LRH : Chauvet, Leitao, Dalecky, and Forget RRH : Mast-fruiting tree in a fragmented landscape

Forest fragmentation and mast-fruiting strategy: consequences for seed and seedling survival in Vouacapoua americana (Caesalpiniaceae) 1

Chauvet S.2, a, A. Leitao b, A. Dalecky c, and P.-M. Forget a

a

Museum National d’Histoire Naturelle, Laboratoire d’Ecologie Générale, 4 avenue du Petit

Château, 91800 Brunoy, France b

Evolutionary and Ecological Sciences, Van der Klaauw Laboratory, P.O. Box 9516, 2300

RA Leiden, Netherlands C

Centre d’Ecologie Fonctionnelle et Evolutive - CNRS, 1919 route de Mende,

34293 Montpellier, France

1

Received ___; revision accepted ___. Corresponding author at: Ecology and Environmental Science, Umeå University, S-901 87 Umeå, Sweden. Fax: 00 46 90 786 67 05 E-mail: [email protected] 2

ABSTRACT Fragmentation of tropical forests rapidly induces a loss of animal forest species richness in favor of resistant and opportunist species, and a limitation in vegetal resource availability to consumers. These changes in animal and vegetal communities consequently affect the preexisting dynamics of ecosystems, and particularly interactions between organisms. Because rodents have major effects on seed and seedling survival, it is expected that modifications in plant-rodent interactions will affect plant regeneration. In this study, we focused on Vouacapoua americana, a mast-fruiting Caesalpiniaceae tree species strictly dependent on rodents for seed dispersal. Using experimental approaches, we examined two major steps of tree regeneration: seed fate and seedling survival through one year. Experiments were conducted in a recently fragmented forest (Saint-Eugène, French Guiana), in four isolated forest patches and two mainland sites. Seed survival showed important variations between patches, being lower on one of the islands and in one of the mainland sites. After recent fragmentation, the survival of V. americana seeds was not related to the isolation status of a site (namely islands vs. mainlands), but was affected by two proximal factors: the number of fruit species available, and the density of conspecific trees. The fraction of seedling mortality due to terrestrial mammals did not show significant differences between patches and were apparently not related to either of the proximal factors investigated. Results are examined in regard to satiation hypothesis, and implications for conservation and management are discussed.

Key words: forest fragmentation, French Guiana, mast fruiting, rodents, satiation, seed dispersal, seed predation, seedling survival, tropical rainforest, Vouacapoua Americana

Forest fragmentation is a widespread phenomenon resulting in the isolation of fragments (Saunders et al. 1991, Zuidema et al. 1996), where dynamics of communities is modified. Plant recruitment on fragments is usually favored for early successional species and limited for core forest species (Laurance et al. 1998), owing to changes in abiotic conditions along the edge (Chen et al. 1992, Didham & Lawton 1999), and to the alteration of the complex trophic relations between the animal and the plant communities (Terborgh et al. 2001). Among animals implicated in these interactions, rodents are essential actors first because they are major seed predators (Howe et al. 1985, Osunkoya 1994, Holl & Lulow 1997), and second because they participate in plant regeneration as seed dispersers when they bury seeds for delayed consumption (Smythe 1989, Hallwachs 1994). By disrupting plant-rodents interactions, forest fragmentation thus affects plant recruitment on fragments via modifications in seed removal (Asquith et al. 1997, 1999; Harrington et al. 1997) and seedling establishment (De Steven & Putz 1984, Benitez-Malvido 1998). For instance, at Gatun Lake (Panama), secondary seed dispersal by rodents does not occur in small islands where large scatterhoarding rodents have disappeared following fragmentation (Asquith et al. 1997), resulting in plant recruitment favored only for plant species that do not require burying prior to seedling establishment (Leigh et al. 1993). In these previous studies, disruptions of plant-rodent interactions have been ascribed to changes in rodent communities, as these latter generally show a reduction in species diversity and the dominance of a few opportunist and resistant species at medium to long term after fragmentation (Laurance 1994, Adler 1996, Dunstan & Fox 1996). However, alteration of plant-rodent interactions can also arise from the direct modification of plant community during fragmentation, owing to the loss of some parts of the forest and to the heterogeneous distribution of tree species (Condit et al. 2002). For instance, when forest fragmentation is due to the flooding of a river, all tree species whose distribution is restricted to the river

3

borders are absent on fragments. The plant community is thus specific of the type of habitat isolated (i.e. top of slopes, in the case of flooding), and shows only a sub-sample of the tree species that were present in the area prior to fragmentation. By affecting the whole seed resources available to rodents, these modifications of the plant community are likely to influence plant-rodents interactions for tree species whose seed removal by rodents is determined by satiation process acting at the community level (Schupp 1990, Chauvet et al. 2003). Furthermore, for tree species whose populations were large prior to fragmentation, the isolation of forest fragments induces a reduction in population size (Zuidema et al. 1996). This reduction results in a decreased amount of conspecific seeds present in the population, and may thus affect the plant-rodent interactions for tree species whose seed removal by rodents is influenced by satiation phenomenon occurring at the population scale (Schupp 1992). This may be particularly conspicuous for mast-fruiting species, for their reproductive strategy (episodic synchronous reproduction events, interspersed by irregular period of low seed production) may have evolved to satiate predators (Janzen 1974, Curran & Leighton 2000). As observed by Curran et al. (1999) for Dipterocarpaceae, predator satiation for mastfruiting trees can occur across the landscape rather than within a local scale owing to synchronous seed production that occurs across extensive areas. This implies that since fragmentation reduces a once-contiguous expanse of a tree population to a number of subpopulations of restricted size, then it reduces the spatial extent of participation in the mastevent, and may thus prevent the mast-fruiting species from satiating seed predators. It is thus expected that mast-fruiting species should be particularly affected by fragmentation, via its direct effect on plant community and thereby on plant-animal interactions. Indeed, if their tree populations are subdivided, and if the seed predators are still present on fragments, then these mast-fruiting species should undergo abnormally high level of seed predation in comparison with un-fragmented populations, owing to the disruption of their predator satiation strategy.

