Results:
There was an initial sharp drop in the relative abundances ofLemna trisulca (Lt) and Wolffia columbiana (Wc) because of transfer shock, with a fraction of plants immediately sinking to the bottom of the mesocosms. Because of this, we have used week 2 as the initial time point. After this initial settling time, community composition continued to change during the 12-week experiment withLemna minor (Lm) coming to dominate most communities, regardless of its initial frequency, at the expense of Lt and Wc which became rare (but not extinct) in most communities (Fig. 2). Spirodela polyrhiza (Sp), maintained moderate abundances in most communities. The overall trajectory of community composition seemed to be largely independent of initial composition and was consistent among replicates. This was the case regardless of whether we relative abundance was weighted by mass (as for the main analysis), or left as un-weighted individual counts (Fig. S1 & S2 in Supporting information).
By the end of the experiment, each community type had diverged substantially from its initial composition. The extent of this divergence has three components: sorting as the source of directional change, drift as idiosyncratic divergence among replicates, and initial state as retention of differences among community types. To quantify the relative contributions of sorting, drift and initial state to final community composition we partitioned the overall sum of squares into components representing these three processes (Fig. 3). This analysis was done for each time point (see Appendix S3 in Supporting information), to evaluate how these three sources of community change varied over the course of the experiment (Fig. 4).
We regressed the normalized change in relative abundance between time points against previous relative abundance for each species to obtain estimates of equilibrium species frequencies (Fig. 5). These regressions are autocorrelated, but they can be used to estimate the equilibrium frequency of a species as the X-intercept, yielding 0.76 for Lm, 0.05 for Lt, 0.27 for Sp and 0.01 for Wc. When constrained to sum to 1, these frequencies are 0.70 for Lm, 0.04 for Lt, 0.25 for Sp and 0.01 for Wc.