Effect of weaver ants on arthropod abundance
Field Methods In June 2014 and May 2015, we measured arthropod abundance at 17 pairs of trees with weaver ants present and trees without weaver ants, where the trees were paired by species, height and girth. To measure arthropod abundance, we beat the foliage of a tree with a stick, and collected all the insects that fell on an upturned umbrella (≈ 100cm diameter), which is similar to the method used by Piñol et al. (2012). We beat the foliage three times before collecting arthropods in vials containing 95% ethanol and repeated the process at another part of the tree. Later in the field camp, we counted all the collected arthropods, measured body length to the nearest mm, and classified them to taxonomic order. To control for observer bias, the person counting the arthropods was unaware of the presence or absence of weaver ants on the source tree. We also estimated leaf damage on the pairs of trees as a longer-term proxy of insect herbivore abundance, with greater arthropod abundance implied by higher leaf damage. We used two methods to assess leaf damage: (1) visual estimation of % absent on 10 leaves from different parts of each tree, and (2) visual estimation of % absent on each leaf on a short (~0.5m) clipped branch of the tree.
In April-June 2015 we carried out a weaver ant removal and exclusion experiment using 15 pairs of trees, paired by species, height and girth. On trees in the experimental treatment group, we removed all weaver ant nests using a tree pruner and then applied a band of TanglefootTM around the trunk of the tree at about 1 m height from the base of the tree. We plugged all the holes in the trunk crevices under the Tanglefoot band with cotton. We were unable to completely remove or exclude weaver ants from 6 out of 15 treatment trees because of the nature of the bark of the tree trunk or the canopy of the tree. On trees receiving the control treatment, we attached a band of brown paper to make them appear similar to trees receiving the experimental treatment and pruned a few branches to mimic the disturbance caused by the removal of weaver ant nests. We measured arthropod abundance at each of these trees using beating and branch-clipping (Ozanne 2005) before and one month after the experimental treatments. We also recorded leaf damage as an index of insect herbivory on the clipped branch and on 16 random leaves at each tree. We recorded arthropod abundances one year later in April-June 2016, but by that time a number of trees had gained or lost weaver ants to an uncertain degree; results are presented in the supplement.
Data analysis We used paired t-tests to compare arthropod abundance and leaf damage between trees with or without weaver ants. Arthropod abundance data were log-transformed with one added to all values to avoid zeroes in the data. We also used paired t-tests to compare the change in arthropod abundance and leaf damage over the period of one month and over the period of one year between trees in the weaver ant exclusion and control treatments. Since we were interested in the effect of weaver ants on other arthropods, we removed the number of all ants from our arthropod count. We also removed the number of insects belonging to the suborder Homoptera, because weaver ants form mutualistic associations with Homopterans (Peng and Christian 2005; Crozier et al. 2009). All analyses were done in the RStudio programming environment (R Core Team 2014).