4

However, for animal dispersed species, one needs to qualify this likely negative effect of forest fragmentation. Indeed, as Herrera et al. (1998) have underlined, besides satiating seed predators, masting also satiates seed dispersal agents, suggesting the existence of a trade-off between the advantages of satiating predators and the disadvantages of satiating dispersers. Thus, if seed dispersers are still present on fragments, then it is expected that fragmentation will have a positive effect on masting species, via the reduction in their population size, by disrupting the seed dispersers satiation and thereby inducing an increase in seed dispersal rate. In this study, we investigated these hypotheses of a disruption of predator and dispersers satiation for Vouacapoua americana Aublet (Caesalpiniaceae), a mast-fruiting canopy tree species, whose seeds are much consumed by all rodent species and only dispersed by scatterhoarding rodents (Forget 1990). This study has been conducted in a recently fragmented landscape in French Guiana, where V. americana seed predators and dispersers are still present on all the fragments under study (Dalecky et al. 2002), and where V. americana populations have been subdivided. To investigate the predictions of higher seed predation and dispersal in comparison to what is known from non fragmented populations, we first assess the proportions of consumed and scatterhoarded seeds after one month using marked seeds experiments. Second, to determine the proportion of efficient seed dispersal among scatterhoarded seeds (namely hoarded and not secondarily consumed seeds) after six and twelve months, we conducted an exclosure experiment with artificially hoarded seeds. In both experiments, we controlled for the effect of proximal factors on seed fate, namely 1) the abundance of Dasyprocta leporina which is the main seed disperser in this area, 2) the abundance and 3) diversity of resources available to rodents, 4) and the density of Vouacapoua americana trees.

5

METHODS STUDY SITE.–– Field work was conducted in French Guiana, at Saint-Eugène (4°51’ N, 53°04’ W) (Claessens et al. 2002) (Figure 1). This area, covered by tropical rainforest, was fragmented in 1994-95 by flooding for a hydroelectric dam (Figure 1). The landscape at the study station is composed of a large number of islands and islets of various shapes, sizes (from 0.1 to 67 ha) and degree of isolation (ca. 20 to 500 m) (Claessens et al. 2002). Mammal communities in areas surrounding the lake remain similar to non-fragmented forests, while they have become impoverished on islands owing notably to the disappearance of large predators (which visit islands only occasionally) and of some monkey species as Ateles paniscus and Cebus spp. (Dalecky et al. 2002). Despite this overall mammal impoverishment, most species involved in trophic relations with V. americana trees were still present and abundant on fragments at time of experiment. For instance, the following species were still observed on most islands larger than 5 ha (Dalecky et al. 2002): red howler monkeys (Alouatta seniculus), peccaries (Tayassu pecari and Pecari tajacu), brocket deers (Mazama americana and M. gouazoubira), and especially large granivore species, i.e. pacas (Agouti paca), agoutis (Dasyprocta leporina) and acouchies (Myoprocta exilis). This recent fragmentation thus provide an unique opportunity to investigate the effects of altered plant community per se on plant-animal interactions, separately from the effects of modified animal communities. This contrasts with studies conducted in old fragmented landscape where animal communities are already much modified on fragments (Terborgh et al. 2001), thus preventing the investigation of the plant component in the plant-animal interactions. However, the absence of quantitative data for periods anterior to fragmentation prevents from investigating possible changes in the densities of mammal species. Furthermore, in contrast to larger rodents, small rodent communities have become impoverished following fragmentation, with only spiny rats (Proechimys spp.) surviving on some islands in 1997

6

(Granjon et al. 2002). However, V. americana being mostly consumed by larger rodents (Jansen 2003), we assumed that this changes in small rodent community should not significantly affect the seed fate of our study species.

STUDY SPECIES.–– Vouacapoua americana Aublet (Caesalpiniaceae) is a hardwood timber tree species present from Amazonian Brazil to French Guiana and Surinam (Schulz 1960). In French Guiana, flowering and fruiting of this species occurs in February-March and AprilJune, respectively. Estimated fruit production ranges from 130 to 3200 fruits per tree (Sabatier 1983), with fruits being pods enclosing one (rarely two) heavy seeds (about 30 g in average) rich in cotyledonous reserves. Among the most likely seed predators are rodent species, peccaries, ants, and beetle larvae (Loubry 1988, Forget 1990). Owing to their large size, seed dispersal is dependent on scatterhoarding rodents, especially agoutis and acouchies that cache seeds under ground and / or under leaves for later consumption (Forget 1990). Germination occurs a few weeks after seed dispersal. Survival of unburied seeds and seedlings below adult trees is very low due to drought and heavy infestation by beetles (Forget 1994). In contrast, cached seeds have a relatively high probability of establishment because burial of seeds promotes hydration and germination and leads to escape from drought, insects and vertebrate predators (Loubry 1988, Forget 1990). This tree species has a spatial distribution that is strongly aggregated (Schulz 1960, Traissac 1998, Forget et al. 1999) and presents a phenology characteristic of mast-fruiting species (Janzen 1974, Curran & Leighton 2000). During episodic reproduction events (interspersed by periods greater than one year with no or few seeds produced), fructification is highly synchronized between trees over large areas, resulting in tens of thousands seeds produced across population (Chauvet 2001). This great amount of seeds is responsible for predator satiation, resulting in low level of seed removal under tree crowns. For instance, P.-

7

M. Forget and P. Jansen estimated in 1995 that seed removal by terrestrial mammals in Nouragues Biological Station (undisturbed forest, French Guiana) were lower than 5 % under tree crowns (N = 16 trees) after one month experiment, for trees within a large tree aggregate (Forget and Jansen, unpublished data). In contrast, trees within a sparser population were unable to satiate predators and thus suffered higher level of seed removal, i.e. ca. 40 % (N = 8 trees) at Nouragues (Forget and Jansen, unpublished data). At our study site, owing to the reduction of V. americana populations size inherent to the landscape fragmentation, we expect that terrestrial mammals are not satiated by V. americana seed production, and thus expect high level of seed removal as observed by Forget and Jansen in the sparse population.

