3303 Ecology, 86(12), 2005, pp. 3303���3311 q 2005 by the Ecological Society of America RAPID RECOVERY OF DUNG BEETLE COMMUNITIES FOLLOWING HABITAT FRAGMENTATION IN CENTRAL AMAZONIA INGRID QUINTERO1,2,4 AND TOMAS ROSLIN3,5 1Biological Dynamics of Forest Fragments (BDFFP), Av. Andre �� Arau ��jo 1753, Aleixo CEP 69060-001, C.P. 478, Manaus, Amazonas, Brazil 2Instituto Nacional de Pesquisas da Amazo ��nia (INPA), Av. Andre �� Arau ��jo 2936, Petro ��polis CEP 69083-000, C.P. 478, Manaus, Amazonas, Brazil 3Metapopulation Research Group, Department of Biological and Environmental Sciences, P.O. Box 65, FI-00014 University of Helsinki, Finland Abstract. Few studies have directly assessed how rapidly functionally important insect communities recover following rain forest loss and fragmentation. In 1986, B. Klein com- pared the dung and carrion beetle assemblages of clearcuts, fragmented, and non-fragmented forests in central Amazonia, reporting drastic short-term changes in community composi- tion. Fifteen years later, we resampled the same sites using identical techniques and found that, with the regrowth of secondary vegetation between forest fragments, the initial dif- ferences had largely disappeared. As the secondary vegetation itself supports dung beetle assemblages similar to those of continuous forest, we conclude that, from the perspective of the dung beetles, the experimentally fragmented area had returned to a continuous state within approximately one decade. These results offer some good news for the conservation of tropical ecosystems, since they suggest that conditions favorable for functionally im- portant arthropods may be quickly restored by secondary regrowth. They also suggest that the preservation of forest fragments and secondary vegetation may provide an important complement to the conservation of intact mature forest. Key words: Amazonas Brazil dung beetle forest fragmentation matrix habitat rain forest recovery secondary vegetation. INTRODUCTION Around the world, human activities are shredding the original forest cover into smaller fragments of var- iable size. Insects are demonstrably susceptible to the adverse effects of deforestation and forest fragmenta- tion (Aizen and Feinsinger 1994, Didham et al. 1996, Didham 1997, Andresen 2003). As these organisms maintain many important ecosystem processes, such as pollination, seed dispersal, and nutrient recycling, the changes ultimately imperil ecosystem functioning (Didham et al. 1996). Insects also form integral parts of many food webs, and the effects of their loss may cascade through larger communities, causing unex- pected changes at higher and lower trophic levels (Ter- borgh 1976, Gibbs and Santon 2001, Koh et al. 2004). While it is now beyond any doubt that habitat mod- ification and fragmentation may induce short-term changes in the abundance and distribution of many in- sect groups (Lovejoy et al. 1986, Powell and Powell 1987, Didham et al. 1996, 1998, Didham 1997), the Manuscript received 30 December 2004 revised and accepted 31 May 2005. Corresponding Editor: M. D. Eubanks. 4 Present address: Inventories Program, Instituto de In- vestigacio ��n de Recursos Biolo ��gicos Alexander von Hum- boldt, Plaza Ricaurte, Claustro San Agust�� ��n, Apartado Ae��reo 048, Villa de Leyva, Boyaca ��, Colombia. E-mail: iquintero@ humboldt.org.co 5 E-mail: tomas.roslin@helsinki.fi duration of such changes has been much less explored. Of particular interest is the rapidity with which func- tionally important insect groups recover after distur- bance. As mature forests are cleared throughout the tropics, much of the deforested land is later abandoned and allowed to regrow (Nepstad et al. 1991). Hence, to conserve local biodiversity, the protection of forest fragments and secondary vegetation may offer an im- portant complement to reserves established in the few remaining tracts of continuous mature forest (Vander- meer and Perfecto 1997, Gascon et al. 1999, Dunn 2004). Nevertheless, the effectiveness of this approach will depend on the rate at which animal communities recover after disturbance. If it is rapid, the conservation of fragments and secondary vegetation provides an ef- fective investment. If it is slow, the conservation of mature forest remains the only tenable alternative. Yet, few earlier studies have rigorously assessed the rate at which insect communities recover following rain forest fragmentation. In a recent review of the literature on faunal recovery with tropical forest regeneration, Dunn (2004) predicted that the complete recovery of species richness in an important insect group���the ants���would typically take approximately 40 years af- ter deforestation and the recovery of species compo- sition ������substantially longer.������ These rates would then be comparable to those observed in birds (Dunn 2004) and plants (Uhl et al. 1988, Guariguata and Ostertag

