Seed fate depends on a complex series of processes, including several phases of dispersal and predation. Primary dispersers remove fruits from plants and usually regurgitate or defecate seeds elsewhere. Secondary dispersers rearrange the resulting seed shadow and sometimes place seeds in beneficial microsites (Janzen1982, Roberts and Heithaus 1986, Forget 1990, 1991b, Levey and Byrne 1993). To complicate seed fate further, secondary dispersal agents like ants and rodents are generally seed predators as well as dispersers. Thus, seed removal of dispersed seeds is not necessarily equivalent to seed predation (O'Dowd and Hay 1980, Horvitz 1981, Janzen 1982, Roberts and Heithaus 1986).
Generalizations about post-dispersal removal and predation in tropical forests are based primarily on studies of large-seeded plants, particularly canopy trees (e.g., DeSteven and Putz 1984, Sork 1987, Smythe 1989, but see Alvarez-Buylla and Martínez-Ramos 1992). Yet,removal agents and subsequent seed fates probably differ for small seeds and large seeds. Large seeds are most likely consumed by rodents, other mammals, and insect larvae, while small seeds are most likely consumed by adult insects such as ants (Hölldobler and Wilson 1990) or, after burial, are infected by fungi (Crist and Friese 1993). For seeds of intermediate size, removal agents may be especially diverse because of overlap between the two major groups of seed harvesters, ants and rodents. In Central American rain forests, seeds ranging from approximately 1.5 mm (easily accessible to ants but still taken by rodents; Perry and Fleming 1980) to approximately 6.7 mm (easily accessible to rodents but still taken by ants; M. Jones, unpublished data) are potentially utilized by both major harvester groups. For seeds in this size range, interactions between seed-harvesting rodents and ants are probably important. For example, temperate zone ants remove and sometimes disperse small seeds before seed-consuming rodents find the seeds, which results in many more seeds surviving than without the ants (O'Dowd and Hay 1980, Heithaus 1981, Pierce and Cowling 1991, Gibson 1993b). Thus, ants and rodents sometimes remove the same seed species. Removal by the different harvester groups can result in different seed fates that have a large effect on plant population dynamics (Heithaus 1981, 1986, Brown and Heske 1990).
In addition to ant-rodent interactions, inter-site variation complicates interpretation of seed removal experiments. At the microhabitat level (e.g., litter vs. no litter), site has been shown to affect seed removal rates (Reader 1991, Whelan et al.1991). Likewise, macrohabitat differences (e.g., field vs. forest) create heterogeneity in removal rates for seeds (Schupp 1988, Schupp and Frost 1989, Willson and Whelan 1990, Whelan et al.1991). Finally, even when macrohabitat is held constant, removal rates can vary. For example, DeSteven and Putz (1984) and Sork(1987) documented removal differences that may reflect differences in the abundance of seed harvesting rodents between two sites within the same rainforest macrohabitat.
I had two objectives:(1) to look for interactions between rodents and ants in the magnitude and rate of secondary removal for two tropical plants with intermediate-sized seeds, and (2) to determine if secondary removal is consistent within and across sites.
I conducted this study on Barro Colorado Island (BCI) and on the Gigante Peninsula (Gigante), both part of the Barro Colorado Nature Monument and classified as Tropical Moist Forest under the Holdridge life-zone system (Croat 1978). One study area on BCI and one on Gigante were chosen (1)near streambeds because these areas had a high abundance of focal plant species and a high diversity of ants (Levings and Franks1982, Levings 1983) and (2) far from the main laboratory buildings (> 1 km) to minimize effect of habituation to scientists as a factor modifying rodent behavior (S. J. Wright, pers. comm.).
Ichose two focal plant species, based on seed size and availability: Virola nobilis (Myristicaceae) and Psychotria marginata (Rubiaceae). Fruiting V. nobilis and P. marginata were present at both sites throughout the course of the experiments. Their fruits are consumed by mammals and birds that either disperse seeds by defecation or regurgitation, or drop them under the parent plant (Howe and Smallwood 1982, Worthington 1982, Howe et al.1985, Levey 1987). Virola nobilis and P. marginata seeds vary in size by more than an order of magnitude (Table 3-1), but both fall in the range of seed sizes potentially consumed by rodents and large ants (see Introduction). Virola nobilis, however, is probably at or near the upper size limit for ants, especially if the seeds are cleaned of aril, which some ants use to drag V. nobilis seeds (M. Jones, pers. obs.).