STUDY PATCHES.–– This study was carried out in six forest patches: two on small islands numbered 14 and 20 (7.9 and 7.5-ha, respectively), two on medium sized islands numbered 2 and 3 (28 and 67-ha, respectively), and two control patches located in continuous forest called CF2 and CF3 according to the nomenclature in Claessens et al. (2002) (Figure 1). In each patch, one 4-ha study plot was delimitated for experiments, where all V. americana trees were censused and mapped. Vouacapoua americana trees are present in all these patches, excepted CF2, and their densities range from 16 (island 2) to 40 (CF3) individual trees with dbh > 25 cm per 4-ha area. Because tree populations are heterogeneously distributed over the studied areas, we further refer to tree density as the mean number of conspecific trees present in a 50m radius around each studied tree (Table 1). At all these patches, agouti (D. leporina) is the large rodent species most frequently encountered. In contrast, acouchies (M. exilis) are more affected by forest perturbations (Dubost 1988), and are thus rarely observed on islands (Dalecky et al. 2002; S. Ratiarison & S. Chauvet pers. obs.). Abundance of agoutis (D. leporina) has been estimated in different patches at Saint-Eugene in April-May 1998 (Chauvet 2001). These estimates for the patches under study are given in Table 1.

8

OVERALL FRUIT RESOURCE.–– We measured overall fruit resources available on the ground during V. americana fruiting period, i.e. in late April-May, using transect line censuses. One 200-m long x 1-m wide transect was laid out across the 4-ha area under study at each patch, and divided into 20 10-meter long sections. Abundance of fruits encountered on the ground was noted for each individual species for each section separately, with an abundance index ranging from I to V, corresponding respectively to 1-10, 10-25, 25-50, 50-100, and >100 fruits. Transect censuses were repeated every week from late April to late May, and fruits present on the transect line were systematically removed after each census. Fruits were sampled for reference, and identified to genus or species using reference books by Sabatier (1983) and Van Roosmalen (1985). From these data we calculated the overall abundance of fruit resources falling on transects during the time of the study. Each abundance index was replaced by the median of its class, namely I by 5, II by 17.5, III by 37.5, IV by 75 and V by 200 fruits. These fruit abundances were then pooled for all sections per transect, for all species and for the whole period, resulting in a single index of fruit abundance per patch (for similar calculation, see Levey 1988). In addition to this abundance index, we also recorded the number of different fruit species that fell on each transect during the study period.

SEED FATE EXPERIMENTS.–– In order to determine the fate of V. americana seeds after removal by terrestrial mammals, we conducted two experiments with marked seeds during fruit production: one under parent tree crowns, and one along the 200-m long transects crossing the 4-ha study plots. In both cases, following Forget (1990) and Brewer and Rejmanek (1999), seeds were marked with 1-m long white nylon thread in order to relocate them after removal by mammals. Following Forget (1990) who retrieved 90 % of removed seeds at less than 10 m from seed set location, we searched for removed seeds within a 10 m-

9

radius area, and recorded located seeds as consumed or dispersed when the thread was found alone or with the seed cached under leaves and / or under soil, respectively. Removed seeds that were not relocated have been assigned to consumed and dispersed categories, according to the ratio of consumed / dispersed observed among retrieved seeds. The first experiment was conducted during the peak of fruit production, between 2-26 May. Among each of the five populations of V. americana (i.e. all patches except CF2), ten fruiting trees were chosen for experiment. We randomly delimited one plot (about 1.1 x 1.3 m) under the crown of each tree, from which all fruits and seeds were removed before setting out 20 thread-marked seeds arranged in four rows of five seeds. During censuses conducted through day 21, we counted the number of intact, consumed and cached seeds. The second experiment was conducted slightly later during fruit production, between 10-28 May. At each of the six forest patches under study (i.e. including CF2), 20 sets (0.5 x 0.5 m) of five marked seeds were placed every 10 meters along the 200-m transects, at five meters on the right and left sides alternatively. During censuses through day 14, we counted the number of intact, consumed and cached seeds. Both marked seed experiments ended at the same period, namely when seed production was over and when seeds started to germinate.

EXCLOSURE EXPERIMENT.–– In order to estimate probability of seed survival and establishment after dispersal by rodents, we conducted seed-seedling survival experiments with and without protection against herbivores by wire exclosures (1.2 m high x 1 m diameter, square mesh = 1 cm). In each of the six patches, ten locations were delimited 20 meters apart along the 200-m transects. At each of these locations, two groups of five germinating seeds (separated from each other by 2-3 meters) were planted beneath ground level at the end of May. One of these groups was placed under exclosure protection, and was thus only exposed to attacks by insects and pathogens. The second one remained unprotected,

10

and was therefore exposed to insect and pathogen attacks, as well as to terrestrial mammals likely to dig up seeds and / or to consume seedlings. Survival until seedling stage was censused after 6 and 12 months.

STATISTICAL ANALYSES.–– To normalize data, all variables expressed as proportions were arcsin square root transformed (Sokal & Rohlf 1981) and analyzed with SYSTAT 8.0 software (SPSS 1998). In experiments with marked seeds, all three variables (i.e. proportions of intact, consumed and cached seeds) were analyzed in two steps. First, we tested for the effect of patch on variations in seed fate variables, using analysis of variance. Second, we investigated which patch characteristics may be responsible for variation in seed fate. Therefore, all seed fate variables were averaged at the patch scale and their relation with patch descriptors were examined using Spearman rank-order correlations. The patch descriptors included in this analysis were the abundance of resource and number of fruit species available, as well as agouti (D. leporina) abundance and mean V. americana tree density (expressed in mean number of conspecifics in 50-m radius around studied trees). In the seed recruitment experiment, we first analysed the effect of protection by wire exclosure on seed-seedling survival using repeated measures analysis of variance, with between factors being (1) protection, (2) patch, and (3) protection x patch, and within factors being time and all interactions including time. Second, we estimated mortality due to terrestrial mammals (i.e. granivory and / or herbivory) as: [mortality when unprotected] – [mortality in wire exclosure]. This mortality, estimated after 6 and 12 months, was analysed in two steps similar to the seed fate experiments: We first tested for an effect of patch on mortality due to terrestrial mammals by using analysis of variance. Then, to investigate whether this mortality was related to habitat characteristics, we computed Spearman rank-

11

order correlations between mean mortality per patch and patch characteristics (i.e. fruit abundance, number of fruit species, agoutis abundance, V. americana density). Finally, to investigate whether mortality due to terrestrial mammals was still acting after seedling establishment, namely between the two controls made after 6 and 12 months, we compared mortality estimates at 6 and 12 months by computing Wilcoxon signed-rank tests for each patch separately.

