Context-dependent forest elephant seed dispersal: Implications for pathways of elephant-driven patterns of biodiversity and carbon storage

Seed dispersal by frugivores facilitates plant reproduction, enhancing germination and seedling survival by reducing risks like diseases and predation under parent trees and potentially directing seeds to favorable sites. Elephants, in particular, are known to strongly impact forests via their seed dispersal patterns because of their ability to disperse many very large seeds across long distances. However, knowledge gaps related to spatio-temporal variability in animal diets, the types and traits of plant species dispersed, and the effectiveness of seed dispersal limit our ability to predict the magnitude of impact that seed dispersers have on plant communities. To better understand how elephant seed dispersal impacts forests, we present a year-long study of fruit availability, forest elephant diet, and seed dispersal across three sites in Gabon showing that: (i) While seeds were prevalent in dung, forest elephants are consuming fewer seeds than in previous studies (ii) There are significant between-site differences in the overall species composition of available fruits and dispersed seeds. (iii) Forest elephants predominately disperse seeds with high carbon storage potential, and the magnitude of carbon storage potential in elephant-dispersed seeds varies between sites. (iv) Forest elephants are effective dispersers of herbs and lianas – growth forms that are often underrepresented in studies of forest elephant seed dispersal – at two out of three of our sites. And (v) seedling emergence and persistence varied significantly across sites. Our results highlight the fundamental role of fruit availability as a driver of forest elephant diet and seed dispersal patterns, and the variability of elephant-dispersed species and functional composition among sites. Additionally, among-site variability in how elephants affect different plant growth types, carbon storage potential, and dispersal effectiveness indicates that predicting the effects of elephants on forest ecosystems depends on site-specific conditions

Methods
METHODS

Data & observational design

Study Areas

We conducted this research in three protected areas in Gabon: Ivindo National Park, Loango National Park, and the Wonga Wongué Presidential Reserve. Ivindo National Park (2990 km2) is covered by primary and secondary lowland humid tropical forests, is home to several major rivers (e.g., Ivindo River and Djidji River), and has many small forest clearings (called bais) throughout the forest (Vande weghe 2009). Loango National Park (1500 km2) has an ocean-to-lowland gradient of coastal savannas to inland forest and features several rivers and a small lagoon. Wonga Wongué Presidential Reserve (4281 km2) is a savanna-forest mosaic ecosystem consisting of 70% secondary forests that are dominated by Aucoumea klaineana and Sacoglottis gabonensis. Overall, the climate in Gabon is bimodal, with a major dry season from June–August and a minor dry season from December–February. The major dry season is cool and cloudy, which is atypical for dry seasons in tropical forests. Annual rainfall at these sites is approximately 1700 mm at Ivindo, 1750-1850 mm at Loango, and 2000-2400 mm at Wonga Wongué.

Data collection

In each park, we conducted focal elephant follows on GPS-collared forest elephants (Fig. 1). We followed 40 elephants, up to three days at a time, for a total of 138, 76, 136 follow days in Ivindo, Loango and Wonga Wongué, respectively. We worked with BaAka forest people to track the forest elephants between the hourly GPS locations ensuring we were following the elephant’s “true” path of movement. To avoid stressing the tracked elephants and to protect the field teams, follows were maintained a minimum of 1 hour behind the focal elephant. Focal follows occurred between February 2017 and April 2018.

To quantify fruit availability, we conducted fruit sampling along transects (fruit walks) at the beginning and end of each day during elephant follows. If the field team moved less than 100m in a day, only one fruit walk was conducted. On each transect, all fallen fleshy fruits encountered on a 2 × 50 m transect, perpendicular to both sides of the elephant’s trail, were recorded and abundance of fruits was noted. Fruits were identified to the lowest taxonomic level possible. Only fallen fruit was recorded as canopy fruit is mostly inaccessible to forest elephants and not representative of what elephants could be eating. If fruits could not be identified in the field, they were photographed or collected as samples, and species identification was verified via published guidebooks and online flora or by consulting a botanist.

We characterized overall elephant diet and seed consumption by destructively searching every third fresh dung during elephant follows to determine the composition of different food sources. Only fresh dung was searched to minimize the possibility that it came from a non-focal elephant group member or that the dung had been disturbed by other animals (e.g., red river hogs, dung beetles). We quantified the relative abundance of fiber, leaf fragments, fruit, and seeds on a 4-point scale. We then identified all seeds greater than 1 cm to the lowest taxonomic level possible and counted their abundances. If seeds could not be identified in the field, they were photographed or collected as samples, and species identification was verified as above.

We characterized community-level seedling germination and survival in elephant dung, we located fresh dung piles accessible from the field stations, independent of focal follows, and monitored dung piles until they disappeared. Upon discovery, half of each dung pile was destructively sampled, and any seeds were identified and counted using the same protocol as during focal follows, described above. The remaining half of the dung pile was left untouched. We revisited the dung piles approximately each week to each month, depending on the site. At each revisit, the abundance of seedlings observed in the dung was recorded. The majority of seedlings were not sufficiently identified (at least to family level), thus taxonomic information on seedlings are not included in this study. 

Statistical analyses

Composition of available fruit and dispersed seeds

​​To assess the variation in composition of fruit observed on transects and seeds observed in elephant dung between sites, we used the function decostand from the vegan package in R to produce Hellinger-transformed species-by-transect or species-by-dung abundance matrices. Then, we created two distance matrices, one for the transects and one for the dung piles, using an abundance-based Euclidean distance metric using the function vegdist from the R package vegan. We tested for between-site differences in the species composition of fruit on transects and seeds in dung, comparing the central location of composition groups for each site using a PERMANOVA test via the adonis2 function and using pairwise PERMANOVA post-hoc tests to identify which sites were different from each other . We created visualizations of the community composition results with non-metric multidimensional scaling (NMDS) plotsusing the ordiplot function from the vegan package. 

Plant growth types and carbon storage potential of dispersed seeds

To characterize the plant growth types of dispersed seeds in dung, we categorized dispersed seeds from our species list as herbs, lianas, or trees using information from published guidebooks and online herbaria. We used this information to explore between-site differences in plant growth types of seeds dispersed in elephant dung, by calculating the proportion and count of seeds from tree, liana, and herbaceous species in each dung pile.

To estimate the carbon storage potential of elephant-dispersed seeds, we compiled information about taxon-specific wood density values for our species list of available fruits and dispersed seeds, using the Global Wood Database. This database contains published wood density information (mass per unit volume) of adult trees. For our dataset, wood density was obtained at the lowest specific taxonomic level possible (family = 5, genus = 15, species = 26) because wood density tends to show a phylogenetic signal. To explore between-site differences in carbon storage potential, we used these taxon-specific wood density values to calculate the community-weighted mean carbon storage potential of available fruits per transect and dispersed seeds per dung pile.

We tested for differences in plant growth types and carbon storage potential between sites using Kruskal–Wallis H tests and post-hoc Dunn’s tests. Differences between sites were considered significant for values of p < 0.05.

Emergence and persistence of seedlings in dung

To explore between-site differences in the germination success of seeds in elephant dung, we calculated the number of seeds and seedlings in each elephant dung pile. We then tested for differences between sites using Kruskal–Wallis H tests and post-hoc Dunn’s tests. Differences between treatments were considered significant for values of p < 0.05. To explore between-site differences in the temporal patterns of seedling germination and survival, we then calculated the number of seedlings in dung on the first monitoring date, the last monitoring date, and the monitoring date with the highest number of seedlings. We identified the average temporal trajectory of seedling flush and persistence for each site, and identified the peak day of seedling flush post-dispersal for each site.