A Preliminary Analysis of a Neotropical Mammal Fauna John F. Eisenberg National Zoological Park, Smithsonian Institution, Washington, D.C. 20009, U.S.A. and Richard W. Thorington, Jr. National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20560, U.S.A. ABSTRACT The abundance and biomass of species of nonvolant mammals on Barro Colorado Island are estimated and compared with data from other areas. The edentates, particularly the sloths, are shown to be dominant elements in neotropical forests. The sloths and howler monkeys together constitute more than 50 percent of the biomass of mammals, em- phasizing the importance of the forest canopy. The bulk of the mammalian biomass is maintained by species which are browsers or frugivores and browsers. The total energy consumption of each species is estimated and the metabolic rate of the whole mammalian fauna is judged to be between 3000 and 4000 kilocalories per hectare per day on Barro Colorado Island. THE OBJECTIVE of this paper is to assess the bio- mass and the numbers of different mammalian spe- cies in a defined area of neotropical forest. For al- most a decade we have been concerned with this assessment and with defining the trophic structure of the major species of nonvolant terrestrial mam- mals within a neotropical rainforest ecosystem. It is most important to determine the biomass and the trophic role of each major species for the following reasons: ( 1 ) These data are necessary for determining the food consumption and the role in nutrient cy- cling of mammalian populations. (2) They provide a measure of the success of different species and orders of mammals within an ecological context. The biomass per taxon is a dif- ferent and perhaps better measure of ecological dominance than is the more conventional measure of number of species or genera per higher taxon in a given geographic area. (3) The information forms the necessary basis for comparison with mammalian faunas elsewhere: in different habitats and geographic areas within the neotropics, in other tropical areas, and in non- tropical habitats. (4) Numerical data from one area are essential for studies of population fluctuations. The tropics are not the stable equitable environments they were once thought to be, and thus we must examine the effects of climatic fluctuations on mammalian popu- lations. (5) It is imperative to consider the biomass and trophic roles of mammalian species in any efforts to conserve them in their native habitats or else- where. Reserves and parks must be large enough for mammals to maintain self-perpetuating breeding populations. The adequacy of such reserves can only be determined when we know the densities and fluctuations of mammalian populations supported by tropical habitats. Within a mature tropical rainforest the under- growth is often sparse and, aside from the fruits, leaves, and other litter which fall from the crowns of trees, most of the plant productivity is either locked in the trunks and root systems or produced and stored in the crowns of the trees themselves. The low biomass estimates for terrestrial mammals in tropical ecosystems were pointed out some years ago by Bourliere (1963). As the result of extensive research on mammalian populations in Ceylon, Ei- senberg, Muckenhirn, and Rudran (1972) called at- tention to the importance of arboreal herbivores in the tropical rainforests of the Old World. The un- suspected abundance of folivorous primates in Old World rainforests was established for the first time. The theoretical implications of the important bio- mass contribution of arboreal forms in tropical rain- forests were amplified in the publication by Eisen- berg and McKay (1973) and reiterated by Bour- liere (1972). In the neotropics, the howler monkey {Alouatta) is a frugivore which also consumes a great deal of foliage and, in some respects, appears to occupy an ecological niche comparable to that occupied by the Old World langurs and colobus monkeys {Presbytis and Colobus). The trophic structure of primate populations on Barro Colorado Island has been clari- fied through the work of Hladik and Hladik (1969). An examination of the presumed density of howlers on Barro Colorado suggested that they 150 BIOTROPICA 5 (3): 150-161 1973 had not reached a density commensurate with those observed for foHvorous primates in the Old World tropics (see Eisenberg, Muckenhirn, and Rudran 1972). It occurred to us that the true folivore niche in the New World tropics could well be partly occu- pied by the tree sloths (Bradypm and Chohepus). On the strength of this hypothesis, a research project was launched in 1970 by G. G. Montgomery and M. Sunquist aimed at clarifying the abundance, ac- tivity, and feeding patterns of sloths on Barro Colo- rado Island (Montgomery and Sunquist 1973). The major results of this work are in preparation, but the conclusions may be summarized here: Sloths are the most numerically abundant large mammal on Barro Colorado Island and account for the bulk of the mammalian biomass. Thus when the combined biomass of sloths and howler monkeys is considered, the arboreal herbivores constitute the major popula- tions of mammals on Barro Colorado Island. An estimate of mammalian biomass in Surinam: During the establishment of a reservoir behind the Afobaka Dam in Surinam, a large lowland area was flooded. A rescue operation was conducted and has been described in a book by Walsh and Gannon (1967) which includes an appendix