. 3), correspond to the approach taken by Woodroffe et al. (1997). Under these conditions, our simula-

Conservation Biology Volume 13, No. 5, October 1999

1180

Extinction Risk in African Wild Dogs

tions showed little effect of disease-induced mortality in pups until the mortality due to disease reached 0.30, or a total pup mortality of 0.71 (baseline annual mortality for pups was 0.41). Disease-induced pup mortality as high as 0.3–0.4 had little effect when the annual probability of an epidemic dropped below 0.5. Collectively, our results and those of Woodroffe et al. (1997) suggest that population persistence is relatively insensitive to juvenile mortality in wild dogs, unless it is severe and persistent. In sharp contrast, however, simulations for variation in litter size revealed rapid declines in population persistence if mean litter size dropped below six pups. Because mean litter size in Selous is only 18% larger than this threshold, any factor that reduces fecundity or the survival of newborns may be a serious threat to viability. In other populations, mean litter size is large enough that this threshold is of less concern. Because wild dogs produce unusually large litters, pup production is not expected to be a demographic limitation. Consistent with this expectation, increased mean litter size above the range seen in the wild (8.5–11.9) had little effect in reducing extinction risk. Our simulations suggest that populations comprising larger packs (up to about 20 adults) have substantially lower extinction risk at high lion density (Fig. 6). Large pack size was also associated with higher per capita rates of net energy intake (Creel & Creel 1996; Creel 1997). The apparent advantages of large pack size are probably offset, however, by disadvantages associated with pack size. For example, social suppression of reproduction may set an upper limit on pack size because subordinates may be unwilling to remain in a pack in which breeding opportunities associated with achieving social dominance are limited (Frame et al. 1979; Malcolm & Marten 1982; Creel et al. 1997a). As pack size increases, subordinates are less likely to achieve dominance and breed. Thus, individual and group selection may be juxtaposed because behavior appearing to increase individual fitness would reduce population viability. Although we did not directly model human effects on population persistence, human-caused mortality is implicitly included in our estimates of age-specific mortality. Moreover, human effects appear to be weak for wild dogs in Selous but are important elsewhere (Woodroffe & Ginsberg 1998). Nevertheless, the apparent extinctionprone nature of wild dogs is not satisfactorily explained by human effects. In the 1800s, prior to widespread effects of humans on the distribution of carnivores in eastern and southern Africa, Selous (1908) reported that wild dogs were widely distributed but nowhere common. Also, wild dogs exhibit much lower densities than sympatric large carnivores, even when human effects on the other species are known to be strong. For example, snaring imposes heavy mortality on spotted hyenas (Crocuta crocuta) in Serengeti National Park (Hofer et al.

Conservation Biology Volume 13, No. 5, October 1999

Vucetich & Creel

1993), but spotted hyenas outnumbered wild dogs by at least an order of magnitude (Creel & Creel 1996) for several decades. These patterns and our simulations suggest that ecological factors are an important contributor to the extinction-prone nature of wild dogs.

Summary and Recommendations Most extant populations of wild dogs consist of fewer than 100 individuals. Our simulations suggest that population of this size face substantial extinction risk within the next 100 years under a range of realistic ecological and demographic conditions. Field studies and our simulations alike suggest that competition with larger carnivores and viral diseases both pose serious problems for the conservation of wild dogs, particularly for populations with small pack size or small litters. Our conclusions, however, which are based on data from a single population, may change to the extent that other populations differ in their demography. Future attempts to model extinction risk for wild dogs would benefit from better data on the degree to which human-caused mortality is additive or compensatory, on age-specific mortality due to infectious diseases, on transmission rates among species, on the demography of dominant competitors, and on the mechanisms by which competitors affect wild dogs. Although our quantitative estimates of extinction risk should be evaluated cautiously, they suggest that the persistence of many wild dog populations will require active management. For example, viability may be adequately raised by managing for dispersal among populations or by augmenting populations with single-sex immigrant groups of 3–4 individuals at intervals of 1–2 years—a potentially feasible rate of translocation for many populations. In small populations comprising one or two packs (e.g., Hluhluwe), however, translocations of single-sex groups are risky because fights between potential immigrants and residents can be fatal (Frame et al. 1979; S. and N. Creel, unpublished data). For such populations, introductions of socially bonded breeding groups might be preferable. Another obstacle to successful translocation is the identification of source populations whose viability would be unaffected by the removal of dogs for translocation. Because the release of dogs bred in captivity has had mixed success due to their poor abilities to hunt and avoid lions (Scheepers & Venzke 1995; Woodroffe et al. 1997), at least some wildborn dogs are desirable for translocations. Source populations might include the Selous Game Reserve (Tanzania), Kruger National Park (S. Africa), and possibly the population of northern Botswana. Effective wild dog conservation is likely to require active management to alleviate the effects of interspecific competition, to promote dispersal among populations, or to augment population size.

