Fig 1.
A schematic presentation of a simple system of competitive release (a) and the design of our experiment (b-d). In the most simple system (a), a releasing factor (C) allows a focal species (A) to persist in a community by releasing it from the negative effect of a superior competitor (B). In our experimental system (b) grazing is treated as the releasing factor, forbs as the focal species, and grasses as the superior competitors. The magnitude of competitive exclusion is quantified by comparing forb richness in plots with (c) and without (d) grasses in the absence of grazing, and the magnitude of competitive release is quantified by comparing plots with (b) and without (c) grazing.
Fig 2.
The CR-CE space as a framework for analyzing the balance between competitive release and competitive exclusion.
Systems in which the releasing factor fully compensates for competitive exclusions fall on the line y = x (the 'compensation line', points A and B in (a)). Systems characterized by partial compensation fall below the compensation line (C, D, E in (a)). Note that points C and D have the same effectiveness although they differ in the magnitude of competitive exclusion. Point E shows a lower effectiveness than points C and D although the magnitude of competitive release is similar to point D. If data on both forces are available for a set of sites within the same system, the effectiveness of the releasing factor can be expressed by the slope of a linear regression fitted to the data (b).
Fig 3.
A map of the experimental system and the sampling design.
Blocks located on the slopes are marked by yellow dashed lines; blocks located in the valleys are marked by white dashed lines. In each habitat there are three blocks with three treatments per block (grazing, grass removal, and control) and four blocks with two treatments (grazing and control). Each plot is 20x20m, but sampling was limited to the inner area of 10x10m. This area was sampled by 25 quadrates of 0.04m2 organized in five clusters of 1m2. Clipping experiments were conducted in the peripheral areas of control plots.
Fig 4.
Effects of grazing and grass removal on forb richness in the study system.
(a) A summary of the experimental results using the CR-CE framework. The dashed line is the compensation line (y = x). Each point represents a certain combination of habitat, block, and scale. Each line is a regression line fitted to a different scale (0.04m2: R2 = 0.63, P = 0.059; 1m2: R2 = 0.82, P = 0.012; 100m2: R2 = 0.54, P = 0.095, all slopes are significantly lower than 1 and all intercepts do not differ significantly from zero, significance levels based on standard errors of the regression coefficients). (b) Effect size (log response ratio) of the grass removal and grazing treatments under each combination of habitat (slopes vs. valleys) and scale (0.04, 1, and 100m2). Log response ratio is quantified as Log(STREATMENT/SCONTROL), where S = mean number of forb species under the relevant combination of treatment, habitat and scale.
Fig 5.
Results of ordination analysis (Nonmetric Multidimensional Scaling, NMDS) showing the effect of treatment (grass removal, grazing, and control) on forb species composition in the two habitats (valleys vs. slopes).
Analyses were performed at the cluster scale (1m2) using the Jaccard index as a measure of dissimilarity.
Fig 6.
Results of a small-scale clipping experiment mimicking the effect of biomass removal by the cattle on the vegetation in the study area (removal of all shoots higher than 7 cm).
The experiment was conducted within experimental units of 0.16m2 protected from grazing (see Appendix B in S1 File for details). (a, e) Clipped biomass of forbs and grasses in the two habitats. (b, f) Seedling mortality. (c, g) Extinction rates. (d, h) Species richness at the end of the growing season. Bars represent 95% confidence levels. Significant differences between clipped biomass of grasses and forbs (a, c) and between clipping treatments (b-d, f-h) are marked by asterisks.