Mammal faunas from northern French Guiana (in northeastern Amazonia) and the Yavarí-Ucayali interfluve (in southwestern Amazonia) are the first Amazonian mammal faunas to be comprehensively described in terms of taxonomic composition and community structure. Bats are the most speciose group in each fauna, followed in order of decreasing species richness by rodents, didelphimorphians, carnivorans, xenarthrans or primates, and ungulates. A total of 180 species are known from northern French Guiana and 199 species from the Yavarí-Ucayali interfluve, but an additional 25 species are possibly missing (“pseudoabsent”) from the Yavarí-Ucayali interfluvial inventory; therefore, the increment by which species richness in northeastern Peru exceeds that of northern French Guiana is on the order of 10–25%. Complementarity (dissimilarity) between these faunas is high (79%–89%) for opossums, primates, and rodents, but complementarity is substantially lower for other groups (especially bats, carnivorans, and ungulates), suggesting taxonomic differences in geographic filtering. Most species in both faunas are nocturnal, as might be expected from the abundance of bats, but even among nonflying mammals nocturnal species outnumber diurnal species by about 2:1. Approximately equal numbers of nonflying species in both faunas are arboreal or terrestrial, whereas much smaller numbers are scansorial or semiaquatic; with one possible exception, none is fossorial. Despite such behavioral similarities, these faunas differ in trophic composition and guild membership: substantially more primary consumers are present in the Yavarí-Ucayali interfluve than in northern French Guiana, whereas numbers of secondary consumers and omnivores are similar. Higher primary productivity in western Amazonia, a geomorphologically dynamic landscape with fertile soils and phenologically diverse habitats, could explain faunal differences in both species richness and trophic structure.
INTRODUCTION
Accurate information about mammalian diversity and community ecology in Neotropical rainforests is important for both research and conservation agendas, but it is hard to obtain. Taxonomically comprehensive inventories require time and intensive effort with multiple methods to detect the presence of elusive species differing in size, diel activity, substrate use, and diet (Voss and Emmons, 1996; Sampaio et al., 2003; Munari et al., 2011; Palmeirim et al., 2019). Additionally, because small mammals (opossums, bats, and rodents) are often difficult to identify in the field, species lists can be problematic if specimens are unavailable to voucher dubious occurrence records. Lastly, sites where inventory fieldwork is logistically feasible have often been partially defaunated by hunting or habitat fragmentation (Voss and Emmons, 1996; Bogoni et al., 2022).
For Amazonia, by far the largest Neotropical rainforest biome, we still lack basic knowledge about sympatric species richness and site-to-site faunal differences. Although richness and turnover can be estimated from stacked range maps (Tognelli and Kelt, 2004), the accuracy of stacked-map estimates cannot be properly assessed without inventory-based ground-truthing. Even so-called expert range maps are likely to be unreliable for small mammals, especially in Amazonia where new species and large range extensions are still routinely reported.
In addition to such Linnaean (taxonomic) and Wallacian (biogeographic) shortfalls (Hortal et al., 2015), our understanding of mammalian diversity in Amazonia is handicapped by the lack of ecobehavioral trait data for many species. As a result, no taxonomically comprehensive analysis of ecological community structure exists for any Amazonian mammal fauna, despite useful attempts that have been made for various faunal subsets (e.g., Janson and Emmons, 1990; Peres, 1999; Kalko and Handley, 2001; Haugaasen and Peres, 2005). Important new sources of natural history information are now available from camera-trapping studies (Gómez et al., 2005; Blake et al., 2012; Gregory et al., 2022) and ethnographic interviews (Voss and Fleck, 2011; Fleck and Voss, 2016), but these data have yet to be combined in any general synthesis of Amazonian mammal community structure.
This report is based on the results of long-term faunal research programs in French Guiana and Peru, results that now provide sufficient materials for assessing the taxonomic and ecological dimensions of mammalian diversity and community structure at opposite biogeographic poles of Amazonia. Importantly, large numbers of specimens were collected in both regions, and taxonomic analyses of those specimens have been summarized in monographs that document species identifications for taxa previously neglected or deemed intractable by many faunal researchers. Although some species remain data deficient, this report also incorporates natural history information from camera-trap surveys, traditional autecological and dietary studies, and ethnographic sources to provide preliminary but taxonomically inclusive analyses of community structure in both regions.
MATERIALS AND METHODS
Species Lists
Most previous discussions of Amazonian mammal diversity have been based on species lists from single-site inventories, but single-site inventories have the key disadvantage of incompleteness. Among other evidence of inventory incompleteness, graphs of species accumulation from single-site inventories are never convincingly asymptotic, and long lists of “missing” species (unobserved species with geographic ranges that overlap the site) are dominated by behaviorally elusive taxa (Voss and Emmons, 1996). Therefore, a more meaningful basis for faunal inference can be had by combining lists from multiple localities within biogeographically homogeneous regions. Such regional lists include species that are potentially sympatric, although not necessarily syntopic.
FIG. 1.
Four Neotropical rainforest biomes (stippled) separated by montane or nonforest vegetation. Primary intercardinal directions (NE, NW, SE, SW) label quadrants of Amazonia defined by the Amazon, Negro, and Madeira rivers (after Wallace, 1854). Insets show the location of important faunal inventory sites in northern French Guiana (in NE Amazonia: Par, Paracou; PS, Petit Saut; Nou, Nouragues) and the Yavarí-Ucayali interfluve (in SW Amazonia: SP, San Pedro; JH, Jenaro Herrera; NSJ, Nuevo San Juan).
![img-z3-1_01.jpg](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z3-1_01.jpg)
I compiled species lists from the published results of inventory projects and ecological research at rainforested sites in northern French Guiana and the Yavarí-Ucayali interfluve of northeastern Peru (fig. 1). These regions lack internal dispersal barriers, and they occupy opposite quadrants of Amazonia as traditionally recognized by biogeographers (Wallace, 1854). Whereas northern French Guiana is in northeastern Amazonia (north of the lower Amazon and east of the Rio Negro), the Yavarí-Ucayali interfluve is in southwestern Amazonia (south of the upper Amazon and west of the Rio Madeira).
