Data and R code for Trait Gap-filling for Global Angiosperms

This record contains the species-level trait and environmental data, processed analysis data for 18,708 angiosperm species, the final Random Forest gap-filled trait dataset, and the R scripts used to compare Random Forest, Support Vector Machine, and Bayesian Hierarchical Probabilistic Matrix Factorization. The underlying trait records were derived from TRY and BIEN, and the climate layers were obtained from WorldClim. The species-level trait and environmental data provided here are merged and cleaned derivatives prepared from these source datasets. Six continuous functional traits were evaluated: plant height, wood density, seed mass, leaf nitrogen, leaf phosphorus, and specific leaf area. Imputation performance was assessed using ten repeated artificial-masking datasets at masking levels of 10–50%. Wood density was imputed only for species classified as woody, while leaf carbon was retained but not imputed.