Skip to main content
Dryad logo

Assessing the effects of elephant foraging on the structure and diversity of an Afrotropical forest

Citation

Rosin, Cooper et al. (2020), Assessing the effects of elephant foraging on the structure and diversity of an Afrotropical forest, Dryad, Dataset, https://doi.org/10.5061/dryad.x95x69pdr

Abstract

African forest elephants (Loxodonta cyclotis) are ecosystem engineers that browse and damage large quantities of vegetation during their foraging and movement. Though elephant trail networks and clearings are conspicuous features of many African forests, the consequences of elephant foraging for forest structure and diversity are poorly documented. In this study in northeastern Gabon, we compare stem size, stem density, proportional damage, species diversity, and species relative abundance of seedlings and saplings in the vicinity of seven tree species that produce elephant-preferred fruits (“elephant trees”) relative to control trees that do not. Across 34 survey trees, with a combined census area of 2.04 ha, we recorded data on 26,128 woody stems in three sizes classes. Compared to control trees, the area around elephant trees had: a) a significantly greater proportion of damaged seedlings and a marginally greater proportion of damaged saplings (with 82% and 24% greater odds of damage, respectively); b) no significant difference in stem density or species diversity; and c) a significantly greater relative abundance of seedlings of elephant tree species. Increasing distance away from focal elephant trees was associated with significantly reduced sapling stem damage, significantly increased sapling stem density, and significantly increased sapling species diversity. Considered in sum, our results suggest that elephants can affect the structure and diversity of Afrotropical forests through their foraging activities, with some variation based on location and plant size class. Developing a more complete understanding of elephants’ ecological effects will require continued research, ideally with manipulative experiments.

Methods

Study Area and Species Selection

            We conducted this study in the Ogooué-Ivindo province of northeastern Gabon.  The region is dominated by lowland forest, and receives approximately 1700 mm of rain annually, with two rainy seasons (September-December and March-June).  The study area includes the northern section of Ivindo National Park and the Ipassa Field Station, located within the park’s buffer zone.

            We selected seven tree species that produce fruits that have been observed with high frequency in elephant dung to serve as our focal elephant trees: Annonidium mannii, Baillonella toxisperma, Dacryodes buettneri, Gambeya lacourtiana, Klainedoxa gabonensis, Mammea africana, and Panda oleosa (White et al. 1993, Poulsen unpublished data).  As many of these fruits are also consumed by apes and other mammals (White et al. 1993), we selected individual adult trees of each species that had visible impacts of elephant browsing around their trunks, in order to maximize the likelihood that the observed effects were caused by elephants rather than by other species.  In total, we conducted plant surveys around three individuals of each elephant tree species (21 total trees), and 13 control trees with the following criteria: a) located within 200 m of one of the elephant trees, b) had similar canopy-level height as the elephant trees, and c) did not produce fruits commonly consumed by elephants.  The control trees included two Alstonia boonei, two Celtis tessmannii, three Dialium pachyphyllum, one Lophira alata, two Newtonia sp., and three Pterocarpus soyauxii.

Survey Methods

            For each of the 34 survey trees, we recorded diameter at breast height (DBH, cm) and measured height (m) using a hypsometer. We also estimated canopy size by measuring the distance from the trunk to the end of the canopy at eight locations, and then calculating the corresponding area. At the base of each tree, we established three 5 x 40 m plots radiating out along the 0°, 120°, and 240° axes.  We marked each plot with ribbon at the corners, and delimitated them with Topofil thread so that the sides of the plot were clearly distinguishable.  We then subdivided each plot into eight 5 x 5 m quadrats.

            For each woody stem within the plots, we identified the plant to genus or species (see below) and recorded it as one of three size class categories for further measurement: seedling (0.5 m – 2 m in height), sapling (>2 m in height but < 6 cm DBH), or adult (≥6 cm DBH). We excluded lianas from our study with the exception of those within the seedling size class, which had not yet exhibited the typical climbing liana growth form and thus were as susceptible to elephant trampling damage as were tree seedlings. For seedlings, we measured the diameter of the plant at 5 cm and measured height from the ground to the terminal bud. For saplings and adults, we recorded only DBH. In the 0° plot, we measured all stems of all three size classes.  In the 120° and 240° plots, we measured all saplings and adult trees, but only counted (and identified) the seedlings. In addition to species identification and stem measurement, we assessed all stems for the presence of damage, including breakages and irregular regrowth that may have occurred after being bent or snapped.  All taxonomic identifications were conducted in the field by the same local botanist, in order to maintain consistency across survey trees. Plants that could not be identified to genus or species were assigned a unique code; all individuals that apparently belonged to the same unknown taxonomic group were given the same code, so that classification could be as specific as possible even when specific taxonomic identification was not possible.

Statistical Analysis

We used linear mixed (LMM) or generalized linear mixed (GLMM) models to test our hypotheses, including the focal tree species as a random effect in all models. For LMMs, we evaluated model fit by examining residuals and selected the best model based on the coefficient of determination, R2.  For GLMMs, we employed the negative binomial distribution when the data were strongly overdispersed and selected among models using the Akaike Information Criterion (AIC).  We performed statistical analyses in R 3.5 (R Development Core Team 2018) using the lme4 package (Bates et al. 2015), and followed the general recommendations for GLMMs outlined by Bolker et al. (2009).

To test whether elephant activity affects stem damage (H1) and density (H2), we modeled the proportion of elephant-damaged seedlings and saplings (binomial distribution) and the number of seedlings and saplings (negative binomial distribution) as functions of tree type (elephant tree vs. control tree) and distance from the tree (by quadrat), with species of the focal tree as a random effect.

To test whether elephant activity affects species diversity (H3), we computed Shannon-Wiener diversity indices (Magurran 2004) for seedlings and saplings around elephant trees and control trees.  We then assessed the effect of tree type (elephant tree vs. control tree) on diversity for both size classes, with species of the focal tree as a random effect. We did not account for differing levels of taxonomic identification in this diversity analysis, as even unidentified stems were given a unique identifying code in the field, and were considered in the analysis as unique “species.”

To test whether elephant activity increases the proportion of elephant-dispersed species (H4), we modeled the proportion of seedlings and saplings of elephant tree species (binomial distribution) as a function of tree type (elephant tree vs. control tree) and distance from the tree (by quadrat), with species of the focal tree as a random effect.  Although there are many other species that can be prominent in elephant diets, we considered only the relative abundance of the seven focal elephant tree species in this analysis.