3304 INGRID QUINTERO AND TOMAS ROSLIN Ecology, Vol. 86, No. 12 2000). However, Dunn (2004) encountered only 17 published data sets on the recovery of insects, and the literature turned out to suffer both from a general lack of accurate sampling design and from a taxonomic bias. In fact, half of the studies were unreplicated, and 11 of the 17 data sets on insects concerned ants. Hence, there is clearly an urgent need for rigorously designed studies addressing the rate of recovery in less known but functionally important taxa. This study builds on the sampling and resampling of dung beetle communities at a replicate set of sites. Dung beetles are functionally important taxa playing a crucial role in tropical and subtropical forests by re- cycling nutrients (Halffter and Matthews 1966, Mittal 1993, Estrada et al. 1998), aerating soil (Mittal 1993), controling vertebrate parasites (Fincher 1973, Bryan 1976), and dispersing seeds (as recently reviewed by Andresen and Feer [2005]). In some instances, dung beetles may even act as plant pollinators (Sakai and Inoue 1999). The first sampling was conducted by Bert Klein in 1986. In his comparison of dung beetle assemblages between three experimentally isolated 1-ha forest frag- ments, three 10-ha fragments, and three continuous for- est areas, Klein (1989) found that forest fragments had fewer and smaller beetle species and lower population densities than intact forest. These differences were clearly reflected in ecosystem functioning: dung de- composed at a slower rate in 1-ha fragments than in larger forest tracts. The second sampling was conducted by Ellen An- dresen in 1996���1997. Andresen (2003) focused on two of Klein���s (1989) original forest fragments (a 1-ha and a 10-ha fragment), two additional forest fragments, and two sites in continuous forest. At this stage, despite some general differences in sampling techniques, An- dresen could still discern differences in the species richness, general abundance, and size of dung beetles among forest fragments of different size, 10 years after Klein���s (1989) initial study was conducted. In addition, she also detected another consequence of forest frag- mentation on ecosystem functioning: dung beetles were quicker to bury seeds in continuous forest than in forest fragments, with potential implications for plant regen- eration. The third sampling was conducted by ourselves in 2000, 15 years after Klein performed his original study and three years after Andresen revisited the sites. We used Klein���s original methods to sample the local dung beetle communities and assessed the changes that had taken place in intervening years. More specifically, we asked: (1) How has the relative abundance of individual species within forest fragments changed over time? (2) How have the beetle assemblages of the matrix habitat between the fragments changed over time? (3) Can the structure of beetle assemblages supported by small for- est fragments still be distinguished from that of the matrix and that of the continuous forests? METHODS Study area Our study was conducted 80 km north of Manaus, Brazil, at the site of the Biological Dynamics of Forest Fragments Project (BDFFP, (formerly the Minimum Critical Size of Ecosystems Project Lovejoy et al. 1986, Bierregaard et al. 1992, Bierregaard and Gascon 2001). This system is one of a very few tropical sites where habitat fragmentation has been experimentally induced: the fragments were created by scheduled log- ging in 1979���1990 (Bierregaard et al. 1992). Subse- quently, the management of the vegetation surrounding the forest fragments has varied substantially in both time and space, with secondary vegetation repeatedly invading the clear-cut areas, only to be cut back and abandoned again. Given the exceptional knowledge about both the dung beetle fauna and fragmentation history of these Brazilian sites, the system offers a unique opportunity to assess the manner in which the assemblage structure develops through time after ex- perimental disturbance. Sampling sites At each of three different sites (Cidade Powell, Co- losso, and Dimona), four habitats were sampled: a 1- ha forest fragment, a 10-ha forest fragment, the matrix habitat between the fragments, and a continuous forest tract (for a map, see Klein [1989]). General conditions at the study site as well as details of each forest frag- ment are described by Klein (1989), Lovejoy et al. (1986), and Bierregaard et al. (1992). More recently, it has been shown that the sizes of the forest fragments slightly differ from their nominal values, with the true size of 1-ha fragments ranging from 1.56 to 2.8 ha and that of 10-ha fragments ranging from 10.7 to 13 ha (M. Santamar�� ��a-Gomez, personal communication). Successional change The forest fragments were originally created by scheduled logging in 1980 (Colosso), 1983 (Cidade Powell), and 1984 (Dimona). However, the vegetation around fragments rapidly regenerates (Bierregaard et al. 1992, 2001), and when Klein sampled the sites in 1986, scattered secondary vegetation had begun to ap- pear. At Cidade Powell, the forest canopy had already closed at a height of 2.5 m. In 1997���1998, when An- dresen (2003) resampled the two sites at Colosso, the vegetation at this site had recently been burnt and cleared (in 1994���1995), and the fragments were sur- rounded by pasture and/or low (2���4 m) secondary growth composed mostly of Vismia spp. (Clusiaceae) and Cecropia spp. (Cecropiaceae Andresen 2003). When we returned to the sites in 2000, successional change had progressed substantially at all sites. At Ci- dade Powell, the vegetation between the forest frag- ments had not been cleared in the intervening years and now formed a tall (25 m) closed-canopy forest