Vucetich & Creel

Acknowledgments This research was supported by grants from the National Science Foundation (IBN-9419452) and the Frankfurt Zoological Society (1112/90). For assistance in the Selous Game Reserve, we thank B. Mbano, G. Bigurube, B. Kibonde, and W. Minja of the Tanzania Wildlife Division. We thank the Tanzania Council for Science and Technology and the Serengeti Wildlife Research Institute for permission to conduct research in Tanzania. J.A.V. was supported by a McIntire-Stennis grant. The computer code used to construct this model was written and compiled with Borland C/C11 software and is available from J.A.V. upon request. Literature Cited Alexander, K. A., P. W. Kat, L. A. Munson, A. Kalake, and M. J. G. Appel. 1996. Canine distemper related mortality among wild dogs (Lycaon pictus) in Chobe National Park, Botswana. Journal of Zoo and Wildlife Medicine 27:426–427. Blackburn, T. M., and K. J. Gaston. 1994. Animal body size distributions: patterns, mechanisms and implications. Trends in Ecology and Evolution 9:471–474. Boyce, M. S. 1992. Population viability analysis. Annual Review of Ecology and Systematics 23:481–506. Burrows, R., H. Hofer, and M. East. 1994. Demography, extinction risk and intervention in a small population: the case of the Serengeti wild dogs. Proceedings of the Royal Society, London, Series B 256: 281–292. Caswell, H., R. M. Nisbet, A. M. de Roos, and S. Tuljapurkar. 1997. Structured-population models: many methods, few basic concepts. Pages 3–18 in S. Tuljapurkar and H. Caswell, editors. Structuredpopulation models in marine, terrestrial and freshwater systems. Chapman and Hall, New York. Caughley, G. 1994. Directions in conservation biology. Journal of Animal Ecology 63:215–244. Clark, T. W., A. P. Curlee, and R. P. Reading. 1996. Crafting effective solutions to the large carnivore conservation problem. Conservation Biology 10:940–948. Creel, S. 1997. Cooperative hunting and group size: assumptions and currencies. Animal Behaviour 54:1319–1324. Creel, S., and N. M. Creel. 1996. Limitation of African wild dogs by competition with larger carnivores. Conservation Biology 10:526–538. Creel, S., and N. M. Creel. 1998. Six ecological factors that may limit African wild dogs. Animal Conservation 1:1–9. Creel, S., N. M. Creel, J. A. Matovelo, M. M. A. Mtambo, E. K. Batamuzi, and J. E. Cooper. 1995. The effects of anthrax on endangered African wild dogs (Lycaon pictus). Journal of Zoology (London) 236: 199–209. Creel, S., N. M. Creel, M. G. L. Mills, and S. L. Monfort. 1997a. Rank and reproduction in cooperatively breeding African wild dogs: behavioral and endocrine correlates. Behavioral Ecology 8:298–306. Creel, S., N. M. Creel, and S. L. Monfort. 1997b. Radiocollaring and stress hormones in African wild dogs. Conservation Biology 11: 544–548. Creel, S., N. M. Creel, L. Munson, D. Sanderlin, and M. J. G. Appel. 1997c. Serosurvey for selected viral diseases and demography of African wild dogs in Tanzania. Journal of Wildlife Diseases 33: 823–832. DeAngelis, D. L., and L. J. Gross, editors. 1992. Individual-based models and approaches in ecology. Chapman and Hall, New York. Dennis, B., P. L. Munholland, and J. M. Scott. 1991. Estimation of

