Search

International audience

The interaction between granivorous scatterhoarding mammals and plants is a conditional mutualism: scatterhoarders consume seeds (acting as predators), but the movement of seed by scatterhoarders may contribute to dispersal (acting as mutualists). Understanding the ecological factors that shape this relationship is highly relevant in anthropogenically disturbed tropical forests where large-bodied frugivores are extirpated. In such forests, large-seeded trees that once depended on these frugivores for dispersal may now only have scatterhoarders as prospective dispersers. We studied Carapa oreophila (Meliaceae) in an Afromontane forest, to test the hypotheses that the proportion of seeds immediately consumed or hoarded (dispersed) would vary over a disturbance gradient. Temporal replication also afforded exploration of how habitat effects might vary with food availability. Using a Bayesian framework, we demonstrate that seeds were more likely to be hoarded in less disturbed forest, irrespective of temporal variation in food abundance. In contrast, forest disturbance only appeared to increase seed predation in temporal replicates that coincided with sustained food availability. These results highlight the potential variability in the dynamics between plants and scatterhoarders over fine temporal scales, elucidating possible ecological scenarios where scatterhoarders might act as mutualists (contributing positively to plant recruitment). Our study also fills important knowledge gaps about the importance of scatterhoarders as dispersers in tropical forests depleted of large-bodied frugivores, particularly in Africa where scatterhoarding mutualisms have not been extensively studied.

Made available in DSpace on 2018-12-11T16:40:04Z (GMT). No. of bitstreams: 0 Previous issue date: 2015-11-01 Species showing mast seeding synchronously produce large amounts of fruits during some scattered years. This massive crop has been hypothesized to improve dispersal effectiveness by a satiation of seed predators, but the consequences for seed dispersers have barely been studied in the tropics. We tested the hypothesis that masting resulted in satiation of frugivorous dispersers using the study case of two Manilkara species growing in an Amazonian forest in French Guiana. Seed dispersal was estimated by means of seed traps in two forest types during a 10-y monitoring. Manilkara huberi and M. bidentata showed three fruiting events in a time span of 10 y (in 2001, 2006 and 2010). Estimates of seed dispersal from 2001 and 2010 showed that satiation of frugivores only occurred in the year with the largest crop of Manilkara (2010) and in the habitat where the diversity of primate-dispersed species retrieved in seed traps was the highest (Grand Plateau, with clay soils), while fruit consumers did not seem to be satiated in other instances. Spatio-temporal variability of seed production and the community-crop context are therefore affecting satiation of frugivores during masting events. CNRS-MNHN Museúm National d'Histoire Naturelle Département d'Écologie et Gestion de la Biodiversité, 1 Av. du Petit Château Departamento de Botânica Plant Phenology and Seed Dispersal Research Group Universidade Estadual Paulista (UNESP), Avenida 24-A CNRS Guyane USR 3456, 2 rue Gustave Charlery Departamento de Botânica Plant Phenology and Seed Dispersal Research Group Universidade Estadual Paulista (UNESP), Avenida 24-A

Plants often suffer heavy mortality in their seed and seedling stages. While some vertebrates play a significant role in seed dispersal, others are predators of seeds and seedlings. In many large-seeded, nutbearing trees, rodents are responsible both for effective seed dispersal by their vigorous activity of seed hoarding (Vander Wall, 2001), and for early-stage mortality through seed and seedling predation (Crawley, 2000; Kitajima and Fenner, 2000; Hulme and Benkman, 2002). Rodents are pivotal in the regeneration and demography of large-seeded trees. Conversely, large-seeded trees can affect rodent populations. Large seeds are a high quality food resource (Vander Wall, 2001), but they are not reliable foods for rodents because the trees often exhibit large interannual variations in seed production (Kelly, 1994; Vander Wall, 2001). The large annual variation in seed crops (e.g. masting followed by low seed crops in subsequent years) has a major influence on the population size of predators, such that the population increases after masting and decreases during low seed-crop years (Wolff, 1996b; Wright et al., 1999; Ostfeld and Keesing, 2000). Variation in rodent numbers may in turn affect seed predation and dispersal in some trees (Vander Wall, 2002; Zhang et al., Chapter 16, this volume). Ingestion of large seeds may have negative effects on rodent physiology; many large seeds, especially Quercus acorns and Aesculus seeds, contain secondary compounds which operate as a chemical defence against seed predators (e.g. Vander Wall, 2001; Shimada and Saitoh, 2003). The interaction between large seeds and rodent populations has been intensively studied (e.g. Crawley, 2000; Vander Wall, 2001; Hulme, 2002; Hulme and Kollmann, Chapter 2, this volume, and references therein). The ecological significance of masting is supposedly that synchronous seed production satiates seed predators andso produces a disproportionately large cohort of seedlings following masting (Janzen, 1971; Kelly, 1994). However, because rodents are polyphagous and generalist seed predators (Hulme and

