As we indicated at the beginning of this chapter, simple predator-prey models inevitably predict an oscillatory interaction between predators and their prey. However, by adding
10 20 30
Number of generations
Figure 10.27 Host-parasitoid interaction with searching efficiency a = 0.03. Result is a stable point.
Figure 10.28 Host-parasitoid interaction with a searching efficiency a = 0.03. Result is a stable point.
a density-dependent factor for the prey population, and some version of a functional response by the predator, depending upon the values of the parameters, these models predict a range of responses, from extinction to cycles to stable points. Adding a carrying-capacity term for the predator, independent of the prey population, may also help stabilize a predator-prey interaction. More realistic models would include the ability of the predator to choose from among several prey species and/or among age classes within a species. With these thoughts in mind we will now examine several examples of real-life predator-prey interactions.
The role of predators in controlling bird and mammal populations has had a controversial history. As discussed earlier, based on his studies of rodent populations in England, and after reviewing data on snowshoe hare and lynx populations, Charles Elton (1924) believed that predator-prey interactions tend to be oscillatory (display a stable limit cycle) and that predators often control prey populations. At one point Peterson et al. (1984), based on their experiences with the wolf and moose populations on Isle Royale in Michigan, seem to have agreed. On the other hand, Errington (1946) and his followers thought that predators mostly took a "doomed surplus" of the prey population. However, for 40 years following Errington's work, there had been few studies in which vertebrate predators were manipulated to discover their effect on vertebrate prey (but see NRC 1997).
Lindstrom et al. (1994) published data based on a "natural experiment" in which a Scandinavian red fox (Vulpes vulpes) population underwent a severe sarcoptic mange epidemic. The mange mite (Sarcoptes scabiei) appeared in Sweden in 1975 and had a dramatic effect on fox populations throughout the 1980s. The mite caused hair loss, skin deterioration, and death, and the fox populations declined by 70%. This allowed Lindstrom et al. to survey the effects on the small-game community of a dramatic drop in a top-carnivore population.
The "alternate prey" hypothesis states that certain prey species are cyclic (3-4 years in voles and lemmings, 10 years in snowshoe hares, for example) and are mostly uncontrolled by predation. Such species are little affected by predation and the oscillations are likely due to interactions with the vegetation (Turchin and Batzli 2001), while other species are controlled by predation. In years when these cyclic prey numbers are in the low phase of their cycle, predators switch to "alternate prey." These alternate prey species are usually limited by predation, especially during the years when the cyclic prey species are in the low phase. In Scandinavia, Lindstrom et al. (1994) proposed that while voles are largely uncontrolled, hare, grouse, and other small mammals and birds are alternate prey species subject to control by predation.
Lindstrom et al. (1994) therefore examined the effect of the reduced fox populations on voles, hares, grouse, and deer. Methods were as follows: (i) rodent populations were surveyed by snap-trapping; (ii) the investigators checked red fox reproduction by entering dens and determining litter sizes; (iii) populations of hare, grouse, and deer were checked by the pellet-count method; (iv) bird populations were estimated by using imitation birdcalls and recording responses; (v) hunters were given questionnaires concerning populations of all of the above and hunting records were consulted. Fawn/doe ratios were estimated from "incidental observations."
In the 1970s, previous to the mange introduction, red fox litter sizes followed the vole (Cricetidae) densities, increasing during the vole peaks. The fox population declined during the mange epidemic and then recovered after 1990 to the level before the mange appeared. Vole populations were not affected by fox numbers. They continued their normal cycles before, during, and after the fox mange epidemic. Lindstrom et al. hypothesized that these cycles were in response to variations in plant productivity and snow depth. Mountain hare
(Lepus timidus) and grouse populations went up 40-100% during the fox mange epidemic and their populations dropped as the fox populations recovered.
Previous to the mange epidemic, black grouse (Tetrao tetrix) populations peaked one year after the peak of vole density. As foxes concentrated on consuming voles, predation on the hares dropped, and hare numbers increased accordingly. European hare (L. europaeus) and black grouse populations followed the same patterns, consistent with the alternate-prey hypothesis.
