Avaliação da identificação botânica de árvores, em planos de manejo florestal no estado do Mato Grosso.

Abstract

The reliability of taxonomic identification in forest inventories is a key component for estimating stock by species and ensuring the traceability of harvested material in timber management operations in the southern Amazon. In this study, we examined the consistency of botanical identification of commercially important trees in operations in northern Mato Grosso State by comparing, on an individual-by-individual basis, the scientific name recorded in the inventory with the identification validated by botanical collections and taxonomic confirmation in a herbarium. Consistency was quantified at the species, genus, and family levels using agreement and performance metrics, including accuracy with bootstrap confidence intervals and Cohen’s κ coefficient, as well as the classification of errors when specific discrepancies occurred (same genus, same family, or different family). The consistency of the naming process was also evaluated in two stages (assignment of the common name and its conversion to the scientific name), along with the concentration of risk by sample substrate and volume. The results indicated moderate agreement at the species level (52.1%), contrasting with high agreement at the genus level (88.5%) and near-total agreement at the family level (99.5%). The κ coefficient confirmed this gradient, with κ = 0.5 at the species level and κ = 0.88 at the genus level. Among individuals with species-level errors, 76.1% remained in the same genus, 22.8% in the same family, and only 1.1% involved a change between families, evidencing Among individuals with species-level errors, 76.1% remained in the same genus, 22.8% in the same family, and only 1.1% involved a change between families, indicating that most inconsistencies result from “subtle” substitutions between closely related species. The distribution across agreement ranges reinforced this pattern: 52.1% of records showed complete agreement (100%), 36.5% showed agreement restricted to the genus (66%), 10.9% showed agreement restricted to the family (33%), and 0.5% showed complete disagreement (0%). In the evaluation of the naming process, high consistency of the common name was observed when compared to regional references (88.0%), but a reduction in the conversion of common names to scientific binomial names (52.1%), indicating that a substantial portion of the error is introduced by name-matching lists and documentary procedures that associate specific scientific names with vernacular labels that are often aggregative. Volume-weighted accuracy (71.2%) was higher than unweighted accuracy, but showed a significant volume associated with specific incorrect attributions and a concentration of risk in a few common names. Botanical identification in 100% inventories demonstrates robustness at broad taxonomic levels but significant limitations at the specific level, with errors systematically concentrated in taxonomic complexes and in highly recurrent common names. The findings support the adoption of targeted mitigation measures, including the local calibration of reference lists, the standardization of diagnostic criteria, and sample validation using vouchers that prioritize critical markers to reduce uncertainty and strengthen the technical basis for disease control.