OUP user menu

Structure of a Neotropical Savanna Bat Community

Luis F. Aguirre
DOI: http://dx.doi.org/10.1644/1545-1542(2002)083<0775:SOANSB>2.0.CO;2 775-784 First published online: 27 August 2002


During the 5-month dry season, I studied the community structure of bats in the Neotropical savanna of Espíritu, Bolivia. This seasonally inundated, grassland-dominated savanna is inhabited by ≥38 species of bats, and diversity is high (H′ = 2.88). Species richness is highest in forest islands (26 species), open woodlands and gallery forests (11 species each), and dense thorny bushes (7 species) and less in other plant associations. Although species from almost all trophic guilds filled by New World bats occur in the area, small medium-size low-flying insectivorous species and medium-to-large fruit-eating bats predominate.

Key words
  • Bolivia
  • chiroptera
  • community structure
  • diversity
  • Neotropical savanna
  • South America
  • trophic guilds

In many tropical systems, bats can represent >50% of the species in a mammalian fauna (Eisenberg 1989; Handley 1966; Wilson 1983). The study of bat ecology in the Neotropical region has been growing rapidly in the last 3 decades (Baker et al. 1976, 1977, 1979; Fleming 1988; Handley 1976; Hill and Smith 1984; Kalko et al. 1996; Kunz 1982, 1988; Medellín 1993; Medellín et al. 1992; Willig et al. 1993, 2000; Wilson 1973). The high diversity of bats, combined with the fact that they are integral components in a number of food webs, suggests that they are important regulators of complex ecological processes in Neotropical forests (Bonaccorso 1979; Fenton et al. 1992; Findley 1993; Fleming et al. 1972; Humphrey et al. 1983; LaVal and Fitch 1977; Medellín 1993; Ochoa 2000; Wilson 1973). Bat communities of Neotropical savanna habitats are poorly known compared with bat communities in other South American habitats, and only a few basic mammalian surveys have been conducted (Eisenberg et al. 1979; Ibañez 1981; Ochoa et al. 1988; but see Medellín and Redford 1992 for discussion). In Bolivia, bats represent 33% of the 345 known mammalian species, including 106 species from 8 families (Aguirre 1999; Anderson 1997). Little or no reliable information is available on structure of bat communities in Bolivia.

When analyzing the structure of bat communities, several attributes can provide information about resource partitioning and habitat use (Fleming et al. 1972; Kalko 1997). The measurement of alpha diversity describes the general assemblage at a local scale. Combining alpha diversity with distribution of individuals within species makes it possible to estimate levels of abundance in particular habitats. However, neither lists of species nor overall measurements of diversity provide complete information about structure of local communities. One way to enhance the study of bat communities is to sort members into more informative groups such as guilds mainly on the basis of diets and morphology (Findley 1976; Kalko 1997).

I examined the structure of the bat community of a Neotropical savanna in northeastern Bolivia. More specifically, I evaluated species richness, relative abundance, and trophic distribution of bats in different plant associations in the savanna and similarities between plant associations. The study of this particular community is interesting because it is located in a transitional zone between wet tropics to the north and arid Chaco to the south, a naturally fragmented forest–savanna matrix.

Materials and Methods


The study was conducted during the dry season from 26 June 1992 to 19 October 1992, and from 29 July to 30 August 1995. The study site (Espíritu, Bolivia) is a temporarily inundated savanna on the Yacuma River, located at 14°08′S, 66°24′W at 170 m elevation. This area is part of the Llanos de Moxos, which is one of the largest flooded Neotropical savannas in South America. It is inundated almost 7 months a year, generally from late December to early June. Average annual temperature is >26°C, and annual precipitation is just below 2,000 mm. The study area includes almost 10,000 ha of grasslands, open woodlands, riparian forests, and forest islands. The main factors determining present distribution of vegetation are extensive periodical inundation, slight changes in elevation, and human influences such as grazing, fire, or drainage control.

