Primary and secondary dispersal of Bertholletia excelsa: Implications for sustainable harvests

Our study tracked intact whole fruits of the giant, large-seeded tropical tree Bertholletia excelsa for the first time during primary (fruitfall) and secondary mammalian fruit dispersal. The scatterhoarding Dasyprocta spp. is particularly important for B. excelsa, because it disperses and often buries seeds into spatially-scattered caches, facilitating seed germination and early seedling establishment. We asked: (1) Given a short and synchronous fruiting period, what patterns of primary and secondary fruit dispersal emerge? (2) Do rainfall and local vegetation affect dispersal? We monitored date of fruitfall and condition (visibly untouched, gnawed open with some seeds removed, or gnawed, but unopened) of individual fruits from 20 trees over a ∼ 20 ha area, sequentially visiting each study tree on average every 3-4 days. We evaluated potential influences of rainfall and subcanopy vegetation (arboreal, liana, bamboo or herbaceous) on dispersal in four 10 m2 plots established under each tree crown. As expected, primary dispersal was fairly synchronous over an approximately 3-month period. Amount of rainfall in the 5 days previous to fruitfall was the best predictor of primary dispersal. Monitoring revealed that while the 6855 observed fruits were falling, scatterhoarders were removing some to outside the projection of the tree crown (4.1%) and opening a much smaller proportion (0.5%) under mother trees. Approximately one quarter of fruit opening or removal occurred within 5 days of fruitfall. More fruits were opened or removed when subcanopy vegetation had higher bamboo density or lower densities of saplings and trees - favorable microenvironments where rodents can gnaw fruits while remaining relatively hidden. Humans directly compete with native dispersers/predators for B. excelsa's seeds traded internationally as Brazil nut, and thus timing and intensity of harvests following fruitfall determines availability of this resource for dispersers. By mid-January when human harvesting typically begins in our study site, 94% of fruits had fallen from the tall trees, and our continuous quantification of disperser activity revealed that 197 fruits (2.9%) or 3341 seeds (assuming an average 17 seeds per fruit) were consumed or dispersed. Thus, human harvesters could safely collect and dispersers had weeks of unlimited resource access. Considering our dispersal findings, cumulative research results and collective understanding of the Bertholletia-Dasyprocta-Homo sapiens interactions, harvests in our study site and region boost incomes and likely do not threaten Brazil nut recruitment or maintenance of agouti populations.