RESULTS OVERALL FRUIT ABUNDANCE.–– Estimations of fruit availability at the community level during April-May 1998 showed large variations between patches. The number of fallen fruits during the entire period varied from 2,766 on island 20 to 5,718 on island 2 for 200-m transects, and the total number of fruit species encountered along transects varied from 22 on island 2 to 49 species on island 14 (Table 1).

SEED FATE EXPERIMENTS.–– When seeds were deposited under tree crowns, there were significant differences between patches for both intact (F4, 45 = 14.106, P < 0.001), consumed (F4, 45 = 5.298, P = 0.001), and cached seeds (F4, 45 = 7.164, P < 0.001) (Figure 2). This patch effect was due to island 2 which greatly differed from all other patches, with 63.0 ± 9.9 % of intact seeds (vs. 97.5 ± 1.1 to 99.0 ± 0.7 % at the other patches), 19.2 ± 7.3 % of consumed seeds (vs. 1 ± 0.7 to 2.5 ± 1.1 % at the other patches), and 17.8 ± 8.7 % of cached seeds (vs. 0 % at the other patches). Among the patch characteristics examined, only the number of fruit species and V. americana tree density were correlated with seed fate (Table 2). Both these factors were positively related to the proportion of intact seeds and negatively related to the proportion of consumed seeds. None of the patch characteristics were correlated with the proportion of cached seeds. 12

When seeds were deposited along transects there were significant differences between patches for intact (F5, 114 = 23.058, P < 0.001), consumed (F5, 114 = 6.142, P < 0.001), and cached seeds (F5, 114 = 14.640, P < 0.001) (Figure 2). Island 2 and forest CF2 greatly differed from other patches with 62.0 ± 6.0 and 38.0 ± 8.8 % of intact seeds, 23.2 ± 6.0 and 21.7 ± 6.5 % of consumed seeds, and 14.8 ± 4.1 and 40.3 ± 8.6 % of cached seeds, respectively. Other patches had between 94.0 ± 2.9 and 99.0 ± 1.0 % of intact seeds, except island 3, which presented a smaller proportion of intact seeds, i.e. 82 ± 5.2 %. When correlations between seed fate and patch descriptors were investigated for all six study patches (i.e. CF2 included in analysis), then only V. americana density was related to seed fate. This factor was positively correlated with proportion of intact seeds, and negatively correlated with proportions of consumed and cached seeds (Table 2). When correlations between seed fate and patch descriptors were investigated only for study patches where V. americana trees were present (i.e. CF2 excluded), then both the number of fruit species (positively with intact; negatively with consumed and cached seeds) and the V. americana tree density (positively with intact and negatively with consumed seeds) were correlated with seed fate.

EXCLOSURE EXPERIMENT.–– Survival of artificially cached seeds until seedling establishment was significantly greater with protection by exclosures against terrestrial mammals than without protection (F1, 104 = 13.537, P < 0.001). This result was observed in all patches under study, except island 14 where both seeds with and without protection had equivalent survival probabilities (Figure 3). Survival also varied among patches (F5, 104 = 7.110, P < 0.001), being greatest at forest CF3 and lowest on islands 14 and 2. The interaction between patch and protection was non significant (F5, 104 = 0.661, P = 0.654). The estimates of mortality due to terrestrial mammals averaged 18.2 ± 4.4 % in the Saint-Eugène area after six months, ranging from 12.0 ± 12.7 % on island 2 to 26.0 ± 6.7 % in 13

forest CF3, except on island 14 where it was negligible. After 12 months, this mortality averaged 21.4 ± 4.4 % for the Saint-Eugène area, ranging from 18.0 ± 12.8 % on island 2 to 30.0 ± 9.1 % in forest CF3, and being also negligible on island 14. Mortality due to terrestrial mammals was not significantly affected by patches, either after six months (F5, 51 = 0.616, P = 0.688) or after 12 months (F5, 51 = 0.564, P = 0.727). It was also not related to any of the patch characteristics examined (Table 2). Estimations of mortality due to terrestrial mammals after 12 months did not differ from those after six months, in any of the study patches (Table 3).

DISCUSSION SEED REMOVAL AND PREDATOR SATIATION.–– In seed fate experiments conducted under parent tree crowns and across V. americana populations, the proportions of seeds that remained untouched by rodents was high (i.e. greater than 80 %) in all patches, except island 2 and the control patch CF2. However we have evidence from other experiments that seed removal can occur in these patches when examined in different context (Chauvet 2001). Indeed, further experiments conducted one year later with an exotic food resource revealed extremely high removal rates along the transects on all islands examined in this study. This suggests that the low removal rates observed for V. americana seeds are not characteristic of the patches under study but rather characteristic of the context of experiment: namely the resource tested and / or the local environment in May 1998. Previous studies of V. americana have shown that seeds of this species are attractive to rodents, and experiments lasting two weeks are sufficiently long for observing seed removal by rodents (Forget 1990). We thus believe that the negligible removal rates observed in our study are not dependent on the seed resource tested but rather on the local environment in May 1998. Indeed, the local environment at the period of experiment was peculiar, owing to the massive fructification of V. americana populations that year. Owing to a fructification that is episodic and highly

14

synchronized between trees (characteristic of mast-fruiting species), V. americana populations produced in 1998 very large amount of seeds that overwhelmed seed predators and thus allowed seeds to escape predation (Chauvet 2001). Thus, low removal rates observed in our seed fate experiments were consequences of the mast-fruiting strategy evolved in this plant species to satiate seed predators (Janzen 1971, 1974; Shupp 1992; Curran & Leighton 2000). The first seed fate experiment was conducted under parent tree crowns (i.e. in conditions of high local seed density), while the second was conducted away from parent trees (i.e. along transects crossing the populations). Both experiments, although conducted at different spatial scales, showed similarly low seed removal rates. This therefore provides evidence that processes underlying predator satiation by this mast-fruiting species did not occur at the level of individual trees but rather at the level of V. americana populations. Thus, seed predator activity does not seem affected by local seed density or distance to parent tree (Howe 1993, Terborgh et al. 1993, Notman et al. 1996), but rather by the amount of seeds produced by V. americana trees at the patch level (Janzen 1971, Schupp 1992). This is corroborated by the analysis of proximal factors related to seed fate, which showed that patches having greater V. americana tree density exhibited lower seed removal rate. These results may be interpreted as a competition between trees for limited seed dispersers, as observed by Manasse and Howe (1983) for Virola surinamensis (Myristicaceae). In both marked seed experiments, seed fate was variable among patches, but did not show any opposition between patches in continuous forest on the one hand and island patches on the other hand. This suggests that four years after fragmentation occurred, seed fate was not related to the fragmentation status of a site (for similar results, see Harrington et al. 1997). In contrast, island 2 differed from the other four patches with V. americana populations, owing to a greater amount of seeds removed through late May. This characteristic of high