Seed removal by rodents was investigated for Manilkara huberi and Chrysophyllum lucentifolium in a French Guianan forest. According to the escape hypothesis, seed survival was expected to be greater in populations of low conspecific adult density, while on the contrary, under the satiation hypothesis, it was expected to be greater in populations of high density. The two plots under study showed opposite densities for the two studied tree species. Therefore, according to both hypotheses, seed survival at each plot was expected to be opposite between species. To assess seed fate, seeds were thread-marked in order to relocate them after removal and to determine whether they were consumed or scatterhoarded by rodents. Contrary to what was expected, our results showed that both M. huberi and C. lucentifoliumhad better survival in the same plot. This suggests that seed fate for both study species was not influenced by the density of conspecific adult trees, but was rather affected by other habitat characteristics, likely the global resource abundance. Variation in seed predation rates of both species seemed largely related to their respective fruiting period, while scatterhoarding rate seemed more affected by intrinsic seed characteristics.

Fruiting phenology, habitat types and proximity to conspecific fruiting adults may influence the degree of escape from pre- and post-dispersal seed predators. Successful predator escape by means of satiation is considered to be especially important for selection of masting and monocarpic reproduction in trees, such as exhibited by Tachigali versicolor, a tropical canopy tree. How pre- and post-dispersal predation rates varied with dispersal timing and among trees in T. versicolor was examined in young and old forests on Barro Colorado Island in Panama during a 4-mo period. Seeds were collected from above-ground traps to assess predispersal predation by bruchid beetles, and from quadrats on the ground to record predation and removal by terrestrial mammals. Proportion of seeds aborted varied greatly among trees (range 6-30 %, mean 16 %), and was especially high for trees on the edge of the island with fruiting conspecifics nearby during the early part of fruiting season. The proportion of seeds killed by bruchid beetles varied less among trees (14-25 %, mean 20 %), and remained constant throughout the fruiting season. Seeds on the ground were attacked mostly by rodents, and possibly by deer (26 % of all seeds and 43 % of intact dispersed seeds). The post- dispersal predation level was higher in the young forest than in the old forest (61 and 26 % of intact dispersed seeds, respectively), and was unaffected by the prox- imity of fruiting conspecifics. Temporal satiation of seed predators was evident only for post-dispersal mammalian predators in the old forest.

1 The establishment probabilities, short-term survival, growth rates and biomass allocation patterns of seedlings in two large-seeded tree species, Carapa procera (Meliaceae) and Vouacapoua americana (Caesalpiniaceae), were studied in contrasting microhabitats at Paracou, French Guiana. 2 Seeds (100 per habitat per species) were buried to simulate rodent caches into gap and understorey microhabitats (n = 20 of each) surrounding mature parent trees. The experiment was replicated in Carapa procera (in early March and May) to investigate the effect of temporal variation in seed dispersal on the establishment process. 3 Surviving seeds and sprouting seedlings were repeatedly censused during 2 years after establishment, and the causes of mortality determined. Growth and biomass parameters were calculated for surviving seedlings after 2.5 years. 4 A greater proportion of seeds and germinating seedlings were dug up but less seedlings suffered from damping-off in C. procera than in V. americana. For both species, short-term survival of seedlings was lower in the understorey where predation by mammals and damping-off occured more frequently than in gaps. 5 Damping off killed more Carapa procera seedlings when buried in March than in May, whereas mammal predation caused more deaths in May planted seeds. 6 Although survival and growth were enhanced in gaps for both species, seeds and seedlings were much more intolerant to the understorey microhabitat in Carapa procera (almost 100% mortality in shaded habitat) than in Vouacapoua americana. 7 Seedling biomass allocation was consistent with previous data for large-seeded species but differed markedly between the two study species.