of all mammals recovered during the rescue operation. From their data, it is possible to calculate the numerical abun- dance and relative biomass contribution of each spe- cies recovered (table 1). Of course, we do not know from what area the sample is drawn nor do we know its bias. We assume, however, that it is biased in favor of larger mammals and that most of the smaller forms either went unnoticed during the rescue operations or did not long survive swim- ming from flooded areas to high ground. In spite of the unknown biases associated with these data, we consider them very informative. The two sloths, Brady pus and Choloepus, constitute 23 percent of the biomass, while Alouatta is another 6 percent. Thus these three arboreal herbivores total 29 percent of the biomass. Tapirus and Mazama ac- count for another 42 percent of the biomass. Since tapirs and deer were the largest and most conspicu- ous (in the water) of the animals rescued, we be- lieve this figure is an overrepresentation of their contribution to the mammalian biomass. Other in- formative data are the estimates of the biomass of Dasypus 9 percent and Coendou 1 percent. These values are higher than we expected and suggest that armadillos and porcupines constitute larger per- centages of the mammalian biomass in the neotropics than is generally conceded. A comparison with the mammalian biomass on Barro Colorado Island: In table 2 we present an analysis of the mammalian populations of Barro Colorado Island as we presently understand them. The densities of marsupials and small rodents are not based upon actual trapping on Barro Colorado but rather are extrapolations from studies carried out on the Panamanian mainland by Fleming (1971, 1972). Other bases for the estimates of population densities are cited by species below. An examination of table 2 will indicate that we do not have any accurate census data for Coendou, TanumduA, Dasypus, and many of the carnivores from Barro Colorado. As noted above, the data from Surinam suggest that Coendou and Dasypus are sig- nificant contributors to the mammalian biomass. This possibility means that the total biomass estimate of Barro Colorado may actually equal only some 80 percent of the possible total. Compensating for this, we have compared the percent contribution to bio- mass of those genera common to the Surinam and Barro Colorado samples in table 3. Since the generic composition is not completely equivalent between the two areas, only 80 percent of the Surinam bio- mass is compared with 95 percent of the known calculated biomass for Barro Colorado. Bearing in mind that this 95 percent representation in table 3 from Barro Colorado is based on perhaps as little as 80 percent of the total, the correspondence be- tween the two samples is remarkably close. In both areas the sloths constitute a large portion of the biomass, although the percentage attributed to Bradypus is three times as large for the BCI estimate as that found in Surinam. A glaring dis- crepancy is the underrepresentation of the tapir (Tapirus bairdii) and the brocket {Mazama) on Bar- ro Colorado, which may result from the isolation of the island or from poaching in the past. Or, as sug- gested above, the tapir biomass from Surinam may be an overestimate. Barro Colorado has a higher percentage of caviomorph rodents than was found in Surinam. This situation may result from a smaller than normal predator complement on the island. The primate data from the two areas compare favorably. After comparing the Surinam and Barro Colo- rado data, we suspect that the density of Coendou, Dasypus, and Tamandu? is much higher on Barro Colorado than previously realized. This is an im- portant hypothesis to test because Coendou and Dasypus could be extremely influential components of the ecosystem, playing significant roles in the consumption and cycling of nutrients. The total biomass of mammals rescued in Suri- nam is about two-thirds the estimated biomass of terrestrial mammals on Barro Colorado. Barro Colo- rado is 15 km^ (Woodring 1958) in area and, pre- A Neotropical Mammal Fauna 151 TABLE 1. Nonvolant terrestrial mammals i ?ecorded from rescue operation in Surinam (Walsh and Gan- non 1967). Ml i z ??^ 1 s c ?S 2.1 ?.O 1800.0 "in + + + + + + + + Frugivore/carnivore Frugivore/carnivore Frugivore/omnivore Frugivore/carnivore Mustelidae Uira barbara > + + + Frugivore/carnivore Felidae ?elis pardalis ? + + Carnivore PERISSODACTYLA Tapiridae Tapirus bairdii 8 <.l 300.0 2400.0 3.6 + + Browser ARTIODACTYLA Tayassuidae Tayassu tajacu 100 .3 23.0 2300.0 3.5 ++ + + Frugivore/omnivore Cervidae Odocoileus virginianus Mazama americana 10? 20? .1 <.l 40 15 400.0 300.0 .6 .4 + + + + Browser Browser Total 21,111 GdyAG'b kg or 4431 kg/km2 ^ Heteromys, Oryzomys, and Zygodontomys probably present. Oryzomys not identified to species with certainty for BCI. ^ Brady pits and Choloepus estimates from Montgomery and Sunquist (1973). sons, although they probably are, but rather we em- phasize that the resources have a stable distribution in the forest which leads to a replicability of the distribution of howler troops. Similarly, our data show persistent distributions of troops of Cebus, Nasua, and Tayassu between 1964 and 1973. For instance, during censusing periods between 1964 and 1970 Eisenberg located coati troops within the core areas of the three bands studied by Kaufmann in 1959-60. Again the argument is strongly reinforced that the distribution of