Invader abundance
Results were largely consistent with predictions (Figs 1 & 5). Of the experimental treatments, fast invader cover was most strongly linked with successional stage (Fig. 2a), which itself was strongly linked with extractable soil nitrate (Fig. S6, Table S7), the dominant limiting resource in these grasslands (Tilman 1990). Limited soil N therefore likely explained the low invader cover in late successional communities (Fig. 2c), an effect that was partially alleviated when plots were disturbed (Fig. 2d), temporarily increasing soil N (Fig. S6). Early successional plots had consistently high levels of invasion through time despite marked declines in soil N (Fig. S6), suggesting that soil N did not limit fast invader cover in early succession, at least not in undisturbed plots. The non-conditional positive effect of seed dose demonstrated that seed dose was important across all conditions, including in early succession where fast invaders were otherwise unconstrained (Fig. 2a-c). Combined with lack of growth rate- or resource-limitation, this suggests that seeds were the key limiting factor in early succession for fast invaders, at least in undisturbed plots (Fig. 2e). Fast invader cover reached intermediate levels in mid succession, with trends consistent with seed- and some resource-limitation.
Slow invader cover was primarily limited by resources in late succession too, as demonstrated by effects of disturbance and trends across the three successional stages (Fig. 3c-d). However, unlike fast invaders, slow invader cover seemed to be limited by species’ growth rates in this short-term experiment, as the strong link between slow invader cover and time (Fig. 3a) and its marked temporal increase in early and mid successional plots demonstrated (Fig. 3c).
Most relationships involved interactions, consistent with effects of an interplay of limiting factors (Fig. 1b). Unless interactions were included, successional stage was not selected in the fast invader model despite it being the most influential treatment for fast invader cover (Tables S4 & S9, Fig. 2a). A similar situation occurred for slow invader cover where invader type and disturbance were absent from the best non-interaction model (Tables S5 & S10). The most marked interactions for both invader groups were between community successional stage and disturbance, and successional stage and time (Fig. 2a,c,d, Fig. 3a,c,d). These interactions reflected the varying importance of resource- and growth rate- limitation in different contexts (Figs 1 & 5). Invader type (both groups or target group only) was the only experimental treatment not included in the best model for fast invaders, which had a marginal R2 of 0.44 and conditional R2 of 0.80. All four experimental treatments plus time were included in the best model for slow invaders, which explained considerable variance (marginal R2 = 0.83; conditional R2 = 0.92).