Northern French Guiana: An overseas department of France, French Guiana lies between the Maroni and Oyapock rivers, which have their headwaters in the remote Tumuc Humac Mountains along the Brazilian border and discharge into the Atlantic Ocean. Other rivers in the department are substantially smaller, have their headwaters in the central highlands (<900 m above sea level), and none is known to limit the distribution of any terrestrial vertebrate. Except for coastal swamps and savannas, low scrub associated with rocky soil on isolated hilltops, a few urban areas, and small agricultural clearings, the department is uniformly covered by tall, closed-canopy, evergreen forest. Physiognomic and floristic descriptions of local rainforest habitats were reported by Mori and Boom (1987), Sabatier (1993), and Poncy et al. (2001). Isohyets based on weather records from 1961 to 1990 suggest that most of this region receives at least 2700 mm of rainfall annually (Freycon et al., 2010).
Faunal data are available from multiyear inventory sites, ecological research stations, and other localities in the northern part of the department (within 100 km of the Atlantic coast), from the right bank of the Maroni to the left bank of the Oyapock. The largest species lists were obtained from Paracou (5.28°N, 52.92°W), a forestry research station where AMNH research teams worked for several years to inventory the entire mammalian fauna (Simmons and Voss, 1998; Voss et al., 2001); Petit Saut (5.07°N, 53.05°W), where thousands of mammals were rescued from rising water behind a newly constructed hydroelectric dam (Vié, 1999); and Nouragues (4.09°N, 52.68°W), an ecological field station where French researchers have studied mammals and their biotic interactions for several decades (Charles-Dominique, 2001; Feer and Charles-Dominique, 2001). Although local populations of some game species may have been depleted by hunting at Paracou, the fauna at Petit Saut was remote from human disturbance prior to dam construction, and the Nouragues site is pristine.
The mammals listed in appendices 1 and 2 include all the rainforest species reported by Simmons and Voss (1998), Vié (1999), Feer and Charles-Dominique (2001), Voss et al. (2001), Catzeflis et al. (2013), Moratelli et al. (2015), Pavan et al. (2018), and Alexandre and Thoisy (2023). Not included in these lists are seven species associated with coastal savannas rather than rainforest (Cryptonanus sp., Natalus tumidirostris, Odocoileus virginiana, Holochilus nanus, Oligoryzomys sp., Sigmodon alstoni, and Zygodontomys brevicauda). Voucher specimens that document taxonomic identifications of mammals from northern French Guiana are at the American Museum of Natural History (AMNH, New York) and the Muséum National d'Histoire Naturelle (MNHN, Paris).
Yavarí-Ucayali Interfluve: The Río Yavarí and the Río Ucayali are right-bank tributaries of the upper Amazon, and the region they delimit is part of Loreto department in northeastern Peru (Voss and Fleck, 2011). All the streams that drain the interfluvial interior are small, have their headwaters in low hills (<200 m above sea level), and are not known to limit the distribution of any terrestrial vertebrate. Except for river beaches, small patches of peatland savanna, and agricultural clearings around several villages, the entire landscape is covered with tall, closed-canopy evergreen forest. Physiognomic and floristic descriptions of local rainforest habitats were reported by Fine et al. (2006), Lopez-Parodi and Freitas (1990), and Pitman et al. (2003). Annual average rainfall throughout the region is thought to be at least 2500 mm (Marengo, 1983).
Information about mammals of the Yavarí-Ucayali interfluve is primarily available from three sites where multiyear faunal inventory projects and/or long-term ecological studies have been carried out: Jenaro Herrera (4.92°S, 73.67°W), a forestry research station on the right bank of the lower Ucayali; Nuevo San Juan (5.25°S, 73.17°W), a Matses Indian village on the Río Gálvez (a left-bank tributary of the Yavarí); and San Pedro (4.33°S, 73.20°W), a ribereño village on the Quebrada Blanco (a tributary of the Río Tahuayo, which is a right-bank tributary of the Amazon). Ecological research, faunal inventory efforts, and historical collecting activities at these and other sites in the Yavarí-Ucayali interfluve were summarized by Voss and Fleck (2011) and Velazco et al. (2021). Species lists from the Yavarí-Ucayali interfluve (appendices 3, 4) were compiled from faunal monographs that treated the regional primates (Voss and Fleck, 2011); xenarthrans, carnivorans, and ungulates (Voss and Fleck, 2017); didelphimorphians (Voss et al., 2019); bats (Velazco et al., 2021); and rodents (Voss et al., 2024). Not included in these lists are two rodent species (Holochilus nanus and Oligoryzomys microtis) that are restricted to open (grassy or marshy) vegetation along river beaches. Voucher specimens that document taxonomic identifications of species from the Yavarí-Ucayali interfluve are at the AMNH and the Museo de Historia Natural de la Universidad Nacional Mayor de San Marcos (MUSM, Lima).
Nomenclature: Scientific names in the source literature cited above have been updated to conform with current usage as summarized by Catzeflis et al. (2013, 2016), Pavan et al. (2018), and Velazco et al. (2021) for bats; Patton et al. (2015), Salazar-Bravo et al. (2023), and Voss et al. (2024) for rodents; Voss (2022) for opossums; Miranda et al. (2018) for pygmy anteaters (Cyclopes); Acosta et al. (2020) for tayassuids; and Morales-Donoso (2023) for cervids. To conserve long-standing binomial usage (Teta, 2019), I treat Sapajus (tufted capuchins) as a subgenus of Cebus, Cebuella (pygmy marmosets) as a subgenus of Callithrix, Leontocebus (a clade of tamarins) as a subgenus of Saguinus, Plecturocebus (a clade of titi monkeys) as a subgenus of Callicebus, and Neogale as a subgenus of Mustela. As discussed by Voss and Fleck (2011), Lima et al. (2018), and Martins et al. (2023), there seems to be no compelling evidence for more than a single Amazonian species of tufted capuchins (Cebus apella). Although analyses of mitochondrial DNA have revealed numerous haplogroups of Amazonian squirrel monkeys (Lynch Alfaro et al., 2015), multiple lines of evidence suggest that all represent a single species (Saimiri sciureus).