For roe deer (Capreolus capreolus) the fox population affected the mortality of fawns. After the mange epidemic the number of fawns per doe increased 30% and the average deer density went up 64%. However, the production of fawns was also correlated with the previous year's winter weather. With a delay of 1.5 years, snow depth explains 48% of the variation in fawn/doe ratios. Roe deer usually go into estrous in the second summer of life and, with delayed implantation, give birth at age two. A fawn undergoing a hard winter, however, would delay puberty by one year. Thus, there are fewer births 1.5 years after a hard winter, introducing a time lag into the deer population response. Still, fox density did have a significant effect on fawn production, even accounting for winter weather.
Summarizing: (i) All prey populations except voles increased in density as the mange struck the fox population. All but roe deer returned to previous population levels as the fox population returned to normal. (ii) Consistent with the alternate-prey hypothesis, whereas voles are not limited by fox predation, hare and grouse are. (iii) Red fox is a "keystone" species in structuring the small-game community in Scandinavia. It conveys the 3-4 year vole cycle to the hare and grouse populations. It shows both a functional response and a numerical response, since vole populations affect fox litter size. (iv) The mange mite, since it affects red fox populations so drastically, could also be called a keystone species. (v) The role of humans cannot be ignored. First, by exterminating wolves from this area, the role of the fox has been enhanced. Second, by opening up the forest through clear-cutting practices, man has also created excellent fox habitat. In a closed forest the role of the fox is likely to be of much less significance.
Are population cycles caused by predation?
Other work, however, seems to indicate that vole cycles are caused by predation. In the paper by Lindstrom et al. (1994), although red fox predation was temporarily eliminated, weasel-type (Mustelidae) and avian predators were not controlled. Recent research suggests that population cycling by prey species is only eliminated when all potential predators or parasites are eliminated (May 1999).
For example, Hudson et al. (1998) studied population cycles in the red grouse (Lagopus lagopus). Red grouse have cycles of population abundance with an average period of seven years. Moreover, these cycles are synchronized over large portions of northern England (Cattadori et al. 2005). These cycles are of special interest because grouse are a traditional English/Scottish shooting bird. About 75% of the British populations of red grouse undergo these cycles. Hypotheses for causation included vertebrate predators, food supplies, territory size, or parasite infections. According to Hudson et al. (1998), the answer is that the cycles are caused by parasites and the synchronization is set by climatic conditions affecting the transmission of the parasite (Cattadori et al. 2005).
The parasite is a nematode (Trichostrongylus tenuis). Hudson et al. (1998) began a study of six grouse populations, in which they predicted cyclic troughs in 1989 and 1993. In two of these populations birds were caught and treated with an anthelminthic medicine that eliminated the parasite. These treatments were in advance of the 1989 and 1993 predicted crashes. Two of the populations were not treated, and two of the populations were treated prior to 1993, but not prior to 1989.
The results were remarkable (May 1999). In the two untreated populations the predicted crashes occurred, on schedule, in 1989 and 1993. The magnitude of these crashes was a drop of more than three orders of magnitude. In the two populations treated twice, one population remained steady throughout the period 1988-96, while the second showed a small drop (threefold) in 1993. The populations treated only in 1989 had a major crash in 1993, but remained relatively steady in 1989. For this vaccination program to work, however, at least 20% of the birds had to be injected. This is based on models of epidemiology, discussed in Chapter 9.
In a second example, Korpimaki and Norrdahl (1998) studied two species of field voles (Microtus agrestis and M. rossiaemeridionalis) and a species of bank vole (Clethrionomys glareolus) that show persistent three-year cycles in Finland. Korpimaki and Norrdahl set up six large study areas of 2-3 km2. They reduced the numbers of weasels (Mustela nivalis) and stoats, the major mustelid predators. They also removed kestrels (Falco tinnunculus) and owls, the major avian predators. In 1992, prior to an anticipated population crash, they were successful in removing most of the weasels but not the stoats, and they did not attempt to remove the birds of prey. In 1995, however, they were able to remove all of the predators. None of the predators were removed in the control areas. In 1992, with only the weasels removed, the vole populations crashed, as did the control populations in both 1992 and 1995. However, in 1995, when all the vertebrate predators were removed, the vole populations remained steady without a crash. In earlier work, Norrdahl and Korpimaki (1995) had removed only avian predators. Here again, the voles crashed in both manipulated and control plots. We can conclude that vole cycles are indeed caused by predators. To eliminate cycles, all of the predators must be removed; removal of only a subset of the predator community leaves other predators to drive the cycle.