| Virola nobilis1 | Psychotria marginata | |||
|---|---|---|---|---|
| Length(mm) | 17.9 | (0.13) | 3.0 | (0.03) |
| Mass (g) | 2.00 | (0.03) | 0.0098 | (0.0004) |
| Volume(mm3)2 | 1,962 | (41) | 3.93 | (0.15) |
| 1 N = 50 seeds for each species, from multiple fruiting adults | ||||
| 2 Seed volumes were estimated by assuming that V. nobilis seeds are ellipsoidal and that P. marginata seeds are half the volume of an ellipsoid. | ||||
In removal experiments I attempted to simulate a clump of seeds as they would have occurred after primary dispersal by a mammal or bird. To simplify experimental design, I placed both P. marginata and V. nobilis seeds in clumps, as would be typical of mammalian dispersal but not necessarily bird dispersal (Howe1989).
Several rodent species are potential seed removers at our sites: Dasyprocta punctata, Agouti paca, Sciurus granatensis, Orozomys spp., Heteromys desmarestianus and Proechimys semispinosus. There is debate whether the abundance of rodents on BCI is similar to mainland sites in Panama and elsewhere (S. J. Wright, pers. comm., Glanz 1990). Glanz(1990) showed that D. punctata and S. granatensis densities are higher on BCI than on the Gigante peninsula and several other Neotropical sites. He also commented, however, that other rodents, including Orozomys spp. and P. semispinosus, are comparable or lower in density on BCI than at these other sites. Thus, rodent assemblages probably differ between our sites; in particular, Gigante seems to have a higher abundance of small rodents and BCI a higher abundance of large rodents (Glanz 1982, 1990, Glanz et al. 1982).
On BCI, litter ants are extremely diverse and abundant (> 127 species and 49 genera; Levings 1983). Ectatomma ruidum, a common litter ant at both sites in this study, was found in 10% of Berlese samples and 94% of baittrap samples across six to eight sites on BCI (Levings 1983). Although common ants on BCI are also common on Gigante (M. Jones,pers. obs.), they have not been studied in detail on Gigante.
For each plant species, I tested three factors in removal experiments: site (BCI or Gigante), rodent access (access or exclosure), and ant access (access or exclosure). These factors were crossed in a full factorial design to give a total of eight treatments for each species, each replicated 10 times. For clarity I will refer to the treatments in terms of potential access by seed removers: (1) all seed removers excluded (hereafter, no access), (2) ants and other crawling insects excluded (rodent access), (3) rodents and other vertebrates excluded (ant access), and (4) no seed removers excluded (total access). I excluded vertebrates from the seeds by securing 10 x 10 x 10 cm hardware cloth (1.27 cm mesh) cages over the seeds. Ants and other crawling insects were excluded by a 10 x 10 x 10 cm sheet metal wall that surrounded the seeds, but was open at the top. The wall was partially buried in the soil and covered with Tanglefoot® (The Tanglefoot Company, Grand Rapids, Michigan, USA). When no exclosure was used (total access), a partially buried 10 x 10 x 1 cm high sheet metal frame delineated the boundaries of the seed area. All exclosures were covered by a 0.5-m diameter circular aluminum roof, 0.5 m off the ground to prevent rain from splashing the seeds out of the exclosure. Although the rodent exclosure also excluded other vertebrates, and the ant exclosure also excluded other crawling insects, rodents and ants were the most likely seed harvesters at these sites.
I placed the exclosures at stations distributed in a stratified random pattern along a grid that was 45 paces (Å70 m) long and 15 paces (Å23 m) wide, with 5 paces (Å7.5 m) between each longitudinal axis. Four stations, one for each exclosure treatment, were randomly placed on each of the 10 longitudinal axes of the grid. Thus, there were forty stations on the grid for each site. With this design I could assess both small scale (within-site) and large scale (between-site) variation in removal by ants. Because all stations within each site could have been within the home range of one or a few individual rodents, our ability to generalize from between-site variation in rodent removal was reduced.
I collected mature fruit from 6 adult V. nobilis and greater than 20 adult P. marginata plants, removed the pulp from the seeds, and placed five seeds at each station. I used equal numbers of seeds (n = 5) for the different plant species to keep seed density constant across species. Note that this resulted in larger seed biomass in the V. nobilis experiment than in the P. marginata experiment. I then monitored seed removal for 4 d for P. marginata and 17 d for V. nobilis, at which times removal rates had declined to almost zero.
Results were analyzed by repeated measures ANOVA to compensate for correlations among the repeated observations within each station over time. The dependent variable was the arcsin square root transformation of the cumulative proportion of seeds within each station removed each day. The independent variables were site, ant access, and rodent access.