Complementarity: The dissimilarity (D) between two species lists, X and Y, can be quantified as the number of species unique to one or the other as a fraction of the species in both. In set notation,
where the brackets for absolute value indicate the number of elements (species) in the corresponding union or intersection. This quantity, which Colwell and Coddington (1994) called complementarity, can vary from zero (for two identical lists) to 1 (for two lists with no species in common). As noted by Levandowski and Winter (1971), D is the one-complement of Jaccard's familiar index of similarity. Because decimal fractions are awkward to discuss in running text, complementarity values are commonly multiplied by 100 and described as percentages.
Although complementarity is a straightforward measure of faunal difference when computed from complete species lists, incomplete faunal sampling tends to overestimate complementarity by inflating the number of unique species (Colwell and Coddington, 1994). Faunal sampling completeness at Paracou was discussed by Simmons and Voss (1998) and Voss et al. (2001), who concluded that numerous species of bats and several species of nonflying mammals remained undetected there, but almost all the missing species they discussed have either been discovered at Paracou by subsequent researchers or are known from adjacent research sites (e.g., Petit Saut and Nouragues), such that the species lists from northern French Guiana (appendices 1, 2) are plausibly complete. Faunal sampling in the Yavarí-Ucayali interfluve (reviewed by Voss and Fleck, 2011, 2017; Voss et al., 2019, 2024; Velazco et al., 2021) is thought to be nearly complete, with the important exception of high-flying insectivorous bats (molossids, diclidurine emballonurids, and Lasiurus spp.), which were ineffectively and haphazardly sampled for methodological reasons, and cave-roosting bats, which might occur in rocky outcrops of the remote and very incompletely surveyed Sierra del Divisor.
I computed complementarity between whole-faunal lists from northern French Guiana and the Yavarí-Ucayali interfluve and for selected taxonomic groups (opossums, xenarthrans, bats, primates, carnivorans, ungulates, and rodents) that were sampled in both regions. Acknowledging the probable undersampling of high-flying insectivorous and cave-roosting bats in the Yavarí-Ucayali interfluve, I also computed complementarity between subsets of the bat fauna by excluding species in those categories.
Missing species: Species with geographic ranges that overlap a region but were not observed there are candidate “missing” species (sensu Voss and Emmons, 1996) or “pseudo-absences” (sensu Bogoni et al., 2022). As explained above, the French Guianan species lists for nonflying mammals and bats are believed to be complete, or nearly so, because they include all the species with geographic ranges that overlap the region; however, this criterion might be unreliable for a coastal area, beyond which no species of terrestrial mammal can occur. The species potentially missing from the Yavarí-Ucayali fauna were discussed by Voss et al. (2019: appendix 4), Velazco et al. (2021: appendix 2), and Voss et al. (2024: appendix 5); they include 3 opossums, 18 bats, and 4 rodents. Because the criterion of geographic overlap might bias the comparison of coastal and interior faunas, I provide comparisons of northern French Guiana with the observed fauna from the Yavarí-Ucayali interfluve and with the “expected” fauna (including candidate missing species) of the latter region.
Body Mass and Ecobehavioral Traits
Body mass: I extracted mass data from the faunal monographs cited above and from other literature (e.g., Richard-Hansen et al., 1999), but I consulted specimen labels or fieldnotes for a few species without published sources of mass data. Insofar as possible, tabulated masses in appendices 1–4 are means of nonpregnant adults (for sexually monomorphic species) or midpoints of male and female means (for size-dimorphic species); for a few species, however, I used the midpoint of the range of values reported by Emmons (1997) for samples of unknown sex composition. Whenever possible, I used mass data of regional origin to allow for geographic variation; however, when regional data were unavailable or were based on excessively small samples, I used data pooled from geographically heterogeneous sources. Masses of nonvolant mammals are rounded to the nearest gram (g), but masses of large mammals were often recorded at coarser intervals (e.g., to the nearest 20 g by Richard-Hansen et al., 1999), and only kilograms (kg) are recorded in the literature for very large species. I rounded bat masses to the nearest 0.1 g following the prevailing custom among bat researchers (e.g., Simmons and Voss, 1998; Velazco et al., 2021). Mass data are unavailable for just two mammals treated in this report (Mustela africana and Coendou ichillus).
Ecobehavioral traits: Information about diel activity (nocturnal, diurnal, cathemeral) and substrate use (arboreal, semiaquatic, scansorial, terrestrial) for nonflying species were obtained from camera-trap surveys (e.g., Gómez et al., 2005; Blake et al., 2012; Bowler et al., 2017; Gregory et al., 2022), conventional trapping studies (e.g., Malcolm, 1991; Patton et al., 2000), direct-observational studies (e.g., Emmons, 1981; Dubost, 1988; Soini and Soini, 1992), interviews with indigenous Amazonians (Voss and Fleck, 2011, 2017; Voss et al., 2024), and fieldnotes. In reviewing fieldnotes, specimens recovered at dawn from traps baited in the late afternoon of the previous day were assumed to have been caught at night. Because game species are known to shift their diel activity near human communities (Mendes et al., 2020), I tabulated activity from the results of camera-trap surveys at remote, unhunted sites whenever possible.
I compiled dietary information for bats from studies that summarized results of fecal analyses or that reviewed such results from previous research (e.g., Giannini and Kalko, 2004, 2005; Delaval et al., 2005; Pellón et al., 2023), and I assumed that species for which such data are lacking have diets resembling those of congeners with known diets (e.g., that unstudied Saccopteryx gymnura trophically resembles well-studied S. bilineata and S. leptura). Additionally, dietarily unstudied species belonging to clades that are morphologically specialized for nectarivory (glossophagines and lonchophyllines) were assumed to be nectarivores. In general, phyllostomids are readily divisible into primarily phytophagous or animalivorous species, but published evidence suggests that Lampronycteris brachyotis, Trinycteris nicefori, and species of Phyllostomus consume substantial quantities of fruit and/or nectar in addition to arthropods (Delaval et al., 2005; Giannini and Kalko, 2005; Pellón et al., 2003) and should be classified as omnivores.