15

removal rate on island 2 has also been observed in other experiments conducted in 1999 (Chauvet 2001). According to the analysis of correlation between patch descriptors and seed fate variables, results observed in 1998 on island 2 are related to the low density of V. americana trees at this patch, as well as to the low diversity of alternative resources available to rodents. This latter result suggests that, in addition to satiation at the population level of V. americana trees, predator satiation also occurs at the larger scale of the plant community, when diversity of alternative resource is great enough (Schupp 1990). When the seed fate experiment was conducted outside the parent tree crown, namely along transects, we observed greater amounts of seeds removed at patch CF2 than at all other patches, despite a very low abundance of agoutis, and a diversity of resource that was similar to other patches. Owing to the absence of any V. americana tree at this patch, this pattern of seed fate may be attributed to the novelty of the food resource tested, that should have made it more attractive to rodents in this patch (anti-apostatic selection, sensu Greenwood 1985) (Hallwachs 1994).

SEEDLING ESTABLISHMENT.–– In exclosure experiment, seed survival through six and twelve months, after being artificially buried, was significantly enhanced when protected by wire exclosure against terrestrial mammals. By comparing seed-seedling survival with and without protection, we evaluated the probability that an artificially dispersed seed was recovered by a terrestrial mammal that was ignorant of the location of the seed, as this may naturally occur when an animal consumes seeds hoarded by another individual (Price & Jenkins 1986, Vander Wall 1990). This source of mortality due to terrestrial mammals remained nonnegligible as about 20 % of seeds were lost that way, at all patches at Saint-Eugene excepted island 14. Because rodents are known to generally search for and locate food using olfaction (Price & Jenkins 1986), this post-dispersal mortality rate that V. americana seeds suffer may result from their large size that result in high emission of chemical components, making them

16

evident and attractive, despite being buried under ground. The surveys conducted after six and twelve months showed comparable levels of mortality due to terrestrial mammals, suggesting that once seedlings were established, these factors of mortality became negligible. This corroborates the observation made by Forget (1990), that post-dispersal predation of V. americana seeds occurs early because of the lack of interest by rodents in germinated seeds, due to the rapid exhaustion of endosperm reserves. Island 14 did not show any mortality due to terrestrial mammals, although the herbivore-granivore mammal community at this patch was similar to that in the other patches (Dalecky et al. 2002, Granjon et al. 2002). Despite this particular case of island 14, our results showed that seed-seedling mortality due to terrestrial mammals did not differ significantly between patches. In contrast, the overall seed-seedling mortality, i.e. including all mortality factors, was affected by patches. This therefore provides evidence that mortality factors other than mammals are variable among patches: namely mortality due to insects, pathogens and abiotic factors, such as light and humidity. Seed-seedling mortality in wire exclosures, due to these latter factors of mortality, was high for most of the patches (except CF3), ranging from 30 to 70 % (see Figure 3). After being artificially buried, seeds were therefore more exposed to insects, pathogens and abiotic mortality factors than to terrestrial mammals. In this recently fragmented landscape, edge-induced effects (Laurance & Yensen 1991, Murcia 1995) may partly explain these high seed-seedling mortality rates. Indeed, higher exposition to sun radiation and higher wind turbulence observed along the edge of fragmented forests result in increased air temperature and evaporative water loss in fragments (Kapos 1989, WilliamsLinera 1990, Camargo & Kapos 1995, Didham & Lawton 1999), which may be responsible for seed desiccation. This phenomenon may have been particularly important on island 14, which is both the most hilly and disturbed patch, and therefore the more exposed to desiccating wind.

17

REGENERATION DYNAMIC AND IMPLICATIONS FOR CONSERVATION.–– We expected that seed fate would be influenced by both communities of rodents and resources available for them (Asquith et al. 1997, Adler 1998). Among rodents that may be involved in removal of large seeds, agoutis may constitute at Saint-Eugène the most important actor owing to their presence and abundance on most islands, in contrast to acouchies that are affected by perturbation (Dubost 1988) and thus rarely encountered (Chauvet 2001, Dalecky et al. 2002). However, among the proximal factors examined, the agoutis abundance was never related to seed fate under or away from parent trees, or to seedling recruitment from artificially dispersed seeds. In contrast, we found evidence that, regardless of agouti abundance, seed survival in the first weeks after production was related to resources: namely to the diversity of all resources available to rodents, and to V. americana tree density (which is related to the amount of seeds produced per patch [Chauvet 2001]). Early stages of regeneration of this tree species were therefore dependent on predator satiation at both community- and populationlevels (Janzen 1974, Schupp 1992). At the population-level, V. americana satiate predators owing to a massive production of seeds that quickly germinate (a few weeks after production), and establish as a seedling bank which are less sensitive to post-dispersal seed predation. Understanding the regeneration dynamics of this tree species in fragmented landscapes would require examining all steps from pollination to adult recruitment. Yet, despite that our results only concern early regeneration stages, they provide implications for conservation and management of this timber tree species. By isolation of forest patches, landscape fragmentation is likely to reduce population size of V. americana. In the same manner, logging reduces population density. Thus, we suggest that fragmentation and logging are likely to affect early stages of regeneration, in case of reduction in V. americana tree density, because it will induce a decrease in the overall fruit production at patch-level, which

18

in turn will alter satiation processes involved in seed survival. The mast-fruiting event of this species would therefore be counterbalanced, resulting in new selective pressures exerted on early stages of regeneration. Consequences of such alterations of processes still need to be investigated, and especially consequences for dispersal processes which are essential to seedling establishment and thus to recruitment.