I sampled the following plant communities (as defined by Beck [1983]). Forest islands: These formations occupy less than 2% of the savanna; the distance between islands ranges from 0.1 to 3 km. Undergrowth is very open and, in larger islands, sometimes undisturbed. Different sizes of forest islands were sampled. Small forest islands (0.4–1.1 ha), corresponding to 2 sites; medium forest islands (1.1–4 ha), 4 sites; and a large forest island (12 ha), 1 site. These communities represent the higher and rarely flooded part of the savanna. Gallery forest: These communities occupy almost 10% of the study area. There are 2 kinds of gallery forests. Along the Yacuma River the forest is dense and high (10–20 m), and by the Carnaval Stream the forest is very dense and lower (2–9 m). Forests at both these sites may be flooded for 2–11 months per year. Tabebuia heptaphylla communities: These communities, locally known as Tajibales, occupy almost 10% of the study area, constituting open woodlands where the canopy coverage is less than 40%. Tajibales range 2–15 m in height and can be flooded during the rainy season for 3–5 months. I sampled 2 of these communities near large forest islands. Machaerium hirtum communities: These communities, locally known as Tusecales, cover less than 1% of the total area. They extend as small islands 5–50 m wide and 3–8 m high and may be flooded for 1–2 months a year. Bats were studied in 1 locality near a large forest island. Grasslands: This type of habitat represents almost 20% of the savanna formation and can be flooded for 1–2 months a year. I sampled bats along a trail in the xerophitic pasture ground. I also recorded some bats by direct observation here. Anthropogenic area: Some bats were sampled around man-made structures using mist nets.

Bats were sampled over a period of 3–5 nights at each of the 14 sites using mist nets 6–12 m long. Some bats were collected directly from roosts, but information on them was not included in analyses of diversity and community structure. Nets were open from 1800 h to 0000 h and occasionally until 0600 h. The sampling effort varied depending on the site: 3 nights in small habitat patches and up to 5 nights in larger patches. The total sampling effort was estimated using the method described by Medellín (1993), as the product of length of all nets in meters multiplied by the total hours worked (M × H). With this number it is possible to estimate relative abundance of bats, dividing the number of individuals collected by M × H (expressed as number of bats per meter per net hour).


A species accumulation curve allows determination of how well the bat community was sampled. It is defined as having reached an asymptote of absolute values when new bats are rarely added to the list. I used an asymptotic equation based on the model of Von Bertalanffy (Soberón and Llorente 1993). This equation is based on the following relation: S = (a/b)(1 − e−bt), where S is the species richness of the community at time t, and a and b are the parameters of the equation to be calculated by iterations. This equation permits an estimation of the total number of species in the community (Soberón and Llorente 1993).

To analyze diversity at each site I used the Shannon–Wiener Index (H′) based on the formula H′ = −Σpi × ln(pi), where pi is the proportion of the total number of individuals constituted by species i. I also used an evenness index (E), based on the formula E = H′/Hmax, where Hmax is the natural logarithm of the number of species (Shannon and Weaver 1963). To compare species composition among sites, I used the Morisita–Horn Index (Magurran 1988). This quantitative similarity index is based on the formula CmH = 2Σ(ani × bni)/(da + db)(aNbN), where aN and bN are the total number of individuals in inventories A and B, respectively, and ani and bni are the numbers of individuals in the ith species in A and B: da = Σani2/aN2 and db = Σbni2/bN2. Finally, I performed a cluster analysis of the Morisita–Horn values using group-average clustering (Crisci and Lopez 1983) to summarize relative similarity of bat community structure across plant formations.

Bats were grouped by trophic guilds using information from Gardner (1977) and Bonaccorso (1979) and my observations of feeding habits. The grouping included fast-flying insectivores, slow-flying insectivores, frugivores, nectarivores, omnivores, carnivores, piscivores, and sanguivores.


During 51 nights of sampling, I used 2,896 m of mist net in sampling for a total of 289 bats caught. On average, nets were open for 5.7 h per night, using an average of 56.8 m per night (Table 1). By multiplying total length of net (in meters) by the total time (in hours) of sampling with nets per locality, I obtained a total of 54,045 M × H (meter-hours), which I used to quantify relative abundances of bats.

View this table:
Table 1.

I recorded and collected 466 bats of 38 species, representing one-third of the bat fauna known to occur in Bolivia (Aguirre 1999; Anderson 1997) and 54% of the mammals occurring in Espíritu (Aguirre and de Urioste 1994; Aguirre et al. 1996). Sampling only with mist nets, I collected 455 bats of 34 species. Another 15 bats of 4 additional species were found by other methods and were included only in the analysis of trophic guilds, not in the analysis of diversity and relative abundance (Appendix I).

The cumulative curve of species captured (Fig. 1) suggests that overall species richness was well sampled. During the first 30 nights I collected 30 species of bats, representing 88% of the total, and in the following 21 nights, only 4 more species were added. The asymptotic curve shows that the number of species on the 102nd night would still be 34. The parameters of the equation are a = 1.94, b = 0.057, and asymptote (a/b) = 34.03. The slight difference between the asymptote, the value for the 102nd night, and observed value (night 51) show the good representation already obtained.

Fig. 1.

