Phase 1: Conditioning phase.
For the conditioning phase, 96 plots of 166 ⋅ 250 cm were each divided into two 83 ⋅ 250 cm experimental subplots. Plots were allocated into four blocks and within each block, each plot was randomly allocated to specific treatment combinations; see below.
To test soil legacy effects in the field, six different conditioning plant communities were designed, each consisting of six species of grasses and forbs. The experimental design of the conditioning phase has been described in full detail in De Long et al. (2019). The plant species were randomly selected from two separate pools of species, so that three communities consisted of random combinations from the first pool and the three remaining communities consisted of random combinations from the second pool.
To test the effects of plant functional types in the conditioning plant community, the ratios between grasses and forbs in each seed mixture were manipulated, and each mixture was prepared in four different forb:grass ratios (i.e., 0:100%, 25:75%, 75:25% 100:0% forb:grass). To test whether soil legacy effects would become stronger when the period of conditioning was longer, the two subplots within each plot were conditioned for different durations (i.e., a one-year or two-year legacy). Each treatment combination was replicated across four blocks, totalling 6 plant communities ⋅ 4 forb:grass ratios x 2 conditioning durations x 4 replicate blocks (=192) subplots.
In May 2015 (i.e., two-year legacy treatments), for each plot, one of the two subplots weas randomly assigned to one of the above treatments. All vegetation of the subplot was removed, while the other subplot was left untouched in that year. Removal included the top soil layer of approximately 5 cm, which generally contains the highest density of roots in this grassland system. This was done to remove the most dominant roots and prevent re-growth of non-target plant species, as well as minimize dominant pre-existing soil legacy effects of the previous plant communities that had grown in the soils. Each stripped subplot was then sown with one of the prepared target communities described above.
In May 2016 (i.e., one-year legacy treatments), all vegetation was removed (as described above) from the remaining untouched subplot in each of the 96 plots. The subplots were sown with the same target community as the other subplot in that plot the year before.