resources in Barro Colorado Island has changed little in the last 10 years and that these resources are exploited in a predictable fashion by mammalian populations. We also recognize that mammalian populations are not static in the tropics. In fact, our censusing data over the years indicate some population trends. During the seven years, agoutis (Dasyprocta) and squirrels {Sciurus) increased in number. The num- ber of Cebus may possibly have declined. The other species do not show any consistent trends, or present problems of interobserver reliability, or were noted with such infrequency as to permit no valid compari- sons. Volant mammals: The preceding analyses do not treat an important element of the neotropical mam- malian fauna, the Order Chiroptera. Bats are abun- dant and ubiquitous in the neotropics. Handley (1966) records 100 species of bats from Panama. Handley (unpublished) lists 60 species recorded from the Canal Zone, of which 34 have been found on Barro Colorado Island. The trophic structure of neotropical bats has been analyzed by MacNab (1971) who finds that of approximately 70 species native to Surinam, 37 percent are frugivorous and nectarivorous, 50 percent insectivorous and carnivor- ous, and 4 percent sanguivorous. The percentages are similar in Panama. For the 100 species listed by Handley (1966), we have used Wilson's (1973) TABLE 3. Biomas s contributions for selected genera. Barro Colorado Island Surinam Contributed % of estima ited Contributed %of Genus biomass (kg) total biomass" biomass (kg) total' Didelphis 1050 1.3 558 .59 Bradypus 31,920 40.0 6732 14.7 Choloepus 9800 12.3 3612 7.9 Alouatta 5500 6.8 2634 5.7 Saguinus 40 <.l 58 .12 Cebus 650 0.8 8 <.l Agouti 4000 5.0 1176 2.5 Dasyprocta 2800 3.5 452 .98 Proechimys' 2400 3.0 83 .18 Tapirus 2400 3.0 9432 20.6 Tayassu 2300 2.8 1633 3.5 Mazama 300 <.4 10,065 22.0 Total 63,160 36,443 represents 95% of represents 80% of the tabulatec 1 total the known total or 76% of the estimated total ?' Percentage calculated from a postulated total of 79.758 kg or 5317 kg/km- which is 20% higher than the known total of 66,465 kg as tabulated in table 2 (see text). *> Percentage calculated from the recorded total of 45,769 kg as quoted in table 1. ^ Echimys from Surinam. 154 Eisenberg and Thorington trophic role values to calculate that 3 percent are carnivorous, 1 percent piscivorous, 3 percent sangui- vorous, 13 percent are foliage gleaners, 39 percent are aerial insectivores, 32 percent are frugivorous, and 9 percent are nectarivorous. Handley (1967) discusses preliminary data concerning the ecological separation of insectivorous bats as a function of their flying height in the forest. Fleming, Hooper, and TABLE 4a. Strip census results." Diurnal and crepuscular Eisenberg sightings Thorington 1964 1965 1970 1971 Genus Number' % Number % Number % Number % Alouatta' 25 25 13 19 Cehus 26 26 17 25 27 28 17 18.3 14 14 4.3 Saguinus 4 4 5 7.5 4 4 1 1.1 Bradypus 1 1 1 1.5 6 6 1 1.1 Dasypus 0 0 0 0 2 2 3 3.2 Nasua 8 8 5 7.5 7 7 4 4.3 Dasyprocta 10 10 9 13.4 23 23 29 31.2 Sciurus 6 6 5 7.5 8 8 33 35.5 Sylvilagus 2 2 0 0 0 0 0 0 Tayassu 8 8 7 10.4 4 4 0 0 Tamandu? 7 7 3 4.5 1 1 0 0 Eira 1 1 2 3.0 0 0 0 0 Mazama 1 1 0 0 0 0 0 0 Odocoileus 0 0 0 0 0 0 1 1.1 99 67 97 93 Based on Based on ] Based on Based on 55 hours 27 hours 43 hours 30 hours * ?teles was not included since it was the object of intensive study by Eisenberg. ^ In group living species, e.g., Alouatta, Cebus, Nasua, Saguinus, Tayassu, one sighting equals the sighting of one or more animals in the same group, thus groups are treated as one to balance com- parisons with solitary species. ^ In 1965, Alouatta troops were counted only once and the same troop was not counted on subsequent days; this lowers the count. The same criterion was applied to no other species. TABLE 4b. Strip census results. Nocturnal sightings Eisenber^ Thoring ton 1964 and 1965 Number % 1971 Genus Number % Didelphis 8 29 13 22 Philander 2 7 3 5 Choloepus 1 3.5 2 3 Dasypus 1 3.5 8 14 Potos or Bassaricyon 1 3.5 5 8 Dasyprocta 0 0 2 3 Agouti 3 10 7 12 Diplomys 0 0 3 5 Misc. small rodents 2 7 5 8 Caluromys 0 0 6 10 Chironectes 0 0 1 1.5 Aotus 2 7 2 3 Proechimys 2 7 0 0 "Tamandu? 1 3.5 0 0 23 57 Based on Based on 21 hours 55 hours Wilson (1972) relate trophic specializations to the breeding cycles of neotropical bats with rather good results. Hence, the analysis of neotropical bat faunas is proceeding, but at present data on abundance or biomass are simply not available. Knowledgeable specialists estimate that the biomass of bats exceeds that of all other mammals in neotropical forests. We doubt that this supposition is true and think the il- lusion results from the rapid movement and large range of bats. They are numerous, but they are small. Referring to Odum (1970), we hypothesize that the bat biomass is not more than 10 percent of the total biomass of nonvolant mammals. An evaluation of census techniques: We have described and reviewed various techniques useful for censusing tropical mammals (Eisenberg, Santiapillai, and Lockhart 1970; Eisen berg and Lockhart 1972; Thorington 1972). Numerous other discussions of censusing techniques exist and will not be repeated here. In 1964, 1965, 1970, and 1971 we conducted visual censuses along the trails of Barro Colorado Island, totaling 155 daylight hours and 66 hours at night, biased toward times of greatest mammalian activity. Table 4 (a and b) summarizes the results of