Resilience of an Australian savanna grassland to selective and non‐selective perturbations

Using a randomized experimental design, plots of a savanna grassland were subjected to two levels of grass tuft removal (50% and 90%) in two ways; non‐selective (all species removed in proportion to abundance) and selective (tufts of the most palatable species removed first, then the next most palatable, etc.). The plots were maintained in their cleared states for three years, then monitored for the next five. In general, the sward was resilient to the disturbance except for the 90% selectively cleared treatment, in which a dominant, palatable species ( Themeda triandra ) failed to recover (though die most palatable species, Sorghum plumosum , did recover). The recovery patterns were dependent on post‐disturbance conditions, and markedly influenced by a particular rainy season and a fire during one of the dry seasons. In addition to species effects, the treatments induced changes in spatial patterning and associated micro‐scale hydrology. These effects persisted in the 90% removal treatment. In this regard the results are scale‐dependent, and the same percentage removals at different scales (e. g. 5 × 5 m patches rather than tuft × tuft scale) would lead to differences in ability to recover. In terms of value to livestock the selective 90% removal treatment was in a poor state at the end of die experiment. In all treatments die trajectory of species changes was back towards the controls, but the selective 90% plots were fully re‐vegetated before this could be achieved. In these plots, the final steps to complete recovery will occur only after death of established new tufts.