Journal of Tropical Ecology (2005) 21:1–5. Copyright © 2005 Cambridge University Press DOI: 10.1017/S0266467404001944 Printed in the United Kingdom

SHORT COMMUNICATION

Edge effects on post-dispersal seed removal in a fragmented rain forest in French Guiana St´ephanie Chauvet1 and Pierre-Michel Forget Mus´eum National d’Histoire Naturelle, D´epartement Ecologie et Gestion de la Biodiversit´e, UMR 5176 CNRS-MNHN, 4 av. du Petit Chˆateau, F-91800 Brunoy, France (Accepted 30 April 2004) T1

Key Words: coconut copra, edge effects, French Guiana, forest fragmentation, land bridge islands, rodents, seed removal, water matrix

Forest fragmentation results in the creation of edge habitats, which can be responsible for large modifications in community dynamics (Murcia 1995, Saunders et al. 1991). Plant recruitment along edges is favoured for early successional species but is limited for core forest species (Fox et al. 1997, Williams-Linera 1990). Because the former generally produce large quantities of small seeds while the latter produce fewer large seeds (Hammond & Brown 1995), seed resource availability along the edge is biased in favour of smaller seeds. Such seeds are more likely to attract small rodents than large scatterhoarding species (Adler 1998, Forget et al. 1998). Moreover, the density and activity of forest animals in edge habitats can also be altered as a consequence of their specific responses to habitat changes, varying from edge avoidance to edge preference (Bowers & Dooley 1993, Goosem 2000). These changes in seed resource availability and rodent communities can have cascading effects on plant–rodent interactions in edge habitats and post-dispersal seed predation processes (Holl & Lulow 1997). However, previous studies focusing on forest edge effects on seed predation provided contradictory results. Some reported no significant difference in seed removal between edge and interior habitats (Diaz et al. 1999, Holl & Lulow 1997, Osunkoya 1994), but two reported higher seed removal rates in the interior (Burkey 1993, Restrepo & Vargas 1999) and one reported higher seed removal along the edge (Jules & Rathcke 1999). Such inconsistency may reflect the confounding influences of other factors. Firstly, dynamics in forest remnants are predominantly driven by factors arising from the

1 Corresponding author, at Ecology and Environmental Science, Ume˚ a University, SE-901 87 Ume˚a, Sweden. Email: chauvet [email protected]

surrounding landscape (Saunders et al. 1991), so that edge effects can be difficult to dissociate from matrix effects that may vary between studies according to the type of fragmentation involved (e.g. savannas, secondary forest, or agricultural lands). Secondly, according to Janzen (1970) and Schupp (1992), seed predation rates might depend on the density of conspecific seeds. If the spatial distribution of the study seeds is not homogeneous over a gradient of distance from edge to interior, conspecific density dependence could obscure the edge effects on rodent foraging activity. Such situations may arise from either a heterogeneous distributions of adult trees (Condit et al. 2002), or differences in levels of seed production between edge and interior of forests (Jules & Rathcke 1999), or greater seed abortion near the edge (Forget et al. 1999). In this study, we aimed to test the effects of edges per se on seed removal by rodents at a fragmented rain forest in French Guiana, by carrying out a field experiment where the above two sources of error are circumvented by the type of fragmentation involved and by using an introduced food item (i.e. coconut copra, see below) rather than a native seed species. The study was conducted at the Mus´eum National d’Histoire Naturelle’s (MNHN) biological station of Saint Eug`ene in French Guiana (4◦ 51 N, 53◦ 04 W) (Claessens et al. 2002, Cosson et al. 1999). The station is located upstream from a storage dam, and encompasses a large number of forest islands and peninsulas of continuous forest isolated by flooding in 1994–95 (Figure 1). As at Gatun Lake (Panama) or Lago Guri (Venezuela), our study site thus presents a water matrix that is an unsuitable habitat for rodents. This implies that no invasive rodents species associated with secondary and disturbed habitats have established in the matrix, so that

2

STE´ PHANIE CHAUVET AND PIERRE-MICHEL FORGET

Figure 1. Location of study patches and transects. Interior and edge transects are represented with dotted lines and solid lines marked with white arrows, respectively.