The fate of the large (mean 12.6 g) seeds of Gustavia superba, an understory tree, was analyzed on Barro Colorado Island (BCI), Panama. Eight hundred thread-marked seeds were placed on the ground at two contrasting 1-ha forest sites (Gustavia-rich vs Gustavia-poor) during two periods (June and July) of the fruiting season of G. superba. These months correspond to maximum food availability on BCI. On average, 85.5 percent of the seeds were removed within 28 days, 47.5 percent and 3.8 percent of them being found scatterhoarded (buried) by agouti (Dasyprocta punctata) and gnawed by rodents, respectively, within 10 m of their origins. The effect of site and the interaction of site with month significantly affected seed removal rates, but not scatterhoarding rates. Proportions of seeds removed were greater where G. superba trees were rare. Because freshly fallen seeds were not infested by bruchid weevils, unburied and germinating seeds provide an abundant short-term food supply for terrestrial mammals. Burying seeds allowed agoutis to later consume cotyledons of germinating seeds from mid-August throughout October, when food is scarce on BCI. Overall seed dispersal effectiveness (% seed dispersal multiplied by % seedling survival) contrasts dramatically between forest sites, being 10.1 percent and 0.75 percent at Gustavia-rich and Gustavia-poor areas, respectively. This result suggests that predator-disperser satiation occurred, maybe due to greater G. superba seed availability and other alternative food supply, allowing greater G. superba seedling survival at the Gustavia-rich area.

In tropical rain forests of Central America, the canopy tree Dipteryx panamensis (Papilionaceae) fruits when overall fruit biomass is low for mammals. Flying and arboreal consumers feed on D. panamensis and drop seeds under the parent or disperse them farther away. Seeds on the ground attract many vertebrate seed-eaters, some of them potential secondary seed dispersers. The fate of seeds artificially distributed to simulate bat dispersal was studied in relation to fruitfall periodicity and the visiting frequency of diurnal rodents at Barro Colorado Island (BCI), Panama. The frequency of visits by agoutis is very high at the beginning of fruitfall, but in the area close (50 m) to fruiting trees (Dipteryx-rich area) it declines throughout fruiting, whereas it remains unchanged farther (50 m) away (Dipteryx-poor and Gustavia-rich area). Squirrels were usually observed in the Dipteryx-rich area. Along with intense post-dispersal seed predation by rodents in the Dipteryx-rich area, a significant proportion of seeds were cached by rodents in the Dipteryx-poor area. Post-dispersal seed predation rate was inversely related to hoarding rate. A significantly greater proportion of seeds was cached in March, especially more than 100 m from the nearest fruiting tree. This correlates with the mid-fruiting period, i.e. during the height of D. panamensis fruiting, when rodents seem to be temporarily satiated with the food supply at parent trees. Hoarding remained high toward April, i.e. late in the fruiting season of D. panamensis. Low survival of scatterhoarded seeds suggests that the alternative food supply over the animal's home-ranges in May-June 1990 was too low to promote survival of cached seeds. Seedlings are assumed to establish in the less-used area of the rodents' home-range when overall food supply is sufficient to satiate post-dispersal predators.

The data presented show thatVirola nobilis (Myristicaceae), a bird/mammal-dispersed tree species in Panama, may also be dispersed by a terrestrial rodent, the agouti (Dasyprocta punctata). Using a thread-marking method, we observed that agoutis scatterhoardedV. nobilis seeds that they found both singly or in clumps. Seed removal and seed burial rates were strongly affected by features of forest habitats, such asV. nobilis tree richness (rich vs poor) and/or forest age (old vs young), but not by seed dispersal treatment (scattered vs clumped). Predation (mostly post-dispersal) of unburied seeds by weevils was independent of habitat and dispersal treatment. Seeds artificially buried in aVirola-rich area were more likely to escape predation and become established than unburied seeds under natural conditions. The food reward for agoutis is in the germinating seedlings. The seed dispersal syndrome ofV. nobilis involves long- and short-distance dispersers which both appear important for tree recruitment.

A possible mutualistic dispersal system between a large-seeded tree of French Guiana, Vouacapoua americana (Caesalpiniaceae), and caviomorph rodents, Myoprocta exilis and Dasyprocta leporina, is described. Mast fruiting of Vouacapoua at the beginning of the wet season coincides with scatter-hoarding seed dispersal. During the wet season, almost 100% of marked seeds on three sites were removed: nearly 70% were buried and the rest were eaten by mammals. Unburied seeds were attacked by insects and/or lost their ability to germinate. Rodents preferred ungerminated seeds, and had no interest in germinated seeds. Seeds were buried individually near natural objects such as palms, branches, logs, lianas, roots and trees. After predation by rodents, seedling distribution did not differ from seed distribution. Most seeds were transported less than 5 m from the feeding plots but some were carried as far as 22.4 m. Between 40 and 85% of dispersed seeds were retrieved during the following month by rodents and eaten. The disinterest of caviomorph rodents in germinated seeds, because of rapid exhaustion of endosperm reserves, prevents feeding from hoarded Vouacapoua during the long dry season when resources are scarce. Seedlings emerging from forgotten or abandoned cached seeds appear to increase the recruitment of Vouacapoua americana.