Turchin et al. (2000) drew a large distinction between vole and lemming populations. They agreed that voles, such as Microtus agrestis, are limited by their predators. They concluded, however, that lemming populations in Norway and other regions of Scandinavia are controlled by their food supply. They make the case that because voles feed on grasses and other herbs with the ability to quickly recover after defoliation by mobilizing underground food reserves, the vole-vegetation interaction is highly stable. According to Turchin et al., lemmings are primarily moss-eaters. Since mosses grow very slowly, the inherent time lag is highly destabilizing. Lemmings tend to deplete their forage in the arctic and alpine habitats where they live before predator populations begin to affect their dynamics. Lemmings are highly mobile during these population peaks, leading to the myths about lemming suicidal behavior (Chitty 1996). Thus, although both voles and lemmings undergo periodic oscillations, the cycles are driven by different ecological mechanisms, a predator-prey interaction in one case (voles) and a food-herbivore interaction in the other case (lemmings). This is a classic top-down (voles) versus bottom-up (lemmings) population control paradigm.
All of these works stress that cyclic prey species are actually preyed upon by a community of predators, again reminding us of the inadequacy of a single predator/single prey model. True cycles appear to be confined to northern temperate and subarctic ecosystems and are unknown in the tropical latitudes or southern hemisphere (Sinclair and Gosline
1997). Furthermore, it is widely believed that the amplitude of these cycles decreases at lower latitudes. However, a review of 700 long-term (25 + years) animal population studies found that this expected latitudinal gradient in population cycles can only be confirmed for North American carnivores, British lagomorphs, and Fennoscandian rodent populations (Kendall et al. 1998). The database used by Kendall et al. did not include studies from the tropics or the southern hemisphere. Nevertheless, they found periodicity in 33% of the mammal and 43% of the fish populations reviewed. Periodicity was less common among bird (13%) and insect (16%) populations, but was particularly common among grouse. Since almost 30% of all populations in their review showed regular cycles, this phenomenon cannot be dismissed (Kendall et al. 1998).
One reason that cycles may be more common at high latitudes among mammals is the relative simplicity of the predator-prey community. For example, when Gilg et al. (2003) studied lemming cycles in the high arctic tundra of Greenland, they found one major prey species, the collared lemming, and four predator species. Three species, as shown in Fig. 10.11, were generalist predators (snowy owl, arctic fox, and long-tailed skua) and one was a specialist predator, the stoat, a member of the weasel (Mustelidae) family. Mustelids are prolific predators and prolific breeders, capable of a very strong numerical response. In their study, Gilg et al. (2003) concluded that the lemming cycle in Greenland is driven by predation and there is no evidence of food or space (nest site) limitation. The specialist predator, the stoat, produces the cycle. There is a one-year delay in the numerical response of the stoat, resulting in a cyclical dynamic between lemmings and stoats. But the authors emphasize that generalist predators are necessary to stabilize the lemming-stoat interaction. The three generalist predators focus on lemmings only when the lemming population is very dense. As seen in Fig. 10.11, all three have strong functional responses to lemming density. Snowy owls and arctic foxes, in addition to stoats, also display a strong numerical response. The generalist predators are thought to be necessary to slow down the growth rate of the lemming at its highest densities, thereby allowing the specialist predator to catch up and begin driving the lemming population downward.
Some Canadian lemming populations, however, do not cycle (Reid et al. 1997). The collared lemmings at Pearce Point in the Northwest Territories are preyed upon by a specialist predator (rough-legged hawk Buteo lagopus), a semi-specialist predator (red fox), and by several generalist predators (golden eagle Aquila chrysaetos, grizzly bear, arctic ground squirrel Spermophilus parryii, peregrine falcon Falco peregrinus, and gyrfalcon Falco rusticolus). The interactions are governed by differences between summer and winter predation rates. The major stabilizing factor, however, is the presence of ground squirrels, which become the primary prey or principal alternative prey for the predators when lemming densities are low. Not only do ground squirrels provide a prey base to keep a diverse array of predators in the ecosystem, but they also prey on lemmings themselves. Without ground squirrels, Reid et al. (1997) hypothesize the loss of three of the generalist predator species. The complex guild of generalist predators plus the ground squirrels found in this ecosystem never allow the lemmings to go through an irruption.
