Human activities have converted mature forests into mosaics of successional vegetation and chronically disturbed habitats, altering the patterns of populations distribution, foraging ecology, and thus, the flow of matter and nutrients through ecosystems. Although the effects of human disturbance are mostly harmful, hyperabundant native generalist species can emerge and increase their populations under disturbance, such as leaf-cutting ants (LCA), prominent herbivores that are considered ecosystem engineers. Here, we examined the population response of two LCA species of the Caatinga dry forest (Acromyrmex balzani and A. rugosus) to increasing levels chronic anthropogenic disturbance and aridity, and assessed the foraging activity, biomass and nutrients harvested by their colonies. We found that colony densities increased at more disturbed habitats, varying considerably from 0 to 81 nests/ha, but aridity had no effect. The two species exhibited markedly different foraging activities (44.66 ± 28.76 and 294.6 ± 260.53 ants foraging daily), with the foraging rate increasing in more arid conditions for a species with smaller nests, but with no response to disturbance. Biomass consumption varied distinctly between species, ranging from 0 to 4.81 g (7.24 kg ha.yr-1, in A. balzani) and from 5.6 to 74 g (174.39 kg ha.yr-1, in A. rugosus). Furthermore, there was no effect of disturbance and aridity on the biomass harvesting of individual colonies. However, there was a considerable increase in the biomass harvested by the populations of colonies in the plots (i.e. accounting for colony densities). Moreover, the species A. balzani foraged upon more nutrient-rich material at more disturbed and arid habitats, with plant material containing higher concentrations of N, Ca, S, Sr, Fe, and Mn, as well as lower C:N ratio in these areas. Our results suggest that Acromyrmex species (1) can achieve larger populations in more disturbed habitats, though not directly associated with aridity, (2) operate as a key herbivore able to fit harvesting/diet through the entire environmental gradient and forage complementary (monocot vs. dicot), (3) reallocate expressive amount of forest biomass, resulting into temporary nutrient sinks with potential impacts on Caatinga resilience.

Colony density

To address the colony density of the two Acromyrmex species, we surveyed four transects of 4 × 100 m (1600 m²) in each of the 19 study plots, totalizing 30,400 m². We started at the centre of the plot and moved straight ahead in four directions separated by a 90º angle. To ensure the transect length and direction we used the GPS tracking function (Global Positioning System, Garmin e-Trex 20), with an estimated resolution < 3 m. Using the GPS device, we marked every colony encountered along the transects. When crossing an active foraging trail, the associated colony was detected by tracing the ants back to their nest entrance. The colony density was calculated by dividing the number of nests found in the total area sampled per plot (1600 m²) by 0.16, thus obtaining colony density per hectare. To fit the data into a poison distribution, decimal values of colony density were rounded to the nearest integer.

Foraging activity

To assess the foraging activity of LCA colonies, we accompanied 12 colonies of A. balzani and five colonies of A. rugosus during a series of 24-hour field campaigns. We conducted 5-min counts of foraging ants (returning to their nest carrying food fragments) passing an imaginary line at hourly intervals (Siqueira et al., 2018; Urbas et al., 2007; Wirth et al., 1997). The field campaigns were conducted twice for all colonies, once during the wet season (March–July) and once again during the dry season (August–February). The objective was to determine the seasonal foraging activity of both LCA species, with a total of four census hours per colony totalling 68 h of field observations. The 12 A. balzani colonies were located in six different plots, each area containing two colonies. We thus used the mean number of foraging ants of the two colonies at each plot. The five A. rugosus were situated in five different plots, and thus each colony constituted a single sample unit. Finally, we determined the peak of daily foraging activity of each colony from both species, which is the time of the day when colonies have the highest number of workers passing by the imaginary line during the 5-min intervals (Wirth et al., 1997, Urbas et al., 2007, Siqueira et al., 2018).

Biomass consumption rates

To quantify the effects of disturbance and water deficit on the colony biomass consumption rates, we marked 32 active colonies of A. balzani in a subset of seven plots, and 21 active colonies of A. rugosus in a subset of eight plots. We sampled each focal colony twice, one time during the wet season (March – July) and a second time at the dry season (August – February). To calculate biomass consumption, we captured all ants carrying plant fragments (leaves, flowers, stalks, and seeds) at 15-min intervals on the colony trail during the peak of daily activity (see the previous section). Thus, each colony was sampled for a total period of 30 minutes, totalling 26 h and 30 minutes of field sampling. Plant fragments were packed in plastic bags for transportation, then transferred to paper bags and dried in a stove to a constant weight at approximately 60°C for dry weight determination. All ants were carefully released back on the trail. Based on fragments weight, we calculated the biomass-based consumption rate per colony by summing daily totals from the 5-minutes hourly observations described above (see also Wirth et al. 1997, Urbas et al. 2007, Siqueira et al. 2018 for more details on methodology). We also determined the biomass harvested per hectare per year (i.e. considering the colony density in the area). Firstly, we calculated the average biomass collected by each colony between the two seasons. Secondly, the biomass collected by all colonies sampled within the same area was added together and the proportion corresponding to one hectare was determined, considering the density of colonies in the area for each LCA species, obtaining the biomass harvested by the colonies per hectare per year for each plot.

Nutrient concentration in plant material harvested by colonies

To determine the nutrient concentrations of plant material harvested by the ants, we collected at least 1 g of dry weight per colony from 21 colonies of A. balzani and 20 colonies of A. rugosus, using the same sampling method as for the above-described biomass harvest at 15-min intervals on the colony’s foraging trails during the peak of daily activity. The harvested plant fragments collected were packed in paper bags and dried at 40-50 °C by 48 h, identified and stored in clean plastic bags. The dry samples were first ground using a planetary ball mill (PM 200, Retsche ®) for 10, 15 or 20 minutes depending on the hardness of the material. We obtained a loose plant powder that was weighed to one gram per sample and packed into a polyethylene pill of 20 mm internal diameter and covered with 6-µm-thick polypropylene film (Mylar®). We placed the polyethylene pills in an energy-dispersive X-ray fluorescence spectrometer, model EDX 720 from Shimadzu, which consists of a rhodium tube for generating X-rays, a sealed chamber for sample analysis in a vacuum atmosphere, and a Si(Li) detector to measure the induced radiation. We selected magnesium, phosphorus, chlorine, potassium, calcium, manganese, iron, zinc, strontium, and sulphur as nutrients to assess their concentrations in the plant material harvested by the focal colonies. We also quantified total carbon and nitrogen in the plant material with an elemental analyser EuroVector (EA3000) coupled to an isotopic ratio mass spectrometer Denta V Advantage (Thermo Scientific). We used the elemental analyser configured with a CHN reactor filled with chromium oxide, reduced copper wires and silver cobalt oxide, a water adsorption trap (magnesium cobalt) and a chromatographic separation column.