Phyllostomid bats of a fragmented landscape in the north-eastern Atlantic forest, Brazil

Journal of Tropical Ecology (2006) 22:531 542. Copyright 2006 Cambridge University Press doi:10.1017/s0266467406003385 Printed in the United Kingdom Phyllostomid bats of a fragmented landscape in the north-eastern Atlantic forest, Brazil Deborah Faria 1 Departamento de Zoologia, Instituto de Biologia, CP6 109, Universidade Estadual de Campinas, CEP 13081-970, Campinas, SP, Brazil (Accepted 20 March 2006) Abstract: This paper addresses the effects of habitat fragmentation on the phyllostomid bats of the Atlantic rain forest in Brazil, by comparing community structure (species richness and capture frequency) and the frequency of three bat species sampled along 36 transects encompassing six habitat categories: interiors and edges of large (>1000 ha) and small fragments (<100 ha), and the surrounding matrix of second-growth forests and areas of shade cocoa plantation. Species composition, richness and total captures were not directly affected by forest size per se, although the frequency of one dominant forest species (Artibeus obscurus) was significantly lower in small fragments compared with larger ones. The high connectivity among forest patches in the study area and the ability of some species to use the surrounding matrix of secondary forests and shade cocoa plantations possibly precludes the insularization effect. Qualitative habitat changes induced by fragmentation, such as edge formation and forest regrowth affected bat community structure; both modified habitats comprised a limited subset of the species assemblage found in the interiors of mature forests. The results presented here provide evidence of impoverished bat assemblages in man-modified habitats linked with deforestation and overall disturbances related with forest fragmentation. Key Words: bats, community ecology, diversity, South America, tropical rain forest INTRODUCTION The conversion of natural habitats into small and isolated remnants is thought to be a major cause of species extinction (Pimm & Raven 2000) and is currently an important force structuring natural communities (Whitmore 1997). An increasing effort has been made to understand the consequences of habitat fragmentation on many biological groups (Turner 1996), particularly in tropical regions (Fahrig 2003, Laurance & Bierregaard 1997). From these studies, it is clear that species vary considerably in their response to fragmentation (Cosson et al. 1999a, Laurance et al. 2002). Bats comprise more than 1000 species worldwide and are especially diverse in the Neotropics, where in some regions they represent nearly half of all locally reported mammal species (Fleming et al. 1972, Patterson et al. 2003). Bats are also considered the most ecologically diverse group of mammals in Neotropical ecosystems, 1 Current address: Departamento de Ciências Biológicas, Universidade Estadual de Santa Cruz, Rodovia Ilhéus Itabuna, km 16, CEP 45650-000, Ilhéus, Bahia, Brazil. Email: deborah@uesc.br representing several distinct trophic groups including carnivores, insectivores, frugivores, nectarivores, piscivores and sanguinivores (Hill & Smith 1984). Many bat species are responsible for the seed dispersal of many plant species (Charles-Dominique 1986) and the pollination of others that are components of pristine forests (Fleming 1988). Presently, the response of Neotropical bats to habitat fragmentation is not well understood. It has been demonstrated that bat communities tend to be less diverse and dominated by fewer species in more deforested habitats (Brosset et al. 1996, Cosson et al. 1999b, Fenton et al. 1992, Reis & Muller 1995) with some bat species being restricted to continuous and less-disturbed forest tracts, and thus, locally vulnerable to deforestation (Bernard & Fenton 2003, Brosset et. al. 1996, Cosson et al. 1999b, Estrada et al. 1993, Fenton et al. 1992, Kalko et al. 1999, Schulze et al. 2000, Stoner et al. 2002). This pattern is of particular concern to bats in the family Phyllostomidae (subfamily Phyllostominae) which appears to be more sensitive to deforestation and overall levels of habitat disturbance (Fenton et al. 1992, Medellin et al. 2000). However, the relationship between bat species richness and fragment size remains controversial, with studies

532 DEBORAH FARIA showing either the absence (Estrada et al. 1993) or the existence of a positive relationship (Cosson et al. 1999b), while others fail to show differences in bat species richness between continuous forested areas and fragments (Estrada et al. 1993, Schulze et al. 2000). Distances among isolated forest patches and species-specific use of modified habitats may play an important role in the structure of bat communities (Cosson et al. 1999b, Estrada et al. 1993). In addition, landscape features such as patch size, forest cover and patch density were important attributes correlated to species and community-level response of phyllostomid bats in fragmented landscapes (Gorresen & Willig 2004). Fragmentation is also known to influence the bat community dynamics in natural forest fragments (Aguirre et al. 2003). Nevertheless, the extent to which bats respond to qualitative changes in habitats modified by fragmentation, such as their use of forest edge or the surrounding landscape matrix, has rarely been addressed for tropical regions (Estrada & Coates-Estrada 2001, Estrada et al. 1993, Medellín et al. 2000). In this study, the response of phyllostomid bats facing habitat fragmentation was investigated in northeastern Brazil. Phyllostomid species richness, capture frequency, species composition and species-specific abundance patterns were studied in six habitat categories representing the landscape mosaic of the area. These habitats included large and small patches of mature forests, and the surrounding matrix of second-growth vegetation and shade cocoa plantations (Theobroma cacao L. Sterculiaceae). Three questions were addressed: (1) Are bats using the surrounding modified forest matrix? (2) Are bats affected by forest reduction, with differences in species richness and capture frequency between small and large forest fragments? (3) Are bats sensitive to qualitative disturbances caused by fragmentation such as edge formation? METHODS Study area Study sites were located in the Una region, of southern Bahia, Brazil (15 17 S, 39 04 W; Figure 1). This region is one of the few remaining forest areas of the Atlantic Forest in north-eastern Brazil and is considered as a high-priority area for biodiversity due to high levels of endemism and species richness (Conservation International do Brasil