ACKNOWLEDGMENTS Thanks to D. Amoïda for tree identification, to D. Sabatier for his help in fruit identification, and to L. Gaume for statistical advice. We would like to thank also S. Ratiarison for her help during field work, and S. Traissac for communicating his results concerning the regeneration dynamics of Vouacapoua americana. Many thanks to G. Dubost and C. Erard who initiated the Saint-Eugène project; to J.-C. Baloup, P. Cerdan and the staff of the Laboratoire Environnement de Petit-Saut (Hydreco) for their priceless help in organization of field session. Comments by C. Brouat, B. Giles, C. Krijger, D. Mc Key and E. Schupp improved earlier versions of this manuscript. This study was supported by E.D.F. (Convention Museum / E.D.F. GP7531) and a pre-doctoral grant of the Ministère de l’Education Nationale, de l’Enseignement Supérieur et de la Recherche.

LITERATURE CITED Adler, G. H. 1996. The island syndrome in isolated populations of a tropical forest rodent. Oecologia 108: 694-700. –––––. 1998. Impacts of resource abundance on populations of a tropical forest rodent. Ecology 79(1): 242-254. Asquith, N. M., S. J. Wright, and M. J. Clauss. 1997. Does mammal community composition control recruitment in neotropical forests? evidence from Panama. Ecology 78(3): 19

941-946. –––––, J. Terborgh, A. E. Arnold, and C. M. Riveros. 1999. The fruits the agouti ate: Hymenaea courbaril seed fate when its disperser is absent. J. Trop. Ecol. 15: 229-235. Benitez-Malvido, J. 1998. Impact of forest fragmentation on seedling abundance in a tropical rain forest. Conserv. Biol. 12(2): 380-389. Brewer, S. W., and M. Rejmanek. 1999. Small rodents as significant dispersers of tree seeds in a neotropical forest. J. Veg. Sci. 10(2): 165-174. Camargo, J. L. C., and V. Kapos. 1995. Complex edge effects on soil moisture and microclimate in central amazonian forest. J. Trop. Ecol. 11: 205−221. Chauvet, S. 2001. Effets de la fragmentation forestière sur les interactions plantes-animaux: conséquences pour la régénération végétale. Thèse de l'Université Paris 6, France. Chen, J., J. F. Franklin & T. A. Spies. 1992. Vegetation responses to edge environments in old-growth Douglas-fir forests. Ecological Applications 2: 387−396. Chiarello, A. G. 1999. Effects of fragmentation of the Atlantic forest on mammal communities in south-eastern Brazil. Biol. Conserv. 89: 71-82. Claessens, O., L. Granjon, J.-C. de Massary, and S. Ringuet. 2002. La station de recherche de Saint-Eugène : situation, environnement et présentation générale. Rev. Ecol. 57: 2137. Cosson, J.-F., S. Ringuet, O. Claessens, J.-C. de Massary, A. Dalecky, J.-F. Villiers, L. Granjon, and J.-M. Pons. 1999. Ecological changes in recent land-bridge islands in French Guiana, with emphasis on vertebrate communities. Biol. Conserv. 91(2-3): 213-222. Curran, L. M., and M. Leighton. 2000. Vertebrate responses to spatiotemporal variation in seed production of mast-fruiting Dipterocarpaceae. Ecol. Monogr. 70(1): 101-128. Dalecky, A., S. Chauvet, S. Ringuet, O. Claessens, J. Judas, M. Larue, and J.-F. Cosson.

20

2002. Large mammals on small islands: short term effects of forest fragmentation on the large mammal fauna in French Guiana. Rev. Ecol. 57: 145-164. De Steven, D., and F. E. Putz. 1984. Impact of mammals on early recruitment of a tropical canopy tree, Dipteryx panamensis, in Panama. Oïkos 43: 207-216. Didham, R. K., and J. H. Lawton. 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. J. Zool. Lond. 214: 107–123. Dunstan, C. E., and B. J. Fox. 1996. The effects of fragmentation and disturbance of rainforest on ground-dwelling small mammals on the Robertson plateau, New South Wales, Australia. J. Biogeogr. 23(2): 187-201. FAO. 1997. State of the world’s forests. Rome: Forestry Department, Food and Agricultural Organization. Rome, Italy. Fischer, M. 2000. Species loss after habitat fragmentation. Trends Ecol. Evol. 15(10): 396. Forget, P.-M. 1990. Seed-dispersal of Vouacapoua americana (Caesalpiniaceae) by caviomorph rodents in French Guiana. J. Trop. Ecol. 6: 459-468. –––––. 1994. Recruitment pattern of Vouacapoua americana (Caesalpiniaceae), a rodentdispersed tree species in French Guiana. Biotropica 26(4): 408-419. Granjon, L., S. Ringuet, and G. Cheylan. 2002. Evolution of small terrestrial mammal species richness on newly formed islands in primary tropical forest of French Guiana: a 6 year study. Rev. Ecol. 57: 131-144. Greenwood, J. J. D. 1985. Frequency-dependent selection by seed predators. Oikos 44: 195210. Hallwachs, W. 1986. Agoutis (Dasyprocta punctata): The inheritors of Guapinol (Hymenaea