rodent communities occurring in fragments consist only of forest-rodent species that were present in the area prior to fragmentation (Dalecky et al. 2002, Ringuet 2000). Because seed removal in our patches is not influenced by rodent communities from the surrounding matrix habitat, this study site allows us to measure edge effects per se, separately from confounding matrix effects. Fieldwork was carried out on three islands differing in size (number 58 [22 ha], 2 [28 ha] and 3 [67 ha]), and two patches in continuous forest that were adjacent to the flooded area (CF2 and CF3) (following nomenclature in Claessens et al. 2002) (Figure 1). All these patches shelter the larger rodents that may be implicated in seed removal: agoutis (Dasyprocta leporina), acouchies (Myoprocta acouchy) and pacas (Agouti paca) (Dalecky et al. 2002), as well as spiny rats (Proechimys spp.) (Ringuet 2000). Since the time of fragmentation, edges have not been affected by any other perturbation, so that the vegetation has begun to close around the fragments. These edges can therefore be referred to as closed, nonfire encroached (Didham & Lawton 1999), and consist of a girdle of dense vegetation c. 10–15 m wide, dominated by pioneer tree species and showing a high density of small trees with low penetration of light to the forest floor. In contrast, interior habitats are dominated by canopy tree species and have an open understorey. For each of the five study patches, two 350-m transects were delimited: one in edge habitat, at c. 5–10 m from the forest boundary (edge

transect), the other in interior forest (interior transect) at least 100 m away from the edge at all patches except island 2, where the minimum distance is 50 m (Figure 1). This edge transect is nevertheless considered to be independent from the interior transect because of the restricted width of edge habitat and the obvious differences in vegetation structure between the two transects. The experiment was performed twice in 1999 at the same locations: in mid-May, i.e. during the wet season after the peak of resource availability, and at the end of November, i.e. early in the next wet season when resources were low. To control for the density and distribution of the resources under study, and therefore avoid confounding effects of conspecific seed density (Janzen 1970, Schupp 1992), we used an exotic resource, i.e. coconut copra, rather than a native one (see Hallwachs 1994). Coconut copra was divided into resource items (hereafter called items), each weighing 5.0 g on average (± SD 0.5 g), and placed in sets of 10 items each in order to simulate primary dispersal and to represent an attractive resource for rodents, i.e. 50 g on average (Hallwachs 1994). Along each 350-m transect, 15 sets of 10 items were placed 25 m apart. Previous studies have demonstrated that influence of microhabitat on rodent foraging activity, and have further revealed that 20 m distances between seed sets constitute independent observations for the assessment of seed removal (Forget et al. 1998, Hallwachs 1994). Thus the distinct item sets were used as replicates in statistical

Edge effects on seed removal

analyses. The duration of experiment was deliberately limited to 3 d in order to avoid bias due to the consumption of entire items by ants, which find coconut copra very attractive. Sets were thus placed early in the morning of day 1, and then checked late in the afternoon of day 3, with the number of items per set remaining untouched by terrestrial mammals recorded. These data were converted to proportions and analysed using analysis of variance for split plot design with repeated measures. To satisfy parametric assumptions, the data were arcsine-squareroot transformed before analyses. In this design, patch and edge were treated as between-block treatments, while period was treated as within-block treatment, i.e. as repeated measures. Because patch was a random effect and edge a fixed effect, these two between-block treatments were analysed as a mixed model ANOVA, by using as error term the interaction patch × edge to test for these effects. This analysis was performed with SYSTAT 8.0 software (SPSS 1999), using the General Linear Model procedure. Both edge and patches significantly affected seed removal (edge, F(1,140) = 11.0, P = 0.001; patch, F(4,140) = 32.6, P < 0.001), while period, and the second and third order interactions were all not significant (0.12 < P < 0.91). Seed removal was variable among patches, with the highest and lowest proportions of intact items observed at islands 58 and 3 respectively, and intermediate proportions observed at the two continuous forests (Figure 2). These results were consistent with observations from other studies at the same site (Chauvet 2001), confirming that seed removal is highly patchy (see also Holl & Lulow 1997, Willson 1988). As observed in previous studies using native seed species (Burkey 1993, Restrepo & Vargas 1999), edges reduced the level of seed removal by terrestrial mammals relative to removal in the interior habitat, as more intact items were found along the edge transects than along the interior transects for all patches in November, and three of the five patches in May (Figure 2). This may reflect greater overall resource availability at the edges compared with interior habitats and/or a lesser degree of attendance by seed consumers. This latter case can result from edge avoidance by seed foragers and/or from habitat preference, e.g. by agoutis (D. leporina) that favour open understorey forest, and avoid habitat with dense undergrowth (Dubost 1988). The magnitude of edge effects on seed removal was variable among patches, since the differences observed between transects within patches ranged from large to small, with for instance 58 vs. 19% and 3 vs. 0% intact items, for CF3 in May and island 3 in November respectively. In fact, neither islands 2 or 3 revealed marked edge effects because nearly all items were removed after 3 d at these patches (independent of the transects and period). Nevertheless, a census of these patches after the first day revealed a greater number of intact items

3

Figure 2. Percentages of seeds (mean ± SE) not removed by rodents, along edge and interior transects, at five forest patches, in May and November 1999. Patches 58, 2 and 3 are islands with respective size of 22, 28 and 67 ha; patches CF2 and CF3 are located in continuous forest. N = 150 seeds per transect.