Some interaction terms in the analyses were significant but not biologically meaningful. In particular, significant interaction terms always occurred when ant access or rodent access was crossed with another significant factor, like time or site. In these cases, significant interactions were artifacts because time or site could only have effects when removers had access to seeds, not when they were excluded. I discuss only biologically meaningful interactions.
Rodent access treatments had a significantly greater proportion of V. nobilis seeds removed than treatments without rodent access (Figure 3-1; F1,73 = 40.5, p = .0001). Comparing the sites, V. nobilis seeds were removed significantly more quickly and more completely from the site on Gigante than the site on BCI (F1,73 = 11.6, p = .0011). At the end of the experiment in the rodent-access treatment, 80% of the seeds were removed from the Gigante site versus 2% from the BCI site.
Figure 3-1. Proportion (mean ± S.E.) of Virola nobilis seeds removed from stations originally containing five seeds as a function of time at each of two sites, BCI and Gigante. Symbols represent exclosure treatments that restrict access to seeds: no access (filled squares), rodent access only (open squares), ant access only (filled circles), both ant and rodent access (open circles).
In the total-access treatment, 96% of the seeds were removed from Gigante versus 38% from BCI. On Gigante, the total-access treatment and the rodent-access treatment were statistically indistinguishable, whereas on BCI the total-access treatment had significantly more removal (38%) than the rodent-access treatment (2%). Seed removal also varied significantly over time (F3,219 = 24.1, p = .0001). Finally, a significant Time-Site interaction indicated that the site factor was only important late in the experiment (F3,219 = 11.1, p = .0001); removal magnitudes were the same on BCI and Gigante through the second day, but then increased substantially on Gigante relative to BCI.
Removal of P. marginata seeds was statistically indistinguishable from the BCI site and the Gigante site throughout the experiment (Figure 3-2; F1,76 = 0.0, p = .9), so I pooled the data across sites for subsequent analyses. In addition, variation in removal rates for ant access treatments between the two sites (C.V = 0.01) was much smaller than within site variation (C.V. = 0.56 for BCI, C.V. = 0.61 for Gigante), indicating that ant removal was patchy on a small but not large scale.
Treatments with ant access had significantly higher removal than treatments without ant access (Figure 3-2; F1,77 = 59.9, p = .0001). Total-access and ant-access treatments did not differ significantly (Figure 3-2). I observed
Figure 3-2. Proportion (mean ± S.E.) of Psychotria marginata seeds removed from stations on BCI and Gigante as a function of time. Symbols as in Figure 3-1.
Ectatomma ruidum palpating and then removing seeds from the ant-access treatments on the first day of the experiment at both sites. Ants typically recruited to seeds within 15 min of seed placement, and seed removal was rapid during the first several hours of the experiment. The removal rate then declined significantly (F3,231 = 12.4, p = .0001); approximately 40% of the seeds were removed in the first 24 h of the experiment, yet only 10% more of the original total were removed during the remaining 3 d. Most seeds were not removed; by the end of the experiment, approximately 45% and 38% of the seeds in the total-access and ant-access treatments were removed, respectively. No removal occurred in the rodent-access treatment. After approximately two weeks, some seeds that had not been removed by ants began to germinate, although I did not quantify seed viability.
Seed removal by both ants and rodents was high but variable for the two plant species. Our experiments demonstrated that seed removal patterns by rodents and ants differ in timing and magnitude, that site differences exist in removal patterns for rodents, and that ants and rodents do not necessarily overlap in their removal of intermediate-sized seeds.
Virola nobilis seeds were primarily removed from rodent-access treatments, whereas the smaller-seeded P. marginata were primarily removed from ant-access treatments. Therefore, V. nobilis seeds were probably too large for ants to remove, even though some ants move these seeds when the aril is intact (M. Jones, pers. obs.). P. marginata seeds were probably too small or perhaps toxic for rodents at these sites, even though much smaller seeds are readily consumed by some tropical rodents (Perry and Fleming 1980). This difference in removal pattern for the two plant species is probably important since seed removal by ants and rodents may lead to different seed fates and therefore different selective pressures on plants.
Seeds removed by Neotropical rodents are likely to be scatterhoarded for later consumption (Glanz et al. 1982, Forget and Milleron 1991, Forget 1992, Galettiet al. 1992). Some of these cached seeds often escape rediscovery by the rodents and germinate (Smythe 1989, Forget 1990, 1991a). They may also escape infestation by insects, a common fate for large tropical seeds (Smythe 1989, Forget and Milleron 1991). Most seeds cached by rodents, however, are rediscovered and consumed (Perry and Fleming 1980, Bond and Breytenbach 1985, Forget 1990, 1991b, 1993). As a result, secondary removal by rodents is usually assumed detrimental for plant fitness.
