Are industrial floristic inventories valuable for resolving taxonomic ambiguity? A test case in the genus Glandularia (Verbenaceae)
Data files
Jul 01, 2026 version files 642.20 KB
-
Glandularia_annotation_data_all.csv
619.75 KB
-
Glandularia_morphology_statistics.csv
15.42 KB
-
README.md
7.04 KB
Abstract
Global commerce has facilitated the widespread introduction of nonnative plant species, contributing to biotic homogenization and posing challenges for accurate species identification, particularly when closely related nonnative taxa are introduced into poorly documented regions. This study addresses taxonomic ambiguity between two closely related species, Glandularia aristigera (S. Moore) Troncoso and G. tenera (Speng.) Cabrera, introduced to the southern United States of America (USA) from South America. A floristic inventory at the Port of Savannah, Georgia, USA, initially misidentified Glandularia spp. specimens due to their morphological similarities and overlapping synonyms. Through herbarium specimen analysis, detailed morphological comparisons, and statistical modeling, we confirmed the presence of G. tenera, a species previously documented from only one location in near another major maritime port in Alabama, USA. Our study identified taxonomic ambiguity among American Glandularia vouchers caused by inconsistent nomenclature and insufficient distinguishing characters in regional botanical literature. Subtle yet diagnostic morphological traits, including seed cluse shape and anther appendage positioning, were crucial for differentiation. Specifically, G. tenera had rostrate seed cluses, and anther appendages surpassing the corolla mouth compared to G. aristigera, forming distinct clusters in multivariate analyses with high predictability (99% accuracy). Statistical modeling effectively differentiates closely related taxa. Resolving taxonomic ambiguity in nonnative plants requires integrating traditional taxonomy, digital collections, statistical analysis, and international collaboration. Increased cooperation with global taxonomic experts, better access to international floras, and broader use of digital herbarium resources are essential for improving invasive species monitoring and biodiversity management.
Dataset DOI: 10.5061/dryad.jdfn2z3rt
Description of the data and file structure
This dataset contains morphological trait measurements and specimen annotation data used to evaluate taxonomic differentiation between two closely related species, Glandularia aristigera and Glandularia tenera (Verbenaceae). Specimens were examined from multiple herbaria as part of a study assessing whether floristic inventories conducted in industrial point-of-entry environments can help resolve taxonomic ambiguity.
Two data files are provided:
- Glandularia_morphology_statistics.csv – Morphological trait dataset used in statistical analyses.
- Glandularia_annotation_data_all.csv – Specimen-level annotation matrix documenting herbarium specimens examined during the study.
Files and variables
This dataset contains morphological trait measurements from herbarium specimens of Glandularia aristigera and Glandularia tenera. The dataset includes 365 specimens with measurements used in multivariate and machine learning analyses to evaluate morphological differentiation and species delimitation.
- cal.len: Calyx length measured in millimeters (mm).
- cor.len: Corolla length measured in millimeters (mm).
- cal.cor: Ratio of calyx length to corolla length.
- anth.appen: Position of anther appendages relative to the corolla mouth (categorical variable).
- seed.apex: Morphology of the seed cluse apex (categorical variable describing apex shape).
- seed.len: Seed cluse length measured in millimeters (mm).
- species: Species identification of the specimen (Glandularia aristigera or Glandularia tenera).
Missing values
Missing values in continuous traits were imputed prior to statistical analysis using k-nearest neighbor imputation (k = 5). Missing values, if present, are indicated as NA.
This dataset contains specimen-level annotation data and morphological character observations recorded during examination of herbarium specimens of Glandularia. Specimens were examined from 21 herbaria. The dataset documents specimen identifiers, original label identifications, updated annotations, and morphological characters used during taxonomic evaluation.
Each row represents one examined specimen.
- Investigator(s): Researcher(s) who examined and annotated the specimen.
- Date of specimen examination: Date when the specimen was examined.
- Re-examination: Indicates whether and when a specimen was re-examined during this study after Long & Marsico annotation to verify or update its taxonomic identification or morphological observations.
- Loan collection: Herbarium acronym of the institution from which the specimen was borrowed. Acronyms follow the Index Herbariorum standard (https://sweetgum.nybg.org/science/ih/).
- Accession number: Unique specimen identifier or herbarium barcode.
- Species ID on original label: Species identification recorded on the specimen label at the time of collection.
- Current annotation: Species identification assigned during this study following morphological evaluation.
- Collector: Name of the collector.
- Collection number: Collector's field number.
- Collection Date: Date when the specimen was collected.
- State / County: Geographic location where the specimen was collected.
- Reason for annotation: Documents the reason for assigning or revising the specimen annotation (e.g., taxonomic ambiguity based on diagnostic morphological characters such as anther appendages or seed cluse morphology).
