The IUCN/SSC Canid Specialist Group's
African Wild Dog Status Survey and Action Plan (1997)

Chapter 5
Extinction Risks Faced by Remaining Wild Dog Populations

Joshua R. Ginsberg & Rosie Woodroffe

In this chapter we use demographic modelling to assess the probability that the threats to wild dog populations outlined in Chapter 4 might cause local extinction of remaining populations. In constructing our model:

• We use real data on wild dog biology to develop a standard model.
• We estimate that the majority of extant wild dog populations contain £50 individuals.
• Rather than attempting to simulate specific wild dog populations, our models reflect the size of remaining wild dog populations (K ~ 20, 50, 100 animals).
• We employ timescales which reflect the true pace of land use change in Africa.
• We examine how population size, fragmentation and inbreeding depression affect the probability of local extinction.
• We assess the degree to which both small and larger populations are affected by changing patterns of adult and juvenile mortality to simulate the impact of threats such as persecution, locally endemic and epidemic disease, road accidents, snaring and lion predation.

It is not our intention to define a minimum size below which populations are likely to become extinct: neither our model, nor the data used to parameterize the model, are adequate to allow such quantitative predictions. The following general conclusions can, however, be valuable for planning management strategies:

Larger populations (~100 individuals) appear remarkably resilient. Wild dogsµ large litters allow them to bounce back from catastrophes which cause temporary declines in population numbers. Given protection from fragmentation and a barrage of multiple threats, these populations should persist over the next 50 years. However, such populations require very large areas (>5000km²). As human populations rise and the African landscape becomes more fragmented, populations of this size will surely disappear without active landscape planning to ensure the integrity and contiguity of current protected areas and wildlife lands.

Smaller populations (~50 individuals) characterize many remaining wild dog populations. Insulated from threats, such populations stand a decent chance of persisting for the next 50 years. They are, however, extremely vulnerable to change: a small increase in either adult or juvenile mortality greatly increases the probability of extinction. Thus direct persecution, disease, road accidents, accidental snaring and lion predation each represents a serious threat to populations of this size. Increasing connectivity to form larger metapopulations will help such populations to persist.

Tiny populations (~20 individuals), consisting of just a few packs, face a high probability of extinction. Whether they are remnants of a once larger population, or populations newly founded by reintroduction, tiny populations will be vulnerable to any threat which increases either adult or juvenile mortality. Such populations may occupy relatively large areas (>500km²) but are constrained in their ability to grow. Connecting these tiny populations to larger populations greatly improves their persistence.

Background

In the previous chapter we outlined factors that may cause wild dog numbers to decline, or even drive them to local extinction. Setting priorities for wild dog management, however, demands an assessment of the relative importance of these threats. For example, if accidental capture of adult wild dogs in snares is a major cause of mortality, then better control of snaring inside and outside protected areas could help to protect them. But how does one rank the risk of such snaring with the threat of disease? And is it more important to invest in controlling epidemic rabies or endemic parvovirus? In this chapter we use demographic modelling (Boyce 1992) to simulate how mortality caused by various threats affects wild dog populations, and use these analyses to assess the extinction risks faced by populations of various sizes.

We have chosen to model wild dog populations by using the computer package VORTEX (Lacy et al. 1995). VORTEX was developed as a tool for conservation biologists to assess the probability of extinction in small populations. The user specifies a series of population parameters, and the program then uses a modified Leslie matrix to simulate population changes over time, incorporating stochastic variation in those parameters. By running each simulation many times, one can measure the probability that a population will persist under a given set of demographic circumstances. By varying the starting conditions, the user can simulate various factors likely to affect the population's viability, such as its size, degree of fragmentation, inbreeding depression, harvesting, consistent changes in mortality or breeding success, and episodic 'catastrophes'.

The use of such simulations to assess the risk of extinction faced by wildlife populations - termed population viability analysis (PVA) - has been criticized recently because it considers only genetic and demographic effects. Such effects may operate on a timescale of centuries, while habitat loss and persecution can drive a species to extinction within a few decades (Harcourt 1995). We have attempted to make our simulations more meaningful by measuring the cumulative probability of extinction per decade, over a total of 50 years for each simulation. This allows us to assess the impact of various threats to wild dogs on a timescale which reflects the true pace of change of land use in Africa. Rather than using simulations to define the size of a minimum viable population, we are concerned with assessing the relative impacts of various threats upon wild dog populations in an attempt to set priorities for their management. Under these circumstances, PVA can provide an extremely valuable tool in conservation biology (Boyce 1992; Caughley 1994; Harcourt 1995).

