Use of multivariate techniques to identify vegetation units and monitor change on a livestock production system in a semi‐arid savanna of the eastern Cape

Abstract

Species composition data of the grass layer were determined from point surveys along permanently demarcated transects. These data were separated into vegetation units using multivariate techniques. The repeatability of this classification was tested on data collected over a period of time. The results indicate that classification based on composition data collected from a single sampling period was adequate for practical purposes provided certain precautions were taken. Change in vegetation was assessed by ordering sites from repeated measurements and tracing site trajectories through ordination space. The trends along the axes were strongly correlated with assessed condition score.