the sightings, and table 5 compares their rank or- der of abundance with those of the Surinam data and the Barro Colorado estimates from tables 1 and 2. From these comparisons we conclude that most of the large diurnal mammals on Barro Colorado are easily censused visually. However, the cryptic, ar- boreal sloth Bradypus is vastly underestimated. The nocturnal animals are not as easily censused by our techniques. We believe the arboreal forms Coendou and Choloepus are grossly underestimated, but that many of the terrestrial forms are seen in approxi- mate proportion to their abundance. We conclude that visual-strip censuses can give useful data about the relative abundance of diurnal terrestrial mam- mals such as Dasyprocta, Nasua, and Tayassu, and diurnal arboreal forms such as Sciurus, Alouatta, and Cebus. PROBLEMS OF CENSUSING PARTICULAR SPECIES A. DIURNAL ARBOREAL FORMS ( 1 ) The three-toed sloth, Bradypus infuscatus: This species has been studied by Montgomery and Sun- quist on Barro Colorado Island, who have found it may reach a density of 7.6 animals per hectare. As- suming an adult weight of 2.8 kg, this is a biomass of almost 2130 kg/km^. It would appear that Bradypus exists at the highest numerical density of any large, arboreal mammal and is therefore a sig- A Neotropical Mammal Fauna 155 TABLE 5. 'Rank order of abundances. BCI estimate Diurnal Censuses 1970 Surinam 1964 1965 1971 Bradypus Alouatta Tamandu? Myoprocta Dasyprocta Tayassu Bradypus Dasyprocta Alouatta Nasua Sciurus Cehus Tayassu Cehus Alouatta Dasyprocta Tayassu Nasua Tamandu? Sciurus Cehus Alouatta Dasyprocta Tayassu Nasua Saguinus Sciurus Alouatta Dasyprocta Cehus Sciurus Nasua Tayassu Saguinus Sciurus Dasyprocta Alouatta Cehus Nasua Dasypus^ BCI Nocturnal Censuses Thoringt Surinam Eisenberg 3n Dasypus Coendou Choloepus Coendou Didelphis Agouti Echimys Choloepus Proechimys Didelphis Agouti Philander Didelphis Proechimys Agouti Didelphis Dasypus Agouti Caluromys ? Animals ^ 250 gm. *> Number of sightings becomes less than three times after Dasypus, nificant component in the cycling of materials in thje rainforest ecosystem (Montgomery and Sunquist 1973). In spite of its high density, the animal is so cryptic that it is rarely seen and impossible to census visually. However, three-toed sloths descend to the ground to defecate every five to eight days. Each digs a small hole at the base of a tree with its tail and deposits its feces in it. A hole may be used in common by a mother and her young. By sweep- ing the forest litter one can find these fecal deposits and from their abundance estimate the density of sloths (see Montgomery and Sunquist 1973). This is the only valid indirect censusing technique for Bradypus available to an observer on foot. (2) The howler monkey, Alouatta palliata: Cen- sused by direct counts, the howler monkey is prob- ably the second most numerically abundant diurnal arboreal mammal on Barro Colorado Island. With an average density of .7 per hectare and with an average adult weight of 5.5 kg, the biomass approximates 400 kg/km^. Censusing techniques for Alouatta have been described elsewhere (Carpenter 1934; Collias and Southwick 1952; Thorington 1972). (3) The white-throated capuchin monkey, Cehus capucinus: Data from Oppenheimer (1968) indi- cate that the Cehus monkey on Barro Colorado Island exists at a moderate density of .17 individuals per hectare; this may be a low estimate. Their biomass would be approximately 44 kg/km^ assuming an av- erage adult weight of 2.6 kg. In walking surveys Eisenberg found Cehus troops almost as frequently as troops of howler monkeys, which are four times as abundant. This situation would appear to result from two factors; First, Cehus frequently emit alarm calls upon sighting a terrestrial observer, which makes them very conspicuous. Secondly, Cehus troops move rapidly through the forest and have home ranges almost 10 times larger than those of howler troops (Oppenheimer 1968; Chivers 1969). Thus, the probability of seeing the same troop sev- eral times on separate surveys is increased and the density of troops is overestimated. (4) The spider monkey. ?teles geoffroyi: This species was introduced on Barro Colorado Island in 1962 and has grown from an established group of 5 adults to 14 animals as of September 1972. Spider monkeys have large home ranges, and troops fre- quently divide into subunits. Thus, one of the prob- lems in censusing spider monkeys is to obtain ac- curate troop counts by recognizing all members of the same troop. Only patient work can ensure that troop composition has been worked out and all animals have been counted (Eisenberg and Kuehn 1966; Klein 1972). (5) The rufous-naped tamarin, Saguinus geof- froyi or Saguinus oedipus geoffroyi: In a mature tropical rainforest, Saguinus exists at rather thin den- sities, since it appears to be adapted to second- growth conditions or forested areas near natural clearings (Moynihan 1970). In recent years, the Saguinus population on BCI has declined. We do not have a reliable estimate but judge that there are fewer than 50 animals on the island at present. We saw them infrequently during our censuses (4% of all sightings ). Marmoset groups are easy to locate by the scold- ing chirps they give when they see an individual on 156 Eisenberg and Thorington foot. They will also respond to the calls of an iso- lated marmoset or imitations of their calls. A caged marmoset can attract any marmoset troop within