Trophic-Guild Membership
Community structure is often discussed in terms of species membership in trophic guilds: sets of species that use the same food resource in similar ways. Unfortunately, many species lack sufficiently detailed information to consistently distinguish (for example) frugivore-granivores from granivore-frugivores—just two among other possible permutations of phytophagous species with mixed diets. Therefore, I defined trophic guilds of nonflying mammals by cross-classifying species by diel activity (diurnal, nocturnal, cathemeral), substrate use (arboreal, terrestrial, scansorial, semiaquatic), and trophic level (primary consumers, secondary consumers, and omnivores). Only a few of the 36 resulting categories (diurnal arboreal primary consumers, nocturnal terrestrial secondary consumers, etc.) are useful for faunal comparisons because some categories are unoccupied—for example, there are no diurnal arboreal secondary consumers in either fauna—whereas other categories (e.g., diurnal semiaquatic secondary consumers) are occupied by the same species throughout Amazonia.
I recognize the following bat trophic guilds: aerial insectivores, gleaning animalivores (including piscivores), gleaning frugivores, gleaning nectarivores, gleaning omnivores, and gleaning sanguivores. More or less the same ecobehavioral categories are widely recognized by Neotropical bat researchers, although there are minor disagreements about species membership. Among other occasionally disputed taxa, I classify Lampronycteris brachyotis and Trinycteris nicefori as omnivores and Phylloderma stenops as a frugivore based on dietary data summarized by Giannini and Kalko (2005: appendix), and I classify species of Thyroptera—formerly considered aerial insectivores (e.g., by Sampaio et al., 2003)—as gleaning animalivores because they appear to mostly eat nonflying arthropods (Dechmann et al., 2006).
RESULTS AND DISCUSSION
Richness and Taxonomic Composition
The observed rainforest mammal faunas of northern French Guiana and the Yavarí-Ucayali interfluve consist of 180 and 199 species, respectively (table 1). The same major groups—opossums, xenarthrans, bats, primates, carnivorans, ungulates, and rodents—are present in both faunas, and the most speciose groups are represented in roughly equivalent proportions: bats are by far the most species-rich ordinal-level clade in each fauna, followed by rodents, opossums, and carnivorans. The same Linnaean families are also present in both faunas (appendices 1–4), with two exceptions: Mormoopidae (present in northern French Guiana but not in the Yavarí-Ucayali interfluve) and Dinomyidae (present in the Yavarí-Ucayali interfluve but not in northern French Guiana). Despite such higher-taxonomic resemblances, there are noteworthy faunal differences in numbers of species. Primates and rodents, in particular, are more speciose in the Yavarí-Ucayali interfluve than in northern French Guiana. Adding missing species to the expected fauna of the Yavarí-Ucayali interfluve (in the rightmost column of table 1) does not affect the diversity ranking of groups but increases the percent representation of bats.
The higher species richness of the Yavarí-Ucayali interfluve by comparison with northern French Guiana—an increment of about 10% based on observed species, but almost 25% if based on expected species)—is consistent with the gradient of increasing species richness from eastern to western Amazonia that Voss and Emmons (1996) hypothesized from geographic range data. Richness contours subsequently generated by computerized stacking of range maps have illustrated this hypothetical gradient for nonflying mammals (Tognelli and Kelt, 2004: fig. 1a) and for various subsets of the nonflying fauna (Fergnani and Ruggiero, 2015: fig. 4). The species lists reported here for northern French Guiana and the Yavarí-Ucayali interfluve—based on fieldwork, specimen collection, and taxonomic analysis—provide empirical confirmation of this predicted gradient and establish a plausible range (180–224 species) to bracket diversity expectations within other faunally homogeneous Amazonian regions.
TABLE 1.
Taxonomic composition of Amazonian rainforest mammal faunas in northern French Guiana (NFG) and the Yavarí-Ucayali interfluve (YUI).
![img-z9-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z9-2_01.gif)
Complementarity
Complementarity between the observed faunas of northern French Guiana and the Yavarí-Ucayali interfluve is 56% (table 2): over half the total species present in these faunas occur in one region or the other, but not in both. Such a large fraction is perhaps unsurprising for faunas that are separated by almost 2500 km, but complementarity values computed for different groups suggest wide taxonomic divergence in geographic filtering. Complementarity is notably high (79%–89%) for opossums, primates, and rodents, only a few species of which are common to both faunas. By contrast, complementarity is much lower (0%–24%) for ungulates and carnivorans, many species of which are found throughout Amazonia. Bat complementarity (42%) is probably biased by incomplete sampling of high-flying insectivorous and cave-roosting species in the Yavarí-Ucayali interfluve (see Methods); for the well-sampled subset of these bat faunas, complementarity is only 32%. Thus, the dissimilarity between these bat faunas seems to be substantially less than the faunal dissimilarities observed for opossums, primates, and rodents. Two possible explanations merit consideration.
TABLE 2.
Complementarity between rainforest mammal faunas in northern French Guiana and the Yavarí-Ucayali interfluve.
![img-z10-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z10-2_01.gif)
First, complementarity estimates are sensitive to taxonomic resolution. The complementarity values recovered in this analysis for marsupials, primates, and rodents are higher than those that would have been estimated just 30 years ago, when (for example) Marmosa constantiae was not distinguished from M. demerarae, Alouatta seniculus was not distinguished from A. macconnelli, and Hylaeamys perenensis was not distinguished from H. megacephalus. If bats were taxonomically neglected by comparison with those groups, bat faunas might exhibit lower complementarity as a consequence.