Seeds removed by ants are probably brought back to nests or dropped along the way to a nest. Ant removal of small, nonmyrmechorous seeds is little studied in tropical forests. For example, of the 127 species of litter ants found on BCI (Levings 1983), only a small fraction have been studied in detail with regards to foraging ecology (Kaspari in press, Tobin in press). Still, recent evidence indicates that secondary removal by ants may benefit adult plant fitness. Levey and Byrne (1993) found that many species of ants consume seeds in a Costa Rican rain forest, contrary to the common belief that most ants in the tropics are primarily carnivores or scavengers (see discussions in Carroll and Janzen 1973, Tobin in press). They demonstrated that a small but consistent proportion of Miconia seeds escaped consumption by Pheidole ants after the seeds were harvested, and that seedlings grown on ant-refuse soils had higher survival and growth than seedlings grown on randomly chosen forest soil (but see Horvitz and Schemske 1986a). Other investigators have also shown germination and survival advantages of secondary removal by ants (Andersen 1988, Kjellsson1991, Gibson 1993a). Finally, plants near ant nests may have higher reproductive output than other plants, and seed shadows rearranged by ants can experience lower seed predation by rodents (Heithaus 1981, Rissing 1986, Nowack et al. 1990, Hughes and Westoby 1992).
In terms of the magnitude of removal, rodents on BCI removed roughly the same proportion of seeds as did ants, but on Gigante rodents removed a much higher proportion of seeds. In both sites, rodent removal rates of V. nobilis seeds were much lower than initial ant removal rates of P. marginata seeds, although ants stopped harvesting seeds after approximately 1 d. Thus, secondary removal by rodents appeared to be more complete but considerably slower than removal by ants.
Seed removal by rodents varied significantly between the two sites, with almost complete V. nobilis removal from the Gigante site but only 38% removal from the BCI site. Because this site difference was not apparent for P. marginata, which was primarily removed by ants, it is probably a result of differences in the rodent communities at the two sites. In support of this interpretation, note that removal of V. nobilis was high in both the total-access and rodent-access treatments on Gigante and totally absent from the rodent-access treatment on BCI. This suggests that the rodent species removing seeds at the two sites were indeed different, either in behavior or density.
Although I cannot generalize our results to include all of Gigante and BCI I note that they disagree with previous comparisons of seed removal from the two sites. DeSteven and Putz (1984) and Sork (1987) found that removal was much higher on BCI than on Gigante. They proposed that this difference was due to higher small mammal densities on BCI, which was purported to be a result of fewer large carnivores there (Terborgh 1988). Agoutis (Dasyprocta punctata), in particular, were thought to be at higher densities on BCI and principally responsible for higher seed removal there. Because V. nobilis is substantially smaller than either Gustavia superba or Dipteryx panamensis and is probably taken by different rodents, its higher removal on Gigante may have been due to site differences in abundance of rodent species other than D. punctata.
Many investigators have assumed that tropical ants are less granivorous than temperate ants (e.g.,Perry and Fleming 1980), largely arising from the perception of tropical ants as scavengers rather than primary consumers (see Tobin in press). I found that ant removal of P. marginata seeds was consistently high across sites. Furthermore, removal rates I observed resembled results from other studies that have examined seed removal by ants from temperate sites (Reichman 1979, Heithaus 1981, Abramsky 1983) and tropical sites (Horvitz 1981, Roberts and Heithaus 1986, Kaspari in press). In particular, removal was very high during the first day of the experiment, but then rapidly declined. This decline could be due to density-dependent foraging, dissipation of chemical cues on seeds as they dried, selection for only viable seeds, or satiation of local ant nests. Because viable seeds remained at the end of the experiment, it is unlikely that the initially high removal rates were due to ants removing only the viable seeds. In addition, ant nests were probably not satiated with seeds because ant nest densities at these sites are high (Levings 1983). That removal was consistent and ant recruitment rapid confirms the idea that ants may be important secondary seed removers in tropical forests (Byrne and Levey 1993, Kaspari in press).
Because both rodents and ants act as secondary seed removers in tropical forests, one must be careful to distinguish between them when designing studies that examine secondary dispersal and postdispersal seed predation. In this study, removal by rodents and ants did not overlap. Note, however, that I selected seeds on the ends of the size range potentially taken by both rodents and ants. With other intermediate-sized seeds both ants and rodents are likely removers. Because the relative contributions to dispersal and predation by ants likely differ from rodents, it is important to identify seed harvesters when intermediate-sized seeds are studied.
[Abstract] [Chapter 1] [Chapter 2] [References] [My Home Page]