hearing range; however, this range probably is less than 200 yards. (6) The lesser anteater, Tamandiui tetradactyla: This animal appears to be active both during the day and at night. It may be encountered on the ground or observed in the trees; hence, its categoriza- tion with respect to activity rhythm and location is somewhat ambiguous. During walking surveys on BCI, Tamandiui accounted for only 4 percent of mammal sightings. Compared with the Surinam data this is a lower than expected density. We would expect it to be at the same density as Cebus on Barro Colorado unless the termite fauna of the is- land differs significantly from that of Surinam. We know Tamandu? is moderately abundant in a tropi- cal rainforest, but as yet we have no adequate cen- susing technique. B. ARBOREAL NOCTURNAL FORMS ( 1 ) The two-toed sloth, Choloepus hofJTnanni: This species is cryptic and almost entirely nocturnal in habits (Sunquist and Montgomery 1973), which render it almost impossible to census directly. An indirect method of censusing utilizing a modified pellet count was developed by Montgomery and Sun- quist (1973). Their preliminary studies suggest that the two-toed sloth has a larger home range than the three-toed sloth and that it exists at roughly one-fourth the density of the three-toed sloth, with a biomass of 650 kg/km^. The data from Surinam differ only slightly, suggesting that Choloepus exists at approximately 40 percent of the density of Brady- pus. Thus Choloepus constitutes a significant pro- portion of the total biomass of arboreal mammals. (2) The prehensile-tailed porcupine, Coendou: The species C. rothschildi occurs on BCI, but its nocturnal habits render it very difficult to census. Males produce a distinctive strong odor which we have detected in the forest, but an olfactory census has not yet been devised. The data from Surinam suggest that Co^endou exists at a density equal to two- thirds that of Bradypus. If this is true on Barro Colorado, then porcupines comprise a significant part of the arboreal mammalian biomass on the island. C. DIURNAL TERRESTRIAL OR SEMI-TERRESTRIAL FORMS (1) The common agouti, Dasyprocta punctata: The density of this species on BCI may be estimated from the data of Smythe (1970 a, b) as roughly one animal per hectare. If we assume 2 kg to be the average weight of an adult, then the biomass estimate would approximate 200 kg/km^. Thus, Dasyprocta is the most numerically abundant diurnal, terrestrial mammal, as was also indicated by the walking surveys. The agouti is conspicuous because it barks when it sees an observer and continues to bark as it re- treats. This anti-predator strategy makes it easy to census by walking transects. The animals tend to live paired but travel separately; hence, they do not show any particular clumping tendency. (2) The coatimundi, Nasua nasua: The coati on BCI is a dominant member of the mammalian com- munity and was studied definitively by Kaufmann (1962). Our density estimates are based on Kauf- mann's data. We assume an average home range of 40 hectares for a female band of 8 animals. Allow- ing overlap in home ranges and accounting for the mostly solitary males, we estimate the average den- sity to be 0.4 animals per hectare. This is half the agouti density, which is approximately the frequency with which we encountered them in the forest. We believe that coatis are amenable to a tran- sect census. They are noisy in the forest. They have squeaking contact notes, and the mothers give a warning squeak when they notice an observer. Thus they are easy to detect in a transect. A rough index of abundance might also be achieved by keeping count of tracks and by determining the relative abun- dance of agouti and coati tracks. (3) The collared peccary, Tayas su tajacu: This species is active in part at night but certainly active enough during the day to be censused by foot sur- veys. Bands often give away their position by bark- ing as they flee. Contact frequency with peccaries was at about the same as that for coatis during the walking surveys by Eisenberg (7%). However, their density is probably only a sixth that of the coati because their greater conspicuousness and larger home range increases the probability for repeated contact with the same band. Computing frequency of tracks along transects might be a useful index to estimating the abundance of this species. (4) The brocket and white-tailed deer, Mazaima and Odocoileus: During walking surveys on BCI, Eisenberg noted Mazama only twice and did not see any Odocoileus; however, Montgomery and Thoring- ton report occasional sightings of Odocoileus. These deer may well be studied by transect walking if spe- cial attention is paid to tracks and if survey plots are set up adjacent to the transects which are cleared bi-weekly of all fecal material. Relative abundance of these deer species could then be estimated in a manner similiar to that employed in Ceylon by Ei- senberg, Santiapillai, and Lockhart (1970). A Neotropical Mammal Fauna 157 D. NOCTURNAL TERRESTRIAL FORMS ( 1 ) The spiny rat, Proechimys semispinosus: Uti- lizing the data from Fleming (1971) and Smythe (1970a), we consider P. semispinosus to be the most numerically abundant nocturnal small mammal on BCI, probably existing at a density of more than 2 per hectare. However, its biomass contribution (1.6 kg per hectare) is not large due to the small size of the animal. The only effective way to esti- mate Proechimys abundance