21

courbaril: Leguminosae). In A. Estrada and T. H. Fleming (Eds.). Frugivores and seed dispersal, pp. 285-304. Dr. W. Junk, Dordrecht, Netherlands. –––––. 1994. The clumsy dance between agoutis and plants: scatterhoarding by costarican dry forest agoutis (Dasyprocta punctata: Dasyproctidae: Rodentia). Ph.D. Dissertation. Cornell University, Ithaca, New York, USA. Harrington, G. N., A. K. Irvine, F. H. J. Crome, and L. A. Moore. 1997. Regeneration of large seeded trees in Australian rainforest fragments: a study of higher-order interactions. In W. F. Laurance and R. O. Bierregaard (Eds.) Tropical forest remnants. Ecology, management, and conservation of fragmented communities, pp. 292-303. The University of Chicago Press, Chicago & London. Holl, K. D., and M. E. Lulow. 1997. Effects of species, habitat, and distance from edge on post-dispersal seed predation in a tropical rainforest. Biotropica 29(4): 459-468. Howe, H. F. 1993. Aspects of variation in a neotropical seed dispersal system. Vegetatio 107/108: 149-162. –––––, E. W. Schupp, and L. C. Westley. 1985. Early consequences of seed dispersal for a neotropical tree (Virola surinamensis). Ecology 66(3): 781-791. Jansen, P. A., and P.-M. Forget. 2001. Scatterhoarding and tree regeneration. In F. Bongers, P. Charles-Dominique, P.-M. Forget and M. Théry (Eds.). Nouragues: dynamics and plant-animal interactions in a neotropical rainforest. Kluwer Academic Publisher, Dordrecht, The Netherlands. –––––, M. Bartholomeus, F. Bongers, J. A. Elzinga, J. Den Ouden, and S. E. Van Wieren. 2002. The role of seed size in dispersal by a scatter-hoarding rodent. In D. levey, W. R. Silva and M. Galetti (Eds.). Seed dispersal and frugivory: Ecology, Evolution and conservation, pp. 209-225. C.A.B.I. Publishing Wallington, UK. Janzen, D. H. 1971. Seed predation by animals. Annu. Rev. Ecol. Syst. 2: 465-492.

22

–––––. 1974. Tropical blackwater rivers, animals, and mast fruiting by the Dipterocarpaceae. Biotropica 6(2): 69-103. Kapos, V. 1989. Effects of isolation on the water status of forest patches in the Brazilian Amazon. J. Trop. Ecol. 5: 173-185. Laurance, W. F. 1994. Rainforest fragmentation and the structure of small mammal communities in tropical Queensland. Biol. Conserv. 69(1): 23-32. –––––, and E. Yensen. 1991. Predicting the impacts of edge effects in fragmented habitats. Biol. Conserv. 55: 77-92. –––––, L. V. Ferreira, J. M. Rankin de Merona, S. G. Laurance, R. W. Hutchings and T. E. Lovejoy. 1998. Effects of forest fragmentation on recruitment patterns in amazonian tree communities. Conservation Biology 12, 460−464. Leigh, E. G. Jr., J. S. Wright, E. A. Herre, and F. E. Putz. 1993. The decline of tree diversity on newly isolated tropical islands: a test of a null hypothesis and some implications. Evol. Ecol. 7: 76-102. Levey, D. J. 1988. Spatial and temporal variation in Costa Rican fruit and fruit-eating bird abundance. Ecol. Monogr. 58(4): 251-269. Loubry, D. 1988. L’impact des insectes séminivores sur la régénération de Vouacapoua americana et Dicorynia guianensis (Caesalpiniaceae). D.E.A. de Biologie Végétale Tropicale. Université Pierre et Marie Curie, Paris, France. Manasse, R. S., and H. F. Howe. 1983. Competition for dispersal agents among tropical trees: influences of neighbors. Oecologia 59: 185-190. Murcia, C. 1995. Edge effects in fragmented forests: implications for conservation. Trends Ecol. Evol. 10(2): 58-62. Myers, N. 1988. Tropical forests and their species. Going, going...? In E. O. Wilson and F. M. Peter (Eds.). Biodiversity, pp. 28-35. National Academy Press, Washington, D. C.

23

Notman, E., D. L. Gorchov, and F. Cornejo. 1996. Effect of distance, aggregation, and habitat on levels of seed predation for two mammal - dispersed neotropical rain forest tree species. Oecologia 106: 221-227. Offerman, H. L., V. H. Dale, S. M. Pearson, R. O. Bierregaard, and R. V. O'Neill. 1995. Effects of forest fragmentation on neotropical fauna: current research and data availability. Environ. Rev. 3: 191-211. Osunkoya, O. O. 1994. Postdispersal survivorship of north Queensland rainforest seeds and fruits: effects of forest, habitat and species. Aust. J. Ecol. 19: 52-64. Price, M. V., and S. H. Jenkins. 1986. Rodents as seed consumers and dispersers. In D. R. Murrau (Ed.). Seed dispersal, pp. 191-235. Academic Press, Washington, D.C. Ringuet, S. 2000. An assessment of the potential influence of rainforest fragmentation on small terrestrial mammal predation in French Guiana. Rev. Ecol. 55: 101-116. –––––, O. Claessens, J.-F. Cosson, J.-C. De Massary, L. Granjon, and J.-M. Pons. 1998. Fragmentation de l'habitat et diversité des petits vertébrés en forêt tropicale humide: l'exemple du barrage de Petit-Saut (Guyane française). JATBA, Rev. Ethnobiol. 40(12): 11-30. Sabatier, D. 1983. Fructification et dissémination en forêt guyanaise. L'exemple de quelques espèces ligneuses. Ph.D. dissertation. Université des Sciences et Techniques du Languedoc, Montpellier, France. Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991. Biological consequences of ecosystem fragmentation: a review. Conserv. Biol. 5(1): 18-32. Schulz, J. P. 1960. Ecological studies on rain forest in Northern Surinam. Verh. K. Ned. Akad. Wet. Afd. Natuurk, Ser. 2, 53, Amsterdam, Netherlands. Schupp, E. W. 1990. Annual variation in seedfall, postdispersal predation, and recruitment of a neotropical tree. Ecology 71(2): 504-515.

24

–––––. 1992. The Janzen-Connell model for tropical tree diversity: population implications and the importance of spatial scale. Am. Nat. 140: 526-530. Smythe, N. 1989. Seed survival in the palm Astrocaryum standleyanum: Evidence for its dependence upon its seed dispersers. Biotropica 21: 50-56. Sokal, R. R., and F. J. Rohlf. 1981. Biometry. W. H. Freeman and Company, New York, USA. SPSS Inc. 1998. SYSTAT 8.0 Statistics. Chicago, Illinois, USA. Swaine, M. D., and T. C. Whitmore. 1988. On the definition of ecological species group in tropical rain forests. Vegetatio 75: 81-86. Terborgh, J., E. Losos, M. P. Riley, and M. Bolanos Riley. 1993. Predation by vertebrates and invertebrates on the seeds of five canopy tree species of an Amazonian forest. Vegetatio 107/108: 375-386. –––––, L. Lopez, J. Tello, D. Yu, and A. R. Bruni. 1997. Transitory states in relaxing ecosystems of land bridge islands. In W. F. Laurance and R. O. Bierregaard (Eds.). Tropical forest remnants. Ecology, management, and conservation of fragmented communities, pp. 256-274. The University of Chicago Press, Chicago & London. –––––, L. Lopez, P. V. Nunez, M. Rao, G. Shahabuddin, G. Orihuela, M. Riveros, R. Ascanio, G. H. Adler, T. D. Lambert, and L. Balbas. 2001. Ecological meltdown in predator-free forest fragments. Science 294: 1923-1926. Vander Wall, S. B. 1990. Food hoarding in animals. The University of Chicago Press, Chicago, Illinois, USA. Van Roosmalen, M. G. M. 1985. Fruits of the Guyanan flora. Utrecht: Institute of systematic botany, Utrecht University, Netherlands. 483 pp. Williams-Linera, G. 1990. Vegetation structure and environmental conditions of forest edges in Panama. J. Ecol. 78: 356−373.

