Trait data
We assigned dispersal mode (abiotic versus animal-dispersed) based on species- or genus-level data in the TRY database or, if such data were missing, based on searching the literature. We defined tree species as ‘large vertebrate dispersed’ if the list of their dispersers did not include any small-bodied rodents (families Muridae, Sciuridae, Heteromyidae, Cricetidae, or Echimyidae) or non-corvid passerine birds. Lists of dispersers were obtained from the TRY database if possible and from the literature if not (information sources for all species presented in Table S2).
Seed mass, wood density, specific leaf area, and adult plant height were gleaned from the TRY Plant Trait Database (Kattge et al. 2011) if available, or otherwise taken from the literature (Table S3). We chose these traits because they are some of the most commonly measured functional traits and ones that are often used in functional trait studies (e.g. Adler et al. 2014; Kunstler et al. 2016). When we could not find species-level data, we used median genus-level values for wood density (10 species; 17% of total) and SLA (17 species; 29%). The TRY database did not include any wood density data for cycads, but we found stem density volume and mass estimates in the literature (Cousins et al. 2011).