Biodiversity and Conservation 13: 2567���2586, 2004. # 2004 Kluwer Academic Publishers. Printed in the Netherlands. Effects of forest fragmentation on small mammals in an Atlantic Forest landscape RENATA PARDINI Departamento de Zoologia, Instituto de Biocie��ncias, Universidade de Sa��o Paulo, Rua do Mata��o ��� travessa 14, 101 CEP-05508-900, Sa��o Paulo, SP, Brazil e-mail: renatapardini@uol.com.br fax: +55-11-30917513 Received 25 March 2003 accepted inrevised form 1 September 2003 Key words: Atlantic forest, Edge effect, Forest fragmentation, Matrix quality, Small mammals Abstract. Recent studies on the effects of tropical forest fragmentation indicate that fragmented landscapes are complex and heterogeneous systems influenced by factors other than the size or degree of isolation of forest remnants: of particular importance are the quality of the matrix and the edge-induced habitat changes. In order to investigate the influence of these factors, small mammals were surveyed in 36 sites in the landscape of Una, a region that encompasses some of the last and largest Atlantic Forest remnants in northeastern Brazil. Six sites were distributed on each of six landscape components ��� the interiors and edges of small remnants, the interiors and edges of large remnants, and the most common forested habitats found in the matrix. The survey comprised 46,656 trap-nights and yielded 1725 individuals of 20 species of rodents and marsupials. Results revealed: an increase in beta-diversity caused by fragmentation the contrasting effects of the altered forested habitats of the matrix, which harbor both forest and disturbance-adapted species a greater importance of edge effect than of patch size to the observed changes in small mammal community in remnants an association among terrestrial forest species and among arboreal forest species in terms of the distribution and abundance in the Una mosaic and a distinctive vulnerability of these two groups of species to fragmentation. Results emphasize the biological importance and con- servation value of both fragmented landscapes and small remnants in the Atlantic Forest, as well as the importance of management techniques to control and attenuate edge effects. Introduction Expansion of human activities and the resulting habitat fragmentation is ex- pected to affect several biological processes and factors, such as population size, species dispersal, structure and quantity of available habitat and the probability of invasion. However, research emphasis has often been placed on the study of the effects of patch size and isolation on species abundance and richness (Davies et al. 2001). Recent studies on the effects of tropical forest fragmentation (e.g., Laurance and Bierregaard 1997 Bierregaard et al. 2001) show that fragmented land- scapes are complex and variable systems mainly because of the altered habitats that surround fragments (Crome 1997 Laurance et al. 1997 Renjifo 2001). The quality or permeability of the matrix is crucial to determine the functional connectivity among remnants. It has been proposed that the occurrence of similar habitat structures in the remnants and in the matrix increases the connectivity among the remnants (Taylor et al. 1993).