25

Wilson, E. O. 1988. The current state of biological diversity. In E. O. Wilson and F. M. Peter (Eds.). Biodiversity, pp. 3-18. National Academy Press, Washington, D. C. Zuidema, P. A., J. A. Sayer, and W. Dijkman. 1996. Forest fragmentation and biodiversity: the case for intermediate-sized conservation areas. Environ. Conserv. 23(4): 290-297.

26

TABLE 1. Biological features of the six patches under study: four islands (numbered 20, 14, 2 and 3) and two control patches in continuous forest (CF2 and CF3).

Area (ha) Fruit abundance a Number of fruit species a V. americana tree density b D. leporina abundance (ind. / km) c

a

b

14

2

3

CF2

CF3

7.5

7.9

28

67

–

–

2766

4721

5718

3596

3948

3342

31

49

22

30

33

39

6.4 (1.0)

5.9 (1.4)

4.2 (0.9)

4.6 (1.1)

0

7.4 (1.0)

0.867

0.470

0.876

0.469

0.213

1.140

Total fruit availability at ground level during the whole fruiting period of Vouacapoua americana, estimated for a 200-m long transect per site. Expressed in mean neighborhood size, i.e. mean number (± SE) of conspecific trees (dbh > 25 cm) present in a 50-m radius around studied

trees. c

20

Source data: Chauvet 2001.

27

TABLE 2. Relations between patch characteristics and seed-seedling fate. These analyses have been conducted for all variables from all experiments using Spearman rank-order correlations. Values of coefficients are presented with their significance (ns: P > 0.05; *: P< 0.05). Fruit abundance

Number of fruit species

V. americana density

Dasyprocta abundance

- 0.50 (ns)

0.90 (*)

0.90 (*)

0.30 (ns)

- Consumed seeds

0.50 (ns)

- 0.90 (*)

- 0.90 (*)

- 0.30 (ns)

- Cached seeds

0.71 (ns)

- 0.71 (ns)

- 0.71 (ns)

0.35 (ns)

- 0.42 (ns)

0.60 (ns)

0.94 (*)

0.60 (ns)

- Consumed seeds

0.54 (ns)

- 0.77 (ns)

- 0.89 (*)

- 0.37 (ns)

- Cached seeds

0.35 (ns)

- 0.64 (ns)

- 0.90 (*)

- 0.52 (ns)

Seed fate experiments: 1. Under parent trees (N = 5): - Intact seeds

2. Along transects: 2.1. Analysis including CF2 (N = 6) - Intact seeds

28

2.2. Analysis excluding CF2 (N = 5) - Intact seeds

- 0.50 (ns)

0.90 (*)

0.90 (*)

0.30 (ns)

- Consumed seeds

0.50 (ns)

- 0.90 (*)

- 0.90 (*)

- 0.30 (ns)

- Cached seeds

0.41 (ns)

- 0.97 (*)

- 0.82 (ns)

- 0.15 (ns)

- 0.60 (ns)

- 0.03 (ns)

0.26 (ns)

0.09 (ns)

- 0.77 (ns)

- 0.09 (ns)

0.49 (ns)

0.26 (ns)

Exclosure experiment: 1. Mortality due to terrestrial mammals after 6 months 2. Mortality due to terrestrial mammals after 12 months

29

TABLE 3. Comparison between levels of seed-seedling mortality due to terrestrial mammals after 6 and 12 months of experiment. Wilcoxon signed-rank tests are given for all patches separately. Z

P

Island 20

1.414

0.157

Island 14

1.000

0.317

Island 2

1.342

0.180

Island 3

0.577

0.564

Control CF2 a

–

–

Control CF3

0.707

0.479

a

Wilcoxon signed rank test could not be computed for patch CF2, because estimates of

mortality due to terrestrial mammals were exactly equivalent after 6 and 12 months.

30

Figure legends FIGURE 1. Study forest patches at Saint-Eugène field station, French Guiana.

FIGURE 2. Seed fate of marked seeds of V. americana, when placed 2a) under parent tree crowns in early May (N = 20 seeds x 10 trees x 5 patches), or 2b) along transects in late May (N = 5 seeds x 20 seed sets x 6 patches). Results are reported after three weeks for 2a, and after two weeks for 2b. Patches 20 and 14 are small islands (7.5 and 7.9 ha, respectively), patches 2 and 3 are medium islands (28 and 67 ha, respectively), and patches CF2 and CF3 are located in continuous forests. Experiment showed in 2a was not conducted at patch CF2 because there was not any V. americana trees at this patch.

FIGURE 3. Seed survival until seedlings stage after 3a) six months and 3b) 12 months, when artificially buried and protected with wire exlosures (N = 5 seeds x 10 locations x 6 patches), or unprotected (N = 5 seeds x 10 locations x 6 patches). Patches 20 and 14 are small islands (7.5 and 7.9 ha, respectively), patches 2 and 3 are medium islands (28 and 67 ha, respectively), and patches CF2 and CF3 are located in continuous forests.

31

Intact

Consumed

Cached

100 75

2a

Percentage of seeds

50 25

No data

0

100 75

2b

50 25 0

20

32

14

2

3

CF2

CF3

Protected

Unprotected

100

3a 75

Percentage of seedlings

50 25 0

100

3b 75 50 25 0

20

33

14

2

3

CF2

CF3