In dryland fields in NE Spain, post-dispersal seed predation by harvester ants, Messor barbarus, can be very high (78%), however variable over time and space. Factors that influence the variability are largely unknown. The purpose of the current research is to quantify temporal and spatial variability in seed predation in dryland cereals to investigate how this variability might influence the efficacy of granivory and provide opportunities to weeds to escape predation. In order to decrease or remove constraints for this natural service, measures to improve conditions for nest establishments in areas with a low presence of harvester ants could be considered.

A short summary of the main results of my PhD research is provided below.

Temporal variability. I found that the reason for the high annual predation rates is that the timing of weed seed shed largely overlapped with the period of highest weed seed demand. However, a small proportion of newly shed seeds could escape predation because of three reasons, namely low preference, seeds were covered by soil and debris during crop harvest, or because seeds were incorporation into the soil matrix, which happens faster for small than for large seeds. For example, I calculated that of the seeds of Bromus diandrus about 15 to 25 % avoided predation because of low preference to ants, while 0–29 % escaped through crop harvest. Of the seeds of Papaver rhoeas, 20–32 % escaped through crop harvest, while another 13–17 % escaped by fast burial into the soil.

Spatial variability. One source of spatial variability may arise from the arrangement of harvester ant nests in the field. Harvester ants are central place foragers, meaning that the area in a circle of approximately 30 m around the nest is searched intensely for food, but outside this circle much less intensely. A regular distribution of nests would ensure a full coverage of the field by foraging ants, while an increasing degree of clumping (over-dispersion) would mean that an ever increasing part of the field is not covered at all. I found that large nests were more or less regularly distributed, but small nests were over-dispersed in 2009, but not in 2010.

The density and distribution of weed seeds over a field constitutes another potential source of spatial variability. The predation rate may either increase (direct density-dependent response) or decrease (inversely density dependent response) when ants are confronted with high density patches. I found that harvester ants responded with extremely high predation rates (99-100 %) to all densities (1000-20000 seeds) of Lolium multiflorum L. seeds applied to 1 m² patches and that the response was density independent. Not all patches were found, but the probability that it was found (encounter rate) was independently of patch density.

Similarly, the size of seed patches could influence predation rates by harvester ants. I found that seed predation rates increased with increasing patch size (0.25 to 9 m²) from 78-94 % in the smallest patches to 99-100% in the largest patches. Small patches had a lower encounter rate than large patches. When patches were found, the exploitation rate was the same, regardless of size.