is a trap-mark-release program, such as that used by Fleming. This same technique could well be applied to the other small rodents which probably exist at reasonable numerical densities but do not contribute significantly to the biomass. (2) The nine-banded armadillo, Dasypus no- vemcinctus: We have no firm estimate of the num- erical density of this species on BCI. However, it was the second most abundant mammal in Thoring- ton's nocturnal counts, and its burrows are common. The data from Surinam suggest that Dasypus may exist at a numerical density roughly half that of the three-toed sloth. If so, it would be a significant con- tributor to terrestrial mammalian biomass. The best census technique of which we know is a capture- mark-release program in which the animals are banded on the tails with colored electrical tape. An index of abundance could be obtained by counting burrows along transects. ( 3 ) The paca. Agouti paca: According to Smythe (1970a), this species lives in pairs like Dasyprocta but with a home range almost three times as large. We calculate from the estimates of Smythe that paca may contribute biomass at the level of 2.7 kg per hectare on BCI, which makes it a significant component of the terrestrial mammalian fauna. The paca is nocturnal, much quieter and more cryptic than Dasyprocta, and therefore it is much more dif- ficult to census directly. If the abundance of the agouti could be estimated by walking surveys, and a ratio of paca tracks to agouti tracks could be ob- tained, then the abundance of the paca could be re- lated to agouti density. (4) The common opossum, Didelphis marsupi- alis: The opossum is reasonably abundant on BCI and is both arboreal and terrestrial. In nocturnal censusing, we encountered it most frequently. The data from Surinam suggest that it exists at roughly half the density of the agouti. This species could also be censused by recording tracks on a transect and relating their abundance to those of agouti tracks. (5) Baird's tapir, Tapirus bairdii: Baird's tapir has been repeatedly reintroduced and is firmly estab- lished on BCI. Most of the individuals are known and named. It has proven difficult if not impossible to introduce new specimens to the island since the intruders are generally attacked by the resident male. This circumstance strongly suggests that mem- bers of a given "community" know one another al- though they do not form cohesive social groupings. Tapir density is very difficult to estimate, but track counts might again prove useful. The data from Surinam suggest it exists at half the density of the peccary, but this may be an overestimate. Even at low numerical densities, the tapir is an important contributor to terrestrial mammalian biomass. CONCLUSIONS AND DISCUSSION In this paper we have examined the importance of the various species of mammals in a neotropical for- est, as judged by the biomass of each. The impact of these species on the forest ecosystem can be viewed in terms of their trophic specializations and their location in the forest. Utilizing the biomass data, as set forth in tables 1 and 2, we offer two biomass pyramids in figure 1, which provide this viewpoint. From 40 percent (Surinam) to 70 per- cent (BCI) of the mammalian biomass is main- tained by the arboreal species. Browsers and grazers, both terrestrial and arboreal, account for 65 percent (Surinam) to 53 percent (BCI) of the total bio- mass; and combination feeders on fruits and browse account for 15 percent (Surinam) to 25 percent (BCI) of the total. The pyramids conform in shape to similar figures published for the Old World trop- ics (Hendrichs 1972; Schallet 1972; Eisenberg and Lockhart 1971). Fleming's (1973) analysis of trophic diversity is not directly comparable with ours, both because he included bats and because his unit of analysis was the species not biomass. We have taken our biomass estimates and com- puted from them an estimate of the energy consump- tion of the different taxa (table 6). This estimate is given in kilocalories per hectare per day and is computed from the estimated basal metabolic rates of average-sized individuals of each species. The basal metabolic rates were computed as 70 Kcal/ kg^/* per day (Kleiber 1961) and the figures for Bradypus and Choloepus were reduced by 51 and 34 percent, respectively (Goffart 1971). These figures are only first approximations of the rates of energy consumption by different mammals in the neotropics and hence of the importance of these in nutrient cycling, but we find them informa- tive. The sloths, Bradypus and Choloepus, are obvi- ously the most important consumers in the canopy and together may use more than four times as many kilocalories as the howler monkeys. We calculate that these three genera account for more than half 158 Eisenberg and Thorington the energy consumed by the total fauna of nonvolant mammals. The rodents account for 23 percent of TABLE 6. Metabolic rates per hectare for species of mam- nuds on Barro Colorado- Island." Average % of Taxon weight (kg) Kcal/H/day total Kcal Marmosa .06 4.6 .3 Philander 1.4 24. 1.5 Didelphis 1.0 49.0 3.0 Aotus .8 1.9 .1 Alouatta 5.5 170. 11. Cebus 2.6 24. 1.5 ?teles 5.0 2.2 .1 Saguinus .8 1.9 .1 Bradypus 2.8 560. 36. Choloepus 3.5 220. 14. Sciurus .25 4.9 .3 Heteromys .10 8.3 .5 Oryzomys .07 19.0 1.2 Agouti 8.0 110. 6.9 Dasyprocta 2.0 110. 6.9 Proechimys .8 120. 7.4 Nasua 3.0 64. 4.0 Tapiras 300. 27. 1.7 Tayassu 23. 49. 3.0 Mazama 15. 