Setting Model Parameters

Our modelling exercise required a set of parameters to describe the characteristic features of wild dog populations. We derived demographic parameters using a combination of published and unpublished data on free-ranging populations of wild dogs. Published data were taken mainly from Fuller et al. (1992); unpublished data were collected from wild dog researchers at the IUCN/SSC Canid Specialist Group's 'Workshop on the Conservation & Recovery of the African Wild Dog', held in Arusha, Tanzania, in 1992, and through subsequent correspondence. These data allowed us to determine both the average values and the degree of variation in population parameters such as adult and juvenile mortality, litter size, birth sex ratios and the proportion of females breeding. These parameters are summarized in Table5.1. While many of the data are straightforward, some deserve further discussion.

Population Size

Because threats vary in both space and time, and because we lacked data on the rôle of known threats in regulating wild dog numbers in known populations, we did not attempt to simulate specific wild dog populations. Taking into account the range of sizes of wild dog populations remaining in Africa, we examined the impact of the various factors in populations of three different sizes chosen to reflect the lower end of the range (and therefore the most threatened) of existing populations: tiny (20); small (50); and larger (100). In Table 5.2 we list our estimates of population size for each known wild dog population in Africa, as a guide to determining how our model results relate to real populations.

Mating System

VORTEX contains no direct provision for the inclusion of social structure within population models. While some have therefore questioned the use of VORTEX for modelling wild dog populations (Heinsohn 1992) many aspects of social structure can be incorporated in the demographic parameters that are defined by the user.

Because only one female usually breeds in each pack (Chapter1), the number of breeding females will be determined, for the most part, by the number of packs in any given population. A factor that increases optimal pack size, thus reducing the number of packs, will therefore reduce the proportion of females breeding in the population as a whole. We simulated the social suppression of reproduction in subordinate female group members by including only 58% of adult (>3 years) females in the breeding pool. This gives a good approximation to the proportion of females breeding in real wild dog populations (Burrows 1995; Fuller et al. 1992). While one might expect this variable to have a relatively strong effect upon population persistence, both sensitivity analyses of VORTEX models (Burrows et al. 1994) and a deterministic Leslie matrix model based upon our parameters (G.Mace pers. comm.) suggest that survivorship of adults and juveniles are far more important.

In contrast with earlier simulation models of wild dog populations (Burrows et al. 1994; Ginsberg et al. 1995), we included 100% of adult males in the breeding pool. All adult male wild dogs are capable of breeding, but usually only the dominant male mates with the dominant female in each pack. Thus, approximately 40-60% of adult males fail to breed because they are socially suppressed (Frame et al. 1979; Girman et al. in press). Our model simulated this situation by assuming that mating was monogamous: the proportion of males breeding was therefore determined by the number of females breeding. The converse situation - where the number of breeding males limits female reproduction - is unlikely to occur because at all times there is a surplus of reproductively capable males waiting for the chance to breed should a dominant male die. Simulation models will ignore this effect if they restrict the proportions of both males and females that breed. Such models will therefore overestimate the probability of extinction, especially in small populations: in a population with a carrying capacity of 50, reducing the proportion of males in the breeding pool from 100% to 40% nearly doubles the estimated probability of extinction within 50 years, from 2% to 5%.

Density Dependence

Our simulations assumed that breeding is independent of population density (although there has been debate about whether this is universally true, Burrows et al. 1995; Ginsberg et al. 1995). Female wild dogs' reproductive success is density dependent at one level: a smaller proportion of females breeds in larger packs. However, in an unconstrained population breeding is unlikely to be density-dependent at the population level because animals which cannot breed disperse and attempt to form new packs (Burrows 1995; Fuller et al. 1992). In small areas this is likely to lead them into unsuitable habitat where they cannot survive. We therefore considered it more appropriate to truncate population size above a certain carrying capacity - denoted by the letter 'K' - than to simulate density-dependent reproduction. We assumed that the population was at carrying capacity at the start of each simulation.