As discussed earlier, snowshoe hare populations undergo 10-year cycles in North America, and, based on pelts returned to the Hudson's Bay Company, Charles Elton proposed that lynx controlled these hare populations. Recently Krebs et al. (1995, 2001), set up 1 km2
plots in the boreal forests of the Yukon in Canada. Fertilizer was added to promote plant growth in some plots, and in some plots predators were excluded. To summarize this very complex study, they found that both food and predation help drive the snowshoe hare cycle, and that the two effects are likely linked. Although Krebs et al. (2001) emphasize that "the hare cycle is not driven primarily by plant-herbivore interactions," food limitation increased predation by forcing hares to search more extensively for food, by reducing their health, and by making them less likely to escape predation. This result reinforces the predator-sensitive food hypothesis described below for wildebeest (Sinclair and Arcese 1995).
When radio-collars were placed on hares, Krebs et al. (2001) found that 95% of the hares died as a result of predation. In the Yukon, predators included lynx, coyotes, goshawks (Accipter gentilis) and great horned owls (Bubo virginianus). In this study, Krebs et al. (2001) could not identify a predominant role for any one predator species. As phrased by Krebs et al. (2001) "the hare cycle is not strictly a lynx-hare cycle, as many textbooks claim." The entire community of predators drives the hare cycles, not just lynx. In fact, in areas such as the Anticosti Island in eastern Canada, where no lynx are found, the hare cycles continue.
Another interesting twist is that when snowshoe hares are in a down portion of their cycle, the predatory species turn on each other (O'Donoghue et al. 1995). These data also come from the Yukon, where hare populations cycle every 8-11 years. Prey species included red squirrels (Tamiasciurus hudsonicus) and voles. The primary predators included lynx, coyotes, great horned owls, and goshawks, as well as red fox, wolverines (Gulo gulo), and wolves. Since the primary predators appear to go through population cycles related to those of the hare, O'Donoghue et al. put radio-collars on lynxes and coyotes and radio-tagged owls, hawks, and other birds of prey in order to determine their fate.
Snowshoe hare populations peaked at 450 per square mile (174 per km2) in 1989-90 and declined to 7 per square mile (3 per km2) by the spring of 1992. The lynx population declined from 60 to 15 per square mile (23 to 6 per km2). Did the lynx leave the area? If they died what were the causes of death?
Of 15 radio-collared lynx in 1991-92, by the spring of 1993 only seven remained. Six individuals emigrated and two deaths were recorded. Collars were found as far away as 800 km in Alaska, British Columbia, and the Northwest Territories.
In 1992-93, of nine radio-collared lynx, all had died by April. One lynx was found healthy and well-fed, but had been killed by a wolf. A female lynx was killed by another lynx. One young male starved. A male and female pair was migrating into the mountains when a wolverine killed the female. The male was killed by either a wolf or a wolverine, based on evidence from blood and tracks in the snow. The investigators actually witnessed a lynx being killed by a coyote. Other scientists have also reported witnessing lynx being killed by other lynx or by wolverines. All of this does not happen when hare are abundant. Furthermore, when hare populations are low, lynx themselves become very aggressive, routinely killing red fox.
Birds of prey followed the same pattern. In 1989 and 1990, when hare were abundant, 10 of 11 goshawk nests succeeded in fledging their young. In 1991, 50% nest failure was reported. All failed because of predation, evidently by great horned owls. The owls ate the adult female birds as well as their young.
In conclusion, snowshoe hare cycles are not simply driven by the lynx population. The vegetation and the entire community of predators are involved in these cycles.
Furthermore, these trophic interactions involve the community of predators itself. When prey populations are scarce, predatory animals turn on each other, and mortality of the predators due to starvation was, in fact, relatively rare.
Isle Royale is a 574 km2 island in Lake Superior that is a protected National Park. Moose populations have been present on the island for about 100 years (Pastor et al. 1988). Shortly after 1900 moose arrived on the island from Minnesota or Canada. They found abundant food and no major predators. The Isle Royale ecosystem is, in fact, relatively simple in terms of large herbivores and carnivores. As described by Smith et al. (2003), when moose arrived, the only mammalian carnivores were coyote and red fox, and the large herbivores included red squirrel, snowshoe hare and beaver (Castor canadensis). Ravens (Corvus corax) were also present. The moose population rapidly increased to some 3000 individuals by the early 1930s (Murie 1934). As they over-browsed their food supply the population declined. Several forest fires in the 1930s resulted in the regeneration of aspen and paper birch (Betula papyrifera), which allowed the moose population to increase again since these are preferred foods. In the late 1940s, gray wolves arrived and evidently quickly extinguished the coyote population, maintaining a relatively simple food web. According to Mech (1966), predation limited the moose population below its food supply. More recent research (Peterson 1999), however, indicates that both vegetation and wolves play a role in moose population dynamics.