7.1 .5 Odocoileus 40. 7.4 .5 ^ See text for bases of estimates. Calculations assume two significant digits. all estimated basal energy consumption. The three genera, Agouti, Dasyprocta, and Proechimys, are the most important terrestrial mammals in considera- tions of nutrient cycling on Barro Colorado Island. The large mammals, Tapirus, Tayassu, Odocoileus, and Mazama, are relatively insignificant consumers of energy accounting for less than 6 percent of the total, approximately the same as the marsupials. Be- cause of their high rates of metabolism, the small rodents and marsupials are potentially significant consumers, and hence it is important to obtain bet- ter estimates of their biomass. The same argument applies equally well to the Chiroptera. The total mammalian population which we have estimated on Barro Colorado consumes energy at the basal rate of 1600 kilocalories per hectare. The ac- tual rate of consumption is greater than this, and the total energy consumption of nonvolant mammals on BCI probably lies between 3000 and 4000 kilo- calories per hectare per day. Obviously much more work needs to be done on populations of tropical mammals. Data on the numbers, biomass, and energy consumption of mam- mals must be collected from many different areas and many different habitats for comparison. This Br/Fr;Br[| C l7 Fr/G M ? Fr/CiFr/0 \^ I/0;Fr/I ? Fr/Br I B.C.I. 20 40 60 80 xlOO Kg Srl?r;Br{_ I i_ 20 40 60 80 X 100 Kg BIOMASS PYRAMIDS FIGURE 1. Biomass pyramids for Barro Colorado Island and Surinam. Br/Fr = primary browser, secondary frugivore; Br = browser; Fr/Br = primary frugivore, secondary browser; I/O = primary invertebrate consumer and secondary omnivore; Fr/I = primary frugivore, secondary insectivore; Fr/C = primary frugivore, secondary carnivore; Fr/O = primary frugivore, secondary omnivore; M =: Myrmecophage; Fr/G = primary frugivore, secondary granivore; C = carnivore. Black bars indicate terrestrial component. Stippled bars indicate scansorial component. White bars indi- cate arboreal component. Question marks refer to incomplete censusing for three trophic levels on Barro Colorado Island. A Neotropical Mammal Fauna 159 procedure will require refinement of censusing tech- niques, which is also obligatory for collecting data on the population dynamics of many species. It is important to obtain this information, because it is basic to an understanding of population fluctuations and the long-term dynamics of survival of mam- malian species in the tropics. Only with such data is it possible to predict the minimal sizes of popula- tions consistent with wise conservation practices, and thereby to establish the minimal sizes of reserves and parks for maintaining the diversity of the neotropi- cal mammalian fauna. Finally, our analysis emphasizes the dominance of the Edentata in the neotropics. This is not clear from faunal lists because a small number of species are involved. The Edentata, according to the Surinam data, account for 36 percent of the terrestrial mam- malian biomass. The Panama data confirm this dominance with sloths alone accounting for more than 50 percent of the estimated mammalian biomass on BCI. The Edentates seem to have evolved three types of feeding specializations, two of which concern very stable, predictable resources. The family Brady- podidae is specialized as an arboreal browser, thus utilizing a predictable and extensive resource. The family Myrmecophagidae includes both arboreal and terrestrial forms, highly specialized for feeding on ants and termites. A significant part of the animal biomass is tied up in ants and termites (Odum 1970), and the niche is apparently an extremely stable one in the tropics. The Dasypodidae are somewhat generalized, although showing many spe- cializations within the family. Dasypm no-vemcinctus is a generalized feeder occupying an insectivore- omnivore niche. The Edentates are definitely holding their own as a dominant mammalian order in the neotropics in spite of the fact that this realm has been invaded since the Pliocene by successive waves of other eutherian immigrants (Simpson 1969). The Eden- tata together with the Marsupialia exhibit the long- est evolutionary history of any surviving mammalian group in the neotropics. Evidently in their long-term occupancy of arboreal browsing and myrmecophagus niches in the neotropics they have become sufficient- ly specialized to withstand competition from subse- quent immigration. The ecological dominance of this order in the neotropics is unique and compara- ble to the ecological dominance of the order Mar- supialia in the rainforests of the York Peninsula in Australia (Harrison 1962). ACKNOWLEDGEMENTS This paper is based on the work of many persons, to whom we are indebted for publications and discussions. In particular we are grateful for the criticism and advice of our colleagues at the Smithsonian. Our recent work in Panama has been funded by the Smithsonian Environmen- tal Sciences Program. Previous work by Thorington was funded by NIH Grant FR 00168 to Harvard University and by Air Force Contract No. F44620-67-C-0063 with the Smithsonian. Eisenberg was previously supported by Grant No. GB-3545 from the National Science Foundation. LITERATURE CITED ALTMANN, S. A. 1959- Field observations on a howling monkey society. J. Mammal. 40: 317-330. BOURLIERE, F. 1963. Observations on the ecology of some large African mammals. In F. C. Howell and F. Bour- liere (Editors). African ecology and human evolution. Pp. 43-54. Aldine, Chicago. . 