2.3 Vegetation sampling
In August of 2004 to 2017, we randomly selected 10 quadrats of 50 cm × 50 cm in each of twelve plots. The standing crop by species was clipped at the soil surface, oven-dried at 65°C, and weighed. To avoid loss of information (e.g., if random sampling inadvertently avoided large individual shrubs), a systematic sampling method with no paddock replication was used to investigate plant species cover. In August from 2017 to 2019, a 40 m × 40 m sampling area was selected randomly in one of the three repeat plots of each of the CK, LG, MG and HG treatments, choosing plots that were similar in terms of topography and species composition. Four plots were selected to represent the LG, MG, HG plots in Block I and the CK plot in Block II was selected in order to avoid edge effects because the CK plot in Block I is located at the edge of the experiment site (Fig. 1). In this representative sampling area, there were 77 quadrats of 50 cm × 50 cm (Fig. 2). The quadrat layout was based on mechanical sampling and cover by species was measured in each quadrat. Simultaneously, 9-15 individuals of Artemisia frigida ,Kochia prostrata , Caragana microphylla andCeratoides latens (shrubs and semi-shrubs, hereinafter, shrubs) were randomly selected from the above grazing plots during 2017 to 2019, and the cover (%), height (cm), the diameter of the base of the cluster (hereinafter, basal diameter, cm), branch number (individuals), and the maximum branch diameter (mm) of the four shrub species were determined. There are five shrub species in experimental region, but Caragana stenophylla has very small individuals and low density and was not found under heavy grazing, and therefore was not systematically sampled.