Open workflow harmonizes 98,020 global non-native species first records

A standardized, open-source workflow has produced a revised global compilation of 98,020 first records for 30,279 non-native species, documenting when alien organisms were first reported outside their native ranges. Applied to source material assembled from scientific papers, checklists, databases and expert communications, the pipeline standardized taxonomic names, historical dates and geographic locations, then attached confidence ratings reflecting the precision of each reported date. The result, version 4.0 of the internal FirstRecords dataset, offers a more transparent and reproducible baseline for studying biological invasions, although it retains substantial geographic and taxonomic biases and cannot determine the true arrival date of a species from its first documentation alone.

The work addresses a fundamental problem in invasion science: first records are assembled from sources that use incompatible names, dates and place descriptions. A record might identify a species by an obsolete name, give an introduction date as a century or broad range, or name a location at the level of an island, county, sea or continent. Processing such evidence by hand is slow, difficult to reproduce and vulnerable to inconsistent decisions. The new workflow divides harmonization into six independently inspectable R scripts covering dataset preparation, taxonomy, dates, locations, remaining terminology and final output. Once source records have been entered in a structured table, a complete rerun takes minutes rather than the hours required by the earlier manual process.

Input must contain at least a taxon name, a year of introduction or naturalization, and a location, with each row representing one taxon-year-location event. The software does not extract records automatically from free text, PDF files or websites; researchers must first retrieve and tabulate the evidence themselves. Column names are converted to Darwin Core terminology where suitable, and custom fields preserve the original date and location alongside standardized values. The earliest record in the wild is preferred over a naturalization date when both are available. Additional information on habitat, establishment, occurrence status, introduction pathway, bibliographic citation and access rights can be processed when supplied, although several ecological traits beyond habitat are not yet standardized.

Taxonomic harmonization uses the Global Biodiversity Information Facility backbone through the rgbif package. Exact matches, synonyms and high-confidence fuzzy matches are handled as separate cases in an explicit decision tree, and every result is labeled in a workflow note. Standardized names receive unique taxon identifiers and broader assignments such as mammals, birds, insects, bryophytes, algae or fungi. All records are retained, including names that cannot be matched, which are exported for manual verification. This is an important safeguard rather than a claim that every fuzzy match is correct: the authors note that automated synonymization can occasionally misidentify older or informal names, particularly for poorly studied taxa. In the application, 28,812 taxa, representing 95.2% of the total, were aligned with the GBIF backbone taxonomy.

Dates are cleaned by removing punctuation, special characters and excess whitespace, and by converting number words into digits. Full dates are reduced to years, references are stripped, two-digit years are expanded, BC years become negative values, and relative expressions such as “20 years ago” are resolved against the publication year or a fixed reference year for highly uncertain statements. Broad ranges and decades are sampled within plausible intervals rather than placed at a single artificial boundary year, while centuries and qualified centuries are converted to representative values. A fixed random seed preserves reproducibility, and the original date remains available in a verbatim field. Each standardized date also receives a confidence category based on its specificity, the width of a range and historical plausibility. In the released data, 85,153 records, or 86.9%, were classified as high confidence; smaller shares were medium-high, medium, medium-low or low. The categories describe the precision of the reported temporal information, not the quality or credibility of its supporting source.

Locations are matched to a supplied reference table aligned with version 2.8 of the Database of Global Administrative Areas. The process normalizes spelling, removes language-specific articles and uses exact and keyword matching to accommodate synonyms and multilingual variants. Recognized places receive standardized names, identifiers and links to broader regional classifications. Unmatched, ambiguous or non-administrative places—including seas and some compound location strings—are retained without identifiers and listed for review, preventing the automated system from silently discarding them. In version 4.0, 273 of 380 locations, representing 71.8% of the locations in the original sources, were harmonized. Users can expand the reference tables when their analyses require a different geographic hierarchy or treatment of islands and marine areas.

The fifth step applies controlled vocabularies to occurrence status, establishment means, degree of establishment, introduction pathway and habitat. Terms indicating absence, confinement in captivity or cultivation, uncertainty and presence are mapped to standardized status categories. Unknown pathway and establishment terms are exported for manual curation, and remaining empty occurrence fields are treated as present because many sources explicitly mention occurrence only when a species is absent, not established in the wild or uncertain. That fallback is documented and should be checked by users for their own source data. Habitat is converted into terrestrial, freshwater, brackish, marine or combined categories, while introduction pathways follow the Convention on Biological Diversity’s hierarchy. The final step removes duplicates, prioritizes published records over other sources when retaining the first record for a taxon and location, and exports the main table together with taxonomy, location and source tables.

Applying the workflow to a large input compilation produced a dataset dominated by Europe, which accounts for 39.9% of standardized records, followed by Oceania at 15.3% and North America at 14.5%. Vascular plants constitute 59.9% of records, insects 18% and birds 4.4%. These distributions are not evidence that Europe has experienced proportionally more introductions or that alien plants are globally dominant; the authors attribute the pattern to unequal research, monitoring and reporting capacity. The temporal series shows low annual counts through much of the nineteenth century, gradual growth from the late 1800s and acceleration after about 1950, reaching more than 1,000 records per year around 2000. The sharp recent decline is attributed to reporting lag, not a demonstrated fall in introductions, because recent observations may not yet have been digitised, verified or published.

Version 4.0 contains 17.8% more records than version 3.1, with 15% more unique taxa and 27% more unique locations. Some changes reflect new source material, while others result from revised harmonization and taxonomic reclassification; record counts for mammals, reptiles, amphibians and algae consequently fell in some groups even as the overall dataset expanded. The confidence system also does not rate the credibility of evidence. A precisely dated but weak source can receive high temporal confidence, and first-record dates generally mark documentation rather than the organism’s actual arrival. Source compilation itself remains manual, dependent on literature monitoring and expert networks, and unmatched taxa and locations still require review. The workflow therefore automates standardization rather than discovery, and its open code, configurable vocabularies and preserved verbatim fields make decisions reproducible while leaving room for future source-quality indicators.

Subject of Research: An open workflow for standardizing global first records of non-native species

Article Title: A standardized workflow for compiling and harmonizing first records of non-native species

Article References: Renard Truong, T., Essl, F., & Seebens, H. (2026). A standardized workflow for compiling and harmonizing first records of non-native species. NeoBiota, 110, 1-20. https://doi.org/10.3897/neobiota.110.198728

Image Credits: AI Generated

DOI: 10.3897/neobiota.110.198728

Keywords: biological invasions, non-native species, first records, data harmonization, biodiversity monitoring, GBIF taxonomy, Darwin Core, temporal uncertainty, open-source workflow, reproducible research, standardized, workflow