Statistical Analysis
All analyses were conducted using R version 3.6.0 (R Development Core Team 2020) and RStudio version 1.2.1335 (R Studio Team 2019) with relevant packages cited below.
To test for differences in plant richness, evenness, diversity, total plant cover and height across the growing season, we constructed a linear mixed-effects model using the lmer function in thelme4 package (Bates et al., 2015). Within our model, herbivore exclusion, survey month, and their interaction were treated as fixed effects and plot was nested within plot pairing to account for the repeated measures and the spatial variation of our site. We performed a Wald test to assess the significance of fixed effects using theAnova function (Fox, 2019). We plotted modeled parameter estimates of the herbivore exclusion treatment after accounting for random effects using the effects package (Fox, 2019) with theggplot2 package (Wickham, 2016).
We used a permutational multivariate analysis of variance (PERMANOVA) to examine the differences in plant species composition due to herbivore exclusion, sampling date, plot pairing (spatial structure) and their interactions using the Adonis function in the veganpackage (Oksanen et al. 2013). We also performed a PERMANOVA, using a Bray-Curtis dissimilarity metric (Bray and Curtis, 1957), to explore the combined effect of species abundance changes and turnover within our plots. Finally, we visualized the results of our PERMANOVA using non-metric multidimensional scaling (NMDS) via the metaMDSfunction in the vegan package.
To understand the contribution of each plant species to compositional change to tortoise exclusion we conducted a similarity percentage analysis using the simper  function in the vegan  package. The simper function calculates the contribution of each species to differences in Bray-Curtis distance between treatment groups and calculates the average abundance per group. The similarity percentage analysis generates a permutation p-value based on the probability of getting a larger or equal average contribution in random permutation of the group factor. We conducted 99 permutations during the similarity percentage analysis.