Modelling Results

In the previous chapter we outlined a series of factors likely to affect wild dog numbers - these are summarized in Table 5.3. We modelled most of these threats by incorporating temporary or sustained changes in adult or juvenile mortality into the VORTEX simulations. We also simulated population fragmentation by using a metapopulation model which broke the population into a number of sub-populations, allowing animals to move between sub-populations, and to re-establish extinct sub-populations.

Inbreeding Depression

Although small populations are expected to face problems associated with inbreeding depression, there is surprisingly little evidence to suggest that inbreeding has deleterious effects in most social carnivores. Indeed, Ralls et al. (1988) found that juvenile survival in captive wild dogs increased with the level of inbreeding. The reasons for this relationship are unknown, although there are alternatives to the interpretation that inbreeding is beneficial.

Figure 5.1. The effect of incorporating inbreeding depression, caused by lethal recessive alleles, into population simulations. The cumulative probability of extinction is given for model populations which either include or exclude inbreeding depression, for carrying capacities of (a) 20, (b) 50, and (c) 100 wild dogs.

The best evidence for a deleterious effect of inbreeding in communally breeding canids comes from a study of wolves held in captivity (Laikre & Ryman 1991). In this study, founders taken from a small wild population were found to carry a deleterious recessive gene for blindness - an allele which would certainly prove fatal in the wild. This study shows that recessive lethal alleles can persist, even in small populations. In the light of these data, we incorporated a recessive lethal model of inbreeding into our simulations, rather than a more general inbreeding depression model to reduce the survival of highly homozygous juveniles (Lacy et al. 1995).   Using this model, our simulations suggest that inbreeding has a small but measurable effect upon the persistence of wild dog populations. Figure 5.1 shows the probability of extinction of populations of three sizes (K=20, 50, 100) simulated using our basic model, including and excluding the effects of inbreeding depression. In tiny populations (K=20, Figure 5.1a) our simulations show that inbreeding depression has a moderate effect on persistence, increasing the probability of extinction within 50 years from 36% to 41%. For a population with a carrying capacity of 50 animals (Figure 5.1b), the addition of inbreeding depression raises the probability of extinction from 2% to 4%. In larger populations, the effects of inbreeding are negligible (Figure 5.1c). A note of caution: when a monogamous mating system is defined in VORTEX, mates are chosen randomly in each year of a simulation, while in wild dog packs, a dominant male and female may breed together for a number of years. VORTEX will therefore underestimate the negative impact of inbreeding, because the proportion of adults contributing to each successive generation will be greater than in the real world. On the other hand, because wild dogs appear to selectively outbreed in the wild (Chapter 2, Girman et al. in press) random assignment of mates may not be too great an overestimate the effect of inbreeding. Other factors will also influence the impact of inbreeding on our simulations: for instance, by allowing 100% of males to breed we further underestimate the potential impact of inbreeding, particularly in small populations. We acknowledge the limitations of VORTEX in this regard, but for the sake of completeness, we retained inbreeding depression in our basic model.

Catastrophes

Catastrophes, as defined by VORTEX, are episodic effects which occasionally depress survival or reproduction. We included two types of catastrophes in our basic model. The first, a 'mild' catastrophe, was devised to simulate the effects of environmental factors such as drought or episodic human persecution. These 'mild' catastrophes reduced adult survival for one year by a factor of 0.85 (i.e. a 15% reduction), and reduced breeding by a factor of 0.5. Our default model included a 5% chance that such a 'mild' catastrophe would occur in any one year (i.e. they occur, on average, every 20 years). Calibrating this type of catastrophe against observed data is difficult, but reproductive failure through environmental effects such as flooding (Malcolm & Marten 1982), through persecution (Ginsberg, Unpublished data), or other causes is not uncommon.

We included a second, 'severe', catastrophe type to simulate the effects of epidemic disease. 'Severe' catastrophes had no effect upon breeding, but reduced adult survival by 50%. Our model included a 3% chance of such a 'severe' catastrophe in any one year. This level of mortality represents an average loss over an array of diseases such as canine distemper and rabies (Chapter4). The cyclicity of such infections will vary with a number of factors (Dobson & Hudson 1995) and, while few empirical data are available, catastrophic die-offs are often of this magnitude (Young 1994).

