Field measurements reveal local limits of imputed functional traits in tropical trees

This study evaluated the reliability of functional data imputed in the classification of ecological strategies according to the CSR model (competitive, stress-tolerant, ruderal) in tree species of a riparian Atlantic Forest in Brazil. Empirical data collected in situ from 68 species were compared with data imputed using the Bayesian Hierarchical Probabilistic Matrix Factorization (BHPMF) technique, considering key functional attributes such as leaf area and specific leaf area. The results showed significant divergences between the classifications based on measured and imputed data, especially for the traits directly involved in the formulation of the CSR model. While imputation overestimated selection for stress tolerance, field data indicated a greater diversity of strategies, with a greater presence of competitive species. The discrepancies point to limitations in the exclusive use of imputed data in local assessments of functional diversity, highlighting the need to integrate empirical collections and improvements in imputation methods for applications in environmental management and restoration.