Cumulative curve of bat species captured at Espíritu, Bolivia. Asymptote (a/b), where a and b are parameters of the equation S = (a/b)(1 − e−bt), shows that the community was well sampled

The Shannon–Wiener diversity index of H′ = 2.88 is slightly higher than the highest reported in literature, 2.85 for La Selva, Costa Rica, by LaVal and Fitch (1977) and 2.82 for Lacandona, Mexico, by Medellín (1993). The higher diversity is because of an increase in species evenness rather an increase in species richness (Table 1). For plant associations, the diversity values vary from 1.22 in small forest islands to 2.58 in large forest islands.

The most abundant species at Espíritu was Myotis nigricans, followed by Noctilio albiventris, Platyrrhinus lineatus, Desmodus rotundus, and Artibeus jamaicensis (Fig. 2). Together, these species represented 52% of the community, with relative abundance ranging from 0.0012 to 0.00059 individuals/(M × H).

Fig. 2.

Relative abundance of bats captured at Espíritu, Bolivia, with species ranked in order of abundance using the standardized method of Medellín (1993). Species ranks are shown in Appendix I

The cluster analysis (Fig. 3) based on the Morisita–Horn similarity index shows 3 main groups. The 1st group is formed by Tusecales, anthropogenic area, and small forest islands; the 2nd group is formed by fairly large habitats such as big forest islands and Tajibales; finally, the 3rd group is formed by medium forest islands and gallery forests.

Fig. 3.

Similarity in bat community structures between plant formations in Espíritu, Bolivia, based on group-average clustering of similarity values calculated using the Morisita–Horn index (see text)

Even though capture effort was higher in large forest islands (15,840 M × H), highest abundance of bats occurred in the anthropogenic area (0.047 individuals/[M × H]), where the least capture effort was made (1,152 M × H). Abundance was lowest in grassland, with 0.00066 bats/(M × H).


A large portion of the community of bats in the savanna and the plant formations of Espíritu depends, at least partially, on the presence of forested ecosystems, and the association with them is very high, as was noted by Medellín and Redford (1992). Moreover, considering that only 12% of the area is covered by forest (the remaining 88% is mostly swamps and grasslands), the species diversity is high, both in richness and in evenness, showing a good proportional distribution (evenness) of individuals per species.

As explained with the cumulative curve of captured species, the community of bats in Espíritu seems to have been well sampled. When plotting number of species against number of individuals to see sampling effort for each plant formation (Fig. 4), it is seen that in Tajibales the number of species is undersampled and that new species are likely to be captured (Fig. 4c). Most of the new species here probably come from neighboring forest islands, so the number of species found in Tajibales may depend in part on the surrounding habitat matrix. A similar situation occurs with gallery forest bats, where the community seems to be underestimated (Fig. 4g). Here, the difficulty of collecting bats because of dense vegetation may be the best explanation of a very low success in capturing bats, and I expect more species will be found.

Fig. 4.

Sampling effort for plant formations of Espíritu, Bolivia: a) anthropogenic area, b) Tusecales, c) Tajibales, d) small forest islands, e) medium forest islands, f) large forest islands, and g) gallery forest. Note that most of the curves are reaching asymptotes, showing that bats were well represented in most cases

The high diversity of bat species found in Espíritu can be explained by the strong relationship between bats and forest because the forest provides suitable roosts and food. For example, Tajibales are important plant formations for roosting for a vast number of insectivorous bats, such as molossids, noctilionids, and vespertilionids (Aguirre 1994). Forest islands also play an important role, not only in containing appropriate roosts for bats, but also in containing a high diversity of food items. Even though only 1 species was captured in the grasslands, resulting in a Shannon–Wiener Index value of 0, more species were seen but were difficult to identify and capture. The highest diversity indexes were in medium and large forest islands. These habitats offer a greater diversity of food and roosts than do other habitats in the study area. In contrast, Tajibales and Tusecales offer only roosts and relatively small amounts of food for a few species, most of them insectivorous. The diversity obtained in gallery forest results mainly from the Carnaval Stream site, where capture is easier than by the Yacuma River, where dense vegetation makes it almost impossible to find appropriate places to set nets. Overall evenness in Espíritu was quite large (0.82), and it was the same for almost every medium-to-large habitat (forest islands and Tajibales; Table 1). Here bats tend to be more uniformly distributed among species, but the opposite was found in small forest islands, where there is a high dominance of few species (1 site contained only 4 species, with 34 individuals captured).