- Overall observations: General notes recorded during specimen examination describing additional morphological features or growth characteristics relevant to taxonomic evaluation but not captured by the measured variables.
- leaf pubescence: Presence or absence of pubescence on leaves.
- floral bract description: Description of floral bract morphology.
- Length of calyx (mm): Calyx length measured in millimeters.
- Calyx with nerves (y/n): Presence or absence of visible nerves on the calyx.
- Outside of corolla pubescent (y/n): Presence or absence of pubescence on the outside of the corolla.
- Length of corolla: Corolla length measured in millimeters.
- anther appendages that exceed corolla mouth: Indicates whether anther appendages extend beyond the corolla mouth.
- seed apex shape: Morphological description of the seed cluse apex.
- seed length: Length of the seed cluse measured in millimeters.
- Flower dissection still needed: Indicates whether additional floral dissection was required.
- additional traits: Additional morphological observations.
- SK Rana annotation: Species annotation assigned by S.K. Rana.
- Long & Marsico annotation: Species annotation assigned by Long and Marsico.
Missing values
Missing data are indicated as blank cells or NA where measurements or observations were unavailable.
Code/software
The dataset is provided as a comma-separated values (CSV) file and can be opened using common spreadsheet software such as Microsoft Excel, LibreOffice Calc, or Google Sheets.
Statistical analyses and data processing were conducted using the R statistical computing environment (version 4.x). The following R packages were used for data analysis and visualization: VIM (for k-nearest neighbor imputation of missing values), ggplot2 (for visualization), FactoMineR (for principal component analysis), plotly (for interactive visualization), randomForest (for classification analysis), and pdp (for partial dependence plots).
Analysis scripts used to generate statistical results and figures are available in the project repository hosted on GitHub (https://github.com/santoshkumarrana/Trait-based-morphological-analysis-of-Glandularia/tree/main)
Access information
Other publicly accessible locations of the data:
- Analysis scripts and workflow used for statistical analyses and figure generation are available at the GitHub repository: https://github.com/santoshkumarrana/Trait-based-morphological-analysis-of-Glandularia/tree/main.
Data was derived from the following sources:
- The dataset was generated from morphological measurements of herbarium specimens examined for this study and from specimens collected during a floristic inventory conducted at the Port of Savannah Garden City Terminal, Savannah, Georgia, USA. The dataset represents processed morphological trait data used for statistical analyses of species differentiation between Glandularia aristigera and Glandularia tenera.
Morphological and taxonomic data were compiled from a floristic inventory conducted at the Port of Savannah Garden City Terminal, Savannah, Georgia, USA, an industrial point-of-entry site. Four collecting trips were carried out to capture seasonal variation in the flora (28 August 2015, 19 May 2016, 14 November 2016, and 27 February 2017). During these surveys, vascular plants that were reproductive or otherwise identifiable were collected as herbarium vouchers, typically in duplicate. Specimens of the genus Glandularia were collected during each survey. Details of the floristic inventory are reported in Lucardi et al. (2020b).
To support species identification and comparative analysis, herbarium specimens were examined from 21 herbaria (APSC, ARIZ, CLEM, CM, FLAS, FSU, FUGR, GA, GAS, GEO, LSU, MISSA, MO, NBYC, NCU, RSA, TENN, TEX, USCH, VSC, and WVA; acronyms following Thiers, continuously updated). Specimens bearing the epithets aristigera, pulchella, tenera, and tenuisecta under either Glandularia or Verbena were requested. A total of 1171 specimens were examined, and 1151 specimens belonging to G. aristigera and G. tenera were annotated. Additional specimens of related taxa were reviewed for comparative purposes but excluded from the statistical analyses.
For trait-based analyses, we retained only specimens collected within the United States that had complete measurements for the diagnostic characters: calyx length, corolla length, seed cluse length, seed cluse apex morphology, and the presence or absence of anther appendages extending beyond the corolla mouth (n = 365). Species identifications were initially performed using traditional taxonomic approaches including dichotomous keys, species descriptions, and comparison with verified herbarium material. Identifications were further refined through morphometric evaluation of diagnostic floral and fruit characters. In cases of uncertainty, additional resources including online herbarium databases and expert consultation were used to confirm taxonomic determinations.
Morphological trait data were analyzed in the R statistical environment. Missing values in continuous traits were imputed using the k-nearest neighbor algorithm (k = 5). Continuous traits were standardized and visualized using boxplots and principal component analysis (PCA). To evaluate the predictive power of morphological characters for species identification, a random forest classifier was implemented, and model performance was assessed using confusion matrices and variable importance metrics. Partial dependence plots were generated to visualize the contribution of individual predictors to species classification. Multidimensional scaling based on the random forest proximity matrix was used to visualize clustering patterns between G. aristigera and G. tenera.