The relative roles of plants competing for resources versus top-down control of vegetation by herbivores, in turn impacted by predators, during early stages of tropical forest succession remain poorly understood. Here we examine the impact of insectivorous birds, bats and ants exclusion on arthropods communities on replicated 5x5 m of pioneering early successional vegetation plots in lowland tropical forest gaps in Papua New Guinea. In plots from which focal taxa of predators were excluded we observed increased biomass of herbivorous and predatory arthropods, and increased density, and decreased diversity of herbivorous insects. However, changes in the biomass of plants, herbivores and arthropod predators were positively correlated or uncorrelated between these three trophic levels and also between individual arthropod orders. Arthropod abundance and biomass correlated strongly with the plant biomass irrespective of the arthropods’ trophic position – a signal of bottom-up control. Patterns in herbivore specialization confirm lack of a strong top-down control and were largely unaffected by the exclusion of insectivorous birds, bats and ants. No changes of plant-herbivore interaction networks were detected except for decrease in modularity of the exclosure plots. Our results suggest weak top-down control of herbivores, limited compensation between arthropod and vertebrate predators, and limited intra-guild predation by birds, bats and ants. Possible explanations are strong bottom-up control, a low activity of the higher order predators, especially birds, possibly also bats, in gaps, and continuous influx of herbivores from surrounding mature forest matrix.

The study was conducted in abandoned food gardens created during slash-and-burn subsistence agriculture practiced by indigenous communities. Within a radius of 3 km from the Wanang village, six experimental gardens (blocks) 3-5 years old were selected, each covering an area of approximately 400 m2, separated by a minimal distance of 200 m. At each garden we initiated regeneration by clearing all aboveground vegetation.

Control and exclosure plots were randomly assigned within each of the block. In the exclosure plots we used a transparent fishing net (1.5-2.0 cm mesh size) to protect vegetation from foraging birds and bats. We used 50 cm plastic fence with insect glue applied every week, and poisonous tuna baits (permethrin 0.5%) exposed in monthly intervals in enclosed traps to reduce the abundance of the predatory ants. Exclosure plots were surrounded by a plastic fence to prevent ants from re-colonizing the plot. A 50 cm high, dummy plastic fence was built around the control plots with a 10 cm gap left open on the ground level to account for possible effect of fencing in the exclosure plots.

Spontaneous regeneration was allowed to proceed undisturbed for 12 months. Exclosures were maintained throughout the entire experiment. After that time, all above-ground vegetation from all plots was destructively sampled. In each plot, all woody plant species with DBH ≥ 1 cm were sorted and identified to species and their leaves and stems were weighed separately with 1 g accuracy. Leaf area was calculated by randomly sampling leaves from the pool of all leaves for each species at a given plot until 50x50 cm area (“leaf frame”) was filled. The frame was then photographed and processed using ImageJ software, measuring the area of individual leaves. Leaf frame samples were then oven dried and weighed.

Before harvesting, individual plants were covered with insect-net and all arthropods were sampled using knock-down insecticide (MORTEIN®, active ingredients: allethrin [2.09 g/kg] and resmethrin [0.39 g/kg]). The arthropods were preserved in 70% ethanol and assigned to trophic guilds (herbivore or arthropod predator), orders, and morpho-species (further referred to as species). We measured the body length of up to 20 individuals per arthropod species and used it to estimate individual species biomass (dry weight) with the help of biomass equations. Arthropod density was calculated as the number of individuals per square meter of the leaf area, estimated for woody plant individuals by multiplying their fresh leaf weight (in g) and specific leaf area values (in cm2/g) that were also based on fresh mass.