Field data: cocoa flower visitors and their ecological functions in agroforestries of China and Brazil

This dataset contains high-resolution observations of cocoa (Theobroma cacao) flower visitors, cocoa management type, and fruit set outcomes collected across 21 plots in two contrasting cocoa production regions: an emerging frontier in Hainan, China, and a long-established production landscape in Bahia, Brazil, during the 2023 flowering season. Data files include: (i) plot-level variables describing agroforestry systems (e.g., monoculture, low-, intermediate-, and high-shade systems), shade-tree species richness, and canopy cover (%); (ii) tree- and flower-level data on flowering intensity, monitored flower identity, and fruit set (binary outcome); (iii) detailed records of flower visitors and classifications (i.e. order, family, and morpho-species where possible), along with their assigned ecological functional roles (e.g., forager, herbivore, predator, and combinations). In addition, the dataset includes two R scripts to facilitate reproducibility and reuse of the analyses: (i) a script implementing generalized linear mixed models and post hoc multiple comparison analyses (Tukey-adjusted contrasts) to evaluate differences in flower visitor functional groups across agroforestry systems, and (ii) a script implementing structural equation models (SEMs) to assess direct and indirect relationships among canopy cover, flower visitor groups, flowering intensity, and fruit set. No personal or sensitive human data are included. All data were collected with permission from farm owners and in compliance with local regulations.