The effects of 'mild' and 'severe' catastrophes, and of the two in combination, are shown in Figure5.2. The effect of either or both catastrophes is surprisingly unimportant in model populations of 50 or above (Figures5.2b & c). Presumably, the remarkable fecundity of wild dogs allows them to recover rapidly from such short-term perturbations. In tiny populations (K=20, Figure5.2a), however, catastrophes can be devastating. As expected, 'severe' catastrophes have a greater impact upon population persistence than do 'mild' catastrophes: the probability of extinction is 13% within 50 years when only 'mild' catastrophes occur, compared with 20% if only 'severe' catastrophes are included in the model, and 40% if both types of catastrophe are incorporated.

VORTEX only allows the user to define a stochastic probability with which catastrophes occur. Clearly, in small populations, the frequency of catastrophes, and the length of the interval between catastrophes, is critical to determining how they will affect the probability of population extinction. Indeed, Ginsberg et al. (1995) found a non-linear increase in the probability of extinction over 25 years as the number of catastrophes increased.

Population Fragmentation

Wild dogs persist only in areas where human population density is low (Chapter3). As a result, many wild dogs have become isolated in parks or other protected areas, with only limited exchange between populations. We investigated the effects of such fragmentation by simulating two sub-populations linked by dispersal. While animals may move between the simulated sub-populations, VORTEX assumes that stochastic effects such as catastrophes influence each sub-population independently. This assumption may be invalid in many circumstances.

In Figure 5.3 we compare the persistence of a single population with that of a fragmented metapopulation. Each metapopulation is composed of two sub-populations, with a combined size equal to that of the single population. For example, we compare the persistence of a single population of 50 animals with that of a metapopulation made up of two populations of 25. Figure 5.3 shows that tiny populations are more likely to become extinct when they are fragmented than when they remain intact: the probability that a population of 20 animals will become extinct within 50 years rises from 41% to 74% when it is divided into two sub-populations of 10 animals each. This is to be expected: since fragmentation reduces the functioning size of each sub-population, it can lead to increases in both inbreeding and the impact of stochastic effects, making sub-populations more likely to die out despite the opportunity for exchange between them.

In contrast, larger populations persist as well - or even marginally better - when they are fragmented. The probability that a population of 50 animals will become extinct falls slightly from 4% to 2% when it is divided into two sub-populations of 25 each (Figure 5.3). This is not entirely surprising. If sub-populations face different threats, or similar threats at different times, then fragmentation may reduce the probability of metapopulation extinction: a series of catastrophes can cause one sub-population to become extinct, but animals from the other sub-population can re-colonize the extinct sub-population. Extinction/recolonization metapopulation dynamics appear to be relatively unimportant in larger metapopulations (K=100) with both fragmented and cohesive populations having high persistence (Figure 5.3).

While the persistence of a larger metapopulation may not be seriously affected by fragmentation (as long as sub-populations remain linked), smaller populations within a metapopulation matrix gain tremendously by being linked together. The value of linking small populations can be seen by examining the persistence of a tiny (K=25) population under three scenarios: alone, linked to another population of K=25, or linked to another population of K=75 (Figure5.4). An isolated population of K=25 has a 13% probability of extinction within 50 years, but this probability falls to 8% if that population is linked to another of K=25, and drops still further to less than 1% when it is linked to a population of K=75. Linking smaller sub-populations into a single metapopulation gives them the persistence profiles of larger populations.

As for all modelling exercises, the value of this finding depends upon the validity of its assumptions. In this case, the important assumption is that catastrophes affect the sub-populations independently. The reason why populations of 50 to 100 individuals persist relatively well when fragmented is that while each sub-population is more likely to become extinct, in most cases the other sub-population persists and re-colonizes the first. However, in the real world, extinction risks within different parts of the same metapopulation are unlikely to be independent. For example, it is very unlikely that linked populations would experience dramatically different weather conditions: a drought that affected one sub-population would also be likely to affect the other. A similar argument can be applied to the effects of epidemic disease. Domestic dogs constitute the reservoir host for many diseases that threaten wild dogs (Chapter 4). Wild dogs may be largely confined to islands of low human population density, but the areas between such sub-populations are likely to contain more-or-less contiguous populations of domestic dogs. If an epidemic disease spread from domestic dogs to one part of a wild dog metapopulation, it would also be likely to affect the other sooner or later. In addition, wild dogs themselves could carry infection from one part of a metapopulation to another (as may have occurred in the last population of black-footed ferrets, Seal et al. 1989). It seems likely, therefore, that absolute size of a population, or metapopulation, is the single most important variable in the persistence of wild dog populations, and we would certainly not advocate population subdivision as a management strategy. Indeed, every effort should be made to maximize the continuity of habitat available to wild dogs.

