Mammalian herbivory indirectly shapes savanna arthropod communities, but only at very low or high levels

Savanna ecosystems support unique biodiversity and provide livelihoods for millions of people. Yet, wild herbivores are in decline due to poaching and land-use change while livestock numbers are increasing. These changes in density and composition alter savanna vegetation. There are likely indirect cascading effects of altered vegetation on savanna arthropods, but our understanding is limited despite their pivotal role in ecosystem functioning. We evaluate how differences in mammalian herbivory affect terrestrial arthropods in a semiarid Kenyan savanna. We sampled ground-active arthropods (focusing on ants) in six herbivory treatments ranging from high-intensity herbivory to complete exclusion of large herbivores. Ant abundance and richness were not affected by herbivory treatments, but the community composition of ants and arthropods differed at extremely high and low levels of herbivory due to indirect impacts on vegetation.
The dataset consists of the raw ant data, the raw arthropod ..., , # Data from: Mammalian herbivory indirectly shapes savanna arthropod communities, but only at very low or high levels

Dataset DOI: [10.5061/dryad.kwh70rzk2](https://doi.org/10.5061/dryad.kwh70rzk2)

## Description of the data and file structure

The ant and arthropod data were collected by pitfall traps. Fifteen pitfall traps were placed in three transects with ten meters distance between traps. The traps were filled with propylene-glycol (50%) and left for three days, until they were retrieved and samples identified under a microscope. All arthropods were identified to order level, ants were identified to genus level, and then assigned to morphospecies.

The vegetation data were collected on 50 x 50 meter sampling plots. Grass and forb richness and percentage of bare ground were assessed on two parallel 50-m transects, using a 1 m^2^ quadrat placed every 5 m along the two transects, i.e., 20 vegetation measurements per variable plot. All trees in each plot with a diameter of 5 cm or g...,