et al. 2001). The dominant vegetation of the Una region is classified as Lowland Atlantic rain forest by Oliveira-Filho & Fontes (2000), with a flora characterized by high levels of endemism (Thomas et al. 1998). The mean annual temperature is 24 C and the average rainfall is 1500 mm y 1, with no defined seasonality, although a rainless period of 1 3 mo may occur from December to March (Mori et al. 1983). The European occupation in southern Bahia began about 500 y ago, although deforestation in the Una region has mainly taken place during the last 40 y with the establishment of the BR 101 highway, connecting the main cities of Brazil along the Atlantic coast. Along with several other aspects of regional economic development, this road opened the frontier for such intensive logging activities, and in 1994 satellite images revealed that only 34.8% of the original forest remained in Una. Timber exploitation is still a major threat to the forest, but squatters and cattle farms have also contributed to deforestation at a regional scale (Alves 1990). Presently, only 11 000 ha of forest are protected by federal law, all of which are located within the Una Biological Reserve, while the majority of the remaining forest is privately owned. Experimental design This study is part of a major project designed to investigate the comparative response of several biological groups faced with habitat fragmentation in the Una region the RestaUna Project. Satellite images, aerial surveys and photo interpretation were used to characterize the landscape of almost 15 000 ha of the Una region. The analysis showed that the fragmentation process in the Una region, although very intense, has not led to the complete isolation of most forest tracts, but rather, created a mosaic of varied habitat types. The largest proportion of the landscape (49%) analysed was covered by mature forests. Other land-cover types in order of the proportion covered included open areas (27%) comprising mainly pastures, secondary forests (15%), shade cocoa plantations (6%) and rubber tree plantations (2%) (Pardini 2004). Most mature forest patches are close to each other (maximum distance of only 480 m between two patches in the three sampling blocks) or connected by other modified forested land-cover types such as shade cocoa plantations, secondary vegetation or by narrow strips of mature forests. These mature forest patches are irregularly shaped, interdigitated with the modified matrix, having many constrictions with a high proportion of edge (contacts between mature forest and other modified habitat). This pattern characterizes a shredded landscape (Feinsinger 1994) where the proportion of edges rather than the isolation of patches seem to be the most striking feature related to the fragmentation process (Faria 2002). After the characterization of the Una landscape, six main habitat categories were selected for sampling (Figure 1). These categories included: IL interiors of large fragments of mature forests (>1000 ha), EL edges

Bats and forest fragmentation in Brazil 533 Figure 1. Map of north-eastern Brazil depicting (a) the distribution of Atlantic Forest remnants, the spatial location of the sampling blocks and the limits (black) of Una Biological Reserve, and (b) the position of the sampling transects in each block (1, 2 and 3), considering six habitat categories: (IL) interior of large forests, (EL) edges of large forests, (IS) interiors of small fragments, (ES) edges of small fragments, (SF) secondary forests and (SCP) shade cocoa plantations (modified after Pardini 2004). of large fragments of mature forests (>1000 ha), IS interiors of small fragments of mature forests (<100 ha), ES edges of small fragments of mature forests (<100 ha), SF secondary forests and SCP shade cocoa plantations, under the management system locally known as cabruca. This latter habitat category represents a traditional crop system in which the forest is thinned as the cocoa shrubs replace the original forest understorey; less than 10% of the canopy trees are left for shade (Alves 1990). Mature forest fragments (large and small) included either isolated patches or those connected by forest cover such as secondary forests, shade plantations or by narrow strips (<100 m wide) of mature forest. Secondary forests consisted of regrowth forest areas clear-cut less than

534 DEBORAH FARIA 20 y ago. These regenerating areas were visibly shorter than the mature forests (usually not exceeding 10 m high), forming a single, thick forest stratum (Faria 2002). Edges were classified as the boundary between forests and pastures; edge sampling sites were established inside the fragments and 20 m from the nearest pasture, while interior sampling locations were placed at least 100 m from the forest-pasture boundary. These habitat categories were located within three replicate sampling blocks (6 6km), each block comprising two replicates of each habitat category, totalling 36 selected sampling sites for the whole study. In each sampling site a transect was established. Each transect was created as a 100-m long, 1.5-m wide trail by partially clearing the vegetation from the ground up to 3 m height. The placement of the sampling transects followed, besides the criteria used to define each habitat category, logistical limitations including the access to the site (e.g. proximity of dirt roads) and the permission of landowners. Sampling procedures From September 1997 to August 2000, bats were mistnetted following a standardized sampling unit consisting of a set of eight, 2.5-m-high ground mist nets of length 12 m (2), 9 m (2) and 6 m (4), comprising a sampling area of 165 m 2 of nets. This sampling unit was placed along each 100-m transect for 5 h after sunset on four moonless, non-consecutive nights, with at least 1 mo between netting in the same transect. Total effort was calculated as the product of the sampling area (165 m 2 ), the number of sampling hours per night (5), the total number of transects (36) and number of netting nights per transect (4) (Straube & Bianconi 2002), which amounted to 118 800 m 2 h of sampling. Nets were checked every 20 30 min and captured bats were kept inside cloth bags and released at the same location. Bats were assigned in the following feeding guilds: aerial insectivore, carnivore, frugivore, gleaning insectivore, nectarivore, omnivore and sanguinivore (Bonaccorso 1979, Handley et al. 1991, Kalko et al. 1996). Bats were not tagged during the present study, and the number of captures probably includes recaptures. However, results from other studies have shown that recapture rates of Neotropical bats are low (Bernard & Fenton 2003, Thomas & La Val 1988) and the capture frequency has been used as an index of abundance (Gorresen & Willig 2004). The use of ground mist nets as the sole sampling technique biases the overall picture of bat assemblages towards understorey phyllostomid bats, as this technique is well known to underestimate the occurrence and capture rates of most canopy phyllostomids (Kunz & Kurta 1988). Therefore, it is important to consider that many species of Phyllostomidae (e.g. those that forage above the height of the ground mist nets, or are more prone to avoid nets in general) are under-represented in the present study. Bat identification Fifty-two specimens not positively identified in the field were collected, quickly killed by cervical dislocation, fixed in 10% formaldehyde and preserved in 70% ethanol. Bats were initially identified using the key of Vizotto & Taddei (1973) and specimens were then sent to Dr Valdir Taddei at the Laboratório de Chiropteros da UNESP de São José do Rio Preto for confirmation of the species identity. Additionally, some specimens were sent or taken to North American collections, both the American Museum of Natural History (AMNH, NY), and at the National Museum of Natural History (USNM, Washington, DC), for a second opinion on dubious specimens and to further confirm the identifications performed in Brazil. Vouchers are deposited in the Mammal Collection of the Universidade Estadual de Santa Cruz (UESC), Ilhéus, Bahia. Nomenclature follows Simmons (2005). Data analysis A two-way ANOVA was used to test for species richness and capture frequency differences among the habitat categories, landscape blocks and possible interactions between these two factors (habitat categories and landscape blocks). Raw data for each transect included the total species richness and capture frequency (number of bats captured, possibly including recaptures) for the four sampling nights. For the species-level response, differences in bat capture frequency were only tested for the dominant species in order to meet the overall assumptions needed to perform parametric tests. Data were tested for normality (Kolmogorov Smirnov test) and homogeneity of variances (Levene test). When required, data were rank-transformed, and differences were considered to be statistically significant at P < 0.05. Prior to the data analysis, five ANOVA comparisons (n treatments minus 1) were chosen by planning the orthogonal contrasts (Montgomery 2001). This procedure allows the partitioning of the variance among contrasts, not increasing the type I error probabilities, a tendency inherent in multiple comparisons. However, shade cocoa plantations showed a strikingly different variance when compared with the remaining five other habitat categories for the mean species richness and capture frequency (see Table 2), so statistical comparisons including this habitat category were unnecessary. Therefore, although the general results included

Bats and forest fragmentation in Brazil 535 data from shade cocoa plantations, the orthogonal comparisons did not include this habitat category. The four remaining orthogonal comparisons (five remaining treatments minus 1) focused on the following three main questions considered in the current study: (1) The use of the matrix component of the secondary forest: mature habitat categories (IL + EL + IS + ES) versus secondary forests (SF) (2) The effect of forest reduction (size of fragments): interiors and edges of large fragments (IL + EL) versus interiors and edges of small fragments (IS + ES), and interiors of large fragments (IL) versus interiors of small fragments (IS) (3) The edge effect: interiors of small fragments (IL + IS) versus edges of large and small fragments (EL + ES). Rarefaction curves were calculated for each habitat category with the package PAST (Hammer et al. 2001), using the algorithm of Krebs (1989) and including standard errors. RESULTS General aspects A total of 39 phyllostomid species from five subfamilies, were reported from 2556 captures during the 36-mo sampling period (Table 1). Frugivores were the most frequent (2352 captures) and speciose group (17 species). From the total sample, Carollia perspicillata and Rhinophylla pumilio were the two most frequently captured bats with 938 and 673 captures each, followed by Artibeus obscurus with 293 captures. Together, these three species of frugivore bats accounted for nearly 74.5% of all captures (n = 1904). Hence, C. perspicillata and R. pumilio were widely distributed among all habitat categories, being recorded in all 36 transects. Pooled data showed that only five bat species (12.8%) were widespread, being observed in all six habitat categories: Artibeus cinereus, Artibeus lituratus, A. obscurus, C. perspicillata and R. pumilio. Twentytwo species (56.5%) were rare, with fewer than 10 captures per species. Among the rare species are the two carnivores and all the gleaning insectivores, both feeding guilds exclusively comprised members of the subfamily Phyllostominae. Bats and habitat categories: the use of the matrix Shade cocoa plantations showed significantly higher mean species richness, total captures and the capture frequency of the three dominant than the other five remaining habitat categories (Tables 1 and 2). Fifty-one per cent of all captures were obtained from shade cocoa plantation transects, rendering 34 species of bat. Only four rare species captured in mature forests, and one from secondary forests, were not sampled in shade plantations (Table 1). This result clearly shows that this component of the matrix harbours most of the species reported to occur in both mature and secondary stands, including the rare carnivores and insect-gleaning species (Table 1). However, the potential of the two components of the matrix (secondary forests and shade plantations) for harbouring forest species contrasted sharply. Only seven species, from a total of 167 captures, were sampled in secondary forests and the results from two-way ANOVAs showed that the average species richness in secondary forest (SF), 4 (SD ± 1.9), was significantly lower than in mature forests (IL + EL + IS + ES = 6.37 ± 1.99) (twoway ANOVA, F= 4.97, P = 0.035). Six out of these seven species were also found in all four mature forest categories, but no Phyllostominae bats were observed in secondary forests. Mean capture frequency reported for the mature forest patches (45.2 ± 19.9) also tended to be higher than for the secondary forests (27.8 ± 14.8), although this difference was not statistically significant (F = 3.66, P = 0.068). Carollia perspicillata and R. pumilio were dominant species in mature