Extinction Risk in African Wild Dogs

1181

growth and extinction parameters for endangered species. Ecological Monographs 61:115–142. Dobson, A. 1995. The ecology and epidemiology of rinderpest virus in Serengeti and Ngorongoro Conservation Area. Pages 485–505 in A. R. E. Sinclair and P. Arcese, editors. Serengeti II: dynamics, management and conservation of an ecosystem. University of Chicago Press, Chicago. Estes, R. D., and J. Goddard. 1967. Prey selection and hunting behavior of the African wild dog. Journal of Wildlife Management 31:52–70. Fanshawe, J. H., and C. D. Fitzgibbon. 1993. Factors influencing the hunting success of an African wild dog pack. Animal Behavior 45: 479–490. Fanshawe, J. H., L. H. Frame, and J. R. Ginsberg. 1991. The wild dog: Africa’s vanishing carnivore. Oryx 25:137–146. Fanshawe, J. H., J. R. Ginsberg, C. Sillero-Zubiri, and R. Woodroffe. 1997. The status and distribution of remaining wild dog populations. Pages 11–57 in R. Woodroffe, J. Ginsberg, and D. Macdonald, editors. World Conservation Union, Gland, Switzerland. Frame, L. H., J. R. Malcolm, G. W. Frame, and H. van Lawick. 1979. Social organization of African wild dogs (Lycaon pictus) on the Serengeti Plains, Tanzania, 1967–1978. Zeitschrift fur Tierpsychologie 50:225–249. Fuller, T. K., P. W. Kat, J. B. Bulger, A. H. Maddock, J. R. Ginsberg, R. Burrows, J. W. McNutt, and M. G. L. Mills. 1992. Population dynamics of African wild dogs. Pages 1125–1139 in D. R. McCullough and R. H. Barret, editors. Wildlife 2001: populations. Elsevier Applied Science, London. Gascoyne, S. C., M. K. Laurenson, S. Lelo, and M. Borner. 1993. Rabies in African wild dogs Lycaon pictus in the Serengeti region, Tanzania. Journal of Wildlife Diseases 29:396–402. Ginsberg, J. R., and R. Woodroffe. 1997. Extinction risks faced by remaining wild dog populations. Pages 11–57 in R. Woodroffe, J. Ginsberg, and D. Macdonald, editors. World Conservation Union, Gland, Switzerland. Ginsberg, J. R., G. M. Mace, and S. Albon. 1995. Local extinction in a small and declining population: wild dogs in the Serengeti. Proceedings of the Royal Society, London, Series B 262:221–228. Gorman, M. L., M. G. L. Mills, J. P. Raath, and J. R. Speakman. 1998. High hunting costs make African wild dogs vulnerable to kleptoparasitism by hyaenas. Nature 391:479–481. Grenfell, B. T., and A. P. Dobson. 1995. Ecology of infectious diseases in natural populations. Cambridge University Press, Cambridge, United Kingdom. Hofer, H., M. East, and K. L. I. Campbell. 1993. Snares, commuting hyenas and migratory herbivores: humans as predators in the Serengeti. Symposia of the Zoological Society of London 65:347–366. Kat, P. W., K. A. Alexander, J. S. Smith, and L. Munson. 1995. Rabies and wild dogs in Kenya. Proceeding of the Royal Society of London, Series B 262:229–233. Kruuk, H. 1972. The spotted hyaena: a study of predation and social behavior. University of Chicago Press, Chicago. Lacy, R. C., and T. Kreeger. 1992. VORTEX users manual. A stochastic simulation of the extinction process. World Conservation Union– Species Survival Commission Captive Breeding Specialist Group, Apple Valley, Minnesota. Maddock, A. H., and M. G. L. Mills. 1994. Population characteristics of African wild dogs Lycaon pictus in the eastern Transvaal Lowveld, South Africa, as revealed through photographic records. Biological Conservation 67:57–62. Malcolm, J. R. 1979. Social organization and the communal rearing of pups in African wild dogs (Lycaon pictus). Ph.D. dissertation. Harvard University, Cambridge, Massachusetts. Malcolm, J. R., and K. Marten. 1982. Natural selection and the communal rearing of pups in African wild dogs (Lycaon pictus). Behavioral Ecology and Sociobiology 10:1–13. McNutt, J. W. 1995. Sociality and dispersal in African wild dogs, Lycaon pictus. Ph.D. dissertation. University of California, Davis.