Second, it is possible that different landscape features or ecological factors limit the distribution of opossums, primates, and rodents on the one hand and bats on the other. The obvious limiting landscape features in Amazonia are large rivers, which have often been identified as dispersal barriers for primates (Ayres and Clutton-Brock, 1992; Fordham et al., 2020; Mouthé et al., 2022) and sometimes also for opossums (Rocha et al., 2015; Voss et al., 2019) and rodents (da Silva et al., 2017; Saldanha et al., 2019; Voss et al., 2024). By contrast, rivers are seldom suggested as distributional limits for Amazonian bats (the unique exception seems to be Carollia benkeithi; Solari and Baker, 2006), and riverine barriers do not explain geographic patterns of Amazonian bat endemism or community composition (Silva et al., 2022). Of course, these alternative explanations for the lower faunal complementarity of bats on the one hand versus opossums, primates, and rodents on the other are not mutually exclusive: bats could be both taxonomically neglected and unaffected by riverine barriers. However, the revisionary taxonomy of Neotropical bats is an active research field—a recent monograph (Velazco et al., 2021) cited 99 publications on this topic in the interval from 2000 to 2020—and numerous analyses of molecular data seem to support the conspecificity of geographically remote Amazonian bat populations. A relevant example is Silva et al.'s (2023) genetic survey of Rhinophylla pumilio, a small (ca. 9–10 g) understory frugivore, which exhibits <2% sequence divergence in protein-coding mitochondrial genes among populations scattered throughout Amazonia (on opposite banks of the Amazon and of all its major tributaries). Although cryptic diversity doubtless exists within some currently recognized species (e.g., Myotis riparius; Novaes et al., 2023), most evidence suggests that Amazonian bats tend to have substantially larger geographic distributions than small and medium-size nonflying Amazonian mammals, consistent with the complementarity values reported here.
TABLE 3.
Body mass statistics for rainforest mammals in northern French Guiana (NFG) and the Yavarí-Ucayali interfluve (YUI).
![img-z11-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z11-2_01.gif)
Size Distributions
Mammalian size distributions have long been scrutinized for evidence of ecological patterns and processes (e.g., by Hutchinson and MacArthur, 1959; Brown and Nicoletto, 1991; Rodríguez, 1999; Bakker and Kelt, 2000). To date, most studies have focused on mass data from nonflying mammals, which are not affected by the physics of flight and echolocation that constrain body size among bats (Norberg and Rayner, 1987; Barclay and Brigham, 1991; Jones, 1999). Rather than ignoring body size in bats, we analyze those data separately.
Nonflying species: Nonflying rainforest mammals in both northern French Guiana and the Yavarí-Ucayali interfluve vary in mass by four orders of magnitude (table 3), but only a few species in each fauna—giant armadillo (Priodontes maximus), giant anteater (Myrmecophaga tridactyla), giant otter (Pteronura brasiliensis), puma (Puma concolor), jaguar (Panthera onca), peccaries (Pecari tajacu, Tayassu pecari), deer (Mazama americana, Passalites nemorivagus), tapir (Tapirus terrestris), pacarana (Dinomys branickii), and capybara (Hydrochoerus hydrochaeris)—exceed 10 kg. Most nonflying mammals are smaller; 75% of the species in both faunas are less than about 6 kg. However, the median mass of nonflying mammals in the Yavarí-Ucayali interfluve (602 g) is substantially less than the median mass in northern French Guiana (1035 g), a difference that is explained by the higher richness of small rodents and opossums in the former region.
TABLE 4.
Descriptive statistics for log2-transformed masses of nonflying rainforest mammals in northern French Guiana (NFG), the Yavarí-Ucayali interfluve (YUI), and seven other Amazonian faunas.
![img-z12-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z12-2_01.gif)
Descriptive statistics of log2-transformed mass data (the usual currency of mammalian size-distributional comparisons) from northern French Guiana and the Yavarí-Ucayali interfluve closely resemble values previously reported for other Amazonian faunas (table 4). Consistent with Bakker and Kelt's (2000) results, the medians for both regions are higher than the medians that those authors reported for the entire South American fauna (7.5 log2 g) and for the Neotropical lowland rainforest biome (8.5 log2 g), whereas the regional interquartile ranges are larger than the corresponding continental (4.0 log2 g) and biome (4.5 log2 g) values. As interpreted by Bakker and Kelt (2000), such results illustrate the scale dependence of mammalian size distributions: medians and interquartile ranges both increase for mass data sequentially compiled from continents, biomes, and smaller regions of uniform habitat. Brown and Nicoletto (1991) suggested several hypotheses that might explain such scaling, but none are uniquely supported or convincingly refuted by data from Amazonian faunas.
Graphs of rank-ordered log-mass data (cenograms) are often used by paleontologists seeking ecological patterns in Recent faunas to interpret fossil data (Rodríguez, 1999; but see Alroy, 2000). Cenograms of nonflying mammals from northern French Guiana and the Yavarí-Ucayali interfluve are smoothly continuous arrays (fig. 2). Neither has any obvious gaps or abrupt changes of slope, consistent with results previously obtained by graphing mass data for other faunas from warm, humid environments (Legendre, 1986; Gingerich, 1989; Croft, 2001). In particular, both faunas include numerous species in the mass interval from 1 to 10 kg (ca. 10–13 log2 g), which other researchers have alleged is suboptimal for mammals (Smith and Lyons, 2011).
Bats: By contrast with sympatric nonflying mammals, bats from both regions range in mass by only two orders of magnitude (table 3). No bat in either fauna—even the largest nonpregnant specimens of the spectral bat (Vampyrum spectrum)—is known to exceed 200 g. The median mass of bats in both regions is in the range of 11–13 g, and the first and third quartiles of mass in both faunas are nearly the same.