and secondary forests. Two-way analysis of variance indicated non-significant differences in the capture frequencies of C. perspicillata (IL + EL + IS + ES = 16.3 ± 13.5, SF = 13.8 ± 13.4) and R. pumilio (IL + EL + IS + ES = 17.2 ± 11.7, SF = 10.2 ± 2.92) between mature and secondary forests (F = 0.18, P = 0.67 and F = 1.92, P = 0.171 for C. perspicillata and R. pumilio respectively). By contrast, A. obscurus was significantly more captured in the mature forests (5.77 ± 3.56) than in the secondary forests (0.66 ± 0.81; F = 11.4, P = 0.003). Small and large fragments Mean species richness did not differ statistically between large (IL + EL = 5.91 ± 1.16) and small forest fragments (IS + ES = 6.83 ± 2.55; F = 0.23, P = 0.631). Bat assemblages in small fragments included 21 different species, whereas 18 species were caught in all transects located in larger forests. Non-significant differences were also reported for total capture frequency between large (41.4 ± 13.6) and small fragments (48.9 ± 24.7; F = 0.81, P = 0.379) and, again, three frugivorous species, A. obscurus, C. perspicillata and R. pumilio were the most commonly captured bats. Artibeus obscurus showed a significantly lower mean capture frequency in small fragments (4.33 ± 2.38) compared with larger ones (7.26 ± 4.02; F = 4.53, P = 0.047), whereas a significant block effect was detected on the mean capture rate of C. perspicillata (F= 3.68, P = 0.045),

536 DEBORAH FARIA Table 1. Capture frequency and feeding guilds of phyllostomid bat species mist netted in six habitat categories at Una region, southern Bahia state, Brazil. Large fragments Small fragments Matrix Feeding guilds 2 Interiors (IL) Edges (EL) Interiors (IS) Edges (ES) Secondary forests (SF) Shade cocoa plantations (SCP) Species 1 Subfamily Carolliinae Carollia brevicauda FR 1 5 4 2 0 14 Carollia perspicillata FR 81 75 142 93 83 464 Rhinophylla pumilio FR 112 84 117 100 61 199 Subfamily Desmodontinae Desmodus rotundus SA 1 0 0 0 0 9 Subfamily Glossophaginae Anoura geoffroyi NE 0 0 0 0 0 3 Choeroniscus minor NE 0 0 0 0 0 2 Glossophaga soricina NE 1 1 3 0 2 27 Lichonycteris obscura NE 0 0 0 0 1 0 Lonchophylla mordax NE 3 1 0 1 0 8 Subfamily Phyllostominae Chrotopterus auritus CA 1 0 0 0 0 1 Lampronycteris brachyotis GI 0 0 0 0 0 1 Lophostoma brasiliense GI 1 0 0 0 0 8 Lophostoma silvicolum GI 0 1 0 0 0 1 Micronycteris hirsuta GI 1 0 1 0 0 2 Micronycteris microtis GI 1 0 0 1 0 0 Micronycteris schmidtorum GI 0 0 1 0 0 1 Mimon crenulatum GI 0 0 0 0 0 7 Phylloderma stenops OM 0 0 0 0 0 2 Phyllostomus discolor OM 0 0 0 0 0 85 Phyllostomus elongatus OM 0 0 2 0 0 0 Phyllostomus hastatus OM 0 0 0 0 0 10 Tonatia saurophila GI 1 0 0 0 0 0 Tonatia sp1 GI 0 0 0 0 0 2 Trachops cirrhosus CA 0 0 1 0 0 2 Trinycteris nicefori GI 0 0 0 0 0 7 Subfamily Stenodermatinae Artibeus cinereus FR 9 11 17 12 11 47 Artibeus fimbriatus FR 0 0 0 1 0 4 Artibeus gnomus FR 1 2 3 0 0 8 Artibeus jamaicensis FR 1 1 1 4 0 22 Artibeus lituratus FR 7 6 12 7 5 38 Artibeus obscurus FR 50 37 27 25 4 150 Chiroderma villosum FR 1 0 2 0 0 5 Platyrrhinus lineatus FR 0 0 0 0 0 15 Platyrrhinus recifinus FR 0 0 2 0 0 21 Sturnira lilium FR 0 0 1 3 0 61 Sturnira tildae FR 0 0 0 0 0 8 Uroderma bilobatum FR 0 0 1 0 0 68 Vampyressa pusilla FR 0 0 0 0 0 3 Vampyrodes caraccioli FR 0 0 1 0 0 0 Total species richness 17 11 18 11 7 34 Total captures 273 224 338 249 167 1305 1 Subfamilies and species listed in alphabetical order. 2 Codes for feeding guilds are: (CA) carnivore, (FR) frugivore, (GI) gleaning insectivore, (NE) nectarivore, (OM) omnivore and (SA) sanguivore. which was more often observed in both large and small fragments located in the block 3, although no significant influence of patch size (F = 2.07, P = 0.167) or interaction between factors were reported for this species (F = 2.97, P = 0.077). Alternatively, it was not possible to separate the independent influence of patch size and landscape blocks on the mean capture frequency of R. pumilio, as a significant interaction between these two factors was reported to occur (F = 6.94, P = 0.006); R. pumilio showed a higher number of captures in the

Bats and forest fragmentation in Brazil 537 Table 2. Phyllostomid species richness, total capture frequency and the capture frequency (mean ± SD) of the three dominant species, Artibeus obscurus, Carollia perspicillata and Rhinophylla pumilio across six habitat categories at Una region, southern Bahia state, Brazil, comprising: interiors of large fragments of mature forests (IL), edges of large fragments of mature forests (EL), interiors of small fragments of mature forests (IS), edges of small fragments of mature forests (ES) and secondary forests (SF). Species richness and capture frequency are, respectively, the averages of the number of species and the number of bats caught over four sampling nights across the six transects per habitat category, possibly including recaptures. Variables IL EL IS ES SF SCP Species richness 6.10 ± 1.17 5.67 ± 1.21 7.67 ± 3.01 6.00 ± 1.90 4.00 ± 1.26 19.5 ± 3.21 Capture frequency 45.5 ± 11.3 37.3 ± 15.4 56.3 ± 29.1 41.5 ± 19.1 27.8 ± 14.8 217 ± 50.7 Artibeus obscurus 8.33 ± 5.16 6.17 ± 2.48 4.50 ± 2.35 4.17 ± 2.64 0.67 ± 0.82 25.0 ± 8.37 Carollia perspicillata 13.5 ± 4.09 12.5 ± 6.83 23.7 ± 22.3 15.5 ± 13.6 13.8 ± 13.4 77.2 ± 44.2 Rhinophylla pumilio 18.7 ± 12.9 14.0 ± 10.9 19.5 ± 10.2 16.7 ± 14.9 10.2 ± 2.93 33.2 ± 15.5 small fragments (edges and interiors considered together) compared with large ones only in one sampling block. Comparison between interiors of large (IL) and small (IS) fragments did not reveal significant differences in species richness (IL = 6.16 ± 1.16, IS = 7.66 ± 3.01; F = 0.16, P = 0.703), overall capture frequency (IL = 45.5 ± 11.2, IS = 56.3 ± 29.1; F = 0.02, P = 0.880), and capture frequency of the three dominant species, C. perspicillata (IL = 13.5 ± 4.08, IS = 23.7 ± 22.3; F = 1.20, P = 0.315), R. pumilio (IL = 18.7 ± 12.9, IS = 19.5 ± 10.2; F = 0.026, P = 0.878) and A. obscurus (IL = 8.33 ± 5.16, IS = 4.55 ± 2.34; F = 2.60, P = 0.158). However, A. obscurus tended to be less frequent in interiors of small fragments than in large ones, with 27 and 50 captures in each habitat category respectively. Edge effect The two-way ANOVAS indicated non-significant differences between interiors (IL + IS) and edges (EL + ES) of large and small fragments for overall capture frequency (IL + IS = 50.9 ± 21.8, EL + ES = 36.4 ± 16.7; F = 1.93, P = 0.181) and the capture frequency of the three dominant species: C. perspicillata (IL + IS = 18.6 ± 16.2, EL + ES = 14 ± 10.3; F = 0.725, P = 0.406), R. pumilio (IL + IS = 19.1 ± 11.1, EL + ES = 15.3 ± 12.5; F = 