Results

Effects of foundation species on mean ecosystem parameters - The presence and absence of Myriophyllum and Dreissenaaffected a wide range of ecosystem parameters. During the first nutrient addition, ponds with Myriophyllum or Dreissena alone had lower chlorophyll fluorescence, i.e. lower algal biomass, than ponds with neither species, consistent with their anticipated negative effects on the phytoplankton community (Fig. 2). However, following both disturbance periods, the co-occurrence of these species had strong non-additive antagonistic effects on algae abundance, illustrated by their positive effects on mean chlorophyll and phycocyanin fluorescence. Furthermore, after the first disturbance period, and throughout the remainder of the experiment, the presence of Myriophyllumincreased the concentration of DOM in the ecosystems, independent ofDreissena presence (i.e. in M and MD treatments). The presence ofMyriophyllum and Dreissena , either alone or in combination, positively affected dissolved oxygen saturation throughout most of the experiment, however, not during most of the perturbation periods: each nutrient addition dramatically increased dissolved oxygen saturation to levels (between 150 and 200 %) that were not significantly different across all species contrasts.
Effects of foundation species on variance of ecosystem parameters - We found only weak effects of Myriophyllum andDreissena presence on variance patterns (CV, Fig. 3) and early warning indicators (Supplement). Overall, we found strong increases in CV across all treatment combinations and ecosystem parameters immediately after the nutrient additions, which reflects the sudden changes in the mean in response to the disturbances. Prior to the first nutrient additions, the pond ecosystems with either Myriophyllumor Dreissena alone were less variable in chlorophyll fluorescence. After the second nutrient pulse, ecosystem where both species were present variance of chlorophyll and phycocyanin fluorescence were significantly higher than when species were alone, or absent. There were almost no effects of foundation species (i.e. their presence or absence) on the variance of DOM fluorescence. There were indications of Myriophyllum presence affecting CV of dissolved oxygen saturation, however, only to weak effect and with variable sign. As expected, each nutrient addition led to a temporary increase in AC across all treatment contrasts and parameters (visible as spikes in the time-series, Fig. S1), but we did not identify treatment specific differences in AC. Similarly, there were no consistent treatment specific differences in GEV or Skewness in any of the ecosystem parameters (Fig. S2 and Fig. S3).
Ecosystem metabolism - GPP and NEP, as well as R were strongly affected by nutrient perturbation and seasonal dynamics, but less so by the presence or absence of foundation species (Fig. 4). Each nutrient addition led to correlated increases of GPP, NEP and R, which reverted within days after the maximum was reached. During each of these peaks, there were only little differences across all species contrasts and all metabolism metrics. During spring 2017, at the beginning of the second Phase, all pond ecosystems containing Myriophyllum orDreissena had lower NEP and higher R than ecosystems devoid of foundation species. We found a similar pattern towards the end of the experiment, after the second nutrient addition in Phase 3, where both GPP and NEP were lower and R higher when foundation species were present. Overall, there were only weak effects on variance patterns of ecosystem metabolism (Fig. 5): there was a tendency for MD ponds to have higher CV of GPP and NEP than ponds without any foundation species, especially around the second perturbation Phase. Interestingly, nutrient perturbation led to increasing CV of GPP and NEP, but not R, which had a relatively static CV of approximately 0.8 throughout the entire experiment.