Mass data for Amazonian bats have not previously been analyzed by descriptive statistics or graphing, so the literature provides no basis for relevant comparisons. In fact, few Amazonian bat researchers have compiled mass data suitable for comparative analyses, with the noteworthy exceptions of Lim and Engstrom's (2001) inventory of Iwokrama Forest (in Guyana) and Hice et al.'s (2004) inventory of the Allpahuayo-Mishana Reserve (in Peru). Despite smaller sample sizes, descriptive statistics of central tendency, dispersion, and skewness for log-mass data from both of those inventories are similar to those from northern French Guiana and the Yavarí-Ucayali interfluve (table 5). Cenograms for Iwokrama and Allpahuayo-Mishana (not shown) are also similar to those from northern French Guiana and the Yavarí-Ucayali interfluve (fig. 3), which show a steep decline in rank-ordered masses among the largest bats, followed by a gradual and almost linear decrease among the smaller species, without any conspicuous gaps. Consistent with theoretical expectations that echolocation constrains the maximum size of aerial hawkers (Barclay and Brigham, 1991; Jones, 1999), the largest bats in these faunas are all gleaning species (e.g., Vampyrum spectrum, Chrotopterus auritus, Phyllostomus spp., and Artibeus lituratus).
TABLE 5.
Descriptive statistics for log2-transformed masses of rainforest bats in northern French Guiana (NFG), the Yavarí-Ucayali interfluve (YUI), Iwokrama Forest (IF), and the Reserva Nacional Allpahuayo-Mishana (RNAM).
![img-z14-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z14-2_01.gif)
Diel Activity
The overwhelming majority of rainforest mammals in northern French Guiana and the Yavarí-Ucayali interfluve are nocturnal (table 6). This result is, of course, strongly influenced by bat richness, but even among nonflying mammals nocturnal species are more numerous than diurnal species by a ratio of about 2:1 in both regions. Only a few species are cathemeral (active both by day and by night); of these, most are large terrestrial forms—e.g., giant anteater, jaguar, puma, red brocket (Mazama americana)—but three-toed sloths (Bradypus spp.) are noteworthy exceptions.
To my knowledge, these are the first Neotropical rainforest faunas to be comprehensively scored for diel activity, but the results do not challenge conventional expectations about trait frequencies. The therian common ancestor is believed to have been nocturnal (Gerkema et al., 2013; Maor et al., 2017), and nocturnality predominates among Recent mammals, especially at low latitudes (Bennie et al., 2014). Most Amazonian species are perhaps nocturnal by direct inheritance from a common therian ancestor, but night monkeys (Aotus spp.) acquired this behavior by evolutionary reversal (Santini et al., 2015).
Substrate Use
Communitywide patterns of substrate use are also similar in northern French Guiana and the Yavarí-Ucayali interfluve (table 6), and not only because bats are so numerous in each fauna. Among nonflying mammals, most species are either arboreal (e.g., monkeys, squirrels, sloths, porcupines) or terrestrial (e.g., armadillos, large cats, ungulates, dasyproctid rodents, Proechimys spp.), with roughly equivalent numbers in both categories. Many fewer species in each fauna are scansorial, a somewhat heterogeneous category that includes species that travel on the ground but often climb into the canopy or subcanopy (e.g., Didelphis marsupialis, Tamandua tetradactyla, Eira barbara, Leopardus wiedii) and others that climb in understory vegetation but often descend to the ground (e.g., Marmosops spp., Oecomys spp.). Still smaller numbers of species in each fauna are semiaquatic (e.g., Chironectes minimus, Nectomys spp., Lontra longicaudis, Pteronura brasiliensis, Hydrochoerus hydrochaeris). Remarkably, no Amazonian mammal is definitely known to be fossorial, although indigenous reports of such behavior for the armadillo Cabassous unicinctus are noteworthy (Voss and Fleck, 2017). Adding missing species to the expected fauna of the Yavarí-Ucayali interfluve brings the proportional representation of several groups (aerial, arboreal, terrestrial) somewhat closer to the percentages observed in northern French Guiana but leaves the proportional representation of others (scansorial, semiaquatic) unchanged.
As for diel activity, these are the first Neotropical rainforest faunas to be comprehensively scored for substrate use. The predominance of aerial species in both faunas follows from the high species richness of bats and requires no other explanation, but the nearly equivalent numbers of arboreal and terrestrial species seem counterintuitive. Most edible primary production (e.g., of leaves, flowers, fruit, and seeds) occurs in the canopy, so it would be reasonable to think that more primary consumers would be arboreal than terrestrial, and that predatory species would pursue them in the treetops. Three observations are probably sufficient to explain why this is not the case. First, a substantial fraction of canopy fruit and seed production eventually falls to the forest floor, where it is accessible to terrestrial frugivores and granivores (Smythe, 1986; Feer et al., 2001). Second, carnivorous mammals large enough to prey on monkeys, sloths, porcupines, and other arboreal primary consumers are too heavy to hunt effectively in trees (although jaguars sometimes do so in flooded forest; Ramalho et al., 2021); instead, the usual predators of arboreal primary consumers are eagles (e.g., Harpia harpyja; Miranda, 2018). Lastly, the forest floor is a rich source of soil invertebrates and detritus-feeding arthropods (Penny and Arias, 1982; Fragoso and Lavelle, 1992), both of which are fodder for mammalian secondary consumers and omnivores (e.g., armadillos and terrestrial opossums; Fleck and Voss, 2016; Voss and Jansa, 2020).
TABLE 6.
Ecobehavioral traits of rainforest mammals in northern French Guiana (NFG) and the Yavarí-Ucayali interfluve (YUI).
![img-z16-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z16-2_01.gif)
Trophic Level and Guild Membership
By contrast with faunal similarities in diel activity and substrate use, trophic comparisons reveal substantial divergence in relative numbers of primary versus secondary consumers (table 6). Whereas secondary consumers are more numerous than primary consumers in the observed fauna of northern French Guiana, primary consumers outnumber secondary consumers in the observed fauna of the Yavarí-Ucayali interfluve. Only a handful of species (<10%) in each fauna are omnivores. Including missing species (many of which are insectivorous bats) to the expected fauna of the Yavarí-Ucayali interfluve reverses the relative numbers of primary versus secondary consumers, suggesting that at least some of the proportional difference between the observed faunas is an artifact of incomplete sampling. However, the increase in primary consumers from northern French Guiana (65 species observed) to the Yavarí-Ucayali interfluve (94 species expected) amounts to a 45% difference, whereas the increase in secondary consumers (97 species observed in northern French Guiana versus 104 species expected in the Yavarí-Ucayali interfluve) amounts to only a 7% difference. Therefore, whether the observed or the expected fauna of the Yavarí-Ucayali interfluve is accepted as a valid basis for comparison, the disparate allocation of species to trophic categories is substantial.