0.554, P = 0.466) and A. obscurus (IL + IS = 6.41 ± 4.31, EL + ES = 5.16 ± 2.65; F = 0.738, P = 0.402). Similarly, no significant differences were reported for species richness (IL + IS = 6.91 ± 2.31, EL + ES = 5.83 ± 1.52; F = 1.619, P = 0.219). However, bat assemblages on forest edges (14 species) were a subset of those found in forest interiors (24 species), and only two species (one capture each) sampled on the edges were not reported from interior transects (Artibeus fimbriatus and Lophostoma silvicolum). The overall patterns of species richness illustrated by rarefaction curves were consistent with the results from the statistical analysis. For a given number of captures, the expected species richness was highest in shade plantations, followed by interiors and edges of mature forests (both large and small fragments), with the secondary forests showing the lowest values of species richness (Figure 2). DISCUSSION Similar to other studies conducted in the Neotropics (Estrada et al. 1993, Schulze et al. 2000), in this study little evidence of fragment-size effects on bat species richness and capture frequency was found. Small and large fragments in Una showed similar species richness and bat capture frequency. In fact, bat assemblages in small mature stands also included species commonly associated with continuous and well-preserved forest, like the species belonging to the subfamily Phyllostominae (Fenton et al. 1992, Medellín et al. 2000). Many bat species appear to persist in heavily fragmented landscapes whereas other biological groups may be more adversely influenced by fragmentation (Laurance et al. 2002, Law & Dickman 1999, Turner 1996). Their ability to explore the mosaic of modified habitats in the matrix and to undertake long-distance movements are two potential key factors influencing the general resistance of bat communities to habitat fragmentation (Bernard & Fenton 2003, Estrada & Coates- Estrada 2002, Law & Dickman 1999, Law et al. 1999, Schulze et al. 2000). In Una, 35 species, or 90% of the local bat assemblage, were reported in the modified habitats of the matrix. However, species reacted differently to the matrix components. Secondary forests were used as habitats only for a limited subset of bats, representing less than 20% of the total species reported. In contrast, only three of the species sampled in mature forests were not reported in the shade cocoa plantations, an agricultural habitat comprising a great variety of bat species (Estrada et al. 1993, Faria & Baumgarten in press, Faria et al. 2006). Shade cocoa plantations in southern Bahia are also known to harbour the forestdwelling species of many biological groups (Alves 1990, Rambaldi & Oliveira 2003, Schroth et al. 2004),

538 DEBORAH FARIA Figure 2. Randomized rarefaction curves depicting the expected number of bat species taken from different sample sizes in six habitat categories in the Una region. possibly due to the fact that shade crops retain vertical stratification, allowing a more complex habitat than deforested areas. The canopy trees that shade the cacao shrubs provide day roost for some bat species, a protective cover for flying bats and, together with the herbaceous plants, also offer food resources that attract frugivore and nectarivore bats (Faria & Baumgarten in press). Secondary forests are even more complex habitats, with a richer and denser plant species community compared with shade plantations, also including pioneer plant species that attract frugivorous bats (D. Faria, unpubl. data). It remains unclear as to why secondary forests showed so few captures and bat species richness compared with shade cocoa plantations. Certainly this differential use of the two matrix elements by bats may rely on qualitative changes in the forest structure as well as in the resource availability for bats. For instance, secondary forests are known to harbour a lower number of mature, tall trees (Faria 2002), possibly resulting in a lower availability of hollow trees and other large tree cavities used as day roosts by many bat species. Together with other general features that characterize a highly disturbed forest (e.g. increased number of clearings, higher solar penetration), secondary forest may represent a suboptimal habitat for bats compared with the other habitat categories. It has been shown that the isolation of remnants is an important factor influencing bat species richness in terrestrial landscapes (Estrada et al. 1993). In addition to the presence of suitable habitats in the matrix, mature forest patches in Una are close to each other or connected by modified forest habitats. Furthermore, the short distance between the few isolated mature forest patches in Una probably does not exceed the potential flight distance of most bat species, as they commute long distances among different patches located within fragmented landscapes (Cosson et al. 1999b, Estrada & Coates-Estrada 2001, Estrada et al. 1993, Gorresen & Willig 2004) including those fragmented naturally (Bernard & Fenton 2003). The response of bats to fragmentation is variable among species (Cosson et al. 1999b, Estrada et al. 1993, Fenton et al. 1992, Schulze et al. 2000). Although most bats show a high dispersal capability among landscape elements including open, disturbed and other man-modified habitats (Bernard & Fenton 2003), the small Phyllostominae, insectivorous foliage-gleaning bats are morphologically constrained to undertake shortdistance flights (Belwood 1988). Furthermore, these species have small home range sizes with feeding sites often located within mature forests, and thus, show greater vulnerability to forest disruption (Bernard & Fenton 2003, Fenton et al. 1992, Kalko et al. 1999). Their apparent natural rarity severely limits any attempt to develop a precise understanding of their distribution pattern among the habitat categories in Una. However, their absence from secondary forest samples and the trend of interior sites (small and large mature forests) to show more species of these feeding guilds than edges is noteworthy. This result is in accordance to the general notion that more disturbed habitats tend to show a lower number of Phyllostominae species than those more preserved tracts (Fenton et al. 1992, Schulze et al. 2000). Cosson et al. (1999b) suggested that large-bodied frugivore species such as A. obscurus, which are highly abundant in mainland forests and canopy foraging habits, tend to be less sensitive to fragmentation, while other small, understorey frugivorous bats such as R. pumilio were less tolerant to habitat reduction and the isolation of islands. Alternatively, Schulze et al.

