Effects of source and exposure environment in the greenhouse
In the analyses of drivers of trait variation in the greenhouse, the best models always included effects of at least one source environmental driver and both light and water exposure treatments (see blue lines in Fig. 3; Table S3; Fig. S2-S6), but results differed between vegetative and reproductive traits. For vegetative traits (biomass, SLA and RSR), light or water treatments showed the strongest effect sizes when compared with source drivers (Fig. 4c,f,i). The effect of source drivers on biomass and SLA frequently changed between positive and negative directions depending on the treatment (see interactions in Fig. 3a,b; Fig. 4a,b,e). For biomass, the effects of Aridity, Vegetation cover and Mowing differed among treatments, although effect sizes were low and had 95% confidence intervals (CI) that mostly overlapped with zero (Fig. 3a). For SLA, all source drivers were selected in the best model except for Mowing. SLA showed two contrasting effects between source and exposure environments: 1) SLA was lower in the Dry treatment, but higher in plants from the most arid populations (Fig. 4d); 2) SLA was higher in the treatment with lowest light, but also higher in populations with lowest source vegetation cover and thus highest light availability (this took place in treatment L33; Fig. 4e). RSR increased with source Aridity, although the effect was smaller than the effect of the Dry treatment (Fig. 4g,i).
For reproductive traits (biomass-corrected probability of flowering and fecundity), exposure treatments exerted equivalent or smaller effects than source drivers (Fig. 4l,o) and the effects of source drivers were consistent in direction across treatments (Fig. 3d,e, 4k,m). Probability of flowering was negatively affected by source Vegetation cover and positively affected by Mowing, and exposure treatments changed the magnitude of these source effects but not their sign (Fig. 3d, 4k). Fecundity was positively affected by source Aridity and Temperature, and showed no interactions between source and exposure environments (Fig. 3e, 4m). When biomass was excluded as a covariate from the analyses, 1) source effects decreased in magnitude and exposure effects generally increased for probability of flowering, and 2) source effects were not included in the best model of fecundity (Table S7).