Statistical analysis
To test whether there were differences in phenotype among sites and microsites in the field, we used a 2-way nested analysis of variance (ANOVA) with microsite nested within site. Both site and microsite were analysed as type II random factors. This was done for both response variables (frond area and root length). Since the environmental correlates were measured at the site level, all 30 measures of phenotype (10 individuals x 3 microsites) were averaged to produce a single value per site for frond area and root length. We then regressed site mean phenotype (both frond area and root length) against the environmental correlates (light availability, TN, TP, DN, DP, and pH) using linear regression and simplified the models by removing non-significant terms. To test whether there were differences in phenotype among sites and microsites after the common garden assay we used a similar nested ANOVA as that used for the field data, but with a 3rd level (replicate flask), nested within microsite, using the 10 individuals per flask as the error variance. Growth rate (fitness) was calculated for each flask over the final 20 days of the common garden assay using the standard formula for exponential growth\(r=\frac{\ln\left(\frac{\text{Nt}}{N0}\right)}{t}\) where N0 is initial population size, t is time in days, and Nt is population size at time t. To test for differences in fitness among sites and microsites we used a similar nested ANOVA with microsites nested within sites. However, since there is only a single measure of fitness per flask, replicate flask was used as the error variance.