Bats and forest fragmentation in Brazil 539 (2000) reported that the most striking effect of fragmentation upon bats in a landscape of Guatemala, Central America, was a decreased proportion of large frugivores in small fragments compared with continuous forests, while the opposite was observed for the small frugivorous bats. Specific features of the landscape in Una region may partially explain the response of two small forest understorey frugivores, R. pumilio and C. perspicillata. The modified habitats surrounding the forest patches at Una provide physical connectivity among patches. Understorey frugivores species, including C. perspicillata and R. pumilio, were highly abundant in both early secondary forests and shade cocoa plantations, being the numerically dominant species. They forage on successional plant species from the genera Piper, Cecropia and Vismia. For these bat species, small and large fragments seem to be truly connected by a permeable matrix that provides reliable foraging areas. Differences in the capture frequency of C. perspicillata and R. pumilio in large and small fragments, however, were influenced by the location of the sampling block considered and by an interaction of the patch size and the landscape block, respectively. Therefore, a more detailed analysis on features of the local landscape may reveal important correlates influencing the capture frequency of these two species of small frugivore in the Una region. Conversely, the low capture frequency of A. obscurus in early secondary forests in Una indicates that this habitat category is not frequently used by this frugivore species. Similarly to other large-bodied Artibeus species (Galetti & Morrelato1994, Handleyet al. 1991, Morrison 1978), A. obscurus is known to feed heavily on canopy trees such as Ficus and other tall plant species characteristic of mature Neotropical forests. The lower capture frequency of A. obscurus in small fragments and its rarity in secondary forest are possibly due to differences in the temporal and spatial distribution, density and availability of fruit resources in these two habitat categories compared with larger forest tracts. The strongest effect of forest fragmentation detected in Una appears to be the edge effect. In contrast to patch size, edge effects showed a strong negative effect on bat communities. The edges of mature forest stands harboured only a subset of the bat community found in mature forest interiors. Many studies have shown that the establishment of edges strongly influences community structure for a variety of taxa (Laurance et al. 2002), including bats (Kalko 1998). Nevertheless, the present study was the first one to demonstrate a decrease in bat species richness in forest edges. Changes in local microclimate (Kapos et al. 1997, Williams-Linera 1990), tree dynamics (Laurance et al. 1998), seedling recruitment (Benitez-Malvido 1998) and forest structure (Laurance et al. 2001, Malcolm 1994, Williams-Linera 1990) qualitatively differentiate edges from interior habitats. Because of the non-linear relationship between perimeter and area and because edge effects are additive (Malcolm 1994), small fragments usually suffer a more intensive edge effect than larger remnants. Although the small fragments sampled in Una can be generally classified as vulnerable to the edge effect (irregularly shaped and smaller than 100 ha), it seems that the lower number of species detected on sites located at forest edges is related to specific habitat changes localized along narrow strips that do not penetrate further than 100 m into the forest. For instance, detailed analysis of forest structure undertaken along the sampling transects in Una revealed that edges from both small and large fragments tend to have a sparser canopy cover and denser understorey vegetation than interior transects (Faria 2002). A denser understorey in edge sites implies a more cluttered habitat for flying bats, which may impose limits upon the flight manoeuvrability of some bat species. Changes in the quality and availability of food and roost resources on edges compared with the forest interior are also important features that might explain the differences in the bat community structure detected between the two habitat categories. Additionally, a buffer to edge effects may be provided in forest fragments by the secondary forests and shade plantations that entirely or partially surround some of the mature forest fragments sampled in this study. This would explain the similar species richness and capture frequency of interiors of large and small stands. It was demonstrated here that bat species assemblages in forest edges and secondary forests comprise a limited number of the species pool reported in the interior forest stands. On the other hand, it is noteworthy that neither edges nor secondary forests showed a shift in community dominance toward species associated with open areas, such those of the genera Sturnira, Platyrrhinus and Uroderma, which were rare in Una. The apparent rarity of Sturnira lilium, a species considered a Solanum (Solanaceae) specialist (Marinho-Filho 1991), in mature and secondary forests from Una may be partially explained by the scarcity of its main food resource in the local forests (E. Mariano-Neto unpubl. data). Species of Solanum are common in shade plantations (pers. obs.), which coincides with the increased representation of both S. lilium and S. tildae in shade crops. For similar reasons, other generalist species such as those from the genera Platyrrhinus and Uroderma may also find shade plantations suitable foraging areas (Faria & Baumgarten in press). Conservation remarks An increasing number of studies have highlighted the importance of small fragments for the maintenance of diverse communities and the conservation of many