Threats which Increase Adult Mortality

Several of the threats summarized in Table 5.3 affect wild dogs by increasing the mortality of animals more than a year old (N.B. in this section we refer to such animals as 'adults', although the model defines separate survival probabilities for yearlings and two-year olds to reflect increased probability of mortality during dispersal). Predation by lions, road traffic accidents, snaring and direct persecution all act in this way. We therefore investigated the effect of sustained changes in adult mortality upon the persistence of simulated wild dog populations.

The results are shown in Figure5.5, and point to some important effects. First, a small drop in adult mortality generates a marked reduction in the probability that very small populations will become extinct: in a population of 20 animals, reducing adult mortality by a step of 5% causes the probability of extinction within 50 years to fall from 41% to 13% (Figure5.5a). This effect essentially disappears in larger population (K=50), where the same reduction in mortality brings the probability of extinction down from 0.3% to zero (Figure5.5b). Perhaps more important, however, is the finding that increasing adult mortality can have dramatic effects upon the probability that even larger populations will become extinct. For example, if adult mortality rises by a step of 10%, the probability that a population of K=50 will become extinct within 50 years increases from close to zero to 7% (Figure5.5c).

These findings have two important implications for the assessment of threats to real wild dog populations. First, small populations are extremely sensitive to changes in adult mortality. Essentially, in a tiny population every adult will be important in ensuring persistence. The management of such populations - which will include those re-established by reintroduction - will therefore demand that factors which kill adults be minimized. This will mean that measures must be taken to control persecution, road kills and snaring. Lion predation may also represent a very serious threat to tiny populations - lions can cause up to 47% of adult mortality (Table5.3). While little can be done to control lion predation in free-ranging wild dogs, reintroduction attempts may be more successful in areas which are free of lions. Indeed, lion predation has foiled at least two reintroduction attempts in the past (Chapter7).

A more important finding, however, is that sustained increases in adult mortality will threaten large populations as well as smaller ones. Thus changes in land use which lead to higher adult mortality - such as the opening of new tarmac roads through national parks, rising human population density generating more intense persecution of wild dogs, or even changes in carnivore management leading to marked increases in lion density - could drive populations of 100 or more wild dogs to extinction.

Threats which Increase Juvenile Mortality

A number of the threats summarized in Table5.2 affect the mortality of wild dog pups. Juvenile mortality varies substantially within and between populations (Fuller et al. 1992). We therefore varied the levels of juvenile mortality in our simulated populations in 5% increments between 50% and 80%. The results - which are shown in Figures5.6 & 5.7 - indicate that persistent changes in juvenile mortality can have a marked effect upon the viability of wild dog populations, even those which are reasonably large.

In all but the smallest populations (Figure 5.6a), varying juvenile mortality in the region 50-70% has little effect upon population persistence. Above 70%, however, small increases in juvenile mortality generate large changes in population persistence. For example, in a population of K=50, increasing juvenile mortality from 70% to 80% raises the probability of population extinction within 50 years from 1% to 24% (Figure5.6b). Likewise, the same increase in juvenile mortality in a population of K=100 causes the extinction probability to rise from less than 1% to 9% (Figure 5.6c). These 'threshold' effects of increasing juvenile mortality on population persistence are shown more clearly in Figure 5.7.

These simulations point to two important conclusions. First, although our 'mild catastrophe' models indicate that episodic reductions in the number of pups born have relatively little impact upon population persistence, a persistent change in juvenile mortality has a much more marked effect. Factors which cause short-term breeding failure, such as epidemic diseases affecting only pups, or flooding of dens, are therefore unlikely to drive populations to extinction, but more long-term effects could be devastating.

A second important finding of our simulations is that average juvenile mortality, at 68%, falls just below the threshold where population persistence starts to decline. This means that even relatively small increases in pup mortality could be sufficient to drive some populations to extinction if new causes of mortality act in addition to existing ones. Changes such as the introduction of diseases which kill pups but rarely adults (e.g. parvovirus), or falling prey densities leading to frequent breeding failure, could therefore contribute to the extinction of even relatively large wild dog populations. Consistent increases in pup mortality would also be generated by opening new high-speed roads in wild dog areas, poor control of snaring, and increasing lion predation (Table5.3). All of these factors would also affect adult mortality, causing even more marked effects upon population persistence.