TABLE 7.
Membership in six guilds of nonflying rainforest mammals in northern French Guiana (NFG) and the Yavarí-Ucayali interfluve (YUI).
![img-z17-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z17-2_01.gif)
The six most populated trophic guilds of nonflying mammals in northern French Guiana and the Yavarí-Ucayali interfluve are permutations of diurnal/nocturnal, arboreal/terrestrial, and primary/secondary consumers (table 7). Of these, four are occupied by the same or similar numbers of species in each fauna, but there are >80% more diurnal-arboreal primary consumers and almost 70% more nocturnal-terrestrial primary consumers in the Yavarí-Ucayali interfluve than there are in northern French Guiana. Including expected species in the Yavarí-Ucayali interfluvial fauna also reveals a modest potential increase in the number of nocturnal-arboreal primary consumers.
The same or similar numbers of species populate most bat trophic guilds in northern French Guiana and the Yavarí-Ucayali interfluve, with two conspicuous exceptions (table 8). Far fewer aerial insectivores are known to occur in the Yavarí-Ucayali interfluve than in northern French Guiana, but, as noted previously, this deficit is probably the result of undersampling high-flying and cave-roosting taxa; including candidate missing species in the expected fauna of the Yavarí-Ucayali interfluve suggests that it hosts almost the same number of aerial insectivores as northern French Guiana. By contrast, there are almost 50% more frugivorous bats in the Yavarí-Ucayali interfluve than there are in northern French Guiana, and this large increment cannot plausibly be explained by sampling bias (see below).
TABLE 8.
Membership in six guilds of rainforest bats in northern French Guiana (NFG) and the Yavarí-Ucayali interfluve (YUI).
![img-z18-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z18-2_01.gif)
Given the higher-taxonomic (ordinal and family-level) similarity of faunas from northern French Guiana and the Yavarí-Ucayali interfluve, these differences in trophic structure necessarily reflect regional differences in generic composition and congeneric species richness. Of the 16 genera present in the Yavarí-Ucayali interfluve that do not occur in northern French Guiana, 11 are primary consumers (Enchisthenes, Sphaeronycteris, Vampyressa, Aotus, Cacajao, Callicebus, Lagothrix, Bassaricyon, Dinomys, Dactylomys, Toromys), whereas of the 10 genera present in northern French Guiana but not observed in the Yavarí-Ucayali interfluve, only 4 are primary consumers (Lichonycteris, Ametrida, Rhipidomys, Echimys). More impressively, the 13 shared genera of primary consumers that differ in richness between these faunas include 45 species in the Yavarí-Ucayali interfluve but only 28 species in northern French Guiana (table 9), and a two-tailed sign test indicates that the median species richness in shared genera of primary consumers differs significantly between these faunas (p = 0.02).
Such trophic disparities cannot be explained by sampling incompleteness because primary consumers are among the easiest species to observe in rainforest faunas using standard inventory methods (Voss and Emmons, 1996). Most frugivorous bats, even those that forage in the canopy, are readily captured in ground-level mistnets; frugivorous and folivorous monkeys are large, noisy, and diurnal; ungulates and large terrestrial rodents leave visible spoor; and most small terrestrial rodents can be taken in commercially available traps with standard kinds of bait. Canopy-dwelling nocturnal rodents are exceptions (silent, hard to see, and hard to trap), but geographic range data suggest that none were missed in northern French Guiana.
A Common Cause for Faunal Differences
As we have seen, mammal faunas from northern French Guiana and the Yavarí-Ucayali interfluve are similar in many respects—including higher-taxonomic composition, size distribution, diel activity, and substrate use—but differ in species richness and trophic structure. Although such differences are logically independent (faunas can presumably differ in species richness but not in trophic structure, and vice versa), they might have a common cause in Amazonia. The observed increase in primary consumers from northern French Guiana to the Yavarí-Ucayali interfluve (24 species; table 6) is more than sufficient to account for the observed increase in species richness (19 species; table 1), whereas the expected increase in primary consumers (29 species) accounts for almost two-thirds of the expected increase in species richness (44 species). These numbers suggest that the east-to-west gradient of increasing species richness of Amazonian mammal faunas (Voss and Emmons, 1996; Peres, 1999) could be largely explained by the surplus of primary consumers in western Amazonia. This hypothesis seems to be supported by stacked range maps, which show conspicuous east-to-west increases in species density for clades with many primary consumers (caviomorph rodents, primates) but no equivalent increases for clades with numerous secondary consumers (xenarthrans, carnivorans; Fergnani and Ruggiero, 2015: fig. 4).
TABLE 9.
Shared genera of primary consumers that differ in numbers of observed species in northern French Guiana (NFG) and the Yavarí-Ucayali interfluve (YUI).