540 DEBORAH FARIA species (Laurance & Bierregaard 1997, Turner & Corlett 1996), including bats (Estrada et al. 1993, Gorresen & Willig 2004). In the Una landscape, small fragments are also important elements of any conservation strategy designed for bats. Although larger forest patches may provide core habitats and metapopulation sources for bats, small fragments support the same species assemblage found in the largest remnants. The same result was also reported for other taxonomic groups investigated in the same sampling areas of Una region, such as frugivorous butterflies, frogs and lizards, birds and non-volant small mammals (Rambaldi & Oliveira 2003). The presence of large forest tracts, the existence of a matrix that includes a mosaic of modified habitats exploited by bats and the close proximity of forest patches to one another are key features of the Una landscape that help to maintain a rich bat assemblage in the region. These general patterns emphasize the possibility of the existence of a rich biological diversity within a mosaic of mature and managed forest. In a highly fragmented landscape in Mexico, the preservation of a constellation of forest tracts of different sizes and levels of disturbance, coupled with the presence of a mosaic of different manmade, agricultural vegetation, also allow the existence of a rich bat assemblage at the landscape scale (Estrada & Coates-Estrada 2001, 2002; Estrada et al. 1993). Species loss in habitats qualitatively different from mature forests (edges and secondary stands), rather than a change in the species richness and abundance pattern related to patch size, was the most striking effect of fragmentation upon bat assemblages in the Una region. These more subtle changes in forest quality may represent an immediate, more important factor rather that the forest loss per se, to take into account in any conservation strategy considered for bats in this landscape. ACKNOWLEDGEMENTS I am grateful to my advisor Dr W. R. Silva, and to Dr W. L. for their valuable contributions during this research. I also thank to B. Campbell, Dr A. Estrada, Dr J. Delabie, Dr E. Vieira and four anonymous referees for valuable comments on the final version of the manuscript. This research was made possible thanks to the collaborative work of the RestaUna team. Financial support was given by the PROBIO (MMA/BIRD/GEF/CNPq), CNPq- Programa Nordeste and the Fapesp grant (processo 97/07075-1). Logistical facilities were provided by UESC, IBAMA and by the local landowners. I specially thank the IESB for the aerial photos and to M. Lima for editing the figures. LITERATURE CITED AGUIRRE, L. F., LENS, L., VAN DAMME, R. & MATTHYSEN, E. 2003. Consistency and variation in the bat assemblages inhabiting two forest islands within a neotropical savanna in Bolivia. Journal of Tropical Ecology 19:367 374. ALVES, M. C. 1990. The role of cacao plantations in the conservation of the Atlantic Forest of Southern Bahia, Brazil. Masters Thesis. University of Florida. Gainesville, Florida. 138 pp. BELWOOD, J. J. 1988. Foraging behavior, prey selection, and echolocation in phyllostomine bats (Phyllostomidae). Pp. 601 605 in Nachtigall, P. E. & Moore, P. W. B. (eds.). Animal sonar. Plenum Press, New York. BENITEZ-MALVIDO, J. 1998. Impact of forest fragmentation on seedling abundance in a tropical rain forest. Conservation Biology 12:380 389. BERNARD, E. & FENTON, B. 2003. Bat mobility in a fragmented landscape in Central Amazonia. Biotropica 35:262 277. BONACCORSO, F. L. 1979. Foraging and reproductive ecology in a Panamanian bat community. Bulletin of the Florida State Museum 24:359 408. BROSSET, A., CHARLES-DOMINIQUE, P., COCKLE, A., COSSON, J. & MASSON, D. 1996. Bat communities and deforestation in French Guiana. Canadian Journal of Zoology 74:1974 1982. CHARLES-DOMINIQUE, P. 1986. Inter-relations between frugivorous vertebrates and pioneer plants: Cecropia, birds and bats in French Guiana. Pp. 119 135 in Estrada, A. & Fleming, T. H. (eds). Frugivorous and seed dispersal. Dr. W. Junk Publisher, Dordrecht. CONSERVATION INTERNATIONAL DO BRASIL, FUNDAÇÃO SOS MATA ATLâNTICA, FUNDAÇÃO BIODIVERSITAS, INSTITUTO DE PESQUISAS ECOLÓGICAS, SECRETARIA DO MEIO AMBIENTE DO ESTADO DE SÃO PAULO & SEMAD INSTITUTO ESTADUAL DE FLORESTAS MG. 2001. Avaliação e ações prioritárias para a conservação da biodiversidade da Mata Atlântica e Campos Sulinos. MMA/SBF, Brasília, 40 pp. COSSON, J.-F., RINGUET, S., CLAESSENS, O., DE MASSARY, J. C., DALECKY, A., VILLIERS, J. F., GRANJON, L. & PONS, J. M. 1999a. Ecological changes in recent land-bridge islands in French Guiana, with emphasis on vertebrate communities. Biological Conservation 91:213 222. COSSON, J.-F., PONS, J.-M. & MASSON, D. 1999b. Effects of forest fragmentation on frugivores and nectarivores bats in French Guiana. Journal of Tropical Ecology 15:515 534. ESTRADA, A. & R. COATES-ESTRADA. 2001. Species composition and reproductive phenology of bats in a tropical landscape at Los Tuxtlas, Mexico. Journal of Tropical Ecology 17:626 646. ESTRADA, A. & COATES ESTRADA, R. 2002. Bats in continuous forest, forest fragments and in an agricultural mosaic habitat island at Los Tuxtlas, Mexico. Biological Conservation 2:237 245. ESTRADA, A., COATES-ESTRADA, R. & MERRITT, D. 1993. Bat species richness and abundance in tropical rain forest fragments and in agricultural habitats at Los Tuxtlas, Mexico. Ecography 16:309 318.