Between 1948 and 1950 four exclosures (each 100 m2 in size) were set up to evaluate the effect of moose on the vegetation and other ecosystem properties. Pastor et al. (1988) concluded that intensive moose browsing led to fewer deciduous trees and a forest dominated by spruce (Picea glauca) and balsam fir (Abies balsamea), which are less palatable to moose. In addition, lower levels of nitrogen remained in the soil. The net result was an ecosystem less able to support moose populations, since moose do less well on this diet.
Although there seems to be some agreement that moose populations are affected by wolves, the question remains, how much of an effect? Peterson et al. (1984), based on about 20 years of data on the wolf population at Isle Royale, concluded that moose and wolves would go through population cycles typical of predator-prey interactions involving rodent and hare populations. However, since the body mass (M) of moose is larger, they concluded that moose-wolf interactions would cycle with considerably longer period lengths. In fact, after noting that the intrinsic rate of increase of mammals scales as M ~0'26, they proposed that vertebrates cycle as the fourth root of the body mass, M025. From this they predicted that moose populations should cycle with a period length of 38 ± 13 years. Given the paucity of data, however, this prediction cannot be confirmed. Moreover, their prediction that "oscillating elephant populations" should cycle with a period of 71 ± 21 years seems without merit.
Making predictions about prey and predator populations based on allometric relationships or on prey-induced vegetation changes also ignores the role of climate. Mech et al. (1987) showed that Isle Royale moose and northeastern Minnesota deer populations were both significantly affected by the snow accumulation of the previous winter. In severe winters moose populations were not able to find browse, and their physical condition deteriorated. Many of the moose and deer that were killed by wolves would have starved to death anyway. Therefore, although wolf predation was a direct mortality agent, Mech et al. believe it was secondary to winter weather in influencing deer and moose populations. In addition, snow accumulation during previous winters affects maternal nutrition to such a degree that fecundity and/or calf survivorship during the next growing seasons is seriously influenced.
The wolf population on Isle Royale decreased from 1981 through the mid-1990s, because of a canine parvovirus, probably introduced from the mainland on the hiking boots of visitors. During this period the moose population increased until it reached a record level in 1995 of 2400. Before the wolf population could recover (there were approximately 50 in 1980 and 24 in 1995), the moose population crashed. Almost 80% died, mostly from starvation, but a severe winter tick infestation contributed to the crash (Peterson et al. 1998). Peterson (1999) concluded that although the moose population density is influenced by wolves, the population level is ultimately set by available food.
Smith et al. (2003) have pointed out how much more complex the food web is at Yellowstone National Park in Wyoming, where wolves have recently been introduced, than is the case at Isle Royale. Whereas the only mammalian predators at Isle Royale are wolf and red fox, at Yellowstone coyote, mountain lion, grizzly bear, and black bear (Ursus americanus) are found along with wolves. Human hunting on American elk (Cervus elaphus, called red deer in Europe) must also be factored in. Avian predators at Yellowstone include bald eagles (Haliaeetus leucocephalus), golden eagles, magpies (Pica pica) and ravens, whereas only ravens are found on Isle Royale. Additional large prey species, beside moose, found at Yellowstone are pronghorn antelopes (Antilocapra americana), bighorn sheep (Ovis canadensis), mule deer, American elk, and bison (Bison bison). At Yellowstone only 26 instances of wolves killing moose have been recorded since the wolves were reintroduced. Wolves mainly prey on elk, and are predicted to reduce elk populations. However, the elk herds have only declined by approximately 18% so far (Smith et al. 2003, Ripple and Beschta 2003).
After reviewing 27 studies on moose-wolf interactions, Messier (1994) was able to generate both functional and numerical for wolves as a function of moose density. He concluded that when wolves are the single predator, a moose population will stabilize at 1.3 moose per km2, whereas the equilibrium density with no predators is 2.0 moose per km2. His analysis is consistent with a model proposing that under these circumstances moose populations are regulated at low densities due to density-dependent wolf predation. However, Messier comments that Isle Royale is an exception: it has a high-density moose population limited by food competition, and wolves are present but do not regulate moose density (Messier 1991, 1994).