![img-z19-2_01.gif](ContentImages/Journals/novi/2024/4019/4019.1/graphic/WebImages/img-z19-2_01.gif)
Geographic variation in mammalian species richness has often been attributed to environmental differences in primary productivity (Emmons, 1984; Voss and Emmons, 1996; Kay et al., 1997; Tognelli and Kelt, 2004; Peres, 2008). In Amazonia, the results of several studies suggest that primary production is limited by soil fertility (Malhi et al., 2004; Aragão et al., 2009; Cunha et al., 2022) and covaries geographically with soil fertility and climate (Quesada et al., 2012). In effect, forests growing on fertile soils in regions with abundant rainfall throughout the year tend to be more productive than those growing on nutrient-impoverished soils in regions with less abundant or more seasonally distributed rainfall. There are no consistent differences in rainfall amount or seasonality between eastern and western Amazonia (Salati and Marques, 1984; Sombroek, 2001), but western Amazonian soils are younger (recently weathered from the Andes) and generally more fertile than eastern Amazonian soils (weathered in situ from ancient geological shields; Hammond, 2005a; Quesada et al., 2010). Although upland soils everywhere in Amazonia tend to be highly weathered and nutrient poor, the Acrisols and Ferralsols that blanket most of eastern Amazonia are among the most nutrient-deficient soils to be found anywhere in the rainforested Neotropics, whereas the Fluvisols that occur in the floodplains of western Amazonian rivers are among the most fertile. Maps of soil phosphorous content—the single most consistent predictor of Amazonian primary productivity—show a clear trend of increase from east to west (Quesada et al., 2010: fig. 2b). Therefore, although direct measurements of edible primary production (bark, leaves, green stems, flowers, fruits, and seeds) are unavailable for relevant comparisons, it seems probable that, averaged across the landscape, the rainforest vegetation of western Amazonia is substantially more productive than that of eastern Amazonia.
Habitat mosaicism may be another factor contributing to the primary productivity of western Amazonia, a geomorphologically dynamic landscape where rivers meandering across broad sedimentary floodplains create a variety of riparian formations with phenological schedules that are out of phase with those of adjacent terra firme forests; such offset phenologies sustain yearlong primary production, even in regions with pronounced annual dry seasons (Terborgh, 1983; Diaz-Martin et al., 2014; Haugaasen and Peres, 2007). By contrast, eastern Amazonia is a geomorphologically static landscape where most rivers, closely constrained in their courses by structural features of the underlying geology, do not have equivalently broad floodplains, and the regional vegetation (except where interrupted by savannas) consists almost entirely of terra firme forest (Hammond, 2005b).
Despite multiple alternative mechanisms by which higher primary productivity might result in higher animal species richness (Evans et al., 2005; Storch et al., 2018), it follows from the inefficiency of energy transfer between trophic levels that any positive effects should be stronger for primary consumers than for secondary consumers (Wright, 1983). In terms of diversification processes, differences in primary production seem unlikely to affect speciation rates, but they could affect extinction rates if consumer population densities are higher in more productive regions, and if larger populations are less likely to go extinct. Because such an effect would be substantially stronger for primary consumers than for secondary consumers, one plausible scenario is that, even if speciation rates were constant across an extensive biome like Amazonia, a gradient of primary productivity could result in geographic variation in both species richness and trophic structure as more species of primary consumers than secondary consumers accumulate in highly productive regions. Given the east-to-west gradients in soil chemistry and geomorphology described above, such a scenario could explain faunal differences not just between northern French Guiana and the Yavarí-Ucayali interfluve, but across the much larger Amazonian landscapes within which these exemplar faunas are embedded.
Intercontinental Comparisons: Two Caveats
Are rainforest mammal faunas on different continents similarly species rich and ecologically structured? Numerous researchers have addressed this question with a variety of data and approaches (e.g., Bourlière, 1973, 1983; Dubost, 1987; Emmons, 1995; Rovero et al., 2020; Tejada et al., 2020; Tanshi et al., 2022). Inevitably, the data provided in this report (appendices 1–4) will be used for intercontinental comparisons, but two caveats are in order. First, no African or Asian rainforest fauna has been inventoried as intensively, nor has any been as thoroughly analyzed taxonomically, as the Amazonian faunas treated here. Although the biases introduced by incomplete faunal sampling and inadequate taxonomic analysis cannot be predicted with certainty, they probably include underestimates of species richness for behaviorally and morphologically cryptic taxa such as bats, eulipotyphlans, and muroid rodents.
The second caveat concerns the end-Pleistocene anthropogenic extinction of megafaunal species, which was severe in South America but not in Africa or tropical Asia (Sandom et al., 2014; Smith et al., 2018). Assessing the impact of such extinctions for Amazonian faunas is difficult because the relevant geological record is sparse. However, Lujanian (latest Pleistocene) fossils from southwestern Amazonia include the gomphotheriid proboscidean Notiomastodon platensis (ca. 6300 kg); the ground sloth Eremotherium laurillardi (ca. 3500 kg); the toxodontid notoungulates Trigodonops lopesi (ca. 1900 kg) and Toxodon platensis (ca. 1800 kg); the camelid Palaeolama major (ca. 280 kg); and the pampatheriid armadillo Holmesina rondoniensis (ca. 120 kg). All these species were herbivores, and stable-isotopic data (δ13C, δ18O) reported by Asevedo et al. (2021) suggest that they could have browsed on understory foliage, a feeding niche that is conspicuously underpopulated in modern Amazonian faunas. Whether these or other extinct taxa were sympatric with the species currently found in northern French Guiana and the Yavarí-Ucayali interfluve is unknown, but it seems certain that the faunas of both regions were taxonomically and ecologically depleted by humans within the last few thousand years.
In summary, meaningful intercontinental comparisons of rainforest mammal faunas are simply not possible without (1) substantial investments of research effort to complete African and Asian inventories, and (2) without taking into account the probable effects of end-Pleistocene extinctions. Both are worthy endeavors because intercontinental comparisons could lead to novel insights about mammalian ecology and evolution in the most biologically diverse landscapes on the planet.
ACKNOWLEDGMENTS
The analyses summarized in this report would have been impossible without the prior publication of faunal monographs coauthored with Nancy B. Simmons, Darrin P. Lunde, David W. Fleck, Sharon A. Jansa, Paúl M. Velazco, and Thomas C. Giarla. Trained as a taxonomist, I particularly valued the ecological and behavioral insights contributed by Louise H. Emmons, who constructively criticized most of our manuscripts over the past several decades. Patricia J. Wynne produced the final versions of the figures for this report. Lastly, I thank Edson F. de Abreu Jr. and Pierre-Olivier Antoine for reviewing the text and making helpful suggestions to improve it.
Copyright © American Museum of Natural History 2024