remaining at the edge than in the interior: namely 74 vs. 46% and 67 vs. 11%, for islands 2 and 3 in May, respectively. This suggests that islands 2 and 3 suffer a limitation in global resource abundance relative to rodent abundances, leading to a high exploitation of seed resources at both the edges and in the interior. Therefore, despite the likely influence of edge habitats on removal rate as revealed after 1 d experiment, the small amount of resource used in this experiment might have been insufficient to reveal any edge effect after 3 d, as ultimately nearly all items were consumed on these two islands. Hammond & Brown (1995) observed that dispersal mode was a powerful explanatory component of seed mass variation among trees, and that heavier seeds were associated with canopy or emergent trees or lianas whose seeds are dispersed by mammals and/or gravity. The size of resource items we used in this experiment (5 g) simulated native seeds belonging to this category of plant, and allowed handling by both small and large rodents (Forget et al. 1998). Furthermore, additional experiments conducted at this study site revealed that 5-g items of coconut copra were partly dispersed by scatterhoarding rodents in interior habitats, as observed for native tree species (Chauvet 2001). Therefore, the edge effects on seed removal (independent of any matrix effects) that

STE´ PHANIE CHAUVET AND PIERRE-MICHEL FORGET

4 we observed with this introduced resource may have implications for seed survival of native tree species, and further suggests that tree recruitment could be indirectly affected in edge habitats through modifications to seed removal processes (see Asquith et al. 1997, Leigh et al. 1993, Terborgh et al. 2001). However, tree recruitment is a complex and multistage process, and so further investigations are needed to assess the global significance of edge effects on plant regeneration. Recommended future directions include further experiments comparing our results with those for seed removal of native tree species, and additional investigations of the effects of forest edges on the subsequent stages of plant regeneration, namely seed germination, seedling survival/growth and juvenile recruitment. ACKNOWLEDGEMENTS We thank G. Dubost and C. Erard who initiated the Saint-Eug`ene project; J.-C. Baloup, P. Cerdan and the staff of Hydreco for their invaluable help in organization of field sessions; Laboratoire G´enˆomes, Populations et Interactions in Montpellier that welcomed S. Chauvet during writing period; A. Blakaman, A. Dalecky, A.-S. Hennion, A. Lyet and S. Ratiarison who participated in field work; A. Bar-Hen, P.-O. Cheptou and J.-D. Lebreton for their advice in statistical analysis; and A. Dalecky, F. G´enin, P. Jansen, D. McKey, B.-G. Mc Kie, E. Schupp and two anonymous reviewers for helpful comments on the manuscript. This study was supported by E.D.F. (Convention Mus´eum/E.D.F. GP7531) and a pre-doctoral grant of the Minist`ere de I’Education Nationale, de I’Enseignement Sup´erieur et de la Recherche.

LITERATURE CITED

CONDIT, R., PITMAN, N., LEIGH, E. G., CHAVE, J., TERBORGH, J., FOSTER, R. B., NUNEZ, P., AGUILAR, S., VALENCIA, R., VILLA, G., MULLER LANDAU, H. C., LOSOS, E. & HUBBELL, S. P. 2002. Beta-diversity in tropical forest trees. Science 295:666–669. COSSON, J. F., RINGUET, S., CLAESSENS, O., DE MASSARY, J. C., DALECKY, A., VILLIERS, J. F., GRANJON, L. & PONS, J. M. 1999. Ecological changes in recent land-bridge islands in French Guiana, with emphasis on vertebrate communities. Biological Conservation 91:213–222. DALECKY, A., CHAUVET, S., RINGUET, S., CLAESSENS, O., JUDAS, J., LARUE, M. & COSSON, J.-F. 2002. Large mammals on small islands: short term effects of forest fragmentation on the large mammal fauna in French Guiana. Revue d’Ecologie 57:145–164. DIAZ, I., PAPIC, C. & ARMESTO, J. J. 1999. An assessment of postdispersal seed predation in temperate rain forest fragments in Chiloe Island, Chile. Oikos 87:228–238. DIDHAM, R. K. & LAWTON, J. H. 1999. Edge structure determines the magnitude of changes in microclimate and vegetation structure in tropical forest fragments. Biotropica 31:17–30. DUBOST, G. 1988. Ecology and social life of the red acouchy, Myoprocta exilis; comparison with the orange-rumped agouti, Dasyprocta leporina. Journal of Zoology 214:107–123. FORGET, P. M., MILLERON, T. & FEER, F. 1998. Patterns in postdispersal seed removal by neotropical rodents and seed fate in relation to seed size. Pp. 25–49 in Newbery, D. M., Prins, H. H. T. & Brown, N. D. (eds). Dynamics of tropical communities. Blackwell Science, Oxford. FORGET, P. M., KITAJIMA, K. & FOSTER, R. B. 1999. Pre- and postdispersal seed predation in Tachigali versicolor (Caesalpiniaceae): effects of timing of fruiting and variation among trees. Journal of Tropical Ecology 15:61–81. FOX, B. J., TAYLOR, J. E., FOX, M. D. & WILLIAMS, C. 1997. Vegetation changes across edges of rainforest remnants. Biological Conservation 82:1–13. GOOSEM, M. 2000. Effects of tropical rainforest roads on small mammals: edge changes in community composition. Wildlife Research 27:151–163. HALLWACHS, W. 1994. The clumsy dance between agoutis and plants:

ADLER, G. H. 1998. Impacts of resource abundance on populations of a tropical forest rodent. Ecology 79:242–254. ASQUITH, N. M., WRIGHT, S. J. & CLAUSS, M. J. 1997. Does mammal

scatterhoarding by Costa Rican dry forest agoutis (Dasyprocta punctata: Dasyproctidae: Rodentia). Ph.D. dissertation, Cornell University, Ithaca.

community composition control recruitment in neotropical forests? Evidence from Panama. Ecology 78:941–946. BOWERS, M. A. & DOOLEY, J. L. 1993. Predation hazard and seed

HAMMOND, D. S. & BROWN, V. K. 1995. Seed size of woody plants in relation to disturbance, dispersal, soil type in wet neotropical forests. Ecology 76:2544–2561.

removal by small mammals: microhabitat versus patch scale effects. Oecologia 94:247–254. BURKEY, T. V. 1993. Edge effects in seed and egg predation at

HOLL, K. D. & LULOW, M. E. 1997. Effects of species, habitat, and distance from edge on post-dispersal seed predation in a tropical rainforest. Biotropica 29:459–468.

two neotropical rainforest sites. Biological Conservation 66:139– 143. CHAUVET, S. 2001. Effets de la fragmentation foresti`ere sur les interactions

JANZEN, D. H. 1970. Herbivores and the number of tree species in tropical forests. American Naturalist 104:501–527. JULES, E. S. & RATHCKE, B. J. 1999. Mechanisms of reduced Trillium

plantes-animaux: cons´equences pour la r´eg´en´eration v´eg´etale. Th`ese de l’Universit´e Pierre et Marie Curie, Paris, France.

recruitment along edges of old-growth forest fragments. Conservation Biology 13:784–793.

CLAESSENS, O., GRANJON, L., DE MASSARY, J.-C. & RINGUET, S. 2002. La station de recherche de Saint-Eug`ene: situation, environnement et pr´esentation g´en´erale. Revue d’Ecologie 57:21–37.

LEIGH, E. G., WRIGHT, J. S., HERRE, E. A. & PUTZ, F. E. 1993. The decline of tree diversity on newly isolated tropical islands: a test of a null hypothesis and some implications. Evolutionary Ecology 7:76–102.

Edge effects on seed removal

5

MURCIA, C. 1995. Edge effects in fragmented forests: implications for

SCHUPP, E. W. 1992. The Janzen–Connell model for tropical tree

conservation. Trends in Ecology and Evolution 10:58–62. OSUNKOYA, O. O. 1994. Postdispersal survivorship of north Queensland rainforest seeds and fruits: effects of forest, habitat and

diversity: population implications and the importance of spatial scale. American Naturalist 140:526–530. SPSS INC. 1999. SYSTAT 9.0 Statistics. Chicago, Illinois.

species. Australian Journal of Ecology 19:52–64. RESTREPO, C. & VARGAS, A. 1999. Seeds and seedlings of two neotropical montane understory shurbs respond differently to

TERBORGH, J., LOPEZ, L., NUNEZ, P. V., RAO, M., SHAHABUDDIN, G., ORIHUELA, G., RIVEROS, M., ASCANIO, R., ADLER, G. H., LAMBERT, T. D. & BALBAS, L. 2001. Ecological melt-

anthropogenic edges and treefall gaps. Oecologia 119:419–426. RINGUET, S. 2000. An assessment of the potential influence of rainforest fragmentation on small terrestrial mammal predation in French

down in predator-free forest fragments. Science 294:1923– 1926. WILLIAMS-LINERA, G. 1990. Origin and early development of forest

Guiana. Revue d’Ecologie 55:101–116. SAUNDERS, D. A., HOBBS, R. J. & MARGULES, C. R. 1991. Biological consequences of ecosystem fragmentation: a review. Conservation

edge vegetation in Panama. Biotropica 22:235–241. WILLSON, M. F. 1988. Spatial heterogeneity of post-dispersal survivorship of Queensland rainforest seeds. Australian Journal of

Biology 5:18–32.

Ecology 13:137–145.