Conservation Biology Volume 13, No. 5, October 1999

1182

Extinction Risk in African Wild Dogs

McNutt, J. W. 1996. Sex-biased dispersal in African wild dogs (Lycaon pictus). Animal Behavior 52:1067–1077. Mills, L. S., S. G. Hayes, C. Baldwin, M. J. Wisdom, J. Citta, D. J. Matson, and K. Murphy. 1996. Factors leading to different viability predictions for a grizzly bear data set. Conservation Biology 10: 863–873. Mills, M. G. L. 1996. Methodological advances in capture, census, and food-habits studies of large African carnivores. Pages 223–242 in J. L. Gittleman, editor. Carnivore behavior, ecology and evolution. Volume 2. Cornell University Press, Ithaca, New York. Mills, M. G. L., and M. L. Gorman. 1997. Factors affecting the density and distribution of wild dogs in the Kruger National Park. Conservation Biology 11:1397–1406. Ogutu, J. O., and H. T. Dublin. 1998. The response of lions and spotted hyaenas to sound playbacks as a technique for estimating population size. African Journal of Ecology 36:83–95. Packer, C., A. E. Pusey, H. Rowley, D. A. Gilbert, J. Martenson, and S. J. O’Brien. 1991. Case study of a population bottleneck: lions of the Ngorongoro Crater. Conservation Biology 5:219–230. Pascual, M. A., P. Kareiva, and R. Hilborn. 1997. The influence of model structure on conclusions about the viability and harvesting of Serengeti wildebeest. Conservation Biology 11:966–976. Rodgers, W. A. 1974. The lion (Panthera leo) population of the eastern Selous Game Reserve. East African Wildlife Journal 12:313–317. Roelke-Parker, M. E., L. Munson, C. Packer, R. Kock, S. Cleaveland, M. Carpenter, S. J. O’Brien, A. Popischil, R. Hofman-Lehman, H. Lutz, G. L. M. Mwamengele, M. N. Mgasa, G. A. Machange, B. A. Summers, and M. J. G. Appel. 1996. A canine distemper virus epidemic in Serengeti lions (Panthera leo). Nature 379:441–445. Schaller, G. B. 1972. The Serengeti lion. University of Chicago Press, Chicago. Schaller, G. B. 1996. Carnivores and conservation biology. Pages 1–10

Conservation Biology Volume 13, No. 5, October 1999

Vucetich & Creel

in J. L. Gittleman, editor. Carnivore behavior, ecology and evolution. Volume 2. Cornell University Press, Ithaca, New York. Scheepers, J. L., and K. A. E. Venzke. 1995. Attempts to reintroduce African wild dogs Lycaon pictus into Etosha National Park, Namibia. South African Journal of Wildlife Research 25:138–140. Selous, F. C. 1908. African nature notes and reminiscences. Macmillan, London. Stander, P. E. 1991. Aspects of the ecology and scientific management of large carnivores in sub-Saharan Africa. Ph.D. dissertation. Cambridge University, Cambridge, United Kingdom. van Heerden, J., M. G. L. Mills, M. J. van Vuuren, P. J. Kelly, and M. J. Dreyer. 1995. An investigation into the health status and diseases of wild dogs (Lycaon pictus) in the Kruger National Park. Journal of the South African Veterinary Association 66:18–27. Vucetich, J. A., R. O. Peterson, and T. Waite. 1997. Effects of social structure and prey dynamics on extinction risk in gray wolves. Conservation Biology 11:957–965. Waser, P. M. 1996. Patterns and consequences of dispersal in gregarious carnivores. Pages 267–295 in J. L. Gittleman, editor. Carnivore behavior, ecology and evolution. Volume 2. Cornell University Press, Ithaca, New York. Waser, P. M., S. R. Creel, and J. R. Lucas. 1994. Death and disappearance: estimating mortality risks associated with philopatry and dispersal. Behavioral Ecology 5:135–141 Wayne, R., E. Geffen, D. J. Girman, K. Koepfli, L. M. Lau, and C. R. Marshall. 1997. Molecular systematics of the Canidae. Systematic Biology 46:622–653. Woodroffe, R., and J. R. Ginsberg. 1998. Edge effects and the extinction of populations inside protected areas. Science 280:2126–2128. Woodroffe, R. B., J. R. Ginsberg, and D. M. Macdonald. 1997. African wild dogs: an action plan for their conservation. World Conservation Union, Gland, Switzerland.