Conclusions

A number of patterns emerge from this modelling exercise. Perhaps the most important conclusion to be drawn is that wild dog populations appear to be remarkably resilient. With their large litters, wild dogs have a high reproductive potential and can, in principal, bounce back from perturbations if their populations are not reduced too far. Our simulations indicate that 'catastrophes' having dramatic short-term effects on breeding and survival affect the persistence of only the smallest populations.

In stark contrast, consistently high mortality of adults or pups can generate an abrupt increase in the probability that simulated populations become extinct. High juvenile mortality negates the effect of high fecundity and prevents wild dogs from bouncing back from perturbations. Thus while wild dog populations are resilient to short-term perturbations, factors which cause consistent increases in adult or juvenile mortality could represent very serious threats.

Our modelling suggests that inbreeding depression is unlikely to have a substantial effect upon most wild dog populations. Indeed, wild dogs have a mechanism for avoiding inbreeding and, probably as a result, large populations show fairly high levels of heterozygosity (Chapter2). While it has been suggested that inbreeding avoidance (rather than inbreeding depression) might halt breeding in small populations (Maddock 1996), this has not been demonstrated: relatives breed together readily in captivity (J.van Heerden pers. comm.), and inbreeding has been recorded once in the wild (Reich 1978). Our simulations suggest that environmental and demographic effects are more important than inbreeding depression in driving small populations to extinction; this appears to be a general pattern in the biology of small populations (Lande 1988).

As expected, larger populations (>100 animals) are best able to persist in the face of threats. Populations of this size remain in extensive tracts of land with low human population density, inside protected areas such as Selous (n>800) and Kruger (n>300), in areas which are either mostly privately or communally held such as north-east Namibia (n>400), or in matrices of protected and communal land as found in northern Botswana (n>400). Since populations of this size are likely to persist if they can be protected adequately, their importance for wild dogs' long-term survival cannot be stated too highly.

Small populations (~50 animals) remain resilient to perturbation, and stand a high chance of persisting if they are well protected. They are, however, very sensitive to consistent increases in adult and juvenile mortality. Factors such as persecution, road accidents, accidental snaring, endemic disease and lion predation will therefore represent very serious threats to such populations. Many of Africa's remaining wild dog populations are about this size (Table5.2), and most inhabit unprotected areas or relatively small protected areas with a correspondingly high perimeter:area ratio. This will bring these animals into contact with human activity. As a result, the populations which are exposed to the most severe threats are likely to be the smaller populations least able to withstand them.

Tiny populations (~20 animals) are still more vulnerable. With so few animals, every individual becomes important in ensuring the survival of the population, so that protection must be intense. All smaller populations stand a much better chance of survival if they can be linked by dispersal to other populations.

These conclusions must be accompanied by a note of caution. Because we have considered each factor independently, in some ways this modelling exercise is extremely conservative and underestimates the extinction risks threatening wild dogs. In the real world, increasing human population density, concomitant increases in the number of domestic dogs and livestock, and resultant reductions in the number of wild prey, would lead simultaneously to increases in threats such as persecution, road casualties and disease. This conservative approach is somewhat mitigated, however, by our assumption that threats themselves are statistically independent of one another, and that an increase in one form of mortality will not lead to a compensatory decrease in another form of mortality. Furthermore, while VORTEX is adequate to enunciate patterns and differences, for modelling to be prescriptive, rather than merely informative, we would advocate a detailed, demographically and spatially structured model be developed for wild dogs.

With these caveats, our results indicate wild dog conservation demands the maintenance of relatively large (>100 individuals) and inter-connected population. To do this, the decline of some populations must be halted through better protection, while ensuring that future development is both zoned and implemented in such a way as to define areas where wild dogs, and other wildlife, can survive. The statement that protecting wild dogs must involve keeping their numbers high may sound like a truism, but this represents a serious conservation challenge for a species that occurs at such low densities. Specific conservation measures for wild dog populations of all sizes are discussed in detail in the next chapter.

Return to Wild Dog Action Plan Table of Contents, AWD Species Account, CSG Publications or CSG Home Page.

© 1997 International Union for the Conservation of Nature and Natural Resources.