The post-dispersal fate of Chrysophyllum lucentifolium (a canopy tree; Sapotaceae) seeds was analyzed in French Guiana over three consecutive years. Experiments using 750 thread-marked seeds were performed to investigate seed removal, predation, and caching by terrestrial vertebrates on howler monkey (Alouatta seniculus) defecation sites, where clumps of intact C. lucentifolium seeds were observed. Year-to-year variations in seed fate during the peak fruiting period were considered in relation to overall fruit and seed resource availability estimated by a raked-trail survey. The effect of two forest areas, which differed in soil and floristic composition, was examined with conspecific fruiting tree density as a covariant. Exclosure versus open treatment was used to discriminate small rodents (not larger than a spiny rat) from other vertebrates. The presence of fresh howler dung did not affect seed fate after 20 days as shown by comparisons between defecation sites and control during the first year. There was a significant effect of year on the percentage of seeds remaining after 20 days. Low seed removal in 1995 and 1996 (compared to 1997) corresponded to higher overall fruiting and higher fruiting of C. lucentifolium, or the presence of alternative resources for rodents. An effect of forest area was observed on the seed removal rate, which varied with years and protection. Comparatively, an effect of forest area on the percentage of seeds lost was observed in 1996 and an effect of treatment on the percentage of seeds eaten was seen in 1995. The mode of seed caching suggested that spiny rats were the main seed remover. Results of this study suggest that greater seedling recruitment may occur when large fruit crop and high howler dispersal co-occur with a lower impact of rodents (i.e., when rodents are saturated by abundant and diversified fruit resources such as in 1995). Such event synchrony, however, is highly unpredictable after only three years of study.

Predation by rodents on seeds of Scheelea zonensis was studied on Barro Colorado Island, Panama, during the latter half of this palm's fruiting season. The number of Scheelea bearing fruit declined steadily during this period. The proportion of both intact and bruchid-infested seeds among those accumulated under fruiting Scheelea declined, while the proportion of gnawed seeds increased. By early October, 87 percent of the Scheelea seeds sampled had been gnawed by rodents, and few intact or bruchid-infested seeds remained below parent palms. From August through October, agoutis, Dasyprocta punctata, visited Scheelea less and less frequently, while visits by squirrels, Sciurus granatensis, did not dedine. To learn the fate of late-falling Scheelea seeds, we placed small piles of thread-marked seeds, each similar to a small Scheelea fruit crop, in the forest during three successive months, a total of 400 seed each month. Of these 1200 thread-marked seeds, 57.2 percent were retrieved within 20 m of their piles 7 days after placement. Of the retrieved seeds, 22.2 percent were gnawed within 1 m of their pile, 70.6 percent were carried between 1 and 20 m before gnawing, and 7.2 percent were cached (not gnawed, but buried in the ground or covered with litter). More seeds were gnawed, and fewer cached, in September and October than in August. On Barro Colorado Island, intense seed predation by rodents on Scheelea correlates with a previously documented forestwide low in availability of fruits of all kinds, and with a decline in egg laying by bruchids.

Long-term studies were conducted in French Guiana on the recruitment pattern of a large-seeded (47.6 g fresh weight) tree species: Eperua grandiflora (Caesalpiniaceae). Fruiting, seed dispersal by gravity, seed predation and seedling damages, and early seedling survival were analyzed from 1987 through 1988 (20-month period) at one tree standing on a slope. Most seeds land within 10-20 m of the parent, and a small fraction (5%) appeared to be secondarily dispersed by scatterhoarding rodents. As predicted by the Janzen-Connell model, clumped seeds and seedlings below the parent suffered post-dispersal predation and damages, respectively, by several families of insects, ants, and mammals. But because of large seed reserves, young seedlings were able to renew leaves or terminal meristems after defoliation or stem damage. Whether early seedling survival (7 months) was distance-dependent, mortality was nowhere near complete, and seedling survival during the next 12-month period (19-months post-germination) was independent of distance. Thus, juveniles persisted in a dumped population under parent tree. Dynamics of juveniles were studied from 1984 through 1989 and two 2500 m2 plots; an understory-environment (mature parent) and a gap-environment (dead parent). The proportion of juveniles greater than 100 cm tall was higher in the gap plot than in the understory plot. Although juveniles were very tolerant of understory conditions, they grew more rapidly in the gap. Occurence of young trees in the closed vicinity of parent trees is likely to result from poor dispersal of seeds, high survival beneath parents and high understory-tolerance of juveniles.