Bats and forest fragmentation in Brazil 541 FAHRIG, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution and Systematics 34:487 515. FARIA, D. 2002. Comunidades de morcegos em uma paisagem fragmentada da mata Atlântica do sul da Bahia, Brasil. Ph.D. thesis, Universidade Estadual de Campinas, São Paulo, Brasil. 134 pp. FARIA, D. & BAUMGARTEN, J. (in press). Shade cacao plantations (Theobroma cacao) and bat conservation in southern Bahia, Brazil. Biodiversity and Conservation. FARIA, D., LAPS, R. R., BAUMGARTEN, J. & CETRA, M. 2006. Bat and bird assemblages from forests and shade cacao plantations in two contrasting landscapes in the Atlantic rainforest of southern Bahia, Brazil. Biodiversity and Conservation 15:587 612. FEINSINGER, P. 1994. Habitat shredding. Pp. 258 260 in Meffe, G. K. & Carroll, C. R. (eds.). Principles of conservation biology. Sinauer Associates, Inc. Sunderland. FENTON, M. B., ACHARYA, L., AUDET, L. D., HICKEY, M. B. C., MERRIMAN, C., OBRIST, M. K., SYME, D. M. & ADKINS, B. 1992. Phyllostomid bats (Chiroptera: Phyllostomidae) as indicators of habitat disruption in the Neotropics. Biotropica 24:440 446. FLEMING, T. H. 1988. The short-tailed fruit bat: a study in plant-animal interactions. University of Chicago Press, Chicago. 365 pp. FLEMING, T. H., HOOPER, E. T. & WILSON, D. E. 1972. Three Central American bat communities: structure, reproductive cycles and movement patterns. Ecology 53:555 569. GALETTI, M. & MORELLATO, L. P. C. 1994. Diet of the large fruiteating bat Artibeus lituratus in a forest fragment in Brazil. Mammalia 58:661 665. GORRESEN, P. M. & WILLIG, M. R. 2004. Landscape response of bats to habitat fragmentation in Atlantic Forest of Paraguay. Journal of Mammalogy 85:688 697. HAMMER, Ø., HARPER, D. A. T. & RYAN, P. D. 2001. PAST: paleontological statistics software package for education and data analysis. Paleontologia Electronica 4:1 9. HANDLEY, C. O., WILSON, D. E. & GARDNER, A. L. 1991. Demography and natural history of the common fruit bat, Artibeus jamaicensis,on Barro Colorado Island, Panama. Smithsonian Contributions to Zoology 511:1 173. HILL, J. E. & SMITH, J. D. 1984. Bats: a natural history. University of Texas Press, Austin. 243 pp. KALKO, E. K. V. 1998. Organization and diversity of tropical bat communities through space and time. Zoology: Analysis of Complex Systems 101:281 297. KALKO, E. K. V., HANDLEY, C. O. & HANDLEY, D. 1996. Organization, diversity, and long-term dynamics of a Neotropical bat community. Pp. 503 551 in Cody, M. & Smallwood, J. (eds). Long term studies in vertebrate communities. Academic Press, New York. KALKO, E. K. V., FRIEMEL, D., HANDLEY, C. O. & SCHNITZLER, H. 1999. Roosting and foraging behavior of two neotropical gleaning bats, Tonatia silvicola and Trachops cirrhosus (Phyllostomidae). Biotropica 31:344 353. KAPOS, V. E., WANDELLI, J. L., CAMARGO, E. G. & GANADE, G. 1997. Edge-related changes in environment and plant responses due to forest fragmentation in Central Amazonia. Pp. 91 110 in Laurance, W. F. & Bierregaard, R. O. (eds.). Tropical forest remnants: ecology, management, and conservation of fragmented communities. University of Chicago Press, Chicago. KREBS, C. J. 1989. Ecological methodology. Harper and Row, New York. 645 pp. KUNZ, T. H. & KURTA, A. 1988. Capture methods and holding devices. Pp. 1 29 in Kunz, T. H. (ed.). Ecological and behavioral methods for the study of bats. Smithsonian Institution Press, Washington, DC. LAURANCE, W. F. & BIERREGAARD, R. O. (eds.) 1997. Tropical forest remnants ecology, management, and conservation of fragmented communities. The University of Chicago Press, Chicago. 616 pp. LAURANCE, W. F., FERREIRA, L. V., RANKIN-DE MERONA, J. M. & LAURANCE, S. G. 1998. Rain forest fragmentation and the dynamics of Amazonian tree communities. Ecology 79:2032 2040. LAURANCE, W. F., PéREZ-SALICRUP, D., DELAMONICA, P., FEARNSIDE, P. M., D ANGELO, S., JEROZOLINSKI, A., POHL, L. & LOVEJOY, T. E. 2001. Rain forest fragmentation and the structure of Amazonian liana communities. Ecology 82:105 116. LAURANCE, W. F., LOVEJOY, T. E., VASCONCELOS, H. L., BRUNA, E. M., DIDHAM, R. K., STOUFFER, P. C., GASCON, C., BIERREGAARD, R. O., LAURANCE, S. G. & SAMPAIO, E. 2002. Ecosystem decay of Amazonian fragments: a 22-year investigation. Conservation Biology 16:604 618. LAW, B. S. & DICKMAN, C. R. 1999. The use of habitat mosaics by terrestrial vertebrate fauna: implications for conservation and management. Biodiversity and Conservation 7:323 333. LAW, B. S., ANDERSON, J. & CHIDEL, M. 1999. Bat communities in a fragmented landscape on the south-west slopes of New South Wales, Australia. Biological Conservation 88:333 345. MALCOLM, J. R. 1994. Edge effects in central Amazonian forest fragments. Ecology 75:2438 2445. MARINHO-FILHO, J. S. 1991. The coexistence of two frugivorous bat species and the phenology of their food plants in Brazil. Journal of Tropical Ecology 7:59 67. MEDELLÍN, R. A., EQUIHUA, M. & ALMIN, M. A. 2000. Bat diversity and abundance as indicators of disturbance in Neotropical rainforests. Conservation Biology 14:1666 1675. MONTGOMERY, D. C. 2001. Design and analysis of experiments. (Fifth edition). John Wiley and Sons, New York. 684 pp. MORI, S. A., BOOM, B. B., CARVALHO, A. M. & SANTOS, T. S. 1983. Southern Bahian moist forests. The Botanical Review 49:155 232. MORRISON, D. W. 1978. Influence of habitat on the foraging distances of the fruit bat, Artibeus jamaicensis. Journal of Mammalogy 59:622 624. OLIVEIRA-FILHO, A. T. & FONTES, M. A. L. 2000. Patterns of floristic differentiation among Atlantic Forests in Southeastern Brazil and the influence of climate. Biotropica 32:793 810. PARDINI, R. 2004. Effects of forest fragmentation on small mammals in an Atlantic Forest landscape. Biodiversity and Conservation 13:2567 2586. PATTERSON, B. D., WILLIG, M. R. & STEVENS, R. D. 2003. Trophic strategies, niche partitioning, and patterns of ecological organization. Pp. 536 579 in Kunz, T. H. & Fenton, M. B. (eds.) Bat ecology. University of Chicago Press, Chicago.