Data for "Climate mediates whether seed predators erode or support tropical tree diversity"

Study site. The Xishuangbanna permanent forest dynamics plot in southern China (101°34′E, 21°36′N) (Fig. 1a) has a mean annual temperature of 21.8  ̊C and there is frequently heavy fog that enables tropical rainforest vegetation to expand further the north than would be otherwise expected (45). There is also variability in topography and microclimates and ravines remain damp throughout most dry seasons. The dominant emergent tree species are Parashorea chinensis and Castanopsis echidnocarpa and the most abundant tree species in the subcanopy layer is Pittosporopsis kerrii (50). Daily precipitation was collected from the meteorological station at the Tropical Rain Forest Ecosystem Studies of Xishuangbanna Tropical Botanical Garden 47.4 km away from the Xishuangbanna forest dynamic plot.  A 20-ha permanent plot (400 m by 500 m) was established in 2007 following the standard protocols of the Smithsonian’s global forest dynamics plots (FDP) and included in the ForestGEO global earth observatory network (51, 52). All free-standing trees of ≥1 cm in diameter at breast height (DBH) were measured, tagged, and identified to species (50). In November 2007, a total of 150 seed-trap stations was established in a random stratified design to monitor seed rain and seedling dynamics throughout FDP (Fig. 1b). We excluded 22 of the 150 seed traps along the plot edges because covariates like tree densities within a 20 m radius of seed traps could not be estimated. Each 0.5-m2 seed trap had three 1-m2 seedling plots placed 2 m away on three sides. Seed censuses were conducted every two weeks, wherein all seeds and litter were collected and sifted by local botanists in the lab. Due to the immense scale of the fieldwork (frequency of collecting seeds and the number of seeds collected), we were unable to do a dissection of each seed, and therefore rely on visual inspection to ascribe seed damage. While we used magnifying glasses, microscopes, and only included tree species with average seed sizes > 0.2 cm to improve confidence around precision ascribing seed predation. We recorded the seed species, maturity, and damage were recorded following (40), with damage defined as having a hole, gall, or other physical damage characteristics. Only species with greater than 20 total seed records were included in our final data analysis. Therefore, while the FDP has recorded a total of 468 tree species to date, of which 169 species had seeds recorded in our experiment, only 146 species had mature seeds >0.2 mm and only 84 species also had 20 total seed records and were included in the analyses presented here (113763 seeds total). Data preparation. To test the JCH and PSH, we included the densities of seeds from conspecifics (conspSeed) and heterospecifics (hetspSeed) per trap per season as covariates mediating seed escape. To test the JC-distance hypothesis, we included the nearby tree densities (individuals ≥1 cm DBH) as covariates mediating seed escape. We estimate nearby tree conspecifics (conspTree) and heterospecifics (hetspTree) from basal areas in the FDP censuses that are conducted every five years (50) using the 2007 FDP census for adult densities mapped to the 2008-2011 seed censuses, the 2012 FDP census for adult densities in the 2012-2016 seed censuses, and the 2017 FDP census for adult densities in 2017-2019 seed censuses. We explored the appropriate spatial scale for including conspTree and hetspTree within a 5-, 10-,15- and 20-m radius using total basal area as well as distance-weighted.  The role of conspecific seeds and adults are inextricably linked, making it difficult to disentangle the relative contribution of JCH and PSH in shaping attack rates. For example, one may test support for the PSH versus JC-density by comparing seed escape rates with treatments of varying seed abundances, but satiation is also dependent on the dispersal of seed enemies from trees. Therefore, we included the distance-weighted basal area (BA) of conspecific and heterospecific trees in all analyses (Extended Data Table S1).