Sinclair and Arcese (1995) studied the interaction between food supply and predation in the regulation of large herbivore populations on the Serengeti Plain in East Africa. They distinguished among three hypotheses:
1 The predation-sensitive food (PSF) hypothesis states that both food and predation limit prey populations. As food becomes limiting, animals take greater risks to obtain food and become victims of predation.
2 The predator regulation hypothesis states that predators hold prey populations well below starvation levels.
3 The surplus predation hypothesis states that predators kill only those prey individuals that are excluded from optimal habitat and are already dying of starvation. This is similar to Errington's (1946) ideas about the "doomed surplus" of muskrat populations.
These three hypotheses were tested by examining the body condition of Serengeti wildebeest (Connochaetes taurinus) over a 24-year period. Two phases of population growth were examined: (i) 1968-73, when food was abundant and prey populations were increasing; and (ii) 1977-91, when the wildebeest population was stationary and partially regulated by competition for food. Sinclair and Arcese examined live animals, predation kills, and non-predation deaths. Body condition was measured by an examination of bone marrow, the last reservoir of fat in these animals.
The predator-regulation hypothesis predicts that bone-marrow condition will be similar in predator-killed and live samples, while the surplus-predation hypothesis predicts that bone marrow will be similar in predator-killed and non-predator deaths. The PSF hypothesis predicts that bone marrow condition of animals killed by predators should be: (i) poorer than that of live animals; (ii) better than that of the animals who die of causes, such as disease, unrelated to predation; (iii) better when food is limiting than when it is abundant (because when food is abundant, predators only kill sick or young animals).
Analyses of the bone marrow from animals dying due to predation or from non-predation-related causes showed that these animals were in poorer health than the live population. In both the increasing and stationary phases, the animals dying from predation were in better condition than the animals dying from non-predation causes. These results are consistent with the PSF hypothesis, and inconsistent with the other two hypotheses. Lions (Panthera leo) and hyenas (Crocuta crocuta) killed animals in similar condition, but lions took younger animals.
The results suggest that: (i) body condition affects vulnerability of individual wildebeest to predation, and (ii) predation jointly limits the population along with intraspecific competition for food resources.
Sinclair et al. (2003) have also analyzed the community-wide patterns of predation in the Serengeti ecosystems of Tanzania and Kenya. Twenty-eight species of ungulates and 10 species of carnivores inhabit these areas, consisting mostly of open grassland (savanna). In any one habitat as many as seven species of canid and felid carnivores are present. The predators range from 8 kg (Golden jackal, Canis aureus) to 150 kg (lions), while prey sizes range from small gazelles (Gazella sp.), which weigh 18-20 kg, to elephants (Loxodonta africana), rhinoceros (Diceros bicornis), and hippopotamus (Hippopotamus amphibius), which come in at over 3000 kg. Long-term studies of the causes of ungulate mortality show that the proportion of adult mortality due to predation is above 80% in the smallest species such as oribi (Ourebia ourebia), impala (Aepyceros melampus), topi (Damaliscus lunatus), and wildebeest. By contrast, in a heavier species such as zebra (Equus quagga), adult mortality due to predation is 70%. There is a "threshold" in body size of about 150 kg, after which deaths due to predation decline substantially. Only about 23% of adult buffalo and 5% of adult giraffe (Giraffa camelopardalis) mortality is caused by predation; rhinos, hippos, and elephants (the "big three") suffer virtually no adult predation. While the smallest ungulates are preyed upon by as many as seven different predators, the number of potential predators falls off linearly as a function of the log of the herbivore weight, to zero for the big three. Thus the smallest ungulates (less than 150 kg)
are limited by the diverse array of predatory species, while the large ungulates are basically food-limited. A "natural experiment" confirmed this generalization. In the northern Serengeti poaching and poisoning eliminated most of the carnivores, including lions, hyenas, and jackals, whereas in the nearby Mara Reserve the predator community remained intact. During the years when predator populations were eliminated five of the species below 150 kg in weight increased their populations conspicuously as compared to the Mara Reserve. Once the predators returned, their populations declined. By comparison, the 800 kg giraffes did not increase in the predator-removal area. Sinclair et al. (2003) concluded that the mammalian herbivore populations in the Serengeti ecosystem are subject to top-down (predation) or bottom-up (food limitation) processes depending on their size.
Was this article helpful?