Edge effects depend on the quality of the matrix, which may function as an alternative habitat for the original species (Medellin and Equihua 1998 Pardini et al. 2000) or as a source of exotic species that may invade remnants (Parsons 1972 Lynam 1997 Goodman and Rakotondravony 2000). The matrix also influences the magnitude of the structural changes in the forest remnants (Didham and Lawton 1999 Mesquita et al. 1999). Fragmentation effects are thus expected to vary not only among species, but also among different landscape structures. Habitat fragmentation studies that aim at improving tropical forest conservation strategies should, consequently, focus on the investigation of the relative importance of different underlying processes in real landscapes, taking into account large spatial scales and different elements of the mosaic (Bierregaard et al. 1997 Crome 1997 Davies et al. 2001). In a review of studies from different landscapes in different parts of the world, Bowers and Matter (1997) found no consistent relationship between the area of habitat and the density of mammalian species. Although an overall patternof positive responses of small mammals to clear-cutting has beenob- served intemperate forests (Kirkland 1990), large-scale studies reveal sig- nificant, persistent and contrasting effects of habitat fragmentation for small mammal species in these forests (Lomolino and Perault 2000 Nupp and Swihart 2000). Intropical forests, several studies showed a clear reductionin species abundance and/or small mammal richness in fragments (Laurance 1994 Grajon et al. 1996 Lynam 1997 Goodman and Rakotondravony 2000), but few studies consider variables other than patch size. Both the quality of the matrix (Gasconet al. 1999) and edge effects (Malcolm 1991, 1995) seem to be important to the structure of small mammal communities in Amazonian forest remnants. The ability to occupy the matrix is one of the most important ecological determinants of extinction proneness for small mammals from Australian rainforest (Laurance 1991, 1994). In this study, I investigate the patterns of variation in composition, abun- dance and richness of the small mammal community among the main land- scape components of Una mosaic, Bahia, northeastern Brazil. These components are the result of the habitat fragmentation process and were de- fined based on the size of the remnants, the intensity of edge effect and the type of forested habitat of the matrix. Methods Study area Approximately 70% of the Brazilianpopulationlives inthe area once occupied by the Atlantic Forest, which has been reduced to less than 8% of its original area extent. Atlantic Forest fauna and flora include a high proportion of en- demic species, comprising 80% of primates, 51% of rodents, and 50% of vascular plants (Conservation International do Brasil et al. 2000). The effects 2568

of this long-standing habitat destruction and fragmentation on mammals are the focus of comparatively few studies (Fonseca and Robinson 1990 Chiarello 1999 Pires et al. 2002). With few exceptions (Pires et al. 2002), these studies compare fragments of different size, without considering other landscape ele- ments in the analysis. The preservationof forest patches inSouth Bahia was partly due to the peculiarities of cocoa farming, the most important regional agricultural pro- duct. Through a system known as cabruca, cocoa shrubs are planted under native forest trees. A large proportion of the South Bahia forest is concentrated in the Una municipality, where the Una Biological Reserve is located. How- ever, the Reserve encompasses just 7022ha and a considerable part of the Una remnants is still privately owned and concentrated around the Reserve. The Una landscape retains 49% of mature forest within a heterogeneous matrix of open areas (mainly pastures ��� 27% of the landscape), secondary forest in the early stages of regeneration (15%), shaded cocoa plantations (6%), and rubber tree plantations (2%) (Pardini 2001). The mature forest has suffered selective logging, and patches are mostly irregularly shaped with a high pro- portion of mature forest in direct contact or near proximity to open areas (Pardini 2001). The soils in Una are sandy and nutrient-poor and the climate is hot and wet with meantemperatures between24 and 25 8C, rainfall between 1660 and 2000mm/year, and a short or nonexistent dry season (Gouvea 1969). The flora presents high levels of endemism: 43% of tree species are endemic to the Atlantic Forest and 28% are endemic to southern Bahia and northern Esp��rito Santo (Thomas et al. 1998). The most common trees (more than 15cm DBH) in Una are Sapotaceae (12% of the individuals), Leguminosae (sensu lato ��� 10%) and Chrisobalanaceae (7%) (Eduardo Mariano ��� Projeto RestaUna, unpublished data). Experimental design This study was developed as part of the RestaUna Project ��� Forest Remnants of Una, South Bahia (Nucleo de Estudos da Mata Atlantica ��� Universidade Estadual de Santa Cruz), which aimed to compare the responses of several biological communities to forest fragmentation in Una. Given the regional landscape configuration, six landscape components re- presenting the main forested habitats associated with fragmentation were de- fined: IL ��� interior of large remnant (mature forest patches, 1000 ha, 100 m from the nearest open area) EL ��� edge of large remnant (mature forest pat- ches, 1000 ha, 20 m from the nearest open area) IS ��� interior of small remnant (mature forest patches, 100 ha, 100 m from the nearest open area) ES ��� edge of small remnant (mature forest patches, 100 ha, 20 m from the nearest open area) SF ��� second growth forest in early stage of re- generation CF ��� cocoa plantations under native forest. 2569