Ant suppression experiments have emerged as a powerful method for assessing the role of ants in ecosystems. However, traditional methods have been limited to canopy ants, and not assessed the role of ants on and below ground. Recent advances have enabled whole-ecosystem ant suppression in large plots, but large-scale experiments are not always feasible. Here, we develop a small-scale, whole-ecosystem suppression method. We compare techniques for monitoring suppression experiments, and assess whether habitat complexity in oil palm influences our method’s effectiveness.

We conducted ant suppression experiments in oil palm agroforestry in Sumatra, Indonesia. We used targeted poison baits, a physical barrier, and canopy isolation to suppress ants in 4m-radius arenas around single palms. We sequentially tested three suppression methods that increased in intensity over 18 months. We sampled ant abundance before and after suppression by fogging, using pitfall traps, and extracting soil monoliths. We also monitored ants throughout the experiment by baiting. We tested the soil for residual poison and monitored other invertebrates (Araneae, Coleoptera, Orthoptera, and Chilopoda) to test for cross-contamination. Plots were established under four oil palm management treatments that varied in their habitat complexity: reduced, intermediate, and high understory complexity treatments in mature plantation, and a recently-replanted plantation.

Post-treatment ant abundance was 92% lower in suppression than control plots. Only the most intensive suppression method, which ran for the final nine months, worked. Baiting rarely reflected the other monitoring methods. The treatment negatively affected Orthoptera, but not other taxa. We detected no residual poison in the soil. Coleoptera abundance increased in suppression plots post-treatment, potentially due to reduced competition with ants. Our findings were consistent across management treatments.

We developed a whole-ecosystem method for suppressing ants on a small scale in oil palm plantations. Our method represents a significant advance; previous reductions in ant abundance have not exceeded 38%. We provide the first example of ants being experimentally suppressed belowground. Baiting is not adequate for assessing suppression effectiveness, and testing a range of taxa for confounding impacts is important. Our study acts as a blueprint for developing suppression methods for other taxa, which offer unique insights into community ecology.

We conducted ant suppression experiments in oil palm agroforestry at the Biodiversity and Ecosystem Function in Tropical Agriculture (BEFTA) programme (oilpalmbiodiversity.com) in Sumatra, Indonesia. We used targeted poison baits, a physical barrier, and canopy isolation to suppress ants in 4m-radius arenas around single palms. We sequentially tested three suppression methods that increased in intensity over 18 months, and the suppression worked for the final method which lasted for nine months.

We established 24 suppression plots and 24 control plots. These were evenly split over four oil palm management treatments that varied in their habitat complexity:

1. Reduced complexity (Reduced): Mature first-generation plantation with all understory vegetation removed by spraying herbicide.
2. Normal complexity (Normal): Mature first-generation plantation with understory vegetation removed from the harvesting paths and harvesting circle (a ~1.5m radius area around the base of each palm) using herbicide, and large woody vegetation removed manually. Other vegetation was allowed to grow. This is standard industry practice in these estates.
3. Enhanced complexity (Enhanced): Mature first-generation plantation with the same understory management as the normal complexity, except harvesting paths and circles were cleared by strimming rather than herbicide.
4. Replanted (Replanted): Young second-generation plantation with the same understory vegetation management as normal complexity plots.

We sampled ant abundance before and after suppression by fogging (six trays per plot), using pitfall traps (three traps per plot left for three days each), and extracting soil monoliths (two per plot). We also monitored ants throughout the experiment by baiting with sugar and tuna baits in the canopy and on the ground (four plates per plot in total). For baiting, we identified two conspicuous species: Yellow Crazy Ant (Anoplolepis gracilipes) and Weaver Ant (Oecophylla smaragdina). We monitored the most abundant non-ant  invertebrate taxa (Araneae and Coleoptera in the canopy, Coleoptera and Orthoptera on the ground, and Chilopoda belowground) to test for cross-contamination. We also tested the soil for residual poison and found that there was none.

There are five datasets which correspond to the different sampling methods (fogging, pitfalls, monoliths, and baiting).

1. Fogging – each row represents one palm (i.e. the values for the six trays were summed)
2. Pitfalls – each row represents one pitfall trap
3. Monoliths (pre-treatment) – each row represents a single observation of an invertebrate
4. Monoliths (post-treatment) – each row represents one monolith
5. Baiting – each row represents one bait plate