Because this study was conducted during the dry season, it should be considered preliminary until similar analyses are conducted during the wet season. With the addition of research and further evaluation during the wet season, it is likely that more species will be added (e.g., migratory species in the genus Lasiurus). There are always seasonal changes in the savanna, both climatic and phenological, but the dry season of 1992 was unusually wet because of the climatic phenomenon of El Niño. Possibly this sample represents a good portion of the bat community of the wet season as well. Fruiting patterns and food availability (of insects and fruits) will modulate diversity and will have influence over species composition and individual abundances. During the wet season more plant species are fruiting (Beck 1983), and the abundance of insects is likely to be higher then than in the dry season. An important finding of the 1995 study was a 36% turnover rate (Aguirre 1996); over 30% of the bat community had changed between 1992 and 1995. Only a long-term study will yield more insight into the dynamics of this community in the savanna.

It is possible that the relative abundance of M. nigricans is overestimated because it was found only in a few places and in large quantities. Medellín (1993) reported on the relative abundance of 43 species in Lacandona (Mexico). The differences are apparently very high when compared with Espíritu (Morisita–Horn similarity index = 41%). However, 1 species (Carollia perspicillata) is similar in abundance between localities (0.000346 for Lacandona and 0.00035 for Espíritu). Overall, there are fewer bat species known from Espíritu during the dry season than from Lacandona.

When analyzing similarities among plant formations, 3 groups were clearly formed. First, a group formed by those environments that provide few resources, especially food, to bats (Tusecales, anthropogenic area, and small forest islands) and that probably are very important roosting places for some species known to occur in disturbed areas, such as M. nigricans and several species of molossids (Molossus rufus and Molossus molossus). Second, environments that are large and very close to each other, where large forest islands could influence the presence of species of bats such as Sturnira lilium, P. lineatus, and N. albiventris in Tajibales. The third group is formed by environments somewhat different (medium forest islands and gallery forests) but with a high dominance of a few species, such as D. rotundus, and a large proportion of species particular to those places, such as Tonatia silvicola and Artibeus anderseni for medium forest islands, and Trachops cirrhosus and Rhynchonycteris naso in gallery forests. Grasslands are the most distinct units, because only 1 species was found there (P. lineatus), and grasslands are exploited mainly by fast-flying insectivorous bats that are difficult to collect and not included in the calculations.

Slow-flying insectivores are the most abundant, with 12 species, followed by fast-flying insectivores and frugivores. The guild of piscivores is the only trophic guild with 1 species (Noctilio leporinus). In Espíritu, 2 of 3 known species of vampire bats are present, but only 2 species of nectarivores were found, 1 of them captured only once (Anoura geoffroyi). The number of nectarivorous species is very low considering that in other parts of the Neotropics 5–9 species are found (Fleming et al. 1972; Graham 1987), but the number is in concordance with numbers present in other savannas, such as in Brazil or Venezuela, with 2 species each (Ibañez 1981; Ojasti 1990). The low number of nectar-feeding bats (Glossophaginae) in Espíritu could be related to restrictions on dietary requirements, lack of suitable roosts, or the proximity of Espíritu to the border of the geographic distribution of nectarivores. On the other hand, the relatively high abundance of vampire bats could be explained by their relationship to cattle and domestic animals, which are common in the area. In the case of M. nigricans it is possible that sampling overestimated the relative abundance of this particular species. Even though the species is very abundant in the anthropogenic area (because of suitable man-made roosts), it is less common in other areas in Espíritu. The most common species across all habitats was P. lineatus. Data recorded on overlapping activities and feeding behavior between bats and crepuscular insectivorous birds (Caprimulgidae) in Espíritu have shown that there is no competition between them and that character displacement and different activity patterns might play an important role in building a noncompetitive scenario (de Urioste 1994).

The patterns of species richness found in Espíritu are consistent with those found in other tropical areas of Central and South America—especially the high richness of phyllostomids. The relatively high number of aerial insectivorous bats suggests that Espíritu is a transitional area between forested areas of the Amazon region and more open and dry areas of the Chaco region (Willig et al. 2000).


This work was partially funded by a grant from the German Cooperation and the Instituto de Ecología, Universidad Mayor de San Andrés. F. and M. Aguirre were very important in making this work possible. I thank B. and P. Bauer for allowing me to work in Espíritu, and W. Hanagarth, supervisor of this work, and R. de Urioste and F. Guerra for their support during the field work. I. Galarza and E. Matthysen made suggestions on the manuscript. P. H. Dilgard was most helpful in translating into English from the original. I offer special thanks to D. E. Wilson, J. Salazar-Bravo, R. Medellín, and an anonymous reviewer for providing constructive comments.


View this table:
Appendix I

Literature Cited

View Abstract