Species abundance declines with increasing environmental and functional niche marginality in subtropical evergreen broad-leaved forests

This dataset contains species-level functional niche marginality values derived from 22 plant functional traits measured across 68 tree species in two permanent 20-hectare subtropical evergreen broadleaved forest dynamics plots in southwestern China: the Jizu plot (400 m × 500 m) and the Ailao plot (500 m × 400 m).Trait measurements correspond to the 2022 census for Jizu, which recorded 44,117 individuals (DBH ≥ 1 cm) from 94 species, and the 2019 census for Ailao, which recorded 79,895 individuals from 103 species.From these censuses, this dataset includes 33 species from Jizu (approximately 83% of individuals) and 35 species from Ailao (approximately 66% of individuals).All raw trait measurements were screened for quality and standardized to ensure comparability across traits.Functional niche marginality was quantified using multivariate analyses in R version 4.3.2 (R Development Core Team, 2023), calculated as the difference between each species’ mean trait value and the community-wide mean, representing the deviation of each species within the functional trait space of the community.The dataset consists of eight comma-separated values (CSV) files, each corresponding to one of four functional niche dimensions (structural construction, photosynthetic capacity, mechanical defense, and hydraulic strategy) for each of the two study plots.Each CSV file contains a table where rows are individual species and columns are the functional niche marginality values for the relevant traits, along with species abundance; column headers indicate trait variable names.All marginality values are unitless standardized distances from the community centroid.Files are encoded in UTF-8 format and are modest in size (~3 KB each).There are no missing data in any of the files.

Because marginality values are deterministically computed from standardized trait means, not from stochastic models or repeated measurements, no error ranges or confidence intervals are provided.The CSV files can be opened with standard spreadsheet software (e.g., Microsoft Excel, LibreOffice Calc) or statistical programs.