1972. Comparative ecology of rainforests mammals. In B. J. Meggers, E. Ayensu, and W. D. Duckworth (Editors). Tropical forest ecosystems in Africa and South America: A comparative review. Pp. 279?292. Smithsonian Press, Washington, D.C. CARPENTER, C. R. 1934. A field study of the behavior and social relations of howling monkeys. Comp. Psychol. Monog. 10(2): Serial No. 48: 168 pp. . 1965. The howlers of Barro Colorado Island. In I. Devore (Editor). Primate behavior. Pp. 250-291. Holt Rinehart & Winston, New York. CHIVHRS, D. J. 1969. On the daily behaviour and spacing of howling monkey groups. Folia Primat. 10: 48-102. COLLIAS, N., AND C. SOUTHWICK. 1952. A field study of population density and social organization in howling monkeys. Proc. Amer. Phil. Soc. 96: 143-156. EISENBERG, J. F., AND R. E. KUEHN. 1966. The behavior of ?teles geoffroyi and related species. Smithsonian Misc. Coll., Pub. 4683, 151(8): 1-63. , AND M. LOCKHART. 1972. An ecological reconnaissance of Wilpattu National Park, Ceylon. Smithsonian Contribs. Zool., No. 101, 118 pp. , AND G. M. MCKAY, 1973. Comparison of ungulate adaptations in the New World and Old World tropi- cal forests with special reference to Ceylon and the rainforests of Central America. In V. Geist and F. Walther ( Editors ). Proceedings of the First International Congress on Ungulate Behavior and Its Relation to Manage- ment. lUCN, Morges. , N. MUCKENHIRN, AND R. RUDRAN. 1972. The relationship between ecology and social strucmre in primates. Science 176: 863-874. 160 Eisenberg and Thorington -, C. SANTIAPILLAI, AND M. LOCKHART. 1970. The study of wildlife populations by indirect methods. Ceylon J. Sei., Biol. Sei. 8(2): 53-62. FLEMING, T. H. 1971. Population ecology of three species of neotropical rodents. Misc. Pubs. Mus. Zoo!., Univer- sity of Michigan, No. 143, 77 pp. . 1972. Aspects of the population dynamics of three species of opossums in the Panama Canal Zone. J. Mammal. 53: 619-623. . 1973. Numbers of mammal species in North and Central American forest communities. Ecology 54: 555- 563. , E. T. HOOPER, AND D. E. WILSON. 1972. Three Central American bat communities: Structure, reproduc- tive cycles, and movement patterns. Ecology 53: 555-570. GOFFART, M. 1971. Function and form in the sloth. 255 pp. Pergamon Press, New York. HANDLEY, C. O. 1966. Checklist of the mammals of Panama. In R. L. Wenzel and V. J. Tipton (Editors). Ecto- parasites of Panama. Pp. 753-795. Field Museum of Natural History, Chicago. . 1967. Bats of the canopy of an Amazonian forest. Atas do Simposio sobre a Biota Amaz?nica vol 5 (Zoolog?a): 211-215. HARRISON, J. L. 1962. The distribution of feeding habits among animals in a tropical rain forest. J. Anim. Ecol 31: 53-64. HENDRICHS, H. 1972. Beobachtungen und Untersuchungen zur ?kologie und Ethologie, insbesondere zur sozialen Organisation, ostafrikanischer S?ugetiere. Z. Tierpsychol. 30: 146-189. HLADIK, A., AND C. M. HLADIK. 1969. Rapports trophiques entre vegetation et primates dans la foret de Barro Colorado (Panama). La Terre et la Vie, No. 1-1969: 25-117. KAUFMANN, J. 1962. Ecology and social behavior of the coati, Nasua narica on Barro Colorado Island Panama. Univ. Calif. Pubs. ZooL 60(3): 95-222. KLEIBER, M. 1961. The fire of life. 454 pp. John Wiley, New York. KLEIN, L. L. 1972. The ecology and social organization of the spider monkey, ?teles helzebuth. Ph.D. disserta- tion. University of California, Berkeley. MACNAB, B. 1971. The structure of tropical bat faunas. Ecology 52: 353-358. MONTGOMERY, G. G., AND M. E. SUNQUIST. 1973. Impact of sloths on neotropical forest energy flow and nutrient cycling. In E. Medina and F. Golley (Editors). Trends in tropical ecology: Ecological Studies IV. Springer, N.Y. (in press). M0YNIHAN, M. 1970. Some behavior patterns of Plaryrrhine monkeys II. Sagainus geoffroyi and some other tam- arins. Smithsonian Contribs. Zoo!. 28: 77 pp. ODUM, H. T. 1970. A tropical rain forest. Division of Technical Information, U.S. Atomic Energy Commission. OPPENHEIMER, J. R. I968. Behavior and ecology of the white-faced monkey, Cebm capucinm, on Barro Colorado Island, C. Z. Ph.D. Thesis, University of Illinois. SCHALLER, G. 1972. The Serengeti lion. University of Chicago Press. SIMPSON, G. G. 1969. South Ameiican mammals. In E. J. Fittkau, J. lilies, H. Klinge, G. H. Schwabe, and H. Sioli. Biogeogiaphy and ecology in South America. Pp. 879-909. Junk, The Hague. SMYTHE, N. 1970a. Ecology and behavior of the agouti (Dasyprocta punctata) and related species on Barro Colo- rado Island, Panama. Ph.D. Thesis, University of Maryland, College Park. . 1970b. Relationships between fruiting seasons and seed dispersal methods in a neotropical forest. Amer. Nat. 104: 25-35. SUNQUIST, M. E., AND G. G. MONTGOMERY. 1973. Activity patterns and rates of movement of two-toed and three- toed sloths. J. Mammal. 54: in press. THORINGTON, R. W., JR. 1972. Censusing wild populations of South American monkeys. II. International sym- posium on health aspects of the international movement of animals. Pan American Health Organization WHO. Sei. pubL 235: 26-32. WALSH, J., AND R. GANNON. 1967. Time is short and the water rises. Thomas Nelson and Sons, Camden, N.J. WILSON, D. E. 1973. Bat faunas: a trophic comparison. Systematic Zoology 22: 14-29. WOODRING, W. P. 1958. Geology of Barro Colorado Island, Canal Zone. Smithsonian Misc. Colls. 135(3): 39 pp. A Neotropical Mammal Fauna 161